MACHINE-LEARNING models can distinguish children with autism spectrum disorder (ASD) from their typically developing (TD) peers using eye-tracking technology with 85% accuracy, a 2026 systematic review and meta-analysis has found.
Autism Spectrum Disorder
Symptoms of ASD typically develop before a child is 3-years-old and diagnosis can be made as early as 18 months.
Standardised screening at 18 and 24 months, alongside ongoing developmental surveillance, continues to be recommended in primary care.
Eye-Tracking Technology
Eye-tracking technology has been increasingly analysed as a potential objective approach to distinguishing people with ASD from TD individuals.
AI and machine-learning methods have been widely applied to support diagnosis and treatment, researchers reported.
Existing evidence points towards the high diagnostic accuracy of eye-tracking data, but evidence on diagnostic performance has been previously limited.
Accuracy, Sensitivity, and Specificity
Researchers analysed more than 2,300 participants across 25 eligible studies.
The pooled accuracy, sensitivity, and specificity of machine-learning models using eye-tracking data to distinguish children with ASD were 85%, 86%, and 86%, respectively.
This suggests that machine-learning approaches centred around eye-tracking, particularly those analysing gaze-pattern features, have strong diagnostic performance for identifying ASD, authors reported.
The performance of the model was influenced, however, by age, stimulus type, task engagement, and machine-learning algorithm.
From Research to Clinical Practice
Whilst eye-tracking-based machine-learning approaches show considerable potential, the robustness and generalisability of the findings were said to be limited.
This followed challenges including a lack of external validation, small sample sizes, and substantial heterogeneity between studies.
Researchers called for standardised eye-tracking methods and large-scale, prospective, multicentre study designs with external validation.
Then, authors said, machine-learning models could be translated into clinical practice as markedly objective and efficient screening tools.
References
Han W et al. Machine learning-based diagnosis of autism spectrum disorder in children and adolescents using eye-tracking data: a systematic review and meta-analysis. Int J Med Inform. 2026;DOI:j.ijmedinf.2025.106235.
Hyman SL et al. Identification, evaluation, and management of children with autism spectrum disorder. Pediatrics. 2020;154(1):DOI:10.1542/peds.2019-3447.
Featured image: Sarah Wilson on Adobe Stock
