Reward network state dynamics track ASD symptom severity but not diagnosis

Abstract

Autism spectrum disorder (ASD) is a heterogeneous developmental condition characterized by repetitive behaviors and social communication difficulties. The reward network has been specifically implicated in the social deficits associated with ASD for several decades. While modern neuroimaging techniques enable investigation of this cortical-subcortical network, task-based fMRI studies have yielded inconsistent findings, largely due to insufficient sample sizes and heterogeneous reward paradigms. More recently the brain at rest has been leveraged to examine reward network connectivity in large-scale datasets. While previous studies have identified associations between network organization, diagnostic group, and individual-level clinical variables, these findings have largely remained at the trend level. However, prior research has focused exclusively on static connectivity patterns, neglecting the temporal dynamics inherent in brain activity. In this study, we analyzed a large multi-site fMRI dataset (ABIDE I) to examine reward network dynamics in individuals with ASD, with particular emphasis on individual-level variability and its relationship to clinical phenotypes. Following an initial assessment of static connectivity, we employed a Hidden Markov Model (HMM) as our primary analysis method, alongside a sliding window approach with subsequent clustering as a secondary validation step, to characterize the temporal properties of reward network activity. We observed a consistent association between greater occupancy of the most sparsely connected network state and milder verbal communication symptoms, as measured by the Autism Diagnostic Interview-Revised (ADI-R), across both methods. Notably, traditional group-level comparisons between ASD and control groups revealed limited differences, underscoring the importance of individual-level characterization in this heterogeneous condition. These findings demonstrate that temporal dynamics of reward network connectivity capture clinically meaningful variation in ASD beyond static connectivity measures, supporting the value of dynamic approaches for understanding neurodevelopmental disorders.