Complementary consistency semi-supervised learning for 3D left atrial image segmentation

by Hejun Huang, Zuguo Chen*, Chaoyang Chen, Ming Lu, Ying Zou

Introduction

This repository is for our paper 'Complementary consistency semi-supervised learning for 3D left atrial image segmentation'.

Requirements

This repository is based on Pytorch 1.9.1, CUDA11.1 and Python 3.6.5

Usage

Install

Clone the repo:

git clone https://github.com/Cuthbert-Huang/CC-Net.git 

Dataset

We use the dataset of 2018 Atrial Segmentation Challenge. The processed h5 datas were provided in googleDrive and baiduNetdisk (password: cuth). Please unzip and put them in the data/LA folder.

Preprocess

If you want to process .nrrd data into .h5 data, you can use code/dataloaders/preprocess.py.

Pretrained models

The pretrained models were provided in googleDrive and baiduNetdisk (password: cuth). Please put them in the pretrained folder.

Train

If you want to train CC-Net for 10% labels on LA.

cd CC-Net
python ./code/train_ccnet_3d_v1.py --dataset_name LA --model ccnet3d_v1 --exp CCNET --labelnum 8 --gpu 0 --temperature 0.1 --max_iteration 10000

If you want to train CC-Net for 20% labels on LA.

cd CC-Net
python ./code/train_ccnet_3d_v1.py --dataset_name LA --model ccnet3d_v1 --exp CCNET --labelnum 16 --gpu 0 --temperature 0.1 --max_iteration 10000

Test

If you want to test CC-Net for 10% labels on LA.

cd CC-Net
python ./code/test.py --dataset_name LA --model ccnet3d_v1 --exp CCNET --labelnum 8 --gpu 0

If you want to test CC-Net for 20% labels on LA.

cd CC-Net
python ./code/test.py --dataset_name LA --model ccnet3d_v1 --exp CCNET --labelnum 16 --gpu 0

Citation

If our CC-Net is useful for your research, please consider citing:

@article{huang2023complementary,
  title = {Complementary consistency semi-supervised learning for 3D left atrial image segmentation},
  journal = {Computers in Biology and Medicine},
  volume = {165},
  pages = {107368},
  year = {2023},
  issn = {0010-4825},
  doi = {https://doi.org/10.1016/j.compbiomed.2023.107368},
  url = {https://www.sciencedirect.com/science/article/pii/S0010482523008338},
  author = {Hejun Huang and Zuguo Chen and Chaoyang Chen and Ming Lu and Ying Zou}
}

If you use the dataset of 2018 Atrial Segmentation Challenge, please consider citing:

@article{Xiong_A_global2021,
  author = {Xiong, Zhaohan and Xia, Qing and Hu, Zhiqiang and Huang, Ning and Bian, Cheng and Zheng, Yefeng and Vesal,          Sulaiman and Ravikumar, Nishant and Maier, Andreas and Yang, Xin},
  title = {A global benchmark of algorithms for segmenting the left atrium from late gadolinium-enhanced cardiac magnetic resonance imaging},
  journal = {Medical Image Analysis},
  year = {2021} }

Acknowledgements

Our code is origin from UAMT, SASSNet, DTC, and MC-Net+. Thanks to these authors for their excellent work.