Unveiling neural activity changes in mild cognitive impairment using microstate analysis and machine learning.

Abstract

Background: Mild cognitive impairment (MCI) is recognized as a condition that may increase the risk of developing Alzheimer's disease (AD). Understanding the neural correlates of MCI is crucial for elucidating its pathophysiology and developing effective interventions. Electroencephalogram (EEG) microstates, reflecting brain activity changes, have shown promise in MCI research. However, current approaches often lack comprehensive characterization of the complex neural dynamics associated with MCI.

Objective: This study aims to investigate neurophysiological changes associated with MCI using a comprehensive set of microstate features, including traditional temporal features and entropy measures.

Methods: Resting-state EEG data were collected from 69 MCI patients and healthy controls (HC). Microstate analysis was performed to extract conventional features (duration, coverage) and entropy measures. Statistical analysis, principal component analysis (PCA), and machine learning (ML) techniques were employed to evaluate neurophysiological patterns associated with MCI.

Results: MCI displayed altered microstate dynamics, with significantly longer coverage and duration in Microstate C but shorter in Microstates A, B, and D compared to HCs. PCA revealed two principal components, primarily composed of microstate dynamics and entropy measures, explaining over 75% of the variance. ML models achieved high accuracy in distinguishing MCI patterns.

Conclusions: Our comprehensive analysis of EEG microstate features provides new insights into neurophysiological changes associated with MCI, highlighting the potential of EEG microstates for investigating complex neural changes in cognitive decline.