Abstract
Sleep entails significant changes in cerebral hemodynamics and metabolism. Yet, the way these processes evolve throughout wakefulness and sleep and their spatiotemporal dependence remain largely unknown. Here, by integrating a functional PET technique with simultaneous EEG-fMRI, we reveal a tightly coupled temporal progression of global hemodynamics and metabolism during the descent into NREM sleep, with large hemodynamic fluctuations emerging as global glucose metabolism declines, both of which track EEG arousal dynamics. Furthermore, we identify two distinct network patterns that emerge during NREM sleep: a ~0.02-Hz oscillating, high-metabolism sensorimotor network remains active and dynamic, whereas hemodynamic and metabolic activity in the default-mode network is suppressed. These results elucidate how sleep diminishes awareness while preserving sensory responses, and uncover a complex, alternating balance of neuronal, hemodynamic, and metabolic dynamics in the sleeping brain. This work also demonstrates the potential of EEG-PET-MRI to explore neuro-metabolic-hemodynamic mechanisms underlying cognition and arousal in humans.
Introduction
Sleep is a complex physiological state characterized by profound changes in brain activity, metabolism, and blood flow. Understanding how these processes interact is crucial for deciphering the functions of sleep, from memory consolidation to metabolic waste clearance. However, the different timescales of neuronal firing, hemodynamics, and energy consumption have historically made it difficult to observe their coordination in real-time. This study addresses this gap by integrating three powerful imaging modalities—EEG, functional PET (fPET), and fMRI—to map the human brain during the transition from wakefulness to non-rapid eye movement (NREM) sleep.
Methodology: Trimodal Imaging and Novel Modeling
The researchers utilized a rare simultaneous EEG-PET-MRI setup with 26 participants. EEG provided high-resolution tracking of sleep stages (N1, N2, N3) and arousal dynamics. fMRI (BOLD) captured oxygenated blood flow fluctuations, while fPET, using a constant infusion of a radioactive glucose tracer ([18F]FDG), allowed for minute-by-minute tracking of glucose metabolism.
To bridge the temporal resolution gap between fast fMRI signals and slower PET kinetics, the team developed a "quasi-metabolic" regressor (BOLD-AV-TAC). This mathematical framework allowed them to directly correlate the amplitude of blood flow "waves" with the rate of glucose consumption across the entire brain, providing a proxy for metabolic state during sleep transitions.
The Metabolic-Hemodynamic Seesaw
A primary finding of the study is the tight temporal coupling between global metabolism and hemodynamics. As participants descended into NREM sleep, a global decline in glucose metabolism was observed, synchronized with a surge in large-scale hemodynamic fluctuations. These "BOLD waves" were precisely locked to EEG arousal indices, suggesting that the brain's energy state and its physical blood flow patterns are two sides of the same coin during sleep transitions.
Spatial Specialization: Sensory Gate vs. DMN Shutdown
The study revealed that the brain does not shut down uniformly. Instead, it exhibits a distinct spatial structure during NREM sleep. The sensorimotor network (including visual and auditory areas) remains metabolically "warm" and active, exhibiting ~0.02 Hz hemodynamic oscillations. This suggests a "sensory gate" that stays primed to respond to external threats even during sleep.
In contrast, the Default-Mode Network (DMN)—associated with self-awareness and internal thought—shows the most significant metabolic suppression and reduced fluctuations. This divergence explains the characteristic "loss of self" during sleep while maintaining the ability to wake up to external stimuli.
Key Results and Implications
The researchers found a significant spatial anti-correlation (r = -0.41) between hemodynamic increases and metabolic decreases across the cortex. Sensory regions showed hemodynamic increases of up to 80%, while the DMN showed the deepest metabolic drops. Furthermore, the "quasi-metabolic" fMRI signal successfully explained a significant portion of the variance in actual PET data (FDR q < 0.05), validating the model.
These findings have significant implications for the theory of sleep as a restorative process. The emergence of large hemodynamic waves exactly when metabolism drops supports the idea that sleep is a "metabolic rinse" cycle. Reduced neuronal activity may allow the brain to prioritize the physical clearance of waste products via cerebrospinal fluid (CSF) flow.
Limitations and Future Outlook
The study primarily focused on N1 and N2 sleep stages. Preliminary data from deep N3 sleep suggests a non-linear relationship where hemodynamic waves may eventually decrease, a phenomenon requiring further investigation. Additionally, the lack of EOG and EMG prevented the definitive isolation of REM sleep. Future work using this trimodal framework could provide critical insights into neurodegenerative conditions like Alzheimer’s disease, where the coordination between metabolism and waste clearance is often compromised.
Summary updated