A Review on Autism Spectrum Disorder Screening by Artificial Intelligence Methods

Area of Science:

• Developmental Pediatrics
• Computational Neuroscience
• Biomedical Engineering

Background:

• Autism spectrum disorder (ASD) prevalence is rising, necessitating early detection.
• Subtle developmental differences in early childhood require advanced diagnostic tools.
• Artificial intelligence (AI) offers potential for automated early screening of ASD.

Purpose of the Study:

• To review research on AI-powered methods for early ASD identification.
• To identify key markers utilized in AI-based ASD screening.
• To assess the efficacy of current AI screening approaches.

Main Methods:

• Systematic literature search across major scientific databases (Web of Science, PubMed, Scopus, etc.).
• Inclusion of 43 articles published between January 2013 and November 2023.
• Categorization of identified recognition markers.

Main Results:

• Recognition markers fall into five categories: gaze behaviors, facial expressions, motor movements, voice features, and task performance.
• AI screening accuracy ranged from 62.13% to 100%.
• Sensitivity and specificity varied from 69.67% to 100% and 54% to 100%, respectively.

Conclusions:

• AI recognition is a promising tool for identifying children with ASD.
• Enhancing AI screening models through multimodal approaches is crucial for improved accuracy.
• Timely intervention and treatment can be facilitated by more accurate AI-driven ASD identification.