A non-neuronal fMRI signal: both a confound and an opportunity for insight

Functional magnetic resonance imaging (fMRI) is widely used to investigate brain function. However, interpretation of the blood oxygen level-dependent (BOLD) signal is complicated by the fact that it reflects both neuronal activity and vascular physiology. This problem is especially relevant in neuropsychopharmacology research because pharmacological and psychiatric effects on brain function are often accompanied by physiological shifts. Here, we discuss recent evidence for an approach to disentangling neuronal and physiological components of the BOLD signal that enhances the interpretability and clinical utility of fMRI. Converging findings suggest that the systemic low-frequency oscillation (sLFO), a major component of the global signal, indexes cardiovascular manifestations of arousal rather than neuronal function. As such, retaining the sLFO in fMRI data can substantially distort functional connectivity estimates, leading to the misinterpretation of physiological fluctuations in arousal as neuronal effects. At the same time, the sLFO itself tracks physiological arousal level, behavioral performance, drug craving, pharmacological modulation, and large-scale brain network organization. Collectively, we suggest that when interpreting fMRI data, sLFO-indexed physiological arousal should both be modeled and considered separately, as extracting this signal from fMRI data provides dual benefits: a cleaner neuronal signal and a complementary index of behaviorally relevant physiology.