Early diagnosis of ASD is critical but often challenging due to subtle initial symptoms. According to the U.S. Centers for Disease Control and Prevention, one in three children with autism in the United States is diagnosed after age 8. Delayed diagnosis can hinder timely interventions, which are crucial for improving outcomes.
The South Korean research team collected data from 1,242 infants and toddlers aged 18 to 48 months across nine hospitals in the country to develop their AI model. The screening process involves parents recording their child’s voice using a smartphone and guiding the child through simple, age-appropriate tasks. These tasks include responding to their name, imitating actions, playing with a ball, engaging in pretend play, or requesting help.
The AI integrates the voice data with results from established autism screening tools, such as the Modified Checklist for Autism in Toddlers (M-CHAT), Social Communication Questionnaire (SCQ), and Social Responsiveness Scale-2 (SRS-2). While traditional screening methods alone achieve an accuracy of about 70%, the addition of AI analysis boosts the accuracy to 94%. The model also distinguishes high-risk children from those with confirmed autism diagnoses with 85% accuracy. Furthermore, its predictions align with the Autism Diagnostic Observation Schedule-2 (ADOS-2), an international gold-standard diagnostic tool, at approximately 80% consistency.
Professor Kim emphasized the tool’s user-friendliness, noting, “We standardized the tasks to make them accessible to everyone. This digital tool can provide reliable insights even before a specialist’s diagnosis.”
The research, supported by the National Center for Mental Health’s digital therapeutics development program, was published in the latest issue of *npj Digital Medicine*, a prestigious peer-reviewed journal.
Lim Hye Jung, HEALTH IN NEWS TEAM
press@hinews.co.kr
