Diagnosis of Alzheimer’s Disease and Frontotemporal Dementia From Electroencephalography Signals

Alzheimer's disease (AD) and frontotemporal dementia (FTD) are the most burdensome neurodegenerative disorders in older populations as they pose significant challenges because of subtle cognitive decline in the initial stages. This study proposes a novel approach to distinguishing AD, FTD, and cognitively normal (CN) subjects using electroencephalography (EEG) time-domain analysis through Hjorth parameters (Activity, Mobility, and Complexity). By exploring various window lengths and moving lengths to determine the optimal segment where dementia features are most pronounced, the proposed method achieved high classification accuracy across multiple diagnostic categories. Using Linear Discriminant Analysis (LDA), the approach attained 92.5% accuracy for AD&FTD/CN classification. In individual group classifications, the method attained 88.6% accuracy for AD/CN using LDA and 96.0% accuracy for FTD/CN using a Support Vector Machine (SVM). AD classification demonstrated 92.6% sensitivity and 83.5% specificity, while FTD classification achieved 85.9% sensitivity and 95.5% specificity. Furthermore, a significant correlation (r =0.70, ${p} =0.01$ ) was found between Hjorth parameters and Mini-Mental State Examination (MMSE) scores, validating the method's potential for cognitive decline assessment. The optimal time segmentation approach and high classification accuracy represent substantial improvements over existing methods, offering a reliable and interpretable tool for clinical dementia diagnosis. These findings suggest that time-domain analysis using Hjorth parameters could be an effective biomarker for early dementia detection, potentially improving diagnostic accuracy in clinical settings.