This repo is the official implementation for: 1) TransBTS: Multimodal Brain Tumor Segmentation Using Transformer.
2) TransBTSV2: Towards Better and More Efficient Volumetric Segmentation of Medical Images.
The details of the our TransBTS and TransBTSV2 can be found at the models directory (TransBTS and TransBTSV2) in this repo or in the original paper.
The multimodal brain tumor datasets (BraTS 2019 & BraTS 2020) could be acquired from here.
The liver tumor dataset LiTS 2017 could be acquired from here.
The kidney tumor dataset KiTS 2019 could be acquired from here.
After downloading the dataset from here, data preprocessing is needed which is to convert the .nii files as .pkl files and realize data normalization.
python3 preprocess.py
Run the training script on BraTS dataset. Distributed training is available for training the proposed TransBTS, where --nproc_per_node decides the numer of gpus and --master_port implys the port number.
python3 -m torch.distributed.launch --nproc_per_node=4 --master_port 20003 train.py
If you want to test the model which has been trained on the BraTS dataset, run the testing script as following.
python3 test.py
After the testing process stops, you can upload the submission file to here for the final Dice_scores.
If you use our code or models in your work or find it is helpful, please cite the corresponding paper:
TransBTS:
@inproceedings{wang2021transbts,
title={TransBTS: Multimodal Brain Tumor Segmentation Using Transformer},
author={Wang, Wenxuan and Chen, Chen and Ding, Meng and Yu, Hong and Zha, Sen and Li, Jiangyun},
booktitle={Medical Image Computing and Computer Assisted Intervention--MICCAI 2021: 24th International Conference, Strasbourg, France, September 27--October 1, 2021, Proceedings, Part I 24},
pages={109--119},
year={2021},
organization={Springer}
}TransBTSV2:
@article{li2022transbtsv2,
title={TransBTSV2: Wider Instead of Deeper Transformer for Medical Image Segmentation},
author={Li, Jiangyun and Wang, Wenxuan and Chen, Chen and Zhang, Tianxiang and Zha, Sen and Yu, Hong and Wang, Jing},
journal={arXiv preprint arXiv:2201.12785},
year={2022}
}