使用卷积神经网络的超声图像对中枢神经和横带的形态表征
Jocelyn L Hawk1, Shalon Walter1, Xiaoxiao Sun2
1Hand Research Laboratory, Department of Orthopaedic Surgery, University of Arizona, AZ, USA.
Medical engineering & physics
|August 20, 2025
概括
在超声波图像中,卷积神经网络 (CNN) 准确地细分了中枢神经和横带. 这种人工智能方法可靠地量化了手腕道解剖, 显示了手腕道综合征的诊断潜力.
科学领域:
- 医学成像分析
- 医疗保健中的人工智能
- 神经学和骨科
背景情况:
- 中枢神经和横带 (TCL) 的准确细分对于诊断手腕道综合征 (CTS) 是至关重要的.
- 超声波图像的手动细分是耗时的,并且受观察者之间的变化影响.
- 使用深度学习的自动化方法为客观和高效的分析提供了潜在的解决方案.
研究的目的:
- 开发和评估卷积神经网络 (CNN) 用于在超声波图像中自动细分和量化中枢神经和手臂.
- 通过CNN模型对健康个体和手腕道综合征患者的解剖学测量进行比较.
主要方法:
- 基于U-Net的CNN模型用于从远端手腕道超声波图像分离中枢神经和TCL.
- 模型的性能是使用Dice分数系数 (DSC),回忆和精度进行评估的.
- 使用统计测试 (Wilcoxon签名等级测试) 和可靠性指标 (平均误差,类内系数) 来比较手动和自动细分的形态参数.
主要成果:
- 在健康受试者中,CNN对中枢神经 (DSC高达0.89) 和TCL (DSC高达0.87) 实现了高分段精度.
- 在手腕道综合征患者中,细分性能略低,但仍然强.
- 自动测量显示与手动细分非常一致,手腕道综合征患者的中枢神经截面面积和手腕高度显著增加.
结论:
- 从超声波图像可以自动细分中枢神经和TCL.
- 开发的模型提供了可靠的手腕道解剖学量化,突出了手腕道综合征的潜在诊断价值.
相关概念视频
Spinal Nerves: Plexus I
1.3K
Nerve plexuses are networks of interlacing nerves that serve as communication hubs to distribute and organize nerve action across various body regions. The nerve plexuses are organized into the cervical plexus located in the neck region, brachial plexus in the shoulder area, lumbar plexus found in the lower back, sacral plexus situated in the pelvis, and coccygeal plexus located in the coccygeal region.
The Cervical Plexus
The cervical plexus, formed by the anterior rami of the first four...
The Cervical Plexus
The cervical plexus, formed by the anterior rami of the first four...
1.3K
Bones of the Upper Limb: Ulna
2.6K
The ulna and radius are parallel bones of the antebrachium or the forearm. The ulna lies medially and consists of a bony tip called the olecranon process at its proximal end. This hook-like projection articulates with the olecranon fossa of the humerus and forms the "hinged" ulnohumeral part of the elbow joint. This joint facilitates forearm extension and flexion while preventing its hyperextension. Similarly, the coronoid process, another bony projection on the proximal/anterior side...
2.6K


