Artificial intelligence in Autism Spectrum Disorder diagnosis of Visual Attention and Facial Recognition: A Scoping Review

Abstract

Background: Autism Spectrum Disorder (ASD) is estimated to affect 1 out of every 100 children globally, with early diagnosis critical for timely intervention. Conventional diagnostic techniques mostly rely on doctor’s subjective behavioural observations, which are labor - intensive, resource- intensive and prone to inter-ratter variability. By identifying tiny abnormal patterns in eye gaze and facial expressions, artificial intelligence (AI), in particular visual attention analysis and facial recognition presents intriguing objective alternatives.

Objectives: With an emphasis on visual attention (such as eye tracking) and facial recognition technologies, this scoping review maps the scope, variety, and kind of AI applications in ASD diagnosis. It also identifies important techniques, performance measures and research needs.

Methods: Using electronic databases from Pubmed, Google Scholar, Web of Science, and Research Gate, a thorough literature analysis was carries out in compliance with PRISMA- ScR guidelines, concentrating on full-text articles published between 2015 and 2025. Analysis was done on eighteen studies that looked at the use of AI in the diagnosis of ASD in areas of visual attention and facial recognition.

Results: There were eighteen studies in the records. While visual attention AI used eye-tracking to identify inappropriate gaze fixation with roughly 85% accuracy, facial recognition models detected micro- expressions with about 80% sensitivity.

Conclusion: AI exhibits strong promise for scalable ASD diagnosis; however, standardization and varied validation are required. Clinical trials for practical implementation and multimodal integration should be given to priority in future research.