Optimisation of a post-processing method to remove the pulse artifact from EEG data recorded during fMRI: An application to P300 recordings during e-fMRI

Abstract

In functional cerebral studies, it has been established that co-registered electroencephalography (EEG) measurements and functional magnetic resonance imaging (fMRI) were complementary. However, EEG data recorded inside an MRI scanner are heavily distorted, mainly by the most prominent artifact, the cardiac pulse artifact (PA). We describe an original algorithm which yields a high-quality PA filter and demonstrates how this tool can be used to improve the quality of P300 ERP measurements during event-related fMRI (e-fMRI) experiments. EEG data were acquired in interleaved mode during e-fMRI while six healthy volunteers performed a visual odd-ball task, involving Distractors, Target and Novel stimuli, to elicit P300 components. The PA was corrected with the original algorithm. The temporal variations in the PA were evidenced using a principal component analysis (PCA), on each EEG channel. The procedure yielded several PA templates, which were regressed from the EEG data. The PA removal procedure was optimised, and then implemented to improve the measured P300 components. Regressing the most adequate PA template resulted in a high-quality reduction in spectral power at frequencies associated with the cardiac PA. More reliable P300 component measurements were obtained, evidencing higher amplitudes for Novels (9.76–11.20 μV) than for to Targets (6.3–9.09 μV) in centro-parietal and prefrontal areas. The improvement of the processing of EEG data acquired simultaneously with fMRI data provides a new tool and casts perspectives to study the functional organisation of the brain.