Quantitative evaluation of simulated functional brain networks in graph theoretical analysis

Abstract

There is increasing interest in the potential of whole-brain computational models to provide mechanistic insights into resting-state brain networks. It is therefore important to determine the degree to which computational models reproduce the topological features of empirical functional brain networks. We used empirical connectivity data derived from diffusion spectrum and resting-state functional magnetic resonance imaging data from healthy individuals. Empirical and simulated functional networks, constrained by structural connectivity, were defined based on 66 brain anatomical regions (nodes). Simulated functional data were generated using the Kuramoto model in which each anatomical region acts as a phase oscillator. Network topology was studied using graph theory in the empirical and simulated data. The difference (relative error) between graph theory measures derived from empirical and simulated data was then estimated. We found that simulated data can be used with confidence to model graph measures of global network organization at different dynamic states and highlight the sensitive dependence of the solutions obtained in simulated data on the specified connection densities. This study provides a method for the quantitative evaluation and external validation of graph theory metrics derived from simulated data that can be used to inform future study designs.

Neuronal Degeneration Impairs Rhythms Between Connected Microcircuits

“…More broadly, this work raises a new dimension of heterogeneity of TBI where the pattern of cellular damage may contribute to the specific outcome and impairment. In future work, this complexity could be explored with a multiscale approach which integrates local, time-varying signal information as inputs to oscillator-based models of macroscale brain connectivity (Váša et al, 2015;Lee et al, 2017).…”

Linking functional connectivity and dynamic properties of resting-state networks

“…There are alternative computational models that could be used to simulate the dynamics of both amplitudes and phases of oscillations, such as the Hopf model 22 – 24 , the Wilson-Cowan model 25 , 26 , the FitzHugh-Nagumo model 27 , and the Jansen-Rit model 28 . We chose the Kuramoto model because of its relative simplicity and ability to capture essential aspects of phase dynamics 12 , 16 , 29 – 33 . Additionally, the Kuramoto model has been successfully used in neuroscience research for the purpose of modeling slow 29 , 34 , 35 and fast 12 , 16 , 33 , 36 – 39 cortical oscillations.…”

“…We chose the Kuramoto model because of its relative simplicity and ability to capture essential aspects of phase dynamics 12 , 16 , 29 – 33 . Additionally, the Kuramoto model has been successfully used in neuroscience research for the purpose of modeling slow 29 , 34 , 35 and fast 12 , 16 , 33 , 36 – 39 cortical oscillations. Here, we sought to evaluate the performance of the model with fast local gamma-band oscillations, in agreement with experimental 40 – 43 and theoretical models of neural networks 44 .…”

Quantifying differences between passive and task-evoked intrinsic functional connectivity in a large-scale brain simulation

“…Global efficiency is calculated as the average inverse shortest path length 740 (Rubinov & Sporns, 2010). Local efficiency is the inverse of the average shortest path 741 connecting a given node to its neighbors (Lee et al, 2017). Clustering coefficient (Watts & 742 Strogatz, 1998) is a measure of "functional segregation" (Rubinov & Sporns, 2010).…”