Diagnosis of Alzheimer's Disease and Frontotemporal Dementia from Electroencephalography Signals

Abstract

Alzheimer’s Disease (AD) and Frontotemporal Dementia (FTD) are progressive neurodegenerative disorders that impair cognition, behaviour, and daily functioning, creating significant challenges for early and accurate diagnosis. Their overlapping clinical symptoms make difficult to distinguish AD, FTD, and Cognitively Normal (CN) individuals during the early stages, delaying appropriate intervention and treatment planning. Existing diagnostic approaches are limited by their reliance on binary classification, lack of direct feature extraction for AD–FTD discrimination, and insufficient capability to identify nonlinear irregularities and frequency redistribution patterns in electroencephalography (EEG) signals. To overcome these limitations, this study proposes a novel EEG-based multiclass classification framework using comprehensive multi-domain feature analysis. EEG signals are preprocessed through filtering, artifact removal, re-referencing, and segmentation to improve signal quality. Discriminative features are extracted from time-domain, frequency-domain, and time-frequency complexity measures, providing a comprehensive representation of neural activity. Classification is performed using nonlinear Support Vector Machine (SVM) model to distinguish AD, FTD, and CN groups effectively. Experimental results demonstrate high classification accuracy, strong class separability, and robust predictive performance with minimal misclassification. The findings confirm that the proposed integrated framework provides a reliable, efficient, and clinically valuable approach for early dementia detection, accurate differential diagnosis, and improved decision-making in clinical practice.