UNet brain tumor image segmentation algorithm based on global and local feature fusion

The invention provides a UNet brain tumor image segmentation algorithm based on global and local feature fusion, and aims at brain tumor magnetic resonance imaging (MRI) multi-modal accurate segmentation. According to the algorithm, a local-global learning strategy is adopted, a UNet network improved by a gated axial attention module is used for forming a local branch of a brain tumor segmentation network model, the local branch pays attention to acquisition of detail information of brain tumors, and spatial information is rich. A UNet network improved by fusing Transform and ResNet modules is used for forming a global branch of a brain tumor segmentation network model, the global branch is large in receptive field and has rich semantic features, the Transform module is fused, the global modeling capacity is high, finally, obtained local and global images are put into a designed self-adaptive feature fusion module for image fusion, and the image fusion efficiency is improved. The adaptive fusion module calcul