github.com/AiEson/CrossMatch @main

CrossMatch

Code for this paper: CrossMatch: Enhance Semi-Supervised Medical Image Segmentation with Perturbation Strategies and Knowledge Distillation

🎉🎉🎉 This paper has been accepted by IEEE Journal of Biomedical and Health Informatics !

CrossMatch Paper:

Requirements

1. Create conda environment: bash conda create -n CrossMatch python=3.11
2. Clone the repo: bash git clone https://github.com/AiEson/CrossMatch.git
3. Activate the environment: bash conda activate CrossMatch
4. Install the requirements: bash cd CrossMatch pip install -r requirements.txt

Usage

LA dataset

One click to run:

cd LA/code
bash train.sh

ACDC dataset

One click to run:

cd ACDC
bash scripts/train.sh gpu_num port
# like `bash scripts/train.sh 4 12333` for 4 GPUs and port 12333

Results

LA dataset results

• The training set consists of 8 labeled scans and 72 unlabeled scans and the testing set includes 20 scans.

• The training set consists of 16 labeled scans and 64 unlabeled scans and the testing set includes 20 scans.

ACDC dataset results

• The training set consists of 3 labeled scans and 67 unlabeled scans and the testing set includes 20 scans.

Qualitative results

Citation

If you find this project useful, please consider citing:

@ARTICLE{CrossMatch,
  author={Zhao, Bin and Wang, Chunshi and Ding, Shuxue},
  journal={IEEE Journal of Biomedical and Health Informatics}, 
  title={CrossMatch: Enhance Semi-Supervised Medical Image Segmentation with Perturbation Strategies and Knowledge Distillation}, 
  year={2024},
  volume={},
  number={},
  pages={1-13},
  keywords={Perturbation methods;Data models;Predictive models;Biomedical imaging;Decoding;Accuracy;Training;Semi-supervised segmentation;Self-knowledge distillation;Image perturbation},
  doi={10.1109/JBHI.2024.3463711}}

Acknowledgement

• This code is adapted from UA-MT, DTC and UniMatch .
• We thank Lequan Yu, Xiangde Luo and Lihe Yang for their elegant and efficient code base.