Connectome-based modeling of real world clinical outcomes in addictions

Sarah Yip
Yale University

Recent advances in predictive modeling and machine learning methods enable data-driven prediction of complex behaviors. By focusing on individual differences and generalization to novel subjects (i.e., cross-validation) these approaches overcome limitations of traditional approaches, increasing the likelihood of replication and potential translation to clinical settings. However, these methods are still relatively new to clinical research and have yet to be fully leveraged within the context of addictions. This talk will present recent data using connectome-based predictive modeling (CPM) to predict real-world clinical outcomes among individuals with polysubstance use. CPM is a machine learning method of generating predictive models of behavioral data (e.g., relapse) based on individual patterns of brain organization. CPM is data-driven and requires no a priori selection of networks. It is therefore both a predictive tool and a method of identifying specific networks underlying behavior. This talk will present (i) CPM work identifying a distributed network that predicts cocaine abstinence during 12 week treatment, with multiple external replications; (ii) more recent work using this approach to predict opioid abstinence; (iii) evidence for network stability over time; and (iv) extension of findings across brain states. Evidence for dissociable anatomical substrates of different types of substance use (e.g., cocaine vs. opioids) will also be presented.

Presentation (PDF File)

Back to Computational Psychiatry