Selecting a pre-processing pipeline for the analysis of EEG event-related rhythms modulation.

Abstract

When deciding how to pre-process EEG data, researchers need to make a choice at each single step of the procedure among different possibilities, equally valid. Therefore, in this work, we illustrate how these decisions may affect the quality of the final cleaned data in an Action Observation/Motor Imagery protocol, using quantitative indices. In particular, we showed the effect of segmenting or not the data in epochs around the stimulus presentation time on the independent component analysis (ICA) used for artifact removal. For ICA analysis, we tested two algorithms (SOBI and Extended Infomax). Finally, three re-reference approaches (Common averaged reference-CAR, robust-CAR and reference electrode standardization technique - REST) were also applied and their effects compared. Results showed that the segmenting method has a prominent effect on the cleaning procedure and consequently on final EEG data quality. Extended Infomax is confirmed as the method of choice for the identification of the artifactual components and, finally, CAR and the REST re-referencing techniques led to similar good results.