Alternative EEG pre-processing pipelines can lead to conflicting conclusions regarding cortical excitation/inhibition ratio

Confluent recent evidence indicates that the spectral slope of 1 /f neurophysiological recordings is correspondent to cortical excitation/inhibition (E/I) ratio. In this framework, a steeper power spectrum (i.e., one with a larger spectral exponent β ) is indicative of stronger inhibitory tone and thus lower E/I ratio, and vice versa. While the tools commonly utilized for estimating β are mostly consistent, there appears to be a lack of standardization among data processing protocols for slope analysis. In this work our goal is to draw attention to a fundamental consequence of this issue, namely that even in a confined, comparative research environment, applying different pre-processing steps to electroencephalography (EEG) data can lead to conflicting conclusions in terms of the E/I ratio. To this end, we analyzed resting-state EEG recordings in two independent datasets, containing data collected with eyes open (EO) and eyes closed (EC), with the latter considered as a physiological state with stronger inhibitory tone. Our analyses confirmed consistently in both cohorts that applying different spatial filtering schemes in an otherwise identical analytical pipeline indicated a decrease in E/I ratio over the prefrontal cortex in one case, but not the other when transitioning from EO to EC. In contrast, this same pattern was apparent over the occipital cortex regardless of the pre-processing scheme. This empirical evidence calls for the development of a standardized data processing protocol for EEG-based analyses of the E/I ratio. Highlights Two EEG spatial filtering schemes - CAR and SL - are compared regarding E/I ratio CAR and SL are shown to substantially affect aperiodic spectral slope estimates CAR and SL yield opposite conclusions about frontal E/I ratio on the same data SL reverses the common β lo < β hi bimodal pattern over frontal cortex Spatial filtering effects are confirmed on two independent datasets