Performance evaluation of brain tumor segmentation based on U-Net hybrid model

Medical image segmentation of brain tumors is a crucial aspect of medical image analysis with direct implications for diagnosis and treatment decisions. In this study, we systematically evaluated the performance of U-Net and its variants (Dense U-Net, DeepResU-Net) on the BraTS18 dataset. Experiments show that the DeepResU-Net model optimally segments healthy tissue, tumor cores, and enhanced tumors. This study provides an essential theoretical basis for automated segmentation of brain tumors, highlighting the importance of choosing an appropriate model architecture to improve segmentation performance.