Scholarlink

<p>Facial paralysis severely changes the individual’s life and enlarges the physician’s need for precise analysis and grading of facial paralysis for appropriate rehabilitation and treatment. A patient with facial paralysis will have asymmetry of their face, challenging their everyday actions like drinking and eating, and having psychosocial and physical distress owing to their physical form. Conventional models are only reliant on the clinician’s decision and so time-consuming and subjective in nature. The notion of intelligent models executing artificial intelligence methods, particularly deep learning (DL), is a significant leap, aiding precision, automation, and scalability. Analysis and treatment planning can be improved by forming a facial paralysis recognition model. Therefore, this study presents a Facial Paralysis Identification Framework using Temporal Convolutional Neural Network (FPIF-TCNN) model to evaluate various facial regions. The objective of this study is to develop a DL-based model for identifying facial paralysis by integrating advanced feature extraction and classification approaches. The FPIF-TCNN model initially preprocesses the input images through the following steps: grayscale conversion, contrast enhancement, face detection, and augmentation, which improve image visibility and support more robust feature extraction. Following that, the multi-enhanced capsule networks are utilized for feature extraction. This model captures rich, multi-scale feature representations for classification. Finally, a temporal convolutional network with attention is employed for effective classification. This hybrid DL model learns temporal patterns and precisely classifies the severity levels based on the extracted features. The experimental result analysis of the FPIF-TCNN methodology takes place against an YFP_Dataset_Updated dataset, and the comparative outcomes demonstrated better performance over existing approaches with maximum accuracy of 99.01%, 99.15%, and 99.14% under eye, eyebrow, and mouth datasets, respectively. The FPIF-TCNN technique allows to automatically and efficiently detect facial paralysis, which exhibits its significance in practical clinical applicability. Therefore, the FPIF-TCNN technique find useful for early detection, clinical decision support, and remote monitoring, specifically in resource constrained clinical settings.</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

An evaluation of various parts of face to identify facial paralysis using temporal convolutional neural network model

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月13日

查看全文
  • Ghalib H. Alshammri,
  • Hadeel Alsolai,
  • Mohammed H. Alghamdi,
  • Afaf Almehmadi,
  • Hind T. AlHashimi,
  • Nadiyah Almutairi,
  • Tawfiq Hasanin,
  • Somia A. Asklany

摘要

Facial paralysis severely changes the individual’s life and enlarges the physician’s need for precise analysis and grading of facial paralysis for appropriate rehabilitation and treatment. A patient with facial paralysis will have asymmetry of their face, challenging their everyday actions like drinking and eating, and having psychosocial and physical distress owing to their physical form. Conventional models are only reliant on the clinician’s decision and so time-consuming and subjective in nature. The notion of intelligent models executing artificial intelligence methods, particularly deep learning (DL), is a significant leap, aiding precision, automation, and scalability. Analysis and treatment planning can be improved by forming a facial paralysis recognition model. Therefore, this study presents a Facial Paralysis Identification Framework using Temporal Convolutional Neural Network (FPIF-TCNN) model to evaluate various facial regions. The objective of this study is to develop a DL-based model for identifying facial paralysis by integrating advanced feature extraction and classification approaches. The FPIF-TCNN model initially preprocesses the input images through the following steps: grayscale conversion, contrast enhancement, face detection, and augmentation, which improve image visibility and support more robust feature extraction. Following that, the multi-enhanced capsule networks are utilized for feature extraction. This model captures rich, multi-scale feature representations for classification. Finally, a temporal convolutional network with attention is employed for effective classification. This hybrid DL model learns temporal patterns and precisely classifies the severity levels based on the extracted features. The experimental result analysis of the FPIF-TCNN methodology takes place against an YFP_Dataset_Updated dataset, and the comparative outcomes demonstrated better performance over existing approaches with maximum accuracy of 99.01%, 99.15%, and 99.14% under eye, eyebrow, and mouth datasets, respectively. The FPIF-TCNN technique allows to automatically and efficiently detect facial paralysis, which exhibits its significance in practical clinical applicability. Therefore, the FPIF-TCNN technique find useful for early detection, clinical decision support, and remote monitoring, specifically in resource constrained clinical settings.