The Left Insula Bridges Cognition, Emotion, and Brain Structure: A Multilayer Network Analysis of the Human Connectome Project-Young Adult

Abstract

Multilayer network analyses allow for the exploration of complex relationships across different modalities. Specifically, this study employed a novel method that integrates psychometric networks with structural covariance networks to explore the relationships between cognition, emotion and the brain. Psychological (NIH Toolbox Cognition Battery and NIH Toolbox Emotion Battery) and anatomical MRI (cortical volume) data were extracted from the Human Connectome Project Young Adult dataset (n = 1109). Partial correlation networks with graphical lasso regularisation and extended Bayesian information criterion tuning were used to model a psychometric bi-layer network consisting of seven cognitive nodes and four emotion nodes, as well as a neuro-psychometric tri-layer network consisting of these same nodes in addition to 24 brain nodes from the central executive and salience networks. Bridge strength centrality was used to identify nodes that bridged between layers. For the bi-layer network, it was found that stress was the only bridge node. For the tri-layer network, six bridge nodes were identified, with the left insula emerging as the most central. These findings demonstrate the utility of multilayer networks in integrating psychological and neurobiological data for the potential identification of targets to improve psychological wellbeing.

Figure 1.

(A) Grey matter ROIs based on the Desikan-Killiany atlas; depicted bilaterally. (B) ROIs according to their network.

Figure 2.

Overview of the analysis steps to generate multilayer networks and identify bridge nodes. First, individual variables for each layer (i.e., cognition, emotion and brain) were entered into a single correlation matrix. Graphical lasso model selection was then applied to generate the multilayer network from this matrix. Bridge strength was subsequently used to determine nodes that were important bridges between layers. “R” = right hemisphere, “L” = left hemisphere. “Z” = z-score. Note: This figure pertains specifically to the neuro-psychometric tri-layer network, while the psychometric bi-layer network followed the same steps but did not include the brain data.

Figure 3.

Bi-layer network (n = 1109). Nodes (circles) are variables from the NIH Toolbox Cognition Battery (cognition; colour coded green) and NIH Toolbox Emotion Battery (emotion; colour coded blue). Undirected edges (lines between nodes) represent partial correlations between variables, after controlling for all other variables, with thicker lines representing stronger relationships. Blue edges represent positive relationships, whereas red edges represent negative relationships. A magenta circle indicates the node with a bridge centrality z-score ≥ 1. Abbreviations for node labels are defined in Table 3.

Figure 4.

Bar graph of bridge strength centrality scores (n =1109). The dashed line indicates a z-score of 1, nodes at or above this line are interpreted as bridge nodes. Abbreviations for node labels are defined in Table 3.

Figure 5.

Tri-layer network (n = 1109). Nodes (circles) are variables from the NIH Toolbox Cognition Battery (cognition; colour coded green), NIH Toolbox Emotion Battery (emotion; colour coded blue), and the cortical volume of ROIs from the central executive and salience networks (Brain; colour coded purple). Undirected edges (lines between nodes) represent partial correlations between variables, after controlling for all other variables, with thicker lines representing stronger relationships. Blue edges represent positive relationships, whereas red edges represent negative relationships. Magenta circles around nodes indicate that the node has a bridge centrality z-score ≥ 1. “l” before a ROI indicates the left hemisphere, “r” before a ROI indicates the right hemisphere. Abbreviations for node labels are defined in Table 3.

Figure 6.

Bar graph of bridge strength centrality scores (n =1109). The dashed line indicates a z-score of 1, nodes at or above this line are interpreted as bridge nodes. “l” before a ROI indicates the left hemisphere, “r” before a ROI indicates the right hemisphere. Abbreviations for node labels are defined in Table 3.