Fig. 4From: EG-TransUNet: a transformer-based U-Net with enhanced and guided models for biomedical image segmentationIllustration of the proposed dilated self-attention convolution (DSA)Back to article page