Physiological Signatures Across the Brain
Poster No:
1978
Submission Type:
Abstract Submission
Authors:
Roza Bayrak1, Nafis Ahmed1, Mara Mather2, Catie Chang1
Institutions:
1Vanderbilt University, Nashville, TN, 2University of Southern California, Los Angeles, CA
First Author:
Co-Author(s):
Introduction:
While fluctuations in low-frequency systemic physiology (e.g., respiration and cardiac activity) are often treated as a nuisance component in fMRI research, an increasing number of studies indicate that they contain meaningful untapped information about brain physiology and autonomic function [1,2,3,4], Yet, the behavioral relevance of systemic, low-frequency BOLD effects, and their large-scale patterns across the brain, are largely underexplored. This study aims to fill this gap by systematically examining the patterns of peripheral physiological influences in fMRI signals and their association to individual differences in behavior.
Methods:
Dataset: A subset of 375 subjects from the HCP S1200 release was utilized based on the quality of their physiological recordings [5,6]. Dataset included: resting-state data (4 scans/subject, 2 days with 2 runs on each day) and 51 cognitive measures based on the exclusion criteria from [7] and their availability in the HCP "unrestricted" behavioral assessments.
Data Prep: The % of temporal variance accounted for by respiratory volume (RV) and heart rate (HR) regressors convolved with respiratory and cardiac response function basis sets [8,9] was calculated at each fMRI voxel. The percent variance explained (PVE) maps were deconfounded for sex, height, weight, intracranial volume, brain size, and average movement in scanner.
Heritability Analysis: First, we assess the spatial similarity of BOLD physiological patterns among family members and other subjects. Next, we use SOLAR-Eclipse imaging genetic analysis package [10] to quantify this similarity. Phenotype values for each individual within the cohort were adjusted for covariates including sex, age, age2, age × sex interaction, age2 × sex interaction, height, weight, intracranial volume, brain size, and average movement in scanner.
Canonical Correlation Analysis: To investigate a linear association between BOLD physiological patterns and behavioral/cognitive variables, we employed CCA. Prior to CCA dimensionality reduction of both brain and behavior data was carried out using PCA with 30 brain and 2 behavioral PCs (a sensitivity analysis was performed on the number of PCs, not shown).
Data Prep: The % of temporal variance accounted for by respiratory volume (RV) and heart rate (HR) regressors convolved with respiratory and cardiac response function basis sets [8,9] was calculated at each fMRI voxel. The percent variance explained (PVE) maps were deconfounded for sex, height, weight, intracranial volume, brain size, and average movement in scanner.
Heritability Analysis: First, we assess the spatial similarity of BOLD physiological patterns among family members and other subjects. Next, we use SOLAR-Eclipse imaging genetic analysis package [10] to quantify this similarity. Phenotype values for each individual within the cohort were adjusted for covariates including sex, age, age2, age × sex interaction, age2 × sex interaction, height, weight, intracranial volume, brain size, and average movement in scanner.
Canonical Correlation Analysis: To investigate a linear association between BOLD physiological patterns and behavioral/cognitive variables, we employed CCA. Prior to CCA dimensionality reduction of both brain and behavior data was carried out using PCA with 30 brain and 2 behavioral PCs (a sensitivity analysis was performed on the number of PCs, not shown).
Results:
BOLD physiological patterns displayed high within-subject reliability compared to a null distribution, both within and between days. Heritability (h2) is computed using SOLAR-eclipse. h2 is the proportion of the total phenotypic variance that can be explained by the genetic effects. In Fig.1, (left) voxel level maps indicates the h2, thresholded at a p<0.05 uncorrected threshold, (right) network-level averages of h2 are shown for left and right hemispheres. CCA yielded a significant first canonical mode between the physiological patterns and the phenotypic profiles of the population (p<0.018). Fig.2 displays (left) the first canonical mode and its maps of the brain CCA weights, and (right) the top 10 positive and negative CCA behavioral weights.
Conclusions:
In this study, we explored the potential of BOLD physiological signatures to serve as predictors of cognitive and behavioral variables. The variability observed in physiological signals captured by BOLD signals may reflect unique characteristics of individual subjects. By considering and analyzing these measures, we gain valuable insights into the complex interplay between brain function, individual differences, and behavioral outcomes. Heritability analysis provides valuable information on the potential influence of genetic and environmental factors on the observed spatial physiological patterns. Physiological signals are closely linked to brain function and have connections to behavior. Removal of these signals should therefore depend on the study.
Modeling and Analysis Methods:
Multivariate Approaches 1
Task-Independent and Resting-State Analysis
Novel Imaging Acquisition Methods:
BOLD fMRI
Physiology, Metabolism and Neurotransmission :
Physiology, Metabolism and Neurotransmission Other 2
Keywords:
Data analysis
Multivariate
Open Data
Open-Source Code
1|2Indicates the priority used for review
Provide references using author date format
[2] Mather & Thayer 2018, PMID: 29333483
[3] Yuan et al. 2013, PMID: 23631982
[4] Shams et al. 2021, PMID: 33516896
[5] Power et al. 2019, PMID: 31589990
[6] Xifra-Porxas et al. 2021, PMID: 34342582
[7] Li et al. 2019, PMID: 31582792
[8] Chang et al. 2009, PMID: 18951982
[9] Birn et al. 2008, PMID: 18234517
[10] Kochunov et al. 2015, PMID: 25812717