A study of 851 children suggests that combining behavioural screening with auditory brainstem responses could support more objective autism assessment, with sound-processing differences varying by brainstem region and stimulus intensity.
A new study suggests that patterns in brainstem hearing responses, combined with behavioural screening, could help improve the diagnosis of autism spectrum disorder in children. Researchers working with 851 children who were suspected of having autism tested a series of machine-learning models built from results of the Autism Behaviour Checklist and auditory brainstem response, or ABR, tests, which measure how sound signals travel through the auditory pathway.
The strongest-performing model was XGBoost, a machine-learning method that builds a prediction by combining many decision trees. It achieved an accuracy of 85.9% and an AUROC, a measure of how well a model separates two groups, of 0.779. The study also used Shapley additive explanation, or SHAP, to show which features mattered most. Wave III and wave V were the most influential, pointing to ABR activity linked to the route from the cochlear nerve towards the inferior colliculus.
The researchers found that differences in ABR results across phenotypes were driven largely by stimulation intensity, with effects varying at 45 and 65 decibels. They said the findings suggest a more complex, region-specific and intensity-dependent pattern in children with autism, rather than a simple uniform difference. In other words, the way sound signals are processed in the brainstem may be tied to autistic behaviour in a more nuanced manner than earlier work has captured.
The study adds to a growing body of research showing that machine learning may help support autism diagnosis using different kinds of data. Other recent studies have reported promising results from neuroimaging, medical history and home-video based tools. However, the new findings are notable because they focus on ABR, a relatively accessible physiological test, and suggest that hearing pathway measurements may offer another route towards earlier and more objective identification of autism in children.
Disclaimer: This content is for informational purposes only and is not intended to be a substitute for professional medical judgment, advice, diagnosis, or treatment.
