Brain-Tumor-Segmentation

Implementation of the BraTS 2019 winning model

BraTS 2018 dataset can be downloaded from this link here

Move the downloaded files to dataset directory

Move the pretrained weights to weights directory

To train the model or to load pretrained weights run train.py

Link to the paper:

https://arxiv.org/pdf/1810.11654#:~:text=In%20this%20work%2C%20we%20describe,won%20the%20BraTS%202018%20challenge.&text=In%20particular%2C%20EMMA%20combined%20DeepMedic,and%20ensembled%20their%20segmentation%20predictions.

Model architecture:

Results:

Dice Loss for Enhancing Tumor: 0.8145

Dice Loss for Whole Tumor: 0.9042

Dice Loss for Tumor Core: 0.8596