The Pytorch implementation of "A 3D Cross-Modality Feature Interaction Network With Volumetric Feature Alignment for Brain Tumor and Tissue Segmentation" on the BrainTS2020 dataset.

Experiments were performed on an Ubuntu 18.04 workstation with two 24G NVIDIA GeForce RTX 3090 GPUs , CUDA 11.1, and install the virtual environment (python3.8) by:

pip install -r requirements.txt

The Ranger optimizer is used in this project, and install it by the following project:

https://github.com/lessw2020/Ranger-Deep-Learning-Optimizer

Download the BraTS2020 dataset and change data config:

vim ./src/path_config.py
# training dataset
BRATS_TRAIN_FOLDERS = "./xx/MICCAI_BraTS2020_TrainingData/"
# validation dataset
BRATS_VAL_FOLDER = "./xx/MICCAI_BraTS2020_ValidationData/"
# test dataset 
BRATS_TEST_FOLDER = "./xx/MICCAI_BraTS2020_Data_Testing/"

Run different models and change the '--fold' parameters to select the cross-validation fold (0 to 4)：

sh train_models.sh

Predicting the validation data by multi-model ensemble:

python step2_inference.py --devices 0 --on val --tta

If our projects are beneficial for your works, please cite:

@ARTICLE{9920184,
author={Zhuang, Yuzhou and Liu, Hong and Song, Enmin and Hung, Chih-Cheng},
journal={IEEE Journal of Biomedical and Health Informatics}, 
title={A 3D Cross-Modality Feature Interaction Network With Volumetric Feature Alignment for Brain Tumor and Tissue Segmentation}, 
year={2023},
volume={27},
number={1},
pages={75-86},
doi={10.1109/JBHI.2022.3214999}}

Our training codes is developed from the Top10 BraTS 2020 open sourced solution, please refer to their paper: https://arxiv.org/abs/2011.01045.