The codes for the work "A novel 3D unsupervised domain adaptation framework for cross-modality medical image segmentation"(https://ieeexplore.ieee.org/abstract/document/9741336). Meanwhile, our work is the 5th solution in CrossMoDA 2021 challenge (https://arxiv.org/pdf/2201.02831.pdf), which did not utilize multiple nnUnet to ensemble results.
- Please go to "./datasets/README.md" for details.
- Please prepare an environment with python=3.7, and then use the command "pip install -r requirements.txt" for the dependencies.
- Train stage one image-to-image translation model for style transfer
python stage_1_i2i_train.py --name sourceAtotargetB
- Generate target-like source domain images
python stage_1.5_i2i_inference.py --ckpt_path YOUR_PATH --source_npy_dirpath SOURCE_PATH --target_npy_dirpath TARGET_PATH --save_npy_dirpath SAVE_PATH --k_means_clusters 6
- Train stage two DAR-UNET model for semantice segmentation
python stage_2_seg_train.py --name experiment --train_dataroot SOURCE2TARGET_PATH --val_dataroot TARGET_VAL_PATH --num_classes NUM_CLASS --epoch_max 100
If you have any questions, please send us email {kai.yao19, zixian.su20}@student.xjtlu.edu.cn.
If you find our work is useful, please cite our work.
@article{yao2022darunet,
title={A novel 3D unsupervised domain adaptation framework for cross-modality medical image segmentation},
author={Yao, Kai and Su, Zixian and Huang, Kaizhu and Yang, Xi and Sun, Jie and Hussain, Amir and Coenen, Frans},
journal={IEEE Journal of Biomedical and Health Informatics},
year={2022},
publisher={IEEE}
}
@article{dorent2022crossmoda,
title={CrossMoDA 2021 challenge: Benchmark of Cross-Modality Domain Adaptation techniques for Vestibular Schwnannoma and Cochlea Segmentation},
author={Dorent, Reuben and Kujawa, Aaron and Ivory, Marina and Bakas, Spyridon and Rieke, Nicola and Joutard, Samuel and Glocker, Ben and Cardoso, Jorge and Modat, Marc and Batmanghelich, Kayhan and others},
journal={arXiv preprint arXiv:2201.02831},
year={2022}
}