Skip to Main Content

Eye-to-Eye Neural Insights Predict Autism Symptom Severity

Publication Title: Support vector machine prediction of individual Autism Diagnostic Observation Schedule (ADOS) scores based on neural responses during live eye-to-eye contact

Summary

Question
This study examined whether machine learning could predict individual Autism Diagnostic Observation Schedule (ADOS) scores—used to measure symptom severity in autism spectrum disorder (ASD)—based on brain activity during live eye-to-eye contact. The researchers aimed to link neural activity patterns to social difficulties in ASD.
Why it Matters
Autism spectrum disorder (ASD) affects about 1% of the global population and is characterized by social communication challenges, such as difficulty making eye contact. Current diagnostic methods rely on behavioral assessments, which may overlook underlying neural mechanisms. Identifying brain activity patterns associated with social challenges could improve diagnosis, provide early intervention strategies, and guide treatments tailored to individual needs.
Methods
The researchers used functional near-infrared spectroscopy (fNIRS), a non-invasive brain imaging method that measures blood flow and oxygenation in the brain, to record neural activity during live eye-to-eye contact between adults with ASD and typically developed (TD) adults. A support vector machine (SVM), a machine learning tool, was trained to distinguish between ASD and TD participants based on these neural responses. ADOS scores were not used in training but were later compared to SVM predictions.
Key Findings
The SVM successfully classified ASD and TD participants with 80.5% accuracy. Importantly, the SVM’s predictions of individual ADOS scores, based solely on brain activity during live eye contact, correlated strongly with observed ADOS scores (r = 0.72). This association was specific to live interactions; when participants viewed pre-recorded video stimuli instead, the correlation dropped significantly (r = 0.14).
Implications
These findings suggest that neural responses during live social interactions may serve as biomarkers for ASD symptom severity. This approach could lead to more objective diagnostic tools, complementing current behavioral assessments, and inform personalized interventions for individuals with ASD. Additionally, it highlights the importance of real-world social contexts in studying the neural basis of autism.
Next Steps

The authors recommend expanding this research to larger and broader populations, including pediatric and community samples, to improve the robustness and applicability of these methods. They also suggest further refinement of machine learning models to enhance accuracy and explore similar approaches for other social disorders.

Funding Information
This research was supported by the National Institute of Mental Health of the National Institutes of Health under Award Number R01MH111629. Yale University also provided funding and support for this research.

Full Citation

Zhang X, Noah J, Singh R, McPartland J, Hirsch J. Support vector machine prediction of individual Autism Diagnostic Observation Schedule (ADOS) scores based on neural responses during live eye-to-eye contact. Scientific Reports 2024, 14: 3232. PMID: 38332184, PMCID: PMC10853508, DOI: 10.1038/s41598-024-53942-z.
This AI-assisted summary has been reviewed and approved by at least one of the study's authors to ensure it accurately reflects the research.

Authors

  • Xian Zhang, PhD

    First Author
    Yale School of Medicine

    Associate Research Scientist in Psychiatry

  • Joy Hirsch, PhD

    Last Author
    Yale School of Medicine

    Elizabeth Mears and House Jameson Professor of Psychiatry and Professor of Comparative Medicine and of Neuroscience

Get Involved With This Research

Media Contact

For media inquiries, please contact us.