EEG-Based Alzheimer's Detection Using Power Spectral Density, Tsallis Entropy, Amplitude Features, and SVM Classification

Abstract

Alzheimer's disease (AD) is a neurodegenerative disease of brain tissue, currently incurable, which leads to the progressive and irreversible loss of mental functions, particularly memory. It is rare to detect Alzheimer to an early stage. However, early diagnosis can allow a faster treatment and thus improve the patient's well-being. Electroencephalogram (EEG) is a non-invasive and cost-effective tool that measures electrical activity in the brain. In this study, we aimed to create an automatic detection method by combining several powerful EEG biomarkers which, to our knowledge, were never put together namely Power Spectral Density, Tsallis entropy and changes in the EEG amplitude. The features were then put into a Support Vector Machine (SVM) for the identifications of Alzheimer patients and healthy controls (CN). Using a five-fold cross-validation strategy across the entire frequency band, the classification accuracy reached 83.08%, with a sensitivity of 78% and a specificity of 90% while it reached 78.46% accuracy ,75% sensitivity and 83% specificity using the leave one subject out cross validation.