Functional network resilience and its relationship to cognition: A cross-sectional lifespan study
Poster No:
1798
Submission Type:
Abstract Submission
Authors:
Georgette Argiris1, Yaakov Stern2, Christian Habeck1
Institutions:
1Columbia University, New York, NY, 2Columbia University Irving Medical Center, New York, NY
First Author:
Co-Author(s):
Introduction:
Functional network studies have shown that healthy brain topology is typically characterized by specific network properties such as high clustering between nodes with a common functionality and short path length between nodal "hubs", purportedly allowing for the efficient and reliable propagation of information across the network [1]. An additional defining feature is resilience, or its ability to retain functionality even when confronted with perturbations, such as lesions, that result in damage or error [2]. Few studies have tested brain network resilience via systematic targeted attack. In the current study, we aimed to assess the impact of virtual nodal lesioning across several network properties of resting BOLD connectivity and to analyze its cross-sectional relationship to age and cognition.
Methods:
Four hundred and twenty-five native English participants from the Reference Ability Neural Network and Cognitive Reserve lifespan cohort (mean age= 50.98 ± 16.53; range= 20-80 years) were included in the study. We generated undirected weighted adjacency matrices based on the time series correlations of all nodal pairings from the 400-region parcellation scheme of the Schaefer atlas [3], with matrices thresholded to retain the top 10% of participant-specific connection strength [4]. As a measure of whole-brain network resilience, we adopted a targeted attack approach described by Albert and colleagues [5], whereby nodes were sequentially removed from the connectome in order of their nodal strength. At each iteration, the nodal strength was recomputed to account for the effect of prior lesioning and a series of network metrics- the largest connected component (LCC), mean clustering coefficient, mean characteristic path length, global efficiency, and modularity- were calculated (see Fig for example schematic). Two properties of each metric were considered: (a) the critical point of lesioning curve; and (b) the total area under the curve (AUC) as the integral of each network measure across lesioning iterations. Linear regression analysis was used to first test for age and brain integrity (i.e., cortical thickness) effects on network measure properties after adjusting for sex, education, NART IQ, scrubbing of the signal, and the "static" network measure (e.g., full connectome metric calculation before lesioning). Models were then created to test the utility of each network measure in predicting out-of-scanner behavioral performance in each of four cognitive domains (i.e., episodic memory, fluid reasoning, processing speed, and vocabulary) considered to comprise the breadth of breadth of age-related cognitive changes [6].
Results:
Results from network measure regression models demonstrated age-related declines in the AUC of the mean clustering coefficient, modularity, and global efficiency. Notably, these effects were found after controlling for static network measures. Additionally, the AUC of modularity but not static modularity displayed age-related effects, suggesting unique network information captured by the AUC and not static measure alone. No significant effects were found for the critical point in the lesioning curve. Results from the behavior models demonstrated a significant positive effect of the mean clustering coefficient, global efficiency, and mean characteristic path length on behavioral performance for the fluid reasoning domain such that higher values of AUC predicted more accurate performance. Additionally, critical point of both the mean characteristic path length and global efficiency both positively predicted fluid reasoning performance, independently of the static network measure.
Conclusions:
Here, we demonstrated age-related declines in specific network metrics that not only partially corroborates previous findings in the aging network literature [4, 7, 8], but additionally supports the inclusion of other network metric properties that may be sensitive to capturing both age-related and cognitive declines beyond static measures alone.
Higher Cognitive Functions:
Reasoning and Problem Solving
Lifespan Development:
Aging 2
Modeling and Analysis Methods:
fMRI Connectivity and Network Modeling 1
Keywords:
Aging
Cognition
FUNCTIONAL MRI
Other - resting state connectivity
1|2Indicates the priority used for review
Provide references using author date format
[2] Gao, J. et al. (2016). Universal resilience patterns in complex networks. Nature, 530(7590), 307-312.
[3] Schaefer, A. et al. (2018). Local-global parcellation of the human cerebral cortex from intrinsic functional connectivity MRI. Cerebral Cortex, 28(9), 3095-3114.
[4] Menardi, A. et al. (2021). Heritability of brain resilience to perturbation in humans. Neuroimage, 235, 118013.
[5] Albert, R. et al. (2000). Error and attack tolerance of complex networks. Nature, 406, 378–382.
[6] Salthouse, T. A. (2009). Decomposing age correlations on neuropsychological and cognitive variables. Journal of the International Neuropsychological Society, 15(5), 650-661.
[7] Deery, H. A., Di Paolo, R., Moran, C., Egan, G. F., & Jamadar, S. D. (2023). The older adult brain is less modular, more integrated, and less efficient at rest: A systematic review of large‐scale resting‐state functional brain networks in aging. Psychophysiology, 60(1), e14159.
[8] Argiris, G., Stern, Y., Lee, S., Ryu, H., & Habeck, C. (2023). Simple Topological Task-Based Functional Connectivity Features Predict Longitudinal Behavioral Change of Fluid Reasoning in the RANN Cohort. NeuroImage, 120237.