Dynamical differential covariance recovers directional network structure in multiscale neural systems

Abstract

Investigating causal neural interactions are essential to understanding sub- sequent behaviors. Many statistical methods have been used for analyzing neural activity, but efficiently and correctly estimating the direction of net- work interactions remains difficult. Here, we derive dynamical differential covariance (DDC), a new method based on dynamical network models that detects directional interactions with low bias and high noise tolerance with- out the stationary assumption. The method is first validated on networks with false positive motifs and multiscale neural simulations where the ground truth connectivity is known. Then, applying DDC to recordings of resting-state functional magnetic resonance imaging (rs-fMRI) from over 1,000 individual subjects, DDC consistently detected regional interactions with strong structural connectivity. DDC can be generalized to a wide range of dynamical models and recording techniques.

Functional connectivity of fMRI using differential covariance predicts structural connectivity and behavioral reaction times

“…Although the computational processes in brains are by no means linear, the averaging of activity in large neural populations, as in the case of fMRI recordings, is approximately linear ( Stephan et al, 2008 ). The current linear dCov estimator could be extended to nonlinear estimators, taking into account the threshold firing properties of individual neurons, to better extract connectivity at the single neuron level ( Chen, Rosen, & Sejnowski, 2021 ).…”

“…Recent efforts have been made toward a dynamic description of directed connectivity, such as dynamic effective connectivity ( Zarghami & Friston, 2020 ), and dCov is another step toward this direction. First, the derivative signal reflects a dynamical description of the system ( Chen et al, 2021 ) and is also used in regression dynamic causal modeling (rDCM) ( Frässle et al, 2017 ). In comparison, dCov is more intuitive and faster than rDCM.…”

“…For the structural connectivity of human subjects, we used the individual level SC matrices constructed in our previous work Chen et al (2021) (Refer to method 4.3 in the reference). In short, we used probabilistic diffusion tractography to evaluate the dMRI measurements Rosen and Halgren (2021).…”