Features of slow rhythmic fMRI activity in the thalamus are linked to changes in arousal state

2590 

Abstract Submission 

Beverly Setzer^1,2, Sydney Bailes^1,2, Carsen Stringer³, Laura Lewis^4,5

¹Boston University, Boston, MA, ²Massachusets Institute of Technology, Cambridge, MA, ³Janelia Research Campus, Ashburn, VA, ⁴Massachusetts Institute of Technology, Cambridge, MA, ⁵Athinoula A. Martinos Center for Biomedical Imaging, Charlestown, MA

Beverly Setzer  
Boston University|Massachusets Institute of Technology
Boston, MA|Cambridge, MA

Sydney Bailes  
Boston University|Massachusets Institute of Technology
Boston, MA|Cambridge, MA

Carsen Stringer, PhD  
Janelia Research Campus
Ashburn, VA

Laura Lewis  
Massachusetts Institute of Technology|Athinoula A. Martinos Center for Biomedical Imaging
Cambridge, MA|Charlestown, MA

In light sleep and drowsiness, the brain exhibits low-frequency oscillations. These can be observed in both electrical neural signals by electroencephalography (EEG) and in blood-oxygen level dependent (BOLD) signals by fast functional magnetic resonance imaging (fMRI) (Horovitz et al., 2008; Watson, 2018). EEG infraslow oscillations have been linked to changes in arousal levels (Monto et al., 2008; Sihn & Kim, 2022), and a peak in BOLD signal locks to changes in physiological arousal state (Raut et al., 2021). However, precisely how rhythmic slow BOLD dynamics are related to behavioral arousal state, the ability to sustain behavior, is unknown.
A key question is whether the distinct spatio-temporal features, like amplitude, frequency, and propagation dynamics of BOLD signal peaks have behavioral significance. Interestingly, peaks in the BOLD signal locked to changes in physiological arousal state propagate across the brain, from unimodal (early sensory processing) to transmodal (higher-order cognition) regions (Gu et al., 2021; Raut et al., 2021). On a fine-grained level, a precise spatio-temporal pattern of BOLD activity propagates across the thalamus with fluctuating arousal (Gu et al., 2021; Raut et al., 2021; Setzer et al., 2021), which is a core deep-brain structure comprised of sub-nuclei with diverse functional roles in cognition. Therefore, slow, rhythmic brain dynamics have some functional and spatial structure. We thus aimed to use fast fMRI to discover how slow BOLD dynamics in the thalamus are linked to arousal state, and investigate if specific spatio-temporal features of these dynamics affect behavior.

We used fMRI to capture activity throughout the brain while subject's arousal state varied across sleep and wakefulness. In Experiment 1, we used ultra-high field (7 Tesla) functional magnetic resonance imaging (fMRI) to capture high resolution (n=15 subjects, TR=247 mS, 2.5 mm isotropic voxels) dynamics. Subjects performed a simple button pressing task to track behavioral state (Prerau et al., 2014), and were allowed to fall asleep inside the scanner. In Experiment 2, we used simultaneous EEG-fMRI at 3 Tesla (n=5 subjects, TR=367 mS, 2.5 mm isotropic voxels) to capture how canonical arousal-related EEG rhythms were coupled to peaks in the BOLD oscillation. First, we evaluated how both behavioral and EEG arousal state was linked to BOLD oscillation peaks by averaging behavior and EEG signals during all BOLD peaks. Then, to group peaks together with similar spatio-temporal features and evaluate arousal differences between groups in a data driven way, we clustered BOLD activity across thalamic voxels using unsupervised machine learning (Stringer et al., 2023). This analysis captured both how temporal dynamics vary across the thalamus spatially and between individual peaks in the rhythmic activity. To assess the relationship directly between peak features with behavior, we averaged behavior across different amplitude and frequencies independently.

We identified two distinct clusters that had different features of thalamic activity dynamics: one with low-amplitude fast oscillations, and one with slower, larger oscillations (Figure 1a). The slow, large amplitude peaks had a stronger coupling to behavior (Figure 1b) and EEG rhythms than the fast, low amplitude peaks, which were associated with a higher baseline EEG arousal state. We found that peak amplitude was directly coupled to transitions in behavior, and that frequency was coupled to overall behavioral arousal state. Therefore, thalamic activity was linked to behavior and EEG rhythms in a state-dependent manner, with changes in behavior and EEG rhythms occurring more strongly in clusters of thalamic peaks with specific features.

This study reveals functional differences in rhythmic BOLD signal activity linked to drowsiness and light sleep. Additionally, we identified distinct states of thalamic activity that coupled to different behavior and EEG arousal levels.

Activation (eg. BOLD task-fMRI)

Subcortical Structures ²

BOLD fMRI

Sleep and Wakefulness ¹

Neurophysiology of Imaging Signals

Cognition

Consciousness

Data analysis

Electroencephaolography (EEG)

FUNCTIONAL MRI

HIGH FIELD MR

Machine Learning

Sleep

Sub-Cortical

Thalamus