{"id": "2606.20250", "title": "Single-Stage Hierarchical Rectification for Weakly Supervised Histopathology Segmentation", "abstract": "Existing weakly supervised semantic segmentation (WSSS) methods in computational pathology rely on a multi-stage paradigm: class activation map (CAM) generation, offline pseudo-mask refinement, and fully supervised retraining. While established, this decoupled approach presents fundamental limitations. The multi-stage process not only incurs high computational training costs but also suffers from error propagation: local texture biases in shallow CNN layers generate false-positive artifacts that subsequent refinement steps often fail to correct. To address these persistent challenges through a simple yet highly effective approach, we propose the Single-Stage Hierarchical Rectification (SSHR) framework. Rather than passively refining CAMs post-hoc, our method proactively purifies intermediate feature representations during the forward pass. We introduce a Hierarchical Feature Rectification Module (HFRM) that utilizes deep global semantic context to filter out local anomalies in shallow layers. This mechanism generates high-fidelity activation maps directly within a single training loop. Experiments on the LUAD-HistoSeg and BCSS datasets demonstrate that SSHR outperforms state-of-the-art multi-stage methods. Furthermore, SSHR reduces training duration by 2 to 5 times. This efficiency minimizes computational overhead and accelerates clinical translation for large-scale histopathology workflows. The code is available at: https://github.com/trongduc-nguyen/SSHR", "date": "2026-06-18", "categories": []}
{"id": "2602.02370", "title": "Uncertainty-Aware Image Classification In Biomedical Imaging Using Spectral-normalized Neural Gaussian Processes", "abstract": "Accurate histopathologic interpretation is key for clinical decision-making; however, current deep learning models for digital pathology are often overconfident and poorly calibrated in out-of-distribution (OOD) settings, which limit trust and clinical adoption. Safety-critical medical imaging workflows benefit from intrinsic uncertainty-aware properties that can accurately reject OOD input. We implement the Spectral-normalized Neural Gaussian Process (SNGP), a set of lightweight modifications that apply spectral normalization and replace the final dense layer with a Gaussian process layer to improve single-model uncertainty estimation and OOD detection. We evaluate SNGP vs. deterministic and MonteCarlo dropout on six datasets across three biomedical classification tasks: white blood cells, amyloid plaques, and colorectal histopathology. SNGP has comparable in-distribution performance while significantly improving uncertainty estimation and OOD detection. Thus, SNGP or related models offer a useful framework for uncertainty-aware classification in digital pathology, supporting safe deployment and building trust with pathologists.", "date": "2026-02-02", "categories": []}
{"id": "2606.15554", "title": "RaLMPH: Reliability-aware Learning for Multi-Pathologist Harmonization in Whole-Slide Image Classification", "abstract": "Multiple Instance Learning (MIL) is a standard paradigm for Whole-Slide Image (WSI) analysis and has achieved strong results in computational pathology. However, most MIL pipelines assume a single \"gold\" label per slide, which conflicts with clinical practice where substantial inter-pathologist variability is common. Existing multi-annotator learning and label-refinement methods typically estimate global annotator reliability or rely on single-instance assumptions, making them poorly suited to MIL and to localized diagnostic contexts where experts disagree. We propose RaLMPH (Reliability-aware Learning for Multi-Pathologist Harmonization), a MIL-based label reconciliation framework for WSIs annotated by multiple pathologists. RaLMPH introduces a reliability field that jointly models (i) local neighborhood structure in WSI feature space and (ii) expert uncertainty (entropy), enabling per-sample identification of trustworthy reference neighborhoods. Leveraging this field, RaLMPH performs sample-wise local annotator ranking to select reliable opinions per slide and applies an adaptive gating mechanism to fuse labels conditioned on local reliability. Experiments on a clinical WSI dataset with labels from six pathologists, as well as controlled simulated benchmarks, show that RaLMPH consistently outperforms existing approaches. Further analyses clarify how our reliability-aware mechanism improves label reconciliation and downstream MIL performance.", "date": "2026-06-14", "categories": []}
{"id": "2606.17412", "title": "Enhancing Pathological VLMs with Cross-scale Reasoning", "abstract": "Pathological images are inherently multi-scale, requiring pathologists to integrate evidence from global tissue architecture at low magnification to cellular morphology at higher magnification for accurate diagnosis. While existing pathological datasets for vision-language model (VLM) include various scales, they often lack an explicit cross-scale reasoning objective. This limitation prevents VLMs from capturing essential cross-scale representations and learning evidence-based reasoning. To bridge this gap, we introduce the first cross-scale training and evaluation paradigm that formulates pathology interpretation as multi-magnification reasoning. However, creating such a task reveals a critical challenge: multi-image visual question answering (VQA) is prone to text-only shortcuts, which allow models to guess answers using magnification-dependent artifacts rather than visual evidence. To address this, we propose a leakage-aware curation pipeline that combines adversarial text-only screening with constraint-guided question design. Using this pipeline, we construct Scale-VQA, a high-quality benchmark with 4,685 multiple-choice questions grounded in 2,537 pathology images across multiple magnification levels. Finally, we present ScaleReasoner-R1, a model trained via reinforcement learning to optimize performance on the cross-scale VQA task. ScaleReasoner-R1 achieves state-of-the-art performance on our cross-scale reasoning benchmark and generalizes to SOTA performance on established single-scale benchmarks. Findings suggest that even the limited cross-scale supervision can significantly improve pathological understanding. The code and demos will be open-sourced.", "date": "2026-06-16", "categories": []}
{"id": "2606.18123", "title": "Predicting Immune Biomarkers with MultiModal Mixture-of-Expert Pathology Foundation Models Empowers Precision Oncology", "abstract": "Predicting immune biomarkers associated with the tumor immune microenvironment (TIME) is critical for advancing precision oncology, yet existing approaches are largely limited to single image modalities and suffer from insufficient resolution and incomplete utilization of complementary clinical and biological information. Here we introduce MixTIME, a multimodal foundation model that leverages a mixture-of-experts (MoE) architecture to integrate pathology foundation models trained across distinct modalities: image only (UNIv2), image text (CONCHv1.5), and image transcriptomic (STPath) representations for pixel-level and slide-level prediction of multiplex immunofluorescence (mIF) protein expression from hematoxylin and eosin (HE) whole-slide images. MixTIME employs a learnable router to dynamically weight expert contributions and is trained with a distribution- and tendency-aware loss function. Benchmarked on two datasets of different scales, MixTIME achieves state-of-the-art performance across 17 protein markers as measured by correlation metrics. The predicted mIF profiles substantially enhance downstream tasks, including spatial domain identification, survival prediction, and AI-assisted pathology report generation validated by expert pathologists from multiple institutes across the world. Furthermore, MixTIME enables longitudinal tracking of protein expression dynamics across clinical time points and reveals protein gene interaction patterns linked to drug resistance and immune suppression in tumor microenvironments. Collectively, MixTIME provides a scalable framework for multimodal biomarker discovery and clinical translation in computational pathology.", "date": "2026-06-16", "categories": []}
{"id": "2606.02877", "title": "Pathway-Structured Privileged Distillation for Deployable Computational Pathology", "abstract": "Integrating transcriptomics and histopathology can improve cancer risk modelling, yet practical use is constrained by the limited availability of RNA profiling in routine settings. Here we introduce Mixture of Pathway Experts (MoPE), a knowledge-distillation framework that reframes multimodal learning as privileged distillation for histology-only inference. MoPE is motivated by the partial observability between RNA profiles and whole-slide images: histology can capture morphology-linked consequences of certain molecular programmes, but cannot be expected to reconstruct the full transcriptomic state. MoPE encodes RNA-derived pathways and transfers the molecular supervision to pathway-indexed pathology experts through memory-usage alignment. Across diverse public benchmarks and two independent breast cancer cohorts, MoPE consistently improved WSI-only inference performance relative to baseline methods. Pathway-usage analyses and human-audited visual inspection provide bounded inspection of model behaviour and candidate morphology-linked readouts. These results support pathway-structured privileged distillation as a promising route to using molecular information during training while preserving RNA-free inference.", "date": "2026-06-01", "categories": []}
{"id": "2606.16477", "title": "AURA: Active-Response Attribution under Treatment Ambiguity in Bacterial Cytological Profiling", "abstract": "When a bacterial sample is exposed to several antibiotics, not every applied drug necessarily acts: if the organism is resistant to one of them, that drug leaves no morphological trace. The clinically meaningful quantity is therefore not which antibiotics were applied, but which ones were active. We show that these two are sharply decoupled in real E. coli microscopy - naively assuming the applied combination equals the active one is correct only about 37% of the time - yet existing computational tools are ill-suited to recovering the active set. Forward perturbation models such as scGen, CPA, and IMPA are designed to predict appearance from treatment, not the reverse, and inverting them degrades sharply; discriminative image classifiers tend to memorise strain- and batch-specific texture and fail to transfer across experimental replicates. We introduce AURA, which reframes the task as constrained, energy-based inverse attribution. Its central inductive bias is that the active set must be a subset of the applied set; this collapses the candidate space and lets AURA infer the active subset of applied antibiotics by decomposing residual morphology into antibiotic response atoms and selecting the subset with the lowest reconstruction energy, using no strain label at test time. AURA-E adds evidence-aware abstention, withholding a prediction when candidate explanations remain near-equally plausible. On cross-replicate transfer in an E. coli cytological profiling dataset, AURA recovers the active antibiotic combination with 95.47% exact-match accuracy.", "date": "2026-06-15", "categories": []}
{"id": "2606.16658", "title": "Vision-Language Models as Zero-Annotation Oracles in Histopathology", "abstract": "Foreground segmentation is the critical first step of every computational pathology pipeline, yet existing methods rely on hand-tuned heuristics or supervised models that overfit to narrow stain and scanner distributions, failing silently on specialised stains such as Jones silver or Elastica van Gieson. We propose a coarse-to-fine approach that recasts foreground segmentation as a visual perception task and leverages general-purpose vision-language models (VLMs) as zero-annotation oracles. Our key insight is that tissue-versus-background discrimination is a natural-image recognition problem, not a histopathological one, so VLMs trained on internet-scale corpora generalise where domain-specific models cannot. We introduce Leica-75, a benchmark of 75 renal transplant whole-slide images spanning three stain families. On Leica-75, our method achieves the highest segmentation quality on out-of-distribution stains (Dice 0.858 +/- 0.027 on Jones, 0.853 +/- 0.041 on EVG) with 7x lower cross-stain variance than the best supervised baseline, while remaining competitive on in-distribution H&E. Few-shot prompting with automatically curated exemplars (Auto-context) rescues hard cases on Stress-32 (n=32), a curated stress-test subset (Dice 0.470 to 0.819 for the 2B model). VLM-based annotation review matches human expert consensus (kappa=0.989 for blur detection; mean precision/recall grading accuracy 0.708 vs. human 0.646 for segmentation mask review). The resulting pseudo-labels are used to distil lightweight student models that are as performant as the teacher model while running for a fraction of the cost. Our framework provides a principled, scalable solution to a persistent infrastructure bottleneck in digital pathology.", "date": "2026-06-15", "categories": []}
{"id": "2606.11846", "title": "SheafStain: Sheaf-Theoretic Schr\\\"odinger Bridge for Spatially and Biologically Coherent Virtual Staining", "abstract": "Current virtual staining approaches offer the potential for time- and cost-efficient biomarker quantification in cancer diagnostics and prognostics. However, patch-wise inference for gigapixel whole slide images (WSIs) fails to maintain spatial continuity, yielding artifacts that cause catastrophic mismatches with ground-truth images. Although pathology Vision Foundation Models (VFMs) offer rich representations, their self-attention causes varying global contexts to produce inconsistent embeddings for the same physical region. We formalize and validate this ``context contamination'' as a sheaf-theoretic problem where these embeddings form a presheaf that violates the gluing axiom. To address this, we propose SheafStain, a new approach that reinterprets VFM features as sheaf-like sections for spatially and biologically coherent virtual staining. Specifically, SheafStain integrates class and patch tokens into a Schr\\\"odinger Bridge framework as sheaf-like sections. While the class token anchors biological consistency, patch tokens form a per-position spatial map. A backbone co-pretrained on Hematoxylin \\& Eosin (H\\&E) and Immunohistochemistry (IHC) yields non-degenerate cross-stain stalks, so a single VFM feature space supervises both input conditioning and output stain alignment. Departing from prior work that evaluates on isolated $256 \\times 256$ patches and either random-crops or resizes the $1024 \\times 1024$ ground truth, we translate at $256 \\times 256$ and evaluate on the stitched $1024 \\times 1024$ outputs across HER2, ER, PR, and Ki-67. SheafStain demonstrates promising results against six prior methods while mitigating patch-boundary stitching artifacts. Code will soon be released.", "date": "2026-06-10", "categories": []}
{"id": "2606.12126", "title": "AGE-MIL: Anchor-Guided Evidence Learning for Patient-Level Prediction", "abstract": "Existing computational pathology methods predominantly operate within whole-slide image (WSI)-level multiple instance learning (MIL) paradigms, while patient-level modeling remains underexplored. In routine pathological practice, however, pathologists derive diagnostic and prognostic conclusions by integrating evidence across multiple WSIs rather than relying on any single slide. This discrepancy creates a fundamental misalignment when patient-level supervision is directly imposed on conventional MIL frameworks, often leading to unstable optimization and degraded predictive reliability. To address this issue, we propose Anchor-Guided Evidence MIL (AGE-MIL), a weakly supervised framework for patient-level prediction. AGE-MIL constructs a patient-level anchor from slide representations to capture global pathological context and guide the retrieval and integration of diagnostically relevant local patches, enabling robust patient-level modeling. Patient-level risk is further modeled as an evidence accumulation process, promoting stable optimization under weak supervision. AGE-MIL is evaluated on six clinically relevant patient-level prediction tasks from two independent cohorts. Experimental results show that the proposed framework consistently outperforms eight state-of-the-art MIL methods. Code is available at https://github.com/wodeniua/AGE-MIL.", "date": "2026-06-10", "categories": []}
{"id": "2606.12346", "title": "Atlas H&E-TME: Scalable AI-Based Tissue Profiling at Expert Pathologist-Level Accuracy", "abstract": "Hematoxylin and eosin (H&E) staining is the cornerstone of histopathology, yet scalable, quantitative analysis of H&E whole-slide images (WSIs) remains a central challenge in computational pathology. We present Atlas H&E-TME, an AI-based system built on the Atlas family of pathology foundation models that predicts tissue quality, tissue region, and cell type labels across multiple cancer types, yielding over 4,500 quantitative readouts per slide at cell-level resolution. A key challenge to validating such systems is overcoming morphological ambiguity inherent to H&E-only ground truth and the limited scalability of more informed references drawing on modalities such as immunohistochemistry (IHC). We address this with a dual validation framework combining biologically grounded depth with technical and morphological breadth. For depth, we propose an IHC-informed multi-pathologist consensus protocol that substantially improves inter-rater agreement over conventional H&E-only annotation. This yields a molecularly grounded reference against which we compare Atlas H&E-TME and pathologists working from H&E alone. For breadth, we benchmark Atlas H&E-TME on over 200,000 high-confidence H&E-only pathologist annotations across 1,500+ cases spanning eight cancer types and their most common metastatic sites, with subtypes covering >90% of clinical cases per cancer type, drawn from 25+ sources and 8+ scanner models. Benchmarked against the IHC-informed consensus, Atlas H&E-TME matches or exceeds pathologist H&E-only performance and generalizes consistently and robustly across this broad morphological and technical scope. In doing so, Atlas H&E-TME turns the H&E slide -- the most ubiquitous data in pathology -- into a scalable, quantitative window into the tumor and its microenvironment, laying a foundation for the next generation of tissue-based biomarkers in translational and clinical research.", "date": "2026-06-10", "categories": []}
{"id": "2606.12407", "title": "How Seemingly Inconsequential Design Choices Dictate Performance of LLMs in Pathology", "abstract": "General-purpose large language models (LLMs) are routinely used as baselines when evaluating specialized pathology models on whole-slide images (WSIs). Because WSIs exceed contemporary model context limits, LLM baselines routinely use small, high-magnification patches processed independently via majority voting, without systematic evaluation of seemingly inconsequential design choices such as patch size, patch count, and magnification. Generalist LLMs have consistently underperformed specialized systems, reinforcing the perception that domain-specific training or architectural adaptation is necessary for pathology tasks involving WSIs. Here, we conduct a systematic factorial analysis of four input design factors: inference mode, patch size, magnification, and patch count. We demonstrate that prior studies have overstated the gap between specialized models and general-purpose LLMs by choosing non-optimized input configurations. On the MultiPathQA benchmark, switching to a single balanced configuration (large patches at lower magnification, processed jointly) raises GPT-5 from 15.1% to 39.5% on cancer-type classification (TCGA) and from 38.1% to 62.9% on organ classification (GTEx). Per-task optimization yields further gains up to 43.9% (TCGA) and 71.6% (GTEx). The same configuration generalizes to two other models and to a fully held-out CPTAC cohort, where it improves Gemini 3 Flash by 23.4 percentage points without any task-specific tuning.", "date": "2026-06-10", "categories": []}
{"id": "2505.23341", "title": "Dual-stream attention-guided learning for weakly supervised whole slide image classification", "abstract": "Whole slide images (WSIs) play a crucial role in cancer diagnosis due to their ultra-high resolution and rich morphological information, and multiple instance learning (MIL) has become a prevalent paradigm to solve the massive size of WSIs and the scarcity of fine-grained annotations of instance. However, most existing MIL methods struggle to accurately identify diagnostically critical local regions (instance) using only slide-level labels, and suffer from modelling the relationship of instances efficiently. To address these defects, we propose a Dual-Stream Attention-Guided Learning (DSAGL) framework. DSAGL bridges slide-level supervision and instance-level learning through a teacher-student dual-stream architecture, and mitigates instance ambiguity by generating attention-guided pseudo labels. The framework employs a shared lightweight encoder to efficiently model long-range dependencies and an attention-based fusion mechanism to enhance sensitivity to sparse, informative regions. Extensive experiments on synthetic benchmarks and real-world pathological WSI datasets demonstrate that DSAGL consistently outperforms state-of-the-art MIL methods, achieving superior discriminative performance and robustness under weak supervision.", "date": "2025-05-29", "categories": []}
{"id": "2602.01951", "title": "Enabling Progressive Whole-slide Image Analysis with Multi-scale Pyramidal Network", "abstract": "Multiple-instance Learning (MIL) is commonly used for computational pathology (CPath), where multi-scale features are essential for capturing both fine cellular details and broad tissue architecture. However, existing multi-scale MIL approaches typically rely on the inflexible multi-magnification inputs or the computationally expensive architectures. As pre-trained foundation models (FMs) become the trend for feature extraction and boost lightweight models, we rethink and explore a more efficient multi-scale MIL method. In this paper, we propose the Multi-scale Pyramidal Network (MSPN), a plug-and-play module for attention-based MIL. MSPN introduces progressive multi-scale whole-slide image analysis using only a single high-magnification input. It consists of (1) grid-based remapping that aggregates high-magnification features to derive spatially-aware coarse feature maps, and (2) the Coarse Guidance Network (CGN) that learns coarse contexts. We benchmark MSPN as an add-on module to 4 attention-based frameworks on 5 clinically relevant tasks with 2 foundation models, and a pre-trained MIL framework. Our results demonstrate that MSPN consistently improves MIL across the compared configurations and tasks, while being lightweight and easy-to-use.", "date": "2026-02-02", "categories": []}
{"id": "2606.09898", "title": "TRAPS: Therapeutic Response Analysis via Pathway-informed Stratification", "abstract": "Cancer treatment planning requires decisions across multiple clinical dimensions at once. Clinicians must determine whether a patient should receive targeted molecular therapy, radiation therapy, and whether they are likely to survive beyond six months. Existing pathway-informed deep learning models have been developed and tested in isolation, making fair comparison across architectures impossible. We present the first unified benchmark for pathway-guided therapy response modeling, evaluating three biologically informed architectures, BINN, GraphPath, and PATH, across five cancer cohorts drawn from The Cancer Genome Atlas, representing 2,622 patients encoded using Reactome pathway activity scores. Each model is trained jointly on all three clinical outcomes under identical data and evaluation conditions, the first study to treat pathway-structured deep learning as a combined therapy and survival prediction problem. Our results show that no single architecture wins across all tasks: PATH performs best for targeted molecular therapy prediction overall, BINN is most reliable for survival prediction, and no model produces useful predictions for radiation therapy, as the key drivers of that decision are clinical variables not captured in gene expression data. Most strikingly, GraphPath achieves an AUROC of 0.92 on prostate targeted molecular therapy prediction, the highest score in the entire benchmark, demonstrating that lateral co-regulation structure produces exceptional discriminative power when matched to a cohort with a narrow targetable driver programme, even under conditions of extreme class imbalance at only 11\\% positive prevalence.", "date": "2026-06-05", "categories": []}
{"id": "2606.10778", "title": "From Patches to Patients: A study of the tile-to-slide performance transferability in Digital Pathology", "abstract": "Foundation Models (FMs) have recently redefined the state-of-the-art in histopathology by providing robust representations for whole-slide image (WSI) analysis. However, selecting the optimal foundation model (FM) for a specific clinical cohort currently requires multiple preprocessing steps, followed by computationally expensive feature extraction and the training of a Multiple Instance Learning (MIL) aggregator for every model. In this work, we investigate whether efficient tile-level linear probing can serve as a reliable proxy for slide-level performance, reducing the need to run full slide-level pipelines for every candidate encoder. We benchmark 19 state-of-the-art FMs on 42 slide-level and 16 tile-level tasks, comparing tile probing metrics against slide-level outcomes using ABMIL and Mean Pooling aggregations. We observe a high correlation between tile and slide performance across varying task difficulties, indicating that encoder representation quality is the primary determinant of WSI success. Sensitivity analyses show that transferability is stable across models and is more influenced by cohort sizes and numbers of tiles per slide than by average task difficulty. We also measure the agreement in best performing models between tile and slide-level tasks, showing tile benchmarks reliably shortlist strong candidates. Overall, our study indicates that tile-level benchmarking provides an efficient and practical first step for narrowing down candidate models, while slide-level evaluation remains essential for final validation on clinical tasks.", "date": "2026-06-09", "categories": []}
{"id": "2606.07549", "title": "PathoSage: Towards Multi-Source Evidence Adjudication in Pathology via Experience-Aware Agentic Workflow", "abstract": "Recent advances in Multimodal Large Language Models (MLLMs) and agent workflows have shown strong promise for computational pathology, yet reliable patch-level reasoning remains challenging. End-to-end pathology MLLMs often hallucinate morphological features, while recent agentic systems usually merge tool outputs and retrieved knowledge into a shared context, making decisions vulnerable to conflicting evidence and context contamination. We propose PathoSage, a three-stage framework that explicitly separates knowledge retrieval, evidence collection, and evidence adjudication for patch-level pathology multimodal reasoning. Its core component, Structured Evidence Deliberation, independently evaluates heterogeneous evidence from tools, performs conflict analysis, and generates the final judgment in a fresh context to reduce anchoring bias. We further introduce a training-free Beta-Bernoulli experience system with continuous credit assignment to model long-term tool reliability and construct similarity-weighted priors for future tool use. Experiments show that PathoSage effectively mitigates VQA hallucinations and classifier disagreement, outperforming strong pathology MLLM and agentic baselines. Our results highlight explicit evidence adjudication and reliability-aware tool modeling as key ingredients for robust pathology agents.", "date": "2026-05-18", "categories": []}
{"id": "2606.08093", "title": "A Multi-modal Agentic Co-pilot for Evidence Grounded Computational Pathology", "abstract": "Pathology is the cornerstone of modern medicine, where accurate decision-making relies heavily on evidence-based practices. While artificial intelligence (AI) has the potential to transform clinical workflows, the intersection of AI and evidence-based medicine remains under-explored, with primitive attempts restricted to text-only general medicine. In this work, we present PathPocket, a multimodal AI agentic co-pilot designed specifically for evidence grounded pathology. We construct the most comprehensive pathology evidence corpus to date, encompassing approximately 110,472 public and authorized documents structured across a rigorous hierarchy of evidence from clinical guideline to expert opinion. From this meticulously graded foundation, we build a large-scale multimodal pathology hypergraph containing over 4.55 million entities and 7.10 million relations. Serving as a robust knowledge engine, this hypergraph provides traceable evidence for a collaborative multi-agent reasoning framework integrating input understanding, evidence retrieval, filtering, and diagnosis generation. This enables PathPocket to seamlessly resolve a wide spectrum of clinical tasks, ranging from text-only queries to complex multimodal diagnostics involving region-of-interest (ROI) and gigapixel whole-slide images (WSIs). We rigorously evaluate the system on a multidimensional benchmark of over 200,000 real-world cases, where it significantly outperforms existing state-of-the-arts. Crucially, extensive user studies demonstrate that PathPocket substantially improves the diagnostic accuracy and confidence of pathologists. By directly grounding pathology interpretations in verifiable literature, PathPocket offers a practical and scalable solution for the future of evidence grounded computational pathology.", "date": "2026-06-06", "categories": []}
{"id": "2606.08745", "title": "Stain-Aware Wavelet Regularization for Instant Adversarial Purification in Histopathology", "abstract": "Deep learning has become prevalent in computational pathology pipelines that support tasks such as cancer screening and digital pathology analysis. However, the susceptibility of neural networks to adversarial perturbations raises safety concerns for reliable deployment in clinical practice. In histopathological images, this challenge is exacerbated by the difficulty of distinguishing high-frequency adversarial noise from subtle and diagnostically relevant tissue structures. To address this issue, we propose Stain-Aware Wavelet Regularization (SAWR), an adversarial purification framework that leverages multi-level wavelet-domain regularization based on Haar transform to hierarchically disentangle adversarial perturbations from diagnostic structural information. This spectral constraint is further extended to individual histological channels, enabling stain-specific frequency regulation consistent with the biological properties of Hematoxylin and Eosin. When integrated into an instant purification framework, SAWR improves adversarial robustness by up to 10.69\\% over the baseline approach, while maintaining texture and spectral fidelity under adversarial perturbations.", "date": "2026-06-07", "categories": []}
{"id": "2606.09453", "title": "GD-MIL: Grade-Disentangled Multiple Instance Learning for Multimodal Biochemical Recurrence Prediction in Prostate Cancer", "abstract": "Biochemical recurrence (BCR) after radical prostatectomy is a critical endpoint in prostate cancer, yet risk stratification relies almost entirely on variables dominated by Gleason grade. Whether H&E whole slide images (WSIs) carry prognostic signal beyond grade, and whether multiple instance learning (MIL) can recover it, remains unsettled. A key obstacle is that many pipelines select model checkpoints on the evaluation fold, artificially inflating concordance. We construct a rigorous benchmark on TCGA-PRAD (487 patients, 101 BCR events) using strict out-of-fold scoring over five-fold cross-validation repeated across five seeds. The choice of MIL aggregator (ABMIL, CLAM, TransMIL, PatchGCN) has little effect (C-index 0.61-0.64 with UNI2-h), while the feature extractor is the dominant factor (ResNet50 0.566 versus pathology foundation models up to 0.639). A clinical Cox model on grade, stage, and age reaches 0.687; no imaging-only model significantly outperforms it (p > 0.10). We introduce Grade-Disentangled MIL (GD-MIL), a gated-attention MIL encoder trained with a gradient-reversal grade adversary that encourages the slide representation to be invariant to Gleason grade before late fusion with clinical variables. GD-MIL achieves C-index 0.704, significantly outperforming both the clinical baseline (delta-c = +0.029, p = 0.0005) and the best imaging-only model (delta-c = +0.062, p = 0.039), suggesting H&E morphology contains prognostic information complementary to grade. A median risk split yields log-rank p < 0.0001 separation in BCR-free survival (~20% vs ~70% at five years).", "date": "2026-06-08", "categories": []}
{"id": "2606.06983", "title": "DaX: Learning General Pathology Representations Across Scales", "abstract": "Computational pathology requires visual representations that transfer across diverse clinical endpoints and remain robust to variation in magnification, staining, scanner type, slide preparation, and input resolution. We present DaX, a pathology vision foundation model that adapts DINOv3-style self-supervised learning to whole-slide histopathology. DaX is initialized from natural-image DINOv3 weights and incorporates continuous magnification training, cross-scale tissue views, orientation-agnostic and acquisition-robust augmentation, multi-input-size training, and Gram-anchored dense consistency. These designs aim to connect local cellular morphology with global tissue architecture while stabilizing dense token-level representations across input scales. We further construct a WSI-level benchmark comprising 161 clinically meaningful tasks from 44 public datasets, covering 28,182 patients and 34,394 slides across four clinical domains and nine task categories. All models are evaluated under a fixed patient-level cross-validation protocol with fold-level statistical ranking, enabling reproducible comparisons that are less sensitive to split-dependent variation. Across this benchmark, DaX achieves the highest mean performance across tasks and consistently strong task-level ranking scores, with gains spanning diagnostic pathology, biomarker and molecular profiling, tissue/specimen context, and risk, response, and prognosis. These results support DaX as a transferable visual encoder for computational pathology and provide a standardized evaluation framework for future pathology foundation models. Project page: https://alibaba-damo-academy.github.io/DaX/benchboard/.", "date": "2026-06-05", "categories": []}
{"id": "2606.07036", "title": "STREAM: Stochastic Riemannian Flow Matching with Anisotropic Decoder for Digital Histopathology Image Generation", "abstract": "Synthetic histopathology image generation addresses critical challenges in computational pathology, including patient privacy and the growing need for large-scale training data for foundation models. Latent diffusion models have dominated the image generation domain, with recent works emphasizing that the choice of latent space is critical to the quality of generated images. Existing state-of-the-art generative models in histopathology use pretrained Vision Foundation Models (VFMs) as conditioning signals, and we observe that this leads to \"conditioning collapse,\" where the conditioning signal dominates the latent space and lowers the quality and diversity of generated samples. Therefore, we instead use pretrained histopathology VFMs as the latent space itself, leveraging their patch-token features that encode rich semantic information. We empirically show that these features are $\\ell_2$-normalized and lie on the unit hypersphere $\\mathcal{S}^{d-1}$ with strong angular dominance and intrinsic curvature, making them naturally suited for a Riemannian formulation. We therefore present STREAM, the first framework to apply Riemannian flow matching in the pathology domain. STREAM consists of two stages: 1) a bridge-type stochastic perturbation that establishes per-token rectifiability on $\\mathcal{S}^{d-1}$ for training a Diffusion Transformer (DiT) in latent space, and 2) a novel anisotropic decoder that allocates robustness to low-energy directions of the velocity-field Jacobian while preserving fidelity along its high-energy directions. Together, STREAM achieves state-of-the-art reconstruction and generation performance on breast and colorectal cancer datasets. The code will be publicly released upon acceptance.", "date": "2026-06-05", "categories": []}
{"id": "2606.07222", "title": "DualGate-Net: A Prior-Gated Dual-Encoder Framework for Histopathology Cell Detection", "abstract": "Cell detection in histopathology images strongly depends on surrounding tissue context, where visually similar cells may belong to different classes under different microenvironments. Recent tissue-aware methods incorporate contextual priors, but often rely on static fusion strategies that may propagate noisy information. In this work, we propose DualGate-Net, a prior-aware dual-encoder framework that combines a ConvNeXtV2-based local encoder and a SegFormer-based global encoder through a learnable prior-gated fusion mechanism. The proposed module adaptively regulates the influence of tissue priors across spatial locations, while an auxiliary foreground reconstruction branch preserves high-frequency cellular structures during training. In addition, auxiliary cellness-guided cues are incorporated to further improve localization robustness. Experiments on the OCELOT benchmark demonstrate consistent improvements, achieving macro F1-scores of 0.7722 on the validation set and 0.7345 on the test set, highlighting the effectiveness of adaptive prior integration for robust histopathology cell detection.", "date": "2026-06-05", "categories": []}
{"id": "2606.07368", "title": "Mitosis Detection in the Wild: Multi-Tumor and Context-Aware Generalization in the MIDOG 2025 Challenge", "abstract": "Automated mitosis detection is a well-established task in computational pathology. While previous benchmarks focused on scanner-induced domain shift, clinical \"real-world\" application requires models to be robust across the vast variance to be expected in the histological landscape. The MItosis DOmain Generalization (MIDOG) 2025 challenge was designed to evaluate algorithmic performance across unprecedented biological and contextual diversity. We curated a test dataset of 365 cases, encompassing 12 distinct human, canine and feline tumor types, digitized across multiple scanning platforms. Moving beyond hand-selected hotspots, the challenge required detection also in random tissue areas (representative of the whole slide detection situation) and challenging areas (areas rich in hard negatives). In the second track, we introduced the classification of atypical mitotic figures (AMFs). There were 18 teams submitting to the detection track, with F1 scores ranging up to 0.740. In the AMF detection track, we had 21 submissions with balanced accuracy values up to 0.908. Our analysis reveals that while most models perform reliably in traditional hotspots, significant performance degradation occurs in challenging ROIs, where false positive rates tripled. Furthermore, performance varied significantly across the 12 tumor types, highlighting \"blind spots\" in current state-of-the-art architectures when encountering rare or highly pleomorphic malignancies. Moreover, we evaluated the effectiveness of ensembling and found a mean increases of 1.5 and 1.3 percentage points in F1 score and balanced accuracy, respectively. In contrast, TTA showed no relevant improvement. MIDOG 2025 demonstrates that \"in the wild\" mitosis detection remains a significant hurdle. The transition from hotspot-only evaluation to a multi-contextual framework provides a more realistic proxy for clinical reliability.", "date": "2026-06-05", "categories": []}
{"id": "2503.10629", "title": "Hierarchical Self-Supervised Adversarial Training for Robust Vision Models in Histopathology", "abstract": "Adversarial attacks pose significant challenges for vision models in critical fields like healthcare, where reliability is essential. Although adversarial training has been well studied in natural images, its application to biomedical and microscopy data remains limited. Existing self-supervised adversarial training methods overlook the hierarchical structure of histopathology images, where patient-slide-patch relationships provide valuable discriminative signals. To address this, we propose Hierarchical Self-Supervised Adversarial Training (HSAT), which exploits these properties to craft adversarial examples using multi-level contrastive learning and integrate it into adversarial training for enhanced robustness. We evaluate HSAT on multiclass histopathology dataset OpenSRH and the results show that HSAT outperforms existing methods from both biomedical and natural image domains. HSAT enhances robustness, achieving an average gain of 54.31% in the white-box setting and reducing performance drops to 3-4% in the black-box setting, compared to 25-30% for the baseline. These results set a new benchmark for adversarial training in this domain, paving the way for more robust models. Our Code for training and evaluation is available at https://github.com/HashmatShadab/HSAT.", "date": "2025-03-13", "categories": []}
{"id": "2606.04764", "title": "Do Foundation Models See Biology? Evaluating Attention Coherence with Spatial Transcriptomics in Glioblastoma", "abstract": "Whether attention maps from pathology foundation models capture genuine biology remains unknown, yet this question is critical for clinical trust and regulatory approval. We propose a spatial transcriptomics-based framework for orthogonal, hypothesis-free evaluation of attention and apply it to five pathology foundation models (CONCH v1.5, UNI v2, Virchow2, GigaPath, H-Optimus-1) and a ResNet50 baseline. Using attention-based multiple instance learning, we train single-task and multi-task models to predict five molecular alterations in glioblastoma on the CPTAC cohort, validate on an independent TCGA cohort, and evaluate biological coherence of attention maps against 87 transcriptional signatures using co-registered Visium spatial transcriptomics data from 18 samples. Internally, no single encoder dominates across all tasks, and external validation inverts internal performance rankings. Attention maps show a five-fold enrichment gradient from pathways (Cohen's d=0.329) to individual genes (d=0.055), indicating that attention captures emergent multi-gene transcriptional programs rather than individual molecular events. Spatially smooth attention maps do not imply biological coherence, and different encoders attend to distinct biological compartments. Our framework provides objective, quantitative assessment of what foundation models learn from histopathology, moving the field beyond qualitative saliency map review.", "date": "2026-06-03", "categories": []}
{"id": "2606.04792", "title": "A Pathology Foundation Model for Gastric Cancer with Real-World Validation", "abstract": "Gastric cancer remains a major cause of cancer mortality, yet its histological and molecular heterogeneity complicates diagnosis and risk stratification. General-purpose pathology foundation models (PFMs) often plateau on fine-grained endpoints central to gastric cancer care, and few have undergone rigorous prospective validation or clinical reader studies. We present GRACE, a Gastric-specific foundation model for Real-world Assessment and Clinical dEcision support. GRACE was developed from multicenter gastric pathology datasets totaling 48,364 primarily HE-stained whole-slide images from 37,493 patients. When evaluated on 28 clinically relevant tasks, GRACE consistently outperformed representative pancancer PFMs, achieving a macro-AUC of 0.9188, with strong performance for precancerous lesion diagnosis (macro-AUC 0.9322), tumor histopathological assessment (macro-AUC 0.9119), molecular profiling (macro-AUC 0.8682), and prognostic prediction. Beyond benchmarking, GRACE's translational value was substantiated through a rigorous evidence chain. Under safety-gated criteria requiring 100% NPV for rule-out and 100% PPV for rule-in, GRACE streamlined review for up to 69.6% of malignancy-diagnosis cases and triaged 46.8% of MMR-IHC follow-up requests. This translational feasibility was further strengthened by a randomized crossover reader study of pathologist-AI collaboration. With GRACE assistance, diagnostic accuracy improved from 82.0% to 89.9%, yielding nearly twofold higher adjusted odds of a correct diagnosis (OR 1.987) alongside concurrent gains in sensitivity and specificity. AI assistance also reduced diagnostic time by 14.9%, elevated diagnostic confidence by 9.0%, and markedly improved inter-rater agreement. When calibrated to maintain non-inferior performance to senior pathologists, the AI-assisted workflow could triage 60.7% of atrophy and 82.7% of intestinal metaplasia cases.", "date": "2026-06-03", "categories": []}
{"id": "2603.00667", "title": "Act Like a Pathologist: Tissue-Aware Whole Slide Image Reasoning", "abstract": "Computational pathology has advanced rapidly in recent years, driven by domain-specific image encoders and growing interest in using vision-language models to answer natural-language questions about diseases. Yet, the core problem behind pathology question-answering remains unsolved, considering that a gigapixel slide contains far more information than necessary for a given question. Pathologists naturally navigate tissue and morphology complexity by scanning broadly, and zooming in selectively according to the clinical questions. Current models, in contrast, rely on uniform patch sampling or broad attention maps, often attending equally to irrelevant regions while overlooking key visual evidence. In this work, we try to bring models closer to how humans actually examine slides. We propose a question-guided, tissue-aware, and coarse-to-fine retrieval framework, HistoSelect, that consists of two key components: a group sampler that identifies question-relevant tissue regions, followed by a patch selector that retrieves the most informative patches within those regions. By selecting only the most informative patches, our method becomes significantly more efficient: reducing visual token usage by 70% on average, while improving accuracy across three pathology QA tasks. Evaluated on 356,000 question-answer pairs, our approach outperforms existing methods and produces answers grounded in interpretable, pathologist-consistent regions. Our results suggest that bringing human-like search and attention patterns into WSI reasoning is a promising direction for building practical and reliable pathology VLMs. Code is available at https://github.com/winston52/HistoSelect.", "date": "2026-02-28", "categories": []}
{"id": "2605.24253", "title": "CRISP -- Clustering-Based Redundancy-Reduced Instance Sampling for Pathology Case Representation and Retrieval", "abstract": "Digital pathology archives increasingly contain multiple whole-slide images (WSIs) per case, capturing spatially distinct tumor regions and reflecting intrinsic morphological heterogeneity. However, most existing approaches rely on a single pathologist-selected slide, thereby discarding potentially informative evidence distributed across the remaining WSIs. To date, no autonomous framework has been proposed for comprehensive multi-WSI case processing. Here, we present an unsupervised framework for case-level analysis that integrates information from all available slides within a case. Rather than relying on a single designated slide, the proposed approach constructs case-level representations by selectively distilling informative patches across WSIs. We introduce Clustering-Based Redundancy-Reduced Instance Sampling for Pathology (CRISP), a two-stage framework that first reduces redundancy within individual WSIs and subsequently applies clustering-based sampling to select a compact yet representative set of patches for the entire case. The resulting patch set captures case-level heterogeneity while avoiding exhaustive processing of gigapixel images, and directly serves as a retrieval index. Using two Mayo Clinic breast cancer datasets for diagnosis and treatment planning, we demonstrate that CRISP consistently matches or surpasses the current standard practice of combined model and pathologist slide selection for patient/case search and retrieval. By automating case-level processing and eliminating subjective WSI selection, CRISP potentially enables the exploitation of clinically relevant information distributed across multiple WSIs that is currently overlooked.", "date": "2026-05-22", "categories": []}
{"id": "2606.02602", "title": "Graph Mamba Survival Analysis Based on Topology-Aware ordering", "abstract": "In computational pathology, Whole Slide Images (WSIs) survival analysis is crucial for patient prognosis assessment, but it faces multiple technical challenges. Although the Transformer captures long-range dependencies through its self-attention mechanism, its $O(N^2)$ time complexity causes a severe computational bottleneck in large-scale WSIs graph structures. The Mamba model breaks through the Transformer's computational bottleneck with linear complexity. But, owing to Mamba's high sensitivity to the order of input data, traditional node sorting methods in Graph Mamba, such as those based on node degree or subgraph size, fail to adequately account for the topological connectivity of graph data. This inadequacy consequently restricts the performance of Mamba's sequential modeling. Moreover, its unidirectional architecture cannot leverage the bidirectional spatial structure of images. To address these challenges, this paper proposes a novel Graph Mamba survival analysis framework based on topology-aware ordering (TopoMamSurv) to adapt to the sequential sensitivity of Mamba. Our visualization experiments further confirmed that the nodes extracted through the topology-aware ordering (TAO) strategy indeed exhibit higher similarity. Furthermore, we designed a bidirectional Mamba module and integrated a Graph Convolutional Network (GCN) to achieve bidirectional spatial context modeling of images, forming a hierarchical feature learning architecture for \"local aggregation - global capture.\" This framework effectively reconciles the contradiction between long-range dependency modeling, computational efficiency, and spatial structure utilization in WSIs analysis through its systematic design of TAO, bidirectional semantic modeling, and hierarchical feature fusion. This framework has been validated for its comprehensive performance advantage on five TCGA datasets.", "date": "2026-05-23", "categories": []}
{"id": "2606.03644", "title": "Spatial Transcriptomics-Guided Alignment Enhances Molecular Profiling in Pathology Foundation Model", "abstract": "Comprehensive molecular profiling is essential for modern precision oncology but remains hindered by prohibitive costs, specimen exhaustion, and protracted turnaround times. While pathology foundation models (PFMs) have demonstrated potential for inferring molecular phenotypes from routine hematoxylin and eosin (H&E) whole-slide images (WSIs), current architectures primarily rely on vision-centric self-supervised learning or vision-language alignment, lacking the spatially resolved molecular supervision required to connect subtle morphological features with underlying genomic alterations. Spatial transcriptomics (ST) emerges as a transformative technology that enables transcriptomic quantification within intact tissue sections, thereby preserving the precise spatial link between histology and molecular profiles. In this study, we present a Spatial Transcriptomics-guided Alignment framework for Molecular Profiling (STAMP), which endows PFMs with intrinsic molecular awareness. To support this paradigm, we curated HumanST-1k, a human ST dataset spanning diverse anatomical organs and sequencing platforms. This atlas yields 1.8 million pairs of H&E patches and corresponding transcriptomic profiles, providing a corpus that links histological structures with their molecular states. To mitigate the technical noise inherent to raw transcriptomics, STAMP applies a pathway-informed alignment strategy that aggregates transcriptomic data into biologically functional pathways, which are subsequently integrated into PFMs via parameter-efficient fine-tuning. This alignment enriches the representation space of PFMs and unlocks their capacity to resolve sub-visual molecular signatures. The clinical utility of these augmented representations was validated through a multi-tier evaluation framework.", "date": "2026-05-29", "categories": []}
{"id": "2606.00092", "title": "Aligning Cellular Sheaves with Classifier Attention for Interpretable Weakly-Supervised Pathology Localization", "abstract": "Weakly-supervised classification of whole-slide images with attention-based multiple instance learning (ABMIL) on top of foundation features now reaches near-saturation on Camelyon16 slide-level performance, but the corresponding attention maps are an imperfect localization signal: in clinical interpretation, a model that classifies correctly without firing on the actual lesion is hard to trust. We address this gap with cellular sheaves, which equip each vertex and edge of a graph with a finite-dimensional vector space and consistent linear maps between them, providing a principled way to detect local disagreement on graph-structured data. We apply cellular sheaves to weakly-supervised tumour localization on whole-slide images, combining a sheaf disagreement field with ABMIL. The natural training objective, encouraging consistency between similar features, produces a disagreement field that tracks tissue-level texture rather than diagnostic content. We propose attention-conditional consistency, which uses the classifier's attention to define which neighbouring patches should agree. Joint training of the classifier and the sheaf under this objective produces a disagreement field with patch-level AUC 0.940 on Camelyon16 and raises the attention head from its ABMIL-alone level of 0.717 to 0.953. Two-stage ablation with the classifier frozen at its ABMIL values reaches only 0.727 on the disagreement field and leaves attention at 0.717, confirming that the gain comes from the projector co-adapting under both objectives, not from the loss change in isolation. The trained model transfers without retraining to annotated slides from Camelyon17, maintaining Delta AUC 0.932 +/- 0.083 and attention AUC 0.955 +/- 0.099. The result is an attention map and a sheaf-disagreement map that fire on the same diagnostic regions, giving clinicians two complementary explanations for each slide-level prediction.", "date": "2026-05-24", "categories": []}
{"id": "2606.00472", "title": "CodeCytos: AI-assisted spatial molecular imaging analysis via code-augmented agent action space", "abstract": "Conventional tissue image analysis software provides foundational capabilities for cellular analysis, including segmentation, basic morphological feature extraction, and spatial organization analysis. However, these tools often require manual intervention and are not well integrated with code-driven automation, limiting efficiency and scalability for complex spatial tissue studies. In addition, they offer limited flexibility for custom analyses, as they typically support only a fixed set of pre-implemented spatial cellular features. To address these limitations, we propose CodeCytos, a coding-based reasoning agent framework that enables dynamic, programmable interaction with spatial molecular imaging data to improve automation and customization. CodeCytos is designed to streamline the exploration of custom spatial cellular features and adapt to diverse research needs. We demonstrate its utility through case studies on four expert-curated datasets from distinct tissue types: frontal cortex, non-small-cell lung cancer, pancreas, and tonsil. We evaluate CodeCytos under a realistic minimal prompt setting, where bioscientists pose simple questions without task-specific instructions or contextual information about spatial cellular analysis, and benchmark multiple LLM backbones with strong coding capabilities. We further show that incorporating tailored, domain-agnostic few-shot in-context coding-reasoning examples (randomly sampled demonstrations outside the spatial analysis domain) can substantially improve performance without requiring costly, expert-crafted in-domain demonstrations. Overall, CodeCytos outperforms baseline approaches, highlighting the potential of code-action agents to assist with custom feature exploration in spatial molecular imaging and to accelerate biomarker discovery.", "date": "2026-05-30", "categories": []}
{"id": "2606.01543", "title": "PathAR: Structure-First Autoregressive Synthesis of Multimodal Pathology Images", "abstract": "Data scarcity in multimodal pathology motivates unified generative models that synthesize modality-specific appearance while preserving anatomically coherent structure. Although modalities differ in appearance statistics, morphological structures such as cellular topology and tissue boundaries are largely preserved across acquisition protocols. However, existing methods often model these factors within a homogeneous token stream, implicitly coupling structure with appearance and weakening structural controllability under modality shifts. To address this, we propose pathology Autorgressive modeling (PathAR), a structure-first autoregressive synthesis framework that explicitly factorizes structure and appearance for modality-label-conditioned pathology generation.PathAR employs a dual vector quantization (Dual-VQ) tokenizer to decompose samples into mask-grounded structure and appearance tokens, and an interleaved autoregressive (IAR) transformer with asymmetric attention visibility to enforce structure-to-appearance dependence. PathAR stabilizes morphology under heterogeneous modality-specific appearances and enables spatially aligned image--mask pair generation. Extensive experiments show that PathAR improves structural consistency and modality fidelity over baselines, maintains sample diversity, supports downstream segmentation in data-scarce regimes, and demonstrates extensibility to finer-grained intra-modality organ-label variation.", "date": "2026-06-01", "categories": []}
{"id": "2606.01871", "title": "Deep Learning for Generating Computational PIN-4 Immunohistochemistry Staining from Prostate Biopsy H&E Images", "abstract": "Immunohistochemistry (IHC)is frequently used to resolve diagnostically ambiguous prostate cancer biopsy findings on hematoxylin and eosin (H&E)-stained tissue. However, PIN-4 IHC staining is typically performed on adjacent tissue sections, limiting direct spatial comparison between the H&E morphology and the corresponding immunophenotypic signal. A paired, registered H&E/PIN-4 dataset was constructed from routine clinical prostate biopsy whole-slide images (WSIs), and a conditional generative adversarial network (cGAN) was trained to synthesize PIN-4 staining patterns directly from native H&E image patches. The final dataset comprised 172 paired WSIs from 93 patients and 27,298 registered 1024x1024 patch pairs, spanning adenocarcinoma-positive and benign cases with representation across age, race, and ethnicity groups. The model was evaluated on a held-out test set of 1,814 patch pairs from 17 WSIs, achieving a mean peak signal-to-noise ratio (PSNR) of 21.88 dB, structural similarity index measure (SSIM) of 0.667, Pearson correlation coefficient (PCC) of 0.684, and learned perceptual image patch similarity (LPIPS) of 0.417. Qualitative review by a board-certified pathologist showed that generated images captured diagnostically relevant PIN-4 staining patterns, including AMACR/racemase expression and basal-cell-associated staining, while preserving spatial correspondence with the source H&E morphology. Accuracy of synthesis varied across morphologically complex regions, including high-grade carcinoma and intraductal carcinoma. These results support the feasibility of supervised PIN-4 synthesis from routinely acquired brightfield H&E prostate biopsy images. The approach enables direct interpretation of predicted PIN-4 marker patterns in the context of the source prostate H&E architecture, addressing a current spatial limitation of conventional adjacent-section IHC.", "date": "2026-06-01", "categories": []}
{"id": "2606.02424", "title": "GC-MoE: Genomics-Guided Cell-Type-Specific Mixture of Experts for Histology-Based Single-Cell Spatial Transcriptomics", "abstract": "Histology-based single-cell spatial transcriptomics (ST) estimation aims to predict gene expression for individual cells from histopathological images and cell locations, reducing the need for costly single-cell ST measurements. Unlike existing histology-to-ST methods that mainly predict spot-level profiles for local regions containing multiple cells, this task requires modeling cell-to-cell expression variability, which is strongly structured by cell type. We propose Genomics-Guided Cell-Type-Specific Mixture-of-Experts (GC-MoE), which estimates cell-type probabilities with a routing network and softly combines cell-type-specific experts for gene expression prediction. To further encode cell-type-dependent gene programs, we introduce the Cell-Type-Specific Co-Expression-Aware Predictor (CAP), together with a lightweight Cell-to-Cell Interaction Attention (C2CA) module for neighboring-cell context. Experiments and ablations on public single-cell ST datasets show consistent improvements over existing single-cell and adapted spot-level baselines.", "date": "2026-06-01", "categories": []}
{"id": "2605.29852", "title": "Parameter-Efficient Subspace Decoupling ViT for Mitigating Multi-Task Negative Transfer in Histological Scoring", "abstract": "Histological scoring is essential for diagnosing Non-Alcoholic Fatty Liver Disease (NAFLD), yet its automation remains challenging due to the high annotation cost and negative transfer among the strongly correlated NAFLD Activity Score (NAS) indicators in multi-task learning. To address this issue, we propose a subspace-decoupled multi-task Vision Transformer (ViT) that integrates lightweight task-specific Adapters with orthogonality-based constraints. This design constructs independent feature subspaces for steatosis, ballooning, and inflammation, effectively reducing task interference while retaining shared representations. We further construct a curated multi-task mouse NAFLD histology dataset with expert annotations for all NAS components. Experimental results demonstrate that the proposed method improves multi-task stability and generalization with substantially reduced computational cost compared to training separate single-task models. The code and the curated dataset have been prepared and will be made publicly available upon acceptance to support reproducibility.", "date": "2026-05-28", "categories": []}
{"id": "2605.24399", "title": "ConceptM$^3$oE: Concept-Guided Multimodal Mixture of Experts for Interpretable Computational Pathology", "abstract": "Healthcare models are transitioning from unimodal prediction toward multimodal reasoning over heterogeneous diagnostic inputs. In computational pathology, for complex tumor subtypes where morphology alone can be challenging to distinguish, pathology reports and molecular measurements may provide additional diagnostic evidence alongside whole-slide images, yet existing models often fail to clarify how diverse signals assemble into recognizable diagnostic concepts. We propose ConceptM$^3$oE (Concept Multimodal MoE), which embeds concept formation directly within interaction-aware mixture-of-experts (MoE) pathways. The architecture decomposes evidence into modality-specific, redundant, and synergistic experts, which are then projected into structured concept bottlenecks mapping latent features to a hierarchy of morphology and biomarker concepts. To prevent the information loss typical of interpretable bottlenecks, we utilize residual pathways within each expert to allow task-relevant signals to flow both through the concepts and directly to the final task prediction, so that high performance is maintained alongside interpretability. Across an institutional pediatric brain tumor cohort and a public glioma cohort, the framework delivers competitive performance to unconstrained models while producing reasoning traces validated by an independent neuropathologist. In data-limited regimes, ConceptM$^3$oE improves limited-data performance, increasing macro-F1 from 56.41% to 66.70% at small training sizes compared to non-concept-informed baselines, while also showing faster training convergence consistent with the regularizing effect of concept learning. This work offers a scalable path toward high-performance medical AI that is inherently verifiable and better aligned with the complex decision-making of clinical practice.", "date": "2026-05-23", "categories": []}
{"id": "2309.07778", "title": "Virchow: A Million-Slide Digital Pathology Foundation Model", "abstract": "The use of artificial intelligence to enable precision medicine and decision support systems through the analysis of pathology images has the potential to revolutionize the diagnosis and treatment of cancer. Such applications will depend on models' abilities to capture the diverse patterns observed in pathology images. To address this challenge, we present Virchow, a foundation model for computational pathology. Using self-supervised learning empowered by the DINOv2 algorithm, Virchow is a vision transformer model with 632 million parameters trained on 1.5 million hematoxylin and eosin stained whole slide images from diverse tissue and specimen types, which is orders of magnitude more data than previous works. The Virchow model enables the development of a pan-cancer detection system with 0.949 overall specimen-level AUC across 17 different cancer types, while also achieving 0.937 AUC on 7 rare cancer types. The Virchow model sets the state-of-the-art on the internal and external image tile level benchmarks and slide level biomarker prediction tasks. The gains in performance highlight the importance of training on massive pathology image datasets, suggesting scaling up the data and network architecture can improve the accuracy for many high-impact computational pathology applications where limited amounts of training data are available.", "date": "2023-09-14", "categories": []}
{"id": "2605.19491", "title": "Thinking in Scales: Accelerating Gigapixel Pathology Image Analysis via Adaptive Continuous Reasoning", "abstract": "Traditional whole slide image (WSI) analysis methods typically rely on the multiple instance learning (MIL) paradigm, which extracts patch-level features at high magnification and aggregates them for slide-level prediction. However, such exhaustive patch-level processing is computationally expensive, severely limiting the efficiency and scalability of WSI analysis. To address this challenge, we propose PathCTM (a Pathology-oriented Continuous Thought Model) that enables token-efficient scale-space continuous reasoning for gigapixel WSIs. PathCTM formulates diagnostic inference as a dynamic sequential information pursuit. It progressively transitions from low-magnification global to high-magnification local inspection, and adaptively terminates inference when sufficient evidence is gathered to effectively bound decision uncertainty. Specifically, it uses conditional computation for dynamic scale switching with attention-guided region pruning, coupled with confidence-aware early stopping. Extensive experiments demonstrate that, compared with standard MIL-based methods, PathCTM reduces the number of required image patches by 95.95% and shortens inference time by approximately 95.62%, while maintaining AUC without degradation. Code is available at https://github.com/JSGe-AI/PathCTM.", "date": "2026-05-19", "categories": []}
{"id": "2605.25764", "title": "Benchmarking Pathology Foundation Models for Spatial Domain Understanding", "abstract": "Pathology foundation models (PFMs) have emerged as a core approach for learning transferable representations from whole slide images (WSIs), and they are typically benchmarked through downstream clinical endpoints. While such task level evaluations are indispensable, they offer limited insight into what the representations themselves encode, particularly whether PFM embeddings can distinguish meaningful tissue regions and capture their spatial relationships. We present SpaPath-Bench, a representation level benchmark designed to diagnose spatial representation capability in PFMs. SpaPath-Bench formulates spatial domain identification (SDI) on paired whole slide image and spatial transcriptomics (ST) data as a diagnostic task. It curates 42 public paired WSI and ST slides, enables large scale evaluation across 19 encoders and seven SDI methods, and measures partition quality using three complementary criteria: unsupervised spatial coherence, transcriptomics referenced agreement, and expert referenced agreement. Across 83K runs, SpaPath-Bench reveals that different pretraining paradigms capture distinct aspects of tissue spatial architecture, and it provides practical guidance for building the next generation of spatially aware computational pathology models. Code and data pipelines are publicly available at https://bokai-zhao.github.io/SpaPath-benchboard/.", "date": "2026-05-25", "categories": []}
{"id": "2605.25878", "title": "A Clinically Validated Foundation Model for Comprehensive Lung Pathology Interpretation", "abstract": "Pathological assessment guides lung cancer diagnosis, treatment selection, and prognostic evaluation, yet current CPath approaches rely on task-specific models for isolated objectives. Although pan-cancer foundation models offer versatility, they lack subspecialty-level depth and have not been evaluated across clinical workflows or prospectively validated in real-world settings. We introduce PulmoFoundation, a multi-center, prospectively validated, randomized controlled trial (RCT)-evaluated foundation model for comprehensive lung pathology assessment across pre-operative, intra-operative, and post-operative care. Built upon Virchow2 via subspecialty-specific pretraining using ~40,000 diagnostic H&E-stained whole-slide images (WSIs), PulmoFoundation was systematically evaluated on ~26,000 WSIs across 32 clinically relevant tasks. In addition to accurately predicting molecular markers and patient survival, our model achieves clinical-grade performance in core diagnostic tasks across biopsy, frozen section, and surgical resection slides. In a registered prospective study of 1,357 patients across 11 diagnostic tasks, our model achieved an average AUC of 92.3%. Using pre-specified triage thresholds, PulmoFoundation could reduce additional second-review burden for 68.8% of biopsies and 83.0% of frozen sections, and defer 44.5% of IHC stain orders, with PPVs of 1.0, 0.991, and 0.966. Beyond prospective validation, we conducted a crossover RCT with eight pathologists, in which AI assistance improved diagnostic accuracy across 4,928 case-reader pairs (91.7% w/ AI vs. 83.8% w/o AI). AI assistance also reduced median diagnostic time by 19.6%, increased diagnostic confidence by 8.7%, and improved inter-rater agreement from moderate (kappa = 0.56) to substantial (kappa = 0.76). Together, these evaluations support PulmoFoundation as a clinically validated decision-support system for lung pathology.", "date": "2026-05-25", "categories": []}
{"id": "2602.05126", "title": "CLEAR-HPV: Interpretable concept discovery for human-papillomavirus-associated morphology in whole-slide histology", "abstract": "Human papillomavirus (HPV) status is a critical determinant of prognosis and treatment response in head and neck and cervical cancers. Although attention-based multiple instance learning (MIL) achieves strong slide-level prediction for HPV-related whole-slide histopathology, it provides limited morphologic interpretability. To address this limitation, we introduce Concept-Level Explainable Attention-guided Representation for HPV (CLEAR-HPV), a framework that restructures the MIL latent space using attention to enable concept discovery without requiring concept labels during training. Operating in an attention-weighted latent space, CLEAR-HPV automatically discovers keratinizing, basaloid, and stromal morphologic concepts, generates spatial concept maps, and represents each slide using a compact concept-fraction vector. CLEAR-HPV's concept-fraction vectors preserve the predictive information of the original MIL embeddings while reducing the high-dimensional feature space (e.g., 1536 dimensions) to only 10 interpretable concepts. CLEAR-HPV generalizes consistently across TCGA-HNSCC, TCGA-CESC, and CPTAC-HNSCC, providing compact, concept-level interpretability through a general, backbone-agnostic framework for attention-based MIL models of whole-slide histopathology.", "date": "2026-02-04", "categories": []}
{"id": "2605.22000", "title": "Virtual 3D H&E Staining from Phase-contrast Back-illumination Interference Tomography", "abstract": "Three-dimensional (3D) histopathology of unprocessed tissues has the potential to transform disease management by enabling volumetric characterization of tissue microarchitecture and in-vivo assessment. Back-illumination Interference Tomography (BIT) is a new phase microscopy technology that provides rapid, non-destructive volumetric imaging of unprocessed tissues. However, translating BIT volumes into clinically interpretable H&E images remains challenging, particularly due to shift-variant contrast and the absence of quantitative validation benchmarks. We introduce HistoBIT3D, the first voxel-wise paired BIT and fluorescence-labeled nuclei dataset, enabling quantitative evaluation of structural preservation in unsupervised virtual staining against ground-truth nuclear distributions. Using this dataset, we present a novel virtual staining framework that translates BIT volumes with shift-variant contrast into realistic H&E volumes by leveraging bidirectional multiscale content consistency and cross-domain style reuse to enhance structural fidelity and perceptual realism. Our method achieves state-of-the-art realism metrics while significantly improving 3D nuclei segmentation accuracy and boundary preservation under zero-shot Cellpose evaluation. Together, these contributions establish a quantitatively validated, structurally faithful, and scalable pipeline for 3D virtual H&E staining, advancing the paradigm of slide-free, volumetric computational histopathology. Our data and code are available at: https://github.com/aasong113/HistoBIT3D_VirtualStaining.", "date": "2026-05-21", "categories": []}
{"id": "2604.21060", "title": "Clinically-Informed Modeling for Pediatric Brain Tumor Classification from Whole-Slide Histopathology Images", "abstract": "Accurate diagnosis of pediatric brain tumors, starting with histopathology, presents unique challenges for deep learning, including severe data scarcity, class imbalance, and fine-grained morphologic overlap across diagnostically distinct subtypes. While pathology foundation models have advanced patch-level representation learning, their effective adaptation to weakly supervised pediatric brain tumor classification under limited data remains underexplored. In this work, we introduce an expert-guided contrastive fine-tuning framework for pediatric brain tumor diagnosis from whole-slide images (WSI). Our approach integrates contrastive learning into slide-level multiple instance learning (MIL) to explicitly regularize the geometry of slide-level representations during downstream fine-tuning. We propose both a general supervised contrastive setting and an expert-guided variant that incorporates clinically informed hard negatives targeting diagnostically confusable subtypes. Through comprehensive experiments on pediatric brain tumor WSI classification under realistic low-sample and class-imbalanced conditions, we demonstrate that contrastive fine-tuning yields measurable improvements in fine-grained diagnostic distinctions. Our experimental analyses reveal complementary strengths across different contrastive strategies, with expert-guided hard negatives promoting more compact intra-class representations and improved inter-class separation. This work highlights the importance of explicitly shaping slide-level representations for robust fine-grained classification in data-scarce pediatric pathology settings.", "date": "2026-04-22", "categories": []}
{"id": "2507.05843", "title": "USIGAN: Unbalanced Self-Information Feature Transport for Weakly Paired Image IHC Virtual Staining", "abstract": "Immunohistochemical (IHC) virtual staining is a task that generates virtual IHC images from H\\&E images while maintaining pathological semantic consistency with adjacent slices. This task aims to achieve cross-domain mapping between morphological structures and staining patterns through generative models, providing an efficient and cost-effective solution for pathological analysis. However, under weakly paired conditions, spatial heterogeneity between adjacent slices presents significant challenges. This can lead to inaccurate one-to-many mappings and generate results that are inconsistent with the pathological semantics of adjacent slices. To address this issue, we propose a novel unbalanced self-information feature transport for IHC virtual staining, named USIGAN, which extracts global morphological semantics without relying on positional correspondence.By removing weakly paired terms in the joint marginal distribution, we effectively mitigate the impact of weak pairing on joint distributions, thereby significantly improving the content consistency and pathological semantic consistency of the generated results. Moreover, we design the Unbalanced Optimal Transport Consistency (UOT-CTM) mechanism and the Pathology Self-Correspondence (PC-SCM) mechanism to construct correlation matrices between H\\&E and generated IHC in image-level and real IHC and generated IHC image sets in intra-group level.. Experiments conducted on two publicly available datasets demonstrate that our method achieves superior performance across multiple clinically significant metrics, such as IoD and Pearson-R correlation, demonstrating better clinical relevance.", "date": "2025-07-08", "categories": []}
{"id": "2510.11344", "title": "MMAP: A Multi-Magnification and Prototype-Aware Architecture for Predicting Spatial Gene Expression", "abstract": "Spatial Transcriptomics (ST) enables the measurement of gene expression while preserving spatial information, offering critical insights into tissue architecture and disease pathology. Recent developments have explored the use of hematoxylin and eosin (H&E)-stained whole-slide images (WSIs) to predict transcriptome-wide gene expression profiles through deep neural networks. This task is commonly framed as a regression problem, where each input corresponds to a localized image patch extracted from the WSI. However, predicting spatial gene expression from histological images remains a challenging problem due to the significant modality gap between visual features and molecular signals. Recent studies have attempted to incorporate both local and global information into predictive models. Nevertheless, existing methods still suffer from two key limitations: (1) insufficient granularity in local feature extraction, and (2) inadequate coverage of global spatial context. In this work, we propose a novel framework, MMAP (Multi-MAgnification and Prototype-enhanced architecture), that addresses both challenges simultaneously. To enhance local feature granularity, MMAP leverages multi-magnification patch representations that capture fine-grained histological details. To improve global contextual understanding, it learns a set of latent prototype embeddings that serve as compact representations of slide-level information. Extensive experimental results demonstrate that MMAP consistently outperforms all existing state-of-the-art methods across multiple evaluation metrics, including Mean Absolute Error (MAE), Mean Squared Error (MSE), and Pearson Correlation Coefficient (PCC).", "date": "2025-10-13", "categories": []}
{"id": "2604.18225", "title": "Is SAM3 ready for pathology segmentation?", "abstract": "Is Segment Anything Model 3 (SAM3) capable in segmenting Any Pathology Images? Digital pathology segmentation spans tissue-level and nuclei-level scales, where traditional methods often suffer from high annotation costs and poor generalization. SAM3 introduces Promptable Concept Segmentation, offering a potential automated interface via text prompts. With this work, we propose a systematic evaluation protocol to explore the capability space of SAM3 in a structured manner. Specifically, we evaluate SAM3 under different supervision settings including zero-shot, few-shot, and supervised with varying prompting strategies. Our extensive evaluation on pathological datasets including NuInsSeg, PanNuke and GlaS, reveals that: (1) text-only prompts poorly activate nuclear concepts; (2) performance is highly sensitive to visual prompt types and budgets; (3) few-shot learning offers gains, but SAM3 lacks robustness against visual prompt noise; and (4) a significant gap persists between prompt-based usage and task-trained adapter-based reference. Our study delineates SAM3's boundaries in pathology image segmentation and provides practical guidance on the necessity of pathology domain adaptation.", "date": "2026-04-20", "categories": []}
{"id": "2605.00578", "title": "Federated Distillation for Whole Slide Image via Gaussian-Mixture Feature Alignment and Curriculum Integration", "abstract": "Federated learning (FL) offers a promising framework for collaborative digital pathology by enabling model training across institutions. However, real-world deployments face heterogeneity arising from diverse multiple instance learning (MIL) architectures and heterogeneous feature extractors across institutions. We propose FedHD, a novel FL framework that performs local Gaussian-mixture feature alignment tailored for WSI analysis. Instead of exchanging model parameters, each client independently distills semantically rich synthetic feature representations aligned with the distribution of real WSIs. To preserve diagnostic diversity, FedHD adopts a one-to-one distillation strategy, generating a synthetic counterpart for each real slide to avoid over-compression. During federation, a curriculum-based integration strategy progressively incorporates cross-site synthetic features into local training once performance plateaus. Furthermore, an optional interpretation module reconstructs pseudo-patches from synthetic embeddings, enhancing transparency. FedHD is architecture-agnostic, privacy-preserving, and supports personalized yet collaborative training across diverse institutions. Experiments on TCGA-IDH, CAMELYON16, and CAMELYON17 show that FedHD consistently outperforms state-of-the-art federated and distillation baselines.", "date": "2026-05-01", "categories": []}
{"id": "2511.19953", "title": "Supervise Less, See More: Training-free Nuclear Instance Segmentation with Prototype-Guided Prompting", "abstract": "Accurate nuclear instance segmentation is a pivotal task in computational pathology, supporting data-driven clinical insights and facilitating downstream translational applications. While large vision foundation models have shown promise for zero-shot biomedical segmentation, most existing approaches still depend on dense supervision and computationally expensive fine-tuning. Consequently, training-free methods present a compelling research direction, yet remain largely unexplored. In this work, we introduce SPROUT, a fully training- and annotation-free prompting framework for nuclear instance segmentation. SPROUT leverages histology-informed priors to construct slide-specific reference prototypes that mitigate domain gaps. These prototypes progressively guide feature alignment through a partial optimal transport scheme. The resulting foreground and background features are transformed into positive and negative point prompts, enabling the Segment Anything Model (SAM) to produce precise nuclear delineations without any parameter updates. Extensive experiments across multiple histopathology benchmarks demonstrate that SPROUT achieves competitive performance without supervision or retraining, establishing a novel paradigm for scalable, training-free nuclear instance segmentation in pathology.", "date": "2025-11-25", "categories": []}
{"id": "2605.16392", "title": "Bridging the Modality Bottleneck in Pathology MIL through Virtual Molecular Staining", "abstract": "Multiple instance learning (MIL) is the dominant framework for whole-slide image analysis in computational pathology, typically combining a frozen patch encoder, a projection layer, and a slide-level aggregator. While encoders and aggregators have been extensively studied, the projection layer remains a largely morphology-only bottleneck. This limits endpoints such as biomarker status and survival, which are governed by a molecular state that is not fully captured by H&E morphology. We introduce Molecularly Informed Staining Transform (MIST), a plug-in replacement for the MIL projection layer that uses paired spatial transcriptomics only during training to construct virtual molecular stains. MIST clusters gene expression profiles into cross-modal prototypes, anchors them in the frozen foundation model feature space, and uses them to reorganize H&E patch features along molecularly guided axes. It requires no transcriptomics at inference and can be inserted before standard MIL aggregators. We evaluate MIST across 23 downstream tasks and 8 MIL aggregators. MIST improves 240 of 256 configurations over the standard projection layer, with an average gain of +3.5%, observed consistently across endpoint types: +5.2% on survival prediction, +3.3% on tissue subtyping, and +2.6% on biomarker prediction. Ablations confirm that gene-derived prototypes are the primary source of the gains, while spatial, biological, and pathological analyses show that cross-modal prototype affinities capture spatially coherent molecular programs from H&E alone.", "date": "2026-05-12", "categories": []}
{"id": "2605.16444", "title": "Diffusion Attention Expert Model for Predicting and Semi-automatic Localizing STAS in Lung Cancer Histopathological Images", "abstract": "Accurate intraoperative and postoperative diagnosis of spread through air spaces (STAS) is essential for guiding surgical decisions and postoperative management in lung cancer. However, histopathological assessment is labor-intensive and is prone to missed or incorrect diagnoses. We propose a Diffusion Attention Expert Model (DAEM) to detect STAS in frozen sections (FSs) and paraffin sections (PSs). Its diffusion attention expert module leverages full attention aggregation to learn multi-scale features from histopathological images, while a dual-branch architecture strengthens multi-scale feature representation. On an internal dataset, DAEM achieves AUCs of 0.8946 for FSs and 0.9112 for PSs. Validation on external multi-center datasets from eight institutions demonstrates strong generalizability and interpretability. Using tumor microenvironment (TME) features in PSs, we further enable semi-automatic measurement of STAS location and its distance from the primary tumor. Several quantitative TME metrics are identified as potential biomarkers for STAS, including micropapillary-type STAS. Overall, DAEM offers a clinically actionable framework for STAS assessment by enabling accurate and interpretable detection on FSs and PSs, supporting postoperative risk stratification through quantitative TME-based analysis.", "date": "2026-05-15", "categories": []}
{"id": "2602.04585", "title": "ImmuVis: Hyperconvolutional Foundation Model for Imaging Mass Cytometry", "abstract": "We present ImmuVis, a family of efficient foundation models for imaging mass cytometry (IMC), a high-throughput multiplex imaging technology that handles molecular marker measurements as image channels and enables large-scale spatial tissue profiling. Unlike natural images, multiplex imaging lacks a fixed channel space, as real-world marker sets vary across studies, violating a core assumption of standard vision backbones. To address this, ImmuVis introduces marker-adaptive hyperconvolutions that generate convolutional kernels from learned marker embeddings, enabling a single model to operate on arbitrary measured marker subsets without retraining. We pretrain ImmuVis on the largest dataset to date, IMC17M (28 cohorts, 24,405 images, 265 markers, over 17M patches), using self-supervised masked reconstruction. ImmuVis outperforms state-of-the-art baselines and ablations in virtual staining and downstream classification tasks at substantially lower compute cost than transformer-based alternatives, and is the sole model that provides calibrated uncertainty via a heteroscedastic likelihood objective. These results position ImmuVis as a practical framework for real-world IMC modeling.", "date": "2026-02-04", "categories": []}
{"id": "2605.13889", "title": "Physics-Grounded Adversarial Stain Augmentation with Calibrated Coverage Guarantees", "abstract": "Stain variation across hospitals degrades histopathology models at deployment. Existing augmentation methods perturb color spaces with arbitrary hyperparameters, lacking both a principled budget and coverage guarantees for unseen centers. We propose \\textbf{C}alibrated \\textbf{A}dversarial \\textbf{S}tain \\textbf{A}ugmentation (\\textbf{CASA}), which performs adversarial augmentation in the Macenko stain parameter space with a budget calibrated from multi-center statistics via the DKW inequality. On Camelyon17-WILDS (5 seeds), CASA achieves $93.9\\% \\pm 1.6\\%$ slide-level accuracy -- outperforming HED-strong ($88.4\\% \\pm 7.3\\%$), RandStainNA ($85.2\\% \\pm 6.7\\%$), and ERM ($63.9\\% \\pm 11.3\\%$) -- with the highest worst-group accuracy ($84.9\\% \\pm 0.9\\%$) among all 10 compared methods.", "date": "2026-05-12", "categories": []}
{"id": "2605.14251", "title": "Generative Deep Learning for Computational Destaining and Restaining of Unregistered Digital Pathology Images", "abstract": "Conditional generative adversarial networks (cGANs) have enabled high-fidelity computational staining and destaining of hematoxylin and eosin (H&E) in digital pathology whole-slide images (WSI). However, their ability to generalize to out-of-distribution WSI across institutions without retraining remains insufficiently characterized. Previously developed cGAN models trained on 102 registered prostate core biopsy WSIs from Brigham and Women's Hospital were evaluated on 82 spatially unregistered WSIs acquired at Stanford University. To mitigate domain shift without retraining, a preprocessing pipeline consisting of histogram-based stain normalization for H&E-stained WSIs and channel-wise intensity calibration for unstained WSIs was developed. Because image registration was intentionally omitted for real-world deployment conditions, the reported quantitative results are conservative lower bounds reflecting both model performance and limited spatial alignment. Under these conditions, virtual destaining achieved a Pearson correlation coefficient (PCC) of 0.854, structural similarity index measure (SSIM) of 0.699, and peak signal-to-noise ratio (PSNR) of 18.41 dB. H&E restaining from computationally destained outputs outperformed direct staining from ground-truth unstained inputs across all metrics (PCC: 0.798 vs. 0.715; SSIM: 0.756 vs. 0.718; PSNR: 20.08 vs. 18.51 dB), suggesting that preprocessing quality may be more limiting than model capacity. Qualitative pathological review indicated preservation of benign glandular structures while showing that malignant glands were often rendered with vessel-like morphologies. These findings support the feasibility of applying cGAN-based computational H&E staining and destaining generative models to external WSI datasets using preprocessing-based adaptation alone while defining specific morphological targets for future domain adaptation.", "date": "2026-05-14", "categories": []}
{"id": "2605.14717", "title": "Towards Label-Free Single-Cell Phenotyping Using Multi-Task Learning", "abstract": "Label-free single-cell imaging offers a scalable, non-invasive alternative to fluorescence-based cytometry, yet inferring molecular phenotypes directly from bright-field morphology remains challenging. We present a unified Deep Learning (DL) framework that jointly performs White Blood Cell (WBC) classification and continuous protein-expression regression from label-free Differential Phase Contrast (DPC) images. Our model employs a Hybrid architecture that fuses convolutional fine-grained texture features with transformer-based global representations through a learnable cross-branch gating module, enabling robust morpho-molecular inference from DPC images. To support downstream interpretability, we further incorporate a Large Language Model (LLM) that generates concise, biologically grounded summaries of the predicted cell states. Experiments on the Berkeley Single Cell Computational Microscopy (BSCCM) and Blood Cells Image benchmarks demonstrate strong performance, achieving a 91.3% WBC classification accuracy and a 0.72 Pearson correlation for CD16 expression regression on BSCCM. These results underscore the promise of label-free single-cell imaging for cost-effective hematological profiling, enabling simultaneous phenotype identification and quantitative biomarker estimation without fluorescent staining. The source code is available at https://github.com/saqibnaziir/Single-Cell-Phenotyping.", "date": "2026-05-14", "categories": []}
{"id": "2605.12575", "title": "Are Compact Rationales Free? Measuring Tile Selection Headroom in Frozen WSI-MIL", "abstract": "Whole-slide image (WSI) multiple instance learning (MIL) classifiers can achieve strong slide-level AUC while leaving the full-bag prediction opaque. Attention scores are widely reused as post-hoc explanations, but high attention can reflect aggregation preference rather than a compact, model-sufficient rationale. We study post-hoc rationale highlighting for frozen WSI-MIL: given a trained classifier, can its slide-level prediction be recovered from a compact, output-consistent tile subset without retraining the backbone? We instantiate this with Finding Optimal Contextual Instances (FOCI), a lightweight rationale-readout layer over a frozen MIL backbone. FOCI is trained with model-output sufficiency and exclusion objectives over keep/drop tile subsets, evaluated with an insertion-style Sequential Reveal Protocol (SRP) adapted to WSI-MIL, and summarized by the Selection Headroom Index (SHI). Across three WSI benchmarks and seven MIL backbones, FOCI reveals that compact rationales are selection-headroom dependent: transformer and multi-branch attention aggregators can admit compact rationales, near-minimal attention-pooling baselines enter a selection-saturation regime, and hard-selection backbones can conflict with an external readout. For TransMIL, relative to its documented CLS-proxy ranking, FOCI reduces the Minimum Sufficient K (MSK) tile count by 32-56% across benchmarks, while ACMIL+FOCI attains the highest mean SHI (+0.465). Deletion-based perturbation and selected-only downstream evaluation provide complementary checks. These results position FOCI as a model-level interpretability and audit layer: selected tiles are not claims of clinical or pathologist-level diagnostic sufficiency, but candidate rationales that offer a compact, reviewable view of when a frozen MIL prediction can be localized to a small output-consistent subset.", "date": "2026-05-12", "categories": []}
{"id": "2602.02282", "title": "MoLF: Mixture-of-Latent-Flow for Pan-Cancer Spatial Gene Expression Prediction from Histology", "abstract": "Inferring spatial transcriptomics (ST) from histology enables scalable histogenomic profiling, yet current methods are largely restricted to single-tissue models. This fragmentation fails to leverage biological principles shared across cancer types and hinders application to data-scarce scenarios. While pan-cancer training offers a solution, the resulting heterogeneity challenges monolithic architectures. To bridge this gap, we introduce MoLF (Mixture-of-Latent-Flow), a generative model for pan-cancer histogenomic prediction. MoLF leverages a conditional Flow Matching objective to map noise to the gene latent manifold, parameterized by a Mixture-of-Experts (MoE) velocity field. By dynamically routing inputs to specialized sub-networks, this architecture effectively decouples the optimization of diverse tissue patterns. Our experiments demonstrate that MoLF establishes a new state-of-the-art, consistently outperforming both specialized and foundation model baselines on pan-cancer benchmarks. Furthermore, MoLF exhibits zero-shot generalization to cross-species data, suggesting it captures fundamental, conserved histo-molecular mechanisms.", "date": "2026-02-02", "categories": []}
{"id": "2602.22347", "title": "Enabling clinical use of foundation models for computational pathology", "abstract": "Foundation models for computational pathology are expected to facilitate the development of high-performing, generalisable deep learning systems. However, in addition to biologically relevant features, current foundation models also capture pre-analytic and scanner-specific variation that bias the predictions made by downstream task-specific models trained on these features. Here we show that introducing novel robustness losses during downstream model training reduces sensitivity to technical variability. A purpose-designed comprehensive experimentation setup with 27,042 whole-slide images from 6,155 patients is used to train thousands of models from the features of eight well-known foundation models for computational pathology. In addition to a substantial improvement in robustness, our approach improves classification accuracy by focusing on biologically relevant features. It mitigates robustness limitations of foundation models for computational pathology without retraining the foundation models themselves, enabling development of models that are more suitable in real-world clinical use.", "date": "2026-02-25", "categories": []}
{"id": "2605.11622", "title": "RNA-FM: Flow-Matching Generative Model for Genome-wide RNA-Seq Prediction", "abstract": "Histopathology whole-slide images (WSIs) are routinely acquired in clinical practice and contain rich tissue morphology but lack direct molecular architecture and functional programs defining pathological states, whereas RNA sequencing (RNA-seq) provides genome-wide transcriptional profiles at substantial cost, thereby motivating WSI-based genome-wide transcriptomic prediction. Existing approaches for predicting gene expression from WSIs predominantly rely on deterministic regression with one-to-one mapping, limiting their ability to capture biological heterogeneity and predictive uncertainty. We propose RNA-FM, a flow-matching generative framework for genome-wide bulk RNA-seq prediction from WSIs. RNA-FM formulates transcriptomic prediction as a continuous-time conditional transport problem, learning a velocity field that maps a simple prior to the target gene expression distribution conditioned on morphologies. By integrating pathway-level structure, RNA-FM enables scalable and biologically interpretable genome-wide gene expression imputation. Extensive experiments demonstrate that RNA-FM consistently outperforms state-of-the-art approaches while maintaining biological meaningfulness. Code is available at https://github.com/YXSong000/RNA-FM.", "date": "2026-05-12", "categories": []}
{"id": "2512.06949", "title": "Can We Go Beyond Visual Features? Neural Tissue Relation Modeling for Relational Graph Analysis in Non-Melanoma Skin Histology", "abstract": "Histopathology image segmentation is essential for delineating tissue structures in skin cancer diagnostics, but modeling spatial context and inter-tissue relationships remains a challenge, especially in regions with overlapping or morphologically similar tissues. Current convolutional neural network (CNN)-based approaches operate primarily on visual texture, often treating tissues as independent regions and failing to encode biological context. To this end, we introduce Neural Tissue Relation Modeling (NTRM), a novel segmentation framework that augments CNNs with a tissue-level graph neural network to model spatial and functional relationships across tissue types. NTRM constructs a graph over predicted regions, propagates contextual information via message passing, and refines segmentation through spatial projection. Unlike prior methods, NTRM explicitly encodes inter-tissue dependencies, enabling structurally coherent predictions in boundary-dense zones. On the benchmark Histopathology Non-Melanoma Skin Cancer Segmentation Dataset, NTRM outperforms state-of-the-art methods, achieving a robust Dice similarity coefficient that is 4.9\\% to 31.25\\% higher than the best-performing models among the evaluated approaches. Our experiments indicate that relational modeling offers a principled path toward more context-aware and interpretable histological segmentation, compared to local receptive-field architectures that lack tissue-level structural awareness. Our code is available at https://github.com/shravan-18/NTRM.", "date": "2025-12-07", "categories": []}
{"id": "2602.00953", "title": "SAGE: Agentic Framework for Interpretable and Clinically Translatable Computational Pathology Biomarker Discovery", "abstract": "Engineered image-based biomarkers offer a clinically interpretable alternative to black-box AI in computational pathology, yet their discovery remains largely intuition-driven, guided by fragmented literature rather than rigorous biological validation. We introduce SAGE (Structured Agentic system for hypothesis Generation and Evaluation), a multi-agent framework that grounds biomarker discovery in biological evidence through three mechanisms: (i) knowledge-graph-anchored hypothesis generation via multi-path ontological reasoning, (ii) a debate-based multi-agent novelty assessment that stress-tests candidate biomarkers against existing literature, and (iii) an end-to-end automated validation pipeline that translates hypotheses directly into executable analyses on multimodal pathology datasets. Together, these components shift biomarker discovery from an intuition-driven, literature-browsing exercise into a structured, traceable reasoning process that clinicians and researchers can inspect, trust, and build upon.", "date": "2026-02-01", "categories": []}
{"id": "2605.08207", "title": "A Breast Vision Pathology Foundation Model for Real-world Clinical Utility", "abstract": "Pathology foundation models have shown strong retrospective performance, but whether such systems can support clinically relevant use remains unclear. This challenge is particularly important in breast cancer, where pathological assessment serves as the gold standard for diagnosis and guides treatment planning, surgical decision-making and risk stratification across pre-, intra- and post-operative stages. Here we present \\textbf{BRAVE}, a breast-adaptive pathology foundation model developed and evaluated using a total resource of 101,638 breast whole-slide images from 32 sources across Asia, Europe and North America. We assessed BRAVE across 34 tasks in 82 cohorts spanning pre-operative biopsy, intra-operative frozen section and post-operative resection, using an evidence chain comprising retrospective benchmarking, clinically challenging scenarios, workflow-oriented clinical impact simulations, prospective observational validation with the thresholds locked in the retrospective cohorts and crossover pathologist-AI interaction studies. Across these settings, BRAVE supported practical roles in the clinical workflow, including safe exclusion of low-risk cases from routine review, AI-assisted second-review rescue of initially missed positives and prioritization of cases for further assessment. In prospective validation across three centres, BRAVE excluded 76.9% of negative biopsy cases (NPV 0.953) and 70.1% of negative frozen-section cases (NPV 0.973), and triaged 78.8% of post-operative subtyping cases as high-confidence clear-cut cases (NPV 1.000). In reader studies, AI assistance improved balanced accuracy from 88.5% to 95.1% (OR 3.14, P<0.001), with better efficiency, confidence and inter-rater agreement. BRAVE-derived scores also independently predicted disease-free survival (adjusted HR 4.79, P<0.001) and overall survival (adjusted HR 8.14, P<0.001).", "date": "2026-05-06", "categories": []}
{"id": "2605.08276", "title": "Beyond ViT Tokens: Masked-Diffusion Pretrained Convolutional Pathology Foundation Model for Cell-Level Dense Prediction", "abstract": "Cell-level dense prediction is central to computational pathology, but remains challenging due to fine-grained histological structures, strong domain shifts, and costly dense annotations. Existing ViT-based pathology foundation models rely on patch tokenization, which can disrupt spatial continuity and weaken local morphological details needed for cell-level prediction. To address this, we propose Masked-Diffusion Convolutional Foundation Models, termed ConvNeXt Masked-Diffusion (CMD), a self-supervised convolutional generative pretraining framework for dense pathology representation learning. CMD uses a fully convolutional ConvNeXt-UNet backbone, performs masked-diffusion pretraining in pixel space, and incorporates frozen pathology foundation model features through adaptive normalization. Experimental results demonstrate that CMD consistently outperforms existing ViT-based pathology foundation models and even surpasses state-of-the-art end-to-end segmentation methods while fine-tuning only a small number of task-specific parameters across multiple pathology dense prediction tasks. The advantage is particularly pronounced under limited annotation settings, where CMD exhibits stronger robustness and generalization ability. Our findings suggest that purely convolutional architectures can also serve as competitive pathology foundation models for cell-level dense prediction, achieving leading performance within the current ViT-dominated paradigm and providing a scalable, high-performance solution that better preserves histological structural priors for fine-grained pathology understanding.", "date": "2026-05-08", "categories": []}
{"id": "2605.10362", "title": "CellDX AI Autopilot: Agent-Guided Training and Deployment of Pathology Classifiers", "abstract": "Training AI models for computational pathology currently requires access to expensive whole-slide-image datasets, GPU infrastructure, deep expertise in machine learning, and substantial engineering effort. We present CellDX AI Autopilot, a platform that lets users -- from pathologists with no ML background to ML practitioners running many parallel experiments -- train, evaluate, and deploy whole-slide image classifiers through natural language interaction with an AI agent. The platform provides a structured set of agent skills that guide the user through dataset curation, automated hyperparameter tuning, multi-strategy model comparison, and human-in-the-loop deployment, all on a pre-built dataset of over 32,000 cases and 66,000 H&E-stained whole-slide images with pre-extracted features. We describe the agent skill architecture, the underlying Multiple Instance Learning (MIL) training framework supporting four classification strategies, and an iterative pairwise hyperparameter search (grid or seeded random) that reduces tuning cost by over 30x compared to exhaustive search. CellDX AI Autopilot is, to our knowledge, the first system to expose pathology-specialized agent skills and a pathology-specialized training platform to general-purpose AI agents (e.g. any LLM-based agent runtime), delivering end-to-end automated model training without requiring the agent itself to be domain-specific. The platform addresses both the ML-expertise bottleneck that limits adoption in diagnostic pathology and the engineering bottleneck that limits how many experiments a researcher can run cost-effectively.", "date": "2026-05-11", "categories": []}
{"id": "2602.07608", "title": "HistoMet: A Pan-Cancer Deep Learning Framework for Prognostic Prediction of Metastatic Progression and Site Tropism from Primary Tumor Histopathology", "abstract": "Metastatic Progression remains the leading cause of cancer-related mortality, yet predicting whether a primary tumor will metastasize and where it will disseminate directly from histopathology remains a fundamental challenge. Although whole-slide images (WSIs) provide rich morphological information, prior computational pathology approaches typically address metastatic status or site prediction as isolated tasks, and do not explicitly model the clinically sequential decision process of metastatic risk assessment followed by downstream site-specific evaluation. To address this research gap, we present a decision-aware, concept-aligned MIL framework, HistoMet, for prognostic metastatic outcome prediction from primary tumor WSIs. Our proposed framework adopts a two-module prediction pipeline in which the likelihood of metastatic progression from the primary tumor is first estimated, followed by conditional prediction of metastatic site for high-risk cases. To guide representation learning and improve clinical interpretability, our framework integrates linguistically defined and data-adaptive metastatic concepts through a pretrained pathology vision-language model. We evaluate HistoMet on a multi-institutional pan-cancer cohort of 6504 patients with metastasis follow-up and site annotations. Under clinically relevant high-sensitivity screening settings (95 percent sensitivity), HistoMet significantly reduces downstream workload while maintaining high metastatic risk recall. Conditional on metastatic cases, HistoMet achieves a macro F1 of 74.6 with a standard deviation of 1.3 and a macro one-vs-rest AUC of 92.1. These results demonstrate that explicitly modeling clinical decision structure enables robust and deployable prognostic prediction of metastatic progression and site tropism directly from primary tumor histopathology.", "date": "2026-02-07", "categories": []}
{"id": "2604.15729", "title": "MambaBack: Bridging Local Features and Global Contexts in Whole Slide Image Analysis", "abstract": "Whole Slide Image (WSI) analysis is pivotal in computational pathology, enabling cancer diagnosis by integrating morphological and architectural cues across magnifications. Multiple Instance Learning (MIL) serves as the standard framework for WSI analysis. Recently, Mamba has become a promising backbone for MIL, overtaking Transformers due to its efficiency and global context modeling capabilities originating from Natural Language Processing (NLP). However, existing Mamba-based MIL approaches face three critical challenges: (1) disruption of 2D spatial locality during 1D sequence flattening; (2) sub-optimal modeling of fine-grained local cellular structures; and (3) high memory peaks during inference on resource-constrained edge devices. Studies like MambaOut reveal that Mamba's SSM component is redundant for local feature extraction, where Gated CNNs suffice. Recognizing that WSI analysis demands both fine-grained local feature extraction akin to natural images, and global context modeling akin to NLP, we propose MambaBack, a novel hybrid architecture that harmonizes the strengths of Mamba and MambaOut. First, we propose the Hilbert sampling strategy to preserve the 2D spatial locality of tiles within 1D sequences, enhancing the model's spatial perception. Second, we design a hierarchical structure comprising a 1D Gated CNN block based on MambaOut to capture local cellular features, and a BiMamba2 block to aggregate global context, jointly enhancing multi-scale representation. Finally, we implement an asymmetric chunking design, allowing parallel processing during training and chunking-streaming accumulation during inference, minimizing peak memory usage for deployment. Experimental results on five datasets demonstrate that MambaBack outperforms seven state-of-the-art methods. Source code and datasets are publicly available.", "date": "2026-04-17", "categories": []}
{"id": "2605.05164", "title": "Geometry-Aware State Space Model: A New Paradigm for Whole-Slide Image Representation", "abstract": "Accurate analysis of histopathological images is critical for disease diagnosis and treatment planning. Whole-slide images (WSIs), which digitize tissue specimens at gigapixel resolution, are fundamental to this process but require aggregating thousands of patches for slide-level predictions. Multiple Instance Learning (MIL) tackles this challenge with a two-stage paradigm, decoupling tile-level embedding and slide-level prediction. However, most existing methods implicitly embed patch representations in homogeneous Euclidean spaces, overlooking the hierarchical organization and regional heterogeneity of pathological tissues. This limits current models' ability to capture global tissue architecture and fine-grained cellular morphology. To address this limitation, we introduce a hybrid hyperbolic-Euclidean representation that embeds WSI features in dual geometric spaces, enabling complementary modeling of hierarchical tissue structures and local morphological details. Building on this formulation, we develop BatMIL, a WSI classification framework that leverages both geometric spaces. To model long-range dependencies among thousands of patches, we employ a structured state space sequence model (S4) backbone that encodes patch sequences with linear computational complexity. Furthermore, to account for regional heterogeneity, we introduce a chunk-level mixture-of-experts (MoE) module that groups patches into regions and dynamically routes them to specialized subnetworks, improving representational capacity while reducing redundant computation. Extensive experiments on seven WSI datasets spanning six cancer types demonstrate that BatMIL consistently outperforms state-of-the-art MIL approaches in slide-level classification tasks. These results indicate that geometry-aware representation learning offers a promising direction for next-generation computational pathology.", "date": "2026-05-06", "categories": []}
{"id": "2605.03144", "title": "NucEval: A Robust Evaluation Framework for Nuclear Instance Segmentation", "abstract": "In computational pathology, nuclear instance segmentation is a fundamental task with many downstream clinical applications. With the advent of deep learning, many approaches, including convolutional neural networks (CNNs) and vision transformers (ViTs), have been proposed for this task, along with both machine learning-based and non-machine learning-based pre- and post-processing techniques to further boost performance. However, one fundamental aspect that has received less attention is the evaluation pipeline. In this study, we identify four key issues associated with nuclear instance segmentation evaluation and propose corresponding solutions. Our proposed modifications, namely handling vague regions, score normalization, overlapping instances, and border uncertainty, are integrated into a unified framework called NucEval, which enables robust evaluation of nuclear instance segmentation. We evaluate this pipeline using the NuInsSeg dataset, which provides unique characteristics that make it particularly suitable for this study, as well as two additional external datasets, with three CNN- and ViT-based nuclear instance segmentation models, to demonstrate the impact of these modifications on instance segmentation metrics. The code, along with complete guidelines and illustrative examples, is publicly available at: https://github.com/masih4/nuc_eval.", "date": "2026-05-04", "categories": []}
{"id": "2605.03544", "title": "DALPHIN: Benchmarking Digital Pathology AI Copilots Against Pathologists on an Open Multicentric Dataset", "abstract": "Foundation models with visual question answering capabilities for digital pathology are emerging. Such unprecedented technology requires independent benchmarking to assess its potential in assisting pathologists in routine diagnostics. We created DALPHIN, the first multicentric open benchmark for pathology AI copilots, comprising 1236 images from 300 cases, spanning 130 rare to common diagnoses, 6 countries, and 14 subspecialties. The DALPHIN design and dataset are introduced alongside a human performance benchmark of 31 pathologists from 10 countries with varying expertise. We report results for two general-purpose (GPT-5, Gemini 2.5 Pro) and one pathology-specific copilot (PathChat+) for sequential and independent answer generation. We observed no statistically significant difference from expert-level performance in four of six tasks for PathChat, 2/6 tasks for Gemini, and 1/6 tasks for GPT. DALPHIN is publicly released with sequestered, indirectly accessible ground truth to foster robust and enduring benchmarking. Data, methods, and the evaluation platform are accessible through dalphin.grand-challenge.org.", "date": "2026-05-05", "categories": []}
{"id": "2605.00902", "title": "Validation of Whole-Slide Foundation Models for Image Retrieval in TCGA Data", "abstract": "Foundation models are reshaping computational histopathology, yet their value for whole-slide image retrieval relative to strong patch-based and supervised aggregation baselines remains unclear. We benchmarked ten pipelines on 9,387 diagnostic slides spanning 17 organs and 60 diagnoses from The Cancer Genome Atlas (TCGA) using patient-level leave-one-patient-out evaluation. Methods included four pre-trained slide foundation models, a supervised attention-based multiple instance learning (ABMIL) aggregator on patch embeddings, and patch-level retrieval across five sampling densities.\n  Performance varied more across organs and diagnoses than across architectures. Although the slide foundation model TITAN achieved the strongest overall results, its advantage was modest; ABMIL and patch-based methods reached comparable Top-1 and Top-3 accuracy, with no model consistently dominant. Morphologically distinctive entities approached ceiling performance, while rare, heterogeneous, and closely related subtypes remained challenging. Misclassifications aligned with organs exhibiting known inter-observer variability, suggesting an intrinsic ceiling for morphology-only retrieval.\n  Performance was driven primarily by patch-level feature representations, with limited benefit from slide-level aggregation, indicating aggregation may be unnecessary in many settings. These findings argue against a universally optimal architecture and instead support organ-resolved benchmarking, diagnosis-aware or ensemble strategies, stronger feature representations, and multimodal retrieval frameworks. Notably, even the best model achieved only $\\approx 68\\% \\pm 21\\%$ retrieval accuracy on TCGA, and some subtypes showed $0\\%$ accuracy across all methods, highlighting fundamental limitations of morphology-based representations and the need for substantial progress before reliable clinical deployment.", "date": "2026-04-28", "categories": []}
{"id": "2605.00925", "title": "Linking spatial biology and clinical histology via Haiku", "abstract": "Integrating molecular, morphological, and clinical data is essential for basic and translational biomedical research, yet systematic frameworks for jointly modeling these modalities remain limited. Here we present Haiku, a tri-modal contrastive learning model trained on multiplexed immunofluorescence (mIF). It comprises 26.7 million spatial proteomics patches from 3,218 tissue sections across 1,606 patients spanning 11 organ types, with matched hematoxylin and eosin (H&E) histology and clinical metadata aligned in a shared embedding space. Haiku enables three-way cross-modal retrieval, improves downstream classification and clinical prediction tasks over unimodal baselines, and supports zero-shot biomarker inference through fusion retrieval conditioned on clinical metadata-only text descriptions. Across tasks, Haiku outperforms competing approaches, achieving cross-modal retrieval (Recall@50 up to 0.611 versus near-zero baseline), survival prediction (C-index 0.737, +7.91% relative improvement), and zero-shot biomarker inference (mean Pearson correlation 0.718 across 52 biomarkers). Furthermore, we introduce a counterfactual prediction framework in which modifying only clinical metadata while fixing tissue morphology surfaces niche-specific molecular shifts associated with breast cancer stage progression and lung cancer survival outcomes. In a lung adenocarcinoma case study, the counterfactual analysis recovers niche-specific shifts characterized by increased CD8 and granzyme B, reduced PD-L1, and decreased Ki67, broadly consistent with patterns reported for favorable outcomes. We present these counterfactual results as exploratory, hypothesis-generating signals rather than mechanistic claims. These capabilities demonstrate that tri-modal alignment via Haiku enables integrative analysis of spatial biology, bridging molecular measurements with clinical context for biological exploration.", "date": "2026-04-30", "categories": []}
{"id": "2605.01144", "title": "Semantic Context-aware mOdality fUsion Transformer (SCOUT): A Context-Aware Multimodal Transformer for Concept-Grounded Pathology Report Generation", "abstract": "Whole-slide images (WSIs) present a fundamental challenge for computational pathology due to their extreme resolution, multi-scale heterogeneity, and the requirement for clinically reliable interpretation. Although recent pathology foundation models have enabled fluent report generation, they often lack clinical grounding, failing to accurately represent key diagnostic concepts and relationships observed by pathologists. This limitation arises from the difficulty of integrating heterogeneous visual evidence spanning fine-grained cellular patterns, slide-level tissue architecture, and high-level diagnostic concepts, while maintaining interpretability and clinical coherence. Here we present SCOUT: Semantic Context-aware mOdality fUsion Transformer, a context-aware concept-grounded multimodal framework for pathology report generation that enables progressive conditioning of image representations by global slide information and explicit diagnostic concepts. The method integrates local histological patterns, whole-slide context, and expert-curated semantic descriptors within a unified learning paradigm, allowing visual features to be dynamically refined throughout the encoding process. By combining depth-aware contextual modulation with adaptive multimodal fusion during text generation, the framework produces clinically coherent reports while preserving complementarity across representational scales. Using CONCH1.5 features, we evaluate SCOUT against WSI-Caption, HistGen, and BiGen on TCGA-BRCA, MICCAI REG, and HistAI. SCOUT achieves the best BLEU-1 to BLEU-4 and METEOR scores on all datasets, plus the best ROUGE-L on TCGA-BRCA and MICCAI REG. On TCGA-BRCA, it reaches 0.436/0.303/0.202/0.156 BLEU-1/2/3/4 and 0.204 METEOR; on REG 2025, it achieves 0.865/0.834/0.805/0.780 and 0.568. These results support progressive contextual conditioning for grounded pathology report generation.", "date": "2026-05-01", "categories": []}
{"id": "2605.02614", "title": "Validation of an AI-based end-to-end model for prostate pathology using long-term archived routine samples", "abstract": "Artificial intelligence (AI) is becoming a clinical tool for prostate pathology, but generalization across variations in sample preparation and preservation over prolonged time periods remains poorly understood. We evaluated GleasonAI, an end-to-end attention-based multiple instance learning model, on an independent validation cohort comprising 10,366 biopsy cores from 1,028 patients across 14 Swedish regions, using archival diagnostic specimens from the ProMort cohorts collected between 1998-2015. The model achieved an overall quadratic-weighted kappa of 0.86 for core-level ISUP grading, comparable to several experienced pathologists and consistent across geographic regions. Notably, performance remained stable across the 17-year collection period, demonstrating robustness to time-related variation in archival material, a property not consistently observed with foundation model-based approaches, with exploratory analysis demonstrating a significant prognostic gradient across AI-assigned grade groups for prostate cancer-specific mortality. These findings support the generalizability of the AI grading model and demonstrate the potential of pathology archives as a large-scale resource for AI development, validation, and retrospective prognostic research.", "date": "2026-05-04", "categories": []}
{"id": "2605.02660", "title": "Biological Spatial Priors Regularize Foundation Model Representations for Cross-Site MSI Generalization in Colorectal Cancer", "abstract": "Predicting microsatellite instability (MSI) status from routine hematoxylin and eosin (H&E) whole slide images (WSIs) offers a practical alternative to molecular testing, but models trained at one institution tend to generalize poorly to slides acquired at a different site. Foundation model representations, despite their generality, still encode site-specific texture alongside the conserved biological morphology underlying MSI. We investigate whether tile-level spatial priors derived from known MSI histology can guide these representations toward more site-invariant features. We introduce a biologically motivated spatial prior based on peripheral distance encoding, reflecting the Crohn's-like peripheral lymphocytic reaction at the tumor invasive margin, and evaluate a secondary local immune neighborhood encoding reflecting the lymphocyte-to-tumor ratio in each tile's immediate spatial neighborhood. Both priors are injected into a TransMIL aggregator before self-attention, allowing the transformer to integrate spatial biological context with UNI2-h or Virchow2 features across all attention layers. We evaluate six foundation model and MIL aggregator combinations as a reference, then assess the effect of each spatial prior. Training on TCGA-COAD (137 slides) and evaluating externally on TCGA-READ (50 slides) without retraining, peripheral distance encoding achieves MSI AUC 0.959 +/- 0.012 on COAD and MSS specificity 1.000 on READ, compared to 0.957 and 0.939 for the strongest reference configuration. Local immune neighborhood encoding achieves comparable internal AUC but lower cross-site specificity, suggesting margin proximity encodes a more site-invariant biological signal than local immune density. Results suggest biologically grounded spatial priors act as regularizers that reduce reliance on site-specific imaging patterns.", "date": "2026-05-04", "categories": []}
{"id": "2410.06723", "title": "Evaluating Computational Pathology Foundation Models for Prostate Cancer Grading under Distribution Shifts", "abstract": "Pathology foundation models (PFMs) have emerged as powerful pretrained encoders for computational pathology, but their robustness under clinically relevant distribution shifts remains insufficiently understood. We benchmark the robustness of recent PFMs in the setting of prostate cancer grading from whole-slide images (WSIs). Using the PANDA dataset, we evaluate PFMs as frozen patch-level feature extractors within weakly supervised slide-level grading models, and assess robustness to two important forms of distribution shift: shifts in WSI image appearance across collection sites, and shifts in the label distribution over cancer grade groups. Across in-distribution settings, PFMs consistently achieve strong performance and clearly outperform a natural-image baseline. Under cross-site transfer from Radboud to Karolinska, however, performance drops substantially for all models, showing that large-scale pretraining alone does not guarantee robust downstream generalization. In contrast, PFMs are less sensitive to label-distribution shift, indicating that visually grounded domain shift is the dominant challenge. Representation analysis further supports these findings by revealing persistent domain separation between sites across all PFMs. While grade-related structure is present, it is comparatively weak, indicating that domain-related variation dominates in the learned feature space. Together, these results provide a comprehensive benchmark of PFMs under distribution shift and highlight an important practical message: although PFMs provide strong representations, generalizability remains constrained by the quality and diversity of the data used to train downstream prediction models.", "date": "2024-10-09", "categories": []}
{"id": "2601.01056", "title": "Enhancing Histopathological Image Classification via Integrated HOG and Deep Features with Robust Noise Performance", "abstract": "The era of digital pathology has advanced histopathological examinations, making automated image analysis essential in clinical practice. This study evaluates the classification performance of machine learning and deep learning models on the LC25000 dataset, which includes five classes of histopathological images. We used the fine-tuned InceptionResNet-v2 network both as a classifier and for feature extraction. Our results show that the fine-tuned InceptionResNet-v2 achieved a classification accuracy of 96.01\\% and an average AUC of 96.8\\%. Models trained on deep features from InceptionResNet-v2 outperformed those using only the pre-trained network, with the Neural Network model achieving an AUC of 99.99\\% and accuracy of 99.84\\%. Evaluating model robustness under varying SNR conditions revealed that models using deep features exhibited greater resilience, particularly GBM and KNN. The combination of HOG and deep features showed enhanced performance, however, less so in noisy environments.", "date": "2026-01-03", "categories": []}
{"id": "2604.01798", "title": "A deep learning pipeline for PAM50 subtype classification using histopathology images and multi-objective patch selection", "abstract": "Breast cancer is a highly heterogeneous disease with diverse molecular profiles. The PAM50 gene signature is widely recognized as a standard for classifying breast cancer into intrinsic subtypes, enabling more personalized treatment strategies. In this study, we introduce a novel optimization-driven deep learning framework that aims to reduce reliance on costly molecular assays by directly predicting PAM50 subtypes from H&E-stained whole-slide images (WSIs). Our method jointly optimizes patch informativeness, spatial diversity, uncertainty, and patch count by combining the non-dominated sorting genetic algorithm II (NSGA-II) with Monte Carlo dropout-based uncertainty estimation. The proposed method can identify a small but highly informative patch subset for classification. We used a ResNet18 backbone for feature extraction and a custom CNN head for classification. For evaluation, we used the internal TCGA-BRCA dataset as the training cohort and the external CPTAC-BRCA dataset as the test cohort. On the internal dataset, an F1-score of 0.8812 and an AUC of 0.9841 using 627 WSIs from the TCGA-BRCA cohort were achieved. The performance of the proposed approach on the external validation dataset showed an F1-score of 0.7952 and an AUC of 0.9512. These findings indicate that the proposed optimization-guided, uncertainty-aware patch selection can achieve high performance and improve the computational efficiency of histopathology-based PAM50 classification compared to existing methods, suggesting a scalable imaging-based replacement that has the potential to support clinical decision-making.", "date": "2026-04-02", "categories": []}
{"id": "2604.25817", "title": "Magnification-Invariant Image Classification via Domain Generalization and Stable Sparse Embedding Signatures", "abstract": "Magnification shift is a major obstacle to robust histopathology classification, because models trained on one imaging scale often generalize poorly to another. Here, we evaluated this problem on the BreaKHis dataset using a strict patient-disjoint leave-one-magnification-out protocol, comparing supervised baseline, baseline augmented with DCGAN-generated patches, and a gradient-reversal domain-general model designed to preserve discriminative information while suppressing magnification-specific variation. Across held-out magnifications, the domain-general model achieved the strongest overall discrimination and its clearest gain was observed when 200X was held out. By contrast, GAN augmentation produced inconsistent effects, improving some folds but degrading others, particularly at 400X. The domain-general model also yielded the lowest Brier score at 0.063 vs 0.089 at baseline. Sparse embedding analysis further revealed that domain-general training reduced average signature size more than three-fold (306 versus 1,074 dimensions) while preserving equivalent predictive performance (AUC: 0.967 vs 0.965; F1: 0.930 vs 0.931). It also increased cross-fold signature reproducibility from near-zero Jaccard overlap in the baseline to 0.99 between the 100X and 200X folds. These findings show that calibrated, compact, and transferable representations can be learned without added architectural complexity, with clear implications for the reliable deployment of computational pathology models across heterogeneous acquisition settings.", "date": "2026-04-28", "categories": []}
{"id": "2604.22846", "title": "Unified Multi-Foundation-Model Slide Representation for Pan-Cancer Recognition and Text-Guided Tumor Localization", "abstract": "The expanding ecosystem of pathology foundation models has produced powerful but fragmented tile-level representations, limiting their use in clinical tasks that require unified slide-level reasoning and interpretable linkage to clinically meaningful information. We present ASTRA, a pan-cancer framework that integrates heterogeneous foundation-model representations into a shared slide-level representation space and semantically grounds that space using structured pathology annotation fields, including classification category, cancer type, and anatomic site. ASTRA combines sparse mixture-of-experts contextualization, masked multi-model reconstruction, and contrastive alignment to structured pathology prompts to learn slide representations that support 4-category classification, 3-class solid tumor typing, 16-class cancer typing, and text-guided tumor localization without pixel-level supervision. Developed on a CHTN cohort of 10,359 whole-slide images (WSIs) spanning 16 tumor types, ASTRA consistently improves pan-cancer classification across four pathology foundation-model backbones, achieving up to 97.8% macro-AUC for 4-category classification, 99.7% for 3-class solid tumor typing, and 99.2% for 16-class cancer typing. For tumor localization, ASTRA achieves a mean Dice of 0.897 on an annotated in-domain CHTN subset (n = 380) spanning 16 cancer types and 0.738 on an external TCGA cohort (n = 1,686) spanning four cancer types. These results demonstrate that minimal structured pathology annotation fields derived from slide-level metadata can provide effective semantic supervision for unified slide representation learning, enabling both pan-cancer prediction and weakly supervised tumor localization within a single framework.", "date": "2026-04-21", "categories": []}
{"id": "2604.22858", "title": "A Digital Pathology Resource for Liver Cancer Quantification with Datasets, Benchmarks, and Tools", "abstract": "Liver cancer, especially hepatocellular carcinoma (HCC), imposes a substantial global disease burden. Accurate diagnosis and prognostic assessment directly influence treatment selection and patient survival, and pathological examination remains the gold standard for liver cancer diagnosis. Identifying diverse tissue components and pathological subtypes on histopathology slides is crucial for estimating postoperative recurrence risk and overall prognosis. However, most publicly available resources are still provided at the whole-slide image (WSI) level, and well-annotated datasets for fine-grained tissue component identification in liver cancer are scarce, which hinders reproducible model development and the deployment of quantitative analysis tools. To address this gap, we release HepatoBench, a patch-level image database for liver cancer with annotations for seven key tissue categories. Based on HepatoBench, we train and open-source a deep learning classification model as a tissue recognition tool. Furthermore, we train a WSI-level tumor/non-tumor segmentation model to automatically localize lesion regions across entire slides. By integrating the patch-level tissue classifier with the WSI-level segmentation model, we build HepatoQuant, an end-to-end, disease-specific regional quantification tool for liver cancer, enabling a unified workflow from WSIs to tissue composition parsing and quantitative statistics. We also open-source HepatoBench, the benchmarking protocol, and supporting tools, providing a solid foundation for automated regional quantification and fair method comparison in liver cancer pathology.", "date": "2026-04-22", "categories": []}
{"id": "2604.23481", "title": "Leveraging Spatial Transcriptomics as Alternative to Manual Annotations for Deep Learning-Based Nuclei Analysis", "abstract": "Deep learning-based nuclei segmentation and classification in pathology images typically rely on large-scale pixel-level manual annotations, which are costly and difficult to obtain across diverse tissues and staining conditions. To address this limitation, we propose a framework that leverages spatial transcriptomics (ST) data as supervision for nuclei segmentation and classification. By incorporating cell-level ST data, we obtain gene expression profiles and corresponding nuclear masks from histopathological images. Gene expression profiles are converted into cell-type labels and used as training data for image-based classification. Because existing gene expression-based cell-type classification methods are not designed for image recognition, we introduce an image-oriented classification approach that bridges gene expression-based cell typing and image-based cell classification. To evaluate generalization, we conduct segmentation experiments on previously unseen organs and compare our method with conventional supervised models. Despite being trained on fewer organ types, our framework achieves higher segmentation accuracy, demonstrating strong transferability. Classification experiments further show consistent improvements over existing approaches.", "date": "2026-04-26", "categories": []}
{"id": "2604.23706", "title": "Weakly Supervised Multicenter Nancy Index Scoring in Ulcerative Colitis Using Foundation Models", "abstract": "Histologic assessment of ulcerative colitis (UC) activity is an important endpoint in clinical trials and routine care, but manual grading with indices such as the Nancy histological index (NHI) is time-consuming and prone to observer variability. While computational pathology methods can automate scoring, many approaches depend on dense region-level annotations, which are costly to obtain, particularly in heterogeneous, multicenter cohorts.\n  We propose a weakly supervised multiple instance learning (MIL) approach for whole-slide images that learns from case- and slide-level NHI labels, leveraging foundation models. Our method targets clinically relevant endpoints, including neutrophilic activity and derived Nancy-low/high groupings, enabling full five-grade NHI prediction.\n  On a multicenter dataset of H&E-stained colon biopsies from three hospitals (2019-2025), we evaluate multiple foundation model encoders and aggregation strategies. We find that foundation model choice and resolution substantially affect performance, with Virchow2 providing the most consistent gains, and that a simple ensembling rule improves five-grade NHI prediction compared to a hierarchical gating baseline.\n  Overall, our results demonstrate that weakly supervised MIL with modern foundation-model representations can provide robust, interpretable UC histology activity assessment in realistic multicenter settings.", "date": "2026-04-26", "categories": []}
{"id": "2604.23799", "title": "VitaminP: cross-modal learning enables whole-cell segmentation from routine histology", "abstract": "Accurate whole-cell and nuclear segmentation is essential for precision pathology and spatial omics, yet routine hematoxylin and eosin (H&E) staining provides limited cytoplasmic contrast, restricting analyses to nuclei. Multiplex immunofluorescence (mIF) facilitates precise whole-cell delineation but remains constrained by cost and accessibility. We introduce VitaminP, a cross-modal learning framework enabling whole cell segmentation from H&E images. By learning from paired H&E-mIF data, VitaminP transfers molecular boundary information from mIF to overcome cytoplasmic contrast in H&E, establishing cross-modal supervision as a general strategy for recovering missing biological structure. We train VitaminP on 14 public datasets covering 34 cancer types and over 7 million instances, integrating publicly available labels with extensive annotations generated in this study, forming one of the largest resources for segmentation. VitaminP outperforms four state-of-the-art methods and generalizes to unseen datasets, including an in-house dataset spanning 24 rare cancer types. We further developed VitaminPScope, an open-source platform providing an interface for scalable inference and enabling broad adoption.", "date": "2026-04-26", "categories": []}
{"id": "2604.24679", "title": "Benchmarking Pathology Foundation Models for Breast Cancer Survival Prediction", "abstract": "Pathology foundation models (PFMs) have recently emerged as powerful pretrained encoders for computational pathology, enabling transfer learning across a wide range of downstream tasks. However, systematic comparisons of these models for clinically meaningful prediction problems remain limited, especially in the context of survival prediction under external validation. In this study, we benchmark widely used and recently proposed PFMs for breast cancer survival prediction from whole-slide histopathology images. Using a standardized pipeline based on patch-level feature extraction and a unified survival modeling framework, we evaluate model representations across three independent clinical cohorts comprising more than 5,400 patients with long-term follow-up. Models are trained on one cohort and evaluated on two independent external cohorts, enabling a rigorous assessment of cross-dataset generalization. Overall, H-optimus-1 achieves the strongest survival prediction performance. More broadly, we observe consistent generational improvements across model families, with second-generation PFMs outperforming their first-generation counterparts. However, absolute performance differences between many recent PFMs remain modest, suggesting diminishing returns from further scaling of pretraining data or model size alone. Notably, the compact distilled model H0-mini slightly outperforms its larger teacher model H-optimus-0, despite using fewer than 8% of the parameters and enabling significantly faster feature extraction. Together, these results provide the first large-scale, externally validated benchmark of PFMs for breast cancer survival prediction, and offer practical guidance for efficient deployment of PFMs in clinical workflows.", "date": "2026-04-27", "categories": []}
{"id": "2512.09003", "title": "Digital Modeling of Spatial Pathway Activity from Histology Reveals Tumor Microenvironment Heterogeneity", "abstract": "Spatial transcriptomics (ST) enables simultaneous mapping of tissue morphology and spatially resolved gene expression, offering unique opportunities to study tumor microenvironment heterogeneity. Here, we introduce a computational framework that predicts spatial pathway activity directly from hematoxylin-and-eosin-stained histology images at microscale resolution 55 and 100 um. Using image features derived from a computational pathology foundation model, we found that TGFb signaling was the most accurately predicted pathway across three independent breast and lung cancer ST datasets. In 87-88% of reliably predicted cases, the resulting spatial TGFb activity maps reflected the expected contrast between tumor and adjacent non-tumor regions, consistent with the known role of TGFb in regulating interactions within the tumor microenvironment. Notably, linear and nonlinear predictive models performed similarly, suggesting that image features may relate to pathway activity in a predominantly linear fashion or that nonlinear structure is small relative to measurement noise. These findings demonstrate that features extracted from routine histopathology may recover spatially coherent and biologically interpretable pathway patterns, offering a scalable strategy for integrating image-based inference with ST information in tumor microenvironment studies.", "date": "2025-12-09", "categories": []}
{"id": "2604.21530", "title": "Attention-based multiple instance learning for predominant growth pattern prediction in lung adenocarcinoma wsi using foundation models", "abstract": "Lung adenocarcinoma (LUAD) grading depends on accurately identifying growth patterns, which are indicators of prognosis and can influence treatment decisions. Common deep learning approaches to determine the predominant pattern rely on patch-level classification or segmentation, requiring extensive annotations. This study proposes an attention-based multiple instance learning (ABMIL) framework to predict the predominant LUAD growth pattern at the whole slide level to reduce annotation burden. Our approach integrates pretrained pathology foundation models as patch encoders, used either frozen or fine-tuned on annotated patches, to extract discriminative features that are aggregated through attention mechanisms. Experiments show that fine-tuned encoders improve performance, with Prov-GigaPath achieving the highest agreement (\\k{appa} = 0.699) under ABMIL. Compared to simple patch-aggregation baselines, ABMIL yields more robust predictions by leveraging slide-level supervision and spatial attention. Future work will extend this framework to estimate the full distribution of growth patterns and validate performance on external cohorts.", "date": "2026-04-23", "categories": []}
{"id": "2510.23807", "title": "Beyond the Failures: Rethinking Foundation Models in Pathology", "abstract": "Despite their successes in vision and language, foundation models have stumbled in pathology, revealing low accuracy, instability, and heavy computational demands. These shortcomings stem not from tuning problems but from deeper conceptual mismatches: dense embeddings cannot represent the combinatorial richness of tissue, and current architectures inherit flaws in self-supervision, patch design, and noise-fragile pretraining. Biological complexity and limited domain innovation further widen the gap. The evidence is clear-pathology requires models explicitly designed for biological images rather than adaptations of large-scale natural-image methods whose assumptions do not hold for tissue.", "date": "2025-10-27", "categories": []}
{"id": "2604.18368", "title": "DSA-CycleGAN: A Domain Shift Aware CycleGAN for Robust Multi-Stain Glomeruli Segmentation", "abstract": "A key challenge in segmentation in digital histopathology is inter- and intra-stain variations as it reduces model performance. Labelling each stain is expensive and time-consuming so methods using stain transfer via CycleGAN, have been developed for training multi-stain segmentation models using labels from a single stain. Nevertheless, CycleGAN tends to introduce noise during translation because of the one-to-many nature of some stain pairs, which conflicts with its cycle consistency loss. To address this, we propose the Domain Shift Aware CycleGAN, which reduces the presence of such noise. Furthermore, we evaluate several advances from the field of machine learning aimed at resolving similar problems and compare their effectiveness against DSA-CycleGAN in the context of multi-stain glomeruli segmentation. Experiments demonstrate that DSA-CycleGAN not only improves segmentation performance in glomeruli segmentation but also outperforms other methods in reducing noise. This is particularly evident when translating between biologically distinct stains. The code is publicly available at https://github.com/zeeshannisar/DSA-CycleGAN.", "date": "2026-04-20", "categories": []}
{"id": "2512.17326", "title": "Democratising Pathology Co-Pilots: An Open Pipeline and Dataset for Whole-Slide Vision-Language Modelling", "abstract": "Vision-language models (VLMs) have the potential to become co-pilots for pathologists. However, most VLMs either focus on small regions of interest within whole-slide images, provide only static slide-level outputs, or rely on data that is not publicly available, limiting reproducibility. Furthermore, training data containing WSIs paired with detailed clinical reports is scarce, restricting progress toward transparent and generalisable VLMs. We address these limitations with three main contributions. First, we introduce Polysome, a standardised tool for synthetic instruction generation. Second, we apply Polysome to the public HISTAI dataset, generating HISTAI-Instruct, a large whole-slide instruction tuning dataset spanning 24,259 slides and over 1.1 million instruction-response pairs. Finally, we use HISTAI-Instruct to train ANTONI-{\\alpha}, a VLM capable of visual-question answering (VQA). We show that ANTONI-{\\alpha} outperforms MedGemma on WSI-level VQA tasks of tissue identification, neoplasm detection, and differential diagnosis. We also compare the performance of multiple incarnations of ANTONI-{\\alpha} trained with different amounts of data. All methods, data, and code are publicly available.", "date": "2025-12-19", "categories": []}
{"id": "2503.24135", "title": "PixelCAM: Pixel Class Activation Mapping for Histology Image Classification and ROI Localization", "abstract": "Weakly supervised object localization (WSOL) methods allow training models to classify images and localize ROIs. WSOL only requires low-cost image-class annotations yet provides a visually interpretable classifier. Standard WSOL methods rely on class activation mapping (CAM) methods to produce spatial localization maps according to a single- or two-step strategy. While both strategies have made significant progress, they still face several limitations with histology images. Single-step methods can easily result in under- or over-activation due to the limited visual ROI saliency in histology images and scarce localization cues. They also face the well-known issue of asynchronous convergence between classification and localization tasks. The two-step approach is sub-optimal because it is constrained to a frozen classifier, limiting the capacity for localization. Moreover, these methods also struggle when applied to out-of-distribution (OOD) datasets. In this paper, a multi-task approach for WSOL is introduced for simultaneous training of both tasks to address the asynchronous convergence problem. In particular, localization is performed in the pixel-feature space of an image encoder that is shared with classification. This allows learning discriminant features and accurate delineation of foreground/background regions to support ROI localization and image classification. We propose PixelCAM, a cost-effective foreground/background pixel-wise classifier in the pixel-feature space that allows for spatial object localization. Using partial-cross entropy, PixelCAM is trained using pixel pseudo-labels collected from a pretrained WSOL model. Both image and pixel-wise classifiers are trained simultaneously using standard gradient descent. In addition, our pixel classifier can easily be integrated into CNN- and transformer-based architectures without any modifications.", "date": "2025-03-31", "categories": []}
{"id": "2604.06390", "title": "MorphDistill: Distilling Unified Morphological Knowledge from Pathology Foundation Models for Colorectal Cancer Survival Prediction", "abstract": "Background: Colorectal cancer (CRC) remains a leading cause of cancer-related mortality worldwide. Accurate survival prediction is essential for treatment stratification, yet existing pathology foundation models often overlook organ-specific features critical for CRC prognostication.\n  Methods: We propose MorphDistill, a two-stage framework that distills complementary knowledge from multiple pathology foundation models into a compact CRC-specific encoder. In Stage I, a student encoder is trained using dimension-agnostic multi-teacher relational distillation with supervised contrastive regularization on large-scale colorectal datasets. This preserves inter-sample relationships from ten foundation models without explicit feature alignment. In Stage II, the encoder extracts patch-level features from whole-slide images, which are aggregated via attention-based multiple instance learning to predict five-year survival.\n  Results: On the Alliance/CALGB 89803 cohort (n=424, stage III CRC), MorphDistill achieves an AUC of 0.68 (SD 0.08), an approximately 8% relative improvement over the strongest baseline (AUC 0.63). It also attains a C-index of 0.661 and a hazard ratio of 2.52 (95% CI: 1.73-3.65), outperforming all baselines. On an external TCGA cohort (n=562), it achieves a C-index of 0.628, demonstrating strong generalization across datasets and robustness across clinical subgroups.\n  Conclusion: MorphDistill enables task-specific representation learning by integrating knowledge from multiple foundation models into a unified encoder. This approach provides an efficient strategy for prognostic modeling in computational pathology, with potential for broader oncology applications. Further validation across additional cohorts and disease stages is warranted.", "date": "2026-04-07", "categories": []}
{"id": "2604.12075", "title": "OpenTME: An Open Dataset of AI-powered H&E Tumor Microenvironment Profiles from TCGA", "abstract": "The tumor microenvironment (TME) plays a central role in cancer progression, treatment response, and patient outcomes, yet large-scale, consistent, and quantitative TME characterization from routine hematoxylin and eosin (H&E)-stained histopathology remains scarce. We introduce OpenTME, an open-access dataset of pre-computed TME profiles derived from 3,634 H&E-stained whole-slide images across five cancer types (bladder, breast, colorectal, liver, and lung cancer) from The Cancer Genome Atlas (TCGA). All outputs were generated using Atlas H&E-TME, an AI-powered application built on the Atlas family of pathology foundation models, which performs tissue quality control, tissue segmentation, cell detection and classification, and spatial neighborhood analysis, yielding over 4,500 quantitative readouts per slide at cell-level resolution. OpenTME is available for non-commercial academic research on Hugging Face. We will continue to expand OpenTME over time and anticipate it will serve as a resource for biomarker discovery, spatial biology research, and the development of computational methods for TME analysis.", "date": "2026-04-13", "categories": []}
{"id": "2604.12100", "title": "PC-MIL: Decoupling Feature Resolution from Supervision Scale in Whole-Slide Learning", "abstract": "Whole-slide image (WSI) classification in computational pathology is commonly formulated as slide-level Multiple Instance Learning (MIL) with a single global bag representation. However, slide-level MIL is fundamentally underconstrained: optimizing only global labels encourages models to aggregate features without learning anatomically meaningful localization. This creates a mismatch between the scale of supervision and the scale of clinical reasoning. Clinicians assess tumor burden, focal lesions, and architectural patterns within millimeter-scale regions, whereas standard MIL is trained only to predict whether \"somewhere in the slide there is cancer.\" As a result, the model's inductive bias effectively erases anatomical structure. We propose Progressive-Context MIL (PC-MIL), a framework that treats the spatial extent of supervision as a first-class design dimension. Rather than altering magnification, patch size, or introducing pixel-level segmentation, we decouple feature resolution from supervision scale. Using fixed 20x features, we vary MIL bag extent in millimeter units and anchor supervision at a clinically motivated 2mm scale to preserve comparable tumor burden and avoid confounding scale with lesion density. PC-MIL progressively mixes slide- and region-level supervision in controlled proportions, enabling explicit train-context x test-context analysis. On 1,476 prostate WSIs from five public datasets for binary cancer detection, we show that anatomical context is an independent axis of generalization in MIL, orthogonal to feature resolution: modest regional supervision improves cross-context performance, and balanced multi-context training stabilizes accuracy across slide and regional evaluation without sacrificing global performance. These results demonstrate that supervision extent shapes MIL inductive bias and support anatomically grounded WSI generalization.", "date": "2026-04-13", "categories": []}
{"id": "2604.07779", "title": "Plug-and-Play Logit Fusion for Heterogeneous Pathology Foundation Models", "abstract": "Pathology foundation models (FMs) have become central to computational histopathology, offering strong transfer performance across a wide range of diagnostic and prognostic tasks. The rapid proliferation of pathology foundation models creates a model-selection bottleneck: no single model is uniformly best, yet exhaustively adapting and validating many candidates for each downstream endpoint is prohibitively expensive. We address this challenge with a lightweight and novel model fusion strategy, LogitProd, which treats independently trained FM-based predictors as fixed experts and learns sample-adaptive fusion weights over their slide-level outputs. The fusion operates purely on logits, requiring no encoder retraining and no feature-space alignment across heterogeneous backbones. We further provide a theoretical analysis showing that the optimal weighted product fusion is guaranteed to perform at least as well as the best individual expert under the training objective. We systematically evaluate LogitProd on \\textbf{22} benchmarks spanning WSI-level classification, tile-level classification, gene mutation prediction, and discrete-time survival modeling. LogitProd ranks first on 20/22 tasks and improves the average performance across all tasks by ~3% over the strongest single expert. LogitProd enables practitioners to upgrade heterogeneous FM-based pipelines in a plug-and-play manner, achieving multi-expert gains with $\\sim$12$\\times$ lower training cost than feature-fusion alternatives.", "date": "2026-04-09", "categories": []}
{"id": "2604.07936", "title": "Shortcut Learning in Glomerular AI: Adversarial Penalties Hurt, Entropy Helps", "abstract": "Stain variability is a pervasive source of distribution shift and potential shortcut learning in renal pathology AI. We ask whether lupus nephritis glomerular lesion classifiers exploit stain as a shortcut, and how to mitigate such bias without stain or site labels. We curate a multi-center, multi-stain dataset of 9,674 glomerular patches (224$\\times$224) from 365 WSIs across three centers and four stains (PAS, H&E, Jones, Trichrome), labeled as proliferative vs. non-proliferative. We evaluate Bayesian CNN and ViT backbones with Monte Carlo dropout in three settings: (1) stain-only classification; (2) a dual-head model jointly predicting lesion and stain with supervised stain loss; and (3) a dual-head model with label-free stain regularization via entropy maximization on the stain head. In (1), stain identity is trivially learnable, confirming a strong candidate shortcut. In (2), varying the strength and sign of stain supervision strongly modulates stain performance but leaves lesion metrics essentially unchanged, indicating no measurable stain-driven shortcut learning on this multi-stain, multi-center dataset, while overly adversarial stain penalties inflate predictive uncertainty. In (3), entropy-based regularization holds stain predictions near chance without degrading lesion accuracy or calibration. Overall, a carefully curated multi-stain dataset can be inherently robust to stain shortcuts, and a Bayesian dual-head architecture with label-free entropy regularization offers a simple, deployment-friendly safeguard against potential stain-related drift in glomerular AI.", "date": "2026-04-09", "categories": []}
{"id": "2604.09076", "title": "Cross-Modal Knowledge Distillation from Spatial Transcriptomics to Histology", "abstract": "Spatial transcriptomics provides a molecularly rich description of tissue organization, enabling unsupervised discovery of tissue niches -- spatially coherent regions of distinct cell-type composition and function that are relevant to both biological research and clinical interpretation. However, spatial transcriptomics remains costly and scarce, while H&E histology is abundant but carries a less granular signal. We propose to leverage paired spatial transcriptomics and H&E data to transfer transcriptomics-derived niche structure to a histology-only model via cross-modal distillation. Across multiple tissue types and disease contexts, the distilled model achieves substantially higher agreement with transcriptomics-derived niche structure than unsupervised morphology-based baselines trained on identical image features, and recovers biologically meaningful neighborhood composition as confirmed by cell-type analysis. The resulting framework leverages paired spatial transcriptomic and H&E data during training, and can then be applied to held-out tissue regions using histology alone, without any transcriptomic input at inference time.", "date": "2026-04-10", "categories": []}
{"id": "2604.08305", "title": "HistDiT: A Structure-Aware Latent Conditional Diffusion Model for High-Fidelity Virtual Staining in Histopathology", "abstract": "Immunohistochemistry (IHC) is essential for assessing specific immune biomarkers like Human Epidermal growth-factor Receptor 2 (HER2) in breast cancer. However, the traditional protocols of obtaining IHC stains are resource-intensive, time-consuming, and prone to structural damages. Virtual staining has emerged as a scalable alternative, but it faces significant challenges in preserving fine-grained cellular structures while accurately translating biochemical expressions. Current state-of-the-art methods still rely on Generative Adversarial Networks (GANs) or standard convolutional U-Net diffusion models that often struggle with \"structure and staining trade-offs\". The generated samples are either structurally relevant but blurry, or texturally realistic but have artifacts that compromise their diagnostic use. In this paper, we introduce HistDiT, a novel latent conditional Diffusion Transformer (DiT) architecture that establishes a new benchmark for visual fidelity in virtual histological staining. The novelty introduced in this work is, a) the Dual-Stream Conditioning strategy that explicitly maintains a balance between spatial constraints via VAE-encoded latents and semantic phenotype guidance via UNI embeddings; b) the multi-objective loss function that contributes to sharper images with clear morphological structure; and c) the use of the Structural Correlation Metric (SCM) to focus on the core morphological structure for precise assessment of sample quality. Consequently, our model outperforms existing baselines, as demonstrated through rigorous quantitative and qualitative evaluations.", "date": "2026-04-09", "categories": []}
{"id": "2604.07298", "title": "Region-Graph Optimal Transport Routing for Mixture-of-Experts Whole-Slide Image Classification", "abstract": "Multiple Instance Learning (MIL) is the dominant framework for gigapixel whole-slide image (WSI) classification in computational pathology. However, current MIL aggregators route all instances through a shared pathway, constraining their capacity to specialise across the pathological heterogeneity inherent in each slide. Mixture-of-Experts (MoE) methods offer a natural remedy by partitioning instances across specialised expert subnetworks; yet unconstrained softmax routing may yield highly imbalanced utilisation, where one or a few experts absorb most routing mass, collapsing the mixture back to a near-single-pathway solution. To address these limitations, we propose ROAM (Region-graph OptimAl-transport Mixture-of-experts), a spatially aware MoE-MIL aggregator that routes region tokens to expert poolers via capacity-constrained entropic optimal transport, promoting balanced expert utilisation by construction. ROAM operates on spatial region tokens, obtained by compressing dense patch bags into spatially binned units that align routing with local tissue neighbourhoods and introduces two key mechanisms: (i) region-to-expert assignment formulated as entropic optimal transport (Sinkhorn) with explicit per slide capacity marginals, enforcing balanced expert utilisation without auxiliary load-balancing losses; and (ii) graph-regularised Sinkhorn iterations that diffuse routing assignments over the spatial region graph, encouraging neighbouring regions to coherently route to the same experts. Evaluated on four WSI benchmarks with frozen foundation-model patch embeddings, ROAM achieves performance competitive against strong MIL and MoE baselines, and on NSCLC generalisation (TCGA-CPTAC) reaches external AUC 0.845 +- 0.019.", "date": "2026-04-08", "categories": []}
{"id": "2511.17652", "title": "TeamPath: Building MultiModal Pathology Experts with Reasoning AI Copilots", "abstract": "Advances in AI have introduced several strong models in computational pathology to usher it into the era of multi-modal diagnosis, analysis, and interpretation. However, the current pathology-specific visual language models still lack capacities in making the diagnosis with rigorous reasoning paths as well as handling divergent tasks, and thus, challenges of building AI Copilots for real scenarios still exist. Here we introduce TeamPath, an AI system powered by reinforcement learning and router-enhanced solutions based on large-scale histopathology multimodal datasets, to work as a virtual assistant for expert-level disease diagnosis, patch-level information summarization, and cross-modality generation to integrate transcriptomic information for clinical usage. We also collaborate with pathologists from Yale School of Medicine to demonstrate that TeamPath can assist them in working more efficiently by identifying and correcting expert conclusions and reasoning paths. We also discuss the human evaluation results to support the reasoning quality from TeamPath. Overall, TeamPath can flexibly choose the best settings according to the needs, and serve as an innovative and reliable system for information communication across different modalities and experts.", "date": "2025-11-20", "categories": []}
{"id": "2601.10073", "title": "ReaMIL: Reasoning- and Evidence-Aware Multiple Instance Learning for Whole-Slide Histopathology", "abstract": "We introduce ReaMIL (Reasoning- and Evidence-Aware MIL), a multiple instance learning approach for whole-slide histopathology that adds a light selection head to a strong MIL backbone. The head produces soft per-tile gates and is trained with a budgeted-sufficiency objective: a hinge loss that enforces the true-class probability to be $\\geq \\tau$ using only the kept evidence, under a sparsity budget on the number of selected tiles. The budgeted-sufficiency objective yields small, spatially compact evidence sets without sacrificing baseline performance. Across TCGA-NSCLC (LUAD vs. LUSC), TCGA-BRCA (IDC vs. Others), and PANDA, ReaMIL matches or slightly improves baseline AUC and provides quantitative evidence-efficiency diagnostics. On NSCLC, it attains AUC 0.983 with a mean minimal sufficient K (MSK) $\\approx 8.2$ tiles at $\\tau = 0.90$ and AUKC $\\approx 0.864$, showing that class confidence rises sharply and stabilizes once a small set of tiles is kept. The method requires no extra supervision, integrates seamlessly with standard MIL training, and naturally yields slide-level overlays. We report accuracy alongside MSK, AUKC, and contiguity for rigorous evaluation of model behavior on WSIs.", "date": "2026-01-15", "categories": []}
{"id": "2604.03630", "title": "A Multimodal Foundation Model of Spatial Transcriptomics and Histology for Biological Discovery and Clinical Prediction", "abstract": "Spatial transcriptomics (ST) enables gene expression mapping within anatomical context but remains costly and low-throughput. Hematoxylin and eosin (H\\&E) staining offers rich morphology yet lacks molecular resolution. We present \\textbf{\\ours} (\\textbf{S}patial \\textbf{T}ranscriptomics and hist\\textbf{O}logy \\textbf{R}epresentation \\textbf{M}odel), a foundation model trained on 1.2 million spatially resolved transcriptomic profiles with matched histology across 18 organs. Using a hierarchical architecture integrating morphological features, gene expression, and spatial context, STORM bridges imaging and omics through robust molecular--morphological representations. STORM enhances spatial domain discovery, producing biologically coherent tissue maps, and outperforms existing methods in predicting spatial gene expression from H\\&E images across 11 tumor types. The model is platform-agnostic, performing consistently across Visium, Xenium, Visium HD, and CosMx. Applied to 23 independent cohorts comprising 7,245 patients, STORM significantly improves immunotherapy response prediction and prognostication over established biomarkers, providing a scalable framework for spatially informed discovery and clinical precision medicine.", "date": "2026-04-04", "categories": []}
{"id": "2604.03635", "title": "A Generative Foundation Model for Multimodal Histopathology", "abstract": "Accurate diagnosis and treatment of complex diseases require integrating histological, molecular, and clinical data, yet in practice these modalities are often incomplete owing to tissue scarcity, assay cost, and workflow constraints. Existing computational approaches attempt to impute missing modalities from available data but rely on task-specific models trained on narrow, single source-target pairs, limiting their generalizability. Here we introduce MuPD (Multimodal Pathology Diffusion), a generative foundation model that embeds hematoxylin and eosin (H&E)-stained histology, molecular RNA profiles, and clinical text into a shared latent space through a diffusion transformer with decoupled cross-modal attention. Pretrained on 100 million histology image patches, 1.6 million text-histology pairs, and 10.8 million RNA-histology pairs spanning 34 human organs, MuPD supports diverse cross-modal synthesis tasks with minimal or no task-specific fine-tuning. For text-conditioned and image-to-image generation, MuPD synthesizes histologically faithful tissue architectures, reducing Fr\\'echet inception distance (FID) scores by 50% relative to domain-specific models and improving few-shot classification accuracy by up to 47% through synthetic data augmentation. For RNA-conditioned histology generation, MuPD reduces FID by 23% compared with the next-best method while preserving cell-type distributions across five cancer types. As a virtual stainer, MuPD translates H&E images to immunohistochemistry and multiplex immunofluorescence, improving average marker correlation by 37% over existing approaches. These results demonstrate that a single, unified generative model pretrained across heterogeneous pathology modalities can substantially outperform specialized alternatives, providing a scalable computational framework for multimodal histopathology.", "date": "2026-04-04", "categories": []}
{"id": "2604.04239", "title": "Good Rankings, Wrong Probabilities: A Calibration Audit of Multimodal Cancer Survival Models", "abstract": "Multimodal deep learning models that fuse whole-slide histopathology images with genomic data have achieved strong discriminative performance for cancer survival prediction, as measured by the concordance index. Yet whether the survival probabilities derived from these models - either directly from native outputs or via standard post-hoc reconstruction - are calibrated remains largely unexamined.\n  We conduct, to our knowledge, the first systematic fold-level 1-calibration audit of multimodal WSI-genomics survival architectures, evaluating native discrete-time survival outputs (Experiment A: 3 models on TCGA-BRCA) and Breslow-reconstructed survival curves from scalar risk scores (Experiment B: 11 architectures across 5 TCGA cancer types). In Experiment A, all three models fail 1-calibration on a majority of folds (12 of 15 fold-level tests reject after Benjamini-Hochberg correction). Across the full 290 fold-level tests, 166 reject the null of correct calibration at the median event time after Benjamini-Hochberg correction (FDR = 0.05). MCAT achieves C-index 0.817 on GBMLGG yet fails 1-calibration on all five folds.\n  Gating-based fusion is associated with better calibration; bilinear and concatenation fusion are not. Post-hoc Platt scaling reduces miscalibration at the evaluated horizon (e.g., MCAT: 5/5 folds failing to 2/5) without affecting discrimination. The concordance index alone is insufficient for evaluating survival models intended for clinical use.", "date": "2026-04-05", "categories": []}
{"id": "2603.27048", "title": "MOOZY: A Patient-First Foundation Model for Computational Pathology", "abstract": "Computational pathology needs whole-slide image (WSI) foundation models that transfer across diverse clinical tasks, yet current approaches remain largely slide-centric, often depend on private data and expensive paired-report supervision, and do not explicitly model relationships among multiple slides from the same patient. We present MOOZY, a patient-first pathology foundation model in which the patient case, not the individual slide, is the core unit of representation. MOOZY explicitly models dependencies across all slides from the same patient via a case transformer during pretraining, combining multi-stage open self-supervision with scaled low-cost task supervision. In Stage 1, we pretrain a vision-only slide encoder on 77,134 public slide feature grids using masked self-distillation. In Stage 2, we align these representations with clinical semantics using a case transformer and multi-task supervision over 333 tasks from 56 public datasets, including 205 classification and 128 survival tasks across four endpoints. Across eight held-out tasks with five-fold frozen-feature probe evaluation, MOOZY achieves best or tied-best performance on most metrics and improves macro averages over TITAN by +7.37%, +5.50%, and +7.83% and over PRISM by +8.83%, +10.70%, and +9.78% for weighted F1, weighted ROC-AUC, and balanced accuracy, respectively. MOOZY is also parameter efficient with 85.77M parameters, 14x smaller than GigaPath. These results demonstrate that open, reproducible patient-level pretraining yields transferable embeddings, providing a practical path toward scalable patient-first histopathology foundation models.", "date": "2026-03-27", "categories": []}
{"id": "2601.11691", "title": "Explainable histomorphology-based survival prediction of glioblastoma, IDH-wildtype", "abstract": "Glioblastoma, IDH-wildtype (GBM-IDHwt) is the most common malignant brain tumor. While histomorphology is a crucial component of GBM-IDHwt diagnosis, it is not further considered for prognosis. Here, we present an explainable artificial intelligence (AI) framework to identify and interpret histomorphological features associated with patient survival. The framework combines an explainable multiple instance learning (MIL) architecture that directly identifies prognostically relevant image tiles with a sparse autoencoder (SAE) that maps these tiles to interpretable visual patterns. The MIL model was trained and evaluated on a new real-world dataset of 720 GBM-IDHwt cases from three hospitals and four cancer registries across Germany. The SAE was trained on 1,878 whole-slide images from five independent public glioblastoma collections. Despite the many factors influencing survival time, our method showed some ability to discriminate between patients living less than 180 days or more than 360 days solely based on histomorphology (AUC: 0.67; 95% CI: 0.63-0.72). Cox proportional hazards regression confirmed a significant survival difference between predicted groups after adjustment for established prognostic factors (hazard ratio: 1.47; 95% CI: 1.26-1.72). Three neuropathologists categorized the identified visual patterns into seven distinct histomorphological groups, revealing both established prognostic features and unexpected associations, the latter being potentially attributable to surgery-related confounders. The presented explainable AI framework facilitates prognostic biomarker discovery in GBM-IDHwt and beyond, highlighting promising histomorphological features for further analysis and exposing potential confounders that would be hidden in black-box models.", "date": "2026-01-16", "categories": []}
{"id": "2603.26809", "title": "Dictionary-based Pathology Mining with Hard-instance-assisted Classifier Debiasing for Genetic Biomarker Prediction from WSIs", "abstract": "Prediction of genetic biomarkers, e.g., microsatellite instability in colorectal cancer is crucial for clinical decision making. But, two primary challenges hamper accurate prediction: (1) It is difficult to construct a pathology-aware representation involving the complex interconnections among pathological components. (2) WSIs contain a large proportion of areas unrelated to genetic biomarkers, which make the model easily overfit simple but irrelative instances. We hereby propose a Dictionary-based hierarchical pathology mining with hard-instance-assisted classifier Debiasing framework to address these challenges, dubbed as D2Bio. Our first module, dictionary-based hierarchical pathology mining, is able to mine diverse and very fine-grained pathological contextual interaction without the limit to the distances between patches. The second module, hard-instance-assisted classfier debiasing, learns a debiased classifier via focusing on hard but task-related features, without any additional annotations. Experimental results on five cohorts show the superiority of our method, with over 4% improvement in AUROC compared with the second best on the TCGA-CRC-MSI cohort. Our analysis further shows the clinical interpretability of D2Bio in genetic biomarker diagnosis and potential clinical utility in survival analysis. Code will be available at https://github.com/DeepMed-Lab-ECNU/D2Bio.", "date": "2026-03-26", "categories": []}
{"id": "2603.27625", "title": "Clore: Interactive Pathology Image Segmentation with Click-based Local Refinement", "abstract": "Recent advancements in deep learning-based interactive segmentation methods have significantly improved pathology image segmentation. Most existing approaches utilize user-provided positive and negative clicks to guide the segmentation process. However, these methods primarily rely on iterative global updates for refinement, which lead to redundant re-prediction and often fail to capture fine-grained structures or correct subtle errors during localized adjustments. To address this limitation, we propose the Click-based Local Refinement (Clore) pipeline, a simple yet efficient method designed to enhance interactive segmentation. The key innovation of Clore lies in its hierarchical interaction paradigm: the initial clicks drive global segmentation to rapidly outline large target regions, while subsequent clicks progressively refine local details to achieve precise boundaries. This approach not only improves the ability to handle fine-grained segmentation tasks but also achieves high-quality results with fewer interactions. Experimental results on four datasets demonstrate that Clore achieves the best balance between segmentation accuracy and interaction cost, making it an effective solution for efficient and accurate interactive pathology image segmentation.", "date": "2026-03-29", "categories": []}
{"id": "2506.20964", "title": "Evidence-based diagnostic reasoning with multi-agent copilot for human pathology", "abstract": "Pathology is experiencing rapid digital transformation driven by whole-slide imaging and artificial intelligence (AI). While deep learning-based computational pathology has achieved notable success, traditional models primarily focus on image analysis without integrating natural language instruction or rich, text-based context. Current multimodal large language models (MLLMs) in computational pathology face limitations, including insufficient training data, inadequate support and evaluation for multi-image understanding, and a lack of autonomous, diagnostic reasoning capabilities. To address these limitations, we introduce PathChat+, a new MLLM specifically designed for human pathology, trained on over 1 million diverse, pathology-specific instruction samples and nearly 5.5 million question answer turns. Extensive evaluations across diverse pathology benchmarks demonstrated that PathChat+ substantially outperforms the prior PathChat copilot, as well as both state-of-the-art (SOTA) general-purpose and other pathology-specific models. Furthermore, we present SlideSeek, a reasoning-enabled multi-agent AI system leveraging PathChat+ to autonomously evaluate gigapixel whole-slide images (WSIs) through iterative, hierarchical diagnostic reasoning, reaching high accuracy on DDxBench, a challenging open-ended differential diagnosis benchmark, while also capable of generating visually grounded, humanly-interpretable summary reports.", "date": "2025-06-26", "categories": []}
{"id": "2603.25802", "title": "LEMON: a foundation model for nuclear morphology in Computational Pathology", "abstract": "Computational pathology relies on effective representation learning to support cancer research and precision medicine. Although self-supervised learning has driven major progress at the patch and whole-slide image levels, representation learning at the single-cell level remains comparatively underexplored, despite its importance for characterizing cell types and cellular phenotypes. We introduce LEMON (Learning Embeddings from Morphology Of Nuclei), a self-supervised foundation model for scalable single-cell image representation learning. Trained on millions of cell images from diverse tissues and cancer types, LEMON learns robust and versatile morphological representations that support large-scale single-cell analyses in pathology. We evaluate LEMON on five benchmark datasets across a range of prediction tasks and show that it provides strong performance, highlighting its potential as a new paradigm for cell-level computational pathology. Model weights are available at https://huggingface.co/aliceblondel/LEMON.", "date": "2026-03-26", "categories": []}
{"id": "2603.23067", "title": "MLLM-HWSI: A Multimodal Large Language Model for Hierarchical Whole Slide Image Understanding", "abstract": "Whole Slide Images (WSIs) exhibit hierarchical structure, where diagnostic information emerges from cellular morphology, regional tissue organization, and global context. Existing Computational Pathology (CPath) Multimodal Large Language Models (MLLMs) typically compress an entire WSI into a single embedding, which hinders fine-grained grounding and ignores how pathologists synthesize evidence across different scales. We introduce \\textbf{MLLM-HWSI}, a Hierarchical WSI-level MLLM that aligns visual features with pathology language at four distinct scales, cell as word, patch as phrase, region as sentence, and WSI as paragraph to support interpretable evidence-grounded reasoning. MLLM-HWSI decomposes each WSI into multi-scale embeddings with scale-specific projectors and jointly enforces (i) a hierarchical contrastive objective and (ii) a cross-scale consistency loss, preserving semantic coherence from cells to the WSI. We compute diagnostically relevant patches and aggregate segmented cell embeddings into a compact cellular token per-patch using a lightweight \\textit{Cell-Cell Attention Fusion (CCAF)} transformer. The projected multi-scale tokens are fused with text tokens and fed to an instruction-tuned LLM for open-ended reasoning, VQA, report, and caption generation tasks. Trained in three stages, MLLM-HWSI achieves new SOTA results on 13 WSI-level benchmarks across six CPath tasks. By aligning language with multi-scale visual evidence, MLLM-HWSI provides accurate, interpretable outputs that mirror diagnostic workflows and advance holistic WSI understanding. Code is available at: \\href{https://github.com/BasitAlawode/HWSI-MLLM}{GitHub}.", "date": "2026-03-24", "categories": []}
{"id": "2602.21395", "title": "Momentum Memory for Knowledge Distillation in Computational Pathology", "abstract": "Multimodal learning that integrates genomics and histopathology has shown strong potential in cancer diagnosis, yet its clinical translation is hindered by the limited availability of paired histology-genomics data. Knowledge distillation (KD) offers a practical solution by transferring genomic supervision into histopathology models, enabling accurate inference using histology alone. However, existing KD methods rely on batch-local alignment, which introduces instability due to limited within-batch comparisons and ultimately degrades performance.\n  To address these limitations, we propose Momentum Memory Knowledge Distillation (MoMKD), a cross-modal distillation framework driven by a momentum-updated memory. This memory aggregates genomic and histopathology information across batches, effectively enlarging the supervisory context available to each mini-batch. Furthermore, we decouple the gradients of the genomics and histology branches, preventing genomic signals from dominating histology feature learning during training and eliminating the modality-gap issue at inference time.\n  Extensive experiments on the TCGA-BRCA benchmark (HER2, PR, and ODX classification tasks) and an independent in-house testing dataset demonstrate that MoMKD consistently outperforms state-of-the-art MIL and multimodal KD baselines, delivering strong performance and generalization under histology-only inference. Overall, MoMKD establishes a robust and generalizable knowledge distillation paradigm for computational pathology.", "date": "2026-02-24", "categories": []}
{"id": "2603.04348", "title": "RANGER: Sparsely-Gated Mixture-of-Experts with Adaptive Retrieval Re-ranking for Pathology Report Generation", "abstract": "Pathology report generation remains a relatively under-explored downstream task, primarily due to the gigapixel scale and complex morphological heterogeneity of Whole Slide Images (WSIs). Existing pathology report generation frameworks typically employ transformer architectures, relying on a homogeneous decoder architecture and static knowledge retrieval integration. Such architectures limit generative specialization and may introduce noisy external guidance during the report generation process. To address these limitations, we propose RANGER, a sparsely-gated Mixture-of-Experts (MoE) framework with adaptive retrieval re-ranking for pathology report generation. Specifically, we integrate a sparsely gated MoE into the decoder, along with noisy top-$k$ routing and load-balancing regularization, to enable dynamic expert specialization across various diagnostic patterns. Additionally, we introduce an adaptive retrieval re-ranking module that selectively refines retrieved memory from a knowledge base before integration, reducing noise and improving semantic alignment based on visual feature representations. We perform extensive experiments on the PathText-BRCA dataset and demonstrate consistent improvements over existing approaches across standard natural language generation metrics. Our full RANGER model achieves optimal performance on PathText dataset, reaching BLEU-1 to BLEU-4 scores of 0.4598, 0.3044, 0.2036, and 0.1435, respectively, with METEOR of 0.1883, and ROUGE-L of 0.3038, validating the effectiveness of dynamic expert routing and adaptive knowledge refinement for semantically grounded pathology report generation.", "date": "2026-03-04", "categories": []}
{"id": "2603.20383", "title": "Multi-Stage Fine-Tuning of Pathology Foundation Models with Head-Diverse Ensembling for White Blood Cell Classification", "abstract": "The classification of white blood cells (WBCs) from peripheral blood smears is critical for the diagnosis of leukemia. However, automated approaches still struggle due to challenges including class imbalance, domain shift, and morphological continuum confusion, where adjacent maturation stages exhibit subtle, overlapping features. We present a multi-stage fine-tuning methodology for 13-class WBC classification in the WBCBench 2026 Challenge (ISBI 2026). Our best-performing model is a fine-tuned DINOBloom-base, on which we train multiple classifier head families (linear, cosine, and multilayer perceptron (MLP)). The cosine head performed best on the mature granulocyte boundary (Band neutrophil (BNE) F1 = 0.470), the linear head on more immature granulocyte classes (Metamyelocyte (MMY) F1 = 0.585), and the MLP head on the most immature granulocyte (Promyelocyte (PMY) F1 = 0.733), revealing class-specific specialization. Based on this specialization, we construct a head-diverse ensemble, where the MLP head acts as the primary predictor, and its predictions within the four predefined confusion pairs are replaced only when two other head families agree. We further show that cases consistently misclassified by all models are substantially enriched for probable labeling errors or inherent morphological ambiguity.", "date": "2026-03-20", "categories": []}
{"id": "2603.20386", "title": "Jigsaw Regularization in Whole-Slide Image Classification", "abstract": "Computational pathology involves the digitization of stained tissues into whole-slide images (WSIs) that contain billions of pixels arranged as contiguous patches. Statistical analysis of WSIs largely focuses on classification via multiple instance learning (MIL), in which slide-level labels are inferred from unlabeled patches. Most MIL methods treat patches as exchangeable, overlooking the rich spatial and topological structure that underlies tissue images. This work builds on recent graph-based methods that aim to incorporate spatial awareness into MIL. Our approach is new in two regards: (1) we deploy vision \\emph{foundation-model embeddings} to incorporate local spatial structure within each patch, and (2) achieve across-patch spatial awareness using graph neural networks together with a novel {\\em jigsaw regularization}. We find that a combination of these two features markedly improves classification over state-of-the-art attention-based MIL approaches on benchmark datasets in breast, head-and-neck, and colon cancer.", "date": "2026-03-20", "categories": []}
{"id": "2603.20848", "title": "GOLDMARK: Governed Outcome-Linked Diagnostic Model Assessment Reference Kit", "abstract": "Computational biomarkers (CBs) are histopathology-derived patterns extracted from hematoxylin-eosin (H&E) whole-slide images (WSIs) using artificial intelligence (AI) to predict therapeutic response or prognosis. Recently, slide-level multiple-instance learning (MIL) with pathology foundation models (PFMs) has become the standard baseline for CB development. While these methods have improved predictive performance, computational pathology lacks standardized intermediate data formats, provenance tracking, checkpointing conventions, and reproducible evaluation metrics required for clinical-grade deployment.\n  We introduce GOLDMARK (https://artificialintelligencepathology.org), a standardized benchmarking framework built on a curated TCGA cohort with clinically actionable OncoKB level 1-3 biomarker labels. GOLDMARK releases structured intermediate representations, including tile coordinate maps, per-slide feature embeddings from canonical PFMs, quality-control metadata, predefined patient-level splits, trained slide-level models, and evaluation outputs. Models are trained on TCGA and evaluated on an independent MSKCC cohort with reciprocal testing.\n  Across 33 tumor-biomarker tasks, mean AUROC was 0.689 (TCGA) and 0.630 (MSKCC). Restricting to the eight highest-performing tasks yielded mean AUROCs of 0.831 and 0.801, respectively. These tasks correspond to established morphologic-genomic associations (e.g., LGG IDH1, COAD MSI/BRAF, THCA BRAF/NRAS, BLCA FGFR3, UCEC PTEN) and showed the most stable cross-site performance. Differences between canonical encoders were modest relative to task-specific variability.\n  GOLDMARK establishes a shared experimental substrate for computational pathology, enabling reproducible benchmarking and direct comparison of methods across datasets and models.", "date": "2026-03-21", "categories": []}
{"id": "2603.21504", "title": "Parameter-efficient Prompt Tuning and Hierarchical Textual Guidance for Few-shot Whole Slide Image Classification", "abstract": "Whole Slide Images (WSIs) are giga-pixel in scale and are typically partitioned into small instances in WSI classification pipelines for computational feasibility. However, obtaining extensive instance level annotations is costly, making few-shot weakly supervised WSI classification (FSWC) crucial for learning from limited slide-level labels. Recently, pre-trained vision-language models (VLMs) have been adopted in FSWC, yet they exhibit several limitations. Existing prompt tuning methods in FSWC substantially increase both the number of trainable parameters and inference overhead. Moreover, current methods discard instances with low alignment to text embeddings from VLMs, potentially leading to information loss. To address these challenges, we propose two key contributions. First, we introduce a new parameter efficient prompt tuning method by scaling and shifting features in text encoder, which significantly reduces the computational cost. Second, to leverage not only the pre-trained knowledge of VLMs, but also the inherent hierarchical structure of WSIs, we introduce a WSI representation learning approach with a soft hierarchical textual guidance strategy without utilizing hard instance filtering. Comprehensive evaluations on pathology datasets covering breast, lung, and ovarian cancer types demonstrate consistent improvements up-to 10.9%, 7.8%, and 13.8% respectively, over the state-of-the-art methods in FSWC. Our method reduces the number of trainable parameters by 18.1% on both breast and lung cancer datasets, and 5.8% on the ovarian cancer dataset, while also excelling at weakly-supervised tumor localization. Code at https://github.com/Jayanie/HIPSS.", "date": "2026-03-23", "categories": []}
{"id": "2603.22198", "title": "Mixture of Mini Experts: Overcoming the Linear Layer Bottleneck in Multiple Instance Learning", "abstract": "Multiple Instance Learning (MIL) is the predominant framework for classifying gigapixel whole-slide images in computational pathology. MIL follows a sequence of 1) extracting patch features, 2) applying a linear layer to obtain task-specific patch features, and 3) aggregating the patches into a slide feature for classification. While substantial efforts have been devoted to optimizing patch feature extraction and aggregation, none have yet addressed the second point, the critical layer which transforms general-purpose features into task-specific features. We hypothesize that this layer constitutes an overlooked performance bottleneck and that stronger representations can be achieved with a low-rank transformation tailored to each patch's phenotype, yielding synergistic effects with any of the existing MIL approaches. To this end, we introduce MAMMOTH, a parameter-efficient, multi-head mixture of experts module designed to improve the performance of any MIL model with minimal alterations to the total number of parameters. Across eight MIL methods and 19 different classification tasks, we find that such task-specific transformation has a larger effect on performance than the choice of aggregation method. For instance, when equipped with MAMMOTH, even simple methods such as max or mean pooling attain higher average performance than any method with the standard linear layer. Overall, MAMMOTH improves performance in 130 of the 152 examined configurations, with an average $+3.8\\%$ change in performance. Code is available at https://github.com/mahmoodlab/mammoth.", "date": "2026-03-23", "categories": []}
{"id": "2603.19925", "title": "ReconMIL: Synergizing Latent Space Reconstruction with Bi-Stream Mamba for Whole Slide Image Analysis", "abstract": "Whole slide image (WSI) analysis heavily relies on multiple instance learning (MIL). While recent methods benefit from large-scale foundation models and advanced sequence modeling to capture long-range dependencies, they still struggle with two critical issues. First, directly applying frozen, task-agnostic features often leads to suboptimal separability due to the domain gap with specific histological tasks. Second, relying solely on global aggregators can cause over-smoothing, where sparse but critical diagnostic signals are overshadowed by the dominant background context. In this paper, we present ReconMIL, a novel framework designed to bridge this domain gap and balance global-local feature aggregation. Our approach introduces a Latent Space Reconstruction module that adaptively projects generic features into a compact, task-specific manifold, improving boundary delineation. To prevent information dilution, we develop a bi-stream architecture combining a Mamba-based global stream for contextual priors and a CNN-based local stream to preserve subtle morphological anomalies. A scale-adaptive selection mechanism dynamically fuses these two streams, determining when to rely on overall architecture versus local saliency. Evaluations across multiple diagnostic and survival prediction benchmarks show that ReconMIL consistently outperforms current state-of-the-art methods, effectively localizing fine-grained diagnostic regions while suppressing background noise. Visualization results confirm the models superior ability to localize diagnostic regions by effectively balancing global structure and local granularity.", "date": "2026-03-20", "categories": []}
{"id": "2603.19957", "title": "HiPath: Hierarchical Vision-Language Alignment for Structured Pathology Report Prediction", "abstract": "Pathology reports are structured, multi-granular documents encoding diagnostic conclusions, histological grades, and ancillary test results across one or more anatomical sites; yet existing pathology vision-language models (VLMs) reduce this output to a flat label or free-form text. We present HiPath, a lightweight VLM framework built on frozen UNI2 and Qwen3 backbones that treats structured report prediction as its primary training objective. Three trainable modules totalling 15M parameters address complementary aspects of the problem: a Hierarchical Patch Aggregator (HiPA) for multi-image visual encoding, Hierarchical Contrastive Learning (HiCL) for cross-modal alignment via optimal transport, and Slot-based Masked Diagnosis Prediction (Slot-MDP) for structured diagnosis generation. Trained on 749K real-world Chinese pathology cases from three hospitals, HiPath achieves 68.9% strict and 74.7% clinically acceptable accuracy with a 97.3% safety rate, outperforming all baselines under the same frozen backbone. Cross-hospital evaluation confirms generalisation with only a 3.4pp drop in strict accuracy while maintaining 97.1% safety.", "date": "2026-03-20", "categories": []}
{"id": "2505.10294", "title": "MIPHEI-ViT: Multiplex Immunofluorescence Prediction from H&E Images using ViT Foundation Models", "abstract": "Histopathological analysis is a cornerstone of cancer diagnosis, with Hematoxylin and Eosin (H&E) staining routinely acquired for every patient to visualize cell morphology and tissue architecture. On the other hand, multiplex immunofluorescence (mIF) enables more precise cell type identification via proteomic markers, but has yet to achieve widespread clinical adoption due to cost and logistical constraints. To bridge this gap, we introduce MIPHEI (Multiplex Immunofluorescence Prediction from H&E Images), a U-Net-inspired architecture that leverages a ViT pathology foundation model as encoder to predict mIF signals from H&E images using rich pretrained representations. MIPHEI targets a comprehensive panel of markers spanning nuclear content, immune lineages (T cells, B cells, myeloid), epithelium, stroma, vasculature, and proliferation. We train our model using the publicly available OrionCRC dataset of restained H&E and mIF images from colorectal cancer tissue, and validate it on five independent datasets: HEMIT, PathoCell, IMMUcan, Lizard and PanNuke. On OrionCRC test set, MIPHEI achieves accurate cell-type classification from H&E alone, with F1 scores of 0.93 for Pan-CK, 0.83 for alpha-SMA, 0.68 for CD3e, 0.36 for CD20, and 0.28 for CD68, substantially outperforming both a state-of-the-art baseline and a random classifier for most markers. Our results indicate that, for some molecular markers, our model captures the complex relationships between nuclear morphologies in their tissue context, as visible in H&E images and molecular markers defining specific cell types. MIPHEI offers a promising step toward enabling cell-type-aware analysis of large-scale H&E datasets, in view of uncovering relationships between spatial cellular organization and patient outcomes.", "date": "2025-05-15", "categories": []}
{"id": "2512.14640", "title": "A Multicenter Benchmark of Multiple Instance Learning Models for Lymphoma Subtyping from HE-stained Whole Slide Images", "abstract": "Timely and accurate lymphoma diagnosis is essential for guiding cancer treatment. Standard diagnostic practice combines hematoxylin and eosin (HE)-stained whole slide images with immunohistochemistry, flow cytometry, and molecular genetic tests to determine lymphoma subtypes, a process requiring costly equipment, and skilled personnel, causing treatment delays. Deep learning methods could assist pathologists by extracting diagnostic information from routinely available HE-stained slides directly, yet comprehensive benchmarks for lymphoma subtyping on multicenter data are lacking. In this work, we present the first multicenter lymphoma benchmark, covering four common lymphoma subtypes and healthy control tissue. We systematically evaluate five publicly available pathology foundation models (H-optimus-1, H0-mini, Virchow2, UNI2, Titan) combined with attention-based (AB-MIL) and transformer-based (TransMIL) multiple instance learning aggregators across three magnifications (10x, 20x, 40x). On in-distribution test sets, models achieve multiclass balanced accuracies exceeding 80% across all magnifications, with foundation models performing similarly, and aggregation methods showing comparable results. The magnification study reveals that 40x resolution is sufficient, with no performance gains from higher resolutions or cross-magnification aggregation. However, on out-of-distribution test sets, performance drops substantially to around 60%, highlighting significant generalization challenges. To advance the field, larger multicenter studies covering additional rare lymphoma subtypes are needed. We provide an automated benchmarking pipeline to facilitate such future research. Our paper codes is publicly available at https://github.com/RaoUmer/LymphomaMIL.", "date": "2025-12-16", "categories": []}
{"id": "2603.18461", "title": "Cell-Type Prototype-Informed Neural Network for Gene Expression Estimation from Pathology Images", "abstract": "Estimating slide- and patch-level gene expression profiles from pathology images enables rapid and low-cost molecular analysis with broad clinical impact. Despite strong results, existing approaches treat gene expression as a mere slide- or spot-level signal and do not incorporate the fact that the measured expression arises from the aggregation of underlying cell-level expression. To explicitly introduce this missing cell-resolved guidance, we propose a Cell-type Prototype-informed Neural Network (CPNN) that leverages publicly available single-cell RNA-sequencing datasets. Since single-cell measurements are noisy and not paired with histology images, we first estimate cell-type prototypes-mean expression profiles that reflect stable gene-gene co-variation patterns.CPNN then learns cell-type compositional weights directly from images and models the relationship between prototypes and observed bulk or spatial expression, providing a biologically grounded and structurally regularized prediction framework. We evaluate CPNN on three slide-level datasets and three patch-level spatial transcriptomics datasets. Across all settings, CPNN achieves the highest performance in terms of Spearman correlation. Moreover, by visualizing the inferred compositional weights, our framework provides interpretable insights into which cell types drive the predicted expression. Code is publicly available at https://github.com/naivete5656/CPNN.", "date": "2026-03-19", "categories": []}
{"id": "2603.18513", "title": "CAFlow: Adaptive-Depth Single-Step Flow Matching for Efficient Histopathology Super-Resolution", "abstract": "In digital pathology, whole-slide images routinely exceed gigapixel resolution, making computationally intensive generative super-resolution (SR) impractical for routine deployment. We introduce CAFlow, an adaptive-depth single-step flow-matching framework that routes each image tile to the shallowest network exit that preserves reconstruction quality. CAFlow performs flow matching in pixel-unshuffled rearranged space, reducing spatial computation by 16x while enabling direct inference. We show that dedicating half of training to exact t=0 samples is essential for single-step quality (-1.5 dB without it). The backbone, FlowResNet (1.90M parameters), mixes convolution and window self-attention blocks across four early exits spanning 3.1 to 13.3 GFLOPs. A lightweight exit classifier (~6K parameters) achieves 33% compute savings at only 0.12 dB cost. On multi-organ histopathology x4 SR, adaptive routing achieves 31.72 dB PSNR versus 31.84 dB at full depth, while the shallowest exit exceeds bicubic by +1.9 dB at 2.8x less compute than SwinIR-light. The method generalizes to held-out colon tissue with minimal quality loss (-0.02 dB), and at x8 upscaling it outperforms all comparable-compute baselines while remaining competitive with the much larger SwinIR-Medium model. Downstream nuclei segmentation confirms preservation of clinically relevant structure. The model trains in under 5 hours on a single GPU, and adaptive routing can reduce whole-slide inference from minutes to seconds.", "date": "2026-03-19", "categories": []}
{"id": "2603.18660", "title": "Multimodal Model for Computational Pathology:Representation Learning and Image Compression", "abstract": "Whole slide imaging (WSI) has transformed digital pathology by enabling computational analysis of gigapixel histopathology images. Recent foundation model advances have accelerated progress in computational pathology, facilitating joint reasoning across pathology images, clinical reports, and structured data. Despite this progress, challenges remain: the extreme resolution of WSIs creates computational hurdles for visual learning; limited expert annotations constrain supervised approaches; integrating multimodal information while preserving biological interpretability remains difficult; and the opacity of modeling ultra-long visual sequences hinders clinical transparency. This review comprehensively surveys recent advances in multimodal computational pathology. We systematically analyze four research directions: (1) self-supervised representation learning and structure-aware token compression for WSIs; (2) multimodal data generation and augmentation; (3) parameter-efficient adaptation and reasoning-enhanced few-shot learning; and (4) multi-agent collaborative reasoning for trustworthy diagnosis. We specifically examine how token compression enables cross-scale modeling and how multi-agent mechanisms simulate a pathologist's \"Chain of Thought\" across magnifications to achieve uncertainty-aware evidence fusion. Finally, we discuss open challenges and argue that future progress depends on unified multimodal frameworks integrating high-resolution visual data with clinical and biomedical knowledge to support interpretable and safe AI-assisted diagnosis.", "date": "2026-03-19", "categories": []}
{"id": "2603.15967", "title": "A Comprehensive Benchmark of Histopathology Foundation Models for Kidney Digital Pathology Images", "abstract": "Histopathology foundation models (HFMs), pretrained on large-scale cancer datasets, have advanced computational pathology. However, their applicability to non-cancerous chronic kidney disease remains underexplored, despite coexistence of renal pathology with malignancies such as renal cell and urothelial carcinoma. We systematically evaluate 11 publicly available HFMs across 11 kidney-specific downstream tasks spanning multiple stains (PAS, H&E, PASM, and IHC), spatial scales (tile and slide-level), task types (classification, regression, and copy detection), and clinical objectives, including detection, diagnosis, and prognosis. Tile-level performance is assessed using repeated stratified group cross-validation, while slide-level tasks are evaluated using repeated nested stratified cross-validation. Statistical significance is examined using Friedman test followed by pairwise Wilcoxon signed-rank testing with Holm-Bonferroni correction and compact letter display visualization. To promote reproducibility, we release an open-source Python package, kidney-hfm-eval, available at https://pypi.org/project/kidney-hfm-eval/ , that reproduces the evaluation pipelines. Results show moderate to strong performance on tasks driven by coarse meso-scale renal morphology, including diagnostic classification and detection of prominent structural alterations. In contrast, performance consistently declines for tasks requiring fine-grained microstructural discrimination, complex biological phenotypes, or slide-level prognostic inference, largely independent of stain type. Overall, current HFMs appear to encode predominantly static meso-scale representations and may have limited capacity to capture subtle renal pathology or prognosis-related signals. Our results highlight the need for kidney-specific, multi-stain, and multimodal foundation models to support clinically reliable decision-making in nephrology.", "date": "2026-03-16", "categories": []}
{"id": "2603.16421", "title": "HGP-Mamba: Integrating Histology and Generated Protein Features for Mamba-based Multimodal Survival Risk Prediction", "abstract": "Recent advances in multimodal learning have significantly improved cancer survival risk prediction. However, the joint prognostic potential of protein markers and histopathology images remains underexplored, largely due to the high cost and limited availability of protein expression profiling. To address this challenge, we propose HGP-Mamba, a Mamba-based multimodal framework that efficiently integrates histological with generated protein features for survival risk prediction. Specifically, we introduce a protein feature extractor (PFE) that leverages pretrained foundation models to derive high-throughput protein embeddings directly from Whole Slide Images (WSIs), enabling data-efficient incorporation of molecular information. Together with histology embeddings that capture morphological patterns, we further introduce the Local Interaction-aware Mamba (LiAM) for fine-grained feature interaction and the Global Interaction-enhanced Mamba (GiEM) to promote holistic modality fusion at the slide level, thus capture complex cross-modal dependencies. Experiments on four public cancer datasets demonstrate that HGP-Mamba achieves state-of-the-art performance while maintaining superior computational efficiency compared with existing methods. Our source code is publicly available at https://github.com/Daijing-ai/HGP-Mamba.git.", "date": "2026-03-17", "categories": []}
{"id": "2602.21637", "title": "CARE: A Molecular-Guided Foundation Model with Adaptive Region Modeling for Whole Slide Image Analysis", "abstract": "Foundation models have recently achieved impressive success in computational pathology, demonstrating strong generalization across diverse histopathology tasks. However, existing models overlook the heterogeneous and non-uniform organization of pathological regions of interest (ROIs) because they rely on natural image backbones not tailored for tissue morphology. Consequently, they often fail to capture the coherent tissue architecture beyond isolated patches, limiting interpretability and clinical relevance. To address these challenges, we present Cross-modal Adaptive Region Encoder (CARE), a foundation model for pathology that automatically partitions WSIs into several morphologically relevant regions. Specifically, CARE employs a two-stage pretraining strategy: (1) a self-supervised unimodal pretraining stage that learns morphological representations from 34,277 whole-slide images (WSIs) without segmentation annotations, and (2) a cross-modal alignment stage that leverages RNA and protein profiles to refine the construction and representation of adaptive regions. This molecular guidance enables CARE to identify biologically relevant patterns and generate irregular yet coherent tissue regions, selecting the most representative area as ROI. CARE supports a broad range of pathology-related tasks, using either the ROI feature or the slide-level feature obtained by aggregating adaptive regions. Based on only one-tenth of the pretraining data typically used by mainstream foundation models, CARE achieves superior average performance across 33 downstream benchmarks, including morphological classification, molecular prediction, and survival analysis, and outperforms other foundation model baselines overall.", "date": "2026-02-25", "categories": []}
{"id": "2603.15774", "title": "Domain Adaptation Without the Compute Burden for Efficient Whole Slide Image Analysis", "abstract": "Computational methods on analyzing Whole Slide Images (WSIs) enable early diagnosis and treatments by supporting pathologists in detection and classification of tumors. However, the extremely high resolution of WSIs makes end-to-end training impractical compared to typical image analysis tasks. To address this, most approaches use pre-trained feature extractors to obtain fixed representations of whole slides, which are then combined with Multiple Instance Learning (MIL) for downstream tasks. These feature extractors are typically pre-trained on natural image datasets such as ImageNet, which fail to capture domain-specific characteristics. Although domain-specific pre-training on histopathology data yields more relevant feature representations, it remains computationally expensive and fail to capture task-specific characteristics within the domain. To address the computational cost and lack of task-specificity in domain-specific pre-training, we propose EfficientWSI (eWSI), a careful integration of Parameter-Efficient-Fine-Tuning (PEFT) and Multiple Instance Learning (MIL) that enables end-to-end training on WSI tasks. We evaluate eWSI on seven WSI-level tasks over Camelyon16, TCGA and BRACS datasets. Our results show that eWSI when applied with ImageNet feature extractors yields strong classification performance, matching or outperforming MILs with in-domain feature extractors, alleviating the need for extensive in-domain pre-training. Furthermore, when eWSI is applied with in-domain feature extractors, it further improves classification performance in most cases, demonstrating its ability to capture task-specific information where beneficial. Our findings suggest that eWSI provides a task-targeted, computationally efficient path for WSI tasks, offering a promising direction for task-specific learning in computational pathology.", "date": "2026-03-16", "categories": []}
{"id": "2603.16160", "title": "Segmentation-before-Staining Improves Structural Fidelity in Virtual IHC-to-Multiplex IF Translation", "abstract": "Multiplex immunofluorescence (mIF) enables simultaneous single-cell quantification of multiple biomarkers within intact tissue architecture, yet its high reagent cost, multi-round staining protocols, and need for specialized imaging platforms limit routine clinical adoption. Virtual staining can synthesize mIF channels from widely available brightfield immunohistochemistry (IHC), but current translators optimize pixel-level fidelity without explicitly constraining nuclear morphology. In pathology, this gap is clinically consequential: subtle distortions in nuclei count, shape, or spatial arrangement propagate directly to quantification endpoints such as the Ki67 proliferation index, where errors of a few percent can shift treatment-relevant risk categories. This work introduces a supervision-free, architecture-agnostic conditioning strategy that injects a continuous cell probability map from a pretrained nuclei segmentation foundation model as an explicit input prior, together with a variance-preserving regularization term that matches local intensity statistics to maintain cell-level heterogeneity in synthesized fluorescence channels. The soft prior retains gradient-level boundary information lost by binary thresholding, providing a richer conditioning signal without task-specific tuning. Controlled experiments across Pix2Pix with U-Net and ResNet generators, deterministic regression U-Net, and conditional diffusion on two independent datasets demonstrate consistent improvements in nuclei count fidelity and perceptual quality, as the sole modifications. Code will be made publicly available upon acceptance.", "date": "2026-03-17", "categories": []}
{"id": "2603.16587", "title": "HistoAtlas: A Pan-Cancer Morphology Atlas Linking Histomics to Molecular Programs and Clinical Outcomes", "abstract": "We present HistoAtlas, a pan-cancer computational atlas that extracts 38 interpretable histomic features from 6,745 diagnostic H&E slides across 21 TCGA cancer types and systematically links every feature to survival, gene expression, somatic mutations, and immune subtypes. All associations are covariate-adjusted, multiple-testing corrected, and classified into evidence-strength tiers. The atlas recovers known biology, from immune infiltration and prognosis to proliferation and kinase signaling, while uncovering compartment-specific immune signals and morphological subtypes with divergent outcomes. Every result is spatially traceable to tissue compartments and individual cells, statistically calibrated, and openly queryable. HistoAtlas enables systematic, large-scale biomarker discovery from routine H&E without specialized staining or sequencing. Data and an interactive web atlas are freely available at https://histoatlas.com .", "date": "2026-03-17", "categories": []}
{"id": "2603.13401", "title": "MAD: Microenvironment-Aware Distillation -- A Pretraining Strategy for Virtual Spatial Omics from Microscopy", "abstract": "Bridging microscopy and omics would allow us to read molecular states from images-at single-cell resolution and tissue scale-without the cost and throughput limits of omics technologies. Self-supervised pretraining offers a scalable approach with minimal labels, yet how to encode single-cell identity within tissue environments-and the extent of biological information such models can capture-remains an open question. Here, we introduce MAD (microenvironment-aware distillation), a pretraining strategy that learns cell-centric embeddings by jointly self-distilling the morphology view and the microenvironment view of the same indexed cell into a unified embedding space. Across diverse tissues and imaging modalities, MAD achieves state-of-the-art prediction performance on downstream tasks including cell subtyping, transcriptomic prediction, and bioinformatic inference. MAD even outperforms foundation models with a similar number of model parameters that have been trained on substantially larger datasets. These results demonstrate that MAD's dual-view joint self-distillation effectively captures the complexity and diversity of cells within tissues. Together, this establishes MAD as a general tool for representation learning in microscopy, enabling virtual spatial omics and biological insights from vast microscopy datasets.", "date": "2026-03-11", "categories": []}
{"id": "2603.14897", "title": "BiTro: Bidirectional Transfer Learning Enhances Bulk and Spatial Transcriptomics Prediction in Cancer Pathological Images", "abstract": "Cancer pathological analysis requires modeling tumor heterogeneity across multiple modalities, primarily through transcriptomics and whole slide imaging (WSI), along with their spatial relations. On one hand, bulk transcriptomics and WSI images are largely available but lack spatial mapping; on the other hand, spatial transcriptomics (ST) data can offer high spatial resolution, yet facing challenges of high cost, low sequencing depth, and limited sample sizes. Therefore, the data foundation of either side is flawed and has its limit in accurately finding the mapping between the two modalities. To this end, we propose BiTro, a bidirectional transfer learning framework that can enhance bulk and spatial transcriptomics prediction from pathological images. Our contributions are twofold. First, we design a universal and transferable model architecture that works for both bulk+WSI and ST data. A major highlight is that we model WSI images on the cellular level to better capture cells' visual features, morphological phenotypes, and their spatial relations; to map cells' features to their transcriptomics measured in bulk or ST, we adopt multiple instance learning. Second, by using LoRA, our model can be efficiently transferred between bulk and ST data to exploit their complementary information. To test our framework, we conducted comprehensive experiments on five cancer datasets. Results demonstrate that 1) our base model can achieve better or competitive performance compared to existing models on bulk or spatial transcriptomics prediction, and 2) transfer learning can further improve the base model's performance.", "date": "2026-03-16", "categories": []}
{"id": "2603.12886", "title": "A protocol for evaluating robustness to H&E staining variation in computational pathology models", "abstract": "Sensitivity to staining variation remains a major barrier to deploying computational pathology (CPath) models as hematoxylin and eosin (H&E) staining varies across laboratories, requiring systematic assessment of how this variability affects model prediction. In this work, we developed a three-step protocol for evaluating robustness to H&E staining variation in CPath models. Step 1: Select reference staining conditions, Step 2: Characterize test set staining properties, Step 3: Apply CPath model(s) under simulated reference staining conditions. Here, we first created a new reference staining library based on the PLISM dataset. As an exemplary use case, we applied the protocol to assess the robustness properties of 306 microsatellite instability (MSI) classification models on the unseen SurGen colorectal cancer dataset (n=738), including 300 attention-based multiple instance learning models trained on the TCGA-COAD/READ datasets across three feature extractors (UNI2-h, H-Optimus-1, Virchow2), alongside six public MSI classification models. Classification performance was measured as AUC, and robustness as the min-max AUC range across four simulated staining conditions (low/high H&E intensity, low/high H&E color similarity). Across models and staining conditions, classification performance ranged from AUC 0.769-0.911 ($\\Delta$ = 0.142). Robustness ranged from 0.007-0.079 ($\\Delta$ = 0.072), and showed a weak inverse correlation with classification performance (Pearson r=-0.22, 95% CI [-0.34, -0.11]). Thus, we show that the proposed evaluation protocol enables robustness-informed CPath model selection and provides insight into performance shifts across H&E staining conditions, supporting the identification of operational ranges for reliable model deployment. Code is available at https://github.com/CTPLab/staining-robustness-evaluation .", "date": "2026-03-13", "categories": []}
{"id": "2509.23097", "title": "Streamline pathology foundation model by cross-magnification distillation", "abstract": "Foundation models (FM) have transformed computational pathology but remain computationally prohibitive for clinical deployment due to their massive parameter counts and high-magnification processing requirements. Here, we introduce XMAG, a lightweight FM developed through corss-magnification distillation that transfers knowledge from state-of-the-art 20x magnification teacher to an efficient 5x magnification student architecture. XMAG employs a compact backbone and operates entirely at 5x, requiring 11.3 times fewer patches per whole slide image (WSI) compared to existing approaches. Our Novel distillation framework incorporates dual-level knowledge transfer, aligning both global image representations and local spatial token mapping. We trained XMAG on 3.49 million images curated from publicly available datasets and evaluated performance across six clinically relevant histopathology analysis tasks spanning multiple cancer types. XMAG achieved diagnostic accuracy within 1% of substantially larger foundation models while delivering 30-fold processing acceleration, reaching 8.8 WSIs per minute processing speed. Our cross-institutional validation confirmed robust generalization. Further, we developed an end-to-end training strategy to further boost our model's performance to approach the larger FMs' performance. These results establish cross-magnification distillation as a viable approach for deploying FM capabilities in resource-constrained clinical environments, potentially enabling real-time pathology AI integration.", "date": "2025-09-27", "categories": []}
{"id": "2603.05884", "title": "Computational Pathology in the Era of Emerging Foundation and Agentic AI -- International Expert Perspectives on Clinical Integration and Translational Readiness", "abstract": "Recent breakthroughs in artificial intelligence through foundation models and agents have accelerated the evolution of computational pathology. Demonstrated performance gains reported across academia in benchmarking datasets in predictive tasks such as diagnosis, prognosis, and treatment response have ignited substantial enthusiasm for clinical application. Despite this development momentum, real world adoption has lagged, as implementation faces economic, technical, and administrative challenges. Beyond existing discussions of technical architectures and comparative performance, this review considers how these emerging AI systems can be responsibly integrated into medical practice by connecting deployable clinical relevance with downstream analytical capabilities and their technical maturity, operational readiness, and economic and regulatory context. Drawing on perspectives from an international group, we provide a practical assessment of current capabilities and barriers to adoption in patient care settings.", "date": "2026-03-06", "categories": []}
{"id": "2603.11403", "title": "DeepHistoViT: An Interpretable Vision Transformer Framework for Histopathological Cancer Classification", "abstract": "Histopathology remains the gold standard for cancer diagnosis because it provides detailed cellular-level assessment of tissue morphology. However, manual histopathological examination is time-consuming, labour-intensive, and subject to inter-observer variability, creating a demand for reliable computer-assisted diagnostic tools. Recent advances in deep learning, particularly transformer-based architectures, have shown strong potential for modelling complex spatial dependencies in medical images. In this work, we propose DeepHistoViT, a transformer-based framework for automated classification of histopathological images. The model employs a customized Vision Transformer architecture with an integrated attention mechanism designed to capture fine-grained cellular structures while improving interpretability through attention-based localization of diagnostically relevant regions. The framework is evaluated on three publicly available histopathology datasets covering lung cancer, colon cancer, and acute lymphoblastic leukaemia. Experimental results demonstrate state-of-the-art performance across all datasets, with classification accuracy, precision, recall, F1-score, and ROC-AUC reaching 100 percent on the lung and colon cancer datasets, and 99.85 percent, 99.84 percent, 99.86 percent, 99.85 percent, and 99.99 percent respectively on the acute lymphoblastic leukaemia dataset. All performance metrics are reported with 95 percent confidence intervals. These results highlight the effectiveness of transformer-based architectures for histopathological image analysis and demonstrate the potential of DeepHistoViT as an interpretable computer-assisted diagnostic tool to support pathologists in clinical decision-making.", "date": "2026-03-12", "categories": []}
{"id": "2603.11818", "title": "Automated Detection of Malignant Lesions in the Ovary Using Deep Learning Models and XAI", "abstract": "The unrestrained proliferation of cells that are malignant in nature is cancer. In recent times, medical professionals are constantly acquiring enhanced diagnostic and treatment abilities by implementing deep learning models to analyze medical data for better clinical decision, disease diagnosis and drug discovery. A majority of cancers are studied and treated by incorporating these technologies. However, ovarian cancer remains a dilemma as it has inaccurate non-invasive detection procedures and a time consuming, invasive procedure for accurate detection. Thus, in this research, several Convolutional Neural Networks such as LeNet-5, ResNet, VGGNet and GoogLeNet/Inception have been utilized to develop 15 variants and choose a model that accurately detects and identifies ovarian cancer. For effective model training, the dataset OvarianCancer&SubtypesDatasetHistopathology from Mendeley has been used. After constructing a model, we utilized Explainable Artificial Intelligence (XAI) models such as LIME, Integrated Gradients and SHAP to explain the black box outcome of the selected model. For evaluating the performance of the model, Accuracy, Precision, Recall, F1-Score, ROC Curve and AUC have been used. From the evaluation, it was seen that the slightly compact InceptionV3 model with ReLu had the overall best result achieving an average score of 94% across all the performance metrics in the augmented dataset. Lastly for XAI, the three aforementioned XAI have been used for an overall comparative analysis. It is the aim of this research that the contributions of the study will help in achieving a better detection method for ovarian cancer.", "date": "2026-03-12", "categories": []}
{"id": "2601.03410", "title": "Inferring Clinically Relevant Molecular Subtypes of Pancreatic Cancer from Routine Histopathology Using Deep Learning", "abstract": "Molecular subtyping of PDAC into basal-like and classical has established prognostic and predictive value. However, its use in clinical practice is limited by cost, turnaround time, and tissue requirements, thereby restricting its application in the management of PDAC. We introduce PanSubNet, an interpretable deep learning framework that predicts therapy-relevant molecular subtypes directly from standard H&E-stained WSIs. PanSubNet was developed using data from 1,055 patients across two multi-institutional cohorts (PANCAN, n=846; TCGA, n=209) with paired histology and RNA-seq data. Ground-truth labels were derived using the validated Moffitt 50-gene signature refined by GATA6 expression. The model employs dual-scale architecture that fuses cellular-level morphology with tissue-level architecture, leveraging attention mechanisms for multi-scale representation learning and transparent feature attribution. On internal validation within PANCAN using five-fold cross-validation, PanSubNet achieved mean AUC of 88.5% with balanced sensitivity and specificity. External validation on the independent TCGA cohort without fine-tuning demonstrated robust generalizability (AUC 84.0%). PanSubNet preserved and, in metastatic disease, strengthened prognostic stratification compared to RNA-seq based labels. Prediction uncertainty linked to intermediate transcriptional states, not classification noise. Model predictions are aligned with established transcriptomic programs, differentiation markers, and DNA damage repair signatures. By enabling rapid, cost-effective molecular stratification from routine H&E-stained slides, PanSubNet offers a clinically deployable and interpretable tool for genetic subtyping. We are gathering data from two institutions to validate and assess real-world performance, supporting integration into digital pathology workflows and advancing precision oncology for PDAC.", "date": "2026-01-06", "categories": []}
{"id": "2603.07170", "title": "Class Visualizations and Activation Atlases for Enhancing Interpretability in Deep Learning-Based Computational Pathology", "abstract": "The rapid adoption of transformer-based models in computational pathology has enabled prediction of molecular and clinical biomarkers from H&E whole-slide images, yet interpretability has not kept pace with model complexity. While attribution- and generative-based methods are common, feature visualization approaches such as class visualizations (CVs) and activation atlases (AAs) have not been systematically evaluated for these models. We developed a visualization framework and assessed CVs and AAs for a transformer-based foundation model across tissue and multi-organ cancer classification tasks with increasing label granularity. Four pathologists annotated real and generated images to quantify inter-observer agreement, complemented by attribution and similarity metrics. CVs preserved recognizability for morphologically distinct tissues but showed reduced separability for overlapping cancer subclasses. In tissue classification, agreement decreased from Fleiss k = 0.75 (scans) to k = 0.31 (CVs), with similar trends in cancer subclass tasks. AAs revealed layer-dependent organization: coarse tissue-level concepts formed coherent regions, whereas finer subclasses exhibited dispersion and overlap. Agreement was moderate for tissue classification (k = 0.58), high for coarse cancer groupings (k = 0.82), and low at subclass level (k = 0.11). Atlas separability closely tracked expert agreement on real images, indicating that representational ambiguity reflects intrinsic pathological complexity. Attribution-based metrics approximated expert variability in low-complexity settings, whereas perceptual and distributional metrics showed limited alignment. Overall, concept-level feature visualization reveals structured morphological manifolds in transformer-based pathology models and provides a framework for expert-centered interrogation of learned representations across label granularities.", "date": "2026-03-07", "categories": []}
{"id": "2603.09953", "title": "Leveraging whole slide difficulty in Multiple Instance Learning to improve prostate cancer grading", "abstract": "Multiple Instance Learning (MIL) has been widely applied in histopathology to classify Whole Slide Images (WSIs) with slide-level diagnoses. While the ground truth is established by expert pathologists, the slides can be difficult to diagnose for non-experts and lead to disagreements between the annotators. In this paper, we introduce the notion of Whole Slide Difficulty (WSD), based on the disagreement between an expert and a non-expert pathologist. We propose two different methods to leverage WSD, a multi-task approach and a weighted classification loss approach, and we apply them to Gleason grading of prostate cancer slides. Results show that integrating WSD during training consistently improves the classification performance across different feature encoders and MIL methods, particularly for higher Gleason grades (i.e. worse diagnosis).", "date": "2026-03-10", "categories": []}
{"id": "2603.06650", "title": "Margin-Consistent Deep Subtyping of Invasive Lung Adenocarcinoma via Perturbation Fidelity in Whole-Slide Image Analysis", "abstract": "Whole-slide image classification for invasive lung adenocarcinoma subtyping remains vulnerable to real-world imaging perturbations that undermine model reliability at the decision boundary. We propose a margin consistency framework evaluated on 203,226 patches from 143 whole-slide images spanning five adenocarcinoma subtypes in the BMIRDS-LUAD dataset. By combining attention-weighted patch aggregation with margin-aware training, our approach achieves robust feature-logit space alignment measured by Kendall correlations of 0.88 during training and 0.64 during validation. Contrastive regularization, while effective at improving class separation, tends to over-cluster features and suppress fine-grained morphological variation; to counteract this, we introduce Perturbation Fidelity (PF) scoring, which imposes structured perturbations through Bayesian-optimized parameters. Vision Transformer-Large achieves 95.20 +/- 4.65% accuracy, representing a 40% error reduction from the 92.00 +/- 5.36% baseline, while ResNet101 with an attention mechanism reaches 95.89 +/- 5.37% from 91.73 +/- 9.23%, a 50% error reduction. All five subtypes exceed an area under the receiver operating characteristic curve (AUC) of 0.99. On the WSSS4LUAD external benchmark, ResNet50 with an attention mechanism attains 80.1% accuracy, demonstrating cross-institutional generalizability despite approximately 15-20% domain-shift-related degradation and identifying opportunities for future adaptation research.", "date": "2026-02-27", "categories": []}
{"id": "2603.06658", "title": "ASMIL: Attention-Stabilized Multiple Instance Learning for Whole Slide Imaging", "abstract": "Attention-based multiple instance learning (MIL) has emerged as a powerful framework for whole slide image (WSI) diagnosis, leveraging attention to aggregate instance-level features into bag-level predictions. Despite this success, we find that such methods exhibit a new failure mode: unstable attention dynamics. Across four representative attention-based MIL methods and two public WSI datasets, we observe that attention distributions oscillate across epochs rather than converging to a consistent pattern, degrading performance. This instability adds to two previously reported challenges: overfitting and over-concentrated attention distribution. To simultaneously overcome these three limitations, we introduce attention-stabilized multiple instance learning (ASMIL), a novel unified framework. ASMIL uses an anchor model to stabilize attention, replaces softmax with a normalized sigmoid function in the anchor to prevent over-concentration, and applies token random dropping to mitigate overfitting. Extensive experiments demonstrate that ASMIL achieves up to a 6.49\\% F1 score improvement over state-of-the-art methods. Moreover, integrating the anchor model and normalized sigmoid into existing attention-based MIL methods consistently boosts their performance, with F1 score gains up to 10.73\\%. All code and data are publicly available at https://github.com/Linfeng-Ye/ASMIL.", "date": "2026-03-01", "categories": []}
{"id": "2603.06936", "title": "Extracting and analyzing 3D histomorphometric features related to perineural and lymphovascular invasion in prostate cancer", "abstract": "Diagnostic grading of prostate cancer (PCa) relies on the examination of 2D histology sections. However, the limited sampling of specimens afforded by 2D histopathology, and ambiguities when viewing 2D cross-sections, can lead to suboptimal treatment decisions. Recent studies have shown that 3D histomorphometric analysis of glands and nuclei can improve PCa risk assessment compared to analogous 2D features. Here, we expand on these efforts by developing an analytical pipeline to extract 3D features related to perineural invasion (PNI) and lymphovascular invasion (LVI), which correlate with poor prognosis for a variety of cancers. A 3D segmentation model (nnU-Net) was trained to segment nerves and vessels in 3D datasets of archived prostatectomy specimens that were optically cleared, labeled with a fluorescent analog of H&E, and imaged with open-top light-sheet (OTLS) microscopy. PNI- and LVI-related features, including metrics describing cancer-nerve and cancer-vessel proximity, were then extracted based on the 3D nerve/vessel segmentation masks in conjunction with 3D masks of cancer-enriched regions. As a preliminary exploration of the prognostic value of these features, we trained a supervised machine learning classifier to predict 5-year biochemical recurrence (BCR) outcomes, finding that 3D PNI-related features are moderately prognostic and outperform 2D PNI-related features (AUC = 0.71 vs. 0.52). Source code is available at https://github.com/sarahrahsl/SegCIA.git.", "date": "2026-03-06", "categories": []}
{"id": "2603.07895", "title": "MINT: Molecularly Informed Training with Spatial Transcriptomics Supervision for Pathology Foundation Models", "abstract": "Pathology foundation models learn morphological representations through self-supervised pretraining on large-scale whole-slide images, yet they do not explicitly capture the underlying molecular state of the tissue. Spatial transcriptomics technologies bridge this gap by measuring gene expression in situ, offering a natural cross-modal supervisory signal. We propose MINT (Molecularly Informed Training), a fine-tuning framework that incorporates spatial transcriptomics supervision into pretrained pathology Vision Transformers. MINT appends a learnable ST token to the ViT input to encode transcriptomic information separately from the morphological CLS token, preventing catastrophic forgetting through DINO self-distillation and explicit feature anchoring to the frozen pretrained encoder. Gene expression regression at both spot-level (Visium) and patch-level (Xenium) resolutions provides complementary supervision across spatial scales. Trained on 577 publicly available HEST samples, MINT achieves the best overall performance on both HEST-Bench for gene expression prediction (mean Pearson r = 0.440) and EVA for general pathology tasks (0.803), demonstrating that spatial transcriptomics supervision complements morphology-centric self-supervised pretraining.", "date": "2026-03-09", "categories": []}
{"id": "2603.05535", "title": "Clinical-Injection Transformer with Domain-Adapted MAE for Lupus Nephritis Prognosis Prediction", "abstract": "Lupus nephritis (LN) is a severe complication of systemic lupus erythematosus that affects pediatric patients with significantly greater severity and worse renal outcomes compared to adults. Despite the urgent clinical need, predicting pediatric LN prognosis remains unexplored in computational pathology. Furthermore, the only existing histopathology-based approach for LN relies on multiple costly staining protocols and fails to integrate complementary clinical data. To address these gaps, we propose the first multimodal computational pathology framework for three-class treatment response prediction (complete remission, partial response, and no response) in pediatric LN, utilizing only routine PAS-stained biopsies and structured clinical data. Our framework introduces two key methodological innovations. First, a Clinical-Injection Transformer (CIT) embeds clinical features as condition tokens into patch-level self-attention, facilitating implicit and bidirectional cross-modal interactions within a unified attention space. Second, we design a decoupled representation-knowledge adaptation strategy using a domain-adapted Masked Autoencoder (MAE). This strategy explicitly separates self-supervised morphological feature learning from pathological knowledge extraction. Additionally, we introduce a multi-granularity morphological type injection mechanism to bridge distilled classification knowledge with downstream prognostic predictions at both the instance and patient levels. Evaluated on a cohort of 71 pediatric LN patients with KDIGO-standardized labels, our method achieves a three-class accuracy of 90.1% and an AUC of 89.4%, demonstrating its potential as a highly accurate and cost-effective prognostic tool.", "date": "2026-03-04", "categories": []}
{"id": "2508.02464", "title": "SAMPO-Path: Segmentation Intent-Aligned Preference Optimization for Pathology Foundation Model Segmentation", "abstract": "Foundation models have shown strong performance in multi-object segmentation with visual prompts, yet histopathology images remain challenging due to high cellular density, heterogeneity, and the gap between pixel-level supervision and clinical segmentation intent (e.g., selectively segmenting nuclei of a specific type). In practice, such intents are expressed through diverse and noisy prompts, causing prompt-intent misalignment and inconsistent predictions. We introduce SAMPO (Segmentation Anything Model with Preference Optimization), a preference-aligned fine-tuning framework that explicitly aligns pathology foundation models with clinical segmentation intent. SAMPO is the first to adapt Direct Preference Optimization (DPO) to pure vision foundation models, enabling accurate segmentation from minimal and imperfect prompts. The framework features three key components: (1) online prompt-centric preference mining to synthesize preference pairs across prompt qualities; (2) multi-mask preference learning to leverage output ambiguity for fine-grained ranking supervision; and (3) a hybrid loss combining preference optimization with pixel-level supervision for stable training. Trained on two datasets covering four tasks and evaluated on corresponding test sets and 12 external validation datasets, SAMPO consistently improves segmentation accuracy, robustness to prompt variations, and clinical intent adherence in dense histopathology images.", "date": "2025-08-04", "categories": []}
{"id": "2603.04405", "title": "Lost in Translation: How Language Re-Aligns Vision for Cross-Species Pathology", "abstract": "Foundation models are increasingly applied to computational pathology, yet their behavior under cross-cancer and cross-species transfer remains unspecified. This study investigated how fine-tuning CPath-CLIP affects cancer detection under same-cancer, cross-cancer, and cross-species conditions using whole-slide image patches from canine and human histopathology. Performance was measured using area under the receiver operating characteristic curve (AUC). Few-shot fine-tuning improved same-cancer (64.9% to 72.6% AUC) and cross-cancer performance (56.84% to 66.31% AUC). Cross-species evaluation revealed that while tissue matching enables meaningful transfer, performance remains below state-of-the-art benchmarks (H-optimus-0: 84.97% AUC), indicating that standard vision-language alignment is suboptimal for cross-species generalization. Embedding space analysis revealed extremely high cosine similarity (greater than 0.99) between tumor and normal prototypes. Grad-CAM shows prototype-based models remain domain-locked, while language-guided models attend to conserved tumor morphology. To address this, we introduce Semantic Anchoring, which uses language to provide a stable coordinate system for visual features. Ablation studies reveal that benefits stem from the text-alignment mechanism itself, regardless of text encoder complexity. Benchmarking against H-optimus-0 shows that CPath-CLIP's failure stems from intrinsic embedding collapse, which text alignment effectively circumvents. Additional gains were observed in same-cancer (8.52%) and cross-cancer classification (5.67%). We identified a previously uncharacterized failure mode: semantic collapse driven by species-dominated alignment rather than missing visual information. These results demonstrate that language acts as a control mechanism, enabling semantic re-interpretation without retraining.", "date": "2026-01-24", "categories": []}
{"id": "2603.04565", "title": "Structure-Guided Histopathology Synthesis via Dual-LoRA Diffusion", "abstract": "Histopathology image synthesis plays an important role in tissue restoration, data augmentation, and modeling of tumor microenvironments. However, existing generative methods typically address restoration and generation as separate tasks, although both share the same objective of structure-consistent tissue synthesis under varying degrees of missingness, and often rely on weak or inconsistent structural priors that limit realistic cellular organization.\n  We propose Dual-LoRA Controllable Diffusion, a unified centroid-guided diffusion framework that jointly supports Local Structure Completion and Global Structure Synthesis within a single model. Multi-class nuclei centroids serve as lightweight and annotation-efficient spatial priors, providing biologically meaningful guidance under both partial and complete image absence. Two task-specific LoRA adapters specialize the shared backbone for local and global objectives without retraining separate diffusion models. Extensive experiments demonstrate consistent improvements over state-of-the-art GAN and diffusion baselines across restoration and synthesis tasks.\n  For local completion, LPIPS computed within the masked region improves from 0.1797 (HARP) to 0.1524, and for global synthesis, FID improves from 225.15 (CoSys) to 76.04, indicating improved structural fidelity and realism. Our approach achieves more faithful structural recovery in masked regions and substantially improved realism and morphology consistency in full synthesis, supporting scalable pan-cancer histopathology modeling.", "date": "2026-03-04", "categories": []}
{"id": "2603.04240", "title": "DeNuC: Decoupling Nuclei Detection and Classification in Histopathology", "abstract": "Pathology Foundation Models (FMs) have shown strong performance across a wide range of pathology image representation and diagnostic tasks. However, FMs do not exhibit the expected performance advantage over traditional specialized models in Nuclei Detection and Classification (NDC). In this work, we reveal that jointly optimizing nuclei detection and classification leads to severe representation degradation in FMs. Moreover, we identify that the substantial intrinsic disparity in task difficulty between nuclei detection and nuclei classification renders joint NDC optimization unnecessarily computationally burdensome for the detection stage. To address these challenges, we propose DeNuC, a simple yet effective method designed to break through existing bottlenecks by Decoupling Nuclei detection and Classification. DeNuC employs a lightweight model for accurate nuclei localization, subsequently leveraging a pathology FM to encode input images and query nucleus-specific features based on the detected coordinates for classification. Extensive experiments on three widely used benchmarks demonstrate that DeNuC effectively unlocks the representational potential of FMs for NDC and significantly outperforms state-of-the-art methods. Notably, DeNuC improves F1 scores by 4.2% and 3.6% (or higher) on the BRCAM2C and PUMA datasets, respectively, while using only 16% (or fewer) trainable parameters compared to other methods. Code is available at https://github.com/ZijiangY1116/DeNuC.", "date": "2026-03-04", "categories": []}
{"id": "2512.20833", "title": "CHAMMI-75: Pre-training multi-channel models with heterogeneous microscopy images", "abstract": "Quantifying cell morphology using images and machine learning has proven to be a powerful tool to study the response of cells to treatments. However, models used to quantify cellular morphology are typically trained with a single microscopy imaging type. This results in specialized models that cannot be reused across biological studies because the technical specifications do not match (e.g., different number of channels). Here, we present CHAMMI-75, an open access dataset of heterogeneous, multi-channel microscopy images from 75 diverse biological studies. We curated this resource from publicly available sources to investigate cellular morphology models that are channel-adaptive and can process any microscopy image type. Our experiments show that training with CHAMMI-75 can improve performance in multi-channel bioimaging tasks primarily because of its high diversity in microscopy modalities. This work paves the way to create the next generation of cellular morphology models for biological studies.", "date": "2025-12-23", "categories": []}
{"id": "2603.02704", "title": "Intelligent Pathological Diagnosis of Gestational Trophoblastic Diseases via Visual-Language Deep Learning Model", "abstract": "The pathological diagnosis of gestational trophoblastic disease(GTD) takes a long time, relies heavily on the experience of pathologists, and the consistency of initial diagnosis is low, which seriously threatens maternal health and reproductive outcomes. We developed an expert model for GTD pathological diagnosis, named GTDoctor. GTDoctor can perform pixel-based lesion segmentation on pathological slides, and output diagnostic conclusions and personalized pathological analysis results. We developed a software system, GTDiagnosis, based on this technology and conducted clinical trials. The retrospective results demonstrated that GTDiagnosis achieved a mean precision of over 0.91 for lesion detection in pathological slides (n=679 slides). In prospective studies, pathologists using GTDiagnosis attained a Positive Predictive Value of 95.59% (n=68 patients). The tool reduced average diagnostic time from 56 to 16 seconds per case (n=285 patients). GTDoctor and GTDiagnosis offer a novel solution for GTD pathological diagnosis, enhancing diagnostic performance and efficiency while maintaining clinical interpretability.", "date": "2026-03-03", "categories": []}
{"id": "2603.02926", "title": "GloPath: An Entity-Centric Foundation Model for Glomerular Lesion Assessment and Clinicopathological Insights", "abstract": "Glomerular pathology is central to the diagnosis and prognosis of renal diseases, yet the heterogeneity of glomerular morphology and fine-grained lesion patterns remain challenging for current AI approaches. We present GloPath, an entity-centric foundation model trained on over one million glomeruli extracted from 14,049 renal biopsy specimens using multi-scale and multi-view self-supervised learning. GloPath addresses two major challenges in nephropathology: glomerular lesion assessment and clinicopathological insights discovery. For lesion assessment, GloPath was benchmarked across three independent cohorts on 52 tasks, including lesion recognition, grading, few-shot classification, and cross-modality diagnosis-outperforming state-of-the-art methods in 42 tasks (80.8%). In the large-scale real-world study, it achieved an ROC-AUC of 91.51% for lesion recognition, demonstrating strong robustness in routine clinical settings. For clinicopathological insights, GloPath systematically revealed statistically significant associations between glomerular morphological parameters and clinical indicators across 224 morphology-clinical variable pairs, demonstrating its capacity to connect tissue-level pathology with patient-level outcomes. Together, these results position GloPath as a scalable and interpretable platform for glomerular lesion assessment and clinicopathological discovery, representing a step toward clinically translatable AI in renal pathology.", "date": "2026-03-03", "categories": []}
{"id": "2603.03030", "title": "BRIGHT: A Collaborative Generalist-Specialist Foundation Model for Breast Pathology", "abstract": "Generalist pathology foundation models (PFMs), pretrained on large-scale multi-organ datasets, have demonstrated remarkable predictive capabilities across diverse clinical applications. However, their proficiency on the full spectrum of clinically essential tasks within a specific organ system remains an open question due to the lack of large-scale validation cohorts for a single organ as well as the absence of a tailored training paradigm that can effectively translate broad histomorphological knowledge into the organ-specific expertise required for specialist-level interpretation. In this study, we propose BRIGHT, the first PFM specifically designed for breast pathology, trained on approximately 210 million histopathology tiles from over 51,000 breast whole-slide images derived from a cohort of over 40,000 patients across 19 hospitals. BRIGHT employs a collaborative generalist-specialist framework to capture both universal and organ-specific features. To comprehensively evaluate the performance of PFMs on breast oncology, we curate the largest multi-institutional cohorts to date for downstream task development and evaluation, comprising over 25,000 WSIs across 10 hospitals. The validation cohorts cover the full spectrum of breast pathology across 24 distinct clinical tasks spanning diagnosis, biomarker prediction, treatment response and survival prediction. Extensive experiments demonstrate that BRIGHT outperforms three leading generalist PFMs, achieving state-of-the-art (SOTA) performance in 21 of 24 internal validation tasks and in 5 of 10 external validation tasks with excellent heatmap interpretability. By evaluating on large-scale validation cohorts, this study not only demonstrates BRIGHT's clinical utility in breast oncology but also validates a collaborative generalist-specialist paradigm, providing a scalable template for developing PFMs on a specific organ system.", "date": "2026-03-03", "categories": []}
{"id": "2508.16479", "title": "Disentangled Multi-modal Learning of Histology and Transcriptomics for Cancer Characterization", "abstract": "Histopathology remains the gold standard for cancer diagnosis and prognosis. With the advent of transcriptome profiling, multi-modal learning combining transcriptomics with histology offers more comprehensive information. However, existing multi-modal approaches are challenged by intrinsic multi-modal heterogeneity, insufficient multi-scale integration, and reliance on paired data, restricting clinical applicability. To address these challenges, we propose a disentangled multi-modal framework with four contributions: 1) To mitigate multi-modal heterogeneity, we decompose WSIs and transcriptomes into tumor and microenvironment subspaces using a disentangled multi-modal fusion module, and introduce a confidence-guided gradient coordination strategy to balance subspace optimization. 2) To enhance multi-scale integration, we propose an inter-magnification gene-expression consistency strategy that aligns transcriptomic signals across WSI magnifications. 3) To reduce dependency on paired data, we propose a subspace knowledge distillation strategy enabling transcriptome-agnostic inference through a WSI-only student model. 4) To improve inference efficiency, we propose an informative token aggregation module that suppresses WSI redundancy while preserving subspace semantics. Extensive experiments on cancer diagnosis, prognosis, and survival prediction demonstrate our superiority over state-of-the-art methods across multiple settings. Code is available at https://github.com/helenypzhang/Disentangled-Multimodal-Learning.", "date": "2025-08-22", "categories": []}
{"id": "2603.01143", "title": "TC-SSA: Token Compression via Semantic Slot Aggregation for Gigapixel Pathology Reasoning", "abstract": "The application of large vision-language models to computational pathology holds great promise for diagnostic assistants but faces a critical computational bottleneck: the gigapixel scale of Whole Slide Images (WSIs). A single WSI typically contains over 105 patches, creating sequence lengths that exceed the constraints of standard Transformer architectures. Existing solutions often resort to spatial sampling, which risks discarding diagnostically critical evidence. To address this, we propose TC-SSA (Token Compression via Semantic Slot Aggregation), a learnable token compression framework that aggregates patch features into a fixed number of semantic slots. A gated routing module assigns patches to slots using sparse Top-2 routing, followed by weighted aggregation, enabling global slide coverage under a strict token budget. The resulting representation retains diagnostically relevant information while reducing the number of visual tokens to 1.7% of the original sequence. On the SlideBench(TCGA), our model achieves 78.34% overall accuracy and 77.14% on the diagnosis subset, outperforming sampling-based baselines under comparable token budgets. The method also generalizes to MIL classification, reaching AUC of 95.83% on TCGA-BRCA, 98.27% on TCGA-NSCLC and 79.80% on PANDA. These results suggest that learnable semantic aggregation provides an effective trade-off between efficiency and diagnostic performance for gigapixel pathology reasoning.", "date": "2026-03-01", "categories": []}
{"id": "2603.01647", "title": "QCAgent: An agentic framework for quality-controllable pathology report generation from whole slide image", "abstract": "Recent methods for pathology report generation from whole-slide image (WSI) are capable of producing slide-level diagnostic descriptions but fail to ground fine-grained statements in localized visual evidence. Furthermore, they lack control over which diagnostic details to include and how to verify them. Inspired by emerging agentic analysis paradigms and the diagnostic workflow of pathologists,who selectively examine multiple fields of view, we propose QCAgent, an agentic framework for quality-controllable WSI report generation. The core innovations of this framework are as follows: (i) it incorporates a customized critique mechanism guided by a user-defined checklist specifying required diagnostic details and constraints; (ii) it re-identifies informative regions in the WSI based on the critique feedback and text-patch semantic retrieval, a process that iteratively enriches and reconciles the report. Experiments demonstrate that by making report requirements explicitly prompt-defined, constraint-aware, and verifiable through evidence-grounded refinement, QCAgent enables controllable generation of clinically meaningful and high-coverage pathology reports from WSI.", "date": "2026-03-02", "categories": []}
{"id": "2603.02162", "title": "Bridging the gap between Performance and Interpretability: An Explainable Disentangled Multimodal Framework for Cancer Survival Prediction", "abstract": "While multimodal survival prediction models are increasingly more accurate, their complexity often reduces interpretability, limiting insight into how different data sources influence predictions. To address this, we introduce DIMAFx, an explainable multimodal framework for cancer survival prediction that produces disentangled, interpretable modality-specific and modality-shared representations from histopathology whole-slide images and transcriptomics data. Across multiple cancer cohorts, DIMAFx achieves state-of-the-art performance and improved representation disentanglement. Leveraging its interpretable design and SHapley Additive exPlanations, DIMAFx systematically reveals key multimodal interactions and the biological information encoded in the disentangled representations. In breast cancer survival prediction, the most predictive features contain modality-shared information, including one capturing solid tumor morphology contextualized primarily by late estrogen response, where higher-grade morphology aligned with pathway upregulation and increased risk, consistent with known breast cancer biology. Key modality-specific features capture microenvironmental signals from interacting adipose and stromal morphologies. These results show that multimodal models can overcome the traditional trade-off between performance and explainability, supporting their application in precision medicine.", "date": "2026-03-02", "categories": []}
{"id": "2602.23817", "title": "Footprint-Guided Exemplar-Free Continual Histopathology Report Generation", "abstract": "Rapid progress in vision-language modeling has enabled pathology report generation from gigapixel whole-slide images, but most approaches assume static training with simultaneous access to all data. In clinical deployment, however, new organs, institutions, and reporting conventions emerge over time, and sequential fine-tuning can cause catastrophic forgetting. We introduce an exemplar-free continual learning framework for WSI-to-report generation that avoids storing raw slides or patch exemplars. The core idea is a compact domain footprint built in a frozen patch-embedding space: a small codebook of representative morphology tokens together with slide-level co-occurrence summaries and lightweight patch-count priors. These footprints support generative replay by synthesizing pseudo-WSI representations that reflect domain-specific morphological mixtures, while a teacher snapshot provides pseudo-reports to supervise the updated model without retaining past data. To address shifting reporting conventions, we distill domain-specific linguistic characteristics into a compact style descriptor and use it to steer generation. At inference, the model identifies the most compatible descriptor directly from the slide signal, enabling domain-agnostic setup without requiring explicit domain identifiers. Evaluated across multiple public continual learning benchmarks, our approach outperforms exemplar-free and limited-buffer rehearsal baselines, highlighting footprint-based generative replay as a practical solution for deployment in evolving clinical settings.", "date": "2026-02-27", "categories": []}
{"id": "2602.24222", "title": "MuViT: Multi-Resolution Vision Transformers for Learning Across Scales in Microscopy", "abstract": "Modern microscopy routinely produces gigapixel images that contain structures across multiple spatial scales, from fine cellular morphology to broader tissue organization. Many analysis tasks require combining these scales, yet most vision models operate at a single resolution or derive multi-scale features from one view, limiting their ability to exploit the inherently multi-resolution nature of microscopy data. We introduce MuViT, a transformer architecture built to fuse true multi-resolution observations from the same underlying image. MuViT embeds all patches into a shared world-coordinate system and extends rotary positional embeddings to these coordinates, enabling attention to integrate wide-field context with high-resolution detail within a single encoder. Across synthetic benchmarks, kidney histopathology, and high-resolution mouse-brain microscopy, MuViT delivers consistent improvements over strong ViT and CNN baselines. Multi-resolution MAE pretraining further produces scale-consistent representations that enhance downstream tasks. These results demonstrate that explicit world-coordinate modelling provides a simple yet powerful mechanism for leveraging multi-resolution information in large-scale microscopy analysis.", "date": "2026-02-27", "categories": []}
{"id": "2602.23292", "title": "PGVMS: A Prompt-Guided Unified Framework for Virtual Multiplex IHC Staining with Pathological Semantic Learning", "abstract": "Immunohistochemical (IHC) staining enables precise molecular profiling of protein expression, with over 200 clinically available antibody-based tests in modern pathology. However, comprehensive IHC analysis is frequently limited by insufficient tissue quantities in small biopsies. Therefore, virtual multiplex staining emerges as an innovative solution to digitally transform H&E images into multiple IHC representations, yet current methods still face three critical challenges: (1) inadequate semantic guidance for multi-staining, (2) inconsistent distribution of immunochemistry staining, and (3) spatial misalignment across different stain modalities. To overcome these limitations, we present a prompt-guided framework for virtual multiplex IHC staining using only uniplex training data (PGVMS). Our framework introduces three key innovations corresponding to each challenge: First, an adaptive prompt guidance mechanism employing a pathological visual language model dynamically adjusts staining prompts to resolve semantic guidance limitations (Challenge 1). Second, our protein-aware learning strategy (PALS) maintains precise protein expression patterns by direct quantification and constraint of protein distributions (Challenge 2). Third, the prototype-consistent learning strategy (PCLS) establishes cross-image semantic interaction to correct spatial misalignments (Challenge 3).", "date": "2026-02-26", "categories": []}
{"id": "2602.23305", "title": "A Proper Scoring Rule for Virtual Staining", "abstract": "Generative virtual staining (VS) models for high-throughput screening (HTS) can provide an estimated posterior distribution of possible biological feature values for each input and cell. However, when evaluating a VS model, the true posterior is unavailable. Existing evaluation protocols only check the accuracy of the marginal distribution over the dataset rather than the predicted posteriors. We introduce information gain (IG) as a cell-wise evaluation framework that enables direct assessment of predicted posteriors. IG is a strictly proper scoring rule and comes with a sound theoretical motivation allowing for interpretability, and for comparing results across models and features. We evaluate diffusion- and GAN-based models on an extensive HTS dataset using IG and other metrics and show that IG can reveal substantial performance differences other metrics cannot.", "date": "2026-02-26", "categories": []}
{"id": "2602.22176", "title": "Mixed Magnification Aggregation for Generalizable Region-Level Representations in Computational Pathology", "abstract": "In recent years, a standard computational pathology workflow has emerged where whole slide images are cropped into tiles, these tiles are processed using a foundation model, and task-specific models are built using the resulting representations. At least 15 different foundation models have been proposed, and the vast majority are trained exclusively with tiles using the 20$\\times$ magnification. However, it is well known that certain histologic features can only be discerned with larger context windows and requires a pathologist to zoom in and out when analyzing a whole slide image. Furthermore, creating 224$\\times$224 pixel crops at 20$\\times$ leads to a large number of tiles per slide, which can be gigapixel in size. To more accurately capture multi-resolution features and investigate the possibility of reducing the number of representations per slide, we propose a region-level mixing encoder. Our approach jointly fuses image tile representations of a mixed magnification foundation model using a masked embedding modeling pretraining step. We explore a design space for pretraining the proposed mixed-magnification region aggregators and evaluate our models on transfer to biomarker prediction tasks representing various cancer types. Results demonstrate cancer dependent improvements in predictive performance, highlighting the importance of spatial context and understanding.", "date": "2026-02-25", "categories": []}
{"id": "2502.13027", "title": "A deep learning framework for efficient pathology image analysis", "abstract": "Artificial intelligence (AI) has transformed digital pathology by enabling biomarker prediction from high-resolution whole-slide images (WSIs). However, current methods are computationally inefficient, processing thousands of redundant tiles per WSI and requiring complex aggregator models. We introduce EAGLE (Efficient Approach for Guided Local Examination), a deep learning framework that emulates pathologists by selectively analyzing informative regions. EAGLE incorporates two foundation models: CHIEF for efficient tile selection and Virchow2 for extracting high-quality features. Benchmarking was conducted against leading slide- and tile-level foundation models across 43 tasks from nine cancer types, spanning morphology, biomarker prediction, treatment response and prognosis. EAGLE outperformed state-of-the-art patch aggregation methods by up to 23% and achieved the highest AUROC overall. It processed a slide in 2.27 seconds, reducing computational time by more than 99% compared to existing models. This efficiency enables real-time workflows, allows rapid review of the exact tiles used for each prediction, and reduces dependence on high-performance computing, making AI-powered pathology more accessible. By reliably identifying meaningful regions and minimizing artifacts, EAGLE provides robust and auditable outputs, supported by systematic negative controls and attention concentration analyses. Its unified embedding enables rapid slide searches, integration into multi-omics pipelines and emerging clinical foundation models.", "date": "2025-02-18", "categories": []}
{"id": "2602.20873", "title": "MUSE: Harnessing Precise and Diverse Semantics for Few-Shot Whole Slide Image Classification", "abstract": "In computational pathology, few-shot whole slide image classification is primarily driven by the extreme scarcity of expert-labeled slides. Recent vision-language methods incorporate textual semantics generated by large language models, but treat these descriptions as static class-level priors that are shared across all samples and lack sample-wise refinement. This limits both the diversity and precision of visual-semantic alignment, hindering generalization under limited supervision. To overcome this, we propose the stochastic MUlti-view Semantic Enhancement (MUSE), a framework that first refines semantic precision via sample-wise adaptation and then enhances semantic richness through retrieval-augmented multi-view generation. Specifically, MUSE introduces Sample-wise Fine-grained Semantic Enhancement (SFSE), which yields a fine-grained semantic prior for each sample through MoE-based adaptive visual-semantic interaction. Guided by this prior, Stochastic Multi-view Model Optimization (SMMO) constructs an LLM-generated knowledge base of diverse pathological descriptions per class, then retrieves and stochastically integrates multiple matched textual views during training. These dynamically selected texts serve as enriched semantic supervisions to stochastically optimize the vision-language model, promoting robustness and mitigating overfitting. Experiments on three benchmark WSI datasets show that MUSE consistently outperforms existing vision-language baselines in few-shot settings, demonstrating that effective few-shot pathology learning requires not only richer semantic sources but also their active and sample-aware semantic optimization. Our code is available at: https://github.com/JiahaoXu-god/CVPR2026_MUSE.", "date": "2026-02-24", "categories": []}
{"id": "2602.18747", "title": "Benchmarking Computational Pathology Foundation Models For Semantic Segmentation", "abstract": "In recent years, foundation models such as CLIP, DINO,and CONCH have demonstrated remarkable domain generalization and unsupervised feature extraction capabilities across diverse imaging tasks. However, systematic and independent evaluations of these models for pixel-level semantic segmentation in histopathology remain scarce. In this study, we propose a robust benchmarking approach to asses 10 foundational models on four histopathological datasets covering both morphological tissue-region and cellular/nuclear segmentation tasks. Our method leverages attention maps of foundation models as pixel-wise features, which are then classified using a machine learning algorithm, XGBoost, enabling fast, interpretable, and model-agnostic evaluation without finetuning. We show that the vision language foundation model, CONCH performed the best across datasets when compared to vision-only foundation models, with PathDino as close second. Further analysis shows that models trained on distinct histopathology cohorts capture complementary morphological representations, and concatenating their features yields superior segmentation performance. Concatenating features from CONCH, PathDino and CellViT outperformed individual models across all the datasets by 7.95% (averaged across the datasets), suggesting that ensembles of foundation models can better generalize to diverse histopathological segmentation tasks.", "date": "2026-02-21", "categories": []}
{"id": "2602.18773", "title": "LAMMI-Pathology: A Tool-Centric Bottom-Up LVLM-Agent Framework for Molecularly Informed Medical Intelligence in Pathology", "abstract": "The emergence of tool-calling-based agent systems introduces a more evidence-driven paradigm for pathology image analysis in contrast to the coarse-grained text-image diagnostic approaches. With the recent large-scale experimental adoption of spatial transcriptomics technologies, molecularly validated pathological diagnosis is becoming increasingly open and accessible. In this work, we propose LAMMI-Pathology (LVLM-Agent System for Molecularly Informed Medical Intelligence in Pathology), a scalable agent framework for domain-specific agent tool-calling. LAMMI-Pathology adopts a tool-centric, bottom-up architecture in which customized domain-adaptive tools serve as the foundation. These tools are clustered by domain style to form component agents, which are then coordinated through a top-level planner hierarchically, avoiding excessively long context lengths that could induce task drift. Based on that, we introduce a novel trajectory construction mechanism based on Atomic Execution Nodes (AENs), which serve as reliable and composable units for building semi-simulated reasoning trajectories that capture credible agent-tool interactions. Building on this foundation, we develop a trajectory-aware fine-tuning strategy that aligns the planner's decision-making process with these multi-step reasoning trajectories, thereby enhancing inference robustness in pathology understanding and its adaptive use of the customized toolset.", "date": "2026-02-21", "categories": []}
{"id": "2602.17793", "title": "LGD-Net: Latent-Guided Dual-Stream Network for HER2 Scoring with Task-Specific Domain Knowledge", "abstract": "It is a critical task to evalaute HER2 expression level accurately for breast cancer evaluation and targeted treatment therapy selection. However, the standard multi-step Immunohistochemistry (IHC) staining is resource-intensive, expensive, and time-consuming, which is also often unavailable in many areas. Consequently, predicting HER2 levels directly from H&E slides has emerged as a potential alternative solution. It has been shown to be effective to use virtual IHC images from H&E images for automatic HER2 scoring. However, the pixel-level virtual staining methods are computationally expensive and prone to reconstruction artifacts that can propagate diagnostic errors. To address these limitations, we propose the Latent-Guided Dual-Stream Network (LGD-Net), a novel framework that employes cross-modal feature hallucination instead of explicit pixel-level image generation. LGD-Net learns to map morphological H&E features directly to the molecular latent space, guided by a teacher IHC encoder during training. To ensure the hallucinated features capture clinically relevant phenotypes, we explicitly regularize the model training with task-specific domain knowledge, specifically nuclei distribution and membrane staining intensity, via lightweight auxiliary regularization tasks. Extensive experiments on the public BCI dataset demonstrate that LGD-Net achieves state-of-the-art performance, significantly outperforming baseline methods while enabling efficient inference using single-modality H&E inputs.", "date": "2026-02-19", "categories": []}
{"id": "2602.03998", "title": "AtlasPatch: Efficient Tissue Detection and High-throughput Patch Extraction for Computational Pathology at Scale", "abstract": "Whole-slide image (WSI) preprocessing, comprising tissue detection followed by patch extraction, is foundational to AI-driven computational pathology but remains a major bottleneck for scaling to large and heterogeneous cohorts. We present AtlasPatch, a scalable framework that couples foundation-model tissue detection with high-throughput patch extraction at minimal computational overhead. Our tissue detector achieves high precision (0.986) and remains robust across varying tissue conditions (e.g., brightness, fragmentation, boundary definition, tissue heterogeneity) and common artifacts (e.g., pen/ink markings, scanner streaks). This robustness is enabled by our annotated, heterogeneous multi-cohort training set of ~30,000 WSI thumbnails combined with efficient adaptation of the Segment-Anything (SAM) model. AtlasPatch also reduces end-to-end WSI preprocessing time by up to 16$\\times$ versus widely used deep-learning pipelines, without degrading downstream task performance. The AtlasPatch tool is open-source, efficiently parallelized for practical deployment, and supports options to save extracted patches or stream them into common feature-extraction models for on-the-fly embedding, making it adaptable to both pathology departments (tissue detection and quality control) and AI researchers (dataset creation and model training). AtlasPatch software package is available at https://github.com/AtlasAnalyticsLab/AtlasPatch.", "date": "2026-02-03", "categories": []}
{"id": "2602.16422", "title": "Automated Histopathology Report Generation via Pyramidal Feature Extraction and the UNI Foundation Model", "abstract": "Generating diagnostic text from histopathology whole slide images (WSIs) is challenging due to the gigapixel scale of the input and the requirement for precise, domain specific language. We propose a hierarchical vision language framework that combines a frozen pathology foundation model with a Transformer decoder for report generation. To make WSI processing tractable, we perform multi resolution pyramidal patch selection (downsampling factors 2^3 to 2^6) and remove background and artifacts using Laplacian variance and HSV based criteria. Patch features are extracted with the UNI Vision Transformer and projected to a 6 layer Transformer decoder that generates diagnostic text via cross attention. To better represent biomedical terminology, we tokenize the output using BioGPT. Finally, we add a retrieval based verification step that compares generated reports with a reference corpus using Sentence BERT embeddings; if a high similarity match is found, the generated report is replaced with the retrieved ground truth reference to improve reliability.", "date": "2026-02-18", "categories": []}
{"id": "2602.15138", "title": "MB-DSMIL-CL-PL: Scalable Weakly Supervised Ovarian Cancer Subtype Classification and Localisation Using Contrastive and Prototype Learning with Frozen Patch Features", "abstract": "The study of histopathological subtypes is valuable for the personalisation of effective treatment strategies for ovarian cancer. However, increasing diagnostic workloads present a challenge for UK pathology departments, leading to the rise in AI approaches. While traditional approaches in this field have relied on pre-computed, frozen image features, recent advances have shifted towards end-to-end feature extraction, providing an improvement in accuracy but at the expense of significantly reduced scalability during training and time-consuming experimentation. In this paper, we propose a new approach for subtype classification and localisation in ovarian cancer histopathology images using contrastive and prototype learning with pre-computed, frozen features via feature-space augmentations. Compared to DSMIL, our method achieves an improvement of 70.4\\% and 15.3\\% in F1 score for instance- and slide-level classification, respectively, along with AUC gains of 16.9\\% for instance localisation and 2.3\\% for slide classification, while maintaining the use of frozen patch features.", "date": "2026-02-16", "categories": []}
{"id": "2602.15783", "title": "Context-aware Skin Cancer Epithelial Cell Classification with Scalable Graph Transformers", "abstract": "Whole-slide images (WSIs) from cancer patients contain rich information that can be used for medical diagnosis or to follow treatment progress. To automate their analysis, numerous deep learning methods based on convolutional neural networks and Vision Transformers have been developed and have achieved strong performance in segmentation and classification tasks. However, due to the large size and complex cellular organization of WSIs, these models rely on patch-based representations, losing vital tissue-level context. We propose using scalable Graph Transformers on a full-WSI cell graph for classification. We evaluate this methodology on a challenging task: the classification of healthy versus tumor epithelial cells in cutaneous squamous cell carcinoma (cSCC), where both cell types exhibit very similar morphologies and are therefore difficult to differentiate for image-based approaches. We first compared image-based and graph-based methods on a single WSI. Graph Transformer models SGFormer and DIFFormer achieved balanced accuracies of $85.2 \\pm 1.5$ ($\\pm$ standard error) and $85.1 \\pm 2.5$ in 3-fold cross-validation, respectively, whereas the best image-based method reached $81.2 \\pm 3.0$. By evaluating several node feature configurations, we found that the most informative representation combined morphological and texture features as well as the cell classes of non-epithelial cells, highlighting the importance of the surrounding cellular context. We then extended our work to train on several WSIs from several patients. To address the computational constraints of image-based models, we extracted four $2560 \\times 2560$ pixel patches from each image and converted them into graphs. In this setting, DIFFormer achieved a balanced accuracy of $83.6 \\pm 1.9$ (3-fold cross-validation), while the state-of-the-art image-based model CellViT256 reached $78.1 \\pm 0.5$.", "date": "2026-02-17", "categories": []}
{"id": "2602.13944", "title": "Fusing Pixels and Genes: Spatially-Aware Learning in Computational Pathology", "abstract": "Recent years have witnessed remarkable progress in multimodal learning within computational pathology. Existing models primarily rely on vision and language modalities; however, language alone lacks molecular specificity and offers limited pathological supervision, leading to representational bottlenecks. In this paper, we propose STAMP, a Spatial Transcriptomics-Augmented Multimodal Pathology representation learning framework that integrates spatially-resolved gene expression profiles to enable molecule-guided joint embedding of pathology images and transcriptomic data. Our study shows that self-supervised, gene-guided training provides a robust and task-agnostic signal for learning pathology image representations. Incorporating spatial context and multi-scale information further enhances model performance and generalizability. To support this, we constructed SpaVis-6M, the largest Visium-based spatial transcriptomics dataset to date, and trained a spatially-aware gene encoder on this resource. Leveraging hierarchical multi-scale contrastive alignment and cross-scale patch localization mechanisms, STAMP effectively aligns spatial transcriptomics with pathology images, capturing spatial structure and molecular variation. We validate STAMP across six datasets and four downstream tasks, where it consistently achieves strong performance. These results highlight the value and necessity of integrating spatially resolved molecular supervision for advancing multimodal learning in computational pathology. The code is included in the supplementary materials. The pretrained weights and SpaVis-6M are available at: https://github.com/Hanminghao/STAMP.", "date": "2026-02-15", "categories": []}
{"id": "2602.14010", "title": "A Deployment-Friendly Foundational Framework for Efficient Computational Pathology", "abstract": "Pathology foundation models (PFMs) have enabled robust generalization in computational pathology through large-scale datasets and expansive architectures, but their substantial computational cost, particularly for gigapixel whole slide images, limits clinical accessibility and scalability. Here, we present LitePath, a deployment-friendly foundational framework designed to mitigate model over-parameterization and patch level redundancy. LitePath integrates LiteFM, a compact model distilled from three large PFMs (Virchow2, H-Optimus-1 and UNI2) using 190 million patches, and the Adaptive Patch Selector (APS), a lightweight component for task-specific patch selection. The framework reduces model parameters by 28x and lowers FLOPs by 403.5x relative to Virchow2, enabling deployment on low-power edge hardware such as the NVIDIA Jetson Orin Nano Super. On this device, LitePath processes 208 slides per hour, 104.5x faster than Virchow2, and consumes 0.36 kWh per 3,000 slides, 171x lower than Virchow2 on an RTX3090 GPU. We validated accuracy using 37 cohorts across four organs and 26 tasks (26 internal, 9 external, and 2 prospective), comprising 15,672 slides from 9,808 patients disjoint from the pretraining data. LitePath ranks second among 19 evaluated models and outperforms larger models including H-Optimus-1, mSTAR, UNI2 and GPFM, while retaining 99.71% of the AUC of Virchow2 on average. To quantify the balance between accuracy and efficiency, we propose the Deployability Score (D-Score), defined as the weighted geometric mean of normalized AUC and normalized FLOP, where LitePath achieves the highest value, surpassing Virchow2 by 10.64%. These results demonstrate that LitePath enables rapid, cost-effective and energy-efficient pathology image analysis on accessible hardware while maintaining accuracy comparable to state-of-the-art PFMs and reducing the carbon footprint of AI deployment.", "date": "2026-02-15", "categories": []}
{"id": "2602.14177", "title": "Towards Spatial Transcriptomics-driven Pathology Foundation Models", "abstract": "Spatial transcriptomics (ST) provides spatially resolved measurements of gene expression, enabling characterization of the molecular landscape of human tissue beyond histological assessment as well as localized readouts that can be aligned with morphology. Concurrently, the success of multimodal foundation models that integrate vision with complementary modalities suggests that morphomolecular coupling between local expression and morphology can be systematically used to improve histological representations themselves. We introduce Spatial Expression-Aligned Learning (SEAL), a vision-omics self-supervised learning framework that infuses localized molecular information into pathology vision encoders. Rather than training new encoders from scratch, SEAL is designed as a parameter-efficient vision-omics finetuning method that can be flexibly applied to widely used pathology foundation models. We instantiate SEAL by training on over 700,000 paired gene expression spot-tissue region examples spanning tumor and normal samples from 14 organs. Tested across 38 slide-level and 15 patch-level downstream tasks, SEAL provides a drop-in replacement for pathology foundation models that consistently improves performance over widely used vision-only and ST prediction baselines on slide-level molecular status, pathway activity, and treatment response prediction, as well as patch-level gene expression prediction tasks. Additionally, SEAL encoders exhibit robust domain generalization on out-of-distribution evaluations and enable new cross-modal capabilities such as gene-to-image retrieval. Our work proposes a general framework for ST-guided finetuning of pathology foundation models, showing that augmenting existing models with localized molecular supervision is an effective and practical step for improving visual representations and expanding their cross-modal utility.", "date": "2026-02-15", "categories": []}
{"id": "2602.14501", "title": "Prototype Instance-semantic Disentanglement with Low-rank Regularized Subspace Clustering for WSIs Explainable Recognition", "abstract": "The tumor region plays a key role in pathological diagnosis. Tumor tissues are highly similar to precancerous lesions and non tumor instances often greatly exceed tumor instances in whole slide images (WSIs). These issues cause instance-semantic entanglement in multi-instance learning frameworks, degrading both model representation capability and interpretability. To address this, we propose an end-to-end prototype instance semantic disentanglement framework with low-rank regularized subspace clustering, PID-LRSC, in two aspects. First, we use secondary instance subspace learning to construct low-rank regularized subspace clustering (LRSC), addressing instance entanglement caused by an excessive proportion of non tumor instances. Second, we employ enhanced contrastive learning to design prototype instance semantic disentanglement (PID), resolving semantic entanglement caused by the high similarity between tumor and precancerous tissues. We conduct extensive experiments on multicentre pathology datasets, implying that PID-LRSC outperforms other SOTA methods. Overall, PID-LRSC provides clearer instance semantics during decision-making and significantly enhances the reliability of auxiliary diagnostic outcomes.", "date": "2026-02-16", "categories": []}
{"id": "2602.14509", "title": "MacNet: An End-to-End Manifold-Constrained Adaptive Clustering Network for Interpretable Whole Slide Image Classification", "abstract": "Whole slide images (WSIs) are the gold standard for pathological diagnosis and sub-typing. Current main-stream two-step frameworks employ offline feature encoders trained without domain-specific knowledge. Among them, attention-based multiple instance learning (MIL) methods are outcome-oriented and offer limited interpretability. Clustering-based approaches can provide explainable decision-making process but suffer from high dimension features and semantically ambiguous centroids. To this end, we propose an end-to-end MIL framework that integrates Grassmann re-embedding and manifold adaptive clustering, where the manifold geometric structure facilitates robust clustering results. Furthermore, we design a prior knowledge guiding proxy instance labeling and aggregation strategy to approximate patch labels and focus on pathologically relevant tumor regions. Experiments on multicentre WSI datasets demonstrate that: 1) our cluster-incorporated model achieves superior performance in both grading accuracy and interpretability; 2) end-to-end learning refines better feature representations and it requires acceptable computation resources.", "date": "2026-02-16", "categories": []}
{"id": "2502.17503", "title": "Doctor-in-the-Loop: An Explainable, Multi-View Deep Learning Framework for Predicting Pathological Response in Non-Small Cell Lung Cancer", "abstract": "Non-small cell lung cancer (NSCLC) remains a major global health challenge, with high post-surgical recurrence rates underscoring the need for accurate pathological response predictions to guide personalized treatments. Although artificial intelligence models show promise in this domain, their clinical adoption is limited by the lack of medically grounded guidance during training, often resulting in non-explainable intrinsic predictions. To address this, we propose Doctor-in-the-Loop, a novel framework that integrates expert-driven domain knowledge with explainable artificial intelligence techniques, directing the model toward clinically relevant anatomical regions and improving both interpretability and trustworthiness. Our approach employs a gradual multi-view strategy, progressively refining the model's focus from broad contextual features to finer, lesion-specific details. By incorporating domain insights at every stage, we enhance predictive accuracy while ensuring that the model's decision-making process aligns more closely with clinical reasoning. Evaluated on a dataset of NSCLC patients, Doctor-in-the-Loop delivers promising predictive performance and provides transparent, justifiable outputs, representing a significant step toward clinically explainable artificial intelligence in oncology.", "date": "2025-02-21", "categories": []}
{"id": "2503.00736", "title": "Unifying Multiple Foundation Models for Advanced Computational Pathology", "abstract": "Foundation models have substantially advanced computational pathology by learning transferable visual representations from large histological datasets, yet their performance varies widely across tasks due to differences in training data composition and reliance on proprietary datasets that cannot be cumulatively expanded. Existing efforts to combine foundation models through offline distillation partially mitigate this issue but require dedicated distillation data and repeated retraining to integrate new models. Here we present Shazam, an online integration model that adaptively combines multiple pretrained pathology foundation models within a unified and scalable representation learning paradigm. Our findings show that fusing multi-level features through adaptive expert weighting and online distillation enables efficient consolidation of complementary model strengths without additional pretraining. Across spatial transcriptomics prediction, survival prognosis, tile-level classification, and visual question answering, Shazam consistently outperforms strong individual models, demonstrating that online model integration provides a practical and extensible strategy for advancing computational pathology.", "date": "2025-03-02", "categories": []}
{"id": "2508.21418", "title": "From slides to AI-ready maps: Standardized multi-layer tissue maps as metadata for artificial intelligence in digital pathology", "abstract": "A Whole Slide Image (WSI) is a high-resolution digital image created by scanning an entire glass slide containing a biological specimen, such as tissue sections or cell samples, at multiple magnifications. These images are digitally viewable, analyzable, and shareable, and are widely used for Artificial Intelligence (AI) algorithm development. WSIs play an important role in pathology for disease diagnosis and oncology for cancer research, but are also applied in neurology, veterinary medicine, hematology, microbiology, dermatology, pharmacology, toxicology, immunology, and forensic science. When assembling cohorts for AI training or validation, it is essential to know the content of a WSI. However, no standard currently exists for this metadata, and such a selection has largely relied on manual inspection, which is not suitable for large collections with millions of objects. We propose a general framework to generate 2D index maps (tissue maps) that describe the morphological content of WSIs using common syntax and semantics to achieve interoperability between catalogs. The tissue maps are structured in three layers: source, tissue type, and pathological alterations. Each layer assigns WSI segments to specific classes, providing AI-ready metadata. We demonstrate the advantages of this standard by applying AI-based metadata extraction from WSIs to generate tissue maps and integrating them into a WSI archive. This integration enhances search capabilities within WSI archives, thereby facilitating the accelerated assembly of high-quality, balanced, and more targeted datasets for AI training, validation, and cancer research.", "date": "2025-08-29", "categories": []}
{"id": "2602.08652", "title": "Deep Learning-Based Fixation Type Prediction for Quality Assurance in Digital Pathology", "abstract": "Accurate annotation of fixation type is a critical step in slide preparation for pathology laboratories. However, this manual process is prone to\n  errors, impacting downstream analyses and diagnostic accuracy. Existing methods for verifying formalin-fixed, paraffin-embedded (FFPE), and frozen\n  section (FS) fixation types typically require full-resolution whole-slide images (WSIs), limiting scalability for high-throughput quality control.\n  We propose a deep-learning model to predict fixation types using low-resolution, pre-scan thumbnail images. The model was trained on WSIs from\n  the TUM Institute of Pathology (n=1,200, Leica GT450DX) and evaluated on a class-balanced subset of The Cancer Genome Atlas dataset (TCGA, n=8,800,\n  Leica AT2), as well as on class-balanced datasets from Augsburg (n=695 [392 FFPE, 303 FS], Philips UFS) and Regensburg (n=202, 3DHISTECH P1000).\n  Our model achieves an AUROC of 0.88 on TCGA, outperforming comparable pre-scan methods by 4.8%. It also achieves AUROCs of 0.72 on Regensburg and\n  Augsburg slides, underscoring challenges related to scanner-induced domain shifts. Furthermore, the model processes each slide in 21 ms, $400\\times$\n  faster than existing high-magnification, full-resolution methods, enabling rapid, high-throughput processing.\n  This approach provides an efficient solution for detecting labelling errors without relying on high-magnification scans, offering a valuable tool for\n  quality control in high-throughput pathology workflows. Future work will improve and evaluate the model's generalisation to additional scanner\n  types. Our findings suggest that this method can increase accuracy and efficiency in digital pathology workflows and may be extended to other\n  low-resolution slide annotations.", "date": "2026-02-09", "categories": []}
{"id": "2602.09477", "title": "Weakly Supervised Contrastive Learning for Histopathology Patch Embeddings", "abstract": "Digital histopathology whole slide images (WSIs) provide gigapixel-scale high-resolution images that are highly useful for disease diagnosis. However, digital histopathology image analysis faces significant challenges due to the limited training labels, since manually annotating specific regions or small patches cropped from large WSIs requires substantial time and effort. Weakly supervised multiple instance learning (MIL) offers a practical and efficient solution by requiring only bag-level (slide-level) labels, while each bag typically contains multiple instances (patches). Most MIL methods directly use frozen image patch features generated by various image encoders as inputs and primarily focus on feature aggregation. However, feature representation learning for encoder pretraining in MIL settings has largely been neglected.\n  In our work, we propose a novel feature representation learning framework called weakly supervised contrastive learning (WeakSupCon) that incorporates bag-level label information during training. Our method does not rely on instance-level pseudo-labeling, yet it effectively separates patches with different labels in the feature space. Experimental results demonstrate that the image features generated by our WeakSupCon method lead to improved downstream MIL performance compared to self-supervised contrastive learning approaches in three datasets. Our related code is available at github.com/BzhangURU/Paper_WeakSupCon_for_MIL", "date": "2026-02-10", "categories": []}
{"id": "2602.09523", "title": "Singpath-VL Technical Report", "abstract": "We present Singpath-VL, a vision-language large model, to fill the vacancy of AI assistant in cervical cytology. Recent advances in multi-modal large language models (MLLMs) have significantly propelled the field of computational pathology. However, their application in cytopathology, particularly cervical cytology, remains underexplored, primarily due to the scarcity of large-scale, high-quality annotated datasets. To bridge this gap, we first develop a novel three-stage pipeline to synthesize a million-scale image-description dataset. The pipeline leverages multiple general-purpose MLLMs as weak annotators, refines their outputs through consensus fusion and expert knowledge injection, and produces high-fidelity descriptions of cell morphology. Using this dataset, we then fine-tune the Qwen3-VL-4B model via a multi-stage strategy to create a specialized cytopathology MLLM. The resulting model, named Singpath-VL, demonstrates superior performance in fine-grained morphological perception and cell-level diagnostic classification. To advance the field, we will open-source a portion of the synthetic dataset and benchmark.", "date": "2026-02-10", "categories": []}
{"id": "2503.05678", "title": "Towards Effective and Efficient Context-aware Nucleus Detection in Histopathology Whole Slide Images", "abstract": "Nucleus detection in histopathology whole slide images (WSIs) is crucial for a broad spectrum of clinical applications. The gigapixel size of WSIs necessitates the use of sliding window methodology for nucleus detection. However, mainstream methods process each sliding window independently, which overlooks broader contextual information and easily leads to inaccurate predictions. To address this limitation, recent studies additionally crop a large Filed-of-View (LFoV) patch centered on each sliding window to extract contextual features. However, such methods substantially increase whole-slide inference latency. In this work, we propose an effective and efficient context-aware nucleus detection approach. Specifically, instead of using LFoV patches, we aggregate contextual clues from off-the-shelf features of historically visited sliding windows, which greatly enhances the inference efficiency. Moreover, compared to LFoV patches used in previous works, the sliding window patches have higher magnification and provide finer-grained tissue details, thereby enhancing the classification accuracy. To develop the proposed context-aware model, we utilize annotated patches along with their surrounding unlabeled patches for training. Beyond exploiting high-level tissue context from these surrounding regions, we design a post-training strategy that leverages abundant unlabeled nucleus samples within them to enhance the model's context adaptability. Extensive experimental results on three challenging benchmarks demonstrate the superiority of our method.", "date": "2025-03-04", "categories": []}
{"id": "2507.05077", "title": "Sequential Attention-based Sampling for Histopathological Analysis", "abstract": "Deep neural networks are increasingly applied in automated histopathology. Yet, whole-slide images (WSIs) are often acquired at gigapixel sizes, rendering them computationally infeasible to analyze entirely at high resolution. Diagnostic labels are largely available only at the slide-level, because expert annotation of images at a finer (patch) level is both laborious and expensive. Moreover, regions with diagnostic information typically occupy only a small fraction of the WSI, making it inefficient to examine the entire slide at full resolution. Here, we propose SASHA -- Sequential Attention-based Sampling for Histopathological Analysis -- a deep reinforcement learning approach for efficient analysis of histopathological images. First, SASHA learns informative features with a lightweight hierarchical, attention-based multiple instance learning (MIL) model. Second, SASHA samples intelligently and zooms selectively into a small fraction (10-20\\%) of high-resolution patches to achieve reliable diagnoses. We show that SASHA matches state-of-the-art methods that analyze the WSI fully at high resolution, albeit at a fraction of their computational and memory costs. In addition, it significantly outperforms competing, sparse sampling methods. We propose SASHA as an intelligent sampling model for medical imaging challenges that involve automated diagnosis with exceptionally large images containing sparsely informative features. Model implementation is available at: https://github.com/coglabiisc/SASHA.", "date": "2025-07-07", "categories": []}
{"id": "2601.21560", "title": "HistoPrism: Unlocking Functional Pathway Analysis from Pan-Cancer Histology via Gene Expression Prediction", "abstract": "Predicting spatial gene expression from H&E histology offers a scalable and clinically accessible alternative to sequencing, but realizing clinical impact requires models that generalize across cancer types and capture biologically coherent signals. Prior work is often limited to per-cancer settings and variance-based evaluation, leaving functional relevance underexplored. We introduce HistoPrism, an efficient transformer-based architecture for pan-cancer prediction of gene expression from histology. To evaluate biological meaning, we introduce a pathway-level benchmark, shifting assessment from isolated gene-level variance to coherent functional pathways. HistoPrism not only surpasses prior state-of-the-art models on highly variable genes , but also more importantly, achieves substantial gains on pathway-level prediction, demonstrating its ability to recover biologically coherent transcriptomic patterns. With strong pan-cancer generalization and improved efficiency, HistoPrism establishes a new standard for clinically relevant transcriptomic modeling from routinely available histology.", "date": "2026-01-29", "categories": []}
{"id": "2505.04672", "title": "Histo-Miner: Deep learning based tissue features extraction pipeline from H&E whole slide images of cutaneous squamous cell carcinoma", "abstract": "Recent advancements in digital pathology have enabled comprehensive analysis of Whole-Slide Images (WSI) from tissue samples, leveraging high-resolution microscopy and computational capabilities. Despite this progress, there is a lack of labeled datasets and open source pipelines specifically tailored for analysis of skin tissue. Here we propose Histo-Miner, a deep learning-based pipeline for analysis of skin WSIs and generate two datasets with labeled nuclei and tumor regions. We develop our pipeline for the analysis of patient samples of cutaneous squamous cell carcinoma (cSCC), a frequent non-melanoma skin cancer. Utilizing the two datasets, comprising 47,392 annotated cell nuclei and 144 tumor-segmented WSIs respectively, both from cSCC patients, Histo-Miner employs convolutional neural networks and vision transformers for nucleus segmentation and classification as well as tumor region segmentation. Performance of trained models positively compares to state of the art with multi-class Panoptic Quality (mPQ) of 0.569 for nucleus segmentation, macro-averaged F1 of 0.832 for nucleus classification and mean Intersection over Union (mIoU) of 0.907 for tumor region segmentation. From these predictions we generate a compact feature vector summarizing tissue morphology and cellular interactions, which can be used for various downstream tasks. Here, we use Histo-Miner to predict cSCC patient response to immunotherapy based on pre-treatment WSIs from 45 patients. Histo-Miner identifies percentages of lymphocytes, the granulocyte to lymphocyte ratio in tumor vicinity and the distances between granulocytes and plasma cells in tumors as predictive features for therapy response. This highlights the applicability of Histo-Miner to clinically relevant scenarios, providing direct interpretation of the classification and insights into the underlying biology.", "date": "2025-05-07", "categories": []}
{"id": "2602.00151", "title": "Investigating the Impact of Histopathological Foundation Models on Regressive Prediction of Homologous Recombination Deficiency", "abstract": "Foundation models pretrained on large-scale histopathology data have found great success in various fields of computational pathology, but their impact on regressive biomarker prediction remains underexplored. In this work, we systematically evaluate histopathological foundation models for regression-based tasks, demonstrated through the prediction of homologous recombination deficiency (HRD) score - a critical biomarker for personalized cancer treatment. Within multiple instance learning frameworks, we extract patch-level features from whole slide images (WSI) using five state-of-the-art foundation models, and evaluate their impact compared to contrastive learning-based features. Models are trained to predict continuous HRD scores based on these extracted features across breast, endometrial, and lung cancer cohorts from two public medical data collections. Extensive experiments demonstrate that models trained on foundation model features consistently outperform the baseline in terms of predictive accuracy and generalization capabilities while exhibiting systematic differences among the foundation models. Additionally, we propose a distribution-based upsampling strategy to mitigate target imbalance in these datasets, significantly improving the recall and balanced accuracy for underrepresented but clinically important patient populations. Furthermore, we investigate the impact of different sampling strategies and instance bagsizes by ablation studies. Our results highlight the benefits of large-scale histopathological pretraining for more precise and transferable regressive biomarker prediction, showcasing its potential to advance AI-driven precision oncology.", "date": "2026-01-29", "categories": []}
{"id": "2602.05420", "title": "Disco: Densely-overlapping Cell Instance Segmentation via Adjacency-aware Collaborative Coloring", "abstract": "Accurate cell instance segmentation is foundational for digital pathology analysis. Existing methods based on contour detection and distance mapping still face significant challenges in processing complex and dense cellular regions. Graph coloring-based methods provide a new paradigm for this task, yet the effectiveness of this paradigm in real-world scenarios with dense overlaps and complex topologies has not been verified. Addressing this issue, we release a large-scale dataset GBC-FS 2025, which contains highly complex and dense sub-cellular nuclear arrangements. We conduct the first systematic analysis of the chromatic properties of cell adjacency graphs across four diverse datasets and reveal an important discovery: most real-world cell graphs are non-bipartite, with a high prevalence of odd-length cycles (predominantly triangles). This makes simple 2-coloring theory insufficient for handling complex tissues, while higher-chromaticity models would cause representational redundancy and optimization difficulties. Building on this observation of complex real-world contexts, we propose Disco (Densely-overlapping Cell Instance Segmentation via Adjacency-aware COllaborative Coloring), an adjacency-aware framework based on the \"divide and conquer\" principle. It uniquely combines a data-driven topological labeling strategy with a constrained deep learning system to resolve complex adjacency conflicts. First, \"Explicit Marking\" strategy transforms the topological challenge into a learnable classification task by recursively decomposing the cell graph and isolating a \"conflict set.\" Second, \"Implicit Disambiguation\" mechanism resolves ambiguities in conflict regions by enforcing feature dissimilarity between different instances, enabling the model to learn separable feature representations.", "date": "2026-02-05", "categories": []}
{"id": "2512.10326", "title": "StainNet: Scaling Self-Supervised Foundation Models on Immunohistochemistry and Special Stains for Computational Pathology", "abstract": "Foundation models trained with self-supervised learning (SSL) on large-scale histological images have significantly accelerated the development of computational pathology. These models can serve as backbones for region-of-interest (ROI) image analysis or patch-level feature extractors in whole-slide images (WSIs) based on multiple instance learning (MIL). Existing pathology foundation models (PFMs) are typically pre-trained on Hematoxylin-Eosin (H\\&E) stained pathology images. However, images such as immunohistochemistry (IHC) and special stains are also frequently used in clinical practice. PFMs pre-trained mainly on H\\&E-stained images may be limited in clinical applications involving these non-H\\&E images. To address this issue, we propose StainNet, a collection of self-supervised foundation models specifically trained for IHC and special stains in pathology images based on the vision transformer (ViT) architecture. StainNet contains a ViT-Small and a ViT-Base model, both of which are trained using a self-distillation SSL approach on over 1.4 million patch images extracted from 20,231 publicly available IHC and special staining WSIs in the HISTAI database. To evaluate StainNet models, we conduct experiments on three in-house slide-level IHC classification tasks, three in-house ROI-level special stain and two public ROI-level IHC classification tasks to demonstrate their strong ability. We also perform ablation studies such as few-ratio learning and retrieval evaluations, and compare StainNet models with recent larger PFMs to further highlight their strengths. The StainNet model weights are available at https://github.com/WonderLandxD/StainNet.", "date": "2025-12-11", "categories": []}
{"id": "2601.20347", "title": "MMSF: Multitask and Multimodal Supervised Framework for WSI Classification and Survival Analysis", "abstract": "Multimodal evidence is critical in computational pathology: gigapixel whole slide images capture tumor morphology, while patient-level clinical descriptors preserve complementary context for prognosis. Integrating such heterogeneous signals remains challenging because feature spaces exhibit distinct statistics and scales. We introduce MMSF, a multitask and multimodal supervised framework built on a linear-complexity MIL backbone that explicitly decomposes and fuses cross-modal information. MMSF comprises a graph feature extraction module embedding tissue topology at the patch level, a clinical data embedding module standardizing patient attributes, a feature fusion module aligning modality-shared and modality-specific representations, and a Mamba-based MIL encoder with multitask prediction heads. Experiments on CAMELYON16 and TCGA-NSCLC demonstrate 2.1--6.6\\% accuracy and 2.2--6.9\\% AUC improvements over competitive baselines, while evaluations on five TCGA survival cohorts yield 7.1--9.8\\% C-index improvements compared with unimodal methods and 5.6--7.1\\% over multimodal alternatives.", "date": "2026-01-28", "categories": []}
{"id": "2602.04063", "title": "iSight: Towards expert-AI co-assessment for improved immunohistochemistry staining interpretation", "abstract": "Immunohistochemistry (IHC) provides information on protein expression in tissue sections and is commonly used to support pathology diagnosis and disease triage. While AI models for H\\&E-stained slides show promise, their applicability to IHC is limited due to domain-specific variations. Here we introduce HPA10M, a dataset that contains 10,495,672 IHC images from the Human Protein Atlas with comprehensive metadata included, and encompasses 45 normal tissue types and 20 major cancer types. Based on HPA10M, we trained iSight, a multi-task learning framework for automated IHC staining assessment. iSight combines visual features from whole-slide images with tissue metadata through a token-level attention mechanism, simultaneously predicting staining intensity, location, quantity, tissue type, and malignancy status. On held-out data, iSight achieved 85.5\\% accuracy for location, 76.6\\% for intensity, and 75.7\\% for quantity, outperforming fine-tuned foundation models (PLIP, CONCH) by 2.5--10.2\\%. In addition, iSight demonstrates well-calibrated predictions with expected calibration errors of 0.0150-0.0408. Furthermore, in a user study with eight pathologists evaluating 200 images from two datasets, iSight outperformed initial pathologist assessments on the held-out HPA dataset (79\\% vs 68\\% for location, 70\\% vs 57\\% for intensity, 68\\% vs 52\\% for quantity). Inter-pathologist agreement also improved after AI assistance in both held-out HPA (Cohen's $\\kappa$ increased from 0.63 to 0.70) and Stanford TMAD datasets (from 0.74 to 0.76), suggesting expert--AI co-assessment can improve IHC interpretation. This work establishes a foundation for AI systems that can improve IHC diagnostic accuracy and highlights the potential for integrating iSight into clinical workflows to enhance the consistency and reliability of IHC assessment.", "date": "2026-02-03", "categories": []}
{"id": "2602.02918", "title": "A Multi-scale Linear-time Encoder for Whole-Slide Image Analysis", "abstract": "We introduce Multi-scale Adaptive Recurrent Biomedical Linear-time Encoder (MARBLE), the first \\textit{purely Mamba-based} multi-state multiple instance learning (MIL) framework for whole-slide image (WSI) analysis. MARBLE processes multiple magnification levels in parallel and integrates coarse-to-fine reasoning within a linear-time state-space model, efficiently capturing cross-scale dependencies with minimal parameter overhead. WSI analysis remains challenging due to gigapixel resolutions and hierarchical magnifications, while existing MIL methods typically operate at a single scale and transformer-based approaches suffer from quadratic attention costs. By coupling parallel multi-scale processing with linear-time sequence modeling, MARBLE provides a scalable and modular alternative to attention-based architectures. Experiments on five public datasets show improvements of up to \\textbf{6.9\\%} in AUC, \\textbf{20.3\\%} in accuracy, and \\textbf{2.3\\%} in C-index, establishing MARBLE as an efficient and generalizable framework for multi-scale WSI analysis.", "date": "2026-02-02", "categories": []}
{"id": "2508.14779", "title": "Hospital-Specific Bias in Patch-Based Pathology Models", "abstract": "Pathology foundation models (PFMs) achieve strong performance on diverse histopathology tasks, but their sensitivity to hospital-specific domain shifts remains underexplored. We systematically evaluate state-of-the-art PFMs on TCGA patch-level datasets and introduce a lightweight adversarial adaptor to remove hospital-related domain information from latent representations. Experiments show that, while disease classification accuracy is largely maintained, the adaptor effectively reduces hospital-specific bias, as confirmed by t-SNE visualizations. Our study establishes a benchmark for assessing cross-hospital robustness in PFMs and provides a practical strategy for enhancing generalization under heterogeneous clinical settings. Our code is available at https://github.com/MengRes/pfm_domain_bias.", "date": "2025-08-20", "categories": []}
{"id": "2602.02124", "title": "Toxicity Assessment in Preclinical Histopathology via Class-Aware Mahalanobis Distance for Known and Novel Anomalies", "abstract": "Drug-induced toxicity remains a leading cause of failure in preclinical development and early clinical trials. Detecting adverse effects at an early stage is critical to reduce attrition and accelerate the development of safe medicines. Histopathological evaluation remains the gold standard for toxicity assessment, but it relies heavily on expert pathologists, creating a bottleneck for large-scale screening. To address this challenge, we introduce an AI-based anomaly detection framework for histopathological whole-slide images (WSIs) in rodent livers from toxicology studies. The system identifies healthy tissue and known pathologies (anomalies) for which training data is available. In addition, it can detect rare pathologies without training data as out-of-distribution (OOD) findings. We generate a novel dataset of pixelwise annotations of healthy tissue and known pathologies and use this data to fine-tune a pre-trained Vision Transformer (DINOv2) via Low-Rank Adaptation (LoRA) in order to do tissue segmentation. Finally, we extract features for OOD detection using the Mahalanobis distance. To better account for class-dependent variability in histological data, we propose the use of class-specific thresholds. We optimize the thresholds using the mean of the false negative and false positive rates, resulting in only 0.16\\% of pathological tissue classified as healthy and 0.35\\% of healthy tissue classified as pathological. Applied to mouse liver WSIs with known toxicological findings, the framework accurately detects anomalies, including rare OOD morphologies. This work demonstrates the potential of AI-driven histopathology to support preclinical workflows, reduce late-stage failures, and improve efficiency in drug development.", "date": "2026-02-02", "categories": []}
{"id": "2412.00678", "title": "2DMamba: Efficient State Space Model for Image Representation with Applications on Giga-Pixel Whole Slide Image Classification", "abstract": "Efficiently modeling large 2D contexts is essential for various fields including Giga-Pixel Whole Slide Imaging (WSI) and remote sensing. Transformer-based models offer high parallelism but face challenges due to their quadratic complexity for handling long sequences. Recently, Mamba introduced a selective State Space Model (SSM) with linear complexity and high parallelism, enabling effective and efficient modeling of wide context in 1D sequences. However, extending Mamba to vision tasks, which inherently involve 2D structures, results in spatial discrepancies due to the limitations of 1D sequence processing. On the other hand, current 2D SSMs inherently model 2D structures but they suffer from prohibitively slow computation due to the lack of efficient parallel algorithms. In this work, we propose 2DMamba, a novel 2D selective SSM framework that incorporates the 2D spatial structure of images into Mamba, with a highly optimized hardware-aware operator, adopting both spatial continuity and computational efficiency. We validate the versatility of our approach on both WSIs and natural images. Extensive experiments on 10 public datasets for WSI classification and survival analysis show that 2DMamba improves up to 2.48% in AUC, 3.11% in F1 score, 2.47% in accuracy and 5.52% in C-index. Additionally, integrating our method with VMamba for natural imaging yields 0.5 to 0.7 improvements in mIoU on the ADE20k semantic segmentation dataset, and 0.2% accuracy improvement on ImageNet-1K classification dataset. Our code is available at https://github.com/AtlasAnalyticsLab/2DMamba.", "date": "2024-12-01", "categories": []}
{"id": "2503.17564", "title": "ModalTune: Fine-Tuning Slide-Level Foundation Models with Multi-Modal Information for Multi-task Learning in Digital Pathology", "abstract": "Prediction tasks in digital pathology are challenging due to the massive size of whole-slide images (WSIs) and the weak nature of training signals. Advances in computing, data availability, and self-supervised learning (SSL) have paved the way for slide-level foundation models (SLFMs) that can improve prediction tasks in low-data regimes. However, current methods under-utilize shared information between tasks and modalities. To overcome this challenge, we propose ModalTune, a novel fine-tuning framework which introduces the Modal Adapter to integrate new modalities without modifying SLFM weights. Additionally, we use large-language models (LLMs) to encode labels as text, capturing semantic relationships across multiple tasks and cancer types in a single training recipe. ModalTune achieves state-of-the-art (SOTA) results against both uni-modal and multi-modal models across four cancer types, jointly improving survival and cancer subtype prediction while remaining competitive in pan-cancer settings. Additionally, we show ModalTune is generalizable to two out-of-distribution (OOD) datasets. To our knowledge, this is the first unified fine-tuning framework for multi-modal, multi-task, and pan-cancer modeling in digital pathology.", "date": "2025-03-21", "categories": []}
{"id": "2601.21022", "title": "AI-based Prediction of Biochemical Recurrence from Biopsy and Prostatectomy Samples", "abstract": "Biochemical recurrence (BCR) after radical prostatectomy (RP) is a surrogate marker for aggressive prostate cancer with adverse outcomes, yet current prognostic tools remain imprecise. We trained an AI-based model on diagnostic prostate biopsy slides from the STHLM3 cohort (n = 676) to predict patient-specific risk of BCR, using foundation models and attention-based multiple instance learning. Generalizability was assessed across three external RP cohorts: LEOPARD (n = 508), CHIMERA (n = 95), and TCGA-PRAD (n = 379). The image-based approach achieved 5-year time-dependent AUCs of 0.64, 0.70, and 0.70, respectively. Integrating clinical variables added complementary prognostic value and enabled statistically significant risk stratification. Compared with guideline-based CAPRA-S, AI incrementally improved postoperative prognostication. These findings suggest biopsy-trained histopathology AI can generalize across specimen types to support preoperative and postoperative decision making, but the added value of AI-based multimodal approaches over simpler predictive models should be critically scrutinized in further studies.", "date": "2026-01-28", "categories": []}
{"id": "2601.21334", "title": "Do Pathology Foundation Models Encode Disease Progression? A Pseudotime Analysis of Visual Representations", "abstract": "Vision foundation models trained on discretely sampled images achieve strong performance on classification benchmarks, yet whether their representations encode the continuous processes underlying their training data remains unclear. This question is especially pertinent in computational pathology, where we posit that models whose latent representations implicitly capture continuous disease progression may better reflect underlying biology, support more robust generalization, and enable quantitative analyses of features associated with disease transitions. Using diffusion pseudotime, a method developed to infer developmental trajectories from single-cell transcriptomics, we probe whether foundation models organize disease states along coherent progression directions in representation space. Across four cancer progressions and six models, we find that all pathology-specific models recover trajectory orderings significantly exceeding null baselines, with vision-only models achieving the highest fidelities $(\\tau > 0.78$ on CRC-Serrated). Model rankings by trajectory fidelity on reference diseases strongly predict few-shot classification performance on held-out diseases ($\\rho = 0.92$), and exploratory analysis shows cell-type composition varies smoothly along inferred trajectories in patterns consistent with known stromal remodeling. Together, these results demonstrate that vision foundation models can implicitly learn to represent continuous processes from independent static observations, and that trajectory fidelity provides a complementary measure of representation quality beyond downstream performance. While demonstrated in pathology, this framework could be applied to other domains where continuous processes are observed through static snapshots.", "date": "2026-01-29", "categories": []}
{"id": "2601.21617", "title": "PathReasoner-R1: Instilling Structured Reasoning into Pathology Vision-Language Model via Knowledge-Guided Policy Optimization", "abstract": "Vision-Language Models (VLMs) are advancing computational pathology with superior visual understanding capabilities. However, current systems often reduce diagnosis to directly output conclusions without verifiable evidence-linked reasoning, which severely limits clinical trust and hinders expert error rectification. To address these barriers, we construct PathReasoner, the first large-scale dataset of whole-slide image (WSI) reasoning. Unlike previous work reliant on unverified distillation, we develop a rigorous knowledge-guided generation pipeline. By leveraging medical knowledge graphs, we explicitly align structured pathological findings and clinical reasoning with diagnoses, generating over 20K high-quality instructional samples. Based on the database, we propose PathReasoner-R1, which synergizes trajectory-masked supervised fine-tuning with reasoning-oriented reinforcement learning to instill structured chain-of-thought capabilities. To ensure medical rigor, we engineer a knowledge-aware multi-granular reward function incorporating an Entity Reward mechanism strictly aligned with knowledge graphs. This effectively guides the model to optimize for logical consistency rather than mere outcome matching, thereby enhancing robustness. Extensive experiments demonstrate that PathReasoner-R1 achieves state-of-the-art performance on both PathReasoner and public benchmarks across various image scales, equipping pathology models with transparent, clinically grounded reasoning capabilities. Dataset and code are available at https://github.com/cyclexfy/PathReasoner-R1.", "date": "2026-01-29", "categories": []}
{"id": "2412.13126", "title": "Knowledge-enhanced Pretraining for Vision-language Pathology Foundation Model on Cancer Diagnosis", "abstract": "Vision-language foundation models have shown great promise in computational pathology but remain primarily data-driven, lacking explicit integration of medical knowledge. We introduce KEEP (KnowledgE-Enhanced Pathology), a foundation model that systematically incorporates disease knowledge into pretraining for cancer diagnosis. KEEP leverages a comprehensive disease knowledge graph encompassing 11,454 diseases and 139,143 attributes to reorganize millions of pathology image-text pairs into 143,000 semantically structured groups aligned with disease ontology hierarchies. This knowledge-enhanced pretraining aligns visual and textual representations within hierarchical semantic spaces, enabling deeper understanding of disease relationships and morphological patterns. Across 18 public benchmarks (over 14,000 whole-slide images) and 4 institutional rare cancer datasets (926 cases), KEEP consistently outperformed existing foundation models, showing substantial gains for rare subtypes. These results establish knowledge-enhanced vision-language modeling as a powerful paradigm for advancing computational pathology.", "date": "2024-12-17", "categories": []}
{"id": "2507.05742", "title": "Whole Slide Concepts: A Supervised Foundation Model For Pathological Images", "abstract": "Foundation models (FMs) are transforming computational pathology by offering new ways to analyze histopathology images. However, FMs typically require weeks of training on large databases, making their creation a resource-intensive process. In this paper, we present a training for foundation models from whole slide images using supervised, end-to-end, multitask learning on slide-level labels. Notably, it is the first model to incorporate cancer subtyping, risk estimation, and genetic mutation prediction into one model. The presented model outperforms self-supervised models on seven benchmark tasks while the training only required 5% of the computational resources. The results not only show that supervised training can outperform self-supervision with less data, but also offer a solution to annotation problems, as patient-based labels are widely available through routine clinical processes. Furthermore, an attention module provides a layer of explainability across different tasks and serves as a tumor detector for unseen cancer types. To address the issue of closed-source datasets, the model was fully trained on openly available data. The code and model weights are made available under https://github.com/FraunhoferMEVIS/MedicalMultitaskModeling.", "date": "2025-07-08", "categories": []}
{"id": "2601.16631", "title": "PanopMamba: Vision State Space Modeling for Nuclei Panoptic Segmentation", "abstract": "Nuclei panoptic segmentation supports cancer diagnostics by integrating both semantic and instance segmentation of different cell types to analyze overall tissue structure and individual nuclei in histopathology images. Major challenges include detecting small objects, handling ambiguous boundaries, and addressing class imbalance. To address these issues, we propose PanopMamba, a novel hybrid encoder-decoder architecture that integrates Mamba and Transformer with additional feature-enhanced fusion via state space modeling. We design a multiscale Mamba backbone and a State Space Model (SSM)-based fusion network to enable efficient long-range perception in pyramid features, thereby extending the pure encoder-decoder framework while facilitating information sharing across multiscale features of nuclei. The proposed SSM-based feature-enhanced fusion integrates pyramid feature networks and dynamic feature enhancement across different spatial scales, enhancing the feature representation of densely overlapping nuclei in both semantic and spatial dimensions. To the best of our knowledge, this is the first Mamba-based approach for panoptic segmentation. Additionally, we introduce alternative evaluation metrics, including image-level Panoptic Quality ($i$PQ), boundary-weighted PQ ($w$PQ), and frequency-weighted PQ ($fw$PQ), which are specifically designed to address the unique challenges of nuclei segmentation and thereby mitigate the potential bias inherent in vanilla PQ. Experimental evaluations on two multiclass nuclei segmentation benchmark datasets, MoNuSAC2020 and NuInsSeg, demonstrate the superiority of PanopMamba for nuclei panoptic segmentation over state-of-the-art methods. Consequently, the robustness of PanopMamba is validated across various metrics, while the distinctiveness of PQ variants is also demonstrated. Code is available at https://github.com/mkang315/PanopMamba.", "date": "2026-01-23", "categories": []}
{"id": "2507.08711", "title": "SGPMIL: Sparse Gaussian Process Multiple Instance Learning", "abstract": "Multiple Instance Learning (MIL) offers a natural solution for settings where only coarse, bag-level labels are available, without having access to instance-level annotations. This is usually the case in digital pathology, which consists of gigapixel-sized images. While deterministic attention-based MIL approaches achieve strong bag-level performance, they often overlook the uncertainty inherent in instance relevance. In this paper, we address the lack of uncertainty quantification in instance-level attention scores by introducing SGPMIL, a new probabilistic attention-based MIL framework grounded in Sparse Gaussian Processes (SGP). By learning a posterior distribution over attention scores, SGPMIL enables principled uncertainty estimation, resulting in more reliable and calibrated instance relevance maps. Our approach not only preserves competitive bag-level performance but also significantly improves the quality and interpretability of instance-level predictions under uncertainty. SGPMIL extends prior work by introducing feature scaling in the SGP predictive mean function, leading to faster training, improved efficiency, and enhanced instance-level performance. Extensive experiments on multiple well-established digital pathology datasets highlight the effectiveness of our approach across both bag- and instance-level evaluations. Our code is available at https://github.com/mandlos/SGPMIL.", "date": "2025-07-11", "categories": []}
{"id": "2601.12082", "title": "Conditional Random Fields for Interactive Refinement of Histopathological Predictions", "abstract": "Assisting pathologists in the analysis of histopathological images has high clinical value, as it supports cancer detection and staging. In this context, histology foundation models have recently emerged. Among them, Vision-Language Models (VLMs) provide strong yet imperfect zero-shot predictions. We propose to refine these predictions by adapting Conditional Random Fields (CRFs) to histopathological applications, requiring no additional model training. We present HistoCRF, a CRF-based framework, with a novel definition of the pairwise potential that promotes label diversity and leverages expert annotations. We consider three experiments: without annotations, with expert annotations, and with iterative human-in-the-loop annotations that progressively correct misclassified patches. Experiments on five patch-level classification datasets covering different organs and diseases demonstrate average accuracy gains of 16.0% without annotations and 27.5% with only 100 annotations, compared to zero-shot predictions. Moreover, integrating a human in the loop reaches a further gain of 32.6% with the same number of annotations. The code will be made available on https://github.com/tgodelaine/HistoCRF.", "date": "2026-01-17", "categories": []}
{"id": "2601.12493", "title": "Histopath-C: Towards Realistic Domain Shifts for Histopathology Vision-Language Adaptation", "abstract": "Medical Vision-language models (VLMs) have shown remarkable performances in various medical imaging domains such as histo\\-pathology by leveraging pre-trained, contrastive models that exploit visual and textual information. However, histopathology images may exhibit severe domain shifts, such as staining, contamination, blurring, and noise, which may severely degrade the VLM's downstream performance. In this work, we introduce Histopath-C, a new benchmark with realistic synthetic corruptions designed to mimic real-world distribution shifts observed in digital histopathology. Our framework dynamically applies corruptions to any available dataset and evaluates Test-Time Adaptation (TTA) mechanisms on the fly. We then propose LATTE, a transductive, low-rank adaptation strategy that exploits multiple text templates, mitigating the sensitivity of histopathology VLMs to diverse text inputs. Our approach outperforms state-of-the-art TTA methods originally designed for natural images across a breadth of histopathology datasets, demonstrating the effectiveness of our proposed design for robust adaptation in histopathology images. Code and data are available at https://github.com/Mehrdad-Noori/Histopath-C.", "date": "2026-01-18", "categories": []}
{"id": "2601.11336", "title": "Beer-Lambert Autoencoder for Unsupervised Stain Representation Learning and Deconvolution in Multi-immunohistochemical Brightfield Histology Images", "abstract": "Separating the contributions of individual chromogenic stains in RGB histology whole slide images (WSIs) is essential for stain normalization, quantitative assessment of marker expression, and cell-level readouts in immunohistochemistry (IHC). Classical Beer-Lambert (BL) color deconvolution is well-established for two- or three-stain settings, but becomes under-determined and unstable for multiplex IHC (mIHC) with K>3 chromogens. We present a simple, data-driven encoder-decoder architecture that learns cohort-specific stain characteristics for mIHC RGB WSIs and yields crisp, well-separated per-stain concentration maps. The encoder is a compact U-Net that predicts K nonnegative concentration channels; the decoder is a differentiable BL forward model with a learnable stain matrix initialized from typical chromogen hues. Training is unsupervised with a perceptual reconstruction objective augmented by loss terms that discourage unnecessary stain mixing. On a colorectal mIHC panel comprising 5 stains (H, CDX2, MUC2, MUC5, CD8) we show excellent RGB reconstruction, and significantly reduced inter-channel bleed-through compared with matrix-based deconvolution. Code and model are available at https://github.com/measty/StainQuant.git.", "date": "2026-01-16", "categories": []}
{"id": "2601.09130", "title": "Equi-ViT: Rotational Equivariant Vision Transformer for Robust Histopathology Analysis", "abstract": "Vision Transformers (ViTs) have gained rapid adoption in computational pathology for their ability to model long-range dependencies through self-attention, addressing the limitations of convolutional neural networks that excel at local pattern capture but struggle with global contextual reasoning. Recent pathology-specific foundation models have further advanced performance by leveraging large-scale pretraining. However, standard ViTs remain inherently non-equivariant to transformations such as rotations and reflections, which are ubiquitous variations in histopathology imaging. To address this limitation, we propose Equi-ViT, which integrates an equivariant convolution kernel into the patch embedding stage of a ViT architecture, imparting built-in rotational equivariance to learned representations. Equi-ViT achieves superior rotation-consistent patch embeddings and stable classification performance across image orientations. Our results on a public colorectal cancer dataset demonstrate that incorporating equivariant patch embedding enhances data efficiency and robustness, suggesting that equivariant transformers could potentially serve as more generalizable backbones for the application of ViT in histopathology, such as digital pathology foundation models.", "date": "2026-01-14", "categories": []}
{"id": "2601.08336", "title": "Tissue Classification and Whole-Slide Images Analysis via Modeling of the Tumor Microenvironment and Biological Pathways", "abstract": "Automatic integration of whole slide images (WSIs) and gene expression profiles has demonstrated substantial potential in precision clinical diagnosis and cancer progression studies. However, most existing studies focus on individual gene sequences and slide level classification tasks, with limited attention to spatial transcriptomics and patch level applications. To address this limitation, we propose a multimodal network, BioMorphNet, which automatically integrates tissue morphological features and spatial gene expression to support tissue classification and differential gene analysis. For considering morphological features, BioMorphNet constructs a graph to model the relationships between target patches and their neighbors, and adjusts the response strength based on morphological and molecular level similarity, to better characterize the tumor microenvironment. In terms of multimodal interactions, BioMorphNet derives clinical pathway features from spatial transcriptomic data based on a predefined pathway database, serving as a bridge between tissue morphology and gene expression. In addition, a novel learnable pathway module is designed to automatically simulate the biological pathway formation process, providing a complementary representation to existing clinical pathways. Compared with the latest morphology gene multimodal methods, BioMorphNet's average classification metrics improve by 2.67%, 5.48%, and 6.29% for prostate cancer, colorectal cancer, and breast cancer datasets, respectively. BioMorphNet not only classifies tissue categories within WSIs accurately to support tumor localization, but also analyzes differential gene expression between tissue categories based on prediction confidence, contributing to the discovery of potential tumor biomarkers.", "date": "2026-01-13", "categories": []}
{"id": "2509.16935", "title": "Parameter-efficient fine-tuning (PEFT) of Vision Foundation Models for Atypical Mitotic Figure Classification", "abstract": "Atypical mitotic figures (AMFs) are rare abnormal cell divisions associated with tumor aggressiveness and poor prognosis. Their detection remains a significant challenge due to subtle morphological cues, class imbalance, and inter-observer variability among pathologists. The MIDOG 2025 challenge introduced a dedicated track for atypical mitosis classification, enabling systematic evaluation of deep learning methods. In this study, we investigated the use of large vision foundation models, including Virchow, Virchow2, and UNI, with Low-Rank Adaptation (LoRA) for parameter-efficient fine-tuning. We conducted extensive experiments with different LoRA ranks, as well as random and group-based data splits, to analyze robustness under varied conditions. Our best approach, Virchow with LoRA rank 8 and ensemble of three-fold cross-validation, achieved a balanced accuracy of 88.37% on the preliminary test set, ranking joint 9th in the challenge leaderboard. These results highlight the promise of foundation models with efficient adaptation strategies for the classification of atypical mitosis, while underscoring the need for improvements in specificity and domain generalization.", "date": "2025-09-21", "categories": []}
{"id": "2601.06360", "title": "Computational Mapping of Reactive Stroma in Prostate Cancer Yields Interpretable, Prognostic Biomarkers", "abstract": "Current histopathological grading of prostate cancer relies primarily on glandular architecture, largely overlooking the tumor microenvironment. Here, we present PROTAS, a deep learning framework that quantifies reactive stroma (RS) in routine hematoxylin and eosin (H&E) slides and links stromal morphology to underlying biology. PROTAS-defined RS is characterized by nuclear enlargement, collagen disorganization, and transcriptomic enrichment of contractile pathways. PROTAS detects RS robustly in the external Prostate, Lung, Colorectal, and Ovarian (PLCO) dataset and, using domain-adversarial training, generalizes to diagnostic biopsies. In head-to-head comparisons, PROTAS outperforms pathologists for RS detection, and spatial RS features predict biochemical recurrence independently of established prognostic variables (c-index 0.80). By capturing subtle stromal phenotypes associated with tumor progression, PROTAS provides an interpretable, scalable biomarker to refine risk stratification.", "date": "2026-01-10", "categories": []}
{"id": "2508.16209", "title": "Deep learning-enabled virtual multiplexed immunostaining of label-free tissue for vascular invasion assessment", "abstract": "Immunohistochemistry (IHC) has transformed clinical pathology by enabling the visualization of specific proteins within tissue sections. However, traditional IHC requires one tissue section per stain, exhibits section-to-section variability, and incurs high costs and laborious staining procedures. While multiplexed IHC (mIHC) techniques enable simultaneous staining with multiple antibodies on a single slide, they are more tedious to perform and are currently unavailable in routine pathology laboratories. Here, we present a deep learning-based virtual multiplexed immunostaining framework to simultaneously generate ERG and PanCK, in addition to H&E virtual staining, enabling accurate localization and interpretation of vascular invasion in thyroid cancers. This virtual mIHC technique is based on the autofluorescence microscopy images of label-free tissue sections, and its output images closely match the histochemical staining counterparts (ERG, PanCK and H&E) of the same tissue sections. Blind evaluation by board-certified pathologists demonstrated that virtual mIHC staining achieved high concordance with the histochemical staining results, accurately highlighting epithelial cells and endothelial cells. Virtual mIHC conducted on the same tissue section also allowed the identification and localization of small vessel invasion. This multiplexed virtual IHC approach can significantly improve diagnostic accuracy and efficiency in the histopathological evaluation of vascular invasion, potentially eliminating the need for traditional staining protocols and mitigating issues related to tissue loss and heterogeneity.", "date": "2025-08-22", "categories": []}
{"id": "2601.05148", "title": "Atlas 2 -- Foundation models for clinical deployment", "abstract": "Pathology foundation models substantially advanced the possibilities in computational pathology -- yet tradeoffs in terms of performance, robustness, and computational requirements remained, which limited their clinical deployment. In this report, we present Atlas 2, Atlas 2-B, and Atlas 2-S, three pathology vision foundation models which bridge these shortcomings by showing state-of-the-art performance in prediction performance, robustness, and resource efficiency in a comprehensive evaluation across eighty public benchmarks. Our models were trained on the largest pathology foundation model dataset to date comprising 5.5 million histopathology whole slide images, collected from three medical institutions Charit\\'e - Universt\\\"atsmedizin Berlin, LMU Munich, and Mayo Clinic.", "date": "2026-01-08", "categories": []}
{"id": "2601.04163", "title": "Scanner-Induced Domain Shifts Undermine the Robustness of Pathology Foundation Models", "abstract": "Pathology foundation models (PFMs) have become central to computational pathology, aiming to offer general encoders for feature extraction from whole-slide images (WSIs). Despite strong benchmark performance, PFM robustness to real-world technical domain shifts, such as variability from whole-slide scanner devices, remains poorly understood. We systematically evaluated the robustness of 14 PFMs to scanner-induced variability, including state-of-the-art models, earlier self-supervised models, and a baseline trained on natural images. Using a multiscanner dataset of 384 breast cancer WSIs scanned on five devices, we isolated scanner effects independently from biological and laboratory confounders. Robustness is assessed via complementary unsupervised embedding analyses and a set of clinicopathological supervised prediction tasks. Our results demonstrate that current PFMs are not invariant to scanner-induced domain shifts. Most models encode pronounced scanner-specific variability in their embedding spaces. While AUC often remains stable, this masks a critical failure mode: scanner variability systematically alters the embedding space and impacts calibration of downstream model predictions, resulting in scanner-dependent bias that can impact reliability in clinical use cases. We further show that robustness is not a simple function of training data scale, model size, or model recency. None of the models provided reliable robustness against scanner-induced variability. While the models trained on the most diverse data, here represented by vision-language models, appear to have an advantage with respect to robustness, they underperformed on downstream supervised tasks. We conclude that development and evaluation of PFMs requires moving beyond accuracy-centric benchmarks toward explicit evaluation and optimisation of embedding stability and calibration under realistic acquisition variability.", "date": "2026-01-07", "categories": []}
{"id": "2508.14681", "title": "Virtual Multiplex Staining for Histological Images using a Marker-wise Conditioned Diffusion Model", "abstract": "Multiplex imaging is revolutionizing pathology by enabling the simultaneous visualization of multiple biomarkers within tissue samples, providing molecular-level insights that traditional hematoxylin and eosin (H&E) staining cannot provide. However, the complexity and cost of multiplex data acquisition have hindered its widespread adoption. Additionally, most existing large repositories of H&E images lack corresponding multiplex images, limiting opportunities for multimodal analysis. To address these challenges, we leverage recent advances in latent diffusion models (LDMs), which excel at modeling complex data distributions by utilizing their powerful priors for fine-tuning to a target domain. In this paper, we introduce a novel framework for virtual multiplex staining that utilizes pretrained LDM parameters to generate multiplex images from H&E images using a conditional diffusion model. Our approach enables marker-by-marker generation by conditioning the diffusion model on each marker, while sharing the same architecture across all markers. To tackle the challenge of varying pixel value distributions across different marker stains and to improve inference speed, we fine-tune the model for single-step sampling, enhancing both color contrast fidelity and inference efficiency through pixel-level loss functions. We validate our framework on two publicly available datasets, notably demonstrating its effectiveness in generating up to 18 different marker types with improved accuracy, a substantial increase over the 2-3 marker types achieved in previous approaches. This validation highlights the potential of our framework, pioneering virtual multiplex staining. Finally, this paper bridges the gap between H&E and multiplex imaging, potentially enabling retrospective studies and large-scale analyses of existing H&E image repositories.", "date": "2025-08-20", "categories": []}
{"id": "2508.16742", "title": "Learning the Language of Histopathology Images reveals Prognostic Subgroups in Invasive Lung Adenocarcinoma Patients", "abstract": "Recurrence remains a major clinical challenge in surgically resected invasive lung adenocarcinoma, where existing grading and staging systems fail to capture the cellular complexity that underlies tumor aggressiveness. We present PathRosetta, a novel AI model that conceptualizes histopathology as a language, where cells serve as words, spatial neighborhoods form syntactic structures, and tissue architecture composes sentences. By learning this language of histopathology, PathRosetta predicts five-year recurrence directly from hematoxylin-and-eosin (H&E) slides, treating them as documents representing the state of the disease. In a multi-cohort dataset of 289 patients (600 slides), PathRosetta achieved an area under the curve (AUC) of 0.78 +- 0.04 on the internal cohort, significantly outperforming IASLC grading (AUC:0.71), AJCC staging (AUC:0.64), and other state-of-the-art AI models (AUC:0.62-0.67). It yielded a hazard ratio of 9.54 and a concordance index of 0.70, generalized robustly to external TCGA (AUC:0.75) and CPTAC (AUC:0.76) cohorts, and performed consistently across demographic and clinical subgroups. Beyond whole-slide prediction, PathRosetta uncovered prognostic subgroups within individual cell types, revealing that even within benign epithelial, stromal, or other cells, distinct morpho-spatial phenotypes correspond to divergent outcomes. Moreover, because the model explicitly understands what it is looking at, including cell types, cellular neighborhoods, and higher-order tissue morphology, it is inherently interpretable and can articulate the rationale behind its predictions. These findings establish that representing histopathology as a language enables interpretable and generalizable prognostication from routine histology.", "date": "2025-08-22", "categories": []}
{"id": "2601.01769", "title": "CTIS-QA: Clinical Template-Informed Slide-level Question Answering for Pathology", "abstract": "In this paper, we introduce a clinical diagnosis template-based pipeline to systematically collect and structure pathological information. In collaboration with pathologists and guided by the the College of American Pathologists (CAP) Cancer Protocols, we design a Clinical Pathology Report Template (CPRT) that ensures comprehensive and standardized extraction of diagnostic elements from pathology reports. We validate the effectiveness of our pipeline on TCGA-BRCA. First, we extract pathological features from reports using CPRT. These features are then used to build CTIS-Align, a dataset of 80k slide-description pairs from 804 WSIs for vision-language alignment training, and CTIS-Bench, a rigorously curated VQA benchmark comprising 977 WSIs and 14,879 question-answer pairs. CTIS-Bench emphasizes clinically grounded, closed-ended questions (e.g., tumor grade, receptor status) that reflect real diagnostic workflows, minimize non-visual reasoning, and require genuine slide understanding. We further propose CTIS-QA, a Slide-level Question Answering model, featuring a dual-stream architecture that mimics pathologists' diagnostic approach. One stream captures global slide-level context via clustering-based feature aggregation, while the other focuses on salient local regions through attention-guided patch perception module. Extensive experiments on WSI-VQA, CTIS-Bench, and slide-level diagnostic tasks show that CTIS-QA consistently outperforms existing state-of-the-art models across multiple metrics. Code and data are available at https://github.com/HLSvois/CTIS-QA.", "date": "2026-01-05", "categories": []}
{"id": "2601.01875", "title": "Toward Auditable Neuro-Symbolic Reasoning in Pathology: SQL as an Explicit Trace of Evidence", "abstract": "Automated pathology image analysis is central to clinical diagnosis, but clinicians still ask which slide features drive a model's decision and why. Vision-language models can produce natural language explanations, but these are often correlational and lack verifiable evidence. In this paper, we introduce an SQL-centered agentic framework that enables both feature measurement and reasoning to be auditable. Specifically, after extracting human-interpretable cellular features, Feature Reasoning Agents compose and execute SQL queries over feature tables to aggregate visual evidence into quantitative findings. A Knowledge Comparison Agent then evaluates these findings against established pathological knowledge, mirroring how pathologists justify diagnoses from measurable observations. Extensive experiments evaluated on two pathology visual question answering datasets demonstrate our method improves interpretability and decision traceability while producing executable SQL traces that link cellular measurements to diagnostic conclusions.", "date": "2026-01-05", "categories": []}
{"id": "2601.02198", "title": "Mind the Gap: Continuous Magnification Sampling for Pathology Foundation Models", "abstract": "In histopathology, pathologists examine both tissue architecture at low magnification and fine-grained morphology at high magnification. Yet, the performance of pathology foundation models across magnifications and the effect of magnification sampling during training remain poorly understood. We model magnification sampling as a multi-source domain adaptation problem and develop a simple theoretical framework that reveals systematic trade-offs between sampling strategies. We show that the widely used discrete uniform sampling of magnifications (0.25, 0.5, 1.0, 2.0 mpp) leads to degradation at intermediate magnifications. We introduce continuous magnification sampling, which removes gaps in magnification coverage while preserving performance at standard scales. Further, we derive sampling distributions that optimize representation quality across magnification scales. To evaluate these strategies, we introduce two new benchmarks (TCGA-MS, BRACS-MS) with appropriate metrics. Our experiments show that continuous sampling substantially improves over discrete sampling at intermediate magnifications, with gains of up to 4 percentage points in balanced classification accuracy, and that optimized distributions can further improve performance. Finally, we evaluate current histopathology foundation models, finding that magnification is a primary driver of performance variation across models. Our work paves the way towards future pathology foundation models that perform reliably across magnifications.", "date": "2026-01-05", "categories": []}
{"id": "2601.00716", "title": "Detecting Performance Degradation under Data Shift in Pathology Vision-Language Model", "abstract": "Vision-Language Models have demonstrated strong potential in medical image analysis and disease diagnosis. However, after deployment, their performance may deteriorate when the input data distribution shifts from that observed during development. Detecting such performance degradation is essential for clinical reliability, yet remains challenging for large pre-trained VLMs operating without labeled data. In this study, we investigate performance degradation detection under data shift in a state-of-the-art pathology VLM. We examine both input-level data shift and output-level prediction behavior to understand their respective roles in monitoring model reliability. To facilitate systematic analysis of input data shift, we develop DomainSAT, a lightweight toolbox with a graphical interface that integrates representative shift detection algorithms and enables intuitive exploration of data shift. Our analysis shows that while input data shift detection is effective at identifying distributional changes and providing early diagnostic signals, it does not always correspond to actual performance degradation. Motivated by this observation, we further study output-based monitoring and introduce a label-free, confidence-based degradation indicator that directly captures changes in model prediction confidence. We find that this indicator exhibits a close relationship with performance degradation and serves as an effective complement to input shift detection. Experiments on a large-scale pathology dataset for tumor classification demonstrate that combining input data shift detection and output confidence-based indicators enables more reliable detection and interpretation of performance degradation in VLMs under data shift. These findings provide a practical and complementary framework for monitoring the reliability of foundation models in digital pathology.", "date": "2026-01-02", "categories": []}
{"id": "2502.07221", "title": "HOMIE: Histopathology Omni-modal Embedding for Pathology Composed Retrieval", "abstract": "The integration of Artificial Intelligence (AI) into pathology faces a fundamental challenge: black-box predictive models lack transparency, while generative approaches risk clinical hallucination. A case-based retrieval paradigm offers a more interpretable alternative for clinical adoption. However, current SOTA models are constrained by dual-encoder architectures that cannot process the composed modality of real-world clinical queries. We formally define the task of Pathology Composed Retrieval (PCR). However, progress in this newly defined task is blocked by two critical challenges: (1) Multimodal Large Language Models (MLLMs) offer the necessary deep-fusion architecture but suffer from a critical Task Mismatch and Domain Mismatch. (2) No benchmark exists to evaluate such compositional queries. To solve these challenges, we propose HOMIE, a systematic framework that transforms a general MLLM into a specialized retrieval expert. HOMIE resolves the dual mismatch via a two-stage process: a retrieval-adaptation stage to solve the task mismatch, and a pathology-specific tuning stage, featuring a progressive knowledge curriculum, pathology specfic stain and native resolution processing, to solve the domain mismatch. We also introduce the PCR Benchmark, a benchmark designed to evaluate composed retrieval in pathology. Experiments show that HOMIE, trained only on public data, matches SOTA performance on traditional retrieval tasks and outperforms all baselines on the newly defined PCR task.", "date": "2025-02-11", "categories": []}
{"id": "2512.22188", "title": "HookMIL: Revisiting Context Modeling in Multiple Instance Learning for Computational Pathology", "abstract": "Multiple Instance Learning (MIL) has enabled weakly supervised analysis of whole-slide images (WSIs) in computational pathology. However, traditional MIL approaches often lose crucial contextual information, while transformer-based variants, though more expressive, suffer from quadratic complexity and redundant computations. To address these limitations, we propose HookMIL, a context-aware and computationally efficient MIL framework that leverages compact, learnable hook tokens for structured contextual aggregation. These tokens can be initialized from (i) key-patch visual features, (ii) text embeddings from vision-language pathology models, and (iii) spatially grounded features from spatial transcriptomics-vision models. This multimodal initialization enables Hook Tokens to incorporate rich textual and spatial priors, accelerating convergence and enhancing representation quality. During training, Hook tokens interact with instances through bidirectional attention with linear complexity. To further promote specialization, we introduce a Hook Diversity Loss that encourages each token to focus on distinct histopathological patterns. Additionally, a hook-to-hook communication mechanism refines contextual interactions while minimizing redundancy. Extensive experiments on four public pathology datasets demonstrate that HookMIL achieves state-of-the-art performance, with improved computational efficiency and interpretability. Codes are available at https://github.com/lingxitong/HookMIL.", "date": "2025-12-20", "categories": []}
{"id": "2512.22262", "title": "INSIGHT: Spatially resolved survival modelling from routine histology crosslinked with molecular profiling reveals prognostic epithelial-immune axes in stage II/III colorectal cancer", "abstract": "Routine histology contains rich prognostic information in stage II/III colorectal cancer, much of which is embedded in complex spatial tissue organisation. We present INSIGHT, a graph neural network that predicts survival directly from routine histology images. Trained and cross-validated on TCGA (n=342) and SURGEN (n=336), INSIGHT produces patient-level spatially resolved risk scores. Large independent validation showed superior prognostic performance compared with pTNM staging (C-index 0.68-0.69 vs 0.44-0.58). INSIGHT spatial risk maps recapitulated canonical prognostic histopathology and identified nuclear solidity and circularity as quantitative risk correlates. Integrating spatial risk with data-driven spatial transcriptomic signatures, spatial proteomics, bulk RNA-seq, and single-cell references revealed an epithelium-immune risk manifold capturing epithelial dedifferentiation and fetal programs, myeloid-driven stromal states including $\\mathrm{SPP1}^{+}$ macrophages and $\\mathrm{LAMP3}^{+}$ dendritic cells, and adaptive immune dysfunction. This analysis exposed patient-specific epithelial heterogeneity, stratification within MSI-High tumours, and high-risk routes of CDX2/HNF4A loss and CEACAM5/6-associated proliferative programs, highlighting coordinated therapeutic vulnerabilities.", "date": "2025-12-24", "categories": []}
{"id": "2512.22335", "title": "Feature Learning with Multi-Stage Vision Transformers on Inter-Modality HER2 Status Scoring and Tumor Classification on Whole Slides", "abstract": "The popular use of histopathology images, such as hematoxylin and eosin (H&E), has proven to be useful in detecting tumors. However, moving such cancer cases forward for treatment requires accurate on the amount of the human epidermal growth factor receptor 2 (HER2) protein expression. Predicting both the lower and higher levels of HER2 can be challenging. Moreover, jointly analyzing H&E and immunohistochemistry (IHC) stained images for HER2 scoring is difficult. Although several deep learning methods have been investigated to address the challenge of HER2 scoring, they suffer from providing a pixel-level localization of HER2 status. In this study, we propose a single end-to-end pipeline using a system of vision transformers with HER2 status scoring on whole slide images of WSIs. The method includes patch-wise processing of H&E WSIs for tumor localization. A novel mapping function is proposed to correspondingly identify correlated IHC WSIs regions with malignant regions on H&E. A clinically inspired HER2 scoring mechanism is embedded in the pipeline and allows for automatic pixel-level annotation of 4-way HER2 scoring (0, 1+, 2+, and 3+). Also, the proposed method accurately returns HER2-negative and HER2-positive. Privately curated datasets were collaboratively extracted from 13 different cases of WSIs of H&E and IHC. A thorough experiment is conducted on the proposed method. Results obtained showed a good classification accuracy during tumor localization. Also, a classification accuracy of 0.94 and a specificity of 0.933 were returned for the prediction of HER2 status, scoring in the 4-way methods. The applicability of the proposed pipeline was investigated using WSIs patches as comparable to human pathologists. Findings from the study showed the usability of jointly evaluated H&E and IHC images on end-to-end ViTs-based models for HER2 scoring", "date": "2025-12-26", "categories": []}
{"id": "2512.23545", "title": "PathFound: An Agentic Multimodal Model Activating Evidence-seeking Pathological Diagnosis", "abstract": "Recent pathological foundation models have substantially advanced visual representation learning and multimodal interaction. However, most models still rely on a static inference paradigm in which whole-slide images are processed once to produce predictions, without reassessment or targeted evidence acquisition under ambiguous diagnoses. This contrasts with clinical diagnostic workflows that refine hypotheses through repeated slide observations and further examination requests. We propose PathFound, an agentic multimodal model designed to support evidence-seeking inference in pathological diagnosis. PathFound integrates the power of pathological visual foundation models, vision-language models, and reasoning models trained with reinforcement learning to perform proactive information acquisition and diagnosis refinement by progressing through the initial diagnosis, evidence-seeking, and final decision stages. Across several large multimodal models, adopting this strategy consistently improves diagnostic accuracy, indicating the effectiveness of evidence-seeking workflows in computational pathology. Among these models, PathFound achieves state-of-the-art diagnostic performance across diverse clinical scenarios and demonstrates strong potential to discover subtle details, such as nuclear features and local invasions.", "date": "2025-12-29", "categories": []}
{"id": "2408.05195", "title": "HistoKernel: Whole Slide Image Level Maximum Mean Discrepancy Kernels for Pan-Cancer Predictive Modelling", "abstract": "Machine learning in computational pathology (CPath) often aggregates patch-level predictions from multi-gigapixel Whole Slide Images (WSIs) to generate WSI-level prediction scores for crucial tasks such as survival prediction and drug effect prediction. However, current methods do not explicitly characterize distributional differences between patch sets within WSIs. We introduce HistoKernel, a novel Maximum Mean Discrepancy (MMD) kernel that measures distributional similarity between WSIs for enhanced prediction performance on downstream prediction tasks.\n  Our comprehensive analysis demonstrates HistoKernel's effectiveness across various machine learning tasks, including retrieval (n = 9,362), drug sensitivity regression (n = 551), point mutation classification (n = 3,419), and survival analysis (n = 2,291), outperforming existing deep learning methods. Additionally, HistoKernel seamlessly integrates multi-modal data and offers a novel perturbation-based method for patch-level explainability. This work pioneers the use of kernel-based methods for WSI-level predictive modeling, opening new avenues for research. Code is available at https://github.com/pkeller00/HistoKernel.", "date": "2024-08-09", "categories": []}
{"id": "2512.21331", "title": "TICON: A Slide-Level Tile Contextualizer for Histopathology Representation Learning", "abstract": "The interpretation of small tiles in large whole slide images (WSI) often needs a larger image context. We introduce TICON, a transformer-based tile representation contextualizer that produces rich, contextualized embeddings for ''any'' application in computational pathology. Standard tile encoder-based pipelines, which extract embeddings of tiles stripped from their context, fail to model the rich slide-level information essential for both local and global tasks. Furthermore, different tile-encoders excel at different downstream tasks. Therefore, a unified model is needed to contextualize embeddings derived from ''any'' tile-level foundation model. TICON addresses this need with a single, shared encoder, pretrained using a masked modeling objective to simultaneously unify and contextualize representations from diverse tile-level pathology foundation models. Our experiments demonstrate that TICON-contextualized embeddings significantly improve performance across many different tasks, establishing new state-of-the-art results on tile-level benchmarks (i.e., HEST-Bench, THUNDER, CATCH) and slide-level benchmarks (i.e., Patho-Bench). Finally, we pretrain an aggregator on TICON to form a slide-level foundation model, using only 11K WSIs, outperforming SoTA slide-level foundation models pretrained with up to 350K WSIs.", "date": "2025-12-24", "categories": []}
{"id": "2512.21792", "title": "AI for Mycetoma Diagnosis in Histopathological Images: The MICCAI 2024 Challenge", "abstract": "Mycetoma is a neglected tropical disease caused by fungi or bacteria leading to severe tissue damage and disabilities. It affects poor and rural communities and presents medical challenges and socioeconomic burdens on patients and healthcare systems in endemic regions worldwide. Mycetoma diagnosis is a major challenge in mycetoma management, particularly in low-resource settings where expert pathologists are limited. To address this challenge, this paper presents an overview of the Mycetoma MicroImage: Detect and Classify Challenge (mAIcetoma) which was organized to advance mycetoma diagnosis through AI solutions. mAIcetoma focused on developing automated models for segmenting mycetoma grains and classifying mycetoma types from histopathological images. The challenge attracted the attention of several teams worldwide to participate and five finalist teams fulfilled the challenge objectives. The teams proposed various deep learning architectures for the ultimate goal of this challenge. Mycetoma database (MyData) was provided to participants as a standardized dataset to run the proposed models. Those models were evaluated using evaluation metrics. Results showed that all the models achieved high segmentation accuracy, emphasizing the necessitate of grain detection as a critical step in mycetoma diagnosis. In addition, the top-performing models show a significant performance in classifying mycetoma types.", "date": "2025-12-25", "categories": []}
{"id": "2511.06433", "title": "Diagnose Like A REAL Pathologist: An Uncertainty-Focused Approach for Trustworthy Multi-Resolution Multiple Instance Learning", "abstract": "With the increasing demand for histopathological specimen examination and diagnostic reporting, Multiple Instance Learning (MIL) has received heightened research focus as a viable solution for AI-centric diagnostic aid. Recently, to improve its performance and make it work more like a pathologist, several MIL approaches based on the use of multiple-resolution images have been proposed, delivering often higher performance than those that use single-resolution images. Despite impressive recent developments of multiple-resolution MIL, previous approaches only focus on improving performance, thereby lacking research on well-calibrated MIL that clinical experts can rely on for trustworthy diagnostic results. In this study, we propose Uncertainty-Focused Calibrated MIL (UFC-MIL), which more closely mimics the pathologists' examination behaviors while providing calibrated diagnostic predictions, using multiple images with different resolutions. UFC-MIL includes a novel patch-wise loss that learns the latent patterns of instances and expresses their uncertainty for classification. Also, the attention-based architecture with a neighbor patch aggregation module collects features for the classifier. In addition, aggregated predictions are calibrated through patch-level uncertainty without requiring multiple iterative inferences, which is a key practical advantage. Against challenging public datasets, UFC-MIL shows superior performance in model calibration while achieving classification accuracy comparable to that of state-of-the-art methods.", "date": "2025-11-09", "categories": []}
{"id": "2512.19954", "title": "HistoWAS: A Pathomics Framework for Large-Scale Feature-Wide Association Studies of Tissue Topology and Patient Outcomes", "abstract": "High-throughput \"pathomic\" analysis of Whole Slide Images (WSIs) offers new opportunities to study tissue characteristics and for biomarker discovery. However, the clinical relevance of the tissue characteristics at the micro- and macro-environment level is limited by the lack of tools that facilitate the measurement of the spatial interaction of individual structure characteristics and their association with clinical parameters. To address these challenges, we introduce HistoWAS (Histology-Wide Association Study), a computational framework designed to link tissue spatial organization to clinical outcomes. Specifically, HistoWAS implements (1) a feature space that augments conventional metrics with 30 topological and spatial features, adapted from Geographic Information Systems (GIS) point pattern analysis, to quantify tissue micro-architecture; and (2) an association study engine, inspired by Phenome-Wide Association Studies (PheWAS), that performs mass univariate regression for each feature with statistical correction. As a proof of concept, we applied HistoWAS to analyze a total of 102 features (72 conventional object-level features and our 30 spatial features) using 385 PAS-stained WSIs from 206 participants in the Kidney Precision Medicine Project (KPMP). The code and data have been released to https://github.com/hrlblab/histoWAS.", "date": "2025-12-23", "categories": []}
{"id": "2512.19982", "title": "WSD-MIL: Window Scale Decay Multiple Instance Learning for Whole Slide Image Classification", "abstract": "In recent years, the integration of pre-trained foundational models with multiple instance learning (MIL) has improved diagnostic accuracy in computational pathology. However, existing MIL methods focus on optimizing feature extractors and aggregation strategies while overlooking the complex semantic relationships among instances within whole slide image (WSI). Although Transformer-based MIL approaches aiming to model instance dependencies, the quadratic computational complexity limits their scalability to large-scale WSIs. Moreover, due to the pronounced variations in tumor region scales across different WSIs, existing Transformer-based methods employing fixed-scale attention mechanisms face significant challenges in precisely capturing local instance correlations and fail to account for the distance-based decay effect of patch relevance. To address these challenges, we propose window scale decay MIL (WSD-MIL), designed to enhance the capacity to model tumor regions of varying scales while improving computational efficiency. WSD-MIL comprises: 1) a window scale decay based attention module, which employs a cluster-based sampling strategy to reduce computational costs while progressively decaying attention window-scale to capture local instance relationships at varying scales; and 2) a squeeze-and-excitation based region gate module, which dynamically adjusts window weights to enhance global information modeling. Experimental results demonstrate that WSD-MIL achieves state-of-the-art performance on the CAMELYON16 and TCGA-BRCA datasets while reducing 62% of the computational memory. The code will be publicly available.", "date": "2025-12-23", "categories": []}
{"id": "2511.16268", "title": "Weakly Supervised Segmentation and Classification of Alpha-Synuclein Aggregates in Brightfield Midbrain Images", "abstract": "Parkinson's disease (PD) is a neurodegenerative disorder associated with the accumulation of misfolded alpha-synuclein aggregates, forming Lewy bodies and neuritic shape used for pathology diagnostics. Automatic analysis of immunohistochemistry histopathological images with Deep Learning provides a promising tool for better understanding the spatial organization of these aggregates. In this study, we develop an automated image processing pipeline to segment and classify these aggregates in whole-slide images (WSIs) of midbrain tissue from PD and incidental Lewy Body Disease (iLBD) cases based on weakly supervised segmentation, robust to immunohistochemical labelling variability, with a ResNet50 classifier. Our approach allows to differentiate between major aggregate morphologies, including Lewy bodies and neurites with a balanced accuracy of $80\\%$. This framework paves the way for large-scale characterization of the spatial distribution and heterogeneity of alpha-synuclein aggregates in brightfield immunohistochemical tissue, and for investigating their poorly understood relationships with surrounding cells such as microglia and astrocytes.", "date": "2025-11-20", "categories": []}
{"id": "2512.17930", "title": "CytoDINO: Risk-Aware and Biologically-Informed Adaptation of DINOv3 for Bone Marrow Cytomorphology", "abstract": "Bone marrow cell cytomorphology analysis is critical for the diagnosis of hematological malignancies but remains a labor-intensive process subject to significant inter-observer variability. While recent foundation models have shown promise in computational pathology, they often require extensive computational resources and fail to account for the asymmetric risks associated with clinical misdiagnosis. We introduce CytoDINO, a framework that achieves state-of-the-art performance on the Munich Leukemia Laboratory (MLL) dataset by fine-tuning DINOv3 using Low-Rank Adaptation (LoRA). Our primary contribution is a novel Hierarchical Focal Loss with Critical Penalties, which encodes biological relationships between cell lineages and explicitly penalizes clinically dangerous misclassifications (e.g., classifying blasts as normal cells). CytoDINO achieves an 88.2% weighted F1 score and 76.5% macro F1 on a held-out test set of 21 cell classes. By utilizing parameter-efficient fine-tuning with only 8% trainable parameters on a single NVIDIA RTX 5080, we demonstrate that consumer-grade hardware can match specialized infrastructure. Furthermore, confidence-based selective prediction yields 99.5% accuracy on 67% of samples, suggesting a viable pathway for clinical deployment where high-uncertainty cases are flagged for expert review", "date": "2025-12-09", "categories": []}
{"id": "2512.18734", "title": "Breast Cancer Recurrence Risk Prediction Based on Multiple Instance Learning", "abstract": "Predicting breast cancer recurrence risk is a critical clinical challenge. This study investigates the potential of computational pathology to stratify patients using deep learning on routine Hematoxylin and Eosin (H&E) stained whole-slide images (WSIs). We developed and compared three Multiple Instance Learning (MIL) frameworks -- CLAM-SB, ABMIL, and ConvNeXt-MIL-XGBoost -- on an in-house dataset of 210 patient cases. The models were trained to predict 5-year recurrence risk, categorized into three tiers (low, medium, high), with ground truth labels established by the 21-gene Recurrence Score. Features were extracted using the UNI and CONCH pre-trained models. In a 5-fold cross-validation, the modified CLAM-SB model demonstrated the strongest performance, achieving a mean Area Under the Curve (AUC) of 0.836 and a classification accuracy of 76.2%. Our findings demonstrate the feasibility of using deep learning on standard histology slides for automated, genomics-correlated risk stratification, highlighting a promising pathway toward rapid and cost-effective clinical decision support.", "date": "2025-12-21", "categories": []}
{"id": "2512.19331", "title": "DeltaMIL: Gated Memory Integration for Efficient and Discriminative Whole Slide Image Analysis", "abstract": "Whole Slide Images (WSIs) are typically analyzed using multiple instance learning (MIL) methods. However, the scale and heterogeneity of WSIs generate highly redundant and dispersed information, making it difficult to identify and integrate discriminative signals. Existing MIL methods either fail to discard uninformative cues effectively or have limited ability to consolidate relevant features from multiple patches, which restricts their performance on large and heterogeneous WSIs. To address this issue, we propose DeltaMIL, a novel MIL framework that explicitly selects semantically relevant regions and integrates the discriminative information from WSIs. Our method leverages the gated delta rule to efficiently filter and integrate information through a block combining forgetting and memory mechanisms. The delta mechanism dynamically updates the memory by removing old values and inserting new ones according to their correlation with the current patch. The gating mechanism further enables rapid forgetting of irrelevant signals. Additionally, DeltaMIL integrates a complementary local pattern mixing mechanism to retain fine-grained pathological locality. Our design enhances the extraction of meaningful cues and suppresses redundant or noisy information, which improves the model's robustness and discriminative power. Experiments demonstrate that DeltaMIL achieves state-of-the-art performance. Specifically, for survival prediction, DeltaMIL improves performance by 3.69\\% using ResNet-50 features and 2.36\\% using UNI features. For slide-level classification, it increases accuracy by 3.09\\% with ResNet-50 features and 3.75\\% with UNI features. These results demonstrate the strong and consistent performance of DeltaMIL across diverse WSI tasks.", "date": "2025-12-22", "categories": []}
{"id": "2512.17499", "title": "Validation of Diagnostic Artificial Intelligence Models for Prostate Pathology in a Middle Eastern Cohort", "abstract": "Background: Artificial intelligence (AI) is improving the efficiency and accuracy of cancer diagnostics. The performance of pathology AI systems has been almost exclusively evaluated on European and US cohorts from large centers. For global AI adoption in pathology, validation studies on currently under-represented populations - where the potential gains from AI support may also be greatest - are needed. We present the first study with an external validation cohort from the Middle East, focusing on AI-based diagnosis and Gleason grading of prostate cancer.\n  Methods: We collected and digitised 339 prostate biopsy specimens from the Kurdistan region, Iraq, representing a consecutive series of 185 patients spanning the period 2013-2024. We evaluated a task-specific end-to-end AI model and two foundation models in terms of their concordance with pathologists and consistency across samples digitised on three scanner models (Hamamatsu, Leica, and Grundium).\n  Findings: Grading concordance between AI and pathologists was similar to pathologist-pathologist concordance with Cohen's quadratically weighted kappa 0.801 vs. 0.799 (p=0.9824). Cross-scanner concordance was high (quadratically weighted kappa > 0.90) for all AI models and scanner pairs, including low-cost compact scanner.\n  Interpretation: AI models demonstrated pathologist-level performance in prostate histopathology assessment. Compact scanners can provide a route for validation studies in non-digitalised settings and enable cost-effective adoption of AI in laboratories with limited sample volumes. This first openly available digital pathology dataset from the Middle East supports further research into globally equitable AI pathology.\n  Funding: SciLifeLab and Wallenberg Data Driven Life Science Program, Instrumentarium Science Foundation, Karolinska Institutet Research Foundation.", "date": "2025-12-19", "categories": []}
{"id": "2512.17517", "title": "PathBench-MIL: A Comprehensive AutoML and Benchmarking Framework for Multiple Instance Learning in Histopathology", "abstract": "We introduce PathBench-MIL, an open-source AutoML and benchmarking framework for multiple instance learning (MIL) in histopathology. The system automates end-to-end MIL pipeline construction, including preprocessing, feature extraction, and MIL-aggregation, and provides reproducible benchmarking of dozens of MIL models and feature extractors. PathBench-MIL integrates visualization tooling, a unified configuration system, and modular extensibility, enabling rapid experimentation and standardization across datasets and tasks. PathBench-MIL is publicly available at https://github.com/Sbrussee/PathBench-MIL", "date": "2025-12-19", "categories": []}
{"id": "2512.17621", "title": "PathFLIP: Fine-grained Language-Image Pretraining for Versatile Computational Pathology", "abstract": "While Vision-Language Models (VLMs) have achieved notable progress in computational pathology (CPath), the gigapixel scale and spatial heterogeneity of Whole Slide Images (WSIs) continue to pose challenges for multimodal understanding. Existing alignment methods struggle to capture fine-grained correspondences between textual descriptions and visual cues across thousands of patches from a slide, compromising their performance on downstream tasks. In this paper, we propose PathFLIP (Pathology Fine-grained Language-Image Pretraining), a novel framework for holistic WSI interpretation. PathFLIP decomposes slide-level captions into region-level subcaptions and generates text-conditioned region embeddings to facilitate precise visual-language grounding. By harnessing Large Language Models (LLMs), PathFLIP can seamlessly follow diverse clinical instructions and adapt to varied diagnostic contexts. Furthermore, it exhibits versatile capabilities across multiple paradigms, efficiently handling slide-level classification and retrieval, fine-grained lesion localization, and instruction following. Extensive experiments demonstrate that PathFLIP outperforms existing large-scale pathological VLMs on four representative benchmarks while requiring significantly less training data, paving the way for fine-grained, instruction-aware WSI interpretation in clinical practice.", "date": "2025-12-19", "categories": []}
{"id": "2512.17726", "title": "MambaMIL+: Modeling Long-Term Contextual Patterns for Gigapixel Whole Slide Image", "abstract": "Whole-slide images (WSIs) are an important data modality in computational pathology, yet their gigapixel resolution and lack of fine-grained annotations challenge conventional deep learning models. Multiple instance learning (MIL) offers a solution by treating each WSI as a bag of patch-level instances, but effectively modeling ultra-long sequences with rich spatial context remains difficult. Recently, Mamba has emerged as a promising alternative for long sequence learning, scaling linearly to thousands of tokens. However, despite its efficiency, it still suffers from limited spatial context modeling and memory decay, constraining its effectiveness to WSI analysis. To address these limitations, we propose MambaMIL+, a new MIL framework that explicitly integrates spatial context while maintaining long-range dependency modeling without memory forgetting. Specifically, MambaMIL+ introduces 1) overlapping scanning, which restructures the patch sequence to embed spatial continuity and instance correlations; 2) a selective stripe position encoder (S2PE) that encodes positional information while mitigating the biases of fixed scanning orders; and 3) a contextual token selection (CTS) mechanism, which leverages supervisory knowledge to dynamically enlarge the contextual memory for stable long-range modeling. Extensive experiments on 20 benchmarks across diagnostic classification, molecular prediction, and survival analysis demonstrate that MambaMIL+ consistently achieves state-of-the-art performance under three feature extractors (ResNet-50, PLIP, and CONCH), highlighting its effectiveness and robustness for large-scale computational pathology", "date": "2025-12-19", "categories": []}
{"id": "2509.02600", "title": "Team Westwood Solution for MIDOG 2025 Challenge: An Ensemble-CNN-Based Approach For Mitosis Detection And Classification", "abstract": "This abstract presents our solution (Team Westwood) for mitosis detection and atypical mitosis classification in the MItosis DOmain Generalization (MIDOG) 2025 challenge. For mitosis detection, we trained an nnUNetV2 for initial mitosis candidate screening with high sensitivity, followed by a random forest classifier ensembling predictions of three convolutional neural networks (CNNs): EfficientNet-b3, EfficientNet-b5, and EfficientNetV2-s. For the atypical mitosis classification, we trained another random forest classifier ensembling the predictions of three CNNs: EfficientNet-b3, EfficientNet-b5, and InceptionV3. On the preliminary test set, our solution achieved an F1 score of 0.7450 for track 1 mitosis detection, and a balanced accuracy of 0.8722 for track 2 atypical mitosis classification. On the final test set, our solution achieved an F1 score of 0.6972 for track 1 mitosis detection, and a balanced accuracy of 0.8242 for track 2 atypical mitosis classification.", "date": "2025-08-29", "categories": []}
{"id": "2512.15171", "title": "Cross-modal ultra-scale learning with tri-modalities of renal biopsy images for glomerular multi-disease auxiliary diagnosis", "abstract": "Constructing a multi-modal automatic classification model based on three types of renal biopsy images can assist pathologists in glomerular multi-disease identification. However, the substantial scale difference between transmission electron microscopy (TEM) image features at the nanoscale and optical microscopy (OM) or immunofluorescence microscopy (IM) images at the microscale poses a challenge for existing multi-modal and multi-scale models in achieving effective feature fusion and improving classification accuracy. To address this issue, we propose a cross-modal ultra-scale learning network (CMUS-Net) for the auxiliary diagnosis of multiple glomerular diseases. CMUS-Net utilizes multiple ultrastructural information to bridge the scale difference between nanometer and micrometer images. Specifically, we introduce a sparse multi-instance learning module to aggregate features from TEM images. Furthermore, we design a cross-modal scale attention module to facilitate feature interaction, enhancing pathological semantic information. Finally, multiple loss functions are combined, allowing the model to weigh the importance among different modalities and achieve precise classification of glomerular diseases. Our method follows the conventional process of renal biopsy pathology diagnosis and, for the first time, performs automatic classification of multiple glomerular diseases including IgA nephropathy (IgAN), membranous nephropathy (MN), and lupus nephritis (LN) based on images from three modalities and two scales. On an in-house dataset, CMUS-Net achieves an ACC of 95.37+/-2.41%, an AUC of 99.05+/-0.53%, and an F1-score of 95.32+/-2.41%. Extensive experiments demonstrate that CMUS-Net outperforms other well-known multi-modal or multi-scale methods and show its generalization capability in staging MN. Code is available at https://github.com/SMU-GL-Group/MultiModal_lkx/tree/main.", "date": "2025-12-17", "categories": []}
{"id": "2512.15410", "title": "Preserving Marker Specificity with Lightweight Channel-Independent Representation Learning", "abstract": "Multiplexed tissue imaging measures dozens of protein markers per cell, yet most deep learning models still apply early channel fusion, assuming shared structure across markers. We investigate whether preserving marker independence, combined with deliberately shallow architectures, provides a more suitable inductive bias for self-supervised representation learning in multiplex data than increasing model scale. Using a Hodgkin lymphoma CODEX dataset with 145,000 cells and 49 markers, we compare standard early-fusion CNNs with channel-separated architectures, including a marker-aware baseline and our novel shallow Channel-Independent Model (CIM-S) with 5.5K parameters. After contrastive pretraining and linear evaluation, early-fusion models show limited ability to retain marker-specific information and struggle particularly with rare-cell discrimination. Channel-independent architectures, and CIM-S in particular, achieve substantially stronger representations despite their compact size. These findings are consistent across multiple self-supervised frameworks, remain stable across augmentation settings, and are reproducible across both the 49-marker and reduced 18-marker settings. These results show that lightweight, channel-independent architectures can match or surpass deep early-fusion CNNs and foundation models for multiplex representation learning. Code is available at https://github.com/SimonBon/CIM-S.", "date": "2025-12-17", "categories": []}
{"id": "2508.00969", "title": "Masked Omics Modeling for Multimodal Representation Learning across Histopathology and Molecular Profiles", "abstract": "Self-supervised learning (SSL) has driven major advances in computational pathology by enabling the learning of rich representations from histopathology data. Yet, tissue analysis alone may fall short in capturing broader molecular complexity, as key complementary information resides in high-dimensional omics profiles such as transcriptomics, methylomics, and genomics. To address this gap, we introduce MORPHEUS, the first multimodal pre-training strategy that integrates histopathology images and multi-omics data within a shared transformer-based architecture. At its core, MORPHEUS relies on a novel masked omics modeling objective that encourages the model to learn meaningful cross-modal relationships. This yields a general-purpose pre-trained encoder that can be applied to histopathology alone or in combination with any subset of omics modalities. Beyond inference, MORPHEUS also supports flexible any-to-any omics reconstruction, enabling one or more omics profiles to be reconstructed from any modality subset that includes histopathology. Pre-trained on a large pan-cancer cohort, MORPHEUS shows substantial improvements over supervised and SSL baselines across diverse tasks and modality combinations. Together, these capabilities position it as a promising direction for the development of multimodal foundation models in oncology. Code is publicly available at https://github.com/Lucas-rbnt/MORPHEUS", "date": "2025-08-01", "categories": []}
{"id": "2511.15464", "title": "SIGMMA: Hierarchical Graph-Based Multi-Scale Multi-modal Contrastive Alignment of Histopathology Image and Spatial Transcriptome", "abstract": "Recent advances in computational pathology have leveraged vision-language models to learn joint representations of Hematoxylin and Eosin (HE) images with spatial transcriptomic (ST) profiles. However, existing approaches typically align HE tiles with their corresponding ST profiles at a single scale, overlooking fine-grained cellular structures and their spatial organization. To address this, we propose Sigmma, a multi-modal contrastive alignment framework for learning hierarchical representations of HE images and spatial transcriptome profiles across multiple scales. Sigmma introduces multi-scale contrastive alignment, ensuring that representations learned at different scales remain coherent across modalities. Furthermore, by representing cell interactions as a graph and integrating inter- and intra-subgraph relationships, our approach effectively captures cell-cell interactions, ranging from fine to coarse, within the tissue microenvironment. We demonstrate that Sigmm learns representations that better capture cross-modal correspondences, leading to an improvement of avg. 9.78\\% in the gene-expression prediction task and avg. 26.93\\% in the cross-modal retrieval task across datasets. We further show that it learns meaningful multi-tissue organization in downstream analyses.", "date": "2025-11-19", "categories": []}
{"id": "2512.14540", "title": "CAPRMIL: Context-Aware Patch Representations for Multiple Instance Learning", "abstract": "In computational pathology, weak supervision has become the standard for deep learning due to the gigapixel scale of WSIs and the scarcity of pixel-level annotations, with Multiple Instance Learning (MIL) established as the principal framework for slide-level model training. In this paper, we introduce a novel setting for MIL methods, inspired by proceedings in Neural Partial Differential Equation (PDE) Solvers. Instead of relying on complex attention-based aggregation, we propose an efficient, aggregator-agnostic framework that removes the complexity of correlation learning from the MIL aggregator. CAPRMIL produces rich context-aware patch embeddings that promote effective correlation learning on downstream tasks. By projecting patch features -- extracted using a frozen patch encoder -- into a small set of global context/morphology-aware tokens and utilizing multi-head self-attention, CAPRMIL injects global context with linear computational complexity with respect to the bag size. Paired with a simple Mean MIL aggregator, CAPRMIL matches state-of-the-art slide-level performance across multiple public pathology benchmarks, while reducing the total number of trainable parameters by 48%-92.8% versus SOTA MILs, lowering FLOPs during inference by 52%-99%, and ranking among the best models on GPU memory efficiency and training time. Our results indicate that learning rich, context-aware instance representations before aggregation is an effective and scalable alternative to complex pooling for whole-slide analysis. Our code is available at https://github.com/mandlos/CAPRMIL", "date": "2025-12-16", "categories": []}
{"id": "2511.20734", "title": "Automated Histopathologic Assessment of Hirschsprung Disease Using a Multi-Stage Vision Transformer Framework", "abstract": "Hirschsprung Disease is characterized by the absence of ganglion cells in the myenteric plexus. Therefore, the correct identification of ganglion cells is crucial for diagnosing Hirschsprung disease. We introduce a three-stage analysis framework that mimics the pathologist's diagnostic approach. The framework, based on a Vision Transformer model (ViT-B/16), sequentially segments the muscularis propria, segments the myenteric plexus, and detects ganglion cells within anatomically valid regions. 30 whole-slide images of colon tissue were used, each containing manual annotations of muscularis, plexus, and ganglion cells. A 5-fold cross-validation scheme was applied to each stage, along with resolution-specific tiling strategies and tailored postprocessing to ensure anatomical consistency. The proposed method achieved a Dice coefficient of 89.9% and a Plexus Inclusion Rate of 100% for muscularis segmentation. Plexus segmentation reached a recall of 94.8%, a precision of 84.2% and a Ganglia Inclusion Rate of 99.7%. For ganglion cells annotated with high certainty, the model achieved 62.1\\% precision and 89.1% recall. When considering all annotated ganglion cells, regardless of certainty level, the overall precision was 67.0%. These results indicate that ViT-based models are effective at leveraging global tissue context and capturing cellular morphology at small scales, even within complex histological tissue structures. This multi-stage methodology has great potential to support digital pathology workflows by reducing inter-observer variability and assisting in the evaluation of Hirschsprung disease. The clinical impact will be evaluated in future work with larger multi-center datasets and additional expert annotations.", "date": "2025-11-25", "categories": []}
{"id": "2512.13440", "title": "IMILIA: interpretable multiple instance learning for inflammation prediction in IBD from H&E whole slide images", "abstract": "As the therapeutic target for Inflammatory Bowel Disease (IBD) shifts toward histologic remission, the accurate assessment of microscopic inflammation has become increasingly central for evaluating disease activity and response to treatment. In this work, we introduce IMILIA (Interpretable Multiple Instance Learning for Inflammation Analysis), an end-to-end framework designed for the prediction of inflammation presence in IBD digitized slides stained with hematoxylin and eosin (H&E), followed by the automated computation of markers characterizing tissue regions driving the predictions. IMILIA is composed of an inflammation prediction module, consisting of a Multiple Instance Learning (MIL) model, and an interpretability module, divided in two blocks: HistoPLUS, for cell instance detection, segmentation and classification; and EpiSeg, for epithelium segmentation. IMILIA achieves a cross-validation ROC-AUC of 0.83 on the discovery cohort, and a ROC-AUC of 0.99 and 0.84 on two external validation cohorts. The interpretability module yields biologically consistent insights: tiles with higher predicted scores show increased densities of immune cells (lymphocytes, plasmocytes, neutrophils and eosinophils), whereas lower-scored tiles predominantly contain normal epithelial cells. Notably, these patterns were consistent across all datasets. Code and models to partially replicate the results on the public IBDColEpi dataset can be found at https://github.com/owkin/imilia.", "date": "2025-12-15", "categories": []}
{"id": "2512.13600", "title": "DA-SSL: self-supervised domain adaptor to leverage foundational models in turbt histopathology slides", "abstract": "Recent deep learning frameworks in histopathology, particularly multiple instance learning (MIL) combined with pathology foundational models (PFMs), have shown strong performance. However, PFMs exhibit limitations on certain cancer or specimen types due to domain shifts - these cancer types were rarely used for pretraining or specimens contain tissue-based artifacts rarely seen within the pretraining population. Such is the case for transurethral resection of bladder tumor (TURBT), which are essential for diagnosing muscle-invasive bladder cancer (MIBC), but contain fragmented tissue chips and electrocautery artifacts and were not widely used in publicly available PFMs. To address this, we propose a simple yet effective domain-adaptive self-supervised adaptor (DA-SSL) that realigns pretrained PFM features to the TURBT domain without fine-tuning the foundational model itself. We pilot this framework for predicting treatment response in TURBT, where histomorphological features are currently underutilized and identifying patients who will benefit from neoadjuvant chemotherapy (NAC) is challenging. In our multi-center study, DA-SSL achieved an AUC of 0.77+/-0.04 in five-fold cross-validation and an external test accuracy of 0.84, sensitivity of 0.71, and specificity of 0.91 using majority voting. Our results demonstrate that lightweight domain adaptation with self-supervision can effectively enhance PFM-based MIL pipelines for clinically challenging histopathology tasks. Code is Available at https://github.com/zhanghaoyue/DA_SSL_TURBT.", "date": "2025-12-15", "categories": []}
{"id": "2409.13115", "title": "Multimodal Learning for Scalable Representation of High-Dimensional Medical Data", "abstract": "Integrating artificial intelligence (AI) with healthcare data is rapidly transforming medical diagnostics and driving progress toward precision medicine. However, effectively leveraging multimodal data, particularly digital pathology whole slide images (WSIs) and genomic sequencing, remains a significant challenge due to the intrinsic heterogeneity of these modalities and the need for scalable and interpretable frameworks. Existing diagnostic models typically operate on unimodal data, overlooking critical cross-modal interactions that can yield richer clinical insights. We introduce MarbliX (Multimodal Association and Retrieval with Binary Latent Indexed matriX), a self-supervised framework that learns to embed WSIs and immunogenomic profiles into compact, scalable binary codes, termed ``monogram.'' By optimizing a triplet contrastive objective across modalities, MarbliX captures high-resolution patient similarity in a unified latent space, enabling efficient retrieval of clinically relevant cases and facilitating case-based reasoning. \\textcolor{black}{In lung cancer, MarbliX achieves 85-89\\% across all evaluation metrics, outperforming histopathology (69-71\\%) and immunogenomics (73-76\\%). In kidney cancer, real-valued monograms yield the strongest performance (F1: 80-83\\%, Accuracy: 87-90\\%), with binary monograms slightly lower (F1: 78-82\\%).", "date": "2024-09-19", "categories": []}
{"id": "2512.11104", "title": "Information-driven Fusion of Pathology Foundation Models for Enhanced Disease Characterization", "abstract": "Foundation models (FMs) have demonstrated strong performance across diverse pathology tasks. While there are similarities in the pre-training objectives of FMs, there is still limited understanding of their complementarity, redundancy in embedding spaces, or biological interpretation of features. In this study, we propose an information-driven, intelligent fusion strategy for integrating multiple pathology FMs into a unified representation and systematically evaluate its performance for cancer grading and staging across three distinct diseases. Diagnostic H&E whole-slide images from kidney (519 slides), prostate (490 slides), and rectal (200 slides) cancers were dichotomized into low versus high grade or stage. Both tile-level FMs (Conch v1.5, MUSK, Virchow2, H-Optimus1, Prov-Gigapath) and slide-level FMs (TITAN, CHIEF, MADELEINE) were considered to train downstream classifiers. We then evaluated three FM fusion schemes at both tile and slide levels: majority-vote ensembling, naive feature concatenation, and intelligent fusion based on correlation-guided pruning of redundant features. Under patient-stratified cross-validation with hold-out testing, intelligent fusion of tile-level embeddings yielded consistent gains in classification performance across all three cancers compared with the best single FMs and naive fusion. Global similarity metrics revealed substantial alignment of FM embedding spaces, contrasted by lower local neighborhood agreement, indicating complementary fine-grained information across FMs. Attention maps showed that intelligent fusion yielded concentrated attention on tumor regions while reducing spurious focus on benign regions. Our findings suggest that intelligent, correlation-guided fusion of pathology FMs can yield compact, task-tailored representations that enhance both predictive performance and interpretability in downstream computational pathology tasks.", "date": "2025-12-11", "categories": []}
{"id": "2512.11745", "title": "mViSE: A Visual Search Engine for Analyzing Multiplex IHC Brain Tissue Images", "abstract": "Whole-slide multiplex imaging of brain tissue generates massive information-dense images that are challenging to analyze and require custom software. We present an alternative query-driven programming-free strategy using a multiplex visual search engine (mViSE) that learns the multifaceted brain tissue chemoarchitecture, cytoarchitecture, and myeloarchitecture. Our divide-and-conquer strategy organizes the data into panels of related molecular markers and uses self-supervised learning to train a multiplex encoder for each panel with explicit visual confirmation of successful learning. Multiple panels can be combined to process visual queries for retrieving similar communities of individual cells or multicellular niches using information-theoretic methods. The retrievals can be used for diverse purposes including tissue exploration, delineating brain regions and cortical cell layers, profiling and comparing brain regions without computer programming. We validated mViSE's ability to retrieve single cells, proximal cell pairs, tissue patches, delineate cortical layers, brain regions and sub-regions. mViSE is provided as an open-source QuPath plug-in.", "date": "2025-12-12", "categories": []}
{"id": "2512.10316", "title": "ConStruct: Structural Distillation of Foundation Models for Prototype-Based Weakly Supervised Histopathology Segmentation", "abstract": "Weakly supervised semantic segmentation (WSSS) in histopathology relies heavily on classification backbones, yet these models often localize only the most discriminative regions and struggle to capture the full spatial extent of tissue structures. Vision-language models such as CONCH offer rich semantic alignment and morphology-aware representations, while modern segmentation backbones like SegFormer preserve fine-grained spatial cues. However, combining these complementary strengths remains challenging, especially under weak supervision and without dense annotations. We propose a prototype learning framework for WSSS in histopathological images that integrates morphology-aware representations from CONCH, multi-scale structural cues from SegFormer, and text-guided semantic alignment to produce prototypes that are simultaneously semantically discriminative and spatially coherent. To effectively leverage these heterogeneous sources, we introduce text-guided prototype initialization that incorporates pathology descriptions to generate more complete and semantically accurate pseudo-masks. A structural distillation mechanism transfers spatial knowledge from SegFormer to preserve fine-grained morphological patterns and local tissue boundaries during prototype learning. Our approach produces high-quality pseudo masks without pixel-level annotations, improves localization completeness, and enhances semantic consistency across tissue types. Experiments on BCSS-WSSS datasets demonstrate that our prototype learning framework outperforms existing WSSS methods while remaining computationally efficient through frozen foundation model backbones and lightweight trainable adapters.", "date": "2025-12-11", "categories": []}
{"id": "2512.10808", "title": "Graph Laplacian Transformer with Progressive Sampling for Prostate Cancer Grading", "abstract": "Prostate cancer grading from whole-slide images (WSIs) remains a challenging task due to the large-scale nature of WSIs, the presence of heterogeneous tissue structures, and difficulty of selecting diagnostically relevant regions. Existing approaches often rely on random or static patch selection, leading to the inclusion of redundant or non-informative regions that degrade performance. To address this, we propose a Graph Laplacian Attention-Based Transformer (GLAT) integrated with an Iterative Refinement Module (IRM) to enhance both feature learning and spatial consistency. The IRM iteratively refines patch selection by leveraging a pretrained ResNet50 for local feature extraction and a foundation model in no-gradient mode for importance scoring, ensuring only the most relevant tissue regions are preserved. The GLAT models tissue-level connectivity by constructing a graph where patches serve as nodes, ensuring spatial consistency through graph Laplacian constraints and refining feature representations via a learnable filtering mechanism that enhances discriminative histological structures. Additionally, a convex aggregation mechanism dynamically adjusts patch importance to generate a robust WSI-level representation. Extensive experiments on five public and one private dataset demonstrate that our model outperforms state-of-the-art methods, achieving higher performance and spatial consistency while maintaining computational efficiency.", "date": "2025-12-11", "categories": []}
{"id": "2409.11110", "title": "Beyond accuracy: quantifying the reliability of Multiple Instance Learning for Whole Slide Image classification", "abstract": "Machine learning models have become integral to many fields, but their reliability, defined as producing dependable, trustworthy, and domain-consistent predictions, remains a critical concern. Multiple Instance Learning (MIL) models designed for Whole Slide Image (WSI) classification in computational pathology are rarely evaluated in terms of reliability, leaving a key gap in understanding their suitability for high-stakes applications like clinical decision-making. In this paper, we address this gap by introducing three quantitative metrics for reliability assessment and applying them to several widely used MIL architectures across three region-wise annotated pathology datasets. Our findings indicate that the mean pooling instance (MEAN-POOL-INS)model demonstrates superior reliability compared to other networks, despite its simple architectural design and computational efficiency. These findings underscore the need of reliability evaluation alongside predictive performance in MIL models and establish MEAN-POOL-INS as a strong, trustworthy baseline for future research.", "date": "2024-09-17", "categories": []}
{"id": "2512.08572", "title": "From Cells to Survival: Hierarchical Analysis of Cell Inter-Relations in Multiplex Microscopy for Lung Cancer Prognosis", "abstract": "The tumor microenvironment (TME) has emerged as a promising source of prognostic biomarkers. To fully leverage its potential, analysis methods must capture complex interactions between different cell types. We propose HiGINE -- a hierarchical graph-based approach to predict patient survival (short vs. long) from TME characterization in multiplex immunofluorescence (mIF) images and enhance risk stratification in lung cancer. Our model encodes both local and global inter-relations in cell neighborhoods, incorporating information about cell types and morphology. Multimodal fusion, aggregating cancer stage with mIF-derived features, further boosts performance. We validate HiGINE on two public datasets, demonstrating improved risk stratification, robustness, and generalizability.", "date": "2025-12-09", "categories": []}
{"id": "2507.14670", "title": "Gene-DML: Dual-Pathway Multi-Level Discrimination for Gene Expression Prediction from Histopathology Images", "abstract": "Accurately predicting gene expression from histopathology images offers a scalable and non-invasive approach to molecular profiling, with significant implications for precision medicine and computational pathology. However, existing methods often underutilize the cross-modal representation alignment between histopathology images and gene expression profiles across multiple representational levels, thereby limiting their prediction performance. To address this, we propose Gene-DML, a unified framework that structures latent space through Dual-pathway Multi-Level discrimination to enhance correspondence between morphological and transcriptional modalities. The multi-scale instance-level discrimination pathway aligns hierarchical histopathology representations extracted at local, neighbor, and global levels with gene expression profiles, capturing scale-aware morphological-transcriptional relationships. In parallel, the cross-level instance-group discrimination pathway enforces structural consistency between individual (image/gene) instances and modality-crossed (gene/image, respectively) groups, strengthening the alignment across modalities. By jointly modeling fine-grained and structural-level discrimination, Gene-DML is able to learn robust cross-modal representations, enhancing both predictive accuracy and generalization across diverse biological contexts. Extensive experiments on public spatial transcriptomics datasets demonstrate that Gene-DML achieves state-of-the-art performance in gene expression prediction. The code and processed datasets are available at https://github.com/YXSong000/Gene-DML.", "date": "2025-07-19", "categories": []}
{"id": "2512.05993", "title": "Domain-Specific Foundation Model Improves AI-Based Analysis of Neuropathology", "abstract": "Foundation models have transformed computational pathology by providing generalizable representations from large-scale histology datasets. However, existing models are predominantly trained on surgical pathology data, which is enriched for non-nervous tissue and overrepresents neoplastic, inflammatory, metabolic, and other non-neurological diseases. Neuropathology represents a markedly different domain of histopathology, characterized by unique cell types (neurons, glia, etc.), distinct cytoarchitecture, and disease-specific pathological features including neurofibrillary tangles, amyloid plaques, Lewy bodies, and pattern-specific neurodegeneration. This domain mismatch may limit the ability of general-purpose foundation models to capture the morphological patterns critical for interpreting neurodegenerative diseases such as Alzheimer's disease, Parkinson's disease, and cerebellar ataxias. To address this gap, we developed NeuroFM, a foundation model trained specifically on whole-slide images of brain tissue spanning diverse neurodegenerative pathologies. NeuroFM demonstrates superior performance compared to general-purpose models across multiple neuropathology-specific downstream tasks, including mixed dementia disease classification, hippocampal region segmentation, and neurodegenerative ataxia identification encompassing cerebellar essential tremor and spinocerebellar ataxia subtypes. This work establishes that domain-specialized foundation models trained on brain tissue can better capture neuropathology-specific features than models trained on general surgical pathology datasets. By tailoring foundation models to the unique morphological landscape of neurodegenerative diseases, NeuroFM enables more accurate and reliable AI-based analysis for brain disease diagnosis and research, setting a precedent for domain-specific model development in specialized areas of digital pathology.", "date": "2025-11-30", "categories": []}
{"id": "2512.06612", "title": "Learning Relative Gene Expression Trends from Pathology Images in Spatial Transcriptomics", "abstract": "Gene expression estimation from pathology images has the potential to reduce the RNA sequencing cost. Point-wise loss functions have been widely used to minimize the discrepancy between predicted and absolute gene expression values. However, due to the complexity of the sequencing techniques and intrinsic variability across cells, the observed gene expression contains stochastic noise and batch effects, and estimating the absolute expression values accurately remains a significant challenge. To mitigate this, we propose a novel objective of learning relative expression patterns rather than absolute levels. We assume that the relative expression levels of genes exhibit consistent patterns across independent experiments, even when absolute expression values are affected by batch effects and stochastic noise in tissue samples. Based on the assumption, we model the relation and propose a novel loss function called STRank that is robust to noise and batch effects. Experiments using synthetic datasets and real datasets demonstrate the effectiveness of the proposed method. The code is available at https://github.com/naivete5656/STRank.", "date": "2025-12-07", "categories": []}
{"id": "2512.07426", "title": "When normalization hallucinates: unseen risks in AI-powered whole slide image processing", "abstract": "Whole slide image (WSI) normalization remains a vital preprocessing step in computational pathology. Increasingly driven by deep learning, these models learn to approximate data distributions from training examples. This often results in outputs that gravitate toward the average, potentially masking diagnostically important features. More critically, they can introduce hallucinated content, artifacts that appear realistic but are not present in the original tissue, posing a serious threat to downstream analysis. These hallucinations are nearly impossible to detect visually, and current evaluation practices often overlook them. In this work, we demonstrate that the risk of hallucinations is real and underappreciated. While many methods perform adequately on public datasets, we observe a concerning frequency of hallucinations when these same models are retrained and evaluated on real-world clinical data. To address this, we propose a novel image comparison measure designed to automatically detect hallucinations in normalized outputs. Using this measure, we systematically evaluate several well-cited normalization methods retrained on real-world data, revealing significant inconsistencies and failures that are not captured by conventional metrics. Our findings underscore the need for more robust, interpretable normalization techniques and stricter validation protocols in clinical deployment.", "date": "2025-12-08", "categories": []}
{"id": "2512.04187", "title": "OnSight Pathology: A real-time platform-agnostic computational pathology companion for histopathology", "abstract": "The microscopic examination of surgical tissue remains a cornerstone of disease classification but relies on subjective interpretations and access to highly specialized experts, which can compromise accuracy and clinical care. While emerging breakthroughs in artificial intelligence (AI) offer promise for automated histological analysis, the growing number of proprietary digital pathology solutions has created barriers to real-world deployment. To address these challenges, we introduce OnSight Pathology, a platform-agnostic computer vision software that uses continuous custom screen captures to provide real-time AI inferences to users as they review digital slide images. Accessible as a single, self-contained executable file (https://onsightpathology.github.io/ ), OnSight Pathology operates locally on consumer-grade personal computers without complex software integration, enabling cost-effective and secure deployment in research and clinical workflows. Here we demonstrate the utility of OnSight Pathology using over 2,500 publicly available whole slide images across different slide viewers, as well as cases from our clinical digital pathology setup. The software's robustness is highlighted across routine histopathological tasks, including the classification of common brain tumor types, mitosis detection, and the quantification of immunohistochemical stains. A built-in multi-modal chat assistant provides verifiable descriptions of images, free of rigid class labels, for added quality control. Lastly, we show compatibility with live microscope camera feeds, including from personal smartphones, offering potential for deployment in more analog, inter-operative, and telepathology settings. Together, we highlight how OnSight Pathology can deliver real-time AI inferences across a broad range of pathology pipelines, removing key barriers to the adoption of AI tools in histopathology.", "date": "2025-12-03", "categories": []}
{"id": "2509.20923", "title": "Revisiting Data Challenges of Computational Pathology: A Pack-based Multiple Instance Learning Training Framework", "abstract": "Computational pathology (CPath) digitizes pathology slides into whole slide images (WSIs), enabling analysis for critical healthcare tasks such as cancer diagnosis and prognosis. However, WSIs possess extremely long sequence lengths (up to 200K), significant length variations (from 200 to 200K), and limited supervision. These extreme variations in sequence length lead to high data heterogeneity and redundancy. Conventional methods often compromise on training efficiency and optimization to preserve such heterogeneity under limited supervision. To comprehensively address these challenges, we propose a pack-based MIL framework. It packs multiple sampled, variable-length feature sequences into fixed-length ones, enabling batched training while preserving data heterogeneity. Moreover, we introduce a residual branch that composes discarded features from multiple slides into a hyperslide which is trained with tailored labels. It offers multi-slide supervision while mitigating feature loss from sampling. Meanwhile, an attention-driven downsampler is introduced to compress features in both branches to reduce redundancy. By alleviating these challenges, our approach achieves an accuracy improvement of up to 8% while using only 12% of the training time in the PANDA(UNI). Extensive experiments demonstrate that focusing data challenges in CPath holds significant potential in the era of foundation models. The code is https://github.com/FangHeng/PackMIL", "date": "2025-09-25", "categories": []}
{"id": "2512.03577", "title": "Cross-Stain Contrastive Learning for Paired Immunohistochemistry and Histopathology Slide Representation Learning", "abstract": "Universal, transferable whole-slide image (WSI) representations are central to computational pathology. Incorporating multiple markers (e.g., immunohistochemistry, IHC) alongside H&E enriches H&E-based features with diverse, biologically meaningful information. However, progress is limited by the scarcity of well-aligned multi-stain datasets. Inter-stain misalignment shifts corresponding tissue across slides, hindering consistent patch-level features and degrading slide-level embeddings. To address this, we curated a slide-level aligned, five-stain dataset (H&E, HER2, KI67, ER, PGR) to enable paired H&E-IHC learning and robust cross-stain representation. Leveraging this dataset, we propose Cross-Stain Contrastive Learning (CSCL), a two-stage pretraining framework with a lightweight adapter trained using patch-wise contrastive alignment to improve the compatibility of H&E features with corresponding IHC-derived contextual cues, and slide-level representation learning with Multiple Instance Learning (MIL), which uses a cross-stain attention fusion module to integrate stain-specific patch features and a cross-stain global alignment module to enforce consistency among slide-level embeddings across different stains. Experiments on cancer subtype classification, IHC biomarker status classification, and survival prediction show consistent gains, yielding high-quality, transferable H&E slide-level representations. The code and data are available at https://github.com/lily-zyz/CSCL.", "date": "2025-12-03", "categories": []}
{"id": "2512.03854", "title": "Prostate biopsy whole slide image dataset from an underrepresented Middle Eastern population", "abstract": "Artificial intelligence (AI) is increasingly used in digital pathology. Publicly available histopathology datasets remain scarce, and those that do exist predominantly represent Western populations. Consequently, the generalizability of AI models to populations from less digitized regions, such as the Middle East, is largely unknown. This motivates the public release of our dataset to support the development and validation of pathology AI models across globally diverse populations. We present 339 whole-slide images of prostate core needle biopsies from a consecutive series of 185 patients collected in Erbil, Iraq. The slides are associated with Gleason scores and International Society of Urological Pathology grades assigned independently by three pathologists. Scanning was performed using two high-throughput scanners (Leica and Hamamatsu) and one compact scanner (Grundium). All slides were de-identified and are provided in their native formats without further conversion. The dataset enables grading concordance analyses, color normalization, and cross-scanner robustness evaluations. Data will be deposited in the Bioimage Archive (BIA) under accession code: to be announced (TBA), and released under a CC BY 4.0 license.", "date": "2025-12-03", "categories": []}
{"id": "2405.02913", "title": "Fast TILs -- A Pipeline for Efficient TILs Estimation in Non-Small Cell Lung Cancer", "abstract": "Addressing the critical need for accurate prognostic biomarkers in cancer treatment, quantifying tumor-infiltrating lymphocytes (TILs) in non-small cell lung cancer (NSCLC) presents considerable challenges. Manual TIL quantification in whole slide images (WSIs) is laborious and subject to variability, potentially undermining patient outcomes. Our study introduces an automated pipeline that utilizes semi-stochastic patch sampling, patch classification to retain prognostically relevant patches, and cell quantification using the HoVer-Net model to streamline the TIL evaluation process. This pipeline efficiently excludes approximately 70% of areas not relevant for prognosis and requires only 5% of the remaining patches to maintain prognostic accuracy (c-index = 0.65). The computational efficiency achieved does not sacrifice prognostic accuracy, as demonstrated by the TILs score's strong association with patient survival, which outperforms traditional CD8 IHC scoring methods. While the pipeline demonstrates potential for enhancing NSCLC prognostication and personalization of treatment, comprehensive clinical validation is still required. Future research should focus on verifying its broader clinical utility and investigating additional biomarkers to improve NSCLC prognosis.", "date": "2024-05-05", "categories": []}
{"id": "2512.01681", "title": "Cross-Domain Validation of a Resection-Trained Self-Supervised Model on Multicentre Mesothelioma Biopsies", "abstract": "Accurate subtype classification and outcome prediction in mesothelioma are essential for guiding therapy and patient care. Most computational pathology models are trained on large tissue images from resection specimens, limiting their use in real-world settings where small biopsies are common. We show that a self-supervised encoder trained on resection tissue can be applied to biopsy material, capturing meaningful morphological patterns. Using these patterns, the model can predict patient survival and classify tumor subtypes. This approach demonstrates the potential of AI-driven tools to support diagnosis and treatment planning in mesothelioma.", "date": "2025-12-01", "categories": []}
{"id": "2503.17983", "title": "Histomorphology-Guided Prototypical Multi-Instance Learning for Breast Cancer WSI Classification", "abstract": "Histomorphology is crucial in cancer diagnosis. However, existing whole slide image (WSI) classification methods struggle to effectively incorporate histomorphology information, limiting their ability to capture key pathological features. Particularly when the number of instances within a bag is large and their features are complex, it becomes challenging to accurately identify instances decisive for the bag label, making these methods prone to interference from ambiguous instances. To address this limitation, we propose a novel Histomorphology-Guided Prototypical Multi-Instance Learning (HGPMIL) framework that explicitly learns histomorphology-guided prototypical representations by incorporating tumor cellularity, cellular morphology, and tissue architecture. Specifically, our approach consists of three key components: (1) estimating the importance of tumor-related histomorphology information at patch-level based on medical prior knowledge; (2) generating representative prototypes through histomorphology-prototypical clustering; and (3) enabling WSI classification through histomorphology-guided prototypical aggregation. HGPMIL adjusts the decision boundary by incorporating histomorphological importance to reduce instance label uncertainty, thereby reversely optimizing the bag-level boundary. Experimental results demonstrate its effectiveness, achieving high diagnostic accuracy for molecular subtyping, cancer subtyping and survival analysis. The code will be made available at https://github.com/Badgewho/HMDMIL.", "date": "2025-03-23", "categories": []}
{"id": "2506.00096", "title": "PathGene: Benchmarking Driver Gene Mutations and Exon Prediction Using Multicenter Lung Cancer Histopathology Image Dataset", "abstract": "Accurately predicting gene mutations, mutation subtypes and their exons in lung cancer is critical for personalized treatment planning and prognostic assessment. Faced with regional disparities in medical resources and the high cost of genomic assays, using artificial intelligence to infer these mutations and exon variants from routine histopathology images could greatly facilitate precision therapy. Although some prior studies have shown that deep learning can accelerate the prediction of key gene mutations from lung cancer pathology slides, their performance remains suboptimal and has so far been limited mainly to early screening tasks. To address these limitations, we have assembled PathGene, which comprises histopathology images paired with next-generation sequencing reports from 1,576 patients at the Second Xiangya Hospital, Central South University, and 448 TCGA-LUAD patients. This multi-center dataset links whole-slide images to driver gene mutation status, mutation subtypes, exon, and tumor mutational burden (TMB) status, with the goal of leveraging pathology images to predict mutations, subtypes, exon locations, and TMB for early genetic screening and to advance precision oncology. Unlike existing datasets, we provide molecular-level information related to histopathology images in PathGene to facilitate the development of biomarker prediction models. We benchmarked 11 multiple-instance learning methods on PathGene for mutation, subtype, exon, and TMB prediction tasks. These experimental methods provide valuable alternatives for early genetic screening of lung cancer patients and assisting clinicians to quickly develop personalized precision targeted treatment plans for patients. Code and data are available at https://github.com/panliangrui/NIPS2025/.", "date": "2025-05-30", "categories": []}
{"id": "2511.22131", "title": "Autonomous labeling of surgical resection margins using a foundation model", "abstract": "Assessing resection margins is central to pathological specimen evaluation and has profound implications for patient outcomes. Current practice employs physical inking, which is applied variably, and cautery artifacts can obscure the true margin on histological sections. We present a virtual inking network (VIN) that autonomously localizes the surgical cut surface on whole-slide images, reducing reliance on inks and standardizing margin-focused review. VIN uses a frozen foundation model as the feature extractor and a compact two-layer multilayer perceptron trained for patch-level classification of cautery-consistent features. The dataset comprised 120 hematoxylin and eosin (H&E) stained slides from 12 human tonsil tissue blocks, resulting in ~2 TB of uncompressed raw image data, where a board-certified pathologist provided boundary annotations. In blind testing with 20 slides from previously unseen blocks, VIN produced coherent margin overlays that qualitatively aligned with expert annotations across serial sections. Quantitatively, region-level accuracy was ~73.3% across the test set, with errors largely confined to limited areas that did not disrupt continuity of the whole-slide margin map. These results indicate that VIN captures cautery-related histomorphology and can provide a reproducible, ink-free margin delineation suitable for integration into routine digital pathology workflows and for downstream measurement of margin distances.", "date": "2025-11-27", "categories": []}
{"id": "2511.22739", "title": "All Centers Are at most a Few Tokens Apart: Knowledge Distillation with Domain Invariant Prompt Tuning", "abstract": "Domain generalization is critical in computational pathology (CPath) due to inherent domain shifts caused by variations in staining protocols, scanner devices, and imaging settings across clinical centers. Vision-language models (VLMs), such as PLIP-a pathology-tuned CLIP-trained on image-text pairs across diverse domains, serve as strong knowledge distillation sources. However, their zero-shot performance with predefined prompts remains limited due to sensitivity to prompt variations. Moreover, unlike natural images, histopathology centers lack semantic descriptors (e.g., 'sketch'), making it difficult to define domain-specific prompts for clinical centers. This requires a data-driven approach for learning domain-specific and ultimately class-generic continuous prompts. We propose Domain Invariant Prompt Tuning (DIPT) for knowledge distillation process, a novel step that learns multiple input tokens for each domain. These tokens are trained separately for each domain and are averaged across domains, leading to domain-invariant prompts. Our student model then distills knowledge from PLIP's text encoder by leveraging the prompts learned by DIPT. This leads to alignment of visual features with domain-invariant embeddings, enhancing generalization by training on multiple domains. Our method adds a significant improvement in average F1-score to existing state-of-the-art (SOTA) knowledge distillation approaches in domain generalization with histopathology datasets. This work helps the way of deploying robust CPath models in real-world clinical problems with heterogeneous data sources.", "date": "2025-11-27", "categories": []}
{"id": "2511.23204", "title": "Pathryoshka: Compressing Pathology Foundation Models via Multi-Teacher Knowledge Distillation with Nested Embeddings", "abstract": "Pathology foundation models (FMs) have driven significant progress in computational pathology. However, these high-performing models can easily exceed a billion parameters and produce high-dimensional embeddings, thus limiting their applicability for research or clinical use when computing resources are tight. Here, we introduce Pathryoshka, a multi-teacher distillation framework inspired by RADIO distillation and Matryoshka Representation Learning to reduce pathology FM sizes while allowing for adaptable embedding dimensions. We evaluate our framework with a distilled model on ten public pathology benchmarks with varying downstream tasks. Compared to its much larger teachers, Pathryoshka reduces the model size by 86-92% at on-par performance. It outperforms state-of-the-art single-teacher distillation models of comparable size by a median margin of 7.0 in accuracy. By enabling efficient local deployment without sacrificing accuracy or representational richness, Pathryoshka democratizes access to state-of-the-art pathology FMs for the broader research and clinical community.", "date": "2025-11-28", "categories": []}
{"id": "2503.01347", "title": "From Spots to Pixels: Dense Spatial Gene Expression Prediction from Histology Images", "abstract": "Spatial transcriptomics (ST) measures gene expression at fine-grained spatial resolution, offering insights into tissue molecular landscapes. Previous methods for spatial gene expression prediction typically crop spots of interest from histopathology slide images, and train models to map each spot to a corresponding gene expression profile. However, these methods inherently lose the spatial resolution in gene expression: 1) each spot often contains multiple cells with distinct gene expression profiles; 2) spots are typically defined at fixed spatial resolutions, limiting the ability to predict gene expression at varying scales. To address these limitations, this paper presents PixNet, a dense prediction network capable of predicting spatially resolved gene expression across spots of varying sizes and scales directly from histopathology slide images. Different from previous methods that map individual spots to gene expression values, we generate a spatially dense continuous gene expression map from the histopathology slide image, and aggregate values within spots of interest to predict the gene expression. Our PixNet outperforms state-of-the-art methods on four common ST datasets in multiple spatial scales. The source code will be publicly available.", "date": "2025-03-03", "categories": []}
{"id": "2511.19751", "title": "Leveraging Foundation Models for Histological Grading in Cutaneous Squamous Cell Carcinoma using PathFMTools", "abstract": "Despite the promise of computational pathology foundation models, adapting them to specific clinical tasks remains challenging due to the complexity of whole-slide image (WSI) processing, the opacity of learned features, and the wide range of potential adaptation strategies. To address these challenges, we introduce PathFMTools, a lightweight, extensible Python package that enables efficient execution, analysis, and visualization of pathology foundation models. We use this tool to interface with and evaluate two state-of-the-art vision-language foundation models, CONCH and MUSK, on the task of histological grading in cutaneous squamous cell carcinoma (cSCC), a critical criterion that informs cSCC staging and patient management. Using a cohort of 440 cSCC H&E WSIs, we benchmark multiple adaptation strategies, demonstrating trade-offs across prediction approaches and validating the potential of using foundation model embeddings to train small specialist models. These findings underscore the promise of pathology foundation models for real-world clinical applications, with PathFMTools enabling efficient analysis and validation.", "date": "2025-11-24", "categories": []}
{"id": "2511.18978", "title": "Zero-shot segmentation of skin tumors in whole-slide images with vision-language foundation models", "abstract": "Accurate annotation of cutaneous neoplasm biopsies represents a major challenge due to their wide morphological variability, overlapping histological patterns, and the subtle distinctions between benign and malignant lesions. Vision-language foundation models (VLMs), pre-trained on paired image-text corpora, learn joint representations that bridge visual features and diagnostic terminology, enabling zero-shot localization and classification of tissue regions without pixel-level labels. However, most existing VLM applications in histopathology remain limited to slide-level tasks or rely on coarse interactive prompts, and they struggle to produce fine-grained segmentations across gigapixel whole-slide images (WSIs). In this work, we introduce a zero-shot visual-language segmentation pipeline for whole-slide images (ZEUS), a fully automated, zero-shot segmentation framework that leverages class-specific textual prompt ensembles and frozen VLM encoders to generate high-resolution tumor masks in WSIs. By partitioning each WSI into overlapping patches, extracting visual embeddings, and computing cosine similarities against text prompts, we generate a final segmentation mask. We demonstrate competitive performance on two in-house datasets, primary spindle cell neoplasms and cutaneous metastases, highlighting the influence of prompt design, domain shifts, and institutional variability in VLMs for histopathology. ZEUS markedly reduces annotation burden while offering scalable, explainable tumor delineation for downstream diagnostic workflows.", "date": "2025-11-24", "categories": []}
{"id": "2511.17052", "title": "PathAgent: Toward Interpretable Analysis of Whole-slide Pathology Images via Large Language Model-based Agentic Reasoning", "abstract": "Analyzing whole-slide images (WSIs) requires an iterative, evidence-driven reasoning process that parallels how pathologists dynamically zoom, refocus, and self-correct while collecting the evidence. However, existing computational pipelines often lack this explicit reasoning trajectory, resulting in inherently opaque and unjustifiable predictions. To bridge this gap, we present PathAgent, a training-free, large language model (LLM)-based agent framework that emulates the reflective, stepwise analytical approach of human experts. PathAgent can autonomously explore WSI, iteratively and precisely locating significant micro-regions using the Navigator module, extracting morphology visual cues using the Perceptor, and integrating these findings into the continuously evolving natural language trajectories in the Executor. The entire sequence of observations and decisions forms an explicit chain-of-thought, yielding fully interpretable predictions. Evaluated across five challenging datasets, PathAgent exhibits strong zero-shot generalization, surpassing task-specific baselines in both open-ended and constrained visual question-answering tasks. Moreover, a collaborative evaluation with human pathologists confirms PathAgent's promise as a transparent and clinically grounded diagnostic assistant.", "date": "2025-11-21", "categories": []}
{"id": "2510.27237", "title": "Fusion of Multi-scale Heterogeneous Pathology Foundation Models for Whole Slide Image Analysis", "abstract": "Whole slide image (WSI) analysis has emerged as an increasingly essential technique in computational pathology. Recent advances in the pathology foundation models (FMs) have demonstrated significant advantages in deriving meaningful patch-level or slide-level multi-scale features from WSIs. However, current pathology FMs have exhibited substantial heterogeneity caused by diverse private training datasets and different network architectures. This heterogeneity introduces performance variability when we utilize the features from different FMs in the downstream tasks. To fully explore the advantages of multiple FMs effectively, in this work, we propose a novel framework for the fusion of multi-scale heterogeneous pathology FMs, called FuseCPath, yielding a model with a superior ensemble performance. The main contributions of our framework can be summarized as follows: (i) To guarantee the representativeness of the training patches, we propose a multi-view clustering-based method to filter out the discriminative patches via multiple FMs' embeddings. (ii) To effectively fuse the patch-level FMs, we devise a cluster-level re-embedding strategy to online capture patch-level local features. (iii) To effectively fuse the slide-level FMs, we devise a collaborative distillation strategy to explore the connections between slide-level FMs. Extensive experiments demonstrate that the proposed FuseCPath achieves state-of-the-art performance across multiple tasks on diverse datasets.", "date": "2025-10-31", "categories": []}
{"id": "2507.16476", "title": "Survival Modeling from Whole Slide Images via Patch-Level Graph Clustering and Mixture Density Experts", "abstract": "We propose a modular framework for predicting cancer specific survival directly from whole slide pathology images (WSIs). The framework consists of four key stages designed to capture prognostic and morphological heterogeneity. First, a Quantile Based Patch Filtering module selects prognostically informative tissue regions through quantile thresholding. Second, Graph Regularized Patch Clustering models phenotype level variations using a k nearest neighbor graph that enforces spatial and morphological coherence. Third, Hierarchical Feature Aggregation learns both intra and inter cluster dependencies to represent multiscale tumor organization. Finally, an Expert Guided Mixture Density Model estimates complex survival distributions via Gaussian mixtures, enabling fine grained risk prediction. Evaluated on TCGA LUAD, TCGA KIRC, and TCGA BRCA cohorts, our model achieves concordance indices of 0.653 ,0.719 ,and 0.733 respectively, surpassing existing state of the art approaches in survival prediction from WSIs.", "date": "2025-07-22", "categories": []}
{"id": "2511.12969", "title": "HiFusion: Hierarchical Intra-Spot Alignment and Regional Context Fusion for Spatial Gene Expression Prediction from Histopathology", "abstract": "Spatial transcriptomics (ST) bridges gene expression and tissue morphology but faces clinical adoption barriers due to technical complexity and prohibitive costs. While computational methods predict gene expression from H&E-stained whole-slide images (WSIs), existing approaches often fail to capture the intricate biological heterogeneity within spots and are susceptible to morphological noise when integrating contextual information from surrounding tissue. To overcome these limitations, we propose HiFusion, a novel deep learning framework that integrates two complementary components. First, we introduce the Hierarchical Intra-Spot Modeling module that extracts fine-grained morphological representations through multi-resolution sub-patch decomposition, guided by a feature alignment loss to ensure semantic consistency across scales. Concurrently, we present the Context-aware Cross-scale Fusion module, which employs cross-attention to selectively incorporate biologically relevant regional context, thereby enhancing representational capacity. This architecture enables comprehensive modeling of both cellular-level features and tissue microenvironmental cues, which are essential for accurate gene expression prediction. Extensive experiments on two benchmark ST datasets demonstrate that HiFusion achieves state-of-the-art performance across both 2D slide-wise cross-validation and more challenging 3D sample-specific scenarios. These results underscore HiFusion's potential as a robust, accurate, and scalable solution for ST inference from routine histopathology.", "date": "2025-11-17", "categories": []}
{"id": "2511.14907", "title": "nnMIL: A generalizable multiple instance learning framework for computational pathology", "abstract": "Computational pathology holds substantial promise for improving diagnosis and guiding treatment decisions. Recent pathology foundation models enable the extraction of rich patch-level representations from large-scale whole-slide images (WSIs), but current approaches for aggregating these features into slide-level predictions remain constrained by design limitations that hinder generalizability and reliability. Here, we developed nnMIL, a simple yet broadly applicable multiple-instance learning framework that connects patch-level foundation models to robust slide-level clinical inference. nnMIL introduces random sampling at both the patch and feature levels, enabling large-batch optimization, task-aware sampling strategies, and efficient and scalable training across datasets and model architectures. A lightweight aggregator performs sliding-window inference to generate ensemble slide-level predictions and supports principled uncertainty estimation. Across 40,000 WSIs encompassing 35 clinical tasks and four pathology foundation models, nnMIL consistently outperformed existing MIL methods for disease diagnosis, histologic subtyping, molecular biomarker detection, and pan- cancer prognosis prediction. It further demonstrated strong cross-model generalization, reliable uncertainty quantification, and robust survival stratification in multiple external cohorts. In conclusion, nnMIL offers a practical and generalizable solution for translating pathology foundation models into clinically meaningful predictions, advancing the development and deployment of reliable AI systems in real-world settings.", "date": "2025-11-18", "categories": []}
{"id": "2511.15067", "title": "Deep Pathomic Learning Defines Prognostic Subtypes and Molecular Drivers in Colorectal Cancer", "abstract": "Precise prognostic stratification of colorectal cancer (CRC) remains a major clinical challenge due to its high heterogeneity. The conventional TNM staging system is inadequate for personalized medicine. We aimed to develop and validate a novel multiple instance learning model TDAM-CRC using histopathological whole-slide images for accurate prognostic prediction and to uncover its underlying molecular mechanisms. We trained the model on the TCGA discovery cohort (n=581), validated it in an independent external cohort (n=1031), and further we integrated multi-omics data to improve model interpretability and identify novel prognostic biomarkers. The results demonstrated that the TDAM-CRC achieved robust risk stratification in both cohorts. Its predictive performance significantly outperformed the conventional clinical staging system and multiple state-of-the-art models. The TDAM-CRC risk score was confirmed as an independent prognostic factor in multivariable analysis. Multi-omics analysis revealed that the high-risk subtype is closely associated with metabolic reprogramming and an immunosuppressive tumor microenvironment. Through interaction network analysis, we identified and validated Mitochondrial Ribosomal Protein L37 (MRPL37) as a key hub gene linking deep pathomic features to clinical prognosis. We found that high expression of MRPL37, driven by promoter hypomethylation, serves as an independent biomarker of favorable prognosis. Finally, we constructed a nomogram incorporating the TDAM-CRC risk score and clinical factors to provide a precise and interpretable clinical decision-making tool for CRC patients. Our AI-driven pathological model TDAM-CRC provides a robust tool for improved CRC risk stratification, reveals new molecular targets, and facilitates personalized clinical decision-making.", "date": "2025-11-19", "categories": []}
{"id": "2503.05933", "title": "Beyond H&E: Unlocking Pathological Insights with Polarization Imaging", "abstract": "Histopathology image analysis is fundamental to digital pathology, with hematoxylin and eosin (H&E) staining as the gold standard for diagnostic and prognostic assessments. While H&E imaging effectively highlights cellular and tissue structures, it lacks sensitivity to birefringence and tissue anisotropy, which are crucial for assessing collagen organization, fiber alignment, and microstructural alterations--key indicators of tumor progression, fibrosis, and other pathological conditions. To bridge this gap, we construct a polarization imaging system and curate a new dataset of over 13,000 paired Polar-H&E images. Visualizations of polarization properties reveal distinctive optical signatures in pathological tissues, underscoring its diagnostic value. Building on this dataset, we propose PolarHE, a dual-modality fusion framework that integrates H&E with polarization imaging, leveraging the latter ability to enhance tissue characterization. Our approach employs a feature decomposition strategy to disentangle common and modality specific features, ensuring effective multimodal representation learning. Through comprehensive validation, our approach significantly outperforms previous methods, achieving an accuracy of 86.70% on the Chaoyang dataset and 89.06% on the MHIST dataset. These results demonstrate that polarization imaging is a powerful and underutilized modality in computational pathology, enriching feature representation and improving diagnostic accuracy. PolarHE establishes a promising direction for multimodal learning, paving the way for more interpretable and generalizable pathology models.", "date": "2025-03-05", "categories": []}
{"id": "2510.02760", "title": "Hierarchical Generalized Category Discovery for Brain Tumor Classification in Digital Pathology", "abstract": "Accurate brain tumor classification is critical for intra-operative decision making in neuro-oncological surgery. However, existing approaches are restricted to a fixed set of predefined classes and are therefore unable to capture patterns of tumor types not available during training. Unsupervised learning can extract general-purpose features, but it lacks the ability to incorporate prior knowledge from labelled data, and semi-supervised methods often assume that all potential classes are represented in the labelled data. Generalized Category Discovery (GCD) aims to bridge this gap by categorizing both known and unknown classes within unlabelled data. To reflect the hierarchical structure of brain tumor taxonomies, in this work, we introduce Hierarchical Generalized Category Discovery for Brain Tumor Classification (HGCD-BT), a novel approach that integrates hierarchical clustering with contrastive learning. Our method extends contrastive learning based GCD by incorporating a novel semi-supervised hierarchical clustering loss. We evaluate HGCD-BT on OpenSRH, a dataset of stimulated Raman histology brain tumor images, achieving a +28% improvement in accuracy over state-of-the-art GCD methods for patch-level classification, particularly in identifying previously unseen tumor categories. Furthermore, we demonstrate the generalizability of HGCD-BT on slide-level classification of hematoxylin and eosin stained whole-slide images from the Digital Brain Tumor Atlas, confirming its utility across imaging modalities.", "date": "2025-10-03", "categories": []}
{"id": "2511.06266", "title": "Spatially-Aware Mixture of Experts with Log-Logistic Survival Modeling for Whole-Slide Images", "abstract": "Accurate survival prediction from histopathology whole-slide images (WSIs) remains challenging due to their gigapixel resolution, strong spatial heterogeneity, and complex survival distributions. We introduce a comprehensive computational pathology framework that addresses these limitations through four complementary innovations: (1) Quantile-Gated Patch Selection for dynamically identifying prognostically relevant regions, (2) Graph-Guided Clustering to group patches by spatial and morphological similarity, (3) Hierarchical Context Attention to model both local tissue interactions and global slide-level context, and (4) an Expert-Driven Mixture of Log-Logistics module that flexibly models complex survival distributions. Across large TCGA cohorts, our method achieves state-of-the-art performance, yielding time-dependent concordance indices of 0.644 on LUAD, 0.751 on KIRC, and 0.752 on BRCA, consistently outperforming both histology-only and multimodal baselines. The framework further provides improved calibration and interpretability, advancing the use of WSIs for personalized cancer prognosis.", "date": "2025-11-09", "categories": []}
{"id": "2511.11984", "title": "From Classification to Cross-Modal Understanding: Leveraging Vision-Language Models for Fine-Grained Renal Pathology", "abstract": "Fine-grained glomerular subtyping is central to kidney biopsy interpretation, but clinically valuable labels are scarce and difficult to obtain. Existing computational pathology approaches instead tend to evaluate coarse diseased classification under full supervision with image-only models, so it remains unclear how vision-language models (VLMs) should be adapted for clinically meaningful subtyping under data constraints. In this work, we model fine-grained glomerular subtyping as a clinically realistic few-shot problem and systematically evaluate both pathology-specialized and general-purpose vision-language models under this setting. We assess not only classification performance (accuracy, AUC, F1) but also the geometry of the learned representations, examining feature alignment between image and text embeddings and the separability of glomerular subtypes. By jointly analyzing shot count, model architecture and domain knowledge, and adaptation strategy, this study provides guidance for future model selection and training under real clinical data constraints. Our results indicate that pathology-specialized vision-language backbones, when paired with the vanilla fine-tuning, are the most effective starting point. Even with only 4-8 labeled examples per glomeruli subtype, these models begin to capture distinctions and show substantial gains in discrimination and calibration, though additional supervision continues to yield incremental improvements. We also find that the discrimination between positive and negative examples is as important as image-text alignment. Overall, our results show that supervision level and adaptation strategy jointly shape both diagnostic performance and multimodal structure, providing guidance for model selection, adaptation strategies, and annotation investment.", "date": "2025-11-15", "categories": []}
{"id": "2511.12044", "title": "FedSDA: Federated Stain Distribution Alignment for Non-IID Histopathological Image Classification", "abstract": "Federated learning (FL) has shown success in collaboratively training a model among decentralized data resources without directly sharing privacy-sensitive training data. Despite recent advances, non-IID (non-independent and identically distributed) data poses an inevitable challenge that hinders the use of FL. In this work, we address the issue of non-IID histopathological images with feature distribution shifts from an intuitive perspective that has only received limited attention. Specifically, we address this issue from the perspective of data distribution by solely adjusting the data distributions of all clients. Building on the success of diffusion models in fitting data distributions and leveraging stain separation to extract the pivotal features that are closely related to the non-IID properties of histopathological images, we propose a Federated Stain Distribution Alignment (FedSDA) method. FedSDA aligns the stain distribution of each client with a target distribution in an FL framework to mitigate distribution shifts among clients. Furthermore, considering that training diffusion models on raw data in FL has been shown to be susceptible to privacy leakage risks, we circumvent this problem while still effectively achieving alignment. Extensive experimental results show that FedSDA is not only effective in improving baselines that focus on mitigating disparities across clients' model updates but also outperforms baselines that address the non-IID data issues from the perspective of data distribution. We show that FedSDA provides valuable and practical insights for the computational pathology community.", "date": "2025-11-15", "categories": []}
{"id": "2511.12588", "title": "Rank-Aware Agglomeration of Foundation Models for Immunohistochemistry Image Cell Counting", "abstract": "Accurate cell counting in immunohistochemistry (IHC) images is critical for quantifying protein expression and aiding cancer diagnosis. However, the task remains challenging due to the chromogen overlap, variable biomarker staining, and diverse cellular morphologies. Regression-based counting methods offer advantages over detection-based ones in handling overlapped cells, yet rarely support end-to-end multi-class counting. Moreover, the potential of foundation models remains largely underexplored in this paradigm. To address these limitations, we propose a rank-aware agglomeration framework that selectively distills knowledge from multiple strong foundation models, leveraging their complementary representations to handle IHC heterogeneity and obtain a compact yet effective student model, CountIHC. Unlike prior task-agnostic agglomeration strategies that either treat all teachers equally or rely on feature similarity, we design a Rank-Aware Teacher Selecting (RATS) strategy that models global-to-local patch rankings to assess each teacher's inherent counting capacity and enable sample-wise teacher selection. For multi-class cell counting, we introduce a fine-tuning stage that reformulates the task as vision-language alignment. Discrete semantic anchors derived from structured text prompts encode both category and quantity information, guiding the regression of class-specific density maps and improving counting for overlapping cells. Extensive experiments demonstrate that CountIHC surpasses state-of-the-art methods across 12 IHC biomarkers and 5 tissue types, while exhibiting high agreement with pathologists' assessments. Its effectiveness on H&E-stained data further confirms the scalability of the proposed method.", "date": "2025-11-16", "categories": []}
{"id": "2511.13170", "title": "THIR: Topological Histopathological Image Retrieval", "abstract": "According to the World Health Organization, breast cancer claimed the lives of approximately 685,000 women in 2020. Early diagnosis and accurate clinical decision making are critical in reducing this global burden. In this study, we propose THIR, a novel Content-Based Medical Image Retrieval (CBMIR) framework that leverages topological data analysis specifically, Betti numbers derived from persistent homology to characterize and retrieve histopathological images based on their intrinsic structural patterns. Unlike conventional deep learning approaches that rely on extensive training, annotated datasets, and powerful GPU resources, THIR operates entirely without supervision. It extracts topological fingerprints directly from RGB histopathological images using cubical persistence, encoding the evolution of loops as compact, interpretable feature vectors. The similarity retrieval is then performed by computing the distances between these topological descriptors, efficiently returning the top-K most relevant matches.\n  Extensive experiments on the BreaKHis dataset demonstrate that THIR outperforms state of the art supervised and unsupervised methods. It processes the entire dataset in under 20 minutes on a standard CPU, offering a fast, scalable, and training free solution for clinical image retrieval.", "date": "2025-11-17", "categories": []}
{"id": "2511.11004", "title": "MeCaMIL: Causality-Aware Multiple Instance Learning for Fair and Interpretable Whole Slide Image Diagnosis", "abstract": "Multiple instance learning (MIL) has emerged as the dominant paradigm for whole slide image (WSI) analysis in computational pathology, achieving strong diagnostic performance through patch-level feature aggregation. However, existing MIL methods face critical limitations: (1) they rely on attention mechanisms that lack causal interpretability, and (2) they fail to integrate patient demographics (age, gender, race), leading to fairness concerns across diverse populations. These shortcomings hinder clinical translation, where algorithmic bias can exacerbate health disparities. We introduce \\textbf{MeCaMIL}, a causality-aware MIL framework that explicitly models demographic confounders through structured causal graphs. Unlike prior approaches treating demographics as auxiliary features, MeCaMIL employs principled causal inference -- leveraging do-calculus and collider structures -- to disentangle disease-relevant signals from spurious demographic correlations. Extensive evaluation on three benchmarks demonstrates state-of-the-art performance across CAMELYON16 (ACC/AUC/F1: 0.939/0.983/0.946), TCGA-Lung (0.935/0.979/0.931), and TCGA-Multi (0.977/0.993/0.970, five cancer types). Critically, MeCaMIL achieves superior fairness -- demographic disparity variance drops by over 65% relative reduction on average across attributes, with notable improvements for underserved populations. The framework generalizes to survival prediction (mean C-index: 0.653, +0.017 over best baseline across five cancer types). Ablation studies confirm causal graph structure is essential -- alternative designs yield 0.048 lower accuracy and 4.2x times worse fairness. These results establish MeCaMIL as a principled framework for fair, interpretable, and clinically actionable AI in digital pathology. Code will be released upon acceptance.", "date": "2025-11-14", "categories": []}
{"id": "2511.11324", "title": "NOVA: An Agentic Framework for Automated Histopathology Analysis and Discovery", "abstract": "Digitized histopathology analysis involves complex, time-intensive workflows and specialized expertise, limiting its accessibility. We introduce NOVA, an agentic framework that translates scientific queries into executable analysis pipelines by iteratively generating and running Python code. NOVA integrates 49 domain-specific tools (e.g., nuclei segmentation, whole-slide encoding) built on open-source software, and can also create new tools ad hoc. To evaluate such systems, we present SlideQuest, a 90-question benchmark -- verified by pathologists and biomedical scientists -- spanning data processing, quantitative analysis, and hypothesis testing. Unlike prior biomedical benchmarks focused on knowledge recall or diagnostic QA, SlideQuest demands multi-step reasoning, iterative coding, and computational problem solving. Quantitative evaluation shows NOVA outperforms coding-agent baselines, and a pathologist-verified case study links morphology to prognostically relevant PAM50 subtypes, demonstrating its scalable discovery potential.", "date": "2025-11-14", "categories": []}
{"id": "2511.11486", "title": "Multimodal Posterior Sampling-based Uncertainty in PD-L1 Segmentation from H&E Images", "abstract": "Accurate assessment of PD-L1 expression is critical for guiding immunotherapy, yet current immunohistochemistry (IHC) based methods are resource-intensive. We present nnUNet-B: a Bayesian segmentation framework that infers PD-L1 expression directly from H&E-stained histology images using Multimodal Posterior Sampling (MPS). Built upon nnUNet-v2, our method samples diverse model checkpoints during cyclic training to approximate the posterior, enabling both accurate segmentation and epistemic uncertainty estimation via entropy and standard deviation. Evaluated on a dataset of lung squamous cell carcinoma, our approach achieves competitive performance against established baselines with mean Dice Score and mean IoU of 0.805 and 0.709, respectively, while providing pixel-wise uncertainty maps. Uncertainty estimates show strong correlation with segmentation error, though calibration remains imperfect. These results suggest that uncertainty-aware H&E-based PD-L1 prediction is a promising step toward scalable, interpretable biomarker assessment in clinical workflows.", "date": "2025-11-14", "categories": []}
{"id": "2511.10432", "title": "Histology-informed tiling of whole tissue sections improves the interpretability and predictability of cancer relapse and genetic alterations", "abstract": "Histopathologists establish cancer grade by assessing histological structures, such as glands in prostate cancer. Yet, digital pathology pipelines often rely on grid-based tiling that ignores tissue architecture. This introduces irrelevant information and limits interpretability. We introduce histology-informed tiling (HIT), which uses semantic segmentation to extract glands from whole slide images (WSIs) as biologically meaningful input patches for multiple-instance learning (MIL) and phenotyping. Trained on 137 samples from the ProMPT cohort, HIT achieved a gland-level Dice score of 0.83 +/- 0.17. By extracting 380,000 glands from 760 WSIs across ICGC-C and TCGA-PRAD cohorts, HIT improved MIL models AUCs by 10% for detecting copy number variation (CNVs) in genes related to epithelial-mesenchymal transitions (EMT) and MYC, and revealed 15 gland clusters, several of which were associated with cancer relapse, oncogenic mutations, and high Gleason. Therefore, HIT improved the accuracy and interpretability of MIL predictions, while streamlining computations by focussing on biologically meaningful structures during feature extraction.", "date": "2025-11-13", "categories": []}
{"id": "2511.08573", "title": "SENCA-st: Integrating Spatial Transcriptomics and Histopathology with Cross Attention Shared Encoder for Region Identification in Cancer Pathology", "abstract": "Spatial transcriptomics is an emerging field that enables the identification of functional regions based on the spatial distribution of gene expression. Integrating this functional information present in transcriptomic data with structural data from histopathology images is an active research area with applications in identifying tumor substructures associated with cancer drug resistance. Current histopathology-spatial-transcriptomic region segmentation methods suffer due to either making spatial transcriptomics prominent by using histopathology features just to assist processing spatial transcriptomics data or using vanilla contrastive learning that make histopathology images prominent due to only promoting common features losing functional information. In both extremes, the model gets either lost in the noise of spatial transcriptomics or overly smoothed, losing essential information. Thus, we propose our novel architecture SENCA-st (Shared Encoder with Neighborhood Cross Attention) that preserves the features of both modalities. More importantly, it emphasizes regions that are structurally similar in histopathology but functionally different on spatial transcriptomics using cross-attention. We demonstrate the superior performance of our model that surpasses state-of-the-art methods in detecting tumor heterogeneity and tumor micro-environment regions, a clinically crucial aspect.", "date": "2025-11-11", "categories": []}
{"id": "2511.08896", "title": "Classifying Histopathologic Glioblastoma Sub-regions with EfficientNet", "abstract": "Glioblastoma (GBM) is the most common aggressive, fast-growing brain tumor, with a grim prognosis. Despite clinical diagnostic advancements, there have not been any substantial improvements to patient prognosis. Histopathological assessment of excised tumors is the first line of clinical diagnostic routine. We hypothesize that automated, robust, and accurate identification of distinct histological sub-regions within GBM could contribute to morphologically understanding this disease at scale. In this study, we designed a four-step deep learning approach to classify six (6) histopathological regions and quantitatively evaluated it on the BraTS-Path 2024 challenge dataset, which includes digitized Hematoxylin \\& Eosin (H\\&E) stained GBM tissue sections annotated for six distinct regions. We used the challenge's publicly available training dataset to develop and evaluate the effectiveness of several variants of EfficientNet architectures (i.e., B0, B1, B2, B3, B4). EfficientNet-B1 and EfficientNet-B4 achieved the best performance, achieving an F1 score of 0.98 in a 5-fold cross-validation configuration using the BraTS-Path training set. The quantitative performance evaluation of our proposed approach with EfficientNet-B1 on the BraTS-Path hold-out validation data and the final hidden testing data yielded F1 scores of 0.546 and 0.517, respectively, for the associated 6-class classification task. The difference in the performance on training, validation, and testing data highlights the challenge of developing models that generalize well to new data, which is crucial for clinical applications. The source code of the proposed approach can be found at the GitHub repository of Indiana University Division of Computational Pathology: https://github.com/IUCompPath/brats-path-2024-enet.", "date": "2025-11-12", "categories": []}
{"id": "2409.17063", "title": "Benchmarking Domain Generalization Algorithms in Computational Pathology", "abstract": "Deep learning models have shown immense promise in computational pathology (CPath) tasks, but their performance often suffers when applied to unseen data due to domain shifts. Addressing this requires domain generalization (DG) algorithms. However, a systematic evaluation of DG algorithms in the CPath context is lacking. This study aims to benchmark the effectiveness of 30 DG algorithms on 3 CPath tasks of varying difficulty through 7,560 cross-validation runs. We evaluate these algorithms using a unified and robust platform, incorporating modality-specific techniques and recent advances like pretrained foundation models. Our extensive cross-validation experiments provide insights into the relative performance of various DG strategies. We observe that self-supervised learning and stain augmentation consistently outperform other methods, highlighting the potential of pretrained models and data augmentation. Furthermore, we introduce a new pan-cancer tumor detection dataset (HISTOPANTUM) as a benchmark for future research. This study offers valuable guidance to researchers in selecting appropriate DG approaches for CPath tasks.", "date": "2024-09-25", "categories": []}
{"id": "2511.02826", "title": "PLUTO-4: Frontier Pathology Foundation Models", "abstract": "Foundation models trained on large-scale pathology image corpora have demonstrated strong transfer capabilities across diverse histopathology tasks. Building on this progress, we introduce PLUTO-4, our next generation of pathology foundation models that extend the Pathology-Universal Transformer (PLUTO) to frontier scale. We share two complementary Vision Transformer architectures in the PLUTO-4 family: a compact and efficient PLUTO-4S model optimized for multi-scale deployment using a FlexiViT setup with 2D-RoPE embeddings, and a frontier-scale PLUTO-4G model trained with a single patch size to maximize representation capacity and stability. Both models are pretrained using a self-supervised objective derived from DINOv2 on a large multi-institutional corpus containing 551,164 WSIs from 137,144 patients across over 50 institutions, spanning over 60 disease types and over 100 stains. Comprehensive evaluation across public and internal benchmarks demonstrates that PLUTO-4 achieves state-of-the-art performance on tasks requiring varying spatial and biological context, including tile classification, segmentation, and slide-level diagnosis. The compact PLUTO-4S provides high-throughput and robust performance for practical deployment, while PLUTO-4G establishes new performance frontiers across multiple pathology benchmarks, including an 11% improvement in dermatopathology diagnosis. These diverse improvements underscore PLUTO-4's potential to transform real-world applications as a backbone for translational research and diagnostic use cases.", "date": "2025-11-04", "categories": []}
{"id": "2511.07560", "title": "EvoPS: Evolutionary Patch Selection for Whole Slide Image Analysis in Computational Pathology", "abstract": "In computational pathology, the gigapixel scale of Whole-Slide Images (WSIs) necessitates their division into thousands of smaller patches. Analyzing these high-dimensional patch embeddings is computationally expensive and risks diluting key diagnostic signals with many uninformative patches. Existing patch selection methods often rely on random sampling or simple clustering heuristics and typically fail to explicitly manage the crucial trade-off between the number of selected patches and the accuracy of the resulting slide representation. To address this gap, we propose EvoPS (Evolutionary Patch Selection), a novel framework that formulates patch selection as a multi-objective optimization problem and leverages an evolutionary search to simultaneously minimize the number of selected patch embeddings and maximize the performance of a downstream similarity search task, generating a Pareto front of optimal trade-off solutions. We validated our framework across four major cancer cohorts from The Cancer Genome Atlas (TCGA) using five pretrained deep learning models to generate patch embeddings, including both supervised CNNs and large self-supervised foundation models. The results demonstrate that EvoPS can reduce the required number of training patch embeddings by over 90% while consistently maintaining or even improving the final classification F1-score compared to a baseline that uses all available patches' embeddings selected through a standard extraction pipeline. The EvoPS framework provides a robust and principled method for creating efficient, accurate, and interpretable WSI representations, empowering users to select an optimal balance between computational cost and diagnostic performance.", "date": "2025-11-10", "categories": []}
{"id": "2511.07941", "title": "Libra-MIL: Multimodal Prototypes Stereoscopic Infused with Task-specific Language Priors for Few-shot Whole Slide Image Classification", "abstract": "While Large Language Models (LLMs) are emerging as a promising direction in computational pathology, the substantial computational cost of giga-pixel Whole Slide Images (WSIs) necessitates the use of Multi-Instance Learning (MIL) to enable effective modeling. A key challenge is that pathological tasks typically provide only bag-level labels, while instance-level descriptions generated by LLMs often suffer from bias due to a lack of fine-grained medical knowledge. To address this, we propose that constructing task-specific pathological entity prototypes is crucial for learning generalizable features and enhancing model interpretability. Furthermore, existing vision-language MIL methods often employ unidirectional guidance, limiting cross-modal synergy. In this paper, we introduce a novel approach, Multimodal Prototype-based Multi-Instance Learning, that promotes bidirectional interaction through a balanced information compression scheme. Specifically, we leverage a frozen LLM to generate task-specific pathological entity descriptions, which are learned as text prototypes. Concurrently, the vision branch learns instance-level prototypes to mitigate the model's reliance on redundant data. For the fusion stage, we employ the Stereoscopic Optimal Transport (SOT) algorithm, which is based on a similarity metric, thereby facilitating broader semantic alignment in a higher-dimensional space. We conduct few-shot classification and explainability experiments on three distinct cancer datasets, and the results demonstrate the superior generalization capabilities of our proposed method.", "date": "2025-11-11", "categories": []}
{"id": "2401.08272", "title": "Siamese Content-based Search Engine for a More Transparent Skin and Breast Cancer Diagnosis through Histological Imaging", "abstract": "Computer Aid Diagnosis (CAD) has developed digital pathology with Deep Learning (DL)-based tools to assist pathologists in decision-making. Content-Based Histopathological Image Retrieval (CBHIR) is a novel tool to seek highly correlated patches in terms of similarity in histopathological features. In this work, we proposed two CBHIR approaches on breast (Breast-twins) and skin cancer (Skin-twins) data sets for robust and accurate patch-level retrieval, integrating a custom-built Siamese network as a feature extractor. The proposed Siamese network is able to generalize for unseen images by focusing on the similar histopathological features of the input pairs. The proposed CBHIR approaches are evaluated on the Breast (public) and Skin (private) data sets with top K accuracy. Finding the optimum amount of K is challenging, but also, as much as K increases, the dissimilarity between the query and the returned images increases which might mislead the pathologists. To the best of the author's belief, this paper is tackling this issue for the first time on histopathological images by evaluating the top first retrieved images. The Breast-twins model achieves 70% of the F1score at the top first, which exceeds the other state-of-the-art methods at a higher amount of K such as 5 and 400. Skin-twins overpasses the recently proposed Convolutional Auto Encoder (CAE) by 67%, increasing the precision. Besides, the Skin-twins model tackles the challenges of Spitzoid Tumors of Uncertain Malignant Potential (STUMP) to assist pathologists with retrieving top K images and their corresponding labels. So, this approach can offer a more explainable CAD tool to pathologists in terms of transparency, trustworthiness, or reliability among other characteristics.", "date": "2024-01-16", "categories": []}
{"id": "2411.00078", "title": "Evaluating Cell AI Foundation Models in Kidney Pathology with Human-in-the-Loop Enrichment", "abstract": "Training AI foundation models has emerged as a promising large-scale learning approach for addressing real-world healthcare challenges, including digital pathology. While many of these models have been developed for tasks like disease diagnosis and tissue quantification using extensive and diverse training datasets, their readiness for deployment on some arguably simplest tasks, such as nuclei segmentation within a single organ (e.g., the kidney), remains uncertain. This paper seeks to answer this key question, \"How good are we?\", by thoroughly evaluating the performance of recent cell foundation models on a curated multi-center, multi-disease, and multi-species external testing dataset. Additionally, we tackle a more challenging question, \"How can we improve?\", by developing and assessing human-in-the-loop data enrichment strategies aimed at enhancing model performance while minimizing the reliance on pixel-level human annotation. To address the first question, we curated a multicenter, multidisease, and multispecies dataset consisting of 2,542 kidney whole slide images (WSIs). Three state-of-the-art (SOTA) cell foundation models-Cellpose, StarDist, and CellViT-were selected for evaluation. To tackle the second question, we explored data enrichment algorithms by distilling predictions from the different foundation models with a human-in-the-loop framework, aiming to further enhance foundation model performance with minimal human efforts. Our experimental results showed that all three foundation models improved over their baselines with model fine-tuning with enriched data. Interestingly, the baseline model with the highest F1 score does not yield the best segmentation outcomes after fine-tuning. This study establishes a benchmark for the development and deployment of cell vision foundation models tailored for real-world data applications.", "date": "2024-10-31", "categories": []}
{"id": "2511.06709", "title": "K-Stain: Keypoint-Driven Correspondence for H&E-to-IHC Virtual Staining", "abstract": "Virtual staining offers a promising method for converting Hematoxylin and Eosin (H&E) images into Immunohistochemical (IHC) images, eliminating the need for costly chemical processes. However, existing methods often struggle to utilize spatial information effectively due to misalignment in tissue slices. To overcome this challenge, we leverage keypoints as robust indicators of spatial correspondence, enabling more precise alignment and integration of structural details in synthesized IHC images. We introduce K-Stain, a novel framework that employs keypoint-based spatial and semantic relationships to enhance synthesized IHC image fidelity. K-Stain comprises three main components: (1) a Hierarchical Spatial Keypoint Detector (HSKD) for identifying keypoints in stain images, (2) a Keypoint-aware Enhancement Generator (KEG) that integrates these keypoints during image generation, and (3) a Keypoint Guided Discriminator (KGD) that improves the discriminator's sensitivity to spatial details. Our approach leverages contextual information from adjacent slices, resulting in more accurate and visually consistent IHC images. Extensive experiments show that K-Stain outperforms state-of-the-art methods in quantitative metrics and visual quality.", "date": "2025-11-10", "categories": []}
{"id": "2511.05150", "title": "From Linear Probing to Joint-Weighted Token Hierarchy: A Foundation Model Bridging Global and Cellular Representations in Biomarker Detection", "abstract": "AI-based biomarkers can infer molecular features directly from hematoxylin & eosin (H&E) slides, yet most pathology foundation models (PFMs) rely on global patch-level embeddings and overlook cell-level morphology. We present a PFM model, JWTH (Joint-Weighted Token Hierarchy), which integrates large-scale self-supervised pretraining with cell-centric post-tuning and attention pooling to fuse local and global tokens. Across four tasks involving four biomarkers and eight cohorts, JWTH achieves up to 8.3% higher balanced accuracy and 1.2% average improvement over prior PFMs, advancing interpretable and robust AI-based biomarker detection in digital pathology.", "date": "2025-11-07", "categories": []}
{"id": "2511.05183", "title": "PySlyde: A Lightweight, Open-Source Toolkit for Pathology Preprocessing", "abstract": "The integration of artificial intelligence (AI) into pathology is advancing precision medicine by improving diagnosis, treatment planning, and patient outcomes. Digitised whole-slide images (WSIs) capture rich spatial and morphological information vital for understanding disease biology, yet their gigapixel scale and variability pose major challenges for standardisation and analysis. Robust preprocessing, covering tissue detection, tessellation, stain normalisation, and annotation parsing is critical but often limited by fragmented and inconsistent workflows. We present PySlyde, a lightweight, open-source Python toolkit built on OpenSlide to simplify and standardise WSI preprocessing. PySlyde provides an intuitive API for slide loading, annotation management, tissue detection, tiling, and feature extraction, compatible with modern pathology foundation models. By unifying these processes, it streamlines WSI preprocessing, enhances reproducibility, and accelerates the generation of AI-ready datasets, enabling researchers to focus on model development and downstream analysis.", "date": "2025-11-07", "categories": []}
{"id": "2509.15482", "title": "Comparing Computational Pathology Foundation Models using Representational Similarity Analysis", "abstract": "Foundation models are increasingly developed in computational pathology (CPath) given their promise in facilitating many downstream tasks. While recent studies have evaluated task performance across models, less is known about the structure and variability of their learned representations. Here, we systematically analyze the representational spaces of six CPath foundation models using techniques popularized in computational neuroscience. The models analyzed span vision-language contrastive learning (CONCH, PLIP, KEEP) and self-distillation (UNI (v2), Virchow (v2), Prov-GigaPath) approaches. Through representational similarity analysis using H&E image patches from TCGA, we find that UNI2 and Virchow2 have the most distinct representational structures, whereas Prov-Gigapath has the highest average similarity across models. Having the same training paradigm (vision-only vs. vision-language) did not guarantee higher representational similarity. The representations of all models showed a high slide-dependence, but relatively low disease-dependence. Stain normalization decreased slide-dependence for all models by a range of 5.5% (CONCH) to 20.5% (PLIP). In terms of intrinsic dimensionality, vision-language models demonstrated relatively compact representations, compared to the more distributed representations of vision-only models. These findings highlight opportunities to improve robustness to slide-specific features, inform model ensembling strategies, and provide insights into how training paradigms shape model representations. Our framework is extendable across medical imaging domains, where probing the internal representations of foundation models can support their effective development and deployment.", "date": "2025-09-18", "categories": []}
{"id": "2511.04171", "title": "Systematic Evaluation of Preprocessing Techniques for Accurate Image Registration in Digital Pathology", "abstract": "Image registration refers to the process of spatially aligning two or more images by mapping them into a common coordinate system, so that corresponding anatomical or tissue structures are matched across images. In digital pathology, registration enables direct comparison and integration of information from different stains or imaging modalities, sup-porting applications such as biomarker analysis and tissue reconstruction. Accurate registration of images from different modalities is an essential step in digital pathology. In this study, we investigated how various color transformation techniques affect image registration between hematoxylin and eosin (H&E) stained images and non-linear multimodal images. We used a dataset of 20 tissue sample pairs, with each pair undergoing several preprocessing steps, including different color transformation (CycleGAN, Macenko, Reinhard, Vahadane), inversion, contrast adjustment, intensity normalization, and denoising. All images were registered using the VALIS registration method, which first applies rigid registration and then performs non-rigid registration in two steps on both low and high-resolution images. Registration performance was evaluated using the relative Target Registration Error (rTRE). We reported the median of median rTRE values (MMrTRE) and the average of median rTRE values (AMrTRE) for each method. In addition, we performed a custom point-based evaluation using ten manually selected key points. Registration was done separately for two scenarios, using either the original or inverted multimodal images. In both scenarios, CycleGAN color transformation achieved the lowest registration errors, while the other methods showed higher errors. These findings show that applying color transformation before registration improves alignment between images from different modalities and supports more reliable analysis in digital pathology.", "date": "2025-11-06", "categories": []}
{"id": "2511.04615", "title": "Building Trust in Virtual Immunohistochemistry: Automated Assessment of Image Quality", "abstract": "Deep learning models can generate virtual immunohistochemistry (IHC) stains from hematoxylin and eosin (H&E) images, offering a scalable and low-cost alternative to laboratory IHC. However, reliable evaluation of image quality remains a challenge as current texture- and distribution-based metrics quantify image fidelity rather than the accuracy of IHC staining. Here, we introduce an automated and accuracy grounded framework to determine image quality across sixteen paired or unpaired image translation models. Using color deconvolution, we generate masks of pixels stained brown (i.e., IHC-positive) as predicted by each virtual IHC model. We use the segmented masks of real and virtual IHC to compute stain accuracy metrics (Dice, IoU, Hausdorff distance) that directly quantify correct pixel - level labeling without needing expert manual annotations. Our results demonstrate that conventional image fidelity metrics, including Frechet Inception Distance (FID), peak signal-to-noise ratio (PSNR), and structural similarity (SSIM), correlate poorly with stain accuracy and pathologist assessment. Paired models such as PyramidPix2Pix and AdaptiveNCE achieve the highest stain accuracy, whereas unpaired diffusion- and GAN-based models are less reliable in providing accurate IHC positive pixel labels. Moreover, whole-slide images (WSI) reveal performance declines that are invisible in patch-based evaluations, emphasizing the need for WSI-level benchmarks. Together, this framework defines a reproducible approach for assessing the quality of virtual IHC models, a critical step to accelerate translation towards routine use by pathologists.", "date": "2025-11-06", "categories": []}
{"id": "2511.03365", "title": "Morpho-Genomic Deep Learning for Ovarian Cancer Subtype and Gene Mutation Prediction from Histopathology", "abstract": "Ovarian cancer remains one of the most lethal gynecological malignancies, largely due to late diagnosis and extensive heterogeneity across subtypes. Current diagnostic methods are limited in their ability to reveal underlying genomic variations essential for precision oncology. This study introduces a novel hybrid deep learning pipeline that integrates quantitative nuclear morphometry with deep convolutional image features to perform ovarian cancer subtype classification and gene mutation inference directly from Hematoxylin and Eosin (H&E) histopathological images. Using $\\sim45,000$ image patches sourced from The Cancer Genome Atlas (TCGA) and public datasets, a fusion model combining a ResNet-50 Convolutional Neural Network (CNN) encoder and a Vision Transformer (ViT) was developed. This model successfully captured both local morphological texture and global tissue context. The pipeline achieved a robust overall subtype classification accuracy of $84.2\\%$ (Macro AUC of $0.87 \\pm 0.03$). Crucially, the model demonstrated the capacity for gene mutation inference with moderate-to-high accuracy: $AUC_{TP53} = 0.82 \\pm 0.02$, $AUC_{BRCA1} = 0.76 \\pm 0.04$, and $AUC_{ARID1A} = 0.73 \\pm 0.05$. Feature importance analysis established direct quantitative links, revealing that nuclear solidity and eccentricity were the dominant predictors for TP53 mutation. These findings validate that quantifiable histological phenotypes encode measurable genomic signals, paving the way for cost-effective, precision histopathology in ovarian cancer triage and diagnosis.", "date": "2025-11-05", "categories": []}
{"id": "2502.04946", "title": "SurGen: 1020 H&E-stained Whole Slide Images With Survival and Genetic Markers", "abstract": "Cancer remains one of the leading causes of morbidity and mortality worldwide. Comprehensive datasets that combine histopathological images with genetic and survival data across various tumour sites are essential for advancing computational pathology and personalised medicine. We present SurGen, a dataset comprising 1,020 H&E-stained whole-slide images (WSIs) from 843 colorectal cancer cases. The dataset includes detailed annotations for key genetic mutations (KRAS, NRAS, BRAF) and mismatch repair status, as well as survival data for 426 cases. We illustrate SurGen's utility with a proof-of-concept model that predicts mismatch repair status directly from WSIs, achieving a test area under the receiver operating characteristic curve of 0.8273. These preliminary results underscore the dataset's potential to facilitate research in biomarker discovery, prognostic modelling, and advanced machine learning applications in colorectal cancer and beyond. SurGen offers a valuable resource for the scientific community, enabling studies that require high-quality WSIs linked with comprehensive clinical and genetic information on colorectal cancer. Our initial findings affirm the dataset's capacity to advance diagnostic precision and foster the development of personalised treatment strategies in colorectal oncology. Data available online: https://doi.org/10.6019/S-BIAD1285.", "date": "2025-02-07", "categories": []}
{"id": "2506.13063", "title": "PRISM2: Unlocking Multi-Modal General Pathology AI with Clinical Dialogue", "abstract": "Recent rapid progress in the field of computational pathology has been enabled by foundation models. These models are beginning to move beyond encoding image patches towards whole-slide understanding but their clinical utility remains limited. In this work, we present PRISM2, a multimodal slide-level foundation model trained on data from 700,000 diagnostic specimen-report pairs, the largest vision (2.3 million whole slide images) and language (14M question-answer pairs) histopathology dataset to date. By learning through clinical-dialogue supervision, PRISM2 aligns histomorphologic features with the language of diagnostic reasoning, producing slide-level representations that support both direct diagnostic question-answering and transferable embeddings for downstream tasks. Without additional training, PRISM2 matches or exceeds the cancer-detection performance of clinical-grade products. This is observed without loss of generality on other tasks, where PRISM2 achieves top performance. Finally, using survival prediction as the example, we show that task-specific finetuning with a large dataset can outperform task-specific models, further improving performance. These results demonstrate how language-supervised pretraining provides a scalable, clinically grounded signal for learning generalizable pathology representations, bridging human diagnostic reasoning and foundation-model performance.", "date": "2025-06-16", "categories": []}
{"id": "2510.15164", "title": "Hyperparameter Optimization and Reproducibility in Deep Learning Model Training", "abstract": "Reproducibility remains a critical challenge in foundation model training for histopathology, often hindered by software randomness, hardware non-determinism, and inconsistent hyperparameter reporting. To investigate these issues, we trained a CLIP model on the QUILT-1M dataset and systematically evaluated the impact of different hyperparameter settings and augmentation strategies across three downstream histopathology datasets (PatchCamelyon, LC25000-Lung, and LC25000-Colon). Despite variability across runs, we identified clear trends: RandomResizedCrop values of 0.7-0.8 outperformed more aggressive (0.6) or conservative (0.9) settings, distributed training without local loss improved stability, and learning rates below 5.0e-5 consistently degraded performance across all datasets. The LC25000 (Colon) dataset consistently provided the most reproducible benchmark. These findings highlight that reproducibility in computational pathology depends not only on transparent documentation but also on carefully chosen experimental configurations, and we provide practical rules to guide future efforts in developing reproducible foundation models for digital pathology.", "date": "2025-10-16", "categories": []}
{"id": "2510.22630", "title": "Robust Atypical Mitosis Classification with DenseNet121: Stain-Aware Augmentation and Hybrid Loss for Domain Generalization", "abstract": "Atypical mitotic figures are important biomarkers of tumor aggressiveness in histopathology, yet reliable recognition remains challenging due to severe class imbalance and variability across imaging domains. We present a DenseNet-121-based framework tailored for atypical mitosis classification in the MIDOG 2025 (Track 2) setting. Our method integrates stain-aware augmentation (Macenko), geometric and intensity transformations, and imbalance-aware learning via weighted sampling with a hybrid objective combining class-weighted binary cross-entropy and focal loss. Trained end-to-end with AdamW and evaluated across multiple independent domains, the model demonstrates strong generalization under scanner and staining shifts, achieving balanced accuracy 85.0%, AUROC 0.927, sensitivity 89.2%, and specificity 80.9% on the official test set. These results indicate that combining DenseNet-121 with stain-aware augmentation and imbalance-adaptive objectives yields a robust, domain-generalizable framework for atypical mitosis classification suitable for real-world computational pathology workflows.", "date": "2025-10-26", "categories": []}
{"id": "2511.00119", "title": "GeneFlow: Translation of Single-cell Gene Expression to Histopathological Images via Rectified Flow", "abstract": "Spatial transcriptomics (ST) technologies can be used to align transcriptomes with histopathological morphology, presenting exciting new opportunities for biomolecular discovery. Using ST data, we construct a novel framework, GeneFlow, to map transcriptomics onto paired cellular images. By combining an attention-based RNA encoder with a conditional UNet guided by rectified flow, we generate high-resolution images with different staining methods (e.g. H&E, DAPI) to highlight various cellular/tissue structures. Rectified flow with high-order ODE solvers creates a continuous, bijective mapping between transcriptomics and image manifolds, addressing the many-to-one relationship inherent in this problem. Our method enables the generation of realistic cellular morphology features and spatially resolved intercellular interactions from observational gene expression profiles, provides potential to incorporate genetic/chemical perturbations, and enables disease diagnosis by revealing dysregulated patterns in imaging phenotypes. Our rectified flow-based method outperforms diffusion-based baseline method in all experiments. Code can be found at https://github.com/wangmengbo/GeneFlow.", "date": "2025-10-31", "categories": []}
{"id": "2511.01698", "title": "Progressive Translation of H&E to IHC with Enhanced Structural Fidelity", "abstract": "Compared to hematoxylin-eosin (H&E) staining, immunohistochemistry (IHC) not only maintains the structural features of tissue samples, but also provides high-resolution protein localization, which is essential for aiding in pathology diagnosis. Despite its diagnostic value, IHC remains a costly and labor-intensive technique. Its limited scalability and constraints in multiplexing further hinder widespread adoption, especially in resource-limited settings. Consequently, researchers are increasingly exploring computational stain translation techniques to synthesize IHC-equivalent images from H&E-stained slides, aiming to extract protein-level information more efficiently and cost-effectively. However, most existing stain translation techniques rely on a linearly weighted summation of multiple loss terms within a single objective function, strategy that often overlooks the interdepedence among these components-resulting in suboptimal image quality and an inability to simultaneously preserve structural authenticity and color fidelity. To address this limitation, we propose a novel network architecture that follows a progressive structure, incorporating color and cell border generation logic, which enables each visual aspect to be optimized in a stage-wise and decoupled manner. To validate the effectiveness of our proposed network architecture, we build upon the Adaptive Supervised PatchNCE (ASP) framework as our baseline. We introduce additional loss functions based on 3,3'-diaminobenzidine (DAB) chromogen concentration and image gradient, enhancing color fidelity and cell boundary clarity in the generated IHC images. By reconstructing the generation pipeline using our structure-color-cell boundary progressive mechanism, experiments on HER2 and ER datasets demonstrated that the model significantly improved visual quality and achieved finer structural details.", "date": "2025-11-03", "categories": []}
{"id": "2510.27158", "title": "How Close Are We? Limitations and Progress of AI Models in Banff Lesion Scoring", "abstract": "The Banff Classification provides the global standard for evaluating renal transplant biopsies, yet its semi-quantitative nature, complex criteria, and inter-observer variability present significant challenges for computational replication. In this study, we explore the feasibility of approximating Banff lesion scores using existing deep learning models through a modular, rule-based framework. We decompose each Banff indicator - such as glomerulitis (g), peritubular capillaritis (ptc), and intimal arteritis (v) - into its constituent structural and inflammatory components, and assess whether current segmentation and detection tools can support their computation. Model outputs are mapped to Banff scores using heuristic rules aligned with expert guidelines, and evaluated against expert-annotated ground truths. Our findings highlight both partial successes and critical failure modes, including structural omission, hallucination, and detection ambiguity. Even when final scores match expert annotations, inconsistencies in intermediate representations often undermine interpretability. These results reveal the limitations of current AI pipelines in replicating computational expert-level grading, and emphasize the importance of modular evaluation and computational Banff grading standard in guiding future model development for transplant pathology.", "date": "2025-10-31", "categories": []}
{"id": "2505.01984", "title": "Lifelong Whole Slide Image Analysis: Online Vision-Language Adaptation and Past-to-Present Gradient Distillation", "abstract": "Whole Slide Images (WSIs) play a crucial role in accurate cancer diagnosis and prognosis, as they provide tissue details at the cellular level. However, the rapid growth of computational tasks involving WSIs poses significant challenges. Given that WSIs are gigapixels in size, they present difficulties in terms of storage, processing, and model training. Therefore, it is essential to develop lifelong learning approaches for WSI analysis. In scenarios where slides are distributed across multiple institutes, we aim to leverage them to develop a unified online model as a computational tool for cancer diagnosis in clinical and hospital settings. In this study, we introduce ADaFGrad, a method designed to enhance lifelong learning for whole-slide image (WSI) analysis. First, we leverage pathology vision-language foundation models to develop a framework that enables interaction between a slide's regional tissue features and a predefined text-based prototype buffer. Additionally, we propose a gradient-distillation mechanism that mimics the gradient of a logit with respect to the classification-head parameters across past and current iterations in a continual-learning setting. We construct a sequence of six TCGA datasets for training and evaluation. Experimental results show that ADaFGrad outperforms both state-of-the-art WSI-specific and conventional continual-learning methods after only a few training epochs, exceeding them by up to +5.068% in the class-incremental learning scenario while exhibiting the least forgetting (i.e., retaining the most knowledge from previous tasks). Moreover, ADaFGrad surpasses its baseline by as much as +40.084% in accuracy, further demonstrating the effectiveness of the proposed modules.", "date": "2025-05-04", "categories": []}
{"id": "2505.20510", "title": "CPathAgent: An Agent-based Foundation Model for Interpretable High-Resolution Pathology Image Analysis Mimicking Pathologists' Diagnostic Logic", "abstract": "Recent advances in computational pathology have led to the emergence of numerous foundation models. These models typically rely on general-purpose encoders with multi-instance learning for whole slide image (WSI) classification or apply multimodal approaches to generate reports directly from images. However, these models cannot emulate the diagnostic approach of pathologists, who systematically examine slides at low magnification to obtain an overview before progressively zooming in on suspicious regions to formulate comprehensive diagnoses. Instead, existing models directly output final diagnoses without revealing the underlying reasoning process. To address this gap, we introduce CPathAgent, an innovative agent-based approach that mimics pathologists' diagnostic workflow by autonomously navigating across WSI based on observed visual features, thereby generating substantially more transparent and interpretable diagnostic summaries. To achieve this, we develop a multi-stage training strategy that unifies patch-level, region-level, and WSI-level capabilities within a single model, which is essential for replicating how pathologists understand and reason across diverse image scales. Additionally, we construct PathMMU-HR2, the first expert-validated benchmark for large region analysis. This represents a critical intermediate scale between patches and whole slides, reflecting a key clinical reality where pathologists typically examine several key large regions rather than entire slides at once. Extensive experiments demonstrate that CPathAgent consistently outperforms existing approaches across benchmarks at three different image scales, validating the effectiveness of our agent-based diagnostic approach and highlighting a promising direction for computational pathology.", "date": "2025-05-26", "categories": []}
{"id": "2508.15256", "title": "Normal and Abnormal Pathology Knowledge-Augmented Vision-Language Model for Anomaly Detection in Pathology Images", "abstract": "Anomaly detection in computational pathology aims to identify rare and scarce anomalies where disease-related data are often limited or missing. Existing anomaly detection methods, primarily designed for industrial settings, face limitations in pathology due to computational constraints, diverse tissue structures, and lack of interpretability. To address these challenges, we propose Ano-NAViLa, a Normal and Abnormal pathology knowledge-augmented Vision-Language model for Anomaly detection in pathology images. Ano-NAViLa is built on a pre-trained vision-language model with a lightweight trainable MLP. By incorporating both normal and abnormal pathology knowledge, Ano-NAViLa enhances accuracy and robustness to variability in pathology images and provides interpretability through image-text associations. Evaluated on two lymph node datasets from different organs, Ano-NAViLa achieves the state-of-the-art performance in anomaly detection and localization, outperforming competing models.", "date": "2025-08-21", "categories": []}
{"id": "2510.24115", "title": "HistoLens: An Interactive XAI Toolkit for Verifying and Mitigating Flaws in Vision-Language Models for Histopathology", "abstract": "For doctors to truly trust artificial intelligence, it can't be a black box. They need to understand its reasoning, almost as if they were consulting a colleague. We created HistoLens1 to be that transparent, collaborative partner. It allows a pathologist to simply ask a question in plain English about a tissue slide--just as they would ask a trainee. Our system intelligently translates this question into a precise query for its AI engine, which then provides a clear, structured report. But it doesn't stop there. If a doctor ever asks, \"Why?\", HistoLens can instantly provide a 'visual proof' for any finding--a heatmap that points to the exact cells and regions the AI used for its analysis. We've also ensured the AI focuses only on the patient's tissue, just like a trained pathologist would, by teaching it to ignore distracting background noise. The result is a workflow where the pathologist remains the expert in charge, using a trustworthy AI assistant to verify their insights and make faster, more confident diagnoses.", "date": "2025-10-28", "categories": []}
{"id": "2510.21083", "title": "Knowledge-Driven Vision-Language Model for Plexus Detection in Hirschsprung's Disease", "abstract": "Hirschsprung's disease is defined as the congenital absence of ganglion cells in some segment(s) of the colon. The muscle cannot make coordinated movements to propel stool in that section, most commonly leading to obstruction. The diagnosis and treatment for this disease require a clear identification of different region(s) of the myenteric plexus, where ganglion cells should be present, on the microscopic view of the tissue slide. While deep learning approaches, such as Convolutional Neural Networks, have performed very well in this task, they are often treated as black boxes, with minimal understanding gained from them, and may not conform to how a physician makes decisions. In this study, we propose a novel framework that integrates expert-derived textual concepts into a Contrastive Language-Image Pre-training-based vision-language model to guide plexus classification. Using prompts derived from expert sources (e.g., medical textbooks and papers) generated by large language models and reviewed by our team before being encoded with QuiltNet, our approach aligns clinically relevant semantic cues with visual features. Experimental results show that the proposed model demonstrated superior discriminative capability across different classification metrics as it outperformed CNN-based models, including VGG-19, ResNet-18, and ResNet-50; achieving an accuracy of 83.9%, a precision of 86.6%, and a specificity of 87.6%. These findings highlight the potential of multi-modal learning in histopathology and underscore the value of incorporating expert knowledge for more clinically relevant model outputs.", "date": "2025-10-24", "categories": []}
{"id": "2510.21479", "title": "ITC-RWKV: Interactive Tissue-Cell Modeling with Recurrent Key-Value Aggregation for Histopathological Subtyping", "abstract": "Accurate interpretation of histopathological images demands integration of information across spatial and semantic scales, from nuclear morphology and cellular textures to global tissue organization and disease-specific patterns. Although recent foundation models in pathology have shown strong capabilities in capturing global tissue context, their omission of cell-level feature modeling remains a key limitation for fine-grained tasks such as cancer subtype classification. To address this, we propose a dual-stream architecture that models the interplay between macroscale tissue features and aggregated cellular representations. To efficiently aggregate information from large cell sets, we propose a receptance-weighted key-value aggregation model, a recurrent transformer that captures inter-cell dependencies with linear complexity. Furthermore, we introduce a bidirectional tissue-cell interaction module to enable mutual attention between localized cellular cues and their surrounding tissue environment. Experiments on four histopathological subtype classification benchmarks show that the proposed method outperforms existing models, demonstrating the critical role of cell-level aggregation and tissue-cell interaction in fine-grained computational pathology.", "date": "2025-10-24", "categories": []}
{"id": "2506.02408", "title": "Revisiting End-to-End Learning with Slide-level Supervision in Computational Pathology", "abstract": "Pre-trained encoders for offline feature extraction followed by multiple instance learning (MIL) aggregators have become the dominant paradigm in computational pathology (CPath), benefiting cancer diagnosis and prognosis. However, performance limitations arise from the absence of encoder fine-tuning for downstream tasks and disjoint optimization with MIL. While slide-level supervised end-to-end (E2E) learning is an intuitive solution to this issue, it faces challenges such as high computational demands and suboptimal results. These limitations motivate us to revisit E2E learning. We argue that prior work neglects inherent E2E optimization challenges, leading to performance disparities compared to traditional two-stage methods. In this paper, we pioneer the elucidation of optimization challenge caused by sparse-attention MIL and propose a novel MIL called ABMILX. It mitigates this problem through global correlation-based attention refinement and multi-head mechanisms. With the efficient multi-scale random patch sampling strategy, an E2E trained ResNet with ABMILX surpasses SOTA foundation models under the two-stage paradigm across multiple challenging benchmarks, while remaining computationally efficient (<10 RTX3090 hours). We show the potential of E2E learning in CPath and calls for greater research focus in this area. The code is https://github.com/DearCaat/E2E-WSI-ABMILX.", "date": "2025-06-03", "categories": []}
{"id": "2510.15138", "title": "Fourier Transform Multiple Instance Learning for Whole Slide Image Classification", "abstract": "Whole Slide Image (WSI) classification relies on Multiple Instance Learning (MIL) with spatial patch features, yet existing methods struggle to capture global dependencies due to the immense size of WSIs and the local nature of patch embeddings. This limitation hinders the modeling of coarse structures essential for robust diagnostic prediction. We propose Fourier Transform Multiple Instance Learning (FFT-MIL), a framework that augments MIL with a frequency-domain branch to provide compact global context. Low-frequency crops are extracted from WSIs via the Fast Fourier Transform and processed through a modular FFT-Block composed of convolutional layers and Min-Max normalization to mitigate the high variance of frequency data. The learned global frequency feature is fused with spatial patch features through lightweight integration strategies, enabling compatibility with diverse MIL architectures. FFT-MIL was evaluated across six state-of-the-art MIL methods on three public datasets (BRACS, LUAD, and IMP). Integration of the FFT-Block improved macro F1 scores by an average of 3.51% and AUC by 1.51%, demonstrating consistent gains across architectures and datasets. These results establish frequency-domain learning as an effective and efficient mechanism for capturing global dependencies in WSI classification, complementing spatial features and advancing the scalability and accuracy of MIL-based computational pathology.", "date": "2025-10-16", "categories": []}
{"id": "2509.02593", "title": "Robust Pan-Cancer Mitotic Figure Detection with YOLOv12", "abstract": "Mitotic figures represent a key histoprognostic feature in tumor pathology, providing crucial insights into tumor aggressiveness and proliferation. However, their identification remains challenging, subject to significant inter-observer variability, even among experienced pathologists. To address this issue, the MItosis DOmain Generalization (MIDOG) 2025 challenge marks the third edition of an international competition aiming to develop robust mitosis detection algorithms. In this paper, we present a mitotic figure detection approach based on the state-of-the-art YOLOv12 object detection architecture. Our method achieved an F1-score of 0.801 on the preliminary test set (hotspots only) and ranked second on the final test leaderboard with an F1-score of 0.7216 across complex and heterogeneous whole-slide regions, without relying on external data.", "date": "2025-08-29", "categories": []}
{"id": "2509.02601", "title": "Foundation Model-Driven Classification of Atypical Mitotic Figures with Domain-Aware Training Strategies", "abstract": "We present a solution for the MIDOG 2025 Challenge Track~2, addressing binary classification of normal mitotic figures (NMFs) versus atypical mitotic figures (AMFs). The approach leverages pathology-specific foundation model H-optimus-0, selected based on recent cross-domain generalization benchmarks and our empirical testing, with Low-Rank Adaptation (LoRA) fine-tuning and MixUp augmentation. Implementation includes soft labels based on multi-expert consensus, hard negative mining, and adaptive focal loss, metric learning and domain adaptation. The method demonstrates both the promise and challenges of applying foundation models to this complex classification task, achieving reasonable performance in the preliminary evaluation phase.", "date": "2025-08-29", "categories": []}
{"id": "2510.14403", "title": "DCMIL: A Progressive Representation Learning of Whole Slide Images for Cancer Prognosis Analysis", "abstract": "The burgeoning discipline of computational pathology shows promise in harnessing whole slide images (WSIs) to quantify morphological heterogeneity and develop objective prognostic modes for human cancers. However, progress is impeded by the computational bottleneck of gigapixel-size inputs and the scarcity of dense manual annotations. Current methods often overlook fine-grained information across multi-magnification WSIs and variations in tumor microenvironments. Here, we propose an easy-to-hard progressive representation learning, termed dual-curriculum contrastive multi-instance learning (DCMIL), to efficiently process WSIs for cancer prognosis. The model does not rely on dense annotations and enables the direct transformation of gigapixel-size WSIs into outcome predictions. Extensive experiments on twelve cancer types (5,954 patients, 12.54 million tiles) demonstrate that DCMIL outperforms standard WSI-based prognostic models. Additionally, DCMIL identifies fine-grained prognosis-salient regions, provides robust instance uncertainty estimation, and captures morphological differences between normal and tumor tissues, with the potential to generate new biological insights. All codes have been made publicly accessible at https://github.com/tuuuc/DCMIL.", "date": "2025-10-16", "categories": []}
{"id": "2510.13995", "title": "Finding Holes: Pathologist Level Performance Using AI for Cribriform Morphology Detection in Prostate Cancer", "abstract": "Background: Cribriform morphology in prostate cancer is a histological feature that indicates poor prognosis and contraindicates active surveillance. However, it remains underreported and subject to significant interobserver variability amongst pathologists. We aimed to develop and validate an AI-based system to improve cribriform pattern detection.\n  Methods: We created a deep learning model using an EfficientNetV2-S encoder with multiple instance learning for end-to-end whole-slide classification. The model was trained on 640 digitised prostate core needle biopsies from 430 patients, collected across three cohorts. It was validated internally (261 slides from 171 patients) and externally (266 slides, 104 patients from three independent cohorts). Internal validation cohorts included laboratories or scanners from the development set, while external cohorts used completely independent instruments and laboratories. Annotations were provided by three expert uropathologists with known high concordance. Additionally, we conducted an inter-rater analysis and compared the model's performance against nine expert uropathologists on 88 slides from the internal validation cohort.\n  Results: The model showed strong internal validation performance (AUC: 0.97, 95% CI: 0.95-0.99; Cohen's kappa: 0.81, 95% CI: 0.72-0.89) and robust external validation (AUC: 0.90, 95% CI: 0.86-0.93; Cohen's kappa: 0.55, 95% CI: 0.45-0.64). In our inter-rater analysis, the model achieved the highest average agreement (Cohen's kappa: 0.66, 95% CI: 0.57-0.74), outperforming all nine pathologists whose Cohen's kappas ranged from 0.35 to 0.62.\n  Conclusion: Our AI model demonstrates pathologist-level performance for cribriform morphology detection in prostate cancer. This approach could enhance diagnostic reliability, standardise reporting, and improve treatment decisions for prostate cancer patients.", "date": "2025-10-15", "categories": []}
{"id": "2510.14800", "title": "Morphology-Aware Prognostic model for Five-Year Survival Prediction in Colorectal Cancer from H&E Whole Slide Images", "abstract": "Colorectal cancer (CRC) remains the third most prevalent malignancy globally, with approximately 154,000 new cases and 54,000 projected deaths anticipated for 2025. The recent advancement of foundation models in computational pathology has been largely propelled by task agnostic methodologies that can overlook organ-specific crucial morphological patterns that represent distinct biological processes that can fundamentally influence tumor behavior, therapeutic response, and patient outcomes. The aim of this study is to develop a novel, interpretable AI model, PRISM (Prognostic Representation of Integrated Spatial Morphology), that incorporates a continuous variability spectrum within each distinct morphology to characterize phenotypic diversity and reflecting the principle that malignant transformation occurs through incremental evolutionary processes rather than abrupt phenotypic shifts. PRISM is trained on 8.74 million histological images extracted from surgical resection specimens of 424 patients with stage III CRC. PRISM achieved superior prognostic performance for five-year OS (AUC = 0.70 +- 0.04; accuracy = 68.37% +- 4.75%; HR = 3.34, 95% CI = 2.28-4.90; p < 0.0001), outperforming existing CRC-specific methods by 15% and AI foundation models by ~23% accuracy. It showed sex-agnostic robustness (AUC delta = 0.02; accuracy delta = 0.15%) and stable performance across clinicopathological subgroups, with minimal accuracy fluctuation (delta = 1.44%) between 5FU/LV and CPT-11/5FU/LV regimens, replicating the Alliance cohort finding of no survival difference between treatments.", "date": "2025-10-16", "categories": []}
{"id": "2508.21041", "title": "Efficient Fine-Tuning of DINOv3 Pretrained on Natural Images for Atypical Mitotic Figure Classification (MIDOG 2025 Task 2 Winner)", "abstract": "Atypical mitotic figures (AMFs) represent abnormal cell division associated with poor prognosis. Yet their detection remains difficult due to low prevalence, subtle morphology, and inter-observer variability. The MIDOG 2025 challenge introduces a benchmark for AMF classification across multiple domains. In this work, we fine-tuned the recently published DINOv3-H+ vision transformer, pretrained on natural images, using low-rank adaptation (LoRA), training only ~1.3M parameters in combination with extensive augmentation and a domain-weighted Focal Loss to handle domain heterogeneity. Despite the domain gap, our fine-tuned DINOv3 transfers effectively to histopathology, reaching first place on the final test set. These results highlight the advantages of DINOv3 pretraining and underline the efficiency and robustness of our fine-tuning strategy, yielding state-of-the-art results for the atypical mitosis classification challenge in MIDOG 2025.", "date": "2025-08-28", "categories": []}
{"id": "2510.04587", "title": "Pathology-CoT: Learning Visual Chain-of-Thought Agent from Expert Whole Slide Image Diagnosis Behavior", "abstract": "Diagnosing a whole-slide image is an interactive, multi-stage process of changing magnification and moving between fields. Although recent pathology foundation models demonstrated superior performances, practical agentic systems that decide what field to examine next, adjust magnification, and deliver explainable diagnoses are still lacking. Such limitation is largely bottlenecked by data: scalable, clinically aligned supervision of expert viewing behavior that is tacit and experience-based, not documented in textbooks or internet, and therefore absent from LLM training. Here we introduce a framework designed to address this challenge through three key breakthroughs. First, the AI Session Recorder seamlessly integrates with standard whole-slide image viewers to unobtrusively record routine navigation and convert the viewer logs into standardized behavioral commands and bounding boxes. Second, a lightweight human-in-the-loop review turns AI-drafted rationales for behavioral commands into the Pathology-CoT dataset, a form of paired \"where to look\" and \"why it matters\", enabling six-fold faster labeling compared to manual constructing such Chain-of-Thought dataset. Using this behavioral data, we build Pathology-o3, a two-stage agent that first proposes important ROIs and then performs behavior-guided reasoning. On the gastrointestinal lymph-node metastasis detection task, our method achieved 100 recall on the internal validation from Stanford Medicine and 97.6 recall on an independent external validation from Sweden, exceeding the state-of-the-art OpenAI o3 model and generalizing across backbones. To our knowledge, Pathology-CoT constitutes one of the first behavior-grounded agentic systems in pathology. Turning everyday viewer logs into scalable, expert-validated supervision, our framework makes agentic pathology practical and establishes a path to human-aligned, upgradeable clinical AI.", "date": "2025-10-06", "categories": []}
{"id": "2505.04192", "title": "ViDRiP-LLaVA: A Dataset and Benchmark for Diagnostic Reasoning from Pathology Videos", "abstract": "We present ViDRiP-LLaVA, the first large multimodal model (LMM) in computational pathology that integrates three distinct image scenarios, including single patch images, automatically segmented pathology video clips, and manually segmented pathology videos. This integration closely mirrors the natural diagnostic process of pathologists. By generating detailed histological descriptions and culminating in a definitive sign-out diagnosis, ViDRiP-LLaVA bridges visual narratives with diagnostic reasoning. Central to our approach is the ViDRiP-Instruct dataset, comprising 4278 video and diagnosis-specific chain-of-thought instructional pairs sourced from educational histopathology videos on YouTube. Although high-quality data is critical for enhancing diagnostic reasoning, its creation is time-intensive and limited in volume. To overcome this challenge, we transfer knowledge from existing single-image instruction datasets to train on weakly annotated, keyframe-extracted clips, followed by fine-tuning on manually segmented videos. ViDRiP-LLaVA establishes a new benchmark in pathology video analysis and offers a promising foundation for future AI systems that support clinical decision-making through integrated visual and diagnostic reasoning. Our code, data, and model are publicly available at: https://github.com/QuIIL/ViDRiP-LLaVA.", "date": "2025-05-07", "categories": []}
{"id": "2510.04861", "title": "A Clinical-grade Universal Foundation Model for Intraoperative Pathology", "abstract": "Intraoperative pathology is pivotal to precision surgery, yet its clinical impact is constrained by diagnostic complexity and the limited availability of high-quality frozen-section data. While computational pathology has made significant strides, the lack of large-scale, prospective validation has impeded its routine adoption in surgical workflows. Here, we introduce CRISP, a clinical-grade foundation model developed on over 100,000 frozen sections from eight medical centers, specifically designed to provide Clinical-grade Robust Intraoperative Support for Pathology (CRISP). CRISP was comprehensively evaluated on more than 15,000 intraoperative slides across nearly 100 retrospective diagnostic tasks, including benign-malignant discrimination, key intraoperative decision-making, and pan-cancer detection, etc. The model demonstrated robust generalization across diverse institutions, tumor types, and anatomical sites-including previously unseen sites and rare cancers. In a prospective cohort of over 2,000 patients, CRISP sustained high diagnostic accuracy under real-world conditions, directly informing surgical decisions in 92.6% of cases. Human-AI collaboration further reduced diagnostic workload by 35%, avoided 105 ancillary tests and enhanced detection of micrometastases with 87.5% accuracy. Together, these findings position CRISP as a clinical-grade paradigm for AI-driven intraoperative pathology, bridging computational advances with surgical precision and accelerating the translation of artificial intelligence into routine clinical practice.", "date": "2025-10-06", "categories": []}
{"id": "2510.11182", "title": "Generalisation of automatic tumour segmentation in histopathological whole-slide images across multiple cancer types", "abstract": "Deep learning is expected to aid pathologists by automating tasks such as tumour segmentation. We aimed to develop one universal tumour segmentation model for histopathological images and examine its performance in different cancer types. The model was developed using over 20 000 whole-slide images from over 4 000 patients with colorectal, endometrial, lung, or prostate carcinoma. Performance was validated in pre-planned analyses on external cohorts with over 3 000 patients across six cancer types. Exploratory analyses included over 1 500 additional patients from The Cancer Genome Atlas. Average Dice coefficient was over 80% in all validation cohorts with en bloc resection specimens and in The Cancer Genome Atlas cohorts. No loss of performance was observed when comparing the universal model with models specialised on single cancer types. In conclusion, extensive and rigorous evaluations demonstrate that generic tumour segmentation by a single model is possible across cancer types, patient populations, sample preparations, and slide scanners.", "date": "2025-10-13", "categories": []}
{"id": "2503.11846", "title": "A Graph-Based Framework for Interpretable Whole Slide Image Analysis", "abstract": "The histopathological analysis of whole-slide images (WSIs) is fundamental to cancer diagnosis but is a time-consuming and expert-driven process. While deep learning methods show promising results, dominant patch-based methods artificially fragment tissue, ignore biological boundaries, and produce black-box predictions. We overcome these limitations with a novel framework that transforms gigapixel WSIs into biologically-informed graph representations and is interpretable by design. Our approach builds graph nodes from tissue regions that respect natural structures, not arbitrary grids. We introduce an adaptive graph coarsening technique, guided by learned embeddings, to efficiently merge homogeneous regions while preserving diagnostically critical details in heterogeneous areas. Each node is enriched with a compact, interpretable feature set capturing clinically-motivated priors. A graph attention network then performs diagnosis on this compact representation. We demonstrate strong performance on challenging cancer staging and survival prediction tasks. Crucially, our resource-efficient model ($>$13x fewer parameters and $>$300x less data) achieves results competitive with a massive foundation model, while offering full interpretability through feature attribution. Our code is publicly available at https://github.com/HistoGraph31/pix2pathology.", "date": "2025-03-14", "categories": []}
{"id": "2510.05315", "title": "DeepAf: One-Shot Spatiospectral Auto-Focus Model for Digital Pathology", "abstract": "While Whole Slide Imaging (WSI) scanners remain the gold standard for digitizing pathology samples, their high cost limits accessibility in many healthcare settings. Other low-cost solutions also face critical limitations: automated microscopes struggle with consistent focus across varying tissue morphology, traditional auto-focus methods require time-consuming focal stacks, and existing deep-learning approaches either need multiple input images or lack generalization capability across tissue types and staining protocols. We introduce a novel automated microscopic system powered by DeepAf, a novel auto-focus framework that uniquely combines spatial and spectral features through a hybrid architecture for single-shot focus prediction. The proposed network automatically regresses the distance to the optimal focal point using the extracted spatiospectral features and adjusts the control parameters for optimal image outcomes. Our system transforms conventional microscopes into efficient slide scanners, reducing focusing time by 80% compared to stack-based methods while achieving focus accuracy of 0.18 {\\mu}m on the same-lab samples, matching the performance of dual-image methods (0.19 {\\mu}m) with half the input requirements. DeepAf demonstrates robust cross-lab generalization with only 0.72% false focus predictions and 90% of predictions within the depth of field. Through an extensive clinical study of 536 brain tissue samples, our system achieves 0.90 AUC in cancer classification at 4x magnification, a significant achievement at lower magnification than typical 20x WSI scans. This results in a comprehensive hardware-software design enabling accessible, real-time digital pathology in resource-constrained settings while maintaining diagnostic accuracy.", "date": "2025-10-06", "categories": []}
{"id": "2510.05657", "title": "A Hierarchical Geometry-guided Transformer for Histological Subtyping of Primary Liver Cancer", "abstract": "Primary liver malignancies are widely recognized as the most heterogeneous and prognostically diverse cancers of the digestive system. Among these, hepatocellular carcinoma (HCC) and intrahepatic cholangiocarcinoma (ICC) emerge as the two principal histological subtypes, demonstrating significantly greater complexity in tissue morphology and cellular architecture than other common tumors. The intricate representation of features in Whole Slide Images (WSIs) encompasses abundant crucial information for liver cancer histological subtyping, regarding hierarchical pyramid structure, tumor microenvironment (TME), and geometric representation. However, recent approaches have not adequately exploited these indispensable effective descriptors, resulting in a limited understanding of histological representation and suboptimal subtyping performance. To mitigate these limitations, ARGUS is proposed to advance histological subtyping in liver cancer by capturing the macro-meso-micro hierarchical information within the TME. Specifically, we first construct a micro-geometry feature to represent fine-grained cell-level pattern via a geometric structure across nuclei, thereby providing a more refined and precise perspective for delineating pathological images. Then, a Hierarchical Field-of-Views (FoVs) Alignment module is designed to model macro- and meso-level hierarchical interactions inherent in WSIs. Finally, the augmented micro-geometry and FoVs features are fused into a joint representation via present Geometry Prior Guided Fusion strategy for modeling holistic phenotype interactions. Extensive experiments on public and private cohorts demonstrate that our ARGUS achieves state-of-the-art (SOTA) performance in histological subtyping of liver cancer, which provide an effective diagnostic tool for primary liver malignancies in clinical practice.", "date": "2025-10-07", "categories": []}
{"id": "2405.15264", "title": "Self-Contrastive Weakly Supervised Learning Framework for Prognostic Prediction Using Whole Slide Images", "abstract": "We present a pioneering investigation into the application of deep learning techniques to analyze histopathological images for addressing the substantial challenge of automated prognostic prediction. Prognostic prediction poses a unique challenge as the ground truth labels are inherently weak, and the model must anticipate future events that are not directly observable in the image. To address this challenge, we propose a novel three-part framework comprising of a convolutional network based tissue segmentation algorithm for region of interest delineation, a contrastive learning module for feature extraction, and a nested multiple instance learning classification module. Our study explores the significance of various regions of interest within the histopathological slides and exploits diverse learning scenarios. The pipeline is initially validated on artificially generated data and a simpler diagnostic task. Transitioning to prognostic prediction, tasks become more challenging. Employing bladder cancer as use case, our best models yield an AUC of 0.721 and 0.678 for recurrence and treatment outcome prediction respectively.", "date": "2024-05-24", "categories": []}
{"id": "2509.23640", "title": "EfficientMIL: Efficient Linear-Complexity MIL Method for WSI Classification", "abstract": "Whole slide images (WSIs) classification represents a fundamental challenge in computational pathology, where multiple instance learning (MIL) has emerged as the dominant paradigm. Current state-of-the-art (SOTA) MIL methods rely on attention mechanisms, achieving good performance but requiring substantial computational resources due to quadratic complexity when processing hundreds of thousands of patches. To address this computational bottleneck, we introduce EfficientMIL, a novel linear-complexity MIL approach for WSIs classification with the patches selection module Adaptive Patch Selector (APS) that we designed, replacing the quadratic-complexity self-attention mechanisms in Transformer-based MIL methods with efficient sequence models including RNN-based GRU, LSTM, and State Space Model (SSM) Mamba. EfficientMIL achieves significant computational efficiency improvements while outperforming other MIL methods across multiple histopathology datasets. On TCGA-Lung dataset, EfficientMIL-Mamba achieved AUC of 0.976 and accuracy of 0.933, while on CAMELYON16 dataset, EfficientMIL-GRU achieved AUC of 0.990 and accuracy of 0.975, surpassing previous state-of-the-art methods. Extensive experiments demonstrate that APS is also more effective for patches selection than conventional selection strategies.", "date": "2025-09-28", "categories": []}
{"id": "2510.03455", "title": "PEaRL: Pathway-Enhanced Representation Learning for Gene and Pathway Expression Prediction from Histology", "abstract": "Integrating histopathology with spatial transcriptomics (ST) provides a powerful opportunity to link tissue morphology with molecular function. Yet most existing multimodal approaches rely on a small set of highly variable genes, which limits predictive scope and overlooks the coordinated biological programs that shape tissue phenotypes. We present PEaRL (Pathway Enhanced Representation Learning), a multimodal framework that represents transcriptomics through pathway activation scores computed with ssGSEA. By encoding biologically coherent pathway signals with a transformer and aligning them with histology features via contrastive learning, PEaRL reduces dimensionality, improves interpretability, and strengthens cross-modal correspondence. Across three cancer ST datasets (breast, skin, and lymph node), PEaRL consistently outperforms SOTA methods, yielding higher accuracy for both gene- and pathway-level expression prediction (up to 58.9 percent and 20.4 percent increase in Pearson correlation coefficient compared to SOTA). These results demonstrate that grounding transcriptomic representation in pathways produces more biologically faithful and interpretable multimodal models, advancing computational pathology beyond gene-level embeddings.", "date": "2025-10-03", "categories": []}
{"id": "2510.03555", "title": "GAS-MIL: Group-Aggregative Selection Multi-Instance Learning for Ensemble of Foundation Models in Digital Pathology Image Analysis", "abstract": "Foundation models (FMs) have transformed computational pathology by providing powerful, general-purpose feature extractors. However, adapting and benchmarking individual FMs for specific diagnostic tasks is often time-consuming and resource-intensive, especially given their scale and diversity. To address this challenge, we introduce Group-Aggregative Selection Multi-Instance Learning (GAS-MIL), a flexible ensemble framework that seamlessly integrates features from multiple FMs, preserving their complementary strengths without requiring manual feature selection or extensive task-specific fine-tuning. Across classification tasks in three cancer datasets-prostate (PANDA), ovarian (UBC-OCEAN), and breast (TCGA-BrCa)-GAS-MIL consistently achieves superior or on-par performance relative to individual FMs and established MIL methods, demonstrating its robustness and generalizability. By enabling efficient integration of heterogeneous FMs, GAS-MIL streamlines model deployment for pathology and provides a scalable foundation for future multimodal and precision oncology applications.", "date": "2025-10-03", "categories": []}
{"id": "2510.04315", "title": "GenAR: Next-Scale Autoregressive Generation for Spatial Gene Expression Prediction", "abstract": "Spatial Transcriptomics (ST) offers spatially resolved gene expression but remains costly. Predicting expression directly from widely available Hematoxylin and Eosin (H&E) stained images presents a cost-effective alternative. However, most computational approaches (i) predict each gene independently, overlooking co-expression structure, and (ii) cast the task as continuous regression despite expression being discrete counts. This mismatch can yield biologically implausible outputs and complicate downstream analyses. We introduce GenAR, a multi-scale autoregressive framework that refines predictions from coarse to fine. GenAR clusters genes into hierarchical groups to expose cross-gene dependencies, models expression as codebook-free discrete token generation to directly predict raw counts, and conditions decoding on fused histological and spatial embeddings. From an information-theoretic perspective, the discrete formulation avoids log-induced biases and the coarse-to-fine factorization aligns with a principled conditional decomposition. Extensive experimental results on four Spatial Transcriptomics datasets across different tissue types demonstrate that GenAR achieves state-of-the-art performance, offering potential implications for precision medicine and cost-effective molecular profiling. Code is publicly available at https://github.com/oyjr/genar.", "date": "2025-10-05", "categories": []}
{"id": "2507.14497", "title": "Efficient Whole Slide Pathology VQA via Token Compression", "abstract": "Whole-slide images (WSIs) in pathology can reach up to 10,000 x 10,000 pixels, posing significant challenges for multimodal large language model (MLLM) due to long context length and high computational demands. Previous methods typically focus on patch-level analysis or slide-level classification using CLIP-based models with multi-instance learning, but they lack the generative capabilities needed for visual question answering (VQA). More recent MLLM-based approaches address VQA by feeding thousands of patch tokens directly into the language model, which leads to excessive resource consumption. To address these limitations, we propose Token Compression Pathology LLaVA (TCP-LLaVA), the first MLLM architecture to perform WSI VQA via token compression. TCP-LLaVA introduces a set of trainable compression tokens that aggregate visual and textual information through a modality compression module, inspired by the [CLS] token mechanism in BERT. Only the compressed tokens are forwarded to the LLM for answer generation, significantly reducing input length and computational cost. Experiments on ten TCGA tumor subtypes show that TCP-LLaVA outperforms existing MLLM baselines in VQA accuracy while reducing training resource consumption by a substantial margin.", "date": "2025-07-19", "categories": []}
{"id": "2510.01287", "title": "Evaluating New AI Cell Foundation Models on Challenging Kidney Pathology Cases Unaddressed by Previous Foundation Models", "abstract": "Accurate cell nuclei segmentation is critical for downstream tasks in kidney pathology and remains a major challenge due to the morphological diversity and imaging variability of renal tissues. While our prior work has evaluated early-generation AI cell foundation models in this domain, the effectiveness of recent cell foundation models remains unclear. In this study, we benchmark advanced AI cell foundation models (2025), including CellViT++ variants and Cellpose-SAM, against three widely used cell foundation models developed prior to 2024, using a diverse large-scale set of kidney image patches within a human-in-the-loop rating framework. We further performed fusion-based ensemble evaluation and model agreement analysis to assess the segmentation capabilities of the different models. Our results show that CellViT++ [Virchow] yields the highest standalone performance with 40.3% of predictions rated as \"Good\" on a curated set of 2,091 challenging samples, outperforming all prior models. In addition, our fused model achieves 62.2% \"Good\" predictions and only 0.4% \"Bad\", substantially reducing segmentation errors. Notably, the fusion model (2025) successfully resolved the majority of challenging cases that remained unaddressed in our previous study. These findings demonstrate the potential of AI cell foundation model development in renal pathology and provide a curated dataset of challenging samples to support future kidney-specific model refinement.", "date": "2025-10-01", "categories": []}
{"id": "2506.18484", "title": "GANs vs. Diffusion Models for virtual staining with the HER2match dataset", "abstract": "Virtual staining is a promising technique that uses deep generative models to recreate histological stains, providing a faster and more cost-effective alternative to traditional tissue chemical staining. Specifically for H&E-HER2 staining transfer, despite a rising trend in publications, the lack of sufficient public datasets has hindered progress in the topic. Additionally, it is currently unclear which model frameworks perform best for this particular task. In this paper, we introduce the HER2match dataset, the first publicly available dataset with the same breast cancer tissue sections stained with both H&E and HER2. Furthermore, we compare the performance of several Generative Adversarial Networks (GANs) and Diffusion Models (DMs), and implement a novel Brownian Bridge Diffusion Model for H&E-HER2 translation. Our findings indicate that, overall, GANs perform better than DMs, with only the BBDM achieving comparable results. Furthermore, we emphasize the importance of data alignment, as all models trained on HER2match produced vastly improved visuals compared to the widely used consecutive-slide BCI dataset. This research provides a new high-quality dataset ([available upon publication acceptance]), improving both model training and evaluation. In addition, our comparison of frameworks offers valuable guidance for researchers working on the topic.", "date": "2025-06-23", "categories": []}
{"id": "2509.06105", "title": "PathoHR: Hierarchical Reasoning for Vision-Language Models in Pathology", "abstract": "Accurate analysis of pathological images is essential for automated tumor diagnosis but remains challenging due to high structural similarity and subtle morphological variations in tissue images. Current vision-language (VL) models often struggle to capture the complex reasoning required for interpreting structured pathological reports. To address these limitations, we propose PathoHR-Bench, a novel benchmark designed to evaluate VL models' abilities in hierarchical semantic understanding and compositional reasoning within the pathology domain. Results of this benchmark reveal that existing VL models fail to effectively model intricate cross-modal relationships, hence limiting their applicability in clinical setting. To overcome this, we further introduce a pathology-specific VL training scheme that generates enhanced and perturbed samples for multimodal contrastive learning. Experimental evaluations demonstrate that our approach achieves state-of-the-art performance on PathoHR-Bench and six additional pathology datasets, highlighting its effectiveness in fine-grained pathology representation.", "date": "2025-09-07", "categories": []}
{"id": "2509.25552", "title": "Evaluating Foundation Models with Pathological Concept Learning for Kidney Cancer", "abstract": "To evaluate the translational capabilities of foundation models, we develop a pathological concept learning approach focused on kidney cancer. By leveraging TNM staging guidelines and pathology reports, we build comprehensive pathological concepts for kidney cancer. Then, we extract deep features from whole slide images using foundation models, construct pathological graphs to capture spatial correlations, and trained graph neural networks to identify these concepts. Finally, we demonstrate the effectiveness of this approach in kidney cancer survival analysis, highlighting its explainability and fairness in identifying low- and high-risk patients. The source code has been released by https://github.com/shangqigao/RadioPath.", "date": "2025-09-29", "categories": []}
{"id": "2509.25863", "title": "MAPLE: Multi-scale Attribute-enhanced Prompt Learning for Few-shot Whole Slide Image Classification", "abstract": "Prompt learning has emerged as a promising paradigm for adapting pre-trained vision-language models (VLMs) to few-shot whole slide image (WSI) classification by aligning visual features with textual representations, thereby reducing annotation cost and enhancing model generalization. Nevertheless, existing methods typically rely on slide-level prompts and fail to capture the subtype-specific phenotypic variations of histological entities (\\emph{e.g.,} nuclei, glands) that are critical for cancer diagnosis. To address this gap, we propose Multi-scale Attribute-enhanced Prompt Learning (\\textbf{MAPLE}), a hierarchical framework for few-shot WSI classification that jointly integrates multi-scale visual semantics and performs prediction at both the entity and slide levels. Specifically, we first leverage large language models (LLMs) to generate entity-level prompts that can help identify multi-scale histological entities and their phenotypic attributes, as well as slide-level prompts to capture global visual descriptions. Then, an entity-guided cross-attention module is proposed to generate entity-level features, followed by aligning with their corresponding subtype-specific attributes for fine-grained entity-level prediction. To enrich entity representations, we further develop a cross-scale entity graph learning module that can update these representations by capturing their semantic correlations within and across scales. The refined representations are then aggregated into a slide-level representation and aligned with the corresponding prompts for slide-level prediction. Finally, we combine both entity-level and slide-level outputs to produce the final prediction results. Results on three cancer cohorts confirm the effectiveness of our approach in addressing few-shot pathology diagnosis tasks.", "date": "2025-09-30", "categories": []}
{"id": "2509.00442", "title": "SemaMIL: Semantic-Aware Multiple Instance Learning with Retrieval-Guided State Space Modeling for Whole Slide Images", "abstract": "Multiple instance learning (MIL) has become the leading approach for extracting discriminative features from whole slide images (WSIs) in computational pathology. Attention-based MIL methods can identify key patches but tend to overlook contextual relationships. Transformer models are able to model interactions but require quadratic computational cost and are prone to overfitting. State space models (SSMs) offer linear complexity, yet shuffling patch order disrupts histological meaning and reduces interpretability. In this work, we introduce SemaMIL, which integrates Semantic Reordering (SR), an adaptive method that clusters and arranges semantically similar patches in sequence through a reversible permutation, with a Semantic-guided Retrieval State Space Module (SRSM) that chooses a representative subset of queries to adjust state space parameters for improved global modeling. Evaluation on four WSI subtype datasets shows that, compared to strong baselines, SemaMIL achieves state-of-the-art accuracy with fewer FLOPs and parameters.", "date": "2025-08-30", "categories": []}
{"id": "2509.02627", "title": "A Two-Stage Strategy for Mitosis Detection Using Improved YOLO11x Proposals and ConvNeXt Classification", "abstract": "MIDOG 2025 Track 1 requires mitosis detection in whole-slideimages (WSIs) containing non-tumor, inflamed, and necrotic re-gions. Due to the complicated and heterogeneous context, aswell as possible artifacts, there are often false positives and falsenegatives, thus degrading the detection F1-score. To addressthis problem, we propose a two-stage framework. Firstly, an im-proved YOLO11x, integrated with EMA attention and LSConv,is employed to generate mitosis candidates. We use a low confi-dence threshold to generate as many proposals as possible, en-suring the detection recall. Then, a ConvNeXt-Tiny classifieris employed to filter out the false positives, ensuring the detec-tion precision. Consequently, the proposed two-stage frame-work can generate a high detection F1-score. Evaluated on afused dataset comprising MIDOG++, MITOS_WSI_CCMCT,and MITOS_WSI_CMC, our framework achieves an F1-scoreof 0.882, which is 0.035 higher than the single-stage YOLO11xbaseline. This performance gain is produced by a significantprecision improvement, from 0.762 to 0.839, and a comparablerecall. On the MIDOG 2025 Track 1 preliminary test set, thealgorithm scores an F1 score of 0.7587. The code is available athttps://github.com/xxiao0304/MIDOG-2025-Track-1-of-SZTU.", "date": "2025-09-01", "categories": []}
{"id": "2502.02707", "title": "LadderMIL: Multiple Instance Learning with Coarse-to-Fine Self-Distillation", "abstract": "Multiple Instance Learning (MIL) for whole slide image (WSI) analysis in computational pathology often neglects instance-level learning as supervision is typically provided only at the bag level, hindering the integrated consideration of instance and bag-level information during the analysis. In this work, we present LadderMIL, a framework designed to improve MIL through two perspectives: (1) employing instance-level supervision and (2) learning inter-instance contextual information at bag level. Firstly, we propose a novel Coarse-to-Fine Self-Distillation (CFSD) paradigm that probes and distils a network trained with bag-level information to adaptively obtain instance-level labels which could effectively provide the instance-level supervision for the same network in a self-improving way. Secondly, to capture inter-instance contextual information in WSI, we propose a Contextual Encoding Generator (CEG), which encodes the contextual appearance of instances within a bag. We also theoretically and empirically prove the instance-level learnability of CFSD. Our LadderMIL is evaluated on multiple clinically relevant benchmarking tasks including breast cancer receptor status classification, multi-class subtype classification, tumour classification, and prognosis prediction. Average improvements of 8.1%, 11% and 2.4% in AUC, F1-score, and C-index, respectively, are demonstrated across the five benchmarks, compared to the best baseline.", "date": "2025-02-04", "categories": []}
{"id": "2509.21239", "title": "SlideMamba: Entropy-Based Adaptive Fusion of GNN and Mamba for Enhanced Representation Learning in Digital Pathology", "abstract": "Advances in computational pathology increasingly rely on extracting meaningful representations from Whole Slide Images (WSIs) to support various clinical and biological tasks. In this study, we propose a generalizable deep learning framework that integrates the Mamba architecture with Graph Neural Networks (GNNs) for enhanced WSI analysis. Our method is designed to capture both local spatial relationships and long-range contextual dependencies, offering a flexible architecture for digital pathology analysis. Mamba modules excels in capturing long-range global dependencies, while GNNs emphasize fine-grained short-range spatial interactions. To effectively combine these complementary signals, we introduce an adaptive fusion strategy that uses an entropy-based confidence weighting mechanism. This approach dynamically balances contributions from both branches by assigning higher weight to the branch with more confident (lower-entropy) predictions, depending on the contextual importance of local versus global information for different downstream tasks. We demonstrate the utility of our approach on a representative task: predicting gene fusion and mutation status from WSIs. Our framework, SlideMamba, achieves an area under the precision recall curve (PRAUC) of 0.751 \\pm 0.05, outperforming MIL (0.491 \\pm 0.042), Trans-MIL (0.39 \\pm 0.017), Mamba-only (0.664 \\pm 0.063), GNN-only (0.748 \\pm 0.091), and a prior similar work GAT-Mamba (0.703 \\pm 0.075). SlideMamba also achieves competitive results across ROC AUC (0.738 \\pm 0.055), sensitivity (0.662 \\pm 0.083), and specificity (0.725 \\pm 0.094). These results highlight the strength of the integrated architecture, enhanced by the proposed entropy-based adaptive fusion strategy, and suggest promising potential for application of spatially-resolved predictive modeling tasks in computational pathology.", "date": "2025-09-25", "categories": []}
{"id": "2509.19896", "title": "Efficient Cell Painting Image Representation Learning via Cross-Well Aligned Masked Siamese Network", "abstract": "Computational models that predict cellular phenotypic responses to chemical and genetic perturbations can accelerate drug discovery by prioritizing therapeutic hypotheses and reducing costly wet-lab iteration. However, extracting biologically meaningful and batch-robust cell painting representations remains challenging. Conventional self-supervised and contrastive learning approaches often require a large-scale model and/or a huge amount of carefully curated data, still struggling with batch effects. We present Cross-Well Aligned Masked Siamese Network (CWA-MSN), a novel representation learning framework that aligns embeddings of cells subjected to the same perturbation across different wells, enforcing semantic consistency despite batch effects. Integrated into a masked siamese architecture, this alignment yields features that capture fine-grained morphology while remaining data- and parameter-efficient. For instance, in a gene-gene relationship retrieval benchmark, CWA-MSN outperforms the state-of-the-art publicly available self-supervised (OpenPhenom) and contrastive learning (CellCLIP) methods, improving the benchmark scores by +29\\% and +9\\%, respectively, while training on substantially fewer data (e.g., 0.2M images for CWA-MSN vs. 2.2M images for OpenPhenom) or smaller model size (e.g., 22M parameters for CWA-MSN vs. 1.48B parameters for CellCLIP). Extensive experiments demonstrate that CWA-MSN is a simple and effective way to learn cell image representation, enabling efficient phenotype modeling even under limited data and parameter budgets.", "date": "2025-09-24", "categories": []}
{"id": "2508.12381", "title": "IPGPhormer: Interpretable Pathology Graph-Transformer for Survival Analysis", "abstract": "Pathological images play an essential role in cancer prognosis, while survival analysis, which integrates computational techniques, can predict critical clinical events such as patient mortality or disease recurrence from whole-slide images (WSIs). Recent advancements in multiple instance learning have significantly improved the efficiency of survival analysis. However, existing methods often struggle to balance the modeling of long-range spatial relationships with local contextual dependencies and typically lack inherent interpretability, limiting their clinical utility. To address these challenges, we propose the Interpretable Pathology Graph-Transformer (IPGPhormer), a novel framework that captures the characteristics of the tumor microenvironment and models their spatial dependencies across the tissue. IPGPhormer uniquely provides interpretability at both tissue and cellular levels without requiring post-hoc manual annotations, enabling detailed analyses of individual WSIs and cross-cohort assessments. Comprehensive evaluations on four public benchmark datasets demonstrate that IPGPhormer outperforms state-of-the-art methods in both predictive accuracy and interpretability. In summary, our method, IPGPhormer, offers a promising tool for cancer prognosis assessment, paving the way for more reliable and interpretable decision-support systems in pathology. The code is publicly available at https://anonymous.4open.science/r/IPGPhormer-6EEB.", "date": "2025-08-17", "categories": []}
{"id": "2509.02957", "title": "Ensemble YOLO Framework for Multi-Domain Mitotic Figure Detection in Histopathology Images", "abstract": "The reliable identification of mitotic figures in whole-slide histopathological images remains difficult, owing to their low prevalence, substantial morphological heterogeneity, and the inconsistencies introduced by tissue processing and staining procedures. The MIDOG competition series provides standardized benchmarks for evaluating detection approaches across diverse domains, thus motivating the development of generalizable deep learning models. In this work, we investigate the performance of two modern one-stage detectors, YOLOv5 and YOLOv8, trained on MIDOG++, CMC, and CCMCT datasets. To enhance robustness, training incorporated stain-invariant color perturbations and texture-preserving augmentations. Ininternal validation, YOLOv5 achieved higher precision (84.3%), while YOLOv8 offered improved recall (82.6%), reflecting architectural trade-offs between anchor-based and anchor-free detections. To capitalize on their complementary strengths, weemployed an ensemble of the two models, which improved sensitivity (85.3%) while maintaining competitive precision, yielding the best F1 score of 83.1%. On the preliminary MIDOG 2025 test leaderboard, our ensemble ranked 5th with an F1 score of 79.2%, precision of 73.6%, and recall of 85.8%, confirming that the proposed strategy generalizes effectively across unseen test data. These findings highlight the effectiveness of combining anchor-based and anchor-free object detectors to advance automated mitosis detection in digital pathology.", "date": "2025-09-03", "categories": []}
{"id": "2505.06793", "title": "HistDiST: Histopathological Diffusion-based Stain Transfer", "abstract": "Hematoxylin and Eosin (H&E) staining is the cornerstone of histopathology but lacks molecular specificity. While Immunohistochemistry (IHC) provides molecular insights, it is costly and complex, motivating H&E-to-IHC translation as a cost-effective alternative. Existing translation methods are mainly GAN-based, often struggling with training instability and limited structural fidelity, while diffusion-based approaches remain underexplored. We propose HistDiST, a Latent Diffusion Model (LDM) based framework for high-fidelity H&E-to-IHC translation. HistDiST introduces a dual-conditioning strategy, utilizing Phikon-extracted morphological embeddings alongside VAE-encoded H&E representations to ensure pathology-relevant context and structural consistency. To overcome brightness biases, we incorporate a rescaled noise schedule, v-prediction, and trailing timesteps, enforcing a zero-SNR condition at the final timestep. During inference, DDIM inversion preserves the morphological structure, while an eta-cosine noise schedule introduces controlled stochasticity, balancing structural consistency and molecular fidelity. Moreover, we propose Molecular Retrieval Accuracy (MRA), a novel pathology-aware metric leveraging GigaPath embeddings to assess molecular relevance. Extensive evaluations on MIST and BCI datasets demonstrate that HistDiST significantly outperforms existing methods, achieving a 28% improvement in MRA on the H&E-to-Ki67 translation task, highlighting its effectiveness in capturing true IHC semantics.", "date": "2025-05-11", "categories": []}
{"id": "2509.15608", "title": "Enhancing WSI-Based Survival Analysis with Report-Auxiliary Self-Distillation", "abstract": "Survival analysis based on Whole Slide Images (WSIs) is crucial for evaluating cancer prognosis, as they offer detailed microscopic information essential for predicting patient outcomes. However, traditional WSI-based survival analysis usually faces noisy features and limited data accessibility, hindering their ability to capture critical prognostic features effectively. Although pathology reports provide rich patient-specific information that could assist analysis, their potential to enhance WSI-based survival analysis remains largely unexplored. To this end, this paper proposes a novel Report-auxiliary self-distillation (Rasa) framework for WSI-based survival analysis. First, advanced large language models (LLMs) are utilized to extract fine-grained, WSI-relevant textual descriptions from original noisy pathology reports via a carefully designed task prompt. Next, a self-distillation-based pipeline is designed to filter out irrelevant or redundant WSI features for the student model under the guidance of the teacher model's textual knowledge. Finally, a risk-aware mix-up strategy is incorporated during the training of the student model to enhance both the quantity and diversity of the training data. Extensive experiments carried out on our collected data (CRC) and public data (TCGA-BRCA) demonstrate the superior effectiveness of Rasa against state-of-the-art methods. Our code is available at https://github.com/zhengwang9/Rasa.", "date": "2025-09-19", "categories": []}
{"id": "2507.20907", "title": "SCORPION: Addressing Scanner-Induced Variability in Histopathology", "abstract": "Ensuring reliable model performance across diverse domains is a critical challenge in computational pathology. A particular source of variability in Whole-Slide Images is introduced by differences in digital scanners, thus calling for better scanner generalization. This is critical for the real-world adoption of computational pathology, where the scanning devices may differ per institution or hospital, and the model should not be dependent on scanner-induced details, which can ultimately affect the patient's diagnosis and treatment planning. However, past efforts have primarily focused on standard domain generalization settings, evaluating on unseen scanners during training, without directly evaluating consistency across scanners for the same tissue. To overcome this limitation, we introduce SCORPION, a new dataset explicitly designed to evaluate model reliability under scanner variability. SCORPION includes 480 tissue samples, each scanned with 5 scanners, yielding 2,400 spatially aligned patches. This scanner-paired design allows for the isolation of scanner-induced variability, enabling a rigorous evaluation of model consistency while controlling for differences in tissue composition. Furthermore, we propose SimCons, a flexible framework that combines augmentation-based domain generalization techniques with a consistency loss to explicitly address scanner generalization. We empirically show that SimCons improves model consistency on varying scanners without compromising task-specific performance. By releasing the SCORPION dataset and proposing SimCons, we provide the research community with a crucial resource for evaluating and improving model consistency across diverse scanners, setting a new standard for reliability testing.", "date": "2025-07-28", "categories": []}
{"id": "2509.02591", "title": "Ensemble of Pathology Foundation Models for MIDOG 2025 Track 2: Atypical Mitosis Classification", "abstract": "Mitotic figures are classified into typical and atypical variants, with atypical counts correlating strongly with tumor aggressiveness. Accurate differentiation is therefore essential for patient prognostication and resource allocation, yet remains challenging even for expert pathologists. Here, we leveraged Pathology Foundation Models (PFMs) pre-trained on large histopathology datasets and applied parameter-efficient fine-tuning via low-rank adaptation. In addition, we incorporated ConvNeXt V2, a state-of-the-art convolutional neural network architecture, to complement PFMs. During training, we employed a fisheye transform to emphasize mitoses and Fourier Domain Adaptation using ImageNet target images. Finally, we ensembled multiple PFMs to integrate complementary morphological insights, achieving competitive balanced accuracy on the Preliminary Evaluation Phase dataset.", "date": "2025-08-29", "categories": []}
{"id": "2509.11526", "title": "Multiple Instance Learning Framework with Masked Hard Instance Mining for Gigapixel Histopathology Image Analysis", "abstract": "Digitizing pathological images into gigapixel Whole Slide Images (WSIs) has opened new avenues for Computational Pathology (CPath). As positive tissue comprises only a small fraction of gigapixel WSIs, existing Multiple Instance Learning (MIL) methods typically focus on identifying salient instances via attention mechanisms. However, this leads to a bias towards easy-to-classify instances while neglecting challenging ones. Recent studies have shown that hard examples are crucial for accurately modeling discriminative boundaries. Applying such an idea at the instance level, we elaborate a novel MIL framework with masked hard instance mining (MHIM-MIL), which utilizes a Siamese structure with a consistency constraint to explore the hard instances. Using a class-aware instance probability, MHIM-MIL employs a momentum teacher to mask salient instances and implicitly mine hard instances for training the student model. To obtain diverse, non-redundant hard instances, we adopt large-scale random masking while utilizing a global recycle network to mitigate the risk of losing key features. Furthermore, the student updates the teacher using an exponential moving average, which identifies new hard instances for subsequent training iterations and stabilizes optimization. Experimental results on cancer diagnosis, subtyping, survival analysis tasks, and 12 benchmarks demonstrate that MHIM-MIL outperforms the latest methods in both performance and efficiency. The code is available at: https://github.com/DearCaat/MHIM-MIL.", "date": "2025-09-15", "categories": []}
{"id": "2508.05084", "title": "AdaFusion: Prompt-Guided Inference with Adaptive Fusion of Pathology Foundation Models", "abstract": "Pathology foundation models (PFMs) have demonstrated strong representational capabilities through self-supervised pre-training on large-scale, unannotated histopathology image datasets. However, their diverse yet opaque pretraining contexts, shaped by both data-related and structural/training factors, introduce latent biases that hinder generalisability and transparency in downstream applications. In this paper, we propose AdaFusion, a novel prompt-guided inference framework that, to our knowledge, is among the very first to dynamically integrate complementary knowledge from multiple PFMs. Our method compresses and aligns tile-level features from diverse models and employs a lightweight attention mechanism to adaptively fuse them based on tissue phenotype context. We evaluate AdaFusion on three real-world benchmarks spanning treatment response prediction, tumour grading, and spatial gene expression inference. Our approach consistently surpasses individual PFMs across both classification and regression tasks, while offering interpretable insights into each model's biosemantic specialisation. These results highlight AdaFusion's ability to bridge heterogeneous PFMs, achieving both enhanced performance and interpretability of model-specific inductive biases.", "date": "2025-08-07", "categories": []}
{"id": "2411.18795", "title": "GloFinder: AI-empowered QuPath Plugin for WSI-level Glomerular Detection, Visualization, and Curation", "abstract": "Artificial intelligence (AI) has demonstrated significant success in automating the detection of glomeruli, the key functional units of the kidney, from whole slide images (WSIs) in kidney pathology. However, existing open-source tools are often distributed as source code or Docker containers, requiring advanced programming skills that hinder accessibility for non-programmers, such as clinicians. Additionally, current models are typically trained on a single dataset and lack flexibility in adjusting confidence levels for predictions. To overcome these challenges, we introduce GloFinder, a QuPath plugin designed for single-click automated glomeruli detection across entire WSIs with online editing through the graphical user interface (GUI). GloFinder employs CircleNet, an anchor-free detection framework utilizing circle representations for precise object localization, with models trained on approximately 160,000 manually annotated glomeruli. To further enhance accuracy, the plugin incorporates Weighted Circle Fusion (WCF), an ensemble method that combines confidence scores from multiple CircleNet models to produce refined predictions, achieving superior performance in glomerular detection. GloFinder enables direct visualization and editing of results in QuPath, facilitating seamless interaction for clinicians and providing a powerful tool for nephropathology research and clinical practice. Code and the QuPath plugin are available at https://github.com/hrlblab/GloFinder", "date": "2024-11-27", "categories": []}
{"id": "2505.10993", "title": "Content Generation Models in Computational Pathology: A Comprehensive Survey on Methods, Applications, and Challenges", "abstract": "Content generation modeling has emerged as a promising direction in computational pathology, offering capabilities such as data-efficient learning, synthetic data augmentation, and task-oriented generation across diverse diagnostic tasks. This review provides a comprehensive synthesis of recent progress in the field, organized into four key domains: image generation, text generation, molecular profile-morphology generation, and other specialized generation applications. By analyzing over 150 representative studies, we trace the evolution of content generation architectures -- from early generative adversarial networks to recent advances in diffusion models and generative vision-language models. We further examine the datasets and evaluation protocols commonly used in this domain and highlight ongoing limitations, including challenges in generating high-fidelity whole slide images, clinical interpretability, and concerns related to the ethical and legal implications of synthetic data. The review concludes with a discussion of open challenges and prospective research directions, with an emphasis on developing integrated and clinically deployable generation systems. This work aims to provide a foundational reference for researchers and practitioners developing content generation models in computational pathology.", "date": "2025-05-16", "categories": []}
{"id": "2508.02431", "title": "Identifying actionable driver mutations in lung cancer using an efficient Asymmetric Transformer Decoder", "abstract": "Identifying actionable driver mutations in non-small cell lung cancer (NSCLC) can impact treatment decisions and significantly improve patient outcomes. Despite guideline recommendations, broader adoption of genetic testing remains challenging due to limited availability and lengthy turnaround times. Machine Learning (ML) methods for Computational Pathology (CPath) offer a potential solution; however, research often focuses on only one or two common mutations, limiting the clinical value of these tools and the pool of patients who can benefit from them. This study evaluates various Multiple Instance Learning (MIL) techniques to detect six key actionable NSCLC driver mutations: ALK, BRAF, EGFR, ERBB2, KRAS, and MET ex14. Additionally, we introduce an Asymmetric Transformer Decoder model that employs queries and key-values of varying dimensions to maintain a low query dimensionality. This approach efficiently extracts information from patch embeddings and minimizes overfitting risks, proving highly adaptable to the MIL setting. Moreover, we present a method to directly utilize tissue type in the model, addressing a typical MIL limitation where either all regions or only some specific regions are analyzed, neglecting biological relevance. Our method outperforms top MIL models by an average of 3%, and over 4% when predicting rare mutations such as ERBB2 and BRAF, moving ML-based tests closer to being practical alternatives to standard genetic testing.", "date": "2025-08-04", "categories": []}
{"id": "2509.02586", "title": "MitoDetect++: A Domain-Robust Pipeline for Mitosis Detection and Atypical Subtyping", "abstract": "Automated detection and classification of mitotic figures especially distinguishing atypical from normal remain critical challenges in computational pathology. We present MitoDetect++, a unified deep learning pipeline designed for the MIDOG 2025 challenge, addressing both mitosis detection and atypical mitosis classification. For detection (Track 1), we employ a U-Net-based encoder-decoder architecture with EfficientNetV2-L as the backbone, enhanced with attention modules, and trained via combined segmentation losses. For classification (Track 2), we leverage the Virchow2 vision transformer, fine-tuned efficiently using Low-Rank Adaptation (LoRA) to minimize resource consumption. To improve generalization and mitigate domain shifts, we integrate strong augmentations, focal loss, and group-aware stratified 5-fold cross-validation. At inference, we deploy test-time augmentation (TTA) to boost robustness. Our method achieves a balanced accuracy of 0.892 across validation domains, highlighting its clinical applicability and scalability across tasks.", "date": "2025-08-28", "categories": []}
{"id": "2509.02640", "title": "Adaptive Learning Strategies for Mitotic Figure Classification in MIDOG2025 Challenge", "abstract": "Atypical mitotic figures (AMFs) are clinically relevant indicators of abnormal cell division, yet their reliable detection remains challenging due to morphological ambiguity and scanner variability. In this work, we investigated three variants of adapting the pathology foundation model UNI2 for the MIDOG2025 Track 2 challenge: (1) LoRA + UNI2, (2) VPT + UNI2 + Vahadane Normalizer, and (3) VPT + UNI2 + GRL + Stain TTA. We observed that the integration of Visual Prompt Tuning (VPT) with stain normalization techniques contributed to improved generalization. The best robustness was achieved by further incorporating test-time augmentation (TTA) with Vahadane and Macenko stain normalization. Our final submission achieved a balanced accuracy of 0.8837 and an ROC-AUC of 0.9513 on the preliminary leaderboard, ranking within the top 10 teams. These results suggest that prompt-based adaptation combined with stain-normalization TTA offers a promising strategy for atypical mitosis classification under diverse imaging conditions.", "date": "2025-09-01", "categories": []}
{"id": "2509.02585", "title": "Pan-Cancer mitotic figures detection and domain generalization: MIDOG 2025 Challenge", "abstract": "This report details our submission to the Mitotic Domain Generalization (MIDOG) 2025 challenge, which addresses the critical task of mitotic figure detection in histopathology for cancer prognostication. Following the \"Bitter Lesson\"\\cite{sutton2019bitterlesson} principle that emphasizes data scale over algorithmic novelty, we have publicly released two new datasets to bolster training data for both conventional \\cite{Shen2024framework} and atypical mitoses \\cite{shen_2025_16780587}. Besides, we implement up-to-date training methodologies for both track and reach a Track-1 F1-Score of 0.8407 on our test set, as well as a Track-2 balanced accuracy of 0.9107 for atypical mitotic cell classification.", "date": "2025-08-28", "categories": []}
{"id": "2509.02588", "title": "Sequential Hard Mining: a data-centric approach for Mitosis Detection", "abstract": "With a continuously growing availability of annotated datasets of mitotic figures in histology images, finding the best way to optimally use with this unprecedented amount of data to optimally train deep learning models has become a new challenge. Here, we build upon previously proposed approaches with a focus on efficient sampling of training data inspired by boosting techniques and present our candidate solutions for the two tracks of the MIDOG 2025 challenge.", "date": "2025-08-28", "categories": []}
{"id": "2509.02589", "title": "Normal and Atypical Mitosis Image Classifier using Efficient Vision Transformer", "abstract": "We tackle atypical versus normal mitosis classification in the MIDOG 2025 challenge using EfficientViT-L2, a hybrid CNN--ViT architecture optimized for accuracy and efficiency. A unified dataset of 13,938 nuclei from seven cancer types (MIDOG++ and AMi-Br) was used, with atypical mitoses comprising ~15. To assess domain generalization, we applied leave-one-cancer-type-out cross-validation with 5-fold ensembles, using stain-deconvolution for image augmentation. For challenge submissions, we trained an ensemble with the same 5-fold split but on all cancer types. In the preliminary evaluation phase, this model achieved balanced accuracy of 0.859, ROC AUC of 0.942, and raw accuracy of 0.85, demonstrating competitive and well-balanced performance across metrics.", "date": "2025-08-28", "categories": []}
{"id": "2509.02595", "title": "ConvNeXt with Histopathology-Specific Augmentations for Mitotic Figure Classification", "abstract": "Accurate mitotic figure classification is crucial in computational pathology, as mitotic activity informs cancer grading and patient prognosis. Distinguishing atypical mitotic figures (AMFs), which indicate higher tumor aggressiveness, from normal mitotic figures (NMFs) remains challenging due to subtle morphological differences and high intra-class variability. This task is further complicated by domain shifts, including variations in organ, tissue type, and scanner, as well as limited annotations and severe class imbalance. To address these challenges in Track 2 of the MIDOG 2025 Challenge, we propose a solution based on the lightweight ConvNeXt architecture, trained on all available datasets (AMi-Br, AtNorM-Br, AtNorM-MD, and OMG-Octo) to maximize domain coverage. Robustness is enhanced through a histopathology-specific augmentation pipeline, including elastic and stain-specific transformations, and balanced sampling to mitigate class imbalance. A grouped 5-fold cross-validation strategy ensures reliable evaluation. On the preliminary leaderboard, our model achieved a balanced accuracy of 0.8961, ranking among the top entries. These results highlight that broad domain exposure combined with targeted augmentation strategies is key to building accurate and generalizable mitotic figure classifiers.", "date": "2025-08-29", "categories": []}
{"id": "2509.02597", "title": "Solutions for Mitotic Figure Detection and Atypical Classification in MIDOG 2025", "abstract": "Deep learning has driven significant advances in mitotic figure analysis within computational pathology. In this paper, we present our approach to the Mitosis Domain Generalization (MIDOG) 2025 Challenge, which consists of two distinct tasks, i.e., mitotic figure detection and atypical mitosis classification. For the mitotic figure detection task, we propose a two-stage detection-classification framework that first localizes candidate mitotic figures and subsequently refines the predictions using a dedicated classification module. For the atypical mitosis classification task, we employ an ensemble strategy that integrates predictions from multiple state-of-the-art deep learning architectures to improve robustness and accuracy. Extensive experiments demonstrate the effectiveness of our proposed methods across both tasks.", "date": "2025-08-29", "categories": []}
{"id": "2509.02599", "title": "RF-DETR for Robust Mitotic Figure Detection: A MIDOG 2025 Track 1 Approach", "abstract": "Mitotic figure detection in histopathology images remains challenging due to significant domain shifts across different scanners, staining protocols, and tissue types. This paper presents our approach for the MIDOG 2025 challenge Track 1, focusing on robust mitotic figure detection across diverse histological contexts. While we initially planned a two-stage approach combining high-recall detection with subsequent classification refinement, time constraints led us to focus on optimizing a single-stage detection pipeline. We employed RF-DETR (Roboflow Detection Transformer) with hard negative mining, trained on MIDOG++ dataset. On the preliminary test set, our method achieved an F1 score of 0.789 with a recall of 0.839 and precision of 0.746, demonstrating effective generalization across unseen domains. The proposed solution offers insights into the importance of training data balance and hard negative mining for addressing domain shift challenges in mitotic figure detection.", "date": "2025-08-29", "categories": []}
{"id": "2509.02612", "title": "Is Synthetic Image Augmentation Useful for Imbalanced Classification Problems? Case-Study on the MIDOG2025 Atypical Cell Detection Competition", "abstract": "The MIDOG 2025 challenge extends prior work on mitotic figure detection by introducing a new Track 2 on atypical mitosis classification. This task aims to distinguish normal from atypical mitotic figures in histopathology images, a clinically relevant but highly imbalanced and cross-domain problem. We investigated two complementary backbones: (i) ConvNeXt-Small, pretrained on ImageNet, and (ii) a histopathology-specific ViT from Lunit trained via self-supervision. To address the strong prevalence imbalance (9408 normal vs. 1741 atypical), we synthesized additional atypical examples to approximate class balance and compared models trained with real-only vs. real+synthetic data. Using five-fold cross-validation, both backbones reached strong performance (mean AUROC approximately 95 percent), with ConvNeXt achieving slightly higher peaks while Lunit exhibited greater fold-to-fold stability. Synthetic balancing, however, did not lead to consistent improvements. On the organizers' preliminary hidden test set, explicitly designed as an out-of-distribution debug subset, ConvNeXt attained the highest AUROC (95.4 percent), whereas Lunit remained competitive on balanced accuracy. These findings suggest that both ImageNet and domain-pretrained backbones are viable for atypical mitosis classification, with domain-pretraining conferring robustness and ImageNet pretraining reaching higher peaks, while naive synthetic balancing has limited benefit. Full hidden test set results will be reported upon challenge completion.", "date": "2025-08-30", "categories": []}
{"id": "2509.02630", "title": "Challenges and Lessons from MIDOG 2025: A Two-Stage Approach to Domain-Robust Mitotic Figure Detection", "abstract": "Mitotic figure detection remains a challenging task in computational pathology due to domain variability and morphological complexity. This paper describes our participation in the MIDOG 2025 challenge, focusing on robust mitotic figure detection across diverse tissue domains. We developed a two-stage pipeline combining Faster R-CNN for candidate detection with an ensemble of three classifiers (DenseNet-121, EfficientNet-v2, InceptionResNet-v2) for false positive reduction. Our best submission achieved F1-score 0.2237 (Recall: 0.9528, Precision: 0.1267) using a Faster R-CNN trained solely on MIDOG++ dataset. While our high recall demonstrates effective mitotic figure detection, the critically low precision (12.67%) reveals fundamental challenges in distinguishing true mitoses from morphologically similar imposters across diverse domains. Analysis of six submission variants showed that subsequent optimization attempts were counterproductive, highlighting the omplexity of domain generalization in histopathology. This work provides valuable insights into the practical challenges of developing robust mitotic figure detection algorithms and emphasizes the importance of effective false positive suppression strategies.", "date": "2025-09-01", "categories": []}
{"id": "2509.02637", "title": "A Single Detect Focused YOLO Framework for Robust Mitotic Figure Detection", "abstract": "Mitotic figure detection is a crucial task in computational pathology, as mitotic activity serves as a strong prognostic marker for tumor aggressiveness. However, domain variability that arises from differences in scanners, tissue types, and staining protocols poses a major challenge to the robustness of automated detection methods. In this study, we introduce SDF-YOLO (Single Detect Focused YOLO), a lightweight yet domain-robust detection framework designed specifically for small, rare targets such as mitotic figures. The model builds on YOLOv11 with task-specific modifications, including a single detection head aligned with mitotic figure scale, coordinate attention to enhance positional sensitivity, and improved cross-channel feature mixing. Experiments were conducted on three datasets that span human and canine tumors: MIDOG ++, canine cutaneous mast cell tumor (CCMCT), and canine mammary carcinoma (CMC). When submitted to the preliminary test set for the MIDOG2025 challenge, SDF-YOLO achieved an average precision (AP) of 0.799, with a precision of 0.758, a recall of 0.775, an F1 score of 0.766, and an FROC-AUC of 5.793, demonstrating both competitive accuracy and computational efficiency. These results indicate that SDF-YOLO provides a reliable and efficient framework for robust mitotic figure detection across diverse domains.", "date": "2025-09-01", "categories": []}
{"id": "2509.03323", "title": "Heatmap Guided Query Transformers for Robust Astrocyte Detection across Immunostains and Resolutions", "abstract": "Astrocytes are critical glial cells whose altered morphology and density are hallmarks of many neurological disorders. However, their intricate branching and stain dependent variability make automated detection of histological images a highly challenging task. To address these challenges, we propose a hybrid CNN Transformer detector that combines local feature extraction with global contextual reasoning. A heatmap guided query mechanism generates spatially grounded anchors for small and faint astrocytes, while a lightweight Transformer module improves discrimination in dense clusters. Evaluated on ALDH1L1 and GFAP stained astrocyte datasets, the model consistently outperformed Faster R-CNN, YOLOv11 and DETR, achieving higher sensitivity with fewer false positives, as confirmed by FROC analysis. These results highlight the potential of hybrid CNN Transformer architectures for robust astrocyte detection and provide a foundation for advanced computational pathology tools.", "date": "2025-09-03", "categories": []}
{"id": "2509.03408", "title": "Scalable and Loosely-Coupled Multimodal Deep Learning for Breast Cancer Subtyping", "abstract": "Healthcare applications are inherently multimodal, benefiting greatly from the integration of diverse data sources. However, the modalities available in clinical settings can vary across different locations and patients. A key area that stands to gain from multimodal integration is breast cancer molecular subtyping, an important clinical task that can facilitate personalized treatment and improve patient prognosis. In this work, we propose a scalable and loosely-coupled multimodal framework that seamlessly integrates data from various modalities, including copy number variation (CNV), clinical records, and histopathology images, to enhance breast cancer subtyping. While our primary focus is on breast cancer, our framework is designed to easily accommodate additional modalities, offering the flexibility to scale up or down with minimal overhead without requiring re-training of existing modalities, making it applicable to other types of cancers as well. We introduce a dual-based representation for whole slide images (WSIs), combining traditional image-based and graph-based WSI representations. This novel dual approach results in significant performance improvements. Moreover, we present a new multimodal fusion strategy, demonstrating its ability to enhance performance across a range of multimodal conditions. Our comprehensive results show that integrating our dual-based WSI representation with CNV and clinical health records, along with our pipeline and fusion strategy, outperforms state-of-the-art methods in breast cancer subtyping.", "date": "2025-09-03", "categories": []}
{"id": "2508.21033", "title": "Mitosis detection in domain shift scenarios: a Mamba-based approach", "abstract": "Mitosis detection in histopathology images plays a key role in tumor assessment. Although machine learning algorithms could be exploited for aiding physicians in accurately performing such a task, these algorithms suffer from significative performance drop when evaluated on images coming from domains that are different from the training ones. In this work, we propose a Mamba-based approach for mitosis detection under domain shift, inspired by the promising performance demonstrated by Mamba in medical imaging segmentation tasks. Specifically, our approach exploits a VM-UNet architecture for carrying out the addressed task, as well as stain augmentation operations for further improving model robustness against domain shift. Our approach has been submitted to the track 1 of the MItosis DOmain Generalization (MIDOG) challenge. Preliminary experiments, conducted on the MIDOG++ dataset, show large room for improvement for the proposed method.", "date": "2025-08-28", "categories": []}
{"id": "2509.01071", "title": "A Unified Low-level Foundation Model for Enhancing Pathology Image Quality", "abstract": "Foundation models have revolutionized computational pathology by achieving remarkable success in high-level diagnostic tasks, yet the critical challenge of low-level image enhancement remains largely unaddressed. Real-world pathology images frequently suffer from degradations such as noise, blur, and low resolution due to slide preparation artifacts, staining variability, and imaging constraints, while the reliance on physical staining introduces significant costs, delays, and inconsistency. Although existing methods target individual problems like denoising or super-resolution, their task-specific designs lack the versatility to handle the diverse low-level vision challenges encountered in practice. To bridge this gap, we propose the first unified Low-level Pathology Foundation Model (LPFM), capable of enhancing image quality in restoration tasks, including super-resolution, deblurring, and denoising, as well as facilitating image translation tasks like virtual staining (H&E and special stains), all through a single adaptable architecture. Our approach introduces a contrastive pre-trained encoder that learns transferable, stain-invariant feature representations from 190 million unlabeled pathology images, enabling robust identification of degradation patterns. A unified conditional diffusion process dynamically adapts to specific tasks via textual prompts, ensuring precise control over output quality. Trained on a curated dataset of 87,810 whole slied images (WSIs) across 34 tissue types and 5 staining protocols, LPFM demonstrates statistically significant improvements (p<0.01) over state-of-the-art methods in most tasks (56/66), achieving Peak Signal-to-Noise Ratio (PSNR) gains of 10-15% for image restoration and Structural Similarity Index Measure (SSIM) improvements of 12-18% for virtual staining.", "date": "2025-09-01", "categories": []}
{"id": "2509.02440", "title": "Efficient Pyramidal Analysis of Gigapixel Images on a Decentralized Modest Computer Cluster", "abstract": "Analyzing gigapixel images is recognized as computationally demanding. In this paper, we introduce PyramidAI, a technique for analyzing gigapixel images with reduced computational cost. The proposed approach adopts a gradual analysis of the image, beginning with lower resolutions and progressively concentrating on regions of interest for detailed examination at higher resolutions. We investigated two strategies for tuning the accuracy-computation performance trade-off when implementing the adaptive resolution selection, validated against the Camelyon16 dataset of biomedical images. Our results demonstrate that PyramidAI substantially decreases the amount of processed data required for analysis by up to 2.65x, while preserving the accuracy in identifying relevant sections on a single computer. To ensure democratization of gigapixel image analysis, we evaluated the potential to use mainstream computers to perform the computation by exploiting the parallelism potential of the approach. Using a simulator, we estimated the best data distribution and load balancing algorithm according to the number of workers. The selected algorithms were implemented and highlighted the same conclusions in a real-world setting. Analysis time is reduced from more than an hour to a few minutes using 12 modest workers, offering a practical solution for efficient large-scale image analysis.", "date": "2025-09-02", "categories": []}
{"id": "2508.20745", "title": "Mix, Align, Distil: Reliable Cross-Domain Atypical Mitosis Classification", "abstract": "Atypical mitotic figures (AMFs) are important histopathological markers yet remain challenging to identify consistently, particularly under domain shift stemming from scanner, stain, and acquisition differences. We present a simple training-time recipe for domain-robust AMF classification in MIDOG 2025 Task 2. The approach (i) increases feature diversity via style perturbations inserted at early and mid backbone stages, (ii) aligns attention-refined features across sites using weak domain labels (Scanner, Origin, Species, Tumor) through an auxiliary alignment loss, and (iii) stabilizes predictions by distilling from an exponential moving average (EMA) teacher with temperature-scaled KL divergence. On the organizer-run preliminary leaderboard for atypical mitosis classification, our submission attains balanced accuracy of 0.8762, sensitivity of 0.8873, specificity of 0.8651, and ROC AUC of 0.9499. The method incurs negligible inference-time overhead, relies only on coarse domain metadata, and delivers strong, balanced performance, positioning it as a competitive submission for the MIDOG 2025 challenge.", "date": "2025-08-28", "categories": []}
{"id": "2508.19804", "title": "A bag of tricks for real-time Mitotic Figure detection", "abstract": "Mitotic figure (MF) detection in histopathology images is challenging due to large variations in slide scanners, staining protocols, tissue types, and the presence of artifacts. This paper presents a collection of training techniques - a bag of tricks - that enable robust, real-time MF detection across diverse domains. We build on the efficient RTMDet single stage object detector to achieve high inference speed suitable for clinical deployment. Our method addresses scanner variability and tumor heterogeneity via extensive multi-domain training data, balanced sampling, and careful augmentation. Additionally, we employ targeted, hard negative mining on necrotic and debris tissue to reduce false positives. In a grouped 5-fold cross-validation across multiple MF datasets, our model achieves an F1 score between 0.78 and 0.84. On the preliminary test set of the MItosis DOmain Generalization (MIDOG) 2025 challenge, our single-stage RTMDet-S based approach reaches an F1 of 0.81, outperforming larger models and demonstrating adaptability to new, unfamiliar domains. The proposed solution offers a practical trade-off between accuracy and speed, making it attractive for real-world clinical adoption.", "date": "2025-08-27", "categories": []}
{"id": "2508.19914", "title": "The Next Layer: Augmenting Foundation Models with Structure-Preserving and Attention-Guided Learning for Local Patches to Global Context Awareness in Computational Pathology", "abstract": "Foundation models have recently emerged as powerful feature extractors in computational pathology, yet they typically omit mechanisms for leveraging the global spatial structure of tissues and the local contextual relationships among diagnostically relevant regions - key elements for understanding the tumor microenvironment. Multiple instance learning (MIL) remains an essential next step following foundation model, designing a framework to aggregate patch-level features into slide-level predictions. We present EAGLE-Net, a structure-preserving, attention-guided MIL architecture designed to augment prediction and interpretability. EAGLE-Net integrates multi-scale absolute spatial encoding to capture global tissue architecture, a top-K neighborhood-aware loss to focus attention on local microenvironments, and background suppression loss to minimize false positives. We benchmarked EAGLE-Net on large pan-cancer datasets, including three cancer types for classification (10,260 slides) and seven cancer types for survival prediction (4,172 slides), using three distinct histology foundation backbones (REMEDIES, Uni-V1, Uni2-h). Across tasks, EAGLE-Net achieved up to 3% higher classification accuracy and the top concordance indices in 6 of 7 cancer types, producing smooth, biologically coherent attention maps that aligned with expert annotations and highlighted invasive fronts, necrosis, and immune infiltration. These results position EAGLE-Net as a generalizable, interpretable framework that complements foundation models, enabling improved biomarker discovery, prognostic modeling, and clinical decision support", "date": "2025-08-27", "categories": []}
{"id": "2407.21604", "title": "MicroMIL: Graph-Based Multiple Instance Learning for Context-Aware Diagnosis with Microscopic Images", "abstract": "Cancer diagnosis has greatly benefited from the integration of whole-slide images (WSIs) with multiple instance learning (MIL), enabling high-resolution analysis of tissue morphology. Graph-based MIL (GNN-MIL) approaches have emerged as powerful solutions for capturing contextual information in WSIs, thereby improving diagnostic accuracy. However, WSIs require significant computational and infrastructural resources, limiting accessibility in resource-constrained settings. Conventional light microscopes offer a cost-effective alternative, but applying GNN-MIL to such data is challenging due to extensive redundant images and missing spatial coordinates, which hinder contextual learning. To address these issues, we introduce MicroMIL, the first weakly-supervised MIL framework specifically designed for images acquired from conventional light microscopes. MicroMIL leverages a representative image extractor (RIE) that employs deep cluster embedding (DCE) and hard Gumbel-Softmax to dynamically reduce redundancy and select representative images. These images serve as graph nodes, with edges computed via cosine similarity, eliminating the need for spatial coordinates while preserving contextual information. Extensive experiments on a real-world colon cancer dataset and the BreakHis dataset demonstrate that MicroMIL achieves state-of-the-art performance, improving both diagnostic accuracy and robustness to redundancy. The code is available at https://github.com/kimjongwoo-cell/MicroMIL", "date": "2024-07-31", "categories": []}
{"id": "2508.18831", "title": "Automated Classification of Normal and Atypical Mitotic Figures Using ConvNeXt V2: MIDOG 2025 Track 2", "abstract": "This paper presents our solution for the MIDOG 2025 Challenge Track 2, which focuses on binary classification of normal mitotic figures (NMFs) versus atypical mitotic figures (AMFs) in histopathological images. Our approach leverages a ConvNeXt V2 base model with center cropping preprocessing and 5-fold cross-validation ensemble strategy. The method addresses key challenges including severe class imbalance, high morphological variability, and domain heterogeneity across different tumor types, species, and scanners. Through strategic preprocessing with 60% center cropping and mixed precision training, our model achieved robust performance on the diverse MIDOG 2025 dataset. The solution demonstrates the effectiveness of modern convolutional architectures for mitotic figure subtyping while maintaining computational efficiency through careful architectural choices and training optimizations.", "date": "2025-08-26", "categories": []}
{"id": "2508.17213", "title": "Multi-modal Knowledge Decomposition based Online Distillation for Biomarker Prediction in Breast Cancer Histopathology", "abstract": "Immunohistochemical (IHC) biomarker prediction benefits from multi-modal data fusion analysis. However, the simultaneous acquisition of multi-modal data, such as genomic and pathological information, is often challenging due to cost or technical limitations. To address this challenge, we propose an online distillation approach based on Multi-modal Knowledge Decomposition (MKD) to enhance IHC biomarker prediction in haematoxylin and eosin (H\\&E) stained histopathology images. This method leverages paired genomic-pathology data during training while enabling inference using either pathology slides alone or both modalities. Two teacher and one student models are developed to extract modality-specific and modality-general features by minimizing the MKD loss. To maintain the internal structural relationships between samples, Similarity-preserving Knowledge Distillation (SKD) is applied. Additionally, Collaborative Learning for Online Distillation (CLOD) facilitates mutual learning between teacher and student models, encouraging diverse and complementary learning dynamics. Experiments on the TCGA-BRCA and in-house QHSU datasets demonstrate that our approach achieves superior performance in IHC biomarker prediction using uni-modal data. Our code is available at https://github.com/qiyuanzz/MICCAI2025_MKD.", "date": "2025-08-24", "categories": []}
{"id": "2508.17666", "title": "M^3-GloDets: Multi-Region and Multi-Scale Analysis of Fine-Grained Diseased Glomerular Detection", "abstract": "Accurate detection of diseased glomeruli is fundamental to progress in renal pathology and underpins the delivery of reliable clinical diagnoses. Although recent advances in computer vision have produced increasingly sophisticated detection algorithms, the majority of research efforts have focused on normal glomeruli or instances of global sclerosis, leaving the wider spectrum of diseased glomerular subtypes comparatively understudied. This disparity is not without consequence; the nuanced and highly variable morphological characteristics that define these disease variants frequently elude even the most advanced computational models. Moreover, ongoing debate surrounds the choice of optimal imaging magnifications and region-of-view dimensions for fine-grained glomerular analysis, adding further complexity to the pursuit of accurate classification and robust segmentation.\n  To bridge these gaps, we present M^3-GloDet, a systematic framework designed to enable thorough evaluation of detection models across a broad continuum of regions, scales, and classes. Within this framework, we evaluate both long-standing benchmark architectures and recently introduced state-of-the-art models that have achieved notable performance, using an experimental design that reflects the diversity of region-of-interest sizes and imaging resolutions encountered in routine digital renal pathology. As the results, we found that intermediate patch sizes offered the best balance between context and efficiency. Additionally, moderate magnifications enhanced generalization by reducing overfitting. Through systematic comparison of these approaches on a multi-class diseased glomerular dataset, our aim is to advance the understanding of model strengths and limitations, and to offer actionable insights for the refinement of automated detection strategies and clinical workflows in the digital pathology domain.", "date": "2025-08-25", "categories": []}
{"id": "2508.15904", "title": "Boosting Pathology Foundation Models via Few-shot Prompt-tuning for Rare Cancer Subtyping", "abstract": "Rare cancers comprise 20-25% of all malignancies but face major diagnostic challenges due to limited expert availability-especially in pediatric oncology, where they represent over 70% of cases. While pathology vision-language (VL) foundation models show promising zero-shot capabilities for common cancer subtyping, their clinical performance for rare cancers remains limited. Existing multi-instance learning (MIL) methods rely only on visual features, overlooking cross-modal knowledge and compromising interpretability critical for rare cancer diagnosis. To address this limitation, we propose PathPT, a novel framework that fully exploits the potential of vision-language pathology foundation models through spatially-aware visual aggregation and task-specific prompt tuning. Unlike conventional MIL, PathPT converts WSI-level supervision into fine-grained tile-level guidance by leveraging the zero-shot capabilities of VL models, thereby preserving localization on cancerous regions and enabling cross-modal reasoning through prompts aligned with histopathological semantics. We benchmark PathPT on eight rare cancer datasets(four adult and four pediatric) spanning 56 subtypes and 2,910 WSIs, as well as three common cancer datasets, evaluating four state-of-the-art VL models and four MIL frameworks under three few-shot settings. Results show that PathPT consistently delivers superior performance, achieving substantial gains in subtyping accuracy and cancerous region grounding ability. This work advances AI-assisted diagnosis for rare cancers, offering a scalable solution for improving subtyping accuracy in settings with limited access to specialized expertise.", "date": "2025-08-21", "categories": []}
{"id": "2508.16085", "title": "Ensemble learning of foundation models for precision oncology", "abstract": "Histopathology is essential for disease diagnosis and treatment decision-making. Recent advances in artificial intelligence (AI) have enabled the development of pathology foundation models that learn rich visual representations from large-scale whole-slide images (WSIs). However, existing models are often trained on disparate datasets using varying strategies, leading to inconsistent performance and limited generalizability. Here, we introduce ELF (Ensemble Learning of Foundation models), a novel framework that integrates five state-of-the-art pathology foundation models to generate unified slide-level representations. Trained on 53,699 WSIs spanning 20 anatomical sites, ELF leverages ensemble learning to capture complementary information from diverse models while maintaining high data efficiency. Unlike traditional tile-level models, ELF's slide-level architecture is particularly advantageous in clinical contexts where data are limited, such as therapeutic response prediction. We evaluated ELF across a wide range of clinical applications, including disease classification, biomarker detection, and response prediction to major anticancer therapies, cytotoxic chemotherapy, targeted therapy, and immunotherapy, across multiple cancer types. ELF consistently outperformed all constituent foundation models and existing slide-level models, demonstrating superior accuracy and robustness. Our results highlight the power of ensemble learning for pathology foundation models and suggest ELF as a scalable and generalizable solution for advancing AI-assisted precision oncology.", "date": "2025-08-22", "categories": []}
{"id": "2410.16671", "title": "NucleiMix: Realistic Data Augmentation for Nuclei Instance Segmentation", "abstract": "Nuclei instance segmentation is an essential task in pathology image analysis, serving as the foundation for many downstream applications. The release of several public datasets has significantly advanced research in this area, yet many existing methods struggle with data imbalance issues. To address this challenge, this study introduces a data augmentation method, called NucleiMix, which is designed to balance the distribution of nuclei types by increasing the number of rare-type nuclei within datasets. NucleiMix operates in two phases. In the first phase, it identifies candidate locations similar to the surroundings of rare-type nuclei and inserts rare-type nuclei into the candidate locations. In the second phase, it employs a progressive inpainting strategy using a pre-trained diffusion model to seamlessly integrate rare-type nuclei into their new environments in replacement of major-type nuclei or background locations. We systematically evaluate the effectiveness of NucleiMix on three public datasets using two popular nuclei instance segmentation models. The results demonstrate the superior ability of NucleiMix to synthesize realistic rare-type nuclei and to enhance the quality of nuclei segmentation and classification in an accurate and robust manner.", "date": "2024-10-22", "categories": []}
{"id": "2508.14393", "title": "Img2ST-Net: Efficient High-Resolution Spatial Omics Prediction from Whole Slide Histology Images via Fully Convolutional Image-to-Image Learning", "abstract": "Recent advances in multi-modal AI have demonstrated promising potential for generating the currently expensive spatial transcriptomics (ST) data directly from routine histology images, offering a means to reduce the high cost and time-intensive nature of ST data acquisition. However, the increasing resolution of ST, particularly with platforms such as Visium HD achieving 8um or finer, introduces significant computational and modeling challenges. Conventional spot-by-spot sequential regression frameworks become inefficient and unstable at this scale, while the inherent extreme sparsity and low expression levels of high-resolution ST further complicate both prediction and evaluation. To address these limitations, we propose Img2ST-Net, a novel histology-to-ST generation framework for efficient and parallel high-resolution ST prediction. Unlike conventional spot-by-spot inference methods, Img2ST-Net employs a fully convolutional architecture to generate dense, HD gene expression maps in a parallelized manner. By modeling HD ST data as super-pixel representations, the task is reformulated from image-to-omics inference into a super-content image generation problem with hundreds or thousands of output channels. This design not only improves computational efficiency but also better preserves the spatial organization intrinsic to spatial omics data. To enhance robustness under sparse expression patterns, we further introduce SSIM-ST, a structural-similarity-based evaluation metric tailored for high-resolution ST analysis. We present a scalable, biologically coherent framework for high-resolution ST prediction. Img2ST-Net offers a principled solution for efficient and accurate ST inference at scale. Our contributions lay the groundwork for next-generation ST modeling that is robust and resolution-aware. The source code has been made publicly available at https://github.com/hrlblab/Img2ST-Net.", "date": "2025-08-20", "categories": []}
{"id": "2508.14537", "title": "WISE-FUSE: Efficient Whole Slide Image Encoding via Coarse-to-Fine Patch Selection with VLM and LLM Knowledge Fusion", "abstract": "Whole slide images (WSIs) in computational pathology (CPath) pose a major computational challenge due to their gigapixel scale, often requiring the processing of tens to hundreds of thousands of high-resolution patches per slide. This results in prohibitive encoding costs, with preprocessing and training times extending to days or even weeks-making WSI encoding the most significant bottleneck in real-world deployment. In this work, we propose WISE-FUSE, an adaptive WSI encoding framework that leverages pathology-domain vision-language models and large language models to address this challenge by selectively processing diagnostically relevant regions. WISE-FUSE first computes similarity scores between low-resolution patches and class-specific textual descriptions using a knowledge distillation mechanism that preserves fine-grained diagnostic features. Based on these similarity scores, we select a small subset of informative regions for the target task, which quickly eliminates irrelevant patches at the coarse level. The corresponding high-resolution patches are then selectively encoded and fused with textual embeddings to reinforce diagnostic context. Extensive experiments demonstrate that WISE-FUSE reduces WSI encoding time by over threefold while achieving diagnostic performance comparable to or surpassing that of exhaustive patch processing, offering a scalable and practical solution for CPath.", "date": "2025-08-20", "categories": []}
{"id": "2507.17303", "title": "A Versatile Pathology Co-pilot via Reasoning Enhanced Multimodal Large Language Model", "abstract": "Multimodal large language models (MLLMs) have emerged as powerful tools for computational pathology, offering unprecedented opportunities to integrate pathological images with language context for comprehensive diagnostic analysis. These models hold particular promise for automating complex tasks that traditionally require expert interpretation of pathologists. However, current MLLM approaches in pathology demonstrate significantly constrained reasoning capabilities, primarily due to their reliance on expensive chain-of-thought annotations. Additionally, existing methods remain limited to simplex application of visual question answering (VQA) at the region-of-interest (ROI) level, failing to address the full spectrum of diagnostic needs such as ROI classification, detection, segmentation, whole-slide-image (WSI) classification and VQA in clinical practice. In this study, we present SmartPath-R1, a versatile MLLM capable of simultaneously addressing both ROI-level and WSI-level tasks while demonstrating robust pathological reasoning capability. Our framework combines scale-dependent supervised fine-tuning and task-aware reinforcement fine-tuning, which circumvents the requirement for chain-of-thought supervision by leveraging the intrinsic knowledge within MLLM. Furthermore, SmartPath-R1 integrates multiscale and multitask analysis through a mixture-of-experts mechanism, enabling dynamic processing for diverse tasks. We curate a large-scale dataset comprising 2.3M ROI samples and 188K WSI samples for training and evaluation. Extensive experiments across 72 tasks validate the effectiveness and superiority of the proposed approach. This work represents a significant step toward developing versatile, reasoning-enhanced AI systems for precision pathology.", "date": "2025-07-23", "categories": []}
{"id": "2508.11181", "title": "HistoViT: Vision Transformer for Accurate and Scalable Histopathological Cancer Diagnosis", "abstract": "Accurate and scalable cancer diagnosis remains a critical challenge in modern pathology, particularly for malignancies such as breast, prostate, bone, and cervical, which exhibit complex histological variability. In this study, we propose a transformer-based deep learning framework for multi-class tumor classification in histopathological images. Leveraging a fine-tuned Vision Transformer (ViT) architecture, our method addresses key limitations of conventional convolutional neural networks, offering improved performance, reduced preprocessing requirements, and enhanced scalability across tissue types. To adapt the model for histopathological cancer images, we implement a streamlined preprocessing pipeline that converts tiled whole-slide images into PyTorch tensors and standardizes them through data normalization. This ensures compatibility with the ViT architecture and enhances both convergence stability and overall classification performance. We evaluate our model on four benchmark datasets: ICIAR2018 (breast), SICAPv2 (prostate), UT-Osteosarcoma (bone), and SipakMed (cervical) dataset -- demonstrating consistent outperformance over existing deep learning methods. Our approach achieves classification accuracies of 99.32%, 96.92%, 95.28%, and 96.94% for breast, prostate, bone, and cervical cancers respectively, with area under the ROC curve (AUC) scores exceeding 99% across all datasets. These results confirm the robustness, generalizability, and clinical potential of transformer-based architectures in digital pathology. Our work represents a significant advancement toward reliable, automated, and interpretable cancer diagnosis systems that can alleviate diagnostic burdens and improve healthcare outcomes.", "date": "2025-08-15", "categories": []}
{"id": "2507.06639", "title": "EXAONE Path 2.0: Pathology Foundation Model with End-to-End Supervision", "abstract": "In digital pathology, whole-slide images (WSIs) are often difficult to handle due to their gigapixel scale, so most approaches train patch encoders via self-supervised learning (SSL) and then aggregate the patch-level embeddings via multiple instance learning (MIL) or slide encoders for downstream tasks. However, patch-level SSL may overlook complex domain-specific features that are essential for biomarker prediction, such as mutation status and molecular characteristics, as SSL methods rely only on basic augmentations selected for natural image domains on small patch-level area. Moreover, SSL methods remain less data efficient than fully supervised approaches, requiring extensive computational resources and datasets to achieve competitive performance. To address these limitations, we present EXAONE Path 2.0, a pathology foundation model that learns patch-level representations under direct slide-level supervision. Using only 37k WSIs for training, EXAONE Path 2.0 achieves state-of-the-art average performance across 10 biomarker prediction tasks, demonstrating remarkable data efficiency.", "date": "2025-07-09", "categories": []}
{"id": "2508.10232", "title": "CellSymphony: Deciphering the molecular and phenotypic orchestration of cells with single-cell pathomics", "abstract": "Xenium, a new spatial transcriptomics platform, enables subcellular-resolution profiling of complex tumor tissues. Despite the rich morphological information in histology images, extracting robust cell-level features and integrating them with spatial transcriptomics data remains a critical challenge. We introduce CellSymphony, a flexible multimodal framework that leverages foundation model-derived embeddings from both Xenium transcriptomic profiles and histology images at true single-cell resolution. By learning joint representations that fuse spatial gene expression with morphological context, CellSymphony achieves accurate cell type annotation and uncovers distinct microenvironmental niches across three cancer types. This work highlights the potential of foundation models and multimodal fusion for deciphering the physiological and phenotypic orchestration of cells within complex tissue ecosystems.", "date": "2025-08-13", "categories": []}
{"id": "2409.01330", "title": "Pediatric brain tumor classification using digital histopathology and deep learning: evaluation of SOTA methods on a multi-center Swedish cohort", "abstract": "Brain tumors are the most common solid tumors in children and young adults, but the scarcity of large histopathology datasets has limited the application of computational pathology in this group. This study implements two weakly supervised multiple-instance learning (MIL) approaches on patch-features obtained from state-of-the-art histology-specific foundation models to classify pediatric brain tumors in hematoxylin and eosin whole slide images (WSIs) from a multi-center Swedish cohort. WSIs from 540 subjects (age 8.5$\\pm$4.9 years) diagnosed with brain tumor were gathered from the six Swedish university hospitals. Instance (patch)-level features were obtained from WSIs using three pre-trained feature extractors: ResNet50, UNI, and CONCH. Instances were aggregated using attention-based MIL (ABMIL) or clustering-constrained attention MIL (CLAM) for patient-level classification. Models were evaluated on three classification tasks based on the hierarchical classification of pediatric brain tumors: tumor category, family, and type. Model generalization was assessed by training on data from two of the centers and testing on data from four other centers. Model interpretability was evaluated through attention mapping. The highest classification performance was achieved using UNI features and ABMIL aggregation, with Matthew's correlation coefficient of 0.76$\\pm$0.04, 0.63$\\pm$0.04, and 0.60$\\pm$0.05 for tumor category, family, and type classification, respectively. When evaluating generalization, models utilizing UNI and CONCH features outperformed those using ResNet50. However, the drop in performance from the in-site to out-of-site testing was similar across feature extractors. These results show the potential of state-of-the-art computational pathology methods in diagnosing pediatric brain tumors at different hierarchical levels with fair generalizability on a multi-center national dataset.", "date": "2024-09-02", "categories": []}
{"id": "2508.09339", "title": "UltraLight Med-Vision Mamba for Classification of Neoplastic Progression in Tubular Adenomas", "abstract": "Identification of precancerous polyps during routine colonoscopy screenings is vital for their excision, lowering the risk of developing colorectal cancer. Advanced deep learning algorithms enable precise adenoma classification and stratification, improving risk assessment accuracy and enabling personalized surveillance protocols that optimize patient outcomes. Ultralight Med-Vision Mamba, a state-space based model (SSM), has excelled in modeling long- and short-range dependencies and image generalization, critical factors for analyzing whole slide images. Furthermore, Ultralight Med-Vision Mamba's efficient architecture offers advantages in both computational speed and scalability, making it a promising tool for real-time clinical deployment.", "date": "2025-08-12", "categories": []}
{"id": "2412.02141", "title": "WSI-LLaVA: A Multimodal Large Language Model for Whole Slide Image", "abstract": "Recent advancements in computational pathology have produced patch-level Multi-modal Large Language Models (MLLMs), but these models are limited by their inability to analyze whole slide images (WSIs) comprehensively and their tendency to bypass crucial morphological features that pathologists rely on for diagnosis. To address these challenges, we first introduce WSI-Bench, a large-scale morphology-aware benchmark containing 180k VQA pairs from 9,850 WSIs across 30 cancer types, designed to evaluate MLLMs' understanding of morphological characteristics crucial for accurate diagnosis. Building upon this benchmark, we present WSI-LLaVA, a novel framework for gigapixel WSI understanding that employs a three-stage training approach: WSI-text alignment, feature space alignment, and task-specific instruction tuning. To better assess model performance in pathological contexts, we develop two specialized WSI metrics: WSI-Precision and WSI-Relevance. Experimental results demonstrate that WSI-LLaVA outperforms existing models across all capability dimensions, with a significant improvement in morphological analysis, establishing a clear correlation between morphological understanding and diagnostic accuracy.", "date": "2024-12-03", "categories": []}
{"id": "2402.10595", "title": "Compact and De-biased Negative Instance Embedding for Multi-Instance Learning on Whole-Slide Image Classification", "abstract": "Whole-slide image (WSI) classification is a challenging task because 1) patches from WSI lack annotation, and 2) WSI possesses unnecessary variability, e.g., stain protocol. Recently, Multiple-Instance Learning (MIL) has made significant progress, allowing for classification based on slide-level, rather than patch-level, annotations. However, existing MIL methods ignore that all patches from normal slides are normal. Using this free annotation, we introduce a semi-supervision signal to de-bias the inter-slide variability and to capture the common factors of variation within normal patches. Because our method is orthogonal to the MIL algorithm, we evaluate our method on top of the recently proposed MIL algorithms and also compare the performance with other semi-supervised approaches. We evaluate our method on two public WSI datasets including Camelyon-16 and TCGA lung cancer and demonstrate that our approach significantly improves the predictive performance of existing MIL algorithms and outperforms other semi-supervised algorithms. We release our code at https://github.com/AITRICS/pathology_mil.", "date": "2024-02-16", "categories": []}
{"id": "2407.14904", "title": "Large-vocabulary forensic pathological analyses via prototypical cross-modal contrastive learning", "abstract": "Forensic pathology is critical in determining the cause and manner of death through post-mortem examinations, both macroscopic and microscopic. The field, however, grapples with issues such as outcome variability, laborious processes, and a scarcity of trained professionals. This paper presents SongCi, an innovative visual-language model (VLM) designed specifically for forensic pathology. SongCi utilizes advanced prototypical cross-modal self-supervised contrastive learning to enhance the accuracy, efficiency, and generalizability of forensic analyses. It was pre-trained and evaluated on a comprehensive multi-center dataset, which includes over 16 million high-resolution image patches, 2,228 vision-language pairs of post-mortem whole slide images (WSIs), and corresponding gross key findings, along with 471 distinct diagnostic outcomes. Our findings indicate that SongCi surpasses existing multi-modal AI models in many forensic pathology tasks, performs comparably to experienced forensic pathologists and significantly better than less experienced ones, and provides detailed multi-modal explainability, offering critical assistance in forensic investigations. To the best of our knowledge, SongCi is the first VLM specifically developed for forensic pathological analysis and the first large-vocabulary computational pathology (CPath) model that directly processes gigapixel WSIs in forensic science.", "date": "2024-07-20", "categories": []}
{"id": "2503.11439", "title": "COIN: Confidence Score-Guided Distillation for Annotation-Free Cell Segmentation", "abstract": "Cell instance segmentation (CIS) is crucial for identifying individual cell morphologies in histopathological images, providing valuable insights for biological and medical research. While unsupervised CIS (UCIS) models aim to reduce the heavy reliance on labor-intensive image annotations, they fail to accurately capture cell boundaries, causing missed detections and poor performance. Recognizing the absence of error-free instances as a key limitation, we present COIN (COnfidence score-guided INstance distillation), a novel annotation-free framework with three key steps: (1) Increasing the sensitivity for the presence of error-free instances via unsupervised semantic segmentation with optimal transport, leveraging its ability to discriminate spatially minor instances, (2) Instance-level confidence scoring to measure the consistency between model prediction and refined mask and identify highly confident instances, offering an alternative to ground truth annotations, and (3) Progressive expansion of confidence with recursive self-distillation. Extensive experiments across six datasets show COIN outperforming existing UCIS methods, even surpassing semi- and weakly-supervised approaches across all metrics on the MoNuSeg and TNBC datasets. The code is available at https://github.com/shjo-april/COIN.", "date": "2025-03-14", "categories": []}
{"id": "2303.09354", "title": "The NCI Imaging Data Commons as a platform for reproducible research in computational pathology", "abstract": "Background and Objectives: Reproducibility is a major challenge in developing machine learning (ML)-based solutions in computational pathology (CompPath). The NCI Imaging Data Commons (IDC) provides >120 cancer image collections according to the FAIR principles and is designed to be used with cloud ML services. Here, we explore its potential to facilitate reproducibility in CompPath research.\n  Methods: Using the IDC, we implemented two experiments in which a representative ML-based method for classifying lung tumor tissue was trained and/or evaluated on different datasets. To assess reproducibility, the experiments were run multiple times with separate but identically configured instances of common ML services.\n  Results: The AUC values of different runs of the same experiment were generally consistent. However, we observed small variations in AUC values of up to 0.045, indicating a practical limit to reproducibility.\n  Conclusions: We conclude that the IDC facilitates approaching the reproducibility limit of CompPath research (i) by enabling researchers to reuse exactly the same datasets and (ii) by integrating with cloud ML services so that experiments can be run in identically configured computing environments.", "date": "2023-03-16", "categories": []}
{"id": "2412.13137", "title": "Unlocking the Potential of Digital Pathology: Novel Baselines for Compression", "abstract": "Digital pathology offers a groundbreaking opportunity to transform clinical practice in histopathological image analysis, yet faces a significant hurdle: the substantial file sizes of pathological Whole Slide Images (WSI). While current digital pathology solutions rely on lossy JPEG compression to address this issue, lossy compression can introduce color and texture disparities, potentially impacting clinical decision-making. While prior research addresses perceptual image quality and downstream performance independently of each other, we jointly evaluate compression schemes for perceptual and downstream task quality on four different datasets. In addition, we collect an initially uncompressed dataset for an unbiased perceptual evaluation of compression schemes. Our results show that deep learning models fine-tuned for perceptual quality outperform conventional compression schemes like JPEG-XL or WebP for further compression of WSI. However, they exhibit a significant bias towards the compression artifacts present in the training data and struggle to generalize across various compression schemes. We introduce a novel evaluation metric based on feature similarity between original files and compressed files that aligns very well with the actual downstream performance on the compressed WSI. Our metric allows for a general and standardized evaluation of lossy compression schemes and mitigates the requirement to independently assess different downstream tasks. Our study provides novel insights for the assessment of lossy compression schemes for WSI and encourages a unified evaluation of lossy compression schemes to accelerate the clinical uptake of digital pathology.", "date": "2024-12-17", "categories": []}
{"id": "2508.04734", "title": "Cross-Domain Image Synthesis: Generating H&E from Multiplex Biomarker Imaging", "abstract": "While multiplex immunofluorescence (mIF) imaging provides deep, spatially-resolved molecular data, integrating this information with the morphological standard of Hematoxylin & Eosin (H&E) can be very important for obtaining complementary information about the underlying tissue. Generating a virtual H&E stain from mIF data offers a powerful solution, providing immediate morphological context. Crucially, this approach enables the application of the vast ecosystem of H&E-based computer-aided diagnosis (CAD) tools to analyze rich molecular data, bridging the gap between molecular and morphological analysis. In this work, we investigate the use of a multi-level Vector-Quantized Generative Adversarial Network (VQGAN) to create high-fidelity virtual H&E stains from mIF images. We rigorously evaluated our VQGAN against a standard conditional GAN (cGAN) baseline on two publicly available colorectal cancer datasets, assessing performance on both image similarity and functional utility for downstream analysis. Our results show that while both architectures produce visually plausible images, the virtual stains generated by our VQGAN provide a more effective substrate for computer-aided diagnosis. Specifically, downstream nuclei segmentation and semantic preservation in tissue classification tasks performed on VQGAN-generated images demonstrate superior performance and agreement with ground-truth analysis compared to those from the cGAN. This work establishes that a multi-level VQGAN is a robust and superior architecture for generating scientifically useful virtual stains, offering a viable pathway to integrate the rich molecular data of mIF into established and powerful H&E-based analytical workflows.", "date": "2025-08-05", "categories": []}
{"id": "2508.05114", "title": "AHDMIL: Asymmetric Hierarchical Distillation Multi-Instance Learning for Fast and Accurate Whole-Slide Image Classification", "abstract": "Although multi-instance learning (MIL) has succeeded in pathological image classification, it faces the challenge of high inference costs due to the need to process thousands of patches from each gigapixel whole slide image (WSI). To address this, we propose AHDMIL, an Asymmetric Hierarchical Distillation Multi-Instance Learning framework that enables fast and accurate classification by eliminating irrelevant patches through a two-step training process. AHDMIL comprises two key components: the Dynamic Multi-Instance Network (DMIN), which operates on high-resolution WSIs, and the Dual-Branch Lightweight Instance Pre-screening Network (DB-LIPN), which analyzes corresponding low-resolution counterparts. In the first step, self-distillation (SD), DMIN is trained for WSI classification while generating per-instance attention scores to identify irrelevant patches. These scores guide the second step, asymmetric distillation (AD), where DB-LIPN learns to predict the relevance of each low-resolution patch. The relevant patches predicted by DB-LIPN have spatial correspondence with patches in high-resolution WSIs, which are used for fine-tuning and efficient inference of DMIN. In addition, we design the first Chebyshev-polynomial-based Kolmogorov-Arnold (CKA) classifier in computational pathology, which improves classification performance through learnable activation layers. Extensive experiments on four public datasets demonstrate that AHDMIL consistently outperforms previous state-of-the-art methods in both classification performance and inference speed. For example, on the Camelyon16 dataset, it achieves a relative improvement of 5.3% in accuracy and accelerates inference by 1.2.times. Across all datasets, area under the curve (AUC), accuracy, f1 score, and brier score show consistent gains, with average inference speedups ranging from 1.2 to 2.1 times. The code is available.", "date": "2025-08-07", "categories": []}
{"id": "2508.05391", "title": "Artificial Intelligence-Based Classification of Spitz Tumors", "abstract": "Spitz tumors are diagnostically challenging due to overlap in atypical histological features with conventional melanomas. We investigated to what extent AI models, using histological and/or clinical features, can: (1) distinguish Spitz tumors from conventional melanomas; (2) predict the underlying genetic aberration of Spitz tumors; and (3) predict the diagnostic category of Spitz tumors. The AI models were developed and validated using a dataset of 393 Spitz tumors and 379 conventional melanomas. Predictive performance was measured using the AUROC and the accuracy. The performance of the AI models was compared with that of four experienced pathologists in a reader study. Moreover, a simulation experiment was conducted to investigate the impact of implementing AI-based recommendations for ancillary diagnostic testing on the workflow of the pathology department. The best AI model based on UNI features reached an AUROC of 0.95 and an accuracy of 0.86 in differentiating Spitz tumors from conventional melanomas. The genetic aberration was predicted with an accuracy of 0.55 compared to 0.25 for randomly guessing. The diagnostic category was predicted with an accuracy of 0.51, where random chance-level accuracy equaled 0.33. On all three tasks, the AI models performed better than the four pathologists, although differences were not statistically significant for most individual comparisons. Based on the simulation experiment, implementing AI-based recommendations for ancillary diagnostic testing could reduce material costs, turnaround times, and examinations. In conclusion, the AI models achieved a strong predictive performance in distinguishing between Spitz tumors and conventional melanomas. On the more challenging tasks of predicting the genetic aberration and the diagnostic category of Spitz tumors, the AI models performed better than random chance.", "date": "2025-08-07", "categories": []}
{"id": "2410.17557", "title": "BlurryScope enables compact, cost-effective scanning microscopy for HER2 scoring using deep learning on blurry images", "abstract": "We developed a rapid scanning optical microscope, termed \"BlurryScope\", that leverages continuous image acquisition and deep learning to provide a cost-effective and compact solution for automated inspection and analysis of tissue sections. This device offers comparable speed to commercial digital pathology scanners, but at a significantly lower price point and smaller size/weight. Using BlurryScope, we implemented automated classification of human epidermal growth factor receptor 2 (HER2) scores on motion-blurred images of immunohistochemically (IHC) stained breast tissue sections, achieving concordant results with those obtained from a high-end digital scanning microscope. Using a test set of 284 unique patient cores, we achieved testing accuracies of 79.3% and 89.7% for 4-class (0, 1+, 2+, 3+) and 2-class (0/1+, 2+/3+) HER2 classification, respectively. BlurryScope automates the entire workflow, from image scanning to stitching and cropping, as well as HER2 score classification.", "date": "2024-10-23", "categories": []}
{"id": "2508.01602", "title": "Enhancing Zero-Shot Brain Tumor Subtype Classification via Fine-Grained Patch-Text Alignment", "abstract": "The fine-grained classification of brain tumor subtypes from histopathological whole slide images is highly challenging due to subtle morphological variations and the scarcity of annotated data. Although vision-language models have enabled promising zero-shot classification, their ability to capture fine-grained pathological features remains limited, resulting in suboptimal subtype discrimination. To address these challenges, we propose the Fine-Grained Patch Alignment Network (FG-PAN), a novel zero-shot framework tailored for digital pathology. FG-PAN consists of two key modules: (1) a local feature refinement module that enhances patch-level visual features by modeling spatial relationships among representative patches, and (2) a fine-grained text description generation module that leverages large language models to produce pathology-aware, class-specific semantic prototypes. By aligning refined visual features with LLM-generated fine-grained descriptions, FG-PAN effectively increases class separability in both visual and semantic spaces. Extensive experiments on multiple public pathology datasets, including EBRAINS and TCGA, demonstrate that FG-PAN achieves state-of-the-art performance and robust generalization in zero-shot brain tumor subtype classification.", "date": "2025-08-03", "categories": []}
{"id": "2507.03532", "title": "PhenoBench: A Comprehensive Benchmark for Cell Phenotyping", "abstract": "Digital pathology has seen the advent of a wealth of foundational models (FM), yet to date their performance on cell phenotyping has not been benchmarked in a unified manner. We therefore propose PhenoBench: A comprehensive benchmark for cell phenotyping on Hematoxylin and Eosin (H&E) stained histopathology images. We provide both PhenoCell, a new H&E dataset featuring 14 granular cell types identified by using multiplexed imaging, and ready-to-use fine-tuning and benchmarking code that allows the systematic evaluation of multiple prominent pathology FMs in terms of dense cell phenotype predictions in different generalization scenarios. We perform extensive benchmarking of existing FMs, providing insights into their generalization behavior under technical vs. medical domain shifts. Furthermore, while FMs achieve macro F1 scores > 0.70 on previously established benchmarks such as Lizard and PanNuke, on PhenoCell, we observe scores as low as 0.20. This indicates a much more challenging task not captured by previous benchmarks, establishing PhenoCell as a prime asset for future benchmarking of FMs and supervised models alike. Code and data are available on GitHub.", "date": "2025-07-04", "categories": []}
{"id": "2508.01352", "title": "Predicting EGFR Mutation in LUAD from Histopathological Whole-Slide Images Using Pretrained Foundation Model and Transfer Learning: An Indian Cohort Study", "abstract": "Lung adenocarcinoma (LUAD) is a subtype of non-small cell lung cancer (NSCLC). LUAD with mutation in the EGFR gene accounts for approximately 46% of LUAD cases. Patients carrying EGFR mutations can be treated with specific tyrosine kinase inhibitors (TKIs). Hence, predicting EGFR mutation status can help in clinical decision making. H&E-stained whole slide imaging (WSI) is a routinely performed screening procedure for cancer staging and subtyping, especially affecting the Southeast Asian populations with significantly higher incidence of the mutation when compared to Caucasians (39-64% vs 7-22%). Recent progress in AI models has shown promising results in cancer detection and classification. In this study, we propose a deep learning (DL) framework built on vision transformers (ViT) based pathology foundation model and attention-based multiple instance learning (ABMIL) architecture to predict EGFR mutation status from H&E WSI. The developed pipeline was trained using data from an Indian cohort (170 WSI) and evaluated across two independent datasets: Internal test (30 WSI from Indian cohort) set, and an external test set from TCGA (86 WSI). The model shows consistent performance across both datasets, with AUCs of 0.933 (+/-0.010), and 0.965 (+/-0.015) for the internal and external test sets respectively. This proposed framework can be efficiently trained on small datasets, achieving superior performance as compared to several prior studies irrespective of training domain. The current study demonstrates the feasibility of accurately predicting EGFR mutation status using routine pathology slides, particularly in resource-limited settings using foundation models and attention-based multiple instance learning.", "date": "2025-08-02", "categories": []}
{"id": "2508.03277", "title": "Efficient Multi-Slide Visual-Language Feature Fusion for Placental Disease Classification", "abstract": "Accurate prediction of placental diseases via whole slide images (WSIs) is critical for preventing severe maternal and fetal complications. However, WSI analysis presents significant computational challenges due to the massive data volume. Existing WSI classification methods encounter critical limitations: (1) inadequate patch selection strategies that either compromise performance or fail to sufficiently reduce computational demands, and (2) the loss of global histological context resulting from patch-level processing approaches. To address these challenges, we propose an Efficient multimodal framework for Patient-level placental disease Diagnosis, named EmmPD. Our approach introduces a two-stage patch selection module that combines parameter-free and learnable compression strategies, optimally balancing computational efficiency with critical feature preservation. Additionally, we develop a hybrid multimodal fusion module that leverages adaptive graph learning to enhance pathological feature representation and incorporates textual medical reports to enrich global contextual understanding. Extensive experiments conducted on both a self-constructed patient-level Placental dataset and two public datasets demonstrating that our method achieves state-of-the-art diagnostic performance. The code is available at https://github.com/ECNU-MultiDimLab/EmmPD.", "date": "2025-08-05", "categories": []}
{"id": "2405.17446", "title": "Comparing ImageNet Pre-training with Digital Pathology Foundation Models for Whole Slide Image-Based Survival Analysis", "abstract": "The abundance of information present in Whole Slide Images (WSIs) renders them an essential tool for survival analysis. Several Multiple Instance Learning frameworks proposed for this task utilize a ResNet50 backbone pre-trained on natural images. By leveraging recenetly released histopathological foundation models such as UNI and Hibou, the predictive prowess of existing MIL networks can be enhanced. Furthermore, deploying an ensemble of digital pathology foundation models yields higher baseline accuracy, although the benefits appear to diminish with more complex MIL architectures. Our code will be made publicly available upon acceptance.", "date": "2024-05-20", "categories": []}
{"id": "2411.13120", "title": "Virtual Staining of Label-Free Tissue in Imaging Mass Spectrometry", "abstract": "Imaging mass spectrometry (IMS) is a powerful tool for untargeted, highly multiplexed molecular mapping of tissue in biomedical research. IMS offers a means of mapping the spatial distributions of molecular species in biological tissue with unparalleled chemical specificity and sensitivity. However, most IMS platforms are not able to achieve microscopy-level spatial resolution and lack cellular morphological contrast, necessitating subsequent histochemical staining, microscopic imaging and advanced image registration steps to enable molecular distributions to be linked to specific tissue features and cell types. Here, we present a virtual histological staining approach that enhances spatial resolution and digitally introduces cellular morphological contrast into mass spectrometry images of label-free human tissue using a diffusion model. Blind testing on human kidney tissue demonstrated that the virtually stained images of label-free samples closely match their histochemically stained counterparts (with Periodic Acid-Schiff staining), showing high concordance in identifying key renal pathology structures despite utilizing IMS data with 10-fold larger pixel size. Additionally, our approach employs an optimized noise sampling technique during the diffusion model's inference process to reduce variance in the generated images, yielding reliable and repeatable virtual staining. We believe this virtual staining method will significantly expand the applicability of IMS in life sciences and open new avenues for mass spectrometry-based biomedical research.", "date": "2024-11-20", "categories": []}
{"id": "2508.01641", "title": "Minimal High-Resolution Patches Are Sufficient for Whole Slide Image Representation via Cascaded Dual-Scale Reconstruction", "abstract": "Whole-slide image (WSI) analysis remains challenging due to the gigapixel scale and sparsely distributed diagnostic regions. Multiple Instance Learning (MIL) mitigates this by modeling the WSI as bags of patches for slide-level prediction. However, most MIL approaches emphasize aggregator design while overlooking the impact of the feature extractor of the feature extraction stage, which is often pretrained on natural images. This leads to domain gap and suboptimal representations. Self-supervised learning (SSL) has shown promise in bridging domain gap via pretext tasks, but it still primarily builds upon generic backbones, thus requiring WSIs to be split into small patches. This inevitably splits histological structures and generates both redundant and interdependent patches, which in turn degrades aggregator performance and drastically increases training costs. To address this challenge, we propose a Cascaded Dual-Scale Reconstruction (CDSR) framework, demonstrating that only an average of 9 high-resolution patches per WSI are sufficient for robust slide-level representation. CDSR employs a two-stage selective sampling strategy that identifies the most informative representative regions from both model-based and semantic perspectives. These patches are then fed into a Local-to-Global Network, which reconstructs spatially coherent high-resolution WSI representations by integrating fine-grained local detail with global contextual information. Unlike existing dense-sampling or SSL pipelines, CDSR is optimized for efficiency and morphological fidelity. Experiments on Camelyon16, TCGA-NSCLC, and TCGA-RCC demonstrate that CDSR achieves improvements of 6.3% in accuracy and 5.5% in area under ROC curve on downstream classification tasks with only 7,070 (4.5% of total) high-resolution patches per dataset on average, outperforming state-of-the-art methods trained on over 10,000,000 patches.", "date": "2025-08-03", "categories": []}
{"id": "2508.02220", "title": "Welcome New Doctor: Continual Learning with Expert Consultation and Autoregressive Inference for Whole Slide Image Analysis", "abstract": "Whole Slide Image (WSI) analysis, with its ability to reveal detailed tissue structures in magnified views, plays a crucial role in cancer diagnosis and prognosis. Due to their giga-sized nature, WSIs require substantial storage and computational resources for processing and training predictive models. With the rapid increase in WSIs used in clinics and hospitals, there is a growing need for a continual learning system that can efficiently process and adapt existing models to new tasks without retraining or fine-tuning on previous tasks. Such a system must balance resource efficiency with high performance. In this study, we introduce COSFormer, a Transformer-based continual learning framework tailored for multi-task WSI analysis. COSFormer is designed to learn sequentially from new tasks wile avoiding the need to revisit full historical datasets. We evaluate COSFormer on a sequence of seven WSI datasets covering seven organs and six WSI-related tasks under both class-incremental and task-incremental settings. The results demonstrate COSFormer's superior generalizability and effectiveness compared to existing continual learning frameworks, establishing it as a robust solution for continual WSI analysis in clinical applications.", "date": "2025-08-04", "categories": []}
{"id": "2508.02528", "title": "From Pixels to Pathology: Restoration Diffusion for Diagnostic-Consistent Virtual IHC", "abstract": "Hematoxylin and eosin (H&E) staining is the clinical standard for assessing tissue morphology, but it lacks molecular-level diagnostic information. In contrast, immunohistochemistry (IHC) provides crucial insights into biomarker expression, such as HER2 status for breast cancer grading, but remains costly and time-consuming, limiting its use in time-sensitive clinical workflows. To address this gap, virtual staining from H&E to IHC has emerged as a promising alternative, yet faces two core challenges: (1) Lack of fair evaluation of synthetic images against misaligned IHC ground truths, and (2) preserving structural integrity and biological variability during translation. To this end, we present an end-to-end framework encompassing both generation and evaluation in this work. We introduce Star-Diff, a structure-aware staining restoration diffusion model that reformulates virtual staining as an image restoration task. By combining residual and noise-based generation pathways, Star-Diff maintains tissue structure while modeling realistic biomarker variability. To evaluate the diagnostic consistency of the generated IHC patches, we propose the Semantic Fidelity Score (SFS), a clinical-grading-task-driven metric that quantifies class-wise semantic degradation based on biomarker classification accuracy. Unlike pixel-level metrics such as SSIM and PSNR, SFS remains robust under spatial misalignment and classifier uncertainty. Experiments on the BCI dataset demonstrate that Star-Diff achieves state-of-the-art (SOTA) performance in both visual fidelity and diagnostic relevance. With rapid inference and strong clinical alignment,it presents a practical solution for applications such as intraoperative virtual IHC synthesis.", "date": "2025-08-04", "categories": []}
{"id": "2508.00383", "title": "$MV_{Hybrid}$: Improving Spatial Transcriptomics Prediction with Hybrid State Space-Vision Transformer Backbone in Pathology Vision Foundation Models", "abstract": "Spatial transcriptomics reveals gene expression patterns within tissue context, enabling precision oncology applications such as treatment response prediction, but its high cost and technical complexity limit clinical adoption. Predicting spatial gene expression (biomarkers) from routine histopathology images offers a practical alternative, yet current vision foundation models (VFMs) in pathology based on Vision Transformer (ViT) backbones perform below clinical standards. Given that VFMs are already trained on millions of diverse whole slide images, we hypothesize that architectural innovations beyond ViTs may better capture the low-frequency, subtle morphological patterns correlating with molecular phenotypes. By demonstrating that state space models initialized with negative real eigenvalues exhibit strong low-frequency bias, we introduce $MV_{Hybrid}$, a hybrid backbone architecture combining state space models (SSMs) with ViT. We compare five other different backbone architectures for pathology VFMs, all pretrained on identical colorectal cancer datasets using the DINOv2 self-supervised learning method. We evaluate all pretrained models using both random split and leave-one-study-out (LOSO) settings of the same biomarker dataset. In LOSO evaluation, $MV_{Hybrid}$ achieves 57% higher correlation than the best-performing ViT and shows 43% smaller performance degradation compared to random split in gene expression prediction, demonstrating superior performance and robustness, respectively. Furthermore, $MV_{Hybrid}$ shows equal or better downstream performance in classification, patch retrieval, and survival prediction tasks compared to that of ViT, showing its promise as a next-generation pathology VFM backbone. Our code is publicly available at: https://github.com/deepnoid-ai/MVHybrid.", "date": "2025-08-01", "categories": []}
{"id": "2506.10006", "title": "HER2 Expression Prediction with Flexible Multi-Modal Inputs via Dynamic Bidirectional Reconstruction", "abstract": "In breast cancer HER2 assessment, clinical evaluation relies on combined H&E and IHC images, yet acquiring both modalities is often hindered by clinical constraints and cost. We propose an adaptive bimodal prediction framework that flexibly supports single- or dual-modality inputs through two core innovations: a dynamic branch selector activating modality completion or joint inference based on input availability, and a cross-modal GAN (CM-GAN) enabling feature-space reconstruction of missing modalities. This design dramatically improves H&E-only accuracy from 71.44% to 94.25%, achieves 95.09% with full dual-modality inputs, and maintains 90.28% reliability under single-modality conditions. The \"dual-modality preferred, single-modality compatible\" architecture delivers near-dual-modality accuracy without mandatory synchronized acquisition, offering a cost-effective solution for resource-limited regions and significantly improving HER2 assessment accessibility.", "date": "2025-04-12", "categories": []}
{"id": "2507.21912", "title": "Predict Patient Self-reported Race from Skin Histological Images", "abstract": "Artificial Intelligence (AI) has demonstrated success in computational pathology (CPath) for disease detection, biomarker classification, and prognosis prediction. However, its potential to learn unintended demographic biases, particularly those related to social determinants of health, remains understudied. This study investigates whether deep learning models can predict self-reported race from digitized dermatopathology slides and identifies potential morphological shortcuts. Using a multisite dataset with a racially diverse population, we apply an attention-based mechanism to uncover race-associated morphological features. After evaluating three dataset curation strategies to control for confounding factors, the final experiment showed that White and Black demographic groups retained high prediction performance (AUC: 0.799, 0.762), while overall performance dropped to 0.663. Attention analysis revealed the epidermis as a key predictive feature, with significant performance declines when these regions were removed. These findings highlight the need for careful data curation and bias mitigation to ensure equitable AI deployment in pathology. Code available at: https://github.com/sinai-computational-pathology/CPath_SAIF.", "date": "2025-07-29", "categories": []}
{"id": "2507.22092", "title": "Pathology Foundation Models are Scanner Sensitive: Benchmark and Mitigation with Contrastive ScanGen Loss", "abstract": "Computational pathology (CPath) has shown great potential in mining actionable insights from Whole Slide Images (WSIs). Deep Learning (DL) has been at the center of modern CPath, and while it delivers unprecedented performance, it is also known that DL may be affected by irrelevant details, such as those introduced during scanning by different commercially available scanners. This may lead to scanner bias, where the model outputs for the same tissue acquired by different scanners may vary. In turn, it hinders the trust of clinicians in CPath-based tools and their deployment in real-world clinical practices. Recent pathology Foundation Models (FMs) promise to provide better domain generalization capabilities. In this paper, we benchmark FMs using a multi-scanner dataset and show that FMs still suffer from scanner bias. Following this observation, we propose ScanGen, a contrastive loss function applied during task-specific fine-tuning that mitigates scanner bias, thereby enhancing the models' robustness to scanner variations. Our approach is applied to the Multiple Instance Learning task of Epidermal Growth Factor Receptor (EGFR) mutation prediction from H\\&E-stained WSIs in lung cancer. We observe that ScanGen notably enhances the ability to generalize across scanners, while retaining or improving the performance of EGFR mutation prediction.", "date": "2025-07-29", "categories": []}
{"id": "2507.16876", "title": "Machine learning-based multimodal prognostic models integrating pathology images and high-throughput omic data for overall survival prediction in cancer: a systematic review", "abstract": "Multimodal machine learning integrating histopathology and molecular data shows promise for cancer prognostication. We systematically reviewed studies combining whole slide images (WSIs) and high-throughput omics to predict overall survival. Searches of EMBASE, PubMed, and Cochrane CENTRAL (12/08/2024), plus citation screening, identified eligible studies. Data extraction used CHARMS; bias was assessed with PROBAST+AI; synthesis followed SWiM and PRISMA 2020. Protocol: PROSPERO (CRD42024594745).\n  Forty-eight studies (all since 2017) across 19 cancer types met criteria; all used The Cancer Genome Atlas. Approaches included regularised Cox regression (n=4), classical ML (n=13), and deep learning (n=31). Reported c-indices ranged 0.550-0.857; multimodal models typically outperformed unimodal ones. However, all studies showed unclear/high bias, limited external validation, and little focus on clinical utility.\n  Multimodal WSI-omics survival prediction is a fast-growing field with promising results but needs improved methodological rigor, broader datasets, and clinical evaluation.\n  Funded by NPIC, Leeds Teaching Hospitals NHS Trust, UK (Project 104687), supported by UKRI Industrial Strategy Challenge Fund.", "date": "2025-07-22", "categories": []}
{"id": "2505.14730", "title": "Predicting Neoadjuvant Chemotherapy Response in Triple-Negative Breast Cancer Using Pre-Treatment Histopathologic Images", "abstract": "Triple-negative breast cancer (TNBC) remains a major clinical challenge due to its aggressive behavior and lack of targeted therapies. Accurate early prediction of response to neoadjuvant chemotherapy (NACT) is essential for guiding personalized treatment strategies and improving patient outcomes. In this study, we present an attention-based multiple instance learning (MIL) framework designed to predict pathologic complete response (pCR) directly from pre-treatment hematoxylin and eosin (H&E)-stained biopsy slides. The model was trained on a retrospective in-house cohort of 174 TNBC patients and externally validated on an independent cohort (n = 30). It achieved a mean area under the curve (AUC) of 0.85 during five-fold cross-validation and 0.78 on external testing, demonstrating robust predictive performance and generalizability. To enhance model interpretability, attention maps were spatially co-registered with multiplex immuno-histochemistry (mIHC) data stained for PD-L1, CD8+ T cells, and CD163+ macrophages. The attention regions exhibited moderate spatial overlap with immune-enriched areas, with mean Intersection over Union (IoU) scores of 0.47 for PD-L1, 0.45 for CD8+ T cells, and 0.46 for CD163+ macrophages. The presence of these biomarkers in high-attention regions supports their biological relevance to NACT response in TNBC. This not only improves model interpretability but may also inform future efforts to identify clinically actionable histological biomarkers directly from H&E-stained biopsy slides, further supporting the utility of this approach for accurate NACT response prediction and advancing precision oncology in TNBC.", "date": "2025-05-20", "categories": []}
{"id": "2502.20144", "title": "Robust sensitivity control in digital pathology via tile score distribution matching", "abstract": "Deploying digital pathology models across medical centers is challenging due to distribution shifts. Recent advances in domain generalization improve model transferability in terms of aggregated performance measured by the Area Under Curve (AUC). However, clinical regulations often require to control the transferability of other metrics, such as prescribed sensitivity levels. We introduce a novel approach to control the sensitivity of whole slide image (WSI) classification models, based on optimal transport and Multiple Instance Learning (MIL). Validated across multiple cohorts and tasks, our method enables robust sensitivity control with only a handful of calibration samples, providing a practical solution for reliable deployment of computational pathology systems.", "date": "2025-02-27", "categories": []}
{"id": "2507.16855", "title": "A tissue and cell-level annotated H&E and PD-L1 histopathology image dataset in non-small cell lung cancer", "abstract": "The tumor immune microenvironment (TIME) in non-small cell lung cancer (NSCLC) histopathology contains morphological and molecular characteristics predictive of immunotherapy response. Computational quantification of TIME characteristics, such as cell detection and tissue segmentation, can support biomarker development. However, currently available digital pathology datasets of NSCLC for the development of cell detection or tissue segmentation algorithms are limited in scope, lack annotations of clinically prevalent metastatic sites, and forgo molecular information such as PD-L1 immunohistochemistry (IHC). To fill this gap, we introduce the IGNITE data toolkit, a multi-stain, multi-centric, and multi-scanner dataset of annotated NSCLC whole-slide images. We publicly release 887 fully annotated regions of interest from 155 unique patients across three complementary tasks: (i) multi-class semantic segmentation of tissue compartments in H&E-stained slides, with 16 classes spanning primary and metastatic NSCLC, (ii) nuclei detection, and (iii) PD-L1 positive tumor cell detection in PD-L1 IHC slides. To the best of our knowledge, this is the first public NSCLC dataset with manual annotations of H&E in metastatic sites and PD-L1 IHC.", "date": "2025-07-21", "categories": []}
{"id": "2410.19166", "title": "DCT-HistoTransformer: Efficient Lightweight Vision Transformer with DCT Integration for histopathological image analysis", "abstract": "In recent years, the integration of advanced imaging techniques and deep learning methods has significantly advanced computer-aided diagnosis (CAD) systems for breast cancer detection and classification. Transformers, which have shown great promise in computer vision, are now being applied to medical image analysis. However, their application to histopathological images presents challenges due to the need for extensive manual annotations of whole-slide images (WSIs), as these models require large amounts of data to work effectively, which is costly and time-consuming. Furthermore, the quadratic computational cost of Vision Transformers (ViTs) is particularly prohibitive for large, high-resolution histopathological images, especially on edge devices with limited computational resources. In this study, we introduce a novel lightweight breast cancer classification approach using transformers that operates effectively without large datasets. By incorporating parallel processing pathways for Discrete Cosine Transform (DCT) Attention and MobileConv, we convert image data from the spatial domain to the frequency domain to utilize the benefits such as filtering out high frequencies in the image, which reduces computational cost. This demonstrates the potential of our approach to improve breast cancer classification in histopathological images, offering a more efficient solution with reduced reliance on extensive annotated datasets. Our proposed model achieves an accuracy of 96.00% $\\pm$ 0.48% for binary classification and 87.85% $\\pm$ 0.93% for multiclass classification, which is comparable to state-of-the-art models while significantly reducing computational costs. This demonstrates the potential of our approach to improve breast cancer classification in histopathological images, offering a more efficient solution with reduced reliance on extensive annotated datasets.", "date": "2024-10-24", "categories": []}
{"id": "2506.18668", "title": "Benchmarking histopathology foundation models in a multi-center dataset for skin cancer subtyping", "abstract": "Pretraining on large-scale, in-domain datasets grants histopathology foundation models (FM) the ability to learn task-agnostic data representations, enhancing transfer learning on downstream tasks. In computational pathology, automated whole slide image analysis requires multiple instance learning (MIL) frameworks due to the gigapixel scale of the slides. The diversity among histopathology FMs has highlighted the need to design real-world challenges for evaluating their effectiveness. To bridge this gap, our work presents a novel benchmark for evaluating histopathology FMs as patch-level feature extractors within a MIL classification framework. For that purpose, we leverage the AI4SkIN dataset, a multi-center cohort encompassing slides with challenging cutaneous spindle cell neoplasm subtypes. We also define the Foundation Model - Silhouette Index (FM-SI), a novel metric to measure model consistency against distribution shifts. Our experimentation shows that extracting less biased features enhances classification performance, especially in similarity-based MIL classifiers.", "date": "2025-06-23", "categories": []}
{"id": "2507.14271", "title": "MiDeSeC: A Dataset for Mitosis Detection and Segmentation in Breast Cancer Histopathology Images", "abstract": "The MiDeSeC dataset is created through H&E stained invasive breast carcinoma, no special type (NST) slides of 25 different patients captured at 40x magnification from the Department of Medical Pathology at Ankara University. The slides have been scanned by 3D Histech Panoramic p250 Flash-3 scanner and Olympus BX50 microscope. As several possible mitosis shapes exist, it is crucial to have a large dataset to cover all the cases. Accordingly, a total of 50 regions is selected from glass slides for 25 patients, each of regions with a size of 1024*1024 pixels. There are more than 500 mitoses in total in these 50 regions. Two-thirds of the regions are reserved for training, the other third for testing.", "date": "2025-07-18", "categories": []}
{"id": "2507.14272", "title": "NuSeC: A Dataset for Nuclei Segmentation in Breast Cancer Histopathology Images", "abstract": "The NuSeC dataset is created by selecting 4 images with the size of 1024*1024 pixels from the slides of each patient among 25 patients. Therefore, there are a total of 100 images in the NuSeC dataset. To carry out a consistent comparative analysis between the methods that will be developed using the NuSeC dataset by the researchers in the future, we divide the NuSeC dataset 75% as the training set and 25% as the testing set. In detail, an image is randomly selected from 4 images of each patient among 25 patients to build the testing set, and then the remaining images are reserved for the training set. While the training set includes 75 images with around 30000 nuclei structures, the testing set includes 25 images with around 6000 nuclei structures.", "date": "2025-07-18", "categories": []}
{"id": "2405.09298", "title": "A Mixture of Experts (MoE) model to improve AI-based computational pathology prediction performance under variable levels of histopathology image blur", "abstract": "AI-based models for histopathology whole slide image (WSI) analysis are increasingly common, but unsharp or blurred areas within WSI can significantly reduce prediction performance. In this study, we investigated the effect of image blur on deep learning models and introduced a mixture of experts (MoE) strategy that combines predictions from multiple expert models trained on data with varying blur levels. Using H&E-stained WSIs from 2,093 breast cancer patients, we benchmarked performance on grade classification and IHC biomarker prediction with both CNN- (CNN_CLAM and MoE-CNN_CLAM) and Vision Transformer-based (UNI_CLAM and MoE-UNI_CLAM) models. Our results show that baseline models' performance consistently decreased with increasing blur, but expert models trained on blurred tiles and especially our proposed MoE approach substantially improved performance, and outperformed baseline models in a range of simulated scenarios. MoE-CNN_CLAM outperformed the baseline CNN_CLAM under moderate (AUC: 0.868 vs. 0.702) and mixed blur conditions (AUC: 0.890 vs. 0.875). MoE-UNI_CLAM outperformed the baseline UNI_CLAM model in both moderate (AUC: 0.950 vs. 0.928) and mixed blur conditions (AUC: 0.944 vs. 0.931). This MoE method has the potential to enhance the reliability of AI-based pathology models under variable image quality, supporting broader application in both research and clinical settings.", "date": "2024-05-15", "categories": []}
{"id": "2412.16425", "title": "Patherea: Cell Detection and Classification for the 2020s", "abstract": "We present Patherea, a unified framework for point-based cell detection and classification that enables the development and fair evaluation of state-of-the-art methods. To support this, we introduce a large-scale dataset that replicates the clinical workflow for Ki-67 proliferation index estimation. Our method directly predicts cell locations and classes without relying on intermediate representations. It incorporates a hybrid Hungarian matching strategy for accurate point assignment and supports flexible backbones and training regimes, including recent pathology foundation models. Patherea achieves state-of-the-art performance on public datasets - Lizard, BRCA-M2C, and BCData - while highlighting performance saturation on these benchmarks. In contrast, our newly proposed Patherea dataset presents a significantly more challenging benchmark. Additionally, we identify and correct common errors in current evaluation protocols and provide an updated benchmarking utility for standardized assessment. The Patherea dataset and code are publicly available to facilitate further research and fair comparisons.", "date": "2024-12-21", "categories": []}
{"id": "2507.10737", "title": "Integrating Biological Knowledge for Robust Microscopy Image Profiling on De Novo Cell Lines", "abstract": "High-throughput screening techniques, such as microscopy imaging of cellular responses to genetic and chemical perturbations, play a crucial role in drug discovery and biomedical research. However, robust perturbation screening for \\textit{de novo} cell lines remains challenging due to the significant morphological and biological heterogeneity across cell lines. To address this, we propose a novel framework that integrates external biological knowledge into existing pretraining strategies to enhance microscopy image profiling models. Our approach explicitly disentangles perturbation-specific and cell line-specific representations using external biological information. Specifically, we construct a knowledge graph leveraging protein interaction data from STRING and Hetionet databases to guide models toward perturbation-specific features during pretraining. Additionally, we incorporate transcriptomic features from single-cell foundation models to capture cell line-specific representations. By learning these disentangled features, our method improves the generalization of imaging models to \\textit{de novo} cell lines. We evaluate our framework on the RxRx database through one-shot fine-tuning on an RxRx1 cell line and few-shot fine-tuning on cell lines from the RxRx19a dataset. Experimental results demonstrate that our method enhances microscopy image profiling for \\textit{de novo} cell lines, highlighting its effectiveness in real-world phenotype-based drug discovery applications.", "date": "2025-07-14", "categories": []}
{"id": "2408.09554", "title": "Screen Them All: High-Throughput Pan-Cancer Genetic and Phenotypic Biomarker Screening from H&E Whole Slide Images", "abstract": "Molecular assays are standard of care for detecting genomic alterations in cancer prognosis and therapy selection but are costly, tissue-destructive and time-consuming. Artificial intelligence (AI) applied to routine hematoxylin and eosin (H&E)-stained whole slide images (WSIs) offers a fast and economical alternative for screening molecular biomarkers. We introduce OmniScreen, a high-throughput AI-based system leveraging Virchow2 embeddings extracted from 60,529 cancer patients with paired 489-gene MSK-IMPACT targeted biomarker panel and WSIs. Unlike conventional approaches that train separate models for each biomarker, OmniScreen employs a unified model to predict a broad range of clinically relevant biomarkers across cancers, including low-prevalence targets impractical to model individually. OmniScreen reliably identifies therapeutic targets and shared phenotypic features across common and rare tumors. We investigate the biomarker prediction probabilities and accuracies of OmniScreen in relation to tumor area, cohort size, histologic subtype alignment, and pathway-level morphological patterns. These findings underscore the potential of OmniScreen for routine clinical screening.", "date": "2024-08-18", "categories": []}
{"id": "2503.20653", "title": "UWarp: A Whole Slide Image Registration Pipeline to Characterize Scanner-Induced Local Domain Shift", "abstract": "Histopathology slide digitization introduces scanner-induced domain shift that can significantly impact computational pathology models based on deep learning methods. In the state-of-the-art, this shift is often characterized at a broad scale (slide-level or dataset-level) but not patch-level, which limits our comprehension of the impact of localized tissue characteristics on the accuracy of the deep learning models. To address this challenge, we present a domain shift analysis framework based on UWarp, a novel registration tool designed to accurately align histological slides scanned under varying conditions. UWarp employs a hierarchical registration approach, combining global affine transformations with fine-grained local corrections to achieve robust tissue patch alignment. We evaluate UWarp using two private datasets, CypathLung and BosomShieldBreast, containing whole slide images scanned by multiple devices. Our experiments demonstrate that UWarp outperforms existing open-source registration methods, achieving a median target registration error (TRE) of less than 4 pixels (<1 micrometer at 40x magnification) while significantly reducing computational time. Additionally, we apply UWarp to characterize scanner-induced local domain shift in the predictions of Breast-NEOprAIdict, a deep learning model for breast cancer pathological response prediction. We find that prediction variability is strongly correlated with tissue density on a given patch. Our findings highlight the importance of localized domain shift analysis and suggest that UWarp can serve as a valuable tool for improving model robustness and domain adaptation strategies in computational pathology.", "date": "2025-03-26", "categories": []}
{"id": "2507.05656", "title": "ADPv2: A Hierarchical Histological Tissue Type-Annotated Dataset for Potential Biomarker Discovery of Colorectal Disease", "abstract": "Computational pathology (CoPath) leverages histopathology images to enhance diagnostic precision and reproducibility in clinical pathology. However, publicly available datasets for CoPath that are annotated with extensive histological tissue type (HTT) taxonomies at a granular level remain scarce due to the significant expertise and high annotation costs required. Existing datasets, such as the Atlas of Digital Pathology (ADP), address this by offering diverse HTT annotations generalized to multiple organs, but limit the capability for in-depth studies on specific organ diseases. Building upon this foundation, we introduce ADPv2, a novel dataset focused on gastrointestinal histopathology. Our dataset comprises 20,004 image patches derived from healthy colon biopsy slides, annotated according to a hierarchical taxonomy of 32 distinct HTTs of 3 levels. Furthermore, we train a multilabel representation learning model following a two-stage training procedure on our ADPv2 dataset. We leverage the VMamba architecture and achieving a mean average precision (mAP) of 0.88 in multilabel classification of colon HTTs. Finally, we show that our dataset is capable of an organ-specific in-depth study for potential biomarker discovery by analyzing the model's prediction behavior on tissues affected by different colon diseases, which reveals statistical patterns that confirm the two pathological pathways of colon cancer development. Our dataset is publicly available at https://zenodo.org/records/15307021", "date": "2025-07-08", "categories": []}
{"id": "2507.07006", "title": "GNN-ViTCap: GNN-Enhanced Multiple Instance Learning with Vision Transformers for Whole Slide Image Classification and Captioning", "abstract": "Microscopic assessment of histopathology images is vital for accurate cancer diagnosis and treatment. Whole Slide Image (WSI) classification and captioning have become crucial tasks in computer-aided pathology. However, microscopic WSI face challenges such as redundant patches and unknown patch positions due to subjective pathologist captures. Moreover, generating automatic pathology captions remains a significant challenge. To address these issues, we introduce a novel GNN-ViTCap framework for classification and caption generation from histopathological microscopic images. First, a visual feature extractor generates patch embeddings. Redundant patches are then removed by dynamically clustering these embeddings using deep embedded clustering and selecting representative patches via a scalar dot attention mechanism. We build a graph by connecting each node to its nearest neighbors in the similarity matrix and apply a graph neural network to capture both local and global context. The aggregated image embeddings are projected into the language model's input space through a linear layer and combined with caption tokens to fine-tune a large language model. We validate our method on the BreakHis and PatchGastric datasets. GNN-ViTCap achieves an F1 score of 0.934 and an AUC of 0.963 for classification, along with a BLEU-4 score of 0.811 and a METEOR score of 0.569 for captioning. Experimental results demonstrate that GNN-ViTCap outperforms state of the art approaches, offering a reliable and efficient solution for microscopy based patient diagnosis.", "date": "2025-07-09", "categories": []}
{"id": "2507.07013", "title": "Integrating Pathology Foundation Models and Spatial Transcriptomics for Cellular Decomposition from Histology Images", "abstract": "The rapid development of digital pathology and modern deep learning has facilitated the emergence of pathology foundation models that are expected to solve general pathology problems under various disease conditions in one unified model, with or without fine-tuning. In parallel, spatial transcriptomics has emerged as a transformative technology that enables the profiling of gene expression on hematoxylin and eosin (H&E) stained histology images. Spatial transcriptomics unlocks the unprecedented opportunity to dive into existing histology images at a more granular, cellular level. In this work, we propose a lightweight and training-efficient approach to predict cellular composition directly from H&E-stained histology images by leveraging information-enriched feature embeddings extracted from pre-trained pathology foundation models. By training a lightweight multi-layer perceptron (MLP) regressor on cell-type abundances derived via cell2location, our method efficiently distills knowledge from pathology foundation models and demonstrates the ability to accurately predict cell-type compositions from histology images, without physically performing the costly spatial transcriptomics. Our method demonstrates competitive performance compared to existing methods such as Hist2Cell, while significantly reducing computational complexity.", "date": "2025-07-09", "categories": []}
{"id": "2507.05383", "title": "Foreground-aware Virtual Staining for Accurate 3D Cell Morphological Profiling", "abstract": "Microscopy enables direct observation of cellular morphology in 3D, with transmitted-light methods offering low-cost, minimally invasive imaging and fluorescence microscopy providing specificity and contrast. Virtual staining combines these strengths by using machine learning to predict fluorescence images from label-free inputs. However, training of existing methods typically relies on loss functions that treat all pixels equally, thus reproducing background noise and artifacts instead of focusing on biologically meaningful signals. We introduce Spotlight, a simple yet powerful virtual staining approach that guides the model to focus on relevant cellular structures. Spotlight uses histogram-based foreground estimation to mask pixel-wise loss and to calculate a Dice loss on soft-thresholded predictions for shape-aware learning. Applied to a 3D benchmark dataset, Spotlight improves morphological representation while preserving pixel-level accuracy, resulting in virtual stains better suited for downstream tasks such as segmentation and profiling.", "date": "2025-07-07", "categories": []}
{"id": "2407.17157", "title": "Establishing Causal Relationship Between Whole Slide Image Predictions and Diagnostic Evidence Subregions in Deep Learning", "abstract": "Due to the lack of fine-grained annotation guidance, current Multiple Instance Learning (MIL) struggles to establish a robust causal relationship between Whole Slide Image (WSI) diagnosis and evidence sub-images, just like fully supervised learning. So many noisy images can undermine the network's prediction. The proposed Causal Inference Multiple Instance Learning (CI-MIL), uses out-of-distribution generalization to reduce the recognition confusion of sub-images by MIL network, without requiring pixelwise annotations. Specifically, feature distillation is introduced to roughly identify the feature representation of lesion patches. Then, in the random Fourier feature space, these features are re-weighted to minimize the cross-correlation, effectively correcting the feature distribution deviation. These processes reduce the uncertainty when tracing the prediction results back to patches. Predicted diagnoses are more direct and reliable because the causal relationship between them and diagnostic evidence images is more clearly recognized by the network. Experimental results demonstrate that CI-MIL outperforms state-of-the-art methods, achieving 92.25% accuracy and 95.28% AUC on the Camelyon16 dataset (breast cancer), while 94.29% accuracy and 98.07% AUC on the TCGA-NSCLC dataset (non-small cell lung cancer). Additionally, CI-MIL exhibits superior interpretability, as its selected regions demonstrate high consistency with ground truth annotations, promising more reliable diagnostic assistance for pathologists.", "date": "2024-07-24", "categories": []}
{"id": "2407.19821", "title": "Distilling High Diagnostic Value Patches for Whole Slide Image Classification Using Attention Mechanism", "abstract": "Multiple Instance Learning (MIL) has garnered widespread attention in the field of Whole Slide Image (WSI) classification as it replaces pixel-level manual annotation with diagnostic reports as labels, significantly reducing labor costs. Recent research has shown that bag-level MIL methods often yield better results because they can consider all patches of the WSI as a whole. However, a drawback of such methods is the incorporation of more redundant patches, leading to interference. To extract patches with high diagnostic value while excluding interfering patches to address this issue, we developed an attention-based feature distillation multi-instance learning (AFD-MIL) approach. This approach proposed the exclusion of redundant patches as a preprocessing operation in weakly supervised learning, directly mitigating interference from extensive noise. It also pioneers the use of attention mechanisms to distill features with high diagnostic value, as opposed to the traditional practice of indiscriminately and forcibly integrating all patches. Additionally, we introduced global loss optimization to finely control the feature distillation module. AFD-MIL is orthogonal to many existing MIL methods, leading to consistent performance improvements. This approach has surpassed the current state-of-the-art method, achieving 91.47% ACC (accuracy) and 94.29% AUC (area under the curve) on the Camelyon16 (Camelyon Challenge 2016, breast cancer), while 93.33% ACC and 98.17% AUC on the TCGA-NSCLC (The Cancer Genome Atlas Program: non-small cell lung cancer). Different feature distillation methods were used for the two datasets, tailored to the specific diseases, thereby improving performance and interpretability.", "date": "2024-07-29", "categories": []}
{"id": "2507.04660", "title": "CP-Dilatation: A Copy-and-Paste Augmentation Method for Preserving the Boundary Context Information of Histopathology Images", "abstract": "Medical AI diagnosis including histopathology segmentation has derived benefits from the recent development of deep learning technology. However, deep learning itself requires a large amount of training data and the medical image segmentation masking, in particular, requires an extremely high cost due to the shortage of medical specialists. To mitigate this issue, we propose a new data augmentation method built upon the conventional Copy and Paste (CP) augmentation technique, called CP-Dilatation, and apply it to histopathology image segmentation. To the well-known traditional CP technique, the proposed method adds a dilation operation that can preserve the boundary context information of the malignancy, which is important in histopathological image diagnosis, as the boundary between the malignancy and its margin is mostly unclear and a significant context exists in the margin. In our experiments using histopathology benchmark datasets, the proposed method was found superior to the other state-of-the-art baselines chosen for comparison.", "date": "2025-07-07", "categories": []}
{"id": "2507.00292", "title": "Reducing Variability of Multiple Instance Learning Methods for Digital Pathology", "abstract": "Digital pathology has revolutionized the field by enabling the digitization of tissue samples into whole slide images (WSIs). However, the high resolution and large size of WSIs present significant challenges when it comes to applying Deep Learning models. As a solution, WSIs are often divided into smaller patches with a global label (\\textit{i.e., diagnostic}) per slide, instead of a (too) costly pixel-wise annotation. By treating each slide as a bag of patches, Multiple Instance Learning (MIL) methods have emerged as a suitable solution for WSI classification. A major drawback of MIL methods is their high variability in performance across different runs, which can reach up to 10-15 AUC points on the test set, making it difficult to compare different MIL methods reliably. This variability mainly comes from three factors: i) weight initialization, ii) batch (shuffling) ordering, iii) and learning rate. To address that, we introduce a Multi-Fidelity, Model Fusion strategy for MIL methods. We first train multiple models for a few epochs and average the most stable and promising ones based on validation scores. This approach can be applied to any existing MIL model to reduce performance variability. It also simplifies hyperparameter tuning and improves reproducibility while maintaining computational efficiency. We extensively validate our approach on WSI classification tasks using 2 different datasets, 3 initialization strategies and 5 MIL methods, for a total of more than 2000 experiments.", "date": "2025-06-30", "categories": []}
{"id": "2502.16889", "title": "Beyond Diagnostic Performance: Revealing and Quantifying Ethical Risks in Pathology Foundation Models", "abstract": "Pathology foundation models (PFMs), as large-scale pre-trained models tailored for computational pathology, have significantly advanced a wide range of applications. Their ability to leverage prior knowledge from massive datasets has streamlined the development of intelligent pathology models. However, we identify several critical and interrelated ethical risks that remain underexplored, yet must be addressed to enable the safe translation of PFMs from lab to clinic. These include the potential leakage of patient-sensitive attributes, disparities in model performance across demographic and institutional subgroups, and the reliance on diagnosis-irrelevant features that undermine clinical reliability. In this study, we pioneer the quantitative analysis for ethical risks in PFMs, including privacy leakage, clinical reliability, and group fairness. Specifically, we propose an evaluation framework that systematically measures key dimensions of ethical concern: the degree to which patient-sensitive attributes can be inferred from model representations, the extent of performance disparities across demographic and institutional subgroups, and the influence of diagnostically irrelevant features on model decisions. We further investigate the underlying causes of these ethical risks in PFMs and empirically validate our findings. Then we offer insights into potential directions for mitigating such risks, aiming to inform the development of more ethically robust PFMs. This work provides the first quantitative and systematic evaluation of ethical risks in PFMs. Our findings highlight the urgent need for ethical safeguards in PFMs and offer actionable insights for building more trustworthy and clinically robust PFMs. To facilitate future research and deployment, we will release the assessment framework as an online toolkit to support the development, auditing, and deployment of ethically robust PFMs.", "date": "2025-02-24", "categories": []}
{"id": "2507.00044", "title": "HistoART: Histopathology Artifact Detection and Reporting Tool", "abstract": "In modern cancer diagnostics, Whole Slide Imaging (WSI) is widely used to digitize tissue specimens for detailed, high-resolution examination; however, other diagnostic approaches, such as liquid biopsy and molecular testing, are also utilized based on the cancer type and clinical context. While WSI has revolutionized digital histopathology by enabling automated, precise analysis, it remains vulnerable to artifacts introduced during slide preparation and scanning. These artifacts can compromise downstream image analysis. To address this challenge, we propose and compare three robust artifact detection approaches for WSIs: (1) a foundation model-based approach (FMA) using a fine-tuned Unified Neural Image (UNI) architecture, (2) a deep learning approach (DLA) built on a ResNet50 backbone, and (3) a knowledge-based approach (KBA) leveraging handcrafted features from texture, color, and frequency-based metrics. The methods target six common artifact types: tissue folds, out-of-focus regions, air bubbles, tissue damage, marker traces, and blood contamination. Evaluations were conducted on 50,000+ image patches from diverse scanners (Hamamatsu, Philips, Leica Aperio AT2) across multiple sites. The FMA achieved the highest patch-wise AUROC of 0.995 (95% CI [0.994, 0.995]), outperforming the ResNet50-based method (AUROC: 0.977, 95% CI [0.977, 0.978]) and the KBA (AUROC: 0.940, 95% CI [0.933, 0.946]). To translate detection into actionable insights, we developed a quality report scorecard that quantifies high-quality patches and visualizes artifact distributions.", "date": "2025-06-23", "categories": []}
{"id": "2506.23184", "title": "Score-based Diffusion Model for Unpaired Virtual Histology Staining", "abstract": "Hematoxylin and eosin (H&E) staining visualizes histology but lacks specificity for diagnostic markers. Immunohistochemistry (IHC) staining provides protein-targeted staining but is restricted by tissue availability and antibody specificity. Virtual staining, i.e., computationally translating the H&E image to its IHC counterpart while preserving the tissue structure, is promising for efficient IHC generation. Existing virtual staining methods still face key challenges: 1) effective decomposition of staining style and tissue structure, 2) controllable staining process adaptable to diverse tissue and proteins, and 3) rigorous structural consistency modelling to handle the non-pixel-aligned nature of paired H&E and IHC images. This study proposes a mutual-information (MI)-guided score-based diffusion model for unpaired virtual staining. Specifically, we design 1) a global MI-guided energy function that disentangles the tissue structure and staining characteristics across modalities, 2) a novel timestep-customized reverse diffusion process for precise control of the staining intensity and structural reconstruction, and 3) a local MI-driven contrastive learning strategy to ensure the cellular level structural consistency between H&E-IHC images. Extensive experiments demonstrate the our superiority over state-of-the-art approaches, highlighting its biomedical potential. Codes will be open-sourced upon acceptance.", "date": "2025-06-29", "categories": []}
{"id": "2506.23827", "title": "Spatially Gene Expression Prediction using Dual-Scale Contrastive Learning", "abstract": "Spatial transcriptomics (ST) provides crucial insights into tissue micro-environments, but is limited to its high cost and complexity. As an alternative, predicting gene expression from pathology whole slide images (WSI) is gaining increasing attention. However, existing methods typically rely on single patches or a single pathology modality, neglecting the complex spatial and molecular interactions between target and neighboring information (e.g., gene co-expression). This leads to a failure in establishing connections among adjacent regions and capturing intricate cross-modal relationships. To address these issues, we propose NH2ST, a framework that integrates spatial context and both pathology and gene modalities for gene expression prediction. Our model comprises a query branch and a neighbor branch to process paired target patch and gene data and their neighboring regions, where cross-attention and contrastive learning are employed to capture intrinsic associations and ensure alignments between pathology and gene expression. Extensive experiments on six datasets demonstrate that our model consistently outperforms existing methods, achieving over 20% in PCC metrics. Codes are available at https://github.com/MCPathology/NH2ST", "date": "2025-06-30", "categories": []}
{"id": "2503.19367", "title": "VGAT: A Cancer Survival Analysis Framework Transitioning from Generative Visual Question Answering to Genomic Reconstruction", "abstract": "Multimodal learning combining pathology images and genomic sequences enhances cancer survival analysis but faces clinical implementation barriers due to limited access to genomic sequencing in under-resourced regions. To enable survival prediction using only whole-slide images (WSI), we propose the Visual-Genomic Answering-Guided Transformer (VGAT), a framework integrating Visual Question Answering (VQA) techniques for genomic modality reconstruction. By adapting VQA's text feature extraction approach, we derive stable genomic representations that circumvent dimensionality challenges in raw genomic data. Simultaneously, a cluster-based visual prompt module selectively enhances discriminative WSI patches, addressing noise from unfiltered image regions. Evaluated across five TCGA datasets, VGAT outperforms existing WSI-only methods, demonstrating the viability of genomic-informed inference without sequencing. This approach bridges multimodal research and clinical feasibility in resource-constrained settings. The code link is https://github.com/CZZZZZZZZZZZZZZZZZ/VGAT.", "date": "2025-03-25", "categories": []}
{"id": "2506.20741", "title": "OTSurv: A Novel Multiple Instance Learning Framework for Survival Prediction with Heterogeneity-aware Optimal Transport", "abstract": "Survival prediction using whole slide images (WSIs) can be formulated as a multiple instance learning (MIL) problem. However, existing MIL methods often fail to explicitly capture pathological heterogeneity within WSIs, both globally -- through long-tailed morphological distributions, and locally through -- tile-level prediction uncertainty. Optimal transport (OT) provides a principled way of modeling such heterogeneity by incorporating marginal distribution constraints. Building on this insight, we propose OTSurv, a novel MIL framework from an optimal transport perspective. Specifically, OTSurv formulates survival predictions as a heterogeneity-aware OT problem with two constraints: (1) global long-tail constraint that models prior morphological distributions to avert both mode collapse and excessive uniformity by regulating transport mass allocation, and (2) local uncertainty-aware constraint that prioritizes high-confidence patches while suppressing noise by progressively raising the total transport mass. We then recast the initial OT problem, augmented by these constraints, into an unbalanced OT formulation that can be solved with an efficient, hardware-friendly matrix scaling algorithm. Empirically, OTSurv sets new state-of-the-art results across six popular benchmarks, achieving an absolute 3.6% improvement in average C-index. In addition, OTSurv achieves statistical significance in log-rank tests and offers high interpretability, making it a powerful tool for survival prediction in digital pathology. Our codes are available at https://github.com/Y-Research-SBU/OTSurv.", "date": "2025-06-25", "categories": []}
{"id": "2410.14388", "title": "Unscrambling disease progression at scale: fast inference of event permutations with optimal transport", "abstract": "Disease progression models infer group-level temporal trajectories of change in patients' features as a chronic degenerative condition plays out. They provide unique insight into disease biology and staging systems with individual-level clinical utility. Discrete models consider disease progression as a latent permutation of events, where each event corresponds to a feature becoming measurably abnormal. However, permutation inference using traditional maximum likelihood approaches becomes prohibitive due to combinatoric explosion, severely limiting model dimensionality and utility. Here we leverage ideas from optimal transport to model disease progression as a latent permutation matrix of events belonging to the Birkhoff polytope, facilitating fast inference via optimisation of the variational lower bound. This enables a factor of 1000 times faster inference than the current state of the art and, correspondingly, supports models with several orders of magnitude more features than the current state of the art can consider. Experiments demonstrate the increase in speed, accuracy and robustness to noise in simulation. Further experiments with real-world imaging data from two separate datasets, one from Alzheimer's disease patients, the other age-related macular degeneration, showcase, for the first time, pixel-level disease progression events in the brain and eye, respectively. Our method is low compute, interpretable and applicable to any progressive condition and data modality, giving it broad potential clinical utility.", "date": "2024-10-18", "categories": []}
{"id": "2506.19106", "title": "Staining normalization in histopathology: Method benchmarking using multicenter dataset", "abstract": "Hematoxylin and Eosin (H&E) has been the gold standard in tissue analysis for decades, however, tissue specimens stained in different laboratories vary, often significantly, in appearance. This variation poses a challenge for both pathologists' and AI-based downstream analysis. Minimizing stain variation computationally is an active area of research. To further investigate this problem, we collected a unique multi-center tissue image dataset, wherein tissue samples from colon, kidney, and skin tissue blocks were distributed to 66 different labs for routine H&E staining. To isolate staining variation, other factors affecting the tissue appearance were kept constant. Further, we used this tissue image dataset to compare the performance of eight different stain normalization methods, including four traditional methods, namely, histogram matching, Macenko, Vahadane, and Reinhard normalization, and two deep learning-based methods namely CycleGAN and Pixp2pix, both with two variants each. We used both quantitative and qualitative evaluation to assess the performance of these methods. The dataset's inter-laboratory staining variation could also guide strategies to improve model generalizability through varied training data", "date": "2025-06-23", "categories": []}
{"id": "2506.19234", "title": "Quantitative Benchmarking of Anomaly Detection Methods in Digital Pathology", "abstract": "Anomaly detection has been widely studied in the context of industrial defect inspection, with numerous methods developed to tackle a range of challenges. In digital pathology, anomaly detection holds significant potential for applications such as rare disease identification, artifact detection, and biomarker discovery. However, the unique characteristics of pathology images, such as their large size, multi-scale structures, stain variability, and repetitive patterns, introduce new challenges that current anomaly detection algorithms struggle to address. In this quantitative study, we benchmark over 20 classical and prevalent anomaly detection methods through extensive experiments. We curated five digital pathology datasets, both real and synthetic, to systematically evaluate these approaches. Our experiments investigate the influence of image scale, anomaly pattern types, and training epoch selection strategies on detection performance. The results provide a detailed comparison of each method's strengths and limitations, establishing a comprehensive benchmark to guide future research in anomaly detection for digital pathology images.", "date": "2025-06-24", "categories": []}
{"id": "2506.19681", "title": "Genome-Anchored Foundation Model Embeddings Improve Molecular Prediction from Histology Images", "abstract": "Precision oncology requires accurate molecular insights, yet obtaining these directly from genomics is costly and time-consuming for broad clinical use. Predicting complex molecular features and patient prognosis directly from routine whole-slide images (WSI) remains a major challenge for current deep learning methods. Here we introduce PathLUPI, which uses transcriptomic privileged information during training to extract genome-anchored histological embeddings, enabling effective molecular prediction using only WSIs at inference. Through extensive evaluation across 49 molecular oncology tasks using 11,257 cases among 20 cohorts, PathLUPI demonstrated superior performance compared to conventional methods trained solely on WSIs. Crucially, it achieves AUC $\\geq$ 0.80 in 14 of the biomarker prediction and molecular subtyping tasks and C-index $\\geq$ 0.70 in survival cohorts of 5 major cancer types. Moreover, PathLUPI embeddings reveal distinct cellular morphological signatures associated with specific genotypes and related biological pathways within WSIs. By effectively encoding molecular context to refine WSI representations, PathLUPI overcomes a key limitation of existing models and offers a novel strategy to bridge molecular insights with routine pathology workflows for wider clinical application.", "date": "2025-06-24", "categories": []}
{"id": "2506.17645", "title": "Histopathology Image Report Generation by Vision Language Model with Multimodal In-Context Learning", "abstract": "Automating medical report generation from histopathology images is a critical challenge requiring effective visual representations and domain-specific knowledge. Inspired by the common practices of human experts, we propose an in-context learning framework called PathGenIC that integrates context derived from the training set with a multimodal in-context learning (ICL) mechanism. Our method dynamically retrieves semantically similar whole slide image (WSI)-report pairs and incorporates adaptive feedback to enhance contextual relevance and generation quality. Evaluated on the HistGen benchmark, the framework achieves state-of-the-art results, with significant improvements across BLEU, METEOR, and ROUGE-L metrics, and demonstrates robustness across diverse report lengths and disease categories. By maximizing training data utility and bridging vision and language with ICL, our work offers a solution for AI-driven histopathology reporting, setting a strong foundation for future advancements in multimodal clinical applications.", "date": "2025-06-21", "categories": []}
{"id": "2506.18523", "title": "Multi-Scale Representation of Follicular Lymphoma Pathology Images in a Single Hyperbolic Space", "abstract": "We propose a method for representing malignant lymphoma pathology images, from high-resolution cell nuclei to low-resolution tissue images, within a single hyperbolic space using self-supervised learning. To capture morphological changes that occur across scales during disease progression, our approach embeds tissue and corresponding nucleus images close to each other based on inclusion relationships. Using the Poincar\\'e ball as the feature space enables effective encoding of this hierarchical structure. The learned representations capture both disease state and cell type variations.", "date": "2025-06-23", "categories": []}
{"id": "2506.15853", "title": "Cross-Modality Learning for Predicting IHC Biomarkers from H&E-Stained Whole-Slide Images", "abstract": "Hematoxylin and Eosin (H&E) staining is a cornerstone of pathological analysis, offering reliable visualization of cellular morphology and tissue architecture for cancer diagnosis, subtyping, and grading. Immunohistochemistry (IHC) staining provides molecular insights by detecting specific proteins within tissues, enhancing diagnostic accuracy, and improving treatment planning. However, IHC staining is costly, time-consuming, and resource-intensive, requiring specialized expertise. To address these limitations, this study proposes HistoStainAlign, a novel deep learning framework that predicts IHC staining patterns directly from H&E whole-slide images (WSIs) by learning joint representations of morphological and molecular features. The framework integrates paired H&E and IHC embeddings through a contrastive training strategy, capturing complementary features across staining modalities without patch-level annotations or tissue registration. The model was evaluated on gastrointestinal and lung tissue WSIs with three commonly used IHC stains: P53, PD-L1, and Ki-67. HistoStainAlign achieved weighted F1 scores of 0.735 [95% Confidence Interval (CI): 0.670-0.799], 0.830 [95% CI: 0.772-0.886], and 0.723 [95% CI: 0.607-0.836], respectively for these three IHC stains. Embedding analyses demonstrated the robustness of the contrastive alignment in capturing meaningful cross-stain relationships. Comparisons with a baseline model further highlight the advantage of incorporating contrastive learning for improved stain pattern prediction. This study demonstrates the potential of computational approaches to serve as a pre-screening tool, helping prioritize cases for IHC staining and improving workflow efficiency.", "date": "2025-06-18", "categories": []}
{"id": "2404.18458", "title": "A robust and scalable framework for hallucination detection in virtual tissue staining and digital pathology", "abstract": "Histopathological staining of human tissue is essential for disease diagnosis. Recent advances in virtual tissue staining technologies using artificial intelligence (AI) alleviate some of the costly and tedious steps involved in traditional histochemical staining processes, permitting multiplexed staining and tissue preservation. However, potential hallucinations and artifacts in these virtually stained tissue images pose concerns, especially for the clinical uses of these approaches. Quality assessment of histology images by experts can be subjective. Here, we present an autonomous quality and hallucination assessment method, AQuA, for virtual tissue staining and digital pathology. AQuA autonomously achieves 99.8% accuracy when detecting acceptable and unacceptable virtually stained tissue images without access to histochemically stained ground truth, and presents an agreement of 98.5% with the manual assessments made by board-certified pathologists, including identifying realistic-looking images that could mislead diagnosticians. We demonstrate the wide adaptability of AQuA across various virtually and histochemically stained human tissue images. This framework enhances the reliability of virtual tissue staining and provides autonomous quality assurance for image generation and transformation tasks in digital pathology and computational imaging.", "date": "2024-04-29", "categories": []}
{"id": "2410.03289", "title": "Semantic Segmentation Based Quality Control of Histopathology Whole Slide Images", "abstract": "We developed a software pipeline for quality control (QC) of histopathology whole slide images (WSIs) that segments various regions, such as blurs of different levels, tissue regions, tissue folds, and pen marks. Given the necessity and increasing availability of GPUs for processing WSIs, the proposed pipeline comprises multiple lightweight deep learning models to strike a balance between accuracy and speed. The pipeline was evaluated in all TCGAs, which is the largest publicly available WSI dataset containing more than 11,000 histopathological images from 28 organs. It was compared to a previous work, which was not based on deep learning, and it showed consistent improvement in segmentation results across organs. To minimize annotation effort for tissue and blur segmentation, annotated images were automatically prepared by mosaicking patches (sub-images) from various WSIs whose labels were identified using a patch classification tool HistoROI. Due to the generality of our trained QC pipeline and its extensive testing the potential impact of this work is broad. It can be used for automated pre-processing any WSI cohort to enhance the accuracy and reliability of large-scale histopathology image analysis for both research and clinical use. We have made the trained models, training scripts, training data, and inference results publicly available at https://github.com/abhijeetptl5/wsisegqc, which should enable the research community to use the pipeline right out of the box or further customize it to new datasets and applications in the future.", "date": "2024-10-04", "categories": []}
{"id": "2506.10916", "title": "Semi-Automated Quality Assurance in Digital Pathology: Tile Classification Approach", "abstract": "Quality assurance is a critical but underexplored area in digital pathology, where even minor artifacts can have significant effects. Artifacts have been shown to negatively impact the performance of AI diagnostic models. In current practice, trained staff manually review digitized images prior to release of these slides to pathologists which are then used to render a diagnosis. Conventional image processing approaches, provide a foundation for detecting artifacts on digital pathology slides. However, current tools do not leverage deep learning, which has the potential to improve detection accuracy and scalability. Despite these advancements, methods for quality assurance in digital pathology remain limited, presenting a gap for innovation.\n  We propose an AI algorithm designed to screen digital pathology slides by analyzing tiles and categorizing them into one of 10 predefined artifact types or as background. This algorithm identifies and localizes artifacts, creating a map that highlights regions of interest. By directing human operators to specific tiles affected by artifacts, the algorithm minimizes the time and effort required to manually review entire slides for quality issues.\n  From internal archives and The Cancer Genome Atlas, 133 whole slide images were selected and 10 artifacts were annotated using an internally developed software ZAPP (Mayo Clinic, Jacksonville, FL). Ablation study of multiple models at different tile sizes and magnification was performed. InceptionResNet was selected. Single artifact models were trained and tested, followed by a limited multiple instance model with artifacts that performed well together (chatter, fold, and pen). From the results of this study we suggest a hybrid design for artifact screening composed of both single artifact binary models as well as multiple instance models to optimize detection of each artifact.", "date": "2025-06-12", "categories": []}
{"id": "2506.09022", "title": "Do Multiple Instance Learning Models Transfer?", "abstract": "Multiple Instance Learning (MIL) is a cornerstone approach in computational pathology (CPath) for generating clinically meaningful slide-level embeddings from gigapixel tissue images. However, MIL often struggles with small, weakly supervised clinical datasets. In contrast to fields such as NLP and conventional computer vision, where transfer learning is widely used to address data scarcity, the transferability of MIL models remains poorly understood. In this study, we systematically evaluate the transfer learning capabilities of pretrained MIL models by assessing 11 models across 21 pretraining tasks for morphological and molecular subtype prediction. Our results show that pretrained MIL models, even when trained on different organs than the target task, consistently outperform models trained from scratch. Moreover, pretraining on pancancer datasets enables strong generalization across organs and tasks, outperforming slide foundation models while using substantially less pretraining data. These findings highlight the robust adaptability of MIL models and demonstrate the benefits of leveraging transfer learning to boost performance in CPath. Lastly, we provide a resource which standardizes the implementation of MIL models and collection of pretrained model weights on popular CPath tasks, available at https://github.com/mahmoodlab/MIL-Lab", "date": "2025-06-10", "categories": []}
{"id": "2506.08854", "title": "Spatial Transcriptomics Expression Prediction from Histopathology Based on Cross-Modal Mask Reconstruction and Contrastive Learning", "abstract": "Spatial transcriptomics is a technology that captures gene expression levels at different spatial locations, widely used in tumor microenvironment analysis and molecular profiling of histopathology, providing valuable insights into resolving gene expression and clinical diagnosis of cancer. Due to the high cost of data acquisition, large-scale spatial transcriptomics data remain challenging to obtain. In this study, we develop a contrastive learning-based deep learning method to predict spatially resolved gene expression from whole-slide images. Evaluation across six different disease datasets demonstrates that, compared to existing studies, our method improves Pearson Correlation Coefficient (PCC) in the prediction of highly expressed genes, highly variable genes, and marker genes by 6.27%, 6.11%, and 11.26% respectively. Further analysis indicates that our method preserves gene-gene correlations and applies to datasets with limited samples. Additionally, our method exhibits potential in cancer tissue localization based on biomarker expression.", "date": "2025-06-10", "categories": []}
{"id": "2506.07652", "title": "FMaMIL: Frequency-Driven Mamba Multi-Instance Learning for Weakly Supervised Lesion Segmentation in Medical Images", "abstract": "Accurate lesion segmentation in histopathology images is essential for diagnostic interpretation and quantitative analysis, yet it remains challenging due to the limited availability of costly pixel-level annotations. To address this, we propose FMaMIL, a novel two-stage framework for weakly supervised lesion segmentation based solely on image-level labels. In the first stage, a lightweight Mamba-based encoder is introduced to capture long-range dependencies across image patches under the MIL paradigm. To enhance spatial sensitivity and structural awareness, we design a learnable frequency-domain encoding module that supplements spatial-domain features with spectrum-based information. CAMs generated in this stage are used to guide segmentation training. In the second stage, we refine the initial pseudo labels via a CAM-guided soft-label supervision and a self-correction mechanism, enabling robust training even under label noise. Extensive experiments on both public and private histopathology datasets demonstrate that FMaMIL outperforms state-of-the-art weakly supervised methods without relying on pixel-level annotations, validating its effectiveness and potential for digital pathology applications.", "date": "2025-06-09", "categories": []}
{"id": "2403.15238", "title": "WEEP: A method for spatial interpretation of weakly supervised CNN models in computational pathology", "abstract": "Deep learning enables the modelling of high-resolution histopathology whole-slide images (WSI). Weakly supervised learning of tile-level data is typically applied for tasks where labels only exist on the patient or WSI level (e.g. patient outcomes or histological grading). In this context, there is a need for improved spatial interpretability of predictions from such models. We propose a novel method, Wsi rEgion sElection aPproach (WEEP), for model interpretation. It provides a principled yet straightforward way to establish the spatial area of WSI required for assigning a particular prediction label. We demonstrate WEEP on a binary classification task in the area of breast cancer computational pathology. WEEP is easy to implement, is directly connected to the model-based decision process, and offers information relevant to both research and diagnostic applications.", "date": "2024-03-22", "categories": []}
{"id": "2506.05361", "title": "Scalable Generation of Spatial Transcriptomics from Histology Images via Whole-Slide Flow Matching", "abstract": "Spatial transcriptomics (ST) has emerged as a powerful technology for bridging histology imaging with gene expression profiling. However, its application has been limited by low throughput and the need for specialized experimental facilities. Prior works sought to predict ST from whole-slide histology images to accelerate this process, but they suffer from two major limitations. First, they do not explicitly model cell-cell interaction as they factorize the joint distribution of whole-slide ST data and predict the gene expression of each spot independently. Second, their encoders struggle with memory constraints due to the large number of spots (often exceeding 10,000) in typical ST datasets. Herein, we propose STFlow, a flow matching generative model that considers cell-cell interaction by modeling the joint distribution of gene expression of an entire slide. It also employs an efficient slide-level encoder with local spatial attention, enabling whole-slide processing without excessive memory overhead. On the recently curated HEST-1k and STImage-1K4M benchmarks, STFlow substantially outperforms state-of-the-art baselines and achieves over 18% relative improvements over the pathology foundation models.", "date": "2025-05-25", "categories": []}
{"id": "2506.05184", "title": "Single GPU Task Adaptation of Pathology Foundation Models for Whole Slide Image Analysis", "abstract": "Pathology foundation models (PFMs) have emerged as powerful tools for analyzing whole slide images (WSIs). However, adapting these pretrained PFMs for specific clinical tasks presents considerable challenges, primarily due to the availability of only weak (WSI-level) labels for gigapixel images, necessitating multiple instance learning (MIL) paradigm for effective WSI analysis. This paper proposes a novel approach for single-GPU \\textbf{T}ask \\textbf{A}daptation of \\textbf{PFM}s (TAPFM) that uses vision transformer (\\vit) attention for MIL aggregation while optimizing both for feature representations and attention weights. The proposed approach maintains separate computational graphs for MIL aggregator and the PFM to create stable training dynamics that align with downstream task objectives during end-to-end adaptation. Evaluated on mutation prediction tasks for bladder cancer and lung adenocarcinoma across institutional and TCGA cohorts, TAPFM consistently outperforms conventional approaches, with H-Optimus-0 (TAPFM) outperforming the benchmarks. TAPFM effectively handles multi-label classification of actionable mutations as well. Thus, TAPFM makes adaptation of powerful pre-trained PFMs practical on standard hardware for various clinical applications.", "date": "2025-06-05", "categories": []}
{"id": "2506.03185", "title": "DLiPath: A Benchmark for the Comprehensive Assessment of Donor Liver Based on Histopathological Image Dataset", "abstract": "Pathologists comprehensive evaluation of donor liver biopsies provides crucial information for accepting or discarding potential grafts. However, rapidly and accurately obtaining these assessments intraoperatively poses a significant challenge for pathologists. Features in donor liver biopsies, such as portal tract fibrosis, total steatosis, macrovesicular steatosis, and hepatocellular ballooning are correlated with transplant outcomes, yet quantifying these indicators suffers from substantial inter- and intra-observer variability. To address this, we introduce DLiPath, the first benchmark for comprehensive donor liver assessment based on a histopathology image dataset. We collected and publicly released 636 whole slide images from 304 donor liver patients at the Department of Pathology, the Third Xiangya Hospital, with expert annotations for key pathological features (including cholestasis, portal tract fibrosis, portal inflammation, total steatosis, macrovesicular steatosis, and hepatocellular ballooning). We selected nine state-of-the-art multiple-instance learning (MIL) models based on the DLiPath dataset as baselines for extensive comparative analysis. The experimental results demonstrate that several MIL models achieve high accuracy across donor liver assessment indicators on DLiPath, charting a clear course for future automated and intelligent donor liver assessment research. Data and code are available at https://github.com/panliangrui/ACM_MM_2025.", "date": "2025-05-30", "categories": []}
{"id": "2504.02628", "title": "Towards Computation- and Communication-efficient Computational Pathology", "abstract": "Despite the impressive performance across a wide range of applications, current computational pathology models face significant diagnostic efficiency challenges due to their reliance on high-magnification whole-slide image analysis. This limitation severely compromises their clinical utility, especially in time-sensitive diagnostic scenarios and situations requiring efficient data transfer. To address these issues, we present a novel computation- and communication-efficient framework called Magnification-Aligned Global-Local Transformer (MAG-GLTrans). Our approach significantly reduces computational time, file transfer requirements, and storage overhead by enabling effective analysis using low-magnification inputs rather than high-magnification ones. The key innovation lies in our proposed magnification alignment (MAG) mechanism, which employs self-supervised learning to bridge the information gap between low and high magnification levels by effectively aligning their feature representations. Through extensive evaluation across various fundamental CPath tasks, MAG-GLTrans demonstrates state-of-the-art classification performance while achieving remarkable efficiency gains: up to 10.7 times reduction in computational time and over 20 times reduction in file transfer and storage requirements. Furthermore, we highlight the versatility of our MAG framework through two significant extensions: (1) its applicability as a feature extractor to enhance the efficiency of any CPath architecture, and (2) its compatibility with existing foundation models and histopathology-specific encoders, enabling them to process low-magnification inputs with minimal information loss. These advancements position MAG-GLTrans as a particularly promising solution for time-sensitive applications, especially in the context of intraoperative frozen section diagnosis where both accuracy and efficiency are paramount.", "date": "2025-04-03", "categories": []}
{"id": "2506.02542", "title": "HIEGNet: A Heterogenous Graph Neural Network Including the Immune Environment in Glomeruli Classification", "abstract": "Graph Neural Networks (GNNs) have recently been found to excel in histopathology. However, an important histopathological task, where GNNs have not been extensively explored, is the classification of glomeruli health as an important indicator in nephropathology. This task presents unique difficulties, particularly for the graph construction, i.e., the identification of nodes, edges, and informative features. In this work, we propose a pipeline composed of different traditional and machine learning-based computer vision techniques to identify nodes, edges, and their corresponding features to form a heterogeneous graph. We then proceed to propose a novel heterogeneous GNN architecture for glomeruli classification, called HIEGNet, that integrates both glomeruli and their surrounding immune cells. Hence, HIEGNet is able to consider the immune environment of each glomerulus in its classification. Our HIEGNet was trained and tested on a dataset of Whole Slide Images from kidney transplant patients. Experimental results demonstrate that HIEGNet outperforms several baseline models and generalises best between patients among all baseline models. Our implementation is publicly available at https://github.com/nklsKrmnn/HIEGNet.git.", "date": "2025-06-03", "categories": []}
{"id": "2505.24141", "title": "The Butterfly Effect in Pathology: Exploring Security in Pathology Foundation Models", "abstract": "With the widespread adoption of pathology foundation models in both research and clinical decision support systems, exploring their security has become a critical concern. However, despite their growing impact, the vulnerability of these models to adversarial attacks remains largely unexplored. In this work, we present the first systematic investigation into the security of pathology foundation models for whole slide image~(WSI) analysis against adversarial attacks. Specifically, we introduce the principle of \\textit{local perturbation with global impact} and propose a label-free attack framework that operates without requiring access to downstream task labels. Under this attack framework, we revise four classical white-box attack methods and redefine the perturbation budget based on the characteristics of WSI. We conduct comprehensive experiments on three representative pathology foundation models across five datasets and six downstream tasks. Despite modifying only 0.1\\% of patches per slide with imperceptible noise, our attack leads to downstream accuracy degradation that can reach up to 20\\% in the worst cases. Furthermore, we analyze key factors that influence attack success, explore the relationship between patch-level vulnerability and semantic content, and conduct a preliminary investigation into potential defence strategies. These findings lay the groundwork for future research on the adversarial robustness and reliable deployment of pathology foundation models. Our code is publicly available at: https://github.com/Jiashuai-Liu-hmos/Attack-WSI-pathology-foundation-models.", "date": "2025-05-30", "categories": []}
{"id": "2504.16979", "title": "Automating tumor-infiltrating lymphocyte assessment in breast cancer histopathology images using QuPath: a transparent and accessible machine learning pipeline", "abstract": "In this study, we built an end-to-end tumor-infiltrating lymphocytes (TILs) assessment pipeline within QuPath, demonstrating the potential of easily accessible tools to perform complex tasks in a fully automatic fashion. First, we trained a pixel classifier to segment tumor, tumor-associated stroma, and other tissue compartments in breast cancer H&E-stained whole-slide images (WSI) to isolate tumor-associated stroma for subsequent analysis. Next, we applied a pre-trained StarDist deep learning model in QuPath for cell detection and used the extracted cell features to train a binary classifier distinguishing TILs from other cells. To evaluate our TILs assessment pipeline, we calculated the TIL density in each WSI and categorized them as low, medium, or high TIL levels. Our pipeline was evaluated against pathologist-assigned TIL scores, achieving a Cohen's kappa of 0.71 on the external test set, corroborating previous research findings. These results confirm that existing software can offer a practical solution for the assessment of TILs in H&E-stained WSIs of breast cancer.", "date": "2025-04-23", "categories": []}
{"id": "2505.12074", "title": "Denoising Mutual Knowledge Distillation in Bi-Directional Multiple Instance Learning", "abstract": "Multiple Instance Learning is the predominant method for Whole Slide Image classification in digital pathology, enabling the use of slide-level labels to supervise model training. Although MIL eliminates the tedious fine-grained annotation process for supervised learning, whether it can learn accurate bag- and instance-level classifiers remains a question. To address the issue, instance-level classifiers and instance masks were incorporated to ground the prediction on supporting patches. These methods, while practically improving the performance of MIL methods, may potentially introduce noisy labels. We propose to bridge the gap between commonly used MIL and fully supervised learning by augmenting both the bag- and instance-level learning processes with pseudo-label correction capabilities elicited from weak to strong generalization techniques. The proposed algorithm improves the performance of dual-level MIL algorithms on both bag- and instance-level predictions. Experiments on public pathology datasets showcase the advantage of the proposed methods.", "date": "2025-05-17", "categories": []}
{"id": "2505.20332", "title": "An Artificial Intelligence Model for Early Stage Breast Cancer Detection from Biopsy Images", "abstract": "Accurate identification of breast cancer types plays a critical role in guiding treatment decisions and improving patient outcomes. This paper presents an artificial intelligence enabled tool designed to aid in the identification of breast cancer types using histopathological biopsy images. Traditionally additional tests have to be done on women who are detected with breast cancer to find out the types of cancer it is to give the necessary cure. Those tests are not only invasive but also delay the initiation of treatment and increase patient burden. The proposed model utilizes a convolutional neural network (CNN) architecture to distinguish between benign and malignant tissues as well as accurate subclassification of breast cancer types. By preprocessing the images to reduce noise and enhance features, the model achieves reliable levels of classification performance. Experimental results on such datasets demonstrate the model's effectiveness, outperforming several existing solutions in terms of accuracy, precision, recall, and F1-score. The study emphasizes the potential of deep learning techniques in clinical diagnostics and offers a promising tool to assist pathologists in breast cancer classification.", "date": "2025-05-24", "categories": []}
{"id": "2501.18984", "title": "Context Matters: Query-aware Dynamic Long Sequence Modeling of Gigapixel Images", "abstract": "Whole slide image (WSI) analysis presents significant computational challenges due to the massive number of patches in gigapixel images. While transformer architectures excel at modeling long-range correlations through self-attention, their quadratic computational complexity makes them impractical for computational pathology applications. Existing solutions like local-global or linear self-attention reduce computational costs but compromise the strong modeling capabilities of full self-attention. In this work, we propose Querent, i.e., the query-aware long contextual dynamic modeling framework, which achieves a theoretically bounded approximation of full self-attention while delivering practical efficiency. Our method adaptively predicts which surrounding regions are most relevant for each patch, enabling focused yet unrestricted attention computation only with potentially important contexts. By using efficient region-wise metadata computation and importance estimation, our approach dramatically reduces computational overhead while preserving global perception to model fine-grained patch correlations. Through comprehensive experiments on biomarker prediction, gene mutation prediction, cancer subtyping, and survival analysis across over 10 WSI datasets, our method demonstrates superior performance compared to the state-of-the-art approaches. Codes are available at https://github.com/dddavid4real/Querent.", "date": "2025-01-31", "categories": []}
{"id": "2505.18745", "title": "C3R: Channel Conditioned Cell Representations for unified evaluation in microscopy imaging", "abstract": "Immunohistochemical (IHC) images reveal detailed information about structures and functions at the subcellular level. However, unlike natural images, IHC datasets pose challenges for deep learning models due to their inconsistencies in channel count and configuration, stemming from varying staining protocols across laboratories and studies. Existing approaches build channel-adaptive models, which unfortunately fail to support out-of-distribution (OOD) evaluation across IHC datasets and cannot be applied in a true zero-shot setting with mismatched channel counts. To address this, we introduce a structured view of cellular image channels by grouping them into either context or concept, where we treat the context channels as a reference to the concept channels in the image. We leverage this context-concept principle to develop Channel Conditioned Cell Representations (C3R), a framework designed for unified evaluation on in-distribution (ID) and OOD datasets. C3R is a two-fold framework comprising a channel-adaptive encoder architecture and a masked knowledge distillation training strategy, both built around the context-concept principle. We find that C3R outperforms existing benchmarks on both ID and OOD tasks, while a trivial implementation of our core idea also outperforms the channel-adaptive methods reported on the CHAMMI benchmark. Our method opens a new pathway for cross-dataset generalization between IHC datasets, without requiring dataset-specific adaptation or retraining.", "date": "2025-05-24", "categories": []}
{"id": "2505.20202", "title": "PathBench: A comprehensive comparison benchmark for pathology foundation models towards precision oncology", "abstract": "The emergence of pathology foundation models has revolutionized computational histopathology, enabling highly accurate, generalized whole-slide image analysis for improved cancer diagnosis, and prognosis assessment. While these models show remarkable potential across cancer diagnostics and prognostics, their clinical translation faces critical challenges including variability in optimal model across cancer types, potential data leakage in evaluation, and lack of standardized benchmarks. Without rigorous, unbiased evaluation, even the most advanced PFMs risk remaining confined to research settings, delaying their life-saving applications. Existing benchmarking efforts remain limited by narrow cancer-type focus, potential pretraining data overlaps, or incomplete task coverage. We present PathBench, the first comprehensive benchmark addressing these gaps through: multi-center in-hourse datasets spanning common cancers with rigorous leakage prevention, evaluation across the full clinical spectrum from diagnosis to prognosis, and an automated leaderboard system for continuous model assessment. Our framework incorporates large-scale data, enabling objective comparison of PFMs while reflecting real-world clinical complexity. All evaluation data comes from private medical providers, with strict exclusion of any pretraining usage to avoid data leakage risks. We have collected 15,888 WSIs from 8,549 patients across 10 hospitals, encompassing over 64 diagnosis and prognosis tasks. Currently, our evaluation of 19 PFMs shows that Virchow2 and H-Optimus-1 are the most effective models overall. This work provides researchers with a robust platform for model development and offers clinicians actionable insights into PFM performance across diverse clinical scenarios, ultimately accelerating the translation of these transformative technologies into routine pathology practice.", "date": "2025-05-26", "categories": []}
{"id": "2501.17822", "title": "Aggregation Schemes for Single-Vector WSI Representation Learning in Digital Pathology", "abstract": "A crucial step to efficiently integrate Whole Slide Images (WSIs) in computational pathology is assigning a single high-quality feature vector, i.e., one embedding, to each WSI. With the existence of many pre-trained deep neural networks and the emergence of foundation models, extracting embeddings for sub-images (i.e., tiles or patches) is straightforward. However, for WSIs, given their high resolution and gigapixel nature, inputting them into existing GPUs as a single image is not feasible. As a result, WSIs are usually split into many patches. Feeding each patch to a pre-trained model, each WSI can then be represented by a set of patches, hence, a set of embeddings. Hence, in such a setup, WSI representation learning reduces to set representation learning where for each WSI we have access to a set of patch embeddings. To obtain a single embedding from a set of patch embeddings for each WSI, multiple set-based learning schemes have been proposed in the literature. In this paper, we evaluate the WSI search performance of multiple recently developed aggregation techniques (mainly set representation learning techniques) including simple average or max pooling operations, Deep Sets, Memory networks, Focal attention, Gaussian Mixture Model (GMM) Fisher Vector, and deep sparse and binary Fisher Vector on four different primary sites including bladder, breast, kidney, and Colon from TCGA. Further, we benchmark the search performance of these methods against the median of minimum distances of patch embeddings, a non-aggregating approach used for WSI retrieval.", "date": "2025-01-29", "categories": []}
{"id": "2504.04045", "title": "A Survey of Pathology Foundation Model: Progress and Future Directions", "abstract": "Computational pathology, which involves analyzing whole slide images for automated cancer diagnosis, relies on multiple instance learning, where performance depends heavily on the feature extractor and aggregator. Recent Pathology Foundation Models (PFMs), pretrained on large-scale histopathology data, have significantly enhanced both the extractor and aggregator, but they lack a systematic analysis framework. In this survey, we present a hierarchical taxonomy organizing PFMs through a top-down philosophy applicable to foundation model analysis in any domain: model scope, model pretraining, and model design. Additionally, we systematically categorize PFM evaluation tasks into slide-level, patch-level, multimodal, and biological tasks, providing comprehensive benchmarking criteria. Our analysis identifies critical challenges in both PFM development (pathology-specific methodology, end-to-end pretraining, data-model scalability) and utilization (effective adaptation, model maintenance), paving the way for future directions in this promising field. Resources referenced in this survey are available at https://github.com/BearCleverProud/AwesomeWSI.", "date": "2025-04-05", "categories": []}
{"id": "2505.14926", "title": "Pathobiological Dictionary Defining Pathomics and Texture Features: Addressing Understandable AI Issues in Personalized Liver Cancer; Dictionary Version LCP1.0", "abstract": "Artificial intelligence (AI) holds strong potential for medical diagnostics, yet its clinical adoption is limited by a lack of interpretability and generalizability. This study introduces the Pathobiological Dictionary for Liver Cancer (LCP1.0), a practical framework designed to translate complex Pathomics and Radiomics Features (PF and RF) into clinically meaningful insights aligned with existing diagnostic workflows. QuPath and PyRadiomics, standardized according to IBSI guidelines, were used to extract 333 imaging features from hepatocellular carcinoma (HCC) tissue samples, including 240 PF-based-cell detection/intensity, 74 RF-based texture, and 19 RF-based first-order features. Expert-defined ROIs from the public dataset excluded artifact-prone areas, and features were aggregated at the case level. Their relevance to the WHO grading system was assessed using multiple classifiers linked with feature selectors. The resulting dictionary was validated by 8 experts in oncology and pathology. In collaboration with 10 domain experts, we developed a Pathobiological dictionary of imaging features such as PFs and RF. In our study, the Variable Threshold feature selection algorithm combined with the SVM model achieved the highest accuracy (0.80, P-value less than 0.05), selecting 20 key features, primarily clinical and pathomics traits such as Centroid, Cell Nucleus, and Cytoplasmic characteristics. These features, particularly nuclear and cytoplasmic, were strongly associated with tumor grading and prognosis, reflecting atypia indicators like pleomorphism, hyperchromasia, and cellular orientation.The LCP1.0 provides a clinically validated bridge between AI outputs and expert interpretation, enhancing model transparency and usability. Aligning AI-derived features with clinical semantics supports the development of interpretable, trustworthy diagnostic tools for liver cancer pathology.", "date": "2025-05-20", "categories": []}
{"id": "2403.11340", "title": "StainDiffuser: MultiTask Dual Diffusion Model for Virtual Staining", "abstract": "Hematoxylin and Eosin (H&E) staining is widely regarded as the standard in pathology for diagnosing diseases and tracking tumor recurrence. While H&E staining shows tissue structures, it lacks the ability to reveal specific proteins that are associated with disease severity and treatment response. Immunohistochemical (IHC) stains use antibodies to highlight the expression of these proteins on their respective cell types, improving diagnostic accuracy, and assisting with drug selection for treatment. Despite their value, IHC stains require additional time and resources, limiting their utilization in some clinical settings. Recent advances in deep learning have positioned Image-to-Image (I2I) translation as a computational, cost-effective alternative for IHC. I2I generates high fidelity stain transformations digitally, potentially replacing manual staining in IHC. Diffusion models, the current state of the art in image generation and conditional tasks, are particularly well suited for virtual IHC due to their ability to produce high quality images and resilience to mode collapse. However, these models require extensive and diverse datasets (often millions of samples) to achieve a robust performance, a challenge in virtual staining applications where only thousands of samples are typically available. Inspired by the success of multitask deep learning models in scenarios with limited data, we introduce STAINDIFFUSER, a novel multitask diffusion architecture tailored to virtual staining that achieves convergence with smaller datasets. STAINDIFFUSER simultaneously trains two diffusion processes: (a) generating cell specific IHC stains from H&E images and (b) performing H&E based cell segmentation, utilizing coarse segmentation labels exclusively during training. STAINDIFFUSER generates high-quality virtual stains for two markers, outperforming over twenty I2I baselines.", "date": "2024-03-17", "categories": []}
{"id": "2505.12711", "title": "Any-to-Any Learning in Computational Pathology via Triplet Multimodal Pretraining", "abstract": "Recent advances in computational pathology and artificial intelligence have significantly enhanced the utilization of gigapixel whole-slide images and and additional modalities (e.g., genomics) for pathological diagnosis. Although deep learning has demonstrated strong potential in pathology, several key challenges persist: (1) fusing heterogeneous data types requires sophisticated strategies beyond simple concatenation due to high computational costs; (2) common scenarios of missing modalities necessitate flexible strategies that allow the model to learn robustly in the absence of certain modalities; (3) the downstream tasks in CPath are diverse, ranging from unimodal to multimodal, cnecessitating a unified model capable of handling all modalities. To address these challenges, we propose ALTER, an any-to-any tri-modal pretraining framework that integrates WSIs, genomics, and pathology reports. The term \"any\" emphasizes ALTER's modality-adaptive design, enabling flexible pretraining with any subset of modalities, and its capacity to learn robust, cross-modal representations beyond WSI-centric approaches. We evaluate ALTER across extensive clinical tasks including survival prediction, cancer subtyping, gene mutation prediction, and report generation, achieving superior or comparable performance to state-of-the-art baselines.", "date": "2025-05-19", "categories": []}
{"id": "2410.18710", "title": "Uncovering the Genetic Basis of Glioblastoma Heterogeneity through Multimodal Analysis of Whole Slide Images and RNA Sequencing Data", "abstract": "Glioblastoma is a highly aggressive form of brain cancer characterized by rapid progression and poor prognosis. Despite advances in treatment, the underlying genetic mechanisms driving this aggressiveness remain poorly understood. In this study, we employed multimodal deep learning approaches to investigate glioblastoma heterogeneity using joint image/RNA-seq analysis. Our results reveal novel genes associated with glioblastoma. By leveraging a combination of whole-slide images and RNA-seq, as well as introducing novel methods to encode RNA-seq data, we identified specific genetic profiles that may explain different patterns of glioblastoma progression. These findings provide new insights into the genetic mechanisms underlying glioblastoma heterogeneity and highlight potential targets for therapeutic intervention. Code and data downloading instructions are available at: https://github.com/ma3oun/gbheterogeneity.", "date": "2024-10-23", "categories": []}
{"id": "2505.12120", "title": "HISTAI: An Open-Source, Large-Scale Whole Slide Image Dataset for Computational Pathology", "abstract": "Recent advancements in Digital Pathology (DP), particularly through artificial intelligence and Foundation Models, have underscored the importance of large-scale, diverse, and richly annotated datasets. Despite their critical role, publicly available Whole Slide Image (WSI) datasets often lack sufficient scale, tissue diversity, and comprehensive clinical metadata, limiting the robustness and generalizability of AI models. In response, we introduce the HISTAI dataset, a large, multimodal, open-access WSI collection comprising over 60,000 slides from various tissue types. Each case in the HISTAI dataset is accompanied by extensive clinical metadata, including diagnosis, demographic information, detailed pathological annotations, and standardized diagnostic coding. The dataset aims to fill gaps identified in existing resources, promoting innovation, reproducibility, and the development of clinically relevant computational pathology solutions. The dataset can be accessed at https://github.com/HistAI/HISTAI.", "date": "2025-05-17", "categories": []}
{"id": "2505.08430", "title": "GNCAF: A GNN-based Neighboring Context Aggregation Framework for Tertiary Lymphoid Structures Semantic Segmentation in WSI", "abstract": "Tertiary lymphoid structures (TLS) are organized clusters of immune cells, whose maturity and area can be quantified in whole slide image (WSI) for various prognostic tasks. Existing methods for assessing these characteristics typically rely on cell proxy tasks and require additional post-processing steps. In this work, We focus on a novel task-TLS Semantic Segmentation (TLS-SS)-which segments both the regions and maturation stages of TLS in WSI in an end-to-end manner. Due to the extensive scale of WSI and patch-based segmentation strategies, TLS-SS necessitates integrating from neighboring patches to guide target patch (target) segmentation. Previous techniques often employ on multi-resolution approaches, constraining the capacity to leverage the broader neighboring context while tend to preserve coarse-grained information. To address this, we propose a GNN-based Neighboring Context Aggregation Framework (GNCAF), which progressively aggregates multi-hop neighboring context from the target and employs a self-attention mechanism to guide the segmentation of the target. GNCAF can be integrated with various segmentation models to enhance their ability to perceive contextual information outside of the patch. We build two TLS-SS datasets, called TCGA-COAD and INHOUSE-PAAD, and make the former (comprising 225 WSIs and 5041 TLSs) publicly available. Experiments on these datasets demonstrate the superiority of GNCAF, achieving a maximum of 22.08% and 26.57% improvement in mF1 and mIoU, respectively. Additionally, we also validate the task scalability of GNCAF on segmentation of lymph node metastases.", "date": "2025-05-13", "categories": []}
{"id": "2505.08590", "title": "Enhancing Thyroid Cytology Diagnosis with RAG-Optimized LLMs and Pa-thology Foundation Models", "abstract": "Advancements in artificial intelligence (AI) are transforming pathology by integrat-ing large language models (LLMs) with retrieval-augmented generation (RAG) and domain-specific foundation models. This study explores the application of RAG-enhanced LLMs coupled with pathology foundation models for thyroid cytology diagnosis, addressing challenges in cytological interpretation, standardization, and diagnostic accuracy. By leveraging a curated knowledge base, RAG facilitates dy-namic retrieval of relevant case studies, diagnostic criteria, and expert interpreta-tion, improving the contextual understanding of LLMs. Meanwhile, pathology foun-dation models, trained on high-resolution pathology images, refine feature extrac-tion and classification capabilities. The fusion of these AI-driven approaches en-hances diagnostic consistency, reduces variability, and supports pathologists in dis-tinguishing benign from malignant thyroid lesions. Our results demonstrate that integrating RAG with pathology-specific LLMs significantly improves diagnostic efficiency and interpretability, paving the way for AI-assisted thyroid cytopathology, with foundation model UNI achieving AUC 0.73-0.93 for correct prediction of surgi-cal pathology diagnosis from thyroid cytology samples.", "date": "2025-05-13", "categories": []}
{"id": "2505.06710", "title": "SimMIL: A Universal Weakly Supervised Pre-Training Framework for Multi-Instance Learning in Whole Slide Pathology Images", "abstract": "Various multi-instance learning (MIL) based approaches have been developed and successfully applied to whole-slide pathological images (WSI). Existing MIL methods emphasize the importance of feature aggregators, but largely neglect the instance-level representation learning. They assume that the availability of a pre-trained feature extractor can be directly utilized or fine-tuned, which is not always the case. This paper proposes to pre-train feature extractor for MIL via a weakly-supervised scheme, i.e., propagating the weak bag-level labels to the corresponding instances for supervised learning. To learn effective features for MIL, we further delve into several key components, including strong data augmentation, a non-linear prediction head and the robust loss function. We conduct experiments on common large-scale WSI datasets and find it achieves better performance than other pre-training schemes (e.g., ImageNet pre-training and self-supervised learning) in different downstream tasks. We further show the compatibility and scalability of the proposed scheme by deploying it in fine-tuning the pathological-specific models and pre-training on merged multiple datasets. To our knowledge, this is the first work focusing on the representation learning for MIL.", "date": "2025-05-10", "categories": []}
{"id": "2505.05041", "title": "ADNP-15: An Open-Source Histopathological Dataset for Neuritic Plaque\n  Segmentation in Human Brain Whole Slide Images with Frequency Domain Image\n  Enhancement for Stain Normalization", "abstract": "  Alzheimer's Disease (AD) is a neurodegenerative disorder characterized by\namyloid-beta plaques and tau neurofibrillary tangles, which serve as key\nhistopathological features. The identification and segmentation of these\nlesions are crucial for understanding AD progression but remain challenging due\nto the lack of large-scale annotated datasets and the impact of staining\nvariations on automated image analysis. Deep learning has emerged as a powerful\ntool for pathology image segmentation; however, model performance is\nsignificantly influenced by variations in staining characteristics,\nnecessitating effective stain normalization and enhancement techniques. In this\nstudy, we address these challenges by introducing an open-source dataset\n(ADNP-15) of neuritic plaques (i.e., amyloid deposits combined with a crown of\ndystrophic tau-positive neurites) in human brain whole slide images. We\nestablish a comprehensive benchmark by evaluating five widely adopted deep\nlearning models across four stain normalization techniques, providing deeper\ninsights into their influence on neuritic plaque segmentation. Additionally, we\npropose a novel image enhancement method that improves segmentation accuracy,\nparticularly in complex tissue structures, by enhancing structural details and\nmitigating staining inconsistencies. Our experimental results demonstrate that\nthis enhancement strategy significantly boosts model generalization and\nsegmentation accuracy. All datasets and code are open-source, ensuring\ntransparency and reproducibility while enabling further advancements in the\nfield.\n", "date": "2025-05-08", "categories": []}
{"id": "2505.05073", "title": "RepSNet: A Nucleus Instance Segmentation model based on Boundary\n  Regression and Structural Re-parameterization", "abstract": "  Pathological diagnosis is the gold standard for tumor diagnosis, and nucleus\ninstance segmentation is a key step in digital pathology analysis and\npathological diagnosis. However, the computational efficiency of the model and\nthe treatment of overlapping targets are the major challenges in the studies of\nthis problem. To this end, a neural network model RepSNet was designed based on\na nucleus boundary regression and a structural re-parameterization scheme for\nsegmenting and classifying the nuclei in H\\&E-stained histopathological images.\nFirst, RepSNet estimates the boundary position information (BPI) of the parent\nnucleus for each pixel. The BPI estimation incorporates the local information\nof the pixel and the contextual information of the parent nucleus. Then, the\nnucleus boundary is estimated by aggregating the BPIs from a series of pixels\nusing a proposed boundary voting mechanism (BVM), and the instance segmentation\nresults are computed from the estimated nucleus boundary using a connected\ncomponent analysis procedure. The BVM intrinsically achieves a kind of\nsynergistic belief enhancement among the BPIs from various pixels. Therefore,\ndifferent from the methods available in literature that obtain nucleus\nboundaries based on a direct pixel recognition scheme, RepSNet computes its\nboundary decisions based on some guidances from macroscopic information using\nan integration mechanism. In addition, RepSNet employs a re-parametrizable\nencoder-decoder structure. This model can not only aggregate features from some\nreceptive fields with various scales which helps segmentation accuracy\nimprovement, but also reduce the parameter amount and computational burdens in\nthe model inference phase through the structural re-parameterization technique.\nExtensive experiments demonstrated the superiorities of RepSNet compared to\nseveral typical benchmark models.\n", "date": "2025-05-08", "categories": []}
{"id": "2505.01109", "title": "Self-Supervision Enhances Instance-based Multiple Instance Learning\n  Methods in Digital Pathology: A Benchmark Study", "abstract": "  Multiple Instance Learning (MIL) has emerged as the best solution for Whole\nSlide Image (WSI) classification. It consists of dividing each slide into\npatches, which are treated as a bag of instances labeled with a global label.\nMIL includes two main approaches: instance-based and embedding-based. In the\nformer, each patch is classified independently, and then the patch scores are\naggregated to predict the bag label. In the latter, bag classification is\nperformed after aggregating patch embeddings. Even if instance-based methods\nare naturally more interpretable, embedding-based MILs have usually been\npreferred in the past due to their robustness to poor feature extractors.\nHowever, recently, the quality of feature embeddings has drastically increased\nusing self-supervised learning (SSL). Nevertheless, many authors continue to\nendorse the superiority of embedding-based MIL. To investigate this further, we\nconduct 710 experiments across 4 datasets, comparing 10 MIL strategies, 6\nself-supervised methods with 4 backbones, 4 foundation models, and various\npathology-adapted techniques. Furthermore, we introduce 4 instance-based MIL\nmethods never used before in the pathology domain. Through these extensive\nexperiments, we show that with a good SSL feature extractor, simple\ninstance-based MILs, with very few parameters, obtain similar or better\nperformance than complex, state-of-the-art (SOTA) embedding-based MIL methods,\nsetting new SOTA results on the BRACS and Camelyon16 datasets. Since simple\ninstance-based MIL methods are naturally more interpretable and explainable to\nclinicians, our results suggest that more effort should be put into\nwell-adapted SSL methods for WSI rather than into complex embedding-based MIL\nmethods.\n", "date": "2025-05-02", "categories": []}
{"id": "2505.00134", "title": "Investigating Zero-Shot Diagnostic Pathology in Vision-Language Models\n  with Efficient Prompt Design", "abstract": "  Vision-language models (VLMs) have gained significant attention in\ncomputational pathology due to their multimodal learning capabilities that\nenhance big-data analytics of giga-pixel whole slide image (WSI). However,\ntheir sensitivity to large-scale clinical data, task formulations, and prompt\ndesign remains an open question, particularly in terms of diagnostic accuracy.\nIn this paper, we present a systematic investigation and analysis of three\nstate of the art VLMs for histopathology, namely Quilt-Net, Quilt-LLAVA, and\nCONCH, on an in-house digestive pathology dataset comprising 3,507 WSIs, each\nin giga-pixel form, across distinct tissue types. Through a structured ablative\nstudy on cancer invasiveness and dysplasia status, we develop a comprehensive\nprompt engineering framework that systematically varies domain specificity,\nanatomical precision, instructional framing, and output constraints. Our\nfindings demonstrate that prompt engineering significantly impacts model\nperformance, with the CONCH model achieving the highest accuracy when provided\nwith precise anatomical references. Additionally, we identify the critical\nimportance of anatomical context in histopathological image analysis, as\nperformance consistently degraded when reducing anatomical precision. We also\nshow that model complexity alone does not guarantee superior performance, as\neffective domain alignment and domain-specific training are critical. These\nresults establish foundational guidelines for prompt engineering in\ncomputational pathology and highlight the potential of VLMs to enhance\ndiagnostic accuracy when properly instructed with domain-appropriate prompts.\n", "date": "2025-04-30", "categories": []}
{"id": "2504.14139", "title": "ThyroidEffi 1.0: A Cost-Effective System for High-Performance\n  Multi-Class Thyroid Carcinoma Classification", "abstract": "  Background: Automated classification of thyroid Fine Needle Aspiration Biopsy\n(FNAB) images faces challenges in limited data, inter-observer variability, and\ncomputational cost. Efficient, interpretable models are crucial for clinical\nsupport.\n  Objective: To develop and externally validate a deep learning system for\nmulti-class thyroid FNAB image classification into three key categories\ndirectly guiding post-biopsy treatment in Vietnam: Benign (Bethesda II),\nIndeterminate/Suspicious (BI, III, IV, V), and Malignant (BVI), achieving high\ndiagnostic accuracy with low computational overhead.\n  Methods: Our pipeline features: (1) YOLOv10 cell cluster detection for\ninformative sub-region extraction/noise reduction; (2) curriculum learning\nsequencing localized crops to full images for multi-scale capture; (3) adaptive\nlightweight EfficientNetB0 (4M parameters) balancing performance/efficiency;\nand (4) a Transformer-inspired module for multi-scale/multi-region analysis.\nExternal validation used 1,015 independent FNAB images.\n  Results: ThyroidEffi Basic achieved macro F1 of 89.19% and AUCs of 0.98\n(Benign), 0.95 (Indeterminate/Suspicious), 0.96 (Malignant) on the internal\ntest set. External validation yielded AUCs of 0.9495 (Benign), 0.7436\n(Indeterminate/Suspicious), 0.8396 (Malignant). ThyroidEffi Premium improved\nmacro F1 to 89.77%. Grad-CAM highlighted key diagnostic regions, confirming\ninterpretability. The system processed 1000 cases in 30 seconds, demonstrating\nfeasibility on widely accessible hardware.\n  Conclusions: This work demonstrates that high-accuracy, interpretable thyroid\nFNAB image classification is achievable with minimal computational demands.\n", "date": "2025-04-19", "categories": []}
{"id": "2504.18856", "title": "Multi-Resolution Pathology-Language Pre-training Model with Text-Guided\n  Visual Representation", "abstract": "  In Computational Pathology (CPath), the introduction of Vision-Language\nModels (VLMs) has opened new avenues for research, focusing primarily on\naligning image-text pairs at a single magnification level. However, this\napproach might not be sufficient for tasks like cancer subtype classification,\ntissue phenotyping, and survival analysis due to the limited level of detail\nthat a single-resolution image can provide. Addressing this, we propose a novel\nmulti-resolution paradigm leveraging Whole Slide Images (WSIs) to extract\nhistology patches at multiple resolutions and generate corresponding textual\ndescriptions through advanced CPath VLM. We introduce visual-textual alignment\nat multiple resolutions as well as cross-resolution alignment to establish more\neffective text-guided visual representations. Cross-resolution alignment using\na multimodal encoder enhances the model's ability to capture context from\nmultiple resolutions in histology images. Our model aims to capture a broader\nrange of information, supported by novel loss functions, enriches feature\nrepresentation, improves discriminative ability, and enhances generalization\nacross different resolutions. Pre-trained on a comprehensive TCGA dataset with\n34 million image-language pairs at various resolutions, our fine-tuned model\noutperforms state-of-the-art (SOTA) counterparts across multiple datasets and\ntasks, demonstrating its effectiveness in CPath. The code is available on\nGitHub at: https://github.com/BasitAlawode/MR-PLIP\n", "date": "2025-04-26", "categories": []}
{"id": "2411.17073", "title": "Path-RAG: Knowledge-Guided Key Region Retrieval for Open-ended Pathology\n  Visual Question Answering", "abstract": "  Accurate diagnosis and prognosis assisted by pathology images are essential\nfor cancer treatment selection and planning. Despite the recent trend of\nadopting deep-learning approaches for analyzing complex pathology images, they\nfall short as they often overlook the domain-expert understanding of tissue\nstructure and cell composition. In this work, we focus on a challenging\nOpen-ended Pathology VQA (PathVQA-Open) task and propose a novel framework\nnamed Path-RAG, which leverages HistoCartography to retrieve relevant domain\nknowledge from pathology images and significantly improves performance on\nPathVQA-Open. Admitting the complexity of pathology image analysis, Path-RAG\nadopts a human-centered AI approach by retrieving domain knowledge using\nHistoCartography to select the relevant patches from pathology images. Our\nexperiments suggest that domain guidance can significantly boost the accuracy\nof LLaVA-Med from 38% to 47%, with a notable gain of 28% for H&E-stained\npathology images in the PathVQA-Open dataset. For longer-form question and\nanswer pairs, our model consistently achieves significant improvements of 32.5%\nin ARCH-Open PubMed and 30.6% in ARCH-Open Books on H\\&E images. Our code and\ndataset is available here (https://github.com/embedded-robotics/path-rag).\n", "date": "2024-11-26", "categories": []}
{"id": "2504.15627", "title": "ZeroSlide: Is Zero-Shot Classification Adequate for Lifelong Learning in\n  Whole-Slide Image Analysis in the Era of Pathology Vision-Language Foundation\n  Models?", "abstract": "  Lifelong learning for whole slide images (WSIs) poses the challenge of\ntraining a unified model to perform multiple WSI-related tasks, such as cancer\nsubtyping and tumor classification, in a distributed, continual fashion. This\nis a practical and applicable problem in clinics and hospitals, as WSIs are\nlarge, require storage, processing, and transfer time. Training new models\nwhenever new tasks are defined is time-consuming. Recent work has applied\nregularization- and rehearsal-based methods to this setting. However, the rise\nof vision-language foundation models that align diagnostic text with pathology\nimages raises the question: are these models alone sufficient for lifelong WSI\nlearning using zero-shot classification, or is further investigation into\ncontinual learning strategies needed to improve performance? To our knowledge,\nthis is the first study to compare conventional continual-learning approaches\nwith vision-language zero-shot classification for WSIs. Our source code and\nexperimental results will be available soon.\n", "date": "2025-04-22", "categories": []}
{"id": "2503.16816", "title": "PH2ST:ST-Prompt Guided Histological Hypergraph Learning for Spatial Gene\n  Expression Prediction", "abstract": "  Spatial Transcriptomics (ST) reveals the spatial distribution of gene\nexpression in tissues, offering critical insights into biological processes and\ndisease mechanisms. However, the high cost, limited coverage, and technical\ncomplexity of current ST technologies restrict their widespread use in clinical\nand research settings, making obtaining high-resolution transcriptomic profiles\nacross large tissue areas challenging. Predicting ST from H\\&E-stained\nhistology images has emerged as a promising alternative to address these\nlimitations but remains challenging due to the heterogeneous relationship\nbetween histomorphology and gene expression, which is affected by substantial\nvariability across patients and tissue sections. In response, we propose PH2ST,\na prompt-guided hypergraph learning framework, which leverages limited ST\nsignals to guide multi-scale histological representation learning for accurate\nand robust spatial gene expression prediction. Extensive evaluations on two\npublic ST datasets and multiple prompt sampling strategies simulating\nreal-world scenarios demonstrate that PH2ST not only outperforms existing\nstate-of-the-art methods, but also shows strong potential for practical\napplications such as imputing missing spots, ST super-resolution, and\nlocal-to-global prediction, highlighting its value for scalable and\ncost-effective spatial gene expression mapping in biomedical contexts.\n", "date": "2025-03-21", "categories": []}
{"id": "2402.03592", "title": "GRASP: GRAph-Structured Pyramidal Whole Slide Image Representation", "abstract": "  Cancer subtyping is one of the most challenging tasks in digital pathology,\nwhere Multiple Instance Learning (MIL) by processing gigapixel whole slide\nimages (WSIs) has been in the spotlight of recent research. However, MIL\napproaches do not take advantage of inter- and intra-magnification information\ncontained in WSIs. In this work, we present GRASP, a novel lightweight\ngraph-structured multi-magnification framework for processing WSIs in digital\npathology. Our approach is designed to dynamically emulate the pathologist's\nbehavior in handling WSIs and benefits from the hierarchical structure of WSIs.\nGRASP, which introduces a convergence-based node aggregation mechanism\nreplacing traditional pooling mechanisms, outperforms state-of-the-art methods\nby a high margin in terms of balanced accuracy, while being significantly\nsmaller than the closest-performing state-of-the-art models in terms of the\nnumber of parameters. Our results show that GRASP is dynamic in finding and\nconsulting with different magnifications for subtyping cancers, is reliable and\nstable across different hyperparameters, and can generalize when using features\nfrom different backbones. The model's behavior has been evaluated by two expert\npathologists confirming the interpretability of the model's dynamic. We also\nprovide a theoretical foundation, along with empirical evidence, for our work,\nexplaining how GRASP interacts with different magnifications and nodes in the\ngraph to make predictions. We believe that the strong characteristics yet\nsimple structure of GRASP will encourage the development of interpretable,\nstructure-based designs for WSI representation in digital pathology. Data and\ncode can be found in https://github.com/AIMLab-UBC/GRASP\n", "date": "2024-02-06", "categories": []}
{"id": "2504.12718", "title": "TUMLS: Trustful Fully Unsupervised Multi-Level Segmentation for Whole\n  Slide Images of Histology", "abstract": "  Digital pathology, augmented by artificial intelligence (AI), holds\nsignificant promise for improving the workflow of pathologists. However,\nchallenges such as the labor-intensive annotation of whole slide images (WSIs),\nhigh computational demands, and trust concerns arising from the absence of\nuncertainty estimation in predictions hinder the practical application of\ncurrent AI methodologies in histopathology. To address these issues, we present\na novel trustful fully unsupervised multi-level segmentation methodology\n(TUMLS) for WSIs. TUMLS adopts an autoencoder (AE) as a feature extractor to\nidentify the different tissue types within low-resolution training data. It\nselects representative patches from each identified group based on an\nuncertainty measure and then does unsupervised nuclei segmentation in their\nrespective higher-resolution space without using any ML algorithms. Crucially,\nthis solution integrates seamlessly into clinicians workflows, transforming the\nexamination of a whole WSI into a review of concise, interpretable cross-level\ninsights. This integration significantly enhances and accelerates the workflow\nwhile ensuring transparency. We evaluated our approach using the UPENN-GBM\ndataset, where the AE achieved a mean squared error (MSE) of 0.0016.\nAdditionally, nucleus segmentation is assessed on the MoNuSeg dataset,\noutperforming all unsupervised approaches with an F1 score of 77.46% and a\nJaccard score of 63.35%. These results demonstrate the efficacy of TUMLS in\nadvancing the field of digital pathology.\n", "date": "2025-04-17", "categories": []}
{"id": "2504.10526", "title": "PathSeqSAM: Sequential Modeling for Pathology Image Segmentation with\n  SAM2", "abstract": "  Current methods for pathology image segmentation typically treat 2D slices\nindependently, ignoring valuable cross-slice information. We present\nPathSeqSAM, a novel approach that treats 2D pathology slices as sequential\nvideo frames using SAM2's memory mechanisms. Our method introduces a\ndistance-aware attention mechanism that accounts for variable physical\ndistances between slices and employs LoRA for domain adaptation. Evaluated on\nthe KPI Challenge 2024 dataset for glomeruli segmentation, PathSeqSAM\ndemonstrates improved segmentation quality, particularly in challenging cases\nthat benefit from cross-slice context. We have publicly released our code at\nhttps://github.com/JackyyyWang/PathSeqSAM.\n", "date": "2025-04-12", "categories": []}
{"id": "2407.18449", "title": "Towards A Generalizable Pathology Foundation Model via Unified Knowledge\n  Distillation", "abstract": "  Foundation models pretrained on large-scale datasets are revolutionizing the\nfield of computational pathology (CPath). The generalization ability of\nfoundation models is crucial for the success in various downstream clinical\ntasks. However, current foundation models have only been evaluated on a limited\ntype and number of tasks, leaving their generalization ability and overall\nperformance unclear. To address this gap, we established a most comprehensive\nbenchmark to evaluate the performance of off-the-shelf foundation models across\nsix distinct clinical task types, encompassing a total of 72 specific tasks,\nincluding slide-level classification, survival prediction, ROI-tissue\nclassification, ROI retrieval, visual question answering, and report\ngeneration. Our findings reveal that existing foundation models excel at\ncertain task types but struggle to effectively handle the full breadth of\nclinical tasks. To improve the generalization of pathology foundation models,\nwe propose a unified knowledge distillation framework consisting of both expert\nand self-knowledge distillation, where the former allows the model to learn\nfrom the knowledge of multiple expert models, while the latter leverages\nself-distillation to enable image representation learning via local-global\nalignment. Based on this framework, we curated a dataset of 96,000 whole slide\nimages (WSIs) and developed a Generalizable Pathology Foundation Model (GPFM).\nThis advanced model was trained on a substantial dataset comprising 190 million\nimages extracted from approximately 72,000 publicly available slides,\nencompassing 34 major tissue types. Evaluated on the established benchmark,\nGPFM achieves an impressive average rank of 1.6, with 42 tasks ranked 1st,\nwhile the second-best model, UNI, attains an average rank of 3.7, with only 6\ntasks ranked 1st.\n", "date": "2024-07-26", "categories": []}
{"id": "2504.09430", "title": "Predicting ulcer in H&E images of inflammatory bowel disease using\n  domain-knowledge-driven graph neural network", "abstract": "  Inflammatory bowel disease (IBD) involves chronic inflammation of the\ndigestive tract, with treatment options often burdened by adverse effects.\nIdentifying biomarkers for personalized treatment is crucial. While immune\ncells play a key role in IBD, accurately identifying ulcer regions in whole\nslide images (WSIs) is essential for characterizing these cells and exploring\npotential therapeutics. Multiple instance learning (MIL) approaches have\nadvanced WSI analysis but they lack spatial context awareness. In this work, we\npropose a weakly-supervised model called DomainGCN that employs a graph\nconvolution neural network (GCN) and incorporates domain-specific knowledge of\nulcer features, specifically, the presence of epithelium, lymphocytes, and\ndebris for WSI-level ulcer prediction in IBD. We demonstrate that DomainGCN\noutperforms various state-of-the-art (SOTA) MIL methods and show the added\nvalue of domain knowledge.\n", "date": "2025-04-13", "categories": []}
{"id": "2504.07061", "title": "Teaching pathology foundation models to accurately predict gene\n  expression with parameter efficient knowledge transfer", "abstract": "  Gene expression profiling provides critical insights into cellular\nheterogeneity, biological processes and disease mechanisms. There has been an\nincreasing interest in computational approaches that can predict gene\nexpression directly from digitalized histopathology images. While image\nfoundation models have shown promise in a variety of pathology downstream\nanalysis, their performances on gene-expression prediction are still limited.\nExplicitly incorporating information from the transcriptomic models can help\nimage models to address domain shift, yet the fine-tuning and alignment of\nfoundation models can be expensive. In the work, we propose Parameter Efficient\nKnowledge trAnsfer (PEKA), a novel framework that leverages Block-Affine\nAdaptation and integrates knowledge distillation and structure alignment losses\nfor cross-modal knowledge transfer. We evaluated PEKA for gene expression\nprediction using multiple spatial transcriptomics datasets (comprising 206,123\nimage tiles with matched gene expression profiles) that encompassed various\ntypes of tissue. PEKA achieved at least 5\\% performance improvement over\nbaseline foundation models while also outperforming alternative\nparameter-efficient fine-tuning strategies. We will release the code, datasets\nand aligned models after peer-review to facilitate broader adoption and further\ndevelopment for parameter efficient model alignment.\n", "date": "2025-04-09", "categories": []}
{"id": "2501.11014", "title": "Transfer Learning Strategies for Pathological Foundation Models: A\n  Systematic Evaluation in Brain Tumor Classification", "abstract": "  Foundation models pretrained on large-scale pathology datasets have shown\npromising results across various diagnostic tasks. Here, we present a\nsystematic evaluation of transfer learning strategies for brain tumor\nclassification using these models. We analyzed 254 cases comprising five major\ntumor types: glioblastoma, astrocytoma, oligodendroglioma, primary central\nnervous system lymphoma, and metastatic tumors. Comparing state-of-the-art\nfoundation models with conventional approaches, we found that foundation models\ndemonstrated robust classification performance with as few as 10 patches per\ncase, despite the traditional assumption that extensive per-case image sampling\nis necessary. Furthermore, our evaluation revealed that simple transfer\nlearning strategies like linear probing were sufficient, while fine-tuning\noften degraded model performance. These findings suggest a paradigm shift from\n\"training encoders on extensive pathological data\" to \"querying pre-trained\nencoders with labeled datasets\", providing practical implications for\nimplementing AI-assisted diagnosis in clinical pathology.\n", "date": "2025-01-19", "categories": []}
{"id": "2504.05403", "title": "A Novel Approach to Linking Histology Images with DNA Methylation", "abstract": "  DNA methylation is an epigenetic mechanism that regulates gene expression by\nadding methyl groups to DNA. Abnormal methylation patterns can disrupt gene\nexpression and have been linked to cancer development. To quantify DNA\nmethylation, specialized assays are typically used. However, these assays are\noften costly and have lengthy processing times, which limits their widespread\navailability in routine clinical practice. In contrast, whole slide images\n(WSIs) for the majority of cancer patients can be more readily available. As\nsuch, given the ready availability of WSIs, there is a compelling need to\nexplore the potential relationship between WSIs and DNA methylation patterns.\nTo address this, we propose an end-to-end graph neural network based weakly\nsupervised learning framework to predict the methylation state of gene groups\nexhibiting coherent patterns across samples. Using data from three cohorts from\nThe Cancer Genome Atlas (TCGA) - TCGA-LGG (Brain Lower Grade Glioma), TCGA-GBM\n(Glioblastoma Multiforme) ($n$=729) and TCGA-KIRC (Kidney Renal Clear Cell\nCarcinoma) ($n$=511) - we demonstrate that the proposed approach achieves\nsignificantly higher AUROC scores than the state-of-the-art (SOTA) methods, by\nmore than $20\\%$. We conduct gene set enrichment analyses on the gene groups\nand show that majority of the gene groups are significantly enriched in\nimportant hallmarks and pathways. We also generate spatially enriched heatmaps\nto further investigate links between histological patterns and DNA methylation\nstates. To the best of our knowledge, this is the first study that explores\nassociation of spatially resolved histological patterns with gene group\nmethylation states across multiple cancer types using weakly supervised deep\nlearning.\n", "date": "2025-04-07", "categories": []}
{"id": "2503.02876", "title": "SPIDER: A Comprehensive Multi-Organ Supervised Pathology Dataset and\n  Baseline Models", "abstract": "  Advancing AI in computational pathology requires large, high-quality, and\ndiverse datasets, yet existing public datasets are often limited in organ\ndiversity, class coverage, or annotation quality. To bridge this gap, we\nintroduce SPIDER (Supervised Pathology Image-DEscription Repository), the\nlargest publicly available patch-level dataset covering multiple organ types,\nincluding Skin, Colorectal, Thorax, and Breast with comprehensive class\ncoverage for each organ. SPIDER provides high-quality annotations verified by\nexpert pathologists and includes surrounding context patches, which enhance\nclassification performance by providing spatial context.\n  Alongside the dataset, we present baseline models trained on SPIDER using the\nHibou-L foundation model as a feature extractor combined with an\nattention-based classification head. The models achieve state-of-the-art\nperformance across multiple tissue categories and serve as strong benchmarks\nfor future digital pathology research. Beyond patch classification, the model\nenables rapid identification of significant areas, quantitative tissue metrics,\nand establishes a foundation for multimodal approaches.\n  Both the dataset and trained models are publicly available to advance\nresearch, reproducibility, and AI-driven pathology development. Access them at:\nhttps://github.com/HistAI/SPIDER\n", "date": "2025-03-04", "categories": []}
{"id": "2504.05186", "title": "Training state-of-the-art pathology foundation models with orders of\n  magnitude less data", "abstract": "  The field of computational pathology has recently seen rapid advances driven\nby the development of modern vision foundation models (FMs), typically trained\non vast collections of pathology images. Recent studies demonstrate that\nincreasing the training data set and model size and integrating domain-specific\nimage processing techniques can significantly enhance the model's performance\non downstream tasks. Building on these insights, our work incorporates several\nrecent modifications to the standard DINOv2 framework from the literature to\noptimize the training of pathology FMs. We also apply a post-training procedure\nfor fine-tuning models on higher-resolution images to further enrich the\ninformation encoded in the embeddings. We present three novel pathology FMs\ntrained on up to two orders of magnitude fewer WSIs than those used to train\nother state-of-the-art FMs while demonstrating a comparable or superior\nperformance on downstream tasks. Even the model trained on TCGA alone (12k\nWSIs) outperforms most existing FMs and, on average, matches Virchow2, the\nsecond-best FM published to date. This suggests that there still remains a\nsignificant potential for further improving the models and algorithms used to\ntrain pathology FMs to take full advantage of the vast data collections.\n", "date": "2025-04-07", "categories": []}
{"id": "2411.07660", "title": "HMIL: Hierarchical Multi-Instance Learning for Fine-Grained Whole Slide\n  Image Classification", "abstract": "  Fine-grained classification of whole slide images (WSIs) is essential in\nprecision oncology, enabling precise cancer diagnosis and personalized\ntreatment strategies. The core of this task involves distinguishing subtle\nmorphological variations within the same broad category of gigapixel-resolution\nimages, which presents a significant challenge. While the multi-instance\nlearning (MIL) paradigm alleviates the computational burden of WSIs, existing\nMIL methods often overlook hierarchical label correlations, treating\nfine-grained classification as a flat multi-class classification task. To\novercome these limitations, we introduce a novel hierarchical multi-instance\nlearning (HMIL) framework. By facilitating on the hierarchical alignment of\ninherent relationships between different hierarchy of labels at instance and\nbag level, our approach provides a more structured and informative learning\nprocess. Specifically, HMIL incorporates a class-wise attention mechanism that\naligns hierarchical information at both the instance and bag levels.\nFurthermore, we introduce supervised contrastive learning to enhance the\ndiscriminative capability for fine-grained classification and a\ncurriculum-based dynamic weighting module to adaptively balance the\nhierarchical feature during training. Extensive experiments on our large-scale\ncytology cervical cancer (CCC) dataset and two public histology datasets, BRACS\nand PANDA, demonstrate the state-of-the-art class-wise and overall performance\nof our HMIL framework. Our source code is available at\nhttps://github.com/ChengJin-git/HMIL.\n", "date": "2024-11-12", "categories": []}
{"id": "2503.20880", "title": "BioX-CPath: Biologically-driven Explainable Diagnostics for Multistain\n  IHC Computational Pathology", "abstract": "  The development of biologically interpretable and explainable models remains\na key challenge in computational pathology, particularly for multistain\nimmunohistochemistry (IHC) analysis. We present BioX-CPath, an explainable\ngraph neural network architecture for whole slide image (WSI) classification\nthat leverages both spatial and semantic features across multiple stains. At\nits core, BioX-CPath introduces a novel Stain-Aware Attention Pooling (SAAP)\nmodule that generates biologically meaningful, stain-aware patient embeddings.\nOur approach achieves state-of-the-art performance on both Rheumatoid Arthritis\nand Sjogren's Disease multistain datasets. Beyond performance metrics,\nBioX-CPath provides interpretable insights through stain attention scores,\nentropy measures, and stain interaction scores, that permit measuring model\nalignment with known pathological mechanisms. This biological grounding,\ncombined with strong classification performance, makes BioX-CPath particularly\nsuitable for clinical applications where interpretability is key. Source code\nand documentation can be found at: https://github.com/AmayaGS/BioX-CPath.\n", "date": "2025-03-26", "categories": []}
{"id": "2504.03146", "title": "Comparative Analysis of Unsupervised and Supervised Autoencoders for\n  Nuclei Classification in Clear Cell Renal Cell Carcinoma Images", "abstract": "  This study explores the application of supervised and unsupervised\nautoencoders (AEs) to automate nuclei classification in clear cell renal cell\ncarcinoma (ccRCC) images, a diagnostic task traditionally reliant on subjective\nvisual grading by pathologists. We evaluate various AE architectures, including\nstandard AEs, contractive AEs (CAEs), and discriminative AEs (DAEs), as well as\na classifier-based discriminative AE (CDAE), optimized using the hyperparameter\ntuning tool Optuna. Bhattacharyya distance is selected from several metrics to\nassess class separability in the latent space, revealing challenges in\ndistinguishing adjacent grades using unsupervised models. CDAE, integrating a\nsupervised classifier branch, demonstrated superior performance in both latent\nspace separation and classification accuracy. Given that CDAE-CNN achieved\nnotable improvements in classification metrics, affirming the value of\nsupervised learning for class-specific feature extraction, F1 score was\nincorporated into the tuning process to optimize classification performance.\nResults show significant improvements in identifying aggressive ccRCC grades by\nleveraging the classification capability of AE through latent clustering\nfollowed by fine-grained classification. Our model outperforms the current\nstate of the art, CHR-Network, across all evaluated metrics. These findings\nsuggest that integrating a classifier branch in AEs, combined with neural\narchitecture search and contrastive learning, enhances grading automation in\nccRCC pathology, particularly in detecting aggressive tumor grades, and may\nimprove diagnostic accuracy.\n", "date": "2025-04-04", "categories": []}
{"id": "2504.01577", "title": "Instance Migration Diffusion for Nuclear Instance Segmentation in\n  Pathology", "abstract": "  Nuclear instance segmentation plays a vital role in disease diagnosis within\ndigital pathology. However, limited labeled data in pathological images\nrestricts the overall performance of nuclear instance segmentation. To tackle\nthis challenge, we propose a novel data augmentation framework Instance\nMigration Diffusion Model (IM-Diffusion), IM-Diffusion designed to generate\nmore varied pathological images by constructing diverse nuclear layouts and\ninternuclear spatial relationships. In detail, we introduce a Nuclear Migration\nModule (NMM) which constructs diverse nuclear layouts by simulating the process\nof nuclear migration. Building on this, we further present an\nInternuclear-regions Inpainting Module (IIM) to generate diverse internuclear\nspatial relationships by structure-aware inpainting. On the basis of the above,\nIM-Diffusion generates more diverse pathological images with different layouts\nand internuclear spatial relationships, thereby facilitating downstream tasks.\nEvaluation on the CoNSeP and GLySAC datasets demonstrate that the images\ngenerated by IM-Diffusion effectively enhance overall instance segmentation\nperformance. Code will be made public later.\n", "date": "2025-04-02", "categories": []}
{"id": "2504.00490", "title": "SCFANet: Style Distribution Constraint Feature Alignment Network For\n  Pathological Staining Translation", "abstract": "  Immunohistochemical (IHC) staining serves as a valuable technique for\ndetecting specific antigens or proteins through antibody-mediated\nvisualization. However, the IHC staining process is both time-consuming and\ncostly. To address these limitations, the application of deep learning models\nfor direct translation of cost-effective Hematoxylin and Eosin (H&E) stained\nimages into IHC stained images has emerged as an efficient solution.\nNevertheless, the conversion from H&E to IHC images presents significant\nchallenges, primarily due to alignment discrepancies between image pairs and\nthe inherent diversity in IHC staining style patterns. To overcome these\nchallenges, we propose the Style Distribution Constraint Feature Alignment\nNetwork (SCFANet), which incorporates two innovative modules: the Style\nDistribution Constrainer (SDC) and Feature Alignment Learning (FAL). The SDC\nensures consistency between the generated and target images' style\ndistributions while integrating cycle consistency loss to maintain structural\nconsistency. To mitigate the complexity of direct image-to-image translation,\nthe FAL module decomposes the end-to-end translation task into two subtasks:\nimage reconstruction and feature alignment. Furthermore, we ensure pathological\nconsistency between generated and target images by maintaining pathological\npattern consistency and Optical Density (OD) uniformity. Extensive experiments\nconducted on the Breast Cancer Immunohistochemical (BCI) dataset demonstrate\nthat our SCFANet model outperforms existing methods, achieving precise\ntransformation of H&E-stained images into their IHC-stained counterparts. The\nproposed approach not only addresses the technical challenges in H&E to IHC\nimage translation but also provides a robust framework for accurate and\nefficient stain conversion in pathological analysis.\n", "date": "2025-04-01", "categories": []}
{"id": "2504.00979", "title": "Artificial Intelligence-Assisted Prostate Cancer Diagnosis for Reduced\n  Use of Immunohistochemistry", "abstract": "  Prostate cancer diagnosis heavily relies on histopathological evaluation,\nwhich is subject to variability. While immunohistochemical staining (IHC)\nassists in distinguishing benign from malignant tissue, it involves increased\nwork, higher costs, and diagnostic delays. Artificial intelligence (AI)\npresents a promising solution to reduce reliance on IHC by accurately\nclassifying atypical glands and borderline morphologies in hematoxylin & eosin\n(H&E) stained tissue sections. In this study, we evaluated an AI model's\nability to minimize IHC use without compromising diagnostic accuracy by\nretrospectively analyzing prostate core needle biopsies from routine\ndiagnostics at three different pathology sites. These cohorts were composed\nexclusively of difficult cases where the diagnosing pathologists required IHC\nto finalize the diagnosis. The AI model demonstrated area under the curve\nvalues of 0.951-0.993 for detecting cancer in routine H&E-stained slides.\nApplying sensitivity-prioritized diagnostic thresholds reduced the need for IHC\nstaining by 44.4%, 42.0%, and 20.7% in the three cohorts investigated, without\na single false negative prediction. This AI model shows potential for\noptimizing IHC use, streamlining decision-making in prostate pathology, and\nalleviating resource burdens.\n", "date": "2025-03-31", "categories": []}
{"id": "2504.01009", "title": "GECKO: Gigapixel Vision-Concept Contrastive Pretraining in\n  Histopathology", "abstract": "  Pretraining a Multiple Instance Learning (MIL) aggregator enables the\nderivation of Whole Slide Image (WSI)-level embeddings from patch-level\nrepresentations without supervision. While recent multimodal MIL pretraining\napproaches leveraging auxiliary modalities have demonstrated performance gains\nover unimodal WSI pretraining, the acquisition of these additional modalities\nnecessitates extensive clinical profiling. This requirement increases costs and\nlimits scalability in existing WSI datasets lacking such paired modalities. To\naddress this, we propose Gigapixel Vision-Concept Knowledge Contrastive\npretraining (GECKO), which aligns WSIs with a Concept Prior derived from the\navailable WSIs. First, we derive an inherently interpretable concept prior by\ncomputing the similarity between each WSI patch and textual descriptions of\npredefined pathology concepts. GECKO then employs a dual-branch MIL network:\none branch aggregates patch embeddings into a WSI-level deep embedding, while\nthe other aggregates the concept prior into a corresponding WSI-level concept\nembedding. Both aggregated embeddings are aligned using a contrastive\nobjective, thereby pretraining the entire dual-branch MIL model. Moreover, when\nauxiliary modalities such as transcriptomics data are available, GECKO\nseamlessly integrates them. Across five diverse tasks, GECKO consistently\noutperforms prior unimodal and multimodal pretraining approaches while also\ndelivering clinically meaningful interpretability that bridges the gap between\ncomputational models and pathology expertise. Code is made available at\nhttps://github.com/bmi-imaginelab/GECKO\n", "date": "2025-04-01", "categories": []}
{"id": "2503.23021", "title": "The impact of tissue detection on diagnostic artificial intelligence\n  algorithms in digital pathology", "abstract": "  Tissue detection is a crucial first step in most digital pathology\napplications. Details of the segmentation algorithm are rarely reported, and\nthere is a lack of studies investigating the downstream effects of a poor\nsegmentation algorithm. Disregarding tissue detection quality could create a\nbottleneck for downstream performance and jeopardize patient safety if\ndiagnostically relevant parts of the specimen are excluded from analysis in\nclinical applications.\n  This study aims to determine whether performance of downstream tasks is\nsensitive to the tissue detection method, and to compare performance of\nclassical and AI-based tissue detection. To this end, we trained an AI model\nfor Gleason grading of prostate cancer in whole slide images (WSIs) using two\ndifferent tissue detection algorithms: thresholding (classical) and UNet++\n(AI). A total of 33,823 WSIs scanned on five digital pathology scanners were\nused to train the tissue detection AI model. The downstream Gleason grading\nalgorithm was trained and tested using 70,524 WSIs from 13 clinical sites\nscanned on 13 different scanners.\n  There was a decrease from 116 (0.43%) to 22 (0.08%) fully undetected tissue\nsamples when switching from thresholding-based tissue detection to AI-based,\nsuggesting an AI model may be more reliable than a classical model for avoiding\ntotal failures on slides with unusual appearance. On the slides where tissue\ncould be detected by both algorithms, no significant difference in overall\nGleason grading performance was observed. However, tissue detection dependent\nclinically significant variations in AI grading were observed in 3.5% of\nmalignant slides, highlighting the importance of robust tissue detection for\noptimal clinical performance of diagnostic AI.\n", "date": "2025-03-29", "categories": []}
{"id": "2503.23042", "title": "MIL vs. Aggregation: Evaluating Patient-Level Survival Prediction\n  Strategies Using Graph-Based Learning", "abstract": "  Oncologists often rely on a multitude of data, including whole-slide images\n(WSIs), to guide therapeutic decisions, aiming for the best patient outcome.\nHowever, predicting the prognosis of cancer patients can be a challenging task\ndue to tumor heterogeneity and intra-patient variability, and the complexity of\nanalyzing WSIs. These images are extremely large, containing billions of\npixels, making direct processing computationally expensive and requiring\nspecialized methods to extract relevant information. Additionally, multiple\nWSIs from the same patient may capture different tumor regions, some being more\ninformative than others. This raises a fundamental question: Should we use all\nWSIs to characterize the patient, or should we identify the most representative\nslide for prognosis? Our work seeks to answer this question by performing a\ncomparison of various strategies for predicting survival at the WSI and patient\nlevel. The former treats each WSI as an independent sample, mimicking the\nstrategy adopted in other works, while the latter comprises methods to either\naggregate the predictions of the several WSIs or automatically identify the\nmost relevant slide using multiple-instance learning (MIL). Additionally, we\nevaluate different Graph Neural Networks architectures under these strategies.\nWe conduct our experiments using the MMIST-ccRCC dataset, which comprises\npatients with clear cell renal cell carcinoma (ccRCC). Our results show that\nMIL-based selection improves accuracy, suggesting that choosing the most\nrepresentative slide benefits survival prediction.\n", "date": "2025-03-29", "categories": []}
{"id": "2503.24345", "title": "PathOrchestra: A Comprehensive Foundation Model for Computational\n  Pathology with Over 100 Diverse Clinical-Grade Tasks", "abstract": "  The complexity and variability inherent in high-resolution pathological\nimages present significant challenges in computational pathology. While\npathology foundation models leveraging AI have catalyzed transformative\nadvancements, their development demands large-scale datasets, considerable\nstorage capacity, and substantial computational resources. Furthermore,\nensuring their clinical applicability and generalizability requires rigorous\nvalidation across a broad spectrum of clinical tasks. Here, we present\nPathOrchestra, a versatile pathology foundation model trained via\nself-supervised learning on a dataset comprising 300K pathological slides from\n20 tissue and organ types across multiple centers. The model was rigorously\nevaluated on 112 clinical tasks using a combination of 61 private and 51 public\ndatasets. These tasks encompass digital slide preprocessing, pan-cancer\nclassification, lesion identification, multi-cancer subtype classification,\nbiomarker assessment, gene expression prediction, and the generation of\nstructured reports. PathOrchestra demonstrated exceptional performance across\n27,755 WSIs and 9,415,729 ROIs, achieving over 0.950 accuracy in 47 tasks,\nincluding pan-cancer classification across various organs, lymphoma subtype\ndiagnosis, and bladder cancer screening. Notably, it is the first model to\ngenerate structured reports for high-incidence colorectal cancer and\ndiagnostically complex lymphoma-areas that are infrequently addressed by\nfoundational models but hold immense clinical potential. Overall, PathOrchestra\nexemplifies the feasibility and efficacy of a large-scale, self-supervised\npathology foundation model, validated across a broad range of clinical-grade\ntasks. Its high accuracy and reduced reliance on extensive data annotation\nunderline its potential for clinical integration, offering a pathway toward\nmore efficient and high-quality medical services.\n", "date": "2025-03-31", "categories": []}
{"id": "2503.21818", "title": "Deep Learning-Based Quantitative Assessment of Renal Chronicity Indices\n  in Lupus Nephritis", "abstract": "  Background: Renal chronicity indices (CI) have been identified as strong\npredictors of long-term outcomes in lupus nephritis (LN) patients. However,\nassessment by pathologists is hindered by challenges such as substantial time\nrequirements, high interobserver variation, and susceptibility to fatigue. This\nstudy aims to develop an effective deep learning (DL) pipeline that automates\nthe assessment of CI and provides valuable prognostic insights from a\ndisease-specific perspective. Methods: We curated a dataset comprising 282\nslides obtained from 141 patients across two independent cohorts with a\ncomplete 10-years follow-up. Our DL pipeline was developed on 60 slides (22,410\npatch images) from 30 patients in the training cohort and evaluated on both an\ninternal testing set (148 slides, 77,605 patch images) and an external testing\nset (74 slides, 27,522 patch images). Results: The study included two cohorts\nwith slight demographic differences, particularly in age and hemoglobin levels.\nThe DL pipeline showed high segmentation performance across tissue compartments\nand histopathologic lesions, outperforming state-of-the-art methods. The DL\npipeline also demonstrated a strong correlation with pathologists in assessing\nCI, significantly improving interobserver agreement. Additionally, the DL\npipeline enhanced prognostic accuracy, particularly in outcome prediction, when\ncombined with clinical parameters and pathologist-assessed CIs Conclusions: The\nDL pipeline demonstrated accuracy and efficiency in assessing CI in LN, showing\npromise in improving interobserver agreement among pathologists. It also\nexhibited significant value in prognostic analysis and enhancing outcome\nprediction in LN patients, offering a valuable tool for clinical\ndecision-making.\n", "date": "2025-03-26", "categories": []}
{"id": "2503.22069", "title": "Contrasting Low and High-Resolution Features for HER2 Scoring using Deep\n  Learning", "abstract": "  Breast cancer, the most common malignancy among women, requires precise\ndetection and classification for effective treatment. Immunohistochemistry\n(IHC) biomarkers like HER2, ER, and PR are critical for identifying breast\ncancer subtypes. However, traditional IHC classification relies on\npathologists' expertise, making it labor-intensive and subject to significant\ninter-observer variability. To address these challenges, this study introduces\nthe India Pathology Breast Cancer Dataset (IPD-Breast), comprising of 1,272 IHC\nslides (HER2, ER, and PR) aimed at automating receptor status classification.\nThe primary focus is on developing predictive models for HER2 3-way\nclassification (0, Low, High) to enhance prognosis. Evaluation of multiple deep\nlearning models revealed that an end-to-end ConvNeXt network utilizing\nlow-resolution IHC images achieved an AUC, F1, and accuracy of 91.79%, 83.52%,\nand 83.56%, respectively, for 3-way classification, outperforming patch-based\nmethods by over 5.35% in F1 score. This study highlights the potential of\nsimple yet effective deep learning techniques to significantly improve accuracy\nand reproducibility in breast cancer classification, supporting their\nintegration into clinical workflows for better patient outcomes.\n", "date": "2025-03-28", "categories": []}
{"id": "2503.20190", "title": "Cross-Modal Prototype Allocation: Unsupervised Slide Representation\n  Learning via Patch-Text Contrast in Computational Pathology", "abstract": "  With the rapid advancement of pathology foundation models (FMs), the\nrepresentation learning of whole slide images (WSIs) attracts increasing\nattention. Existing studies develop high-quality patch feature extractors and\nemploy carefully designed aggregation schemes to derive slide-level\nrepresentations. However, mainstream weakly supervised slide representation\nlearning methods, primarily based on multiple instance learning (MIL), are\ntailored to specific downstream tasks, which limits their generalizability. To\naddress this issue, some studies explore unsupervised slide representation\nlearning. However, these approaches focus solely on the visual modality of\npatches, neglecting the rich semantic information embedded in textual data. In\nthis work, we propose ProAlign, a cross-modal unsupervised slide representation\nlearning framework. Specifically, we leverage a large language model (LLM) to\ngenerate descriptive text for the prototype types present in a WSI, introducing\npatch-text contrast to construct initial prototype embeddings. Furthermore, we\npropose a parameter-free attention aggregation strategy that utilizes the\nsimilarity between patches and these prototypes to form unsupervised slide\nembeddings applicable to a wide range of downstream tasks. Extensive\nexperiments on four public datasets show that ProAlign outperforms existing\nunsupervised frameworks and achieves performance comparable to some weakly\nsupervised models.\n", "date": "2025-03-26", "categories": []}
{"id": "2407.15362", "title": "A Multimodal Knowledge-enhanced Whole-slide Pathology Foundation Model", "abstract": "  Remarkable strides in computational pathology have been made in the\ntask-agnostic foundation model that advances the performance of a wide array of\ndownstream clinical tasks. Despite the promising performance, there are still\nseveral challenges. First, prior works have resorted to either vision-only or\nimage-caption data, disregarding pathology reports with more clinically\nauthentic information from pathologists and gene expression profiles which\nrespectively offer distinct knowledge for versatile clinical applications.\nSecond, the current progress in pathology FMs predominantly concentrates on the\npatch level, where the restricted context of patch-level pretraining fails to\ncapture whole-slide patterns. Even recent slide-level FMs still struggle to\nprovide whole-slide context for patch representation. In this study, for the\nfirst time, we develop a pathology foundation model incorporating three levels\nof modalities: pathology slides, pathology reports, and gene expression data,\nwhich resulted in 26,169 slide-level modality pairs from 10,275 patients across\n32 cancer types, amounting to over 116 million pathological patch images. To\nleverage these data for CPath, we propose a novel whole-slide pretraining\nparadigm that injects the multimodal whole-slide context into the patch\nrepresentation, called Multimodal Self-TAught PRetraining (mSTAR). The proposed\nparadigm revolutionizes the pretraining workflow for CPath, enabling the\npathology FM to acquire the whole-slide context. To the best of our knowledge,\nthis is the first attempt to incorporate three modalities at the whole-slide\ncontext for enhancing pathology FMs. To systematically evaluate the\ncapabilities of mSTAR, we built the largest spectrum of oncological benchmark,\nspanning 7 categories of oncological applications in 15 types of 97 practical\noncological tasks.\n", "date": "2024-07-22", "categories": []}
{"id": "2503.06056", "title": "Pathological Prior-Guided Multiple Instance Learning For Mitigating\n  Catastrophic Forgetting in Breast Cancer Whole Slide Image Classification", "abstract": "  In histopathology, intelligent diagnosis of Whole Slide Images (WSIs) is\nessential for automating and objectifying diagnoses, reducing the workload of\npathologists. However, diagnostic models often face the challenge of forgetting\npreviously learned data during incremental training on datasets from different\nsources. To address this issue, we propose a new framework PaGMIL to mitigate\ncatastrophic forgetting in breast cancer WSI classification. Our framework\nintroduces two key components into the common MIL model architecture. First, it\nleverages microscopic pathological prior to select more accurate and diverse\nrepresentative patches for MIL. Secondly, it trains separate classification\nheads for each task and uses macroscopic pathological prior knowledge, treating\nthe thumbnail as a prompt guide (PG) to select the appropriate classification\nhead. We evaluate the continual learning performance of PaGMIL across several\npublic breast cancer datasets. PaGMIL achieves a better balance between the\nperformance of the current task and the retention of previous tasks,\noutperforming other continual learning methods. Our code will be open-sourced\nupon acceptance.\n", "date": "2025-03-08", "categories": []}
{"id": "2503.19606", "title": "Single Shot AI-assisted quantification of KI-67 proliferation index in\n  breast cancer", "abstract": "  Reliable quantification of Ki-67, a key proliferation marker in breast\ncancer, is essential for molecular subtyping and informed treatment planning.\nConventional approaches, including visual estimation and manual counting,\nsuffer from interobserver variability and limited reproducibility. This study\nintroduces an AI-assisted method using the YOLOv8 object detection framework\nfor automated Ki-67 scoring. High-resolution digital images (40x magnification)\nof immunohistochemically stained tumor sections were captured from Ki-67\nhotspot regions and manually annotated by a domain expert to distinguish\nKi-67-positive and negative tumor cells. The dataset was augmented and divided\ninto training (80%), validation (10%), and testing (10%) subsets. Among the\nYOLOv8 variants tested, the Medium model achieved the highest performance, with\na mean Average Precision at 50% Intersection over Union (mAP50) exceeding 85%\nfor Ki-67-positive cells. The proposed approach offers an efficient, scalable,\nand objective alternative to conventional scoring methods, supporting greater\nconsistency in Ki-67 evaluation. Future directions include developing\nuser-friendly clinical interfaces and expanding to multi-institutional datasets\nto enhance generalizability and facilitate broader adoption in diagnostic\npractice.\n", "date": "2025-03-25", "categories": []}
{"id": "2503.17970", "title": "PathoHR: Breast Cancer Survival Prediction on High-Resolution\n  Pathological Images", "abstract": "  Breast cancer survival prediction in computational pathology presents a\nremarkable challenge due to tumor heterogeneity. For instance, different\nregions of the same tumor in the pathology image can show distinct\nmorphological and molecular characteristics. This makes it difficult to extract\nrepresentative features from whole slide images (WSIs) that truly reflect the\ntumor's aggressive potential and likely survival outcomes. In this paper, we\npresent PathoHR, a novel pipeline for accurate breast cancer survival\nprediction that enhances any size of pathological images to enable more\neffective feature learning. Our approach entails (1) the incorporation of a\nplug-and-play high-resolution Vision Transformer (ViT) to enhance patch-wise\nWSI representation, enabling more detailed and comprehensive feature\nextraction, (2) the systematic evaluation of multiple advanced similarity\nmetrics for comparing WSI-extracted features, optimizing the representation\nlearning process to better capture tumor characteristics, (3) the demonstration\nthat smaller image patches enhanced follow the proposed pipeline can achieve\nequivalent or superior prediction accuracy compared to raw larger patches,\nwhile significantly reducing computational overhead. Experimental findings\nvalid that PathoHR provides the potential way of integrating enhanced image\nresolution with optimized feature learning to advance computational pathology,\noffering a promising direction for more accurate and efficient breast cancer\nsurvival prediction. Code will be available at\nhttps://github.com/AIGeeksGroup/PathoHR.\n", "date": "2025-03-23", "categories": []}
{"id": "2503.18567", "title": "Advancing Cross-Organ Domain Generalization with Test-Time Style\n  Transfer and Diversity Enhancement", "abstract": "  Deep learning has made significant progress in addressing challenges in\nvarious fields including computational pathology (CPath). However, due to the\ncomplexity of the domain shift problem, the performance of existing models will\ndegrade, especially when it comes to multi-domain or cross-domain tasks. In\nthis paper, we propose a Test-time style transfer (T3s) that uses a\nbidirectional mapping mechanism to project the features of the source and\ntarget domains into a unified feature space, enhancing the generalization\nability of the model. To further increase the style expression space, we\nintroduce a Cross-domain style diversification module (CSDM) to ensure the\northogonality between style bases. In addition, data augmentation and low-rank\nadaptation techniques are used to improve feature alignment and sensitivity,\nenabling the model to adapt to multi-domain inputs effectively. Our method has\ndemonstrated effectiveness on three unseen datasets.\n", "date": "2025-03-24", "categories": []}
{"id": "2408.09278", "title": "Cross-Species Data Integration for Enhanced Layer Segmentation in Kidney\n  Pathology", "abstract": "  Accurate delineation of the boundaries between the renal cortex and medulla\nis crucial for subsequent functional structural analysis and disease diagnosis.\nTraining high-quality deep-learning models for layer segmentation relies on the\navailability of large amounts of annotated data. However, due to the patient's\nprivacy of medical data and scarce clinical cases, constructing pathological\ndatasets from clinical sources is relatively difficult and expensive. Moreover,\nusing external natural image datasets introduces noise during the domain\ngeneralization process. Cross-species homologous data, such as mouse kidney\ndata, which exhibits high structural and feature similarity to human kidneys,\nhas the potential to enhance model performance on human datasets. In this\nstudy, we incorporated the collected private Periodic Acid-Schiff (PAS) stained\nmouse kidney dataset into the human kidney dataset for joint training. The\nresults showed that after introducing cross-species homologous data, the\nsemantic segmentation models based on CNN and Transformer architectures\nachieved an average increase of 1.77% and 1.24% in mIoU, and 1.76% and 0.89% in\nDice score for the human renal cortex and medulla datasets, respectively. This\napproach is also capable of enhancing the model's generalization ability. This\nindicates that cross-species homologous data, as a low-noise trainable data\nsource, can help improve model performance under conditions of limited clinical\nsamples. Code is available at https://github.com/hrlblab/layer_segmentation.\n", "date": "2024-08-17", "categories": []}
{"id": "2411.14743", "title": "FOCUS: Knowledge-enhanced Adaptive Visual Compression for Few-shot Whole\n  Slide Image Classification", "abstract": "  Few-shot learning presents a critical solution for cancer diagnosis in\ncomputational pathology (CPath), addressing fundamental limitations in data\navailability, particularly the scarcity of expert annotations and patient\nprivacy constraints. A key challenge in this paradigm stems from the inherent\ndisparity between the limited training set of whole slide images (WSIs) and the\nenormous number of contained patches, where a significant portion of these\npatches lacks diagnostically relevant information, potentially diluting the\nmodel's ability to learn and focus on critical diagnostic features. While\nrecent works attempt to address this by incorporating additional knowledge,\nseveral crucial gaps hinder further progress: (1) despite the emergence of\npowerful pathology foundation models (FMs), their potential remains largely\nuntapped, with most approaches limiting their use to basic feature extraction;\n(2) current language guidance mechanisms attempt to align text prompts with\nvast numbers of WSI patches all at once, struggling to leverage rich\npathological semantic information. To this end, we introduce the\nknowledge-enhanced adaptive visual compression framework, dubbed FOCUS, which\nuniquely combines pathology FMs with language prior knowledge to enable a\nfocused analysis of diagnostically relevant regions by prioritizing\ndiscriminative WSI patches. Our approach implements a progressive three-stage\ncompression strategy: we first leverage FMs for global visual redundancy\nelimination, and integrate compressed features with language prompts for\nsemantic relevance assessment, then perform neighbor-aware visual token\nfiltering while preserving spatial coherence. Extensive experiments on\npathological datasets spanning breast, lung, and ovarian cancers demonstrate\nits superior performance in few-shot pathology diagnosis. Codes are available\nat https://github.com/dddavid4real/FOCUS.\n", "date": "2024-11-22", "categories": []}
{"id": "2412.02601", "title": "MERGE: Multi-faceted Hierarchical Graph-based GNN for Gene Expression\n  Prediction from Whole Slide Histopathology Images", "abstract": "  Recent advances in Spatial Transcriptomics (ST) pair histology images with\nspatially resolved gene expression profiles, enabling predictions of gene\nexpression across different tissue locations based on image patches. This opens\nup new possibilities for enhancing whole slide image (WSI) prediction tasks\nwith localized gene expression. However, existing methods fail to fully\nleverage the interactions between different tissue locations, which are crucial\nfor accurate joint prediction. To address this, we introduce MERGE\n(Multi-faceted hiErarchical gRaph for Gene Expressions), which combines a\nmulti-faceted hierarchical graph construction strategy with graph neural\nnetworks (GNN) to improve gene expression predictions from WSIs. By clustering\ntissue image patches based on both spatial and morphological features, and\nincorporating intra- and inter-cluster edges, our approach fosters interactions\nbetween distant tissue locations during GNN learning. As an additional\ncontribution, we evaluate different data smoothing techniques that are\nnecessary to mitigate artifacts in ST data, often caused by technical\nimperfections. We advocate for adopting gene-aware smoothing methods that are\nmore biologically justified. Experimental results on gene expression prediction\nshow that our GNN method outperforms state-of-the-art techniques across\nmultiple metrics.\n", "date": "2024-12-03", "categories": []}
{"id": "2501.15598", "title": "Diffusion Generative Modeling for Spatially Resolved Gene Expression\n  Inference from Histology Images", "abstract": "  Spatial Transcriptomics (ST) allows a high-resolution measurement of RNA\nsequence abundance by systematically connecting cell morphology depicted in\nHematoxylin and Eosin (H&E) stained histology images to spatially resolved gene\nexpressions. ST is a time-consuming, expensive yet powerful experimental\ntechnique that provides new opportunities to understand cancer mechanisms at a\nfine-grained molecular level, which is critical for uncovering new approaches\nfor disease diagnosis and treatments. Here, we present $\\textbf{Stem}$\n($\\textbf{S}$pa$\\textbf{T}$ially resolved gene $\\textbf{E}$xpression inference\nwith diffusion $\\textbf{M}$odel), a novel computational tool that leverages a\nconditional diffusion generative model to enable in silico gene expression\ninference from H&E stained images. Through better capturing the inherent\nstochasticity and heterogeneity in ST data, $\\textbf{Stem}$ achieves\nstate-of-the-art performance on spatial gene expression prediction and\ngenerates biologically meaningful gene profiles for new H&E stained images at\ntest time. We evaluate the proposed algorithm on datasets with various tissue\nsources and sequencing platforms, where it demonstrates clear improvement over\nexisting approaches. $\\textbf{Stem}$ generates high-fidelity gene expression\npredictions that share similar gene variation levels as ground truth data,\nsuggesting that our method preserves the underlying biological heterogeneity.\nOur proposed pipeline opens up the possibility of analyzing existing, easily\naccessible H&E stained histology images from a genomics point of view without\nphysically performing gene expression profiling and empowers potential\nbiological discovery from H&E stained histology images.\n", "date": "2025-01-26", "categories": []}
{"id": "2503.09091", "title": "Multi-Modal Foundation Models for Computational Pathology: A Survey", "abstract": "  Foundation models have emerged as a powerful paradigm in computational\npathology (CPath), enabling scalable and generalizable analysis of\nhistopathological images. While early developments centered on uni-modal models\ntrained solely on visual data, recent advances have highlighted the promise of\nmulti-modal foundation models that integrate heterogeneous data sources such as\ntextual reports, structured domain knowledge, and molecular profiles. In this\nsurvey, we provide a comprehensive and up-to-date review of multi-modal\nfoundation models in CPath, with a particular focus on models built upon\nhematoxylin and eosin (H&E) stained whole slide images (WSIs) and tile-level\nrepresentations. We categorize 32 state-of-the-art multi-modal foundation\nmodels into three major paradigms: vision-language, vision-knowledge graph, and\nvision-gene expression. We further divide vision-language models into\nnon-LLM-based and LLM-based approaches. Additionally, we analyze 28 available\nmulti-modal datasets tailored for pathology, grouped into image-text pairs,\ninstruction datasets, and image-other modality pairs. Our survey also presents\na taxonomy of downstream tasks, highlights training and evaluation strategies,\nand identifies key challenges and future directions. We aim for this survey to\nserve as a valuable resource for researchers and practitioners working at the\nintersection of pathology and AI.\n", "date": "2025-03-12", "categories": []}
{"id": "2410.11761", "title": "SlideChat: A Large Vision-Language Assistant for Whole-Slide Pathology\n  Image Understanding", "abstract": "  Despite the progress made by multimodal large language models (MLLMs) in\ncomputational pathology, they remain limited by a predominant focus on\npatch-level analysis, missing essential contextual information at the\nwhole-slide level. The lack of large-scale instruction datasets and the\ngigapixel scale of whole slide images (WSIs) pose significant developmental\nchallenges. In this paper, we present SlideChat, the first vision-language\nassistant capable of understanding gigapixel whole-slide images, exhibiting\nexcellent multimodal conversational capability and response complex instruction\nacross diverse pathology scenarios. To support its development, we created\nSlideInstruction, the largest instruction-following dataset for WSIs consisting\nof 4.2K WSI captions and 176K VQA pairs with multiple categories. Furthermore,\nwe propose SlideBench, a multimodal benchmark that incorporates captioning and\nVQA tasks to assess SlideChat's capabilities in varied clinical settings such\nas microscopy, diagnosis. Compared to both general and specialized MLLMs,\nSlideChat exhibits exceptional capabilities achieving state-of-the-art\nperformance on 18 of 22 tasks. For example, it achieved an overall accuracy of\n81.17% on SlideBench-VQA (TCGA), and 54.15% on SlideBench-VQA (BCNB). Our code,\ndata, and model is publicly accessible at\nhttps://uni-medical.github.io/SlideChat.github.io.\n", "date": "2024-10-15", "categories": []}
{"id": "2503.13799", "title": "SMILE: a Scale-aware Multiple Instance Learning Method for Multicenter\n  STAS Lung Cancer Histopathology Diagnosis", "abstract": "  Spread through air spaces (STAS) represents a newly identified aggressive\npattern in lung cancer, which is known to be associated with adverse prognostic\nfactors and complex pathological features. Pathologists currently rely on time\nconsuming manual assessments, which are highly subjective and prone to\nvariation. This highlights the urgent need for automated and precise diag\nnostic solutions. 2,970 lung cancer tissue slides are comprised from multiple\ncenters, re-diagnosed them, and constructed and publicly released three lung\ncancer STAS datasets: STAS CSU (hospital), STAS TCGA, and STAS CPTAC. All STAS\ndatasets provide corresponding pathological feature diagnoses and related\nclinical data. To address the bias, sparse and heterogeneous nature of STAS, we\npropose an scale-aware multiple instance learning(SMILE) method for STAS\ndiagnosis of lung cancer. By introducing a scale-adaptive attention mechanism,\nthe SMILE can adaptively adjust high attention instances, reducing\nover-reliance on local regions and promoting consistent detection of STAS\nlesions. Extensive experiments show that SMILE achieved competitive diagnostic\nresults on STAS CSU, diagnosing 251 and 319 STAS samples in CPTAC\nandTCGA,respectively, surpassing clinical average AUC. The 11 open baseline\nresults are the first to be established for STAS research, laying the\nfoundation for the future expansion, interpretability, and clinical integration\nof computational pathology technologies. The datasets and code are available at\nhttps://anonymous.4open.science/r/IJCAI25-1DA1.\n", "date": "2025-03-18", "categories": []}
{"id": "2403.08967", "title": "PathM3: A Multimodal Multi-Task Multiple Instance Learning Framework for\n  Whole Slide Image Classification and Captioning", "abstract": "  In the field of computational histopathology, both whole slide images (WSIs)\nand diagnostic captions provide valuable insights for making diagnostic\ndecisions. However, aligning WSIs with diagnostic captions presents a\nsignificant challenge. This difficulty arises from two main factors: 1)\nGigapixel WSIs are unsuitable for direct input into deep learning models, and\nthe redundancy and correlation among the patches demand more attention; and 2)\nAuthentic WSI diagnostic captions are extremely limited, making it difficult to\ntrain an effective model. To overcome these obstacles, we present PathM3, a\nmultimodal, multi-task, multiple instance learning (MIL) framework for WSI\nclassification and captioning. PathM3 adapts a query-based transformer to\neffectively align WSIs with diagnostic captions. Given that histopathology\nvisual patterns are redundantly distributed across WSIs, we aggregate each\npatch feature with MIL method that considers the correlations among instances.\nFurthermore, our PathM3 overcomes data scarcity in WSI-level captions by\nleveraging limited WSI diagnostic caption data in the manner of multi-task\njoint learning. Extensive experiments with improved classification accuracy and\ncaption generation demonstrate the effectiveness of our method on both WSI\nclassification and captioning task.\n", "date": "2024-03-13", "categories": []}
{"id": "2410.23642", "title": "Development and prospective validation of a prostate cancer detection,\n  grading, and workflow optimization system at an academic medical center", "abstract": "  Artificial intelligence may assist healthcare systems in meeting increasing\ndemand for pathology services while maintaining diagnostic quality and reducing\nturnaround time and costs. We aimed to investigate the performance of an\ninstitutionally developed system for prostate cancer detection, grading, and\nworkflow optimization and to contrast this with commercial alternatives. From\nAugust 2021 to March 2023, we scanned 21,396 slides from 1,147 patients\nreceiving prostate biopsy. We developed models for cancer detection, grading,\nand screening of equivocal cases for IHC ordering. We compared the performance\nof task-specific prostate models with general-purpose foundation models in a\nprospectively collected dataset that reflects our patient population. We also\nevaluated the contributions of a bespoke model designed to improve sensitivity\nto small cancer foci and perception of low-resolution patterns. We found high\nconcordance with pathologist ground-truth in detection (area under curve 98.5%,\nsensitivity 95.0%, and specificity 97.8%), ISUP grading (Cohen's kappa 0.869),\ngrade group 3 or higher classification (area under curve 97.5%, sensitivity\n94.9%, specificity 96.6%). Screening models could correctly classify 55% of\nbiopsy blocks where immunohistochemistry was ordered with a 1.4% error rate. No\nstatistically significant differences were observed between task-specific and\nfoundation models in cancer detection, although the task-specific model is\nsignificantly smaller and faster. Institutions like academic medical centers\nthat have high scanning volumes and report abstraction capabilities can develop\nhighly accurate computational pathology models for internal use. These models\nhave the potential to aid in quality control role and to improve resource\nallocation and workflow in the pathology lab to help meet future challenges in\nprostate cancer diagnosis.\n", "date": "2024-10-31", "categories": []}
{"id": "2503.12269", "title": "Cracking the PUMA Challenge in 24 Hours with CellViT++ and nnU-Net", "abstract": "  Automatic tissue segmentation and nuclei detection is an important task in\npathology, aiding in biomarker extraction and discovery. The panoptic\nsegmentation of nuclei and tissue in advanced melanoma (PUMA) challenge aims to\nimprove tissue segmentation and nuclei detection in melanoma histopathology.\nUnlike many challenge submissions focusing on extensive model tuning, our\napproach emphasizes delivering a deployable solution within a 24-hour\ndevelopment timeframe, using out-of-the-box frameworks. The pipeline combines\ntwo models, namely CellViT++ for nuclei detection and nnU-Net for tissue\nsegmentation. Our results demonstrate a significant improvement in tissue\nsegmentation, achieving a Dice score of 0.750, surpassing the baseline score of\n0.629. For nuclei detection, we obtained results comparable to the baseline in\nboth challenge tracks. The code is publicly available at\nhttps://github.com/TIO-IKIM/PUMA.\n", "date": "2025-03-15", "categories": []}
{"id": "2503.12990", "title": "How Good is my Histopathology Vision-Language Foundation Model? A\n  Holistic Benchmark", "abstract": "  Recently, histopathology vision-language foundation models (VLMs) have gained\npopularity due to their enhanced performance and generalizability across\ndifferent downstream tasks. However, most existing histopathology benchmarks\nare either unimodal or limited in terms of diversity of clinical tasks, organs,\nand acquisition instruments, as well as their partial availability to the\npublic due to patient data privacy. As a consequence, there is a lack of\ncomprehensive evaluation of existing histopathology VLMs on a unified benchmark\nsetting that better reflects a wide range of clinical scenarios. To address\nthis gap, we introduce HistoVL, a fully open-source comprehensive benchmark\ncomprising images acquired using up to 11 various acquisition tools that are\npaired with specifically crafted captions by incorporating class names and\ndiverse pathology descriptions. Our Histo-VL includes 26 organs, 31 cancer\ntypes, and a wide variety of tissue obtained from 14 heterogeneous patient\ncohorts, totaling more than 5 million patches obtained from over 41K WSIs\nviewed under various magnification levels. We systematically evaluate existing\nhistopathology VLMs on Histo-VL to simulate diverse tasks performed by experts\nin real-world clinical scenarios. Our analysis reveals interesting findings,\nincluding large sensitivity of most existing histopathology VLMs to textual\nchanges with a drop in balanced accuracy of up to 25% in tasks such as\nMetastasis detection, low robustness to adversarial attacks, as well as\nimproper calibration of models evident through high ECE values and low model\nprediction confidence, all of which can affect their clinical implementation.\n", "date": "2025-03-17", "categories": []}
{"id": "2503.10731", "title": "Leveraging Vision-Language Embeddings for Zero-Shot Learning in\n  Histopathology Images", "abstract": "  Zero-shot learning holds tremendous potential for histopathology image\nanalysis by enabling models to generalize to unseen classes without extensive\nlabeled data. Recent advancements in vision-language models (VLMs) have\nexpanded the capabilities of ZSL, allowing models to perform tasks without\ntask-specific fine-tuning. However, applying VLMs to histopathology presents\nconsiderable challenges due to the complexity of histopathological imagery and\nthe nuanced nature of diagnostic tasks. In this paper, we propose a novel\nframework called Multi-Resolution Prompt-guided Hybrid Embedding (MR-PHE) to\naddress these challenges in zero-shot histopathology image classification.\nMR-PHE leverages multiresolution patch extraction to mimic the diagnostic\nworkflow of pathologists, capturing both fine-grained cellular details and\nbroader tissue structures critical for accurate diagnosis. We introduce a\nhybrid embedding strategy that integrates global image embeddings with weighted\npatch embeddings, effectively combining local and global contextual\ninformation. Additionally, we develop a comprehensive prompt generation and\nselection framework, enriching class descriptions with domain-specific synonyms\nand clinically relevant features to enhance semantic understanding. A\nsimilarity-based patch weighting mechanism assigns attention-like weights to\npatches based on their relevance to class embeddings, emphasizing\ndiagnostically important regions during classification. Our approach utilizes\npretrained VLM, CONCH for ZSL without requiring domain-specific fine-tuning,\noffering scalability and reducing dependence on large annotated datasets.\nExperimental results demonstrate that MR-PHE not only significantly improves\nzero-shot classification performance on histopathology datasets but also often\nsurpasses fully supervised models.\n", "date": "2025-03-13", "categories": []}
{"id": "2501.04467", "title": "Histologic Dataset of Normal and Atypical Mitotic Figures on Human\n  Breast Cancer (AMi-Br)", "abstract": "  Assessment of the density of mitotic figures (MFs) in histologic tumor\nsections is an important prognostic marker for many tumor types, including\nbreast cancer. Recently, it has been reported in multiple works that the\nquantity of MFs with an atypical morphology (atypical MFs, AMFs) might be an\nindependent prognostic criterion for breast cancer. AMFs are an indicator of\nmutations in the genes regulating the cell cycle and can lead to aberrant\nchromosome constitution (aneuploidy) of the tumor cells. To facilitate further\nresearch on this topic using pattern recognition, we present the first ever\npublicly available dataset of atypical and normal MFs (AMi-Br). For this, we\nutilized two of the most popular MF datasets (MIDOG 2021 and TUPAC) and\nsubclassified all MFs using a three expert majority vote. Our final dataset\nconsists of 3,720 MFs, split into 832 AMFs (22.4%) and 2,888 normal MFs (77.6%)\nacross all 223 tumor cases in the combined set. We provide baseline\nclassification experiments to investigate the consistency of the dataset, using\na Monte Carlo cross-validation and different strategies to combat class\nimbalance. We found an averaged balanced accuracy of up to 0.806 when using a\npatch-level data set split, and up to 0.713 when using a patient-level split.\n", "date": "2025-01-08", "categories": []}
{"id": "2503.10510", "title": "Extreme Learning Machines for Attention-based Multiple Instance Learning\n  in Whole-Slide Image Classification", "abstract": "  Whole-slide image classification represents a key challenge in computational\npathology and medicine. Attention-based multiple instance learning (MIL) has\nemerged as an effective approach for this problem. However, the effect of\nattention mechanism architecture on model performance is not well-documented\nfor biomedical imagery. In this work, we compare different methods and\nimplementations of MIL, including deep learning variants. We introduce a new\nmethod using higher-dimensional feature spaces for deep MIL. We also develop a\nnovel algorithm for whole-slide image classification where extreme machine\nlearning is combined with attention-based MIL to improve sensitivity and reduce\ntraining complexity. We apply our algorithms to the problem of detecting\ncirculating rare cells (CRCs), such as erythroblasts, in peripheral blood. Our\nresults indicate that nonlinearities play a key role in the classification, as\nremoving them leads to a sharp decrease in stability in addition to a decrease\nin average area under the curve (AUC) of over 4%. We also demonstrate a\nconsiderable increase in robustness of the model with improvements of over 10%\nin average AUC when higher-dimensional feature spaces are leveraged. In\naddition, we show that extreme learning machines can offer clear improvements\nin terms of training efficiency by reducing the number of trained parameters by\na factor of 5 whilst still maintaining the average AUC to within 1.5% of the\ndeep MIL model. Finally, we discuss options of enriching the classical\ncomputing framework with quantum algorithms in the future. This work can thus\nhelp pave the way towards more accurate and efficient single-cell diagnostics,\none of the building blocks of precision medicine.\n", "date": "2025-03-13", "categories": []}
{"id": "2503.10520", "title": "CountPath: Automating Fragment Counting in Digital Pathology", "abstract": "  Quality control of medical images is a critical component of digital\npathology, ensuring that diagnostic images meet required standards. A\npre-analytical task within this process is the verification of the number of\nspecimen fragments, a process that ensures that the number of fragments on a\nslide matches the number documented in the macroscopic report. This step is\nimportant to ensure that the slides contain the appropriate diagnostic material\nfrom the grossing process, thereby guaranteeing the accuracy of subsequent\nmicroscopic examination and diagnosis. Traditionally, this assessment is\nperformed manually, requiring significant time and effort while being subject\nto significant variability due to its subjective nature. To address these\nchallenges, this study explores an automated approach to fragment counting\nusing the YOLOv9 and Vision Transformer models. Our results demonstrate that\nthe automated system achieves a level of performance comparable to expert\nassessments, offering a reliable and efficient alternative to manual counting.\nAdditionally, we present findings on interobserver variability, showing that\nthe automated approach achieves an accuracy of 86%, which falls within the\nrange of variation observed among experts (82-88%), further supporting its\npotential for integration into routine pathology workflows.\n", "date": "2025-03-13", "categories": []}
{"id": "2503.08581", "title": "MsaMIL-Net: An End-to-End Multi-Scale Aware Multiple Instance Learning\n  Network for Efficient Whole Slide Image Classification", "abstract": "  Bag-based Multiple Instance Learning (MIL) approaches have emerged as the\nmainstream methodology for Whole Slide Image (WSI) classification. However,\nmost existing methods adopt a segmented training strategy, which first extracts\nfeatures using a pre-trained feature extractor and then aggregates these\nfeatures through MIL. This segmented training approach leads to insufficient\ncollaborative optimization between the feature extraction network and the MIL\nnetwork, preventing end-to-end joint optimization and thereby limiting the\noverall performance of the model. Additionally, conventional methods typically\nextract features from all patches of fixed size, ignoring the multi-scale\nobservation characteristics of pathologists. This not only results in\nsignificant computational resource waste when tumor regions represent a minimal\nproportion (as in the Camelyon16 dataset) but may also lead the model to\nsuboptimal solutions.\n  To address these limitations, this paper proposes an end-to-end multi-scale\nWSI classification framework that integrates multi-scale feature extraction\nwith multiple instance learning. Specifically, our approach includes: (1) a\nsemantic feature filtering module to reduce interference from non-lesion areas;\n(2) a multi-scale feature extraction module to capture pathological information\nat different levels; and (3) a multi-scale fusion MIL module for global\nmodeling and feature integration. Through an end-to-end training strategy, we\nsimultaneously optimize both the feature extractor and MIL network, ensuring\nmaximum compatibility between them.\n  Experiments were conducted on three cross-center datasets (DigestPath2019,\nBCNB, and UBC-OCEAN). Results demonstrate that our proposed method outperforms\nexisting state-of-the-art approaches in terms of both accuracy (ACC) and AUC\nmetrics.\n", "date": "2025-03-11", "categories": []}
{"id": "2503.06482", "title": "PathVQ: Reforming Computational Pathology Foundation Model for Whole\n  Slide Image Analysis via Vector Quantization", "abstract": "  Computational pathology and whole-slide image (WSI) analysis are pivotal in\ncancer diagnosis and prognosis. However, the ultra-high resolution of WSIs\npresents significant modeling challenges. Recent advancements in pathology\nfoundation models have improved performance, yet most approaches rely on [CLS]\ntoken representation of tile ViT as slide-level inputs (16x16 pixels is\nrefereed as patch and 224x224 pixels as tile). This discards critical spatial\ndetails from patch tokens, limiting downstream WSI analysis tasks. We find that\nleveraging all spatial patch tokens benefits WSI analysis but incurs nearly\n200x higher storage and training costs (e.g., 196 tokens in ViT$_{224}$). To\naddress this, we introduce vector quantized (VQ) distillation on patch feature,\nwhich efficiently compresses spatial patch tokens using discrete indices and a\ndecoder. Our method reduces token dimensionality from 1024 to 16, achieving a\n64x compression rate while preserving reconstruction fidelity. Furthermore, we\nemploy a multi-scale VQ (MSVQ) strategy, which not only enhances VQ\nreconstruction performance but also serves as a Self-supervised Learning (SSL)\nsupervision for a seamless slide-level pretraining objective. Built upon the\nquantized patch features and supervision targets of tile via MSVQ, we develop a\nprogressive convolutional module and slide-level SSL to extract representations\nwith rich spatial-information for downstream WSI tasks. Extensive evaluations\non multiple datasets demonstrate the effectiveness of our approach, achieving\nstate-of-the-art performance in WSI analysis. Code will be available soon.\n", "date": "2025-03-09", "categories": []}
{"id": "2503.07173", "title": "Towards Spatial Transcriptomics-guided Pathological Image Recognition\n  with Batch-Agnostic Encoder", "abstract": "  Spatial transcriptomics (ST) is a novel technique that simultaneously\ncaptures pathological images and gene expression profiling with spatial\ncoordinates. Since ST is closely related to pathological features such as\ndisease subtypes, it may be valuable to augment image representation with\npathological information. However, there are no attempts to leverage ST for\nimage recognition ({\\it i.e,} patch-level classification of subtypes of\npathological image.). One of the big challenges is significant batch effects in\nspatial transcriptomics that make it difficult to extract pathological features\nof images from ST. In this paper, we propose a batch-agnostic contrastive\nlearning framework that can extract consistent signals from gene expression of\nST in multiple patients. To extract consistent signals from ST, we utilize the\nbatch-agnostic gene encoder that is trained in a variational inference manner.\nExperiments demonstrated the effectiveness of our framework on a publicly\navailable dataset. Code is publicly available at\nhttps://github.com/naivete5656/TPIRBAE\n", "date": "2025-03-10", "categories": []}
{"id": "2408.01167", "title": "Rethinking Pre-Trained Feature Extractor Selection in Multiple Instance\n  Learning for Whole Slide Image Classification", "abstract": "  Multiple instance learning (MIL) has become a preferred method for gigapixel\nwhole slide image (WSI) classification without requiring patch-level\nannotations. Current MIL research primarily relies on embedding-based\napproaches, which extract patch features using a pre-trained feature extractor\nand aggregate them for slide-level prediction. Despite the critical role of\nfeature extraction, there is limited guidance on selecting optimal feature\nextractors to maximize WSI performance. This study addresses this gap by\nsystematically evaluating MIL feature extractors across three dimensions:\npre-training dataset, backbone model, and pre-training method. Extensive\nexperiments were conducted on two public WSI datasets (TCGA-NSCLC and\nCamelyon16) using four state-of-the-art (SOTA) MIL models. Our findings reveal\nthat: 1) selecting a robust self-supervised learning (SSL) method has a greater\nimpact on performance than relying solely on an in-domain pre-training dataset;\n2) prioritizing Transformer-based backbones with deeper architectures over\nCNN-based models; and 3) using larger, more diverse pre-training datasets\nsignificantly enhances classification outcomes. We hope that these insights can\nprovide practical guidance for optimizing WSI classification and explain the\nreasons behind the performance advantages of the current SOTA pathology\nfoundation models. Furthermore, this work may inform the development of more\neffective pathology foundation models. Our code is publicly available at\nhttps://github.com/bryanwong17/MIL-Feature-Extractor-Selection\n", "date": "2024-08-02", "categories": []}
{"id": "2401.04720", "title": "Low-resource finetuning of foundation models beats state-of-the-art in\n  histopathology", "abstract": "  To handle the large scale of whole slide images in computational pathology,\nmost approaches first tessellate the images into smaller patches, extract\nfeatures from these patches, and finally aggregate the feature vectors with\nweakly-supervised learning. The performance of this workflow strongly depends\non the quality of the extracted features. Recently, foundation models in\ncomputer vision showed that leveraging huge amounts of data through supervised\nor self-supervised learning improves feature quality and generalizability for a\nvariety of tasks. In this study, we benchmark the most popular vision\nfoundation models as feature extractors for histopathology data. We evaluate\nthe models in two settings: slide-level classification and patch-level\nclassification. We show that foundation models are a strong baseline. Our\nexperiments demonstrate that by finetuning a foundation model on a single GPU\nfor only two hours or three days depending on the dataset, we can match or\noutperform state-of-the-art feature extractors for computational pathology.\nThese findings imply that even with little resources one can finetune a feature\nextractor tailored towards a specific downstream task and dataset. This is a\nconsiderable shift from the current state, where only few institutions with\nlarge amounts of resources and datasets are able to train a feature extractor.\nWe publish all code used for training and evaluation as well as the finetuned\nmodels.\n", "date": "2024-01-09", "categories": []}
{"id": "2503.02064", "title": "CrossFusion: A Multi-Scale Cross-Attention Convolutional Fusion Model\n  for Cancer Survival Prediction", "abstract": "  Cancer survival prediction from whole slide images (WSIs) is a challenging\ntask in computational pathology due to the large size, irregular shape, and\nhigh granularity of the WSIs. These characteristics make it difficult to\ncapture the full spectrum of patterns, from subtle cellular abnormalities to\ncomplex tissue interactions, which are crucial for accurate prognosis. To\naddress this, we propose CrossFusion, a novel multi-scale feature integration\nframework that extracts and fuses information from patches across different\nmagnification levels. By effectively modeling both scale-specific patterns and\ntheir interactions, CrossFusion generates a rich feature set that enhances\nsurvival prediction accuracy. We validate our approach across six cancer types\nfrom public datasets, demonstrating significant improvements over existing\nstate-of-the-art methods. Moreover, when coupled with domain-specific feature\nextraction backbones, our method shows further gains in prognostic performance\ncompared to general-purpose backbones. The source code is available at:\nhttps://github.com/RustinS/CrossFusion\n", "date": "2025-03-03", "categories": []}
{"id": "2503.02849", "title": "Multimodal Deep Learning for Subtype Classification in Breast Cancer\n  Using Histopathological Images and Gene Expression Data", "abstract": "  Molecular subtyping of breast cancer is crucial for personalized treatment\nand prognosis. Traditional classification approaches rely on either\nhistopathological images or gene expression profiling, limiting their\npredictive power. In this study, we propose a deep multimodal learning\nframework that integrates histopathological images and gene expression data to\nclassify breast cancer into BRCA.Luminal and BRCA.Basal / Her2 subtypes. Our\napproach employs a ResNet-50 model for image feature extraction and fully\nconnected layers for gene expression processing, with a cross-attention fusion\nmechanism to enhance modality interaction. We conduct extensive experiments\nusing five-fold cross-validation, demonstrating that our multimodal integration\noutperforms unimodal approaches in terms of classification accuracy,\nprecision-recall AUC, and F1-score. Our findings highlight the potential of\ndeep learning for robust and interpretable breast cancer subtype\nclassification, paving the way for improved clinical decision-making.\n", "date": "2025-03-04", "categories": []}
{"id": "2502.21130", "title": "Fast and Accurate Gigapixel Pathological Image Classification with\n  Hierarchical Distillation Multi-Instance Learning", "abstract": "  Although multi-instance learning (MIL) has succeeded in pathological image\nclassification, it faces the challenge of high inference costs due to\nprocessing numerous patches from gigapixel whole slide images (WSIs). To\naddress this, we propose HDMIL, a hierarchical distillation multi-instance\nlearning framework that achieves fast and accurate classification by\neliminating irrelevant patches. HDMIL consists of two key components: the\ndynamic multi-instance network (DMIN) and the lightweight instance\npre-screening network (LIPN). DMIN operates on high-resolution WSIs, while LIPN\noperates on the corresponding low-resolution counterparts. During training,\nDMIN are trained for WSI classification while generating attention-score-based\nmasks that indicate irrelevant patches. These masks then guide the training of\nLIPN to predict the relevance of each low-resolution patch. During testing,\nLIPN first determines the useful regions within low-resolution WSIs, which\nindirectly enables us to eliminate irrelevant regions in high-resolution WSIs,\nthereby reducing inference time without causing performance degradation. In\naddition, we further design the first Chebyshev-polynomials-based\nKolmogorov-Arnold classifier in computational pathology, which enhances the\nperformance of HDMIL through learnable activation layers. Extensive experiments\non three public datasets demonstrate that HDMIL outperforms previous\nstate-of-the-art methods, e.g., achieving improvements of 3.13% in AUC while\nreducing inference time by 28.6% on the Camelyon16 dataset.\n", "date": "2025-02-28", "categories": []}
{"id": "2502.21264", "title": "Foundation Models -- A Panacea for Artificial Intelligence in Pathology?", "abstract": "  The role of artificial intelligence (AI) in pathology has evolved from aiding\ndiagnostics to uncovering predictive morphological patterns in whole slide\nimages (WSIs). Recently, foundation models (FMs) leveraging self-supervised\npre-training have been widely advocated as a universal solution for diverse\ndownstream tasks. However, open questions remain about their clinical\napplicability and generalization advantages over end-to-end learning using\ntask-specific (TS) models. Here, we focused on AI with clinical-grade\nperformance for prostate cancer diagnosis and Gleason grading. We present the\nlargest validation of AI for this task, using over 100,000 core needle biopsies\nfrom 7,342 patients across 15 sites in 11 countries. We compared two FMs with a\nfully end-to-end TS model in a multiple instance learning framework. Our\nfindings challenge assumptions that FMs universally outperform TS models. While\nFMs demonstrated utility in data-scarce scenarios, their performance converged\nwith - and was in some cases surpassed by - TS models when sufficient labeled\ntraining data were available. Notably, extensive task-specific training\nmarkedly reduced clinically significant misgrading, misdiagnosis of challenging\nmorphologies, and variability across different WSI scanners. Additionally, FMs\nused up to 35 times more energy than the TS model, raising concerns about their\nsustainability. Our results underscore that while FMs offer clear advantages\nfor rapid prototyping and research, their role as a universal solution for\nclinically applicable medical AI remains uncertain. For high-stakes clinical\napplications, rigorous validation and consideration of task-specific training\nremain critically important. We advocate for integrating the strengths of FMs\nand end-to-end learning to achieve robust and resource-efficient AI pathology\nsolutions fit for clinical use.\n", "date": "2025-02-28", "categories": []}
{"id": "2503.00915", "title": "Multimodal Distillation-Driven Ensemble Learning for Long-Tailed\n  Histopathology Whole Slide Images Analysis", "abstract": "  Multiple Instance Learning (MIL) plays a significant role in computational\npathology, enabling weakly supervised analysis of Whole Slide Image (WSI)\ndatasets. The field of WSI analysis is confronted with a severe long-tailed\ndistribution problem, which significantly impacts the performance of\nclassifiers. Long-tailed distributions lead to class imbalance, where some\nclasses have sparse samples while others are abundant, making it difficult for\nclassifiers to accurately identify minority class samples. To address this\nissue, we propose an ensemble learning method based on MIL, which employs\nexpert decoders with shared aggregators and consistency constraints to learn\ndiverse distributions and reduce the impact of class imbalance on classifier\nperformance. Moreover, we introduce a multimodal distillation framework that\nleverages text encoders pre-trained on pathology-text pairs to distill\nknowledge and guide the MIL aggregator in capturing stronger semantic features\nrelevant to class information. To ensure flexibility, we use learnable prompts\nto guide the distillation process of the pre-trained text encoder, avoiding\nlimitations imposed by specific prompts. Our method, MDE-MIL, integrates\nmultiple expert branches focusing on specific data distributions to address\nlong-tailed issues. Consistency control ensures generalization across classes.\nMultimodal distillation enhances feature extraction. Experiments on\nCamelyon+-LT and PANDA-LT datasets show it outperforms state-of-the-art\nmethods.\n", "date": "2025-03-02", "categories": []}
{"id": "2406.02618", "title": "Immunocto: a massive immune cell database auto-generated for\n  histopathology", "abstract": "  With the advent of novel cancer treatment options such as immunotherapy,\nstudying the tumour immune micro-environment (TIME) is crucial to inform on\nprognosis and understand potential response to therapeutic agents. A key\napproach to characterising the TIME may be through combining (1) digitised\nmicroscopic high-resolution optical images of hematoxylin and eosin (H&E)\nstained tissue sections obtained in routine histopathology examinations with\n(2) automated immune cell detection and classification methods. In this work,\nwe introduce a workflow to automatically generate robust single cell contours\nand labels from dually stained tissue sections with H&E and multiplexed\nimmunofluorescence (IF) markers. The approach harnesses the Segment Anything\nModel and requires minimal human intervention compared to existing single cell\ndatabases. With this methodology, we create Immunocto, a massive, multi-million\nautomatically generated database of 6,848,454 human cells and objects,\nincluding 2,282,818 immune cells distributed across 4 subtypes: CD4$^+$ T cell\nlymphocytes, CD8$^+$ T cell lymphocytes, CD20$^+$ B cell lymphocytes, and\nCD68$^+$/CD163$^+$ macrophages. For each cell, we provide a 64$\\times$64\npixels$^2$ H&E image at $\\mathbf{40}\\times$ magnification, along with a binary\nmask of the nucleus and a label. The database, which is made publicly\navailable, can be used to train models to study the TIME on routine H&E slides.\nWe show that deep learning models trained on Immunocto result in\nstate-of-the-art performance for lymphocyte detection. The approach\ndemonstrates the benefits of using matched H&E and IF data to generate robust\ndatabases for computational pathology applications.\n", "date": "2024-06-03", "categories": []}
{"id": "2412.03084", "title": "Hybrid deep learning-based strategy for the hepatocellular carcinoma\n  cancer grade classification of H&E stained liver histopathology images", "abstract": "  Hepatocellular carcinoma (HCC) is a common type of liver cancer whose\nearly-stage diagnosis is a common challenge, mainly due to the manual\nassessment of hematoxylin and eosin-stained whole slide images, which is a\ntime-consuming process and may lead to variability in decision-making. For\naccurate detection of HCC, we propose a hybrid deep learning-based architecture\nthat uses transfer learning to extract the features from pre-trained\nconvolutional neural network (CNN) models and a classifier made up of a\nsequence of fully connected layers. This study uses a publicly available The\nCancer Genome Atlas Hepatocellular Carcinoma (TCGA-LIHC)database (n=491) for\nmodel development and database of Kasturba Gandhi Medical College (KMC), India\nfor validation. The pre-processing step involves patch extraction, colour\nnormalization, and augmentation that results in 3920 patches for the TCGA\ndataset. The developed hybrid deep neural network consisting of a CNN-based\npre-trained feature extractor and a customized artificial neural network-based\nclassifier is trained using five-fold cross-validation. For this study, eight\ndifferent state-of-the-art models are trained and tested as feature extractors\nfor the proposed hybrid model. The proposed hybrid model with ResNet50-based\nfeature extractor provided the sensitivity, specificity, F1-score, accuracy,\nand AUC of 100.00%, 100.00%, 100.00%, 100.00%, and 1.00, respectively on the\nTCGA database. On the KMC database, EfficientNetb3 resulted in the optimal\nchoice of the feature extractor giving sensitivity, specificity, F1-score,\naccuracy, and AUC of 96.97, 98.85, 96.71, 96.71, and 0.99, respectively. The\nproposed hybrid models showed improvement in accuracy of 2% and 4% over the\npre-trained models in TCGA-LIHC and KMC databases.\n", "date": "2024-12-04", "categories": []}
{"id": "2501.16239", "title": "Distilling foundation models for robust and efficient models in digital\n  pathology", "abstract": "  In recent years, the advent of foundation models (FM) for digital pathology\nhas relied heavily on scaling the pre-training datasets and the model size,\nyielding large and powerful models. While it resulted in improving the\nperformance on diverse downstream tasks, it also introduced increased\ncomputational cost and inference time. In this work, we explore the\ndistillation of a large foundation model into a smaller one, reducing the\nnumber of parameters by several orders of magnitude. Leveraging distillation\ntechniques, our distilled model, H0-mini, achieves nearly comparable\nperformance to large FMs at a significantly reduced inference cost. It is\nevaluated on several public benchmarks, achieving 3rd place on the HEST\nbenchmark and 5th place on the EVA benchmark. Additionally, a robustness\nanalysis conducted on the PLISM dataset demonstrates that our distilled model\nreaches excellent robustness to variations in staining and scanning conditions,\nsignificantly outperforming other state-of-the art models. This opens new\nperspectives to design lightweight and robust models for digital pathology,\nwithout compromising on performance.\n", "date": "2025-01-27", "categories": []}
{"id": "2502.20823", "title": "Can We Simplify Slide-level Fine-tuning of Pathology Foundation Models?", "abstract": "  The emergence of foundation models in computational pathology has transformed\nhistopathological image analysis, with whole slide imaging (WSI) diagnosis\nbeing a core application. Traditionally, weakly supervised fine-tuning via\nmultiple instance learning (MIL) has been the primary method for adapting\nfoundation models to WSIs. However, in this work we present a key experimental\nfinding: a simple nonlinear mapping strategy combining mean pooling and a\nmultilayer perceptron, called SiMLP, can effectively adapt patch-level\nfoundation models to slide-level tasks without complex MIL-based learning.\nThrough extensive experiments across diverse downstream tasks, we demonstrate\nthe superior performance of SiMLP with state-of-the-art methods. For instance,\non a large-scale pan-cancer classification task, SiMLP surpasses popular\nMIL-based methods by 3.52%. Furthermore, SiMLP shows strong learning ability in\nfew-shot classification and remaining highly competitive with slide-level\nfoundation models pretrained on tens of thousands of slides. Finally, SiMLP\nexhibits remarkable robustness and transferability in lung cancer subtyping.\nOverall, our findings challenge the conventional MIL-based fine-tuning\nparadigm, demonstrating that a task-agnostic representation strategy alone can\neffectively adapt foundation models to WSI analysis. These insights offer a\nunique and meaningful perspective for future research in digital pathology,\npaving the way for more efficient and broadly applicable methodologies.\n", "date": "2025-02-28", "categories": []}
{"id": "2502.20850", "title": "VLEER: Vision and Language Embeddings for Explainable Whole Slide Image\n  Representation", "abstract": "  Recent advances in vision-language models (VLMs) have shown remarkable\npotential in bridging visual and textual modalities. In computational\npathology, domain-specific VLMs, which are pre-trained on extensive\nhistopathology image-text datasets, have succeeded in various downstream tasks.\nHowever, existing research has primarily focused on the pre-training process\nand direct applications of VLMs on the patch level, leaving their great\npotential for whole slide image (WSI) applications unexplored. In this study,\nwe hypothesize that pre-trained VLMs inherently capture informative and\ninterpretable WSI representations through quantitative feature extraction. To\nvalidate this hypothesis, we introduce Vision and Language Embeddings for\nExplainable WSI Representation (VLEER), a novel method designed to leverage\nVLMs for WSI representation. We systematically evaluate VLEER on three\npathological WSI datasets, proving its better performance in WSI analysis\ncompared to conventional vision features. More importantly, VLEER offers the\nunique advantage of interpretability, enabling direct human-readable insights\ninto the results by leveraging the textual modality for detailed pathology\nannotations, providing clear reasoning for WSI-level pathology downstream\ntasks.\n", "date": "2025-02-28", "categories": []}
{"id": "2502.19293", "title": "Pathology Report Generation and Multimodal Representation Learning for\n  Cutaneous Melanocytic Lesions", "abstract": "  Millions of melanocytic skin lesions are examined by pathologists each year,\nthe majority of which concern common nevi (i.e., ordinary moles). While most of\nthese lesions can be diagnosed in seconds, writing the corresponding pathology\nreport is much more time-consuming. Automating part of the report writing\ncould, therefore, alleviate the increasing workload of pathologists. In this\nwork, we develop a vision-language model specifically for the pathology domain\nof cutaneous melanocytic lesions. The model follows the Contrastive Captioner\nframework and was trained and evaluated using a melanocytic lesion dataset of\n42,512 H&E-stained whole slide images and 19,645 corresponding pathology\nreports. Our results show that the quality scores of model-generated reports\nwere on par with pathologist-written reports for common nevi, assessed by an\nexpert pathologist in a reader study. While report generation revealed to be\nmore difficult for rare melanocytic lesion subtypes, the cross-modal retrieval\nperformance for these cases was considerably better.\n", "date": "2025-02-26", "categories": []}
{"id": "2502.19568", "title": "PhenoProfiler: Advancing Phenotypic Learning for Image-based Drug\n  Discovery", "abstract": "  In the field of image-based drug discovery, capturing the phenotypic response\nof cells to various drug treatments and perturbations is a crucial step.\nHowever, existing methods require computationally extensive and complex\nmulti-step procedures, which can introduce inefficiencies, limit\ngeneralizability, and increase potential errors. To address these challenges,\nwe present PhenoProfiler, an innovative model designed to efficiently and\neffectively extract morphological representations, enabling the elucidation of\nphenotypic changes induced by treatments. PhenoProfiler is designed as an\nend-to-end tool that processes whole-slide multi-channel images directly into\nlow-dimensional quantitative representations, eliminating the extensive\ncomputational steps required by existing methods. It also includes a\nmulti-objective learning module to enhance robustness, accuracy, and\ngeneralization in morphological representation learning. PhenoProfiler is\nrigorously evaluated on large-scale publicly available datasets, including over\n230,000 whole-slide multi-channel images in end-to-end scenarios and more than\n8.42 million single-cell images in non-end-to-end settings. Across these\nbenchmarks, PhenoProfiler consistently outperforms state-of-the-art methods by\nup to 20%, demonstrating substantial improvements in both accuracy and\nrobustness. Furthermore, PhenoProfiler uses a tailored phenotype correction\nstrategy to emphasize relative phenotypic changes under treatments,\nfacilitating the detection of biologically meaningful signals. UMAP\nvisualizations of treatment profiles demonstrate PhenoProfiler ability to\neffectively cluster treatments with similar biological annotations, thereby\nenhancing interpretability. These findings establish PhenoProfiler as a\nscalable, generalizable, and robust tool for phenotypic learning.\n", "date": "2025-02-26", "categories": []}
{"id": "2502.20156", "title": "Adaptive H&E-IHC information fusion staining framework based on feature\n  extra", "abstract": "  Immunohistochemistry (IHC) staining plays a significant role in the\nevaluation of diseases such as breast cancer. The H&E-to-IHC transformation\nbased on generative models provides a simple and cost-effective method for\nobtaining IHC images. Although previous models can perform digital coloring\nwell, they still suffer from (i) coloring only through the pixel features that\nare not prominent in HE, which is easy to cause information loss in the\ncoloring process; (ii) The lack of pixel-perfect H&E-IHC groundtruth pairs\nposes a challenge to the classical L1 loss.To address the above challenges, we\npropose an adaptive information enhanced coloring framework based on feature\nextractors. We first propose the VMFE module to effectively extract the color\ninformation features using multi-scale feature extraction and wavelet transform\nconvolution, while combining the shared decoder for feature fusion. The\nhigh-performance dual feature extractor of H&E-IHC is trained by contrastive\nlearning, which can effectively perform feature alignment of HE-IHC in high\nlatitude space. At the same time, the trained feature encoder is used to\nenhance the features and adaptively adjust the loss in the HE section staining\nprocess to solve the problems related to unclear and asymmetric information. We\nhave tested on different datasets and achieved excellent performance.Our code\nis available at https://github.com/babyinsunshine/CEFF\n", "date": "2025-02-27", "categories": []}
{"id": "2501.15724", "title": "A Survey on Computational Pathology Foundation Models: Datasets,\n  Adaptation Strategies, and Evaluation Tasks", "abstract": "  Computational pathology foundation models (CPathFMs) have emerged as a\npowerful approach for analyzing histopathological data, leveraging\nself-supervised learning to extract robust feature representations from\nunlabeled whole-slide images. These models, categorized into uni-modal and\nmulti-modal frameworks, have demonstrated promise in automating complex\npathology tasks such as segmentation, classification, and biomarker discovery.\nHowever, the development of CPathFMs presents significant challenges, such as\nlimited data accessibility, high variability across datasets, the necessity for\ndomain-specific adaptation, and the lack of standardized evaluation benchmarks.\nThis survey provides a comprehensive review of CPathFMs in computational\npathology, focusing on datasets, adaptation strategies, and evaluation tasks.\nWe analyze key techniques, such as contrastive learning and multi-modal\nintegration, and highlight existing gaps in current research. Finally, we\nexplore future directions from four perspectives for advancing CPathFMs. This\nsurvey serves as a valuable resource for researchers, clinicians, and AI\npractitioners, guiding the advancement of CPathFMs toward robust and clinically\napplicable AI-driven pathology solutions.\n", "date": "2025-01-27", "categories": []}
{"id": "2502.19123", "title": "From Traditional to Deep Learning Approaches in Whole Slide Image\n  Registration: A Methodological Review", "abstract": "  Whole slide image (WSI) registration is an essential task for analysing the\ntumour microenvironment (TME) in histopathology. It involves the alignment of\nspatial information between WSIs of the same section or serial sections of a\ntissue sample. The tissue sections are usually stained with single or multiple\nbiomarkers before imaging, and the goal is to identify neighbouring nuclei\nalong the Z-axis for creating a 3D image or identifying subclasses of cells in\nthe TME. This task is considerably more challenging compared to radiology image\nregistration, such as magnetic resonance imaging or computed tomography, due to\nvarious factors. These include gigapixel size of images, variations in\nappearance between differently stained tissues, changes in structure and\nmorphology between non-consecutive sections, and the presence of artefacts,\ntears, and deformations. Currently, there is a noticeable gap in the literature\nregarding a review of the current approaches and their limitations, as well as\nthe challenges and opportunities they present. We aim to provide a\ncomprehensive understanding of the available approaches and their application\nfor various purposes. Furthermore, we investigate current deep learning methods\nused for WSI registration, emphasising their diverse methodologies. We examine\nthe available datasets and explore tools and software employed in the field.\nFinally, we identify open challenges and potential future trends in this area\nof research.\n", "date": "2025-02-26", "categories": []}
{"id": "2502.16162", "title": "Patch Stitching Data Augmentation for Cancer Classification in Pathology\n  Images", "abstract": "  Computational pathology, integrating computational methods and digital\nimaging, has shown to be effective in advancing disease diagnosis and\nprognosis. In recent years, the development of machine learning and deep\nlearning has greatly bolstered the power of computational pathology. However,\nthere still remains the issue of data scarcity and data imbalance, which can\nhave an adversarial effect on any computational method. In this paper, we\nintroduce an efficient and effective data augmentation strategy to generate new\npathology images from the existing pathology images and thus enrich datasets\nwithout additional data collection or annotation costs. To evaluate the\nproposed method, we employed two sets of colorectal cancer datasets and\nobtained improved classification results, suggesting that the proposed simple\napproach holds the potential for alleviating the data scarcity and imbalance in\ncomputational pathology.\n", "date": "2025-02-22", "categories": []}
{"id": "2412.10853", "title": "SEW: Self-calibration Enhanced Whole Slide Pathology Image Analysis", "abstract": "  Pathology images are considered the ``gold standard\" for cancer diagnosis and\ntreatment, with gigapixel images providing extensive tissue and cellular\ninformation. Existing methods fail to simultaneously extract global structural\nand local detail features for comprehensive pathology image analysis\nefficiently. To address these limitations, we propose a self-calibration\nenhanced framework for whole slide pathology image analysis, comprising three\ncomponents: a global branch, a focus predictor, and a detailed branch. The\nglobal branch initially classifies using the pathological thumbnail, while the\nfocus predictor identifies relevant regions for classification based on the\nlast layer features of the global branch. The detailed extraction branch then\nassesses whether the magnified regions correspond to the lesion area. Finally,\na feature consistency constraint between the global and detail branches ensures\nthat the global branch focuses on the appropriate region and extracts\nsufficient discriminative features for final identification. These focused\ndiscriminative features prove invaluable for uncovering novel prognostic tumor\nmarkers from the perspective of feature cluster uniqueness and tissue spatial\ndistribution. Extensive experiment results demonstrate that the proposed\nframework can rapidly deliver accurate and explainable results for pathological\ngrading and prognosis tasks.\n", "date": "2024-12-14", "categories": []}
{"id": "2502.09662", "title": "Generalizable Cervical Cancer Screening via Large-scale Pretraining and\n  Test-Time Adaptation", "abstract": "  Cervical cancer is a leading malignancy in female reproductive system. While\nAI-assisted cytology offers a cost-effective and non-invasive screening\nsolution, current systems struggle with generalizability in complex clinical\nscenarios. To address this issue, we introduced Smart-CCS, a generalizable\nCervical Cancer Screening paradigm based on pretraining and adaptation to\ncreate robust and generalizable screening systems. To develop and validate\nSmart-CCS, we first curated a large-scale, multi-center dataset named CCS-127K,\nwhich comprises a total of 127,471 cervical cytology whole-slide images\ncollected from 48 medical centers. By leveraging large-scale self-supervised\npretraining, our CCS models are equipped with strong generalization capability,\npotentially generalizing across diverse scenarios. Then, we incorporated\ntest-time adaptation to specifically optimize the trained CCS model for complex\nclinical settings, which adapts and refines predictions, improving real-world\napplicability. We conducted large-scale system evaluation among various\ncohorts. In retrospective cohorts, Smart-CCS achieved an overall area under the\ncurve (AUC) value of 0.965 and sensitivity of 0.913 for cancer screening on 11\ninternal test datasets. In external testing, system performance maintained high\nat 0.950 AUC across 6 independent test datasets. In prospective cohorts, our\nSmart-CCS achieved AUCs of 0.947, 0.924, and 0.986 in three prospective\ncenters, respectively. Moreover, the system demonstrated superior sensitivity\nin diagnosing cervical cancer, confirming the accuracy of our cancer screening\nresults by using histology findings for validation. Interpretability analysis\nwith cell and slide predictions further indicated that the system's\ndecision-making aligns with clinical practice. Smart-CCS represents a\nsignificant advancement in cancer screening across diverse clinical contexts.\n", "date": "2025-02-12", "categories": []}
{"id": "2502.07979", "title": "Joint Modelling Histology and Molecular Markers for Cancer\n  Classification", "abstract": "  Cancers are characterized by remarkable heterogeneity and diverse prognosis.\nAccurate cancer classification is essential for patient stratification and\nclinical decision-making. Although digital pathology has been advancing cancer\ndiagnosis and prognosis, the paradigm in cancer pathology has shifted from\npurely relying on histology features to incorporating molecular markers. There\nis an urgent need for digital pathology methods to meet the needs of the new\nparadigm. We introduce a novel digital pathology approach to jointly predict\nmolecular markers and histology features and model their interactions for\ncancer classification. Firstly, to mitigate the challenge of\ncross-magnification information propagation, we propose a multi-scale\ndisentangling module, enabling the extraction of multi-scale features from\nhigh-magnification (cellular-level) to low-magnification (tissue-level) whole\nslide images. Further, based on the multi-scale features, we propose an\nattention-based hierarchical multi-task multi-instance learning framework to\nsimultaneously predict histology and molecular markers. Moreover, we propose a\nco-occurrence probability-based label correlation graph network to model the\nco-occurrence of molecular markers. Lastly, we design a cross-modal interaction\nmodule with the dynamic confidence constrain loss and a cross-modal gradient\nmodulation strategy, to model the interactions of histology and molecular\nmarkers. Our experiments demonstrate that our method outperforms other\nstate-of-the-art methods in classifying glioma, histology features and\nmolecular markers. Our method promises to promote precise oncology with the\npotential to advance biomedical research and clinical applications. The code is\navailable at https://github.com/LHY1007/M3C2\n", "date": "2025-02-11", "categories": []}
{"id": "2502.08333", "title": "Foundation Models in Computational Pathology: A Review of Challenges,\n  Opportunities, and Impact", "abstract": "  From self-supervised, vision-only models to contrastive visual-language\nframeworks, computational pathology has rapidly evolved in recent years.\nGenerative AI \"co-pilots\" now demonstrate the ability to mine subtle,\nsub-visual tissue cues across the cellular-to-pathology spectrum, generate\ncomprehensive reports, and respond to complex user queries. The scale of data\nhas surged dramatically, growing from tens to millions of multi-gigapixel\ntissue images, while the number of trainable parameters in these models has\nrisen to several billion. The critical question remains: how will this new wave\nof generative and multi-purpose AI transform clinical diagnostics? In this\narticle, we explore the true potential of these innovations and their\nintegration into clinical practice. We review the rapid progress of foundation\nmodels in pathology, clarify their applications and significance. More\nprecisely, we examine the very definition of foundational models, identifying\nwhat makes them foundational, general, or multipurpose, and assess their impact\non computational pathology. Additionally, we address the unique challenges\nassociated with their development and evaluation. These models have\ndemonstrated exceptional predictive and generative capabilities, but\nestablishing global benchmarks is crucial to enhancing evaluation standards and\nfostering their widespread clinical adoption. In computational pathology, the\nbroader impact of frontier AI ultimately depends on widespread adoption and\nsocietal acceptance. While direct public exposure is not strictly necessary, it\nremains a powerful tool for dispelling misconceptions, building trust, and\nsecuring regulatory support.\n", "date": "2025-02-12", "categories": []}
{"id": "2409.09369", "title": "Interpretable Vision-Language Survival Analysis with Ordinal Inductive\n  Bias for Computational Pathology", "abstract": "  Histopathology Whole-Slide Images (WSIs) provide an important tool to assess\ncancer prognosis in computational pathology (CPATH). While existing survival\nanalysis (SA) approaches have made exciting progress, they are generally\nlimited to adopting highly-expressive network architectures and only\ncoarse-grained patient-level labels to learn visual prognostic representations\nfrom gigapixel WSIs. Such learning paradigm suffers from critical performance\nbottlenecks, when facing present scarce training data and standard\nmulti-instance learning (MIL) framework in CPATH. To overcome it, this paper,\nfor the first time, proposes a new Vision-Language-based SA (VLSA) paradigm.\nConcretely, (1) VLSA is driven by pathology VL foundation models. It no longer\nrelies on high-capability networks and shows the advantage of data efficiency.\n(2) In vision-end, VLSA encodes textual prognostic prior and then employs it as\nauxiliary signals to guide the aggregating of visual prognostic features at\ninstance level, thereby compensating for the weak supervision in MIL. Moreover,\ngiven the characteristics of SA, we propose i) ordinal survival prompt learning\nto transform continuous survival labels into textual prompts; and ii) ordinal\nincidence function as prediction target to make SA compatible with VL-based\nprediction. Notably, VLSA's predictions can be interpreted intuitively by our\nShapley values-based method. The extensive experiments on five datasets confirm\nthe effectiveness of our scheme. Our VLSA could pave a new way for SA in CPATH\nby offering weakly-supervised MIL an effective means to learn valuable\nprognostic clues from gigapixel WSIs. Our source code is available at\nhttps://github.com/liupei101/VLSA.\n", "date": "2024-09-14", "categories": []}
{"id": "2411.04782", "title": "A General Pipeline for Glomerulus Whole-Slide Image Segmentation", "abstract": "  Whole-slide images (WSI) glomerulus segmentation is essential for accurately\ndiagnosing kidney diseases. In this work, we propose a general and practical\npipeline for glomerulus segmentation that effectively enhances both patch-level\nand WSI-level segmentation tasks. Our approach leverages stitching on\noverlapping patches, increasing the detection coverage, especially when\nglomeruli are located near patch image borders. In addition, we conduct\ncomprehensive evaluations from different segmentation models across two large\nand diverse datasets with over 30K glomerulus annotations. Experimental results\ndemonstrate that models using our pipeline outperform the previous\nstate-of-the-art method, achieving superior results across both datasets and\nsetting a new benchmark for glomerulus segmentation in WSIs. The code and\npre-trained models are available at\nhttps://github.com/huuquan1994/wsi_glomerulus_seg.\n", "date": "2024-11-07", "categories": []}
{"id": "2502.07288", "title": "KPIs 2024 Challenge: Advancing Glomerular Segmentation from Patch- to\n  Slide-Level", "abstract": "  Chronic kidney disease (CKD) is a major global health issue, affecting over\n10% of the population and causing significant mortality. While kidney biopsy\nremains the gold standard for CKD diagnosis and treatment, the lack of\ncomprehensive benchmarks for kidney pathology segmentation hinders progress in\nthe field. To address this, we organized the Kidney Pathology Image\nSegmentation (KPIs) Challenge, introducing a dataset that incorporates\npreclinical rodent models of CKD with over 10,000 annotated glomeruli from 60+\nPeriodic Acid Schiff (PAS)-stained whole slide images. The challenge includes\ntwo tasks, patch-level segmentation and whole slide image segmentation and\ndetection, evaluated using the Dice Similarity Coefficient (DSC) and F1-score.\nBy encouraging innovative segmentation methods that adapt to diverse CKD models\nand tissue conditions, the KPIs Challenge aims to advance kidney pathology\nanalysis, establish new benchmarks, and enable precise, large-scale\nquantification for disease research and diagnosis.\n", "date": "2025-02-11", "categories": []}
{"id": "2409.04175", "title": "CISCA and CytoDArk0: a Cell Instance Segmentation and Classification\n  method for histo(patho)logical image Analyses and a new, open, Nissl-stained\n  dataset for brain cytoarchitecture studies", "abstract": "  Delineating and classifying individual cells in microscopy tissue images is\ninherently challenging yet remains essential for advancements in medical and\nneuroscientific research. In this work, we propose a new deep learning\nframework, CISCA, for automatic cell instance segmentation and classification\nin histological slices. At the core of CISCA is a network architecture\nfeaturing a lightweight U-Net with three heads in the decoder. The first head\nclassifies pixels into boundaries between neighboring cells, cell bodies, and\nbackground, while the second head regresses four distance maps along four\ndirections. The outputs from the first and second heads are integrated through\na tailored post-processing step, which ultimately produces the segmentation of\nindividual cells. The third head enables the simultaneous classification of\ncells into relevant classes, if required. We demonstrate the effectiveness of\nour method using four datasets, including CoNIC, PanNuke, and MoNuSeg, which\nare publicly available H&Estained datasets that cover diverse tissue types and\nmagnifications. In addition, we introduce CytoDArk0, the first annotated\ndataset of Nissl-stained histological images of the mammalian brain, containing\nnearly 40k annotated neurons and glia cells, aimed at facilitating advancements\nin digital neuropathology and brain cytoarchitecture studies. We evaluate CISCA\nagainst other state-of-the-art methods, demonstrating its versatility,\nrobustness, and accuracy in segmenting and classifying cells across diverse\ntissue types, magnifications, and staining techniques. This makes CISCA\nwell-suited for detailed analyses of cell morphology and efficient cell\ncounting in both digital pathology workflows and brain cytoarchitecture\nresearch.\n", "date": "2024-09-06", "categories": []}
{"id": "2502.06750", "title": "Accelerating Data Processing and Benchmarking of AI Models for Pathology", "abstract": "  Advances in foundation modeling have reshaped computational pathology.\nHowever, the increasing number of available models and lack of standardized\nbenchmarks make it increasingly complex to assess their strengths, limitations,\nand potential for further development. To address these challenges, we\nintroduce a new suite of software tools for whole-slide image processing,\nfoundation model benchmarking, and curated publicly available tasks. We\nanticipate that these resources will promote transparency, reproducibility, and\ncontinued progress in the field.\n", "date": "2025-02-10", "categories": []}
{"id": "2408.06381", "title": "Assessment of Cell Nuclei AI Foundation Models in Kidney Pathology", "abstract": "  Cell nuclei instance segmentation is a crucial task in digital kidney\npathology. Traditional automatic segmentation methods often lack\ngeneralizability when applied to unseen datasets. Recently, the success of\nfoundation models (FMs) has provided a more generalizable solution, potentially\nenabling the segmentation of any cell type. In this study, we perform a\nlarge-scale evaluation of three widely used state-of-the-art (SOTA) cell nuclei\nfoundation models (Cellpose, StarDist, and CellViT). Specifically, we created a\nhighly diverse evaluation dataset consisting of 2,542 kidney whole slide images\n(WSIs) collected from both human and rodent sources, encompassing various\ntissue types, sizes, and staining methods. To our knowledge, this is the\nlargest-scale evaluation of its kind to date. Our quantitative analysis of the\nprediction distribution reveals a persistent performance gap in kidney\npathology. Among the evaluated models, CellViT demonstrated superior\nperformance in segmenting nuclei in kidney pathology. However, none of the\nfoundation models are perfect; a performance gap remains in general nuclei\nsegmentation for kidney pathology.\n", "date": "2024-08-09", "categories": []}
{"id": "2501.18055", "title": "Current Pathology Foundation Models are unrobust to Medical Center\n  Differences", "abstract": "  Pathology Foundation Models (FMs) hold great promise for healthcare. Before\nthey can be used in clinical practice, it is essential to ensure they are\nrobust to variations between medical centers. We measure whether pathology FMs\nfocus on biological features like tissue and cancer type, or on the well known\nconfounding medical center signatures introduced by staining procedure and\nother differences. We introduce the Robustness Index. This novel robustness\nmetric reflects to what degree biological features dominate confounding\nfeatures. Ten current publicly available pathology FMs are evaluated. We find\nthat all current pathology foundation models evaluated represent the medical\ncenter to a strong degree. Significant differences in the robustness index are\nobserved. Only one model so far has a robustness index greater than one,\nmeaning biological features dominate confounding features, but only slightly. A\nquantitative approach to measure the influence of medical center differences on\nFM-based prediction performance is described. We analyze the impact of\nunrobustness on classification performance of downstream models, and find that\ncancer-type classification errors are not random, but specifically attributable\nto same-center confounders: images of other classes from the same medical\ncenter. We visualize FM embedding spaces, and find these are more strongly\norganized by medical centers than by biological factors. As a consequence, the\nmedical center of origin is predicted more accurately than the tissue source\nand cancer type. The robustness index introduced here is provided with the aim\nof advancing progress towards clinical adoption of robust and reliable\npathology FMs.\n", "date": "2025-01-29", "categories": []}
{"id": "2501.19048", "title": "The Role of Graph-based MIL and Interventional Training in the\n  Generalization of WSI Classifiers", "abstract": "  Whole Slide Imaging (WSI), which involves high-resolution digital scans of\npathology slides, has become the gold standard for cancer diagnosis, but its\ngigapixel resolution and the scarcity of annotated datasets present challenges\nfor deep learning models. Multiple Instance Learning (MIL), a widely-used\nweakly supervised approach, bypasses the need for patch-level annotations.\nHowever, conventional MIL methods overlook the spatial relationships between\npatches, which are crucial for tasks such as cancer grading and diagnosis. To\naddress this, graph-based approaches have gained prominence by incorporating\nspatial information through node connections. Despite their potential, both MIL\nand graph-based models are vulnerable to learning spurious associations, like\ncolor variations in WSIs, affecting their robustness. In this dissertation, we\nconduct an extensive comparison of multiple graph construction techniques, MIL\nmodels, graph-MIL approaches, and interventional training, introducing a new\nframework, Graph-based Multiple Instance Learning with Interventional Training\n(GMIL-IT), for WSI classification. We evaluate their impact on model\ngeneralization through domain shift analysis and demonstrate that graph-based\nmodels alone achieve the generalization initially anticipated from\ninterventional training. Our code is available here:\ngithub.com/ritamartinspereira/GMIL-IT\n", "date": "2025-01-31", "categories": []}
{"id": "2501.16652", "title": "Molecular-driven Foundation Model for Oncologic Pathology", "abstract": "  Foundation models are reshaping computational pathology by enabling transfer\nlearning, where models pre-trained on vast datasets can be adapted for\ndownstream diagnostic, prognostic, and therapeutic response tasks. Despite\nthese advances, foundation models are still limited in their ability to encode\nthe entire gigapixel whole-slide images without additional training and often\nlack complementary multimodal data. Here, we introduce Threads, a slide-level\nfoundation model capable of generating universal representations of whole-slide\nimages of any size. Threads was pre-trained using a multimodal learning\napproach on a diverse cohort of 47,171 hematoxylin and eosin (H&E)-stained\ntissue sections, paired with corresponding genomic and transcriptomic profiles\n- the largest such paired dataset to be used for foundation model development\nto date. This unique training paradigm enables Threads to capture the tissue's\nunderlying molecular composition, yielding powerful representations applicable\nto a wide array of downstream tasks. In extensive benchmarking across 54\noncology tasks, including clinical subtyping, grading, mutation prediction,\nimmunohistochemistry status determination, treatment response prediction, and\nsurvival prediction, Threads outperformed all baselines while demonstrating\nremarkable generalizability and label efficiency. It is particularly well\nsuited for predicting rare events, further emphasizing its clinical utility. We\nintend to make the model publicly available for the broader community.\n", "date": "2025-01-28", "categories": []}
{"id": "2501.15743", "title": "Z-Stack Scanning can Improve AI Detection of Mitosis: A Case Study of\n  Meningiomas", "abstract": "  Z-stack scanning is an emerging whole slide imaging technology that captures\nmultiple focal planes alongside the z-axis of a glass slide. Because z-stacking\ncan offer enhanced depth information compared to the single-layer whole slide\nimaging, this technology can be particularly useful in analyzing small-scaled\nhistopathological patterns. However, its actual clinical impact remains debated\nwith mixed results. To clarify this, we investigate the effect of z-stack\nscanning on artificial intelligence (AI) mitosis detection of meningiomas. With\nthe same set of 22 Hematoxylin and Eosin meningioma glass slides scanned by\nthree different digital pathology scanners, we tested the performance of three\nAI pipelines on both single-layer and z-stacked whole slide images (WSIs).\nResults showed that in all scanner-AI combinations, z-stacked WSIs\nsignificantly increased AI's sensitivity (+17.14%) on the mitosis detection\nwith only a marginal impact on precision. Our findings provide quantitative\nevidence that highlights z-stack scanning as a promising technique for AI\nmitosis detection, paving the way for more reliable AI-assisted pathology\nworkflows, which can ultimately benefit patient management.\n", "date": "2025-01-27", "categories": []}
{"id": "2501.14056", "title": "Prior Knowledge Injection into Deep Learning Models Predicting Gene\n  Expression from Whole Slide Images", "abstract": "  Cancer diagnosis and prognosis primarily depend on clinical parameters such\nas age and tumor grade, and are increasingly complemented by molecular data,\nsuch as gene expression, from tumor sequencing. However, sequencing is costly\nand delays oncology workflows. Recent advances in Deep Learning allow to\npredict molecular information from morphological features within Whole Slide\nImages (WSIs), offering a cost-effective proxy of the molecular markers. While\npromising, current methods lack the robustness to fully replace direct\nsequencing. Here we aim to improve existing methods by introducing a\nmodel-agnostic framework that allows to inject prior knowledge on gene-gene\ninteractions into Deep Learning architectures, thereby increasing accuracy and\nrobustness. We design the framework to be generic and flexibly adaptable to a\nwide range of architectures. In a case study on breast cancer, our strategy\nleads to an average increase of 983 significant genes (out of 25,761) across\nall 18 experiments, with 14 generalizing to an increase on an independent\ndataset. Our findings reveal a high potential for injection of prior knowledge\nto increase gene expression prediction performance from WSIs across a wide\nrange of architectures.\n", "date": "2025-01-23", "categories": []}
{"id": "2501.14379", "title": "ECTIL: Label-efficient Computational Tumour Infiltrating Lymphocyte\n  (TIL) assessment in breast cancer: Multicentre validation in 2,340 patients\n  with breast cancer", "abstract": "  The level of tumour-infiltrating lymphocytes (TILs) is a prognostic factor\nfor patients with (triple-negative) breast cancer (BC). Computational TIL\nassessment (CTA) has the potential to assist pathologists in this\nlabour-intensive task, but current CTA models rely heavily on many detailed\nannotations. We propose and validate a fundamentally simpler deep learning\nbased CTA that can be trained in only ten minutes on hundredfold fewer\npathologist annotations. We collected whole slide images (WSIs) with TILs\nscores and clinical data of 2,340 patients with BC from six cohorts including\nthree randomised clinical trials. Morphological features were extracted from\nwhole slide images (WSIs) using a pathology foundation model. Our\nlabel-efficient Computational stromal TIL assessment model (ECTIL) directly\nregresses the TILs score from these features. ECTIL trained on only a few\nhundred samples (ECTIL-TCGA) showed concordance with the pathologist over five\nheterogeneous external cohorts (r=0.54-0.74, AUROC=0.80-0.94). Training on all\nslides of five cohorts (ECTIL-combined) improved results on a held-out test set\n(r=0.69, AUROC=0.85). Multivariable Cox regression analyses indicated that\nevery 10% increase of ECTIL scores was associated with improved overall\nsurvival independent of clinicopathological variables (HR 0.86, p<0.01),\nsimilar to the pathologist score (HR 0.87, p<0.001). We demonstrate that ECTIL\nis highly concordant with an expert pathologist and obtains a similar hazard\nratio. ECTIL has a fundamentally simpler design than existing methods and can\nbe trained on orders of magnitude fewer annotations. Such a CTA may be used to\npre-screen patients for, e.g., immunotherapy clinical trial inclusion, or as a\ntool to assist clinicians in the diagnostic work-up of patients with BC. Our\nmodel is available under an open source licence\n(https://github.com/nki-ai/ectil).\n", "date": "2025-01-24", "categories": []}
{"id": "2501.08042", "title": "Exploring visual language models as a powerful tool in the diagnosis of\n  Ewing Sarcoma", "abstract": "  Ewing's sarcoma (ES), characterized by a high density of small round blue\ncells without structural organization, presents a significant health concern,\nparticularly among adolescents aged 10 to 19. Artificial intelligence-based\nsystems for automated analysis of histopathological images are promising to\ncontribute to an accurate diagnosis of ES. In this context, this study explores\nthe feature extraction ability of different pre-training strategies for\ndistinguishing ES from other soft tissue or bone sarcomas with similar\nmorphology in digitized tissue microarrays for the first time, as far as we\nknow. Vision-language supervision (VLS) is compared to fully-supervised\nImageNet pre-training within a multiple instance learning paradigm. Our\nfindings indicate a substantial improvement in diagnostic accuracy with the\nadaption of VLS using an in-domain dataset. Notably, these models not only\nenhance the accuracy of predicted classes but also drastically reduce the\nnumber of trainable parameters and computational costs.\n", "date": "2025-01-14", "categories": []}
{"id": "2501.08094", "title": "CellOMaps: A Compact Representation for Robust Classification of Lung\n  Adenocarcinoma Growth Patterns", "abstract": "  Lung adenocarcinoma (LUAD) is a morphologically heterogeneous disease,\ncharacterized by five primary histological growth patterns. The classification\nof such patterns is crucial due to their direct relation to prognosis but the\nhigh subjectivity and observer variability pose a major challenge. Although\nseveral studies have developed machine learning methods for growth pattern\nclassification, they either only report the predominant pattern per slide or\nlack proper evaluation. We propose a generalizable machine learning pipeline\ncapable of classifying lung tissue into one of the five patterns or as\nnon-tumor. The proposed pipeline's strength lies in a novel compact Cell\nOrganization Maps (cellOMaps) representation that captures the cellular spatial\npatterns from Hematoxylin and Eosin whole slide images (WSIs). The proposed\npipeline provides state-of-the-art performance on LUAD growth pattern\nclassification when evaluated on both internal unseen slides and external\ndatasets, significantly outperforming the current approaches. In addition, our\npreliminary results show that the model's outputs can be used to predict\npatients Tumor Mutational Burden (TMB) levels.\n", "date": "2025-01-14", "categories": []}
{"id": "2312.04746", "title": "Quilt-LLaVA: Visual Instruction Tuning by Extracting Localized\n  Narratives from Open-Source Histopathology Videos", "abstract": "  Diagnosis in histopathology requires a global whole slide images (WSIs)\nanalysis, requiring pathologists to compound evidence from different WSI\npatches. The gigapixel scale of WSIs poses a challenge for histopathology\nmulti-modal models. Training multi-model models for histopathology requires\ninstruction tuning datasets, which currently contain information for individual\nimage patches, without a spatial grounding of the concepts within each patch\nand without a wider view of the WSI. Therefore, they lack sufficient diagnostic\ncapacity for histopathology. To bridge this gap, we introduce Quilt-Instruct, a\nlarge-scale dataset of 107,131 histopathology-specific instruction\nquestion/answer pairs, grounded within diagnostically relevant image patches\nthat make up the WSI. Our dataset is collected by leveraging educational\nhistopathology videos from YouTube, which provides spatial localization of\nnarrations by automatically extracting the narrators' cursor positions.\nQuilt-Instruct supports contextual reasoning by extracting diagnosis and\nsupporting facts from the entire WSI. Using Quilt-Instruct, we train\nQuilt-LLaVA, which can reason beyond the given single image patch, enabling\ndiagnostic reasoning across patches. To evaluate Quilt-LLaVA, we propose a\ncomprehensive evaluation dataset created from 985 images and 1283\nhuman-generated question-answers. We also thoroughly evaluate Quilt-LLaVA using\npublic histopathology datasets, where Quilt-LLaVA significantly outperforms\nSOTA by over 10% on relative GPT-4 score and 4% and 9% on open and closed set\nVQA. Our code, data, and model are publicly accessible at\nquilt-llava.github.io.\n", "date": "2023-12-07", "categories": []}
{"id": "2409.03519", "title": "Tissue Concepts: supervised foundation models in computational pathology", "abstract": "  Due to the increasing workload of pathologists, the need for automation to\nsupport diagnostic tasks and quantitative biomarker evaluation is becoming more\nand more apparent. Foundation models have the potential to improve\ngeneralizability within and across centers and serve as starting points for\ndata efficient development of specialized yet robust AI models. However, the\ntraining foundation models themselves is usually very expensive in terms of\ndata, computation, and time. This paper proposes a supervised training method\nthat drastically reduces these expenses. The proposed method is based on\nmulti-task learning to train a joint encoder, by combining 16 different\nclassification, segmentation, and detection tasks on a total of 912,000\npatches. Since the encoder is capable of capturing the properties of the\nsamples, we term it the Tissue Concepts encoder. To evaluate the performance\nand generalizability of the Tissue Concepts encoder across centers,\nclassification of whole slide images from four of the most prevalent solid\ncancers - breast, colon, lung, and prostate - was used. The experiments show\nthat the Tissue Concepts model achieve comparable performance to models trained\nwith self-supervision, while requiring only 6% of the amount of training\npatches. Furthermore, the Tissue Concepts encoder outperforms an ImageNet\npre-trained encoder on both in-domain and out-of-domain data.\n", "date": "2024-09-05", "categories": []}
{"id": "2501.05409", "title": "Atlas: A Novel Pathology Foundation Model by Mayo Clinic, Charit\\'e, and\n  Aignostics", "abstract": "  Recent advances in digital pathology have demonstrated the effectiveness of\nfoundation models across diverse applications. In this report, we present\nAtlas, a novel vision foundation model based on the RudolfV approach. Our model\nwas trained on a dataset comprising 1.2 million histopathology whole slide\nimages, collected from two medical institutions: Mayo Clinic and Charit\\'e -\nUniverst\\\"atsmedizin Berlin. Comprehensive evaluations show that Atlas achieves\nstate-of-the-art performance across twenty-one public benchmark datasets, even\nthough it is neither the largest model by parameter count nor by training\ndataset size.\n", "date": "2025-01-09", "categories": []}
{"id": "2501.05945", "title": "Reusable specimen-level inference in computational pathology", "abstract": "  Foundation models for computational pathology have shown great promise for\nspecimen-level tasks and are increasingly accessible to researchers. However,\nspecimen-level models built on these foundation models remain largely\nunavailable, hindering their broader utility and impact. To address this gap,\nwe developed SpinPath, a toolkit designed to democratize specimen-level deep\nlearning by providing a zoo of pretrained specimen-level models, a Python-based\ninference engine, and a JavaScript-based inference platform. We demonstrate the\nutility of SpinPath in metastasis detection tasks across nine foundation\nmodels. SpinPath may foster reproducibility, simplify experimentation, and\naccelerate the adoption of specimen-level deep learning in computational\npathology research.\n", "date": "2025-01-10", "categories": []}
{"id": "2501.06151", "title": "PySpatial: A High-Speed Whole Slide Image Pathomics Toolkit", "abstract": "  Whole Slide Image (WSI) analysis plays a crucial role in modern digital\npathology, enabling large-scale feature extraction from tissue samples.\nHowever, traditional feature extraction pipelines based on tools like\nCellProfiler often involve lengthy workflows, requiring WSI segmentation into\npatches, feature extraction at the patch level, and subsequent mapping back to\nthe original WSI. To address these challenges, we present PySpatial, a\nhigh-speed pathomics toolkit specifically designed for WSI-level analysis.\nPySpatial streamlines the conventional pipeline by directly operating on\ncomputational regions of interest, reducing redundant processing steps.\nUtilizing rtree-based spatial indexing and matrix-based computation, PySpatial\nefficiently maps and processes computational regions, significantly\naccelerating feature extraction while maintaining high accuracy. Our\nexperiments on two datasets-Perivascular Epithelioid Cell (PEC) and data from\nthe Kidney Precision Medicine Project (KPMP)-demonstrate substantial\nperformance improvements. For smaller and sparse objects in PEC datasets,\nPySpatial achieves nearly a 10-fold speedup compared to standard CellProfiler\npipelines. For larger objects, such as glomeruli and arteries in KPMP datasets,\nPySpatial achieves a 2-fold speedup. These results highlight PySpatial's\npotential to handle large-scale WSI analysis with enhanced efficiency and\naccuracy, paving the way for broader applications in digital pathology.\n", "date": "2025-01-10", "categories": []}
{"id": "2501.03592", "title": "A Value Mapping Virtual Staining Framework for Large-scale Histological\n  Imaging", "abstract": "  The emergence of virtual staining technology provides a rapid and efficient\nalternative for researchers in tissue pathology. It enables the utilization of\nunlabeled microscopic samples to generate virtual replicas of chemically\nstained histological slices, or facilitate the transformation of one staining\ntype into another. The remarkable performance of generative networks, such as\nCycleGAN, offers an unsupervised learning approach for virtual coloring,\novercoming the limitations of high-quality paired data required in supervised\nlearning. Nevertheless, large-scale color transformation necessitates\nprocessing large field-of-view images in patches, often resulting in\nsignificant boundary inconsistency and artifacts. Additionally, the\ntransformation between different colorized modalities typically needs further\nefforts to modify loss functions and tune hyperparameters for independent\ntraining of networks. In this study, we introduce a general virtual staining\nframework that is adaptable to various conditions. We propose a loss function\nbased on the value mapping constraint to ensure the accuracy of virtual\ncoloring between different pathological modalities, termed the Value Mapping\nGenerative Adversarial Network (VM-GAN). Meanwhile, we present a\nconfidence-based tiling method to address the challenge of boundary\ninconsistency arising from patch-wise processing. Experimental results on\ndiverse data with varying staining protocols demonstrate that our method\nachieves superior quantitative indicators and improved visual perception.\n", "date": "2025-01-07", "categories": []}
{"id": "2501.02922", "title": "Label-free Concept Based Multiple Instance Learning for Gigapixel\n  Histopathology", "abstract": "  Multiple Instance Learning (MIL) methods allow for gigapixel Whole-Slide\nImage (WSI) analysis with only slide-level annotations. Interpretability is\ncrucial for safely deploying such algorithms in high-stakes medical domains.\nTraditional MIL methods offer explanations by highlighting salient regions.\nHowever, such spatial heatmaps provide limited insights for end users. To\naddress this, we propose a novel inherently interpretable WSI-classification\napproach that uses human-understandable pathology concepts to generate\nexplanations. Our proposed Concept MIL model leverages recent advances in\nvision-language models to directly predict pathology concepts based on image\nfeatures. The model's predictions are obtained through a linear combination of\nthe concepts identified on the top-K patches of a WSI, enabling inherent\nexplanations by tracing each concept's influence on the prediction. In contrast\nto traditional concept-based interpretable models, our approach eliminates the\nneed for costly human annotations by leveraging the vision-language model. We\nvalidate our method on two widely used pathology datasets: Camelyon16 and\nPANDA. On both datasets, Concept MIL achieves AUC and accuracy scores over 0.9,\nputting it on par with state-of-the-art models. We further find that 87.1\\%\n(Camelyon16) and 85.3\\% (PANDA) of the top 20 patches fall within the tumor\nregion. A user study shows that the concepts identified by our model align with\nthe concepts used by pathologists, making it a promising strategy for\nhuman-interpretable WSI classification.\n", "date": "2025-01-06", "categories": []}
{"id": "2501.01932", "title": "Bridging Classification and Segmentation in Osteosarcoma Assessment via\n  Foundation and Discrete Diffusion Models", "abstract": "  Osteosarcoma, the most common primary bone cancer, often requires accurate\nnecrosis assessment from whole slide images (WSIs) for effective treatment\nplanning and prognosis. However, manual assessments are subjective and prone to\nvariability. In response, we introduce FDDM, a novel framework bridging the gap\nbetween patch classification and region-based segmentation. FDDM operates in\ntwo stages: patch-based classification, followed by region-based refinement,\nenabling cross-patch information intergation. Leveraging a newly curated\ndataset of osteosarcoma images, FDDM demonstrates superior segmentation\nperformance, achieving up to a 10% improvement mIOU and a 32.12% enhancement in\nnecrosis rate estimation over state-of-the-art methods. This framework sets a\nnew benchmark in osteosarcoma assessment, highlighting the potential of\nfoundation models and diffusion-based refinements in complex medical imaging\ntasks.\n", "date": "2025-01-03", "categories": []}
{"id": "2501.00053", "title": "Implementing Trust in Non-Small Cell Lung Cancer Diagnosis with a\n  Conformalized Uncertainty-Aware AI Framework in Whole-Slide Images", "abstract": "  Ensuring trustworthiness is fundamental to the development of artificial\nintelligence (AI) that is considered societally responsible, particularly in\ncancer diagnostics, where a misdiagnosis can have dire consequences. Current\ndigital pathology AI models lack systematic solutions to address\ntrustworthiness concerns arising from model limitations and data discrepancies\nbetween model deployment and development environments. To address this issue,\nwe developed TRUECAM, a framework designed to ensure both data and model\ntrustworthiness in non-small cell lung cancer subtyping with whole-slide\nimages. TRUECAM integrates 1) a spectral-normalized neural Gaussian process for\nidentifying out-of-scope inputs and 2) an ambiguity-guided elimination of tiles\nto filter out highly ambiguous regions, addressing data trustworthiness, as\nwell as 3) conformal prediction to ensure controlled error rates. We\nsystematically evaluated the framework across multiple large-scale cancer\ndatasets, leveraging both task-specific and foundation models, illustrate that\nan AI model wrapped with TRUECAM significantly outperforms models that lack\nsuch guidance, in terms of classification accuracy, robustness,\ninterpretability, and data efficiency, while also achieving improvements in\nfairness. These findings highlight TRUECAM as a versatile wrapper framework for\ndigital pathology AI models with diverse architectural designs, promoting their\nresponsible and effective applications in real-world settings.\n", "date": "2024-12-28", "categories": []}
{"id": "2412.20430", "title": "Unlocking adaptive digital pathology through dynamic feature learning", "abstract": "  Foundation models have revolutionized the paradigm of digital pathology, as\nthey leverage general-purpose features to emulate real-world pathological\npractices, enabling the quantitative analysis of critical histological patterns\nand the dissection of cancer-specific signals. However, these static general\nfeatures constrain the flexibility and pathological relevance in the\never-evolving needs of clinical applications, hindering the broad use of the\ncurrent models. Here we introduce PathFiT, a dynamic feature learning method\nthat can be effortlessly plugged into various pathology foundation models to\nunlock their adaptability. Meanwhile, PathFiT performs seamless implementation\nacross diverse pathology applications regardless of downstream specificity. To\nvalidate PathFiT, we construct a digital pathology benchmark with over 20\nterabytes of Internet and real-world data comprising 28 H\\&E-stained tasks and\n7 specialized imaging tasks including Masson's Trichrome staining and\nimmunofluorescence images. By applying PathFiT to the representative pathology\nfoundation models, we demonstrate state-of-the-art performance on 34 out of 35\ntasks, with significant improvements on 23 tasks and outperforming by 10.20% on\nspecialized imaging tasks. The superior performance and versatility of PathFiT\nopen up new avenues in computational pathology.\n", "date": "2024-12-29", "categories": []}
{"id": "2412.20571", "title": "Segmentation of Muscularis Propria in Colon Histopathology Images Using\n  Vision Transformers for Hirschsprung's Disease", "abstract": "  Hirschsprung's disease (HD) is a congenital birth defect diagnosed by\nidentifying the lack of ganglion cells within the colon's muscularis propria,\nspecifically within the myenteric plexus regions. There may be advantages for\nquantitative assessments of histopathology images of the colon, such as\ncounting the ganglion and assessing their spatial distribution; however, this\nwould be time-intensive for pathologists, costly, and subject to inter- and\nintra-rater variability. Previous research has demonstrated the potential for\ndeep learning approaches to automate histopathology image analysis, including\nsegmentation of the muscularis propria using convolutional neural networks\n(CNNs). Recently, Vision Transformers (ViTs) have emerged as a powerful deep\nlearning approach due to their self-attention. This study explores the\napplication of ViTs for muscularis propria segmentation in calretinin-stained\nhistopathology images and compares their performance to CNNs and shallow\nlearning methods. The ViT model achieved a DICE score of 89.9% and Plexus\nInclusion Rate (PIR) of 100%, surpassing the CNN (DICE score of 89.2%; PIR of\n96.0%) and k-means clustering method (DICE score of 80.7%; PIR 77.4%). Results\nassert that ViTs are a promising tool for advancing HD-related image analysis.\n", "date": "2024-12-29", "categories": []}
{"id": "2409.15092", "title": "M2OST: Many-to-one Regression for Predicting Spatial Transcriptomics\n  from Digital Pathology Images", "abstract": "  The advancement of Spatial Transcriptomics (ST) has facilitated the\nspatially-aware profiling of gene expressions based on histopathology images.\nAlthough ST data offers valuable insights into the micro-environment of tumors,\nits acquisition cost remains expensive. Therefore, directly predicting the ST\nexpressions from digital pathology images is desired. Current methods usually\nadopt existing regression backbones along with patch-sampling for this task,\nwhich ignores the inherent multi-scale information embedded in the pyramidal\ndata structure of digital pathology images, and wastes the inter-spot visual\ninformation crucial for accurate gene expression prediction. To address these\nlimitations, we propose M2OST, a many-to-one regression Transformer that can\naccommodate the hierarchical structure of the pathology images via a decoupled\nmulti-scale feature extractor. Unlike traditional models that are trained with\none-to-one image-label pairs, M2OST uses multiple images from different levels\nof the digital pathology image to jointly predict the gene expressions in their\ncommon corresponding spot. Built upon our many-to-one scheme, M2OST can be\neasily scaled to fit different numbers of inputs, and its network structure\ninherently incorporates nearby inter-spot features, enhancing regression\nperformance. We have tested M2OST on three public ST datasets and the\nexperimental results show that M2OST can achieve state-of-the-art performance\nwith fewer parameters and floating-point operations (FLOPs).\n", "date": "2024-09-23", "categories": []}
{"id": "2309.17166", "title": "Advances in Kidney Biopsy Lesion Assessment through Dense Instance\n  Segmentation", "abstract": "  Renal biopsies are the gold standard for the diagnosis of kidney diseases.\nLesion scores made by renal pathologists are semi-quantitative and exhibit high\ninter-observer variability. Automating lesion classification within segmented\nanatomical structures can provide decision support in quantification analysis,\nthereby reducing inter-observer variability. Nevertheless, classifying lesions\nin regions-of-interest (ROIs) is clinically challenging due to (a) a large\namount of densely packed anatomical objects, (b) class imbalance across\ndifferent compartments (at least 3), (c) significant variation in size and\nshape of anatomical objects and (d) the presence of multi-label lesions per\nanatomical structure. Existing models cannot address these complexities in an\nefficient and generic manner. This paper presents an analysis for a\n\\textbf{generalized solution} to datasets from various sources (pathology\ndepartments) with different types of lesions. Our approach utilizes two\nsub-networks: dense instance segmentation and lesion classification. We\nintroduce \\textbf{DiffRegFormer}, an end-to-end dense instance segmentation\nsub-network designed for multi-class, multi-scale objects within ROIs.\nCombining diffusion models, transformers, and RCNNs, DiffRegFormer {is a\ncomputational-friendly framework that can efficiently recognize over 500\nobjects across three anatomical classes, i.e., glomeruli, tubuli, and arteries,\nwithin ROIs.} In a dataset of 303 ROIs from 148 Jones' silver-stained renal\nWhole Slide Images (WSIs), our approach outperforms previous methods, achieving\nan Average Precision of 52.1\\% (detection) and 46.8\\% (segmentation). Moreover,\nour lesion classification sub-network achieves 89.2\\% precision and 64.6\\%\nrecall on 21889 object patches out of the 303 ROIs. Lastly, our model\ndemonstrates direct domain transfer to PAS-stained renal WSIs without\nfine-tuning.\n", "date": "2023-09-29", "categories": []}
{"id": "2412.13857", "title": "Diagnosising Helicobacter pylori using AutoEncoders and Limited\n  Annotations through Anomalous Staining Patterns in IHC Whole Slide Images", "abstract": "  Purpose: This work addresses the detection of Helicobacter pylori (H. pylori)\nin histological images with immunohistochemical staining. This analysis is a\ntime demanding task, currently done by an expert pathologist that visually\ninspects the samples. Given the effort required to localise the pathogen in\nimages, a limited number of annotations might be available in an initial\nsetting. Our goal is to design an approach that, using a limited set of\nannotations, is capable of obtaining results good enough to be used as a\nsupport tool. Methods: We propose to use autoencoders to learn the latent\npatterns of healthy patches and formulate a specific measure of the\nreconstruction error of the image in HSV space. ROC analysis is used to set the\noptimal threshold of this measure and the percentage of positive patches in a\nsample that determines the presence of H. pylori. Results: Our method has been\ntested on an own database of 245 Whole Slide Images (WSI) having 117 cases\nwithout H. pylori and different density of the bacteria in the remaining ones.\nThe database has 1211 annotated patches, with only 163 positive patches. This\ndataset of positive annotations was used to train a baseline thresholding and\nan SVM using the features of a pre-trained RedNet18 and ViT models. A 10-fold\ncross-validation shows that our method has better performance with 91%\naccuracy, 86% sensitivity, 96% specificity and 0.97 AUC in the diagnosis of H.\npylori. Conclusion: Unlike classification approaches, our shallow autoencoder\nwith threshold adaptation for the detection of anomalous staining is able to\nachieve competitive results with a limited set of annotated data. This initial\napproach is good enough to be used as a guide for fast annotation of infected\npatches.\n", "date": "2024-12-18", "categories": []}
{"id": "2407.07841", "title": "Benchmarking Embedding Aggregation Methods in Computational Pathology: A\n  Clinical Data Perspective", "abstract": "  Recent advances in artificial intelligence (AI), in particular\nself-supervised learning of foundation models (FMs), are revolutionizing\nmedical imaging and computational pathology (CPath). A constant challenge in\nthe analysis of digital Whole Slide Images (WSIs) is the problem of aggregating\ntens of thousands of tile-level image embeddings to a slide-level\nrepresentation. Due to the prevalent use of datasets created for genomic\nresearch, such as TCGA, for method development, the performance of these\ntechniques on diagnostic slides from clinical practice has been inadequately\nexplored. This study conducts a thorough benchmarking analysis of ten\nslide-level aggregation techniques across nine clinically relevant tasks,\nincluding diagnostic assessment, biomarker classification, and outcome\nprediction. The results yield following key insights: (1) Embeddings derived\nfrom domain-specific (histological images) FMs outperform those from generic\nImageNet-based models across aggregation methods. (2) Spatial-aware aggregators\nenhance the performance significantly when using ImageNet pre-trained models\nbut not when using FMs. (3) No single model excels in all tasks and\nspatially-aware models do not show general superiority as it would be expected.\nThese findings underscore the need for more adaptable and universally\napplicable aggregation techniques, guiding future research towards tools that\nbetter meet the evolving needs of clinical-AI in pathology. The code used in\nthis work is available at\n\\url{https://github.com/fuchs-lab-public/CPath_SABenchmark}.\n", "date": "2024-07-10", "categories": []}
{"id": "2407.10785", "title": "Learning biologically relevant features in a pathology foundation model\n  using sparse autoencoders", "abstract": "  Pathology plays an important role in disease diagnosis, treatment\ndecision-making and drug development. Previous works on interpretability for\nmachine learning models on pathology images have revolved around methods such\nas attention value visualization and deriving human-interpretable features from\nmodel heatmaps. Mechanistic interpretability is an emerging area of model\ninterpretability that focuses on reverse-engineering neural networks. Sparse\nAutoencoders (SAEs) have emerged as a promising direction in terms of\nextracting monosemantic features from polysemantic model activations. In this\nwork, we trained a Sparse Autoencoder on the embeddings of a pathology\npretrained foundation model. We found that Sparse Autoencoder features\nrepresent interpretable and monosemantic biological concepts. In particular,\nindividual SAE dimensions showed strong correlations with cell type counts such\nas plasma cells and lymphocytes. These biological representations were unique\nto the pathology pretrained model and were not found in a self-supervised model\npretrained on natural images. We demonstrated that such biologically-grounded\nmonosemantic representations evolved across the model's depth, and the\npathology foundation model eventually gained robustness to non-biological\nfactors such as scanner type. The emergence of biologically relevant SAE\nfeatures was generalizable to an out-of-domain dataset. Our work paves the way\nfor further exploration around interpretable feature dimensions and their\nutility for medical and clinical applications.\n", "date": "2024-07-15", "categories": []}
{"id": "2412.10392", "title": "Computational Methods for Breast Cancer Molecular Profiling through\n  Routine Histopathology: A Review", "abstract": "  Precision medicine has become a central focus in breast cancer management,\nadvancing beyond conventional methods to deliver more precise and\nindividualized therapies. Traditionally, histopathology images have been used\nprimarily for diagnostic purposes; however, they are now recognized for their\npotential in molecular profiling, which provides deeper insights into cancer\nprognosis and treatment response. Recent advancements in artificial\nintelligence (AI) have enabled digital pathology to analyze histopathologic\nimages for both targeted molecular and broader omic biomarkers, marking a\npivotal step in personalized cancer care. These technologies offer the\ncapability to extract various biomarkers such as genomic, transcriptomic,\nproteomic, and metabolomic markers directly from the routine hematoxylin and\neosin (H&E) stained images, which can support treatment decisions without the\nneed for costly molecular assays. In this work, we provide a comprehensive\nreview of AI-driven techniques for biomarker detection, with a focus on diverse\nomic biomarkers that allow novel biomarker discovery. Additionally, we analyze\nthe major challenges faced in this field for robust algorithm development.\nThese challenges highlight areas where further research is essential to bridge\nthe gap between AI research and clinical application.\n", "date": "2024-12-01", "categories": []}
{"id": "2412.11106", "title": "Unpaired Multi-Domain Histopathology Virtual Staining using Dual Path\n  Prompted Inversion", "abstract": "  Virtual staining leverages computer-aided techniques to transfer the style of\nhistochemically stained tissue samples to other staining types. In virtual\nstaining of pathological images, maintaining strict structural consistency is\ncrucial, as these images emphasize structural integrity more than natural\nimages. Even slight structural alterations can lead to deviations in diagnostic\nsemantic information. Furthermore, the unpaired characteristic of virtual\nstaining data may compromise the preservation of pathological diagnostic\ncontent. To address these challenges, we propose a dual-path inversion virtual\nstaining method using prompt learning, which optimizes visual prompts to\ncontrol content and style, while preserving complete pathological diagnostic\ncontent. Our proposed inversion technique comprises two key components: (1)\nDual Path Prompted Strategy, we utilize a feature adapter function to generate\nreference images for inversion, providing style templates for input image\ninversion, called Style Target Path. We utilize the inversion of the input\nimage as the Structural Target path, employing visual prompt images to maintain\nstructural consistency in this path while preserving style information from the\nstyle Target path. During the deterministic sampling process, we achieve\ncomplete content-style disentanglement through a plug-and-play embedding visual\nprompt approach. (2) StainPrompt Optimization, where we only optimize the null\nvisual prompt as ``operator'' for dual path inversion, rather than fine-tune\npre-trained model. We optimize null visual prompt for structual and style\ntrajectory around pivotal noise on each timestep, ensuring accurate dual-path\ninversion reconstruction. Extensive evaluations on publicly available\nmulti-domain unpaired staining datasets demonstrate high structural consistency\nand accurate style transfer results.\n", "date": "2024-12-15", "categories": []}
{"id": "2412.11938", "title": "Are the Latent Representations of Foundation Models for Pathology\n  Invariant to Rotation?", "abstract": "  Self-supervised foundation models for digital pathology encode small patches\nfrom H\\&E whole slide images into latent representations used for downstream\ntasks. However, the invariance of these representations to patch rotation\nremains unexplored. This study investigates the rotational invariance of latent\nrepresentations across twelve foundation models by quantifying the alignment\nbetween non-rotated and rotated patches using mutual $k$-nearest neighbours and\ncosine distance. Models that incorporated rotation augmentation during\nself-supervised training exhibited significantly greater invariance to\nrotations. We hypothesise that the absence of rotational inductive bias in the\ntransformer architecture necessitates rotation augmentation during training to\nachieve learned invariance. Code:\nhttps://github.com/MatousE/rot-invariance-analysis.\n", "date": "2024-12-16", "categories": []}
{"id": "2412.12077", "title": "CPath-Omni: A Unified Multimodal Foundation Model for Patch and Whole\n  Slide Image Analysis in Computational Pathology", "abstract": "  The emergence of large multimodal models (LMMs) has brought significant\nadvancements to pathology. Previous research has primarily focused on\nseparately training patch-level and whole-slide image (WSI)-level models,\nlimiting the integration of learned knowledge across patches and WSIs, and\nresulting in redundant models. In this work, we introduce CPath-Omni, the first\n15-billion-parameter LMM designed to unify both patch and WSI level image\nanalysis, consolidating a variety of tasks at both levels, including\nclassification, visual question answering, captioning, and visual referring\nprompting. Extensive experiments demonstrate that CPath-Omni achieves\nstate-of-the-art (SOTA) performance across seven diverse tasks on 39 out of 42\ndatasets, outperforming or matching task-specific models trained for individual\ntasks. Additionally, we develop a specialized pathology CLIP-based visual\nprocessor for CPath-Omni, CPath-CLIP, which, for the first time, integrates\ndifferent vision models and incorporates a large language model as a text\nencoder to build a more powerful CLIP model, which achieves SOTA performance on\nnine zero-shot and four few-shot datasets. Our findings highlight CPath-Omni's\nability to unify diverse pathology tasks, demonstrating its potential to\nstreamline and advance the field of foundation model in pathology.\n", "date": "2024-12-16", "categories": []}
{"id": "2308.16561", "title": "MoMA: Momentum Contrastive Learning with Multi-head Attention-based\n  Knowledge Distillation for Histopathology Image Analysis", "abstract": "  There is no doubt that advanced artificial intelligence models and high\nquality data are the keys to success in developing computational pathology\ntools. Although the overall volume of pathology data keeps increasing, a lack\nof quality data is a common issue when it comes to a specific task due to\nseveral reasons including privacy and ethical issues with patient data. In this\nwork, we propose to exploit knowledge distillation, i.e., utilize the existing\nmodel to learn a new, target model, to overcome such issues in computational\npathology. Specifically, we employ a student-teacher framework to learn a\ntarget model from a pre-trained, teacher model without direct access to source\ndata and distill relevant knowledge via momentum contrastive learning with\nmulti-head attention mechanism, which provides consistent and context-aware\nfeature representations. This enables the target model to assimilate\ninformative representations of the teacher model while seamlessly adapting to\nthe unique nuances of the target data. The proposed method is rigorously\nevaluated across different scenarios where the teacher model was trained on the\nsame, relevant, and irrelevant classification tasks with the target model.\nExperimental results demonstrate the accuracy and robustness of our approach in\ntransferring knowledge to different domains and tasks, outperforming other\nrelated methods. Moreover, the results provide a guideline on the learning\nstrategy for different types of tasks and scenarios in computational pathology.\nCode is available at: \\url{https://github.com/trinhvg/MoMA}.\n", "date": "2023-08-31", "categories": []}
{"id": "2408.15823", "title": "Benchmarking foundation models as feature extractors for\n  weakly-supervised computational pathology", "abstract": "  Advancements in artificial intelligence have driven the development of\nnumerous pathology foundation models capable of extracting clinically relevant\ninformation. However, there is currently limited literature independently\nevaluating these foundation models on truly external cohorts and\nclinically-relevant tasks to uncover adjustments for future improvements. In\nthis study, we benchmarked 19 histopathology foundation models on 13 patient\ncohorts with 6,818 patients and 9,528 slides from lung, colorectal, gastric,\nand breast cancers. The models were evaluated on weakly-supervised tasks\nrelated to biomarkers, morphological properties, and prognostic outcomes. We\nshow that a vision-language foundation model, CONCH, yielded the highest\nperformance when compared to vision-only foundation models, with Virchow2 as\nclose second. The experiments reveal that foundation models trained on distinct\ncohorts learn complementary features to predict the same label, and can be\nfused to outperform the current state of the art. An ensemble combining CONCH\nand Virchow2 predictions outperformed individual models in 55% of tasks,\nleveraging their complementary strengths in classification scenarios. Moreover,\nour findings suggest that data diversity outweighs data volume for foundation\nmodels. Our work highlights actionable adjustments to improve pathology\nfoundation models.\n", "date": "2024-08-28", "categories": []}
{"id": "2412.06129", "title": "GCUNet: A GNN-Based Contextual Learning Network for Tertiary Lymphoid\n  Structure Semantic Segmentation in Whole Slide Image", "abstract": "  We focus on tertiary lymphoid structure (TLS) semantic segmentation in whole\nslide image (WSI). Unlike TLS binary segmentation, TLS semantic segmentation\nidentifies boundaries and maturity, which requires integrating contextual\ninformation to discover discriminative features. Due to the extensive scale of\nWSI (e.g., 100,000 \\times 100,000 pixels), the segmentation of TLS is usually\ncarried out through a patch-based strategy. However, this prevents the model\nfrom accessing information outside of the patches, limiting the performance. To\naddress this issue, we propose GCUNet, a GNN-based contextual learning network\nfor TLS semantic segmentation. Given an image patch (target) to be segmented,\nGCUNet first progressively aggregates long-range and fine-grained context\noutside the target. Then, a Detail and Context Fusion block (DCFusion) is\ndesigned to integrate the context and detail of the target to predict the\nsegmentation mask. We build four TLS semantic segmentation datasets, called\nTCGA-COAD, TCGA-LUSC, TCGA-BLCA and INHOUSE-PAAD, and make the former three\ndatasets (comprising 826 WSIs and 15,276 TLSs) publicly available to promote\nthe TLS semantic segmentation. Experiments on these datasets demonstrate the\nsuperiority of GCUNet, achieving at least 7.41% improvement in mF1 compared\nwith SOTA.\n", "date": "2024-12-09", "categories": []}
{"id": "2412.04260", "title": "Enhancing Whole Slide Image Classification through Supervised\n  Contrastive Domain Adaptation", "abstract": "  Domain shift in the field of histopathological imaging is a common phenomenon\ndue to the intra- and inter-hospital variability of staining and digitization\nprotocols. The implementation of robust models, capable of creating generalized\ndomains, represents a need to be solved. In this work, a new domain adaptation\nmethod to deal with the variability between histopathological images from\nmultiple centers is presented. In particular, our method adds a training\nconstraint to the supervised contrastive learning approach to achieve domain\nadaptation and improve inter-class separability. Experiments performed on\ndomain adaptation and classification of whole-slide images of six skin cancer\nsubtypes from two centers demonstrate the method's usefulness. The results\nreflect superior performance compared to not using domain adaptation after\nfeature extraction or staining normalization.\n", "date": "2024-12-05", "categories": []}
{"id": "2404.08027", "title": "SurvMamba: State Space Model with Multi-grained Multi-modal Interaction\n  for Survival Prediction", "abstract": "  Multi-modal learning that combines pathological images with genomic data has\nsignificantly enhanced the accuracy of survival prediction. Nevertheless,\nexisting methods have not fully utilized the inherent hierarchical structure\nwithin both whole slide images (WSIs) and transcriptomic data, from which\nbetter intra-modal representations and inter-modal integration could be\nderived. Moreover, many existing studies attempt to improve multi-modal\nrepresentations through attention mechanisms, which inevitably lead to high\ncomplexity when processing high-dimensional WSIs and transcriptomic data.\nRecently, a structured state space model named Mamba emerged as a promising\napproach for its superior performance in modeling long sequences with low\ncomplexity. In this study, we propose Mamba with multi-grained multi-modal\ninteraction (SurvMamba) for survival prediction. SurvMamba is implemented with\na Hierarchical Interaction Mamba (HIM) module that facilitates efficient\nintra-modal interactions at different granularities, thereby capturing more\ndetailed local features as well as rich global representations. In addition, an\nInteraction Fusion Mamba (IFM) module is used for cascaded inter-modal\ninteractive fusion, yielding more comprehensive features for survival\nprediction. Comprehensive evaluations on five TCGA datasets demonstrate that\nSurvMamba outperforms other existing methods in terms of performance and\ncomputational cost.\n", "date": "2024-04-11", "categories": []}
{"id": "2412.00651", "title": "Towards Unified Molecule-Enhanced Pathology Image Representation\n  Learning via Integrating Spatial Transcriptomics", "abstract": "  Recent advancements in multimodal pre-training models have significantly\nadvanced computational pathology. However, current approaches predominantly\nrely on visual-language models, which may impose limitations from a molecular\nperspective and lead to performance bottlenecks. Here, we introduce a Unified\nMolecule-enhanced Pathology Image REpresentationn Learning framework (UMPIRE).\nUMPIRE aims to leverage complementary information from gene expression profiles\nto guide the multimodal pre-training, enhancing the molecular awareness of\npathology image representation learning. We demonstrate that this molecular\nperspective provides a robust, task-agnostic training signal for learning\npathology image embeddings. Due to the scarcity of paired data, approximately 4\nmillion entries of spatial transcriptomics gene expression were collected to\ntrain the gene encoder. By leveraging powerful pre-trained encoders, UMPIRE\naligns the encoders across over 697K pathology image-gene expression pairs. The\nperformance of UMPIRE is demonstrated across various molecular-related\ndownstream tasks, including gene expression prediction, spot classification,\nand mutation state prediction in whole slide images. Our findings highlight the\neffectiveness of multimodal data integration and open new avenues for exploring\ncomputational pathology enhanced by molecular perspectives. The code and\npre-trained weights are available at https://github.com/Hanminghao/UMPIRE.\n", "date": "2024-12-01", "categories": []}
{"id": "2412.01182", "title": "MeasureNet: Measurement Based Celiac Disease Identification", "abstract": "  Celiac disease is an autoimmune disorder triggered by the consumption of\ngluten. It causes damage to the villi, the finger-like projections in the small\nintestine that are responsible for nutrient absorption. Additionally, the\ncrypts, which form the base of the villi, are also affected, impairing the\nregenerative process. The deterioration in villi length, computed as the\nvilli-to-crypt length ratio, indicates the severity of celiac disease. However,\nmanual measurement of villi-crypt length can be both time-consuming and\nsusceptible to inter-observer variability, leading to inconsistencies in\ndiagnosis. While some methods can perform measurement as a post-hoc process,\nthey are prone to errors in the initial stages. This gap underscores the need\nfor pathologically driven solutions that enhance measurement accuracy and\nreduce human error in celiac disease assessments.\n  Our proposed method, MeasureNet, is a pathologically driven polyline\ndetection framework incorporating polyline localization and object-driven\nlosses specifically designed for measurement tasks. Furthermore, we leverage\nsegmentation model to provide auxiliary guidance about crypt location when\ncrypt are partially visible. To ensure that model is not overdependent on\nsegmentation mask we enhance model robustness through a mask feature mixup\ntechnique. Additionally, we introduce a novel dataset for grading celiac\ndisease, consisting of 750 annotated duodenum biopsy images. MeasureNet\nachieves an 82.66% classification accuracy for binary classification and 81%\naccuracy for multi-class grading of celiac disease. Code:\nhttps://github.com/dair-iitd/MeasureNet\n", "date": "2024-12-02", "categories": []}
{"id": "2411.19666", "title": "Multimodal Whole Slide Foundation Model for Pathology", "abstract": "  The field of computational pathology has been transformed with recent\nadvances in foundation models that encode histopathology region-of-interests\n(ROIs) into versatile and transferable feature representations via\nself-supervised learning (SSL). However, translating these advancements to\naddress complex clinical challenges at the patient and slide level remains\nconstrained by limited clinical data in disease-specific cohorts, especially\nfor rare clinical conditions. We propose TITAN, a multimodal whole slide\nfoundation model pretrained using 335,645 WSIs via visual self-supervised\nlearning and vision-language alignment with corresponding pathology reports and\n423,122 synthetic captions generated from a multimodal generative AI copilot\nfor pathology. Without any finetuning or requiring clinical labels, TITAN can\nextract general-purpose slide representations and generate pathology reports\nthat generalize to resource-limited clinical scenarios such as rare disease\nretrieval and cancer prognosis. We evaluate TITAN on diverse clinical tasks and\nfind that TITAN outperforms both ROI and slide foundation models across machine\nlearning settings such as linear probing, few-shot and zero-shot\nclassification, rare cancer retrieval and cross-modal retrieval, and pathology\nreport generation.\n", "date": "2024-11-29", "categories": []}
{"id": "2411.18225", "title": "PATHS: A Hierarchical Transformer for Efficient Whole Slide Image\n  Analysis", "abstract": "  Computational analysis of whole slide images (WSIs) has seen significant\nresearch progress in recent years, with applications ranging across important\ndiagnostic and prognostic tasks such as survival or cancer subtype prediction.\nMany state-of-the-art models process the entire slide - which may be as large\nas $150,000 \\times 150,000$ pixels - as a bag of many patches, the size of\nwhich necessitates computationally cheap feature aggregation methods. However,\na large proportion of these patches are uninformative, such as those containing\nonly healthy or adipose tissue, adding significant noise and size to the bag.\nWe propose Pathology Transformer with Hierarchical Selection (PATHS), a novel\ntop-down method for hierarchical weakly supervised representation learning on\nslide-level tasks in computational pathology. PATHS is inspired by the\ncross-magnification manner in which a human pathologist examines a slide,\nrecursively filtering patches at each magnification level to a small subset\nrelevant to the diagnosis. Our method overcomes the complications of processing\nthe entire slide, enabling quadratic self-attention and providing a simple\ninterpretable measure of region importance. We apply PATHS to five datasets of\nThe Cancer Genome Atlas (TCGA), and achieve superior performance on slide-level\nprediction tasks when compared to previous methods, despite processing only a\nsmall proportion of the slide.\n", "date": "2024-11-27", "categories": []}
{"id": "2411.18391", "title": "GeneQuery: A General QA-based Framework for Spatial Gene Expression\n  Predictions from Histology Images", "abstract": "  Gene expression profiling provides profound insights into molecular\nmechanisms, but its time-consuming and costly nature often presents significant\nchallenges. In contrast, whole-slide hematoxylin and eosin (H&E) stained\nhistological images are readily accessible and allow for detailed examinations\nof tissue structure and composition at the microscopic level. Recent\nadvancements have utilized these histological images to predict spatially\nresolved gene expression profiles. However, state-of-the-art works treat gene\nexpression prediction as a multi-output regression problem, where each gene is\nlearned independently with its own weights, failing to capture the shared\ndependencies and co-expression patterns between genes. Besides, existing works\ncan only predict gene expression values for genes seen during training,\nlimiting their ability to generalize to new, unseen genes.\n  To address the above limitations, this paper presents GeneQuery, which aims\nto solve this gene expression prediction task in a question-answering (QA)\nmanner for better generality and flexibility. Specifically, GeneQuery takes\ngene-related texts as queries and whole-slide images as contexts and then\npredicts the queried gene expression values. With such a transformation,\nGeneQuery can implicitly estimate the gene distribution by introducing the gene\nrandom variable. Besides, the proposed GeneQuery consists of two architecture\nimplementations, i.e., spot-aware GeneQuery for capturing patterns between\nimages and gene-aware GeneQuery for capturing patterns between genes.\nComprehensive experiments on spatial transcriptomics datasets show that the\nproposed GeneQuery outperforms existing state-of-the-art methods on known and\nunseen genes. More results also demonstrate that GeneQuery can potentially\nanalyze the tissue structure.\n", "date": "2024-11-27", "categories": []}
{"id": "2411.16956", "title": "Contrastive Deep Learning Reveals Age Biomarkers in Histopathological\n  Skin Biopsies", "abstract": "  As global life expectancy increases, so does the burden of chronic diseases,\nyet individuals exhibit considerable variability in the rate at which they age.\nIdentifying biomarkers that distinguish fast from slow ageing is crucial for\nunderstanding the biology of ageing, enabling early disease detection, and\nimproving prevention strategies. Using contrastive deep learning, we show that\nskin biopsy images alone are sufficient to determine an individual's age. We\nthen use visual features in histopathology slides of the skin biopsies to\nconstruct a novel biomarker of ageing. By linking with comprehensive health\nregisters in Denmark, we demonstrate that visual features in histopathology\nslides of skin biopsies predict mortality and the prevalence of chronic\nage-related diseases. Our work highlights how routinely collected health data\ncan provide additional value when used together with deep learning, by creating\na new biomarker for ageing which can be actively used to determine mortality\nover time.\n", "date": "2024-11-25", "categories": []}
{"id": "2411.16961", "title": "Glo-In-One-v2: Holistic Identification of Glomerular Cells, Tissues, and\n  Lesions in Human and Mouse Histopathology", "abstract": "  Segmenting glomerular intraglomerular tissue and lesions traditionally\ndepends on detailed morphological evaluations by expert nephropathologists, a\nlabor-intensive process susceptible to interobserver variability. Our group\npreviously developed the Glo-In-One toolkit for integrated detection and\nsegmentation of glomeruli. In this study, we leverage the Glo-In-One toolkit to\nversion 2 with fine-grained segmentation capabilities, curating 14 distinct\nlabels for tissue regions, cells, and lesions across a dataset of 23,529\nannotated glomeruli across human and mouse histopathology data. To our\nknowledge, this dataset is among the largest of its kind to date.In this study,\nwe present a single dynamic head deep learning architecture designed to segment\n14 classes within partially labeled images of human and mouse pathology data.\nOur model was trained using a training set derived from 368 annotated kidney\nwhole-slide images (WSIs) to identify 5 key intraglomerular tissues covering\nBowman's capsule, glomerular tuft, mesangium, mesangial cells, and podocytes.\nAdditionally, the network segments 9 glomerular lesion classes including\nadhesion, capsular drop, global sclerosis, hyalinosis, mesangial lysis,\nmicroaneurysm, nodular sclerosis, mesangial expansion, and segmental sclerosis.\nThe glomerulus segmentation model achieved a decent performance compared with\nbaselines, and achieved a 76.5 % average Dice Similarity Coefficient (DSC).\nAdditional, transfer learning from rodent to human for glomerular lesion\nsegmentation model has enhanced the average segmentation accuracy across\ndifferent types of lesions by more than 3 %, as measured by Dice scores. The\nGlo-In-One-v2 model and trained weight have been made publicly available at\nhttps: //github.com/hrlblab/Glo-In-One_v2.\n", "date": "2024-11-25", "categories": []}
{"id": "2411.11487", "title": "Look a Group at Once: Multi-Slide Modeling for Survival Prediction", "abstract": "  Survival prediction is a critical task in pathology. In clinical practice,\npathologists often examine multiple cases, leveraging a broader spectrum of\ncancer phenotypes to enhance pathological assessment. Despite significant\nadvancements in deep learning, current solutions typically model each slide as\na sample, struggling to effectively capture comparable and slide-agnostic\npathological features. In this paper, we introduce GroupMIL, a novel framework\ninspired by the clinical practice of collective analysis, which models multiple\nslides as a single sample and organizes groups of patches and slides\nsequentially to capture cross-slide prognostic features. We also present\nGPAMamba, a model designed to facilitate intra- and inter-slide feature\ninteractions, effectively capturing local micro-environmental characteristics\nwithin slide-level graphs while uncovering essential prognostic patterns across\nan extended patch sequence within the group framework. Furthermore, we develop\na dual-head predictor that delivers comprehensive survival risk and probability\nassessments for each patient. Extensive empirical evaluations demonstrate that\nour model significantly outperforms state-of-the-art approaches across five\ndatasets from The Cancer Genome Atlas.\n", "date": "2024-11-18", "categories": []}
{"id": "2411.15237", "title": "Stain-Invariant Representation for Tissue Classification in Histology\n  Images", "abstract": "  The process of digitising histology slides involves multiple factors that can\naffect a whole slide image's (WSI) final appearance, including the staining\nprotocol, scanner, and tissue type. This variability constitutes a domain shift\nand results in significant problems when training and testing deep learning\n(DL) algorithms in multi-cohort settings. As such, developing robust and\ngeneralisable DL models in computational pathology (CPath) remains an open\nchallenge. In this regard, we propose a framework that generates\nstain-augmented versions of the training images using stain matrix\nperturbation. Thereafter, we employed a stain regularisation loss to enforce\nconsistency between the feature representations of the source and augmented\nimages. Doing so encourages the model to learn stain-invariant and,\nconsequently, domain-invariant feature representations. We evaluate the\nperformance of the proposed model on cross-domain multi-class tissue type\nclassification of colorectal cancer images and have achieved improved\nperformance compared to other state-of-the-art methods.\n", "date": "2024-11-21", "categories": []}
{"id": "2411.15274", "title": "Feature-interactive Siamese graph encoder-based image analysis to\n  predict STAS from histopathology images in lung cancer", "abstract": "  Spread through air spaces (STAS) is a distinct invasion pattern in lung\ncancer, crucial for prognosis assessment and guiding surgical decisions.\nHistopathology is the gold standard for STAS detection, yet traditional methods\nare subjective, time-consuming, and prone to misdiagnosis, limiting large-scale\napplications. We present VERN, an image analysis model utilizing a\nfeature-interactive Siamese graph encoder to predict STAS from lung cancer\nhistopathological images. VERN captures spatial topological features with\nfeature sharing and skip connections to enhance model training. Using 1,546\nhistopathology slides, we built a large single-cohort STAS lung cancer dataset.\nVERN achieved an AUC of 0.9215 in internal validation and AUCs of 0.8275 and\n0.8829 in frozen and paraffin-embedded test sections, respectively,\ndemonstrating clinical-grade performance. Validated on a single-cohort and\nthree external datasets, VERN showed robust predictive performance and\ngeneralizability, providing an open platform (http://plr.20210706.xyz:5000/) to\nenhance STAS diagnosis efficiency and accuracy.\n", "date": "2024-11-22", "categories": []}
{"id": "2411.15719", "title": "Comparative Analysis of Diffusion Generative Models in Computational\n  Pathology", "abstract": "  Diffusion Generative Models (DGM) have rapidly surfaced as emerging topics in\nthe field of computer vision, garnering significant interest across a wide\narray of deep learning applications. Despite their high computational demand,\nthese models are extensively utilized for their superior sample quality and\nrobust mode coverage. While research in diffusion generative models is\nadvancing, exploration within the domain of computational pathology and its\nlarge-scale datasets has been comparatively gradual. Bridging the gap between\nthe high-quality generation capabilities of Diffusion Generative Models and the\nintricate nature of pathology data, this paper presents an in-depth comparative\nanalysis of diffusion methods applied to a pathology dataset. Our analysis\nextends to datasets with varying Fields of View (FOV), revealing that DGMs are\nhighly effective in producing high-quality synthetic data. An ablative study is\nalso conducted, followed by a detailed discussion on the impact of various\nmethods on the synthesized histopathology images. One striking observation from\nour experiments is how the adjustment of image size during data generation can\nsimulate varying fields of view. These findings underscore the potential of\nDGMs to enhance the quality and diversity of synthetic pathology data,\nespecially when used with real data, ultimately increasing accuracy of deep\nlearning models in histopathology. Code is available from\nhttps://github.com/AtlasAnalyticsLab/Diffusion4Path\n", "date": "2024-11-24", "categories": []}
{"id": "2411.15942", "title": "Cross-organ Deployment of EOS Detection AI without Retraining:\n  Feasibility and Limitation", "abstract": "  Chronic rhinosinusitis (CRS) is characterized by persistent inflammation in\nthe paranasal sinuses, leading to typical symptoms of nasal congestion, facial\npressure, olfactory dysfunction, and discolored nasal drainage, which can\nsignificantly impact quality-of-life. Eosinophils (Eos), a crucial component in\nthe mucosal immune response, have been linked to disease severity in CRS. The\ndiagnosis of eosinophilic CRS typically uses a threshold of 10-20 eos per\nhigh-power field (HPF). However, manually counting Eos in histological samples\nis laborious and time-intensive, making the use of AI-driven methods for\nautomated evaluations highly desirable. Interestingly, eosinophils are\npredominantly located in the gastrointestinal (GI) tract, which has prompted\nthe release of numerous deep learning models trained on GI data. This study\nleverages a CircleSnake model initially trained on upper-GI data to segment Eos\ncells in whole slide images (WSIs) of nasal tissues. It aims to determine the\nextent to which Eos segmentation models developed for the GI tract can be\nadapted to nasal applications without retraining. The experimental results show\npromising accuracy in some WSIs, although, unsurprisingly, the performance\nvaries across cases. This paper details these performance outcomes, delves into\nthe reasons for such variations, and aims to provide insights that could guide\nfuture development of deep learning models for eosinophilic CRS.\n", "date": "2024-11-24", "categories": []}
{"id": "2401.03271", "title": "Analysis and Validation of Image Search Engines in Histopathology", "abstract": "  Searching for similar images in archives of histology and histopathology\nimages is a crucial task that may aid in patient matching for various purposes,\nranging from triaging and diagnosis to prognosis and prediction. Whole slide\nimages (WSIs) are highly detailed digital representations of tissue specimens\nmounted on glass slides. Matching WSI to WSI can serve as the critical method\nfor patient matching. In this paper, we report extensive analysis and\nvalidation of four search methods bag of visual words (BoVW), Yottixel, SISH,\nRetCCL, and some of their potential variants. We analyze their algorithms and\nstructures and assess their performance. For this evaluation, we utilized four\ninternal datasets ($1269$ patients) and three public datasets ($1207$\npatients), totaling more than $200,000$ patches from $38$ different\nclasses/subtypes across five primary sites. Certain search engines, for\nexample, BoVW, exhibit notable efficiency and speed but suffer from low\naccuracy. Conversely, search engines like Yottixel demonstrate efficiency and\nspeed, providing moderately accurate results. Recent proposals, including SISH,\ndisplay inefficiency and yield inconsistent outcomes, while alternatives like\nRetCCL prove inadequate in both accuracy and efficiency. Further research is\nimperative to address the dual aspects of accuracy and minimal storage\nrequirements in histopathological image search.\n", "date": "2024-01-06", "categories": []}
{"id": "2411.14975", "title": "Exploring Foundation Models Fine-Tuning for Cytology Classification", "abstract": "  Cytology slides are essential tools in diagnosing and staging cancer, but\ntheir analysis is time-consuming and costly. Foundation models have shown great\npotential to assist in these tasks. In this paper, we explore how existing\nfoundation models can be applied to cytological classification. More\nparticularly, we focus on low-rank adaptation, a parameter-efficient\nfine-tuning method suited to few-shot learning. We evaluated five foundation\nmodels across four cytological classification datasets. Our results demonstrate\nthat fine-tuning the pre-trained backbones with LoRA significantly improves\nmodel performance compared to fine-tuning only the classifier head, achieving\nstate-of-the-art results on both simple and complex classification tasks while\nrequiring fewer data samples.\n", "date": "2024-11-22", "categories": []}
{"id": "2302.09682", "title": "Dual Attention Model with Reinforcement Learning for Classification of\n  Histology Whole-Slide Images", "abstract": "  Digital whole slide images (WSIs) are generally captured at microscopic\nresolution and encompass extensive spatial data. Directly feeding these images\nto deep learning models is computationally intractable due to memory\nconstraints, while downsampling the WSIs risks incurring information loss.\nAlternatively, splitting the WSIs into smaller patches may result in a loss of\nimportant contextual information. In this paper, we propose a novel dual\nattention approach, consisting of two main components, both inspired by the\nvisual examination process of a pathologist: The first soft attention model\nprocesses a low magnification view of the WSI to identify relevant regions of\ninterest, followed by a custom sampling method to extract diverse and spatially\ndistinct image tiles from the selected ROIs. The second component, the hard\nattention classification model further extracts a sequence of multi-resolution\nglimpses from each tile for classification. Since hard attention is\nnon-differentiable, we train this component using reinforcement learning to\npredict the location of the glimpses. This approach allows the model to focus\non essential regions instead of processing the entire tile, thereby aligning\nwith a pathologist's way of diagnosis. The two components are trained in an\nend-to-end fashion using a joint loss function to demonstrate the efficacy of\nthe model. The proposed model was evaluated on two WSI-level classification\nproblems: Human epidermal growth factor receptor 2 scoring on breast cancer\nhistology images and prediction of Intact/Loss status of two Mismatch Repair\nbiomarkers from colorectal cancer histology images. We show that the proposed\nmodel achieves performance better than or comparable to the state-of-the-art\nmethods while processing less than 10% of the WSI at the highest magnification\nand reducing the time required to infer the WSI-level label by more than 75%.\n", "date": "2023-02-19", "categories": []}
{"id": "2411.14001", "title": "Graph Domain Adaptation with Dual-branch Encoder and Two-level Alignment\n  for Whole Slide Image-based Survival Prediction", "abstract": "  In recent years, histopathological whole slide image (WSI)- based survival\nanalysis has attracted much attention in medical image analysis. In practice,\nWSIs usually come from different hospitals or laboratories, which can be seen\nas different domains, and thus may have significant differences in imaging\nequipment, processing procedures, and sample sources. These differences\ngenerally result in large gaps in distribution between different WSI domains,\nand thus the survival analysis models trained on one domain may fail to\ntransfer to another. To address this issue, we propose a Dual-branch Encoder\nand Two-level Alignment (DETA) framework to explore both feature and\ncategory-level alignment between different WSI domains. Specifically, we first\nformulate the concerned problem as graph domain adaptation (GDA) by virtue the\ngraph representation of WSIs. Then we construct a dual-branch graph encoder,\nincluding the message passing branch and the shortest path branch, to\nexplicitly and implicitly extract semantic information from the\ngraph-represented WSIs. To realize GDA, we propose a two-level alignment\napproach: at the category level, we develop a coupling technique by virtue of\nthe dual-branch structure, leading to reduced divergence between the category\ndistributions of the two domains; at the feature level, we introduce an\nadversarial perturbation strategy to better augment source domain feature,\nresulting in improved alignment in feature distribution. To the best of our\nknowledge, our work is the first attempt to alleviate the domain shift issue\nfor WSI data analysis. Extensive experiments on four TCGA datasets have\nvalidated the effectiveness of our proposed DETA framework and demonstrated its\nsuperior performance in WSI-based survival analysis.\n", "date": "2024-11-21", "categories": []}
{"id": "2411.11613", "title": "Leveraging Computational Pathology AI for Noninvasive Optical Imaging\n  Analysis Without Retraining", "abstract": "  Noninvasive optical imaging modalities can probe patient's tissue in 3D and\nover time generate gigabytes of clinically relevant data per sample. There is a\nneed for AI models to analyze this data and assist clinical workflow. The lack\nof expert labelers and the large dataset required (>100,000 images) for model\ntraining and tuning are the main hurdles in creating foundation models. In this\npaper we introduce FoundationShift, a method to apply any AI model from\ncomputational pathology without retraining. We show our method is more accurate\nthan state of the art models (SAM, MedSAM, SAM-Med2D, CellProfiler, Hover-Net,\nPLIP, UNI and ChatGPT), with multiple imaging modalities (OCT and RCM). This is\nachieved without the need for model retraining or fine-tuning. Applying our\nmethod to noninvasive in vivo images could enable physicians to readily\nincorporate optical imaging modalities into their clinical practice, providing\nreal time tissue analysis and improving patient care.\n", "date": "2024-11-18", "categories": []}
{"id": "2411.10709", "title": "Diagnostic Text-guided Representation Learning in Hierarchical\n  Classification for Pathological Whole Slide Image", "abstract": "  With the development of digital imaging in medical microscopy, artificial\nintelligent-based analysis of pathological whole slide images (WSIs) provides a\npowerful tool for cancer diagnosis. Limited by the expensive cost of\npixel-level annotation, current research primarily focuses on representation\nlearning with slide-level labels, showing success in various downstream tasks.\nHowever, given the diversity of lesion types and the complex relationships\nbetween each other, these techniques still deserve further exploration in\naddressing advanced pathology tasks. To this end, we introduce the concept of\nhierarchical pathological image classification and propose a representation\nlearning called PathTree. PathTree considers the multi-classification of\ndiseases as a binary tree structure. Each category is represented as a\nprofessional pathological text description, which messages information with a\ntree-like encoder. The interactive text features are then used to guide the\naggregation of hierarchical multiple representations. PathTree uses slide-text\nsimilarity to obtain probability scores and introduces two extra tree specific\nlosses to further constrain the association between texts and slides. Through\nextensive experiments on three challenging hierarchical classification\ndatasets: in-house cryosectioned lung tissue lesion identification, public\nprostate cancer grade assessment, and public breast cancer subtyping, our\nproposed PathTree is consistently competitive compared to the state-of-the-art\nmethods and provides a new perspective on the deep learning-assisted solution\nfor more complex WSI classification.\n", "date": "2024-11-16", "categories": []}
{"id": "2411.10752", "title": "Towards a Comprehensive Benchmark for Pathological Lymph Node Metastasis\n  in Breast Cancer Sections", "abstract": "  Advances in optical microscopy scanning have significantly contributed to\ncomputational pathology (CPath) by converting traditional histopathological\nslides into whole slide images (WSIs). This development enables comprehensive\ndigital reviews by pathologists and accelerates AI-driven diagnostic support\nfor WSI analysis. Recent advances in foundational pathology models have\nincreased the need for benchmarking tasks. The Camelyon series is one of the\nmost widely used open-source datasets in computational pathology. However, the\nquality, accessibility, and clinical relevance of the labels have not been\ncomprehensively evaluated. In this study, we reprocessed 1,399 WSIs and labels\nfrom the Camelyon-16 and Camelyon-17 datasets, removing low-quality slides,\ncorrecting erroneous labels, and providing expert pixel annotations for tumor\nregions in the previously unreleased test set. Based on the sizes of\nre-annotated tumor regions, we upgraded the binary cancer screening task to a\nfour-class task: negative, micro-metastasis, macro-metastasis, and Isolated\nTumor Cells (ITC). We reevaluated pre-trained pathology feature extractors and\nmultiple instance learning (MIL) methods using the cleaned dataset, providing a\nbenchmark that advances AI development in histopathology.\n", "date": "2024-11-16", "categories": []}
{"id": "2411.11262", "title": "Cross-Patient Pseudo Bags Generation and Curriculum Contrastive Learning\n  for Imbalanced Multiclassification of Whole Slide Image", "abstract": "  Pathology computing has dramatically improved pathologists' workflow and\ndiagnostic decision-making processes. Although computer-aided diagnostic\nsystems have shown considerable value in whole slide image (WSI) analysis, the\nproblem of multi-classification under sample imbalance remains an intractable\nchallenge. To address this, we propose learning fine-grained information by\ngenerating sub-bags with feature distributions similar to the original WSIs.\nAdditionally, we utilize a pseudo-bag generation algorithm to further leverage\nthe abundant and redundant information in WSIs, allowing efficient training in\nunbalanced-sample multi-classification tasks. Furthermore, we introduce an\naffinity-based sample selection and curriculum contrastive learning strategy to\nenhance the stability of model representation learning. Unlike previous\napproaches, our framework transitions from learning bag-level representations\nto understanding and exploiting the feature distribution of multi-instance\nbags. Our method demonstrates significant performance improvements on three\ndatasets, including tumor classification and lymph node metastasis. On average,\nit achieves a 4.39-point improvement in F1 score compared to the second-best\nmethod across the three tasks, underscoring its superior performance.\n", "date": "2024-11-18", "categories": []}
{"id": "2406.01613", "title": "QuST: QuPath Extension for Integrative Whole Slide Image and Spatial\n  Transcriptomics Analysis", "abstract": "  The integration of AI in digital pathology, particularly in whole slide image\n(WSI) and spatial transcriptomics (ST) analysis, holds immense potential for\nenhancing our understanding of diseases. Despite challenges such as training\npattern preparation and resolution disparities, the convergence of these\ntechnologies can unlock new insights. We introduce QuST, a tool that bridges\nthe gap between WSI and ST, underscoring the transformative power of this\nintegrated approach in disease biology.\n", "date": "2024-05-31", "categories": []}
{"id": "2411.09767", "title": "Deep Learning for Fetal Inflammatory Response Diagnosis in the Umbilical\n  Cord", "abstract": "  Inflammation of the umbilical cord can be seen as a result of ascending\nintrauterine infection or other inflammatory stimuli. Acute fetal inflammatory\nresponse (FIR) is characterized by infiltration of the umbilical cord by fetal\nneutrophils, and can be associated with neonatal sepsis or fetal inflammatory\nresponse syndrome. Recent advances in deep learning in digital pathology have\ndemonstrated favorable performance across a wide range of clinical tasks, such\nas diagnosis and prognosis. In this study we classified FIR from whole slide\nimages (WSI). We digitized 4100 histological slides of umbilical cord stained\nwith hematoxylin and eosin(H&E) and extracted placental diagnoses from the\nelectronic health record. We build models using attention-based whole slide\nlearning models. We compared strategies between features extracted by a model\n(ConvNeXtXLarge) pretrained on non-medical images (ImageNet), and one\npretrained using histopathology images (UNI). We trained multiple iterations of\neach model and combined them into an ensemble. The predictions from the\nensemble of models trained using UNI achieved an overall balanced accuracy of\n0.836 on the test dataset. In comparison, the ensembled predictions using\nConvNeXtXLarge had a lower balanced accuracy of 0.7209. Heatmaps generated from\ntop accuracy model appropriately highlighted arteritis in cases of FIR 2. In\nFIR 1, the highest performing model assigned high attention to areas of\nactivated-appearing stroma in Wharton's Jelly. However, other high-performing\nmodels assigned attention to umbilical vessels. We developed models for\ndiagnosis of FIR from placental histology images, helping reduce interobserver\nvariability among pathologists. Future work may examine the utility of these\nmodels for identifying infants at risk of systemic inflammatory response or\nearly onset neonatal sepsis.\n", "date": "2024-11-14", "categories": []}
{"id": "2411.09894", "title": "Free Lunch in Pathology Foundation Model: Task-specific Model Adaptation\n  with Concept-Guided Feature Enhancement", "abstract": "  Whole slide image (WSI) analysis is gaining prominence within the medical\nimaging field. Recent advances in pathology foundation models have shown the\npotential to extract powerful feature representations from WSIs for downstream\ntasks. However, these foundation models are usually designed for\ngeneral-purpose pathology image analysis and may not be optimal for specific\ndownstream tasks or cancer types. In this work, we present Concept\nAnchor-guided Task-specific Feature Enhancement (CATE), an adaptable paradigm\nthat can boost the expressivity and discriminativeness of pathology foundation\nmodels for specific downstream tasks. Based on a set of task-specific concepts\nderived from the pathology vision-language model with expert-designed prompts,\nwe introduce two interconnected modules to dynamically calibrate the generic\nimage features extracted by foundation models for certain tasks or cancer\ntypes. Specifically, we design a Concept-guided Information Bottleneck module\nto enhance task-relevant characteristics by maximizing the mutual information\nbetween image features and concept anchors while suppressing superfluous\ninformation. Moreover, a Concept-Feature Interference module is proposed to\nutilize the similarity between calibrated features and concept anchors to\nfurther generate discriminative task-specific features. The extensive\nexperiments on public WSI datasets demonstrate that CATE significantly enhances\nthe performance and generalizability of MIL models. Additionally, heatmap and\numap visualization results also reveal the effectiveness and interpretability\nof CATE. The source code is available at https://github.com/HKU-MedAI/CATE.\n", "date": "2024-11-15", "categories": []}
{"id": "2411.08936", "title": "Clustered Patch Embeddings for Permutation-Invariant Classification of\n  Whole Slide Images", "abstract": "  Whole Slide Imaging (WSI) is a cornerstone of digital pathology, offering\ndetailed insights critical for diagnosis and research. Yet, the gigapixel size\nof WSIs imposes significant computational challenges, limiting their practical\nutility. Our novel approach addresses these challenges by leveraging various\nencoders for intelligent data reduction and employing a different\nclassification model to ensure robust, permutation-invariant representations of\nWSIs. A key innovation of our method is the ability to distill the complex\ninformation of an entire WSI into a single vector, effectively capturing the\nessential features needed for accurate analysis. This approach significantly\nenhances the computational efficiency of WSI analysis, enabling more accurate\npathological assessments without the need for extensive computational\nresources. This breakthrough equips us with the capability to effectively\naddress the challenges posed by large image resolutions in whole-slide imaging,\npaving the way for more scalable and effective utilization of WSIs in medical\ndiagnostics and research, marking a significant advancement in the field.\n", "date": "2024-11-13", "categories": []}
{"id": "2411.08975", "title": "Fluoroformer: Scaling multiple instance learning to multiplexed images\n  via attention-based channel fusion", "abstract": "  Though multiple instance learning (MIL) has been a foundational strategy in\ncomputational pathology for processing whole slide images (WSIs), current\napproaches are designed for traditional hematoxylin and eosin (H&E) slides\nrather than emerging multiplexed technologies. Here, we present an MIL\nstrategy, the Fluoroformer module, that is specifically tailored to multiplexed\nWSIs by leveraging scaled dot-product attention (SDPA) to interpretably fuse\ninformation across disparate channels. On a cohort of 434 non-small cell lung\ncancer (NSCLC) samples, we show that the Fluoroformer both obtains strong\nprognostic performance and recapitulates immuno-oncological hallmarks of NSCLC.\nOur technique thereby provides a path for adapting state-of-the-art AI\ntechniques to emerging spatial biology assays.\n", "date": "2024-11-13", "categories": []}
{"id": "2406.18054", "title": "Leveraging Pre-trained Models for FF-to-FFPE Histopathological Image\n  Translation", "abstract": "  The two primary types of Hematoxylin and Eosin (H&E) slides in histopathology\nare Formalin-Fixed Paraffin-Embedded (FFPE) and Fresh Frozen (FF). FFPE slides\noffer high quality histopathological images but require a labor-intensive\nacquisition process. In contrast, FF slides can be prepared quickly, but the\nimage quality is relatively poor. Our task is to translate FF images into FFPE\nstyle, thereby improving the image quality for diagnostic purposes. In this\npaper, we propose Diffusion-FFPE, a method for FF-to-FFPE histopathological\nimage translation using a pre-trained diffusion model. Specifically, we utilize\na one-step diffusion model as the generator, which we fine-tune using LoRA\nadapters within an adversarial learning framework. To enable the model to\neffectively capture both global structural patterns and local details, we\nintroduce a multi-scale feature fusion module that leverages two VAE encoders\nto extract features at different image resolutions, performing feature fusion\nbefore inputting them into the UNet. Additionally, a pre-trained\nvision-language model for histopathology serves as the backbone for the\ndiscriminator, enhancing model performance. Our FF-to-FFPE translation\nexperiments on the TCGA-NSCLC dataset demonstrate that the proposed approach\noutperforms existing methods. The code and models are released at\nhttps://github.com/QilaiZhang/Diffusion-FFPE.\n", "date": "2024-06-26", "categories": []}
{"id": "2411.08531", "title": "Classification and Morphological Analysis of DLBCL Subtypes in\n  H\\&E-Stained Slides", "abstract": "  We address the challenge of automated classification of diffuse large B-cell\nlymphoma (DLBCL) into its two primary subtypes: activated B-cell-like (ABC) and\ngerminal center B-cell-like (GCB). Accurate classification between these\nsubtypes is essential for determining the appropriate therapeutic strategy,\ngiven their distinct molecular profiles and treatment responses. Our proposed\ndeep learning model demonstrates robust performance, achieving an average area\nunder the curve (AUC) of (87.4 pm 5.7)\\% during cross-validation. It shows a\nhigh positive predictive value (PPV), highlighting its potential for clinical\napplication, such as triaging for molecular testing. To gain biological\ninsights, we performed an analysis of morphological features of ABC and GCB\nsubtypes. We segmented cell nuclei using a pre-trained deep neural network and\ncompared the statistics of geometric and color features for ABC and GCB. We\nfound that the distributions of these features were not very different for the\ntwo subtypes, which suggests that the visual differences between them are more\nsubtle. These results underscore the potential of our method to assist in more\nprecise subtype classification and can contribute to improved treatment\nmanagement and outcomes for patients of DLBCL.\n", "date": "2024-11-13", "categories": []}
{"id": "2401.16131", "title": "CIMIL-CRC: a clinically-informed multiple instance learning framework\n  for patient-level colorectal cancer molecular subtypes classification from\n  H\\&E stained images", "abstract": "  Treatment approaches for colorectal cancer (CRC) are highly dependent on the\nmolecular subtype, as immunotherapy has shown efficacy in cases with\nmicrosatellite instability (MSI) but is ineffective for the microsatellite\nstable (MSS) subtype. There is promising potential in utilizing deep neural\nnetworks (DNNs) to automate the differentiation of CRC subtypes by analyzing\nHematoxylin and Eosin (H\\&E) stained whole-slide images (WSIs). Due to the\nextensive size of WSIs, Multiple Instance Learning (MIL) techniques are\ntypically explored. However, existing MIL methods focus on identifying the most\nrepresentative image patches for classification, which may result in the loss\nof critical information. Additionally, these methods often overlook clinically\nrelevant information, like the tendency for MSI class tumors to predominantly\noccur on the proximal (right side) colon. We introduce `CIMIL-CRC', a DNN\nframework that: 1) solves the MSI/MSS MIL problem by efficiently combining a\npre-trained feature extraction model with principal component analysis (PCA) to\naggregate information from all patches, and 2) integrates clinical priors,\nparticularly the tumor location within the colon, into the model to enhance\npatient-level classification accuracy. We assessed our CIMIL-CRC method using\nthe average area under the curve (AUC) from a 5-fold cross-validation\nexperimental setup for model development on the TCGA-CRC-DX cohort, contrasting\nit with a baseline patch-level classification, MIL-only approach, and\nClinically-informed patch-level classification approach. Our CIMIL-CRC\noutperformed all methods (AUROC: $0.92\\pm0.002$ (95\\% CI 0.91-0.92), vs.\n$0.79\\pm0.02$ (95\\% CI 0.76-0.82), $0.86\\pm0.01$ (95\\% CI 0.85-0.88), and\n$0.87\\pm0.01$ (95\\% CI 0.86-0.88), respectively). The improvement was\nstatistically significant.\n", "date": "2024-01-29", "categories": []}
{"id": "2411.05959", "title": "Efficient Self-Supervised Barlow Twins from Limited Tissue Slide Cohorts\n  for Colonic Pathology Diagnostics", "abstract": "  Colorectal cancer (CRC) is one of the few cancers that have an established\ndysplasia-carcinoma sequence that benefits from screening. Everyone over 50\nyears of age in Canada is eligible for CRC screening. About 20\\% of those\npeople will undergo a biopsy for a pre-neoplastic polyp and, in many cases,\nmultiple polyps. As such, these polyp biopsies make up the bulk of a\npathologist's workload. Developing an efficient computational model to help\nscreen these polyp biopsies can improve the pathologist's workflow and help\nguide their attention to critical areas on the slide. DL models face\nsignificant challenges in computational pathology (CPath) because of the\ngigapixel image size of whole-slide images and the scarcity of detailed\nannotated datasets. It is, therefore, crucial to leverage self-supervised\nlearning (SSL) methods to alleviate the burden and cost of data annotation.\nHowever, current research lacks methods to apply SSL frameworks to analyze\npathology data effectively. This paper aims to propose an optimized Barlow\nTwins framework for colorectal polyps screening. We adapt its hyperparameters,\naugmentation strategy and encoder to the specificity of the pathology data to\nenhance performance. Additionally, we investigate the best Field of View (FoV)\nfor colorectal polyps screening and propose a new benchmark dataset for CRC\nscreening, made of four types of colorectal polyps and normal tissue, by\nperforming downstream tasking on MHIST and NCT-CRC-7K datasets. Furthermore, we\nshow that the SSL representations are more meaningful and qualitative than the\nsupervised ones and that Barlow Twins benefits from the Swin Transformer when\napplied to pathology data. Codes are avaialble from\nhttps://github.com/AtlasAnalyticsLab/PathBT.\n", "date": "2024-11-08", "categories": []}
{"id": "2411.05489", "title": "Do Histopathological Foundation Models Eliminate Batch Effects? A\n  Comparative Study", "abstract": "  Deep learning has led to remarkable advancements in computational\nhistopathology, e.g., in diagnostics, biomarker prediction, and outcome\nprognosis. Yet, the lack of annotated data and the impact of batch effects,\ne.g., systematic technical data differences across hospitals, hamper model\nrobustness and generalization. Recent histopathological foundation models --\npretrained on millions to billions of images -- have been reported to improve\ngeneralization performances on various downstream tasks. However, it has not\nbeen systematically assessed whether they fully eliminate batch effects. In\nthis study, we empirically show that the feature embeddings of the foundation\nmodels still contain distinct hospital signatures that can lead to biased\npredictions and misclassifications. We further find that the signatures are not\nremoved by stain normalization methods, dominate distances in feature space,\nand are evident across various principal components. Our work provides a novel\nperspective on the evaluation of medical foundation models, paving the way for\nmore robust pretraining strategies and downstream predictors.\n", "date": "2024-11-08", "categories": []}
{"id": "2408.00738", "title": "Virchow2: Scaling Self-Supervised Mixed Magnification Models in\n  Pathology", "abstract": "  Foundation models are rapidly being developed for computational pathology\napplications. However, it remains an open question which factors are most\nimportant for downstream performance with data scale and diversity, model size,\nand training algorithm all playing a role. In this work, we propose algorithmic\nmodifications, tailored for pathology, and we present the result of scaling\nboth data and model size, surpassing previous studies in both dimensions. We\nintroduce three new models: Virchow2, a 632 million parameter vision\ntransformer, Virchow2G, a 1.9 billion parameter vision transformer, and\nVirchow2G Mini, a 22 million parameter distillation of Virchow2G, each trained\nwith 3.1 million histopathology whole slide images, with diverse tissues,\noriginating institutions, and stains. We achieve state of the art performance\non 12 tile-level tasks, as compared to the top performing competing models. Our\nresults suggest that data diversity and domain-specific methods can outperform\nmodels that only scale in the number of parameters, but, on average,\nperformance benefits from the combination of domain-specific methods, data\nscale, and model scale.\n", "date": "2024-08-01", "categories": []}
{"id": "2408.13800", "title": "BCDNet: A Fast Residual Neural Network For Invasive Ductal Carcinoma\n  Detection", "abstract": "  It is of great significance to diagnose Invasive Ductal Carcinoma (IDC) in\nearly stage, which is the most common subtype of breast cancer. Although the\npowerful models in the Computer-Aided Diagnosis (CAD) systems provide promising\nresults, it is still difficult to integrate them into other medical devices or\nuse them without sufficient computation resource. In this paper, we propose\nBCDNet, which firstly upsamples the input image by the residual block and use\nsmaller convolutional block and a special MLP to learn features. BCDNet is\nproofed to effectively detect IDC in histopathological RGB images with an\naverage accuracy of 91.6% and reduce training consumption effectively compared\nto ResNet 50 and ViT-B-16.\n", "date": "2024-08-25", "categories": []}
{"id": "2411.03341", "title": "Interpretable Embeddings for Segmentation-Free Single-Cell Analysis in\n  Multiplex Imaging", "abstract": "  Multiplex Imaging (MI) enables the simultaneous visualization of multiple\nbiological markers in separate imaging channels at subcellular resolution,\nproviding valuable insights into cell-type heterogeneity and spatial\norganization. However, current computational pipelines rely on cell\nsegmentation algorithms, which require laborious fine-tuning and can introduce\ndownstream errors due to inaccurate single-cell representations. We propose a\nsegmentation-free deep learning approach that leverages grouped convolutions to\nlearn interpretable embedded features from each imaging channel, enabling\nrobust cell-type identification without manual feature selection. Validated on\nan Imaging Mass Cytometry dataset of 1.8 million cells from neuroblastoma\npatients, our method enables the accurate identification of known cell types,\nshowcasing its scalability and suitability for high-dimensional MI data.\n", "date": "2024-11-02", "categories": []}
{"id": "2406.19899", "title": "On the Value of PHH3 for Mitotic Figure Detection on H&E-stained Images", "abstract": "  The count of mitotic figures (MFs) observed in hematoxylin and eosin\n(H&E)-stained slides is an important prognostic marker as it is a measure for\ntumor cell proliferation. However, the identification of MFs has a known low\ninter-rater agreement. Deep learning algorithms can standardize this task, but\nthey require large amounts of annotated data for training and validation.\nFurthermore, label noise introduced during the annotation process may impede\nthe algorithm's performance. Unlike H&E, the mitosis-specific antibody\nphospho-histone H3 (PHH3) specifically highlights MFs. Counting MFs on slides\nstained against PHH3 leads to higher agreement among raters and has therefore\nrecently been used as a ground truth for the annotation of MFs in H&E. However,\nas PHH3 facilitates the recognition of cells indistinguishable from H&E stain\nalone, the use of this ground truth could potentially introduce noise into the\nH&E-related dataset, impacting model performance. This study analyzes the\nimpact of PHH3-assisted MF annotation on inter-rater reliability and object\nlevel agreement through an extensive multi-rater experiment. We found that the\nannotators' object-level agreement increased when using PHH3-assisted labeling.\nSubsequently, MF detectors were evaluated on the resulting datasets to\ninvestigate the influence of PHH3-assisted labeling on the models' performance.\nAdditionally, a novel dual-stain MF detector was developed to investigate the\ninterpretation-shift of PHH3-assisted labels used in H&E, which clearly\noutperformed single-stain detectors. However, the PHH3-assisted labels did not\nhave a positive effect on solely H&E-based models. The high performance of our\ndual-input detector reveals an information mismatch between the H&E and\nPHH3-stained images as the cause of this effect.\n", "date": "2024-06-28", "categories": []}
{"id": "2411.00948", "title": "Multiplex Imaging Analysis in Pathology: a Comprehensive Review on\n  Analytical Approaches and Digital Toolkits", "abstract": "  Conventional histopathology has long been essential for disease diagnosis,\nrelying on visual inspection of tissue sections. Immunohistochemistry aids in\ndetecting specific biomarkers but is limited by its single-marker approach,\nrestricting its ability to capture the full tissue environment. The advent of\nmultiplexed imaging technologies, like multiplexed immunofluorescence and\nspatial transcriptomics, allows for simultaneous visualization of multiple\nbiomarkers in a single section, enhancing morphological data with molecular and\nspatial information. This provides a more comprehensive view of the tissue\nmicroenvironment, cellular interactions, and disease mechanisms - crucial for\nunderstanding disease progression, prognosis, and treatment response. However,\nthe extensive data from multiplexed imaging necessitates sophisticated\ncomputational methods for preprocessing, segmentation, feature extraction, and\nspatial analysis. These tools are vital for managing large, multidimensional\ndatasets, converting raw imaging data into actionable insights. By automating\nlabor-intensive tasks and enhancing reproducibility and accuracy, computational\ntools are pivotal in diagnostics and research. This review explores the current\nlandscape of multiplexed imaging in pathology, detailing workflows and key\ntechnologies like PathML, an AI-powered platform that streamlines image\nanalysis, making complex dataset interpretation accessible for clinical and\nresearch settings.\n", "date": "2024-11-01", "categories": []}
{"id": "2403.05160", "title": "MamMIL: Multiple Instance Learning for Whole Slide Images with State\n  Space Models", "abstract": "  Recently, pathological diagnosis has achieved superior performance by\ncombining deep learning models with the multiple instance learning (MIL)\nframework using whole slide images (WSIs). However, the giga-pixeled nature of\nWSIs poses a great challenge for efficient MIL. Existing studies either do not\nconsider global dependencies among instances, or use approximations such as\nlinear attentions to model the pair-to-pair instance interactions, which\ninevitably brings performance bottlenecks. To tackle this challenge, we propose\na framework named MamMIL for WSI analysis by cooperating the selective\nstructured state space model (i.e., Mamba) with MIL, enabling the modeling of\nglobal instance dependencies while maintaining linear complexity. Specifically,\nconsidering the irregularity of the tissue regions in WSIs, we represent each\nWSI as an undirected graph. To address the problem that Mamba can only process\n1D sequences, we further propose a topology-aware scanning mechanism to\nserialize the WSI graphs while preserving the topological relationships among\nthe instances. Finally, in order to further perceive the topological structures\namong the instances and incorporate short-range feature interactions, we\npropose an instance aggregation block based on graph neural networks.\nExperiments show that MamMIL can achieve advanced performance than the\nstate-of-the-art frameworks. The code can be accessed at\nhttps://github.com/Vison307/MamMIL.\n", "date": "2024-03-08", "categories": []}
{"id": "2209.01380", "title": "Classification of Breast Tumours Based on Histopathology Images Using\n  Deep Features and Ensemble of Gradient Boosting Methods", "abstract": "  Breast cancer is the most common cancer among women worldwide. Early-stage\ndiagnosis of breast cancer can significantly improve the efficiency of\ntreatment. Computer-aided diagnosis (CAD) systems are widely adopted in this\nissue due to their reliability, accuracy and affordability. There are different\nimaging techniques for a breast cancer diagnosis; one of the most accurate ones\nis histopathology which is used in this paper. Deep feature transfer learning\nis used as the main idea of the proposed CAD system's feature extractor.\nAlthough 16 different pre-trained networks have been tested in this study, our\nmain focus is on the classification phase. The Inception-ResNet-v2 which has\nboth residual and inception networks profits together has shown the best\nfeature extraction capability in the case of breast cancer histopathology\nimages among all tested CNNs. In the classification phase, the ensemble of\nCatBoost, XGBoost and LightGBM has provided the best average accuracy. The\nBreakHis dataset was used to evaluate the proposed method. BreakHis contains\n7909 histopathology images (2,480 benign and 5,429 malignant) in four\nmagnification factors. The proposed method's accuracy (IRv2-CXL) using 70% of\nBreakHis dataset as training data in 40x, 100x, 200x and 400x magnification is\n96.82%, 95.84%, 97.01% and 96.15%, respectively. Most studies on automated\nbreast cancer detection have focused on feature extraction, which made us\nattend to the classification phase. IRv2-CXL has shown better or comparable\nresults in all magnifications due to using the soft voting ensemble method\nwhich could combine the advantages of CatBoost, XGBoost and LightGBM together.\n", "date": "2022-09-03", "categories": []}
{"id": "2211.09566", "title": "Enabling Collagen Quantification on HE-stained Slides Through Stain\n  Deconvolution and Restained HE-HES", "abstract": "  In histology, the presence of collagen in the extra-cellular matrix has both\ndiagnostic and prognostic value for cancer malignancy, and can be highlighted\nby adding Saffron (S) to a routine Hematoxylin and Eosin (HE) staining.\nHowever, Saffron is not usually added because of the additional cost and\nbecause pathologists are accustomed to HE, with the exception of France-based\nlaboratories. In this paper, we show that it is possible to quantify the\ncollagen content from the HE image alone and to digitally create an HES image.\nTo do so, we trained a UNet to predict the Saffron densities from HE images. We\ncreated a dataset of registered, restained HE-HES slides and we extracted the\nSaffron concentrations as ground truth using stain deconvolution on the HES\nimages. Our model reached a Mean Absolute Error of 0.0668 $\\pm$ 0.0002 (Saffron\nvalues between 0 and 1) on a 3-fold testing set. We hope our approach can aid\nin improving the clinical workflow while reducing reagent costs for\nlaboratories.\n", "date": "2022-11-17", "categories": []}
{"id": "2410.19690", "title": "Deep Learning for Classification of Inflammatory Bowel Disease Activity\n  in Whole Slide Images of Colonic Histopathology", "abstract": "  Grading inflammatory bowel disease (IBD) activity using standardized\nhistopathological scoring systems remains challenging due to resource\nconstraints and inter-observer variability. In this study, we developed a deep\nlearning model to classify activity grades in hematoxylin and eosin-stained\nwhole slide images (WSIs) from patients with IBD, offering a robust approach\nfor general pathologists. We utilized 2,077 WSIs from 636 patients treated at\nDartmouth-Hitchcock Medical Center in 2018 and 2019, scanned at 40x\nmagnification (0.25 micron/pixel). Board-certified gastrointestinal\npathologists categorized the WSIs into four activity classes: inactive, mildly\nactive, moderately active, and severely active. A transformer-based model was\ndeveloped and validated using five-fold cross-validation to classify IBD\nactivity. Using HoVerNet, we examined neutrophil distribution across activity\ngrades. Attention maps from our model highlighted areas contributing to its\nprediction. The model classified IBD activity with weighted averages of 0.871\n[95% Confidence Interval (CI): 0.860-0.883] for the area under the curve, 0.695\n[95% CI: 0.674-0.715] for precision, 0.697 [95% CI: 0.678-0.716] for recall,\nand 0.695 [95% CI: 0.674-0.714] for F1-score. Neutrophil distribution was\nsignificantly different across activity classes. Qualitative evaluation of\nattention maps by a gastrointestinal pathologist suggested their potential for\nimproved interpretability. Our model demonstrates robust diagnostic performance\nand could enhance consistency and efficiency in IBD activity assessment.\n", "date": "2024-10-25", "categories": []}
{"id": "2406.04377", "title": "Combining Graph Neural Network and Mamba to Capture Local and Global\n  Tissue Spatial Relationships in Whole Slide Images", "abstract": "  In computational pathology, extracting spatial features from gigapixel whole\nslide images (WSIs) is a fundamental task, but due to their large size, WSIs\nare typically segmented into smaller tiles. A critical aspect of this analysis\nis aggregating information from these tiles to make predictions at the WSI\nlevel. We introduce a model that combines a message-passing graph neural\nnetwork (GNN) with a state space model (Mamba) to capture both local and global\nspatial relationships among the tiles in WSIs. The model's effectiveness was\ndemonstrated in predicting progression-free survival among patients with\nearly-stage lung adenocarcinomas (LUAD). We compared the model with other\nstate-of-the-art methods for tile-level information aggregation in WSIs,\nincluding tile-level information summary statistics-based aggregation, multiple\ninstance learning (MIL)-based aggregation, GNN-based aggregation, and\nGNN-transformer-based aggregation. Additional experiments showed the impact of\ndifferent types of node features and different tile sampling strategies on the\nmodel performance. This work can be easily extended to any WSI-based analysis.\nCode: https://github.com/rina-ding/gat-mamba.\n", "date": "2024-06-05", "categories": []}
{"id": "2407.00664", "title": "SCMIL: Sparse Context-aware Multiple Instance Learning for Predicting\n  Cancer Survival Probability Distribution in Whole Slide Images", "abstract": "  Cancer survival prediction is a challenging task that involves analyzing of\nthe tumor microenvironment within Whole Slide Image (WSI). Previous methods\ncannot effectively capture the intricate interaction features among instances\nwithin the local area of WSI. Moreover, existing methods for cancer survival\nprediction based on WSI often fail to provide better clinically meaningful\npredictions. To overcome these challenges, we propose a Sparse Context-aware\nMultiple Instance Learning (SCMIL) framework for predicting cancer survival\nprobability distributions. SCMIL innovatively segments patches into various\nclusters based on their morphological features and spatial location\ninformation, subsequently leveraging sparse self-attention to discern the\nrelationships between these patches with a context-aware perspective.\nConsidering many patches are irrelevant to the task, we introduce a learnable\npatch filtering module called SoftFilter, which ensures that only interactions\nbetween task-relevant patches are considered. To enhance the clinical relevance\nof our prediction, we propose a register-based mixture density network to\nforecast the survival probability distribution for individual patients. We\nevaluate SCMIL on two public WSI datasets from the The Cancer Genome Atlas\n(TCGA) specifically focusing on lung adenocarcinom (LUAD) and kidney renal\nclear cell carcinoma (KIRC). Our experimental results indicate that SCMIL\noutperforms current state-of-the-art methods for survival prediction, offering\nmore clinically meaningful and interpretable outcomes. Our code is accessible\nat https://github.com/yang-ze-kang/SCMIL.\n", "date": "2024-06-30", "categories": []}
{"id": "2410.15886", "title": "Foundation Models for Slide-level Cancer Subtyping in Digital Pathology", "abstract": "  Since the emergence of the ImageNet dataset, the pretraining and fine-tuning\napproach has become widely adopted in computer vision due to the ability of\nImageNet-pretrained models to learn a wide variety of visual features. However,\na significant challenge arises when adapting these models to domain-specific\nfields, such as digital pathology, due to substantial gaps between domains. To\naddress this limitation, foundation models (FM) have been trained on\nlarge-scale in-domain datasets to learn the intricate features of\nhistopathology images. In cancer diagnosis, whole-slide image (WSI) prediction\nis essential for patient prognosis, and multiple instance learning (MIL) has\nbeen implemented to handle the giga-pixel size of WSI. As MIL frameworks rely\non patch-level feature aggregation, this work aims to compare the performance\nof various feature extractors developed under different pretraining strategies\nfor cancer subtyping on WSI under a MIL framework. Results demonstrate the\nability of foundation models to surpass ImageNet-pretrained models for the\nprediction of six skin cancer subtypes\n", "date": "2024-10-21", "categories": []}
{"id": "2410.16038", "title": "Benchmarking Pathology Foundation Models: Adaptation Strategies and\n  Scenarios", "abstract": "  In computational pathology, several foundation models have recently emerged\nand demonstrated enhanced learning capability for analyzing pathology images.\nHowever, adapting these models to various downstream tasks remains challenging,\nparticularly when faced with datasets from different sources and acquisition\nconditions, as well as limited data availability. In this study, we benchmark\nfour pathology-specific foundation models across 14 datasets and two\nscenarios-consistency assessment and flexibility assessment-addressing diverse\nadaptation scenarios and downstream tasks. In the consistency assessment\nscenario, involving five fine-tuning methods, we found that the\nparameter-efficient fine-tuning approach was both efficient and effective for\nadapting pathology-specific foundation models to diverse datasets within the\nsame downstream task. In the flexibility assessment scenario under data-limited\nenvironments, utilizing five few-shot learning methods, we observed that the\nfoundation models benefited more from the few-shot learning methods that\ninvolve modification during the testing phase only. These findings provide\ninsights that could guide the deployment of pathology-specific foundation\nmodels in real clinical settings, potentially improving the accuracy and\nreliability of pathology image analysis. The code for this study is available\nat: https://github.com/QuIIL/BenchmarkingPathologyFoundationModels.\n", "date": "2024-10-21", "categories": []}
{"id": "2410.14195", "title": "Rethinking Transformer for Long Contextual Histopathology Whole Slide\n  Image Analysis", "abstract": "  Histopathology Whole Slide Image (WSI) analysis serves as the gold standard\nfor clinical cancer diagnosis in the daily routines of doctors. To develop\ncomputer-aided diagnosis model for WSIs, previous methods typically employ\nMulti-Instance Learning to enable slide-level prediction given only slide-level\nlabels. Among these models, vanilla attention mechanisms without pairwise\ninteractions have traditionally been employed but are unable to model\ncontextual information. More recently, self-attention models have been utilized\nto address this issue. To alleviate the computational complexity of long\nsequences in large WSIs, methods like HIPT use region-slicing, and TransMIL\nemploys approximation of full self-attention. Both approaches suffer from\nsuboptimal performance due to the loss of key information. Moreover, their use\nof absolute positional embedding struggles to effectively handle long\ncontextual dependencies in shape-varying WSIs. In this paper, we first analyze\nhow the low-rank nature of the long-sequence attention matrix constrains the\nrepresentation ability of WSI modelling. Then, we demonstrate that the rank of\nattention matrix can be improved by focusing on local interactions via a local\nattention mask. Our analysis shows that the local mask aligns with the\nattention patterns in the lower layers of the Transformer. Furthermore, the\nlocal attention mask can be implemented during chunked attention calculation,\nreducing the quadratic computational complexity to linear with a small local\nbandwidth. Building on this, we propose a local-global hybrid Transformer for\nboth computational acceleration and local-global information interactions\nmodelling. Our method, Long-contextual MIL (LongMIL), is evaluated through\nextensive experiments on various WSI tasks to validate its superiority. Our\ncode will be available at github.com/invoker-LL/Long-MIL.\n", "date": "2024-10-18", "categories": []}
{"id": "2410.14365", "title": "Impact of imperfect annotations on CNN training and performance for\n  instance segmentation and classification in digital pathology", "abstract": "  Segmentation and classification of large numbers of instances, such as cell\nnuclei, are crucial tasks in digital pathology for accurate diagnosis. However,\nthe availability of high-quality datasets for deep learning methods is often\nlimited due to the complexity of the annotation process. In this work, we\ninvestigate the impact of noisy annotations on the training and performance of\na state-of-the-art CNN model for the combined task of detecting, segmenting and\nclassifying nuclei in histopathology images. In this context, we investigate\nthe conditions for determining an appropriate number of training epochs to\nprevent overfitting to annotation noise during training. Our results indicate\nthat the utilisation of a small, correctly annotated validation set is\ninstrumental in avoiding overfitting and maintaining model performance to a\nlarge extent. Additionally, our findings underscore the beneficial role of\npre-training.\n", "date": "2024-10-18", "categories": []}
{"id": "2410.12833", "title": "MyData: A Comprehensive Database of Mycetoma Tissue Microscopic Images\n  for Histopathological Analysis", "abstract": "  Mycetoma is a chronic and neglected inflammatory disease prevalent in\ntropical and subtropical regions. It can lead to severe disability and social\nstigma. The disease is classified into two types based on the causative\nmicroorganisms: eumycetoma (fungal) and actinomycetoma (bacterial). Effective\ntreatment strategies depend on accurately identifying the causative agents.\nCurrent identification methods include molecular, cytological, and\nhistopathological techniques, as well as grain culturing. Among these,\nhistopathological techniques are considered optimal for use in endemic areas,\nbut they require expert pathologists for accurate identification, which can be\nchallenging in rural areas lacking such expertise. The advent of digital\npathology and automated image analysis algorithms offers a potential solution.\nThis report introduces a novel dataset designed for the automated detection and\nclassification of mycetoma using histopathological images. It includes the\nfirst database of microscopic images of mycetoma tissue, detailing the entire\npipeline from species distribution and patient sampling to acquisition\nprotocols through histological procedures. The dataset consists of images from\n142 patients, totalling 864 images, each annotated with binary masks indicating\nthe presence of grains, facilitating both detection and segmentation tasks.\n", "date": "2024-10-02", "categories": []}
{"id": "2406.11538", "title": "Improving Quality Control of Whole Slide Images by Explicit Artifact\n  Augmentation", "abstract": "  The problem of artifacts in whole slide image acquisition, prevalent in both\nclinical workflows and research-oriented settings, necessitates human\nintervention and re-scanning. Overcoming this challenge requires developing\nquality control algorithms, that are hindered by the limited availability of\nrelevant annotated data in histopathology. The manual annotation of\nground-truth for artifact detection methods is expensive and time-consuming.\nThis work addresses the issue by proposing a method dedicated to augmenting\nwhole slide images with artifacts. The tool seamlessly generates and blends\nartifacts from an external library to a given histopathology dataset. The\naugmented datasets are then utilized to train artifact classification methods.\nThe evaluation shows their usefulness in classification of the artifacts, where\nthey show an improvement from 0.10 to 0.01 AUROC depending on the artifact\ntype. The framework, model, weights, and ground-truth annotations are freely\nreleased to facilitate open science and reproducible research.\n", "date": "2024-06-17", "categories": []}
{"id": "2410.09818", "title": "TopOC: Topological Deep Learning for Ovarian and Breast Cancer Diagnosis", "abstract": "  Microscopic examination of slides prepared from tissue samples is the primary\ntool for detecting and classifying cancerous lesions, a process that is\ntime-consuming and requires the expertise of experienced pathologists. Recent\nadvances in deep learning methods hold significant potential to enhance medical\ndiagnostics and treatment planning by improving accuracy, reproducibility, and\nspeed, thereby reducing clinicians' workloads and turnaround times. However,\nthe necessity for vast amounts of labeled data to train these models remains a\nmajor obstacle to the development of effective clinical decision support\nsystems.\n  In this paper, we propose the integration of topological deep learning\nmethods to enhance the accuracy and robustness of existing histopathological\nimage analysis models. Topological data analysis (TDA) offers a unique approach\nby extracting essential information through the evaluation of topological\npatterns across different color channels. While deep learning methods capture\nlocal information from images, TDA features provide complementary global\nfeatures. Our experiments on publicly available histopathological datasets\ndemonstrate that the inclusion of topological features significantly improves\nthe differentiation of tumor types in ovarian and breast cancers.\n", "date": "2024-10-13", "categories": []}
{"id": "2410.10260", "title": "Slide-based Graph Collaborative Training for Histopathology Whole Slide\n  Image Analysis", "abstract": "  The development of computational pathology lies in the consensus that\npathological characteristics of tumors are significant guidance for cancer\ndiagnostics. Most existing research focuses on the inner-contextual information\nwithin each WSI yet ignores the possible inter-correlations between slides. As\nthe development of tumors is a continuous process involving a series of\nhistological, morphological, and genetic changes that accumulate over time, the\nsimilarities and differences between WSIs across various stages, grades,\nlocations and patients should potentially contribute to the representation of\nWSIs and deserve to be taken into account in WSI modeling. To verify the\nadvancement of introducing the slide inter-correlations into the representation\nlearning of WSIs, we proposed a generic WSI analysis pipeline SlideGCD that can\nbe adapted to any existing Multiple Instance Learning (MIL) frameworks and\nimprove their performance. With the new paradigm, the prior knowledge of cancer\ndevelopment can participate in the end-to-end workflow, which concurrently\ninitializes and refines the slide representation, as a guide for message\npassing in the slide-based graph. Extensive comparisons and experiments are\nconducted to validate the effectiveness and robustness of the proposed pipeline\nacross 4 different tasks, including cancer subtyping, cancer staging, survival\nprediction, and gene mutation prediction, with 7 representative SOTA WSI\nanalysis frameworks as backbones.\n", "date": "2024-10-14", "categories": []}
{"id": "2305.05314", "title": "CAMIL: Context-Aware Multiple Instance Learning for Cancer Detection and\n  Subtyping in Whole Slide Images", "abstract": "  The visual examination of tissue biopsy sections is fundamental for cancer\ndiagnosis, with pathologists analyzing sections at multiple magnifications to\ndiscern tumor cells and their subtypes. However, existing attention-based\nmultiple instance learning (MIL) models used for analyzing Whole Slide Images\n(WSIs) in cancer diagnostics often overlook the contextual information of tumor\nand neighboring tiles, leading to misclassifications. To address this, we\npropose the Context-Aware Multiple Instance Learning (CAMIL) architecture.\nCAMIL incorporates neighbor-constrained attention to consider dependencies\namong tiles within a WSI and integrates contextual constraints as prior\nknowledge into the MIL model. We evaluated CAMIL on subtyping non-small cell\nlung cancer (TCGA-NSCLC) and detecting lymph node (CAMELYON16 and CAMELYON17)\nmetastasis, achieving test AUCs of 97.5\\%, 95.9\\%, and 88.1\\%, respectively,\noutperforming other state-of-the-art methods. Additionally, CAMIL enhances\nmodel interpretability by identifying regions of high diagnostic value.\n", "date": "2023-05-09", "categories": []}
{"id": "2410.04507", "title": "MECFormer: Multi-task Whole Slide Image Classification with Expert\n  Consultation Network", "abstract": "  Whole slide image (WSI) classification is a crucial problem for cancer\ndiagnostics in clinics and hospitals. A WSI, acquired at gigapixel size, is\ncommonly tiled into patches and processed by multiple-instance learning (MIL)\nmodels. Previous MIL-based models designed for this problem have only been\nevaluated on individual tasks for specific organs, and the ability to handle\nmultiple tasks within a single model has not been investigated. In this study,\nwe propose MECFormer, a generative Transformer-based model designed to handle\nmultiple tasks within one model. To leverage the power of learning multiple\ntasks simultaneously and to enhance the model's effectiveness in focusing on\neach individual task, we introduce an Expert Consultation Network, a projection\nlayer placed at the beginning of the Transformer-based model. Additionally, to\nenable flexible classification, autoregressive decoding is incorporated by a\nlanguage decoder for WSI classification. Through extensive experiments on five\ndatasets involving four different organs, one cancer classification task, and\nfour cancer subtyping tasks, MECFormer demonstrates superior performance\ncompared to individual state-of-the-art multiple-instance learning models.\n", "date": "2024-10-06", "categories": []}
{"id": "2410.06974", "title": "Diagnosis of Malignant Lymphoma Cancer Using Hybrid Optimized Techniques\n  Based on Dense Neural Networks", "abstract": "  Lymphoma diagnosis, particularly distinguishing between subtypes, is critical\nfor effective treatment but remains challenging due to the subtle morphological\ndifferences in histopathological images. This study presents a novel hybrid\ndeep learning framework that combines DenseNet201 for feature extraction with a\nDense Neural Network (DNN) for classification, optimized using the Harris Hawks\nOptimization (HHO) algorithm. The model was trained on a dataset of 15,000\nbiopsy images, spanning three lymphoma subtypes: Chronic Lymphocytic Leukemia\n(CLL), Follicular Lymphoma (FL), and Mantle Cell Lymphoma (MCL). Our approach\nachieved a testing accuracy of 99.33\\%, demonstrating significant improvements\nin both accuracy and model interpretability. Comprehensive evaluation using\nprecision, recall, F1-score, and ROC-AUC underscores the model's robustness and\npotential for clinical adoption. This framework offers a scalable solution for\nimproving diagnostic accuracy and efficiency in oncology.\n", "date": "2024-10-09", "categories": []}
{"id": "2410.07124", "title": "Cross-Task Pretraining for Cross-Organ Cross-Scanner Adenocarcinoma\n  Segmentation", "abstract": "  This short abstract describes a solution to the COSAS 2024 competition on\nCross-Organ and Cross-Scanner Adenocarcinoma Segmentation from\nhistopathological image patches. The main challenge in the task of segmenting\nthis type of cancer is a noticeable domain shift encountered when changing\nacquisition devices (microscopes) and also when tissue comes from different\norgans. The two tasks proposed in COSAS were to train on a dataset of images\nfrom three different organs, and then predict segmentations on data from unseen\norgans (dataset T1), and to train on a dataset of images acquired on three\ndifferent scanners and then segment images acquired with another unseen\nmicroscope. We attempted to bridge the domain shift gap by experimenting with\nthree different strategies: standard training for each dataset, pretraining on\ndataset T1 and then fine-tuning on dataset T2 (and vice-versa, a strategy we\ncall \\textit{Cross-Task Pretraining}), and training on the combination of\ndataset A and B. Our experiments showed that Cross-Task Pre-training is a more\npromising approach to domain generalization.\n", "date": "2024-09-20", "categories": []}
{"id": "2409.16205", "title": "Segmentation Strategies in Deep Learning for Prostate Cancer Diagnosis:\n  A Comparative Study of Mamba, SAM, and YOLO", "abstract": "  Accurate segmentation of prostate cancer histopathology images is crucial for\ndiagnosis and treatment planning. This study presents a comparative analysis of\nthree deep learning-based methods, Mamba, SAM, and YOLO, for segmenting\nprostate cancer histopathology images. We evaluated the performance of these\nmodels on two comprehensive datasets, Gleason 2019 and SICAPv2, using Dice\nscore, precision, and recall metrics. Our results show that the High-order\nVision Mamba UNet (H-vmunet) model outperforms the other two models, achieving\nthe highest scores across all metrics on both datasets. The H-vmunet model's\nadvanced architecture, which integrates high-order visual state spaces and\n2D-selective-scan operations, enables efficient and sensitive lesion detection\nacross different scales. Our study demonstrates the potential of the H-vmunet\nmodel for clinical applications and highlights the importance of robust\nvalidation and comparison of deep learning-based methods for medical image\nanalysis. The findings of this study contribute to the development of accurate\nand reliable computer-aided diagnosis systems for prostate cancer. The code is\navailable at http://github.com/alibdz/prostate-segmentation.\n", "date": "2024-09-24", "categories": []}
{"id": "2409.17122", "title": "Classification of Gleason Grading in Prostate Cancer Histopathology\n  Images Using Deep Learning Techniques: YOLO, Vision Transformers, and Vision\n  Mamba", "abstract": "  Prostate cancer ranks among the leading health issues impacting men, with the\nGleason scoring system serving as the primary method for diagnosis and\nprognosis. This system relies on expert pathologists to evaluate samples of\nprostate tissue and assign a Gleason grade, a task that requires significant\ntime and manual effort. To address this challenge, artificial intelligence (AI)\nsolutions have been explored to automate the grading process. In light of these\nchallenges, this study evaluates and compares the effectiveness of three deep\nlearning methodologies, YOLO, Vision Transformers, and Vision Mamba, in\naccurately classifying Gleason grades from histopathology images. The goal is\nto enhance diagnostic precision and efficiency in prostate cancer management.\nThis study utilized two publicly available datasets, Gleason2019 and SICAPv2,\nto train and test the performance of YOLO, Vision Transformers, and Vision\nMamba models. Each model was assessed based on its ability to classify Gleason\ngrades accurately, considering metrics such as false positive rate, false\nnegative rate, precision, and recall. The study also examined the computational\nefficiency and applicability of each method in a clinical setting. Vision Mamba\ndemonstrated superior performance across all metrics, achieving high precision\nand recall rates while minimizing false positives and negatives. YOLO showed\npromise in terms of speed and efficiency, particularly beneficial for real-time\nanalysis. Vision Transformers excelled in capturing long-range dependencies\nwithin images, although they presented higher computational complexity compared\nto the other models. Vision Mamba emerges as the most effective model for\nGleason grade classification in histopathology images, offering a balance\nbetween accuracy and computational efficiency.\n", "date": "2024-09-25", "categories": []}
{"id": "2410.00945", "title": "Evaluating Deep Regression Models for WSI-Based Gene-Expression\n  Prediction", "abstract": "  Prediction of mRNA gene-expression profiles directly from routine whole-slide\nimages (WSIs) using deep learning models could potentially offer cost-effective\nand widely accessible molecular phenotyping. While such WSI-based\ngene-expression prediction models have recently emerged within computational\npathology, the high-dimensional nature of the corresponding regression problem\noffers numerous design choices which remain to be analyzed in detail. This\nstudy provides recommendations on how deep regression models should be trained\nfor WSI-based gene-expression prediction. For example, we conclude that\ntraining a single model to simultaneously regress all 20530 genes is a\ncomputationally efficient yet very strong baseline.\n", "date": "2024-10-01", "categories": []}
{"id": "2410.01408", "title": "SHAP-CAT: A interpretable multi-modal framework enhancing WSI\n  classification via virtual staining and shapley-value-based multimodal fusion", "abstract": "  The multimodal model has demonstrated promise in histopathology. However,\nmost multimodal models are based on H\\&E and genomics, adopting increasingly\ncomplex yet black-box designs. In our paper, we propose a novel interpretable\nmultimodal framework named SHAP-CAT, which uses a Shapley-value-based dimension\nreduction technique for effective multimodal fusion. Starting with two paired\nmodalities -- H\\&E and IHC images, we employ virtual staining techniques to\nenhance limited input data by generating a new clinical-related modality.\nLightweight bag-level representations are extracted from image modalities and a\nShapley-value-based mechanism is used for dimension reduction. For each\ndimension of the bag-level representation, attribution values are calculated to\nindicate how changes in the specific dimensions of the input affect the model\noutput. In this way, we select a few top important dimensions of bag-level\nrepresentation for each image modality to late fusion. Our experimental results\ndemonstrate that the proposed SHAP-CAT framework incorporating synthetic\nmodalities significantly enhances model performance, yielding a 5\\% increase in\naccuracy for the BCI, an 8\\% increase for IHC4BC-ER, and an 11\\% increase for\nthe IHC4BC-PR dataset.\n", "date": "2024-10-02", "categories": []}
{"id": "2410.00152", "title": "Multimodal Alignment of Histopathological Images Using Cell Segmentation\n  and Point Set Matching for Integrative Cancer Analysis", "abstract": "  Histopathological imaging is vital for cancer research and clinical practice,\nwith multiplexed Immunofluorescence (MxIF) and Hematoxylin and Eosin (H&E)\nproviding complementary insights. However, aligning different stains at the\ncell level remains a challenge due to modality differences. In this paper, we\npresent a novel framework for multimodal image alignment using cell\nsegmentation outcomes. By treating cells as point sets, we apply Coherent Point\nDrift (CPD) for initial alignment and refine it with Graph Matching (GM).\nEvaluated on ovarian cancer tissue microarrays (TMAs), our method achieves high\nalignment accuracy, enabling integration of cell-level features across\nmodalities and generating virtual H&E images from MxIF data for enhanced\nclinical interpretation.\n", "date": "2024-09-30", "categories": []}
{"id": "2407.06979", "title": "Can virtual staining for high-throughput screening generalize?", "abstract": "  The large volume and variety of imaging data from high-throughput screening\n(HTS) in the pharmaceutical industry present an excellent resource for training\nvirtual staining models. However, the potential of models trained under one set\nof experimental conditions to generalize to other conditions remains\nunderexplored. This study systematically investigates whether data from three\ncell types (lung, ovarian, and breast) and two phenotypes (toxic and non-toxic\nconditions) commonly found in HTS can effectively train virtual staining models\nto generalize across three typical HTS distribution shifts: unseen phenotypes,\nunseen cell types, and the combination of both. Utilizing a dataset of 772,416\npaired bright-field, cytoplasm, nuclei, and DNA-damage stain images, we\nevaluate the generalization capabilities of models across pixel-based,\ninstance-wise, and biological-feature-based levels. Our findings indicate that\ntraining virtual nuclei and cytoplasm models on non-toxic condition samples not\nonly generalizes to toxic condition samples but leads to improved performance\nacross all evaluation levels compared to training on toxic condition samples.\nGeneralization to unseen cell types shows variability depending on the cell\ntype; models trained on ovarian or lung cell samples often perform well under\nother conditions, while those trained on breast cell samples consistently show\npoor generalization. Generalization to unseen cell types and phenotypes shows\ngood generalization across all levels of evaluation compared to addressing\nunseen cell types alone. This study represents the first large-scale,\ndata-centric analysis of the generalization capability of virtual staining\nmodels trained on diverse HTS datasets, providing valuable strategies for\nexperimental training data generation.\n", "date": "2024-07-09", "categories": []}
{"id": "2407.13027", "title": "SpaRED benchmark: Enhancing Gene Expression Prediction from Histology\n  Images with Spatial Transcriptomics Completion", "abstract": "  Spatial Transcriptomics is a novel technology that aligns histology images\nwith spatially resolved gene expression profiles. Although groundbreaking, it\nstruggles with gene capture yielding high corruption in acquired data. Given\npotential applications, recent efforts have focused on predicting\ntranscriptomic profiles solely from histology images. However, differences in\ndatabases, preprocessing techniques, and training hyperparameters hinder a fair\ncomparison between methods. To address these challenges, we present a\nsystematically curated and processed database collected from 26 public sources,\nrepresenting an 8.6-fold increase compared to previous works. Additionally, we\npropose a state-of-the-art transformer based completion technique for inferring\nmissing gene expression, which significantly boosts the performance of\ntranscriptomic profile predictions across all datasets. Altogether, our\ncontributions constitute the most comprehensive benchmark of gene expression\nprediction from histology images to date and a stepping stone for future\nresearch on spatial transcriptomics.\n", "date": "2024-07-17", "categories": []}
{"id": "2408.13818", "title": "HER2 and FISH Status Prediction in Breast Biopsy H&E-Stained Images\n  Using Deep Learning", "abstract": "  The current standard for detecting human epidermal growth factor receptor 2\n(HER2) status in breast cancer patients relies on HER2 amplification,\nidentified through fluorescence in situ hybridization (FISH) or\nimmunohistochemistry (IHC). However, hematoxylin and eosin (H\\&E) tumor stains\nare more widely available, and accurately predicting HER2 status using H\\&E\ncould reduce costs and expedite treatment selection. Deep Learning algorithms\nfor H&E have shown effectiveness in predicting various cancer features and\nclinical outcomes, including moderate success in HER2 status prediction. In\nthis work, we employed a customized weak supervision classification technique\ncombined with MoCo-v2 contrastive learning to predict HER2 status. We trained\nour pipeline on 182 publicly available H&E Whole Slide Images (WSIs) from The\nCancer Genome Atlas (TCGA), for which annotations by the pathology team at Yale\nSchool of Medicine are publicly available. Our pipeline achieved an Area Under\nthe Curve (AUC) of 0.85 across four different test folds. Additionally, we\ntested our model on 44 H&E slides from the TCGA-BRCA dataset, which had an HER2\nscore of 2+ and included corresponding HER2 status and FISH test results. These\ncases are considered equivocal for IHC, requiring an expensive FISH test on\ntheir IHC slides for disambiguation. Our pipeline demonstrated an AUC of 0.81\non these challenging H&E slides. Reducing the need for FISH test can have\nsignificant implications in cancer treatment equity for underserved\npopulations.\n", "date": "2024-08-25", "categories": []}
{"id": "2310.16954", "title": "Improving Performance in Colorectal Cancer Histology Decomposition using\n  Deep and Ensemble Machine Learning", "abstract": "  In routine colorectal cancer management, histologic samples stained with\nhematoxylin and eosin are commonly used. Nonetheless, their potential for\ndefining objective biomarkers for patient stratification and treatment\nselection is still being explored. The current gold standard relies on\nexpensive and time-consuming genetic tests. However, recent research highlights\nthe potential of convolutional neural networks (CNNs) in facilitating the\nextraction of clinically relevant biomarkers from these readily available\nimages. These CNN-based biomarkers can predict patient outcomes comparably to\ngolden standards, with the added advantages of speed, automation, and minimal\ncost. The predictive potential of CNN-based biomarkers fundamentally relies on\nthe ability of convolutional neural networks (CNNs) to classify diverse tissue\ntypes from whole slide microscope images accurately. Consequently, enhancing\nthe accuracy of tissue class decomposition is critical to amplifying the\nprognostic potential of imaging-based biomarkers. This study introduces a\nhybrid Deep and ensemble machine learning model that surpassed all preceding\nsolutions for this classification task. Our model achieved 96.74% accuracy on\nthe external test set and 99.89% on the internal test set. Recognizing the\npotential of these models in advancing the task, we have made them publicly\navailable for further research and development.\n", "date": "2023-10-25", "categories": []}
{"id": "2405.03642", "title": "Classification of Breast Cancer Histopathology Images using a Modified\n  Supervised Contrastive Learning Method", "abstract": "  Deep neural networks have reached remarkable achievements in medical image\nprocessing tasks, specifically in classifying and detecting various diseases.\nHowever, when confronted with limited data, these networks face a critical\nvulnerability, often succumbing to overfitting by excessively memorizing the\nlimited information available. This work addresses the challenge mentioned\nabove by improving the supervised contrastive learning method leveraging both\nimage-level labels and domain-specific augmentations to enhance model\nrobustness. This approach integrates self-supervised pre-training with a\ntwo-stage supervised contrastive learning strategy. In the first stage, we\nemploy a modified supervised contrastive loss that not only focuses on reducing\nfalse negatives but also introduces an elimination effect to address false\npositives. In the second stage, a relaxing mechanism is introduced that refines\npositive and negative pairs based on similarity, ensuring that only relevant\nimage representations are aligned. We evaluate our method on the BreakHis\ndataset, which consists of breast cancer histopathology images, and demonstrate\nan increase in classification accuracy by 1.45% in the image level, compared to\nthe state-of-the-art method. This improvement corresponds to 93.63% absolute\naccuracy, highlighting the effectiveness of our approach in leveraging\nproperties of data to learn more appropriate representation space.\n", "date": "2024-05-06", "categories": []}
{"id": "2409.15491", "title": "Computational Pathology for Accurate Prediction of Breast Cancer\n  Recurrence: Development and Validation of a Deep Learning-based Tool", "abstract": "  Accurate recurrence risk stratification is crucial for optimizing treatment\nplans for breast cancer patients. Current prognostic tools like Oncotype DX\n(ODX) offer valuable genomic insights for HR+/HER2- patients but are limited by\ncost and accessibility, particularly in underserved populations. In this study,\nwe present Deep-BCR-Auto, a deep learning-based computational pathology\napproach that predicts breast cancer recurrence risk from routine H&E-stained\nwhole slide images (WSIs). Our methodology was validated on two independent\ncohorts: the TCGA-BRCA dataset and an in-house dataset from The Ohio State\nUniversity (OSU). Deep-BCR-Auto demonstrated robust performance in stratifying\npatients into low- and high-recurrence risk categories. On the TCGA-BRCA\ndataset, the model achieved an area under the receiver operating characteristic\ncurve (AUROC) of 0.827, significantly outperforming existing weakly supervised\nmodels (p=0.041). In the independent OSU dataset, Deep-BCR-Auto maintained\nstrong generalizability, achieving an AUROC of 0.832, along with 82.0%\naccuracy, 85.0% specificity, and 67.7% sensitivity. These findings highlight\nthe potential of computational pathology as a cost-effective alternative for\nrecurrence risk assessment, broadening access to personalized treatment\nstrategies. This study underscores the clinical utility of integrating deep\nlearning-based computational pathology into routine pathological assessment for\nbreast cancer prognosis across diverse clinical settings.\n", "date": "2024-09-23", "categories": []}
{"id": "2409.15574", "title": "Clinical-grade Multi-Organ Pathology Report Generation for Multi-scale\n  Whole Slide Images via a Semantically Guided Medical Text Foundation Model", "abstract": "  Vision language models (VLM) have achieved success in both natural language\ncomprehension and image recognition tasks. However, their use in pathology\nreport generation for whole slide images (WSIs) is still limited due to the\nhuge size of multi-scale WSIs and the high cost of WSI annotation. Moreover, in\nmost of the existing research on pathology report generation, sufficient\nvalidation regarding clinical efficacy has not been conducted. Herein, we\npropose a novel Patient-level Multi-organ Pathology Report Generation (PMPRG)\nmodel, which utilizes the multi-scale WSI features from our proposed\nmulti-scale regional vision transformer (MR-ViT) model and their real pathology\nreports to guide VLM training for accurate pathology report generation. The\nmodel then automatically generates a report based on the provided key features\nattended regional features. We assessed our model using a WSI dataset\nconsisting of multiple organs, including the colon and kidney. Our model\nachieved a METEOR score of 0.68, demonstrating the effectiveness of our\napproach. This model allows pathologists to efficiently generate pathology\nreports for patients, regardless of the number of WSIs involved.\n", "date": "2024-09-23", "categories": []}
{"id": "2409.13720", "title": "Efficient Classification of Histopathology Images", "abstract": "  This work addresses how to efficiently classify challenging histopathology\nimages, such as gigapixel whole-slide images for cancer diagnostics with\nimage-level annotation. We use images with annotated tumor regions to identify\na set of tumor patches and a set of benign patches in a cancerous slide. Due to\nthe variable nature of region of interest the tumor positive regions may refer\nto an extreme minority of the pixels. This creates an important problem during\npatch-level classification, where the majority of patches from an image labeled\nas 'cancerous' are actually tumor-free. This problem is different from semantic\nsegmentation which associates a label to every pixel in an image, because after\npatch extraction we are only dealing with patch-level labels.Most existing\napproaches address the data imbalance issue by mitigating the data shortage in\nminority classes in order to prevent the model from being dominated by the\nmajority classes. These methods include data re-sampling, loss re-weighting,\nmargin modification, and data augmentation. In this work, we mitigate the\npatch-level class imbalance problem by taking a divide-and-conquer approach.\nFirst, we partition the data into sub-groups, and define three separate\nclassification sub-problems based on these data partitions. Then, using an\ninformation-theoretic cluster-based sampling of deep image patch features, we\nsample discriminative patches from the sub-groups. Using these sampled patches,\nwe build corresponding deep models to solve the new classification\nsub-problems. Finally, we integrate information learned from the respective\nmodels to make a final decision on the patches. Our result shows that the\nproposed approach can perform competitively using a very low percentage of the\navailable patches in a given whole-slide image.\n", "date": "2024-09-08", "categories": []}
{"id": "2409.13987", "title": "Holistic and Historical Instance Comparison for Cervical Cell Detection", "abstract": "  Cytology screening from Papanicolaou (Pap) smears is a common and effective\ntool for the preventive clinical management of cervical cancer, where abnormal\ncell detection from whole slide images serves as the foundation for reporting\ncervical cytology. However, cervical cell detection remains challenging due to\n1) hazily-defined cell types (e.g., ASC-US) with subtle morphological\ndiscrepancies caused by the dynamic cancerization process, i.e., cell class\nambiguity, and 2) imbalanced class distributions of clinical data may cause\nmissed detection, especially for minor categories, i.e., cell class imbalance.\nTo this end, we propose a holistic and historical instance comparison approach\nfor cervical cell detection. Specifically, we first develop a holistic instance\ncomparison scheme enforcing both RoI-level and class-level cell discrimination.\nThis coarse-to-fine cell comparison encourages the model to learn\nforeground-distinguishable and class-wise representations. To emphatically\nimprove the distinguishability of minor classes, we then introduce a historical\ninstance comparison scheme with a confident sample selection-based memory bank,\nwhich involves comparing current embeddings with historical embeddings for\nbetter cell instance discrimination. Extensive experiments and analysis on two\nlarge-scale cytology datasets including 42,592 and 114,513 cervical cells\ndemonstrate the effectiveness of our method. The code is available at\nhttps://github.com/hjiangaz/HERO.\n", "date": "2024-09-21", "categories": []}
{"id": "2408.14496", "title": "A New Era in Computational Pathology: A Survey on Foundation and\n  Vision-Language Models", "abstract": "  Recent advances in deep learning have completely transformed the domain of\ncomputational pathology (CPath). More specifically, it has altered the\ndiagnostic workflow of pathologists by integrating foundation models (FMs) and\nvision-language models (VLMs) in their assessment and decision-making process.\nThe limitations of existing deep learning approaches in CPath can be overcome\nby FMs through learning a representation space that can be adapted to a wide\nvariety of downstream tasks without explicit supervision. Deploying VLMs allow\npathology reports written in natural language be used as rich semantic\ninformation sources to improve existing models as well as generate predictions\nin natural language form. In this survey, a holistic and systematic overview of\nrecent innovations in FMs and VLMs in CPath is presented. Furthermore, the\ntools, datasets and training schemes for these models are summarized in\naddition to categorizing them into distinct groups. This extensive survey\nhighlights the current trends in CPath and its possible revolution through the\nuse of FMs and VLMs in the future.\n", "date": "2024-08-23", "categories": []}
{"id": "2409.11119", "title": "Multi-Cohort Framework with Cohort-Aware Attention and Adversarial\n  Mutual-Information Minimization for Whole Slide Image Classification", "abstract": "  Whole Slide Images (WSIs) are critical for various clinical applications,\nincluding histopathological analysis. However, current deep learning approaches\nin this field predominantly focus on individual tumor types, limiting model\ngeneralization and scalability. This relatively narrow focus ultimately stems\nfrom the inherent heterogeneity in histopathology and the diverse morphological\nand molecular characteristics of different tumors. To this end, we propose a\nnovel approach for multi-cohort WSI analysis, designed to leverage the\ndiversity of different tumor types. We introduce a Cohort-Aware Attention\nmodule, enabling the capture of both shared and tumor-specific pathological\npatterns, enhancing cross-tumor generalization. Furthermore, we construct an\nadversarial cohort regularization mechanism to minimize cohort-specific biases\nthrough mutual information minimization. Additionally, we develop a\nhierarchical sample balancing strategy to mitigate cohort imbalances and\npromote unbiased learning. Together, these form a cohesive framework for\nunbiased multi-cohort WSI analysis. Extensive experiments on a uniquely\nconstructed multi-cancer dataset demonstrate significant improvements in\ngeneralization, providing a scalable solution for WSI classification across\ndiverse cancer types. Our code for the experiments is publicly available at\n<link>.\n", "date": "2024-09-17", "categories": []}
{"id": "2404.09942", "title": "Knowledge-enhanced Visual-Language Pretraining for Computational\n  Pathology", "abstract": "  In this paper, we consider the problem of visual representation learning for\ncomputational pathology, by exploiting large-scale image-text pairs gathered\nfrom public resources, along with the domain-specific knowledge in pathology.\nSpecifically, we make the following contributions: (i) We curate a pathology\nknowledge tree that consists of 50,470 informative attributes for 4,718\ndiseases requiring pathology diagnosis from 32 human tissues. To our knowledge,\nthis is the first comprehensive structured pathology knowledge base; (ii) We\ndevelop a knowledge-enhanced visual-language pretraining approach, where we\nfirst project pathology-specific knowledge into latent embedding space via a\nlanguage model, and use it to guide the visual representation learning; (iii)\nWe conduct thorough experiments to validate the effectiveness of our proposed\ncomponents, demonstrating significant performance improvement on various\ndownstream tasks, including cross-modal retrieval, zero-shot classification on\npathology patches, and zero-shot tumor subtyping on whole slide images (WSIs).\n", "date": "2024-04-15", "categories": []}
{"id": "2409.09053", "title": "Deep learning-based classification of breast cancer molecular subtypes\n  from H&E whole-slide images", "abstract": "  Classifying breast cancer molecular subtypes is crucial for tailoring\ntreatment strategies. While immunohistochemistry (IHC) and gene expression\nprofiling are standard methods for molecular subtyping, IHC can be subjective,\nand gene profiling is costly and not widely accessible in many regions.\nPrevious approaches have highlighted the potential application of deep learning\nmodels on H&E-stained whole slide images (WSI) for molecular subtyping, but\nthese efforts vary in their methods, datasets, and reported performance. In\nthis work, we investigated whether H&E-stained WSIs could be solely leveraged\nto predict breast cancer molecular subtypes (luminal A, B, HER2-enriched, and\nBasal). We used 1,433 WSIs of breast cancer in a two-step pipeline: first,\nclassifying tumor and non-tumor tiles to use only the tumor regions for\nmolecular subtyping; and second, employing a One-vs-Rest (OvR) strategy to\ntrain four binary OvR classifiers and aggregating their results using an\neXtreme Gradient Boosting (XGBoost) model. The pipeline was tested on 221\nhold-out WSIs, achieving an overall macro F1 score of 0.95 for tumor detection\nand 0.73 for molecular subtyping. Our findings suggest that, with further\nvalidation, supervised deep learning models could serve as supportive tools for\nmolecular subtyping in breast cancer. Our codes are made available to\nfacilitate ongoing research and development.\n", "date": "2024-08-30", "categories": []}
{"id": "2409.09173", "title": "Phikon-v2, A large and public feature extractor for biomarker prediction", "abstract": "  Gathering histopathology slides from over 100 publicly available cohorts, we\ncompile a diverse dataset of 460 million pathology tiles covering more than 30\ncancer sites. Using this dataset, we train a large self-supervised vision\ntransformer using DINOv2 and publicly release one iteration of this model for\nfurther experimentation, coined Phikon-v2. While trained on publicly available\nhistology slides, Phikon-v2 surpasses our previously released model (Phikon)\nand performs on par with other histopathology foundation models (FM) trained on\nproprietary data. Our benchmarks include eight slide-level tasks with results\nreported on external validation cohorts avoiding any data contamination between\npre-training and evaluation datasets. Our downstream training procedure follows\na simple yet robust ensembling strategy yielding a +1.75 AUC increase across\ntasks and models compared to one-shot retraining (p<0.001). We compare Phikon\n(ViT-B) and Phikon-v2 (ViT-L) against 14 different histology feature\nextractors, making our evaluation the most comprehensive to date. Our result\nsupport evidences that DINOv2 handles joint model and data scaling better than\niBOT. Also, we show that recent scaling efforts are overall beneficial to\ndownstream performance in the context of biomarker prediction with GigaPath and\nH-Optimus-0 (two ViT-g with 1.1B parameters each) standing out. However, the\nstatistical margins between the latest top-performing FMs remain mostly\nnon-significant; some even underperform on specific indications or tasks such\nas MSI prediction - deposed by a 13x smaller model developed internally. While\nlatest foundation models may exhibit limitations for clinical deployment, they\nnonetheless offer excellent grounds for the development of more specialized and\ncost-efficient histology encoders fueling AI-guided diagnostic tools.\n", "date": "2024-09-13", "categories": []}
{"id": "2409.08331", "title": "Digital Volumetric Biopsy Cores Improve Gleason Grading of Prostate\n  Cancer Using Deep Learning", "abstract": "  Prostate cancer (PCa) was the most frequently diagnosed cancer among American\nmen in 2023. The histological grading of biopsies is essential for diagnosis,\nand various deep learning-based solutions have been developed to assist with\nthis task. Existing deep learning frameworks are typically applied to\nindividual 2D cross-sections sliced from 3D biopsy tissue specimens. This\nprocess impedes the analysis of complex tissue structures such as glands, which\ncan vary depending on the tissue slice examined. We propose a novel digital\npathology data source called a \"volumetric core,\" obtained via the extraction\nand co-alignment of serially sectioned tissue sections using a novel\nmorphology-preserving alignment framework. We trained an attention-based\nmultiple-instance learning (ABMIL) framework on deep features extracted from\nvolumetric patches to automatically classify the Gleason Grade Group (GGG). To\nhandle volumetric patches, we used a modified video transformer with a deep\nfeature extractor pretrained using self-supervised learning. We ran our\nmorphology-preserving alignment framework to construct 10,210 volumetric cores,\nleaving out 30% for pretraining. The rest of the dataset was used to train\nABMIL, which resulted in a 0.958 macro-average AUC, 0.671 F1 score, 0.661\nprecision, and 0.695 recall averaged across all five GGG significantly\noutperforming the 2D baselines.\n", "date": "2024-09-12", "categories": []}
{"id": "2303.14703", "title": "Exploring the Interplay Between Colorectal Cancer Subtypes Genomic\n  Variants and Cellular Morphology: A Deep-Learning Approach", "abstract": "  Molecular subtypes of colorectal cancer (CRC) significantly influence\ntreatment decisions. While convolutional neural networks (CNNs) have recently\nbeen introduced for automated CRC subtype identification using H&E stained\nhistopathological images, the correlation between CRC subtype genomic variants\nand their corresponding cellular morphology expressed by their imaging\nphenotypes is yet to be fully explored. The goal of this study was to determine\nsuch correlations by incorporating genomic variants in CNN models for CRC\nsubtype classification from H&E images. We utilized the publicly available\nTCGA-CRC-DX dataset, which comprises whole slide images from 360 CRC-diagnosed\npatients (260 for training and 100 for testing). This dataset also provides\ninformation on CRC subtype classifications and genomic variations. We trained\nCNN models for CRC subtype classification that account for potential\ncorrelation between genomic variations within CRC subtypes and their\ncorresponding cellular morphology patterns. We assessed the interplay between\nCRC subtypes' genomic variations and cellular morphology patterns by evaluating\nthe CRC subtype classification accuracy of the different models in a stratified\n5-fold cross-validation experimental setup using the area under the ROC curve\n(AUROC) and average precision (AP) as the performance metrics. Combining the\nCNN models account for variations in CIMP and SNP further improved\nclassification accuracy (AUROC: 0.847$\\pm$0.01 vs. 0.787$\\pm$0.03, p$=$0.01,\nAP: 0.68$\\pm$0.02 vs. 0.64$\\pm$0.05).\n", "date": "2023-03-26", "categories": []}
{"id": "2409.04631", "title": "Zero-Shot Whole Slide Image Retrieval in Histopathology Using Embeddings\n  of Foundation Models", "abstract": "  We have tested recently published foundation models for histopathology for\nimage retrieval. We report macro average of F1 score for top-1 retrieval,\nmajority of top-3 retrievals, and majority of top-5 retrievals. We perform\nzero-shot retrievals, i.e., we do not alter embeddings and we do not train any\nclassifier. As test data, we used diagnostic slides of TCGA, The Cancer Genome\nAtlas, consisting of 23 organs and 117 cancer subtypes. As a search platform we\nused Yottixel that enabled us to perform WSI search using patches. Achieved F1\nscores show low performance, e.g., for top-5 retrievals, 27% +/- 13%\n(Yottixel-DenseNet), 42% +/- 14% (Yottixel-UNI), 40%+/-13% (Yottixel-Virchow),\n41%+/-13% (Yottixel-GigaPath), and 41%+/-14% (GigaPath WSI).\n", "date": "2024-09-06", "categories": []}
{"id": "2409.06455", "title": "Continual Domain Incremental Learning for Privacy-aware Digital\n  Pathology", "abstract": "  In recent years, there has been remarkable progress in the field of digital\npathology, driven by the ability to model complex tissue patterns using\nadvanced deep-learning algorithms. However, the robustness of these models is\noften severely compromised in the presence of data shifts (e.g., different\nstains, organs, centers, etc.). Alternatively, continual learning (CL)\ntechniques aim to reduce the forgetting of past data when learning new data\nwith distributional shift conditions. Specifically, rehearsal-based CL\ntechniques, which store some past data in a buffer and then replay it with new\ndata, have proven effective in medical image analysis tasks. However, privacy\nconcerns arise as these approaches store past data, prompting the development\nof our novel Generative Latent Replay-based CL (GLRCL) approach. GLRCL captures\nthe previous distribution through Gaussian Mixture Models instead of storing\npast samples, which are then utilized to generate features and perform latent\nreplay with new data. We systematically evaluate our proposed framework under\ndifferent shift conditions in histopathology data, including stain and organ\nshift. Our approach significantly outperforms popular buffer-free CL approaches\nand performs similarly to rehearsal-based CL approaches that require large\nbuffers causing serious privacy violations.\n", "date": "2024-09-10", "categories": []}
{"id": "2409.04615", "title": "A Short Survey on Set-Based Aggregation Techniques for Single-Vector WSI\n  Representation in Digital Pathology", "abstract": "  Digital pathology is revolutionizing the field of pathology by enabling the\ndigitization, storage, and analysis of tissue samples as whole slide images\n(WSIs). WSIs are gigapixel files that capture the intricate details of tissue\nsamples, providing a rich source of information for diagnostic and research\npurposes. However, due to their enormous size, representing these images as one\ncompact vector is essential for many computational pathology tasks, such as\nsearch and retrieval, to ensure efficiency and scalability. Most current\nmethods are \"patch-oriented,\" meaning they divide WSIs into smaller patches for\nprocessing, which prevents a holistic analysis of the entire slide.\nAdditionally, the necessity for compact representation is driven by the\nexpensive high-performance storage required for WSIs. Not all hospitals have\naccess to such extensive storage solutions, leading to potential disparities in\nhealthcare quality and accessibility. This paper provides an overview of\nexisting set-based approaches to single-vector WSI representation, highlighting\nthe innovations that allow for more efficient and effective use of these\ncomplex images in digital pathology, thus addressing both computational\nchallenges and storage limitations.\n", "date": "2024-09-06", "categories": []}
{"id": "2409.05697", "title": "Segmentation by Factorization: Unsupervised Semantic Segmentation for\n  Pathology by Factorizing Foundation Model Features", "abstract": "  We introduce Segmentation by Factorization (F-SEG), an unsupervised\nsegmentation method for pathology that generates segmentation masks from\npre-trained deep learning models. F-SEG allows the use of pre-trained deep\nneural networks, including recently developed pathology foundation models, for\nsemantic segmentation. It achieves this without requiring additional training\nor finetuning, by factorizing the spatial features extracted by the models into\nsegmentation masks and their associated concept features. We create generic\ntissue phenotypes for H&E images by training clustering models for multiple\nnumbers of clusters on features extracted from several deep learning models on\nThe Cancer Genome Atlas Program (TCGA), and then show how the clusters can be\nused for factorizing corresponding segmentation masks using off-the-shelf deep\nlearning models. Our results show that F-SEG provides robust unsupervised\nsegmentation capabilities for H&E pathology images, and that the segmentation\nquality is greatly improved by utilizing pathology foundation models. We\ndiscuss and propose methods for evaluating the performance of unsupervised\nsegmentation in pathology.\n", "date": "2024-09-09", "categories": []}
{"id": "2409.02284", "title": "Biochemical Prostate Cancer Recurrence Prediction: Thinking Fast & Slow", "abstract": "  Time to biochemical recurrence in prostate cancer is essential for prognostic\nmonitoring of the progression of patients after prostatectomy, which assesses\nthe efficacy of the surgery. In this work, we proposed to leverage multiple\ninstance learning through a two-stage ``thinking fast \\& slow'' strategy for\nthe time to recurrence (TTR) prediction. The first (``thinking fast'') stage\nfinds the most relevant WSI area for biochemical recurrence and the second\n(``thinking slow'') stage leverages higher resolution patches to predict TTR.\nOur approach reveals a mean C-index ($Ci$) of 0.733 ($\\theta=0.059$) on our\ninternal validation and $Ci=0.603$ on the LEOPARD challenge validation set.\nPost hoc attention visualization shows that the most attentive area contributes\nto the TTR prediction.\n", "date": "2024-09-03", "categories": []}
{"id": "2406.11799", "title": "Mix-Domain Contrastive Learning for Unpaired H&E-to-IHC Stain\n  Translation", "abstract": "  H&E-to-IHC stain translation techniques offer a promising solution for\nprecise cancer diagnosis, especially in low-resource regions where there is a\nshortage of health professionals and limited access to expensive equipment.\nConsidering the pixel-level misalignment of H&E-IHC image pairs, current\nresearch explores the pathological consistency between patches from the same\npositions of the image pair. However, most of them overemphasize the\ncorrespondence between domains or patches, overlooking the side information\nprovided by the non-corresponding objects. In this paper, we propose a\nMix-Domain Contrastive Learning (MDCL) method to leverage the supervision\ninformation in unpaired H&E-to-IHC stain translation. Specifically, the\nproposed MDCL method aggregates the inter-domain and intra-domain pathology\ninformation by estimating the correlation between the anchor patch and all the\npatches from the matching images, encouraging the network to learn additional\ncontrastive knowledge from mixed domains. With the mix-domain pathology\ninformation aggregation, MDCL enhances the pathological consistency between the\ncorresponding patches and the component discrepancy of the patches from the\ndifferent positions of the generated IHC image. Extensive experiments on two\nH&E-to-IHC stain translation datasets, namely MIST and BCI, demonstrate that\nthe proposed method achieves state-of-the-art performance across multiple\nmetrics.\n", "date": "2024-06-17", "categories": []}
{"id": "2409.00649", "title": "DeReStainer: H&E to IHC Pathological Image Translation via Decoupled\n  Staining Channels", "abstract": "  Breast cancer is a highly fatal disease among cancers in women, and early\ndetection is crucial for treatment. HER2 status, a valuable diagnostic marker\nbased on Immunohistochemistry (IHC) staining, is instrumental in determining\nbreast cancer status. The high cost of IHC staining and the ubiquity of\nHematoxylin and Eosin (H&E) staining make the conversion from H&E to IHC\nstaining essential. In this article, we propose a destain-restain framework for\nconverting H&E staining to IHC staining, leveraging the characteristic that H&E\nstaining and IHC staining of the same tissue sections share the Hematoxylin\nchannel. We further design loss functions specifically for Hematoxylin and\nDiaminobenzidin (DAB) channels to generate IHC images exploiting insights from\nseparated staining channels. Beyond the benchmark metrics on BCI contest, we\nhave developed semantic information metrics for the HER2 level. The\nexperimental results demonstrated that our method outperforms previous\nopen-sourced methods in terms of image intrinsic property and semantic\ninformation.\n", "date": "2024-09-01", "categories": []}
{"id": "2408.16008", "title": "Novel Methods for Analyzing Cellular Interactions in Deep Learning-Based\n  Image Cytometry: Spatial Interaction Potential and Co-Localization Index", "abstract": "  The study presents a novel approach for quantifying cellular interactions in\ndigital pathology using deep learning-based image cytometry. Traditional\nmethods struggle with the diversity and heterogeneity of cells within tissues.\nTo address this, we introduce the Spatial Interaction Potential (SIP) and the\nCo-Localization Index (CLI), leveraging deep learning classification\nprobabilities. SIP assesses the potential for cell-to-cell interactions,\nsimilar to an electric field, while CLI incorporates distances between cells,\naccounting for dynamic cell movements. Our approach enhances traditional\nmethods, providing a more sophisticated analysis of cellular interactions. We\nvalidate SIP and CLI through simulations and apply them to colorectal cancer\nspecimens, demonstrating strong correlations with actual biological data. This\ninnovative method offers significant improvements in understanding cellular\ninteractions and has potential applications in various fields of digital\npathology.\n", "date": "2024-08-14", "categories": []}
{"id": "2408.15032", "title": "Mamba2MIL: State Space Duality Based Multiple Instance Learning for\n  Computational Pathology", "abstract": "  Computational pathology (CPath) has significantly advanced the clinical\npractice of pathology. Despite the progress made, Multiple Instance Learning\n(MIL), a promising paradigm within CPath, continues to face challenges,\nparticularly related to incomplete information utilization. Existing\nframeworks, such as those based on Convolutional Neural Networks (CNNs),\nattention, and selective scan space state sequential model (SSM), lack\nsufficient flexibility and scalability in fusing diverse features, and cannot\neffectively fuse diverse features. Additionally, current approaches do not\nadequately exploit order-related and order-independent features, resulting in\nsuboptimal utilization of sequence information. To address these limitations,\nwe propose a novel MIL framework called Mamba2MIL. Our framework utilizes the\nstate space duality model (SSD) to model long sequences of patches of whole\nslide images (WSIs), which, combined with weighted feature selection, supports\nthe fusion processing of more branching features and can be extended according\nto specific application needs. Moreover, we introduce a sequence transformation\nmethod tailored to varying WSI sizes, which enhances sequence-independent\nfeatures while preserving local sequence information, thereby improving\nsequence information utilization. Extensive experiments demonstrate that\nMamba2MIL surpasses state-of-the-art MIL methods. We conducted extensive\nexperiments across multiple datasets, achieving improvements in nearly all\nperformance metrics. Specifically, on the NSCLC dataset, Mamba2MIL achieves a\nbinary tumor classification AUC of 0.9533 and an accuracy of 0.8794. On the\nBRACS dataset, it achieves a multiclass classification AUC of 0.7986 and an\naccuracy of 0.4981. The code is available at\nhttps://github.com/YuqiZhang-Buaa/Mamba2MIL.\n", "date": "2024-08-27", "categories": []}
{"id": "2408.13978", "title": "Histology Virtual Staining with Mask-Guided Adversarial Transfer\n  Learning for Tertiary Lymphoid Structure Detection", "abstract": "  Histological Tertiary Lymphoid Structures (TLSs) are increasingly recognized\nfor their correlation with the efficacy of immunotherapy in various solid\ntumors. Traditionally, the identification and characterization of TLSs rely on\nimmunohistochemistry (IHC) staining techniques, utilizing markers such as CD20\nfor B cells. Despite the specificity of IHC, Hematoxylin-Eosin (H&E) staining\noffers a more accessible and cost-effective choice. Capitalizing on the\nprevalence of H&E staining slides, we introduce a novel Mask-Guided Adversarial\nTransfer Learning method designed for virtual pathological staining. This\nmethod adeptly captures the nuanced color variations across diverse tissue\ntypes under various staining conditions, such as nucleus, red blood cells,\npositive reaction regions, without explicit label information, and adeptly\nsynthesizes realistic IHC-like virtual staining patches, even replicating the\npositive reaction. Further, we propose the Virtual IHC Pathology Analysis\nNetwork (VIPA-Net), an integrated framework encompassing a Mask-Guided Transfer\nModule and an H&E-Based Virtual Staining TLS Detection Module. VIPA-Net\nsynergistically harnesses both H\\&E staining slides and the synthesized virtual\nIHC patches to enhance the detection of TLSs within H&E Whole Slide Images\n(WSIs). We evaluate the network with a comprehensive dataset comprising 1019\nannotated slides from The Cancer Genome Atlas (TCGA). Experimental results\ncompellingly illustrate that the VIPA-Net substantially elevates TLS detection\naccuracy, effectively circumventing the need for actual CD20 staining across\nthe public dataset.\n", "date": "2024-08-26", "categories": []}
{"id": "2408.12825", "title": "MergeUp-augmented Semi-Weakly Supervised Learning for WSI Classification", "abstract": "  Recent advancements in computational pathology and artificial intelligence\nhave significantly improved whole slide image (WSI) classification. However,\nthe gigapixel resolution of WSIs and the scarcity of manual annotations present\nsubstantial challenges. Multiple instance learning (MIL) is a promising weakly\nsupervised learning approach for WSI classification. Recently research revealed\nemploying pseudo bag augmentation can encourage models to learn various data,\nthus bolstering models' performance. While directly inheriting the parents'\nlabels can introduce more noise by mislabeling in training. To address this\nissue, we translate the WSI classification task from weakly supervised learning\nto semi-weakly supervised learning, termed SWS-MIL, where adaptive pseudo bag\naugmentation (AdaPse) is employed to assign labeled and unlabeled data based on\na threshold strategy. Using the \"student-teacher\" pattern, we introduce a\nfeature augmentation technique, MergeUp, which merges bags with low-priority\nbags to enhance inter-category information, increasing training data diversity.\nExperimental results on the CAMELYON-16, BRACS, and TCGA-LUNG datasets\ndemonstrate the superiority of our method over existing state-of-the-art\napproaches, affirming its efficacy in WSI classification.\n", "date": "2024-08-23", "categories": []}
{"id": "2408.00380", "title": "EXAONEPath 1.0 Patch-level Foundation Model for Pathology", "abstract": "  Recent advancements in digital pathology have led to the development of\nnumerous foundational models that utilize self-supervised learning on patches\nextracted from gigapixel whole slide images (WSIs). While this approach\nleverages vast amounts of unlabeled data, we have discovered a significant\nissue: features extracted from these self-supervised models tend to cluster by\nindividual WSIs, a phenomenon we term WSI-specific feature collapse. This\nproblem can potentially limit the model's generalization ability and\nperformance on various downstream tasks. To address this issue, we introduce\nEXAONEPath, a novel foundational model trained on patches that have undergone\nstain normalization. Stain normalization helps reduce color variability arising\nfrom different laboratories and scanners, enabling the model to learn more\nconsistent features. EXAONEPath is trained using 285,153,903 patches extracted\nfrom a total of 34,795 WSIs. Our experiments demonstrate that EXAONEPath\nsignificantly mitigates the feature collapse problem, indicating that the model\nhas learned more generalized features rather than overfitting to individual WSI\ncharacteristics. We compared EXAONEPath with state-of-the-art models across six\ndownstream task datasets, and our results show that EXAONEPath achieves\nsuperior performance relative to the number of WSIs used and the model's\nparameter count. This suggests that the application of stain normalization has\nsubstantially improved the model's efficiency and generalization capabilities.\n", "date": "2024-08-01", "categories": []}
{"id": "2406.05074", "title": "Hibou: A Family of Foundational Vision Transformers for Pathology", "abstract": "  Pathology, the microscopic examination of diseased tissue, is critical for\ndiagnosing various medical conditions, particularly cancers. Traditional\nmethods are labor-intensive and prone to human error. Digital pathology, which\nconverts glass slides into high-resolution digital images for analysis by\ncomputer algorithms, revolutionizes the field by enhancing diagnostic accuracy,\nconsistency, and efficiency through automated image analysis and large-scale\ndata processing. Foundational transformer pretraining is crucial for developing\nrobust, generalizable models as it enables learning from vast amounts of\nunannotated data.\n  This paper introduces the Hibou family of foundational vision transformers\nfor pathology, leveraging the DINOv2 framework to pretrain two model variants,\nHibou-B and Hibou-L, on a proprietary dataset of over 1 million whole slide\nimages (WSIs) representing diverse tissue types and staining techniques. Our\npretrained models demonstrate superior performance on both patch-level and\nslide-level benchmarks, surpassing existing state-of-the-art methods. Notably,\nHibou-L achieves the highest average accuracy across multiple benchmark\ndatasets. To support further research and application in the field, we have\nopen-sourced the Hibou models, which can be accessed at\nhttps://github.com/HistAI/hibou.\n", "date": "2024-06-07", "categories": []}
{"id": "2408.09476", "title": "Advances in Multiple Instance Learning for Whole Slide Image Analysis:\n  Techniques, Challenges, and Future Directions", "abstract": "  Whole slide images (WSIs) are gigapixel-scale digital images of H\\&E-stained\ntissue samples widely used in pathology. The substantial size and complexity of\nWSIs pose unique analytical challenges. Multiple Instance Learning (MIL) has\nemerged as a powerful approach for addressing these challenges, particularly in\ncancer classification and detection. This survey provides a comprehensive\noverview of the challenges and methodologies associated with applying MIL to\nWSI analysis, including attention mechanisms, pseudo-labeling, transformers,\npooling functions, and graph neural networks. Additionally, it explores the\npotential of MIL in discovering cancer cell morphology, constructing\ninterpretable machine learning models, and quantifying cancer grading. By\nsummarizing the current challenges, methodologies, and potential applications\nof MIL in WSI analysis, this survey aims to inform researchers about the state\nof the field and inspire future research directions.\n", "date": "2024-08-18", "categories": []}
{"id": "2408.08527", "title": "Focus on Focus: Focus-oriented Representation Learning and Multi-view\n  Cross-modal Alignment for Glioma Grading", "abstract": "  Recently, multimodal deep learning, which integrates histopathology slides\nand molecular biomarkers, has achieved a promising performance in glioma\ngrading. Despite great progress, due to the intra-modality complexity and\ninter-modality heterogeneity, existing studies suffer from inadequate\nhistopathology representation learning and inefficient molecular-pathology\nknowledge alignment. These two issues hinder existing methods to precisely\ninterpret diagnostic molecular-pathology features, thereby limiting their\ngrading performance. Moreover, the real-world applicability of existing\nmultimodal approaches is significantly restricted as molecular biomarkers are\nnot always available during clinical deployment. To address these problems, we\nintroduce a novel Focus on Focus (FoF) framework with paired pathology-genomic\ntraining and applicable pathology-only inference, enhancing molecular-pathology\nrepresentation effectively. Specifically, we propose a Focus-oriented\nRepresentation Learning (FRL) module to encourage the model to identify regions\npositively or negatively related to glioma grading and guide it to focus on the\ndiagnostic areas with a consistency constraint. To effectively link the\nmolecular biomarkers to morphological features, we propose a Multi-view\nCross-modal Alignment (MCA) module that projects histopathology representations\ninto molecular subspaces, aligning morphological features with corresponding\nmolecular biomarker status by supervised contrastive learning. Experiments on\nthe TCGA GBM-LGG dataset demonstrate that our FoF framework significantly\nimproves the glioma grading. Remarkably, our FoF achieves superior performance\nusing only histopathology slides compared to existing multimodal methods. The\nsource code is available at https://github.com/peterlipan/FoF.\n", "date": "2024-08-16", "categories": []}
{"id": "2408.08847", "title": "HistoGym: A Reinforcement Learning Environment for Histopathological\n  Image Analysis", "abstract": "  In pathological research, education, and clinical practice, the\ndecision-making process based on pathological images is critically important.\nThis significance extends to digital pathology image analysis: its adequacy is\ndemonstrated by the extensive information contained within tissue structures,\nwhich is essential for accurate cancer classification and grading.\nAdditionally, its necessity is highlighted by the inherent requirement for\ninterpretability in the conclusions generated by algorithms. For humans,\ndetermining tumor type and grade typically involves multi-scale analysis, which\npresents a significant challenge for AI algorithms. Traditional patch-based\nmethods are inadequate for modeling such complex structures, as they fail to\ncapture the intricate, multi-scale information inherent in whole slide images.\nConsequently, there is a pressing need for advanced AI techniques capable of\nefficiently and accurately replicating this complex analytical process. To\naddress this issue, we introduce HistoGym, an open-source reinforcement\nlearning environment for histopathological image analysis. Following OpenAI Gym\nAPIs, HistoGym aims to foster whole slide image diagnosis by mimicking the\nreal-life processes of doctors. Leveraging the pyramid feature of WSIs and the\nOpenSlide API, HistoGym provides a unified framework for various clinical\ntasks, including tumor detection and classification. We detail the observation,\naction, and reward specifications tailored for the histopathological image\nanalysis domain and provide an open-source Python-based interface for both\nclinicians and researchers. To accommodate different clinical demands, we offer\nvarious scenarios for different organs and cancers, including both WSI-based\nand selected region-based scenarios, showcasing several noteworthy results.\n", "date": "2024-08-16", "categories": []}
{"id": "2408.07860", "title": "A Novel Generative Artificial Intelligence Method for Interference Study\n  on Multiplex Brightfield Immunohistochemistry Images", "abstract": "  Multiplex brightfield imaging offers the advantage of simultaneously\nanalyzing multiple biomarkers on a single slide, as opposed to single biomarker\nlabeling on multiple consecutive slides. To accurately analyze multiple\nbiomarkers localized at the same cellular compartment, two representative\nbiomarker sets were selected as assay models - cMET-PDL1-EGFR and\nCD8-LAG3-PDL1, where all three biomarkers can co-localize on the cell membrane.\nOne of the most crucial preliminary stages for analyzing such assay is\nidentifying each unique chromogen on individual cells. This is a challenging\nproblem due to the co-localization of membrane stains from all the three\nbiomarkers. It requires advanced color unmixing for creating the equivalent\nsingleplex images from each triplex image for each biomarker.\n  In this project, we developed a cycle-Generative Adversarial Network\n(cycle-GAN) method for unmixing the triplex images generated from the\nabove-mentioned assays. Three different models were designed to generate the\nsingleplex image for each of the three stains Tamra (purple), QM-Dabsyl\n(yellow) and Green. A notable novelty of our approach was that the input to the\nnetwork were images in the optical density domain instead of conventionally\nused RGB images. The use of the optical density domain helped in reducing the\nblurriness of the synthetic singleplex images, which was often observed when\nthe network was trained on RGB images.\n  The cycle-GAN models were validated on 10,800 lung, gastric and colon images\nfor the cMET-PDL1-EGFR assay and 3600 colon images for the CD8-LAG3-PDL1 assay.\nVisual as well as quantified assessments demonstrated that the proposed method\nis effective and efficient when compared with the manual reviewing results and\nis readily applicable to various multiplex assays.\n", "date": "2024-08-15", "categories": []}
{"id": "2308.10994", "title": "Switched auxiliary loss for robust training of transformer models for\n  histopathological image segmentation", "abstract": "  Functional tissue Units (FTUs) are cell population neighborhoods local to a\nparticular organ performing its main function.The FTUs provide crucial\ninformation to the pathologist in understanding the disease affecting a\nparticular organ by providing information at the cellular level.In our\nresearch, we have developed a model to segment multi-organ FTUs across 5 organs\nnamely: the kidney, large intestine, lung, prostate and spleen by utilizing the\n'HuBMAP + HPA - Hacking the Human Body' competition dataset.We propose adding\nswitched auxiliary loss for training models like the transformers to overcome\nthe diminishing gradient problem which poses a challenge towards optimal\ntraining of deep models.Overall, our model achieved a dice score of 0.793 on\nthe public dataset and 0.778 on the private dataset.The results supports the\nrobustness of the proposed training methodology.The findings also bolster the\nuse of transformers models for dense prediction tasks in the field of medical\nimage analysis.The study assists in understanding the relationships between\ncell and tissue organization thereby providing a useful medium to look at the\nimpact of cellular functions on human health.\n", "date": "2023-08-21", "categories": []}
{"id": "2408.01797", "title": "NuLite -- Lightweight and Fast Model for Nuclei Instance Segmentation\n  and Classification", "abstract": "  In pathology, accurate and efficient analysis of Hematoxylin and Eosin (H\\&E)\nslides is crucial for timely and effective cancer diagnosis. Although many deep\nlearning solutions for nuclei instance segmentation and classification exist in\nthe literature, they often entail high computational costs and resource\nrequirements, thus limiting their practical usage in medical applications. To\naddress this issue, we introduce a novel convolutional neural network, NuLite,\na U-Net-like architecture designed explicitly on Fast-ViT, a state-of-the-art\n(SOTA) lightweight CNN. We obtained three versions of our model, NuLite-S,\nNuLite-M, and NuLite-H, trained on the PanNuke dataset. The experimental\nresults prove that our models equal CellViT (SOTA) in terms of panoptic quality\nand detection. However, our lightest model, NuLite-S, is 40 times smaller in\nterms of parameters and about 8 times smaller in terms of GFlops, while our\nheaviest model is 17 times smaller in terms of parameters and about 7 times\nsmaller in terms of GFlops. Moreover, our model is up to about 8 times faster\nthan CellViT. Lastly, to prove the effectiveness of our solution, we provide a\nrobust comparison of external datasets, namely CoNseP, MoNuSeg, and GlySAC. Our\nmodel is publicly available at https://github.com/CosmoIknosLab/NuLite\n", "date": "2024-08-03", "categories": []}
{"id": "2408.03954", "title": "Histopathology image embedding based on foundation models features\n  aggregation for patient treatment response prediction", "abstract": "  Predicting the response of a patient to a cancer treatment is of high\ninterest. Nonetheless, this task is still challenging from a medical point of\nview due to the complexity of the interaction between the patient organism and\nthe considered treatment. Recent works on foundation models pre-trained with\nself-supervised learning on large-scale unlabeled histopathology datasets have\nopened a new direction towards the development of new methods for cancer\ndiagnosis related tasks. In this article, we propose a novel methodology for\npredicting Diffuse Large B-Cell Lymphoma patients treatment response from Whole\nSlide Images. Our method exploits several foundation models as feature\nextractors to obtain a local representation of the image corresponding to a\nsmall region of the tissue, then, a global representation of the image is\nobtained by aggregating these local representations using attention-based\nMultiple Instance Learning. Our experimental study conducted on a dataset of\n152 patients, shows the promising results of our methodology, notably by\nhighlighting the advantage of using foundation models compared to conventional\nImageNet pre-training. Moreover, the obtained results clearly demonstrates the\npotential of foundation models for characterizing histopathology images and\ngenerating more suited semantic representation for this task.\n", "date": "2024-07-23", "categories": []}
{"id": "2408.02859", "title": "Multistain Pretraining for Slide Representation Learning in Pathology", "abstract": "  Developing self-supervised learning (SSL) models that can learn universal and\ntransferable representations of H&E gigapixel whole-slide images (WSIs) is\nbecoming increasingly valuable in computational pathology. These models hold\nthe potential to advance critical tasks such as few-shot classification, slide\nretrieval, and patient stratification. Existing approaches for slide\nrepresentation learning extend the principles of SSL from small images (e.g.,\n224 x 224 patches) to entire slides, usually by aligning two different\naugmentations (or views) of the slide. Yet the resulting representation remains\nconstrained by the limited clinical and biological diversity of the views.\nInstead, we postulate that slides stained with multiple markers, such as\nimmunohistochemistry, can be used as different views to form a rich\ntask-agnostic training signal. To this end, we introduce Madeleine, a\nmultimodal pretraining strategy for slide representation learning. Madeleine is\ntrained with a dual global-local cross-stain alignment objective on large\ncohorts of breast cancer samples (N=4,211 WSIs across five stains) and kidney\ntransplant samples (N=12,070 WSIs across four stains). We demonstrate the\nquality of slide representations learned by Madeleine on various downstream\nevaluations, ranging from morphological and molecular classification to\nprognostic prediction, comprising 21 tasks using 7,299 WSIs from multiple\nmedical centers. Code is available at https://github.com/mahmoodlab/MADELEINE.\n", "date": "2024-08-05", "categories": []}
{"id": "2311.17804", "title": "The Importance of Downstream Networks in Digital Pathology Foundation\n  Models", "abstract": "  Digital pathology has significantly advanced disease detection and\npathologist efficiency through the analysis of gigapixel whole-slide images\n(WSI). In this process, WSIs are first divided into patches, for which a\nfeature extractor model is applied to obtain feature vectors, which are\nsubsequently processed by an aggregation model to predict the respective WSI\nlabel. With the rapid evolution of representation learning, numerous new\nfeature extractor models, often termed foundational models, have emerged.\nTraditional evaluation methods rely on a static downstream aggregation model\nsetup, encompassing a fixed architecture and hyperparameters, a practice we\nidentify as potentially biasing the results. Our study uncovers a sensitivity\nof feature extractor models towards aggregation model configurations,\nindicating that performance comparability can be skewed based on the chosen\nconfigurations. By accounting for this sensitivity, we find that the\nperformance of many current feature extractor models is notably similar. We\nsupport this insight by evaluating seven feature extractor models across three\ndifferent datasets with 162 different aggregation model configurations. This\ncomprehensive approach provides a more nuanced understanding of the feature\nextractors' sensitivity to various aggregation model configurations, leading to\na fairer and more accurate assessment of new foundation models in digital\npathology.\n", "date": "2023-11-29", "categories": []}
{"id": "2312.02111", "title": "TriDeNT: Triple Deep Network Training for Privileged Knowledge\n  Distillation in Histopathology", "abstract": "  Computational pathology models rarely utilise data that will not be available\nfor inference. This means most models cannot learn from highly informative data\nsuch as additional immunohistochemical (IHC) stains and spatial\ntranscriptomics. We present TriDeNT, a novel self-supervised method for\nutilising privileged data that is not available during inference to improve\nperformance. We demonstrate the efficacy of this method for a range of\ndifferent paired data including immunohistochemistry, spatial transcriptomics\nand expert nuclei annotations. In all settings, TriDeNT outperforms other\nstate-of-the-art methods in downstream tasks, with observed improvements of up\nto 101%. Furthermore, we provide qualitative and quantitative measurements of\nthe features learned by these models and how they differ from baselines.\nTriDeNT offers a novel method to distil knowledge from scarce or costly data\nduring training, to create significantly better models for routine inputs.\n", "date": "2023-12-04", "categories": []}
{"id": "2407.20518", "title": "High-Resolution Spatial Transcriptomics from Histology Images using\n  HisToSGE", "abstract": "  Spatial transcriptomics (ST) is a groundbreaking genomic technology that\nenables spatial localization analysis of gene expression within tissue\nsections. However, it is significantly limited by high costs and sparse spatial\nresolution. An alternative, more cost-effective strategy is to use deep\nlearning methods to predict high-density gene expression profiles from\nhistological images. However, existing methods struggle to capture rich image\nfeatures effectively or rely on low-dimensional positional coordinates, making\nit difficult to accurately predict high-resolution gene expression profiles. To\naddress these limitations, we developed HisToSGE, a method that employs a\nPathology Image Large Model (PILM) to extract rich image features from\nhistological images and utilizes a feature learning module to robustly generate\nhigh-resolution gene expression profiles. We evaluated HisToSGE on four ST\ndatasets, comparing its performance with five state-of-the-art baseline\nmethods. The results demonstrate that HisToSGE excels in generating\nhigh-resolution gene expression profiles and performing downstream tasks such\nas spatial domain identification. All code and public datasets used in this\npaper are available at https://github.com/wenwenmin/HisToSGE and\nhttps://zenodo.org/records/12792163.\n", "date": "2024-07-30", "categories": []}
{"id": "2407.20596", "title": "Benchmarking Histopathology Foundation Models for Ovarian Cancer\n  Bevacizumab Treatment Response Prediction from Whole Slide Images", "abstract": "  Bevacizumab is a widely studied targeted therapeutic drug used in conjunction\nwith standard chemotherapy for the treatment of recurrent ovarian cancer. While\nits administration has shown to increase the progression-free survival (PFS) in\npatients with advanced stage ovarian cancer, the lack of identifiable\nbiomarkers for predicting patient response has been a major roadblock in its\neffective adoption towards personalized medicine. In this work, we leverage the\nlatest histopathology foundation models trained on large-scale whole slide\nimage (WSI) datasets to extract ovarian tumor tissue features for predicting\nbevacizumab response from WSIs. Our extensive experiments across a combination\nof different histopathology foundation models and multiple instance learning\n(MIL) strategies demonstrate capability of these large models in predicting\nbevacizumab response in ovarian cancer patients with the best models achieving\nan AUC score of 0.86 and an accuracy score of 72.5%. Furthermore, our survival\nmodels are able to stratify high- and low-risk cases with statistical\nsignificance (p < 0.05) even among the patients with the aggressive subtype of\nhigh-grade serous ovarian carcinoma. This work highlights the utility of\nhistopathology foundation models for the task of ovarian bevacizumab response\nprediction from WSIs. The high-attention regions of the WSIs highlighted by\nthese models not only aid the model explainability but also serve as promising\nimaging biomarkers for treatment prognosis.\n", "date": "2024-07-30", "categories": []}
{"id": "2407.20643", "title": "Generalizing AI-driven Assessment of Immunohistochemistry across\n  Immunostains and Cancer Types: A Universal Immunohistochemistry Analyzer", "abstract": "  Despite advancements in methodologies, immunohistochemistry (IHC) remains the\nmost utilized ancillary test for histopathologic and companion diagnostics in\ntargeted therapies. However, objective IHC assessment poses challenges.\nArtificial intelligence (AI) has emerged as a potential solution, yet its\ndevelopment requires extensive training for each cancer and IHC type, limiting\nversatility. We developed a Universal IHC (UIHC) analyzer, an AI model for\ninterpreting IHC images regardless of tumor or IHC types, using training\ndatasets from various cancers stained for PD-L1 and/or HER2. This multi-cohort\ntrained model outperforms conventional single-cohort models in interpreting\nunseen IHCs (Kappa score 0.578 vs. up to 0.509) and consistently shows superior\nperformance across different positive staining cutoff values. Qualitative\nanalysis reveals that UIHC effectively clusters patches based on expression\nlevels. The UIHC model also quantitatively assesses c-MET expression with MET\nmutations, representing a significant advancement in AI application in the era\nof personalized medicine and accumulating novel biomarkers.\n", "date": "2024-07-30", "categories": []}
{"id": "2407.03655", "title": "Pathological Semantics-Preserving Learning for H&E-to-IHC Virtual\n  Staining", "abstract": "  Conventional hematoxylin-eosin (H&E) staining is limited to revealing cell\nmorphology and distribution, whereas immunohistochemical (IHC) staining\nprovides precise and specific visualization of protein activation at the\nmolecular level. Virtual staining technology has emerged as a solution for\nhighly efficient IHC examination, which directly transforms H&E-stained images\nto IHC-stained images. However, virtual staining is challenged by the\ninsufficient mining of pathological semantics and the spatial misalignment of\npathological semantics. To address these issues, we propose the Pathological\nSemantics-Preserving Learning method for Virtual Staining (PSPStain), which\ndirectly incorporates the molecular-level semantic information and enhances\nsemantics interaction despite any spatial inconsistency. Specifically, PSPStain\ncomprises two novel learning strategies: 1) Protein-Aware Learning Strategy\n(PALS) with Focal Optical Density (FOD) map maintains the coherence of protein\nexpression level, which represents molecular-level semantic information; 2)\nPrototype-Consistent Learning Strategy (PCLS), which enhances cross-image\nsemantic interaction by prototypical consistency learning. We evaluate PSPStain\non two public datasets using five metrics: three clinically relevant metrics\nand two for image quality. Extensive experiments indicate that PSPStain\noutperforms current state-of-the-art H&E-to-IHC virtual staining methods and\ndemonstrates a high pathological correlation between the staging of real and\nvirtual stains.\n", "date": "2024-07-04", "categories": []}
{"id": "2407.19113", "title": "VIMs: Virtual Immunohistochemistry Multiplex staining via Text-to-Stain\n  Diffusion Trained on Uniplex Stains", "abstract": "  This paper introduces a Virtual Immunohistochemistry Multiplex staining\n(VIMs) model designed to generate multiple immunohistochemistry (IHC) stains\nfrom a single hematoxylin and eosin (H&E) stained tissue section. IHC stains\nare crucial in pathology practice for resolving complex diagnostic questions\nand guiding patient treatment decisions. While commercial laboratories offer a\nwide array of up to 400 different antibody-based IHC stains, small biopsies\noften lack sufficient tissue for multiple stains while preserving material for\nsubsequent molecular testing. This highlights the need for virtual IHC\nstaining. Notably, VIMs is the first model to address this need, leveraging a\nlarge vision-language single-step diffusion model for virtual IHC multiplexing\nthrough text prompts for each IHC marker. VIMs is trained on uniplex paired H&E\nand IHC images, employing an adversarial training module. Testing of VIMs\nincludes both paired and unpaired image sets. To enhance computational\nefficiency, VIMs utilizes a pre-trained large latent diffusion model fine-tuned\nwith small, trainable weights through the Low-Rank Adapter (LoRA) approach.\nExperiments on nuclear and cytoplasmic IHC markers demonstrate that VIMs\noutperforms the base diffusion model and achieves performance comparable to\nPix2Pix, a standard generative model for paired image translation. Multiple\nevaluation methods, including assessments by two pathologists, are used to\ndetermine the performance of VIMs. Additionally, experiments with different\nprompts highlight the impact of text conditioning. This paper represents the\nfirst attempt to accelerate histopathology research by demonstrating the\ngeneration of multiple IHC stains from a single H&E input using a single model\ntrained solely on uniplex data.\n", "date": "2024-07-26", "categories": []}
{"id": "2402.17228", "title": "Feature Re-Embedding: Towards Foundation Model-Level Performance in\n  Computational Pathology", "abstract": "  Multiple instance learning (MIL) is the most widely used framework in\ncomputational pathology, encompassing sub-typing, diagnosis, prognosis, and\nmore. However, the existing MIL paradigm typically requires an offline instance\nfeature extractor, such as a pre-trained ResNet or a foundation model. This\napproach lacks the capability for feature fine-tuning within the specific\ndownstream tasks, limiting its adaptability and performance. To address this\nissue, we propose a Re-embedded Regional Transformer (R$^2$T) for re-embedding\nthe instance features online, which captures fine-grained local features and\nestablishes connections across different regions. Unlike existing works that\nfocus on pre-training powerful feature extractor or designing sophisticated\ninstance aggregator, R$^2$T is tailored to re-embed instance features online.\nIt serves as a portable module that can seamlessly integrate into mainstream\nMIL models. Extensive experimental results on common computational pathology\ntasks validate that: 1) feature re-embedding improves the performance of MIL\nmodels based on ResNet-50 features to the level of foundation model features,\nand further enhances the performance of foundation model features; 2) the\nR$^2$T can introduce more significant performance improvements to various MIL\nmodels; 3) R$^2$T-MIL, as an R$^2$T-enhanced AB-MIL, outperforms other latest\nmethods by a large margin.The code is available at:\nhttps://github.com/DearCaat/RRT-MIL.\n", "date": "2024-02-27", "categories": []}
{"id": "2407.17857", "title": "Mew: Multiplexed Immunofluorescence Image Analysis through an Efficient\n  Multiplex Network", "abstract": "  Recent advancements in graph-based approaches for multiplexed\nimmunofluorescence (mIF) images have significantly propelled the field forward,\noffering deeper insights into patient-level phenotyping. However, current\ngraph-based methodologies encounter two primary challenges: (1) Cellular\nHeterogeneity, where existing approaches fail to adequately address the\ninductive biases inherent in graphs, particularly the homophily characteristic\nobserved in cellular connectivity and; (2) Scalability, where handling cellular\ngraphs from high-dimensional images faces difficulties in managing a high\nnumber of cells. To overcome these limitations, we introduce Mew, a novel\nframework designed to efficiently process mIF images through the lens of\nmultiplex network. Mew innovatively constructs a multiplex network comprising\ntwo distinct layers: a Voronoi network for geometric information and a\nCell-type network for capturing cell-wise homogeneity. This framework equips a\nscalable and efficient Graph Neural Network (GNN), capable of processing the\nentire graph during training. Furthermore, Mew integrates an interpretable\nattention module that autonomously identifies relevant layers for image\nclassification. Extensive experiments on a real-world patient dataset from\nvarious institutions highlight Mew's remarkable efficacy and efficiency,\nmarking a significant advancement in mIF image analysis. The source code of Mew\ncan be found here: \\url{https://github.com/UNITES-Lab/Mew}\n", "date": "2024-07-25", "categories": []}
{"id": "2407.18054", "title": "LKCell: Efficient Cell Nuclei Instance Segmentation with Large\n  Convolution Kernels", "abstract": "  The segmentation of cell nuclei in tissue images stained with the blood dye\nhematoxylin and eosin (H$\\&$E) is essential for various clinical applications\nand analyses. Due to the complex characteristics of cellular morphology, a\nlarge receptive field is considered crucial for generating high-quality\nsegmentation. However, previous methods face challenges in achieving a balance\nbetween the receptive field and computational burden. To address this issue, we\npropose LKCell, a high-accuracy and efficient cell segmentation method. Its\ncore insight lies in unleashing the potential of large convolution kernels to\nachieve computationally efficient large receptive fields. Specifically, (1) We\ntransfer pre-trained large convolution kernel models to the medical domain for\nthe first time, demonstrating their effectiveness in cell segmentation. (2) We\nanalyze the redundancy of previous methods and design a new segmentation\ndecoder based on large convolution kernels. It achieves higher performance\nwhile significantly reducing the number of parameters. We evaluate our method\non the most challenging benchmark and achieve state-of-the-art results (0.5080\nmPQ) in cell nuclei instance segmentation with only 21.6% FLOPs compared with\nthe previous leading method. Our source code and models are available at\nhttps://github.com/hustvl/LKCell.\n", "date": "2024-07-25", "categories": []}
{"id": "2407.17267", "title": "M4: Multi-Proxy Multi-Gate Mixture of Experts Network for Multiple\n  Instance Learning in Histopathology Image Analysis", "abstract": "  Multiple instance learning (MIL) has been successfully applied for whole\nslide images (WSIs) analysis in computational pathology, enabling a wide range\nof prediction tasks from tumor subtyping to inferring genetic mutations and\nmulti-omics biomarkers. However, existing MIL methods predominantly focus on\nsingle-task learning, resulting in not only overall low efficiency but also the\noverlook of inter-task relatedness. To address these issues, we proposed an\nadapted architecture of Multi-gate Mixture-of-experts with Multi-proxy for\nMultiple instance learning (M4), and applied this framework for simultaneous\nprediction of multiple genetic mutations from WSIs. The proposed M4 model has\ntwo main innovations: (1) utilizing a mixture of experts with multiple gating\nstrategies for multi-genetic mutation prediction on a single pathological\nslide; (2) constructing multi-proxy expert network and gate network for\ncomprehensive and effective modeling of pathological image information. Our\nmodel achieved significant improvements across five tested TCGA datasets in\ncomparison to current state-of-the-art single-task methods. The code is\navailable at:https://github.com/Bigyehahaha/M4.\n", "date": "2024-07-24", "categories": []}
{"id": "2407.16464", "title": "Lymphoid Infiltration Assessment of the Tumor Margins in H&E Slides", "abstract": "  Lymphoid infiltration at tumor margins is a key prognostic marker in solid\ntumors, playing a crucial role in guiding immunotherapy decisions. Current\nassessment methods, heavily reliant on immunohistochemistry (IHC), face\nchallenges in tumor margin delineation and are affected by tissue preservation\nconditions. In contrast, we propose a Hematoxylin and Eosin (H&E)\nstaining-based approach, underpinned by an advanced lymphocyte segmentation\nmodel trained on a public dataset for the precise detection of CD3+ and CD20+\nlymphocytes. In our colorectal cancer study, we demonstrate that our H&E-based\nmethod offers a compelling alternative to traditional IHC, achieving comparable\nresults in many cases. Our method's validity is further explored through a\nTuring test, involving blinded assessments by a pathologist of anonymized\ncurves from H&E and IHC slides. This approach invites the medical community to\nconsider Turing tests as a standard for evaluating medical applications\ninvolving expert human evaluation, thereby opening new avenues for enhancing\ncancer management and immunotherapy planning.\n", "date": "2024-07-23", "categories": []}
{"id": "2407.14628", "title": "Advancing Melanoma Diagnosis with Self-Supervised Neural Networks:\n  Evaluating the Effectiveness of Different Techniques", "abstract": "  We investigate the potential of self-supervision in improving the accuracy of\ndeep learning models trained to classify melanoma patches. Various\nself-supervision techniques such as rotation prediction, missing patch\nprediction, and corruption removal were implemented and assessed for their\nimpact on the convolutional neural network's performance. Preliminary results\nsuggest a positive influence of self-supervision methods on the model's\naccuracy. The study notably demonstrates the efficacy of the corruption removal\nmethod in enhancing model performance. Despite observable improvements, we\nconclude that the self-supervised models have considerable potential for\nfurther enhancement, achievable through training over more epochs or expanding\nthe dataset. We suggest exploring other self-supervision methods like Bootstrap\nYour Own Latent (BYOL) and contrastive learning in future research, emphasizing\nthe cost-benefit trade-off due to their resource-intensive nature. The findings\nunderline the promise of self-supervision in augmenting melanoma detection\ncapabilities of deep learning models.\n", "date": "2024-07-19", "categories": []}
{"id": "2407.15816", "title": "Efficient and generalizable prediction of molecular alterations in\n  multiple cancer cohorts using H&E whole slide images", "abstract": "  Molecular testing of tumor samples for targetable biomarkers is restricted by\na lack of standardization, turnaround-time, cost, and tissue availability\nacross cancer types. Additionally, targetable alterations of low prevalence may\nnot be tested in routine workflows. Algorithms that predict DNA alterations\nfrom routinely generated hematoxylin and eosin (H&E)-stained images could\nprioritize samples for confirmatory molecular testing. Costs and the necessity\nof a large number of samples containing mutations limit approaches that train\nindividual algorithms for each alteration. In this work, models were trained\nfor simultaneous prediction of multiple DNA alterations from H&E images using a\nmulti-task approach. Compared to biomarker-specific models, this approach\nperformed better on average, with pronounced gains for rare mutations. The\nmodels reasonably generalized to independent temporal-holdout,\nexternally-stained, and multi-site TCGA test sets. Additionally, whole slide\nimage embeddings derived using multi-task models demonstrated strong\nperformance in downstream tasks that were not a part of training. Overall, this\nis a promising approach to develop clinically useful algorithms that provide\nmultiple actionable predictions from a single slide.\n", "date": "2024-07-22", "categories": []}
{"id": "2407.08968", "title": "SlideGCD: Slide-based Graph Collaborative Training with Knowledge\n  Distillation for Whole Slide Image Classification", "abstract": "  Existing WSI analysis methods lie on the consensus that histopathological\ncharacteristics of tumors are significant guidance for cancer diagnostics.\nParticularly, as the evolution of cancers is a continuous process, the\ncorrelations and differences across various stages, anatomical locations and\npatients should be taken into account. However, recent research mainly focuses\non the inner-contextual information in a single WSI, ignoring the correlations\nbetween slides. To verify whether introducing the slide inter-correlations can\nbring improvements to WSI representation learning, we propose a generic WSI\nanalysis pipeline SlideGCD that considers the existing multi-instance learning\n(MIL) methods as the backbone and forge the WSI classification task as a node\nclassification problem. More specifically, SlideGCD declares a node buffer that\nstores previous slide embeddings for subsequent extensive slide-based graph\nconstruction and conducts graph learning to explore the inter-correlations\nimplied in the slide-based graph. Moreover, we frame the MIL classifier and\ngraph learning into two parallel workflows and deploy the knowledge\ndistillation to transfer the differentiable information to the graph neural\nnetwork. The consistent performance boosting, brought by SlideGCD, of four\nprevious state-of-the-art MIL methods is observed on two TCGA benchmark\ndatasets. The code is available at https://github.com/HFUT-miaLab/SlideGCD.\n", "date": "2024-07-12", "categories": []}
{"id": "2407.13277", "title": "URCDM: Ultra-Resolution Image Synthesis in Histopathology", "abstract": "  Diagnosing medical conditions from histopathology data requires a thorough\nanalysis across the various resolutions of Whole Slide Images (WSI). However,\nexisting generative methods fail to consistently represent the hierarchical\nstructure of WSIs due to a focus on high-fidelity patches. To tackle this, we\npropose Ultra-Resolution Cascaded Diffusion Models (URCDMs) which are capable\nof synthesising entire histopathology images at high resolutions whilst\nauthentically capturing the details of both the underlying anatomy and\npathology at all magnification levels. We evaluate our method on three separate\ndatasets, consisting of brain, breast and kidney tissue, and surpass existing\nstate-of-the-art multi-resolution models. Furthermore, an expert evaluation\nstudy was conducted, demonstrating that URCDMs consistently generate outputs\nacross various resolutions that trained evaluators cannot distinguish from real\nimages. All code and additional examples can be found on GitHub.\n", "date": "2024-07-18", "categories": []}
{"id": "2407.13708", "title": "Are We Ready for Out-of-Distribution Detection in Digital Pathology?", "abstract": "  The detection of semantic and covariate out-of-distribution (OOD) examples is\na critical yet overlooked challenge in digital pathology (DP). Recently,\nsubstantial insight and methods on OOD detection were presented by the ML\ncommunity, but how do they fare in DP applications? To this end, we establish a\nbenchmark study, our highlights being: 1) the adoption of proper evaluation\nprotocols, 2) the comparison of diverse detectors in both a single and\nmulti-model setting, and 3) the exploration into advanced ML settings like\ntransfer learning (ImageNet vs. DP pre-training) and choice of architecture\n(CNNs vs. transformers). Through our comprehensive experiments, we contribute\nnew insights and guidelines, paving the way for future research and discussion.\n", "date": "2024-07-18", "categories": []}
{"id": "2404.00837", "title": "Automated HER2 Scoring in Breast Cancer Images Using Deep Learning and\n  Pyramid Sampling", "abstract": "  Human epidermal growth factor receptor 2 (HER2) is a critical protein in\ncancer cell growth that signifies the aggressiveness of breast cancer (BC) and\nhelps predict its prognosis. Accurate assessment of immunohistochemically (IHC)\nstained tissue slides for HER2 expression levels is essential for both\ntreatment guidance and understanding of cancer mechanisms. Nevertheless, the\ntraditional workflow of manual examination by board-certified pathologists\nencounters challenges, including inter- and intra-observer inconsistency and\nextended turnaround times. Here, we introduce a deep learning-based approach\nutilizing pyramid sampling for the automated classification of HER2 status in\nIHC-stained BC tissue images. Our approach analyzes morphological features at\nvarious spatial scales, efficiently managing the computational load and\nfacilitating a detailed examination of cellular and larger-scale tissue-level\ndetails. This method addresses the tissue heterogeneity of HER2 expression by\nproviding a comprehensive view, leading to a blind testing classification\naccuracy of 84.70%, on a dataset of 523 core images from tissue microarrays.\nOur automated system, proving reliable as an adjunct pathology tool, has the\npotential to enhance diagnostic precision and evaluation speed, and might\nsignificantly impact cancer treatment planning.\n", "date": "2024-04-01", "categories": []}
{"id": "2407.11486", "title": "An efficient framework based on large foundation model for cervical\n  cytopathology whole slide image screening", "abstract": "  Current cervical cytopathology whole slide image (WSI) screening primarily\nrelies on detection-based approaches, which are limited in performance due to\nthe expense and time-consuming annotation process. Multiple Instance Learning\n(MIL), a weakly supervised approach that relies solely on bag-level labels, can\neffectively alleviate these challenges. Nonetheless, MIL commonly employs\nfrozen pretrained models or self-supervised learning for feature extraction,\nwhich suffers from low efficacy or inefficiency. In this paper, we propose an\nefficient framework for cervical cytopathology WSI classification using only\nWSI-level labels through unsupervised and weakly supervised learning. Given the\nsparse and dispersed nature of abnormal cells within cytopathological WSIs, we\npropose a strategy that leverages the pretrained foundation model to filter the\ntop$k$ high-risk patches. Subsequently, we suggest parameter-efficient\nfine-tuning (PEFT) of a large foundation model using contrastive learning on\nthe filtered patches to enhance its representation ability for task-specific\nsignals. By training only the added linear adapters, we enhance the learning of\npatch-level features with substantially reduced time and memory consumption.\nExperiments conducted on the CSD and FNAC 2019 datasets demonstrate that the\nproposed method enhances the performance of various MIL methods and achieves\nstate-of-the-art (SOTA) performance. The code and trained models are publicly\navailable at https://github.com/CVIU-CSU/TCT-InfoNCE.\n", "date": "2024-07-16", "categories": []}
{"id": "2403.16497", "title": "PathoTune: Adapting Visual Foundation Model to Pathological Specialists", "abstract": "  As natural image understanding moves towards the pretrain-finetune era,\nresearch in pathology imaging is concurrently evolving. Despite the predominant\nfocus on pretraining pathological foundation models, how to adapt foundation\nmodels to downstream tasks is little explored. For downstream adaptation, we\npropose the existence of two domain gaps, i.e., the Foundation-Task Gap and the\nTask-Instance Gap. To mitigate these gaps, we introduce PathoTune, a framework\ndesigned to efficiently adapt pathological or even visual foundation models to\npathology-specific tasks via multi-modal prompt tuning. The proposed framework\nleverages Task-specific Visual Prompts and Task-specific Textual Prompts to\nidentify task-relevant features, along with Instance-specific Visual Prompts\nfor encoding single pathological image features. Results across multiple\ndatasets at both patch-level and WSI-level demonstrate its superior performance\nover single-modality prompt tuning approaches. Significantly, PathoTune\nfacilitates the direct adaptation of natural visual foundation models to\npathological tasks, drastically outperforming pathological foundation models\nwith simple linear probing. The code is available at\nhttps://github.com/openmedlab/PathoDuet.\n", "date": "2024-03-25", "categories": []}
{"id": "2407.09540", "title": "Prompting Whole Slide Image Based Genetic Biomarker Prediction", "abstract": "  Prediction of genetic biomarkers, e.g., microsatellite instability and BRAF\nin colorectal cancer is crucial for clinical decision making. In this paper, we\npropose a whole slide image (WSI) based genetic biomarker prediction method via\nprompting techniques. Our work aims at addressing the following challenges: (1)\nextracting foreground instances related to genetic biomarkers from gigapixel\nWSIs, and (2) the interaction among the fine-grained pathological components in\nWSIs.Specifically, we leverage large language models to generate medical\nprompts that serve as prior knowledge in extracting instances associated with\ngenetic biomarkers. We adopt a coarse-to-fine approach to mine biomarker\ninformation within the tumor microenvironment. This involves extracting\ninstances related to genetic biomarkers using coarse medical prior knowledge,\ngrouping pathology instances into fine-grained pathological components and\nmining their interactions. Experimental results on two colorectal cancer\ndatasets show the superiority of our method, achieving 91.49% in AUC for MSI\nclassification. The analysis further shows the clinical interpretability of our\nmethod. Code is publicly available at\nhttps://github.com/DeepMed-Lab-ECNU/PromptBio.\n", "date": "2024-06-26", "categories": []}
{"id": "2407.10274", "title": "Enhancing Weakly-Supervised Histopathology Image Segmentation with\n  Knowledge Distillation on MIL-Based Pseudo-Labels", "abstract": "  Segmenting tumors in histological images is vital for cancer diagnosis. While\nfully supervised models excel with pixel-level annotations, creating such\nannotations is labor-intensive and costly. Accurate histopathology image\nsegmentation under weakly-supervised conditions with coarse-grained image\nlabels is still a challenging problem. Although multiple instance learning\n(MIL) has shown promise in segmentation tasks, surprisingly, no previous\npseudo-supervision methods have used MIL-based outputs as pseudo-masks for\ntraining. We suspect this stems from concerns over noises in MIL results\naffecting pseudo supervision quality. To explore the potential of leveraging\nMIL-based segmentation for pseudo supervision, we propose a novel distillation\nframework for histopathology image segmentation. This framework introduces a\niterative fusion-knowledge distillation strategy, enabling the student model to\nlearn directly from the teacher's comprehensive outcomes. Through dynamic role\nreversal between the fixed teacher and learnable student models and the\nincorporation of weighted cross-entropy loss for model optimization, our\napproach prevents performance deterioration and noise amplification during\nknowledge distillation. Experimental results on public histopathology datasets,\nCamelyon16 and Digestpath2019, demonstrate that our approach not only\ncomplements various MIL-based segmentation methods but also significantly\nenhances their performance. Additionally, our method achieves new SOTA in the\nfield.\n", "date": "2024-07-14", "categories": []}
{"id": "2407.10814", "title": "Pathology-knowledge Enhanced Multi-instance Prompt Learning for Few-shot\n  Whole Slide Image Classification", "abstract": "  Current multi-instance learning algorithms for pathology image analysis often\nrequire a substantial number of Whole Slide Images for effective training but\nexhibit suboptimal performance in scenarios with limited learning data. In\nclinical settings, restricted access to pathology slides is inevitable due to\npatient privacy concerns and the prevalence of rare or emerging diseases. The\nemergence of the Few-shot Weakly Supervised WSI Classification accommodates the\nsignificant challenge of the limited slide data and sparse slide-level labels\nfor diagnosis. Prompt learning based on the pre-trained models (\\eg, CLIP)\nappears to be a promising scheme for this setting; however, current research in\nthis area is limited, and existing algorithms often focus solely on patch-level\nprompts or confine themselves to language prompts. This paper proposes a\nmulti-instance prompt learning framework enhanced with pathology knowledge,\n\\ie, integrating visual and textual prior knowledge into prompts at both patch\nand slide levels. The training process employs a combination of static and\nlearnable prompts, effectively guiding the activation of pre-trained models and\nfurther facilitating the diagnosis of key pathology patterns. Lightweight\nMessenger (self-attention) and Summary (attention-pooling) layers are\nintroduced to model relationships between patches and slides within the same\npatient data. Additionally, alignment-wise contrastive losses ensure the\nfeature-level alignment between visual and textual learnable prompts for both\npatches and slides. Our method demonstrates superior performance in three\nchallenging clinical tasks, significantly outperforming comparative few-shot\nmethods.\n", "date": "2024-07-15", "categories": []}
{"id": "2407.08153", "title": "Lifelong Histopathology Whole Slide Image Retrieval via Distance\n  Consistency Rehearsal", "abstract": "  Content-based histopathological image retrieval (CBHIR) has gained attention\nin recent years, offering the capability to return histopathology images that\nare content-wise similar to the query one from an established database.\nHowever, in clinical practice, the continuously expanding size of WSI databases\nlimits the practical application of the current CBHIR methods. In this paper,\nwe propose a Lifelong Whole Slide Retrieval (LWSR) framework to address the\nchallenges of catastrophic forgetting by progressive model updating on\ncontinuously growing retrieval database. Our framework aims to achieve the\nbalance between stability and plasticity during continuous learning. To\npreserve system plasticity, we utilize local memory bank with reservoir\nsampling method to save instances, which can comprehensively encompass the\nfeature spaces of both old and new tasks. Furthermore, A distance consistency\nrehearsal (DCR) module is designed to ensure the retrieval queue's consistency\nfor previous tasks, which is regarded as stability within a lifelong CBHIR\nsystem. We evaluated the proposed method on four public WSI datasets from TCGA\nprojects. The experimental results have demonstrated the proposed method is\neffective and is superior to the state-of-the-art methods.\n", "date": "2024-07-11", "categories": []}
{"id": "2407.09030", "title": "CAMP: Continuous and Adaptive Learning Model in Pathology", "abstract": "  There exist numerous diagnostic tasks in pathology. Conventional\ncomputational pathology formulates and tackles them as independent and\nindividual image classification problems, thereby resulting in computational\ninefficiency and high costs. To address the challenges, we propose a generic,\nunified, and universal framework, called a continuous and adaptive learning\nmodel in pathology (CAMP), for pathology image classification. CAMP is a\ngenerative, efficient, and adaptive classification model that can continuously\nadapt to any classification task by leveraging pathology-specific prior\nknowledge and learning taskspecific knowledge with minimal computational cost\nand without forgetting the knowledge from the existing tasks. We evaluated CAMP\non 22 datasets, including 1,171,526 patches and 11,811 pathology slides, across\n17 classification tasks. CAMP achieves state-of-theart classification\nperformance on a wide range of datasets and tasks at both patch- and\nslide-levels and reduces up to 94% of computation time and 85% of storage\nmemory in comparison to the conventional classification models. Our results\ndemonstrate that CAMP can offer a fundamental transformation in pathology image\nclassification, paving the way for the fully digitized and computerized\npathology practice.\n", "date": "2024-07-12", "categories": []}
{"id": "2407.09035", "title": "GPC: Generative and General Pathology Image Classifier", "abstract": "  Deep learning has been increasingly incorporated into various computational\npathology applications to improve its efficiency, accuracy, and robustness.\nAlthough successful, most previous approaches for image classification have\ncrucial drawbacks. There exist numerous tasks in pathology, but one needs to\nbuild a model per task, i.e., a task-specific model, thereby increasing the\nnumber of models, training resources, and cost. Moreover, transferring\narbitrary task-specific model to another task is still a challenging problem.\nHerein, we propose a task-agnostic generative and general pathology image\nclassifier, so called GPC, that aims at learning from diverse kinds of\npathology images and conducting numerous classification tasks in a unified\nmodel. GPC, equipped with a convolutional neural network and a\nTransformer-based language model, maps pathology images into a high-dimensional\nfeature space and generates pertinent class labels as texts via the\nimage-to-text classification mechanism. We evaluate GPC on six datasets for\nfour different pathology image classification tasks. Experimental results show\nthat GPC holds considerable potential for developing an effective and efficient\nuniversal model for pathology image analysis.\n", "date": "2024-07-12", "categories": []}
{"id": "2407.06508", "title": "A Clinical Benchmark of Public Self-Supervised Pathology Foundation\n  Models", "abstract": "  The use of self-supervised learning (SSL) to train pathology foundation\nmodels has increased substantially in the past few years. Notably, several\nmodels trained on large quantities of clinical data have been made publicly\navailable in recent months. This will significantly enhance scientific research\nin computational pathology and help bridge the gap between research and\nclinical deployment. With the increase in availability of public foundation\nmodels of different sizes, trained using different algorithms on different\ndatasets, it becomes important to establish a benchmark to compare the\nperformance of such models on a variety of clinically relevant tasks spanning\nmultiple organs and diseases. In this work, we present a collection of\npathology datasets comprising clinical slides associated with clinically\nrelevant endpoints including cancer diagnoses and a variety of biomarkers\ngenerated during standard hospital operation from two medical centers. We\nleverage these datasets to systematically assess the performance of public\npathology foundation models and provide insights into best practices for\ntraining new foundation models and selecting appropriate pretrained models.\n", "date": "2024-07-09", "categories": []}
{"id": "2407.08625", "title": "Histopathological Image Classification with Cell Morphology Aware Deep\n  Neural Networks", "abstract": "  Histopathological images are widely used for the analysis of diseased (tumor)\ntissues and patient treatment selection. While the majority of microscopy image\nprocessing was previously done manually by pathologists, recent advances in\ncomputer vision allow for accurate recognition of lesion regions with deep\nlearning-based solutions. Such models, however, usually require extensive\nannotated datasets for training, which is often not the case in the considered\ntask, where the number of available patient data samples is very limited. To\ndeal with this problem, we propose a novel DeepCMorph model pre-trained to\nlearn cell morphology and identify a large number of different cancer types.\nThe model consists of two modules: the first one performs cell nuclei\nsegmentation and annotates each cell type, and is trained on a combination of 8\npublicly available datasets to ensure its high generalizability and robustness.\nThe second module combines the obtained segmentation map with the original\nmicroscopy image and is trained for the downstream task. We pre-trained this\nmodule on the Pan-Cancer TCGA dataset consisting of over 270K tissue patches\nextracted from 8736 diagnostic slides from 7175 patients. The proposed solution\nachieved a new state-of-the-art performance on the dataset under consideration,\ndetecting 32 cancer types with over 82% accuracy and outperforming all\npreviously proposed solutions by more than 4%. We demonstrate that the\nresulting pre-trained model can be easily fine-tuned on smaller microscopy\ndatasets, yielding superior results compared to the current top solutions and\nmodels initialized with ImageNet weights. The codes and pre-trained models\npresented in this paper are available at: https://github.com/aiff22/DeepCMorph\n", "date": "2024-07-11", "categories": []}
{"id": "2407.07360", "title": "Towards a text-based quantitative and explainable histopathology image\n  analysis", "abstract": "  Recently, vision-language pre-trained models have emerged in computational\npathology. Previous works generally focused on the alignment of image-text\npairs via the contrastive pre-training paradigm. Such pre-trained models have\nbeen applied to pathology image classification in zero-shot learning or\ntransfer learning fashion. Herein, we hypothesize that the pre-trained\nvision-language models can be utilized for quantitative histopathology image\nanalysis through a simple image-to-text retrieval. To this end, we propose a\nText-based Quantitative and Explainable histopathology image analysis, which we\ncall TQx. Given a set of histopathology images, we adopt a pre-trained\nvision-language model to retrieve a word-of-interest pool. The retrieved words\nare then used to quantify the histopathology images and generate understandable\nfeature embeddings due to the direct mapping to the text description. To\nevaluate the proposed method, the text-based embeddings of four histopathology\nimage datasets are utilized to perform clustering and classification tasks. The\nresults demonstrate that TQx is able to quantify and analyze histopathology\nimages that are comparable to the prevalent visual models in computational\npathology.\n", "date": "2024-07-10", "categories": []}
{"id": "2406.06517", "title": "Genomics-guided Representation Learning for Pathologic Pan-cancer Tumor\n  Microenvironment Subtype Prediction", "abstract": "  The characterization of Tumor MicroEnvironment (TME) is challenging due to\nits complexity and heterogeneity. Relatively consistent TME characteristics\nembedded within highly specific tissue features, render them difficult to\npredict. The capability to accurately classify TME subtypes is of critical\nsignificance for clinical tumor diagnosis and precision medicine. Based on the\nobservation that tumors with different origins share similar microenvironment\npatterns, we propose PathoTME, a genomics-guided Siamese representation\nlearning framework employing Whole Slide Image (WSI) for pan-cancer TME\nsubtypes prediction. Specifically, we utilize Siamese network to leverage\ngenomic information as a regularization factor to assist WSI embeddings\nlearning during the training phase. Additionally, we employ Domain Adversarial\nNeural Network (DANN) to mitigate the impact of tissue type variations. To\neliminate domain bias, a dynamic WSI prompt is designed to further unleash the\nmodel's capabilities. Our model achieves better performance than other\nstate-of-the-art methods across 23 cancer types on TCGA dataset. Our code is\navailable at https://github.com/Mengflz/PathoTME.\n", "date": "2024-06-10", "categories": []}
{"id": "2407.06116", "title": "Data-driven Nucleus Subclassification on Colon H&E using\n  Style-transferred Digital Pathology", "abstract": "  Understanding the way cells communicate, co-locate, and interrelate is\nessential to furthering our understanding of how the body functions. H&E is\nwidely available, however, cell subtyping often requires expert knowledge and\nthe use of specialized stains. To reduce the annotation burden, AI has been\nproposed for the classification of cells on H&E. For example, the recent Colon\nNucleus Identification and Classification (CoNIC) Challenge focused on labeling\n6 cell types on H&E of the colon. However, the CoNIC Challenge was unable to\nclassify epithelial subtypes (progenitor, enteroendocrine, goblet), lymphocyte\nsubtypes (B, helper T, cytotoxic T), and connective subtypes (fibroblasts). We\nuse inter-modality learning to label previously un-labelable cell types on H&E.\nWe take advantage of multiplexed immunofluorescence (MxIF) histology to label\n14 cell subclasses. We performed style transfer on the same MxIF tissues to\nsynthesize realistic virtual H&E which we paired with the MxIF-derived cell\nsubclassification labels. We evaluated the efficacy of using a supervised\nlearning scheme where the input was realistic-quality virtual H&E and the\nlabels were MxIF-derived cell subclasses. We assessed our model on private\nvirtual H&E and public real H&E. On virtual H&E, we were able to classify\nhelper T cells and epithelial progenitors with positive predictive values of\n$0.34 \\pm 0.15$ (prevalence $0.03 \\pm 0.01$) and $0.47 \\pm 0.1$ (prevalence\n$0.07 \\pm 0.02$) respectively, when using ground truth centroid information. On\nreal H&E we could classify helper T cells and epithelial progenitors with upper\nbound positive predictive values of $0.43 \\pm 0.03$ (parent class prevalence\n0.21) and $0.94 \\pm 0.02$ (parent class prevalence 0.49) when using ground\ntruth centroid information. This is the first work to provide cell type\nclassification for helper T and epithelial progenitor nuclei on H&E.\n", "date": "2024-05-15", "categories": []}
{"id": "2406.19081", "title": "Unsupervised Latent Stain Adaptation for Computational Pathology", "abstract": "  In computational pathology, deep learning (DL) models for tasks such as\nsegmentation or tissue classification are known to suffer from domain shifts\ndue to different staining techniques. Stain adaptation aims to reduce the\ngeneralization error between different stains by training a model on source\nstains that generalizes to target stains. Despite the abundance of target stain\ndata, a key challenge is the lack of annotations. To address this, we propose a\njoint training between artificially labeled and unlabeled data including all\navailable stained images called Unsupervised Latent Stain Adaptation (ULSA).\nOur method uses stain translation to enrich labeled source images with\nsynthetic target images in order to increase the supervised signals. Moreover,\nwe leverage unlabeled target stain images using stain-invariant feature\nconsistency learning. With ULSA we present a semi-supervised strategy for\nefficient stain adaptation without access to annotated target stain data.\nRemarkably, ULSA is task agnostic in patch-level analysis for whole slide\nimages (WSIs). Through extensive evaluation on external datasets, we\ndemonstrate that ULSA achieves state-of-the-art (SOTA) performance in kidney\ntissue segmentation and breast cancer classification across a spectrum of\nstaining variations. Our findings suggest that ULSA is an important framework\nfor stain adaptation in computational pathology.\n", "date": "2024-06-27", "categories": []}
{"id": "2407.03307", "title": "HoloHisto: End-to-end Gigapixel WSI Segmentation with 4K Resolution\n  Sequential Tokenization", "abstract": "  In digital pathology, the traditional method for deep learning-based image\nsegmentation typically involves a two-stage process: initially segmenting\nhigh-resolution whole slide images (WSI) into smaller patches (e.g., 256x256,\n512x512, 1024x1024) and subsequently reconstructing them to their original\nscale. This method often struggles to capture the complex details and vast\nscope of WSIs. In this paper, we propose the holistic histopathology\n(HoloHisto) segmentation method to achieve end-to-end segmentation on gigapixel\nWSIs, whose maximum resolution is above 80,000$\\times$70,000 pixels. HoloHisto\nfundamentally shifts the paradigm of WSI segmentation to an end-to-end learning\nfashion with 1) a large (4K) resolution base patch for elevated visual\ninformation inclusion and efficient processing, and 2) a novel sequential\ntokenization mechanism to properly model the contextual relationships and\nefficiently model the rich information from the 4K input. To our best\nknowledge, HoloHisto presents the first holistic approach for gigapixel\nresolution WSI segmentation, supporting direct I/O of complete WSI and their\ncorresponding gigapixel masks. Under the HoloHisto platform, we unveil a random\n4K sampler that transcends ultra-high resolution, delivering 31 and 10 times\nmore pixels than standard 2D and 3D patches, respectively, for advancing\ncomputational capabilities. To facilitate efficient 4K resolution dense\nprediction, we leverage sequential tokenization, utilizing a pre-trained image\ntokenizer to group image features into a discrete token grid. To assess the\nperformance, our team curated a new kidney pathology image segmentation (KPIs)\ndataset with WSI-level glomeruli segmentation from whole mouse kidneys. From\nthe results, HoloHisto-4K delivers remarkable performance gains over previous\nstate-of-the-art models.\n", "date": "2024-07-03", "categories": []}
{"id": "2407.00098", "title": "Scalable, Trustworthy Generative Model for Virtual Multi-Staining from\n  H&E Whole Slide Images", "abstract": "  Chemical staining methods are dependable but require extensive time,\nexpensive chemicals, and raise environmental concerns. These challenges\nhighlight the need for alternative solutions like virtual staining, which\naccelerates the diagnostic process and enhances stain application flexibility.\nGenerative AI technologies are pivotal in addressing these issues. However, the\nhigh-stakes nature of healthcare decisions, especially in computational\npathology, complicates the adoption of these tools due to their opaque\nprocesses. Our work introduces the use of generative AI for virtual staining,\naiming to enhance performance, trustworthiness, scalability, and adaptability\nin computational pathology. The methodology centers on a singular H&E encoder\nsupporting multiple stain decoders. This design focuses on critical regions in\nthe latent space of H&E, enabling precise synthetic stain generation. Our\nmethod, tested to generate 8 different stains from a single H&E slide, offers\nscalability by loading only necessary model components during production. We\nintegrate label-free knowledge in training, using loss functions and\nregularization to minimize artifacts, thus improving paired/unpaired virtual\nstaining accuracy. To build trust, we use real-time self-inspection with\ndiscriminators for each stain type, providing pathologists with confidence\nheat-maps. Automatic quality checks on new H&E slides ensure conformity to the\ntrained distribution, ensuring accurate synthetic stains. Recognizing\npathologists' challenges with new technologies, we have developed an\nopen-source, cloud-based system, that allows easy virtual staining of H&E\nslides through a browser, addressing hardware/software issues and facilitating\nreal-time user feedback. We also curated a novel dataset of 8 paired H&E/stains\nrelated to pediatric Crohn's disease, comprising 480 WSIs to further stimulate\ncomputational pathology research.\n", "date": "2024-06-26", "categories": []}
{"id": "2407.00596", "title": "HATs: Hierarchical Adaptive Taxonomy Segmentation for Panoramic\n  Pathology Image Analysis", "abstract": "  Panoramic image segmentation in computational pathology presents a remarkable\nchallenge due to the morphologically complex and variably scaled anatomy. For\ninstance, the intricate organization in kidney pathology spans multiple layers,\nfrom regions like the cortex and medulla to functional units such as glomeruli,\ntubules, and vessels, down to various cell types. In this paper, we propose a\nnovel Hierarchical Adaptive Taxonomy Segmentation (HATs) method, which is\ndesigned to thoroughly segment panoramic views of kidney structures by\nleveraging detailed anatomical insights. Our approach entails (1) the\ninnovative HATs technique which translates spatial relationships among 15\ndistinct object classes into a versatile \"plug-and-play\" loss function that\nspans across regions, functional units, and cells, (2) the incorporation of\nanatomical hierarchies and scale considerations into a unified simple matrix\nrepresentation for all panoramic entities, (3) the adoption of the latest AI\nfoundation model (EfficientSAM) as a feature extraction tool to boost the\nmodel's adaptability, yet eliminating the need for manual prompt generation in\nconventional segment anything model (SAM). Experimental findings demonstrate\nthat the HATs method offers an efficient and effective strategy for integrating\nclinical insights and imaging precedents into a unified segmentation model\nacross more than 15 categories. The official implementation is publicly\navailable at https://github.com/hrlblab/HATs.\n", "date": "2024-06-30", "categories": []}
{"id": "2406.19578", "title": "PathAlign: A vision-language model for whole slide images in\n  histopathology", "abstract": "  Microscopic interpretation of histopathology images underlies many important\ndiagnostic and treatment decisions. While advances in vision-language modeling\nraise new opportunities for analysis of such images, the gigapixel-scale size\nof whole slide images (WSIs) introduces unique challenges. Additionally,\npathology reports simultaneously highlight key findings from small regions\nwhile also aggregating interpretation across multiple slides, often making it\ndifficult to create robust image-text pairs. As such, pathology reports remain\na largely untapped source of supervision in computational pathology, with most\nefforts relying on region-of-interest annotations or self-supervision at the\npatch-level. In this work, we develop a vision-language model based on the\nBLIP-2 framework using WSIs paired with curated text from pathology reports.\nThis enables applications utilizing a shared image-text embedding space, such\nas text or image retrieval for finding cases of interest, as well as\nintegration of the WSI encoder with a frozen large language model (LLM) for\nWSI-based generative text capabilities such as report generation or\nAI-in-the-loop interactions. We utilize a de-identified dataset of over 350,000\nWSIs and diagnostic text pairs, spanning a wide range of diagnoses, procedure\ntypes, and tissue types. We present pathologist evaluation of text generation\nand text retrieval using WSI embeddings, as well as results for WSI\nclassification and workflow prioritization (slide-level triaging).\nModel-generated text for WSIs was rated by pathologists as accurate, without\nclinically significant error or omission, for 78% of WSIs on average. This work\ndemonstrates exciting potential capabilities for language-aligned WSI\nembeddings.\n", "date": "2024-06-27", "categories": []}
{"id": "2406.18556", "title": "Renal digital pathology visual knowledge search platform based on\n  language large model and book knowledge", "abstract": "  Large models have become mainstream, yet their applications in digital\npathology still require exploration. Meanwhile renal pathology images play an\nimportant role in the diagnosis of renal diseases. We conducted image\nsegmentation and paired corresponding text descriptions based on 60 books for\nrenal pathology, clustering analysis for all image and text description\nfeatures based on large models, ultimately building a retrieval system based on\nthe semantic features of large models. Based above analysis, we established a\nknowledge base of 10,317 renal pathology images and paired corresponding text\ndescriptions, and then we evaluated the semantic feature capabilities of 4\nlarge models, including GPT2, gemma, LLma and Qwen, and the image-based feature\ncapabilities of dinov2 large model. Furthermore, we built a semantic retrieval\nsystem to retrieve pathological images based on text descriptions, and named\nRppD (aidp.zjsru.edu.cn).\n", "date": "2024-05-27", "categories": []}
{"id": "2406.17225", "title": "Multimodal Cross-Task Interaction for Survival Analysis in Whole Slide\n  Pathological Images", "abstract": "  Survival prediction, utilizing pathological images and genomic profiles, is\nincreasingly important in cancer analysis and prognosis. Despite significant\nprogress, precise survival analysis still faces two main challenges: (1) The\nmassive pixels contained in whole slide images (WSIs) complicate the process of\npathological images, making it difficult to generate an effective\nrepresentation of the tumor microenvironment (TME). (2) Existing multimodal\nmethods often rely on alignment strategies to integrate complementary\ninformation, which may lead to information loss due to the inherent\nheterogeneity between pathology and genes. In this paper, we propose a\nMultimodal Cross-Task Interaction (MCTI) framework to explore the intrinsic\ncorrelations between subtype classification and survival analysis tasks.\nSpecifically, to capture TME-related features in WSIs, we leverage the subtype\nclassification task to mine tumor regions. Simultaneously, multi-head attention\nmechanisms are applied in genomic feature extraction, adaptively performing\ngenes grouping to obtain task-related genomic embedding. With the joint\nrepresentation of pathological images and genomic data, we further introduce a\nTransport-Guided Attention (TGA) module that uses optimal transport theory to\nmodel the correlation between subtype classification and survival analysis\ntasks, effectively transferring potential information. Extensive experiments\ndemonstrate the superiority of our approaches, with MCTI outperforming\nstate-of-the-art frameworks on three public benchmarks.\n\\href{https://github.com/jsh0792/MCTI}{https://github.com/jsh0792/MCTI}.\n", "date": "2024-06-25", "categories": []}
{"id": "2311.11772", "title": "Benchmarking Pathology Feature Extractors for Whole Slide Image\n  Classification", "abstract": "  Weakly supervised whole slide image classification is a key task in\ncomputational pathology, which involves predicting a slide-level label from a\nset of image patches constituting the slide. Constructing models to solve this\ntask involves multiple design choices, often made without robust empirical or\nconclusive theoretical justification. To address this, we conduct a\ncomprehensive benchmarking of feature extractors to answer three critical\nquestions: 1) Is stain normalisation still a necessary preprocessing step? 2)\nWhich feature extractors are best for downstream slide-level classification? 3)\nHow does magnification affect downstream performance? Our study constitutes the\nmost comprehensive evaluation of publicly available pathology feature\nextractors to date, involving more than 10,000 training runs across 14 feature\nextractors, 9 tasks, 5 datasets, 3 downstream architectures, 2 levels of\nmagnification, and various preprocessing setups. Our findings challenge\nexisting assumptions: 1) We observe empirically, and by analysing the latent\nspace, that skipping stain normalisation and image augmentations does not\ndegrade performance, while significantly reducing memory and computational\ndemands. 2) We develop a novel evaluation metric to compare relative downstream\nperformance, and show that the choice of feature extractor is the most\nconsequential factor for downstream performance. 3) We find that\nlower-magnification slides are sufficient for accurate slide-level\nclassification. Contrary to previous patch-level benchmarking studies, our\napproach emphasises clinical relevance by focusing on slide-level biomarker\nprediction tasks in a weakly supervised setting with external validation\ncohorts. Our findings stand to streamline digital pathology workflows by\nminimising preprocessing needs and informing the selection of feature\nextractors.\n", "date": "2023-11-20", "categories": []}
{"id": "2406.06393", "title": "STimage-1K4M: A histopathology image-gene expression dataset for spatial\n  transcriptomics", "abstract": "  Recent advances in multi-modal algorithms have driven and been driven by the\nincreasing availability of large image-text datasets, leading to significant\nstrides in various fields, including computational pathology. However, in most\nexisting medical image-text datasets, the text typically provides high-level\nsummaries that may not sufficiently describe sub-tile regions within a large\npathology image. For example, an image might cover an extensive tissue area\ncontaining cancerous and healthy regions, but the accompanying text might only\nspecify that this image is a cancer slide, lacking the nuanced details needed\nfor in-depth analysis. In this study, we introduce STimage-1K4M, a novel\ndataset designed to bridge this gap by providing genomic features for sub-tile\nimages. STimage-1K4M contains 1,149 images derived from spatial transcriptomics\ndata, which captures gene expression information at the level of individual\nspatial spots within a pathology image. Specifically, each image in the dataset\nis broken down into smaller sub-image tiles, with each tile paired with\n15,000-30,000 dimensional gene expressions. With 4,293,195 pairs of sub-tile\nimages and gene expressions, STimage-1K4M offers unprecedented granularity,\npaving the way for a wide range of advanced research in multi-modal data\nanalysis an innovative applications in computational pathology, and beyond.\n", "date": "2024-06-10", "categories": []}
{"id": "2406.14287", "title": "Segmentation of Non-Small Cell Lung Carcinomas: Introducing DRU-Net and\n  Multi-Lens Distortion", "abstract": "  Considering the increased workload in pathology laboratories today, automated\ntools such as artificial intelligence models can help pathologists with their\ntasks and ease the workload. In this paper, we are proposing a segmentation\nmodel (DRU-Net) that can provide a delineation of human non-small cell lung\ncarcinomas and an augmentation method that can improve classification results.\nThe proposed model is a fused combination of truncated pre-trained DenseNet201\nand ResNet101V2 as a patch-wise classifier followed by a lightweight U-Net as a\nrefinement model. We have used two datasets (Norwegian Lung Cancer Biobank and\nHaukeland University Hospital lung cancer cohort) to create our proposed model.\nThe DRU-Net model achieves an average of 0.91 Dice similarity coefficient. The\nproposed spatial augmentation method (multi-lens distortion) improved the\nnetwork performance by 3%. Our findings show that choosing image patches that\nspecifically include regions of interest leads to better results for the\npatch-wise classifier compared to other sampling methods. The qualitative\nanalysis showed that the DRU-Net model is generally successful in detecting the\ntumor. On the test set, some of the cases showed areas of false positive and\nfalse negative segmentation in the periphery, particularly in tumors with\ninflammatory and reactive changes.\n", "date": "2024-06-20", "categories": []}
{"id": "2406.10893", "title": "Development and Validation of Fully Automatic Deep Learning-Based\n  Algorithms for Immunohistochemistry Reporting of Invasive Breast Ductal\n  Carcinoma", "abstract": "  Immunohistochemistry (IHC) analysis is a well-accepted and widely used method\nfor molecular subtyping, a procedure for prognosis and targeted therapy of\nbreast carcinoma, the most common type of tumor affecting women. There are four\nmolecular biomarkers namely progesterone receptor (PR), estrogen receptor (ER),\nantigen Ki67, and human epidermal growth factor receptor 2 (HER2) whose\nassessment is needed under IHC procedure to decide prognosis as well as\npredictors of response to therapy. However, IHC scoring is based on subjective\nmicroscopic examination of tumor morphology and suffers from poor\nreproducibility, high subjectivity, and often incorrect scoring in low-score\ncases. In this paper, we present, a deep learning-based semi-supervised\ntrained, fully automatic, decision support system (DSS) for IHC scoring of\ninvasive ductal carcinoma. Our system automatically detects the tumor region\nremoving artifacts and scores based on Allred standard. The system is developed\nusing 3 million pathologist-annotated image patches from 300 slides, fifty\nthousand in-house cell annotations, and forty thousand pixels marking of HER2\nmembrane. We have conducted multicentric trials at four centers with three\ndifferent types of digital scanners in terms of percentage agreement with\ndoctors. And achieved agreements of 95, 92, 88 and 82 percent for Ki67, HER2,\nER, and PR stain categories, respectively. In addition to overall accuracy, we\nfound that there is 5 percent of cases where pathologist have changed their\nscore in favor of algorithm score while reviewing with detailed algorithmic\nanalysis. Our approach could improve the accuracy of IHC scoring and subsequent\ntherapy decisions, particularly where specialist expertise is unavailable. Our\nsystem is highly modular. The proposed algorithm modules can be used to develop\nDSS for other cancer types.\n", "date": "2024-06-16", "categories": []}
{"id": "2401.02651", "title": "Benchmarking PathCLIP for Pathology Image Analysis", "abstract": "  Accurate image classification and retrieval are of importance for clinical\ndiagnosis and treatment decision-making. The recent contrastive language-image\npretraining (CLIP) model has shown remarkable proficiency in understanding\nnatural images. Drawing inspiration from CLIP, PathCLIP is specifically\ndesigned for pathology image analysis, utilizing over 200,000 image and text\npairs in training. While the performance the PathCLIP is impressive, its\nrobustness under a wide range of image corruptions remains unknown. Therefore,\nwe conduct an extensive evaluation to analyze the performance of PathCLIP on\nvarious corrupted images from the datasets of Osteosarcoma and WSSS4LUAD. In\nour experiments, we introduce seven corruption types including brightness,\ncontrast, Gaussian blur, resolution, saturation, hue, and markup at four\nseverity levels. Through experiments, we find that PathCLIP is relatively\nrobustness to image corruptions and surpasses OpenAI-CLIP and PLIP in zero-shot\nclassification. Among the seven corruptions, blur and resolution can cause\nserver performance degradation of the PathCLIP. This indicates that ensuring\nthe quality of images is crucial before conducting a clinical test.\nAdditionally, we assess the robustness of PathCLIP in the task of image-image\nretrieval, revealing that PathCLIP performs less effectively than PLIP on\nOsteosarcoma but performs better on WSSS4LUAD under diverse corruptions.\nOverall, PathCLIP presents impressive zero-shot classification and retrieval\nperformance for pathology images, but appropriate care needs to be taken when\nusing it. We hope this study provides a qualitative impression of PathCLIP and\nhelps understand its differences from other CLIP models.\n", "date": "2024-01-05", "categories": []}
{"id": "2401.04079", "title": "RudolfV: A Foundation Model by Pathologists for Pathologists", "abstract": "  Artificial intelligence has started to transform histopathology impacting\nclinical diagnostics and biomedical research. However, while many computational\npathology approaches have been proposed, most current AI models are limited\nwith respect to generalization, application variety, and handling rare\ndiseases. Recent efforts introduced self-supervised foundation models to\naddress these challenges, yet existing approaches do not leverage pathologist\nknowledge by design. In this study, we present a novel approach to designing\nfoundation models for computational pathology, incorporating pathologist\nexpertise, semi-automated data curation, and a diverse dataset from over 15\nlaboratories, including 58 tissue types, and encompassing 129 different\nhistochemical and immunohistochemical staining modalities. We demonstrate that\nour model \"RudolfV\" surpasses existing state-of-the-art foundation models\nacross different benchmarks focused on tumor microenvironment profiling,\nbiomarker evaluation, and reference case search while exhibiting favorable\nrobustness properties. Our study shows how domain-specific knowledge can\nincrease the efficiency and performance of pathology foundation models and\nenable novel application areas.\n", "date": "2024-01-08", "categories": []}
{"id": "2406.02077", "title": "Multi-target stain normalization for histology slides", "abstract": "  Traditional staining normalization approaches, e.g. Macenko, typically rely\non the choice of a single representative reference image, which may not\nadequately account for the diverse staining patterns of datasets collected in\npractical scenarios. In this study, we introduce a novel approach that\nleverages multiple reference images to enhance robustness against stain\nvariation. Our method is parameter-free and can be adopted in existing\ncomputational pathology pipelines with no significant changes. We evaluate the\neffectiveness of our method through experiments using a deep-learning pipeline\nfor automatic nuclei segmentation on colorectal images. Our results show that\nby leveraging multiple reference images, better results can be achieved when\ngeneralizing to external data, where the staining can widely differ from the\ntraining set.\n", "date": "2024-06-04", "categories": []}
{"id": "2406.05205", "title": "CPLIP: Zero-Shot Learning for Histopathology with Comprehensive\n  Vision-Language Alignment", "abstract": "  This paper proposes Comprehensive Pathology Language Image Pre-training\n(CPLIP), a new unsupervised technique designed to enhance the alignment of\nimages and text in histopathology for tasks such as classification and\nsegmentation. This methodology enriches vision-language models by leveraging\nextensive data without needing ground truth annotations. CPLIP involves\nconstructing a pathology-specific dictionary, generating textual descriptions\nfor images using language models, and retrieving relevant images for each text\nsnippet via a pre-trained model. The model is then fine-tuned using a\nmany-to-many contrastive learning method to align complex interrelated concepts\nacross both modalities. Evaluated across multiple histopathology tasks, CPLIP\nshows notable improvements in zero-shot learning scenarios, outperforming\nexisting methods in both interpretability and robustness and setting a higher\nbenchmark for the application of vision-language models in the field. To\nencourage further research and replication, the code for CPLIP is available on\nGitHub at https://cplip.github.io/\n", "date": "2024-06-07", "categories": []}
{"id": "2406.05828", "title": "Multi-Stain Multi-Level Convolutional Network for Multi-Tissue Breast\n  Cancer Image Segmentation", "abstract": "  Digital pathology and microscopy image analysis are widely employed in the\nsegmentation of digitally scanned IHC slides, primarily to identify cancer and\npinpoint regions of interest (ROI) indicative of tumor presence. However,\ncurrent ROI segmentation models are either stain-specific or suffer from the\nissues of stain and scanner variance due to different staining protocols or\nmodalities across multiple labs. Also, tissues like Ductal Carcinoma in Situ\n(DCIS), acini, etc. are often classified as Tumors due to their structural\nsimilarities and color compositions. In this paper, we proposed a novel\nconvolutional neural network (CNN) based Multi-class Tissue Segmentation model\nfor histopathology whole-slide Breast slides which classify tumors and segments\nother tissue regions such as Ducts, acini, DCIS, Squamous epithelium, Blood\nVessels, Necrosis, etc. as a separate class. Our unique pixel-aligned\nnon-linear merge across spatial resolutions empowers models with both local and\nglobal fields of view for accurate detection of various classes. Our proposed\nmodel is also able to separate bad regions such as folds, artifacts, blurry\nregions, bubbles, etc. from tissue regions using multi-level context from\ndifferent resolutions of WSI. Multi-phase iterative training with context-aware\naugmentation and increasing noise was used to efficiently train a multi-stain\ngeneric model with partial and noisy annotations from 513 slides. Our training\npipeline used 12 million patches generated using context-aware augmentations\nwhich made our model stain and scanner invariant across data sources. To\nextrapolate stain and scanner invariance, our model was evaluated on 23000\npatches which were for a completely new stain (Hematoxylin and Eosin) from a\ncompletely new scanner (Motic) from a different lab. The mean IOU was 0.72\nwhich is on par with model performance on other data sources and scanners.\n", "date": "2024-06-09", "categories": []}
{"id": "2406.02990", "title": "Predicting Genetic Mutation from Whole Slide Images via\n  Biomedical-Linguistic Knowledge Enhanced Multi-label Classification", "abstract": "  Predicting genetic mutations from whole slide images is indispensable for\ncancer diagnosis. However, existing work training multiple binary\nclassification models faces two challenges: (a) Training multiple binary\nclassifiers is inefficient and would inevitably lead to a class imbalance\nproblem. (b) The biological relationships among genes are overlooked, which\nlimits the prediction performance. To tackle these challenges, we innovatively\ndesign a Biological-knowledge enhanced PathGenomic multi-label Transformer to\nimprove genetic mutation prediction performances. BPGT first establishes a\nnovel gene encoder that constructs gene priors by two carefully designed\nmodules: (a) A gene graph whose node features are the genes' linguistic\ndescriptions and the cancer phenotype, with edges modeled by genes' pathway\nassociations and mutation consistencies. (b) A knowledge association module\nthat fuses linguistic and biomedical knowledge into gene priors by\ntransformer-based graph representation learning, capturing the intrinsic\nrelationships between different genes' mutations. BPGT then designs a label\ndecoder that finally performs genetic mutation prediction by two tailored\nmodules: (a) A modality fusion module that firstly fuses the gene priors with\ncritical regions in WSIs and obtains gene-wise mutation logits. (b) A\ncomparative multi-label loss that emphasizes the inherent comparisons among\nmutation status to enhance the discrimination capabilities. Sufficient\nexperiments on The Cancer Genome Atlas benchmark demonstrate that BPGT\noutperforms the state-of-the-art.\n", "date": "2024-06-05", "categories": []}
{"id": "2406.00555", "title": "Length-scale study in deep learning prediction for non-small cell lung\n  cancer brain metastasis", "abstract": "  Deep learning assisted digital pathology has the potential to impact clinical\npractice in significant ways. In recent studies, deep neural network (DNN)\nenabled analysis outperforms human pathologists. Increasing sizes and\ncomplexity of the DNN architecture generally improves performance at the cost\nof DNN's explainability. For pathology, this lack of DNN explainability is\nparticularly problematic as it hinders the broader clinical interpretation of\nthe pathology features that may provide physiological disease insights. To\nbetter assess the features that DNN uses in developing predictive algorithms to\ninterpret digital microscopic images, we sought to understand the role of\nresolution and tissue scale and here describe a novel method for studying the\npredictive feature length-scale that underpins a DNN's predictive power. We\napplied the method to study a DNN's predictive capability in the case example\nof brain metastasis prediction from early-stage non-small-cell lung cancer\nbiopsy slides. The study highlights the DNN attention in the brain metastasis\nprediction targeting both cellular scale (resolution) and tissue scale features\non H&E-stained histological whole slide images. At the cellular scale, we see\nthat DNN's predictive power is progressively increased at higher resolution\n(i.e., lower resolvable feature length) and is largely lost when the resolvable\nfeature length is longer than 5 microns. In addition, DNN uses more macro-scale\nfeatures (maximal feature length) associated with tissue\norganization/architecture and is optimized when assessing visual fields larger\nthan 41 microns. This study for the first time demonstrates the length-scale\nrequirements necessary for optimal DNN learning on digital whole slide images.\n", "date": "2024-06-01", "categories": []}
{"id": "2405.18327", "title": "Histopathology Based AI Model Predicts Anti-Angiogenic Therapy Response\n  in Renal Cancer Clinical Trial", "abstract": "  Predictive biomarkers of treatment response are lacking for metastatic clear\ncell renal cell carcinoma (ccRCC), a tumor type that is treated with\nangiogenesis inhibitors, immune checkpoint inhibitors, mTOR inhibitors and a\nHIF2 inhibitor. The Angioscore, an RNA-based quantification of angiogenesis, is\narguably the best candidate to predict anti-angiogenic (AA) response. However,\nthe clinical adoption of transcriptomic assays faces several challenges\nincluding standardization, time delay, and high cost. Further, ccRCC tumors are\nhighly heterogenous, and sampling multiple areas for sequencing is impractical.\nHere we present a novel deep learning (DL) approach to predict the Angioscore\nfrom ubiquitous histopathology slides. To overcome the lack of\ninterpretability, one of the biggest limitations of typical DL models, our\nmodel produces a visual vascular network which is the basis of the model's\nprediction. To test its reliability, we applied this model to multiple cohorts\nincluding a clinical trial dataset. Our model accurately predicts the RNA-based\nAngioscore on multiple independent cohorts (spearman correlations of 0.77 and\n0.73). Further, the predictions help unravel meaningful biology such as\nassociation of angiogenesis with grade, stage, and driver mutation status.\nFinally, we find our model can predict response to AA therapy, in both a\nreal-world cohort and the IMmotion150 clinical trial. The predictive power of\nour model vastly exceeds that of CD31, a marker of vasculature, and nearly\nrivals the performance (c-index 0.66 vs 0.67) of the ground truth RNA-based\nAngioscore at a fraction of the cost. By providing a robust yet interpretable\nprediction of the Angioscore from histopathology slides alone, our approach\noffers insights into angiogenesis biology and AA treatment response.\n", "date": "2024-05-28", "categories": []}
{"id": "2404.13222", "title": "Vim4Path: Self-Supervised Vision Mamba for Histopathology Images", "abstract": "  Representation learning from Gigapixel Whole Slide Images (WSI) poses a\nsignificant challenge in computational pathology due to the complicated nature\nof tissue structures and the scarcity of labeled data. Multi-instance learning\nmethods have addressed this challenge, leveraging image patches to classify\nslides utilizing pretrained models using Self-Supervised Learning (SSL)\napproaches. The performance of both SSL and MIL methods relies on the\narchitecture of the feature encoder. This paper proposes leveraging the Vision\nMamba (Vim) architecture, inspired by state space models, within the DINO\nframework for representation learning in computational pathology. We evaluate\nthe performance of Vim against Vision Transformers (ViT) on the Camelyon16\ndataset for both patch-level and slide-level classification. Our findings\nhighlight Vim's enhanced performance compared to ViT, particularly at smaller\nscales, where Vim achieves an 8.21 increase in ROC AUC for models of similar\nsize. An explainability analysis further highlights Vim's capabilities, which\nreveals that Vim uniquely emulates the pathologist workflow-unlike ViT. This\nalignment with human expert analysis highlights Vim's potential in practical\ndiagnostic settings and contributes significantly to developing effective\nrepresentation-learning algorithms in computational pathology. We release the\ncodes and pretrained weights at\n\\url{https://github.com/AtlasAnalyticsLab/Vim4Path}.\n", "date": "2024-04-20", "categories": []}
{"id": "2402.06188", "title": "A self-supervised framework for learning whole slide representations", "abstract": "  Whole slide imaging is fundamental to biomedical microscopy and computational\npathology. Previously, learning representations for gigapixel-sized whole slide\nimages (WSIs) has relied on multiple instance learning with weak labels, which\ndo not annotate the diverse morphologic features and spatial heterogeneity of\nWSIs. A high-quality self-supervised learning method for WSIs would provide\ntransferable visual representations for downstream computational pathology\ntasks, without the need for dense annotations. We present Slide Pre-trained\nTransformers (SPT) for gigapixel-scale self-supervision of WSIs. Treating WSI\npatches as tokens, SPT combines data transformation strategies from language\nand vision modeling into a general and unified framework to generate views of\nWSIs for self-supervised pretraining. SPT leverages the inherent regional\nheterogeneity, histologic feature variability, and information redundancy\nwithin WSIs to learn high-quality whole slide representations. We benchmark SPT\nvisual representations on five diagnostic tasks across three biomedical\nmicroscopy datasets. SPT significantly outperforms baselines for\nhistopathologic diagnosis, cancer subtyping, and genetic mutation prediction.\nFinally, we demonstrate that SPT consistently improves whole slide\nrepresentations when using off-the-shelf, in-domain, and foundational patch\nencoders for whole slide multiple instance learning.\n", "date": "2024-02-09", "categories": []}
{"id": "2307.01098", "title": "Automated identification and quantification of myocardial inflammatory\n  infiltration in digital histological images to diagnose myocarditis", "abstract": "  This study aims to develop a new computational pathology approach that\nautomates the identification and quantification of myocardial inflammatory\ninfiltration in digital HE-stained images to provide a quantitative\nhistological diagnosis of myocarditis.898 HE-stained whole slide images (WSIs)\nof myocardium from 154 heart transplant patients diagnosed with myocarditis or\ndilated cardiomyopathy (DCM) were included in this study. An automated DL-based\ncomputational pathology approach was developed to identify nuclei and detect\nmyocardial inflammatory infiltration, enabling the quantification of the\nlymphocyte nuclear density (LND) on myocardial WSIs. A cutoff value based on\nthe quantification of LND was proposed to determine if the myocardial\ninflammatory infiltration was present. The performance of our approach was\nevaluated with a five-fold cross-validation experiment, tested with an internal\ntest set from the myocarditis group, and confirmed by an external test from a\ndouble-blind trial group. An LND of 1.02/mm2 could distinguish WSIs with\nmyocarditis from those without. The accuracy, sensitivity, specificity, and\narea under the receiver operating characteristic curve (AUC) in the five-fold\ncross-validation experiment were 0.899 plus or minus 0.035, 0.971 plus or minus\n0.017, 0.728 plus or minus 0.073 and 0.849 plus or minus 0.044, respectively.\nFor the internal test set, the accuracy, sensitivity, specificity, and AUC were\n0.887, 0.971, 0.737, and 0.854, respectively. The accuracy, sensitivity,\nspecificity, and AUC for the external test set reached 0.853, 0.846, 0.858, and\n0.852, respectively. Our new approach provides accurate and reliable\nquantification of the LND of myocardial WSIs, facilitating automated\nquantitative diagnosis of myocarditis with HE-stained images.\n", "date": "2023-07-03", "categories": []}
{"id": "2403.04865", "title": "Beyond Multiple Instance Learning: Full Resolution All-In-Memory\n  End-To-End Pathology Slide Modeling", "abstract": "  Artificial Intelligence (AI) has great potential to improve health outcomes\nby training systems on vast digitized clinical datasets. Computational\nPathology, with its massive amounts of microscopy image data and impact on\ndiagnostics and biomarkers, is at the forefront of this development. Gigapixel\npathology slides pose a unique challenge due to their enormous size and are\nusually divided into tens of thousands of smaller tiles for analysis. This\nresults in a discontinuity in the machine learning process by separating the\ntraining of tile-level encoders from slide-level aggregators and the need to\nadopt weakly supervised learning strategies. Training models from entire\npathology slides end-to-end has been largely unexplored due to its\ncomputational challenges. To overcome this problem, we propose a novel approach\nto jointly train both a tile encoder and a slide-aggregator fully in memory and\nend-to-end at high-resolution, bridging the gap between input and slide-level\nsupervision. While more computationally expensive, detailed quantitative\nvalidation shows promise for large-scale pre-training and fine-tuning of\npathology foundation models.\n", "date": "2024-03-07", "categories": []}
{"id": "2403.07743", "title": "Equipping Computational Pathology Systems with Artifact Processing\n  Pipelines: A Showcase for Computation and Performance Trade-offs", "abstract": "  Histopathology is a gold standard for cancer diagnosis under a microscopic\nexamination. However, histological tissue processing procedures result in\nartifacts, which are ultimately transferred to the digitized version of glass\nslides, known as whole slide images (WSIs). Artifacts are diagnostically\nirrelevant areas and may result in wrong deep learning (DL) algorithms\npredictions. Therefore, detecting and excluding artifacts in the computational\npathology (CPATH) system is essential for reliable automated diagnosis. In this\npaper, we propose a mixture of experts (MoE) scheme for detecting five notable\nartifacts, including damaged tissue, blur, folded tissue, air bubbles, and\nhistologically irrelevant blood from WSIs. First, we train independent binary\nDL models as experts to capture particular artifact morphology. Then, we\nensemble their predictions using a fusion mechanism. We apply probabilistic\nthresholding over the final probability distribution to improve the sensitivity\nof the MoE. We developed DL pipelines using two MoEs and two multiclass models\nof state-of-the-art deep convolutional neural networks (DCNNs) and vision\ntransformers (ViTs). DCNNs-based MoE and ViTs-based MoE schemes outperformed\nsimpler multiclass models and were tested on datasets from different hospitals\nand cancer types, where MoE using DCNNs yielded the best results. The proposed\nMoE yields 86.15% F1 and 97.93% sensitivity scores on unseen data, retaining\nless computational cost for inference than MoE using ViTs. This best\nperformance of MoEs comes with relatively higher computational trade-offs than\nmulticlass models. The proposed artifact detection pipeline will not only\nensure reliable CPATH predictions but may also provide quality control.\n", "date": "2024-03-12", "categories": []}
{"id": "2405.10254", "title": "PRISM: A Multi-Modal Generative Foundation Model for Slide-Level\n  Histopathology", "abstract": "  Foundation models in computational pathology promise to unlock the\ndevelopment of new clinical decision support systems and models for precision\nmedicine. However, there is a mismatch between most clinical analysis, which is\ndefined at the level of one or more whole slide images, and foundation models\nto date, which process the thousands of image tiles contained in a whole slide\nimage separately. The requirement to train a network to aggregate information\nacross a large number of tiles in multiple whole slide images limits these\nmodels' impact. In this work, we present a slide-level foundation model for\nH&E-stained histopathology, PRISM, that builds on Virchow tile embeddings and\nleverages clinical report text for pre-training. Using the tile embeddings,\nPRISM produces slide-level embeddings with the ability to generate clinical\nreports, resulting in several modes of use. Using text prompts, PRISM achieves\nzero-shot cancer detection and sub-typing performance approaching and\nsurpassing that of a supervised aggregator model. Using the slide embeddings\nwith linear classifiers, PRISM surpasses supervised aggregator models.\nFurthermore, we demonstrate that fine-tuning of the PRISM slide encoder yields\nlabel-efficient training for biomarker prediction, a task that typically\nsuffers from low availability of training data; an aggregator initialized with\nPRISM and trained on as little as 10% of the training data can outperform a\nsupervised baseline that uses all of the data.\n", "date": "2024-05-16", "categories": []}
{"id": "2405.13710", "title": "Optimizing Lymphocyte Detection in Breast Cancer Whole Slide Imaging\n  through Data-Centric Strategies", "abstract": "  Efficient and precise quantification of lymphocytes in histopathology slides\nis imperative for the characterization of the tumor microenvironment and\nimmunotherapy response insights. We developed a data-centric optimization\npipeline that attain great lymphocyte detection performance using an\noff-the-shelf YOLOv5 model, without any architectural modifications. Our\ncontribution that rely on strategic dataset augmentation strategies, includes\nnovel biological upsampling and custom visual cohesion transformations tailored\nto the unique properties of tissue imagery, and enables to dramatically improve\nmodel performances. Our optimization reveals a pivotal realization: given\nintensive customization, standard computational pathology models can achieve\nhigh-capability biomarker development, without increasing the architectural\ncomplexity. We showcase the interest of this approach in the context of breast\ncancer where our strategies lead to good lymphocyte detection performances,\nechoing a broadly impactful paradigm shift. Furthermore, our data curation\ntechniques enable crucial histological analysis benchmarks, highlighting\nimproved generalizable potential.\n", "date": "2024-05-22", "categories": []}
{"id": "2405.11618", "title": "Transcriptomics-guided Slide Representation Learning in Computational\n  Pathology", "abstract": "  Self-supervised learning (SSL) has been successful in building patch\nembeddings of small histology images (e.g., 224x224 pixels), but scaling these\nmodels to learn slide embeddings from the entirety of giga-pixel whole-slide\nimages (WSIs) remains challenging. Here, we leverage complementary information\nfrom gene expression profiles to guide slide representation learning using\nmultimodal pre-training. Expression profiles constitute highly detailed\nmolecular descriptions of a tissue that we hypothesize offer a strong\ntask-agnostic training signal for learning slide embeddings. Our slide and\nexpression (S+E) pre-training strategy, called Tangle, employs\nmodality-specific encoders, the outputs of which are aligned via contrastive\nlearning. Tangle was pre-trained on samples from three different organs: liver\n(n=6,597 S+E pairs), breast (n=1,020), and lung (n=1,012) from two different\nspecies (Homo sapiens and Rattus norvegicus). Across three independent test\ndatasets consisting of 1,265 breast WSIs, 1,946 lung WSIs, and 4,584 liver\nWSIs, Tangle shows significantly better few-shot performance compared to\nsupervised and SSL baselines. When assessed using prototype-based\nclassification and slide retrieval, Tangle also shows a substantial performance\nimprovement over all baselines. Code available at\nhttps://github.com/mahmoodlab/TANGLE.\n", "date": "2024-05-19", "categories": []}
{"id": "2405.11643", "title": "Morphological Prototyping for Unsupervised Slide Representation Learning\n  in Computational Pathology", "abstract": "  Representation learning of pathology whole-slide images (WSIs) has been has\nprimarily relied on weak supervision with Multiple Instance Learning (MIL).\nHowever, the slide representations resulting from this approach are highly\ntailored to specific clinical tasks, which limits their expressivity and\ngeneralization, particularly in scenarios with limited data. Instead, we\nhypothesize that morphological redundancy in tissue can be leveraged to build a\ntask-agnostic slide representation in an unsupervised fashion. To this end, we\nintroduce PANTHER, a prototype-based approach rooted in the Gaussian mixture\nmodel that summarizes the set of WSI patches into a much smaller set of\nmorphological prototypes. Specifically, each patch is assumed to have been\ngenerated from a mixture distribution, where each mixture component represents\na morphological exemplar. Utilizing the estimated mixture parameters, we then\nconstruct a compact slide representation that can be readily used for a wide\nrange of downstream tasks. By performing an extensive evaluation of PANTHER on\nsubtyping and survival tasks using 13 datasets, we show that 1) PANTHER\noutperforms or is on par with supervised MIL baselines and 2) the analysis of\nmorphological prototypes brings new qualitative and quantitative insights into\nmodel interpretability.\n", "date": "2024-05-19", "categories": []}
{"id": "2405.10871", "title": "BraTS-Path Challenge: Assessing Heterogeneous Histopathologic Brain\n  Tumor Sub-regions", "abstract": "  Glioblastoma is the most common primary adult brain tumor, with a grim\nprognosis - median survival of 12-18 months following treatment, and 4 months\notherwise. Glioblastoma is widely infiltrative in the cerebral hemispheres and\nwell-defined by heterogeneous molecular and micro-environmental histopathologic\nprofiles, which pose a major obstacle in treatment. Correctly diagnosing these\ntumors and assessing their heterogeneity is crucial for choosing the precise\ntreatment and potentially enhancing patient survival rates. In the\ngold-standard histopathology-based approach to tumor diagnosis, detecting\nvarious morpho-pathological features of distinct histology throughout digitized\ntissue sections is crucial. Such \"features\" include the presence of cellular\ntumor, geographic necrosis, pseudopalisading necrosis, areas abundant in\nmicrovascular proliferation, infiltration into the cortex, wide extension in\nsubcortical white matter, leptomeningeal infiltration, regions dense with\nmacrophages, and the presence of perivascular or scattered lymphocytes. With\nthese features in mind and building upon the main aim of the BraTS Cluster of\nChallenges https://www.synapse.org/brats2024, the goal of the BraTS-Path\nchallenge is to provide a systematically prepared comprehensive dataset and a\nbenchmarking environment to develop and fairly compare deep-learning models\ncapable of identifying tumor sub-regions of distinct histologic profile. These\nmodels aim to further our understanding of the disease and assist in the\ndiagnosis and grading of conditions in a consistent manner.\n", "date": "2024-05-17", "categories": []}
{"id": "2405.09851", "title": "Region of Interest Detection in Melanocytic Skin Tumor Whole Slide\n  Images -- Nevus & Melanoma", "abstract": "  Automated region of interest detection in histopathological image analysis is\na challenging and important topic with tremendous potential impact on clinical\npractice. The deep-learning methods used in computational pathology may help us\nto reduce costs and increase the speed and accuracy of cancer diagnosis. We\nstarted with the UNC Melanocytic Tumor Dataset cohort that contains 160\nhematoxylin and eosin whole-slide images of primary melanomas (86) and nevi\n(74). We randomly assigned 80% (134) as a training set and built an in-house\ndeep-learning method to allow for classification, at the slide level, of nevi\nand melanomas. The proposed method performed well on the other 20% (26) test\ndataset; the accuracy of the slide classification task was 92.3% and our model\nalso performed well in terms of predicting the region of interest annotated by\nthe pathologists, showing excellent performance of our model on melanocytic\nskin tumors. Even though we tested the experiments on the skin tumor dataset,\nour work could also be extended to other medical image detection problems to\nbenefit the clinical evaluation and diagnosis of different tumors.\n", "date": "2024-05-16", "categories": []}
{"id": "2401.15638", "title": "Cyto R-CNN and CytoNuke Dataset: Towards reliable whole-cell\n  segmentation in bright-field histological images", "abstract": "  Background: Cell segmentation in bright-field histological slides is a\ncrucial topic in medical image analysis. Having access to accurate segmentation\nallows researchers to examine the relationship between cellular morphology and\nclinical observations. Unfortunately, most segmentation methods known today are\nlimited to nuclei and cannot segmentate the cytoplasm.\n  Material & Methods: We present a new network architecture Cyto R-CNN that is\nable to accurately segment whole cells (with both the nucleus and the\ncytoplasm) in bright-field images. We also present a new dataset CytoNuke,\nconsisting of multiple thousand manual annotations of head and neck squamous\ncell carcinoma cells. Utilizing this dataset, we compared the performance of\nCyto R-CNN to other popular cell segmentation algorithms, including QuPath's\nbuilt-in algorithm, StarDist and Cellpose. To evaluate segmentation\nperformance, we calculated AP50, AP75 and measured 17 morphological and\nstaining-related features for all detected cells. We compared these\nmeasurements to the gold standard of manual segmentation using the\nKolmogorov-Smirnov test.\n  Results: Cyto R-CNN achieved an AP50 of 58.65% and an AP75 of 11.56% in\nwhole-cell segmentation, outperforming all other methods (QuPath\n$19.46/0.91\\%$; StarDist $45.33/2.32\\%$; Cellpose $31.85/5.61\\%$). Cell\nfeatures derived from Cyto R-CNN showed the best agreement to the gold standard\n($\\bar{D} = 0.15$) outperforming QuPath ($\\bar{D} = 0.22$), StarDist ($\\bar{D}\n= 0.25$) and Cellpose ($\\bar{D} = 0.23$).\n  Conclusion: Our newly proposed Cyto R-CNN architecture outperforms current\nalgorithms in whole-cell segmentation while providing more reliable cell\nmeasurements than any other model. This could improve digital pathology\nworkflows, potentially leading to improved diagnosis. Moreover, our published\ndataset can be used to develop further models in the future.\n", "date": "2024-01-28", "categories": []}
{"id": "2404.19113", "title": "Source-Free Domain Adaptation of Weakly-Supervised Object Localization\n  Models for Histology", "abstract": "  Given the emergence of deep learning, digital pathology has gained popularity\nfor cancer diagnosis based on histology images. Deep weakly supervised object\nlocalization (WSOL) models can be trained to classify histology images\naccording to cancer grade and identify regions of interest (ROIs) for\ninterpretation, using inexpensive global image-class annotations. A WSOL model\ninitially trained on some labeled source image data can be adapted using\nunlabeled target data in cases of significant domain shifts caused by\nvariations in staining, scanners, and cancer type. In this paper, we focus on\nsource-free (unsupervised) domain adaptation (SFDA), a challenging problem\nwhere a pre-trained source model is adapted to a new target domain without\nusing any source domain data for privacy and efficiency reasons. SFDA of WSOL\nmodels raises several challenges in histology, most notably because they are\nnot intended to adapt for both classification and localization tasks. In this\npaper, 4 state-of-the-art SFDA methods, each one representative of a main SFDA\nfamily, are compared for WSOL in terms of classification and localization\naccuracy. They are the SFDA-Distribution Estimation, Source HypOthesis\nTransfer, Cross-Domain Contrastive Learning, and Adaptively Domain Statistics\nAlignment. Experimental results on the challenging Glas (smaller, breast\ncancer) and Camelyon16 (larger, colon cancer) histology datasets indicate that\nthese SFDA methods typically perform poorly for localization after adaptation\nwhen optimized for classification.\n", "date": "2024-04-29", "categories": []}
{"id": "2401.02565", "title": "Demonstration of an Adversarial Attack Against a Multimodal Vision\n  Language Model for Pathology Imaging", "abstract": "  In the context of medical artificial intelligence, this study explores the\nvulnerabilities of the Pathology Language-Image Pretraining (PLIP) model, a\nVision Language Foundation model, under targeted attacks. Leveraging the Kather\nColon dataset with 7,180 H&E images across nine tissue types, our investigation\nemploys Projected Gradient Descent (PGD) adversarial perturbation attacks to\ninduce misclassifications intentionally. The outcomes reveal a 100% success\nrate in manipulating PLIP's predictions, underscoring its susceptibility to\nadversarial perturbations. The qualitative analysis of adversarial examples\ndelves into the interpretability challenges, shedding light on nuanced changes\nin predictions induced by adversarial manipulations. These findings contribute\ncrucial insights into the interpretability, domain adaptation, and\ntrustworthiness of Vision Language Models in medical imaging. The study\nemphasizes the pressing need for robust defenses to ensure the reliability of\nAI models. The source codes for this experiment can be found at\nhttps://github.com/jaiprakash1824/VLM_Adv_Attack.\n", "date": "2024-01-04", "categories": []}
{"id": "2405.04815", "title": "Proportion Estimation by Masked Learning from Label Proportion", "abstract": "  The PD-L1 rate, the number of PD-L1 positive tumor cells over the total\nnumber of all tumor cells, is an important metric for immunotherapy. This\nmetric is recorded as diagnostic information with pathological images. In this\npaper, we propose a proportion estimation method with a small amount of\ncell-level annotation and proportion annotation, which can be easily collected.\nSince the PD-L1 rate is calculated from only `tumor cells' and not using\n`non-tumor cells', we first detect tumor cells with a detection model. Then, we\nestimate the PD-L1 proportion by introducing a masking technique to `learning\nfrom label proportion.' In addition, we propose a weighted focal proportion\nloss to address data imbalance problems. Experiments using clinical data\ndemonstrate the effectiveness of our method. Our method achieved the best\nperformance in comparisons.\n", "date": "2024-05-08", "categories": []}
{"id": "2405.01688", "title": "Adapting Self-Supervised Learning for Computational Pathology", "abstract": "  Self-supervised learning (SSL) has emerged as a key technique for training\nnetworks that can generalize well to diverse tasks without task-specific\nsupervision. This property makes SSL desirable for computational pathology, the\nstudy of digitized images of tissues, as there are many target applications and\noften limited labeled training samples. However, SSL algorithms and models have\nbeen primarily developed in the field of natural images and whether their\nperformance can be improved by adaptation to particular domains remains an open\nquestion. In this work, we present an investigation of modifications to SSL for\npathology data, specifically focusing on the DINOv2 algorithm. We propose\nalternative augmentations, regularization functions, and position encodings\nmotivated by the characteristics of pathology images. We evaluate the impact of\nthese changes on several benchmarks to demonstrate the value of tailored\napproaches.\n", "date": "2024-05-02", "categories": []}
{"id": "2405.00908", "title": "Transformer-Based Self-Supervised Learning for Histopathological\n  Classification of Ischemic Stroke Clot Origin", "abstract": "  Background and Purpose: Identifying the thromboembolism source in ischemic\nstroke is crucial for treatment and secondary prevention yet is often\nundetermined. This study describes a self-supervised deep learning approach in\ndigital pathology of emboli for classifying ischemic stroke clot origin from\nhistopathological images. Methods: The dataset included whole slide images\n(WSI) from the STRIP AI Kaggle challenge, consisting of retrieved clots from\nischemic stroke patients following mechanical thrombectomy. Transformer-based\ndeep learning models were developed using transfer learning and self-supervised\npretraining for classifying WSI. Customizations included an attention pooling\nlayer, weighted loss function, and threshold optimization. Various model\narchitectures were tested and compared, and model performances were primarily\nevaluated using weighted logarithmic loss. Results: The model achieved a\nlogloss score of 0.662 in cross-validation and 0.659 on the test set. Different\nmodel backbones were compared, with the swin_large_patch4_window12_384 showed\nhigher performance. Thresholding techniques for clot origin classification were\nemployed to balance false positives and negatives. Conclusion: The study\ndemonstrates the extent of efficacy of transformer-based deep learning models\nin identifying ischemic stroke clot origins from histopathological images and\nemphasizes the need for refined modeling techniques specifically adapted to\nthrombi WSI. Further research is needed to improve model performance,\ninterpretability, validate its effectiveness. Future enhancement could include\nintegrating larger patient cohorts, advanced preprocessing strategies, and\nexploring ensemble multimodal methods for enhanced diagnostic accuracy.\n", "date": "2024-05-01", "categories": []}
{"id": "2301.02608", "title": "An interpretable machine learning system for colorectal cancer diagnosis\n  from pathology slides", "abstract": "  Considering the profound transformation affecting pathology practice, we\naimed to develop a scalable artificial intelligence (AI) system to diagnose\ncolorectal cancer from whole-slide images (WSI). For this, we propose a deep\nlearning (DL) system that learns from weak labels, a sampling strategy that\nreduces the number of training samples by a factor of six without compromising\nperformance, an approach to leverage a small subset of fully annotated samples,\nand a prototype with explainable predictions, active learning features and\nparallelisation. Noting some problems in the literature, this study is\nconducted with one of the largest WSI colorectal samples dataset with\napproximately 10,500 WSIs. Of these samples, 900 are testing samples.\nFurthermore, the robustness of the proposed method is assessed with two\nadditional external datasets (TCGA and PAIP) and a dataset of samples collected\ndirectly from the proposed prototype. Our proposed method predicts, for the\npatch-based tiles, a class based on the severity of the dysplasia and uses that\ninformation to classify the whole slide. It is trained with an interpretable\nmixed-supervision scheme to leverage the domain knowledge introduced by\npathologists through spatial annotations. The mixed-supervision scheme allowed\nfor an intelligent sampling strategy effectively evaluated in several different\nscenarios without compromising the performance. On the internal dataset, the\nmethod shows an accuracy of 93.44% and a sensitivity between positive\n(low-grade and high-grade dysplasia) and non-neoplastic samples of 0.996. On\nthe external test samples varied with TCGA being the most challenging dataset\nwith an overall accuracy of 84.91% and a sensitivity of 0.996.\n", "date": "2023-01-06", "categories": []}
{"id": "2404.17704", "title": "SPLICE -- Streamlining Digital Pathology Image Processing", "abstract": "  Digital pathology and the integration of artificial intelligence (AI) models\nhave revolutionized histopathology, opening new opportunities. With the\nincreasing availability of Whole Slide Images (WSIs), there's a growing demand\nfor efficient retrieval, processing, and analysis of relevant images from vast\nbiomedical archives. However, processing WSIs presents challenges due to their\nlarge size and content complexity. Full computer digestion of WSIs is\nimpractical, and processing all patches individually is prohibitively\nexpensive. In this paper, we propose an unsupervised patching algorithm,\nSequential Patching Lattice for Image Classification and Enquiry (SPLICE). This\nnovel approach condenses a histopathology WSI into a compact set of\nrepresentative patches, forming a \"collage\" of WSI while minimizing redundancy.\nSPLICE prioritizes patch quality and uniqueness by sequentially analyzing a WSI\nand selecting non-redundant representative features. We evaluated SPLICE for\nsearch and match applications, demonstrating improved accuracy, reduced\ncomputation time, and storage requirements compared to existing\nstate-of-the-art methods. As an unsupervised method, SPLICE effectively reduces\nstorage requirements for representing tissue images by 50%. This reduction\nenables numerous algorithms in computational pathology to operate much more\nefficiently, paving the way for accelerated adoption of digital pathology.\n", "date": "2024-04-26", "categories": []}
{"id": "2403.07592", "title": "Accurate Spatial Gene Expression Prediction by integrating\n  Multi-resolution features", "abstract": "  Recent advancements in Spatial Transcriptomics (ST) technology have\nfacilitated detailed gene expression analysis within tissue contexts. However,\nthe high costs and methodological limitations of ST necessitate a more robust\npredictive model. In response, this paper introduces TRIPLEX, a novel deep\nlearning framework designed to predict spatial gene expression from Whole Slide\nImages (WSIs). TRIPLEX uniquely harnesses multi-resolution features, capturing\ncellular morphology at individual spots, the local context around these spots,\nand the global tissue organization. By integrating these features through an\neffective fusion strategy, TRIPLEX achieves accurate gene expression\nprediction. Our comprehensive benchmark study, conducted on three public ST\ndatasets and supplemented with Visium data from 10X Genomics, demonstrates that\nTRIPLEX outperforms current state-of-the-art models in Mean Squared Error\n(MSE), Mean Absolute Error (MAE), and Pearson Correlation Coefficient (PCC).\nThe model's predictions align closely with ground truth gene expression\nprofiles and tumor annotations, underscoring TRIPLEX's potential in advancing\ncancer diagnosis and treatment.\n", "date": "2024-03-12", "categories": []}
{"id": "2404.16397", "title": "Deep Learning-based Prediction of Breast Cancer Tumor and Immune\n  Phenotypes from Histopathology", "abstract": "  The interactions between tumor cells and the tumor microenvironment (TME)\ndictate therapeutic efficacy of radiation and many systemic therapies in breast\ncancer. However, to date, there is not a widely available method to\nreproducibly measure tumor and immune phenotypes for each patient's tumor.\nGiven this unmet clinical need, we applied multiple instance learning (MIL)\nalgorithms to assess activity of ten biologically relevant pathways from the\nhematoxylin and eosin (H&E) slide of primary breast tumors. We employed\ndifferent feature extraction approaches and state-of-the-art model\narchitectures. Using binary classification, our models attained area under the\nreceiver operating characteristic (AUROC) scores above 0.70 for nearly all gene\nexpression pathways and on some cases, exceeded 0.80. Attention maps suggest\nthat our trained models recognize biologically relevant spatial patterns of\ncell sub-populations from H&E. These efforts represent a first step towards\ndeveloping computational H&E biomarkers that reflect facets of the TME and hold\npromise for augmenting precision oncology.\n", "date": "2024-04-25", "categories": []}
{"id": "2404.14434", "title": "DeeperHistReg: Robust Whole Slide Images Registration Framework", "abstract": "  DeeperHistReg is a software framework dedicated to registering whole slide\nimages (WSIs) acquired using multiple stains. It allows one to perform the\npreprocessing, initial alignment, and nonrigid registration of WSIs acquired\nusing multiple stains (e.g. hematoxylin \\& eosin, immunochemistry). The\nframework implements several state-of-the-art registration algorithms and\nprovides an interface to operate on arbitrary resolution of the WSIs (up to\n200k x 200k). The framework is extensible and new algorithms can be easily\nintegrated by other researchers. The framework is available both as a PyPI\npackage and as a Docker container.\n", "date": "2024-04-19", "categories": []}
{"id": "2304.06652", "title": "ProtoDiv: Prototype-guided Division of Consistent Pseudo-bags for\n  Whole-slide Image Classification", "abstract": "  Due to the limitations of inadequate Whole-Slide Image (WSI) samples with\nweak labels, pseudo-bag-based multiple instance learning (MIL) appears as a\nvibrant prospect in WSI classification. However, the pseudo-bag dividing\nscheme, often crucial for classification performance, is still an open topic\nworth exploring. Therefore, this paper proposes a novel scheme, ProtoDiv, using\na bag prototype to guide the division of WSI pseudo-bags. Rather than designing\ncomplex network architecture, this scheme takes a plugin-and-play approach to\nsafely augment WSI data for effective training while preserving sample\nconsistency. Furthermore, we specially devise an attention-based prototype that\ncould be optimized dynamically in training to adapt to a classification task.\nWe apply our ProtoDiv scheme on seven baseline models, and then carry out a\ngroup of comparison experiments on two public WSI datasets. Experiments confirm\nour ProtoDiv could usually bring obvious performance improvements to WSI\nclassification.\n", "date": "2023-04-13", "categories": []}
{"id": "2404.13097", "title": "DISC: Latent Diffusion Models with Self-Distillation from Separated\n  Conditions for Prostate Cancer Grading", "abstract": "  Latent Diffusion Models (LDMs) can generate high-fidelity images from noise,\noffering a promising approach for augmenting histopathology images for training\ncancer grading models. While previous works successfully generated\nhigh-fidelity histopathology images using LDMs, the generation of image tiles\nto improve prostate cancer grading has not yet been explored. Additionally,\nLDMs face challenges in accurately generating admixtures of multiple cancer\ngrades in a tile when conditioned by a tile mask. In this study, we train\nspecific LDMs to generate synthetic tiles that contain multiple Gleason Grades\n(GGs) by leveraging pixel-wise annotations in input tiles. We introduce a novel\nframework named Self-Distillation from Separated Conditions (DISC) that\ngenerates GG patterns guided by GG masks. Finally, we deploy a training\nframework for pixel-level and slide-level prostate cancer grading, where\nsynthetic tiles are effectively utilized to improve the cancer grading\nperformance of existing models. As a result, this work surpasses previous works\nin two domains: 1) our LDMs enhanced with DISC produce more accurate tiles in\nterms of GG patterns, and 2) our training scheme, incorporating synthetic data,\nsignificantly improves the generalization of the baseline model for prostate\ncancer grading, particularly in challenging cases of rare GG5, demonstrating\nthe potential of generative models to enhance cancer grading when data is\nlimited.\n", "date": "2024-04-19", "categories": []}
{"id": "2404.12986", "title": "Nuclei Instance Segmentation of Cryosectioned H&E Stained Histological\n  Images using Triple U-Net Architecture", "abstract": "  Nuclei instance segmentation is crucial in oncological diagnosis and cancer\npathology research. H&E stained images are commonly used for medical diagnosis,\nbut pre-processing is necessary before using them for image processing tasks.\nTwo principal pre-processing methods are formalin-fixed paraffin-embedded\nsamples (FFPE) and frozen tissue samples (FS). While FFPE is widely used, it is\ntime-consuming, while FS samples can be processed quickly. Analyzing H&E\nstained images derived from fast sample preparation, staining, and scanning can\npose difficulties due to the swift process, which can result in the degradation\nof image quality. This paper proposes a method that leverages the unique\noptical characteristics of H&E stained images. A three-branch U-Net\narchitecture has been implemented, where each branch contributes to the final\nsegmentation results. The process includes applying watershed algorithm to\nseparate overlapping regions and enhance accuracy. The Triple U-Net\narchitecture comprises an RGB branch, a Hematoxylin branch, and a Segmentation\nbranch. This study focuses on a novel dataset named CryoNuSeg. The results\nobtained through robust experiments outperform the state-of-the-art results\nacross various metrics. The benchmark score for this dataset is AJI 52.5 and PQ\n47.7, achieved through the implementation of U-Net Architecture. However, the\nproposed Triple U-Net architecture achieves an AJI score of 67.41 and PQ of\n50.56. The proposed architecture improves more on AJI than other evaluation\nmetrics, which further justifies the superiority of the Triple U-Net\narchitecture over the baseline U-Net model, as AJI is a more strict evaluation\nmetric. The use of the three-branch U-Net model, followed by watershed\npost-processing, significantly surpasses the benchmark scores, showing\nsubstantial improvement in the AJI score\n", "date": "2024-04-19", "categories": []}
{"id": "2404.11052", "title": "Supervised Contrastive Vision Transformer for Breast Histopathological\n  Image Classification", "abstract": "  Invasive ductal carcinoma (IDC) is the most prevalent form of breast cancer.\nBreast tissue histopathological examination is critical in diagnosing and\nclassifying breast cancer. Although existing methods have shown promising\nresults, there is still room for improvement in the classification accuracy and\ngeneralization of IDC using histopathology images. We present a novel approach,\nSupervised Contrastive Vision Transformer (SupCon-ViT), for improving the\nclassification of invasive ductal carcinoma in terms of accuracy and\ngeneralization by leveraging the inherent strengths and advantages of both\ntransfer learning, i.e., pre-trained vision transformer, and supervised\ncontrastive learning. Our results on a benchmark breast cancer dataset\ndemonstrate that SupCon-Vit achieves state-of-the-art performance in IDC\nclassification, with an F1-score of 0.8188, precision of 0.7692, and\nspecificity of 0.8971, outperforming existing methods. In addition, the\nproposed model demonstrates resilience in scenarios with minimal labeled data,\nmaking it highly efficient in real-world clinical settings where labelled data\nis limited. Our findings suggest that supervised contrastive learning in\nconjunction with pre-trained vision transformers appears to be a viable\nstrategy for an accurate classification of IDC, thus paving the way for a more\nefficient and reliable diagnosis of breast cancer through histopathological\nimage analysis.\n", "date": "2024-04-17", "categories": []}
{"id": "2404.01446", "title": "Finding Regions of Interest in Whole Slide Images Using Multiple\n  Instance Learning", "abstract": "  Whole Slide Images (WSI), obtained by high-resolution digital scanning of\nmicroscope slides at multiple scales, are the cornerstone of modern Digital\nPathology. However, they represent a particular challenge to\nAI-based/AI-mediated analysis because pathology labeling is typically done at\nslide-level, instead of tile-level. It is not just that medical diagnostics is\nrecorded at the specimen level, the detection of oncogene mutation is also\nexperimentally obtained, and recorded by initiatives like The Cancer Genome\nAtlas (TCGA), at the slide level. This configures a dual challenge: a)\naccurately predicting the overall cancer phenotype and b) finding out what\ncellular morphologies are associated with it at the tile level. To address\nthese challenges, a weakly supervised Multiple Instance Learning (MIL) approach\nwas explored for two prevalent cancer types, Invasive Breast Carcinoma\n(TCGA-BRCA) and Lung Squamous Cell Carcinoma (TCGA-LUSC). This approach was\nexplored for tumor detection at low magnification levels and TP53 mutations at\nvarious levels. Our results show that a novel additive implementation of MIL\nmatched the performance of reference implementation (AUC 0.96), and was only\nslightly outperformed by Attention MIL (AUC 0.97). More interestingly from the\nperspective of the molecular pathologist, these different AI architectures\nidentify distinct sensitivities to morphological features (through the\ndetection of Regions of Interest, RoI) at different amplification levels.\nTellingly, TP53 mutation was most sensitive to features at the higher\napplications where cellular morphology is resolved.\n", "date": "2024-04-01", "categories": []}
{"id": "2305.05006", "title": "Synthesis of Annotated Colorectal Cancer Tissue Images from Gland Layout", "abstract": "  Generating realistic tissue images with annotations is a challenging task\nthat is important in many computational histopathology applications.\nSynthetically generated images and annotations are valuable for training and\nevaluating algorithms in this domain. To address this, we propose an\ninteractive framework generating pairs of realistic colorectal cancer histology\nimages with corresponding glandular masks from glandular structure layouts. The\nframework accurately captures vital features like stroma, goblet cells, and\nglandular lumen. Users can control gland appearance by adjusting parameters\nsuch as the number of glands, their locations, and sizes. The generated images\nexhibit good Frechet Inception Distance (FID) scores compared to the\nstate-of-the-art image-to-image translation model. Additionally, we demonstrate\nthe utility of our synthetic annotations for evaluating gland segmentation\nalgorithms. Furthermore, we present a methodology for constructing glandular\nmasks using advanced deep generative models, such as latent diffusion models.\nThese masks enable tissue image generation through a residual encoder-decoder\nnetwork.\n", "date": "2023-05-08", "categories": []}
{"id": "2404.03188", "title": "Classification of Nasopharyngeal Cases using DenseNet Deep Learning\n  Architecture", "abstract": "  Nasopharyngeal carcinoma (NPC) is one of the understudied yet deadliest\ncancers in South East Asia. In Malaysia, the prevalence is identified mainly in\nSarawak, among the ethnic of Bidayuh. NPC is often late-diagnosed because it is\nasymptomatic at the early stage. There are several tissue representations from\nthe nasopharynx biopsy, such as nasopharyngeal inflammation (NPI), lymphoid\nhyperplasia (LHP), nasopharyngeal carcinoma (NPC) and normal tissue. This paper\nis our first initiative to identify the difference between NPC, NPI and normal\ncases. Seven whole slide images (WSIs) with gigapixel resolutions from seven\ndifferent patients and two hospitals were experimented with using two test\nsetups, consisting of a different set of images. The tissue regions are patched\ninto smaller blocks and classified using DenseNet architecture with 21 dense\nlayers. Two tests are carried out, each for proof of concept (Test 1) and\nreal-test scenario (Test 2). The accuracy achieved for NPC class is 94.8% for\nTest 1 and 67.0% for Test 2.\n", "date": "2024-04-04", "categories": []}
{"id": "2302.06089", "title": "Federated attention consistent learning models for prostate cancer\n  diagnosis and Gleason grading", "abstract": "  Artificial intelligence (AI) holds significant promise in transforming\nmedical imaging, enhancing diagnostics, and refining treatment strategies.\nHowever, the reliance on extensive multicenter datasets for training AI models\nposes challenges due to privacy concerns. Federated learning provides a\nsolution by facilitating collaborative model training across multiple centers\nwithout sharing raw data. This study introduces a federated\nattention-consistent learning (FACL) framework to address challenges associated\nwith large-scale pathological images and data heterogeneity. FACL enhances\nmodel generalization by maximizing attention consistency between local clients\nand the server model. To ensure privacy and validate robustness, we\nincorporated differential privacy by introducing noise during parameter\ntransfer. We assessed the effectiveness of FACL in cancer diagnosis and Gleason\ngrading tasks using 19,461 whole-slide images of prostate cancer from multiple\ncenters. In the diagnosis task, FACL achieved an area under the curve (AUC) of\n0.9718, outperforming seven centers with an average AUC of 0.9499 when\ncategories are relatively balanced. For the Gleason grading task, FACL attained\na Kappa score of 0.8463, surpassing the average Kappa score of 0.7379 from six\ncenters. In conclusion, FACL offers a robust, accurate, and cost-effective AI\ntraining model for prostate cancer pathology while maintaining effective data\nsafeguards.\n", "date": "2023-02-13", "categories": []}
{"id": "2403.18134", "title": "Integrative Graph-Transformer Framework for Histopathology Whole Slide\n  Image Representation and Classification", "abstract": "  In digital pathology, the multiple instance learning (MIL) strategy is widely\nused in the weakly supervised histopathology whole slide image (WSI)\nclassification task where giga-pixel WSIs are only labeled at the slide level.\nHowever, existing attention-based MIL approaches often overlook contextual\ninformation and intrinsic spatial relationships between neighboring tissue\ntiles, while graph-based MIL frameworks have limited power to recognize the\nlong-range dependencies. In this paper, we introduce the integrative\ngraph-transformer framework that simultaneously captures the context-aware\nrelational features and global WSI representations through a novel Graph\nTransformer Integration (GTI) block. Specifically, each GTI block consists of a\nGraph Convolutional Network (GCN) layer modeling neighboring relations at the\nlocal instance level and an efficient global attention model capturing\ncomprehensive global information from extensive feature embeddings. Extensive\nexperiments on three publicly available WSI datasets: TCGA-NSCLC, TCGA-RCC and\nBRIGHT, demonstrate the superiority of our approach over current\nstate-of-the-art MIL methods, achieving an improvement of 1.0% to 2.6% in\naccuracy and 0.7%-1.6% in AUROC.\n", "date": "2024-03-26", "categories": []}
{"id": "2403.17255", "title": "Decoding the visual attention of pathologists to reveal their level of\n  expertise", "abstract": "  We present a method for classifying the expertise of a pathologist based on\nhow they allocated their attention during a cancer reading. We engage this\ndecoding task by developing a novel method for predicting the attention of\npathologists as they read whole-slide Images (WSIs) of prostate and make cancer\ngrade classifications. Our ground truth measure of a pathologists' attention is\nthe x, y and z (magnification) movement of their viewport as they navigated\nthrough WSIs during readings, and to date we have the attention behavior of 43\npathologists reading 123 WSIs. These data revealed that specialists have higher\nagreement in both their attention and cancer grades compared to general\npathologists and residents, suggesting that sufficient information may exist in\ntheir attention behavior to classify their expertise level. To attempt this, we\ntrained a transformer-based model to predict the visual attention heatmaps of\nresident, general, and specialist (GU) pathologists during Gleason grading.\nBased solely on a pathologist's attention during a reading, our model was able\nto predict their level of expertise with 75.3%, 56.1%, and 77.2% accuracy,\nrespectively, better than chance and baseline models. Our model therefore\nenables a pathologist's expertise level to be easily and objectively evaluated,\nimportant for pathology training and competency assessment. Tools developed\nfrom our model could also be used to help pathology trainees learn how to read\nWSIs like an expert.\n", "date": "2024-03-25", "categories": []}
{"id": "2403.16695", "title": "Assessing the Performance of Deep Learning for Automated Gleason Grading\n  in Prostate Cancer", "abstract": "  Prostate cancer is a dominant health concern calling for advanced diagnostic\ntools. Utilizing digital pathology and artificial intelligence, this study\nexplores the potential of 11 deep neural network architectures for automated\nGleason grading in prostate carcinoma focusing on comparing traditional and\nrecent architectures. A standardized image classification pipeline, based on\nthe AUCMEDI framework, facilitated robust evaluation using an in-house dataset\nconsisting of 34,264 annotated tissue tiles. The results indicated varying\nsensitivity across architectures, with ConvNeXt demonstrating the strongest\nperformance. Notably, newer architectures achieved superior performance, even\nthough with challenges in differentiating closely related Gleason grades. The\nConvNeXt model was capable of learning a balance between complexity and\ngeneralizability. Overall, this study lays the groundwork for enhanced Gleason\ngrading systems, potentially improving diagnostic efficiency for prostate\ncancer.\n", "date": "2024-03-25", "categories": []}
{"id": "2401.08699", "title": "On Image Search in Histopathology", "abstract": "  Pathology images of histopathology can be acquired from camera-mounted\nmicroscopes or whole slide scanners. Utilizing similarity calculations to match\npatients based on these images holds significant potential in research and\nclinical contexts. Recent advancements in search technologies allow for\nimplicit quantification of tissue morphology across diverse primary sites,\nfacilitating comparisons and enabling inferences about diagnosis, and\npotentially prognosis, and predictions for new patients when compared against a\ncurated database of diagnosed and treated cases. In this paper, we\ncomprehensively review the latest developments in image search technologies for\nhistopathology, offering a concise overview tailored for computational\npathology researchers seeking effective, fast and efficient image search\nmethods in their work.\n", "date": "2024-01-14", "categories": []}
{"id": "2403.15068", "title": "Integrating multiscale topology in digital pathology with pyramidal\n  graph convolutional networks", "abstract": "  Graph convolutional networks (GCNs) have emerged as a powerful alternative to\nmultiple instance learning with convolutional neural networks in digital\npathology, offering superior handling of structural information across various\nspatial ranges - a crucial aspect of learning from gigapixel H&E-stained whole\nslide images (WSI). However, graph message-passing algorithms often suffer from\noversmoothing when aggregating a large neighborhood. Hence, effective modeling\nof multi-range interactions relies on the careful construction of the graph.\nOur proposed multi-scale GCN (MS-GCN) tackles this issue by leveraging\ninformation across multiple magnification levels in WSIs. MS-GCN enables the\nsimultaneous modeling of long-range structural dependencies at lower\nmagnifications and high-resolution cellular details at higher magnifications,\nakin to analysis pipelines usually conducted by pathologists. The\narchitecture's unique configuration allows for the concurrent modeling of\nstructural patterns at lower magnifications and detailed cellular features at\nhigher ones, while also quantifying the contribution of each magnification\nlevel to the prediction. Through testing on different datasets, MS-GCN\ndemonstrates superior performance over existing single-magnification GCN\nmethods. The enhancement in performance and interpretability afforded by our\nmethod holds promise for advancing computational pathology models, especially\nin tasks requiring extensive spatial context.\n", "date": "2024-03-22", "categories": []}
{"id": "2402.19286", "title": "PrPSeg: Universal Proposition Learning for Panoramic Renal Pathology\n  Segmentation", "abstract": "  Understanding the anatomy of renal pathology is crucial for advancing disease\ndiagnostics, treatment evaluation, and clinical research. The complex kidney\nsystem comprises various components across multiple levels, including regions\n(cortex, medulla), functional units (glomeruli, tubules), and cells (podocytes,\nmesangial cells in glomerulus). Prior studies have predominantly overlooked the\nintricate spatial interrelations among objects from clinical knowledge. In this\nresearch, we introduce a novel universal proposition learning approach, called\npanoramic renal pathology segmentation (PrPSeg), designed to segment\ncomprehensively panoramic structures within kidney by integrating extensive\nknowledge of kidney anatomy.\n  In this paper, we propose (1) the design of a comprehensive universal\nproposition matrix for renal pathology, facilitating the incorporation of\nclassification and spatial relationships into the segmentation process; (2) a\ntoken-based dynamic head single network architecture, with the improvement of\nthe partial label image segmentation and capability for future data\nenlargement; and (3) an anatomy loss function, quantifying the inter-object\nrelationships across the kidney.\n", "date": "2024-02-29", "categories": []}
{"id": "2403.14489", "title": "Adversary-Robust Graph-Based Learning of WSIs", "abstract": "  Enhancing the robustness of deep learning models against adversarial attacks\nis crucial, especially in critical domains like healthcare where significant\nfinancial interests heighten the risk of such attacks. Whole slide images\n(WSIs) are high-resolution, digitized versions of tissue samples mounted on\nglass slides, scanned using sophisticated imaging equipment. The digital\nanalysis of WSIs presents unique challenges due to their gigapixel size and\nmulti-resolution storage format. In this work, we aim at improving the\nrobustness of cancer Gleason grading classification systems against adversarial\nattacks, addressing challenges at both the image and graph levels. As regards\nthe proposed algorithm, we develop a novel and innovative graph-based model\nwhich utilizes GNN to extract features from the graph representation of WSIs. A\ndenoising module, along with a pooling layer is incorporated to manage the\nimpact of adversarial attacks on the WSIs. The process concludes with a\ntransformer module that classifies various grades of prostate cancer based on\nthe processed data. To assess the effectiveness of the proposed method, we\nconducted a comparative analysis using two scenarios. Initially, we trained and\ntested the model without the denoiser using WSIs that had not been exposed to\nany attack. We then introduced a range of attacks at either the image or graph\nlevel and processed them through the proposed network. The performance of the\nmodel was evaluated in terms of accuracy and kappa scores. The results from\nthis comparison showed a significant improvement in cancer diagnosis accuracy,\nhighlighting the robustness and efficiency of the proposed method in handling\nadversarial challenges in the context of medical imaging.\n", "date": "2024-03-21", "categories": []}
{"id": "2401.16355", "title": "PathMMU: A Massive Multimodal Expert-Level Benchmark for Understanding\n  and Reasoning in Pathology", "abstract": "  The emergence of large multimodal models has unlocked remarkable potential in\nAI, particularly in pathology. However, the lack of specialized, high-quality\nbenchmark impeded their development and precise evaluation. To address this, we\nintroduce PathMMU, the largest and highest-quality expert-validated pathology\nbenchmark for Large Multimodal Models (LMMs). It comprises 33,428 multimodal\nmulti-choice questions and 24,067 images from various sources, each accompanied\nby an explanation for the correct answer. The construction of PathMMU harnesses\nGPT-4V's advanced capabilities, utilizing over 30,000 image-caption pairs to\nenrich captions and generate corresponding Q&As in a cascading process.\nSignificantly, to maximize PathMMU's authority, we invite seven pathologists to\nscrutinize each question under strict standards in PathMMU's validation and\ntest sets, while simultaneously setting an expert-level performance benchmark\nfor PathMMU. We conduct extensive evaluations, including zero-shot assessments\nof 14 open-sourced and 4 closed-sourced LMMs and their robustness to image\ncorruption. We also fine-tune representative LMMs to assess their adaptability\nto PathMMU. The empirical findings indicate that advanced LMMs struggle with\nthe challenging PathMMU benchmark, with the top-performing LMM, GPT-4V,\nachieving only a 49.8% zero-shot performance, significantly lower than the\n71.8% demonstrated by human pathologists. After fine-tuning, significantly\nsmaller open-sourced LMMs can outperform GPT-4V but still fall short of the\nexpertise shown by pathologists. We hope that the PathMMU will offer valuable\ninsights and foster the development of more specialized, next-generation LMMs\nfor pathology.\n", "date": "2024-01-29", "categories": []}
{"id": "2403.12090", "title": "Foundation Models and Information Retrieval in Digital Pathology", "abstract": "  The paper reviews the state-of-the-art of foundation models, LLMs, generative\nAI, information retrieval and CBIR in digital pathology\n", "date": "2024-03-13", "categories": []}
{"id": "2403.12537", "title": "Prompt-Guided Adaptive Model Transformation for Whole Slide Image\n  Classification", "abstract": "  Multiple instance learning (MIL) has emerged as a popular method for\nclassifying histopathology whole slide images (WSIs). Existing approaches\ntypically rely on frozen pre-trained models to extract instance features,\nneglecting the substantial domain shift between pre-training natural and\nhistopathological images. To address this issue, we propose PAMT, a novel\nPrompt-guided Adaptive Model Transformation framework that enhances MIL\nclassification performance by seamlessly adapting pre-trained models to the\nspecific characteristics of histopathology data. To capture the intricate\nhistopathology distribution, we introduce Representative Patch Sampling (RPS)\nand Prototypical Visual Prompt (PVP) to reform the input data, building a\ncompact while informative representation. Furthermore, to narrow the domain\ngap, we introduce Adaptive Model Transformation (AMT) that integrates adapter\nblocks within the feature extraction pipeline, enabling the pre-trained models\nto learn domain-specific features. We rigorously evaluate our approach on two\npublicly available datasets, Camelyon16 and TCGA-NSCLC, showcasing substantial\nimprovements across various MIL models. Our findings affirm the potential of\nPAMT to set a new benchmark in WSI classification, underscoring the value of a\ntargeted reprogramming approach.\n", "date": "2024-03-19", "categories": []}
{"id": "2403.10858", "title": "RetMIL: Retentive Multiple Instance Learning for Histopathological Whole\n  Slide Image Classification", "abstract": "  Histopathological whole slide image (WSI) analysis with deep learning has\nbecome a research focus in computational pathology. The current paradigm is\nmainly based on multiple instance learning (MIL), in which approaches with\nTransformer as the backbone are well discussed. These methods convert WSI tasks\ninto sequence tasks by representing patches as tokens in the WSI sequence.\nHowever, the feature complexity brought by high heterogeneity and the\nultra-long sequences brought by gigapixel size makes Transformer-based MIL\nsuffer from the challenges of high memory consumption, slow inference speed,\nand lack of performance. To this end, we propose a retentive MIL method called\nRetMIL, which processes WSI sequences through hierarchical feature propagation\nstructure. At the local level, the WSI sequence is divided into multiple\nsubsequences. Tokens of each subsequence are updated through a parallel linear\nretention mechanism and aggregated utilizing an attention layer. At the global\nlevel, subsequences are fused into a global sequence, then updated through a\nserial retention mechanism, and finally the slide-level representation is\nobtained through a global attention pooling. We conduct experiments on two\npublic CAMELYON and BRACS datasets and an public-internal LUNG dataset,\nconfirming that RetMIL not only achieves state-of-the-art performance but also\nsignificantly reduces computational overhead. Our code will be accessed\nshortly.\n", "date": "2024-03-16", "categories": []}
{"id": "2403.11135", "title": "A lightweight deep learning pipeline with DRDA-Net and MobileNet for\n  breast cancer classification", "abstract": "  Accurate and early detection of breast cancer is essential for successful\ntreatment. This paper introduces a novel deep-learning approach for improved\nbreast cancer classification in histopathological images, a crucial step in\ndiagnosis. Our method hinges on the Dense Residual Dual-Shuffle Attention\nNetwork (DRDA-Net), inspired by ShuffleNet's efficient architecture. DRDA-Net\nachieves exceptional accuracy across various magnification levels on the\nBreaKHis dataset, a breast cancer histopathology analysis benchmark. However,\nfor real-world deployment, computational efficiency is paramount. We integrate\na pre-trained MobileNet model renowned for its lightweight design to address\ncomputational. MobileNet ensures fast execution even on devices with limited\nresources without sacrificing performance. This combined approach offers a\npromising solution for accurate breast cancer diagnosis, paving the way for\nfaster and more accessible screening procedures.\n", "date": "2024-03-17", "categories": []}
{"id": "2403.11211", "title": "RCdpia: A Renal Carcinoma Digital Pathology Image Annotation dataset\n  based on pathologists", "abstract": "  The annotation of digital pathological slide data for renal cell carcinoma is\nof paramount importance for correct diagnosis of artificial intelligence models\ndue to the heterogeneous nature of the tumor. This process not only facilitates\na deeper understanding of renal cell cancer heterogeneity but also aims to\nminimize noise in the data for more accurate studies. To enhance the\napplicability of the data, two pathologists were enlisted to meticulously\ncurate, screen, and label a kidney cancer pathology image dataset from The\nCancer Genome Atlas Program (TCGA) database. Subsequently, a Resnet model was\ndeveloped to validate the annotated dataset against an additional dataset from\nthe First Affiliated Hospital of Zhejiang University. Based on these results,\nwe have meticulously compiled the TCGA digital pathological dataset with\nindependent labeling of tumor regions and adjacent areas (RCdpia), which\nincludes 109 cases of kidney chromophobe cell carcinoma, 486 cases of kidney\nclear cell carcinoma, and 292 cases of kidney papillary cell carcinoma. This\ndataset is now publicly accessible at http://39.171.241.18:8888/RCdpia/.\nFurthermore, model analysis has revealed significant discrepancies in\npredictive outcomes when applying the same model to datasets from different\ncenters. Leveraging the RCdpia, we can now develop more precise digital\npathology artificial intelligence models for tasks such as normalization,\nclassification, and segmentation. These advancements underscore the potential\nfor more nuanced and accurate AI applications in the field of digital\npathology.\n", "date": "2024-03-17", "categories": []}
{"id": "2312.01435", "title": "Automatic Report Generation for Histopathology images using pre-trained\n  Vision Transformers and BERT", "abstract": "  Deep learning for histopathology has been successfully used for disease\nclassification, image segmentation and more. However, combining image and text\nmodalities using current state-of-the-art (SOTA) methods has been a challenge\ndue to the high resolution of histopathology images. Automatic report\ngeneration for histopathology images is one such challenge. In this work, we\nshow that using an existing pre-trained Vision Transformer (ViT) to encode\n4096x4096 sized patches of the Whole Slide Image (WSI) and a pre-trained\nBidirectional Encoder Representations from Transformers (BERT) model for\nlanguage modeling-based decoder for report generation, we can build a\nperformant and portable report generation mechanism that takes into account the\nwhole high resolution image. Our method allows us to not only generate and\nevaluate captions that describe the image, but also helps us classify the image\ninto tissue types and the gender of the patient as well. Our best performing\nmodel achieves a 89.52% accuracy in Tissue Type classification with a BLEU-4\nscore of 0.12 in our caption generation task.\n", "date": "2023-12-03", "categories": []}
{"id": "2211.00646", "title": "Learning Melanocytic Cell Masks from Adjacent Stained Tissue", "abstract": "  Melanoma is one of the most aggressive forms of skin cancer, causing a large\nproportion of skin cancer deaths. However, melanoma diagnoses by pathologists\nshows low interrater reliability. As melanoma is a cancer of the melanocyte,\nthere is a clear need to develop a melanocytic cell segmentation tool that is\nagnostic to pathologist variability and automates pixel-level annotation.\nGigapixel-level pathologist labeling, however, is impractical. Herein, we\npropose a means to train deep neural networks for melanocytic cell segmentation\nfrom hematoxylin and eosin (H&E) stained sections and paired\nimmunohistochemistry (IHC) of adjacent tissue sections, achieving a mean IOU of\n0.64 despite imperfect ground-truth labels.\n", "date": "2022-11-01", "categories": []}
{"id": "2401.01386", "title": "Tissue Artifact Segmentation and Severity Analysis for Automated\n  Diagnosis Using Whole Slide Images", "abstract": "  Traditionally, pathological analysis and diagnosis are performed by manually\neyeballing glass slide specimens under a microscope by an expert. The whole\nslide image is the digital specimen produced from the glass slide. Whole slide\nimage enabled specimens to be observed on a computer screen and led to\ncomputational pathology where computer vision and artificial intelligence are\nutilized for automated analysis and diagnosis. With the current computational\nadvancement, the entire whole slide image can be analyzed autonomously without\nhuman supervision. However, the analysis could fail or lead to wrong diagnosis\nif the whole slide image is affected by tissue artifacts such as tissue fold or\nair bubbles depending on the severity. Existing artifact detection methods rely\non experts for severity assessment to eliminate artifact affected regions from\nthe analysis. This process is time consuming, exhausting and undermines the\ngoal of automated analysis or removal of artifacts without evaluating their\nseverity, which could result in the loss of diagnostically important data.\nTherefore, it is necessary to detect artifacts and then assess their severity\nautomatically. In this paper, we propose a system that incorporates severity\nevaluation with artifact detection utilizing convolutional neural networks. The\nproposed system uses DoubleUNet to segment artifacts and an ensemble network of\nsix fine tuned convolutional neural network models to determine severity. This\nmethod outperformed current state of the art in accuracy by 9 percent for\nartifact segmentation and achieved a strong correlation of 97 percent with the\nevaluation of pathologists for severity assessment. The robustness of the\nsystem was demonstrated using our proposed heterogeneous dataset and practical\nusability was ensured by integrating it with an automated analysis system.\n", "date": "2024-01-01", "categories": []}
{"id": "2403.07939", "title": "Dynamic Policy-Driven Adaptive Multi-Instance Learning for Whole Slide\n  Image Classification", "abstract": "  Multi-Instance Learning (MIL) has shown impressive performance for\nhistopathology whole slide image (WSI) analysis using bags or pseudo-bags. It\ninvolves instance sampling, feature representation, and decision-making.\nHowever, existing MIL-based technologies at least suffer from one or more of\nthe following problems: 1) requiring high storage and intensive pre-processing\nfor numerous instances (sampling); 2) potential over-fitting with limited\nknowledge to predict bag labels (feature representation); 3) pseudo-bag counts\nand prior biases affect model robustness and generalizability\n(decision-making). Inspired by clinical diagnostics, using the past sampling\ninstances can facilitate the final WSI analysis, but it is barely explored in\nprior technologies. To break free these limitations, we integrate the dynamic\ninstance sampling and reinforcement learning into a unified framework to\nimprove the instance selection and feature aggregation, forming a novel Dynamic\nPolicy Instance Selection (DPIS) scheme for better and more credible\ndecision-making. Specifically, the measurement of feature distance and reward\nfunction are employed to boost continuous instance sampling. To alleviate the\nover-fitting, we explore the latent global relations among instances for more\nrobust and discriminative feature representation while establishing reward and\npunishment mechanisms to correct biases in pseudo-bags using contrastive\nlearning. These strategies form the final Dynamic Policy-Driven Adaptive\nMulti-Instance Learning (PAMIL) method for WSI tasks. Extensive experiments\nreveal that our PAMIL method outperforms the state-of-the-art by 3.8\\% on\nCAMELYON16 and 4.4\\% on TCGA lung cancer datasets.\n", "date": "2024-03-09", "categories": []}
{"id": "2311.08359", "title": "Rotation-Agnostic Image Representation Learning for Digital Pathology", "abstract": "  This paper addresses complex challenges in histopathological image analysis\nthrough three key contributions. Firstly, it introduces a fast patch selection\nmethod, FPS, for whole-slide image (WSI) analysis, significantly reducing\ncomputational cost while maintaining accuracy. Secondly, it presents PathDino,\na lightweight histopathology feature extractor with a minimal configuration of\nfive Transformer blocks and only 9 million parameters, markedly fewer than\nalternatives. Thirdly, it introduces a rotation-agnostic representation\nlearning paradigm using self-supervised learning, effectively mitigating\noverfitting. We also show that our compact model outperforms existing\nstate-of-the-art histopathology-specific vision transformers on 12 diverse\ndatasets, including both internal datasets spanning four sites (breast, liver,\nskin, and colorectal) and seven public datasets (PANDA, CAMELYON16, BRACS,\nDigestPath, Kather, PanNuke, and WSSS4LUAD). Notably, even with a training\ndataset of 6 million histopathology patches from The Cancer Genome Atlas\n(TCGA), our approach demonstrates an average 8.5% improvement in patch-level\nmajority vote performance. These contributions provide a robust framework for\nenhancing image analysis in digital pathology, rigorously validated through\nextensive evaluation. Project Page:\nhttps://kimialabmayo.github.io/PathDino-Page/\n", "date": "2023-11-14", "categories": []}
{"id": "2401.15886", "title": "Grey Level Texture Features for Segmentation of Chromogenic Dye RNAscope\n  From Breast Cancer Tissue", "abstract": "  Chromogenic RNAscope dye and haematoxylin staining of cancer tissue\nfacilitates diagnosis of the cancer type and subsequent treatment, and fits\nwell into existing pathology workflows. However, manual quantification of the\nRNAscope transcripts (dots), which signify gene expression, is prohibitively\ntime consuming. In addition, there is a lack of verified supporting methods for\nquantification and analysis. This paper investigates the usefulness of grey\nlevel texture features for automatically segmenting and classifying the\npositions of RNAscope transcripts from breast cancer tissue. Feature analysis\nshowed that a small set of grey level features, including Grey Level Dependence\nMatrix and Neighbouring Grey Tone Difference Matrix features, were well suited\nfor the task. The automated method performed similarly to expert annotators at\nidentifying the positions of RNAscope transcripts, with an F1-score of 0.571\ncompared to the expert inter-rater F1-score of 0.596. These results demonstrate\nthe potential of grey level texture features for automated quantification of\nRNAscope in the pathology workflow.\n", "date": "2024-01-29", "categories": []}
{"id": "2403.04558", "title": "Reducing self-supervised learning complexity improves weakly-supervised\n  classification performance in computational pathology", "abstract": "  Deep Learning models have been successfully utilized to extract clinically\nactionable insights from routinely available histology data. Generally, these\nmodels require annotations performed by clinicians, which are scarce and costly\nto generate. The emergence of self-supervised learning (SSL) methods remove\nthis barrier, allowing for large-scale analyses on non-annotated data. However,\nrecent SSL approaches apply increasingly expansive model architectures and\nlarger datasets, causing the rapid escalation of data volumes, hardware\nprerequisites, and overall expenses, limiting access to these resources to few\ninstitutions. Therefore, we investigated the complexity of contrastive SSL in\ncomputational pathology in relation to classification performance with the\nutilization of consumer-grade hardware. Specifically, we analyzed the effects\nof adaptations in data volume, architecture, and algorithms on downstream\nclassification tasks, emphasizing their impact on computational resources. We\ntrained breast cancer foundation models on a large public patient cohort and\nvalidated them on various downstream classification tasks in a weakly\nsupervised manner on two external public patient cohorts. Our experiments\ndemonstrate that we can improve downstream classification performance whilst\nreducing SSL training duration by 90%. In summary, we propose a set of\nadaptations which enable the utilization of SSL in computational pathology in\nnon-resource abundant environments.\n", "date": "2024-03-07", "categories": []}
{"id": "2403.07719", "title": "Dynamic Graph Representation with Knowledge-aware Attention for\n  Histopathology Whole Slide Image Analysis", "abstract": "  Histopathological whole slide images (WSIs) classification has become a\nfoundation task in medical microscopic imaging processing. Prevailing\napproaches involve learning WSIs as instance-bag representations, emphasizing\nsignificant instances but struggling to capture the interactions between\ninstances. Additionally, conventional graph representation methods utilize\nexplicit spatial positions to construct topological structures but restrict the\nflexible interaction capabilities between instances at arbitrary locations,\nparticularly when spatially distant. In response, we propose a novel dynamic\ngraph representation algorithm that conceptualizes WSIs as a form of the\nknowledge graph structure. Specifically, we dynamically construct neighbors and\ndirected edge embeddings based on the head and tail relationships between\ninstances. Then, we devise a knowledge-aware attention mechanism that can\nupdate the head node features by learning the joint attention score of each\nneighbor and edge. Finally, we obtain a graph-level embedding through the\nglobal pooling process of the updated head, serving as an implicit\nrepresentation for the WSI classification. Our end-to-end graph representation\nlearning approach has outperformed the state-of-the-art WSI analysis methods on\nthree TCGA benchmark datasets and in-house test sets. Our code is available at\nhttps://github.com/WonderLandxD/WiKG.\n", "date": "2024-03-12", "categories": []}
{"id": "2403.06349", "title": "MOAB: Multi-Modal Outer Arithmetic Block For Fusion Of Histopathological\n  Images And Genetic Data For Brain Tumor Grading", "abstract": "  Brain tumors are an abnormal growth of cells in the brain. They can be\nclassified into distinct grades based on their growth. Often grading is\nperformed based on a histological image and is one of the most significant\npredictors of a patients prognosis, the higher the grade, the more aggressive\nthe tumor. Correct diagnosis of a tumor grade remains challenging. Though\nhistopathological grading has been shown to be prognostic, results are subject\nto interobserver variability, even among experienced pathologists. Recently,\nthe World Health Organization reported that advances in molecular genetics have\nled to improvements in tumor classification. This paper seeks to integrate\nhistological images and genetic data for improved computer-aided diagnosis. We\npropose a novel Multi-modal Outer Arithmetic Block (MOAB) based on arithmetic\noperations to combine latent representations of the different modalities for\npredicting the tumor grade (Grade \\rom{2}, \\rom{3} and \\rom{4}). Extensive\nexperiments evaluate the effectiveness of our approach. By applying MOAB to The\nCancer Genome Atlas (TCGA) glioma dataset, we show that it can improve\nseparation between similar classes (Grade \\rom{2} and \\rom{3}) and outperform\nprior state-of-the-art grade classification techniques.\n", "date": "2024-03-11", "categories": []}
{"id": "2403.06545", "title": "ReStainGAN: Leveraging IHC to IF Stain Domain Translation for in-silico\n  Data Generation", "abstract": "  The creation of in-silico datasets can expand the utility of existing\nannotations to new domains with different staining patterns in computational\npathology. As such, it has the potential to significantly lower the cost\nassociated with building large and pixel precise datasets needed to train\nsupervised deep learning models. We propose a novel approach for the generation\nof in-silico immunohistochemistry (IHC) images by disentangling morphology\nspecific IHC stains into separate image channels in immunofluorescence (IF)\nimages. The proposed approach qualitatively and quantitatively outperforms\nbaseline methods as proven by training nucleus segmentation models on the\ncreated in-silico datasets.\n", "date": "2024-03-11", "categories": []}
{"id": "2403.06800", "title": "MambaMIL: Enhancing Long Sequence Modeling with Sequence Reordering in\n  Computational Pathology", "abstract": "  Multiple Instance Learning (MIL) has emerged as a dominant paradigm to\nextract discriminative feature representations within Whole Slide Images (WSIs)\nin computational pathology. Despite driving notable progress, existing MIL\napproaches suffer from limitations in facilitating comprehensive and efficient\ninteractions among instances, as well as challenges related to time-consuming\ncomputations and overfitting. In this paper, we incorporate the Selective Scan\nSpace State Sequential Model (Mamba) in Multiple Instance Learning (MIL) for\nlong sequence modeling with linear complexity, termed as MambaMIL. By\ninheriting the capability of vanilla Mamba, MambaMIL demonstrates the ability\nto comprehensively understand and perceive long sequences of instances.\nFurthermore, we propose the Sequence Reordering Mamba (SR-Mamba) aware of the\norder and distribution of instances, which exploits the inherent valuable\ninformation embedded within the long sequences. With the SR-Mamba as the core\ncomponent, MambaMIL can effectively capture more discriminative features and\nmitigate the challenges associated with overfitting and high computational\noverhead. Extensive experiments on two public challenging tasks across nine\ndiverse datasets demonstrate that our proposed framework performs favorably\nagainst state-of-the-art MIL methods. The code is released at\nhttps://github.com/isyangshu/MambaMIL.\n", "date": "2024-03-11", "categories": []}
{"id": "2402.15832", "title": "Multiple Instance Learning for Glioma Diagnosis using Hematoxylin and\n  Eosin Whole Slide Images: An Indian Cohort Study", "abstract": "  The effective management of brain tumors relies on precise typing, subtyping,\nand grading. This study advances patient care with findings from rigorous\nmultiple instance learning experimentations across various feature extractors\nand aggregators in brain tumor histopathology. It establishes new performance\nbenchmarks in glioma subtype classification across multiple datasets, including\na novel dataset focused on the Indian demographic (IPD- Brain), providing a\nvaluable resource for existing research. Using a ResNet-50, pretrained on\nhistopathology datasets for feature extraction, combined with the Double-Tier\nFeature Distillation (DTFD) feature aggregator, our approach achieves\nstate-of-the-art AUCs of 88.08 on IPD-Brain and 95.81 on the TCGA-Brain\ndataset, respectively, for three-way glioma subtype classification. Moreover,\nit establishes new benchmarks in grading and detecting IHC molecular biomarkers\n(IDH1R132H, TP53, ATRX, Ki-67) through H&E stained whole slide images for the\nIPD-Brain dataset. The work also highlights a significant correlation between\nthe model decision-making processes and the diagnostic reasoning of\npathologists, underscoring its capability to mimic professional diagnostic\nprocedures.\n", "date": "2024-02-24", "categories": []}
{"id": "2403.05351", "title": "Multiple Instance Learning with random sampling for Whole Slide Image\n  Classification", "abstract": "  In computational pathology, random sampling of patches during training of\nMultiple Instance Learning (MIL) methods is computationally efficient and\nserves as a regularization strategy. Despite its promising benefits, questions\nconcerning performance trends for varying sample sizes and its influence on\nmodel interpretability remain. Addressing these, we reach an optimal\nperformance enhancement of 1.7% using thirty percent of patches on the\nCAMELYON16 dataset, and 3.7% with only eight samples on the TUPAC16 dataset. We\nalso find interpretability effects are strongly dataset-dependent, with\ninterpretability impacted on CAMELYON16, while remaining unaffected on TUPAC16.\nThis reinforces that both the performance and interpretability relationships\nwith sampling are closely task-specific. End-to-end training with 1024 samples\nreveals improvements across both datasets compared to pre-extracted features,\nfurther highlighting the potential of this efficient approach.\n", "date": "2024-03-08", "categories": []}
{"id": "2403.03891", "title": "Joint multi-task learning improves weakly-supervised biomarker\n  prediction in computational pathology", "abstract": "  Deep Learning (DL) can predict biomarkers directly from digitized cancer\nhistology in a weakly-supervised setting. Recently, the prediction of\ncontinuous biomarkers through regression-based DL has seen an increasing\ninterest. Nonetheless, clinical decision making often requires a categorical\noutcome. Consequently, we developed a weakly-supervised joint multi-task\nTransformer architecture which has been trained and evaluated on four public\npatient cohorts for the prediction of two key predictive biomarkers,\nmicrosatellite instability (MSI) and homologous recombination deficiency (HRD),\ntrained with auxiliary regression tasks related to the tumor microenvironment.\nMoreover, we perform a comprehensive benchmark of 16 approaches of task\nbalancing for weakly-supervised joint multi-task learning in computational\npathology. Using our novel approach, we improve over the state-of-the-art area\nunder the receiver operating characteristic by +7.7% and +4.1%, as well as\nyielding better clustering of latent embeddings by +8% and +5% for the\nprediction of MSI and HRD in external cohorts, respectively.\n", "date": "2024-03-06", "categories": []}
{"id": "2402.19296", "title": "An AI based Digital Score of Tumour-Immune Microenvironment Predicts\n  Benefit to Maintenance Immunotherapy in Advanced Oesophagogastric\n  Adenocarcinoma", "abstract": "  Gastric and oesophageal (OG) cancers are the leading causes of cancer\nmortality worldwide. In OG cancers, recent studies have showed that PDL1 immune\ncheckpoint inhibitors (ICI) in combination with chemotherapy improves patient\nsurvival. However, our understanding of the tumour immune microenvironment in\nOG cancers remains limited. In this study, we interrogate multiplex\nimmunofluorescence (mIF) images taken from patients with advanced\nOesophagogastric Adenocarcinoma (OGA) who received first-line fluoropyrimidine\nand platinum-based chemotherapy in the PLATFORM trial (NCT02678182) to predict\nthe efficacy of the treatment and to explore the biological basis of patients\nresponding to maintenance durvalumab (PDL1 inhibitor). Our proposed Artificial\nIntelligence (AI) based marker successfully identified responder from\nnon-responder (p < 0.05) as well as those who could potentially benefit from\nICI with statistical significance (p < 0.05) for both progression free and\noverall survival. Our findings suggest that T cells that express FOXP3 seem to\nheavily influence the patient treatment response and survival outcome. We also\nobserved that higher levels of CD8+PD1+ cells are consistently linked to poor\nprognosis for both OS and PFS, regardless of ICI.\n", "date": "2024-02-29", "categories": []}
{"id": "2308.00920", "title": "Virtual histological staining of unlabeled autopsy tissue", "abstract": "  Histological examination is a crucial step in an autopsy; however, the\ntraditional histochemical staining of post-mortem samples faces multiple\nchallenges, including the inferior staining quality due to autolysis caused by\ndelayed fixation of cadaver tissue, as well as the resource-intensive nature of\nchemical staining procedures covering large tissue areas, which demand\nsubstantial labor, cost, and time. These challenges can become more pronounced\nduring global health crises when the availability of histopathology services is\nlimited, resulting in further delays in tissue fixation and more severe\nstaining artifacts. Here, we report the first demonstration of virtual staining\nof autopsy tissue and show that a trained neural network can rapidly transform\nautofluorescence images of label-free autopsy tissue sections into brightfield\nequivalent images that match hematoxylin and eosin (H&E) stained versions of\nthe same samples, eliminating autolysis-induced severe staining artifacts\ninherent in traditional histochemical staining of autopsied tissue. Our virtual\nH&E model was trained using >0.7 TB of image data and a data-efficient\ncollaboration scheme that integrates the virtual staining network with an image\nregistration network. The trained model effectively accentuated nuclear,\ncytoplasmic and extracellular features in new autopsy tissue samples that\nexperienced severe autolysis, such as COVID-19 samples never seen before, where\nthe traditional histochemical staining failed to provide consistent staining\nquality. This virtual autopsy staining technique can also be extended to\nnecrotic tissue, and can rapidly and cost-effectively generate artifact-free\nH&E stains despite severe autolysis and cell death, also reducing labor, cost\nand infrastructure requirements associated with the standard histochemical\nstaining.\n", "date": "2023-08-02", "categories": []}
{"id": "2402.13932", "title": "Tumor segmentation on whole slide images: training or prompting?", "abstract": "  Tumor segmentation stands as a pivotal task in cancer diagnosis. Given the\nimmense dimensions of whole slide images (WSI) in histology, deep learning\napproaches for WSI classification mainly operate at patch-wise or\nsuperpixel-wise level. However, these solutions often struggle to capture\nglobal WSI information and cannot directly generate the binary mask.\nDownsampling the WSI and performing semantic segmentation is another possible\napproach. While this method offers computational efficiency, it necessitates a\nlarge amount of annotated data since resolution reduction may lead to\ninformation loss. Visual prompting is a novel paradigm that allows the model to\nperform new tasks by making subtle modifications to the input space, rather\nthan adapting the model itself. Such approach has demonstrated promising\nresults on many computer vision tasks. In this paper, we show the efficacy of\nvisual prompting in the context of tumor segmentation for three distinct\norgans. In comparison to classical methods trained for this specific task, our\nfindings reveal that, with appropriate prompt examples, visual prompting can\nachieve comparable or better performance without extensive fine-tuning.\n", "date": "2024-02-21", "categories": []}
{"id": "2305.15072", "title": "PathAsst: A Generative Foundation AI Assistant Towards Artificial\n  General Intelligence of Pathology", "abstract": "  As advances in large language models (LLMs) and multimodal techniques\ncontinue to mature, the development of general-purpose multimodal large\nlanguage models (MLLMs) has surged, offering significant applications in\ninterpreting natural images. However, the field of pathology has largely\nremained untapped, particularly in gathering high-quality data and designing\ncomprehensive model frameworks. To bridge the gap in pathology MLLMs, we\npresent PathAsst, a multimodal generative foundation AI assistant to\nrevolutionize diagnostic and predictive analytics in pathology. The development\nof PathAsst involves three pivotal steps: data acquisition, CLIP model\nadaptation, and the training of PathAsst's multimodal generative capabilities.\nFirstly, we collect over 207K high-quality pathology image-text pairs from\nauthoritative sources. Leveraging the advanced power of ChatGPT, we generate\nover 180K instruction-following samples. Furthermore, we devise additional\ninstruction-following data specifically tailored for invoking eight\npathology-specific sub-models we prepared, allowing the PathAsst to effectively\ncollaborate with these models, enhancing its diagnostic ability. Secondly, by\nleveraging the collected data, we construct PathCLIP, a pathology-dedicated\nCLIP, to enhance PathAsst's capabilities in interpreting pathology images.\nFinally, we integrate PathCLIP with the Vicuna-13b and utilize\npathology-specific instruction-tuning data to enhance the multimodal generation\ncapacity of PathAsst and bolster its synergistic interactions with sub-models.\nThe experimental results of PathAsst show the potential of harnessing\nAI-powered generative foundation model to improve pathology diagnosis and\ntreatment processes.\n", "date": "2023-05-24", "categories": []}
{"id": "2304.00216", "title": "Cross-scale Multi-instance Learning for Pathological Image Diagnosis", "abstract": "  Analyzing high resolution whole slide images (WSIs) with regard to\ninformation across multiple scales poses a significant challenge in digital\npathology. Multi-instance learning (MIL) is a common solution for working with\nhigh resolution images by classifying bags of objects (i.e. sets of smaller\nimage patches). However, such processing is typically performed at a single\nscale (e.g., 20x magnification) of WSIs, disregarding the vital inter-scale\ninformation that is key to diagnoses by human pathologists. In this study, we\npropose a novel cross-scale MIL algorithm to explicitly aggregate inter-scale\nrelationships into a single MIL network for pathological image diagnosis. The\ncontribution of this paper is three-fold: (1) A novel cross-scale MIL (CS-MIL)\nalgorithm that integrates the multi-scale information and the inter-scale\nrelationships is proposed; (2) A toy dataset with scale-specific morphological\nfeatures is created and released to examine and visualize differential\ncross-scale attention; (3) Superior performance on both in-house and public\ndatasets is demonstrated by our simple cross-scale MIL strategy. The official\nimplementation is publicly available at https://github.com/hrlblab/CS-MIL.\n", "date": "2023-04-01", "categories": []}
{"id": "2402.09990", "title": "TIAViz: A Browser-based Visualization Tool for Computational Pathology\n  Models", "abstract": "  Digital pathology has gained significant traction in modern healthcare\nsystems. This shift from optical microscopes to digital imagery brings with it\nthe potential for improved diagnosis, efficiency, and the integration of AI\ntools into the pathologists workflow. A critical aspect of this is\nvisualization. Throughout the development of a machine learning (ML) model in\ndigital pathology, it is crucial to have flexible, openly available tools to\nvisualize models, from their outputs and predictions to the underlying\nannotations and images used to train or test a model. We introduce TIAViz, a\nPython-based visualization tool built into TIAToolbox which allows flexible,\ninteractive, fully zoomable overlay of a wide variety of information onto whole\nslide images, including graphs, heatmaps, segmentations, annotations and other\nWSIs. The UI is browser-based, allowing use either locally, on a remote\nmachine, or on a server to provide publicly available demos. This tool is open\nsource and is made available at:\nhttps://github.com/TissueImageAnalytics/tiatoolbox and via pip installation\n(pip install tiatoolbox) and conda as part of TIAToolbox.\n", "date": "2024-02-15", "categories": []}
{"id": "2302.09833", "title": "Domain-Specific Pre-training Improves Confidence in Whole Slide Image\n  Classification", "abstract": "  Whole Slide Images (WSIs) or histopathology images are used in digital\npathology. WSIs pose great challenges to deep learning models for clinical\ndiagnosis, owing to their size and lack of pixel-level annotations. With the\nrecent advancements in computational pathology, newer multiple-instance\nlearning-based models have been proposed. Multiple-instance learning for WSIs\nnecessitates creating patches and uses the encoding of these patches for\ndiagnosis. These models use generic pre-trained models (ResNet-50 pre-trained\non ImageNet) for patch encoding. The recently proposed KimiaNet, a DenseNet121\nmodel pre-trained on TCGA slides, is a domain-specific pre-trained model. This\npaper shows the effect of domain-specific pre-training on WSI classification.\nTo investigate the effect of domain-specific pre-training, we considered the\ncurrent state-of-the-art multiple-instance learning models, 1) CLAM, an\nattention-based model, and 2) TransMIL, a self-attention-based model, and\nevaluated the models' confidence and predictive performance in detecting\nprimary brain tumors - gliomas. Domain-specific pre-training improves the\nconfidence of the models and also achieves a new state-of-the-art performance\nof WSI-based glioma subtype classification, showing a high clinical\napplicability in assisting glioma diagnosis. We will publicly share our code\nand experimental results at\nhttps://github.com/soham-chitnis10/WSI-domain-specific.\n", "date": "2023-02-20", "categories": []}
{"id": "2402.08333", "title": "Scribble-based fast weak-supervision and interactive corrections for\n  segmenting whole slide images", "abstract": "  This paper proposes a dynamic interactive and weakly supervised segmentation\nmethod with minimal user interactions to address two major challenges in the\nsegmentation of whole slide histopathology images. First, the lack of\nhand-annotated datasets to train algorithms. Second, the lack of interactive\nparadigms to enable a dialogue between the pathologist and the machine, which\ncan be a major obstacle for use in clinical routine.\n  We therefore propose a fast and user oriented method to bridge this gap by\ngiving the pathologist control over the final result while limiting the number\nof interactions needed to achieve a good result (over 90\\% on all our metrics\nwith only 4 correction scribbles).\n", "date": "2024-02-13", "categories": []}
{"id": "2401.09773", "title": "SEINE: Structure Encoding and Interaction Network for Nuclei Instance\n  Segmentation", "abstract": "  Nuclei instance segmentation in histopathological images is of great\nimportance for biological analysis and cancer diagnosis but remains challenging\nfor two reasons. (1) Similar visual presentation of intranuclear and\nextranuclear regions of chromophobe nuclei often causes under-segmentation, and\n(2) current methods lack the exploration of nuclei structure, resulting in\nfragmented instance predictions. To address these problems, this paper proposes\na structure encoding and interaction network, termed SEINE, which develops the\nstructure modeling scheme of nuclei and exploits the structure similarity\nbetween nuclei to improve the integrality of each segmented instance.\nConcretely, SEINE introduces a contour-based structure encoding (SE) that\nconsiders the correlation between nuclei structure and semantics, realizing a\nreasonable representation of the nuclei structure. Based on the encoding, we\npropose a structure-guided attention (SGA) module that takes the clear nuclei\nas prototypes to enhance the structure learning for the fuzzy nuclei. To\nstrengthen the structural learning ability, a semantic feature fusion (SFF) is\npresented to boost the semantic consistency of semantic and structure branches.\nFurthermore, a position enhancement (PE) method is applied to suppress\nincorrect nuclei boundary predictions. Extensive experiments demonstrate the\nsuperiority of our approaches, and SEINE achieves state-of-the-art (SOTA)\nperformance on four datasets. The code is available at\nhttps://github.com/zhangye-zoe/SEINE.\n", "date": "2024-01-18", "categories": []}
{"id": "2304.05482", "title": "Computational Pathology: A Survey Review and The Way Forward", "abstract": "  Computational Pathology CPath is an interdisciplinary science that augments\ndevelopments of computational approaches to analyze and model medical\nhistopathology images. The main objective for CPath is to develop\ninfrastructure and workflows of digital diagnostics as an assistive CAD system\nfor clinical pathology, facilitating transformational changes in the diagnosis\nand treatment of cancer that are mainly address by CPath tools. With\nevergrowing developments in deep learning and computer vision algorithms, and\nthe ease of the data flow from digital pathology, currently CPath is witnessing\na paradigm shift. Despite the sheer volume of engineering and scientific works\nbeing introduced for cancer image analysis, there is still a considerable gap\nof adopting and integrating these algorithms in clinical practice. This raises\na significant question regarding the direction and trends that are undertaken\nin CPath. In this article we provide a comprehensive review of more than 800\npapers to address the challenges faced in problem design all-the-way to the\napplication and implementation viewpoints. We have catalogued each paper into a\nmodel-card by examining the key works and challenges faced to layout the\ncurrent landscape in CPath. We hope this helps the community to locate relevant\nworks and facilitate understanding of the field's future directions. In a\nnutshell, we oversee the CPath developments in cycle of stages which are\nrequired to be cohesively linked together to address the challenges associated\nwith such multidisciplinary science. We overview this cycle from different\nperspectives of data-centric, model-centric, and application-centric problems.\nWe finally sketch remaining challenges and provide directions for future\ntechnical developments and clinical integration of CPath\n(https://github.com/AtlasAnalyticsLab/CPath_Survey).\n", "date": "2023-04-11", "categories": []}
{"id": "2305.17891", "title": "The Rise of AI Language Pathologists: Exploring Two-level Prompt\n  Learning for Few-shot Weakly-supervised Whole Slide Image Classification", "abstract": "  This paper introduces the novel concept of few-shot weakly supervised\nlearning for pathology Whole Slide Image (WSI) classification, denoted as FSWC.\nA solution is proposed based on prompt learning and the utilization of a large\nlanguage model, GPT-4. Since a WSI is too large and needs to be divided into\npatches for processing, WSI classification is commonly approached as a Multiple\nInstance Learning (MIL) problem. In this context, each WSI is considered a bag,\nand the obtained patches are treated as instances. The objective of FSWC is to\nclassify both bags and instances with only a limited number of labeled bags.\nUnlike conventional few-shot learning problems, FSWC poses additional\nchallenges due to its weak bag labels within the MIL framework. Drawing\ninspiration from the recent achievements of vision-language models (V-L models)\nin downstream few-shot classification tasks, we propose a two-level prompt\nlearning MIL framework tailored for pathology, incorporating language prior\nknowledge. Specifically, we leverage CLIP to extract instance features for each\npatch, and introduce a prompt-guided pooling strategy to aggregate these\ninstance features into a bag feature. Subsequently, we employ a small number of\nlabeled bags to facilitate few-shot prompt learning based on the bag features.\nOur approach incorporates the utilization of GPT-4 in a question-and-answer\nmode to obtain language prior knowledge at both the instance and bag levels,\nwhich are then integrated into the instance and bag level language prompts.\nAdditionally, a learnable component of the language prompts is trained using\nthe available few-shot labeled data. We conduct extensive experiments on three\nreal WSI datasets encompassing breast cancer, lung cancer, and cervical cancer,\ndemonstrating the notable performance of the proposed method in bag and\ninstance classification. All codes will be available.\n", "date": "2023-05-29", "categories": []}
{"id": "2401.05602", "title": "Nucleus subtype classification using inter-modality learning", "abstract": "  Understanding the way cells communicate, co-locate, and interrelate is\nessential to understanding human physiology. Hematoxylin and eosin (H&E)\nstaining is ubiquitously available both for clinical studies and research. The\nColon Nucleus Identification and Classification (CoNIC) Challenge has recently\ninnovated on robust artificial intelligence labeling of six cell types on H&E\nstains of the colon. However, this is a very small fraction of the number of\npotential cell classification types. Specifically, the CoNIC Challenge is\nunable to classify epithelial subtypes (progenitor, endocrine, goblet),\nlymphocyte subtypes (B, helper T, cytotoxic T), or connective subtypes\n(fibroblasts, stromal). In this paper, we propose to use inter-modality\nlearning to label previously un-labelable cell types on virtual H&E. We\nleveraged multiplexed immunofluorescence (MxIF) histology imaging to identify\n14 subclasses of cell types. We performed style transfer to synthesize virtual\nH&E from MxIF and transferred the higher density labels from MxIF to these\nvirtual H&E images. We then evaluated the efficacy of learning in this\napproach. We identified helper T and progenitor nuclei with positive predictive\nvalues of $0.34 \\pm 0.15$ (prevalence $0.03 \\pm 0.01$) and $0.47 \\pm 0.1$\n(prevalence $0.07 \\pm 0.02$) respectively on virtual H&E. This approach\nrepresents a promising step towards automating annotation in digital pathology.\n", "date": "2024-01-11", "categories": []}
{"id": "2210.10981", "title": "MGTUNet: An new UNet for colon nuclei instance segmentation and\n  quantification", "abstract": "  Colorectal cancer (CRC) is among the top three malignant tumor types in terms\nof morbidity and mortality. Histopathological images are the gold standard for\ndiagnosing colon cancer. Cellular nuclei instance segmentation and\nclassification, and nuclear component regression tasks can aid in the analysis\nof the tumor microenvironment in colon tissue. Traditional methods are still\nunable to handle both types of tasks end-to-end at the same time, and have poor\nprediction accuracy and high application costs. This paper proposes a new UNet\nmodel for handling nuclei based on the UNet framework, called MGTUNet, which\nuses Mish, Group normalization and transposed convolution layer to improve the\nsegmentation model, and a ranger optimizer to adjust the SmoothL1Loss values.\nSecondly, it uses different channels to segment and classify different types of\nnucleus, ultimately completing the nuclei instance segmentation and\nclassification task, and the nuclei component regression task simultaneously.\nFinally, we did extensive comparison experiments using eight segmentation\nmodels. By comparing the three evaluation metrics and the parameter sizes of\nthe models, MGTUNet obtained 0.6254 on PQ, 0.6359 on mPQ, and 0.8695 on R2.\nThus, the experiments demonstrated that MGTUNet is now a state-of-the-art\nmethod for quantifying histopathological images of colon cancer.\n", "date": "2022-10-20", "categories": []}
{"id": "2307.04596", "title": "Distill-SODA: Distilling Self-Supervised Vision Transformer for\n  Source-Free Open-Set Domain Adaptation in Computational Pathology", "abstract": "  Developing computational pathology models is essential for reducing manual\ntissue typing from whole slide images, transferring knowledge from the source\ndomain to an unlabeled, shifted target domain, and identifying unseen\ncategories. We propose a practical setting by addressing the above-mentioned\nchallenges in one fell swoop, i.e., source-free open-set domain adaptation. Our\nmethodology focuses on adapting a pre-trained source model to an unlabeled\ntarget dataset and encompasses both closed-set and open-set classes. Beyond\naddressing the semantic shift of unknown classes, our framework also deals with\na covariate shift, which manifests as variations in color appearance between\nsource and target tissue samples. Our method hinges on distilling knowledge\nfrom a self-supervised vision transformer (ViT), drawing guidance from either\nrobustly pre-trained transformer models or histopathology datasets, including\nthose from the target domain. In pursuit of this, we introduce a novel\nstyle-based adversarial data augmentation, serving as hard positives for\nself-training a ViT, resulting in highly contextualized embeddings. Following\nthis, we cluster semantically akin target images, with the source model\noffering weak pseudo-labels, albeit with uncertain confidence. To enhance this\nprocess, we present the closed-set affinity score (CSAS), aiming to correct the\nconfidence levels of these pseudo-labels and to calculate weighted class\nprototypes within the contextualized embedding space. Our approach establishes\nitself as state-of-the-art across three public histopathological datasets for\ncolorectal cancer assessment. Notably, our self-training method seamlessly\nintegrates with open-set detection methods, resulting in enhanced performance\nin both closed-set and open-set recognition tasks.\n", "date": "2023-07-10", "categories": []}
{"id": "2401.14248", "title": "On generalisability of segment anything model for nuclear instance\n  segmentation in histology images", "abstract": "  Pre-trained on a large and diverse dataset, the segment anything model (SAM)\nis the first promptable foundation model in computer vision aiming at object\nsegmentation tasks. In this work, we evaluate SAM for the task of nuclear\ninstance segmentation performance with zero-shot learning and finetuning. We\ncompare SAM with other representative methods in nuclear instance segmentation,\nespecially in the context of model generalisability. To achieve automatic\nnuclear instance segmentation, we propose using a nuclei detection model to\nprovide bounding boxes or central points of nu-clei as visual prompts for SAM\nin generating nuclear instance masks from histology images.\n", "date": "2024-01-25", "categories": []}
{"id": "2401.10608", "title": "M2ORT: Many-To-One Regression Transformer for Spatial Transcriptomics\n  Prediction from Histopathology Images", "abstract": "  The advancement of Spatial Transcriptomics (ST) has facilitated the\nspatially-aware profiling of gene expressions based on histopathology images.\nAlthough ST data offers valuable insights into the micro-environment of tumors,\nits acquisition cost remains expensive. Therefore, directly predicting the ST\nexpressions from digital pathology images is desired. Current methods usually\nadopt existing regression backbones for this task, which ignore the inherent\nmulti-scale hierarchical data structure of digital pathology images. To address\nthis limit, we propose M2ORT, a many-to-one regression Transformer that can\naccommodate the hierarchical structure of the pathology images through a\ndecoupled multi-scale feature extractor. Different from traditional models that\nare trained with one-to-one image-label pairs, M2ORT accepts multiple pathology\nimages of different magnifications at a time to jointly predict the gene\nexpressions at their corresponding common ST spot, aiming at learning a\nmany-to-one relationship through training. We have tested M2ORT on three public\nST datasets and the experimental results show that M2ORT can achieve\nstate-of-the-art performance with fewer parameters and floating-point\noperations (FLOPs). The code is available at:\nhttps://github.com/Dootmaan/M2ORT/.\n", "date": "2024-01-19", "categories": []}
{"id": "2401.11062", "title": "Learned Image resizing with efficient training (LRET) facilitates\n  improved performance of large-scale digital histopathology image\n  classification models", "abstract": "  Histologic examination plays a crucial role in oncology research and\ndiagnostics. The adoption of digital scanning of whole slide images (WSI) has\ncreated an opportunity to leverage deep learning-based image classification\nmethods to enhance diagnosis and risk stratification. Technical limitations of\ncurrent approaches to training deep convolutional neural networks (DCNN) result\nin suboptimal model performance and make training and deployment of\ncomprehensive classification models unobtainable. In this study, we introduce a\nnovel approach that addresses the main limitations of traditional\nhistopathology classification model training. Our method, termed Learned\nResizing with Efficient Training (LRET), couples efficient training techniques\nwith image resizing to facilitate seamless integration of larger histology\nimage patches into state-of-the-art classification models while preserving\nimportant structural information.\n  We used the LRET method coupled with two distinct resizing techniques to\ntrain three diverse histology image datasets using multiple diverse DCNN\narchitectures. Our findings demonstrate a significant enhancement in\nclassification performance and training efficiency. Across the spectrum of\nexperiments, LRET consistently outperforms existing methods, yielding a\nsubstantial improvement of 15-28% in accuracy for a large-scale, multiclass\ntumor classification task consisting of 74 distinct brain tumor types. LRET not\nonly elevates classification accuracy but also substantially reduces training\ntimes, unlocking the potential for faster model development and iteration. The\nimplications of this work extend to broader applications within medical imaging\nand beyond, where efficient integration of high-resolution images into deep\nlearning pipelines is paramount for driving advancements in research and\nclinical practice.\n", "date": "2024-01-19", "categories": []}
{"id": "2309.04631", "title": "Open and reusable deep learning for pathology with WSInfer and QuPath", "abstract": "  The field of digital pathology has seen a proliferation of deep learning\nmodels in recent years. Despite substantial progress, it remains rare for other\nresearchers and pathologists to be able to access models published in the\nliterature and apply them to their own images. This is due to difficulties in\nboth sharing and running models. To address these concerns, we introduce\nWSInfer: a new, open-source software ecosystem designed to make deep learning\nfor pathology more streamlined and accessible. WSInfer comprises three main\nelements: 1) a Python package and command line tool to efficiently apply\npatch-based deep learning inference to whole slide images; 2) a QuPath\nextension that provides an alternative inference engine through user-friendly\nand interactive software, and 3) a model zoo, which enables pathology models\nand metadata to be easily shared in a standardized form. Together, these\ncontributions aim to encourage wider reuse, exploration, and interrogation of\ndeep learning models for research purposes, by putting them into the hands of\npathologists and eliminating a need for coding experience when accessed through\nQuPath. The WSInfer source code is hosted on GitHub and documentation is\navailable at https://wsinfer.readthedocs.io.\n", "date": "2023-09-08", "categories": []}
{"id": "2305.11728", "title": "Towards More Transparent and Accurate Cancer Diagnosis with an\n  Unsupervised CAE Approach", "abstract": "  Digital pathology has revolutionized cancer diagnosis by leveraging\nContent-Based Medical Image Retrieval (CBMIR) for analyzing histopathological\nWhole Slide Images (WSIs). CBMIR enables searching for similar content,\nenhancing diagnostic reliability and accuracy. In 2020, breast and prostate\ncancer constituted 11.7% and 14.1% of cases, respectively, as reported by the\nGlobal Cancer Observatory (GCO). The proposed Unsupervised CBMIR (UCBMIR)\nreplicates the traditional cancer diagnosis workflow, offering a dependable\nmethod to support pathologists in WSI-based diagnostic conclusions. This\napproach alleviates pathologists' workload, potentially enhancing diagnostic\nefficiency. To address the challenge of the lack of labeled histopathological\nimages in CBMIR, a customized unsupervised Convolutional Auto Encoder (CAE) was\ndeveloped, extracting 200 features per image for the search engine component.\nUCBMIR was evaluated using widely-used numerical techniques in CBMIR, alongside\nvisual evaluation and comparison with a classifier. The validation involved\nthree distinct datasets, with an external evaluation demonstrating its\neffectiveness. UCBMIR outperformed previous studies, achieving a top 5 recall\nof 99% and 80% on BreaKHis and SICAPv2, respectively, using the first\nevaluation technique. Precision rates of 91% and 70% were achieved for BreaKHis\nand SICAPv2, respectively, using the second evaluation technique. Furthermore,\nUCBMIR demonstrated the capability to identify various patterns in patches,\nachieving an 81% accuracy in the top 5 when tested on an external image from\nArvaniti.\n", "date": "2023-05-19", "categories": []}
{"id": "2301.13128", "title": "Standardized CycleGAN training for unsupervised stain adaptation in\n  invasive carcinoma classification for breast histopathology", "abstract": "  Generalization is one of the main challenges of computational pathology.\nSlide preparation heterogeneity and the diversity of scanners lead to poor\nmodel performance when used on data from medical centers not seen during\ntraining. In order to achieve stain invariance in breast invasive carcinoma\npatch classification, we implement a stain translation strategy using cycleGANs\nfor unsupervised image-to-image translation. We compare three cycleGAN-based\napproaches to a baseline classification model obtained without any stain\ninvariance strategy. Two of the proposed approaches use cycleGAN's translations\nat inference or training in order to build stain-specific classification\nmodels. The last method uses them for stain data augmentation during training.\nThis constrains the classification model to learn stain-invariant features.\nBaseline metrics are set by training and testing the baseline classification\nmodel on a reference stain. We assessed performances using three medical\ncenters with H&E and H&E&S staining. Every approach tested in this study\nimproves baseline metrics without needing labels on target stains. The stain\naugmentation-based approach produced the best results on every stain. Each\nmethod's pros and cons are studied and discussed in this paper. However,\ntraining highly performing cycleGANs models in itself represents a challenge.\nIn this work, we introduce a systematical method for optimizing cycleGAN\ntraining by setting a novel stopping criterion. This method has the benefit of\nnot requiring any visual inspection of cycleGAN results and proves superiority\nto methods using a predefined number of training epochs. In addition, we also\nstudy the minimal amount of data required for cycleGAN training.\n", "date": "2023-01-30", "categories": []}
{"id": "2304.13775", "title": "Advancing Ischemic Stroke Diagnosis: A Novel Two-Stage Approach for\n  Blood Clot Origin Identification", "abstract": "  An innovative two-stage methodology for categorizing blood clot origins is\npresented in this paper, which is important for the diagnosis and treatment of\nischemic stroke. First, a background classifier based on MobileNetV3 segments\nbig whole-slide digital pathology images into numerous tiles to detect the\npresence of cellular material. After that, different pre-trained image\nclassification algorithms are fine-tuned to determine the origin of blood\nclots. Due to complex blood flow dynamics and limitations in conventional\nimaging methods such as computed tomography (CT), magnetic resonance imaging\n(MRI), and ultrasound, identifying the sources of blood clots is a challenging\ntask. Although these techniques are useful for identifying blood clots, they\nare not very good at determining how they originated. To address these\nchallenges, our method makes use of robust computer vision models that have\nbeen refined using information from whole-slide digital pathology images. Out\nof all the models tested, the PoolFormer \\cite{yu2022metaformer} performs\nbetter than the others, with 93.4\\% accuracy, 93.4\\% precision, 93.4\\% recall,\nand 93.4\\% F1-score. Moreover, it achieves the good weighted multi-class\nlogarithmic loss (WMCLL) of 0.4361, which emphasizes how effective it is in\nthis particular application. These encouraging findings suggest that our\napproach can successfully identify the origin of blood clots in a variety of\nvascular locations, potentially advancing ischemic stroke diagnosis and\ntreatment approaches.\n", "date": "2023-04-26", "categories": []}
{"id": "2306.17019", "title": "Histopathology Slide Indexing and Search: Are We There Yet?", "abstract": "  The search and retrieval of digital histopathology slides is an important\ntask that has yet to be solved. In this case study, we investigate the clinical\nreadiness of three state-of-the-art histopathology slide search engines,\nYottixel, SISH, and RetCCL, on three patients with solid tumors. We provide a\nqualitative assessment of each model's performance in providing retrieval\nresults that are reliable and useful to pathologists. We found that all three\nimage search engines fail to produce consistently reliable results and have\ndifficulties in capturing granular and subtle features of malignancy, limiting\ntheir diagnostic accuracy. Based on our findings, we also propose a minimal set\nof requirements to further advance the development of accurate and reliable\nhistopathology image search engines for successful clinical adoption.\n", "date": "2023-06-29", "categories": []}
{"id": "2207.14650", "title": "SYNTA: A novel approach for deep learning-based image analysis in muscle\n  histopathology using photo-realistic synthetic data", "abstract": "  Artificial intelligence (AI), machine learning, and deep learning (DL)\nmethods are becoming increasingly important in the field of biomedical image\nanalysis. However, to exploit the full potential of such methods, a\nrepresentative number of experimentally acquired images containing a\nsignificant number of manually annotated objects is needed as training data.\nHere we introduce SYNTA (synthetic data) as a novel approach for the generation\nof synthetic, photo-realistic, and highly complex biomedical images as training\ndata for DL systems. We show the versatility of our approach in the context of\nmuscle fiber and connective tissue analysis in histological sections. We\ndemonstrate that it is possible to perform robust and expert-level segmentation\ntasks on previously unseen real-world data, without the need for manual\nannotations using synthetic training data alone. Being a fully parametric\ntechnique, our approach poses an interpretable and controllable alternative to\nGenerative Adversarial Networks (GANs) and has the potential to significantly\naccelerate quantitative image analysis in a variety of biomedical applications\nin microscopy and beyond.\n", "date": "2022-07-29", "categories": []}
{"id": "2312.02225", "title": "Digital Histopathology with Graph Neural Networks: Concepts and\n  Explanations for Clinicians", "abstract": "  To address the challenge of the ``black-box\" nature of deep learning in\nmedical settings, we combine GCExplainer - an automated concept discovery\nsolution - along with Logic Explained Networks to provide global explanations\nfor Graph Neural Networks. We demonstrate this using a generally applicable\ngraph construction and classification pipeline, involving panoptic segmentation\nwith HoVer-Net and cancer prediction with Graph Convolution Networks. By\ntraining on H&E slides of breast cancer, we show promising results in offering\nexplainable and trustworthy AI tools for clinicians.\n", "date": "2023-12-04", "categories": []}
{"id": "2312.15825", "title": "Comparative Analysis of Radiomic Features and Gene Expression Profiles\n  in Histopathology Data Using Graph Neural Networks", "abstract": "  This study leverages graph neural networks to integrate MELC data with\nRadiomic-extracted features for melanoma classification, focusing on cell-wise\nanalysis. It assesses the effectiveness of gene expression profiles and\nRadiomic features, revealing that Radiomic features, particularly when combined\nwith UMAP for dimensionality reduction, significantly enhance classification\nperformance. Notably, using Radiomics contributes to increased diagnostic\naccuracy and computational efficiency, as it allows for the extraction of\ncritical data from fewer stains, thereby reducing operational costs. This\nmethodology marks an advancement in computational dermatology for melanoma cell\nclassification, setting the stage for future research and potential\ndevelopments.\n", "date": "2023-12-25", "categories": []}
{"id": "2312.12619", "title": "Hierarchical Vision Transformers for Context-Aware Prostate Cancer\n  Grading in Whole Slide Images", "abstract": "  Vision Transformers (ViTs) have ushered in a new era in computer vision,\nshowcasing unparalleled performance in many challenging tasks. However, their\npractical deployment in computational pathology has largely been constrained by\nthe sheer size of whole slide images (WSIs), which result in lengthy input\nsequences. Transformers faced a similar limitation when applied to long\ndocuments, and Hierarchical Transformers were introduced to circumvent it.\nGiven the analogous challenge with WSIs and their inherent hierarchical\nstructure, Hierarchical Vision Transformers (H-ViTs) emerge as a promising\nsolution in computational pathology. This work delves into the capabilities of\nH-ViTs, evaluating their efficiency for prostate cancer grading in WSIs. Our\nresults show that they achieve competitive performance against existing\nstate-of-the-art solutions.\n", "date": "2023-12-19", "categories": []}
{"id": "2312.12824", "title": "FedSODA: Federated Cross-assessment and Dynamic Aggregation for\n  Histopathology Segmentation", "abstract": "  Federated learning (FL) for histopathology image segmentation involving\nmultiple medical sites plays a crucial role in advancing the field of accurate\ndisease diagnosis and treatment. However, it is still a task of great\nchallenges due to the sample imbalance across clients and large data\nheterogeneity from disparate organs, variable segmentation tasks, and diverse\ndistribution. Thus, we propose a novel FL approach for histopathology nuclei\nand tissue segmentation, FedSODA, via synthetic-driven cross-assessment\noperation (SO) and dynamic stratified-layer aggregation (DA). Our SO constructs\na cross-assessment strategy to connect clients and mitigate the representation\nbias under sample imbalance. Our DA utilizes layer-wise interaction and dynamic\naggregation to diminish heterogeneity and enhance generalization. The\neffectiveness of our FedSODA has been evaluated on the most extensive\nhistopathology image segmentation dataset from 7 independent datasets. The code\nis available at https://github.com/yuanzhang7/FedSODA.\n", "date": "2023-12-20", "categories": []}
{"id": "2308.13304", "title": "Rapid Artefact Removal and H&E-Stained Tissue Segmentation", "abstract": "  We present an innovative method for rapidly segmenting hematoxylin and eosin\n(H&E)-stained tissue in whole-slide images (WSIs) that eliminates a wide range\nof undesirable artefacts such as pen marks and scanning artefacts. Our method\ninvolves taking a single-channel representation of a lowmagnification RGB\noverview of the WSI in which the pixel values are bimodally distributed such\nthat H&E-stained tissue is easily distinguished from both background and a wide\nvariety of artefacts. We demonstrate our method on 30 WSIs prepared from a wide\nrange of institutions and WSI digital scanners, each containing substantial\nartefacts, and compare it to segmentations provided by Otsu thresholding and\nHistolab tissue segmentation and pen filtering tools. We found that our method\nsegmented the tissue and fully removed all artefacts in 29 out of 30 WSIs,\nwhereas Otsu thresholding failed to remove any artefacts, and the Histolab pen\nfiltering tools only partially removed the pen marks. The beauty of our\napproach lies in its simplicity: manipulating RGB colour space and using Otsu\nthresholding allows for the segmentation of H&E-stained tissue and the rapid\nremoval of artefacts without the need for machine learning or parameter tuning.\n", "date": "2023-08-25", "categories": []}
{"id": "2312.12151", "title": "SoftCTM: Cell detection by soft instance segmentation and consideration\n  of cell-tissue interaction", "abstract": "  Detecting and classifying cells in histopathology H\\&E stained whole-slide\nimages is a core task in computational pathology, as it provides valuable\ninsight into the tumor microenvironment. In this work we investigate the impact\nof ground truth formats on the models performance. Additionally, cell-tissue\ninteractions are considered by providing tissue segmentation predictions as\ninput to the cell detection model. We find that a \"soft\", probability-map\ninstance segmentation ground truth leads to best model performance. Combined\nwith cell-tissue interaction and test-time augmentation our Soft\nCell-Tissue-Model (SoftCTM) achieves 0.7172 mean F1-Score on the Overlapped\nCell On Tissue (OCELOT) test set, achieving the third best overall score in the\nOCELOT 2023 Challenge. The source code for our approach is made publicly\navailable at https://github.com/lely475/ocelot23algo.\n", "date": "2023-12-19", "categories": []}
{"id": "2312.10944", "title": "From Whole-slide Image to Biomarker Prediction: A Protocol for\n  End-to-End Deep Learning in Computational Pathology", "abstract": "  Hematoxylin- and eosin (H&E) stained whole-slide images (WSIs) are the\nfoundation of diagnosis of cancer. In recent years, development of deep\nlearning-based methods in computational pathology enabled the prediction of\nbiomarkers directly from WSIs. However, accurately linking tissue phenotype to\nbiomarkers at scale remains a crucial challenge for democratizing complex\nbiomarkers in precision oncology. This protocol describes a practical workflow\nfor solid tumor associative modeling in pathology (STAMP), enabling prediction\nof biomarkers directly from WSIs using deep learning. The STAMP workflow is\nbiomarker agnostic and allows for genetic- and clinicopathologic tabular data\nto be included as an additional input, together with histopathology images. The\nprotocol consists of five main stages which have been successfully applied to\nvarious research problems: formal problem definition, data preprocessing,\nmodeling, evaluation and clinical translation. The STAMP workflow\ndifferentiates itself through its focus on serving as a collaborative framework\nthat can be used by clinicians and engineers alike for setting up research\nprojects in the field of computational pathology. As an example task, we\napplied STAMP to the prediction of microsatellite instability (MSI) status in\ncolorectal cancer, showing accurate performance for the identification of\nMSI-high tumors. Moreover, we provide an open-source codebase which has been\ndeployed at several hospitals across the globe to set up computational\npathology workflows. The STAMP workflow requires one workday of hands-on\ncomputational execution and basic command line knowledge.\n", "date": "2023-12-18", "categories": []}
{"id": "2312.07489", "title": "NearbyPatchCL: Leveraging Nearby Patches for Self-Supervised Patch-Level\n  Multi-Class Classification in Whole-Slide Images", "abstract": "  Whole-slide image (WSI) analysis plays a crucial role in cancer diagnosis and\ntreatment. In addressing the demands of this critical task, self-supervised\nlearning (SSL) methods have emerged as a valuable resource, leveraging their\nefficiency in circumventing the need for a large number of annotations, which\ncan be both costly and time-consuming to deploy supervised methods.\nNevertheless, patch-wise representation may exhibit instability in performance,\nprimarily due to class imbalances stemming from patch selection within WSIs. In\nthis paper, we introduce Nearby Patch Contrastive Learning (NearbyPatchCL), a\nnovel self-supervised learning method that leverages nearby patches as positive\nsamples and a decoupled contrastive loss for robust representation learning.\nOur method demonstrates a tangible enhancement in performance for downstream\ntasks involving patch-level multi-class classification. Additionally, we curate\na new dataset derived from WSIs sourced from the Canine Cutaneous Cancer\nHistology, thus establishing a benchmark for the rigorous evaluation of\npatch-level multi-class classification methodologies. Intensive experiments\nshow that our method significantly outperforms the supervised baseline and\nstate-of-the-art SSL methods with top-1 classification accuracy of 87.56%. Our\nmethod also achieves comparable results while utilizing a mere 1% of labeled\ndata, a stark contrast to the 100% labeled data requirement of other\napproaches. Source code: https://github.com/nvtien457/NearbyPatchCL\n", "date": "2023-12-12", "categories": []}
{"id": "2312.03812", "title": "Seeing the random forest through the decision trees. Supporting learning\n  health systems from histopathology with machine learning models: Challenges\n  and opportunities", "abstract": "  This paper discusses some overlooked challenges faced when working with\nmachine learning models for histopathology and presents a novel opportunity to\nsupport \"Learning Health Systems\" with them. Initially, the authors elaborate\non these challenges after separating them according to their mitigation\nstrategies: those that need innovative approaches, time, or future\ntechnological capabilities and those that require a conceptual reappraisal from\na critical perspective. Then, a novel opportunity to support \"Learning Health\nSystems\" by integrating hidden information extracted by ML models from\ndigitalized histopathology slides with other healthcare big data is presented.\n", "date": "2023-12-06", "categories": []}
{"id": "2306.01546", "title": "Publicly available datasets of breast histopathology H&E whole-slide\n  images: A scoping review", "abstract": "  Advancements in digital pathology and computing resources have made a\nsignificant impact in the field of computational pathology for breast cancer\ndiagnosis and treatment. However, access to high-quality labeled\nhistopathological images of breast cancer is a big challenge that limits the\ndevelopment of accurate and robust deep learning models. In this scoping\nreview, we identified the publicly available datasets of breast H&E stained\nwhole-slide images (WSI) that can be used to develop deep learning algorithms.\nWe systematically searched nine scientific literature databases and nine\nresearch data repositories and found 17 publicly available datasets containing\n10385 H&E WSIs of breast cancer. Moreover, we reported image metadata and\ncharacteristics for each dataset to assist researchers in selecting proper\ndatasets for specific tasks in breast cancer computational pathology. In\naddition, we compiled two lists of breast H&E patches and private datasets as\nsupplementary resources for researchers. Notably, only 28% of the included\narticles utilized multiple datasets, and only 14% used an external validation\nset, suggesting that the performance of other developed models may be\nsusceptible to overestimation. The TCGA-BRCA was used in 52% of the selected\nstudies. This dataset has a considerable selection bias that can impact the\nrobustness and generalizability of the trained algorithms. There is also a lack\nof consistent metadata reporting of breast WSI datasets that can be an issue in\ndeveloping accurate deep learning models, indicating the necessity of\nestablishing explicit guidelines for documenting breast WSI dataset\ncharacteristics and metadata.\n", "date": "2023-06-02", "categories": []}
{"id": "2312.03558", "title": "When an Image is Worth 1,024 x 1,024 Words: A Case Study in\n  Computational Pathology", "abstract": "  This technical report presents LongViT, a vision Transformer that can process\ngigapixel images in an end-to-end manner. Specifically, we split the gigapixel\nimage into a sequence of millions of patches and project them linearly into\nembeddings. LongNet is then employed to model the extremely long sequence,\ngenerating representations that capture both short-range and long-range\ndependencies. The linear computation complexity of LongNet, along with its\ndistributed algorithm, enables us to overcome the constraints of both\ncomputation and memory. We apply LongViT in the field of computational\npathology, aiming for cancer diagnosis and prognosis within gigapixel\nwhole-slide images. Experimental results demonstrate that LongViT effectively\nencodes gigapixel images and outperforms previous state-of-the-art methods on\ncancer subtyping and survival prediction. Code and models will be available at\nhttps://aka.ms/LongViT.\n", "date": "2023-12-06", "categories": []}
{"id": "2312.03647", "title": "Editable Stain Transformation Of Histological Images Using Unpaired GANs", "abstract": "  Double staining in histopathology, particularly for metaplastic breast\ncancer, typically employs H&E and P63 dyes. However, P63's tissue damage and\nhigh cost necessitate alternative methods. This study introduces xAI-CycleGAN,\nan advanced architecture combining Mask CycleGAN with explainability features\nand structure-preserving capabilities for transforming H&E stained breast\ntissue images into P63-like images. The architecture allows for output editing,\nenhancing resemblance to actual images and enabling further model refinement.\nWe showcase xAI-CycleGAN's efficacy in maintaining structural integrity and\ngenerating high-quality images. Additionally, a histopathologist survey\nindicates the generated images' realism is often comparable to actual images,\nvalidating our model's high-quality output.\n", "date": "2023-12-06", "categories": []}
{"id": "2312.01152", "title": "Ultra-Resolution Cascaded Diffusion Model for Gigapixel Image Synthesis\n  in Histopathology", "abstract": "  Diagnoses from histopathology images rely on information from both high and\nlow resolutions of Whole Slide Images. Ultra-Resolution Cascaded Diffusion\nModels (URCDMs) allow for the synthesis of high-resolution images that are\nrealistic at all magnification levels, focusing not only on fidelity but also\non long-distance spatial coherency. Our model beats existing methods, improving\nthe pFID-50k [2] score by 110.63 to 39.52 pFID-50k. Additionally, a human\nexpert evaluation study was performed, reaching a weighted Mean Absolute Error\n(MAE) of 0.11 for the Lower Resolution Diffusion Models and a weighted MAE of\n0.22 for the URCDM.\n", "date": "2023-12-02", "categories": []}
{"id": "2311.14971", "title": "Segmentation of diagnostic tissue compartments on whole slide images\n  with renal thrombotic microangiopathies (TMAs)", "abstract": "  The thrombotic microangiopathies (TMAs) manifest in renal biopsy histology\nwith a broad spectrum of acute and chronic findings. Precise diagnostic\ncriteria for a renal biopsy diagnosis of TMA are missing. As a first step\ntowards a machine learning- and computer vision-based analysis of wholes slide\nimages from renal biopsies, we trained a segmentation model for the decisive\ndiagnostic kidney tissue compartments artery, arteriole, glomerulus on a set of\nwhole slide images from renal biopsies with TMAs and Mimickers (distinct\ndiseases with a similar nephropathological appearance as TMA like severe benign\nnephrosclerosis, various vasculitides, Bevacizumab-plug glomerulopathy,\narteriolar light chain deposition disease). Our segmentation model combines a\nU-Net-based tissue detection with a Shifted windows-transformer architecture to\nreach excellent segmentation results for even the most severely altered\nglomeruli, arterioles and arteries, even on unseen staining domains from a\ndifferent nephropathology lab. With accurate automatic segmentation of the\ndecisive renal biopsy compartments in human renal vasculopathies, we have laid\nthe foundation for large-scale compartment-specific machine learning and\ncomputer vision analysis of renal biopsy repositories with TMAs.\n", "date": "2023-11-25", "categories": []}
{"id": "2311.12885", "title": "Long-MIL: Scaling Long Contextual Multiple Instance Learning for\n  Histopathology Whole Slide Image Analysis", "abstract": "  Histopathology image analysis is the golden standard of clinical diagnosis\nfor Cancers. In doctors daily routine and computer-aided diagnosis, the Whole\nSlide Image (WSI) of histopathology tissue is used for analysis. Because of the\nextremely large scale of resolution, previous methods generally divide the WSI\ninto a large number of patches, then aggregate all patches within a WSI by\nMulti-Instance Learning (MIL) to make the slide-level prediction when\ndeveloping computer-aided diagnosis tools. However, most previous WSI-MIL\nmodels using global-attention without pairwise interaction and any positional\ninformation, or self-attention with absolute position embedding can not well\nhandle shape varying large WSIs, e.g. testing WSIs after model deployment may\nbe larger than training WSIs, since the model development set is always limited\ndue to the difficulty of histopathology WSIs collection. To deal with the\nproblem, in this paper, we propose to amend position embedding for shape\nvarying long-contextual WSI by introducing Linear Bias into Attention, and\nadapt it from 1-d long sequence into 2-d long-contextual WSI which helps model\nextrapolate position embedding to unseen or under-fitted positions. We further\nutilize Flash-Attention module to tackle the computational complexity of\nTransformer, which also keep full self-attention performance compared to\nprevious attention approximation work. Our method, Long-contextual MIL\n(Long-MIL) are evaluated on extensive experiments including 4 dataset including\nWSI classification and survival prediction tasks to validate the superiority on\nshape varying WSIs. The source code will be open-accessed soon.\n", "date": "2023-11-21", "categories": []}
{"id": "2311.06185", "title": "An Automated Pipeline for Tumour-Infiltrating Lymphocyte Scoring in\n  Breast Cancer", "abstract": "  Tumour-infiltrating lymphocytes (TILs) are considered as a valuable\nprognostic markers in both triple-negative and human epidermal growth factor\nreceptor 2 (HER2) positive breast cancer. In this study, we introduce an\ninnovative deep learning pipeline based on the Efficient-UNet architecture to\npredict the TILs score for breast cancer whole-slide images (WSIs). We first\nsegment tumour and stromal regions in order to compute a tumour bulk mask. We\nthen detect TILs within the tumour-associated stroma, generating a TILs score\nby closely mirroring the pathologist's workflow. Our method exhibits\nstate-of-the-art performance in segmenting tumour/stroma areas and TILs\ndetection, as demonstrated by internal cross-validation on the TiGER Challenge\ntraining dataset and evaluation on the final leaderboards. Additionally, our\nTILs score proves competitive in predicting survival outcomes within the same\nchallenge, underscoring the clinical relevance and potential of our automated\nTILs scoring pipeline as a breast cancer prognostic tool.\n", "date": "2023-11-10", "categories": []}
{"id": "2311.09574", "title": "LymphoML: An interpretable artificial intelligence-based method\n  identifies morphologic features that correlate with lymphoma subtype", "abstract": "  The accurate classification of lymphoma subtypes using hematoxylin and eosin\n(H&E)-stained tissue is complicated by the wide range of morphological features\nthese cancers can exhibit. We present LymphoML - an interpretable machine\nlearning method that identifies morphologic features that correlate with\nlymphoma subtypes. Our method applies steps to process H&E-stained tissue\nmicroarray cores, segment nuclei and cells, compute features encompassing\nmorphology, texture, and architecture, and train gradient-boosted models to\nmake diagnostic predictions. LymphoML's interpretable models, developed on a\nlimited volume of H&E-stained tissue, achieve non-inferior diagnostic accuracy\nto pathologists using whole-slide images and outperform black box deep-learning\non a dataset of 670 cases from Guatemala spanning 8 lymphoma subtypes. Using\nSHapley Additive exPlanation (SHAP) analysis, we assess the impact of each\nfeature on model prediction and find that nuclear shape features are most\ndiscriminative for DLBCL (F1-score: 78.7%) and classical Hodgkin lymphoma\n(F1-score: 74.5%). Finally, we provide the first demonstration that a model\ncombining features from H&E-stained tissue with features from a standardized\npanel of 6 immunostains results in a similar diagnostic accuracy (85.3%) to a\n46-stain panel (86.1%).\n", "date": "2023-11-16", "categories": []}
{"id": "2311.11659", "title": "MGCT: Mutual-Guided Cross-Modality Transformer for Survival Outcome\n  Prediction using Integrative Histopathology-Genomic Features", "abstract": "  The rapidly emerging field of deep learning-based computational pathology has\nshown promising results in utilizing whole slide images (WSIs) to objectively\nprognosticate cancer patients. However, most prognostic methods are currently\nlimited to either histopathology or genomics alone, which inevitably reduces\ntheir potential to accurately predict patient prognosis. Whereas integrating\nWSIs and genomic features presents three main challenges: (1) the enormous\nheterogeneity of gigapixel WSIs which can reach sizes as large as\n150,000x150,000 pixels; (2) the absence of a spatially corresponding\nrelationship between histopathology images and genomic molecular data; and (3)\nthe existing early, late, and intermediate multimodal feature fusion strategies\nstruggle to capture the explicit interactions between WSIs and genomics. To\nameliorate these issues, we propose the Mutual-Guided Cross-Modality\nTransformer (MGCT), a weakly-supervised, attention-based multimodal learning\nframework that can combine histology features and genomic features to model the\ngenotype-phenotype interactions within the tumor microenvironment. To validate\nthe effectiveness of MGCT, we conduct experiments using nearly 3,600 gigapixel\nWSIs across five different cancer types sourced from The Cancer Genome Atlas\n(TCGA). Extensive experimental results consistently emphasize that MGCT\noutperforms the state-of-the-art (SOTA) methods.\n", "date": "2023-11-20", "categories": []}
{"id": "2307.07482", "title": "Dual-Query Multiple Instance Learning for Dynamic Meta-Embedding based\n  Tumor Classification", "abstract": "  Whole slide image (WSI) assessment is a challenging and crucial step in\ncancer diagnosis and treatment planning. WSIs require high magnifications to\nfacilitate sub-cellular analysis. Precise annotations for patch- or even\npixel-level classifications in the context of gigapixel WSIs are tedious to\nacquire and require domain experts. Coarse-grained labels, on the other hand,\nare easily accessible, which makes WSI classification an ideal use case for\nmultiple instance learning (MIL). In our work, we propose a novel\nembedding-based Dual-Query MIL pipeline (DQ-MIL). We contribute to both the\nembedding and aggregation steps. Since all-purpose visual feature\nrepresentations are not yet available, embedding models are currently limited\nin terms of generalizability. With our work, we explore the potential of\ndynamic meta-embedding based on cutting-edge self-supervised pre-trained models\nin the context of MIL. Moreover, we propose a new MIL architecture capable of\ncombining MIL-attention with correlated self-attention. The Dual-Query\nPerceiver design of our approach allows us to leverage the concept of\nself-distillation and to combine the advantages of a small model in the context\nof a low data regime with the rich feature representation of a larger model. We\ndemonstrate the superior performance of our approach on three histopathological\ndatasets, where we show improvement of up to 10% over state-of-the-art\napproaches.\n", "date": "2023-07-14", "categories": []}
{"id": "2311.09902", "title": "Selection of Distinct Morphologies to Divide & Conquer Gigapixel\n  Pathology Images", "abstract": "  Whole slide images (WSIs) are massive digital pathology files illustrating\nintricate tissue structures. Selecting a small, representative subset of\npatches from each WSI is essential yet challenging. Therefore, following the\n\"Divide & Conquer\" approach becomes essential to facilitate WSI analysis\nincluding the classification and the WSI matching in computational pathology.\nTo this end, we propose a novel method termed \"Selection of Distinct\nMorphologies\" (SDM) to choose a subset of WSI patches. The aim is to encompass\nall inherent morphological variations within a given WSI while simultaneously\nminimizing the number of selected patches to represent these variations,\nensuring a compact yet comprehensive set of patches. This systematically\ncurated patch set forms what we term a \"montage\". We assess the\nrepresentativeness of the SDM montage across various public and private\nhistopathology datasets. This is conducted by using the leave-one-out WSI\nsearch and matching evaluation method, comparing it with the state-of-the-art\nYottixel's mosaic. SDM demonstrates remarkable efficacy across all datasets\nduring its evaluation. Furthermore, SDM eliminates the necessity for empirical\nparameterization, a crucial aspect of Yottixel's mosaic, by inherently\noptimizing the selection process to capture the distinct morphological features\nwithin the WSI.\n", "date": "2023-11-16", "categories": []}
{"id": "2311.08949", "title": "Automated Volume Corrected Mitotic Index Calculation Through\n  Annotation-Free Deep Learning using Immunohistochemistry as Reference\n  Standard", "abstract": "  The volume-corrected mitotic index (M/V-Index) was shown to provide\nprognostic value in invasive breast carcinomas. However, despite its prognostic\nsignificance, it is not established as the standard method for assessing\naggressive biological behaviour, due to the high additional workload associated\nwith determining the epithelial proportion. In this work, we show that using a\ndeep learning pipeline solely trained with an annotation-free,\nimmunohistochemistry-based approach, provides accurate estimations of\nepithelial segmentation in canine breast carcinomas. We compare our automatic\nframework with the manually annotated M/V-Index in a study with three\nboard-certified pathologists. Our results indicate that the deep learning-based\npipeline shows expert-level performance, while providing time efficiency and\nreproducibility.\n", "date": "2023-11-15", "categories": []}
{"id": "2310.12334", "title": "Improving Representation Learning for Histopathologic Images with\n  Cluster Constraints", "abstract": "  Recent advances in whole-slide image (WSI) scanners and computational\ncapabilities have significantly propelled the application of artificial\nintelligence in histopathology slide analysis. While these strides are\npromising, current supervised learning approaches for WSI analysis come with\nthe challenge of exhaustively labeling high-resolution slides - a process that\nis both labor-intensive and time-consuming. In contrast, self-supervised\nlearning (SSL) pretraining strategies are emerging as a viable alternative,\ngiven that they don't rely on explicit data annotations. These SSL strategies\nare quickly bridging the performance disparity with their supervised\ncounterparts. In this context, we introduce an SSL framework. This framework\naims for transferable representation learning and semantically meaningful\nclustering by synergizing invariance loss and clustering loss in WSI analysis.\nNotably, our approach outperforms common SSL methods in downstream\nclassification and clustering tasks, as evidenced by tests on the Camelyon16\nand a pancreatic cancer dataset.\n", "date": "2023-10-18", "categories": []}
{"id": "2311.06552", "title": "Stain Consistency Learning: Handling Stain Variation for Automatic\n  Digital Pathology Segmentation", "abstract": "  Stain variation is a unique challenge associated with automated analysis of\ndigital pathology. Numerous methods have been developed to improve the\nrobustness of machine learning methods to stain variation, but comparative\nstudies have demonstrated limited benefits to performance. Moreover, methods to\nhandle stain variation were largely developed for H&E stained data, with\nevaluation generally limited to classification tasks. Here we propose Stain\nConsistency Learning, a novel framework combining stain-specific augmentation\nwith a stain consistency loss function to learn stain colour invariant\nfeatures. We perform the first, extensive comparison of methods to handle stain\nvariation for segmentation tasks, comparing ten methods on Masson's trichrome\nand H&E stained cell and nuclei datasets, respectively. We observed that stain\nnormalisation methods resulted in equivalent or worse performance, while stain\naugmentation or stain adversarial methods demonstrated improved performance,\nwith the best performance consistently achieved by our proposed approach. The\ncode is available at: https://github.com/mlyg/stain_consistency_learning\n", "date": "2023-11-11", "categories": []}
{"id": "2311.05452", "title": "Transformer-based Model for Oral Epithelial Dysplasia Segmentation", "abstract": "  Oral epithelial dysplasia (OED) is a premalignant histopathological diagnosis\ngiven to lesions of the oral cavity. OED grading is subject to large\ninter/intra-rater variability, resulting in the under/over-treatment of\npatients. We developed a new Transformer-based pipeline to improve detection\nand segmentation of OED in haematoxylin and eosin (H&E) stained whole slide\nimages (WSIs). Our model was trained on OED cases (n = 260) and controls (n =\n105) collected using three different scanners, and validated on test data from\nthree external centres in the United Kingdom and Brazil (n = 78). Our internal\nexperiments yield a mean F1-score of 0.81 for OED segmentation, which reduced\nslightly to 0.71 on external testing, showing good generalisability, and\ngaining state-of-the-art results. This is the first externally validated study\nto use Transformers for segmentation in precancerous histology images. Our\npublicly available model shows great promise to be the first step of a\nfully-integrated pipeline, allowing earlier and more efficient OED diagnosis,\nultimately benefiting patient outcomes.\n", "date": "2023-11-09", "categories": []}
{"id": "2302.03014", "title": "Detection and Localization of Melanoma Skin Cancer in Histopathological\n  Whole Slide Images", "abstract": "  Melanoma diagnosed and treated in its early stages can increase the survival\nrate. A projected increase in skin cancer incidents and a dearth of\ndermatopathologists have emphasized the need for computational pathology\n(CPATH) systems. CPATH systems with deep learning (DL) models have the\npotential to identify the presence of melanoma by exploiting underlying\nmorphological and cellular features. This paper proposes a DL method to detect\nmelanoma and distinguish between normal skin and benign/malignant melanocytic\nlesions in Whole Slide Images (WSI). Our method detects lesions with high\naccuracy and localizes them on a WSI to identify potential regions of interest\nfor pathologists. Interestingly, our DL method relies on using a single CNN\nnetwork to create localization maps first and use them to perform slide-level\npredictions to determine patients who have melanoma. Our best model provides\nfavorable patch-wise classification results with a 0.992 F1 score and 0.99\nsensitivity on unseen data. The source code is\nhttps://github.com/RogerAmundsen/Melanoma-Diagnosis-and-Localization-from-Whole-Slide-Images-using-Convolutional-Neural-Networks.\n", "date": "2023-02-06", "categories": []}
{"id": "2212.06515", "title": "AdvMIL: Adversarial Multiple Instance Learning for the Survival Analysis\n  on Whole-Slide Images", "abstract": "  The survival analysis on histological whole-slide images (WSIs) is one of the\nmost important means to estimate patient prognosis. Although many\nweakly-supervised deep learning models have been developed for gigapixel WSIs,\ntheir potential is generally restricted by classical survival analysis rules\nand fully-supervised learning requirements. As a result, these models provide\npatients only with a completely-certain point estimation of time-to-event, and\nthey could only learn from the labeled WSI data currently at a small scale. To\ntackle these problems, we propose a novel adversarial multiple instance\nlearning (AdvMIL) framework. This framework is based on adversarial\ntime-to-event modeling, and integrates the multiple instance learning (MIL)\nthat is much necessary for WSI representation learning. It is a plug-and-play\none, so that most existing MIL-based end-to-end methods can be easily upgraded\nby applying this framework, gaining the improved abilities of survival\ndistribution estimation and semi-supervised learning. Our extensive experiments\nshow that AdvMIL not only could often bring performance improvement to\nmainstream WSI survival analysis methods at a relatively low computational\ncost, but also enables these methods to effectively utilize unlabeled data via\nsemi-supervised learning. Moreover, it is observed that AdvMIL could help\nimproving the robustness of models against patch occlusion and two\nrepresentative image noises. The proposed AdvMIL framework could promote the\nresearch of survival analysis in computational pathology with its novel\nadversarial MIL paradigm.\n", "date": "2022-12-13", "categories": []}
{"id": "2206.06862", "title": "Evaluating histopathology transfer learning with ChampKit", "abstract": "  Histopathology remains the gold standard for diagnosis of various cancers.\nRecent advances in computer vision, specifically deep learning, have\nfacilitated the analysis of histopathology images for various tasks, including\nimmune cell detection and microsatellite instability classification. The\nstate-of-the-art for each task often employs base architectures that have been\npretrained for image classification on ImageNet. The standard approach to\ndevelop classifiers in histopathology tends to focus narrowly on optimizing\nmodels for a single task, not considering the aspects of modeling innovations\nthat improve generalization across tasks. Here we present ChampKit\n(Comprehensive Histopathology Assessment of Model Predictions toolKit): an\nextensible, fully reproducible benchmarking toolkit that consists of a broad\ncollection of patch-level image classification tasks across different cancers.\nChampKit enables a way to systematically document the performance impact of\nproposed improvements in models and methodology. ChampKit source code and data\nare freely accessible at https://github.com/kaczmarj/champkit .\n", "date": "2022-06-14", "categories": []}
{"id": "2310.20427", "title": "Assessing and Enhancing Robustness of Deep Learning Models with\n  Corruption Emulation in Digital Pathology", "abstract": "  Deep learning in digital pathology brings intelligence and automation as\nsubstantial enhancements to pathological analysis, the gold standard of\nclinical diagnosis. However, multiple steps from tissue preparation to slide\nimaging introduce various image corruptions, making it difficult for deep\nneural network (DNN) models to achieve stable diagnostic results for clinical\nuse. In order to assess and further enhance the robustness of the models, we\nanalyze the physical causes of the full-stack corruptions throughout the\npathological life-cycle and propose an Omni-Corruption Emulation (OmniCE)\nmethod to reproduce 21 types of corruptions quantified with 5-level severity.\nWe then construct three OmniCE-corrupted benchmark datasets at both patch level\nand slide level and assess the robustness of popular DNNs in classification and\nsegmentation tasks. Further, we explore to use the OmniCE-corrupted datasets as\naugmentation data for training and experiments to verify that the\ngeneralization ability of the models has been significantly enhanced.\n", "date": "2023-10-31", "categories": []}
{"id": "2310.19656", "title": "Domain Generalization in Computational Pathology: Survey and Guidelines", "abstract": "  Deep learning models have exhibited exceptional effectiveness in\nComputational Pathology (CPath) by tackling intricate tasks across an array of\nhistology image analysis applications. Nevertheless, the presence of\nout-of-distribution data (stemming from a multitude of sources such as\ndisparate imaging devices and diverse tissue preparation methods) can cause\n\\emph{domain shift} (DS). DS decreases the generalization of trained models to\nunseen datasets with slightly different data distributions, prompting the need\nfor innovative \\emph{domain generalization} (DG) solutions. Recognizing the\npotential of DG methods to significantly influence diagnostic and prognostic\nmodels in cancer studies and clinical practice, we present this survey along\nwith guidelines on achieving DG in CPath. We rigorously define various DS\ntypes, systematically review and categorize existing DG approaches and\nresources in CPath, and provide insights into their advantages, limitations,\nand applicability. We also conduct thorough benchmarking experiments with 28\ncutting-edge DG algorithms to address a complex DG problem. Our findings\nsuggest that careful experiment design and CPath-specific Stain Augmentation\ntechnique can be very effective. However, there is no one-size-fits-all\nsolution for DG in CPath. Therefore, we establish clear guidelines for\ndetecting and managing DS depending on different scenarios. While most of the\nconcepts, guidelines, and recommendations are given for applications in CPath,\nwe believe that they are applicable to most medical image analysis tasks as\nwell.\n", "date": "2023-10-30", "categories": []}
{"id": "2306.01859", "title": "Spatially Resolved Gene Expression Prediction from H&E Histology Images\n  via Bi-modal Contrastive Learning", "abstract": "  Histology imaging is an important tool in medical diagnosis and research,\nenabling the examination of tissue structure and composition at the microscopic\nlevel. Understanding the underlying molecular mechanisms of tissue architecture\nis critical in uncovering disease mechanisms and developing effective\ntreatments. Gene expression profiling provides insight into the molecular\nprocesses underlying tissue architecture, but the process can be time-consuming\nand expensive. We present BLEEP (Bi-modaL Embedding for Expression Prediction),\na bi-modal embedding framework capable of generating spatially resolved gene\nexpression profiles of whole-slide Hematoxylin and eosin (H&E) stained\nhistology images. BLEEP uses contrastive learning to construct a\nlow-dimensional joint embedding space from a reference dataset using paired\nimage and expression profiles at micrometer resolution. With this approach, the\ngene expression of any query image patch can be imputed using the expression\nprofiles from the reference dataset. We demonstrate BLEEP's effectiveness in\ngene expression prediction by benchmarking its performance on a human liver\ntissue dataset captured using the 10x Visium platform, where it achieves\nsignificant improvements over existing methods. Our results demonstrate the\npotential of BLEEP to provide insights into the molecular mechanisms underlying\ntissue architecture, with important implications in diagnosis and research of\nvarious diseases. The proposed approach can significantly reduce the time and\ncost associated with gene expression profiling, opening up new avenues for\nhigh-throughput analysis of histology images for both research and clinical\napplications.\n", "date": "2023-06-02", "categories": []}
{"id": "2308.00806", "title": "Addressing Uncertainty in Imbalanced Histopathology Image Classification\n  of HER2 Breast Cancer: An interpretable Ensemble Approach with Threshold\n  Filtered Single Instance Evaluation (SIE)", "abstract": "  Breast Cancer (BC) is among women's most lethal health concerns. Early\ndiagnosis can alleviate the mortality rate by helping patients make efficient\ntreatment decisions. Human Epidermal Growth Factor Receptor (HER2) has become\none the most lethal subtype of BC. According to the College of American\nPathologists American Society of Clinical Oncology (CAP/ASCO), the severity\nlevel of HER2 expression can be classified between 0 and 3+ range. HER2 can be\ndetected effectively from immunohistochemical (IHC) and, hematoxylin & eosin\n(HE) images of different classes such as 0, 1+, 2+, and 3+. An ensemble\napproach integrated with threshold filtered single instance evaluation (SIE)\ntechnique has been proposed in this study to diagnose BC from the\nmulti-categorical expression of HER2 subtypes. Initially, DenseNet201 and\nXception have been ensembled into a single classifier as feature extractors\nwith an effective combination of global average pooling, dropout layer, dense\nlayer with a swish activation function, and l2 regularizer, batch\nnormalization, etc. After that, extracted features has been processed through\nsingle instance evaluation (SIE) to determine different confidence levels and\nadjust decision boundary among the imbalanced classes. This study has been\nconducted on the BC immunohistochemical (BCI) dataset, which is classified by\npathologists into four stages of HER2 BC. This proposed approach known as\nDenseNet201-Xception-SIE with a threshold value of 0.7 surpassed all other\nexisting state-of-art models with an accuracy of 97.12%, precision of 97.15%,\nand recall of 97.68% on H&E data and, accuracy of 97.56%, precision of 97.57%,\nand recall of 98.00% on IHC data respectively, maintaining momentous\nimprovement. Finally, Grad-CAM and Guided Grad-CAM have been employed in this\nstudy to interpret, how TL-based model works on the histopathology dataset and\nmake decisions from the data.\n", "date": "2023-08-01", "categories": []}
{"id": "2310.18192", "title": "Artifact-Robust Graph-Based Learning in Digital Pathology", "abstract": "  Whole slide images~(WSIs) are digitized images of tissues placed in glass\nslides using advanced scanners. The digital processing of WSIs is challenging\nas they are gigapixel images and stored in multi-resolution format. A common\nchallenge with WSIs is that perturbations/artifacts are inevitable during\nstoring the glass slides and digitizing them. These perturbations include\nmotion, which often arises from slide movement during placement, and changes in\nhue and brightness due to variations in staining chemicals and the quality of\ndigitizing scanners. In this work, a novel robust learning approach to account\nfor these artifacts is presented. Due to the size and resolution of WSIs and to\naccount for neighborhood information, graph-based methods are called for. We\nuse graph convolutional network~(GCN) to extract features from the graph\nrepresenting WSI. Through a denoiser {and pooling layer}, the effects of\nperturbations in WSIs are controlled and the output is followed by a\ntransformer for the classification of different grades of prostate cancer. To\ncompare the efficacy of the proposed approach, the model without denoiser is\ntrained and tested with WSIs without any perturbation and then different\nperturbations are introduced in WSIs and passed through the network with the\ndenoiser. The accuracy and kappa scores of the proposed model with prostate\ncancer dataset compared with non-robust algorithms show significant improvement\nin cancer diagnosis.\n", "date": "2023-10-27", "categories": []}
{"id": "2310.12769", "title": "Mixing Histopathology Prototypes into Robust Slide-Level Representations\n  for Cancer Subtyping", "abstract": "  Whole-slide image analysis via the means of computational pathology often\nrelies on processing tessellated gigapixel images with only slide-level labels\navailable. Applying multiple instance learning-based methods or transformer\nmodels is computationally expensive as, for each image, all instances have to\nbe processed simultaneously. The MLP-Mixer is an under-explored alternative\nmodel to common vision transformers, especially for large-scale datasets. Due\nto the lack of a self-attention mechanism, they have linear computational\ncomplexity to the number of input patches but achieve comparable performance on\nnatural image datasets. We propose a combination of feature embedding and\nclustering to preprocess the full whole-slide image into a reduced prototype\nrepresentation which can then serve as input to a suitable MLP-Mixer\narchitecture. Our experiments on two public benchmarks and one inhouse\nmalignant lymphoma dataset show comparable performance to current\nstate-of-the-art methods, while achieving lower training costs in terms of\ncomputational time and memory load. Code is publicly available at\nhttps://github.com/butkej/ProtoMixer.\n", "date": "2023-10-19", "categories": []}
{"id": "2310.12866", "title": "Predicting Ovarian Cancer Treatment Response in Histopathology using\n  Hierarchical Vision Transformers and Multiple Instance Learning", "abstract": "  For many patients, current ovarian cancer treatments offer limited clinical\nbenefit. For some therapies, it is not possible to predict patients' responses,\npotentially exposing them to the adverse effects of treatment without any\ntherapeutic benefit. As part of the automated prediction of treatment\neffectiveness in ovarian cancer using histopathological images (ATEC23)\nchallenge, we evaluated the effectiveness of deep learning to predict whether a\ncourse of treatment including the antiangiogenic drug bevacizumab could\ncontribute to remission or prevent disease progression for at least 6 months in\na set of 282 histopathology whole slide images (WSIs) from 78 ovarian cancer\npatients. Our approach used a pretrained Hierarchical Image Pyramid Transformer\n(HIPT) to extract region-level features and an attention-based multiple\ninstance learning (ABMIL) model to aggregate features and classify whole\nslides. The optimal HIPT-ABMIL model had an internal balanced accuracy of 60.2%\n+- 2.9% and an AUC of 0.646 +- 0.033. Histopathology-specific model pretraining\nwas found to be beneficial to classification performance, though hierarchical\ntransformers were not, with a ResNet feature extractor achieving similar\nperformance. Due to the dataset being small and highly heterogeneous,\nperformance was variable across 5-fold cross-validation folds, and there were\nsome extreme differences between validation and test set performance within\nfolds. The model did not generalise well to tissue microarrays, with accuracy\nworse than random chance. It is not yet clear whether ovarian cancer WSIs\ncontain information that can be used to accurately predict treatment response,\nwith further validation using larger, higher-quality datasets required.\n", "date": "2023-10-19", "categories": []}
{"id": "2310.07682", "title": "Prediction of MET Overexpression in Non-Small Cell Lung Adenocarcinomas\n  from Hematoxylin and Eosin Images", "abstract": "  MET protein overexpression is a targetable event in non-small cell lung\ncancer (NSCLC) and is the subject of active drug development. Challenges in\nidentifying patients for these therapies include lack of access to validated\ntesting, such as standardized immunohistochemistry (IHC) assessment, and\nconsumption of valuable tissue for a single gene/protein assay. Development of\npre-screening algorithms using routinely available digitized hematoxylin and\neosin (H&E)-stained slides to predict MET overexpression could promote testing\nfor those who will benefit most. While assessment of MET expression using IHC\nis currently not routinely performed in NSCLC, next-generation sequencing is\ncommon and in some cases includes RNA expression panel testing. In this work,\nwe leveraged a large database of matched H&E slides and RNA expression data to\ntrain a weakly supervised model to predict MET RNA overexpression directly from\nH&E images. This model was evaluated on an independent holdout test set of 300\nover-expressed and 289 normal patients, demonstrating an ROC-AUC of 0.70 (95th\npercentile interval: 0.66 - 0.74) with stable performance characteristics\nacross different patient clinical variables and robust to synthetic noise on\nthe test set. These results suggest that H&E-based predictive models could be\nuseful to prioritize patients for confirmatory testing of MET protein or MET\ngene expression status.\n", "date": "2023-10-11", "categories": []}
{"id": "2310.08743", "title": "Development and Validation of a Deep Learning-Based Microsatellite\n  Instability Predictor from Prostate Cancer Whole-Slide Images", "abstract": "  Microsatellite instability-high (MSI-H) is a tumor agnostic biomarker for\nimmune checkpoint inhibitor therapy. However, MSI status is not routinely\ntested in prostate cancer, in part due to low prevalence and assay cost. As\nsuch, prediction of MSI status from hematoxylin and eosin (H&E) stained\nwhole-slide images (WSIs) could identify prostate cancer patients most likely\nto benefit from confirmatory testing and becoming eligible for immunotherapy.\nProstate biopsies and surgical resections from de-identified records of\nconsecutive prostate cancer patients referred to our institution were analyzed.\nTheir MSI status was determined by next generation sequencing. Patients before\na cutoff date were split into an algorithm development set (n=4015, MSI-H 1.8%)\nand a paired validation set (n=173, MSI-H 19.7%) that consisted of two serial\nsections from each sample, one stained and scanned internally and the other at\nan external site. Patients after the cutoff date formed the temporal validation\nset (n=1350, MSI-H 2.3%). Attention-based multiple instance learning models\nwere trained to predict MSI-H from H&E WSIs. The MSI-H predictor achieved area\nunder the receiver operating characteristic curve values of 0.78 (95% CI\n[0.69-0.86]), 0.72 (95% CI [0.63-0.81]), and 0.72 (95% CI [0.62-0.82]) on the\ninternally prepared, externally prepared, and temporal validation sets,\nrespectively. While MSI-H status is significantly correlated with Gleason\nscore, the model remained predictive within each Gleason score subgroup. In\nsummary, we developed and validated an AI-based MSI-H diagnostic model on a\nlarge real-world cohort of routine H&E slides, which effectively generalized to\nexternally stained and scanned samples and a temporally independent validation\ncohort. This algorithm has the potential to direct prostate cancer patients\ntoward immunotherapy and to identify MSI-H cases secondary to Lynch syndrome.\n", "date": "2023-10-12", "categories": []}
{"id": "2310.07033", "title": "Computational Pathology at Health System Scale -- Self-Supervised\n  Foundation Models from Three Billion Images", "abstract": "  Recent breakthroughs in self-supervised learning have enabled the use of\nlarge unlabeled datasets to train visual foundation models that can generalize\nto a variety of downstream tasks. While this training paradigm is well suited\nfor the medical domain where annotations are scarce, large-scale pre-training\nin the medical domain, and in particular pathology, has not been extensively\nstudied. Previous work in self-supervised learning in pathology has leveraged\nsmaller datasets for both pre-training and evaluating downstream performance.\nThe aim of this project is to train the largest academic foundation model and\nbenchmark the most prominent self-supervised learning algorithms by\npre-training and evaluating downstream performance on large clinical pathology\ndatasets. We collected the largest pathology dataset to date, consisting of\nover 3 billion images from over 423 thousand microscopy slides. We compared\npre-training of visual transformer models using the masked autoencoder (MAE)\nand DINO algorithms. We evaluated performance on six clinically relevant tasks\nfrom three anatomic sites and two institutions: breast cancer detection,\ninflammatory bowel disease detection, breast cancer estrogen receptor\nprediction, lung adenocarcinoma EGFR mutation prediction, and lung cancer\nimmunotherapy response prediction. Our results demonstrate that pre-training on\npathology data is beneficial for downstream performance compared to\npre-training on natural images. Additionally, the DINO algorithm achieved\nbetter generalization performance across all tasks tested. The presented\nresults signify a phase change in computational pathology research, paving the\nway into a new era of more performant models based on large-scale, parallel\npre-training at the billion-image scale.\n", "date": "2023-10-10", "categories": []}
{"id": "2310.03893", "title": "Characterizing the Features of Mitotic Figures Using a Conditional\n  Diffusion Probabilistic Model", "abstract": "  Mitotic figure detection in histology images is a hard-to-define, yet\nclinically significant task, where labels are generated with pathologist\ninterpretations and where there is no ``gold-standard'' independent\nground-truth. However, it is well-established that these interpretation based\nlabels are often unreliable, in part, due to differences in expertise levels\nand human subjectivity. In this paper, our goal is to shed light on the\ninherent uncertainty of mitosis labels and characterize the mitotic figure\nclassification task in a human interpretable manner. We train a probabilistic\ndiffusion model to synthesize patches of cell nuclei for a given mitosis label\ncondition. Using this model, we can then generate a sequence of synthetic\nimages that correspond to the same nucleus transitioning into the mitotic\nstate. This allows us to identify different image features associated with\nmitosis, such as cytoplasm granularity, nuclear density, nuclear irregularity\nand high contrast between the nucleus and the cell body. Our approach offers a\nnew tool for pathologists to interpret and communicate the features driving the\ndecision to recognize a mitotic figure.\n", "date": "2023-10-05", "categories": []}
{"id": "2310.03106", "title": "Creating an Atlas of Normal Tissue for Pruning WSI Patching Through\n  Anomaly Detection", "abstract": "  Patching gigapixel whole slide images (WSIs) is an important task in\ncomputational pathology. Some methods have been proposed to select a subset of\npatches as WSI representation for downstream tasks. While most of the\ncomputational pathology tasks are designed to classify or detect the presence\nof pathological lesions in each WSI, the confounding role and redundant nature\nof normal histology in tissue samples are generally overlooked in WSI\nrepresentations. In this paper, we propose and validate the concept of an\n\"atlas of normal tissue\" solely using samples of WSIs obtained from normal\ntissue biopsies. Such atlases can be employed to eliminate normal fragments of\ntissue samples and hence increase the representativeness collection of patches.\nWe tested our proposed method by establishing a normal atlas using 107 normal\nskin WSIs and demonstrated how established indexes and search engines like\nYottixel can be improved. We used 553 WSIs of cutaneous squamous cell carcinoma\n(cSCC) to show the advantage. We also validated our method applied to an\nexternal dataset of 451 breast WSIs. The number of selected WSI patches was\nreduced by 30% to 50% after utilizing the proposed normal atlas while\nmaintaining the same indexing and search performance in leave-one-patinet-out\nvalidation for both datasets. We show that the proposed normal atlas shows\npromise for unsupervised selection of the most representative patches of the\nabnormal/malignant WSI lesions.\n", "date": "2023-10-04", "categories": []}
{"id": "2310.03279", "title": "Classifying Whole Slide Images: What Matters?", "abstract": "  Recently there have been many algorithms proposed for the classification of\nvery high resolution whole slide images (WSIs). These new algorithms are mostly\nfocused on finding novel ways to combine the information from small local\npatches extracted from the slide, with an emphasis on effectively aggregating\nmore global information for the final predictor. In this paper we thoroughly\nexplore different key design choices for WSI classification algorithms to\ninvestigate what matters most for achieving high accuracy. Surprisingly, we\nfound that capturing global context information does not necessarily mean\nbetter performance. A model that captures the most global information\nconsistently performs worse than a model that captures less global information.\nIn addition, a very simple multi-instance learning method that captures no\nglobal information performs almost as well as models that capture a lot of\nglobal information. These results suggest that the most important features for\neffective WSI classification are captured at the local small patch level, where\ncell and tissue micro-environment detail is most pronounced. Another surprising\nfinding was that unsupervised pre-training on a larger set of 33 cancers gives\nsignificantly worse performance compared to pre-training on a smaller dataset\nof 7 cancers (including the target cancer). We posit that pre-training on a\nsmaller, more focused dataset allows the feature extractor to make better use\nof the limited feature space to better discriminate between subtle differences\nin the input patch.\n", "date": "2023-10-05", "categories": []}
{"id": "2309.10650", "title": "MUSTANG: Multi-Stain Self-Attention Graph Multiple Instance Learning\n  Pipeline for Histopathology Whole Slide Images", "abstract": "  Whole Slide Images (WSIs) present a challenging computer vision task due to\ntheir gigapixel size and presence of numerous artefacts. Yet they are a\nvaluable resource for patient diagnosis and stratification, often representing\nthe gold standard for diagnostic tasks. Real-world clinical datasets tend to\ncome as sets of heterogeneous WSIs with labels present at the patient-level,\nwith poor to no annotations. Weakly supervised attention-based multiple\ninstance learning approaches have been developed in recent years to address\nthese challenges, but can fail to resolve both long and short-range\ndependencies. Here we propose an end-to-end multi-stain self-attention graph\n(MUSTANG) multiple instance learning pipeline, which is designed to solve a\nweakly-supervised gigapixel multi-image classification task, where the label is\nassigned at the patient-level, but no slide-level labels or region annotations\nare available. The pipeline uses a self-attention based approach by restricting\nthe operations to a highly sparse k-Nearest Neighbour Graph of embedded WSI\npatches based on the Euclidean distance. We show this approach achieves a\nstate-of-the-art F1-score/AUC of 0.89/0.92, outperforming the widely used CLAM\nmodel. Our approach is highly modular and can easily be modified to suit\ndifferent clinical datasets, as it only requires a patient-level label without\nannotations and accepts WSI sets of different sizes, as the graphs can be of\nvarying sizes and structures. The source code can be found at\nhttps://github.com/AmayaGS/MUSTANG.\n", "date": "2023-09-19", "categories": []}
{"id": "2310.01924", "title": "RoFormer for Position Aware Multiple Instance Learning in Whole Slide\n  Image Classification", "abstract": "  Whole slide image (WSI) classification is a critical task in computational\npathology. However, the gigapixel-size of such images remains a major challenge\nfor the current state of deep-learning. Current methods rely on\nmultiple-instance learning (MIL) models with frozen feature extractors. Given\nthe the high number of instances in each image, MIL methods have long assumed\nindependence and permutation-invariance of patches, disregarding the tissue\nstructure and correlation between patches. Recent works started studying this\ncorrelation between instances but the computational workload of such a high\nnumber of tokens remained a limiting factor. In particular, relative position\nof patches remains unaddressed. We propose to apply a straightforward encoding\nmodule, namely a RoFormer layer , relying on memory-efficient exact\nself-attention and relative positional encoding. This module can perform full\nself-attention with relative position encoding on patches of large and\narbitrary shaped WSIs, solving the need for correlation between instances and\nspatial modeling of tissues. We demonstrate that our method outperforms\nstate-of-the-art MIL models on three commonly used public datasets (TCGA-NSCLC,\nBRACS and Camelyon16)) on weakly supervised classification tasks. Code is\navailable at https://github.com/Sanofi-Public/DDS-RoFormerMIL\n", "date": "2023-10-03", "categories": []}
{"id": "2211.07384", "title": "Language models are good pathologists: using attention-based sequence\n  reduction and text-pretrained transformers for efficient WSI classification", "abstract": "  In digital pathology, Whole Slide Image (WSI) analysis is usually formulated\nas a Multiple Instance Learning (MIL) problem. Although transformer-based\narchitectures have been used for WSI classification, these methods require\nmodifications to adapt them to specific challenges of this type of image data.\nAmong these challenges is the amount of memory and compute required by deep\ntransformer models to process long inputs, such as the thousands of image\npatches that can compose a WSI at $\\times 10$ or $\\times 20$ magnification. We\nintroduce \\textit{SeqShort}, a multi-head attention-based sequence shortening\nlayer to summarize each WSI in a fixed- and short-sized sequence of instances,\nthat allows us to reduce the computational costs of self-attention on long\nsequences, and to include positional information that is unavailable in other\nMIL approaches. Furthermore, we show that WSI classification performance can be\nimproved when the downstream transformer architecture has been pre-trained on a\nlarge corpus of text data, and only fine-tuning less than 0.1\\% of its\nparameters. We demonstrate the effectiveness of our method in lymph node\nmetastases classification and cancer subtype classification tasks, without the\nneed of designing a WSI-specific transformer nor doing in-domain pre-training,\nkeeping a reduced compute budget and low number of trainable parameters.\n", "date": "2022-11-14", "categories": []}
{"id": "2309.07394", "title": "Nucleus-aware Self-supervised Pretraining Using Unpaired Image-to-image\n  Translation for Histopathology Images", "abstract": "  Self-supervised pretraining attempts to enhance model performance by\nobtaining effective features from unlabeled data, and has demonstrated its\neffectiveness in the field of histopathology images. Despite its success, few\nworks concentrate on the extraction of nucleus-level information, which is\nessential for pathologic analysis. In this work, we propose a novel\nnucleus-aware self-supervised pretraining framework for histopathology images.\nThe framework aims to capture the nuclear morphology and distribution\ninformation through unpaired image-to-image translation between histopathology\nimages and pseudo mask images. The generation process is modulated by both\nconditional and stochastic style representations, ensuring the reality and\ndiversity of the generated histopathology images for pretraining. Further, an\ninstance segmentation guided strategy is employed to capture instance-level\ninformation. The experiments on 7 datasets show that the proposed pretraining\nmethod outperforms supervised ones on Kather classification, multiple instance\nlearning, and 5 dense-prediction tasks with the transfer learning protocol, and\nyields superior results than other self-supervised approaches on 8\nsemi-supervised tasks. Our project is publicly available at\nhttps://github.com/zhiyuns/UNITPathSSL.\n", "date": "2023-09-14", "categories": []}
{"id": "2306.03407", "title": "LESS: Label-efficient Multi-scale Learning for Cytological Whole Slide\n  Image Screening", "abstract": "  In computational pathology, multiple instance learning (MIL) is widely used\nto circumvent the computational impasse in giga-pixel whole slide image (WSI)\nanalysis. It usually consists of two stages: patch-level feature extraction and\nslide-level aggregation. Recently, pretrained models or self-supervised\nlearning have been used to extract patch features, but they suffer from low\neffectiveness or inefficiency due to overlooking the task-specific supervision\nprovided by slide labels. Here we propose a weakly-supervised Label-Efficient\nWSI Screening method, dubbed LESS, for cytological WSI analysis with only\nslide-level labels, which can be effectively applied to small datasets. First,\nwe suggest using variational positive-unlabeled (VPU) learning to uncover\nhidden labels of both benign and malignant patches. We provide appropriate\nsupervision by using slide-level labels to improve the learning of patch-level\nfeatures. Next, we take into account the sparse and random arrangement of cells\nin cytological WSIs. To address this, we propose a strategy to crop patches at\nmultiple scales and utilize a cross-attention vision transformer (CrossViT) to\ncombine information from different scales for WSI classification. The\ncombination of our two steps achieves task-alignment, improving effectiveness\nand efficiency. We validate the proposed label-efficient method on a urine\ncytology WSI dataset encompassing 130 samples (13,000 patches) and FNAC 2019\ndataset with 212 samples (21,200 patches). The experiment shows that the\nproposed LESS reaches 84.79%, 85.43%, 91.79% and 78.30% on a urine cytology WSI\ndataset, and 96.88%, 96.86%, 98.95%, 97.06% on FNAC 2019 dataset in terms of\naccuracy, AUC, sensitivity and specificity. It outperforms state-of-the-art MIL\nmethods on pathology WSIs and realizes automatic cytological WSI cancer\nscreening.\n", "date": "2023-06-06", "categories": []}
{"id": "2303.09987", "title": "Breast Cancer Histopathology Image based Gene Expression Prediction\n  using Spatial Transcriptomics data and Deep Learning", "abstract": "  Tumour heterogeneity in breast cancer poses challenges in predicting outcome\nand response to therapy. Spatial transcriptomics technologies may address these\nchallenges, as they provide a wealth of information about gene expression at\nthe cell level, but they are expensive, hindering their use in large-scale\nclinical oncology studies. Predicting gene expression from hematoxylin and\neosin stained histology images provides a more affordable alternative for such\nstudies. Here we present BrST-Net, a deep learning framework for predicting\ngene expression from histopathology images using spatial transcriptomics data.\nUsing this framework, we trained and evaluated 10 state-of-the-art deep\nlearning models without utilizing pretrained weights for the prediction of 250\ngenes. To enhance the generalisation performance of the main network, we\nintroduce an auxiliary network into the framework. Our methodology outperforms\nprevious studies, with 237 genes identified with positive correlation,\nincluding 24 genes with a median correlation coefficient greater than 0.50.\nThis is a notable improvement over previous studies, which could predict only\n102 genes with positive correlation, with the highest correlation values\nranging from 0.29 to 0.34.\n", "date": "2023-03-17", "categories": []}
{"id": "2309.06421", "title": "AGMDT: Virtual Staining of Renal Histology Images with Adjacency-Guided\n  Multi-Domain Transfer", "abstract": "  Renal pathology, as the gold standard of kidney disease diagnosis, requires\ndoctors to analyze a series of tissue slices stained by H&E staining and\nspecial staining like Masson, PASM, and PAS, respectively. These special\nstaining methods are costly, time-consuming, and hard to standardize for wide\nuse especially in primary hospitals. Advances of supervised learning methods\nhave enabled the virtually conversion of H&E images into special staining\nimages, but achieving pixel-to-pixel alignment for training remains\nchallenging. In contrast, unsupervised learning methods regarding different\nstains as different style transfer domains can utilize unpaired data, but they\nignore the spatial inter-domain correlations and thus decrease the\ntrustworthiness of structural details for diagnosis. In this paper, we propose\na novel virtual staining framework AGMDT to translate images into other domains\nby avoiding pixel-level alignment and meanwhile utilizing the correlations\namong adjacent tissue slices. We first build a high-quality multi-domain renal\nhistological dataset where each specimen case comprises a series of slices\nstained in various ways. Based on it, the proposed framework AGMDT discovers\npatch-level aligned pairs across the serial slices of multi-domains through\nglomerulus detection and bipartite graph matching, and utilizes such\ncorrelations to supervise the end-to-end model for multi-domain staining\ntransformation. Experimental results show that the proposed AGMDT achieves a\ngood balance between the precise pixel-level alignment and unpaired domain\ntransfer by exploiting correlations across multi-domain serial pathological\nslices, and outperforms the state-of-the-art methods in both quantitative\nmeasure and morphological details.\n", "date": "2023-09-12", "categories": []}
{"id": "2309.08259", "title": "BROW: Better featuRes fOr Whole slide image based on self-distillation", "abstract": "  Whole slide image (WSI) processing is becoming part of the key components of\nstandard clinical diagnosis for various diseases. However, the direct\napplication of conventional image processing algorithms to WSI faces certain\nobstacles because of WSIs' distinct property: the super-high resolution. The\nperformance of most WSI-related tasks relies on the efficacy of the backbone\nwhich extracts WSI patch feature representations. Hence, we proposed BROW, a\nfoundation model for extracting better feature representations for WSIs, which\ncan be conveniently adapted to downstream tasks without or with slight\nfine-tuning. The model takes transformer architecture, pretrained using\nself-distillation framework. To improve model's robustness, techniques such as\npatch shuffling have been employed. Additionally, the model leverages the\nunique properties of WSIs, utilizing WSI's multi-scale pyramid to incorporate\nan additional global view, thereby further enhancing its performance. We used\nboth private and public data to make up a large pretraining dataset, containing\nmore than 11000 slides, over 180M extracted patches, encompassing WSIs related\nto various organs and tissues. To assess the effectiveness of \\ourmodel, we run\na wide range of downstream tasks, including slide-level subtyping, patch-level\nclassification and nuclei instance segmentation. The results confirmed the\nefficacy, robustness and good generalization ability of the proposed model.\nThis substantiates its potential as foundation model for WSI feature extraction\nand highlights promising prospects for its application in WSI processing.\n", "date": "2023-09-15", "categories": []}
{"id": "2309.07255", "title": "Automated segmentation of rheumatoid arthritis immunohistochemistry\n  stained synovial tissue", "abstract": "  Rheumatoid Arthritis (RA) is a chronic, autoimmune disease which primarily\naffects the joint's synovial tissue. It is a highly heterogeneous disease, with\nwide cellular and molecular variability observed in synovial tissues. Over the\nlast two decades, the methods available for their study have advanced\nconsiderably. In particular, Immunohistochemistry stains are well suited to\nhighlighting the functional organisation of samples. Yet, analysis of\nIHC-stained synovial tissue samples is still overwhelmingly done manually and\nsemi-quantitatively by expert pathologists. This is because in addition to the\nfragmented nature of IHC stained synovial tissue, there exist wide variations\nin intensity and colour, strong clinical centre batch effect, as well as the\npresence of many undesirable artefacts present in gigapixel Whole Slide Images\n(WSIs), such as water droplets, pen annotation, folded tissue, blurriness, etc.\nThere is therefore a strong need for a robust, repeatable automated tissue\nsegmentation algorithm which can cope with this variability and provide support\nto imaging pipelines. We train a UNET on a hand-curated, heterogeneous\nreal-world multi-centre clinical dataset R4RA, which contains multiple types of\nIHC staining. The model obtains a DICE score of 0.865 and successfully segments\ndifferent types of IHC staining, as well as dealing with variance in colours,\nintensity and common WSIs artefacts from the different clinical centres. It can\nbe used as the first step in an automated image analysis pipeline for synovial\ntissue samples stained with IHC, increasing speed, reproducibility and\nrobustness.\n", "date": "2023-09-13", "categories": []}
{"id": "2309.07400", "title": "HIGT: Hierarchical Interaction Graph-Transformer for Whole Slide Image\n  Analysis", "abstract": "  In computation pathology, the pyramid structure of gigapixel Whole Slide\nImages (WSIs) has recently been studied for capturing various information from\nindividual cell interactions to tissue microenvironments. This hierarchical\nstructure is believed to be beneficial for cancer diagnosis and prognosis\ntasks. However, most previous hierarchical WSI analysis works (1) only\ncharacterize local or global correlations within the WSI pyramids and (2) use\nonly unidirectional interaction between different resolutions, leading to an\nincomplete picture of WSI pyramids. To this end, this paper presents a novel\nHierarchical Interaction Graph-Transformer (i.e., HIGT) for WSI analysis. With\nGraph Neural Network and Transformer as the building commons, HIGT can learn\nboth short-range local information and long-range global representation of the\nWSI pyramids. Considering that the information from different resolutions is\ncomplementary and can benefit each other during the learning process, we\nfurther design a novel Bidirectional Interaction block to establish\ncommunication between different levels within the WSI pyramids. Finally, we\naggregate both coarse-grained and fine-grained features learned from different\nlevels together for slide-level prediction. We evaluate our methods on two\npublic WSI datasets from TCGA projects, i.e., kidney carcinoma (KICA) and\nesophageal carcinoma (ESCA). Experimental results show that our HIGT\noutperforms both hierarchical and non-hierarchical state-of-the-art methods on\nboth tumor subtyping and staging tasks.\n", "date": "2023-09-14", "categories": []}
{"id": "2208.13902", "title": "Radial Prediction Domain Adaption Classifier for the MIDOG 2022\n  Challenge", "abstract": "  This paper describes our contribution to the MIDOG 2022 challenge for\ndetecting mitotic cells. One of the major problems to be addressed in the MIDOG\n2022 challenge is the robustness under the natural variance that appears for\nreal-life data in the histopathology field. To address the problem, we use an\nadapted YOLOv5s model for object detection in conjunction with a new Domain\nAdaption Classifier (DAC) variant, the Radial-Prediction-DAC, to achieve\nrobustness under domain shifts. In addition, we increase the variability of the\navailable training data using stain augmentation in HED color space. Using the\nsuggested method, we obtain a test set F1-score of 0.6658.\n", "date": "2022-08-29", "categories": []}
{"id": "2303.13405", "title": "SC-MIL: Supervised Contrastive Multiple Instance Learning for Imbalanced\n  Classification in Pathology", "abstract": "  Multiple Instance learning (MIL) models have been extensively used in\npathology to predict biomarkers and risk-stratify patients from gigapixel-sized\nimages. Machine learning problems in medical imaging often deal with rare\ndiseases, making it important for these models to work in a label-imbalanced\nsetting. In pathology images, there is another level of imbalance, where given\na positively labeled Whole Slide Image (WSI), only a fraction of pixels within\nit contribute to the positive label. This compounds the severity of imbalance\nand makes imbalanced classification in pathology challenging. Furthermore,\nthese imbalances can occur in out-of-distribution (OOD) datasets when the\nmodels are deployed in the real-world. We leverage the idea that decoupling\nfeature and classifier learning can lead to improved decision boundaries for\nlabel imbalanced datasets. To this end, we investigate the integration of\nsupervised contrastive learning with multiple instance learning (SC-MIL).\nSpecifically, we propose a joint-training MIL framework in the presence of\nlabel imbalance that progressively transitions from learning bag-level\nrepresentations to optimal classifier learning. We perform experiments with\ndifferent imbalance settings for two well-studied problems in cancer pathology:\nsubtyping of non-small cell lung cancer and subtyping of renal cell carcinoma.\nSC-MIL provides large and consistent improvements over other techniques on both\nin-distribution (ID) and OOD held-out sets across multiple imbalanced settings.\n", "date": "2023-03-23", "categories": []}
{"id": "2309.01007", "title": "Comparative Analysis of Deep Learning Architectures for Breast Cancer\n  Diagnosis Using the BreaKHis Dataset", "abstract": "  Cancer is an extremely difficult and dangerous health problem because it\nmanifests in so many different ways and affects so many different organs and\ntissues. The primary goal of this research was to evaluate deep learning\nmodels' ability to correctly identify breast cancer cases using the BreakHis\ndataset. The BreakHis dataset covers a wide range of breast cancer subtypes\nthrough its huge collection of histopathological pictures. In this study, we\nuse and compare the performance of five well-known deep learning models for\ncancer classification: VGG, ResNet, Xception, Inception, and InceptionResNet.\nThe results placed the Xception model at the top, with an F1 score of 0.9 and\nan accuracy of 89%. At the same time, the Inception and InceptionResNet models\nboth hit accuracy of 87% . However, the F1 score for the Inception model was\n87, while that for the InceptionResNet model was 86. These results demonstrate\nthe importance of deep learning methods in making correct breast cancer\ndiagnoses. This highlights the potential to provide improved diagnostic\nservices to patients. The findings of this study not only improve current\nmethods of cancer diagnosis, but also make significant contributions to the\ncreation of new and improved cancer treatment strategies. In a nutshell, the\nresults of this study represent a major advancement in the direction of\nachieving these vital healthcare goals.\n", "date": "2023-09-02", "categories": []}
{"id": "2309.02670", "title": "Progressive Attention Guidance for Whole Slide Vulvovaginal Candidiasis\n  Screening", "abstract": "  Vulvovaginal candidiasis (VVC) is the most prevalent human candidal\ninfection, estimated to afflict approximately 75% of all women at least once in\ntheir lifetime. It will lead to several symptoms including pruritus, vaginal\nsoreness, and so on. Automatic whole slide image (WSI) classification is highly\ndemanded, for the huge burden of disease control and prevention. However, the\nWSI-based computer-aided VCC screening method is still vacant due to the scarce\nlabeled data and unique properties of candida. Candida in WSI is challenging to\nbe captured by conventional classification models due to its distinctive\nelongated shape, the small proportion of their spatial distribution, and the\nstyle gap from WSIs. To make the model focus on the candida easier, we propose\nan attention-guided method, which can obtain a robust diagnosis classification\nmodel. Specifically, we first use a pre-trained detection model as prior\ninstruction to initialize the classification model. Then we design a Skip\nSelf-Attention module to refine the attention onto the fined-grained features\nof candida. Finally, we use a contrastive learning method to alleviate the\noverfitting caused by the style gap of WSIs and suppress the attention to false\npositive regions. Our experimental results demonstrate that our framework\nachieves state-of-the-art performance. Code and example data are available at\nhttps://github.com/cjdbehumble/MICCAI2023-VVC-Screening.\n", "date": "2023-09-06", "categories": []}
{"id": "2309.02702", "title": "Gene-induced Multimodal Pre-training for Image-omic Classification", "abstract": "  Histology analysis of the tumor micro-environment integrated with genomic\nassays is the gold standard for most cancers in modern medicine. This paper\nproposes a Gene-induced Multimodal Pre-training (GiMP) framework, which jointly\nincorporates genomics and Whole Slide Images (WSIs) for classification tasks.\nOur work aims at dealing with the main challenges of multi-modality image-omic\nclassification w.r.t. (1) the patient-level feature extraction difficulties\nfrom gigapixel WSIs and tens of thousands of genes, and (2) effective fusion\nconsidering high-order relevance modeling. Concretely, we first propose a group\nmulti-head self-attention gene encoder to capture global structured features in\ngene expression cohorts. We design a masked patch modeling paradigm (MPM) to\ncapture the latent pathological characteristics of different tissues. The mask\nstrategy is randomly masking a fixed-length contiguous subsequence of patch\nembeddings of a WSI. Finally, we combine the classification tokens of paired\nmodalities and propose a triplet learning module to learn high-order relevance\nand discriminative patient-level information.After pre-training, a simple\nfine-tuning can be adopted to obtain the classification results. Experimental\nresults on the TCGA dataset show the superiority of our network architectures\nand our pre-training framework, achieving 99.47% in accuracy for image-omic\nclassification. The code is publicly available at\nhttps://github.com/huangwudiduan/GIMP.\n", "date": "2023-09-06", "categories": []}
{"id": "2306.05029", "title": "Multi-level Multiple Instance Learning with Transformer for Whole Slide\n  Image Classification", "abstract": "  Whole slide image (WSI) refers to a type of high-resolution scanned tissue\nimage, which is extensively employed in computer-assisted diagnosis (CAD). The\nextremely high resolution and limited availability of region-level annotations\nmake employing deep learning methods for WSI-based digital diagnosis\nchallenging. Recently integrating multiple instance learning (MIL) and\nTransformer for WSI analysis shows very promising results. However, designing\neffective Transformers for this weakly-supervised high-resolution image\nanalysis is an underexplored yet important problem. In this paper, we propose a\nMulti-level MIL (MMIL) scheme by introducing a hierarchical structure to MIL,\nwhich enables efficient handling of MIL tasks involving a large number of\ninstances. Based on MMIL, we instantiated MMIL-Transformer, an efficient\nTransformer model with windowed exact self-attention for large-scale MIL tasks.\nTo validate its effectiveness, we conducted a set of experiments on WSI\nclassification tasks, where MMIL-Transformer demonstrate superior performance\ncompared to existing state-of-the-art methods, i.e., 96.80% test AUC and 97.67%\ntest accuracy on the CAMELYON16 dataset, 99.04% test AUC and 94.37% test\naccuracy on the TCGA-NSCLC dataset, respectively. All code and pre-trained\nmodels are available at: https://github.com/hustvl/MMIL-Transformer\n", "date": "2023-06-08", "categories": []}
{"id": "2305.18033", "title": "The ACROBAT 2022 Challenge: Automatic Registration Of Breast Cancer\n  Tissue", "abstract": "  The alignment of tissue between histopathological whole-slide-images (WSI) is\ncrucial for research and clinical applications. Advances in computing, deep\nlearning, and availability of large WSI datasets have revolutionised WSI\nanalysis. Therefore, the current state-of-the-art in WSI registration is\nunclear. To address this, we conducted the ACROBAT challenge, based on the\nlargest WSI registration dataset to date, including 4,212 WSIs from 1,152\nbreast cancer patients. The challenge objective was to align WSIs of tissue\nthat was stained with routine diagnostic immunohistochemistry to its\nH&E-stained counterpart. We compare the performance of eight WSI registration\nalgorithms, including an investigation of the impact of different WSI\nproperties and clinical covariates. We find that conceptually distinct WSI\nregistration methods can lead to highly accurate registration performances and\nidentify covariates that impact performances across methods. These results\nestablish the current state-of-the-art in WSI registration and guide\nresearchers in selecting and developing methods.\n", "date": "2023-05-29", "categories": []}
{"id": "2308.15474", "title": "A General-Purpose Self-Supervised Model for Computational Pathology", "abstract": "  Tissue phenotyping is a fundamental computational pathology (CPath) task in\nlearning objective characterizations of histopathologic biomarkers in anatomic\npathology. However, whole-slide imaging (WSI) poses a complex computer vision\nproblem in which the large-scale image resolutions of WSIs and the enormous\ndiversity of morphological phenotypes preclude large-scale data annotation.\nCurrent efforts have proposed using pretrained image encoders with either\ntransfer learning from natural image datasets or self-supervised pretraining on\npublicly-available histopathology datasets, but have not been extensively\ndeveloped and evaluated across diverse tissue types at scale. We introduce UNI,\na general-purpose self-supervised model for pathology, pretrained using over\n100 million tissue patches from over 100,000 diagnostic haematoxylin and\neosin-stained WSIs across 20 major tissue types, and evaluated on 33\nrepresentative CPath clinical tasks in CPath of varying diagnostic\ndifficulties. In addition to outperforming previous state-of-the-art models, we\ndemonstrate new modeling capabilities in CPath such as resolution-agnostic\ntissue classification, slide classification using few-shot class prototypes,\nand disease subtyping generalization in classifying up to 108 cancer types in\nthe OncoTree code classification system. UNI advances unsupervised\nrepresentation learning at scale in CPath in terms of both pretraining data and\ndownstream evaluation, enabling data-efficient AI models that can generalize\nand transfer to a gamut of diagnostically-challenging tasks and clinical\nworkflows in anatomic pathology.\n", "date": "2023-08-29", "categories": []}
{"id": "2303.02602", "title": "DPA-P2PNet: Deformable Proposal-aware P2PNet for Accurate Point-based\n  Cell Detection", "abstract": "  Point-based cell detection (PCD), which pursues high-performance cell sensing\nunder low-cost data annotation, has garnered increased attention in\ncomputational pathology community. Unlike mainstream PCD methods that rely on\nintermediate density map representations, the Point-to-Point network (P2PNet)\nhas recently emerged as an end-to-end solution for PCD, demonstrating\nimpressive cell detection accuracy and efficiency. Nevertheless, P2PNet is\nlimited to decoding from a single-level feature map due to the scale-agnostic\nproperty of point proposals, which is insufficient to leverage multi-scale\ninformation. Moreover, the spatial distribution of pre-set point proposals is\nbiased from that of cells, leading to inaccurate cell localization. To lift\nthese limitations, we present DPA-P2PNet in this work. The proposed method\ndirectly extracts multi-scale features for decoding according to the\ncoordinates of point proposals on hierarchical feature maps. On this basis, we\nfurther devise deformable point proposals to mitigate the positional bias\nbetween proposals and potential cells to promote cell localization. Inspired by\npractical pathological diagnosis that usually combines high-level tissue\nstructure and low-level cell morphology for accurate cell classification, we\npropose a multi-field-of-view (mFoV) variant of DPA-P2PNet to accommodate\nadditional large FoV images with tissue information as model input. Finally, we\nexecute the first self-supervised pre-training on immunohistochemistry\nhistopathology image data and evaluate the suitability of four representative\nself-supervised methods on the PCD task. Experimental results on three\nbenchmarks and a large-scale and real-world interval dataset demonstrate the\nsuperiority of our proposed models over the state-of-the-art counterparts.\nCodes and pre-trained weights will be available.\n", "date": "2023-03-05", "categories": []}
{"id": "2308.14030", "title": "Forensic Histopathological Recognition via a Context-Aware MIL Network\n  Powered by Self-Supervised Contrastive Learning", "abstract": "  Forensic pathology is critical in analyzing death manner and time from the\nmicroscopic aspect to assist in the establishment of reliable factual bases for\ncriminal investigation. In practice, even the manual differentiation between\ndifferent postmortem organ tissues is challenging and relies on expertise,\nconsidering that changes like putrefaction and autolysis could significantly\nchange typical histopathological appearance. Developing AI-based computational\npathology techniques to assist forensic pathologists is practically meaningful,\nwhich requires reliable discriminative representation learning to capture\ntissues' fine-grained postmortem patterns. To this end, we propose a framework\ncalled FPath, in which a dedicated self-supervised contrastive learning\nstrategy and a context-aware multiple-instance learning (MIL) block are\ndesigned to learn discriminative representations from postmortem\nhistopathological images acquired at varying magnification scales. Our\nself-supervised learning step leverages multiple complementary contrastive\nlosses and regularization terms to train a double-tier backbone for\nfine-grained and informative patch/instance embedding. Thereafter, the\ncontext-aware MIL adaptively distills from the local instances a holistic\nbag/image-level representation for the recognition task. On a large-scale\ndatabase of $19,607$ experimental rat postmortem images and $3,378$ real-world\nhuman decedent images, our FPath led to state-of-the-art accuracy and promising\ncross-domain generalization in recognizing seven different postmortem tissues.\nThe source code will be released on\n\\href{https://github.com/ladderlab-xjtu/forensic_pathology}{https://github.com/ladderlab-xjtu/forensic\\_pathology}.\n", "date": "2023-08-27", "categories": []}
{"id": "2308.13182", "title": "Structural Cycle GAN for Virtual Immunohistochemistry Staining of Gland\n  Markers in the Colon", "abstract": "  With the advent of digital scanners and deep learning, diagnostic operations\nmay move from a microscope to a desktop. Hematoxylin and Eosin (H&E) staining\nis one of the most frequently used stains for disease analysis, diagnosis, and\ngrading, but pathologists do need different immunohistochemical (IHC) stains to\nanalyze specific structures or cells. Obtaining all of these stains (H&E and\ndifferent IHCs) on a single specimen is a tedious and time-consuming task.\nConsequently, virtual staining has emerged as an essential research direction.\nHere, we propose a novel generative model, Structural Cycle-GAN (SC-GAN), for\nsynthesizing IHC stains from H&E images, and vice versa. Our method expressly\nincorporates structural information in the form of edges (in addition to color\ndata) and employs attention modules exclusively in the decoder of the proposed\ngenerator model. This integration enhances feature localization and preserves\ncontextual information during the generation process. In addition, a structural\nloss is incorporated to ensure accurate structure alignment between the\ngenerated and input markers. To demonstrate the efficacy of the proposed model,\nexperiments are conducted with two IHC markers emphasizing distinct structures\nof glands in the colon: the nucleus of epithelial cells (CDX2) and the\ncytoplasm (CK818). Quantitative metrics such as FID and SSIM are frequently\nused for the analysis of generative models, but they do not correlate\nexplicitly with higher-quality virtual staining results. Therefore, we propose\ntwo new quantitative metrics that correlate directly with the virtual staining\nspecificity of IHC markers.\n", "date": "2023-08-25", "categories": []}
{"id": "2308.11992", "title": "Critical Evaluation of Artificial Intelligence as Digital Twin of\n  Pathologist for Prostate Cancer Pathology", "abstract": "  Prostate cancer pathology plays a crucial role in clinical management but is\ntime-consuming. Artificial intelligence (AI) shows promise in detecting\nprostate cancer and grading patterns. We tested an AI-based digital twin of a\npathologist, vPatho, on 2,603 histology images of prostate tissue stained with\nhematoxylin and eosin. We analyzed various factors influencing tumor-grade\ndisagreement between vPatho and six human pathologists. Our results\ndemonstrated that vPatho achieved comparable performance in prostate cancer\ndetection and tumor volume estimation, as reported in the literature.\nConcordance levels between vPatho and human pathologists were examined.\nNotably, moderate to substantial agreement was observed in identifying\ncomplementary histological features such as ductal, cribriform, nerve, blood\nvessels, and lymph cell infiltrations. However, concordance in tumor grading\nshowed a decline when applied to prostatectomy specimens (kappa = 0.44)\ncompared to biopsy cores (kappa = 0.70). Adjusting the decision threshold for\nthe secondary Gleason pattern from 5% to 10% improved the concordance level\nbetween pathologists and vPatho for tumor grading on prostatectomy specimens\n(kappa from 0.44 to 0.64). Potential causes of grade discordance included the\nvertical extent of tumors toward the prostate boundary and the proportions of\nslides with prostate cancer. Gleason pattern 4 was particularly associated with\ndiscordance. Notably, grade discordance with vPatho was not specific to any of\nthe six pathologists involved in routine clinical grading. In conclusion, our\nstudy highlights the potential utility of AI in developing a digital twin of a\npathologist. This approach can help uncover limitations in AI adoption and the\ncurrent grading system for prostate cancer pathology.\n", "date": "2023-08-23", "categories": []}
{"id": "2308.11562", "title": "EndoNet: model for automatic calculation of H-score on histological\n  slides", "abstract": "  H-score is a semi-quantitative method used to assess the presence and\ndistribution of proteins in tissue samples by combining the intensity of\nstaining and percentage of stained nuclei. It is widely used but time-consuming\nand can be limited in accuracy and precision. Computer-aided methods may help\novercome these limitations and improve the efficiency of pathologists'\nworkflows. In this work, we developed a model EndoNet for automatic calculation\nof H-score on histological slides. Our proposed method uses neural networks and\nconsists of two main parts. The first is a detection model which predicts\nkeypoints of centers of nuclei. The second is a H-score module which calculates\nthe value of the H-score using mean pixel values of predicted keypoints. Our\nmodel was trained and validated on 1780 annotated tiles with a shape of 100x100\n$\\mu m$ and performed 0.77 mAP on a test dataset. Moreover, the model can be\nadjusted to a specific specialist or whole laboratory to reproduce the manner\nof calculating the H-score. Thus, EndoNet is effective and robust in the\nanalysis of histology slides, which can improve and significantly accelerate\nthe work of pathologists.\n", "date": "2023-08-22", "categories": []}
{"id": "2308.10166", "title": "Cell Spatial Analysis in Crohn's Disease: Unveiling Local Cell\n  Arrangement Pattern with Graph-based Signatures", "abstract": "  Crohn's disease (CD) is a chronic and relapsing inflammatory condition that\naffects segments of the gastrointestinal tract. CD activity is determined by\nhistological findings, particularly the density of neutrophils observed on\nHematoxylin and Eosin stains (H&E) imaging. However, understanding the broader\nmorphometry and local cell arrangement beyond cell counting and tissue\nmorphology remains challenging. To address this, we characterize six distinct\ncell types from H&E images and develop a novel approach for the local spatial\nsignature of each cell. Specifically, we create a 10-cell neighborhood matrix,\nrepresenting neighboring cell arrangements for each individual cell. Utilizing\nt-SNE for non-linear spatial projection in scatter-plot and Kernel Density\nEstimation contour-plot formats, our study examines patterns of differences in\nthe cellular environment associated with the odds ratio of spatial patterns\nbetween active CD and control groups. This analysis is based on data collected\nat the two research institutes. The findings reveal heterogeneous\nnearest-neighbor patterns, signifying distinct tendencies of cell clustering,\nwith a particular focus on the rectum region. These variations underscore the\nimpact of data heterogeneity on cell spatial arrangements in CD patients.\nMoreover, the spatial distribution disparities between the two research sites\nhighlight the significance of collaborative efforts among healthcare\norganizations. All research analysis pipeline tools are available at\nhttps://github.com/MASILab/cellNN.\n", "date": "2023-08-20", "categories": []}
{"id": "2308.08974", "title": "Eosinophils Instance Object Segmentation on Whole Slide Imaging Using\n  Multi-label Circle Representation", "abstract": "  Eosinophilic esophagitis (EoE) is a chronic and relapsing disease\ncharacterized by esophageal inflammation. Symptoms of EoE include difficulty\nswallowing, food impaction, and chest pain which significantly impact the\nquality of life, resulting in nutritional impairments, social limitations, and\npsychological distress. The diagnosis of EoE is typically performed with a\nthreshold (15 to 20) of eosinophils (Eos) per high-power field (HPF). Since the\ncurrent counting process of Eos is a resource-intensive process for human\npathologists, automatic methods are desired. Circle representation has been\nshown as a more precise, yet less complicated, representation for automatic\ninstance cell segmentation such as CircleSnake approach. However, the\nCircleSnake was designed as a single-label model, which is not able to deal\nwith multi-label scenarios. In this paper, we propose the multi-label\nCircleSnake model for instance segmentation on Eos. It extends the original\nCircleSnake model from a single-label design to a multi-label model, allowing\nsegmentation of multiple object types. Experimental results illustrate the\nCircleSnake model's superiority over the traditional Mask R-CNN model and\nDeepSnake model in terms of average precision (AP) in identifying and\nsegmenting eosinophils, thereby enabling enhanced characterization of EoE. This\nautomated approach holds promise for streamlining the assessment process and\nimproving diagnostic accuracy in EoE analysis. The source code has been made\npublicly available at https://github.com/yilinliu610730/EoE.\n", "date": "2023-08-17", "categories": []}
{"id": "2308.06288", "title": "Spatial Pathomics Toolkit for Quantitative Analysis of Podocyte Nuclei\n  with Histology and Spatial Transcriptomics Data in Renal Pathology", "abstract": "  Podocytes, specialized epithelial cells that envelop the glomerular\ncapillaries, play a pivotal role in maintaining renal health. The current\ndescription and quantification of features on pathology slides are limited,\nprompting the need for innovative solutions to comprehensively assess diverse\nphenotypic attributes within Whole Slide Images (WSIs). In particular,\nunderstanding the morphological characteristics of podocytes, terminally\ndifferentiated glomerular epithelial cells, is crucial for studying glomerular\ninjury. This paper introduces the Spatial Pathomics Toolkit (SPT) and applies\nit to podocyte pathomics. The SPT consists of three main components: (1)\ninstance object segmentation, enabling precise identification of podocyte\nnuclei; (2) pathomics feature generation, extracting a comprehensive array of\nquantitative features from the identified nuclei; and (3) robust statistical\nanalyses, facilitating a comprehensive exploration of spatial relationships\nbetween morphological and spatial transcriptomics features.The SPT successfully\nextracted and analyzed morphological and textural features from podocyte\nnuclei, revealing a multitude of podocyte morphomic features through\nstatistical analysis. Additionally, we demonstrated the SPT's ability to\nunravel spatial information inherent to podocyte distribution, shedding light\non spatial patterns associated with glomerular injury. By disseminating the\nSPT, our goal is to provide the research community with a powerful and\nuser-friendly resource that advances cellular spatial pathomics in renal\npathology. The implementation and its complete source code of the toolkit are\nmade openly accessible at https://github.com/hrlblab/spatial_pathomics.\n", "date": "2023-08-10", "categories": []}
{"id": "2308.06333", "title": "Deep Learning-Based Open Source Toolkit for Eosinophil Detection in\n  Pediatric Eosinophilic Esophagitis", "abstract": "  Eosinophilic Esophagitis (EoE) is a chronic, immune/antigen-mediated\nesophageal disease, characterized by symptoms related to esophageal dysfunction\nand histological evidence of eosinophil-dominant inflammation. Owing to the\nintricate microscopic representation of EoE in imaging, current methodologies\nwhich depend on manual identification are not only labor-intensive but also\nprone to inaccuracies. In this study, we develop an open-source toolkit, named\nOpen-EoE, to perform end-to-end whole slide image (WSI) level eosinophil (Eos)\ndetection using one line of command via Docker. Specifically, the toolkit\nsupports three state-of-the-art deep learning-based object detection models.\nFurthermore, Open-EoE further optimizes the performance by implementing an\nensemble learning strategy, and enhancing the precision and reliability of our\nresults. The experimental results demonstrated that the Open-EoE toolkit can\nefficiently detect Eos on a testing set with 289 WSIs. At the widely accepted\nthreshold of >= 15 Eos per high power field (HPF) for diagnosing EoE, the\nOpen-EoE achieved an accuracy of 91%, showing decent consistency with\npathologist evaluations. This suggests a promising avenue for integrating\nmachine learning methodologies into the diagnostic process for EoE. The docker\nand source code has been made publicly available at\nhttps://github.com/hrlblab/Open-EoE.\n", "date": "2023-08-11", "categories": []}
{"id": "2308.05782", "title": "Multi-scale Multi-site Renal Microvascular Structures Segmentation for\n  Whole Slide Imaging in Renal Pathology", "abstract": "  Segmentation of microvascular structures, such as arterioles, venules, and\ncapillaries, from human kidney whole slide images (WSI) has become a focal\npoint in renal pathology. Current manual segmentation techniques are\ntime-consuming and not feasible for large-scale digital pathology images. While\ndeep learning-based methods offer a solution for automatic segmentation, most\nsuffer from a limitation: they are designed for and restricted to training on\nsingle-site, single-scale data. In this paper, we present Omni-Seg, a novel\nsingle dynamic network method that capitalizes on multi-site, multi-scale\ntraining data. Unique to our approach, we utilize partially labeled images,\nwhere only one tissue type is labeled per training image, to segment\nmicrovascular structures. We train a singular deep network using images from\ntwo datasets, HuBMAP and NEPTUNE, across different magnifications (40x, 20x,\n10x, and 5x). Experimental results indicate that Omni-Seg outperforms in terms\nof both the Dice Similarity Coefficient (DSC) and Intersection over Union\n(IoU). Our proposed method provides renal pathologists with a powerful\ncomputational tool for the quantitative analysis of renal microvascular\nstructures.\n", "date": "2023-08-10", "categories": []}
{"id": "2304.09534", "title": "Realistic Data Enrichment for Robust Image Segmentation in\n  Histopathology", "abstract": "  Poor performance of quantitative analysis in histopathological Whole Slide\nImages (WSI) has been a significant obstacle in clinical practice. Annotating\nlarge-scale WSIs manually is a demanding and time-consuming task, unlikely to\nyield the expected results when used for fully supervised learning systems.\nRarely observed disease patterns and large differences in object scales are\ndifficult to model through conventional patient intake. Prior methods either\nfall back to direct disease classification, which only requires learning a few\nfactors per image, or report on average image segmentation performance, which\nis highly biased towards majority observations. Geometric image augmentation is\ncommonly used to improve robustness for average case predictions and to enrich\nlimited datasets. So far no method provided sampling of a realistic posterior\ndistribution to improve stability, e.g. for the segmentation of imbalanced\nobjects within images. Therefore, we propose a new approach, based on diffusion\nmodels, which can enrich an imbalanced dataset with plausible examples from\nunderrepresented groups by conditioning on segmentation maps. Our method can\nsimply expand limited clinical datasets making them suitable to train machine\nlearning pipelines, and provides an interpretable and human-controllable way of\ngenerating histopathology images that are indistinguishable from real ones to\nhuman experts. We validate our findings on two datasets, one from the public\ndomain and one from a Kidney Transplant study.\n", "date": "2023-04-19", "categories": []}
{"id": "2308.02597", "title": "Designing a Deep Learning-Driven Resource-Efficient Diagnostic System\n  for Metastatic Breast Cancer: Reducing Long Delays of Clinical Diagnosis and\n  Improving Patient Survival in Developing Countries", "abstract": "  Breast cancer is one of the leading causes of cancer mortality. Breast cancer\npatients in developing countries, especially sub-Saharan Africa, South Asia,\nand South America, suffer from the highest mortality rate in the world. One\ncrucial factor contributing to the global disparity in mortality rate is long\ndelay of diagnosis due to a severe shortage of trained pathologists, which\nconsequently has led to a large proportion of late-stage presentation at\ndiagnosis. The delay between the initial development of symptoms and the\nreceipt of a diagnosis could stretch upwards 15 months. To tackle this critical\nhealthcare disparity, this research has developed a deep learning-based\ndiagnosis system for metastatic breast cancer that can achieve high diagnostic\naccuracy as well as computational efficiency. Based on our evaluation, the\nMobileNetV2-based diagnostic model outperformed the more complex VGG16,\nResNet50 and ResNet101 models in diagnostic accuracy, model generalization, and\nmodel training efficiency. The visual comparisons between the model prediction\nand ground truth have demonstrated that the MobileNetV2 diagnostic models can\nidentify very small cancerous nodes embedded in a large area of normal cells\nwhich is challenging for manual image analysis. Equally Important, the light\nweighted MobleNetV2 models were computationally efficient and ready for mobile\ndevices or devices of low computational power. These advances empower the\ndevelopment of a resource-efficient and high performing AI-based metastatic\nbreast cancer diagnostic system that can adapt to under-resourced healthcare\nfacilities in developing countries. This research provides an innovative\ntechnological solution to address the long delays in metastatic breast cancer\ndiagnosis and the consequent disparity in patient survival outcome in\ndeveloping countries.\n", "date": "2023-08-04", "categories": []}
{"id": "2308.03076", "title": "Study for Performance of MobileNetV1 and MobileNetV2 Based on Breast\n  Cancer", "abstract": "  Artificial intelligence is constantly evolving and can provide effective help\nin all aspects of people's lives. The experiment is mainly to study the use of\nartificial intelligence in the field of medicine. The purpose of this\nexperiment was to compare which of MobileNetV1 and MobileNetV2 models was\nbetter at detecting histopathological images of the breast downloaded at\nKaggle. When the doctor looks at the pathological image, there may be errors\nthat lead to errors in judgment, and the observation speed is slow. Rational\nuse of artificial intelligence can effectively reduce the error of doctor\ndiagnosis in breast cancer judgment and speed up doctor diagnosis. The dataset\nwas downloaded from Kaggle and then normalized. The basic principle of the\nexperiment is to let the neural network model learn the downloaded data set.\nThen find the pattern and be able to judge on your own whether breast tissue is\ncancer. In the dataset, benign tumor pictures and malignant tumor pictures have\nbeen classified, of which 198738 are benign tumor pictures and 78, 786 are\nmalignant tumor pictures. After calling MobileNetV1 and MobileNetV2, the\ndataset is trained separately, the training accuracy and validation accuracy\nrate are obtained, and the image is drawn. It can be observed that MobileNetV1\nhas better validation accuracy and overfit during MobileNetV2 training. From\nthe experimental results, it can be seen that in the case of processing this\ndataset, MobileNetV1 is much better than MobileNetV2.\n", "date": "2023-08-06", "categories": []}
{"id": "2308.01982", "title": "Predicting Ki67, ER, PR, and HER2 Statuses from H&E-stained Breast\n  Cancer Images", "abstract": "  Despite the advances in machine learning and digital pathology, it is not yet\nclear if machine learning methods can accurately predict molecular information\nmerely from histomorphology. In a quest to answer this question, we built a\nlarge-scale dataset (185538 images) with reliable measurements for Ki67, ER,\nPR, and HER2 statuses. The dataset is composed of mirrored images of H\\&E and\ncorresponding images of immunohistochemistry (IHC) assays (Ki67, ER, PR, and\nHER2. These images are mirrored through registration. To increase reliability,\nindividual pairs were inspected and discarded if artifacts were present (tissue\nfolding, bubbles, etc). Measurements for Ki67, ER and PR were determined by\ncalculating H-Score from image analysis. HER2 measurement is based on binary\nclassification: 0 and 1+ (IHC scores representing a negative subset) vs 3+ (IHC\nscore positive subset). Cases with IHC equivocal score (2+) were excluded. We\nshow that a standard ViT-based pipeline can achieve prediction performances\naround 90% in terms of Area Under the Curve (AUC) when trained with a proper\nlabeling protocol. Finally, we shed light on the ability of the trained\nclassifiers to localize relevant regions, which encourages future work to\nimprove the localizations. Our proposed dataset is publicly available:\nhttps://ihc4bc.github.io/\n", "date": "2023-08-03", "categories": []}
{"id": "2307.14603", "title": "A Weakly Supervised Segmentation Network Embedding Cross-scale Attention\n  Guidance and Noise-sensitive Constraint for Detecting Tertiary Lymphoid\n  Structures of Pancreatic Tumors", "abstract": "  The presence of tertiary lymphoid structures (TLSs) on pancreatic\npathological images is an important prognostic indicator of pancreatic tumors.\nTherefore, TLSs detection on pancreatic pathological images plays a crucial\nrole in diagnosis and treatment for patients with pancreatic tumors. However,\nfully supervised detection algorithms based on deep learning usually require a\nlarge number of manual annotations, which is time-consuming and\nlabor-intensive. In this paper, we aim to detect the TLSs in a manner of\nfew-shot learning by proposing a weakly supervised segmentation network. We\nfirstly obtain the lymphocyte density maps by combining a pretrained model for\nnuclei segmentation and a domain adversarial network for lymphocyte nuclei\nrecognition. Then, we establish a cross-scale attention guidance mechanism by\njointly learning the coarse-scale features from the original histopathology\nimages and fine-scale features from our designed lymphocyte density attention.\nA noise-sensitive constraint is introduced by an embedding signed distance\nfunction loss in the training procedure to reduce tiny prediction errors.\nExperimental results on two collected datasets demonstrate that our proposed\nmethod significantly outperforms the state-of-the-art segmentation-based\nalgorithms in terms of TLSs detection accuracy. Additionally, we apply our\nmethod to study the congruent relationship between the density of TLSs and\nperipancreatic vascular invasion and obtain some clinically statistical\nresults.\n", "date": "2023-07-27", "categories": []}
{"id": "2307.14907", "title": "Weakly Supervised AI for Efficient Analysis of 3D Pathology Samples", "abstract": "  Human tissue and its constituent cells form a microenvironment that is\nfundamentally three-dimensional (3D). However, the standard-of-care in\npathologic diagnosis involves selecting a few two-dimensional (2D) sections for\nmicroscopic evaluation, risking sampling bias and misdiagnosis. Diverse methods\nfor capturing 3D tissue morphologies have been developed, but they have yet had\nlittle translation to clinical practice; manual and computational evaluations\nof such large 3D data have so far been impractical and/or unable to provide\npatient-level clinical insights. Here we present Modality-Agnostic Multiple\ninstance learning for volumetric Block Analysis (MAMBA), a deep-learning-based\nplatform for processing 3D tissue images from diverse imaging modalities and\npredicting patient outcomes. Archived prostate cancer specimens were imaged\nwith open-top light-sheet microscopy or microcomputed tomography and the\nresulting 3D datasets were used to train risk-stratification networks based on\n5-year biochemical recurrence outcomes via MAMBA. With the 3D block-based\napproach, MAMBA achieves an area under the receiver operating characteristic\ncurve (AUC) of 0.86 and 0.74, superior to 2D traditional single-slice-based\nprognostication (AUC of 0.79 and 0.57), suggesting superior prognostication\nwith 3D morphological features. Further analyses reveal that the incorporation\nof greater tissue volume improves prognostic performance and mitigates risk\nprediction variability from sampling bias, suggesting the value of capturing\nlarger extents of heterogeneous 3D morphology. With the rapid growth and\nadoption of 3D spatial biology and pathology techniques by researchers and\nclinicians, MAMBA provides a general and efficient framework for 3D weakly\nsupervised learning for clinical decision support and can help to reveal novel\n3D morphological biomarkers for prognosis and therapeutic response.\n", "date": "2023-07-27", "categories": []}
{"id": "2307.13947", "title": "Centroid-aware feature recalibration for cancer grading in pathology\n  images", "abstract": "  Cancer grading is an essential task in pathology. The recent developments of\nartificial neural networks in computational pathology have shown that these\nmethods hold great potential for improving the accuracy and quality of cancer\ndiagnosis. However, the issues with the robustness and reliability of such\nmethods have not been fully resolved yet. Herein, we propose a centroid-aware\nfeature recalibration network that can conduct cancer grading in an accurate\nand robust manner. The proposed network maps an input pathology image into an\nembedding space and adjusts it by using centroids embedding vectors of\ndifferent cancer grades via attention mechanism. Equipped with the recalibrated\nembedding vector, the proposed network classifiers the input pathology image\ninto a pertinent class label, i.e., cancer grade. We evaluate the proposed\nnetwork using colorectal cancer datasets that were collected under different\nenvironments. The experimental results confirm that the proposed network is\nable to conduct cancer grading in pathology images with high accuracy\nregardless of the environmental changes in the datasets.\n", "date": "2023-07-26", "categories": []}
{"id": "2307.12914", "title": "Towards a Visual-Language Foundation Model for Computational Pathology", "abstract": "  The accelerated adoption of digital pathology and advances in deep learning\nhave enabled the development of powerful models for various pathology tasks\nacross a diverse array of diseases and patient cohorts. However, model training\nis often difficult due to label scarcity in the medical domain and the model's\nusage is limited by the specific task and disease for which it is trained.\nAdditionally, most models in histopathology leverage only image data, a stark\ncontrast to how humans teach each other and reason about histopathologic\nentities. We introduce CONtrastive learning from Captions for Histopathology\n(CONCH), a visual-language foundation model developed using diverse sources of\nhistopathology images, biomedical text, and notably over 1.17 million\nimage-caption pairs via task-agnostic pretraining. Evaluated on a suite of 13\ndiverse benchmarks, CONCH can be transferred to a wide range of downstream\ntasks involving either or both histopathology images and text, achieving\nstate-of-the-art performance on histology image classification, segmentation,\ncaptioning, text-to-image and image-to-text retrieval. CONCH represents a\nsubstantial leap over concurrent visual-language pretrained systems for\nhistopathology, with the potential to directly facilitate a wide array of\nmachine learning-based workflows requiring minimal or no further supervised\nfine-tuning.\n", "date": "2023-07-24", "categories": []}
{"id": "2307.13199", "title": "An Investigation into Glomeruli Detection in Kidney H&E and PAS Images\n  using YOLO", "abstract": "  Context: Analyzing digital pathology images is necessary to draw diagnostic\nconclusions by investigating tissue patterns and cellular morphology. However,\nmanual evaluation can be time-consuming, expensive, and prone to inter- and\nintra-observer variability. Objective: To assist pathologists using\ncomputerized solutions, automated tissue structure detection and segmentation\nmust be proposed. Furthermore, generating pixel-level object annotations for\nhistopathology images is expensive and time-consuming. As a result, detection\nmodels with bounding box labels may be a feasible solution. Design: This paper\nstudies. YOLO-v4 (You-Only-Look-Once), a real-time object detector for\nmicroscopic images. YOLO uses a single neural network to predict several\nbounding boxes and class probabilities for objects of interest. YOLO can\nenhance detection performance by training on whole slide images. YOLO-v4 has\nbeen used in this paper. for glomeruli detection in human kidney images.\nMultiple experiments have been designed and conducted based on different\ntraining data of two public datasets and a private dataset from the University\nof Michigan for fine-tuning the model. The model was tested on the private\ndataset from the University of Michigan, serving as an external validation of\ntwo different stains, namely hematoxylin and eosin (H&E) and periodic\nacid-Schiff (PAS). Results: Average specificity and sensitivity for all\nexperiments, and comparison of existing segmentation methods on the same\ndatasets are discussed. Conclusions: Automated glomeruli detection in human\nkidney images is possible using modern AI models. The design and validation for\ndifferent stains still depends on variability of public multi-stain datasets.\n", "date": "2023-07-25", "categories": []}
{"id": "2301.11513", "title": "CellMix: A General Instance Relationship based Method for Data\n  Augmentation Towards Pathology Image Classification", "abstract": "  In pathology image analysis, obtaining and maintaining high-quality annotated\nsamples is an extremely labor-intensive task. To overcome this challenge,\nmixing-based methods have emerged as effective alternatives to traditional\npreprocessing data augmentation techniques. Nonetheless, these methods fail to\nfully consider the unique features of pathology images, such as local\nspecificity, global distribution, and inner/outer-sample instance\nrelationships. To better comprehend these characteristics and create valuable\npseudo samples, we propose the CellMix framework, which employs a novel\ndistribution-oriented in-place shuffle approach. By dividing images into\npatches based on the granularity of pathology instances and shuffling them\nwithin the same batch, the absolute relationships between instances can be\neffectively preserved when generating new samples. Moreover, we develop a\ncurriculum learning-inspired, loss-driven strategy to handle perturbations and\ndistribution-related noise during training, enabling the model to adaptively\nfit the augmented data. Our experiments in pathology image classification tasks\ndemonstrate state-of-the-art (SOTA) performance on 7 distinct datasets. This\ninnovative instance relationship-centered method has the potential to inform\ngeneral data augmentation approaches for pathology image classification. The\nassociated codes are available at https://github.com/sagizty/CellMix.\n", "date": "2023-01-27", "categories": []}
{"id": "2307.10696", "title": "SLPD: Slide-level Prototypical Distillation for WSIs", "abstract": "  Improving the feature representation ability is the foundation of many whole\nslide pathological image (WSIs) tasks. Recent works have achieved great success\nin pathological-specific self-supervised learning (SSL). However, most of them\nonly focus on learning patch-level representations, thus there is still a gap\nbetween pretext and slide-level downstream tasks, e.g., subtyping, grading and\nstaging. Aiming towards slide-level representations, we propose Slide-Level\nPrototypical Distillation (SLPD) to explore intra- and inter-slide semantic\nstructures for context modeling on WSIs. Specifically, we iteratively perform\nintra-slide clustering for the regions (4096x4096 patches) within each WSI to\nyield the prototypes and encourage the region representations to be closer to\nthe assigned prototypes. By representing each slide with its prototypes, we\nfurther select similar slides by the set distance of prototypes and assign the\nregions by cross-slide prototypes for distillation. SLPD achieves\nstate-of-the-art results on multiple slide-level benchmarks and demonstrates\nthat representation learning of semantic structures of slides can make a\nsuitable proxy task for WSI analysis. Code will be available at\nhttps://github.com/Carboxy/SLPD.\n", "date": "2023-07-20", "categories": []}
{"id": "2307.09570", "title": "SAM-Path: A Segment Anything Model for Semantic Segmentation in Digital\n  Pathology", "abstract": "  Semantic segmentations of pathological entities have crucial clinical value\nin computational pathology workflows. Foundation models, such as the Segment\nAnything Model (SAM), have been recently proposed for universal use in\nsegmentation tasks. SAM shows remarkable promise in instance segmentation on\nnatural images. However, the applicability of SAM to computational pathology\ntasks is limited due to the following factors: (1) lack of comprehensive\npathology datasets used in SAM training and (2) the design of SAM is not\ninherently optimized for semantic segmentation tasks. In this work, we adapt\nSAM for semantic segmentation by introducing trainable class prompts, followed\nby further enhancements through the incorporation of a pathology encoder,\nspecifically a pathology foundation model. Our framework, SAM-Path enhances\nSAM's ability to conduct semantic segmentation in digital pathology without\nhuman input prompts. Through experiments on two public pathology datasets, the\nBCSS and the CRAG datasets, we demonstrate that the fine-tuning with trainable\nclass prompts outperforms vanilla SAM with manual prompts and post-processing\nby 27.52% in Dice score and 71.63% in IOU. On these two datasets, the proposed\nadditional pathology foundation model further achieves a relative improvement\nof 5.07% to 5.12% in Dice score and 4.50% to 8.48% in IOU.\n", "date": "2023-07-12", "categories": []}
{"id": "2301.08125", "title": "Diagnose Like a Pathologist: Transformer-Enabled Hierarchical\n  Attention-Guided Multiple Instance Learning for Whole Slide Image\n  Classification", "abstract": "  Multiple Instance Learning (MIL) and transformers are increasingly popular in\nhistopathology Whole Slide Image (WSI) classification. However, unlike human\npathologists who selectively observe specific regions of histopathology tissues\nunder different magnifications, most methods do not incorporate multiple\nresolutions of the WSIs, hierarchically and attentively, thereby leading to a\nloss of focus on the WSIs and information from other resolutions. To resolve\nthis issue, we propose a Hierarchical Attention-Guided Multiple Instance\nLearning framework to fully exploit the WSIs. This framework can dynamically\nand attentively discover the discriminative regions across multiple resolutions\nof the WSIs. Within this framework, an Integrated Attention Transformer is\nproposed to further enhance the performance of the transformer and obtain a\nmore holistic WSI (bag) representation. This transformer consists of multiple\nIntegrated Attention Modules, which is the combination of a transformer layer\nand an aggregation module that produces a bag representation based on every\ninstance representation in that bag. The experimental results show that our\nmethod achieved state-of-the-art performances on multiple datasets, including\nCamelyon16, TCGA-RCC, TCGA-NSCLC, and an in-house IMGC dataset. The code is\navailable at https://github.com/BearCleverProud/HAG-MIL.\n", "date": "2023-01-19", "categories": []}
{"id": "2305.14301", "title": "A Laplacian Pyramid Based Generative H&E Stain Augmentation Network", "abstract": "  Hematoxylin and Eosin (H&E) staining is a widely used sample preparation\nprocedure for enhancing the saturation of tissue sections and the contrast\nbetween nuclei and cytoplasm in histology images for medical diagnostics.\nHowever, various factors, such as the differences in the reagents used, result\nin high variability in the colors of the stains actually recorded. This\nvariability poses a challenge in achieving generalization for machine-learning\nbased computer-aided diagnostic tools. To desensitize the learned models to\nstain variations, we propose the Generative Stain Augmentation Network (G-SAN)\n-- a GAN-based framework that augments a collection of cell images with\nsimulated yet realistic stain variations. At its core, G-SAN uses a novel and\nhighly computationally efficient Laplacian Pyramid (LP) based generator\narchitecture, that is capable of disentangling stain from cell morphology.\nThrough the task of patch classification and nucleus segmentation, we show that\nusing G-SAN-augmented training data provides on average 15.7% improvement in F1\nscore and 7.3% improvement in panoptic quality, respectively. Our code is\navailable at https://github.com/lifangda01/GSAN-Demo.\n", "date": "2023-05-23", "categories": []}
{"id": "2307.07528", "title": "PatchSorter: A High Throughput Deep Learning Digital Pathology Tool for\n  Object Labeling", "abstract": "  The discovery of patterns associated with diagnosis, prognosis, and therapy\nresponse in digital pathology images often requires intractable labeling of\nlarge quantities of histological objects. Here we release an open-source\nlabeling tool, PatchSorter, which integrates deep learning with an intuitive\nweb interface. Using >100,000 objects, we demonstrate a >7x improvement in\nlabels per second over unaided labeling, with minimal impact on labeling\naccuracy, thus enabling high-throughput labeling of large datasets.\n", "date": "2023-07-13", "categories": []}
{"id": "2307.07168", "title": "Adaptive Region Selection for Active Learning in Whole Slide Image\n  Semantic Segmentation", "abstract": "  The process of annotating histological gigapixel-sized whole slide images\n(WSIs) at the pixel level for the purpose of training a supervised segmentation\nmodel is time-consuming. Region-based active learning (AL) involves training\nthe model on a limited number of annotated image regions instead of requesting\nannotations of the entire images. These annotation regions are iteratively\nselected, with the goal of optimizing model performance while minimizing the\nannotated area. The standard method for region selection evaluates the\ninformativeness of all square regions of a specified size and then selects a\nspecific quantity of the most informative regions. We find that the efficiency\nof this method highly depends on the choice of AL step size (i.e., the\ncombination of region size and the number of selected regions per WSI), and a\nsuboptimal AL step size can result in redundant annotation requests or inflated\ncomputation costs. This paper introduces a novel technique for selecting\nannotation regions adaptively, mitigating the reliance on this AL\nhyperparameter. Specifically, we dynamically determine each region by first\nidentifying an informative area and then detecting its optimal bounding box, as\nopposed to selecting regions of a uniform predefined shape and size as in the\nstandard method. We evaluate our method using the task of breast cancer\nmetastases segmentation on the public CAMELYON16 dataset and show that it\nconsistently achieves higher sampling efficiency than the standard method\nacross various AL step sizes. With only 2.6\\% of tissue area annotated, we\nachieve full annotation performance and thereby substantially reduce the costs\nof annotating a WSI dataset. The source code is available at\nhttps://github.com/DeepMicroscopy/AdaptiveRegionSelection.\n", "date": "2023-07-14", "categories": []}
{"id": "2307.06344", "title": "The Whole Pathological Slide Classification via Weakly Supervised\n  Learning", "abstract": "  Due to its superior efficiency in utilizing annotations and addressing\ngigapixel-sized images, multiple instance learning (MIL) has shown great\npromise as a framework for whole slide image (WSI) classification in digital\npathology diagnosis. However, existing methods tend to focus on advanced\naggregators with different structures, often overlooking the intrinsic features\nof H\\&E pathological slides. To address this limitation, we introduced two\npathological priors: nuclear heterogeneity of diseased cells and spatial\ncorrelation of pathological tiles. Leveraging the former, we proposed a data\naugmentation method that utilizes stain separation during extractor training\nvia a contrastive learning strategy to obtain instance-level representations.\nWe then described the spatial relationships between the tiles using an\nadjacency matrix. By integrating these two views, we designed a multi-instance\nframework for analyzing H\\&E-stained tissue images based on pathological\ninductive bias, encompassing feature extraction, filtering, and aggregation.\nExtensive experiments on the Camelyon16 breast dataset and TCGA-NSCLC Lung\ndataset demonstrate that our proposed framework can effectively handle tasks\nrelated to cancer detection and differentiation of subtypes, outperforming\nstate-of-the-art medical image classification methods based on MIL. The code\nwill be released later.\n", "date": "2023-07-12", "categories": []}
{"id": "2307.05519", "title": "Physical Color Calibration of Digital Pathology Scanners for Robust\n  Artificial Intelligence Assisted Cancer Diagnosis", "abstract": "  The potential of artificial intelligence (AI) in digital pathology is limited\nby technical inconsistencies in the production of whole slide images (WSIs),\nleading to degraded AI performance and posing a challenge for widespread\nclinical application as fine-tuning algorithms for each new site is\nimpractical. Changes in the imaging workflow can also lead to compromised\ndiagnoses and patient safety risks. We evaluated whether physical color\ncalibration of scanners can standardize WSI appearance and enable robust AI\nperformance. We employed a color calibration slide in four different\nlaboratories and evaluated its impact on the performance of an AI system for\nprostate cancer diagnosis on 1,161 WSIs. Color standardization resulted in\nconsistently improved AI model calibration and significant improvements in\nGleason grading performance. The study demonstrates that physical color\ncalibration provides a potential solution to the variation introduced by\ndifferent scanners, making AI-based cancer diagnostics more reliable and\napplicable in clinical settings.\n", "date": "2023-07-07", "categories": []}
{"id": "2307.05889", "title": "Rethinking Mitosis Detection: Towards Diverse Data and Feature\n  Representation", "abstract": "  Mitosis detection is one of the fundamental tasks in computational pathology,\nwhich is extremely challenging due to the heterogeneity of mitotic cell. Most\nof the current studies solve the heterogeneity in the technical aspect by\nincreasing the model complexity. However, lacking consideration of the\nbiological knowledge and the complex model design may lead to the overfitting\nproblem while limited the generalizability of the detection model. In this\npaper, we systematically study the morphological appearances in different\nmitotic phases as well as the ambiguous non-mitotic cells and identify that\nbalancing the data and feature diversity can achieve better generalizability.\nBased on this observation, we propose a novel generalizable framework (MitDet)\nfor mitosis detection. The data diversity is considered by the proposed\ndiversity-guided sample balancing (DGSB). And the feature diversity is\npreserved by inter- and intra- class feature diversity-preserved module\n(InCDP). Stain enhancement (SE) module is introduced to enhance the\ndomain-relevant diversity of both data and features simultaneously. Extensive\nexperiments have demonstrated that our proposed model outperforms all the SOTA\napproaches in several popular mitosis detection datasets in both internal and\nexternal test sets using minimal annotation efforts with point annotations\nonly. Comprehensive ablation studies have also proven the effectiveness of the\nrethinking of data and feature diversity balancing. By analyzing the results\nquantitatively and qualitatively, we believe that our proposed model not only\nachieves SOTA performance but also might inspire the future studies in new\nperspectives. Source code is at https://github.com/Onehour0108/MitDet.\n", "date": "2023-07-12", "categories": []}
{"id": "2307.03757", "title": "A Fully Automated and Explainable Algorithm for the Prediction of\n  Malignant Transformation in Oral Epithelial Dysplasia", "abstract": "  Oral epithelial dysplasia (OED) is a premalignant histopathological diagnosis\ngiven to lesions of the oral cavity. Its grading suffers from significant\ninter-/intra- observer variability, and does not reliably predict malignancy\nprogression, potentially leading to suboptimal treatment decisions. To address\nthis, we developed a novel artificial intelligence algorithm that can assign an\nOral Malignant Transformation (OMT) risk score, based on histological patterns\nin the in Haematoxylin and Eosin stained whole slide images, to quantify the\nrisk of OED progression. The algorithm is based on the detection and\nsegmentation of nuclei within (and around) the epithelium using an in-house\nsegmentation model. We then employed a shallow neural network fed with\ninterpretable morphological/spatial features, emulating histological markers.\nWe conducted internal cross-validation on our development cohort (Sheffield; n\n= 193 cases) followed by independent validation on two external cohorts\n(Birmingham and Belfast; n = 92 cases). The proposed OMTscore yields an AUROC =\n0.74 in predicting whether an OED progresses to malignancy or not. Survival\nanalyses showed the prognostic value of our OMTscore for predicting malignancy\ntransformation, when compared to the manually-assigned WHO and binary grades.\nAnalysis of the correctly predicted cases elucidated the presence of\nperi-epithelial and epithelium-infiltrating lymphocytes in the most predictive\npatches of cases that transformed (p < 0.0001). This is the first study to\npropose a completely automated algorithm for predicting OED transformation\nbased on interpretable nuclear features, whilst being validated on external\ndatasets. The algorithm shows better-than-human-level performance for\nprediction of OED malignant transformation and offers a promising solution to\nthe challenges of grading OED in routine clinical practice.\n", "date": "2023-07-06", "categories": []}
{"id": "2307.02308", "title": "Multi-Scale Prototypical Transformer for Whole Slide Image\n  Classification", "abstract": "  Whole slide image (WSI) classification is an essential task in computational\npathology. Despite the recent advances in multiple instance learning (MIL) for\nWSI classification, accurate classification of WSIs remains challenging due to\nthe extreme imbalance between the positive and negative instances in bags, and\nthe complicated pre-processing to fuse multi-scale information of WSI. To this\nend, we propose a novel multi-scale prototypical Transformer (MSPT) for WSI\nclassification, which includes a prototypical Transformer (PT) module and a\nmulti-scale feature fusion module (MFFM). The PT is developed to reduce\nredundant instances in bags by integrating prototypical learning into the\nTransformer architecture. It substitutes all instances with cluster prototypes,\nwhich are then re-calibrated through the self-attention mechanism of the\nTrans-former. Thereafter, an MFFM is proposed to fuse the clustered prototypes\nof different scales, which employs MLP-Mixer to enhance the information\ncommunication between prototypes. The experimental results on two public WSI\ndatasets demonstrate that the proposed MSPT outperforms all the compared\nalgorithms, suggesting its potential applications.\n", "date": "2023-07-05", "categories": []}
{"id": "2306.17373", "title": "HVTSurv: Hierarchical Vision Transformer for Patient-Level Survival\n  Prediction from Whole Slide Image", "abstract": "  Survival prediction based on whole slide images (WSIs) is a challenging task\nfor patient-level multiple instance learning (MIL). Due to the vast amount of\ndata for a patient (one or multiple gigapixels WSIs) and the irregularly shaped\nproperty of WSI, it is difficult to fully explore spatial, contextual, and\nhierarchical interaction in the patient-level bag. Many studies adopt random\nsampling pre-processing strategy and WSI-level aggregation models, which\ninevitably lose critical prognostic information in the patient-level bag. In\nthis work, we propose a hierarchical vision Transformer framework named\nHVTSurv, which can encode the local-level relative spatial information,\nstrengthen WSI-level context-aware communication, and establish patient-level\nhierarchical interaction. Firstly, we design a feature pre-processing strategy,\nincluding feature rearrangement and random window masking. Then, we devise\nthree layers to progressively obtain patient-level representation, including a\nlocal-level interaction layer adopting Manhattan distance, a WSI-level\ninteraction layer employing spatial shuffle, and a patient-level interaction\nlayer using attention pooling. Moreover, the design of hierarchical network\nhelps the model become more computationally efficient. Finally, we validate\nHVTSurv with 3,104 patients and 3,752 WSIs across 6 cancer types from The\nCancer Genome Atlas (TCGA). The average C-Index is 2.50-11.30% higher than all\nthe prior weakly supervised methods over 6 TCGA datasets. Ablation study and\nattention visualization further verify the superiority of the proposed HVTSurv.\nImplementation is available at: https://github.com/szc19990412/HVTSurv.\n", "date": "2023-06-30", "categories": []}
{"id": "2306.15789", "title": "Structured State Space Models for Multiple Instance Learning in Digital\n  Pathology", "abstract": "  Multiple instance learning is an ideal mode of analysis for histopathology\ndata, where vast whole slide images are typically annotated with a single\nglobal label. In such cases, a whole slide image is modelled as a collection of\ntissue patches to be aggregated and classified. Common models for performing\nthis classification include recurrent neural networks and transformers.\nAlthough powerful compression algorithms, such as deep pre-trained neural\nnetworks, are used to reduce the dimensionality of each patch, the sequences\narising from whole slide images remain excessively long, routinely containing\ntens of thousands of patches. Structured state space models are an emerging\nalternative for sequence modelling, specifically designed for the efficient\nmodelling of long sequences. These models invoke an optimal projection of an\ninput sequence into memory units that compress the entire sequence. In this\npaper, we propose the use of state space models as a multiple instance learner\nto a variety of problems in digital pathology. Across experiments in metastasis\ndetection, cancer subtyping, mutation classification, and multitask learning,\nwe demonstrate the competitiveness of this new class of models with existing\nstate of the art approaches. Our code is available at\nhttps://github.com/MICS-Lab/s4_digital_pathology.\n", "date": "2023-06-27", "categories": []}
{"id": "2303.14845", "title": "Multi-task Learning of Histology and Molecular Markers for Classifying\n  Diffuse Glioma", "abstract": "  Most recently, the pathology diagnosis of cancer is shifting to integrating\nmolecular makers with histology features. It is a urgent need for digital\npathology methods to effectively integrate molecular markers with histology,\nwhich could lead to more accurate diagnosis in the real world scenarios. This\npaper presents a first attempt to jointly predict molecular markers and\nhistology features and model their interactions for classifying diffuse glioma\nbases on whole slide images. Specifically, we propose a hierarchical multi-task\nmulti-instance learning framework to jointly predict histology and molecular\nmarkers. Moreover, we propose a co-occurrence probability-based label\ncorrection graph network to model the co-occurrence of molecular markers.\nLastly, we design an inter-omic interaction strategy with the dynamical\nconfidence constraint loss to model the interactions of histology and molecular\nmarkers. Our experiments show that our method outperforms other\nstate-of-the-art methods in classifying diffuse glioma,as well as related\nhistology and molecular markers on a multi-institutional dataset.\n", "date": "2023-03-26", "categories": []}
{"id": "2304.08297", "title": "Comments on 'Fast and scalable search of whole-slide images via\n  self-supervised deep learning'", "abstract": "  Chen et al. [Chen2022] recently published the article 'Fast and scalable\nsearch of whole-slide images via self-supervised deep learning' in Nature\nBiomedical Engineering. The authors call their method 'self-supervised image\nsearch for histology', short SISH. We express our concerns that SISH is an\nincremental modification of Yottixel, has used MinMax binarization but does not\ncite the original works, and is based on a misnomer 'self-supervised image\nsearch'. As well, we point to several other concerns regarding experiments and\ncomparisons performed by Chen et al.\n", "date": "2023-04-07", "categories": []}
{"id": "2306.08198", "title": "Explainable and Position-Aware Learning in Digital Pathology", "abstract": "  Encoding whole slide images (WSI) as graphs is well motivated since it makes\nit possible for the gigapixel resolution WSI to be represented in its entirety\nfor the purpose of graph learning. To this end, WSIs can be broken into smaller\npatches that represent the nodes of the graph. Then, graph-based learning\nmethods can be utilized for the grading and classification of cancer. Message\npassing among neighboring nodes is the foundation of graph-based learning\nmethods. However, they do not take into consideration any positional\ninformation for any of the patches, and if two patches are found in\ntopologically isomorphic neighborhoods, their embeddings are nearly similar to\none another. In this work, classification of cancer from WSIs is performed with\npositional embedding and graph attention. In order to represent the positional\nembedding of the nodes in graph classification, the proposed method makes use\nof spline convolutional neural networks (CNN). The algorithm is then tested\nwith the WSI dataset for grading prostate cancer and kidney cancer. A\ncomparison of the proposed method with leading approaches in cancer diagnosis\nand grading verify improved performance. The identification of cancerous\nregions in WSIs is another critical task in cancer diagnosis. In this work, the\nexplainability of the proposed model is also addressed. A gradient-based\nexplainbility approach is used to generate the saliency mapping for the WSIs.\nThis can be used to look into regions of WSI that are responsible for cancer\ndiagnosis thus rendering the proposed model explainable.\n", "date": "2023-06-14", "categories": []}
{"id": "2303.15693", "title": "Large-scale pretraining on pathological images for fine-tuning of small\n  pathological benchmarks", "abstract": "  Pretraining a deep learning model on large image datasets is a standard step\nbefore fine-tuning the model on small targeted datasets. The large dataset is\nusually general images (e.g. imagenet2012) while the small dataset can be\nspecialized datasets that have different distributions from the large dataset.\nHowever, this 'large-to-small' strategy is not well-validated when the large\ndataset is specialized and has a similar distribution to small datasets. We\nnewly compiled three hematoxylin and eosin-stained image datasets, one large\n(PTCGA200) and two magnification-adjusted small datasets (PCam200 and\nsegPANDA200). Major deep learning models were trained with supervised and\nself-supervised learning methods and fine-tuned on the small datasets for tumor\nclassification and tissue segmentation benchmarks. ResNet50 pretrained with\nMoCov2, SimCLR, and BYOL on PTCGA200 was better than imagenet2012 pretraining\nwhen fine-tuned on PTCGA200 (accuracy of 83.94%, 86.41%, 84.91%, and 82.72%,\nrespectively). ResNet50 pre-trained on PTCGA200 with MoCov2 exceeded the\nCOCOtrain2017-pretrained baseline and was the best in ResNet50 for the tissue\nsegmentation benchmark (mIoU of 63.53% and 63.22%). We found re-training\nimagenet-pretrained models (ResNet50, BiT-M-R50x1, and ViT-S/16) on PTCGA200\nimproved downstream benchmarks.\n", "date": "2023-03-28", "categories": []}
{"id": "2306.07831", "title": "Visual Language Pretrained Multiple Instance Zero-Shot Transfer for\n  Histopathology Images", "abstract": "  Contrastive visual language pretraining has emerged as a powerful method for\neither training new language-aware image encoders or augmenting existing\npretrained models with zero-shot visual recognition capabilities. However,\nexisting works typically train on large datasets of image-text pairs and have\nbeen designed to perform downstream tasks involving only small to medium\nsized-images, neither of which are applicable to the emerging field of\ncomputational pathology where there are limited publicly available paired\nimage-text datasets and each image can span up to 100,000 x 100,000 pixels. In\nthis paper we present MI-Zero, a simple and intuitive framework for unleashing\nthe zero-shot transfer capabilities of contrastively aligned image and text\nmodels on gigapixel histopathology whole slide images, enabling multiple\ndownstream diagnostic tasks to be carried out by pretrained encoders without\nrequiring any additional labels. MI-Zero reformulates zero-shot transfer under\nthe framework of multiple instance learning to overcome the computational\nchallenge of inference on extremely large images. We used over 550k pathology\nreports and other available in-domain text corpora to pre-train our text\nencoder. By effectively leveraging strong pre-trained encoders, our best model\npretrained on over 33k histopathology image-caption pairs achieves an average\nmedian zero-shot accuracy of 70.2% across three different real-world cancer\nsubtyping tasks. Our code is available at:\nhttps://github.com/mahmoodlab/MI-Zero.\n", "date": "2023-06-13", "categories": []}
{"id": "2306.06780", "title": "Multimodal Pathology Image Search Between H&E Slides and Multiplexed\n  Immunofluorescent Images", "abstract": "  We present an approach for multimodal pathology image search, using dynamic\ntime warping (DTW) on Variational Autoencoder (VAE) latent space that is fed\ninto a ranked choice voting scheme to retrieve multiplexed immunofluorescent\nimaging (mIF) that is most similar to a query H&E slide. Through training the\nVAE and applying DTW, we align and compare mIF and H&E slides. Our method\nimproves differential diagnosis and therapeutic decisions by integrating\nmorphological H&E data with immunophenotyping from mIF, providing clinicians a\nrich perspective of disease states. This facilitates an understanding of the\nspatial relationships in tissue samples and could revolutionize the diagnostic\nprocess, enhancing precision and enabling personalized therapy selection. Our\ntechnique demonstrates feasibility using colorectal cancer and healthy tonsil\nsamples. An exhaustive ablation study was conducted on a search engine designed\nto explore the correlation between multiplexed Immunofluorescence (mIF) and\nHematoxylin and Eosin (H&E) staining, in order to validate its ability to map\nthese distinct modalities into a unified vector space. Despite extreme class\nimbalance, the system demonstrated robustness and utility by returning similar\nresults across various data features, which suggests potential for future use\nin multimodal histopathology data analysis.\n", "date": "2023-06-11", "categories": []}
{"id": "2306.04709", "title": "Improved statistical benchmarking of digital pathology models using\n  pairwise frames evaluation", "abstract": "  Nested pairwise frames is a method for relative benchmarking of cell or\ntissue digital pathology models against manual pathologist annotations on a set\nof sampled patches. At a high level, the method compares agreement between a\ncandidate model and pathologist annotations with agreement among pathologists'\nannotations. This evaluation framework addresses fundamental issues of data\nsize and annotator variability in using manual pathologist annotations as a\nsource of ground truth for model validation. We implemented nested pairwise\nframes evaluation for tissue classification, cell classification, and cell\ncount prediction tasks and show results for cell and tissue models deployed on\nan H&E-stained melanoma dataset.\n", "date": "2023-06-07", "categories": []}
{"id": "2306.04763", "title": "Context-Aware Self-Supervised Learning of Whole Slide Images", "abstract": "  Presenting whole slide images (WSIs) as graph will enable a more efficient\nand accurate learning framework for cancer diagnosis. Due to the fact that a\nsingle WSI consists of billions of pixels and there is a lack of vast annotated\ndatasets required for computational pathology, the problem of learning from\nWSIs using typical deep learning approaches such as convolutional neural\nnetwork (CNN) is challenging. Additionally, WSIs down-sampling may lead to the\nloss of data that is essential for cancer detection. A novel two-stage learning\ntechnique is presented in this work. Since context, such as topological\nfeatures in the tumor surroundings, may hold important information for cancer\ngrading and diagnosis, a graph representation capturing all dependencies among\nregions in the WSI is very intuitive. Graph convolutional network (GCN) is\ndeployed to include context from the tumor and adjacent tissues, and\nself-supervised learning is used to enhance training through unlabeled data.\nMore specifically, the entire slide is presented as a graph, where the nodes\ncorrespond to the patches from the WSI. The proposed framework is then tested\nusing WSIs from prostate and kidney cancers. To assess the performance\nimprovement through self-supervised mechanism, the proposed context-aware model\nis tested with and without use of pre-trained self-supervised layer. The\noverall model is also compared with multi-instance learning (MIL) based and\nother existing approaches.\n", "date": "2023-06-07", "categories": []}
{"id": "2306.02691", "title": "Cyclic Learning: Bridging Image-level Labels and Nuclei Instance\n  Segmentation", "abstract": "  Nuclei instance segmentation on histopathology images is of great clinical\nvalue for disease analysis. Generally, fully-supervised algorithms for this\ntask require pixel-wise manual annotations, which is especially time-consuming\nand laborious for the high nuclei density. To alleviate the annotation burden,\nwe seek to solve the problem through image-level weakly supervised learning,\nwhich is underexplored for nuclei instance segmentation. Compared with most\nexisting methods using other weak annotations (scribble, point, etc.) for\nnuclei instance segmentation, our method is more labor-saving. The obstacle to\nusing image-level annotations in nuclei instance segmentation is the lack of\nadequate location information, leading to severe nuclei omission or overlaps.\nIn this paper, we propose a novel image-level weakly supervised method, called\ncyclic learning, to solve this problem. Cyclic learning comprises a front-end\nclassification task and a back-end semi-supervised instance segmentation task\nto benefit from multi-task learning (MTL). We utilize a deep learning\nclassifier with interpretability as the front-end to convert image-level labels\nto sets of high-confidence pseudo masks and establish a semi-supervised\narchitecture as the back-end to conduct nuclei instance segmentation under the\nsupervision of these pseudo masks. Most importantly, cyclic learning is\ndesigned to circularly share knowledge between the front-end classifier and the\nback-end semi-supervised part, which allows the whole system to fully extract\nthe underlying information from image-level labels and converge to a better\noptimum. Experiments on three datasets demonstrate the good generality of our\nmethod, which outperforms other image-level weakly supervised methods for\nnuclei instance segmentation, and achieves comparable performance to\nfully-supervised methods.\n", "date": "2023-06-05", "categories": []}
{"id": "2305.16465", "title": "An AI-Ready Multiplex Staining Dataset for Reproducible and Accurate\n  Characterization of Tumor Immune Microenvironment", "abstract": "  We introduce a new AI-ready computational pathology dataset containing\nrestained and co-registered digitized images from eight head-and-neck squamous\ncell carcinoma patients. Specifically, the same tumor sections were stained\nwith the expensive multiplex immunofluorescence (mIF) assay first and then\nrestained with cheaper multiplex immunohistochemistry (mIHC). This is a first\npublic dataset that demonstrates the equivalence of these two staining methods\nwhich in turn allows several use cases; due to the equivalence, our cheaper\nmIHC staining protocol can offset the need for expensive mIF staining/scanning\nwhich requires highly-skilled lab technicians. As opposed to subjective and\nerror-prone immune cell annotations from individual pathologists (disagreement\n> 50%) to drive SOTA deep learning approaches, this dataset provides objective\nimmune and tumor cell annotations via mIF/mIHC restaining for more reproducible\nand accurate characterization of tumor immune microenvironment (e.g. for\nimmunotherapy). We demonstrate the effectiveness of this dataset in three use\ncases: (1) IHC quantification of CD3/CD8 tumor-infiltrating lymphocytes via\nstyle transfer, (2) virtual translation of cheap mIHC stains to more expensive\nmIF stains, and (3) virtual tumor/immune cellular phenotyping on standard\nhematoxylin images. The dataset is available at\n\\url{https://github.com/nadeemlab/DeepLIIF}.\n", "date": "2023-05-25", "categories": []}
{"id": "2305.15773", "title": "Multi-scale Efficient Graph-Transformer for Whole Slide Image\n  Classification", "abstract": "  The multi-scale information among the whole slide images (WSIs) is essential\nfor cancer diagnosis. Although the existing multi-scale vision Transformer has\nshown its effectiveness for learning multi-scale image representation, it still\ncannot work well on the gigapixel WSIs due to their extremely large image\nsizes. To this end, we propose a novel Multi-scale Efficient Graph-Transformer\n(MEGT) framework for WSI classification. The key idea of MEGT is to adopt two\nindependent Efficient Graph-based Transformer (EGT) branches to process the\nlow-resolution and high-resolution patch embeddings (i.e., tokens in a\nTransformer) of WSIs, respectively, and then fuse these tokens via a\nmulti-scale feature fusion module (MFFM). Specifically, we design an EGT to\nefficiently learn the local-global information of patch tokens, which\nintegrates the graph representation into Transformer to capture spatial-related\ninformation of WSIs. Meanwhile, we propose a novel MFFM to alleviate the\nsemantic gap among different resolution patches during feature fusion, which\ncreates a non-patch token for each branch as an agent to exchange information\nwith another branch by cross-attention. In addition, to expedite network\ntraining, a novel token pruning module is developed in EGT to reduce the\nredundant tokens. Extensive experiments on TCGA-RCC and CAMELYON16 datasets\ndemonstrate the effectiveness of the proposed MEGT.\n", "date": "2023-05-25", "categories": []}
{"id": "2305.11968", "title": "An End-to-end Pipeline for 3D Slide-wise Multi-stain Renal Pathology\n  Registration", "abstract": "  Tissue examination and quantification in a 3D context on serial section whole\nslide images (WSIs) were laborintensive and time-consuming tasks. Our previous\nstudy proposed a novel registration-based method (Map3D) to automatically align\nWSIs to the same physical space, reducing the human efforts of screening serial\nsections from WSIs. However, the registration performance of our Map3D method\nwas only evaluated on single-stain WSIs with large-scale kidney tissue samples.\nIn this paper, we provide a Docker for an end-to-end 3D slide-wise registration\npipeline on needle biopsy serial sections in a multi-stain paradigm. The\ncontribution of this study is three-fold: (1) We release a containerized Docker\nfor an end-to-end multi-stain WSI registration. (2) We prove that the Map3D\npipeline is capable of sectional registration from multi-stain WSI. (3) We\nverify that the Map3D pipeline can also be applied to needle biopsy tissue\nsamples. The source code and the Docker have been made publicly available at\nhttps://github.com/hrlblab/Map3D.\n", "date": "2023-05-19", "categories": []}
{"id": "2303.12095", "title": "Interpretable histopathology-based prediction of disease relevant\n  features in Inflammatory Bowel Disease biopsies using weakly-supervised deep\n  learning", "abstract": "  Crohn's Disease (CD) and Ulcerative Colitis (UC) are the two main\nInflammatory Bowel Disease (IBD) types. We developed deep learning models to\nidentify histological disease features for both CD and UC using only endoscopic\nlabels. We explored fine-tuning and end-to-end training of two state-of-the-art\nself-supervised models for predicting three different endoscopic categories (i)\nCD vs UC (AUC=0.87), (ii) normal vs lesional (AUC=0.81), (iii) low vs high\ndisease severity score (AUC=0.80). We produced visual attention maps to\ninterpret what the models learned and validated them with the support of a\npathologist, where we observed a strong association between the models'\npredictions and histopathological inflammatory features of the disease.\nAdditionally, we identified several cases where the model incorrectly predicted\nnormal samples as lesional but were correct on the microscopic level when\nreviewed by the pathologist. This tendency of histological presentation to be\nmore severe than endoscopic presentation was previously published in the\nliterature. In parallel, we utilised a model trained on the Colon Nuclei\nIdentification and Counting (CoNIC) dataset to predict and explore 6 cell\npopulations. We observed correlation between areas enriched with the predicted\nimmune cells in biopsies and the pathologist's feedback on the attention maps.\nFinally, we identified several cell level features indicative of disease\nseverity in CD and UC. These models can enhance our understanding about the\npathology behind IBD and can shape our strategies for patient stratification in\nclinical trials.\n", "date": "2023-03-20", "categories": []}
{"id": "2302.00669", "title": "Detecting Histologic & Clinical Glioblastoma Patterns of Prognostic\n  Relevance", "abstract": "  Glioblastoma is the most common and aggressive malignant adult tumor of the\ncentral nervous system, with a grim prognosis and heterogeneous morphologic and\nmolecular profiles. Since adopting the current standard-of-care treatment 18\nyears ago, no substantial prognostic improvement has been noticed. Accurate\nprediction of patient overall survival (OS) from histopathology whole slide\nimages (WSI) integrated with clinical data using advanced computational methods\ncould optimize clinical decision-making and patient management. Here, we focus\non identifying prognostically relevant glioblastoma characteristics from H&E\nstained WSI & clinical data relating to OS. The exact approach for WSI\ncapitalizes on the comprehensive curation of apparent artifactual content and\nan interpretability mechanism via a weakly supervised attention-based\nmultiple-instance learning algorithm that further utilizes clustering to\nconstrain the search space. The automatically placed patterns of high\ndiagnostic value classify each WSI as representative of short or\nlong-survivors. Further assessment of the prognostic relevance of the\nassociated clinical patient data is performed both in isolation and in an\nintegrated manner, using XGBoost and SHapley Additive exPlanations (SHAP).\nIdentifying tumor morphological & clinical patterns associated with short and\nlong OS will enable the clinical neuropathologist to provide additional\nrelevant prognostic information to the treating team and suggest avenues of\nbiological investigation for understanding and potentially treating\nglioblastoma.\n", "date": "2023-02-01", "categories": []}
{"id": "2305.07404", "title": "Color Deconvolution applied to Domain Adaptation in HER2\n  histopathological images", "abstract": "  Breast cancer early detection is crucial for improving patient outcomes. The\nInstitut Catal\\`a de la Salut (ICS) has launched the DigiPatICS project to\ndevelop and implement artificial intelligence algorithms to assist with the\ndiagnosis of cancer. In this paper, we propose a new approach for facing the\ncolor normalization problem in HER2-stained histopathological images of breast\ncancer tissue, posed as an style transfer problem. We combine the Color\nDeconvolution technique with the Pix2Pix GAN network to present a novel\napproach to correct the color variations between different HER2 stain brands.\nOur approach focuses on maintaining the HER2 score of the cells in the\ntransformed images, which is crucial for the HER2 analysis. Results demonstrate\nthat our final model outperforms the state-of-the-art image style transfer\nmethods in maintaining the cell classes in the transformed images and is as\neffective as them in generating realistic images.\n", "date": "2023-05-12", "categories": []}
{"id": "2305.05100", "title": "Adaptive Domain Generalization for Digital Pathology Images", "abstract": "  In AI-based histopathology, domain shifts are common and well-studied.\nHowever, this research focuses on stain and scanner variations, which do not\nshow the full picture -- shifts may be combinations of other shifts, or\n\"invisible\" shifts that are not obvious but still damage performance of machine\nlearning models. Furthermore, it is important for models to generalize to these\nshifts without expensive or scarce annotations, especially in the\nhistopathology space and if wanting to deploy models on a larger scale. Thus,\nthere is a need for \"reactive\" domain generalization techniques: ones that\nadapt to domain shifts at test-time rather than requiring predictions of or\nexamples of the shifts at training time. We conduct a literature review and\nintroduce techniques that react to domain shifts rather than requiring a\nprediction of them in advance. We investigate test time training, a technique\nfor domain generalization that adapts model parameters at test-time through\noptimization of a secondary self-supervised task.\n", "date": "2023-05-09", "categories": []}
{"id": "2304.04507", "title": "hist2RNA: An efficient deep learning architecture to predict gene\n  expression from breast cancer histopathology images", "abstract": "  Gene expression can be used to subtype breast cancer with improved prediction\nof risk of recurrence and treatment responsiveness over that obtained using\nroutine immunohistochemistry (IHC). However, in the clinic, molecular profiling\nis primarily used for ER+ breast cancer, which is costly, tissue destructive,\nrequires specialized platforms and takes several weeks to obtain a result. Deep\nlearning algorithms can effectively extract morphological patterns in digital\nhistopathology images to predict molecular phenotypes quickly and\ncost-effectively. We propose a new, computationally efficient approach called\nhist2RNA inspired by bulk RNA-sequencing techniques to predict the expression\nof 138 genes (incorporated from six commercially available molecular profiling\ntests), including luminal PAM50 subtype, from hematoxylin and eosin (H&E)\nstained whole slide images (WSIs). The training phase involves the aggregation\nof extracted features for each patient from a pretrained model to predict gene\nexpression at the patient level using annotated H&E images from The Cancer\nGenome Atlas (TCGA, n=335). We demonstrate successful gene prediction on a\nheld-out test set (n = 160, corr = 0.82 across patients, corr = 0.29 across\ngenes) and perform exploratory analysis on an external tissue microarray (TMA)\ndataset (n = 498) with known IHC and survival information. Our model is able to\npredict gene expression and luminal PAM50 subtype (Luminal A versus Luminal B)\non the TMA dataset with prognostic significance for overall survival in\nunivariate analysis (c-index = 0.56, hazard ratio = 2.16 (95% CI 1.12-3.06), p\n< 5 x 10-3), and independent significance in multivariate analysis\nincorporating standard clinicopathological variables (c-index = 0.65, hazard\nratio = 1.85 (95% CI 1.30-2.68), p < 5 x 10-3).\n", "date": "2023-04-10", "categories": []}
{"id": "2305.02032", "title": "Unsupervised Mutual Transformer Learning for Multi-Gigapixel Whole Slide\n  Image Classification", "abstract": "  Classification of gigapixel Whole Slide Images (WSIs) is an important\nprediction task in the emerging area of computational pathology. There has been\na surge of research in deep learning models for WSI classification with\nclinical applications such as cancer detection or prediction of molecular\nmutations from WSIs. Most methods require expensive and labor-intensive manual\nannotations by expert pathologists. Weakly supervised Multiple Instance\nLearning (MIL) methods have recently demonstrated excellent performance;\nhowever, they still require large slide-level labeled training datasets that\nneed a careful inspection of each slide by an expert pathologist. In this work,\nwe propose a fully unsupervised WSI classification algorithm based on mutual\ntransformer learning. Instances from gigapixel WSI (i.e., image patches) are\ntransformed into a latent space and then inverse-transformed to the original\nspace. Using the transformation loss, pseudo-labels are generated and cleaned\nusing a transformer label-cleaner. The proposed transformer-based pseudo-label\ngeneration and cleaning modules mutually train each other iteratively in an\nunsupervised manner. A discriminative learning mechanism is introduced to\nimprove normal versus cancerous instance labeling. In addition to unsupervised\nclassification, we demonstrate the effectiveness of the proposed framework for\nweak supervision for cancer subtype classification as downstream analysis.\nExtensive experiments on four publicly available datasets show excellent\nperformance compared to the state-of-the-art methods. We intend to make the\nsource code of our algorithm publicly available soon.\n", "date": "2023-05-03", "categories": []}
{"id": "2305.00223", "title": "PathRTM: Real-time prediction of KI-67 and tumor-infiltrated lymphocytes", "abstract": "  In this paper, we introduce PathRTM, a novel deep neural network detector\nbased on RTMDet, for automated KI-67 proliferation and tumor-infiltrated\nlymphocyte estimation. KI-67 proliferation and tumor-infiltrated lymphocyte\nestimation play a crucial role in cancer diagnosis and treatment. PathRTM is an\nextension of the PathoNet work, which uses single pixel keypoints for within\neach cell. We demonstrate that PathRTM, with higher-level supervision in the\nform of bounding box labels generated automatically from the keypoints using\nNuClick, can significantly improve KI-67 proliferation and tumorinfiltrated\nlymphocyte estimation. Experiments on our custom dataset show that PathRTM\nachieves state-of-the-art performance in KI-67 immunopositive, immunonegative,\nand lymphocyte detection, with an average precision (AP) of 41.3%. Our results\nsuggest that PathRTM is a promising approach for accurate KI-67 proliferation\nand tumor-infiltrated lymphocyte estimation, offering annotation efficiency,\naccurate predictive capabilities, and improved runtime. The method also enables\nestimation of cell sizes of interest, which was previously unavailable, through\nthe bounding box predictions.\n", "date": "2023-04-23", "categories": []}
{"id": "2305.00696", "title": "TPMIL: Trainable Prototype Enhanced Multiple Instance Learning for Whole\n  Slide Image Classification", "abstract": "  Digital pathology based on whole slide images (WSIs) plays a key role in\ncancer diagnosis and clinical practice. Due to the high resolution of the WSI\nand the unavailability of patch-level annotations, WSI classification is\nusually formulated as a weakly supervised problem, which relies on multiple\ninstance learning (MIL) based on patches of a WSI. In this paper, we aim to\nlearn an optimal patch-level feature space by integrating prototype learning\nwith MIL. To this end, we develop a Trainable Prototype enhanced deep MIL\n(TPMIL) framework for weakly supervised WSI classification. In contrast to the\nconventional methods which rely on a certain number of selected patches for\nfeature space refinement, we softly cluster all the instances by allocating\nthem to their corresponding prototypes. Additionally, our method is able to\nreveal the correlations between different tumor subtypes through distances\nbetween corresponding trained prototypes. More importantly, TPMIL also enables\nto provide a more accurate interpretability based on the distance of the\ninstances from the trained prototypes which serves as an alternative to the\nconventional attention score-based interpretability. We test our method on two\nWSI datasets and it achieves a new SOTA. GitHub repository:\nhttps://github.com/LitaoYang-Jet/TPMIL\n", "date": "2023-05-01", "categories": []}
{"id": "2304.11445", "title": "Improving Stain Invariance of CNNs for Segmentation by Fusing Channel\n  Attention and Domain-Adversarial Training", "abstract": "  Variability in staining protocols, such as different slide preparation\ntechniques, chemicals, and scanner configurations, can result in a diverse set\nof whole slide images (WSIs). This distribution shift can negatively impact the\nperformance of deep learning models on unseen samples, presenting a significant\nchallenge for developing new computational pathology applications. In this\nstudy, we propose a method for improving the generalizability of convolutional\nneural networks (CNNs) to stain changes in a single-source setting for semantic\nsegmentation. Recent studies indicate that style features mainly exist as\ncovariances in earlier network layers. We design a channel attention mechanism\nbased on these findings that detects stain-specific features and modify the\npreviously proposed stain-invariant training scheme. We reweigh the outputs of\nearlier layers and pass them to the stain-adversarial training branch. We\nevaluate our method on multi-center, multi-stain datasets and demonstrate its\neffectiveness through interpretability analysis. Our approach achieves\nsubstantial improvements over baselines and competitive performance compared to\nother methods, as measured by various evaluation metrics. We also show that\ncombining our method with stain augmentation leads to mutually beneficial\nresults and outperforms other techniques. Overall, our study makes significant\ncontributions to the field of computational pathology.\n", "date": "2023-04-22", "categories": []}
{"id": "2304.12149", "title": "Exploring shared memory architectures for end-to-end gigapixel deep\n  learning", "abstract": "  Deep learning has made great strides in medical imaging, enabled by hardware\nadvances in GPUs. One major constraint for the development of new models has\nbeen the saturation of GPU memory resources during training. This is especially\ntrue in computational pathology, where images regularly contain more than 1\nbillion pixels. These pathological images are traditionally divided into small\npatches to enable deep learning due to hardware limitations. In this work, we\nexplore whether the shared GPU/CPU memory architecture on the M1 Ultra\nsystems-on-a-chip (SoCs) recently released by Apple, Inc. may provide a\nsolution. These affordable systems (less than \\$5000) provide access to 128 GB\nof unified memory (Mac Studio with M1 Ultra SoC). As a proof of concept for\ngigapixel deep learning, we identified tissue from background on gigapixel\nareas from whole slide images (WSIs). The model was a modified U-Net (4492\nparameters) leveraging large kernels and high stride. The M1 Ultra SoC was able\nto train the model directly on gigapixel images (16000$\\times$64000 pixels,\n1.024 billion pixels) with a batch size of 1 using over 100 GB of unified\nmemory for the process at an average speed of 1 minute and 21 seconds per batch\nwith Tensorflow 2/Keras. As expected, the model converged with a high Dice\nscore of 0.989 $\\pm$ 0.005. Training up until this point took 111 hours and 24\nminutes over 4940 steps. Other high RAM GPUs like the NVIDIA A100 (largest\ncommercially accessible at 80 GB, $\\sim$\\$15000) are not yet widely available\n(in preview for select regions on Amazon Web Services at \\$40.96/hour as a\ngroup of 8). This study is a promising step towards WSI-wise end-to-end deep\nlearning with prevalent network architectures.\n", "date": "2023-04-24", "categories": []}
{"id": "2210.09021", "title": "Histopathological Image Classification based on Self-Supervised Vision\n  Transformer and Weak Labels", "abstract": "  Whole Slide Image (WSI) analysis is a powerful method to facilitate the\ndiagnosis of cancer in tissue samples. Automating this diagnosis poses various\nissues, most notably caused by the immense image resolution and limited\nannotations. WSIs commonly exhibit resolutions of 100Kx100K pixels. Annotating\ncancerous areas in WSIs on the pixel level is prohibitively labor-intensive and\nrequires a high level of expert knowledge. Multiple instance learning (MIL)\nalleviates the need for expensive pixel-level annotations. In MIL, learning is\nperformed on slide-level labels, in which a pathologist provides information\nabout whether a slide includes cancerous tissue. Here, we propose Self-ViT-MIL,\na novel approach for classifying and localizing cancerous areas based on\nslide-level annotations, eliminating the need for pixel-wise annotated training\ndata. Self-ViT- MIL is pre-trained in a self-supervised setting to learn rich\nfeature representation without relying on any labels. The recent Vision\nTransformer (ViT) architecture builds the feature extractor of Self-ViT-MIL.\nFor localizing cancerous regions, a MIL aggregator with global attention is\nutilized. To the best of our knowledge, Self-ViT- MIL is the first approach to\nintroduce self-supervised ViTs in MIL-based WSI analysis tasks. We showcase the\neffectiveness of our approach on the common Camelyon16 dataset. Self-ViT-MIL\nsurpasses existing state-of-the-art MIL-based approaches in terms of accuracy\nand area under the curve (AUC).\n", "date": "2022-10-17", "categories": []}
{"id": "2212.04690", "title": "Benchmarking Self-Supervised Learning on Diverse Pathology Datasets", "abstract": "  Computational pathology can lead to saving human lives, but models are\nannotation hungry and pathology images are notoriously expensive to annotate.\nSelf-supervised learning has shown to be an effective method for utilizing\nunlabeled data, and its application to pathology could greatly benefit its\ndownstream tasks. Yet, there are no principled studies that compare SSL methods\nand discuss how to adapt them for pathology. To address this need, we execute\nthe largest-scale study of SSL pre-training on pathology image data, to date.\nOur study is conducted using 4 representative SSL methods on diverse downstream\ntasks. We establish that large-scale domain-aligned pre-training in pathology\nconsistently out-performs ImageNet pre-training in standard SSL settings such\nas linear and fine-tuning evaluations, as well as in low-label regimes.\nMoreover, we propose a set of domain-specific techniques that we experimentally\nshow leads to a performance boost. Lastly, for the first time, we apply SSL to\nthe challenging task of nuclei instance segmentation and show large and\nconsistent performance improvements under diverse settings.\n", "date": "2022-12-09", "categories": []}
{"id": "2303.00943", "title": "Evolutionary Computation in Action: Feature Selection for Deep Embedding\n  Spaces of Gigapixel Pathology Images", "abstract": "  One of the main obstacles of adopting digital pathology is the challenge of\nefficient processing of hyperdimensional digitized biopsy samples, called whole\nslide images (WSIs). Exploiting deep learning and introducing compact WSI\nrepresentations are urgently needed to accelerate image analysis and facilitate\nthe visualization and interpretability of pathology results in a postpandemic\nworld. In this paper, we introduce a new evolutionary approach for WSI\nrepresentation based on large-scale multi-objective optimization (LSMOP) of\ndeep embeddings. We start with patch-based sampling to feed KimiaNet , a\nhistopathology-specialized deep network, and to extract a multitude of feature\nvectors. Coarse multi-objective feature selection uses the reduced search space\nstrategy guided by the classification accuracy and the number of features. In\nthe second stage, the frequent features histogram (FFH), a novel WSI\nrepresentation, is constructed by multiple runs of coarse LSMOP. Fine\nevolutionary feature selection is then applied to find a compact (short-length)\nfeature vector based on the FFH and contributes to a more robust deep-learning\napproach to digital pathology supported by the stochastic power of evolutionary\nalgorithms. We validate the proposed schemes using The Cancer Genome Atlas\n(TCGA) images in terms of WSI representation, classification accuracy, and\nfeature quality. Furthermore, a novel decision space for multicriteria decision\nmaking in the LSMOP field is introduced. Finally, a patch-level visualization\napproach is proposed to increase the interpretability of deep features. The\nproposed evolutionary algorithm finds a very compact feature vector to\nrepresent a WSI (almost 14,000 times smaller than the original feature vectors)\nwith 8% higher accuracy compared to the codes provided by the state-of-the-art\nmethods.\n", "date": "2023-03-02", "categories": []}
{"id": "2304.07434", "title": "Masked Pre-Training of Transformers for Histology Image Analysis", "abstract": "  In digital pathology, whole slide images (WSIs) are widely used for\napplications such as cancer diagnosis and prognosis prediction. Visual\ntransformer models have recently emerged as a promising method for encoding\nlarge regions of WSIs while preserving spatial relationships among patches.\nHowever, due to the large number of model parameters and limited labeled data,\napplying transformer models to WSIs remains challenging. Inspired by masked\nlanguage models, we propose a pretext task for training the transformer model\nwithout labeled data to address this problem. Our model, MaskHIT, uses the\ntransformer output to reconstruct masked patches and learn representative\nhistological features based on their positions and visual features. The\nexperimental results demonstrate that MaskHIT surpasses various multiple\ninstance learning approaches by 3% and 2% on survival prediction and cancer\nsubtype classification tasks, respectively. Furthermore, MaskHIT also\noutperforms two of the most recent state-of-the-art transformer-based methods.\nFinally, a comparison between the attention maps generated by the MaskHIT model\nwith pathologist's annotations indicates that the model can accurately identify\nclinically relevant histological structures in each task.\n", "date": "2023-04-14", "categories": []}
{"id": "2301.06304", "title": "LYSTO: The Lymphocyte Assessment Hackathon and Benchmark Dataset", "abstract": "  We introduce LYSTO, the Lymphocyte Assessment Hackathon, which was held in\nconjunction with the MICCAI 2019 Conference in Shenzen (China). The competition\nrequired participants to automatically assess the number of lymphocytes, in\nparticular T-cells, in histopathological images of colon, breast, and prostate\ncancer stained with CD3 and CD8 immunohistochemistry. Differently from other\nchallenges setup in medical image analysis, LYSTO participants were solely\ngiven a few hours to address this problem. In this paper, we describe the goal\nand the multi-phase organization of the hackathon; we describe the proposed\nmethods and the on-site results. Additionally, we present post-competition\nresults where we show how the presented methods perform on an independent set\nof lung cancer slides, which was not part of the initial competition, as well\nas a comparison on lymphocyte assessment between presented methods and a panel\nof pathologists. We show that some of the participants were capable to achieve\npathologist-level performance at lymphocyte assessment. After the hackathon,\nLYSTO was left as a lightweight plug-and-play benchmark dataset on\ngrand-challenge website, together with an automatic evaluation platform. LYSTO\nhas supported a number of research in lymphocyte assessment in oncology. LYSTO\nwill be a long-lasting educational challenge for deep learning and digital\npathology, it is available at https://lysto.grand-challenge.org/.\n", "date": "2023-01-16", "categories": []}
{"id": "2304.05153", "title": "Regression-based Deep-Learning predicts molecular biomarkers from\n  pathology slides", "abstract": "  Deep Learning (DL) can predict biomarkers from cancer histopathology. Several\nclinically approved applications use this technology. Most approaches, however,\npredict categorical labels, whereas biomarkers are often continuous\nmeasurements. We hypothesized that regression-based DL outperforms\nclassification-based DL. Therefore, we developed and evaluated a new\nself-supervised attention-based weakly supervised regression method that\npredicts continuous biomarkers directly from images in 11,671 patients across\nnine cancer types. We tested our method for multiple clinically and\nbiologically relevant biomarkers: homologous repair deficiency (HRD) score, a\nclinically used pan-cancer biomarker, as well as markers of key biological\nprocesses in the tumor microenvironment. Using regression significantly\nenhances the accuracy of biomarker prediction, while also improving the\ninterpretability of the results over classification. In a large cohort of\ncolorectal cancer patients, regression-based prediction scores provide a higher\nprognostic value than classification-based scores. Our open-source regression\napproach offers a promising alternative for continuous biomarker analysis in\ncomputational pathology.\n", "date": "2023-04-11", "categories": []}
{"id": "2304.04142", "title": "Slideflow: Deep Learning for Digital Histopathology with Real-Time\n  Whole-Slide Visualization", "abstract": "  Deep learning methods have emerged as powerful tools for analyzing\nhistopathological images, but current methods are often specialized for\nspecific domains and software environments, and few open-source options exist\nfor deploying models in an interactive interface. Experimenting with different\ndeep learning approaches typically requires switching software libraries and\nreprocessing data, reducing the feasibility and practicality of experimenting\nwith new architectures. We developed a flexible deep learning library for\nhistopathology called Slideflow, a package which supports a broad array of deep\nlearning methods for digital pathology and includes a fast whole-slide\ninterface for deploying trained models. Slideflow includes unique tools for\nwhole-slide image data processing, efficient stain normalization and\naugmentation, weakly-supervised whole-slide classification, uncertainty\nquantification, feature generation, feature space analysis, and explainability.\nWhole-slide image processing is highly optimized, enabling whole-slide tile\nextraction at 40X magnification in 2.5 seconds per slide. The\nframework-agnostic data processing pipeline enables rapid experimentation with\nnew methods built with either Tensorflow or PyTorch, and the graphical user\ninterface supports real-time visualization of slides, predictions, heatmaps,\nand feature space characteristics on a variety of hardware devices, including\nARM-based devices such as the Raspberry Pi.\n", "date": "2023-04-09", "categories": []}
{"id": "2304.04155", "title": "Segment Anything Model (SAM) for Digital Pathology: Assess Zero-shot\n  Segmentation on Whole Slide Imaging", "abstract": "  The segment anything model (SAM) was released as a foundation model for image\nsegmentation. The promptable segmentation model was trained by over 1 billion\nmasks on 11M licensed and privacy-respecting images. The model supports\nzero-shot image segmentation with various segmentation prompts (e.g., points,\nboxes, masks). It makes the SAM attractive for medical image analysis,\nespecially for digital pathology where the training data are rare. In this\nstudy, we evaluate the zero-shot segmentation performance of SAM model on\nrepresentative segmentation tasks on whole slide imaging (WSI), including (1)\ntumor segmentation, (2) non-tumor tissue segmentation, (3) cell nuclei\nsegmentation. Core Results: The results suggest that the zero-shot SAM model\nachieves remarkable segmentation performance for large connected objects.\nHowever, it does not consistently achieve satisfying performance for dense\ninstance object segmentation, even with 20 prompts (clicks/boxes) on each\nimage. We also summarized the identified limitations for digital pathology: (1)\nimage resolution, (2) multiple scales, (3) prompt selection, and (4) model\nfine-tuning. In the future, the few-shot fine-tuning with images from\ndownstream pathological segmentation tasks might help the model to achieve\nbetter performance in dense object segmentation.\n", "date": "2023-04-09", "categories": []}
{"id": "2304.04600", "title": "Rotation-Scale Equivariant Steerable Filters", "abstract": "  Incorporating either rotation equivariance or scale equivariance into CNNs\nhas proved to be effective in improving models' generalization performance.\nHowever, jointly integrating rotation and scale equivariance into CNNs has not\nbeen widely explored. Digital histology imaging of biopsy tissue can be\ncaptured at arbitrary orientation and magnification and stored at different\nresolutions, resulting in cells appearing in different scales. When\nconventional CNNs are applied to histopathology image analysis, the\ngeneralization performance of models is limited because 1) a part of the\nparameters of filters are trained to fit rotation transformation, thus\ndecreasing the capability of learning other discriminative features; 2)\nfixed-size filters trained on images at a given scale fail to generalize to\nthose at different scales. To deal with these issues, we propose the\nRotation-Scale Equivariant Steerable Filter (RSESF), which incorporates\nsteerable filters and scale-space theory. The RSESF contains copies of filters\nthat are linear combinations of Gaussian filters, whose direction is controlled\nby directional derivatives and whose scale parameters are trainable but\nconstrained to span disjoint scales in successive layers of the network.\nExtensive experiments on two gland segmentation datasets demonstrate that our\nmethod outperforms other approaches, with much fewer trainable parameters and\nfewer GPU resources required. The source code is available at:\nhttps://github.com/ynulonger/RSESF.\n", "date": "2023-04-10", "categories": []}
{"id": "2304.02255", "title": "Topology-Guided Multi-Class Cell Context Generation for Digital\n  Pathology", "abstract": "  In digital pathology, the spatial context of cells is important for cell\nclassification, cancer diagnosis and prognosis. To model such complex cell\ncontext, however, is challenging. Cells form different mixtures, lineages,\nclusters and holes. To model such structural patterns in a learnable fashion,\nwe introduce several mathematical tools from spatial statistics and topological\ndata analysis. We incorporate such structural descriptors into a deep\ngenerative model as both conditional inputs and a differentiable loss. This\nway, we are able to generate high quality multi-class cell layouts for the\nfirst time. We show that the topology-rich cell layouts can be used for data\naugmentation and improve the performance of downstream tasks such as cell\nclassification.\n", "date": "2023-04-05", "categories": []}
{"id": "2303.13110", "title": "OCELOT: Overlapped Cell on Tissue Dataset for Histopathology", "abstract": "  Cell detection is a fundamental task in computational pathology that can be\nused for extracting high-level medical information from whole-slide images. For\naccurate cell detection, pathologists often zoom out to understand the\ntissue-level structures and zoom in to classify cells based on their morphology\nand the surrounding context. However, there is a lack of efforts to reflect\nsuch behaviors by pathologists in the cell detection models, mainly due to the\nlack of datasets containing both cell and tissue annotations with overlapping\nregions. To overcome this limitation, we propose and publicly release OCELOT, a\ndataset purposely dedicated to the study of cell-tissue relationships for cell\ndetection in histopathology. OCELOT provides overlapping cell and tissue\nannotations on images acquired from multiple organs. Within this setting, we\nalso propose multi-task learning approaches that benefit from learning both\ncell and tissue tasks simultaneously. When compared against a model trained\nonly for the cell detection task, our proposed approaches improve cell\ndetection performance on 3 datasets: proposed OCELOT, public TIGER, and\ninternal CARP datasets. On the OCELOT test set in particular, we show up to\n6.79 improvement in F1-score. We believe the contributions of this paper,\nincluding the release of the OCELOT dataset at\nhttps://lunit-io.github.io/research/publications/ocelot are a crucial starting\npoint toward the important research direction of incorporating cell-tissue\nrelationships in computation pathology.\n", "date": "2023-03-23", "categories": []}
{"id": "2303.13122", "title": "Exploring Visual Prompts for Whole Slide Image Classification with\n  Multiple Instance Learning", "abstract": "  Multiple instance learning (MIL) has emerged as a popular method for\nclassifying histopathology whole slide images (WSIs). However, existing\napproaches typically rely on pre-trained models from large natural image\ndatasets, such as ImageNet, to generate instance features, which can be\nsub-optimal due to the significant differences between natural images and\nhistopathology images that lead to a domain shift. In this paper, we present a\nnovel, simple yet effective method for learning domain-specific knowledge\ntransformation from pre-trained models to histopathology images. Our approach\nentails using a prompt component to assist the pre-trained model in discerning\ndifferences between the pre-trained dataset and the target histopathology\ndataset, resulting in improved performance of MIL models. We validate our\nmethod on two publicly available datasets, Camelyon16 and TCGA-NSCLC. Extensive\nexperimental results demonstrate the significant performance improvement of our\nmethod for different MIL models and backbones. Upon publication of this paper,\nwe will release the source code for our method.\n", "date": "2023-03-23", "categories": []}
{"id": "2303.10172", "title": "Hematoxylin and eosin stained oral squamous cell carcinoma histological\n  images dataset", "abstract": "  Computer-aided diagnosis (CAD) can be used as an important tool to aid and\nenhance pathologists' diagnostic decision-making. Deep learning techniques,\nsuch as convolutional neural networks (CNN) and fully convolutional networks\n(FCN), have been successfully applied in medical and biological research.\nUnfortunately, histological image segmentation is often constrained by the\navailability of labeled training data once labeling histological images for\nsegmentation purposes is a highly-skilled, complex, and time-consuming task.\nThis paper presents the hematoxylin and eosin (H&E) stained oral cavity-derived\ncancer (OCDC) dataset, a labeled dataset containing H&E-stained histological\nimages of oral squamous cell carcinoma (OSCC) cases. The tumor regions in our\ndataset are labeled manually by a specialist and validated by a pathologist.\nThe OCDC dataset presents 1,020 histological images of size 640x640 pixels\ncontaining tumor regions fully annotated for segmentation purposes. All the\nhistological images are digitized at 20x magnification.\n", "date": "2023-01-13", "categories": []}
{"id": "2303.10342", "title": "Whole-slide-imaging Cancer Metastases Detection and Localization with\n  Limited Tumorous Data", "abstract": "  Recently, various deep learning methods have shown significant successes in\nmedical image analysis, especially in the detection of cancer metastases in\nhematoxylin and eosin (H&E) stained whole-slide images (WSIs). However, in\norder to obtain good performance, these research achievements rely on hundreds\nof well-annotated WSIs. In this study, we tackle the tumor localization and\ndetection problem under the setting of few labeled whole slide images and\nintroduce a patch-based analysis pipeline based on the latest reverse knowledge\ndistillation architecture. To address the extremely unbalanced normal and\ntumorous samples in training sample collection, we applied the focal loss\nformula to the representation similarity metric for model optimization.\nCompared with prior arts, our method achieves similar performance by less than\nten percent of training samples on the public Camelyon16 dataset. In addition,\nthis is the first work that show the great potential of the knowledge\ndistillation models in computational histopathology.\n", "date": "2023-03-18", "categories": []}
{"id": "2303.10730", "title": "Pretrained Vision Models for Predicting High-Risk Breast Cancer Stage", "abstract": "  Cancer is increasingly a global health issue. Seconding cardiovascular\ndiseases, cancers are the second biggest cause of death in the world with\nmillions of people succumbing to the disease every year. According to the World\nHealth Organization (WHO) report, by the end of 2020, more than 7.8 million\nwomen have been diagnosed with breast cancer, making it the world's most\nprevalent cancer. In this paper, using the Nightingale Open Science dataset of\ndigital pathology (breast biopsy) images, we leverage the capabilities of\npre-trained computer vision models for the breast cancer stage prediction task.\nWhile individual models achieve decent performances, we find out that the\npredictions of an ensemble model are more efficient, and offer a winning\nsolution\\footnote{https://www.nightingalescience.org/updates/hbc1-results}. We\nalso provide analyses of the results and explore pathways for better\ninterpretability and generalization. Our code is open-source at\n\\url{https://github.com/bonaventuredossou/nightingale_winning_solution}\n", "date": "2023-03-19", "categories": []}
{"id": "2303.09956", "title": "GNNFormer: A Graph-based Framework for Cytopathology Report Generation", "abstract": "  Cytopathology report generation is a necessary step for the standardized\nexamination of pathology images. However, manually writing detailed reports\nbrings heavy workloads for pathologists. To improve efficiency, some existing\nworks have studied automatic generation of cytopathology reports, mainly by\napplying image caption generation frameworks with visual encoders originally\nproposed for natural images. A common weakness of these works is that they do\nnot explicitly model the structural information among cells, which is a key\nfeature of pathology images and provides significant information for making\ndiagnoses. In this paper, we propose a novel graph-based framework called\nGNNFormer, which seamlessly integrates graph neural network (GNN) and\nTransformer into the same framework, for cytopathology report generation. To\nthe best of our knowledge, GNNFormer is the first report generation method that\nexplicitly models the structural information among cells in pathology images.\nIt also effectively fuses structural information among cells, fine-grained\nmorphology features of cells and background features to generate high-quality\nreports. Experimental results on the NMI-WSI dataset show that GNNFormer can\noutperform other state-of-the-art baselines.\n", "date": "2023-03-17", "categories": []}
{"id": "2303.08446", "title": "Task-specific Fine-tuning via Variational Information Bottleneck for\n  Weakly-supervised Pathology Whole Slide Image Classification", "abstract": "  While Multiple Instance Learning (MIL) has shown promising results in digital\nPathology Whole Slide Image (WSI) classification, such a paradigm still faces\nperformance and generalization problems due to challenges in high computational\ncosts on Gigapixel WSIs and limited sample size for model training. To deal\nwith the computation problem, most MIL methods utilize a frozen pretrained\nmodel from ImageNet to obtain representations first. This process may lose\nessential information owing to the large domain gap and hinder the\ngeneralization of model due to the lack of image-level training-time\naugmentations. Though Self-supervised Learning (SSL) proposes viable\nrepresentation learning schemes, the improvement of the downstream task still\nneeds to be further explored in the conversion from the task-agnostic features\nof SSL to the task-specifics under the partial label supervised learning. To\nalleviate the dilemma of computation cost and performance, we propose an\nefficient WSI fine-tuning framework motivated by the Information Bottleneck\ntheory. The theory enables the framework to find the minimal sufficient\nstatistics of WSI, thus supporting us to fine-tune the backbone into a\ntask-specific representation only depending on WSI-level weak labels. The\nWSI-MIL problem is further analyzed to theoretically deduce our fine-tuning\nmethod. Our framework is evaluated on five pathology WSI datasets on various\nWSI heads. The experimental results of our fine-tuned representations show\nsignificant improvements in both accuracy and generalization compared with\nprevious works. Source code will be available at\nhttps://github.com/invoker-LL/WSI-finetuning.\n", "date": "2023-03-15", "categories": []}
{"id": "2208.12506", "title": "EGFR Mutation Prediction of Lung Biopsy Images using Deep Learning", "abstract": "  The standard diagnostic procedures for targeted therapies in lung cancer\ntreatment involve histological subtyping and subsequent detection of key driver\nmutations, such as EGFR. Even though molecular profiling can uncover the driver\nmutation, the process is often expensive and time-consuming. Deep\nlearning-oriented image analysis offers a more economical alternative for\ndiscovering driver mutations directly from whole slide images (WSIs). In this\nwork, we used customized deep learning pipelines with weak supervision to\nidentify the morphological correlates of EGFR mutation from hematoxylin and\neosin-stained WSIs, in addition to detecting tumor and histologically subtyping\nit. We demonstrate the effectiveness of our pipeline by conducting rigorous\nexperiments and ablation studies on two lung cancer datasets - TCGA and a\nprivate dataset from India. With our pipeline, we achieved an average area\nunder the curve (AUC) of 0.964 for tumor detection, and 0.942 for histological\nsubtyping between adenocarcinoma and squamous cell carcinoma on the TCGA\ndataset. For EGFR detection, we achieved an average AUC of 0.864 on the TCGA\ndataset and 0.783 on the dataset from India. Our key learning points include\nthe following. Firstly, there is no particular advantage of using a feature\nextractor layers trained on histology, if one is going to fine-tune the feature\nextractor on the target dataset. Secondly, selecting patches with high\ncellularity, presumably capturing tumor regions, is not always helpful, as the\nsign of a disease class may be present in the tumor-adjacent stroma.\n", "date": "2022-08-26", "categories": []}
{"id": "2303.06727", "title": "Increasing the usefulness of already existing annotations through WSI\n  registration", "abstract": "  Computational pathology methods have the potential to improve access to\nprecision medicine, as well as the reproducibility and accuracy of pathological\ndiagnoses. Particularly the analysis of whole-slide-images (WSIs) of\nimmunohistochemically (IHC) stained tissue sections could benefit from\ncomputational pathology methods. However, scoring biomarkers such as KI67 in\nIHC WSIs often necessitates the detection of areas of invasive cancer. Training\ncancer detection models often requires annotations, which is time-consuming and\ntherefore costly. Currently, cancer regions are typically annotated in WSIs of\nhaematoxylin and eosin (H&E) stained tissue sections. In this study, we\ninvestigate the possibility to register annotations that were made in H&E WSIs\nto their IHC counterparts. Two pathologists annotated regions of invasive\ncancer in WSIs of 272 breast cancer cases. For each case, a matched H&E and\nKI67 WSI are available, resulting in 544 WSIs with invasive cancer annotations.\nWe find that cancer detection CNNs that were trained with annotations\nregistered from the H&E to the KI67 WSIs only differ slightly in calibration\nbut not in performance compared to cancer detection models trained on\nannotations made directly in the KI67 WSIs in a test set consisting of 54\ncases. The mean slide-level AUROC is 0.974 [0.964, 0.982] for models trained\nwith the KI67 annotations and 0.974 [0.965, 0.982] for models trained using\nregistered annotations. This indicates that WSI registration has the potential\nto reduce the need for IHC-specific annotations. This could significantly\nincrease the usefulness of already existing annotations.\n", "date": "2023-03-12", "categories": []}
{"id": "2303.06873", "title": "Interventional Bag Multi-Instance Learning On Whole-Slide Pathological\n  Images", "abstract": "  Multi-instance learning (MIL) is an effective paradigm for whole-slide\npathological images (WSIs) classification to handle the gigapixel resolution\nand slide-level label. Prevailing MIL methods primarily focus on improving the\nfeature extractor and aggregator. However, one deficiency of these methods is\nthat the bag contextual prior may trick the model into capturing spurious\ncorrelations between bags and labels. This deficiency is a confounder that\nlimits the performance of existing MIL methods. In this paper, we propose a\nnovel scheme, Interventional Bag Multi-Instance Learning (IBMIL), to achieve\ndeconfounded bag-level prediction. Unlike traditional likelihood-based\nstrategies, the proposed scheme is based on the backdoor adjustment to achieve\nthe interventional training, thus is capable of suppressing the bias caused by\nthe bag contextual prior. Note that the principle of IBMIL is orthogonal to\nexisting bag MIL methods. Therefore, IBMIL is able to bring consistent\nperformance boosting to existing schemes, achieving new state-of-the-art\nperformance. Code is available at https://github.com/HHHedo/IBMIL.\n", "date": "2023-03-13", "categories": []}
{"id": "2211.14491", "title": "Human-machine Interactive Tissue Prototype Learning for Label-efficient\n  Histopathology Image Segmentation", "abstract": "  Recently, deep neural networks have greatly advanced histopathology image\nsegmentation but usually require abundant annotated data. However, due to the\ngigapixel scale of whole slide images and pathologists' heavy daily workload,\nobtaining pixel-level labels for supervised learning in clinical practice is\noften infeasible. Alternatively, weakly-supervised segmentation methods have\nbeen explored with less laborious image-level labels, but their performance is\nunsatisfactory due to the lack of dense supervision. Inspired by the recent\nsuccess of self-supervised learning methods, we present a label-efficient\ntissue prototype dictionary building pipeline and propose to use the obtained\nprototypes to guide histopathology image segmentation. Particularly, taking\nadvantage of self-supervised contrastive learning, an encoder is trained to\nproject the unlabeled histopathology image patches into a discriminative\nembedding space where these patches are clustered to identify the tissue\nprototypes by efficient pathologists' visual examination. Then, the encoder is\nused to map the images into the embedding space and generate pixel-level pseudo\ntissue masks by querying the tissue prototype dictionary. Finally, the pseudo\nmasks are used to train a segmentation network with dense supervision for\nbetter performance. Experiments on two public datasets demonstrate that our\nhuman-machine interactive tissue prototype learning method can achieve\ncomparable segmentation performance as the fully-supervised baselines with less\nannotation burden and outperform other weakly-supervised methods. Codes will be\navailable upon publication.\n", "date": "2022-11-26", "categories": []}
{"id": "2303.05180", "title": "Classification in Histopathology: A unique deep embeddings extractor for\n  multiple classification tasks", "abstract": "  In biomedical imaging, deep learning-based methods are state-of-the-art for\nevery modality (virtual slides, MRI, etc.) In histopathology, these methods can\nbe used to detect certain biomarkers or classify lesions. However, such\ntechniques require large amounts of data to train high-performing models which\ncan be intrinsically difficult to acquire, especially when it comes to scarce\nbiomarkers. To address this challenge, we use a single, pre-trained, deep\nembeddings extractor to convert images into deep features and train small,\ndedicated classification head on these embeddings for each classification task.\nThis approach offers several benefits such as the ability to reuse a single\npre-trained deep network for various tasks; reducing the amount of labeled data\nneeded as classification heads have fewer parameters; and accelerating training\ntime by up to 1000 times, which allows for much more tuning of the\nclassification head. In this work, we perform an extensive comparison of\nvarious open-source backbones and assess their fit to the target histological\nimage domain. This is achieved using a novel method based on a proxy\nclassification task. We demonstrate that thanks to this selection method, an\noptimal feature extractor can be selected for different tasks on the target\ndomain. We also introduce a feature space augmentation strategy which proves to\nsubstantially improve the final metrics computed for the different tasks\nconsidered. To demonstrate the benefit of such backbone selection and\nfeature-space augmentation, our experiments are carried out on three separate\nclassification tasks and show a clear improvement on each of them:\nmicrocalcifications (29.1% F1-score increase), lymph nodes metastasis (12.5%\nF1-score increase), mitosis (15.0% F1-score increase).\n", "date": "2023-03-09", "categories": []}
{"id": "2303.04604", "title": "Simple and Efficient Confidence Score for Grading Whole Slide Images", "abstract": "  Grading precancerous lesions on whole slide images is a challenging task: the\ncontinuous space of morphological phenotypes makes clear-cut decisions between\ndifferent grades often difficult, leading to low inter- and intra-rater\nagreements. More and more Artificial Intelligence (AI) algorithms are developed\nto help pathologists perform and standardize their diagnosis. However, those\nmodels can render their prediction without consideration of the ambiguity of\nthe classes and can fail without notice which prevent their wider acceptance in\na clinical context. In this paper, we propose a new score to measure the\nconfidence of AI models in grading tasks. Our confidence score is specifically\nadapted to ordinal output variables, is versatile and does not require extra\ntraining or additional inferences nor particular architecture changes.\nComparison to other popular techniques such as Monte Carlo Dropout and deep\nensembles shows that our method provides state-of-the art results, while being\nsimpler, more versatile and less computationally intensive. The score is also\neasily interpretable and consistent with real life hesitations of pathologists.\nWe show that the score is capable of accurately identifying mispredicted slides\nand that accuracy for high confidence decisions is significantly higher than\nfor low-confidence decisions (gap in AUC of 17.1% on the test set). We believe\nthat the proposed confidence score could be leveraged by pathologists directly\nin their workflow and assist them on difficult tasks such as grading\nprecancerous lesions.\n", "date": "2023-03-08", "categories": []}
{"id": "2303.01605", "title": "Hierarchical discriminative learning improves visual representations of\n  biomedical microscopy", "abstract": "  Learning high-quality, self-supervised, visual representations is essential\nto advance the role of computer vision in biomedical microscopy and clinical\nmedicine. Previous work has focused on self-supervised representation learning\n(SSL) methods developed for instance discrimination and applied them directly\nto image patches, or fields-of-view, sampled from gigapixel whole-slide images\n(WSIs) used for cancer diagnosis. However, this strategy is limited because it\n(1) assumes patches from the same patient are independent, (2) neglects the\npatient-slide-patch hierarchy of clinical biomedical microscopy, and (3)\nrequires strong data augmentations that can degrade downstream performance.\nImportantly, sampled patches from WSIs of a patient's tumor are a diverse set\nof image examples that capture the same underlying cancer diagnosis. This\nmotivated HiDisc, a data-driven method that leverages the inherent\npatient-slide-patch hierarchy of clinical biomedical microscopy to define a\nhierarchical discriminative learning task that implicitly learns features of\nthe underlying diagnosis. HiDisc uses a self-supervised contrastive learning\nframework in which positive patch pairs are defined based on a common ancestry\nin the data hierarchy, and a unified patch, slide, and patient discriminative\nlearning objective is used for visual SSL. We benchmark HiDisc visual\nrepresentations on two vision tasks using two biomedical microscopy datasets,\nand demonstrate that (1) HiDisc pretraining outperforms current\nstate-of-the-art self-supervised pretraining methods for cancer diagnosis and\ngenetic mutation prediction, and (2) HiDisc learns high-quality visual\nrepresentations using natural patch diversity without strong data\naugmentations.\n", "date": "2023-03-02", "categories": []}
{"id": "2303.01342", "title": "Active Learning Enhances Classification of Histopathology Whole Slide\n  Images with Attention-based Multiple Instance Learning", "abstract": "  In many histopathology tasks, sample classification depends on morphological\ndetails in tissue or single cells that are only visible at the highest\nmagnification. For a pathologist, this implies tedious zooming in and out,\nwhile for a computational decision support algorithm, it leads to the analysis\nof a huge number of small image patches per whole slide image (WSI).\nAttention-based multiple instance learning (MIL), where attention estimation is\nlearned in a weakly supervised manner, has been successfully applied in\ncomputational histopathology, but it is challenged by large numbers of\nirrelevant patches, reducing its accuracy. Here, we present an active learning\napproach to the problem. Querying the expert to annotate regions of interest in\na WSI guides the formation of high-attention regions for MIL. We train an\nattention-based MIL and calculate a confidence metric for every image in the\ndataset to select the most uncertain WSIs for expert annotation. We test our\napproach on the CAMELYON17 dataset classifying metastatic lymph node sections\nin breast cancer. With a novel attention guiding loss, this leads to an\naccuracy boost of the trained models with few regions annotated for each class.\nActive learning thus improves WSIs classification accuracy, leads to faster and\nmore robust convergence, and speeds up the annotation process. It may in the\nfuture serve as an important contribution to train MIL models in the clinically\nrelevant context of cancer classification in histopathology.\n", "date": "2023-03-02", "categories": []}
{"id": "2301.09617", "title": "Fully transformer-based biomarker prediction from colorectal cancer\n  histology: a large-scale multicentric study", "abstract": "  Background: Deep learning (DL) can extract predictive and prognostic\nbiomarkers from routine pathology slides in colorectal cancer. For example, a\nDL test for the diagnosis of microsatellite instability (MSI) in CRC has been\napproved in 2022. Current approaches rely on convolutional neural networks\n(CNNs). Transformer networks are outperforming CNNs and are replacing them in\nmany applications, but have not been used for biomarker prediction in cancer at\na large scale. In addition, most DL approaches have been trained on small\npatient cohorts, which limits their clinical utility. Methods: In this study,\nwe developed a new fully transformer-based pipeline for end-to-end biomarker\nprediction from pathology slides. We combine a pre-trained transformer encoder\nand a transformer network for patch aggregation, capable of yielding single and\nmulti-target prediction at patient level. We train our pipeline on over 9,000\npatients from 10 colorectal cancer cohorts. Results: A fully transformer-based\napproach massively improves the performance, generalizability, data efficiency,\nand interpretability as compared with current state-of-the-art algorithms.\nAfter training on a large multicenter cohort, we achieve a sensitivity of 0.97\nwith a negative predictive value of 0.99 for MSI prediction on surgical\nresection specimens. We demonstrate for the first time that resection\nspecimen-only training reaches clinical-grade performance on endoscopic biopsy\ntissue, solving a long-standing diagnostic problem. Interpretation: A fully\ntransformer-based end-to-end pipeline trained on thousands of pathology slides\nyields clinical-grade performance for biomarker prediction on surgical\nresections and biopsies. Our new methods are freely available under an open\nsource license.\n", "date": "2023-01-23", "categories": []}
{"id": "2212.07724", "title": "Attention-based Multiple Instance Learning for Survival Prediction on\n  Lung Cancer Tissue Microarrays", "abstract": "  Attention-based multiple instance learning (AMIL) algorithms have proven to\nbe successful in utilizing gigapixel whole-slide images (WSIs) for a variety of\ndifferent computational pathology tasks such as outcome prediction and cancer\nsubtyping problems. We extended an AMIL approach to the task of survival\nprediction by utilizing the classical Cox partial likelihood as a loss\nfunction, converting the AMIL model into a nonlinear proportional hazards\nmodel. We applied the model to tissue microarray (TMA) slides of 330 lung\ncancer patients. The results show that AMIL approaches can handle very small\namounts of tissue from a TMA and reach similar C-index performance compared to\nestablished survival prediction methods trained with highly discriminative\nclinical factors such as age, cancer grade, and cancer stage\n", "date": "2022-12-15", "categories": []}
{"id": "2302.08867", "title": "Efficient subtyping of ovarian cancer histopathology whole slide images\n  using active sampling in multiple instance learning", "abstract": "  Weakly-supervised classification of histopathology slides is a\ncomputationally intensive task, with a typical whole slide image (WSI)\ncontaining billions of pixels to process. We propose Discriminative Region\nActive Sampling for Multiple Instance Learning (DRAS-MIL), a computationally\nefficient slide classification method using attention scores to focus sampling\non highly discriminative regions. We apply this to the diagnosis of ovarian\ncancer histological subtypes, which is an essential part of the patient care\npathway as different subtypes have different genetic and molecular profiles,\ntreatment options, and patient outcomes. We use a dataset of 714 WSIs acquired\nfrom 147 epithelial ovarian cancer patients at Leeds Teaching Hospitals NHS\nTrust to distinguish the most common subtype, high-grade serous carcinoma, from\nthe other four subtypes (low-grade serous, endometrioid, clear cell, and\nmucinous carcinomas) combined. We demonstrate that DRAS-MIL can achieve similar\nclassification performance to exhaustive slide analysis, with a 3-fold\ncross-validated AUC of 0.8679 compared to 0.8781 with standard attention-based\nMIL classification. Our approach uses at most 18% as much memory as the\nstandard approach, while taking 33% of the time when evaluating on a GPU and\nonly 14% on a CPU alone. Reducing prediction time and memory requirements may\nbenefit clinical deployment and the democratisation of AI, reducing the extent\nto which computational hardware limits end-user adoption.\n", "date": "2023-02-17", "categories": []}
{"id": "2302.10406", "title": "Time to Embrace Natural Language Processing (NLP)-based Digital\n  Pathology: Benchmarking NLP- and Convolutional Neural Network-based Deep\n  Learning Pipelines", "abstract": "  NLP-based computer vision models, particularly vision transformers, have been\nshown to outperform CNN models in many imaging tasks. However, most digital\npathology artificial-intelligence models are based on CNN architectures,\nprobably owing to a lack of data regarding NLP models for pathology images. In\nthis study, we developed digital pathology pipelines to benchmark the five most\nrecently proposed NLP models (vision transformer (ViT), Swin Transformer,\nMobileViT, CMT, and Sequencer2D) and four popular CNN models (ResNet18,\nResNet50, MobileNetV2, and EfficientNet) to predict biomarkers in colorectal\ncancer (microsatellite instability, CpG island methylator phenotype, and BRAF\nmutation). Hematoxylin and eosin-stained whole-slide images from Molecular and\nCellular Oncology and The Cancer Genome Atlas were used as training and\nexternal validation datasets, respectively. Cross-study external validations\nrevealed that the NLP-based models significantly outperformed the CNN-based\nmodels in biomarker prediction tasks, improving the overall prediction and\nprecision up to approximately 10% and 26%, respectively. Notably, compared with\nexisting models in the current literature using large training datasets, our\nNLP models achieved state-of-the-art predictions for all three biomarkers using\na relatively small training dataset, suggesting that large training datasets\nare not a prerequisite for NLP models or transformers, and NLP may be more\nsuitable for clinical studies in which small training datasets are commonly\ncollected. The superior performance of Sequencer2D suggests that further\nresearch and innovation on both transformer and bidirectional long short-term\nmemory architectures are warranted in the field of digital pathology. NLP\nmodels can replace classic CNN architectures and become the new workhorse\nbackbone in the field of digital pathology.\n", "date": "2023-02-21", "categories": []}
{"id": "2302.08511", "title": "Visual deep learning-based explanation for neuritic plaques segmentation\n  in Alzheimer's Disease using weakly annotated whole slide histopathological\n  images", "abstract": "  Quantifying the distribution and morphology of tau protein structures in\nbrain tissues is key to diagnosing Alzheimer's Disease (AD) and its subtypes.\nRecently, deep learning (DL) models such as UNet have been successfully used\nfor automatic segmentation of histopathological whole slide images (WSI) of\nbiological tissues. In this study, we propose a DL-based methodology for\nsemantic segmentation of tau lesions (i.e., neuritic plaques) in WSI of\npostmortem patients with AD. The state of the art in semantic segmentation of\nneuritic plaques in human WSI is very limited. Our study proposes a baseline\nable to generate a significant advantage for morphological analysis of these\ntauopathies for further stratification of AD patients. Essential discussions\nconcerning biomarkers (ALZ50 versus AT8 tau antibodies), the imaging modality\n(different slide scanner resolutions), and the challenge of weak annotations\nare addressed within this seminal study. The analysis of the impact of context\nin plaque segmentation is important to understand the role of the\nmicro-environment for reliable tau protein segmentation. In addition, by\nintegrating visual interpretability, we are able to explain how the network\nfocuses on a region of interest (ROI), giving additional insights to\npathologists. Finally, the release of a new expert-annotated database and the\ncode (\\url{https://github.com/aramis-lab/miccai2022-stratifiad.git}) will be\nhelpful for the scientific community to accelerate the development of new\npipelines for human WSI processing in AD.\n", "date": "2023-01-13", "categories": []}
{"id": "2302.05830", "title": "NephroNet: A Novel Program for Identifying Renal Cell Carcinoma and\n  Generating Synthetic Training Images with Convolutional Neural Networks and\n  Diffusion Models", "abstract": "  Renal cell carcinoma (RCC) is a type of cancer that originates in the kidneys\nand is the most common type of kidney cancer in adults. It can be classified\ninto several subtypes, including clear cell RCC, papillary RCC, and chromophobe\nRCC. In this study, an artificial intelligence model was developed and trained\nfor classifying different subtypes of RCC using ResNet-18, a convolutional\nneural network that has been widely used for image classification tasks. The\nmodel was trained on a dataset of RCC histopathology images, which consisted of\ndigital images of RCC surgical resection slides that were annotated with the\ncorresponding subtype labels. The performance of the trained model was\nevaluated using several metrics, including accuracy, precision, and recall.\nAdditionally, in this research, a novel synthetic image generation tool,\nNephroNet, is developed on diffusion models that are used to generate original\nimages of RCC surgical resection slides. Diffusion models are a class of\ngenerative models capable of synthesizing high-quality images from noise.\nSeveral diffusers such as Stable Diffusion, Dreambooth Text-to-Image, and\nTextual Inversion were trained on a dataset of RCC images and were used to\ngenerate a series of original images that resembled RCC surgical resection\nslides, all within the span of fewer than four seconds. The generated images\nwere visually realistic and could be used for creating new training datasets,\ntesting the performance of image analysis algorithms, and training medical\nprofessionals. NephroNet is provided as an open-source software package and\ncontains files for data preprocessing, training, and visualization. Overall,\nthis study demonstrates the potential of artificial intelligence and diffusion\nmodels for classifying and generating RCC images, respectively. These methods\ncould be useful for improving the diagnosis and treatment of RCC and more.\n", "date": "2023-02-12", "categories": []}
{"id": "2302.03120", "title": "Studying Therapy Effects and Disease Outcomes in Silico using Artificial\n  Counterfactual Tissue Samples", "abstract": "  Understanding the interactions of different cell types inside the immune\ntumor microenvironment (iTME) is crucial for the development of immunotherapy\ntreatments as well as for predicting their outcomes. Highly multiplexed tissue\nimaging (HMTI) technologies offer a tool which can capture cell properties of\ntissue samples by measuring expression of various proteins and storing them in\nseparate image channels. HMTI technologies can be used to gain insights into\nthe iTME and in particular how the iTME differs for different patient outcome\ngroups of interest (e.g., treatment responders vs. non-responders).\nUnderstanding the systematic differences in the iTME of different patient\noutcome groups is crucial for developing better treatments and personalising\nexisting treatments. However, such analyses are inherently limited by the fact\nthat any two tissue samples vary due to a large number of factors unrelated to\nthe outcome. Here, we present CF-HistoGAN, a machine learning framework that\nemploys generative adversarial networks (GANs) to create artificial\ncounterfactual tissue samples that resemble the original tissue samples as\nclosely as possible but capture the characteristics of a different patient\noutcome group. Specifically, we learn to \"translate\" HMTI samples from one\npatient group to create artificial paired samples. We show that this approach\nallows to directly study the effects of different patient outcomes on the iTMEs\nof individual tissue samples. We demonstrate that CF-HistoGAN can be employed\nas an explorative tool for understanding iTME effects on the pixel level.\nMoreover, we show that our method can be used to identify statistically\nsignificant differences in the expression of different proteins between patient\ngroups with greater sensitivity compared to conventional approaches.\n", "date": "2023-02-06", "categories": []}
{"id": "2206.08182", "title": "Nucleus Segmentation and Analysis in Breast Cancer with the MIScnn\n  Framework", "abstract": "  The NuCLS dataset contains over 220.000 annotations of cell nuclei in breast\ncancers. We show how to use these data to create a multi-rater model with the\nMIScnn Framework to automate the analysis of cell nuclei. For the model\ncreation, we use the widespread U-Net approach embedded in a pipeline. This\npipeline provides besides the high performance convolution neural network,\nseveral preprocessor techniques and a extended data exploration. The final\nmodel is tested in the evaluation phase using a wide variety of metrics with a\nsubsequent visualization. Finally, the results are compared and interpreted\nwith the results of the NuCLS study. As an outlook, indications are given which\nare important for the future development of models in the context of cell\nnuclei.\n", "date": "2022-06-16", "categories": []}
{"id": "2206.03003", "title": "Transformer-based Personalized Attention Mechanism for Medical Images\n  with Clinical Records", "abstract": "  In medical image diagnosis, identifying the attention region, i.e., the\nregion of interest for which the diagnosis is made, is an important task.\nVarious methods have been developed to automatically identify target regions\nfrom given medical images. However, in actual medical practice, the diagnosis\nis made based not only on the images but also on a variety of clinical records.\nThis means that pathologists examine medical images with some prior knowledge\nof the patients and that the attention regions may change depending on the\nclinical records. In this study, we propose a method called the Personalized\nAttention Mechanism (PersAM), by which the attention regions in medical images\nare adaptively changed according to the clinical records. The primary idea of\nthe PersAM method is to encode the relationships between the medical images and\nclinical records using a variant of Transformer architecture. To demonstrate\nthe effectiveness of the PersAM method, we applied it to a large-scale digital\npathology problem of identifying the subtypes of 842 malignant lymphoma\npatients based on their gigapixel whole slide images and clinical records.\n", "date": "2022-06-07", "categories": []}
{"id": "2301.09985", "title": "Navigating the reporting guideline environment for computational\n  pathology: A review", "abstract": "  The application of new artificial intelligence (AI) discoveries is\ntransforming healthcare research. However, the standards of reporting are\nvariable in this still evolving field, leading to potential research waste. The\naim of this work is to highlight resources and reporting guidelines available\nto researchers working in computational pathology. The EQUATOR Network library\nof reporting guidelines and extensions was systematically searched up to August\n2022 to identify applicable resources. Inclusion and exclusion criteria were\nused and guidance was screened for utility at different stages of research and\nfor a range of study types. Items were compiled to create a summary for easy\nidentification of useful resources and guidance. Over 70 published resources\napplicable to pathology AI research were identified. Guidelines were divided\ninto key categories, reflecting current study types and target areas for AI\nresearch: Literature & Research Priorities, Discovery, Clinical Trial,\nImplementation and Post-Implementation & Guidelines. Guidelines useful at\nmultiple stages of research and those currently in development were also\nhighlighted. Summary tables with links to guidelines for these groups were\ndeveloped, to assist those working in cancer AI research with complete\nreporting of research. Issues with replication and research waste are\nrecognised problems in AI research. Reporting guidelines can be used as\ntemplates to ensure the essential information needed to replicate research is\nincluded within journal articles and abstracts. Reporting guidelines are\navailable and useful for many study types, but greater awareness is needed to\nencourage researchers to utilise them and for journals to adopt them. This\nreview and summary of resources highlights guidance to researchers, aiming to\nimprove completeness of reporting.\n", "date": "2023-01-03", "categories": []}
{"id": "2301.09624", "title": "Maximum Mean Discrepancy Kernels for Predictive and Prognostic Modeling\n  of Whole Slide Images", "abstract": "  How similar are two images? In computational pathology, where Whole Slide\nImages (WSIs) of digitally scanned tissue samples from patients can be\nmulti-gigapixels in size, determination of degree of similarity between two\nWSIs is a challenging task with a number of practical applications. In this\nwork, we explore a novel strategy based on kernelized Maximum Mean Discrepancy\n(MMD) analysis for determination of pairwise similarity between WSIs. The\nproposed approach works by calculating MMD between two WSIs using kernels over\ndeep features of image patches. This allows representation of an entire dataset\nof WSIs as a kernel matrix for WSI level clustering, weakly-supervised\nprediction of TP-53 mutation status in breast cancer patients from their\nroutine WSIs as well as survival analysis with state of the art prediction\nperformance. We believe that this work will open up further avenues for\napplication of WSI-level kernels for predictive and prognostic tasks in\ncomputational pathology.\n", "date": "2023-01-23", "categories": []}
{"id": "2301.07627", "title": "A novel dataset and a two-stage mitosis nuclei detection method based on\n  hybrid anchor branch", "abstract": "  Mitosis detection is one of the challenging problems in computational\npathology, and mitotic count is an important index of cancer grading for\npathologists. However, current counts of mitotic nuclei rely on pathologists\nlooking microscopically at the number of mitotic nuclei in hot spots, which is\nsubjective and time-consuming. In this paper, we propose a two-stage cascaded\nnetwork, named FoCasNet, for mitosis detection. In the first stage, a detection\nnetwork named M_det is proposed to detect as many mitoses as possible. In the\nsecond stage, a classification network M_class is proposed to refine the\nresults of the first stage. In addition, the attention mechanism, normalization\nmethod, and hybrid anchor branch classification subnet are introduced to\nimprove the overall detection performance. Our method achieves the current\nhighest F1-score of 0.888 on the public dataset ICPR 2012. We also evaluated\nour method on the GZMH dataset released by our research team for the first time\nand reached the highest F1-score of 0.563, which is also better than multiple\nclassic detection networks widely used at present. It confirmed the\neffectiveness and generalization of our method. The code will be available at:\nhttps://github.com/antifen/mitosis-nuclei-detection.\n", "date": "2023-01-18", "categories": []}
{"id": "2208.06445", "title": "CCRL: Contrastive Cell Representation Learning", "abstract": "  Cell identification within the H&E slides is an essential prerequisite that\ncan pave the way towards further pathology analyses including tissue\nclassification, cancer grading, and phenotype prediction. However, performing\nsuch a task using deep learning techniques requires a large cell-level\nannotated dataset. Although previous studies have investigated the performance\nof contrastive self-supervised methods in tissue classification, the utility of\nthis class of algorithms in cell identification and clustering is still\nunknown. In this work, we investigated the utility of Self-Supervised Learning\n(SSL) in cell clustering by proposing the Contrastive Cell Representation\nLearning (CCRL) model. Through comprehensive comparisons, we show that this\nmodel can outperform all currently available cell clustering models by a large\nmargin across two datasets from different tissue types. More interestingly, the\nresults show that our proposed model worked well with a few number of cell\ncategories while the utility of SSL models has been mainly shown in the context\nof natural image datasets with large numbers of classes (e.g., ImageNet). The\nunsupervised representation learning approach proposed in this research\neliminates the time-consuming step of data annotation in cell classification\ntasks, which enables us to train our model on a much larger dataset compared to\nprevious methods. Therefore, considering the promising outcome, this approach\ncan open a new avenue to automatic cell representation learning.\n", "date": "2022-08-12", "categories": []}
{"id": "2301.07030", "title": "Computational Pathology for Brain Disorders", "abstract": "  Non-invasive brain imaging techniques allow understanding the behavior and\nmacro changes in the brain to determine the progress of a disease. However,\ncomputational pathology provides a deeper understanding of brain disorders at\ncellular level, able to consolidate a diagnosis and make the bridge between the\nmedical image and the omics analysis. In traditional histopathology, histology\nslides are visually inspected, under the microscope, by trained pathologists.\nThis process is time-consuming and labor-intensive; therefore, the emergence of\nComputational Pathology has triggered great hope to ease this tedious task and\nmake it more robust. This chapter focuses on understanding the state-of-the-art\nmachine learning techniques used to analyze whole slide images within the\ncontext of brain disorders. We present a selective set of remarkable machine\nlearning algorithms providing discriminative approaches and quality results on\nbrain disorders. These methodologies are applied to different tasks, such as\nmonitoring mechanisms contributing to disease progression and patient survival\nrates, analyzing morphological phenotypes for classification and quantitative\nassessment of disease, improving clinical care, diagnosing tumor specimens, and\nintraoperative interpretation. Thanks to the recent progress in machine\nlearning algorithms for high-content image processing, computational pathology\nmarks the rise of a new generation of medical discoveries and clinical\nprotocols, including in brain disorders.\n", "date": "2023-01-13", "categories": []}
{"id": "2301.03418", "title": "Nuclear Segmentation and Classification: On Color & Compression\n  Generalization", "abstract": "  Since the introduction of digital and computational pathology as a field, one\nof the major problems in the clinical application of algorithms has been the\nstruggle to generalize well to examples outside the distribution of the\ntraining data. Existing work to address this in both pathology and natural\nimages has focused almost exclusively on classification tasks. We explore and\nevaluate the robustness of the 7 best performing nuclear segmentation and\nclassification models from the largest computational pathology challenge for\nthis problem to date, the CoNIC challenge. We demonstrate that existing\nstate-of-the-art (SoTA) models are robust towards compression artifacts but\nsuffer substantial performance reduction when subjected to shifts in the color\ndomain. We find that using stain normalization to address the domain shift\nproblem can be detrimental to the model performance. On the other hand, neural\nstyle transfer is more consistent in improving test performance when presented\nwith large color variations in the wild.\n", "date": "2023-01-09", "categories": []}
{"id": "2212.07721", "title": "Deep Learning-Based Automatic Assessment of AgNOR-scores in\n  Histopathology Images", "abstract": "  Nucleolar organizer regions (NORs) are parts of the DNA that are involved in\nRNA transcription. Due to the silver affinity of associated proteins,\nargyrophilic NORs (AgNORs) can be visualized using silver-based staining. The\naverage number of AgNORs per nucleus has been shown to be a prognostic factor\nfor predicting the outcome of many tumors. Since manual detection of AgNORs is\nlaborious, automation is of high interest. We present a deep learning-based\npipeline for automatically determining the AgNOR-score from histopathological\nsections. An additional annotation experiment was conducted with six\npathologists to provide an independent performance evaluation of our approach.\nAcross all raters and images, we found a mean squared error of 0.054 between\nthe AgNOR- scores of the experts and those of the model, indicating that our\napproach offers performance comparable to humans.\n", "date": "2022-12-15", "categories": []}
{"id": "2212.06834", "title": "Deep Neural Networks integrating genomics and histopathological images\n  for predicting stages and survival time-to-event in colon cancer", "abstract": "  There exists unexplained diverse variation within the predefined colon cancer\nstages using only features either from genomics or histopathological whole\nslide images as prognostic factors. Unraveling this variation will bring about\nimproved in staging and treatment outcome, hence motivated by the advancement\nof Deep Neural Network libraries and different structures and factors within\nsome genomic dataset, we aggregate atypical patterns in histopathological\nimages with diverse carcinogenic expression from mRNA, miRNA and DNA\nMethylation as an integrative input source into an ensemble deep neural network\nfor colon cancer stages classification and samples stratification into low or\nhigh risk survival groups. The results of our Ensemble Deep Convolutional\nNeural Network model show an improved performance in stages classification on\nthe integrated dataset. The fused input features return Area under curve\nReceiver Operating Characteristic curve (AUC ROC) of 0.95 compared with AUC ROC\nof 0.71 and 0.68 obtained when only genomics and images features are used for\nthe stage's classification, respectively. Also, the extracted features were\nused to split the patients into low or high risk survival groups. Among the\n2548 fused features, 1695 features showed a statistically significant survival\nprobability differences between the two risk groups defined by the extracted\nfeatures.\n", "date": "2022-12-13", "categories": []}
{"id": "2212.01737", "title": "RLogist: Fast Observation Strategy on Whole-slide Images with Deep\n  Reinforcement Learning", "abstract": "  Whole-slide images (WSI) in computational pathology have high resolution with\ngigapixel size, but are generally with sparse regions of interest, which leads\nto weak diagnostic relevance and data inefficiency for each area in the slide.\nMost of the existing methods rely on a multiple instance learning framework\nthat requires densely sampling local patches at high magnification. The\nlimitation is evident in the application stage as the heavy computation for\nextracting patch-level features is inevitable. In this paper, we develop\nRLogist, a benchmarking deep reinforcement learning (DRL) method for fast\nobservation strategy on WSIs. Imitating the diagnostic logic of human\npathologists, our RL agent learns how to find regions of observation value and\nobtain representative features across multiple resolution levels, without\nhaving to analyze each part of the WSI at the high magnification. We benchmark\nour method on two whole-slide level classification tasks, including detection\nof metastases in WSIs of lymph node sections, and subtyping of lung cancer.\nExperimental results demonstrate that RLogist achieves competitive\nclassification performance compared to typical multiple instance learning\nalgorithms, while having a significantly short observation path. In addition,\nthe observation path given by RLogist provides good decision-making\ninterpretability, and its ability of reading path navigation can potentially be\nused by pathologists for educational/assistive purposes. Our code is available\nat: \\url{https://github.com/tencent-ailab/RLogist}.\n", "date": "2022-12-04", "categories": []}
{"id": "2212.05900", "title": "Deep learning-based Subtyping of Atypical and Normal Mitoses using a\n  Hierarchical Anchor-Free Object Detector", "abstract": "  Mitotic activity is key for the assessment of malignancy in many tumors.\nMoreover, it has been demonstrated that the proportion of abnormal mitosis to\nnormal mitosis is of prognostic significance. Atypical mitotic figures (MF) can\nbe identified morphologically as having segregation abnormalities of the\nchromatids. In this work, we perform, for the first time, automatic subtyping\nof mitotic figures into normal and atypical categories according to\ncharacteristic morphological appearances of the different phases of mitosis.\nUsing the publicly available MIDOG21 and TUPAC16 breast cancer mitosis\ndatasets, two experts blindly subtyped mitotic figures into five morphological\ncategories. Further, we set up a state-of-the-art object detection pipeline\nextending the anchor-free FCOS approach with a gated hierarchical\nsubclassification branch. Our labeling experiment indicated that subtyping of\nmitotic figures is a challenging task and prone to inter-rater disagreement,\nwhich we found in 24.89% of MF. Using the more diverse MIDOG21 dataset for\ntraining and TUPAC16 for testing, we reached a mean overall average precision\nscore of 0.552, a ROC AUC score of 0.833 for atypical/normal MF and a mean\nclass-averaged ROC-AUC score of 0.977 for discriminating the different phases\nof cells undergoing mitosis.\n", "date": "2022-12-12", "categories": []}
{"id": "2212.03273", "title": "Giga-SSL: Self-Supervised Learning for Gigapixel Images", "abstract": "  Whole slide images (WSI) are microscopy images of stained tissue slides\nroutinely prepared for diagnosis and treatment selection in medical practice.\nWSI are very large (gigapixel size) and complex (made of up to millions of\ncells). The current state-of-the-art (SoTA) approach to classify WSI subdivides\nthem into tiles, encodes them by pre-trained networks and applies Multiple\nInstance Learning (MIL) to train for specific downstream tasks. However,\nannotated datasets are often small, typically a few hundred to a few thousand\nWSI, which may cause overfitting and underperforming models. Conversely, the\nnumber of unannotated WSI is ever increasing, with datasets of tens of\nthousands (soon to be millions) of images available. While it has been\npreviously proposed to use these unannotated data to identify suitable tile\nrepresentations by self-supervised learning (SSL), downstream classification\ntasks still require full supervision because parts of the MIL architecture is\nnot trained during tile level SSL pre-training. Here, we propose a strategy of\nslide level SSL to leverage the large number of WSI without annotations to\ninfer powerful slide representations. Applying our method to The Cancer-Genome\nAtlas, one of the most widely used data resources in cancer research (16 TB\nimage data), we are able to downsize the dataset to 23 MB without any loss in\npredictive power: we show that a linear classifier trained on top of these\nembeddings maintains or improves previous SoTA performances on various\nbenchmark WSI classification tasks. Finally, we observe that training a\nclassifier on these representations with tiny datasets (e.g. 50 slides)\nimproved performances over SoTA by an average of +6.3 AUC points over all\ndownstream tasks.\n", "date": "2022-12-06", "categories": []}
{"id": "2212.00573", "title": "A perspective on the use of health digital twins in computational\n  pathology", "abstract": "  A digital health twin can be defined as a virtual model of a physical person,\nin this specific case, a patient. This virtual model is constituted by\nmultidimensional data that can host from clinical, molecular and therapeutic\nparameters to sensor data and living conditions. Given that in computational\npathology, it is very important to have the information from image donors to\ncreate computational models, the integration of digital twins in this field\ncould be crucial. However, since these virtual entities collect sensitive data\nfrom physical people, privacy safeguards must also be considered and\nimplemented. With these data safeguards in place, health digital twins could\nintegrate digital clinical trials and be necessary participants in the\ngeneration of real-world evidence, which could positively change both fields.\n", "date": "2022-11-30", "categories": []}
{"id": "2211.16632", "title": "Hierarchical Transformer for Survival Prediction Using Multimodality\n  Whole Slide Images and Genomics", "abstract": "  Learning good representation of giga-pixel level whole slide pathology images\n(WSI) for downstream tasks is critical. Previous studies employ multiple\ninstance learning (MIL) to represent WSIs as bags of sampled patches because,\nfor most occasions, only slide-level labels are available, and only a tiny\nregion of the WSI is disease-positive area. However, WSI representation\nlearning still remains an open problem due to: (1) patch sampling on a higher\nresolution may be incapable of depicting microenvironment information such as\nthe relative position between the tumor cells and surrounding tissues, while\npatches at lower resolution lose the fine-grained detail; (2) extracting\npatches from giant WSI results in large bag size, which tremendously increases\nthe computational cost. To solve the problems, this paper proposes a\nhierarchical-based multimodal transformer framework that learns a hierarchical\nmapping between pathology images and corresponding genes. Precisely, we\nrandomly extract instant-level patch features from WSIs with different\nmagnification. Then a co-attention mapping between imaging and genomics is\nlearned to uncover the pairwise interaction and reduce the space complexity of\nimaging features. Such early fusion makes it computationally feasible to use\nMIL Transformer for the survival prediction task. Our architecture requires\nfewer GPU resources compared with benchmark methods while maintaining better\nWSI representation ability. We evaluate our approach on five cancer types from\nthe Cancer Genome Atlas database and achieved an average c-index of $0.673$,\noutperforming the state-of-the-art multimodality methods.\n", "date": "2022-11-29", "categories": []}
{"id": "2211.15667", "title": "Artificial Intelligence-based Eosinophil Counting in Gastrointestinal\n  Biopsies", "abstract": "  Normally eosinophils are present in the gastrointestinal (GI) tract of\nhealthy individuals. When the eosinophils increase beyond their usual amount in\nthe GI tract, a patient gets varied symptoms. Clinicians find it difficult to\ndiagnose this condition called eosinophilia. Early diagnosis can help in\ntreating patients. Histopathology is the gold standard in the diagnosis for\nthis condition. As this is an under-diagnosed condition, counting eosinophils\nin the GI tract biopsies is important. In this study, we trained and tested a\ndeep neural network based on UNet to detect and count eosinophils in GI tract\nbiopsies. We used connected component analysis to extract the eosinophils. We\nstudied correlation of eosinophilic infiltration counted by AI with a manual\ncount. GI tract biopsy slides were stained with H&E stain. Slides were scanned\nusing a camera attached to a microscope and five high-power field images were\ntaken per slide. Pearson correlation coefficient was 85% between the\nmachine-detected and manual eosinophil counts on 300 held-out (test) images.\n", "date": "2022-11-25", "categories": []}
{"id": "2211.14847", "title": "Deep Learning-Based Prediction of Molecular Tumor Biomarkers from H&E: A\n  Practical Review", "abstract": "  Molecular and genomic properties are critical in selecting cancer treatments\nto target individual tumors, particularly for immunotherapy. However, the\nmethods to assess such properties are expensive, time-consuming, and often not\nroutinely performed. Applying machine learning to H&E images can provide a more\ncost-effective screening method. Dozens of studies over the last few years have\ndemonstrated that a variety of molecular biomarkers can be predicted from H&E\nalone using the advancements of deep learning: molecular alterations, genomic\nsubtypes, protein biomarkers, and even the presence of viruses. This article\nreviews the diverse applications across cancer types and the methodology to\ntrain and validate these models on whole slide images. From bottom-up to\npathologist-driven to hybrid approaches, the leading trends include a variety\nof weakly supervised deep learning-based approaches, as well as mechanisms for\ntraining strongly supervised models in select situations. While results of\nthese algorithms look promising, some challenges still persist, including small\ntraining sets, rigorous validation, and model explainability. Biomarker\nprediction models may yield a screening method to determine when to run\nmolecular tests or an alternative when molecular tests are not possible. They\nalso create new opportunities in quantifying intratumoral heterogeneity and\npredicting patient outcomes.\n", "date": "2022-11-27", "categories": []}
{"id": "2211.13621", "title": "ACROBAT -- a multi-stain breast cancer histological whole-slide-image\n  data set from routine diagnostics for computational pathology", "abstract": "  The analysis of FFPE tissue sections stained with haematoxylin and eosin\n(H&E) or immunohistochemistry (IHC) is an essential part of the pathologic\nassessment of surgically resected breast cancer specimens. IHC staining has\nbeen broadly adopted into diagnostic guidelines and routine workflows to\nmanually assess status and scoring of several established biomarkers, including\nER, PGR, HER2 and KI67. However, this is a task that can also be facilitated by\ncomputational pathology image analysis methods. The research in computational\npathology has recently made numerous substantial advances, often based on\npublicly available whole slide image (WSI) data sets. However, the field is\nstill considerably limited by the sparsity of public data sets. In particular,\nthere are no large, high quality publicly available data sets with WSIs of\nmatching IHC and H&E-stained tissue sections. Here, we publish the currently\nlargest publicly available data set of WSIs of tissue sections from surgical\nresection specimens from female primary breast cancer patients with matched\nWSIs of corresponding H&E and IHC-stained tissue, consisting of 4,212 WSIs from\n1,153 patients. The primary purpose of the data set was to facilitate the\nACROBAT WSI registration challenge, aiming at accurately aligning H&E and IHC\nimages. For research in the area of image registration, automatic quantitative\nfeedback on registration algorithm performance remains available through the\nACROBAT challenge website, based on more than 37,000 manually annotated\nlandmark pairs from 13 annotators. Beyond registration, this data set has the\npotential to enable many different avenues of computational pathology research,\nincluding stain-guided learning, virtual staining, unsupervised pre-training,\nartefact detection and stain-independent models.\n", "date": "2022-11-24", "categories": []}
{"id": "2206.08885", "title": "Incorporating intratumoral heterogeneity into weakly-supervised deep\n  learning models via variance pooling", "abstract": "  Supervised learning tasks such as cancer survival prediction from gigapixel\nwhole slide images (WSIs) are a critical challenge in computational pathology\nthat requires modeling complex features of the tumor microenvironment. These\nlearning tasks are often solved with deep multi-instance learning (MIL) models\nthat do not explicitly capture intratumoral heterogeneity. We develop a novel\nvariance pooling architecture that enables a MIL model to incorporate\nintratumoral heterogeneity into its predictions. Two interpretability tools\nbased on representative patches are illustrated to probe the biological signals\ncaptured by these models. An empirical study with 4,479 gigapixel WSIs from the\nCancer Genome Atlas shows that adding variance pooling onto MIL frameworks\nimproves survival prediction performance for five cancer types.\n", "date": "2022-06-17", "categories": []}
{"id": "2211.07692", "title": "Self-training of Machine Learning Models for Liver Histopathology:\n  Generalization under Clinical Shifts", "abstract": "  Histopathology images are gigapixel-sized and include features and\ninformation at different resolutions. Collecting annotations in histopathology\nrequires highly specialized pathologists, making it expensive and\ntime-consuming. Self-training can alleviate annotation constraints by learning\nfrom both labeled and unlabeled data, reducing the amount of annotations\nrequired from pathologists. We study the design of teacher-student\nself-training systems for Non-alcoholic Steatohepatitis (NASH) using clinical\nhistopathology datasets with limited annotations. We evaluate the models on\nin-distribution and out-of-distribution test data under clinical data shifts.\nWe demonstrate that through self-training, the best student model statistically\noutperforms the teacher with a $3\\%$ absolute difference on the macro F1 score.\nThe best student model also approaches the performance of a fully supervised\nmodel trained with twice as many annotations.\n", "date": "2022-11-14", "categories": []}
{"id": "2211.04649", "title": "Gold-standard of HER2 breast cancer biopsies using supervised learning\n  based on multiple pathologist annotations", "abstract": "  Breast cancer is one of the most common cancer in women around the world. For\ndiagnosis, pathologists evaluate biomarkers such as HER2 protein using\nimmunohistochemistry over tissue extracted by a biopsy. Through microscopic\ninspection, this assessment estimates the intensity and integrity of the\nmembrane cells' staining and scores the sample as 0, 1+, 2+, or 3+: a\nsubjective decision that depends on the interpretation of the pathologist. This\npaper presents the preliminary data analysis of the annotations of three\npathologists over the same set of samples obtained using 20x magnification and\nincluding $1,252$ non-overlapping biopsy patches. We evaluate the intra- and\ninter-expert variability achieving substantial and moderate agreement,\nrespectively, according to Fleiss' Kappa coefficient, as a previous stage\ntowards a generation of a HER2 breast cancer biopsy gold-standard using\nsupervised learning from multiple pathologist annotations.\n", "date": "2022-11-09", "categories": []}
{"id": "2206.08439", "title": "OpenSRH: optimizing brain tumor surgery using intraoperative stimulated\n  Raman histology", "abstract": "  Accurate intraoperative diagnosis is essential for providing safe and\neffective care during brain tumor surgery. Our standard-of-care diagnostic\nmethods are time, resource, and labor intensive, which restricts access to\noptimal surgical treatments. To address these limitations, we propose an\nalternative workflow that combines stimulated Raman histology (SRH), a rapid\noptical imaging method, with deep learning-based automated interpretation of\nSRH images for intraoperative brain tumor diagnosis and real-time surgical\ndecision support. Here, we present OpenSRH, the first public dataset of\nclinical SRH images from 300+ brain tumors patients and 1300+ unique whole\nslide optical images. OpenSRH contains data from the most common brain tumors\ndiagnoses, full pathologic annotations, whole slide tumor segmentations, raw\nand processed optical imaging data for end-to-end model development and\nvalidation. We provide a framework for patch-based whole slide SRH\nclassification and inference using weak (i.e. patient-level) diagnostic labels.\nFinally, we benchmark two computer vision tasks: multiclass histologic brain\ntumor classification and patch-based contrastive representation learning. We\nhope OpenSRH will facilitate the clinical translation of rapid optical imaging\nand real-time ML-based surgical decision support in order to improve the\naccess, safety, and efficacy of cancer surgery in the era of precision\nmedicine. Dataset access, code, and benchmarks are available at\nopensrh.mlins.org.\n", "date": "2022-06-16", "categories": []}
{"id": "2211.00749", "title": "ViT-DeiT: An Ensemble Model for Breast Cancer Histopathological Images\n  Classification", "abstract": "  Breast cancer is the most common cancer in the world and the second most\ncommon type of cancer that causes death in women. The timely and accurate\ndiagnosis of breast cancer using histopathological images is crucial for\npatient care and treatment. Pathologists can make more accurate diagnoses with\nthe help of a novel approach based on image processing. This approach is an\nensemble model of two types of pre-trained vision transformer models, namely,\nVision Transformer and Data-Efficient Image Transformer. The proposed ensemble\nmodel classifies breast cancer histopathology images into eight classes, four\nof which are categorized as benign, whereas the others are categorized as\nmalignant. A public dataset was used to evaluate the proposed model. The\nexperimental results showed 98.17% accuracy, 98.18% precision, 98.08% recall,\nand a 98.12% F1 score.\n", "date": "2022-11-01", "categories": []}
{"id": "2211.01256", "title": "An Aggregation of Aggregation Methods in Computational Pathology", "abstract": "  Image analysis and machine learning algorithms operating on multi-gigapixel\nwhole-slide images (WSIs) often process a large number of tiles (sub-images)\nand require aggregating predictions from the tiles in order to predict\nWSI-level labels. In this paper, we present a review of existing literature on\nvarious types of aggregation methods with a view to help guide future research\nin the area of computational pathology (CPath). We propose a general CPath\nworkflow with three pathways that consider multiple levels and types of data\nand the nature of computation to analyse WSIs for predictive modelling. We\ncategorize aggregation methods according to the context and representation of\nthe data, features of computational modules and CPath use cases. We compare and\ncontrast different methods based on the principle of multiple instance\nlearning, perhaps the most commonly used aggregation method, covering a wide\nrange of CPath literature. To provide a fair comparison, we consider a specific\nWSI-level prediction task and compare various aggregation methods for that\ntask. Finally, we conclude with a list of objectives and desirable attributes\nof aggregation methods in general, pros and cons of the various approaches,\nsome recommendations and possible future directions.\n", "date": "2022-11-02", "categories": []}
{"id": "2208.12437", "title": "Detecting Mitoses with a Convolutional Neural Network for MIDOG 2022\n  Challenge", "abstract": "  This work presents a mitosis detection method with only one vanilla\nConvolutional Neural Network (CNN). Our method consists of two steps: given an\nimage, we first apply a CNN using a sliding window technique to extract patches\nthat have mitoses; we then calculate each extracted patch's class activation\nmap to obtain the mitosis's precise location. To increase the model performance\non high-domain-variance pathology images, we train the CNN with a data\naugmentation pipeline, a noise-tolerant loss that copes with unlabeled images,\nand a multi-rounded active learning strategy. In the MIDOG 2022 challenge, our\napproach, with an EfficientNet-b3 CNN model, achieved an overall F1 score of\n0.7323 in the preliminary test phase, and 0.6847 in the final test phase (task\n1). Our approach sheds light on the broader applicability of class activation\nmaps for object detections in pathology images.\n", "date": "2022-08-26", "categories": []}
{"id": "2210.16457", "title": "Region of Interest Detection in Melanocytic Skin Tumor Whole Slide\n  Images", "abstract": "  Automated region of interest detection in histopathological image analysis is\na challenging and important topic with tremendous potential impact on clinical\npractice. The deep-learning methods used in computational pathology help us to\nreduce costs and increase the speed and accuracy of regions of interest\ndetection and cancer diagnosis. In this work, we propose a patch-based region\nof interest detection method for melanocytic skin tumor whole-slide images. We\nwork with a dataset that contains 165 primary melanomas and nevi Hematoxylin\nand Eosin whole-slide images and build a deep-learning method. The proposed\nmethod performs well on a hold-out test data set including five TCGA-SKCM\nslides (accuracy of 93.94\\% in slide classification task and intersection over\nunion rate of 41.27\\% in the region of interest detection task), showing the\noutstanding performance of our model on melanocytic skin tumor. Even though we\ntest the experiments on the skin tumor dataset, our work could also be extended\nto other medical image detection problems, such as various tumors'\nclassification and prediction, to help and benefit the clinical evaluation and\ndiagnosis of different tumors.\n", "date": "2022-10-29", "categories": []}
{"id": "2210.16728", "title": "ISG: I can See Your Gene Expression", "abstract": "  This paper aims to predict gene expression from a histology slide image\nprecisely. Such a slide image has a large resolution and sparsely distributed\ntextures. These obstruct extracting and interpreting discriminative features\nfrom the slide image for diverse gene types prediction. Existing gene\nexpression methods mainly use general components to filter textureless regions,\nextract features, and aggregate features uniformly across regions. However,\nthey ignore gaps and interactions between different image regions and are\ntherefore inferior in the gene expression task. Instead, we present ISG\nframework that harnesses interactions among discriminative features from\ntexture-abundant regions by three new modules: 1) a Shannon Selection module,\nbased on the Shannon information content and Solomonoff's theory, to filter out\ntextureless image regions; 2) a Feature Extraction network to extract\nexpressive low-dimensional feature representations for efficient region\ninteractions among a high-resolution image; 3) a Dual Attention network attends\nto regions with desired gene expression features and aggregates them for the\nprediction task. Extensive experiments on standard benchmark datasets show that\nthe proposed ISG framework outperforms state-of-the-art methods significantly.\n", "date": "2022-10-30", "categories": []}
{"id": "2207.02946", "title": "Virtual staining of defocused autofluorescence images of unlabeled\n  tissue using deep neural networks", "abstract": "  Deep learning-based virtual staining was developed to introduce image\ncontrast to label-free tissue sections, digitally matching the histological\nstaining, which is time-consuming, labor-intensive, and destructive to tissue.\nStandard virtual staining requires high autofocusing precision during the whole\nslide imaging of label-free tissue, which consumes a significant portion of the\ntotal imaging time and can lead to tissue photodamage. Here, we introduce a\nfast virtual staining framework that can stain defocused autofluorescence\nimages of unlabeled tissue, achieving equivalent performance to virtual\nstaining of in-focus label-free images, also saving significant imaging time by\nlowering the microscope's autofocusing precision. This framework incorporates a\nvirtual-autofocusing neural network to digitally refocus the defocused images\nand then transforms the refocused images into virtually stained images using a\nsuccessive network. These cascaded networks form a collaborative inference\nscheme: the virtual staining model regularizes the virtual-autofocusing network\nthrough a style loss during the training. To demonstrate the efficacy of this\nframework, we trained and blindly tested these networks using human lung\ntissue. Using 4x fewer focus points with 2x lower focusing precision, we\nsuccessfully transformed the coarsely-focused autofluorescence images into\nhigh-quality virtually stained H&E images, matching the standard virtual\nstaining framework that used finely-focused autofluorescence input images.\nWithout sacrificing the staining quality, this framework decreases the total\nimage acquisition time needed for virtual staining of a label-free whole-slide\nimage (WSI) by ~32%, together with a ~89% decrease in the autofocusing time,\nand has the potential to eliminate the laborious and costly histochemical\nstaining process in pathology.\n", "date": "2022-07-06", "categories": []}
{"id": "2209.00585", "title": "Adversarial Stain Transfer to Study the Effect of Color Variation on\n  Cell Instance Segmentation", "abstract": "  Stain color variation in histological images, caused by a variety of factors,\nis a challenge not only for the visual diagnosis of pathologists but also for\ncell segmentation algorithms. To eliminate the color variation, many stain\nnormalization approaches have been proposed. However, most were designed for\nhematoxylin and eosin staining images and performed poorly on\nimmunohistochemical staining images. Current cell segmentation methods\nsystematically apply stain normalization as a preprocessing step, but the\nimpact brought by color variation has not been quantitatively investigated yet.\nIn this paper, we produced five groups of NeuN staining images with different\ncolors. We applied a deep learning image-recoloring method to perform color\ntransfer between histological image groups. Finally, we altered the color of a\nsegmentation set and quantified the impact of color variation on cell\nsegmentation. The results demonstrated the necessity of color normalization\nprior to subsequent analysis.\n", "date": "2022-09-01", "categories": []}
{"id": "2210.12938", "title": "GradMix for nuclei segmentation and classification in imbalanced\n  pathology image datasets", "abstract": "  An automated segmentation and classification of nuclei is an essential task\nin digital pathology. The current deep learning-based approaches require a vast\namount of annotated datasets by pathologists. However, the existing datasets\nare imbalanced among different types of nuclei in general, leading to a\nsubstantial performance degradation. In this paper, we propose a simple but\neffective data augmentation technique, termed GradMix, that is specifically\ndesigned for nuclei segmentation and classification. GradMix takes a pair of a\nmajor-class nucleus and a rare-class nucleus, creates a customized mixing mask,\nand combines them using the mask to generate a new rare-class nucleus. As it\ncombines two nuclei, GradMix considers both nuclei and the neighboring\nenvironment by using the customized mixing mask. This allows us to generate\nrealistic rare-class nuclei with varying environments. We employed two datasets\nto evaluate the effectiveness of GradMix. The experimental results suggest that\nGradMix is able to improve the performance of nuclei segmentation and\nclassification in imbalanced pathology image datasets.\n", "date": "2022-10-24", "categories": []}
{"id": "2210.10162", "title": "Computational pathology in renal disease: a comprehensive perspective", "abstract": "  Computational pathology is a field that has complemented various\nsubspecialties of diagnostic pathology over the last few years. In this article\na brief analyzis the different applications in nephrology is developed. To\nbegin, an overview of the different forms of image production is provided. To\ncontinue, the most frequent applications of computer vision models, the salient\nfeatures of the different clinical applications, and the data protection\nconsiderations encountered are described. To finish the development, I delve\ninto the interpretability of these applications, expanding in depth on the\nthree dimensions of this area.\n", "date": "2022-10-18", "categories": []}
{"id": "2210.09909", "title": "Uncertainty estimation for out-of-distribution detection in\n  computational histopathology", "abstract": "  In computational histopathology algorithms now outperform humans on a range\nof tasks, but to date none are employed for automated diagnoses in the clinic.\nBefore algorithms can be involved in such high-stakes decisions they need to\n\"know when they don't know\", i.e., they need to estimate their predictive\nuncertainty. This allows them to defer potentially erroneous predictions to a\nhuman pathologist, thus increasing their safety. Here, we evaluate the\npredictive performance and calibration of several uncertainty estimation\nmethods on clinical histopathology data. We show that a distance-aware\nuncertainty estimation method outperforms commonly used approaches, such as\nMonte Carlo dropout and deep ensembles. However, we observe a drop in\npredictive performance and calibration on novel samples across all uncertainty\nestimation methods tested. We also investigate the use of uncertainty\nthresholding to reject out-of-distribution samples for selective prediction. We\ndemonstrate the limitations of this approach and suggest areas for future\nresearch.\n", "date": "2022-10-18", "categories": []}
{"id": "2206.01794", "title": "Additive MIL: Intrinsically Interpretable Multiple Instance Learning for\n  Pathology", "abstract": "  Multiple Instance Learning (MIL) has been widely applied in pathology towards\nsolving critical problems such as automating cancer diagnosis and grading,\npredicting patient prognosis, and therapy response. Deploying these models in a\nclinical setting requires careful inspection of these black boxes during\ndevelopment and deployment to identify failures and maintain physician trust.\nIn this work, we propose a simple formulation of MIL models, which enables\ninterpretability while maintaining similar predictive performance. Our Additive\nMIL models enable spatial credit assignment such that the contribution of each\nregion in the image can be exactly computed and visualized. We show that our\nspatial credit assignment coincides with regions used by pathologists during\ndiagnosis and improves upon classical attention heatmaps from attention MIL\nmodels. We show that any existing MIL model can be made additive with a simple\nchange in function composition. We also show how these models can debug model\nfailures, identify spurious features, and highlight class-wise regions of\ninterest, enabling their use in high-stakes environments such as clinical\ndecision-making.\n", "date": "2022-06-03", "categories": []}
{"id": "2210.05175", "title": "Variability Matters : Evaluating inter-rater variability in\n  histopathology for robust cell detection", "abstract": "  Large annotated datasets have been a key component in the success of deep\nlearning. However, annotating medical images is challenging as it requires\nexpertise and a large budget. In particular, annotating different types of\ncells in histopathology suffer from high inter- and intra-rater variability due\nto the ambiguity of the task. Under this setting, the relation between\nannotators' variability and model performance has received little attention. We\npresent a large-scale study on the variability of cell annotations among 120\nboard-certified pathologists and how it affects the performance of a deep\nlearning model. We propose a method to measure such variability, and by\nexcluding those annotators with low variability, we verify the trade-off\nbetween the amount of data and its quality. We found that naively increasing\nthe data size at the expense of inter-rater variability does not necessarily\nlead to better-performing models in cell detection. Instead, decreasing the\ninter-rater variability with the expense of decreasing dataset size increased\nthe model performance. Furthermore, models trained from data annotated with\nlower inter-labeler variability outperform those from higher inter-labeler\nvariability. These findings suggest that the evaluation of the annotators may\nhelp tackle the fundamental budget issues in the histopathology domain\n", "date": "2022-10-11", "categories": []}
{"id": "2210.05258", "title": "EOCSA: Predicting Prognosis of Epithelial Ovarian Cancer with Whole\n  Slide Histopathological Images", "abstract": "  Ovarian cancer is one of the most serious cancers that threaten women around\nthe world. Epithelial ovarian cancer (EOC), as the most commonly seen subtype\nof ovarian cancer, has rather high mortality rate and poor prognosis among\nvarious gynecological cancers. Survival analysis outcome is able to provide\ntreatment advices to doctors. In recent years, with the development of medical\nimaging technology, survival prediction approaches based on pathological images\nhave been proposed. In this study, we designed a deep framework named EOCSA\nwhich analyzes the prognosis of EOC patients based on pathological whole slide\nimages (WSIs). Specifically, we first randomly extracted patches from WSIs and\ngrouped them into multiple clusters. Next, we developed a survival prediction\nmodel, named DeepConvAttentionSurv (DCAS), which was able to extract\npatch-level features, removed less discriminative clusters and predicted the\nEOC survival precisely. Particularly, channel attention, spatial attention, and\nneuron attention mechanisms were used to improve the performance of feature\nextraction. Then patient-level features were generated from our weight\ncalculation method and the survival time was finally estimated using LASSO-Cox\nmodel. The proposed EOCSA is efficient and effective in predicting prognosis of\nEOC and the DCAS ensures more informative and discriminative features can be\nextracted. As far as we know, our work is the first to analyze the survival of\nEOC based on WSIs and deep neural network technologies. The experimental\nresults demonstrate that our proposed framework has achieved state-of-the-art\nperformance of 0.980 C-index. The implementation of the approach can be found\nat https://github.com/RanSuLab/EOCprognosis.\n", "date": "2022-10-11", "categories": []}
{"id": "2209.13167", "title": "A Morphology Focused Diffusion Probabilistic Model for Synthesis of\n  Histopathology Images", "abstract": "  Visual microscopic study of diseased tissue by pathologists has been the\ncornerstone for cancer diagnosis and prognostication for more than a century.\nRecently, deep learning methods have made significant advances in the analysis\nand classification of tissue images. However, there has been limited work on\nthe utility of such models in generating histopathology images. These synthetic\nimages have several applications in pathology including utilities in education,\nproficiency testing, privacy, and data sharing. Recently, diffusion\nprobabilistic models were introduced to generate high quality images. Here, for\nthe first time, we investigate the potential use of such models along with\nprioritized morphology weighting and color normalization to synthesize high\nquality histopathology images of brain cancer. Our detailed results show that\ndiffusion probabilistic models are capable of synthesizing a wide range of\nhistopathology images and have superior performance compared to generative\nadversarial networks.\n", "date": "2022-09-27", "categories": []}
{"id": "2207.08267", "title": "Gigapixel Whole-Slide Images Classification using Locally Supervised\n  Learning", "abstract": "  Histopathology whole slide images (WSIs) play a very important role in\nclinical studies and serve as the gold standard for many cancer diagnoses.\nHowever, generating automatic tools for processing WSIs is challenging due to\ntheir enormous sizes. Currently, to deal with this issue, conventional methods\nrely on a multiple instance learning (MIL) strategy to process a WSI at patch\nlevel. Although effective, such methods are computationally expensive, because\ntiling a WSI into patches takes time and does not explore the spatial relations\nbetween these tiles. To tackle these limitations, we propose a locally\nsupervised learning framework which processes the entire slide by exploring the\nentire local and global information that it contains. This framework divides a\npre-trained network into several modules and optimizes each module locally\nusing an auxiliary model. We also introduce a random feature reconstruction\nunit (RFR) to preserve distinguishing features during training and improve the\nperformance of our method by 1% to 3%. Extensive experiments on three publicly\navailable WSI datasets: TCGA-NSCLC, TCGA-RCC and LKS, highlight the superiority\nof our method on different classification tasks. Our method outperforms the\nstate-of-the-art MIL methods by 2% to 5% in accuracy, while being 7 to 10 times\nfaster. Additionally, when dividing it into eight modules, our method requires\nas little as 20% of the total gpu memory required by end-to-end training. Our\ncode is available at https://github.com/cvlab-stonybrook/local_learning_wsi.\n", "date": "2022-07-17", "categories": []}
{"id": "2209.13155", "title": "Ki-67 Index Measurement in Breast Cancer Using Digital Image Analysis", "abstract": "  Ki-67 is a nuclear protein that can be produced during cell proliferation.\nThe Ki67 index is a valuable prognostic variable in several kinds of cancer. In\nbreast cancer, the index is even routinely checked in many patients. Currently,\npathologists use the immunohistochemistry method to calculate the percentage of\nKi-67 positive malignant cells as Ki-67 index. The higher score usually means\nmore aggressive tumor behavior. In clinical practice, the measurement of Ki-67\nindex relies on visual identifying method and manual counting. However, visual\nand manual assessment method is timeconsuming and leads to poor reproducibility\nbecause of different scoring standards or limited tumor area under assessment.\nHere, we use digital image processing technics including image binarization and\nimage morphological operations to create a digital image analysis method to\ninterpretate Ki-67 index. Then, 10 breast cancer specimens are used as\nvalidation with high accuracy (correlation efficiency r = 0.95127). With the\nassistance of digital image analysis, pathologists can interpretate the Ki67\nindex more efficiently, precisely with excellent reproducibility.\n", "date": "2022-09-27", "categories": []}
{"id": "2209.13408", "title": "A Pathologist-Informed Workflow for Classification of Prostate Glands in\n  Histopathology", "abstract": "  Pathologists diagnose and grade prostate cancer by examining tissue from\nneedle biopsies on glass slides. The cancer's severity and risk of metastasis\nare determined by the Gleason grade, a score based on the organization and\nmorphology of prostate cancer glands. For diagnostic work-up, pathologists\nfirst locate glands in the whole biopsy core, and -- if they detect cancer --\nthey assign a Gleason grade. This time-consuming process is subject to errors\nand significant inter-observer variability, despite strict diagnostic criteria.\nThis paper proposes an automated workflow that follows pathologists'\n\\textit{modus operandi}, isolating and classifying multi-scale patches of\nindividual glands in whole slide images (WSI) of biopsy tissues using distinct\nsteps: (1) two fully convolutional networks segment epithelium versus stroma\nand gland boundaries, respectively; (2) a classifier network separates benign\nfrom cancer glands at high magnification; and (3) an additional classifier\npredicts the grade of each cancer gland at low magnification. Altogether, this\nprocess provides a gland-specific approach for prostate cancer grading that we\ncompare against other machine-learning-based grading methods.\n", "date": "2022-09-27", "categories": []}
{"id": "2209.08770", "title": "S$^3$R: Self-supervised Spectral Regression for Hyperspectral\n  Histopathology Image Classification", "abstract": "  Benefited from the rich and detailed spectral information in hyperspectral\nimages (HSI), HSI offers great potential for a wide variety of medical\napplications such as computational pathology. But, the lack of adequate\nannotated data and the high spatiospectral dimensions of HSIs usually make\nclassification networks prone to overfit. Thus, learning a general\nrepresentation which can be transferred to the downstream tasks is imperative.\nTo our knowledge, no appropriate self-supervised pre-training method has been\ndesigned for histopathology HSIs. In this paper, we introduce an efficient and\neffective Self-supervised Spectral Regression (S$^3$R) method, which exploits\nthe low rank characteristic in the spectral domain of HSI. More concretely, we\npropose to learn a set of linear coefficients that can be used to represent one\nband by the remaining bands via masking out these bands. Then, the band is\nrestored by using the learned coefficients to reweight the remaining bands. Two\npre-text tasks are designed: (1)S$^3$R-CR, which regresses the linear\ncoefficients, so that the pre-trained model understands the inherent structures\nof HSIs and the pathological characteristics of different morphologies;\n(2)S$^3$R-BR, which regresses the missing band, making the model to learn the\nholistic semantics of HSIs. Compared to prior arts i.e., contrastive learning\nmethods, which focuses on natural images, S$^3$R converges at least 3 times\nfaster, and achieves significant improvements up to 14% in accuracy when\ntransferring to HSI classification tasks.\n", "date": "2022-09-19", "categories": []}
{"id": "2209.09193", "title": "Improving Mitosis Detection Via UNet-based Adversarial Domain\n  Homogenizer", "abstract": "  The effective localization of mitosis is a critical precursory task for\ndeciding tumor prognosis and grade. Automated mitosis detection through deep\nlearning-oriented image analysis often fails on unseen patient data due to\ninherent domain biases. This paper proposes a domain homogenizer for mitosis\ndetection that attempts to alleviate domain differences in histology images via\nadversarial reconstruction of input images. The proposed homogenizer is based\non a U-Net architecture and can effectively reduce domain differences commonly\nseen with histology imaging data. We demonstrate our domain homogenizer's\neffectiveness by observing the reduction in domain differences between the\npreprocessed images. Using this homogenizer, along with a subsequent retina-net\nobject detector, we were able to outperform the baselines of the 2021 MIDOG\nchallenge in terms of average precision of the detected mitotic figures.\n", "date": "2022-09-15", "categories": []}
{"id": "2209.03041", "title": "Multi-Scale Attention-based Multiple Instance Learning for\n  Classification of Multi-Gigapixel Histology Images", "abstract": "  Histology images with multi-gigapixel of resolution yield rich information\nfor cancer diagnosis and prognosis. Most of the time, only slide-level label is\navailable because pixel-wise annotation is labour intensive task. In this\npaper, we propose a deep learning pipeline for classification in histology\nimages. Using multiple instance learning, we attempt to predict the latent\nmembrane protein 1 (LMP1) status of nasopharyngeal carcinoma (NPC) based on\nhaematoxylin and eosin-stain (H&E) histology images. We utilised attention\nmechanism with residual connection for our aggregation layers. In our 3-fold\ncross-validation experiment, we achieved average accuracy, AUC and F1-score\n0.936, 0.995 and 0.862, respectively. This method also allows us to examine the\nmodel interpretability by visualising attention scores. To the best of our\nknowledge, this is the first attempt to predict LMP1 status on NPC using deep\nlearning.\n", "date": "2022-09-07", "categories": []}
{"id": "2208.03526", "title": "Multiplex-detection Based Multiple Instance Learning Network for Whole\n  Slide Image Classification", "abstract": "  Multiple instance learning (MIL) is a powerful approach to classify whole\nslide images (WSIs) for diagnostic pathology. A fundamental challenge of MIL on\nWSI classification is to discover the \\textit{critical instances} that trigger\nthe bag label. However, previous methods are primarily designed under the\nindependent and identical distribution hypothesis (\\textit{i.i.d}), ignoring\neither the correlations between instances or heterogeneity of tumours. In this\npaper, we propose a novel multiplex-detection-based multiple instance learning\n(MDMIL) to tackle the issues above. Specifically, MDMIL is constructed by the\ninternal query generation module (IQGM) and the multiplex detection module\n(MDM) and assisted by the memory-based contrastive loss during training.\nFirstly, IQGM gives the probability of instances and generates the internal\nquery (IQ) for the subsequent MDM by aggregating highly reliable features after\nthe distribution analysis. Secondly, the multiplex-detection cross-attention\n(MDCA) and multi-head self-attention (MHSA) in MDM cooperate to generate the\nfinal representations for the WSI. In this process, the IQ and trainable\nvariational query (VQ) successfully build up the connections between instances\nand significantly improve the model's robustness toward heterogeneous tumours.\nAt last, to further enforce constraints in the feature space and stabilize the\ntraining process, we adopt a memory-based contrastive loss, which is\npracticable for WSI classification even with a single sample as input in each\niteration. We conduct experiments on three computational pathology datasets,\ne.g., CAMELYON16, TCGA-NSCLC, and TCGA-RCC datasets. The superior accuracy and\nAUC demonstrate the superiority of our proposed MDMIL over other\nstate-of-the-art methods.\n", "date": "2022-08-06", "categories": []}
{"id": "2208.13776", "title": "Attention-based Interpretable Regression of Gene Expression in Histology", "abstract": "  Interpretability of deep learning is widely used to evaluate the reliability\nof medical imaging models and reduce the risks of inaccurate patient\nrecommendations. For models exceeding human performance, e.g. predicting RNA\nstructure from microscopy images, interpretable modelling can be further used\nto uncover highly non-trivial patterns which are otherwise imperceptible to the\nhuman eye. We show that interpretability can reveal connections between the\nmicroscopic appearance of cancer tissue and its gene expression profiling.\nWhile exhaustive profiling of all genes from the histology images is still\nchallenging, we estimate the expression values of a well-known subset of genes\nthat is indicative of cancer molecular subtype, survival, and treatment\nresponse in colorectal cancer. Our approach successfully identifies meaningful\ninformation from the image slides, highlighting hotspots of high gene\nexpression. Our method can help characterise how gene expression shapes tissue\nmorphology and this may be beneficial for patient stratification in the\npathology unit. The code is available on GitHub.\n", "date": "2022-08-29", "categories": []}
{"id": "2208.10730", "title": "Ultra-high-resolution unpaired stain transformation via Kernelized\n  Instance Normalization", "abstract": "  While hematoxylin and eosin (H&E) is a standard staining procedure,\nimmunohistochemistry (IHC) staining further serves as a diagnostic and\nprognostic method. However, acquiring special staining results requires\nsubstantial costs.\n  Hence, we proposed a strategy for ultra-high-resolution unpaired\nimage-to-image translation: Kernelized Instance Normalization (KIN), which\npreserves local information and successfully achieves seamless stain\ntransformation with constant GPU memory usage. Given a patch, corresponding\nposition, and a kernel, KIN computes local statistics using convolution\noperation. In addition, KIN can be easily plugged into most currently developed\nframeworks without re-training.\n  We demonstrate that KIN achieves state-of-the-art stain transformation by\nreplacing instance normalization (IN) layers with KIN layers in three popular\nframeworks and testing on two histopathological datasets. Furthermore, we\nmanifest the generalizability of KIN with high-resolution natural images.\nFinally, human evaluation and several objective metrics are used to compare the\nperformance of different approaches.\n  Overall, this is the first successful study for the ultra-high-resolution\nunpaired image-to-image translation with constant space complexity. Code is\navailable at: https://github.com/Kaminyou/URUST\n", "date": "2022-08-23", "categories": []}
{"id": "2208.12847", "title": "Region-guided CycleGANs for Stain Transfer in Whole Slide Images", "abstract": "  In whole slide imaging, commonly used staining techniques based on\nhematoxylin and eosin (H&E) and immunohistochemistry (IHC) stains accentuate\ndifferent aspects of the tissue landscape. In the case of detecting metastases,\nIHC provides a distinct readout that is readily interpretable by pathologists.\nIHC, however, is a more expensive approach and not available at all medical\ncenters. Virtually generating IHC images from H&E using deep neural networks\nthus becomes an attractive alternative. Deep generative models such as\nCycleGANs learn a semantically-consistent mapping between two image domains,\nwhile emulating the textural properties of each domain. They are therefore a\nsuitable choice for stain transfer applications. However, they remain fully\nunsupervised, and possess no mechanism for enforcing biological consistency in\nstain transfer. In this paper, we propose an extension to CycleGANs in the form\nof a region of interest discriminator. This allows the CycleGAN to learn from\nunpaired datasets where, in addition, there is a partial annotation of objects\nfor which one wishes to enforce consistency. We present a use case on whole\nslide images, where an IHC stain provides an experimentally generated signal\nfor metastatic cells. We demonstrate the superiority of our approach over prior\nart in stain transfer on histopathology tiles over two datasets. Our code and\nmodel are available at https://github.com/jcboyd/miccai2022-roigan.\n", "date": "2022-08-26", "categories": []}
{"id": "2208.12635", "title": "A Two Step Approach for Whole Slide Image Registration", "abstract": "  Multi-stain whole-slide-image (WSI) registration is an active field of\nresearch. It is unclear, however, how the current WSI registration methods\nwould perform on a real-world data set. AutomatiC Registration Of Breast cAncer\nTissue (ACROBAT) challenge is held to verify the performance of the current WSI\nregistration methods by using a new dataset that originates from routine\ndiagnostics to assess real-world applicability. In this report, we present our\nsolution for the ACROBAT challenge. We employ a two-step approach including\nrigid and non-rigid transforms. The experimental results show that the median\n90th percentile is 1,250 um for the validation dataset.\n", "date": "2022-08-25", "categories": []}
{"id": "2208.11052", "title": "IMPaSh: A Novel Domain-shift Resistant Representation for Colorectal\n  Cancer Tissue Classification", "abstract": "  The appearance of histopathology images depends on tissue type, staining and\ndigitization procedure. These vary from source to source and are the potential\ncauses for domain-shift problems. Owing to this problem, despite the great\nsuccess of deep learning models in computational pathology, a model trained on\na specific domain may still perform sub-optimally when we apply them to another\ndomain. To overcome this, we propose a new augmentation called PatchShuffling\nand a novel self-supervised contrastive learning framework named IMPaSh for\npre-training deep learning models. Using these, we obtained a ResNet50 encoder\nthat can extract image representation resistant to domain-shift. We compared\nour derived representation against those acquired based on other\ndomain-generalization techniques by using them for the cross-domain\nclassification of colorectal tissue images. We show that the proposed method\noutperforms other traditional histology domain-adaptation and state-of-the-art\nself-supervised learning methods. Code is available at:\nhttps://github.com/trinhvg/IMPash .\n", "date": "2022-08-23", "categories": []}
{"id": "2208.10429", "title": "Patient-level Microsatellite Stability Assessment from Whole Slide\n  Images By Combining Momentum Contrast Learning and Group Patch Embeddings", "abstract": "  Assessing microsatellite stability status of a patient's colorectal cancer is\ncrucial in personalizing treatment regime. Recently,\nconvolutional-neural-networks (CNN) combined with transfer-learning approaches\nwere proposed to circumvent traditional laboratory testing for determining\nmicrosatellite status from hematoxylin and eosin stained biopsy whole slide\nimages (WSI). However, the high resolution of WSI practically prevent direct\nclassification of the entire WSI. Current approaches bypass the WSI high\nresolution by first classifying small patches extracted from the WSI, and then\naggregating patch-level classification logits to deduce the patient-level\nstatus. Such approaches limit the capacity to capture important information\nwhich resides at the high resolution WSI data. We introduce an effective\napproach to leverage WSI high resolution information by momentum contrastive\nlearning of patch embeddings along with training a patient-level classifier on\ngroups of those embeddings. Our approach achieves up to 7.4\\% better accuracy\ncompared to the straightforward patch-level classification and patient level\naggregation approach with a higher stability (AUC, $0.91 \\pm 0.01$ vs. $0.85\n\\pm 0.04$, p-value$<0.01$). Our code can be found at\nhttps://github.com/TechnionComputationalMRILab/colorectal_cancer_ai.\n", "date": "2022-08-22", "categories": []}
{"id": "2207.06578", "title": "Virtual stain transfer in histology via cascaded deep neural networks", "abstract": "  Pathological diagnosis relies on the visual inspection of histologically\nstained thin tissue specimens, where different types of stains are applied to\nbring contrast to and highlight various desired histological features. However,\nthe destructive histochemical staining procedures are usually irreversible,\nmaking it very difficult to obtain multiple stains on the same tissue section.\nHere, we demonstrate a virtual stain transfer framework via a cascaded deep\nneural network (C-DNN) to digitally transform hematoxylin and eosin (H&E)\nstained tissue images into other types of histological stains. Unlike a single\nneural network structure which only takes one stain type as input to digitally\noutput images of another stain type, C-DNN first uses virtual staining to\ntransform autofluorescence microscopy images into H&E and then performs stain\ntransfer from H&E to the domain of the other stain in a cascaded manner. This\ncascaded structure in the training phase allows the model to directly exploit\nhistochemically stained image data on both H&E and the target special stain of\ninterest. This advantage alleviates the challenge of paired data acquisition\nand improves the image quality and color accuracy of the virtual stain transfer\nfrom H&E to another stain. We validated the superior performance of this C-DNN\napproach using kidney needle core biopsy tissue sections and successfully\ntransferred the H&E-stained tissue images into virtual PAS (periodic\nacid-Schiff) stain. This method provides high-quality virtual images of special\nstains using existing, histochemically stained slides and creates new\nopportunities in digital pathology by performing highly accurate stain-to-stain\ntransformations.\n", "date": "2022-07-14", "categories": []}
{"id": "2208.06833", "title": "Shuffle Instances-based Vision Transformer for Pancreatic Cancer ROSE\n  Image Classification", "abstract": "  The rapid on-site evaluation (ROSE) technique can signifi-cantly accelerate\nthe diagnosis of pancreatic cancer by im-mediately analyzing the fast-stained\ncytopathological images. Computer-aided diagnosis (CAD) can potentially address\nthe shortage of pathologists in ROSE. However, the cancerous patterns vary\nsignificantly between different samples, making the CAD task extremely\nchallenging. Besides, the ROSE images have complicated perturbations regarding\ncolor distribution, brightness, and contrast due to different staining\nqualities and various acquisition device types. To address these challenges, we\nproposed a shuffle instances-based Vision Transformer (SI-ViT) approach, which\ncan reduce the perturbations and enhance the modeling among the instances. With\nthe regrouped bags of shuffle instances and their bag-level soft labels, the\napproach utilizes a regression head to make the model focus on the cells rather\nthan various perturbations. Simultaneously, combined with a classification\nhead, the model can effectively identify the general distributive patterns\namong different instances. The results demonstrate significant improvements in\nthe classification accuracy with more accurate attention regions, indicating\nthat the diverse patterns of ROSE images are effectively extracted, and the\ncomplicated perturbations are significantly reduced. It also suggests that the\nSI-ViT has excellent potential in analyzing cytopathological images. The code\nand experimental results are available at https://github.com/sagizty/MIL-SI.\n", "date": "2022-08-14", "categories": []}
{"id": "2208.07322", "title": "Cross-scale Attention Guided Multi-instance Learning for Crohn's Disease\n  Diagnosis with Pathological Images", "abstract": "  Multi-instance learning (MIL) is widely used in the computer-aided\ninterpretation of pathological Whole Slide Images (WSIs) to solve the lack of\npixel-wise or patch-wise annotations. Often, this approach directly applies\n\"natural image driven\" MIL algorithms which overlook the multi-scale (i.e.\npyramidal) nature of WSIs. Off-the-shelf MIL algorithms are typically deployed\non a single-scale of WSIs (e.g., 20x magnification), while human pathologists\nusually aggregate the global and local patterns in a multi-scale manner (e.g.,\nby zooming in and out between different magnifications). In this study, we\npropose a novel cross-scale attention mechanism to explicitly aggregate\ninter-scale interactions into a single MIL network for Crohn's Disease (CD),\nwhich is a form of inflammatory bowel disease. The contribution of this paper\nis two-fold: (1) a cross-scale attention mechanism is proposed to aggregate\nfeatures from different resolutions with multi-scale interaction; and (2)\ndifferential multi-scale attention visualizations are generated to localize\nexplainable lesion patterns. By training ~250,000 H&E-stained Ascending Colon\n(AC) patches from 20 CD patient and 30 healthy control samples at different\nscales, our approach achieved a superior Area under the Curve (AUC) score of\n0.8924 compared with baseline models. The official implementation is publicly\navailable at https://github.com/hrlblab/CS-MIL.\n", "date": "2022-08-15", "categories": []}
{"id": "2208.04910", "title": "Deep Learning-Based Objective and Reproducible Osteosarcoma Chemotherapy\n  Response Assessment and Outcome Prediction", "abstract": "  Osteosarcoma is the most common primary bone cancer whose standard treatment\nincludes pre-operative chemotherapy followed by resection. Chemotherapy\nresponse is used for predicting prognosis and further management of patients.\nNecrosis is routinely assessed post-chemotherapy from histology slides on\nresection specimens where necrosis ratio is defined as the ratio of necrotic\ntumor to overall tumor. Patients with necrosis ratio >=90% are known to have\nbetter outcome. Manual microscopic review of necrosis ratio from multiple glass\nslides is semi-quantitative and can have intra- and inter-observer variability.\nWe propose an objective and reproducible deep learning-based approach to\nestimate necrosis ratio with outcome prediction from scanned hematoxylin and\neosin whole slide images. We collected 103 osteosarcoma cases with 3134 WSIs to\ntrain our deep learning model, to validate necrosis ratio assessment, and to\nevaluate outcome prediction. We trained Deep Multi-Magnification Network to\nsegment multiple tissue subtypes including viable tumor and necrotic tumor in\npixel-level and to calculate case-level necrosis ratio from multiple WSIs. We\nshowed necrosis ratio estimated by our segmentation model highly correlates\nwith necrosis ratio from pathology reports manually assessed by experts where\nmean absolute differences for Grades IV (100%), III (>=90%), and II (>=50% and\n<90%) necrosis response are 4.4%, 4.5%, and 17.8%, respectively. We\nsuccessfully stratified patients to predict overall survival with p=10^-6 and\nprogression-free survival with p=0.012. Our reproducible approach without\nvariability enabled us to tune cutoff thresholds, specifically for our model\nand our data set, to 80% for OS and 60% for PFS. Our study indicates deep\nlearning can support pathologists as an objective tool to analyze osteosarcoma\nfrom histology for assessing treatment response and predicting patient outcome.\n", "date": "2022-08-09", "categories": []}
{"id": "2208.03609", "title": "Continual Learning for Tumor Classification in Histopathology Images", "abstract": "  Recent years have seen great advancements in the development of deep learning\nmodels for histopathology image analysis in digital pathology applications,\nevidenced by the increasingly common deployment of these models in both\nresearch and clinical settings. Although such models have shown unprecedented\nperformance in solving fundamental computational tasks in DP applications, they\nsuffer from catastrophic forgetting when adapted to unseen data with transfer\nlearning. With an increasing need for deep learning models to handle ever\nchanging data distributions, including evolving patient population and new\ndiagnosis assays, continual learning models that alleviate model forgetting\nneed to be introduced in DP based analysis. However, to our best knowledge,\nthere is no systematic study of such models for DP-specific applications. Here,\nwe propose CL scenarios in DP settings, where histopathology image data from\ndifferent sources/distributions arrive sequentially, the knowledge of which is\nintegrated into a single model without training all the data from scratch. We\nthen established an augmented dataset for colorectal cancer H&E classification\nto simulate shifts of image appearance and evaluated CL model performance in\nthe proposed CL scenarios. We leveraged a breast tumor H&E dataset along with\nthe colorectal cancer to evaluate CL from different tumor types. In addition,\nwe evaluated CL methods in an online few-shot setting under the constraints of\nannotation and computational resources. We revealed promising results of CL in\nDP applications, potentially paving the way for application of these methods in\nclinical practice.\n", "date": "2022-08-07", "categories": []}
{"id": "2208.01674", "title": "Diagnosis of Paratuberculosis in Histopathological Images Based on\n  Explainable Artificial Intelligence and Deep Learning", "abstract": "  Artificial intelligence holds great promise in medical imaging, especially\nhistopathological imaging. However, artificial intelligence algorithms cannot\nfully explain the thought processes during decision-making. This situation has\nbrought the problem of explainability, i.e., the black box problem, of\nartificial intelligence applications to the agenda: an algorithm simply\nresponds without stating the reasons for the given images. To overcome the\nproblem and improve the explainability, explainable artificial intelligence\n(XAI) has come to the fore, and piqued the interest of many researchers.\nAgainst this backdrop, this study examines a new and original dataset using the\ndeep learning algorithm, and visualizes the output with gradient-weighted class\nactivation mapping (Grad-CAM), one of the XAI applications. Afterwards, a\ndetailed questionnaire survey was conducted with the pathologists on these\nimages. Both the decision-making processes and the explanations were verified,\nand the accuracy of the output was tested. The research results greatly help\npathologists in the diagnosis of paratuberculosis.\n", "date": "2022-08-02", "categories": []}
{"id": "2208.00594", "title": "Breast Cancer Classification Based on Histopathological Images Using a\n  Deep Learning Capsule Network", "abstract": "  Breast cancer is one of the most serious types of cancer that can occur in\nwomen. The automatic diagnosis of breast cancer by analyzing histological\nimages (HIs) is important for patients and their prognosis. The classification\nof HIs provides clinicians with an accurate understanding of diseases and\nallows them to treat patients more efficiently. Deep learning (DL) approaches\nhave been successfully employed in a variety of fields, particularly medical\nimaging, due to their capacity to extract features automatically. This study\naims to classify different types of breast cancer using HIs. In this research,\nwe present an enhanced capsule network that extracts multi-scale features using\nthe Res2Net block and four additional convolutional layers. Furthermore, the\nproposed method has fewer parameters due to using small convolutional kernels\nand the Res2Net block. As a result, the new method outperforms the old ones\nsince it automatically learns the best possible features. The testing results\nshow that the model outperformed the previous DL methods.\n", "date": "2022-08-01", "categories": []}
{"id": "2207.00095", "title": "End-to-end Learning for Image-based Detection of Molecular Alterations\n  in Digital Pathology", "abstract": "  Current approaches for classification of whole slide images (WSI) in digital\npathology predominantly utilize a two-stage learning pipeline. The first stage\nidentifies areas of interest (e.g. tumor tissue), while the second stage\nprocesses cropped tiles from these areas in a supervised fashion. During\ninference, a large number of tiles are combined into a unified prediction for\nthe entire slide. A major drawback of such approaches is the requirement for\ntask-specific auxiliary labels which are not acquired in clinical routine. We\npropose a novel learning pipeline for WSI classification that is trainable\nend-to-end and does not require any auxiliary annotations. We apply our\napproach to predict molecular alterations for a number of different use-cases,\nincluding detection of microsatellite instability in colorectal tumors and\nprediction of specific mutations for colon, lung, and breast cancer cases from\nThe Cancer Genome Atlas. Results reach AUC scores of up to 94% and are shown to\nbe competitive with state of the art two-stage pipelines. We believe our\napproach can facilitate future research in digital pathology and contribute to\nsolve a large range of problems around the prediction of cancer phenotypes,\nhopefully enabling personalized therapies for more patients in future.\n", "date": "2022-06-30", "categories": []}
{"id": "2207.06193", "title": "Domain adaptation strategies for cancer-independent detection of lymph\n  node metastases", "abstract": "  Recently, large, high-quality public datasets have led to the development of\nconvolutional neural networks that can detect lymph node metastases of breast\ncancer at the level of expert pathologists. Many cancers, regardless of the\nsite of origin, can metastasize to lymph nodes. However, collecting and\nannotating high-volume, high-quality datasets for every cancer type is\nchallenging. In this paper we investigate how to leverage existing high-quality\ndatasets most efficiently in multi-task settings for closely related tasks.\nSpecifically, we will explore different training and domain adaptation\nstrategies, including prevention of catastrophic forgetting, for colon and\nhead-and-neck cancer metastasis detection in lymph nodes.\n  Our results show state-of-the-art performance on both cancer metastasis\ndetection tasks. Furthermore, we show the effectiveness of repeated adaptation\nof networks from one cancer type to another to obtain multi-task metastasis\ndetection networks. Last, we show that leveraging existing high-quality\ndatasets can significantly boost performance on new target tasks and that\ncatastrophic forgetting can be effectively mitigated using regularization.\n", "date": "2022-07-13", "categories": []}
{"id": "2207.05290", "title": "Trusted Multi-Scale Classification Framework for Whole Slide Image", "abstract": "  Despite remarkable efforts been made, the classification of gigapixels\nwhole-slide image (WSI) is severely restrained from either the constrained\ncomputing resources for the whole slides, or limited utilizing of the knowledge\nfrom different scales. Moreover, most of the previous attempts lacked of the\nability of uncertainty estimation. Generally, the pathologists often jointly\nanalyze WSI from the different magnifications. If the pathologists are\nuncertain by using single magnification, then they will change the\nmagnification repeatedly to discover various features of the tissues. Motivated\nby the diagnose process of the pathologists, in this paper, we propose a\ntrusted multi-scale classification framework for the WSI. Leveraging the Vision\nTransformer as the backbone for multi branches, our framework can jointly\nclassification modeling, estimating the uncertainty of each magnification of a\nmicroscope and integrate the evidence from different magnification. Moreover,\nto exploit discriminative patches from WSIs and reduce the requirement for\ncomputation resources, we propose a novel patch selection schema using\nattention rollout and non-maximum suppression. To empirically investigate the\neffectiveness of our approach, empirical experiments are conducted on our WSI\nclassification tasks, using two benchmark databases. The obtained results\nsuggest that the trusted framework can significantly improve the WSI\nclassification performance compared with the state-of-the-art methods.\n", "date": "2022-07-12", "categories": []}
{"id": "2207.00431", "title": "Stain Isolation-based Guidance for Improved Stain Translation", "abstract": "  Unsupervised and unpaired domain translation using generative adversarial\nneural networks, and more precisely CycleGAN, is state of the art for the stain\ntranslation of histopathology images. It often, however, suffers from the\npresence of cycle-consistent but non structure-preserving errors. We propose an\nalternative approach to the set of methods which, relying on segmentation\nconsistency, enable the preservation of pathology structures. Focusing on\nimmunohistochemistry (IHC) and multiplexed immunofluorescence (mIF), we\nintroduce a simple yet effective guidance scheme as a loss function that\nleverages the consistency of stain translation with stain isolation.\nQualitative and quantitative experiments show the ability of the proposed\napproach to improve translation between the two domains.\n", "date": "2022-06-28", "categories": []}
{"id": "2206.10033", "title": "Test Time Transform Prediction for Open Set Histopathological Image\n  Recognition", "abstract": "  Tissue typology annotation in Whole Slide histological images is a complex\nand tedious, yet necessary task for the development of computational pathology\nmodels. We propose to address this problem by applying Open Set Recognition\ntechniques to the task of jointly classifying tissue that belongs to a set of\nannotated classes, e.g. clinically relevant tissue categories, while rejecting\nin test time Open Set samples, i.e. images that belong to categories not\npresent in the training set. To this end, we introduce a new approach for Open\nSet histopathological image recognition based on training a model to accurately\nidentify image categories and simultaneously predict which data augmentation\ntransform has been applied. In test time, we measure model confidence in\npredicting this transform, which we expect to be lower for images in the Open\nSet. We carry out comprehensive experiments in the context of colorectal cancer\nassessment from histological images, which provide evidence on the strengths of\nour approach to automatically identify samples from unknown categories. Code is\nreleased at https://github.com/agaldran/t3po .\n", "date": "2022-06-20", "categories": []}
{"id": "2206.12798", "title": "Multiple Instance Learning with Mixed Supervision in Gleason Grading", "abstract": "  With the development of computational pathology, deep learning methods for\nGleason grading through whole slide images (WSIs) have excellent prospects.\nSince the size of WSIs is extremely large, the image label usually contains\nonly slide-level label or limited pixel-level labels. The current mainstream\napproach adopts multi-instance learning to predict Gleason grades. However,\nsome methods only considering the slide-level label ignore the limited\npixel-level labels containing rich local information. Furthermore, the method\nof additionally considering the pixel-level labels ignores the inaccuracy of\npixel-level labels. To address these problems, we propose a mixed supervision\nTransformer based on the multiple instance learning framework. The model\nutilizes both slide-level label and instance-level labels to achieve more\naccurate Gleason grading at the slide level. The impact of inaccurate\ninstance-level labels is further reduced by introducing an efficient random\nmasking strategy in the mixed supervision training process. We achieve the\nstate-of-the-art performance on the SICAPv2 dataset, and the visual analysis\nshows the accurate prediction results of instance level. The source code is\navailable at https://github.com/bianhao123/Mixed_supervision.\n", "date": "2022-06-26", "categories": []}
{"id": "2206.11943", "title": "TIAger: Tumor-Infiltrating Lymphocyte Scoring in Breast Cancer for the\n  TiGER Challenge", "abstract": "  The quantification of tumor-infiltrating lymphocytes (TILs) has been shown to\nbe an independent predictor for prognosis of breast cancer patients. Typically,\npathologists give an estimate of the proportion of the stromal region that\ncontains TILs to obtain a TILs score. The Tumor InfiltratinG lymphocytes in\nbreast cancER (TiGER) challenge, aims to assess the prognostic significance of\ncomputer-generated TILs scores for predicting survival as part of a Cox\nproportional hazards model. For this challenge, as the TIAger team, we have\ndeveloped an algorithm to first segment tumor vs. stroma, before localising the\ntumor bulk region for TILs detection. Finally, we use these outputs to generate\na TILs score for each case. On preliminary testing, our approach achieved a\ntumor-stroma weighted Dice score of 0.791 and a FROC score of 0.572 for\nlymphocytic detection. For predicting survival, our model achieved a C-index of\n0.719. These results achieved first place across the preliminary testing\nleaderboards of the TiGER challenge.\n", "date": "2022-06-23", "categories": []}
{"id": "2206.11520", "title": "ICOS Protein Expression Segmentation: Can Transformer Networks Give\n  Better Results?", "abstract": "  Biomarkers identify a patients response to treatment. With the recent\nadvances in artificial intelligence based on the Transformer networks, there is\nonly limited research has been done to measure the performance on challenging\nhistopathology images. In this paper, we investigate the efficacy of the\nnumerous state-of-the-art Transformer networks for immune-checkpoint biomarker,\nInducible Tcell COStimulator (ICOS) protein cell segmentation in colon cancer\nfrom immunohistochemistry (IHC) slides. Extensive and comprehensive\nexperimental results confirm that MiSSFormer achieved the highest Dice score of\n74.85% than the rest evaluated Transformer and Efficient U-Net methods.\n", "date": "2022-06-23", "categories": []}
{"id": "2206.10573", "title": "H&E-based Computational Biomarker Enables Universal EGFR Screening for\n  Lung Adenocarcinoma", "abstract": "  Lung cancer is the leading cause of cancer death worldwide, with lung\nadenocarcinoma being the most prevalent form of lung cancer. EGFR positive lung\nadenocarcinomas have been shown to have high response rates to TKI therapy,\nunderlying the essential nature of molecular testing for lung cancers. Despite\ncurrent guidelines consider testing necessary, a large portion of patients are\nnot routinely profiled, resulting in millions of people not receiving the\noptimal treatment for their lung cancer. Sequencing is the gold standard for\nmolecular testing of EGFR mutations, but it can take several weeks for results\nto come back, which is not ideal in a time constrained scenario. The\ndevelopment of alternative screening tools capable of detecting EGFR mutations\nquickly and cheaply while preserving tissue for sequencing could help reduce\nthe amount of sub-optimally treated patients. We propose a multi-modal approach\nwhich integrates pathology images and clinical variables to predict EGFR\nmutational status achieving an AUC of 84% on the largest clinical cohort to\ndate. Such a computational model could be deployed at large at little\nadditional cost. Its clinical application could reduce the number of patients\nwho receive sub-optimal treatments by 53.1% in China, and up to 96.6% in the\nUS.\n", "date": "2022-06-21", "categories": []}
{"id": "2206.08861", "title": "DGMIL: Distribution Guided Multiple Instance Learning for Whole Slide\n  Image Classification", "abstract": "  Multiple Instance Learning (MIL) is widely used in analyzing\nhistopathological Whole Slide Images (WSIs). However, existing MIL methods do\nnot explicitly model the data distribution, and instead they only learn a\nbag-level or instance-level decision boundary discriminatively by training a\nclassifier. In this paper, we propose DGMIL: a feature distribution guided deep\nMIL framework for WSI classification and positive patch localization. Instead\nof designing complex discriminative network architectures, we reveal that the\ninherent feature distribution of histopathological image data can serve as a\nvery effective guide for instance classification. We propose a\ncluster-conditioned feature distribution modeling method and a pseudo\nlabel-based iterative feature space refinement strategy so that in the final\nfeature space the positive and negative instances can be easily separated.\nExperiments on the CAMELYON16 dataset and the TCGA Lung Cancer dataset show\nthat our method achieves new SOTA for both global classification and positive\npatch localization tasks.\n", "date": "2022-06-17", "categories": []}
{"id": "2206.08275", "title": "Rank the triplets: A ranking-based multiple instance learning framework\n  for detecting HPV infection in head and neck cancers using routine H&E images", "abstract": "  The aetiology of head and neck squamous cell carcinoma (HNSCC) involves\nmultiple carcinogens such as alcohol, tobacco and infection with human\npapillomavirus (HPV). As the HPV infection influences the prognosis, treatment\nand survival of patients with HNSCC, it is important to determine the HPV\nstatus of these tumours. In this paper, we propose a novel triplet-ranking loss\nfunction and a multiple instance learning pipeline for HPV status prediction.\nThis achieves a new state-of-the-art performance in HPV detection using only\nthe routine H&E stained WSIs on two HNSCC cohorts. Furthermore, a comprehensive\ntumour microenvironment profiling was performed, which characterised the unique\npatterns between HPV+/- HNSCC from genomic, immunology and cellular\nperspectives. Positive correlations of the proposed score with different\nsubtypes of T cells (e.g. T cells follicular helper, CD8+ T cells), and\nnegative correlations with macrophages and connective cells (e.g. fibroblast)\nwere identified, which is in line with clinical findings. Unique gene\nexpression profiles were also identified with respect to HPV infection status,\nand is in line with existing findings.\n", "date": "2022-06-16", "categories": []}
{"id": "2206.07573", "title": "AI and Pathology: Steering Treatment and Predicting Outcomes", "abstract": "  The combination of data analysis methods, increasing computing capacity, and\nimproved sensors enable quantitative granular, multi-scale, cell-based\nanalyses. We describe the rich set of application challenges related to tissue\ninterpretation and survey AI methods currently used to address these\nchallenges. We focus on a particular class of targeted human tissue analysis -\nhistopathology - aimed at quantitative characterization of disease state,\npatient outcome prediction and treatment steering.\n", "date": "2022-06-15", "categories": []}
{"id": "2206.03368", "title": "IL-MCAM: An interactive learning and multi-channel attention\n  mechanism-based weakly supervised colorectal histopathology image\n  classification approach", "abstract": "  In recent years, colorectal cancer has become one of the most significant\ndiseases that endanger human health. Deep learning methods are increasingly\nimportant for the classification of colorectal histopathology images. However,\nexisting approaches focus more on end-to-end automatic classification using\ncomputers rather than human-computer interaction. In this paper, we propose an\nIL-MCAM framework. It is based on attention mechanisms and interactive\nlearning. The proposed IL-MCAM framework includes two stages: automatic\nlearning (AL) and interactivity learning (IL). In the AL stage, a multi-channel\nattention mechanism model containing three different attention mechanism\nchannels and convolutional neural networks is used to extract multi-channel\nfeatures for classification. In the IL stage, the proposed IL-MCAM framework\ncontinuously adds misclassified images to the training set in an interactive\napproach, which improves the classification ability of the MCAM model. We\ncarried out a comparison experiment on our dataset and an extended experiment\non the HE-NCT-CRC-100K dataset to verify the performance of the proposed\nIL-MCAM framework, achieving classification accuracies of 98.98% and 99.77%,\nrespectively. In addition, we conducted an ablation experiment and an\ninterchangeability experiment to verify the ability and interchangeability of\nthe three channels. The experimental results show that the proposed IL-MCAM\nframework has excellent performance in the colorectal histopathological image\nclassification tasks.\n", "date": "2022-06-07", "categories": []}
{"id": "2206.02647", "title": "Scaling Vision Transformers to Gigapixel Images via Hierarchical\n  Self-Supervised Learning", "abstract": "  Vision Transformers (ViTs) and their multi-scale and hierarchical variations\nhave been successful at capturing image representations but their use has been\ngenerally studied for low-resolution images (e.g. - 256x256, 384384). For\ngigapixel whole-slide imaging (WSI) in computational pathology, WSIs can be as\nlarge as 150000x150000 pixels at 20X magnification and exhibit a hierarchical\nstructure of visual tokens across varying resolutions: from 16x16 images\ncapture spatial patterns among cells, to 4096x4096 images characterizing\ninteractions within the tissue microenvironment. We introduce a new ViT\narchitecture called the Hierarchical Image Pyramid Transformer (HIPT), which\nleverages the natural hierarchical structure inherent in WSIs using two levels\nof self-supervised learning to learn high-resolution image representations.\nHIPT is pretrained across 33 cancer types using 10,678 gigapixel WSIs, 408,218\n4096x4096 images, and 104M 256x256 images. We benchmark HIPT representations on\n9 slide-level tasks, and demonstrate that: 1) HIPT with hierarchical\npretraining outperforms current state-of-the-art methods for cancer subtyping\nand survival prediction, 2) self-supervised ViTs are able to model important\ninductive biases about the hierarchical structure of phenotypes in the tumor\nmicroenvironment.\n", "date": "2022-06-06", "categories": []}
{"id": "2206.00389", "title": "A comparative study between vision transformers and CNNs in digital\n  pathology", "abstract": "  Recently, vision transformers were shown to be capable of outperforming\nconvolutional neural networks when pretrained on sufficient amounts of data. In\ncomparison to convolutional neural networks, vision transformers have a weaker\ninductive bias and therefore allow a more flexible feature detection. Due to\ntheir promising feature detection, this work explores vision transformers for\ntumor detection in digital pathology whole slide images in four tissue types,\nand for tissue type identification. We compared the patch-wise classification\nperformance of the vision transformer DeiT-Tiny to the state-of-the-art\nconvolutional neural network ResNet18. Due to the sparse availability of\nannotated whole slide images, we further compared both models pretrained on\nlarge amounts of unlabeled whole-slide images using state-of-the-art\nself-supervised approaches. The results show that the vision transformer\nperformed slightly better than the ResNet18 for three of four tissue types for\ntumor detection while the ResNet18 performed slightly better for the remaining\ntasks. The aggregated predictions of both models on slide level were\ncorrelated, indicating that the models captured similar imaging features. All\ntogether, the vision transformer models performed on par with the ResNet18\nwhile requiring more effort to train. In order to surpass the performance of\nconvolutional neural networks, vision transformers might require more\nchallenging tasks to benefit from their weak inductive bias.\n", "date": "2022-06-01", "categories": []}
{"id": "2606.21174", "title": "HERO: Hypothesis-Driven Evidence Retrieval from Omics for Multi-Task Breast Cancer Analysis", "abstract": "Matched multi-omics can improve WSI-based biomarker and prognosis prediction, but most existing pipelines use omics as a paral lel feature stream or textual context rather than as an explicit retrieval constraint. HERO asks whether observed omics can be a testable mor phology hypothesis: a sparse pathway-to-morphology prior maps DNA methylation and miRNA into a K-dimensional intent vector m (K=16), TF-IDF retrieval over structured 10 captions selects endpoint-relevant regions, and a cosine gate c=cos(m,v) triggers deterministic deficit driven repair when c<{\\tau}c. This closed-loop design bounds VLM calls, reduces reliance on embedding-based semantic matching, and makes every retrieval and verification step lexically auditable. On TCGA-BRCA (930WSIs, patient-level 5-fold CV), HERO sets new state-of-the-art across ER, PR, HER2, subtype, and risk prediction, outperforming both multimodal fusion and VLM-based baselines.", "date": "2026-06-19", "categories": []}
{"id": "2606.28395", "title": "JASPR: Joint Spatial Representation learning of histology and spatial genomics for improved virtual genomic screening and clinical prognostication", "abstract": "Recent studies have shown that spatial properties of tumors are critical for understanding disease biology and predicting patient outcomes. These spatial properties are increasingly uncovered through complementary modalities: spatial transcriptomics (ST) captures spatially-resolved molecular states, while hematoxylin and eosin-stained whole slide images (HE) reveal tissue morphology. While approaches are emerging to fuse these modalities, effective methods that learn not only joint representations but also incorporate spatial context across modalities are lacking. Here, we present JASPR (Joint Spatial Representation learning), a self-supervised deep learning framework that integrates HE images and ST data through a cross-modal reconstruction objective that incorporates spatial context within HE images and ST profiles. It employs shared modules to capture universal spatial properties across modalities, while modality-specific experts encode features unique to morphological and genomic data. We train and validate JASPR on breast cancer datasets, demonstrating that its learned joint representation substantially improves HE-based prediction of 9,248 genes and provides prognostic value for breast cancer outcomes.", "date": "2026-06-24", "categories": []}
{"id": "2606.28676", "title": "Predicting Metastatic Risk from Primary Tissue Architecture via Distance-Aware Spatial Modeling", "abstract": "Predicting the risk of distant metastasis from primary tumor tissue histology is a critical yet challenging task in computational pathology. Multiple Instance Learning (MIL) approaches can attend to subdomains in tumor regions that harbor features of metastatic cancer progression. However MIL models treat tissue patches as unordered bags, discarding the spatial layout that defines the metastatic potential. We propose that metastatic risk is inherently dictated by the geometric arrangement of the tumor microenvironment at the interface with tumor cells. Our model is designed to explicitly capture the spatial relationships between tumor cells, tumor associated fibroblasts and infiltrating lymphocytes. For this purpose, we propose Distance aware Tissue Modeling for Multiple Instance Learning(DTMf-MIL), a novel method that reinforces visual features with explicit spatial priors. By computing signed distance functions (SDF) relative to tissue phenotypes, our model learns to recognize structural signatures of metastatic risk. This geometric awareness translates directly to superior clinical performance as DTMf-MIL significantly outperforms state-of-the-art methods that ignore spatial layout on metastasis prediction from tissue in the primary tumor. We further validate our approach on public benchmarks, demonstrating that spatial awareness consistently improves diagnostic accuracy across diverse clinical tasks.", "date": "2026-06-27", "categories": []}
{"id": "2606.28697", "title": "Mitigating Batch Effects in Histopathology via Language-Mediated Robust Embedding Generation", "abstract": "Pathology foundation models (PFMs) have demonstrated strong potential across clinical and scientific applications, yet their performance is often hindered by batch effects, which are non-biological variations across tissue source institutions (TSIs) that distort learned feature representations and impair generalization. Conventional mitigation strategies, such as stain normalization, offer limited success in addressing these high-dimensional, complex artifacts. We present GLMP (General-purpose LLM-Mediated Pathology model), a novel framework that generates robust numerical embeddings from histology image patches through an intermediate textual representation. By leveraging pretrained general-purpose multimodal large language models (MLLMs) and text encoders, GLMP effectively prioritizes biologically meaningful signals over TSI-specific artifacts, thereby improving cross-institutional generalization. To our knowledge, GLMP is the first pathology model to use text descriptions of histological features as an intermediate representation for generating numerical embeddings from histology images. Our results highlight the untapped potential of broad-domain, non-specialized MLLMs in computational pathology and introduce a new paradigm for building versatile, generalizable, and robust pathology models.", "date": "2026-06-27", "categories": []}
{"id": "2606.29463", "title": "CellDETR: A Detection-Guided Framework for Scalable Cell Representation Learning from Histopathology Images", "abstract": "Recent advances in pathology foundation models have substantially improved patch and slide level representation learning from whole-slide images (WSIs).However, cell-level representations learning remain underexplored, limiting cell resolved interpretability, biological discovery, and clinical translation. We propose CellDETR, a detection-guided framework built on Deformable DETR for scalable cell representation learning from WSIs. By introducing location feature decoupling and box-constrained attention mechanism, CellDETR enables automated extraction of cell-level embeddings, and outperform existing state-of-the-art methods in supervised cell classification on PanNuke data. In addition, by incorporating contrastive learning design, we build a CellDETR-based pretraining model for scalable cell representation learning from unlabeled WSIs, which improves downstream cell classification performance. Furthermore, we show that after pretraining with Xenium spatial transcriptomics-derived cell annotations, CellDETR achieves accurate cross-dataset cell classification, demonstrating the transferability and biological relevance of the learned cell embeddings. Together, CellDETR provides a scalable route toward general cell-level representation learning framework for interpretable computational patholog", "date": "2026-06-28", "categories": []}
{"id": "2606.29883", "title": "Building artificial intelligence virtual tissue (AIVT) for tissue state representation, feature prediction, and dynamic simulation", "abstract": "Modeling tissue states and their transitions is essential for understanding tissue homeostasis in health and pathological remodeling in disease. However, conventional computational modeling approaches are inadequate to capture the complexity of tissues as spatially organized, multiscale biological systems. Artificial intelligence (AI) has shown a remarkable ability for representing intricate systems, creating new opportunities to characterize tissue states and their transitions. Here, we propose the concept of AI virtual tissue (AIVT), an AI framework grounded in spatial multimodal data for modeling tissues in health and disease. AIVT is designed to learn unified, spatially resolved, and dynamically manipulatable representations of tissue state, enabling tissue state representation and analysis, molecular and morphological feature prediction, and simulation of spatiotemporal tissue dynamics. We outline the fundamental assumptions, core capabilities, architectural components, as well as data and algorithm foundations of AIVT as a framework for AI-driven tissue modeling.", "date": "2026-06-29", "categories": []}
{"id": "2606.29949", "title": "Data-Efficient Multimodal Alignment for Histopathology-based Molecular Prediction", "abstract": "H&E-stained whole-slide images offer cohort-scale availability and rich spatial context but lack molecular specificity, whereas bulk RNA-seq provides transcriptome-wide resolution at high cost with limited archival availability. We show that training a lightweight alignment module atop frozen histopathology and RNA-Seq foundation models enables open-vocabulary molecular prompting -- querying H&E slides with gene-set signatures to predict pathway activity without sequencing or end-to-end retraining. Using contrastive learning on a multi-cancer cohort (N=1,720), we achieve a 25-fold improvement in retrieval over baseline methods. Systematic analysis reveals a graduated predictability spectrum: morphologically grounded programs (cell-cycle programs, immune-related) are most reliably predicted (R^2>0.5), while predicting pathways with no morphological footprint remains challenging as expected. We validate clinical utility on the POSEIDON clinical trial: H&E-predicted squamous cell carcinoma scores recapitulate NSCLC subtype identity and predicted IFN-gamma mirror PD-L1 tumor-cell expression groups. Furthermore, genesets describing immune activation and fibrosis predict known tumor microenvironment archetypes from histology alone. We further validate generalization of our approach across unseen cohorts and demonstrate data-efficient domain adaptation, establishing a slide-native framework for molecular analysis on H&E images.", "date": "2026-06-29", "categories": []}
{"id": "2606.30020", "title": "Uncertainty Estimation in Pathology Foundation Models via Deep Mutual Learning", "abstract": "Pathology foundation models (PFMs) offer generalizable representations for whole-slide image (WSI) analysis, yet their clinical adoption remains limited. Specifically, their predictions lack reliable confidence estimates, and no single PFM is universally best across tasks, which severely undermines trust in medical settings. To overcome this, we propose $\\mathtt{DICE}$, a plug-and-play framework that ensembles $K$ frozen PFMs and models their disagreement as a proxy for uncertainty estimation. To ensure this proxy yields meaningful estimates, we align the ensemble members via deep mutual learning, and theoretically show that this objective upper-bounds the model uncertainty. Additionally, we demonstrate that the ensemble's consensus localizes abnormalities at the patch level without any explicit supervision. We evaluate $\\mathtt{DICE}$ on three challenging WSI benchmarks. Notably, our framework provides reliable uncertainty estimates that accurately flag failure-prone cases under in- and out-of-distribution settings, while matching or outperforming SOTA baselines in classification, calibration, and localization. Overall, $\\mathtt{DICE}$ takes a crucial step toward translating PFMs into uncertainty-aware decision-support systems.", "date": "2026-06-29", "categories": []}
{"id": "2606.30209", "title": "A Multi Center Breast FNAC Whole-Slide Cytology Dataset for AI-Assisted Patch-Wise Classification Using C1 to C5 Reporting Categories", "abstract": "We present a multi center breast fine needle aspiration cytology (FNAC) dataset designed for patch wise classification using C1 to C5 reporting labels. The prospective dataset includes 321 patients and 470 whole-slide images (WSIs) collected from participating tertiary medical centers in India between May 2023 and March 2026. Slides were stained using Papanicolaou (190 WSIs) or MayGrunwald Giemsa (280 WSIs), scanned on a Hamamatsu NanoZoomer S360 at 40X magnification and 0.25 microns per pixel, and stored directly in NDPI format. Across the 470 WSIs, 446 WSIs contain annotated patch regions, yielding 7,398 PNG image patches with expert-verified C1 to C5 labels. The release includes NDPI WSIs, WSI-level GeoJSON annotation files, extracted patch images, deidentified metadata, a data dictionary, a validation summary, a manifest linking WSIs to Zenodo records, and code for dataset inspection and reuse. The complete dataset is approximately 950 GB and is available through Zenodo.", "date": "2026-06-29", "categories": []}
{"id": "2606.31100", "title": "TaxoMIL: Taxonomy-Constrained Learning for Hierarchical Whole Slide Image Analysis", "abstract": "Whole slide image (WSI) analysis is central to computational pathology, with multiple instance learning (MIL) emerging as the standard pipeline for slide-level diagnosis. However, conventional approaches formulate WSI diagnosis as a flat classification task over discrete labels, contradicting the inherently hierarchical, coarse-to-fine nature of clinical reasoning. Although recent hierarchical classifiers and vision-language models (VLMs) have sought to address this structural gap, they either fail to capture semantic continuity between related diagnoses or suffer from unconstrained text generation that produces taxonomic hallucinations and parent-child label violations. To address these limitations, we propose TaxoMIL, a taxonomy-constrained framework that reformulates WSI diagnosis as a multi-granularity text generation task. TaxoMIL utilizes a dual-head Transformer decoder to generate coarse- and fine-level diagnostic text, and introduces taxonomy-guided objectives that explicitly structure the label embedding space and strictly ground slide-level visual representations within the clinical taxonomy. Extensive experiments across three diverse WSI datasets demonstrate that TaxoMIL consistently outperforms state-of-the-art MIL classifiers and VLM-based generative methods, yielding accurate and hierarchy-aware diagnostic predictions. The code is released at https://github.com/QuIIL/TaxoMIL", "date": "2026-06-30", "categories": []}