Development and external validation of AI-ECG models in athlete pre-participation screening: Performance, limitations, and clinical implications.
AI-ECG models show modest accuracy for detecting valvular heart disease in athletes.
Development and external validation of AI-ECG models in athlete pre-participation screening: Performance, limitations, and clinical implications.
Pre-participation screening (pre-participation screening) in competitive athletes aims to identify cardiovascular diseases associated with sudden cardiac death (sudden cardiac death).
To develop and externally validate a deep learning (deep learning)-based AI-ECG ensemble model for detecting imaging-confirmed structural heart disease in competitive athletes undergoing pre-participation screening.
A convolutional neural network (convolutional neural network) ensemble was trained using hospital-derived electrocardiogram images from Beth Israel Deaconess Medical Center (BIDMC, Boston, USA) and externally validated in the Italian Team for Athlete CARDiac evaluation and artificial intelligence-based Risk prediction (ITACARD-artificial intelligence) registry.
how accurately the AI-ECG could detect heart valve problems in athletes
The ITACARD-AI cohort included 1115 competitive athletes (mean age 26 ± 13 years; 70% male), including 48 athletes (4.3%) with valvular heart disease and 30 (2.7%) with cardiomyopathies.
External validation demonstrated substantial performance degradation compared with hospital-based internal validation.
Threshold analyses showed high negative predictive values (~99%) but persistently low positive predictive values (≤8%), reflecting limited disease enrichment and strong prevalence dependence.
Hospital-trained AI-electrocardiogram models demonstrated limited transportability to real-world athlete screening populations.
The marked reduction in external performance suggests that low disease prevalence, heterogeneous physiological remodeling, and incomplete electrocardiogram expression of structural abnormalities may substantially limit class separability in pre-participation screening environments.