Associations Between Neural Signature Scores and ROI-Based GLM Beta Weights in Alcohol Cue Reactivity
Roundy, Gwenyth
Roundy, Gwenyth
Citations
Altmetric:
License
License
Abstract
Alcohol use disorder (AUD) is the compulsive use of alcohol and heightened sensitivity to alcohol related cues, including neural dysregulation in reward salience, and cognitive control circuits. Understanding the brain�s response to alcohol cues is necessary for identifying neural mechanisms underlying craving and relapse risk. Region of interest (ROI) mean general linear model (GLM) provides a strong approach to examine neural responses by averaging voxel-level task effects within anatomically or functionally defined regions, while reducing noise effects and giving interpretable comparisons across networks. In parallel to this, neural signature modeling can capture disturbed patterns of brain activity by generating subject-level scores that reflect brain responses to task conditions. When combined, these approaches give a complementary perspective on alcohol cue reactivity. Neural responses to alcohol cues were examined in 1184 participants (409 females, 790 AUD) across nineteen datasets from the ENIGMA Addiction Working Group. ROI-mean GLM activation was calculated using combined cortical and subcortical parcellations from the Schafer 2018 17-network, 400-parcel atlas and the Tian S4 3T atlas. Activation estimates were extracted for alcohol cues and alcohol minus neutral contrasts, enabling assessment of regional activation and stimulus specific reactivity. Neural signature scores were derived from beta weights for neutral cues (condition 0) and alcohol cues (condition 1) using cross-validated multivariate modeling using the Schaefer 2018 atlas. Model performance was evaluated using cross-validation metrics and area under the ROC curve (AUC). Subject-level scores were calculated as condition 1 probability minus condition 0 probability. ROI-mean GLM estimates significantly correlated with neural signature scores. This shows a relationship between regionally averaged activation and distributed multivariate representation. This suggests that neural signatures capture variance reflected in ROI analyses while integrating information across networks. These results highlight the value of combining univariate and multivariate approaches to better characterize neural mechanisms of alcohol cue reactivity in AUD.
Description
Date
1/1/2026
Student Status
Graduate Student
Journal Title
Journal ISSN
Volume Title
Type of presentation
Poster
Collections
Research Projects
Organizational Units
Journal Issue
Citation
DOI
Advisor(s)
Department
Program/Major
Public Health
College/School
Larner College of Medicine
Organization
Research Category
Clinical Science