The role of the dopamine system in autism spectrum disorder revealed using machine learning: an ABIDE database-based study.

Abstract

This study explores the diagnostic value of dopamine system imaging characteristics in children with autism spectrum disorder. Functional magnetic resonance data from 551 children in the Autism Brain Imaging Data Exchange database were analyzed, focusing on six dopamine-related brain regions as regions of interest. Functional connectivity between these ROIs and across the whole brain was assessed. Machine learning techniques then evaluated the ability of the dopamine system's imaging features to predict autism spectrum disorder. Functional connectivity was significantly higher in autism spectrum disorder children between the ventral tegmental area and substantia nigra, prefrontal cortex, nucleus accumbens, and between the substantia nigra and hypothalamus compared to typically developing children. Additionally, clustering methods identified two autism spectrum disorder subtypes, achieving over 0.8 accuracy. Subtype 1 showed higher stereotyped behavior scores than subtype 2 in both genders, with subtype-specific functional connectivity differences between male and female autism spectrum disorder groups. These findings suggest that abnormal functional connectivity in the dopamine system serves as a diagnostic biomarker for autism spectrum disorder and can support clinical decision-making and personalized treatment optimization.