Traditional U-Net architectures rely heavily on local convolution operations, which struggle to model long-range spatial relationships essential for segmenting large or dispersed anatomical structures ...