Decoding Dementia: Leveraging the Tunable Q Factor Wavelet Transform to Classify EEG Signals in Alzheimer’s and Frontotemporal Dementia

Abstract

Early diagnosis of neurodegenerative diseases such as Alzheimer’s Disease (AD) and Frontotemporal Dementia (FTD) is essential for improving patient care and reducing healthcare burden. This study proposes a machine learning-based framework for the classification of EEG signals using the Tunable Q-Factor Wavelet Transform (TQWT). EEG recordings obtained from 88 participants (36 AD, 23 FTD, and 29 cognitively normal subjects) were analyzed under resting-state conditions using 19 EEG channels. The signals were decomposed using multi-level TQWT to extract statistical and rhythm-based features from EEG frequency bands. A total of 1881 features were obtained from both standard and rhythm-based decompositions. Several machine learning classifiers including Decision Trees, K-Nearest Neighbors (k-NN), Support Vector Machines (SVM), Neural Networks, and Ensemble Learning models were evaluated. Experimental results show that rhythm-based TQWT features provide a compact and discriminative representation of EEG signals. The highest classification accuracy (92.7%) was achieved using the Ensemble Learning (Bagged Trees) classifier. The results demonstrate that TQWT-based EEG feature extraction combined with machine learning algorithms can effectively distinguish AD, FTD, and cognitively normal subjects, suggesting strong potential for supporting non-invasive dementia diagnosis.