Glioma Segmentation on 3D MRI Using 2D U-Net: A Study on the BraTS Dataset

Gliomas are highly aggressive brain tumors, and their precise segmentation in MRI scans is important for treatment planning. In this study, we employ a 2D U-Net model for automatic segmentation of brain tumors using the BraTS dataset. Our technique segments sub-regions such as the enhancing tumor, tumor core, and entire tumor from four MRI sequences (T1, T1CE, T2, FLAIR). The best-performing model achieved a mean Intersection over Union (IoU) of 81% and a Dice score of 65.5%, showing the viability of 2D U-Net for real-world neuroimaging applications.