U-Net Variants for Brain Tumor Segmentation: Performance and Limitations

Brain Tumor (BT) segmentation is a vital aspect of medical imaging essential for precise identification and subsequent treatment. Deep Learning (DL)-based methods, including U-Net and its variations, have demonstrated potential in BT segmentation. This paper reviews recent research on U-Net and its variants for BT segmentation, highlighting their strengths and limitations. The study also summarizes recent results from different BraTS datasets, which establish the effectiveness of U-Net in segmentation tasks. The high modularity and versatility of U-Net make it an indispensable tool in the medical imaging community, widely adopted for various segmentation tasks. The study's findings will benefit medical image analysis researchers, practitioners, and those interested in DL techniques for image segmentation.