Artificial intelligence support for diagnosis of neurodevelopmental disorders during childhood: an umbrella review
-
By
-
March 18, 2026
Objective:
To synthesize evidence from systematic reviews and meta-analyses on the use of AI for diagnosing neurodevelopmental disorders in children, focusing on studies published in any language.
Key Findings:
- 148 records were identified, and 64 studies were included, primarily focusing on ASD (31 studies) and ADHD (14 studies).
- AI models, especially deep learning, achieved diagnostic accuracy from 66% to 99%.
- The majority of studies had critically low methodological quality (80% rated low).
Interpretation:
AI shows potential for enhancing diagnostic accuracy for neurodevelopmental disorders, but current research lacks the necessary methodological rigor for reliable clinical application.
Limitations:
- Insufficient external validation of AI models, impacting their reliability.
- Lack of standardization in data collection and model development, which affects comparability.
- Reporting inconsistencies across studies that hinder the synthesis of findings.
Conclusion:
While AI holds promise for diagnosing neurodevelopmental disorders, further methodologically sound research is essential for clinical implementation.
