Abstract
Microsleep episodes—brief, unintended intrusions of sleep—frequently occur during resting-state fMRI (rs-fMRI) scans. This study investigated how these episodes influence neural activity, autonomic signals (respiration and heart rate), and the resulting BOLD signal. Using simultaneous EEG, PPG, and respiration monitoring, the researchers identified a consistent, multiphasic BOLD response across microsleeps of varying durations. These changes are tightly coupled with autonomic shifts and significantly distort functional connectivity (FC) estimates, even after applying standard denoising techniques like global signal regression.
Introduction
Resting-state functional MRI is a foundational tool for mapping the human functional connectome. However, the dark, noisy environment of an fMRI scanner often induces drowsiness, leading to "microsleeps" that last only a few seconds. While these episodes are often dismissed as noise or ignored, sleep is a systemic physiological event. This research characterizes the complex BOLD dynamics and autonomic changes associated with microsleeps to determine their impact on the validity of functional connectivity measurements.
Methods
Ten participants underwent six 20-minute rs-fMRI runs. To establish a "ground truth" for microsleep episodes, the researchers employed a multi-modal monitoring approach:
- Neural Activity: Concurrent EEG tracked arousal levels via the ratio of alpha to delta and theta bands.
- Autonomic Monitoring: Photoplethysmography (PPG) tracked heart rate, while respiration belts monitored breathing patterns.
- Behavioral Validation: Real-time eye-closure monitoring via camera and post-scan self-reports confirmed sleep episodes.
- Analysis: BOLD signal dynamics were analyzed for microsleep durations ranging from 4 to 44 seconds to evaluate consistency and the impact on functional connectivity (FC) estimates.
Results
The Multiphasic BOLD Response
The study identified a remarkably consistent four-stage BOLD signal signature during microsleep episodes, regardless of their duration:
- Initial Sleep-Onset Dip: A signal reduction reaching its lowest point (nadir) approximately 6 seconds after sleep onset.
- The Plateau: An increase above the wake baseline that persists until the participant awakens.
- Awakening Bump: A transient positive spike in the signal lasting up to 6 seconds upon waking.
- Post-Awakening Undershoot: A prolonged dip below the wake baseline lasting roughly 30 seconds.
Regional and Autonomic Variations
While this multiphasic pattern was observed across most of the brain, the thalamus—the brain's sensory relay station—served as a notable exception. During the plateau phase, the thalamic BOLD signal remained below the wake baseline, highlighting its unique role in gating consciousness during sleep transitions.
Furthermore, BOLD changes were "intimately intertwined" with autonomic shifts. Sleep onset was associated with reductions in respiration and heart rate, while awakening triggered immediate increases. These autonomic fluctuations affect blood oxygen levels, suggesting that the BOLD signal during drowsiness is a composite of neural activity and systemic physiological changes.
Impact on Functional Connectivity
The frequency of microsleep episodes significantly altered functional connectivity estimates. Crucially, the study found that Global Signal Regression (GSR), a standard denoising technique, was unable to resolve these connectivity distortions. This indicates that microsleep-related signals have complex regional structures rather than being simple global noise.
Conclusion
Falling asleep and awakening trigger large, widespread BOLD signal changes that are consistent across varied microsleep durations. These findings emphasize that autonomic contributions to the BOLD signal are integral to the sleep-wake transition. For the neuroimaging community, these results highlight the urgent need to minimize sleep during resting-state scans and to integrate both autonomic and central nervous system contributions into frameworks for understanding brain function. Simply treating the brain as an isolated processor without accounting for the heart and lungs may lead to fundamental biases in our understanding of the functional connectome.
Summary updated