CereTS: [Unsupervised Domain Adaptation for Cross-Modality Cerebrovascular Segmentation]

This repository provides a PyTorch implementation of our work. The paper is published on IEEE JBHI.

Overview

• Model: A novel unsupervised domain adaptation (UDA) framework with image-level, patch-level and feature-level alignments
• Task: Cross-Modality Cerebrovascular Segmentation
• Data: TOF-MRA and CTA

Updates

• 2023.08: Code released.
• 2024.02.06: Repo improved.
• The complete code and tutorial will be released soon.

1. Data preparation

• The whole dataset consists of two parts, CTA dataset and TOF-MRA dataset. The TOF-MRA dataset comes from the IXI Dataset, collected from the Institute of Psychiatry. The CTA dataset is private.
• If you want to use the data, please contact us ([email protected]). We will share the CTA dataset after you fill out a questionnaire on data usage intention. Next we will update this process.

2. Environment

• Please prepare an environment with python=3.8, and then use the command "pip install -r requirements.txt" for more dependencies.

3. Train

If you have already arranged your data, you can start training your model.

cd "/home/...  .../CereTS/"
python train.py -name <your path to save>

4. Test

After finishing training, you can start testing your model.

python test.py -ckpt_path <your checkpoint path> -save_path <your save path> -target_npy_dirpath <your input path>

Citation

If our paper or code is helpful to you, please consider citing our paper:

Y. Wang, C. Meng, Z. Tang, X. Bai, P. Ji and X. Bai, "Unsupervised Domain Adaptation for Cross-Modality Cerebrovascular Segmentation," in IEEE Journal of Biomedical and Health Informatics, doi: 10.1109/JBHI.2024.3523103.