Scholarlink

Abstract <p>Ultrasound (US) imaging enables the evaluation of vascular structures in real time, and it can provide morphological and pathological information during US-guided procedures. Automatic prediction of vascular structure boundaries can help clinicians in locating and measuring target structures more accurately and efficiently. Most existing US segmentation methods use per-pixel classification or regression, which require post-processing to obtain contour coordinates. In this work, we present ACE-Net, a novel approach that directly predicts the contour coordinates for every scanning line (A-line) in US images. ACE-Net combines two main modules: a boundary regression module that predicts the upper and lower coordinates of the target area for each A-line, and an A-line classification module that determines whether an A-line belongs to the target area or not. We evaluated our method on three clinical US datasets using, among others, dice similarity coefficient (DSC) and inference time as performance metrics. Our method outperformed state-of-the-art segmentation methods in inference time while achieving superior or comparable performance in DSC. ACE-Net is publicly available at <a href="https://github.com/bfarolabarata/ace-net">https://github.com/bfarolabarata/ace-net</a>.</p> Graphical abstract <p>ACE-Net is a deep CNN architecture designed for direct contour segmentation of vascular structures in ultrasound images. It formulates segmentation as a boundary regression process combined with the prediction of vascular structure presence. To achieve this, ACE-Net introduces an A-line-based encoding scheme, representing contour coordinates of each vascular structure as two boundary vectors (upper and lower). The network consists of a backbone module, an A-line feature extractor, a multi-scale fusion module, and two final head encoders, which output boundary coordinates and presence probabilities. By leveraging a presence vector, ACE-Net enhances segmentation accuracy and effectively handles common ultrasound artifacts without requiring pre-processing. Its architecture is specifically designed to address the imbalance between lateral and axial information in ultrasound images, and it enables real-time clinical metric computation.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

ACE-Net: A-line coordinates encoding network for vascular structure segmentation in ultrasound images

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年4月2日

查看全文
  • Beatriz Farola Barata,
  • Guiqiu Liao,
  • Gianni Borghesan,
  • Keir McCutcheon,
  • Johan Bennett,
  • Benoit Rosa,
  • Michel de Mathelin,
  • Florent Nageotte,
  • Michalina J. Gora,
  • Jos Vander Sloten,
  • Emmanuel Vander Poorten,
  • Diego Dall’Alba

摘要

Abstract

Ultrasound (US) imaging enables the evaluation of vascular structures in real time, and it can provide morphological and pathological information during US-guided procedures. Automatic prediction of vascular structure boundaries can help clinicians in locating and measuring target structures more accurately and efficiently. Most existing US segmentation methods use per-pixel classification or regression, which require post-processing to obtain contour coordinates. In this work, we present ACE-Net, a novel approach that directly predicts the contour coordinates for every scanning line (A-line) in US images. ACE-Net combines two main modules: a boundary regression module that predicts the upper and lower coordinates of the target area for each A-line, and an A-line classification module that determines whether an A-line belongs to the target area or not. We evaluated our method on three clinical US datasets using, among others, dice similarity coefficient (DSC) and inference time as performance metrics. Our method outperformed state-of-the-art segmentation methods in inference time while achieving superior or comparable performance in DSC. ACE-Net is publicly available at https://github.com/bfarolabarata/ace-net.

Graphical abstract

ACE-Net is a deep CNN architecture designed for direct contour segmentation of vascular structures in ultrasound images. It formulates segmentation as a boundary regression process combined with the prediction of vascular structure presence. To achieve this, ACE-Net introduces an A-line-based encoding scheme, representing contour coordinates of each vascular structure as two boundary vectors (upper and lower). The network consists of a backbone module, an A-line feature extractor, a multi-scale fusion module, and two final head encoders, which output boundary coordinates and presence probabilities. By leveraging a presence vector, ACE-Net enhances segmentation accuracy and effectively handles common ultrasound artifacts without requiring pre-processing. Its architecture is specifically designed to address the imbalance between lateral and axial information in ultrasound images, and it enables real-time clinical metric computation.