[MICCAI 2025] ConStyX

This repo is the PyTorch implementation of our paper:

"ConStyX: Content Style Augmentation for Generalizable Medical Image Segmentation"

Content Style Augmentation (ConStyX)

Usage

🔥🔥 Code for generalizable medical image segmentation with ConStyX. 🔥🔥

1. Data Preparation

Download the datasets from this link: OD/OC Segmentation
Then, the dataset is arranged in the following format:

dataset/
|-- ORIGA
|   |-- train
|   |   |-- images
|   |   |-- mask
|   |-- test
|   |   |-- image
|   |   |-- mask
...

2. OD/OC Segmentation

We take the scenario using BinRushed (target domain) and other four datasets (source domains) as the example.

# Training
CUDA_VISIBLE_DEVICES=0 python main.py --mode train_DG --num_epochs 100 --Source_Dataset BinRushed
# Test
CUDA_VISIBLE_DEVICES=0 python test.py --mode multi_test --load_time TIME_OF_MODEL --Source_Dataset BinRushed

Citation

If you find this project useful, please consider citing:

@inproceedings{chen2025constyx,
  title={Constyx: Content style augmentation for generalizable medical image segmentation},
  author={Chen, Xi and Shen, Zhiqiang and Cao, Peng and Yang, Jinzhu and Zaiane, Osmar R},
  booktitle={International Conference on Medical Image Computing and Computer-Assisted Intervention},
  pages={100--110},
  year={2025},
  organization={Springer}
}

Acknowledgements

Part of the code is revised from the Pytorch implementation of TriD

Thanks to the authors for providing the processed data.