Early detection of Alzheimer’s and Parkinson’s diseases using multiband nonlinear EEG analysis

Objective: This study aims to distinguish, at early stages, patients with Alzheimer’s disease (AD), patients with Parkinson’s disease (PD) and healthy control (CTL) subjects using a nonlinear multiband electroencephalography (EEG)-based procedure. Method: An EEG nonlinear multiband analysis was performed in order to extract features for feeding several classifiers (such as decision trees, support vector machines, K-nearestneighbours, and logistic regression) within a leave-one-out cross-validation procedure. Results: The comparison between (a) PD versus AD groups reached classification accuracies of 100% on at least one channel of δ, θ, α, and β bands; (b) AD versus CTL groups provided accuracies of 100% on at least one channel in δ band; (c) PD versus CTL groups reached 100% of accuracy on at least one channel of γ and δ bands; and (d) the combination of PD and AD groups versus CTL group provided an accuracy of 90.6% in the θ band. Conclusions: Our findings suggest that patients may benefit from careful monitoring for early signs of change in the EEG pattern in both AD and PD.