Impact of the pulse artifact on evoked activity in human wakefulness and sleep

Abstract

The investigation of the neural evoked response to the heartbeat quantified using electroencephalography (i.e., heartbeat evoked potentials or HEPs), has gained recent attention in neuroscience, notably as a measure of interoception, the sensory system responding to internal bodily states. One main challenge in measuring HEPs is their susceptibility to cardiac artifacts contamination, including the cardiac field artifact and the pulse artifact (PA), the latter possibly caused by the mechanical pressure of pulsating vessels. Here, we aimed at assessing the impact of PAs on HEPs and auditory evoked potentials (AEPs, a proxy of the neural responses to exteroceptive sensory stimuli). To this aim, we compared two pre-processing pipelines using independent component analysis in healthy volunteers (N=30). The first, standard, pipeline excluded ocular, muscle, sweating-related activity and cardiac-related activity stemming from the cardiac field artifact. The second, pairwise phase consistency (PPC) pipeline, in addition to the removal of the aforementioned components, used the quantitative metric of PPC between independent components and the ECG to remove the cardiac-related PA. We tested how these two pre-processing approaches influenced HEPs and AEPs recorded under diverse neurophysiological conditions (wakefulness, N2, N3, and REM sleep). Comparing the HEPs from the standard and the PPC approaches (cluster-based permutation statistics, p<0.05, two-tailed) revealed a significant effect of PAs, particularly in wakefulness, followed by REM and N2 sleep, with Cohens d effect sizes of 1.92, 0.95 and 0.88, respectively. By contrast, PA correction had a negligible effect (p>0.05, two-tailed) on the HEPs in N3 sleep and on the AEPs in all vigilance states. Our results emphasize the need to account for the PA as a significant confounding factor when comparing HEPs across groups with varying vascular or cardiac conditions.