Scholarlink

<p>Alzheimer’s disease (AD) and frontotemporal dementia (FTD) are neurodegenerative disorders causing cognitive decline. Despite distinct pathologies, their overlapping symptoms often lead to FTD misdiagnosis as AD, resulting in inappropriate treatment. This study combines temporal and spectral features in an electroencephalography (EEG) framework to accurately diagnose AD and FTD. We systematically evaluated five machine learning and six deep learning models, and optimized feature-algorithm pairs to achieve superior performance. Using resting-state EEG data from 88 participants, comprising 36 AD patients, 23 FTD patients, and 29 healthy controls (CN), our convolutional neural network (CNN) trained on wavelet feature achieved outstanding binary classification accuracies of 97.05% (AD vs. CN), 96.55% (AD vs. FTD), and 97.84% (FTD vs. CN). For multiclass classification (AD/FTD/CN), the model attained an accuracy of 94.94%, surpassing results reported in previous studies. The proposed framework shows promise as a non-invasive tool for the early diagnosis of Alzheimer’s and frontotemporal dementia, potentially addressing critical screening challenges. However, its translation to clinical practice is contingent upon future validation in larger, independent, and diverse cohorts to confirm generalizability and equity.</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

Deep learning techniques for early diagnosis of Alzheimer’s disease and frontotemporal dementia using EEG signals

  • 期刊论文

  • 发表日期：

    2026年4月29日

查看全文
  • Amir Hossein Hachamnia,
  • Ali Mehri,
  • Maryam Jamaati

摘要

Alzheimer’s disease (AD) and frontotemporal dementia (FTD) are neurodegenerative disorders causing cognitive decline. Despite distinct pathologies, their overlapping symptoms often lead to FTD misdiagnosis as AD, resulting in inappropriate treatment. This study combines temporal and spectral features in an electroencephalography (EEG) framework to accurately diagnose AD and FTD. We systematically evaluated five machine learning and six deep learning models, and optimized feature-algorithm pairs to achieve superior performance. Using resting-state EEG data from 88 participants, comprising 36 AD patients, 23 FTD patients, and 29 healthy controls (CN), our convolutional neural network (CNN) trained on wavelet feature achieved outstanding binary classification accuracies of 97.05% (AD vs. CN), 96.55% (AD vs. FTD), and 97.84% (FTD vs. CN). For multiclass classification (AD/FTD/CN), the model attained an accuracy of 94.94%, surpassing results reported in previous studies. The proposed framework shows promise as a non-invasive tool for the early diagnosis of Alzheimer’s and frontotemporal dementia, potentially addressing critical screening challenges. However, its translation to clinical practice is contingent upon future validation in larger, independent, and diverse cohorts to confirm generalizability and equity.