haibinswe / mtcl

[MICCAI21, TMI22] Mean-Teacher-Assisted Confident Learning

Mean-Teacher-Assisted Confident Learning (MTCL)

• Implementation of our work Mean-Teacher-assisted Confident Learning for learning segmentation from mixed-quality labeled data.
• Note that, our label denoising scheme aims at the binary task.
• [Note] This's an initial report, code will be further re-organized but the core implementation is included that can be easily adapted to your own application.

Citation

If our work brings insights to you, or you use the codebase, please cite our paper as:

@article{xu2022anti,
  title={Anti-interference from Noisy Labels: Mean-Teacher-assisted Confident Learning for Medical Image Segmentation},
  author={Xu, Zhe and Lu, Donghuan and Luo, Jie and Wang, Yixin and Yan, Jiangpeng and Ma, Kai and Zheng, Yefeng and Tong, Raymond Kai-yu},
  journal={IEEE Transactions on Medical Imaging},
  year={2022},
  publisher={IEEE}
}

@artical{xu2021noisylabel,
  title={Noisy Labels are Treasure: Mean-Teacher-Assisted Confident Learning for Hepatic Vessel Segmentation},
  author={Zhe Xu, Donghuan Lu, Yixin Wang, Jie Luo, Jagadeesan Jayender, Kai Ma, Yefeng Zheng and Xiu Li},
  booktitle={International Conference on Medical Image Computing and Computer Assisted Intervention},
  year={2021}
}

Acknowledgement

The scripts are mainly based on the project SSL4MIS and the API cleanlab.