Scholarlink

<p>Mild Cognitive Impairment (MCI) is a clinical condition characterized by noticeable cognitive decline that is greater than expected for an individual’s age, yet not severe enough to interfere significantly with daily life. Early detection of MCI is critical, as it offers the opportunity to intervene before progression to more severe neurodegenerative diseases such as Alzheimer’s. While traditional diagnostic methods such as the Clock Drawing Test, Trail Making Test, and Cube Copying Test are widely used by clinicians, their manual assessment process can be subjective and time-consuming. This research addresses the automation of MCI detection using deep learning techniques applied to these neuro-psychological drawing tasks. A hybrid deep learning architecture—ResViT, which integrates ResNet50 for local feature extraction and a Vision Transformer (ViT) for capturing global context within the drawings, is being proposed. The ResViT architecture showed improved generalization and robustness across test cases, achieving a classification accuracy of 74.09% and an F1 score of 0.6716. Our results demonstrate that integrating Vision Transformer and ResNet architectures into a unified hybrid model enhances performance in cognitive disorder classification tasks, offering more accurate and measurable outcomes for early dementia detection through neuropsychological screening tools.</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

MCI detection from handwritten drawing test using residual vision transformer

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年2月24日

查看全文
  • Mehreen Sirshar,
  • Irum Matloob,
  • Ayesha Tayyabah,
  • Faiza Syed,
  • Aliya Ashraf,
  • Hessa Alfraihi

摘要

Mild Cognitive Impairment (MCI) is a clinical condition characterized by noticeable cognitive decline that is greater than expected for an individual’s age, yet not severe enough to interfere significantly with daily life. Early detection of MCI is critical, as it offers the opportunity to intervene before progression to more severe neurodegenerative diseases such as Alzheimer’s. While traditional diagnostic methods such as the Clock Drawing Test, Trail Making Test, and Cube Copying Test are widely used by clinicians, their manual assessment process can be subjective and time-consuming. This research addresses the automation of MCI detection using deep learning techniques applied to these neuro-psychological drawing tasks. A hybrid deep learning architecture—ResViT, which integrates ResNet50 for local feature extraction and a Vision Transformer (ViT) for capturing global context within the drawings, is being proposed. The ResViT architecture showed improved generalization and robustness across test cases, achieving a classification accuracy of 74.09% and an F1 score of 0.6716. Our results demonstrate that integrating Vision Transformer and ResNet architectures into a unified hybrid model enhances performance in cognitive disorder classification tasks, offering more accurate and measurable outcomes for early dementia detection through neuropsychological screening tools.