Problem
Research questions and friction points this paper is trying to address.
Ultrasound image segmentation
boundary delineation
Segment Anything Model
speckle noise
contour ambiguity
An Edge-aware Prompt-enhanced SAM for Ultrasound Image Segmentation
🤖 AI Summary
This work addresses the limitations of general-purpose segmentation models, such as SAM, in ultrasound imaging due to blurred boundaries and speckle noise. To overcome these challenges, the authors propose EP-SAM, which integrates multi-level feature extraction and edge-aware supervision into the SAM encoder for the first time. Additionally, a prompt enhancement strategy is designed to fuse complementary cues and guide precise segmentation. The proposed method significantly improves boundary localization accuracy and robustness to noise, outperforming existing SAM-based approaches across multiple ultrasound benchmark datasets. Consequently, EP-SAM achieves more accurate segmentation of both anatomical structures and pathological lesions in ultrasound images.
📝 Abstract
Ultrasound image segmentation is essential for delineating anatomical structures and lesions, providing the foundation for accurate diagnosis. While the Segment Anything Model (SAM) has demonstrated remarkable success on natural images, its performance on ultrasound data is often hindered by poor boundary delineation. To address this limitation, we propose EP-SAM, an edge-aware and prompt-enhanced adaptation of SAM. Specifically, we leverage multi-block feature extraction from the image encoder to enrich coarse-to-fine semantic representations, while edge-aware supervision of the image encoder improves robustness to contour ambiguity and speckle noise. By integrating these complementary cues, EP-SAM generates high-quality prompts that effectively guide the model toward target regions of interest. Experimental results on multiple benchmarks demonstrate that EP-SAM consistently outperforms existing SAM-based methods.
Innovation
Methods, ideas, or system contributions that make the work stand out.
edge-aware supervision
prompt enhancement
multi-block feature extraction
ultrasound image segmentation
Segment Anything Model
🔎 Similar Papers
No similar papers found.
💼 Related Jobs
