An interpretable connectivity-based decoding model for classification of chronic marijuana use

Abstract

BackgroundPsychiatric neuroimaging typically proceeds with one of two approaches: encoding models, which aim to model neural mechanisms, and decoding models, which aim to predict behavioral or clinical features from brain data. In this study, we seek to combine these aims by developing interpretable decoding models that offer both accurate prediction and novel neural insight, using substance use disorder as a test case.MethodsChronic marijuana (MJ) users (n=195) and non-using healthy controls (n=128) completed a cue-elicited craving task during functional magnetic resonance imaging. Linear machine learning algorithms were used to classify MJ use based on task-evoked, whole-brain functional connectivity. We then used graph theoretical measures to explore ‘predictive functional connectivity’ and to elucidate whole-brain regional and network involvement implicated in chronic marijuana use.ResultsWe obtained high accuracy (~80% out-of-sample) across four different linear models, demonstrating that task-evoked, whole-brain functional connectivity can successfully differentiate chronic marijuana users from non-users. Subsequent network analysis revealed key predictive regions (e.g., anterior cingulate cortex, dorsolateral prefrontal cortex, precuneus) that are often found in neuroimaging studies of substance use disorders, as well as some key exceptions – such as sensorimotor and visual areas. We also identified a core set of networks of brain regions that contributed to successful classification, comprised of many of the same predictive regions.ConclusionsOur dual aims of accurate prediction and interpretability were successful, producing a predictive model that also provides interpretability at the neural level. This novel approach may complement other predictive-exploratory approaches for a more complete understanding of neural mechanisms in drug use disorders and other neuropsychiatric disorders.