Scholarlink

<p>Posterior cruciate ligament (PCL) injuries often present subtle clinical signs and can be overlooked during standard diagnostic procedures, leading to missed or delayed treatment. Although magnetic resonance imaging (MRI) is highly accurate for acute injuries, it is expensive, time-consuming, has low sensitivity for chronic and graft tears, and is limited for patients with metallic implants. Using ultrasound imaging can clearly visualize the intra-articular PCL and holds great promise as an accessible, cost-effective, and real-time alternative with proven high accuracy in both acute and chronic injuries. However, its interpretation is operator-dependent and lacks streamlined quantitative assessment tools. We propose a novel deep learning-based framework for instantaneous, operator-independent quantification of PCL morphology directly from ultrasound images. We employ a Real-Time Detection Transformer that bypasses the need for complex segmentation models, instead deriving clinically meaningful metrics—PCL location, width, and angle through an object detection model and novel pre- and post-processing steps.</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

A novel deep learning based automatic ultrasonic posterior cruciate ligament clinical assessment tool

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年1月7日

查看全文
  • Jyun-Ping Kao,
  • Jiajing Zhang,
  • Wei-Ning Lee,
  • Chung-Ping Chen,
  • Lawrence Chun-Man Lau

摘要

Posterior cruciate ligament (PCL) injuries often present subtle clinical signs and can be overlooked during standard diagnostic procedures, leading to missed or delayed treatment. Although magnetic resonance imaging (MRI) is highly accurate for acute injuries, it is expensive, time-consuming, has low sensitivity for chronic and graft tears, and is limited for patients with metallic implants. Using ultrasound imaging can clearly visualize the intra-articular PCL and holds great promise as an accessible, cost-effective, and real-time alternative with proven high accuracy in both acute and chronic injuries. However, its interpretation is operator-dependent and lacks streamlined quantitative assessment tools. We propose a novel deep learning-based framework for instantaneous, operator-independent quantification of PCL morphology directly from ultrasound images. We employ a Real-Time Detection Transformer that bypasses the need for complex segmentation models, instead deriving clinically meaningful metrics—PCL location, width, and angle through an object detection model and novel pre- and post-processing steps.