by Quande Liu, Hongzheng Yang, Qi Dou, Pheng-Ann Heng.

Edited by Yasin Tarakçı

Pytorch implementation for MICCAI 2021 paper "Federated Semi-supervised Medical Image Classification via Inter-client Relation Matching"

1. Clone the repository and go to the repository root directory:

   git clone https://github.com/ysntrkc/FedIRM.git
   cd FedIRM
   

2. Create a conda environment and activate it:

   conda env create -f environment.yml
   conda activate fedIRM
   

3. Download datasets from following sources:

   1. ISIC 2019 (Kaggle)
   2. HAM10000 (Kaggle)
   3. RSNA Intracranial Hemorrhage Detection (drive)

4. Extract the datasets to data/ folder and place the content into folders like this:

   1. ISIC 2019: data/isic2019/
   2. HAM10000: data/ham10000/
   3. RSNA Intracranial Hemorrhage Detection: data/rsna/

5. Go to data folder from terminal with cd data and run python prepare_data.py to prepare the datasets.

1. Go to the src folder with cd src and run python main.py to train the model.

2. You can see the different options in src/options.py and change the parameters as you want.

3. You can see the example run commands in src/example_run.sh file.

If this repository is useful for your research, please cite:

@article{liu2021federated,
  title={Federated Semi-supervised Medical Image Classification via Inter-client Relation Matching},
  author={Liu, Quande and Yang, Hongzheng and Dou, Qi and Heng, Pheng-Ann},
  journal={International Conference on Medical Image Computing and Computer Assisted Intervention},
  year={2021}
}

Please contact '[email protected]' or '[email protected]'

If you have questions about this version of the code, you can also contact '[email protected]'