
Evaluating Artificial Intelligence in Airway Management for Neonates and Pediatric: A Systematic Review
Parnian Safikhani 1, Parisa Akbarpour 2 ©, Siamak Nazari 2, Bahareh Mahdood 3, Maryam Bakhshaei 1 ℗, Shaghayegh Bamian 1, Hanieh Hashemi 1
Abstract
Introduction: Difficult airway management in pediatric and neonates remains a critical challenge due to anatomical variability and the limitations of traditional assessment tools. Recent advances in artificial intelligence (AI) and machine learning (ML) offer new opportunities for accurate, individualized airway risk prediction. Search Strategy: This systematic review was conducted by PRISMA 2020 guidelines between January 2024 and May 2025. A comprehensive search was performed across PubMed, Scopus, Embase, IEEE Xplore, Web of Science, and Google Scholar. Search terms included combinations of "Airway," "Machine learning," "Deep learning," and "Pediatric" and "Neonate" "Anesthesia." Studies were included if they applied AI/ML to assess or predict airway difficulty in pediatric patients. Exclusion criteria were non-original works, animal studies, abstracts without full text, and articles not focused on AI or airway-related outcomes. All titles, abstracts, and full-text articles were independently reviewed by two researchers. Methodological quality was assessed using the AMSTAR 2 tool, and potential sources of bias were evaluated based on model validation and data transparency. Results: From an initial pool of 1,374 articles, 436 duplicates were removed. After title and abstract screening of 938 articles, 841 were excluded for irrelevance or failing to meet inclusion criteria. A Full-text review of 97 articles led to the exclusion of 82 additional studies due to a lack of original data, poor methodological rigor, or unclear AI application. Ultimately, 15 studies were included. These studies applied diverse ML techniques—including CNNs, SVMs, random forests, and ensemble models—to pediatric airway assessment, with sample sizes ranging from 48 to 53,637. Several models reported high diagnostic performance (AUC 0.85) for predicting difficult intubation or airway challenges. Applications also included robotic-assisted intubation, facial morphometrics, and AI-based training simulators. However, most studies lacked external validation and presented methodological heterogeneity. Conclusion and Discussion: AI and ML technologies show strong potential to improve the prediction and management of difficult pediatric and neonates airways. While early results are promising, broader clinical integration will require robust validation, standardized methodologies, and careful ethical oversight.
Keywords: Artificial intelligence, Machine learning, Pediatric, Intubation, Deep learning, Neonates