Dual-Part Multi-Lateral Branched Network for Multi-Class Segmentation in Cardiovascular Catheterization Angiograms

MLBNet utilizes multi-lateral encoder blocks for repeated feature extraction alongside multi-head decoder branches to handle class-skewed structural properties. This dual-part architecture enables simultaneous multi-class segmentation of diverse cardiovascular structures in catheterization angiograms, moving beyond traditional binary segmentation methods to improve class-specific accuracy.