Directed Network Discovery with Dynamic Network Modeling

Abstract

Cognitive tasks recruit multiple brain regions. Understanding how these regions influence each other (the network structure) is an important step to characterize the neural basis of cognitive processes. Often, limited evidence is available to restrict the range of hypotheses a priori, and techniques that sift efficiently through a large number of possible network structures are needed (network discovery). This article introduces a novel modeling technique for network discovery (Dynamic Network Modeling or DNM) that builds on ideas from Granger Causality and Dynamic Causal Modeling introducing three key changes: 1) regularization is exploited for efficient network discovery, 2) the magnitude and sign of each influence are tested with a random effects model across participants, and 3) variance explained in independent data is used as an absolute (rather than relative) measure of the quality of the network model. In this article, we outline the functioning of DNM and we report an example of its application to the investigation of influences between regions during emotion recognition. Across two experiments, DNM individuates a stable set of influences between face-selective regions during emotion recognition.\n\nNew and NoteworthyIn this article we introduce a new analysis method (Dynamic Network Mod- elling or DNM) which exploits{ell} 1 regularization to perform efficient for network discovery. DNM provides information about the direction and sign (inhibitory vs excitatory) of influences between brain regions, and generates measures of variance explained in independent data to evaluate quality of fit. The method is applied to brain regions engaged in emotion recognition, individuating a similar network structure across two separate experiments.