A Hierarchical Aggregation Network with Edge Guidance for Sturdy Medical Image Segmentation

Convolutional neural networks (CNNs) are essential for medical picture segmentation in clinical analysis and illness detection. On the other hand, noise levels and intraclass variability are considerable in medical photos, but interclass differences are negligible. It is difficult to extract strong contextual information and combine discriminative characteristics for fine-grained segmentation. Current techniques are unable to produce precise mask edges, resulting in fuzzy borders and decreased segmentation precision. In this research, a new A Hierarchical Aggregation Network with Edge Guidance that maintains precise spatial features and gathers rich contextual information is presented. The Efficient Fusion Attention Module is used by the proposed model to highlight important spatial and channel information, the Adaptive Hierarchical Feature Aggregation Module to maximize feature fusion, and the Inter-layer Edge-aware Module to improve edge prediction accuracy. Experimentally and qualitatively, the proposed method will perform better than six other cutting-edge segmentation techniques.