EEMS: Edge-Prompt Enhanced Medical Image Segmentation Based on Learnable Gating Mechanism

TL;DR

EEMS tackles the challenge of precise medical image segmentation under edge ambiguity and noise by concurrently enhancing edges and guiding segmentation with multi-scale prompts. It introduces two main components, the Edge-Aware Enhancement Unit and the Multi-scale Prompt Generation Unit, whose outputs are adaptively fused by a learnable Dual-Source Adaptive Gated Fusion Unit. Key innovations include the Frequency Feature Combination and Enhancement, Multi-Scale Channel Attention Feature Aggregation, and Deformable Feature Refinement within EAEU, plus MSFF and PGFI in MSPGU for robust prompt guidance. Evaluations on ISIC2018, Kvasir-SEG, Monu-Seg, and COVID-19 CT datasets show state-of-the-art performance and improved boundary delineation, demonstrating strong potential for clinical deployment.

Abstract

Medical image segmentation is vital for diagnosis, treatment planning, and disease monitoring but is challenged by complex factors like ambiguous edges and background noise. We introduce EEMS, a new model for segmentation, combining an Edge-Aware Enhancement Unit (EAEU) and a Multi-scale Prompt Generation Unit (MSPGU). EAEU enhances edge perception via multi-frequency feature extraction, accurately defining boundaries. MSPGU integrates high-level semantic and low-level spatial features using a prompt-guided approach, ensuring precise target localization. The Dual-Source Adaptive Gated Fusion Unit (DAGFU) merges edge features from EAEU with semantic features from MSPGU, enhancing segmentation accuracy and robustness. Tests on datasets like ISIC2018 confirm EEMS's superior performance and reliability as a clinical tool.

Figures (4)

• Figure 1: Effect of segmentation models applied to different body parts
• Figure 2: EEMS Model Architecture Diagram
• Figure 3: EEMS Model Visual Segmentation Results
• Figure 4: Ablation experiment results of EEMS on ISIC2018 dataset