Theta-alpha aperiodic dynamics covary with persistence homological scaffolds of hemodynamic networks
Poster No:
1808
Submission Type:
Abstract Submission
Authors:
Nghi Nguyen1, Tao Hou2, Li Shen3,4, Duy Duong-Tran4,5
Institutions:
1Gonda Multidisciplinary Brain Research Center, Bar-Ilan University, Ramat Gan, Israel, 2Department of Computer Science, DePaul University, Chicago, IL, 3Department of Biostatistics, Epidemiology and Informatics, University of Pennsylvania, Philadelphia, PA, 4Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA, 5Department of Mathematics, United States Naval Academy, Annapolis, MD
First Author:
Co-Author(s):
Li Shen
Department of Biostatistics, Epidemiology and Informatics, University of Pennsylvania|Perelman School of Medicine, University of Pennsylvania
Philadelphia, PA|Philadelphia, PA
Department of Biostatistics, Epidemiology and Informatics, University of Pennsylvania|Perelman School of Medicine, University of Pennsylvania
Philadelphia, PA|Philadelphia, PA
Duy Duong-Tran, Dr.
Perelman School of Medicine, University of Pennsylvania|Department of Mathematics, United States Naval Academy
Philadelphia, PA|Annapolis, MD
Perelman School of Medicine, University of Pennsylvania|Department of Mathematics, United States Naval Academy
Philadelphia, PA|Annapolis, MD
Introduction:
Cortical aperiodic dynamics have recently gained growing interest due to their implicated roles in cognition, consciousness, aging, diseases, and their relevance in theoretical neuronal network models (Gerster, 2022). Several studies have associated such dynamics with resting-state BOLD signals (Jacob, 2021); however, mechanistic models explaining how that association emerges across cortical regions are still inconclusive. This study proposes that the missing element might lie in regional contributions to maintaining the topological structure of the hemodynamic network, especially how these contributions change between resting and excited states. To quantify these contributions, we used persistence strength scaffold centrality (PC) (Lord, 2016) derived from weight-optimal cycles in persistent homology.
Methods:
We used the resting-state and task-based 3T fMRI data from the Human Connectome Project Young Adult dataset (Van Essen, 2013) and curated and preprocessed by the Neuromatch Academy ('t Hart, 2022). We focused on motor and working memory tasks and the following task conditions: left-hand, right-hand, left-foot, right-foot for motor, and 0-back faces, 0-back tools, 2-back faces, and 2-back tools for working memory. Each sample (n=318) represents the Glasser-parcellated BOLD signals of each subject recorded either at rest or during tasks. Functional connectivity maps are obtained by taking the Pearson correlation coefficients between BOLD time series of every pair of regions.
From each functional connectivity map, we produced a homological scaffold matrix following the pipeline described by Lord (2016) with Vietoris-Rips filtration to track the persistence of cycles. The group-average scaffold matrix for each task condition (and one for the resting state) was used to calculate regional PCs under that task condition (and the resting state). We also computed degree centrality (DC) vectors from the group-average functional connectivity maps for comparative purposes.
To test the relationship between the centrality models and their electromagnetic correlates while ensuring generalizability, we use MEG data from the same HCP dataset under the same task conditions but with different subjects (n = 20). We performed source reconstruction using an LCMV beamformer on BEM surfaces, applied Glasser parcellation, epoched by stimulus onset, and averaged the signals across trials. For the signal from each region, we ran the IRASA algorithm (Wen, 2016) to extract aperiodic activity from 1 Hz to 90 Hz.
From each functional connectivity map, we produced a homological scaffold matrix following the pipeline described by Lord (2016) with Vietoris-Rips filtration to track the persistence of cycles. The group-average scaffold matrix for each task condition (and one for the resting state) was used to calculate regional PCs under that task condition (and the resting state). We also computed degree centrality (DC) vectors from the group-average functional connectivity maps for comparative purposes.
To test the relationship between the centrality models and their electromagnetic correlates while ensuring generalizability, we use MEG data from the same HCP dataset under the same task conditions but with different subjects (n = 20). We performed source reconstruction using an LCMV beamformer on BEM surfaces, applied Glasser parcellation, epoched by stimulus onset, and averaged the signals across trials. For the signal from each region, we ran the IRASA algorithm (Wen, 2016) to extract aperiodic activity from 1 Hz to 90 Hz.
Results:
By comparing task-evoked changes in PC and those in DC, or ΔPC and ΔDC, respectively, we observed that ΔPC was more correlated with changes in cognitive demand than with the nature of the task, while DC was task-specific (Figure 1). Spatial distributions of ΔPC could also be inferred from resting-state PC while exhibiting robust patterns, such as "sinks" concentrated in the Dorsal-Attention network and "sources" in the Auditory network and part of the Cingulo-Opercular network previously identified as the salience network (Ji, 2019). These observations coincide with aperiodic dynamic patterns outlined in previous fMRI-EEG studies (Jacob, 2021). Further corroborating this correlation, cosine similarities between ΔPC and changes in bandpower ratios (ΔBPRs) within the θ-α range were significantly higher than those between ΔBPRs and ΔDC (Figure 2).
·Figure 1. Resting state-subtracted PC (Δ PC) captures task-evoked hemodynamic patterns nonspecific to task conditions, whereas resting state-subtracted DC (Δ DC) is more condition-specific.
·Figure 2. Regional changes in bandpower ratio (Δ BPR) of cortical aperiodic activity between the theta-alpha range (4 - 12 Hz) show a stronger correlation with Δ PC than with Δ DC.
Conclusions:
This study underscores the spatial and temporal correlation between cortical aperiodic dynamics in the θ-α range and PCs, i.e., levels of participation in cycles scaffolding the hemodynamic networks. This insight helps inform more precise BOLD deconvolution models that account for aperiodic activity, as well as a better mechanistic understanding of lower-level processes simultaneously giving rise to neuronal aperiodic dynamics and brain hemodynamics.
Modeling and Analysis Methods:
EEG/MEG Modeling and Analysis 2
fMRI Connectivity and Network Modeling 1
Methods Development
Perception, Attention and Motor Behavior:
Perception and Attention Other
Physiology, Metabolism and Neurotransmission :
Cerebral Metabolism and Hemodynamics
Keywords:
FUNCTIONAL MRI
MEG
Other - Fractal Dynamics; Persistence Homology
1|2Indicates the priority used for review
Provide references using author date format
Jacob, M. S. (2021). Aperiodic measures of neural excitability are associated with anticorrelated hemodynamic networks at rest: a combined EEG-fMRI study. NeuroImage, 245, 118705.
Ji, J. L. (2019). Mapping the human brain's cortical-subcortical functional network organization. Neuroimage, 185, 35-57.
Lord, L. D. (2016). Insights into brain architectures from the homological scaffolds of functional connectivity networks. Frontiers in systems neuroscience, 10, 85.
't Hart, B. (2022). Neuromatch Academy: a 3-week, online summer school in computational neuroscience. Journal of Open Source Education, 5(49), 118.
Van Essen, D. C. (2013). The WU-Minn human connectome project: an overview. Neuroimage, 80, 62-79.
Wen, H. (2016). Separating fractal and oscillatory components in the power spectrum of neurophysiological signal. Brain topography, 29, 13-26.