This is the code for MICCAI 2022 paper: "Multimodal Brain Tumor Segmentation Using Contrastive Learning based Feature Comparison with Monomodal Normal Brain Images". The training pipeline is motivated by nnUNet
If you find it useful in your own research, please cite our paper:
@inproceedings{liu2022multimodal,
title={Multimodal Brain Tumor Segmentation Using Contrastive Learning Based Feature Comparison with Monomodal Normal Brain Images},
author={Liu, Huabing and Nie, Dong and Shen, Dinggang and Wang, Jinda and Tang, Zhenyu},
booktitle={International Conference on Medical Image Computing and Computer-Assisted Intervention},
pages={118--127},
year={2022},
organization={Springer}
}
The following environments/libraries are required
- pytorch
- SimpleITK
- yacs
- scikit-learn
- medpy
- torchio
- For data preprocessing, please refer to nnUNet, the preprocessed data should be placed in DATA/preprocessed folder
- run train.py / test.py
- Note that, the reconstructed images are loaded and augmented as the extra modality before sent into the segmentation network, so you can see code like this in Segmentation/core/function.py:
x, rec = data[:, :-1], data[:, -1:]