The virtual ageing brain: structure-function relationship and cognitive decline in longitudinal data
Poster No:
1202
Submission Type:
Abstract Submission
Authors:
Jan Fousek1, Christiane Jockwitz2, Nora Bittner2, Meysam Hashemi3, Spase Petkoski3, Svenja Caspers2, Viktor Jirsa3
Institutions:
1CEITEC, Masaryk University, Brno, Czech Republic, 2Institute for Anatomy I, Medical Faculty, Heinrich-Heine University, Düsseldorf, Germany, 3Institut de Neurosciences des Systèmes, Aix-Marseille University, Marseille, France
First Author:
Co-Author(s):
Christiane Jockwitz
Institute for Anatomy I, Medical Faculty, Heinrich-Heine University
Düsseldorf, Germany
Institute for Anatomy I, Medical Faculty, Heinrich-Heine University
Düsseldorf, Germany
Nora Bittner
Institute for Anatomy I, Medical Faculty, Heinrich-Heine University
Düsseldorf, Germany
Institute for Anatomy I, Medical Faculty, Heinrich-Heine University
Düsseldorf, Germany
Svenja Caspers
Institute for Anatomy I, Medical Faculty, Heinrich-Heine University
Düsseldorf, Germany
Institute for Anatomy I, Medical Faculty, Heinrich-Heine University
Düsseldorf, Germany
Introduction:
The human brain changes during healthy aging with large individual variation in the cognitive decline (Oschwald 2020). Sources and mechanisms of this variability have been previously linked with whole-brain reorganisation, specifically in terms of the white-matter fibre tracts (structural connectivity, SC), and functional co-activations (functional connectivity, FC) of brain regions (Suárez 2020, Cabeza 2002), e.g. hemispheric asymmetry reduction. Recently, a causal framework employing whole-brain modelling where the individual SC informs a computational brain network model (Lavanga 2023) was developed on cross-sectional data of a large ageing cohort. Here, we validate the virtual ageing brain on the longitudinal data, and use it to unfold the variability of the individual cognitive decline.
The prediction from the cross-sectional study was such that the subjects with stronger cognitive decline will move away from their optimal working point in terms of the network coupling, that is have a smaller slope of the change in parameter G than the subjects with well maintained cognitive performance.
The prediction from the cross-sectional study was such that the subjects with stronger cognitive decline will move away from their optimal working point in terms of the network coupling, that is have a smaller slope of the change in parameter G than the subjects with well maintained cognitive performance.
·Figure 1: Prediction from the cross-sectional study.
Methods:
We used the SC and regional BOLD (Schaefer parcellation with 100 regions) signal of older subjects from the 1000BRAINS dataset (Caspers 2014) (full cohort: n=649, age range [51.1−85.4], nfemales=317; subjects with a follow-up: n=220, mean Δt=3.8 years). In particular, we have focused on data features with significant cross-sectional trends: from the SC we computed the average weight of the inter-hemispheric connections, from the FC we used the mean of weights of the homotopic connections, and from the FC dynamics (FCD, that is FC in a sliding window) we computed fluidity: variance of the upper triangle of the FCD (Lavanga 2023).
The brain network model implemented in The Virtual Brain (Schirner 2022) was constructed from individual SC with the neural mass model (Montbrió 2015) governing the node dynamics enabling simulation of resting-state BOLD data. For each subject we employed Simulation Based Inference (SBI) to compute the posterior estimate of the global modulation from the empirical functional data (Gonçalves 2020). To assess the longitudinal change of the empirical data features and estimated parameters of the model, we computed the rate of change for the individual variables as X(t2)-X(t1) / (t2-t1).
The brain network model implemented in The Virtual Brain (Schirner 2022) was constructed from individual SC with the neural mass model (Montbrió 2015) governing the node dynamics enabling simulation of resting-state BOLD data. For each subject we employed Simulation Based Inference (SBI) to compute the posterior estimate of the global modulation from the empirical functional data (Gonçalves 2020). To assess the longitudinal change of the empirical data features and estimated parameters of the model, we computed the rate of change for the individual variables as X(t2)-X(t1) / (t2-t1).
Results:
The cross-sectional study relied on three pillars: structural and functional data features, individualized model parameters, and cognitive scores. Compared to the cross-sectional trends, the rates of change in the longitudinal dataset (Figure 2A,B) were preserved for the structural data (decrease of interhemispheric connections), estimated model parameters (increase of the global coupling strength), and in the cognitive scores (increased time to complete the selective attention task). Furthermore, for the older subjects (age >67) the change in estimated global coupling between the two sessions was higher for the subjects with good cognitive performance (Figure 2C).
·Figure 2: Longitudinal trends in SC, estimated coupling parameter and cognitive score (A,B), lower coupling parameter G for cognitively lower performing subjects (C).
Conclusions:
Our results indicate that the deterioration of the interhemispheric SC is accompanied by increased modulation of the functional brain dynamics. The SC reorganisation might reflect a potential scaffolding of the brain during the ageing process. This effect is weaker for the cognitively well performing subjects, which suggests a process of brain maintenance. The decline in functional data features was not significant in the relatively short time span between the two visits, however it was sufficiently informative on the individual level for the estimation of the global coupling. The confrontation of this framework with the longitudinal dataset provides an important validity check as well as it opens an opportunity for further extensions by including other factors beyond 'age' and more nuanced relationships between aspects of cognitive decline and the brain changes.
Lifespan Development:
Aging 1
Modeling and Analysis Methods:
Connectivity (eg. functional, effective, structural)
fMRI Connectivity and Network Modeling 2
Keywords:
Aging
Computational Neuroscience
Data analysis
FUNCTIONAL MRI
Machine Learning
Modeling
1|2Indicates the priority used for review
Provide references using author date format
Suárez, L. E. (2020). Linking Structure and Function in Macroscale Brain Networks. Trends in Cognitive Sciences, 24(4), 302–315.
Cabeza, R. (2002). Hemispheric asymmetry reduction in older adults: The HAROLD model. Psychology and Aging, 17(1), 85–100.
Lavanga, M. (2023). The virtual aging brain: Causal inference supports interhemispheric dedifferentiation in healthy aging. NeuroImage, 283, 120403.
Caspers, S. (2014). Studying variability in human brain aging in a population-based German cohort-rationale and design of 1000BRAINS. Frontiers in Aging Neuroscience, 6, 149.
Schirner, M. (2022). Brain simulation as a cloud service: The Virtual Brain on EBRAINS. NeuroImage, 251, 118973.
Montbrió, E. (2015). Macroscopic Description for Networks of Spiking Neurons. Physical Review X, 5(2), 021028.
Gonçalves, P. J. (2020). Training deep neural density estimators to identify mechanistic models of neural dynamics. eLife, 9.