Machine Learning-Based EEG Analysis for Early Detection of Alzheimer’s Disease in Aging Populations

Abstract

The assessment of Alzheimer’s disease (AD), the most common form of dementia, which is especially prevalent in aging populations, is addressed. However, current diagnostic tools are either expensive, invasive and/or not sensitive enough for preclinical stages and early diagnosis is critical for timely intervention. In this work, we propose a machine learning based framework based on electroencephalography (EEG) data for early detection of Alzheimer's disease. EEG provides an inexpensive, noninvasive means of recording neurophysiological changes related to cognitive decline. Feature extraction methods like power spectral density (PSD), entropy measures and connectivity metrics, together with supervised learning models such as support vector machines (SVM), random forests and deep neural networks are incorporated in the study. Classification accuracies of over 90% are achieved on benchmark EEG datasets, and the strong potential for clinical deployment of the EEG deep attention network is demonstrated via cross validation. The work describes a scalable approach for non-invasively screening for Alzheimer’s in the elderly and could help accelerate AI-driven precision neurology.