AG-HFFNet: Attention-Guided Hierarchical Feature Fusion for Ultrasound Image Segmentation

Abstract

Accurate ultrasound image segmentation is essential for computer-aided diagnosis, enabling reliable localization and characterization of anatomical structures and lesions. However, it remains challenging due to speckle noise, low contrast, and complex structural variations that hinder precise feature extraction. To address these challenges, we propose AG-HFFNet, an attention-guided hierarchical feature fusion network for ultrasound image segmentation. The proposed framework employs an EfficientNet-B5 encoder to extract multi-scale feature representations, which are further refined through a Block-wise Feature Enhancement Layer (BWFEL) incorporating dilated convolutions along with channel and spatial attention mechanisms. In addition, an attention-guided hierarchical feature fusion module progressively aggregates features in a top-down manner, enabling effective interaction between low-level spatial details and high-level semantic information. Extensive experiments on two benchmark ultrasound datasets, BUSI and DDTI, demonstrate that AG-HFFNet consistently outperforms state-of-the-art methods in terms of Dice score and Intersection over Union (IoU), highlighting its effectiveness and generalization capability for robust ultrasound image segmentation.