An Edge-aware Prompt-enhanced SAM for Ultrasound Image Segmentation

2026-07-08 · Source: Computer Vision and Pattern Recognition · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Computer Vision · Depth: Expert, quick

Summary

EP-SAM, an edge-aware and prompt-enhanced adaptation of the Segment Anything Model (SAM), is proposed to improve ultrasound image segmentation. While SAM excels on natural images, its performance on ultrasound data is often limited by poor boundary delineation. EP-SAM addresses this by leveraging multi-block feature extraction from its image encoder, enriching coarse-to-fine semantic representations. Additionally, edge-aware supervision of the image encoder enhances robustness against contour ambiguity and speckle noise, common issues in ultrasound. By integrating these complementary cues, EP-SAM generates high-quality prompts that effectively guide the model to target regions of interest. Experimental results on multiple benchmarks demonstrate EP-SAM consistently outperforms existing SAM-based methods for delineating anatomical structures and lesions.

Key takeaway

For Computer Vision Engineers developing medical imaging solutions, EP-SAM offers a significant advancement for ultrasound image segmentation. If you are struggling with SAM's performance on ambiguous boundaries or speckle noise in ultrasound data, consider integrating EP-SAM's multi-block feature extraction and edge-aware supervision. This approach can yield more accurate delineations of anatomical structures and lesions, enhancing diagnostic foundations in your applications.

Key insights

EP-SAM enhances SAM for ultrasound segmentation by integrating multi-block feature extraction and edge-aware supervision to improve boundary delineation and robustness.

Principles

• Multi-block features enrich semantic representations.
• Edge-aware supervision improves contour robustness.
• High-quality prompts guide segmentation effectively.

Method

EP-SAM adapts SAM by using multi-block feature extraction and applying edge-aware supervision to the image encoder, generating enhanced prompts for ultrasound image segmentation.

In practice

• Delineate anatomical structures in ultrasound.
• Segment lesions in medical imaging.
• Improve robustness to speckle noise.

Topics

• Ultrasound Image Segmentation
• Segment Anything Model
• EP-SAM
• Edge-aware Supervision
• Multi-block Features
• Medical Diagnostics

Best for: AI Scientist, Computer Vision Engineer, Research Scientist

Related on AIssential

• From Boundaries to Semantics: Prompt-Guided Multi-Task Learning for Petrographic Thin-section Segmentation — Petro-SAM enables high-quality joint grain-edge and lithology segmentation in petrographic images by adapting SAM.
• MNet++: Extended 2D/3D Networks for Anisotropic Medical Image Segmentation — Adaptive 2D-3D feature fusion and state-space models enhance MNet's anisotropic medical image segmentation reliability and reproducibility.
• SAM 3.1: Faster and More Accessible Real-Time Video Detection and Tracking With Multiplexing and Global Reasoning - AI at Meta — SAM 3.1 enhances real-time video object tracking via multiplexing, doubling processing speed and reducing GPU demands.
• Domain-Guided Prompting of the Segment Anything Model for Seismic Interpretation: The Role of Attributes, Visualization, and Hybrid Prompts — Zero-shot SAM adaptation for seismic interpretation is achieved by combining domain-guided visualization with hybrid prompting.
• A 3D SAM-Based Progressive Prompting Framework for Multi-Task Segmentation of Radiotherapy-induced Normal Tissue Injuries in Limited-Data Settings — A 3D SAM-based framework improves radiotherapy injury segmentation using progressive prompting and a small-target focus loss.
• Look-Closer-Then-Diagnose: Confidence-Aware Ultrasound VQA via Active Zooming — A novel VLM framework integrates lesion-focused reasoning and uncertainty-aware rewards for improved ultrasound diagnosis.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Computer Vision and Pattern Recognition.