This repository is the official implementation of Gradient-Aware for Class-Imbalanced Semi-supervised Medical Image Segmentation, ECCV2024.

Requirements:

• Python 3.7.11
• torch 1.10.0
• torchvision 0.11.0
• opencv-python 4.1.1.26
• numpy 1.20.3
• h5py 3.7.0
• scipy 1.7.1

Datasets

Dataset I Synapse. Following DHC, 20 samples were split for training, 4 samples for validation, and 6 samples for testing. We use the processed data by MagicNet.

Dataset II AMOS. The processed dataset can be downloaded via BaiduPan: https://pan.baidu.com/s/1TVZO_Ebx0t6helGOhuIF2Q password：zb96. Download and place the datasets in ./data/

Dataset III ACDC. We use the code and preprocessed data by SSLMIS.

Running

CUDA_VISIBLE_DEVICES=0 python train_Synapse_CPS.py --seed 1337 --labelnum 4

Reference

• MagicNet
• DHC

Citations

@inproceedings{qi2024gradient,
  title={Gradient-Aware for Class-Imbalanced Semi-supervised Medical Image Segmentation},
  author={Qi, Wenbo and Wu, Jiafei and Chan, Shing Chow},
  booktitle={European Conference on Computer Vision},
  pages={473--490},
  year={2024},
  organization={Springer}
}