Machine learning models can distinguish children with autism spectrum disorder (ASD) from typically developing (TD) children with 85% accuracy using eye-tracking technology, a 2026 systematic review and meta-analysis found.
autism spectrum disorder
Symptoms of ASD usually begin before a child is 3 years old and are diagnosed as early as 18 months.
Standardized screening at 18 and 24 months continues to be recommended in primary care, alongside ongoing developmental monitoring.
eye tracking technology
Eye-tracking techniques are increasingly being analyzed as a potential objective approach to distinguish between ASD and TD patients.
AI and machine learning techniques are widely applied to support diagnosis and treatment, the researchers reported.
Although existing evidence indicates high diagnostic accuracy of eye-tracking data, evidence regarding diagnostic performance has so far been limited.
Accuracy, sensitivity, specificity
Researchers analyzed more than 2,300 participants across 25 included studies.
The pooled accuracy, sensitivity, and specificity of the machine learning model using eye-tracking data to differentiate children with ASD were 85%, 86%, and 86%, respectively.
The authors reported that this suggests that machine learning approaches centered on eye tracking, especially those that analyze features of gaze patterns, have strong diagnostic performance for identifying ASD.
However, model performance was influenced by age, stimulus type, task engagement, and machine learning algorithm.
From research to clinical practice
Although eye-tracking-based machine learning approaches have shown great potential, the robustness and generalizability of their results have been said to be limited.
This was fraught with challenges such as lack of external validation, small sample sizes, and significant heterogeneity between studies.
The researchers called for a standardized eye-tracking method and a large, prospective, multicenter study design with external validation.
The authors said the machine learning model could be applied in clinical settings as a highly objective and efficient screening tool.
References
Han W et al. Machine learning-based diagnosis of autism spectrum disorders 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, assessment, and management of children with autism spectrum disorders. Pediatrics. 2020;154(1):DOI:10.1542/peds.2019-3447.
Featured image: Sarah Wilson from Adobe Stock
