Semi-supervised Learning for Medical Image Segmentation (SSL4MIS)

Requirements

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.

Usage

1. Clone the repo:

git clone https://https://github.com/HiLab-git/SSL4MIS.git 
cd SSL4MIS

2. Download the processed data and put the data in ../data/BraTS2019 or ../data/ACDC, please read and follow the README.

3. Train the model

cd code
python train_XXXXX_3D.py or python train_XXXXX_2D.py or bash train_acdc_XXXXX.sh

4. Test the model

python test_XXXXX.py

Reimplemented methods

• Mean Teacher
• Entropy Minimization
• Deep Adversarial Networks
• Uncertainty Aware Mean Teacher
• Interpolation Consistency Training
• Uncertainty Rectified Pyramid Consistency
• Cross Pseudo Supervision

Acknowledgement

• Part of the code is adapted from open-source codebase and original implementations of algorithms, we thank these author for their fantastic and efficient codebase, such as, UA-MT, Attention-Gated-Networks and segmentatic_segmentation.pytorch .