Multicenter validation of AI-enabled ECG for pediatric biological sex prediction

Area of Science:

• Cardiology
• Artificial Intelligence
• Pediatric Health

Background:

• Biological sex influences electrocardiogram (ECG) patterns, impacting health and disease.
• Existing AI models show potential for sex prediction via ECG.

Purpose of the Study:

• To externally validate an AI-enabled ECG model for biological sex prediction in pediatric populations.
• To assess the model's performance across different pubertal stages.

Main Methods:

• Multicenter external validation of a previously developed AI ECG model.
• Analysis of ECG data from pediatric patients at Texas Children's Hospital and Boston Children's Hospital.
• Saliency mapping to identify sex-related electrophysiologic patterns.

Main Results:

• The AI ECG model successfully predicted biological sex across pediatric development.
• Model accuracy demonstrated a significant gradient correlating with pubertal stage (pre-puberty AUROC 0.64, peri-puberty AUROC 0.84, post-puberty AUROC 0.94).
• Replication of findings across multiple institutions confirmed model robustness.

Conclusions:

• AI-enabled ECG analysis is a viable tool for non-invasive biological sex prediction in children.
• The model's performance variation with puberty underscores developmental electrophysiologic differences.
• Established sex-related ECG patterns are identifiable by AI, offering insights into sex-specific cardiac physiology.