Artificial intelligence in diagnosis of pediatric neurodevelopmental disorders: a scoping review

Neurodevelopmental disorders are a group of conditions that affect key areas of development and may significantly impact a child's quality of life. This underscores the importance of accurate diagnostic tools to improve outcomes. Artificial intelligence (AI) has shown measurable effectiveness for enhancing the diagnosis and monitoring of neurodevelopmental disorders. This scoping review aims to summarize the current evidence on the use of AI technologies, including deep learning, supervised machine learning, decision support systems, and biosignal analysis, in improving diagnostic accuracy for pediatric neurodevelopmental disorders.

A systematic search was conducted across PubMed, LILACS, MEDLINE, Google Scholar, and psychology-indexed journals, covering publications from 2000 to January 2025. Keywords and Medical Subject Headings terms were used to search for and select studies, applying specific inclusion and exclusion criteria. Selection followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews guidelines and included clinical studies, reviews, and validation research. The data were extracted and synthesized descriptively.

Twenty-two studies were included. Deep learning models achieved diagnostic accuracies exceeding 85% in most studies in neuroimaging interpretation, whereas supervised machine learning improved the subtype classification of autism spectrum disorder and attention deficit hyperactivity disorder. Decision support systems have increased diagnostic efficiency, and biosignal-based AI has shown potential in identifying physiological markers related to neurodevelopmental disorders.

AI technologies may significantly contribute to improving early diagnosis and clinical decision-making in pediatric neurodevelopment. However, variability in study design, population, and algorithm standardization remains a challenge. AI technologies are also facing ethical concerns such as data privacy and security, interpretability, equity and access, and algorithmic bias. Further multicenter validation and regulatory frameworks are essential for clinical translation.