Discover BEFUnet, a hybrid CNN-Transformer architecture that promises precise medical image segmentation. Embracing CNN’s success and addressing their limitations, this network merges body and edge information with innovative modules:

• Local Cross-Attention Feature (LCAF) for spatially-focused edge and body feature fusion
• Double-Level Fusion (DLF) module for enhanced feature integration
• Dual-branch encoder with an edge encoder and a body encoder to extract comprehensive semantic information

In-depth evaluations reveal BEFUnet’s superior performance across various datasets:

• Superior segmentation accuracy
• Efficient computational complexity
• Scalability across shapes, scales, and textures
• Compatibility with existing segmentation frameworks

BEFUnet is a significant stride towards advanced medical diagnostics. Its design not only demonstrates the synergy between CNNs and Transformers but also opens a pathway for future research in long-range relational features within medical imagery.