Early Autism Detection Using Machine Learning: A Multimodal Behavioral Analysis Approach
Posted: Oct 16, 2025
Abstract
This research presents a novel machine learning framework for early autism spectrum disorder (ASD) detection using multimodal behavioral data. We collected and analyzed video recordings, audio patterns, and structured behavioral observations from 450 children aged 18-36 months, including 225 diagnosed with ASD and 225 typically developing controls. Our approach integrates computer vision techniques for facial expression analysis, audio processing for vocal pattern recognition, and temporal modeling of behavioral sequences. The proposed ensemble model achieved 92.3% accuracy in distinguishing ASD from typically developing children, significantly outperforming single-modality approaches. Feature importance analysis revealed that gaze patterns, response latency, and social smile frequency were the most discriminative behavioral markers. This study demonstrates the potential of automated machine learning systems to support early ASD identification, potentially reducing diagnostic delays and improving intervention outcomes.
Downloads: 900
Abstract Views: 1418
Rank: 22409