Duo-SegNet: Adversarial Dual-Views for Semi-Supervised Medical Image Segmentation

This repo contains the supported pytorch code and configuration files to reproduce medical image segmentaion results of Duo-SegNet.

and denote Segmentation networks and Critic network. Here, Critic criticizes between prediction masks and the ground truth masks to perform the min-max game.

Environment

Please prepare an environment with python=3.8, and then run the command “pip install -r requirements.txt” for the dependencies.

Data Preparation

• For experiments we used three datasets:

  • Nuclei (2018 Data Science Bowl)
  • Spleen (Medical segmentation decathlon - MSD)
  • Heart ([Medical segmentation decathlon - MSD)

• File structure

     data
      ├── nuclei
      |   ├── train
      │   │   ├── image
      │   │   │   └── 00ae65...
      │   │   └── mask
      │   │       └── 00ae65...       
      ├── spleen
      ├── heart
      │   
      |
     Duo-SegNet
      ├──train.py
      ...

  • Use Med2Image to convert NIFTI to PNG.

Train/Test

• Train : Run the train script on nuclei dataset for 5% of labeled data.

  python train.py --dataset nuclei --ratio 0.05 --epoch 200

• Test : Run the test script on nuclei dataset.

  python test.py --dataset nuclei

Acknowledgements

This repository makes liberal use of code from Deep Co-training and pytorch-CycleGAN-and-pix2pix

References

• 2018 Data Science Bowl
• Medical segmentation decathlon

Citing Duo-SegNet

    @inproceedings{peiris2021duo,
      title={Duo-SegNet: Adversarial Dual-Views for Semi-Supervised Medical Image Segmentation},
      author={Peiris, Himashi and Chen, Zhaolin and Egan, Gary and Harandi, Mehrtash},
      booktitle={International Conference on Medical Image Computing and Computer-Assisted Intervention},
      pages={428--438},
      year={2021},
      organization={Springer}
    }