text stringlengths 0 5.79k |
|---|
Pathological score available | 1 | 3 | 1 | 1 |
Age, mean (minimum, maximum) | 68.43 (37, 92) | 67.66 (31, 86) | 65.48 (42, 85) | – |
Sex, n (%) |
Female | 111 (53.88) | 444 (52.54) | 102 (57.30) | – |
Male | 95 (46.12) | 401 (47.46) | 76 (42.70) | – |
Tumor stage, n (%) |
I | 108 (52.43) | 337 (39.88) | 92 (51.69) | – |
II | 54 (26.21) | 202 (23.91) | 40 (22.47) | – |
III | 44 (21.36) | 190 (22.49) | 33 (18.54) | – |
IV | 0 (0) | 0 (0) | 12 (6.74) | – |
Not applicable | 0 (0) | 116 (13.73) | 1 (0.56) | – |
Smoking status, n (%) |
Current smoker | 88 (42.72) | 259 (30.65) | – | – |
Ex-smoker | 101 (49.03) | 419 (49.59) | – | – |
Never smoker | 17 (8.25) | 64 (7.57) | – | – |
Not applicable | – | 103 (12.19) | | |
Adjuvant treatment, n (%) |
Yes | 64 (31.07) | 134 (15.86) | – | – |
No | 142 (68.93) | 711 (84.14) | – | – |
Type of surgery, n (%) |
Lobectomy or greater | 180 (83.98) | 640 (70.53) | – | – |
Sublobar resection | 26 (16.02) | 89 (29.47) | – | – |
ANORAK generated promising outputs for growth pattern segmentation (Fig. 2a and Extended Data Figs. 2a,b and 3a). To validate the effectiveness of the developed model, we conducted the ablation study at the patch level (Extended Data Fig. 3b). Overall, multi-stream variants were more promising than single-stream ones, ... |
In all four cohorts, AI-predicted growth pattern proportions were highly correlated with the pathologists’ estimates (Fig. 2b and Supplementary Table 1), notably for the solid pattern (TRACERx 421, Spearman’s rho = 0.79; LATTICe-A correlations against each pathologist’s scoring, rho1 = 0.80, rho2 = 0.77, rho3 = 0.78; T... |
### AI grading consistently improves patient risk stratification |
Patients with IASLC grade 1 and 2 tumors as identified by AI had notably favorable disease-free survival (DFS) compared to patients with IASLC grade 3 tumors in TRACERx 421 (n = 206, P = 0.003, hazard ratio (HR) = 0.48, 95% confidence interval (CI) = 0.30–0.78) and LATTICe-A (n = 729, P = 1.73 × 10−7, HR = 0.53, 95% CI... |
To determine the prognostic information provided by AI compared to manual scoring and the clinical baseline characteristics, we focused on the large LATTICe-A cohort. While manual IASLC grading from all three pathologists was prognostic (Extended Data Fig. 5b–d), AI grading achieved a comparable performance with all th... |
Taken together, these data suggest that AI grading adds independent prognostic value for patient stratification, particularly for stage I disease in which clinical decision-making regarding adjuvant therapy following surgery can be challenging in the absence of evidence for outcome benefit. |
### Assisting pathologists in challenging scenarios |
To evaluate the utility of our AI method to assist pathologists with LUAD grading, we identified four specific scenarios and used the large LATTICe-A cohort with manual grading available from three pathologists. We focused on stage I LUAD tumors, a group of patients with an unmet need for predicting which patients are ... |
The first scenario consisted of cases with highly diversified growth patterns indicated by the Shannon diversity index (Fig. 4a), which was notably higher in cases with discrepant predominant patterns between AI and pathologists (Fig. 2d). When evaluated in cases with high growth pattern diversity based on the Shannon ... |
Second, we focused on tumors scored predominantly as lepidic or acinar by each pathologist, excluding any morphologically homogeneous tumor that received a score of 90% or more for either pattern22. There is an ongoing difficulty in the histopathological discrimination between in situ and invasive disease4, and the dis... |
The third scenario was the detection of aggressive, high-grade patterns. Although there was a high concordance rate for cases composed predominantly of high-grade patterns (Extended Data Fig. 4e), the proposed IASLC grading system sets a 20% cutoff for high-grade patterns to qualify as grade 3, adding challenges to ide... |
Finally, we considered cases with high numbers of diagnostic slides per tumor (Fig. 4d), defined as four or more slides (n = 233, decreased kappa index in Fig. 4d). In these cases, AI grading achieved a C-index higher than average for the manual grading but lower than pathologist 1 in predicting DFS (AI = 0.638, 95% CI... |
These data indicated that our proposed AI method was not inferior to pathological grading and could assist pathologists to grade growth patterns in certain challenging scenarios. |
### Acinar morphology and spatial heterogeneity |
Precise spatial delineations of growth patterns allowed us to study the spatial configuration of tumors as morphologically distinct pattern islands (Fig. 2a and Extended Data Figs. 2a,b and 3a). Acinar growth, often considered as an intermediate state during the transition of morphological patterns6,23, was also the mo... |
We used area and shape measured using pixel number and solidity index (Extended Data Fig. 7a) to represent the morphological features of individual acinar islands. Acinar island area and shape were notably different in tumors (≥5% of acinar) with different predominant patterns (TRACERx 421 n = 173; LATTICe-A n = 654; E... |
To investigate the spatial arrangement of acinar patterns, we developed an acinar scattering score that measured the degree of acinus dispersion. A low score indicated locally clustered acinar islands, while a high score implied a dispersion of acinar islands throughout the tissue (Extended Data Fig. 7f). Low acinar sc... |
## Discussion |
We have developed an AI method ANORAK for the precise classification of growth patterns in LUAD. To the best of our knowledge, this is the first AI method to dissect LUAD growth patterns at the pixel level and be tested in over 1,000 cases, setting a benchmark in automated grading of LUAD. Our method can automatically ... |
The AI method was evaluated in four cohorts, comprising a total of 1,372 tumors. The overall agreement of predominant pattern at the tumor level between AI and pathologists across four cohorts was moderate, which is consistent with the inter-pathologist agreement in the LATTICe-A and DHMC cohorts13. Similar results wer... |
The proposed IASLC grading system was originally introduced to improve prognostication using tumor morphology2. In our study, AI grading improved the performance of predicting DFS compared to the baseline and pathological grading for stage I tumors, and be comparable for stage I–III tumors. Moreover, the prognostic val... |
The LATTICe-A cohort, consisting of 845 tumors with scores from three pathologists, allowed a comprehensive investigation of the clinical impact of the AI method and showed its benefit as a morphological biomarker. This benefit was slightly higher than that brought by an additional manual grading for stage I tumors, an... |
The AI method enables the spatial profiling of growth patterns at the pixel level, allowing morphological and spatial heterogeneity analyses at the growth pattern island level. This would be unattainable with alternative manual or patch-wise classification methods. We used the area and solidity index to measure acinar ... |
This study has some limitations. The Dice coefficient of ANORAK is still limited, indicating that error modes exist. Intratumoral and tumor microenvironment heterogeneity may result in variations in growth pattern morphology, making segmentation more challenging, specifically among lepidic, papillary and acinar pattern... |
In summary, the AI method we developed can automate the predominant growth pattern and IASLC grading for LUAD tumors, achieving a moderate agreement with pathologists; this was validated in four cohorts consisting of 1,372 cases. In the TRACERx 421 and LATTICe-A cohort, AI grading was an independent prognostic indicato... |
## Methods |
### Study cohorts |
TRACERx is a multi-center, prospective study, which began recruitment in April 2014 (https://clinicaltrials.gov/ct2/show/NCT01888601, approved by an independent research ethics committee, ref. no. 13/LO/1546). Formalin-fixed paraffin-embedded and H&E-stained histopathology diagnostic slides were scanned using the NanoZ... |
The training set of the AI method consisted of 49 WSIs from 49 patients in the TRACERx 100 cohort28,29. The WSIs were sparsely annotated by three independent thoracic subspeciality pathologists, yielding 3,662 patches (768 × 768 pixels at ×20, approximately 0.45 μm per pixel) of annotations for six typical growth patte... |
The AI method was then applied and evaluated on a total of 5,540 WSIs from four cohorts, which were collected, processed and scanned independently. This included patients with invasive non-mucinous LUAD as primary diagnosis (excluding adenocarcinoma in situ, minimally invasive adenocarcinomas and other variants) from t... |
Manual pathological grading of growth patterns, as well as individual pattern proportion scoring, were available for the TRACERx 421, LATTICe-A and TCGA cohorts. The DHMC cohort only had predominant pattern data for each slide. In the LATTICe-A cohort, three independent consultant-level thoracic subspeciality pathologi... |
In the TRACERx 421 cohort, DFS was defined as the period from the date of registration to the time of radiological confirmation of the recurrence of the primary tumor registered for the TRACERx or the time of death by any cause. During the follow-up, three participants with LUAD (CRUK0512, CRUK0428 and CRUK0511) develo... |
In the LATTICE-A cohort, recurrence data were obtained from the examination of patient records, notably paper notes and radiological databases, to identify the date of radiologically or biopsy-confirmed recurrence. Cancer-specific death was determined by the presence of lung cancer in the cause of death in the death ce... |
### Deep learning model architecture |
We developed a deep learning-based model14 ANORAK which leveraged cross-stream interaction to recognize and segment six histological patterns (lepidic, acinar, papillary, micropapillary, cribriform and solid) on WSIs at the pixel level. The model applied ResNet50 (ref. 31) as the backbone with customized modifications ... |
### Implementation and evaluation |
Before training, the annotated tiles were divided into nonoverlapping patches, except for patches at the bottom and right edges, with a size of 768 × 768 pixels at ×20. During training, four data augmentation strategies were used to mitigate overfitting: random rotation within 90 degrees; random width-shift and height-... |
The ablation experiments at the patch level included comparisons with the baseline method (single-stream), multi-stream with the element-wise add combination (multi-ADD), multi-stream with first-order attention alone (multi-FO), multi-stream with second-order attention alone (multi-SO) and the proposed ANORAK model (mu... |
### Growth pattern and grading inference |
Each WSI was divided into tiles of 2,000 × 2,000 pixels with the magnification downsampled to ×20 (approximately 0.45 μm per pixel)29. Each tile was then normalized to a target image to align the color before feeding it to the well-trained deep learning model, which, in turn, generated corresponding masks for all growt... |
The predominant pattern and grading were inferred from a stitched and downsampled mask (approximately 7.2 μm per pixel). The growth pattern proportion for each tumor was computed as the proportion across all slides of a given tumor: |
where gj is a proportion for the j pattern, j represents lepidic, acinar, papillary, cribriform, micropapillary and solid, i is the i-th slide, m is the number of slides per tumor, n is the number of patterns and Sij is the number of pixels identified for the j pattern with the i-th slide. The predominant pattern, P, i... |
### Agreement between AI and pathological scores with regard to predominant patterns |
The strongest correlation for growth pattern proportion between the AI and manual estimates was observed for the solid pattern (TRACERx 421, rho = 0.79; LATTICe-A correlations against each pathologist’s scoring, rho1 = 0.80, rho2 = 0.77, rho3 = 0.78; TCGA, rho = 0.67; Fig. 2b and Supplementary Table 1), followed by aci... |
### C-index measuring prognostic ability |
We used the C-index to measure the prognostic ability of the survival models. Cox regression models were considered for predicting DFS; specifically, the baseline model included age, sex, tumor stage (excluded for stage I tumors as the stage information remains the same). The AI grading-based model included clinical ba... |
### Acinar morphological features |
The pixel number and solidity index, that is, the proportion of pixels in the convex hull that were also in a region of interest, were used to measure the individual acinar island area and shape generated by the AI method. A higher solidity index indicated a more regular shape. The average area and solidity index of al... |
### Acinar scattering score |
We adapted an established score, standard distance32, to measure the spatial distribution of acinar patterns, which we termed ‘acinar scattering’: |
where d is the standard distance, n is the number of isolated acinar islands within the tissue identified by the proposed AI method, N is the area of the tissue, (xi, yi) is the centroid of an acinar island and (x0, y0) is the mean center of all the acinar islands. |
A higher acinar scattering score indicated a more scattered distribution of acini across the tissue. The median value of all available slides for a given tumor was taken as the tumor-level score. The optimal cutoff (0.36) separating tumors into low-scattering and high-scattering groups was selected from the discovery c... |
In a univariable model, acinar scattering was prognostic of DFS for LATTICe-A in grade 2 and 3 tumors, respectively (grade 2, n = 212, P = 1.95 × 10−5, HR = 2.48, 95% CI = 1.63–3.76; grade 3, n = 570, P = 0.007, HR = 1.35, 95% CI = 1.08–1.68; Extended Data Fig. 8b,c), but not in grade 1 tumors (Extended Data Fig. 8a). ... |
### Statistics and reproducibility |
Correlation tests used Spearman’s method and were generated using the function cor.test from the stats v.4.1.2R package. Confusion matrices were obtained using the function confusionMatrix from the caret v.6.0-93R package. Fleiss’ kappa was computed to assess the agreement among observers using the function kappam.flei... |
### Reporting summary |
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. |
### Supplementary information |
### Source data |
## Acknowledgements |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.