Detection of Alzheimer’s Disease Using a Four-Channel EEG Montage

Abstract

Alzheimer’s disease (AD) represents the most common form of dementia, hence, its diagnosis and treatment have a heavy socioeconomic impact. Among the medical procedures for AD diagnosis, those based on biochemical markers and medical images stand out. However, data acquisition and processing times associated to these procedures are long, what contributes to growing clinical waiting lists. On the other hand, traditional approaches like neuro-psychological tests, which evaluate cognitive performance, are less powerful since they are influenced by external factors like schooling level or visual health. As an alternative, researchers have proposed the analysis of resting state electroencephalography (EEG), since it is an inexpensive and non-invasive technique to acquire endogenous brain information. Nevertheless, EEG recordings are typically performed using cumbersome electrical setups with many channels (usually 16–64), what hinders the usability of these approaches. The aim of this work is to demonstrate that early-stage AD can be detected using a simple EEG montage, and to elucidate the most relevant EEG channels in terms of classification performance. To this aim, we recorded the resting state EEG of eight patients with early-stage AD and eight healthy controls, and we implemented a binary classifier based on a multi-layer perceptron. The cross validation results we obtained suggest that early-stage AD can be detected (F1 score = 0.89 ± 0.06 ) using an EEG montage consisting of four channels (Fz, C3, Cz, and C4). This opens the possibility for early-stage AD detection in early care services within minutes, using a wearable plug-and-play EEG headset.