Awesome

Generalizable Medical Image Segmentation via Random Amplitude Mixup and Domain Specific Image Restoration

Introduction

This repository is for our ECCV 2022 paper: Generalizable Medical Image Segmentation via Random Amplitude Mixup and Domain-Specific Image Restoration.

Pytorch Implementation

Clone this repository

git clone https://github.com/zzzqzhou/RAM-DSIR.git
cd RAM-DSIR

Download Dataset

Fundus

Download dataset Fundus (Provided by DoFE) and put images in ./dataset/fundus/

Prostate

Download our pre-processed dataset Prostate (Originally Provided by SAML) and put data in ./dataset/prostate/

Training and Testing

The training and testing process can all be done on one Nvidia RTX 2080Ti GPU with 11 GB memory.

Train on Fundus Dataset (Target Domain 0)

cd code
python -W ignore train.py --data_root ../dataset --dataset fundus --domain_idxs 1,2,3 --test_domain_idx 0 --ram --rec --is_out_domain --consistency --consistency_type kd --save_path ../outdir/fundus/target0 --gpu 0

Test on Fundus Dataset (Target Domain 0)

cd code
python -W ignore test_fundus_slice.py --model_file ../outdir/fundus/target0/final_model.pth --dataset fundus --data_dir ../dataset --datasetTest 0 --test_prediction_save_path ../results/fundus/target0 --save_result --gpu 0

Acknowledgement

Our implementation is heavily drived from Fed-DG and DoFE. Thanks to their great work.

Citation

If you find this project useful for your research, please consider citing:

@inproceedings{zhou2022ram_dsir,
  title={Generalizable Medical Image Segmentation via Random Amplitude Mixup and Domain Specific Image Restoration},
  author={Zhou, Ziqi and Qi, Lei and Shi, Yinghuan},
  booktitle={ECCV},
  year={2022}
}