EEG is better left alone

Abstract

Automated preprocessing methods are critically needed to process the large publicly-available EEG databases, but the optimal approach remains unknown because we lack data quality metrics to compare them. Here, we designed a simple yet robust EEG data quality metric assessing the percentage of significant channels between two experimental conditions within a 100 ms post-stimulus time range. Because of volume conduction in EEG, given no noise, most brain-evoked related potentials (ERP) should be visible on every single channel. Using three publicly available collections of EEG data, we showed that, with the exceptions of high-pass filtering and bad channel interpolation, automated data corrections had no effect on or significantly decreased the percentage of significant channels. Referencing and advanced baseline removal methods were significantly detrimental to performance. Rejecting bad data segments or trials could not compensate for the loss in statistical power. Automated Independent Component Analysis rejection of eyes and muscles failed to increase performance reliably. We compared optimized pipelines for preprocessing EEG data maximizing ERP significance using the leading open-source EEG software: EEGLAB, FieldTrip, MNE, and Brainstorm. Only one pipeline performed significantly better than high-pass filtering the data.

Methods for motion artifact reduction in online brain-computer interface experiments: a systematic review

“…Within the reviewed papers, we found that many characteristics of phase-locked EEG signals are often only visible by averaging across multiple gaits, trials, and even subjects. The authors argue that these mean characteristics are “typical brain patterns” suited to be used as commands for BCI systems (Delorme, 2022 ). However, their high inter-subject and inter-trial variability shows that approaches based on averaged characteristic patterns might not help to build robust BCI systems (Kline et al, 2015 ; Nathan and Contreras-Vidal, 2016 ).…”

“…The field of EEG data analysis faces a significant challenge in separating non-neuronal motion artifacts from neuronal activity, as noted in several studies (Gwin et al, 2010 ; Snyder et al, 2015 ; Symeonidou et al, 2018 ; Delorme, 2022 ). The primary reason for this difficulty is the lack of a ground truth measurement of the pure brain signals, which makes it challenging to create and evaluate artifact removal methods.…”

“…As a result, there is a lack of consensus on benchmarks for comparing the performance and usability of EEG systems, and no reliable quantitative metric for evaluating artifact removal methods has emerged (Delorme, 2022 ). In practice, many authors rely on the visual comparison, signal-to-noise ratio (SNR), or correlation to validate the quality of the artifact reduction method, but these practices have questionable scrutiny (Oliveira et al, 2016 ; Mur et al, 2019 ; Delorme, 2022 ). A comprehensive summary of the most common metrics applied in the reviewed papers is provided in Figure 7 .…”

“…This suggests further investigating whether correctly detected artifacts can even always lead to better classification accuracy. Recently, Delorme ( 2022 ) showed that almost none evaluated pipelines for artifact reduction increased the performance significantly over multiple datasets.…”

Measurement of the N170 during facial neuromuscular electrical stimulation (fNMES)

“…adaptive filtering (Mancini et al, 2015) and linear decomposition (Bai et al, 2016;Hernandez-Pavon et al, 2022), and so such methods might prove useful for reducing any potential artefacts introduced by fNMES. It should be noted however, that a liberal cleaning of EEG data can have minimal, and even detrimental effects, on the measurement of brain components (Delorme, 2022).…”