C3R_UDA

Implementation for the paper entitled "C³R: Category Contrastive Adaptation and Consistency Regularization for Cross-Modality Medical Image Segmentation" (Accepted by ESWA).

Table of Contents

• C3R_UDA
  • Table of Contents
  • Environment requirements
  • Dataset
  • Project Tree
  • Training
  • Evaluation
  • Citation
  • Acknowledgement

Environment requirements

We test the code on 1 single V100-32GB GPU, or 4 TITAN-XP-12GB GPU.

Note that the code is tested on python 3.7.10, and the cuda version is 10.2.

imgaug==0.4.0
matplotlib==3.4.2
MedPy==0.3.0
numpy==1.20.3
Pillow==8.4.0
PyYAML==6.0
scikit_learn==1.0.2
tensorboardX==2.4.1
torch==1.3.1
torchsummary==1.5.1
torchvision==0.4.2
tqdm==4.46.0

Dataset

Some steps to prepare the dataset:

1. Download the preprocessed data from SIFA.
2. Decode the tfrecords data to npz data using the preprocessing code as:

    python preprocess.py --data_dir./data --save_dir./data

3. Please move the decoded data to the data folder.

Project Tree

C3R_UDA
├─ configs                # hyperparameters
│  └─ config_mr2ct.yml    
│  └─ config_ct2mr.yml    
├─ data                   # mmwhs dataset or other datasets
│  ├─ training_mr
│  │  ├─ xxxx.npz
│  │  └─ xxxx.npz
│  ├─ training_ct
│  │  ├─ xxxx.npz
│  │  └─ xxxx.npz
│  ├─ validation_mr
│  │  ├─ xxxx.npz
│  │  └─ xxxx.npz
│  ├─ validation_ct
│  │  ├─ xxxx.npz
│  │  └─ xxxx.npz
│  ├─ test_mr
│  │  ├─ xxxx.npz
│  │  └─ xxxx.npz
│  └─ test_ct
│  │  ├─ xxxx.npz
│  │  └─ xxxx.npz
├─ datasets               # data loader
│  ├─ __init__.py
│  ├─ augmentations.py
│  ├─ ct_dataset.py
│  └─ mr_dataset.py
├─ losses
│  ├─ __init__.py         
│  ├─ gan_loss.py
│  └─ seg_loss.py
├─ models                 # model architecture and training details
│  ├─ __init__.py
│  ├─ c3r_trainer.py
│  ├─ discriminator.py
│  ├─ res_parts.py
│  ├─ seg_model.py
│  └─ unet_parts.py
├─ utils
│  ├─ __init__.py
│  ├─ utils.py
│  └─ visualizer.py
├─ train.py               # training entry
└─ evaluate.py            # evaluation entry

Training

python train.py --config configs/config_mr2ct.yml  # for mr2ct
python train.py --config configs/config_ct2mr.yml  # for ct2mr

Evaluation

python run_eval.py --target ct  --model_path pretrained_mr2ct.pt  # for mr2ct
python run_eval.py --target mr  --model_path pretrained_ct2mr.pt  # for ct2mr

Citation

If you find the code useful for your research, please cite:

@article{ding2025c3r,
  title={C3R: Category contrastive adaptation and consistency regularization for cross-modality medical image segmentation},
  author={Ding, Shaodong and Liu, Ziyang and Liu, Pan and Zhu, Wanlin and Xu, Hong and Li, Zixiao and Niu, Haijun and Cheng, Jian and Liu, Tao},
  journal={Expert Systems with Applications},
  volume={269},
  pages={126304},
  year={2025},
  publisher={Elsevier}
}

Acknowledgement

• We thank the repo of SIFA, you can find the preprocessed data in SIFA.
• The style of the project organization is borrowed from CAG_UDA.