Skip to content
/ CDMA Public

offical code for: Semi-supervised Pathological Image Segmentation via Cross Distillation of Multiple Attentions. MICCAI 2023.

Notifications You must be signed in to change notification settings

HiLab-git/CDMA

Repository files navigation

CDMA

official code for: Semi-supervised Pathological Image Segmentation via Cross Distillation of Multiple Attentions. MICCAI 2023, early accept arxiv. And the extension is published on the Pattern Recognition 2024.

Overall Framework

There are three branches based on different attention mechanisms and two losses in our framework overall

usage

First, split the dataset into train, val and test sets, then crop WSIs into patches for computational feasibility.

python utils.move_file.py
python slide_window.py

Then, just use the run.sh script to run the code.

sh run.sh

Data Acquisition

The DigestPath dataset can be downloaded in: DigestPath

The dataset dir is like this after splitting and cropping:

digestpath2019
-----tissue-train-100
-----tissue-train-100-patch
-----tissue-train-5
-----tissue-train-5-patch
-----tissue-val
-----tissue-val-patch
-----tissue-test

You can get data lists in data/digestpath

Citation

@inproceedings{zhong2023semi,
  title={Semi-supervised Pathological Image Segmentation via Cross Distillation of Multiple Attentions},
  author={Zhong, Lanfeng and Liao, Xin and Zhang, Shaoting and Wang, Guotai},
  booktitle={International Conference on Medical Image Computing and Computer-Assisted Intervention},
  pages={570--579},
  year={2023},
  organization={Springer}
}

@article{zhong2024semi,
  title={Semi-supervised pathological image segmentation via cross distillation of multiple attentions and Seg-CAM consistency},
  author={Zhong, Lanfeng and Luo, Xiangde and Liao, Xin and Zhang, Shaoting and Wang, Guotai},
  journal={Pattern Recognition},
  pages={110492},
  year={2024},
  publisher={Elsevier}
}

Acknowledgement

The code of semi-supervised learning framework is borrowed from SSL4MIS

About

offical code for: Semi-supervised Pathological Image Segmentation via Cross Distillation of Multiple Attentions. MICCAI 2023.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published