Restate the reference for EEG microstate analysis

TL;DR

This study compares EEG reference methods for microstate analysis, finding that REST provides more reliable and objective microstate features than the traditional average reference, especially in simulated data.

Contribution

It introduces a microstate-based EEG forward model and evaluates the effectiveness of REST versus AR in microstate analysis using simulations and real data.

Findings

REST recovers more similar cluster maps in simulations

REST yields more objective microstates than AR in real EEG data

Cluster maps differ significantly between REST and AR on real data

Abstract

Despite the decades of efforts, the choice of EEG reference is still a debated fundamental issue. Non-neutral reference can inevitably inject the uncontrolled temporal biases into all EEG recordings, which may influence the spatiotemporal analysis of brain activity. A method, termed microstates, identifying spatiotemporal EEG features as the quasi-stable topography states in milliseconds, suggests its potential as biomarkers of neurophysiological disease. As reference electrode standardization technique (REST) could reconstruct an infinity reference approximately, it is a question whether REST or the other references will be more reliable than average reference (AR) for the microstates analysis. In this study, we design the microstate-based EEG forward model, and apply different references for microstates analysis. The spatial similarity between the generated and assumed cluster maps is mainly investigated. Furthermore, the real EEG data by the parametric bootstrap method is used to validate the performance of the references. Finally, we find that REST is robust to recover more similar cluster maps to the assumption than AR in the simulation, and the cluster maps between REST and AR on the real EEG data are quite different. This study may indicate that REST contributes to identifying more objective microstates features than AR.