Comparison of different spatial transformations applied to EEG data: A case study of error processing

• Published in: International journal of psychophysiology Vol. 97; no. 3; pp. 245 - 257
• Main Author: Cohen, Michael X
• Format: Journal Article
• Language: English
• Published: Netherlands Elsevier B.V 01.09.2015
• Subjects: Adolescent; Adult; Analysis of Variance; Beamforming; Brain - physiology; Brain Mapping; Connectivity; Current source density; EEG; Electroencephalography; Evoked Potentials - physiology; Female; Humans; Independent components analysis; Laplacian; Magnetic Resonance Imaging; Male; Physical Stimulation; Reaction Time - physiology; Signal Processing, Computer-Assisted; Spatial transformations; Spectrum Analysis; Time Factors; Time–frequency; Young Adult; Independent components analysis; Beamforming; Spatial transformations; EEG; Current source density; Connectivity; Time–frequency; Laplacian
• ISSN: 0167-8760; 1872-7697; 1872-7697
• DOI: 10.1016/j.ijpsycho.2014.09.013

The purpose of this paper is to compare the effects of different spatial transformations applied to the same scalp-recorded EEG data. The spatial transformations applied are two referencing schemes (average and linked earlobes), the surface Laplacian, and beamforming (a distributed source localization procedure). EEG data were collected during a speeded reaction time task that provided a comparison of activity between error vs. correct responses. Analyses focused on time–frequency power, frequency band-specific inter-electrode connectivity, and within-subject cross-trial correlations between EEG activity and reaction time. Time–frequency power analyses showed similar patterns of midfrontal delta–theta power for errors compared to correct responses across all spatial transformations. Beamforming additionally revealed error-related anterior and lateral prefrontal beta-band activity. Within-subject brain–behavior correlations showed similar patterns of results across the spatial transformations, with the correlations being the weakest after beamforming. The most striking difference among the spatial transformations was seen in connectivity analyses: linked earlobe reference produced weak inter-site connectivity that was attributable to volume conduction (zero phase lag), while the average reference and Laplacian produced more interpretable connectivity results. Beamforming did not reveal any significant condition modulations of connectivity. Overall, these analyses show that some findings are robust to spatial transformations, while other findings, particularly those involving cross-trial analyses or connectivity, are more sensitive and may depend on the use of appropriate spatial transformations. •The effects of different spatial transformations applied to the same EEG data were evaluated.•Spatial transformations included Laplacian, beamforming, and referencing schemes.•Analyses included time–frequency power, connectivity, and brain–behavior correlations.•Power condition effects (response accuracy) were robust to the choice of transformation.•Connectivity and brain–behavior correlation results depended on transformations.