Integrating multimodal neuroimaging and physical-health markers outperformed single-modality models for ASD classification in a population-based cohort, yielding an AUC-ROC of 0.68, with sleep function, right superior temporal gyrus cortical thickness, and cingulo-opercular/default mode network connectivity as the strongest predictors.
Multimodal integration of brain and physical-health markers outperformed single-modality models for ASD classification.
Among physical-health markers, sleep function contributed most strongly to ASD classification.
Cortical thickness in the right superior temporal gyrus was among the top neuroimaging predictors of ASD classification.
Connectivity between the cingulo-opercular and default mode networks was a significant neuroimaging predictor of ASD classification.
Propensity-score matching was used to create demographically balanced ASD and non-ASD groups from the ABCD Study cohort.
The study provides proof of concept that combining multimodal MRI and physical-health data within a large, demographically representative cohort can enhance ASD classification and yield biologically interpretable features.
This research suggests that combining brain imaging data with physical health information — such as sleep patterns, growth measures, and early developmental history — can improve the ability to distinguish adolescents with autism spectrum disorder (ASD) from those without it, compared to using any single type of data alone. The study used data from the Adolescent Brain Cognitive Development (ABCD) Study, a large, nationally representative cohort of children and adolescents, and carefully matched participants with and without ASD to ensure fair comparisons. The combined model achieved moderate classification performance (AUC-ROC of 0.68), which, while not clinically diagnostic on its own, represents meaningful signal above chance in a real-world, community-based population. Among the physical health factors examined, sleep was the most informative for identifying ASD-related differences. On the brain imaging side, the thickness of a region called the right superior temporal gyrus — an area involved in social processing and language — and the connectivity between two brain networks (the cingulo-opercular and default mode networks) were the strongest neural predictors. These findings are consistent with prior research highlighting social-communicative brain regions and network differences in ASD. This research suggests that no single biological marker captures the full picture of ASD, and that combining information from the brain and body together offers a more complete view. The study is presented as a proof-of-concept rather than a clinical tool, and the authors emphasize that the population-based design means results reflect community-level variation in ASD rather than only severely affected clinical cases. Future large-scale studies could build on this framework to better understand the biological underpinnings of neurodevelopmental diversity.
Zeevi D, Acosta-Rodriguez H, Bobba P, Stephan A, Lin H, Malhotra A, et al.. (2026). Integrating Multimodal Neuroimaging and Physical-Health Markers for Autism Spectrum Disorder in the ABCD Study.. Journal of integrative neuroscience. https://doi.org/10.31083/JIN48212