Modifications of Microstates in Resting-State EEG Associated to Cognitive Decline in Early Alzheimer's Disease Assessed by a Machine Learning Approach

Electroencephalography (EEG) is a neuroimaging technique able to measure the brain electrical activity. The spontaneous cerebral activity at rest exhibits dynamic alterations of brain states called functional microstates. It is well known that EEG microstates are altered in Alzheimer's disease (AD), suggesting that EEG microstates might be indicative of the cognitive decline associated with AD. The present study investigated the capabilities of a machine learning (ML) approach to predict the Mini-Mental Score (MMSE) of AD patients and healthy controls (HC) from the EEG microstates parameters. Furthermore, a classification of AD and HC was implemented with a Receiver Operating Curve (ROC) employing the predicted MMSE score of the ML framework as input. The correlation coefficient between the MMSE and the multivariate metric estimated by the ML approach was 0.53. Furthermore, the area under the ROC Curve was 0.80 in discriminating AD from HC. The results demonstrated that EEG microstates may represent a powerful tool for clinical evaluation of cognitive decline in early AD.