How do the resting EEG preprocessing states affect the outcomes of postprocessing?

Abstract

Plenty of artifact removal tools and pipelines have been developed to correct the resting EEG waves and discover scientific values behind. Without expertised visual inspection, it is susceptible to derive improper preprocessing, resulting in either insufficient preprocessed EEG (IPE) or excessive preprocessed EEG (EPE). However, little is known about the impacts of IPE or EPE on postprocessing in the temporal, frequency, and spatial domains, particularly as to the spectra and the functional connectivity analysis. Here, the clean EEG (CE) with linear and quasi-stationary assumption was synthesized as ground truth based on the New-York head model and the multivariate autoregressive model. Later, IPE and EPE were simulated by injecting Gaussian noise and losing brain components, respectively. Spectral homogeneities of all EEGs were evaluated by the proposed Parallel LOg Spectra index (PaLOSi). Then, the impacts on postprocessing were quantified by the IPE/EPE deviation from CE as to the temporal statistics, multichannel power, cross spectra, scalp EEG network properties, and source dispersion. Lastly, the association between PaLOSi and varying trends of postprocessing outcomes was analyzed with evolutionary preprocessing states. We found that compared with CE: 1) IPE (EPE) temporal statistics deviated more greatly with more noise injected (brain activities discarded); 2) IPE (EPE) power was higher (lower), and IPE power was almost parallel to that of CE across frequencies, while EPE power deviation decreased with higher frequencies; IPE cross spectra deviated more greatly than EPE, except for β band; 3) derived from 7 coupling measures, IPE (EPE) network had lower (higher) transmission efficiency and worse (better) integration ability; 4) IPE sources distributed more dispersedly with greater strength while EPE sources activated more focally with lower amplitudes; 5) PaLOSi was consistently correlated with varying trends of investigated postprocessing for both simulated and real data. This study shed light on how the postprocessing outcomes are affected by the preprocessing states and PaLOSi is a promising quality control metric for creating normative EEG databases.