Artificial Intelligence in Pediatric Surgery: From Diagnostics and Preoperative Planning to Risk Stratification: A Comprehensive Review of Current Applications.
Authors
Affiliations (2)
Affiliations (2)
- Department of Pediatric Surgery, Leipzig University, Leipzig, Germany.
- Department of Pediatric Radiology, Leipzig University, Leipzig, Germany.
Abstract
Artificial intelligence (AI) is increasingly explored in pediatric surgical care, yet its translation into diagnostics and preoperative planning lags behind adult surgery. Unlike prior reviews, this study provides a comprehensive synthesis across four domains, diagnostics, preoperative planning, risk stratification, and surgical error prevention, highlighting recent advances and unmet challenges.A narrative review of PubMed/MEDLINE (2020-2025) identified peer-reviewed studies on AI in pediatric surgery. Eligible articles addressed one of the four domains and were assessed for methodology, clinical applicability, and relevance to pediatric surgical patients.Diagnostic imaging is the most advanced field, with deep learning models for fracture detection and bone age assessment achieving accuracies up to 95% and near-expert agreement, though external validation is scarce. Preoperative planning benefits from AI-driven segmentation, 3D reconstruction, and virtual reality, with reports of altered surgical strategy in up to 8% of oncology cases, but evidence of outcome benefit is limited. Risk models for appendicitis and congenital heart surgery often surpass clinical scores, yet fewer than 10% have undergone external validation. Tools for error prevention, such as intelligent checklists and workflow monitoring, remain at the proof-of-concept stage. Across domains, most studies are retrospective, single-center, and methodologically heterogeneous.AI demonstrates tangible potential to improve pediatric surgical diagnostics, planning, and safety. However, translation into clinical practice requires multicenter pediatric datasets, prospective validation, and transparent, interpretable models. By consolidating the most recent evidence across four domains, this review outlines both the opportunities and critical gaps that should be addressed for safe and effective adoption.