文章：

高速视频内窥镜与刚度映射在人工智能辅助声带病变鉴别中的应用

High-Speed Videoendoscopy and Stiffness Mapping for AI-Assisted Glottic Lesion Differentiation

原文发布日期：21 April 2025

DOI: 10.3390/cancers17081376

类型: Article

开放获取: 是

英文摘要：

Objectives: This study evaluates the potential of high-speed videoendoscopy (HSV) in differentiating between benign and malignant glottic lesions, offering a non-invasive diagnostic tool for clinicians. Moreover, a new parameter derived from high-speed videoendoscopy (HSV) had been proposed and implemented in the analysis for an objective assessment of the vocal fold stiffness.Methods: High-speed videoendoscopy (HSV) was conducted on 102 participants, including 21 normophonic individuals, 39 patients with benign vocal fold lesions, and 42 with glottic cancer. Laryngotopographic parameter describing the stiffness of vocal fold (SAI) and kymographic parameters describing amplitude, symmetry, and glottal dynamics were quantified. Statistical differences between groups were assessed using receiver operating characteristic (ROC) analysis and lesion classification was performed using a machine learning model.Results: Univariate receiver operating characteristic (ROC) analysis revealed that SAI (AUC = 0.91, 95% CI: 0.839–0.962) and weighted amplitude asymmetry (AUC = 0.92, 95% CI: 0.85–0.974) were highly effective in distinguishing between normophonic and organic lesions (p< 0.01). Further multivariate analysis using machine learning models demonstrated improved accuracy, with the SVM classifier achieving an AUC of 0.93 for detecting organic lesions and 0.83 for distinguishing benign from malignant lesions.Conclusions: The study demonstrates the potential value of parameter describing the pliability of infiltrated vocal fold (SAI) as a non-invasive tool to support histopathological evaluation in laryngeal lesions, with machine learning models enhancing diagnostic performance.

摘要翻译： 

目的：本研究旨在评估高速视频喉镜在鉴别声门区良恶性病变中的潜力，为临床医生提供一种无创诊断工具。此外，研究提出并应用了一种基于高速视频喉镜的新参数，用于声带硬度的客观评估。 方法：对102名参与者进行高速视频喉镜检查，包括21名声带功能正常者、39名声带良性病变患者和42名声门癌患者。量化了描述声带硬度的喉地形参数（SAI）以及描述振幅、对称性和声门动态的声门图参数。采用受试者工作特征曲线分析评估组间统计学差异，并利用机器学习模型进行病变分类。 结果：单变量受试者工作特征曲线分析显示，SAI（曲线下面积=0.91，95%置信区间：0.839–0.962）和加权振幅不对称性（曲线下面积=0.92，95%置信区间：0.85–0.974）在区分声带功能正常者与器质性病变方面效果显著（p<0.01）。进一步采用机器学习模型进行多变量分析显示诊断准确性有所提升，支持向量机分类器在检测器质性病变时曲线下面积达到0.93，在区分良恶性病变时达到0.83。 结论：本研究证实了描述浸润性声带柔韧性的参数（SAI）作为无创工具辅助喉部病变组织病理学评估的潜在价值，机器学习模型的应用进一步提升了诊断性能。

原文链接：

High-Speed Videoendoscopy and Stiffness Mapping for AI-Assisted Glottic Lesion Differentiation