ARTIFICIAL INTELLIGENCE IN PEDIATRIC ELECTROCARDIOGRAPHY: DIAGNOSTIC APPLICATIONS, PREDICTIVE POTENTIAL, AND CLINICAL LIMITATIONS - A LITERATURE REVIEW
TL;DR
Findings suggest that AI models can extract complex information from ECG recordings beyond individual conventional parameters, which supports its role as a tool for automated ECG interpretation, disease detection, risk stratification, and indirect assessment of cardiac structure and function.
Abstract
Introduction Electrocardiography (ECG) remains a fundamental diagnostic tool in pediatric cardiology; however, its interpretation is challenging because normal electrocardiographic parameters vary considerably with age and cardiovascular maturation. The development of artificial intelligence (AI), particularly machine learning and deep learning, has created new possibilities for automated ECG analysis and the detection of subtle patterns that may be difficult to identify using conventional interpretation. This review aims to summarize current applications of AI in pediatric electrocardiography, with particular emphasis on diagnostic and predictive applications and their clinical potential. Current State of Knowledge Current evidence indicates that AI-based models can support the automated interpretation of pediatric ECGs and the detection of rhythm and conduction disorders, including Wolff-Parkinson-White syndrome, long QT syndrome, and congenital heart disease. Deep learning models have also demonstrated the ability to estimate corrected QT intervals and to identify ECG patterns associated with left ventricular dysfunction, hypertrophy, and dilation. Beyond conventional diagnosis, AI-ECG has shown potential for prognostic applications, including the prediction of adverse cardiovascular events in children. These findings suggest that AI models can extract complex information from ECG recordings beyond individual conventional parameters. However, the available evidence remains limited by relatively small and heterogeneous pediatric datasets, age-related variability in ECG characteristics, data heterogeneity, limited external validation, and the predominance of retrospective studies. Moreover, evidence regarding the impact of AI-ECG on clinical decision-making and patient outcomes remains limited. Conclusions AI represents a promising approach to expanding the diagnostic and prognostic potential of pediatric electrocardiography. Current findings support its role as a tool for automated ECG interpretation, disease detection, risk stratification, and indirect assessment of cardiac structure and function. Nevertheless, AI-ECG should currently be regarded as a clinical decision-support tool rather than a replacement for conventional cardiac assessment. Further multicenter, prospective, and externally validated studies are needed to establish the generalizability, safety, interpretability, and clinical utility of AI-based ECG models in pediatric practice.