AI-Assisted Autism Screening Tool for Pediatric and School-Based Early Interventions: Enhancing Early Detection Through Multimodal Behavioral Analysis
Posted: Nov 19, 2022
Abstract
Early detection of autism spectrum disorder (ASD) is crucial for effective intervention, yet significant disparities exist in access to timely screening, particularly
in school and primary care settings. This research presents the development and
validation of an AI-assisted autism screening tool specifically designed for pediatric
and school-based early intervention programs. The multimodal framework integrates computer vision analysis of behavioral cues, natural language processing of
developmental histories, and machine learning evaluation of standardized screening
instruments. Our approach leverages a diverse dataset of 8,450 children aged 18-60
months from 32 school districts and pediatric clinics, incorporating video recordings
of social interactions, parent-reported developmental questionnaires, and educator
observations. The AI system achieves 94.2% sensitivity and 91.8% specificity in
identifying children requiring comprehensive ASD evaluation, significantly outperforming traditional screening methods while maintaining clinical interpretability.
Implementation in 15 school-based health centers demonstrated a 67% reduction in
screening wait times and a 42% increase in early intervention referrals before age 36
months. The tool’s adaptive learning capability enables continuous improvement
through deployment feedback, while its web-based interface ensures accessibility
across diverse educational and healthcare settings. This research addresses critical
national child health needs by providing scalable, evidence-based autism screening
that bridges gaps in early detection and facilitates timely access to intervention
services, ultimately improving developmental outcomes for children across socioeconomic and geographic spectra.
Downloads: 42
Abstract Views: 2053
Rank: 50010