Multimodal Brain Tumor Segmentation
The task for this project is to provide segmentation of gliomas in pre-operative MRI scans. We use the provided clinically-acquired training data from BraTS 2018 data set to produce segmentation labels.
In the project, I take inspiration from U-Net model architecture and modify the model and present it which can be seen in this link. For checking the success rate of segmentation I use Dice Coefficient & Dice Coefficient Loss Function.Â
The proposed model in this project has been used as a reference in the works like https://github.com/carinanorre/Brain-Tumour-Segmentation-Dissertation and https://info.cs.st-andrews.ac.uk/student-handbook/files/project-library/cs5098/dst1-Final_report.pdf
More details on the project can be found on GitHub
Below are some glimpses of the resultant prediction and actual data in test set:
