Functional MRI signals exhibit stronger covariation with peripheral autonomic measures as vigilance decreases

Vigilance naturally drifts over time, coinciding with marked changes in brain-wide functional magnetic resonance imaging (fMRI) signals. Though the precise origins of these hemodynamic changes are unclear, largely separate lines of research have linked different vigilance levels not only to changes in fMRI signal fluctuation amplitudes and functional connectivity, but also to significant variations in autonomic physiology. These findings raise the possibility that vigilance-related modulations in fMRI signals may arise in part from changes in autonomic physiology and their effects on cerebral hemodynamics. Here, using simultaneous recordings of fMRI, EEG-indexed vigilance, respiration, and pulse oximetry, we investigate how the relationship between autonomic and fMRI signals varies systematically as vigilance gradually drifts. Regression analyses indicated that the strength and extent of fMRI-autonomic covariation increased as vigilance diminished, during both resting state and an auditory vigilance task. Spatiotemporally, autonomic signals exhibited early positive correlations and delayed negative correlations with fMRI signals throughout much of the grey matter, accompanied by late positive correlations in the ventricles and periventricular white matter. Low-frequency EEG power fluctuations also demonstrated state-dependent associations with both fMRI and autonomic signals, with effects in fMRI that partially overlapped with those of peripheral autonomic variations. Functional connectivity between most brain networks strengthened as vigilance decreased, especially during resting-state scans, and removing autonomic variance from fMRI signals largely attenuated this effect. Together, these results demonstrate interactions between vigilance levels, autonomic physiology, and brain hemodynamics, showing that the physiological constituents of fMRI signals vary markedly over vigilance levels and brain regions. These findings contribute to knowledge of human brain physiology and toward the accurate parsing, analysis, and interpretation of fMRI data.