208-generalizable-cross-modality-medical-image-segmentation-via-style-augmentation-and-dual-normaliz

https://github.com/szu-advtech-2023/208-generalizable-cross-modality-medical-image-segmentation-via-style-augmentation-and-dual-normaliz

Repository

https://github.com/SZU-AdvTech-2023/208-Generalizable-Cross-Modality-Medical-Image-Segmentation-via-Style-Augmentation-and-Dual-Normaliz/blob/main/

~~# Generalizable Cross-modality Medical Image Segmentation via Style Augmentation and Dual Normalization~~
by [Ziqi Zhou](https://zzzqzhou.github.io/), [Lei Qi](http://palm.seu.edu.cn/qilei/), [Xin Yang](https://xy0806.github.io/), Dong Ni, [Yinghuan Shi](https://cs.nju.edu.cn/shiyh/index.htm). 

## Introduction

This repository is for our CVPR 2022 paper '[Generalizable Cross-modality Medical Image Segmentation via Style Augmentation and Dual Normalization](https://arxiv.org/abs/2112.11177)'. 


![](./picture/cvpr22_dn.PNG)

## Data Preparation

### Dataset
[BraTS 2018](https://www.med.upenn.edu/sbia/brats2018/data.html) | [MMWHS](http://www.sdspeople.fudan.edu.cn/zhuangxiahai/0/mmwhs/) | [Abdominal-MRI](https://chaos.grand-challenge.org/) | [Abdominal-CT](https://www.synapse.org/#!Synapse:syn3193805/wiki/217789)

### File Organization

T2 as source domain
``` 
 [Your BraTS2018 Path]
     npz_data
         train
             t2_ss
                 sample1.npz, sample2.npz, xxx
             t2_sd
         test
             t1
                 test_sample1.npz, test_sample2.npz, xxx
             t1ce
             flair
```

## Training and Testing

Train on source domain T2.
```
python -W ignore train_dn_unet.py \
  --train_domain_list_1 t2_ss --train_domain_list_2 t2_sd --n_classes 2 \
  --batch_size 16 --n_epochs 50 --save_step 10 --lr 0.001 --gpu_ids 0 \
  --result_dir ./results/unet_dn_t2 --data_dir [Your BraTS2018 Path]/npz_data
```

Test on target domains (T1, T1ce and Flair).

```
python -W ignore test_dn_unet.py \
  --test_domain_list t1 t1ce flair --model_dir ./results/unet_dn_t2/model
  --batch_size 32 --save_label --label_dir ./vis/unet_dn_t2 --gpu_ids 0 \
  --num_classes 2 --data_dir [Your BraTS2018 Path]/npz_data
```

## Acknowledgement
The U-Net model is borrowed from [Fed-DG](https://github.com/liuquande/FedDG-ELCFS). The Style Augmentation (SA) module is based on the nonlinear transformation in [Models Genesis](https://github.com/MrGiovanni/ModelsGenesis). The Dual-Normalizaiton is borrow from [DSBN](https://github.com/wgchang/DSBN). We thank all of them for their great contributions.

## Citation

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

```bibtex
@inproceedings{zhou2022dn,
  title={Generalizable Cross-modality Medical Image Segmentation via Style Augmentation and Dual Normalization},
  author={Zhou, Ziqi and Qi, Lei and Yang, Xin and Ni, Dong and Shi, Yinghuan},
  booktitle={CVPR},
  year={2022}
}
```

Owner

GitHub Events