Resting state networks in empirical and simulated dynamic functional connectivity
Katharina Glomb1, Adrián Ponce-Alvarez1, Matthieu Gilson1
1Center for Brain and Cognition, Dept. of Technology and Information, Universitat Pompeu Fabra, Carrer Ramon Trias Fargas, 25-27, 08005 Barcelona, Spain.
Neuroimage
|August 8, 2017
Summary
Stationary brain dynamics can explain the emergence of resting state networks (RSNs). Our novel tensor decomposition method reveals these patterns in functional connectivity (FC) data.
Area of Science:
- Neuroimaging
- Computational Neuroscience
- Network Science
Background:
- Resting-state functional connectivity (FC) patterns in the human brain are dynamic, not static.
- Understanding the temporal dynamics governing FC transitions is crucial but challenging.
- Distinguishing between stationary and non-stationary dynamics in brain activity remains an active research area.
Purpose of the Study:
- To investigate whether stationary dynamics can explain the emergence of distinct functional connectivity patterns (RSNs) and their temporal dynamics.
- To develop and apply an innovative whole-brain approach combining tensor decomposition and a mean-field model.
- To characterize spatio-temporal dynamics in resting-state fMRI data.
Main Methods:
- Applied tensor decomposition to resting-state fMRI data from 24 healthy controls, analyzing FC within sliding windows.
- Created tensors combining temporal and spatial information to identify brain regions with similar temporal dynamics (communities).
- Simulated data using a stationary mean-field model with DTI-based connectivity and analyzed it using the same tensor decomposition method.
Main Results:
- Identified four distinct communities resembling known resting state networks (default mode, visual, control, somatomotor) in empirical data.
- Demonstrated that the stationary mean-field model, when connected via effective connectivity, can reproduce these four RSNs and their time courses.
- Found that only the strongest functional connectivity values across time and ROI pairs are necessary to explain RSN emergence.
Conclusions:
- Stationary dynamics, driven by noisy fluctuations around average functional connectivity, can account for the emergence of resting state networks.
- The developed tensor decomposition method offers an innovative, assumption-light approach for analyzing spatio-temporal dynamics in brain imaging data.
- The findings have broad applicability to resting-state and task-based fMRI data across diverse subject populations.
Keywords:
Dynamic functional connectivityFeature extractionFunctional connectivityHumanMean field modelsTensor decompositionWhole-brain modelsfMRI
