Decoding reappraisal and suppression from neural circuits: A combined supervised and unsupervised machine learning approach

Emotion regulation is a core construct of mental health and deficits in emotion regulation abilities lead to psychological disorders. Reappraisal and suppression are two widely studied emotion regulation strategies but, possibly due to methodo- logical limitations in previous studies, a consistent picture of the neural correlates related to the individual differences in their habitual use remains elusive. To address these issues, the present study applied a combination of unsupervised and supervised machine learning algorithms to the structural MRI scans of 128 individuals. First, unsupervised machine learning was used to separate the brain into naturally grouping grey matter circuits. Then, supervised machine learning was applied to predict individual differences in the use of different strategies of emotion regulation. Two predictive models, including structural brain features and psychological ones, were tested. Results showed that a temporo-parahippocampal-orbitofrontal network successfully predicted the individual differences in the use of reappraisal. Differently, insular and fronto-temporo- cerebellar networks successfully predicted suppression. In both predictive models, anxiety, the opposite strategy, and specific emotional intelligence factors played a role in predicting the use of reappraisal and suppression. This work provides new insights regarding the decoding of individual differences from structural features and other psychologically relevant variables while extending previous observations on the neural bases of emotion regulation strategies.