Scholarlink

<p>This paper addresses the detection and identification of vocal fold (VF) disorders through the analysis of high-speed endoscopic videos, with the goal of assisting otorhinolaryngologists in clinical diagnosis. In this respect, different types of pathologies such as polyps, paresis and Reinke edema are considered. The proposed recognition framework relies on a set of explicitly defined features computed from high-speed VF sequences. The first contribution of this work lies in the investigation of a diverse set of features that capture not only the shape and dynamic behavior of the vocal folds but also the textural characteristics of the VF tissue. The second contribution concerns the analysis of the impact of each feature in the recognition task. Third, the influence of the region of interest (ROI) delineation (from where the features are computed), whether manual or automatic on the overall classification performance is thoroughly examined. Experimental results demonstrate the effectiveness of using selected features in order to derive an explainable classifier.</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

Vocal fold disorders recognition based on machine learning using spatial and temporal features

  • 期刊论文

  • 发表日期：

    2026年3月5日

查看全文
  • Dhouha Attia,
  • Amel Benazza-Benyahia

摘要

This paper addresses the detection and identification of vocal fold (VF) disorders through the analysis of high-speed endoscopic videos, with the goal of assisting otorhinolaryngologists in clinical diagnosis. In this respect, different types of pathologies such as polyps, paresis and Reinke edema are considered. The proposed recognition framework relies on a set of explicitly defined features computed from high-speed VF sequences. The first contribution of this work lies in the investigation of a diverse set of features that capture not only the shape and dynamic behavior of the vocal folds but also the textural characteristics of the VF tissue. The second contribution concerns the analysis of the impact of each feature in the recognition task. Third, the influence of the region of interest (ROI) delineation (from where the features are computed), whether manual or automatic on the overall classification performance is thoroughly examined. Experimental results demonstrate the effectiveness of using selected features in order to derive an explainable classifier.