A Connectome-based Predictive Model of Affective Experience During Naturalistic Viewing

Abstract

: Our thoughts and actions are guided by our ongoing affective experience. Affective states are often measured using self-report ratings, which are labor intensive to collect and can also disrupt ongoing cognition if obtained while performing a task. In this study, we aim to 1) derive a continuous and non-intrusive measure of affective experience based on dynamic functional connectivity (FC), and 2) characterize the interaction between brain regions underlying changes in affective states. We trained a connectome-based predictive model to predict subjective ratings of valence, arousal and dominance from fMRI data of participants watching a TV episode. All three models achieved reasonable accuracy (valence: r = .486, p = .018; arousal: r = .519, p = .002; dominance: r = .602, p = .008). FC edges within and between multiple large-scale functional brain networks reliably contributed to model predictions, suggesting that affective states are encoded in the interactions between brain regions. Taken together, our work presents a promising approach to probe affective experience based on brain imaging data.