Pytorch implementation of our Weakly Supervised Medical Image Segmentation via Superpixel-guided Scribble Walking and Class-wise Contrastive Regularization (MICCAI 2023).

Weakly Supervised Medical Image Segmentation via Superpixel-guided Scribble Walking and Class-wise Contrastive Regularization MICCAI 2023

• The ACDC dataset with mask annotations can be downloaded from: ACDC.
• The Scribble annotations of ACDC can be downloaded from: Scribble.
• The data processing code Link the pre-processed ACDC data Link.

Some important required packages include:

• Pytorch version >=0.4.1.
• TensorBoardX
• Python >= 3.6
• Efficientnet-Pytorch pip install efficientnet_pytorch
• Some basic python packages such as Numpy, Scikit-image, SimpleITK, Scipy ......

Follow official guidance to install Pytorch.

1. Clone this project.

git clone https://github.com/Lemonzhoumeng/SC-Net cd SC-Net 

2. Data pre-processing or directly download the pre-processed data.

cd code python dataloaders/acdc_data_processing.py 

3. Superpixel-guided Scribble Walking to augment the scribble labels

python add_super.py 

Or download our augmented labels from Google drive.

4. Train the model

python train_superpixel_dual_contrastive.py --fold {} 

5. Test the model

python test_2D_contrastive_superpixel.py 

The code is modified from WSL4MIS.

If you use this codebase in your research, please cite the following paper:

@InProceedings{Zhou2023scnet,	author={Meng Zhou, Zhe Xu, Kang Zhou, Kai-yu Tong},	title={Weakly Supervised Medical Image Segmentation via Superpixel-guided Scribble Walking and Class-wise Contrastive Regularization},	booktitle={MICCAI},	year={2023}} 

• If you have any questions, feel free to contact Meng at (1155156866@link.cuhk.edu.hk)