Validation of a Mobile App for Remote Autism Screening in Toddlers

Area of Science:

• Developmental Pediatrics
• Computational Psychiatry
• Human-Computer Interaction

Background:

• Early detection of autism spectrum disorder (ASD) is crucial for timely intervention and improved child outcomes.
• Despite clinical ability to diagnose autism in toddlers, diagnosis is often delayed.
• Mobile health applications offer potential for objective, quantitative autism screening.

Purpose of the Study:

• To evaluate the diagnostic accuracy of the SenseToKnow mobile application for autism detection when administered remotely by caregivers.
• To assess the feasibility and performance of a smartphone-based autism screening tool utilizing computer vision and machine learning.

Main Methods:

• Caregivers administered the SenseToKnow app on personal smartphones/tablets to 620 toddlers (16-40 months).
• The app recorded child behavior via device camera and sensors during interactive stimuli (movies, games).
• Computer vision and machine learning algorithms analyzed behavioral data for autism prediction.

Main Results:

• SenseToKnow achieved high diagnostic accuracy (AUC=0.92), with sensitivity of 83.0% and specificity of 93.3%.
• Positive predictive value was 84.3% and negative predictive value was 92.6%.
• App performance was consistent across different devices (iPhone, iPad).

Conclusions:

• A remote, caregiver-administered mobile app using computer vision shows high accuracy for autism detection in toddlers.
• Remote screening via mobile apps can potentially reduce barriers to early diagnosis and intervention, mitigating disparities.
• This technology supports improved access to services and better outcomes for children with autism.