Test-retest reliability of EEG microstate metrics for evaluating noise reductions in simultaneous EEG-fMRI

Abstract

Simultaneous EEG-fMRI has potential for elucidating brain activities but suffers from severe noise/artifacts in EEG. While several countermeasures have been developed, it remains difficult to evaluate noise reductions in the absence of ground truth in EEG. We introduce a new evaluation method which takes advantage of high test-retest reliability of EEG microstate metrics. We assumed, if the reliability is high for a pair of EEG recorded outside an MR scanner on two different days, then it should also be high for a pair of EEG recorded inside and outside the scanner on the same day if MR-induced noise is absent. Thus, noise should be removed in a way that the reliability increases. Accordingly, we obtained EEG both inside and outside the scanner on two different days. Using ICC as an index, we examined test-retest reliability for 1) a pair of EEG outside the scanner across the days, 2) a pair of EEG inside and outside the scanner on the same day, and 3) a pair of EEG inside the scanner across the days. MR-induced noise, BCG artifact in particular, was reduced with joint decorrelation with varying thresholds. We obtained moderately high reliability in all the three pairs (ICCs > 0.5), suggesting sufficient noise reductions. Taking these steps, the quality of EEG improved as assessed with its traces, power spectra density, and microstate templates in resting state as well as event-related potentials in a visual oddball task. We discuss advantages and limitations of this new evaluation method.