Improved quality of auditory event-related potentials recorded simultaneously with 3-T fMRI: removal of the ballistocardiogram artefact

Abstract

EEG signals recorded simultaneously with fMRI are massively compromised by severe artefacts, among them the cardiac cycle-related ballistocardiogram (BCG) artefact. Different methods have been proposed to remove the BCG artefact focusing on channel-wise template subtraction procedures or spatial filtering approaches such as independent component analysis (ICA). Here we systematically compared the performance of the optimal basis set (OBS), a channel-wise correction approach, with ICA and a recently proposed combination of both (OBS-ICA). The three different procedures were applied to 60-channel EEG data from 12 subjects recorded during fMRI acquisition in a 3-T scanner. In addition to examination of the residual BCG artefact, the signal-to-noise ratio (SNR) and the topography of the resulting auditory evoked potential component N1 were compared. Whereas all three approaches led to a significant artefact reduction, the ICA procedure resulted in a significantly reduced N1 SNR and amplitude when compared to BCG-uncorrected data, indicating a rather poor performance. In contrast to ICA, OBS and OBS-ICA corrected data substantially improved the SNR of the N1. The quality of the auditory evoked potential N1 topography was investigated by means of equivalent current dipole modelling. On a descriptive level, all three correction procedures led to a reduced localization error when compared to BCG-uncorrected data. This improvement was significant for OBS-ICA. We conclude that OBS and OBS-ICA can efficiently remove BCG artefacts and substantially improve the quality of EEG signals recorded inside the scanner, a prerequisite for the successful integration of simultaneously recorded EEG and fMRI.