EchoNet-Peds

This study developed EchoNet-Peds, a pediatric-specific, video-based deep learning model trained on 4,467 echocardiograms to automate left ventricular (LV) segmentation and ejection fraction (EF) estimation. The model achieved strong performance with a Dice coefficient of 0.89 for LV segmentation, a mean absolute EF error of 3.66%, and excellent detection of systolic dysfunction (AUC 0.95), outperforming adult-trained models on pediatric data. These findings demonstrate that accurate, rapid AI-based cardiac function assessment in children is feasible and highlight the availability of a large annotated pediatric echocardiography dataset to support future research.