Domain Composition and Attention for Unseen-Domain Generalizable Medical Image Segmentation

Pytorch implementation of our 'Domain Composition and Attention for Unseen-Domain Generalizable Medical Image Segmentation', which is accepted in MICCAI 2021.

Usage

1. Training StyleGAN

StyleGAN was implemented by following (https://github.com/NVlabs/stylegan2-ada-pytorch), and it was trained for each domain respectively, as image_gan.

2. Prepare Dataset.

Prepare the following files: image, label, image_gan.

 ├── images
 ├── label
 └── image_gan

3. Train the model.

1. Modify the configuration settings in settings.ini according to your requirements.
2. Run the training script:

python train_multi_fundus.py

4. Test the model.

1. Not using TTFA test model.

python test_multi_fundus.py

2. Using TTFA test models.

python test_multi_fundus_ttfa.py