File size: 2,752 Bytes
266cae6
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
---
library_name: diffusers
license: mit
pipeline_tag: image-to-image
tags:
- computed-tomography
- ct-reconstruction
- diffusion-model
- inverse-problems
- dm4ct
- sparse-view-ct
---

# Pixel Diffusion UNet – LoDoChallenge (DM4CT)

This repository contains the pretrained **pixel-space diffusion UNet** presented in the paper [DM4CT: Benchmarking Diffusion Models for Computed Tomography Reconstruction](https://huggingface.co/papers/2602.18589) (ICLR 2026).

- **Project Page:** [https://dm4ct.github.io/DM4CT/](https://dm4ct.github.io/DM4CT/)
- **Codebase:** [https://github.com/DM4CT/DM4CT](https://github.com/DM4CT/DM4CT)
- **Paper:** [DM4CT: Benchmarking Diffusion Models for Computed Tomography Reconstruction](https://huggingface.co/papers/2602.18589)

---

## 🔬 Model Overview

This model learns a **prior over CT reconstruction images** using a denoising diffusion probabilistic model (DDPM). It operates directly in **pixel space** (not latent space).

- **Architecture**: 2D UNet (Diffusers `UNet2DModel`)
- **Input resolution**: 512 × 512
- **Channels**: 1 (grayscale CT slice)
- **Training objective**: ε-prediction (standard DDPM formulation)
- **Noise schedule**: Linear beta schedule
- **Training dataset**: Low Dose Grand Challenge (LoDoChallenge)
- **Intensity normalization**: Rescaled to (-1, 1)

This model is intended to be combined with data-consistency correction for CT reconstruction tasks as detailed in the DM4CT benchmark.

---

## 📊 Dataset: Low Dose Grand Challenge

Source: [AAPM Low Dose CT Grand Challenge](https://www.aapm.org/grandchallenge/lowdosect/)

Preprocessing steps:
- Train/test split.
- Rescale reconstructed slices to (-1, 1).
- No geometry information is embedded in the model.

The model learns an unconditional image prior over medical CT slices.

---

## 🧠 Training Details

- **Optimizer**: AdamW
- **Learning rate**: 1e-4
- **Hardware**: NVIDIA A100 GPU
- **Training script**: [train_pixel.py](https://github.com/DM4CT/DM4CT/blob/main/train_pixel.py)

---

## 🚀 Usage

You can use this model with the `diffusers` library:

```python
from diffusers import DDPMPipeline

# Load the pretrained pipeline
pipeline = DDPMPipeline.from_pretrained("jiayangshi/lodochallenge_pixel_diffusion")

# Generate a sample (unconditional CT slice prior)
image = pipeline().images[0]
image.save("generated_ct_slice.png")
```

---

## Citation

```bibtex
@inproceedings{
  shi2026dmct,
  title={{DM}4{CT}: Benchmarking Diffusion Models for Computed Tomography Reconstruction},
  author={Shi, Jiayang and Pelt, Dani{\"e}l M and Batenburg, K Joost},
  booktitle={The Fourteenth International Conference on Learning Representations},
  year={2026},
  url={https://openreview.net/forum?id=YE5scJekg5}
}
```