Functional MRI signals at and beyond 1 Hz are coupled to brain states and predict spontaneous neural activity

Abstract

Technological advances have enabled fMRI acquisition with high temporal resolution, enabling brainwide imaging in just a few hundreds of milliseconds. However, the relationship between fast hemodynamic signals and spontaneous neural activity in the resting state is not yet well understood, limiting our ability to infer neural processes from these fast data. We hypothesized that high-frequency fMRI signals are linked to spontaneous neural activity tied to vigilance states, and that these high-frequency signals could be used to infer the dynamic variations in neuronal activity indexed by EEG neural rhythms. Using fast fMRI (TR=378 ms) and simultaneous EEG in 27 humans drifting between sleep and wakefulness, we found that fMRI power increased during NREM sleep (compared to wakefulness) in frequencies up to 1.3 Hz. High-frequency fMRI power was also correlated to canonical arousal-linked EEG rhythms (alpha and delta), with spatiotemporal cross-correlation patterns reflecting both shared arousal dynamics and rhythm-specific signatures. Using machine learning, we found that EEG alpha and delta power can be decoded from high-frequency fMRI signals in subjects held-out from the training set, showing that the high-frequency components of fMRI signals contain neurally-coupled information robust enough to generalize across individuals. These results reveal that high-frequency fMRI signals are coupled to dynamically varying brain states, and that fast fMRI allows for temporally precise quantification of spontaneous neural activity. Significance StatementFunctional MRI enables non-invasive brainwide neuroimaging in humans, but classically had limited temporal resolution. Recently, studies demonstrated that fast fMRI sampling can detect stimulus-evoked neural activity at subsecond timescales. However, it was unknown whether spontaneous neural activity is represented in fast fMRI signals. Using fMRI and simultaneous EEG in subjects who fell asleep, we found that EEG neural activity associated with sleep and wakefulness is represented in high-frequency fMRI signals. The results reveal that fast fMRI sampling can be used to investigate the spontaneous neural activity underlying brain states with high temporal precision.