Functional Brain Network of Trait Impulsivity: Whole-Brain Functional Connectivity Predicts Self-Reported Impulsivity

Abstract

Given impulsivity's multidimensional nature and its implications across various aspects of human behavior, a comprehensive understanding of functional brain circuits associated with this trait is warranted. In the current study, we utilized whole-brain resting-state functional connectivity data of healthy males (n = 156) to identify a network of connections predictive of an individual's impulsivity, as assessed by the Barratt Impulsiveness Scale (BIS)-11. Our participants were selected, in part, based on their self-reported BIS-11 impulsivity scores. Specifically, individuals who reported high or low trait impulsivity scores during screening were selected first, followed by those with intermediate impulsivity levels. This enabled us to include participants with rare, extreme scores and to cover the entire BIS-11 impulsivity spectrum. We employed repeated K-fold cross-validation for feature-selection and used stratified 10-fold cross-validation to train and test our models. Our findings revealed a widespread neural network associated with trait impulsivity and a notable correlation between predicted and observed scores. Feature importance and node degree were assessed to highlight specific nodes and edges within the impulsivity network, revealing previously overlooked key brain regions, such as the cerebellum, brainstem, and temporal lobe, while supporting previous findings on the basal ganglia-thalamo-prefrontal network and the prefrontal-motor strip network in relation to impulsiveness. This deepened understanding establishes a foundation for identifying alterations in functional brain networks associated with dysfunctional impulsivity.

Keywords: impulsivity; individual differences; machine learning; resting‐state functional connectivity.

FIGURE 1

Distribution of BIS‐11 scores in the present sample.

FIGURE 2

Proposed pipeline for construction of trait impulsivity brain network and prediction of individual BIS‐11 scores using whole‐brain functional connectivity data.

FIGURE 3

Quantile Binning of BIS‐11 scores for stratified k‐fold cross‐validation.

FIGURE 4

Functional connections predicting individual trait impulsivity. (A) all common edges contributing to the final model. (B) edges with a positive association with the BIS‐11. (C) edges with a negative association with the BIS‐11. (D and E) number of edges connecting each pair of macroscale brain regions contributing to the BIS‐11 network, for the positive (red) and negative (blue) associations, respectively.

FIGURE 5

Model evaluation (A) predicted BIS‐11 scores and their residuals. (B) observed and predicted BIS‐11 scores.

FIGURE 6

Top 20 node pairs of the model as determined by their feature importance (A). (B) Macroscale brain regions of most important features. Red lines indicate positive BIS‐11 associations and turquoise lines indicate negative BIS‐11 associations.