Scholarlink

Alzheimer’s disease (AD) is a progressive neurodegenerative disorder requiring reliable early diagnostic tools. We propose PSOs-Net, a Bayesian-optimized 1DCNN-BLSTM model for automated eye movement detection, alongside the novel AD-VSEMD dataset capturing visual search behaviors in AD patients and controls. The model achieves state-of-the-art performance, with event-level Kappa scores of 0.972 (fixations), 0.968 (saccades), and 0.786 (post-saccadic oscillations) on Lund2013, surpassing existing methods. Results show AD-specific impairments, including prolonged fixations, increased saccadic activity, and altered search patterns linked to cognitive decline. This demonstrates that automated eye movement analysis offers a sensitive, non-invasive AD assessment, with our Bayesian-optimized framework enhancing neurodegenerative disease detection and progression monitoring through eye-tracking and machine learning.

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

Bayesian-Optimized Eye Movement Event Detection for Assisted Alzheimer’s Diagnosis

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Xin Wang,
  • Yujie Nie,
  • Jia Zhao,
  • Yahan Wang,
  • Yibin Li,
  • Xiangqing Xu,
  • Xin Ma

摘要

Alzheimer’s disease (AD) is a progressive neurodegenerative disorder requiring reliable early diagnostic tools. We propose PSOs-Net, a Bayesian-optimized 1DCNN-BLSTM model for automated eye movement detection, alongside the novel AD-VSEMD dataset capturing visual search behaviors in AD patients and controls. The model achieves state-of-the-art performance, with event-level Kappa scores of 0.972 (fixations), 0.968 (saccades), and 0.786 (post-saccadic oscillations) on Lund2013, surpassing existing methods. Results show AD-specific impairments, including prolonged fixations, increased saccadic activity, and altered search patterns linked to cognitive decline. This demonstrates that automated eye movement analysis offers a sensitive, non-invasive AD assessment, with our Bayesian-optimized framework enhancing neurodegenerative disease detection and progression monitoring through eye-tracking and machine learning.