This repository is the official implementation of the CIBM paper Cross-Domain Attention-Guided Generative Data Augmentation for Medical Image Analysis with Limited Data.

To install the necessary requirements:

pip install -r requirements.txt

To download the necessary datasets, visit the following links:

• BraTS 2020
• TCIA

To train the CDAGAN model, execute those commands:

cd CDAGAN
bash train_cda.sh

To generate data for data augmentation on medical image analysis tasks based on the trained CDAGAN model, execute those commands:

cd CDAGAN
bash infer_cda.sh

To perform data augmentation on classification tasks, execute those commands:

cd ResNet
bash classification.sh 

To evaluate the performance of the trained classification model, execute those commands:

cd ResNet
python infer.py

To perform data augmentation on segmentation tasks, execute those commands:

cd UNet
python train.py

To evaluate the performance of the trained segmentation model, execute those commands:

cd UNet
python test.py

If you find CDAGAN helpful for your research, please consider citing:

@article{xu2023cross,
title={Cross-domain attention-guided generative data augmentation for medical image analysis with limited data},
author={Xu, Zhenghua and Tang, Jiaqi and Qi, Chang and Yao, Dan and Liu, Caihua and Zhan, Yuefu and Lukasiewicz, Thomas},
journal={Computers in Biology and Medicine},
pages={107744},
year={2023},
publisher={Elsevier}
}

For any issues or queries, please contact chang.qi@student.tuwien.ac.at.