Left Ventricular Dysfunction Prediction

A convolutional neural network was trained on ECG-echocardiogram pairs from pediatric patients to detect left ventricular (LV) dysfunction, hypertrophy, and dilation, showing strong performance across internal and external validation cohorts. The model matched or exceeded pediatric cardiologist benchmarks, particularly in identifying LV hypertrophy, and performed consistently even without additional demographic data. This AI-enhanced ECG tool could enable broad, cost-effective screening for cardiac remodeling in children, improving early detection and access to care.