Characterizing dynamic brain states in subjects across states of wakefulness
Poster No:
2028
Submission Type:
Abstract Submission
Authors:
Lauren Daley1, Harrison Watters2, Theodore LaGrow3, Shella Keilholz3
Institutions:
1Emory University-Georgia Institute of Technology, Atlanta, GA, 2Emory University, Atlanta, GA, 3Georgia Institute of Technology, Atlanta, GA
First Author:
Co-Author(s):
Introduction:
It is widely accepted that brain activity, and therefore brain state, differ between conscious and unconscious (sleeping) conditions; fMRI is used to identify and quantify these changes across space and time. Dynamic analysis methods, such as complex principle component analysis (cPCA) (Bolt et al., 2022) and quasi-periodic pattern (QPP) analysis (Majeed et al., 2011), preserve temporal information, allowing spatiotemporal network detection. There are several QPPs previously detected in BOLD signal, including global signal and patterns linked to arousal. Similar spatiotemporal patterns were found from the BOLD signal using cPCA. Components of the BOLD signal are known to correlate to infraslow neural activity, and recent studies validating this principle have also discovered considerable variations in the delays related to different states, spanning both cortical and sub-cortical areas during sleep stages, from wakefulness to deep sleep (Mitra et al, 2015). The current study utilizes these dynamic methods to detect and analyze how this infraslow activity correlate propagates spatially and temporally in healthy individuals in resting-state versus in active sleep.
Methods:
The data was acquired at 3T by Yameng Gu's group and shared on OpenNeuro (Gu et al., 2023). Of the 31 participants, each had 1 scan session, which consisted of 1 anatomical scan, 2 resting-state scans, and at least 2 sleeping scans, with simultaneous EEG recordings. Functional scans were acquired using an EPI sequence: TR=2100ms, TE=25ms, FOV=240mm, slice thickness=2mm. The raw data were preprocessed using CPAC (Craddock et al., 2013), and Brainnetome parcellations (Fan et al., 2016). The parcellated and processed data was input into a pattern-finding algorithm, similar to that used in Majeed et al, to iteratively detect and converge upon a QPP template. The data was also input into cPCA, introduced by Bolt et al. Correlation values between networks were calculated on a network-level and tested for significance across conditions using a p-value<0.05, and then corrected for multiple comparisons.
Results:
The results from the two different dynamic analysis techniques used are shown in figures 1 and 2. The QPP results (fig. 1) yield some expected observations, including clear differences across limbic, visual, and somato-motor networks between resting-state and sleep. Unexpectedly, there were observed differences in VAN phase relative to other networks and the drastic decrease in the subcortical network seen in the sleep group. Figure 2, results from cPCA, is more difficult to parse out differences, showing similar expected variance, probability of components, and overall proportion for both conditions. However, using the simultaneously acquired EEG recordings, the sleeping scans were divided into three subgroups – sleep1 (score<0.5, mostly awake), sleep2 (0.5<score<0.9, possibly sleep) and sleep3 (score>0.9, likely asleep). Interestingly, the second pattern (PC2) has a higher probability of occurring vs pattern three (PC3) in sleep2 and sleep3 groups, a trend not seen in resting-state nor sleep1.
Conclusions:
This study utilizes dynamic methods to identify differences in spatiotemporal functional patterns across resting-state and sleep in healthy individuals, including key nodes such as VAN and subcortical structures. Typically, VAN is weakly correlated with DMN and in similar phase; however, during sleep, it appears more anticorrelated and in opposite phase. Subcortical network also changes phase relative to DMN during sleep, implying these 2 networks are heavily impacted by consciousness. The cPCA results reveal distinct differences across conditions, particularly for PC2, a network linked to arousal and DMN/TPN anticorrelation. PC2 is typically more active during task, not known for high activation in sleep (Hong, et al., 2021), but is more likely to occur than PC3. Further studies are needed on the sleep stage data and its relation to QPP and component proportionality.
Modeling and Analysis Methods:
Connectivity (eg. functional, effective, structural)
fMRI Connectivity and Network Modeling
Task-Independent and Resting-State Analysis 1
Novel Imaging Acquisition Methods:
BOLD fMRI
Perception, Attention and Motor Behavior:
Sleep and Wakefulness 2
Keywords:
Data analysis
Electroencephaolography (EEG)
FUNCTIONAL MRI
MRI
Sleep
Statistical Methods
Other - dynamic analysis
1|2Indicates the priority used for review
Provide references using author date format
Thompson, G. J., et al. (2015). ‘Different dynamic resting state fMRI patterns are linked to different frequencies of neural activity.’ Journal of Neurophysiology, 114(1), 114–124.
Majeed, W., et al. (2011). Spatiotemporal dynamics of low frequency BOLD fluctuations in rats and humans. NeuroImage, 54(2), 1140–1150.
Majeed, W., et al. (2009). ‘Spatiotemporal dynamics of low frequency fluctuations in BOLD fMRI of the rat.’ Journal of Magnetic Resonance Imaging, 30(2), 384–393.
Mitra, A., et al. (2015) ‘Propagated infra-slow intrinsic brain activity reorganizes across wake and slow wave sleep’. eLife 4:e10781.
Fan, L., et al. (2016). ‘The Human Brainnetome Atlas: A New Brain Atlas Based on Connectional Architecture.’ Cerebral Cortex, 26(8), 3508–3526.
Thomas Yeo, B. T., et al. (2011). The organization of the human cerebral cortex estimated by intrinsic functional connectivity. Journal of Neurophysiology, 106(3), 1125–1165.
Gu, Y., et al. ‘Simultaneous EEG and fMRI signals during sleep from humans.’ OpenNeuro. [Dataset] doi: doi:10.18112/openneuro.ds003768.v1.0.11
Hong, C., et al. (2021). ‘fMRI evidence for default mode network deactivation associated with rapid eye movements in sleep.’ Brain Sciences 11.11 (2021): 1528.