Awesome

Personalizing Federated Medical Image Segmentation via Local Calibration

Introduction

This is an official release of the paper Personalizing Federated Medical Image Segmentation via Local Calibration, including the network implementation and the training scripts.

Personalizing Federated Medical Image Segmentation via Local Calibration, <br/> Jiacheng Wang, Yueming Jin, Liansheng Wang <br/> In: European Conference on Computer Vision (ECCV), 2022 <br/> [arXiv][Bibetex][Supp]

News

• [3/1 2023] Codes for Head Calibration are tuned.
• [7/12 2022] We have released the training codes.
• [7/25 2022] We have uploaded the test scripts.
• [7/12 2022] We have released the pre-print manuscript.
• [7/11 2022] We have released the pre-trained weights on the polyp segmentation.
• [7/4 2022] We have released the pre-processing scripts.
• [7/4 2022] We have created this repo.

Code List

• Network
• Pre-processing
• Training Codes
• Pretrained Weights

For more details or any questions, please feel easy to contact us by email (jiachengw@stu.xmu.edu.cn).

Usage

Dataset

In this paper, we perform the experiments using three imaging modalities, including the polyp images, fundus images, and prostate MR images. They could be downloaded from the public websites, or copied from FedDG and PraNet.

Pre-processing

After downloading the data resources, please run the file utils/prepare_dataset.py. Note that the file directory should be replaced with yours.

Training

Run the train script $ python scripts/train_lcfed.py.

Testing

Please download the pre-trained weights from Baidu Disk (https://pan.baidu.com/s/10HkQ90xeFcHMaNgfIyT0iw, a1sm) and put them in the project directory.

Rename the directory as logs/{dataset}/{exp_name}/model/.

Run the test script $ python scripts/test.py.

Result

The test IoU scores and ASSD scores on the Polyp dataset are:

Citation

If you find LC-Fed useful in your research, please consider citing:

@inproceedings{wang2022personalizing,
  title={Personalizing Federated Medical Image Segmentation via Local Calibration},
  author={Wang, Jiacheng and Jin, Yueming and Wang, Liansheng},
  booktitle={European Conference on Computer Vision},
  pages={456--472},
  year={2022},
  organization={Springer}
}