fMRI prediction of the severity of alcohol use disorder

PROJECT SUMMARY Alcohol use disorder (AUD) remains a leading cause of morbidity and mortality in the United States, and most people who attempt to quit drinking relapse within six months. Since treatment efficacy seems in part to be a function of alcohol use severity, existing treatments could benefit from better brain measures that explicitly

characterize the severity of AUD. Unfortunately, most imaging studies compare individuals with AUD to healthy controls – or if they do assess severity, only do so in one cognitive domain and in one sample. Therefore, we propose two avenues to investigate AUD severity in fMRI data. Both lines of research are influenced by the

observation that the default mode network (DMN) is altered in AUD and in addiction generally. Although the DMN is typically more prominent in resting-state fMRI, analysis of task-based fMRI has suggested a plausible functional role for DMN activity in addiction. High DMN activity seems to attribute high value to commodities like

alcohol, driven by preoccupation during the unsatiated and craving states. Based on these observations, Aim 1 predicts AUD severity from both task-based fMRI maps and resting-state-derived DMN maps. We will examine the relative level at which each task or DMN map encodes AUD severity, as assayed by machine-

learning predictions of AUDIT score. To accomplish this Aim, we will use multiple machine-learning techniques with a secondary goal of performing a head-to-head comparison of different algorithms. Aim 2 directly tests the ability to modulate DMN activity as a function of AUD severity using real-time fMRI. The real-time system

created by the sponsor of this application has been used to demonstrate that healthy participants can successfully learn to gain volitional control over their own DMN activity, and that this ability appears to be impaired in psychiatric conditions. In this Aim, individuals with AUD will see their DMN activity and attempt to

increase and decrease its level prompted by neurofeedback. We will assess whether this ability is a function of AUD severity by correlating the DMN activity with the experimentally-controlled “increase” and “decrease” cues, and then comparing this ability to AUDIT score. This experiment constitutes an experimental medicine approach

to AUD because it may reveal a novel target of severity-informed treatments for AUD. This project is driven by a unique and comprehensive training plan designed to integrate expertise in real-time neuroimaging, behavioral training in AUD, neuroeconomic and computational modeling, and advanced statistical methodology. It

emphasizes the development of technical and programming skills, written and oral communication skills, grant writing, and undergraduate mentorship. Further, this proposal is supported by a sponsor (Dr. LaConte) and cosponsor (Dr. Bickel) whose labs actively collaborate to study neural and behavioral models of addiction.

Therefore, the project will be conducted in an ideal environment to study the severity of AUD and its effects on the brain.