Deep learning model for prenatal congenital heart disease (CHD) screening

This study applies a deep learning (DL) model, previously trained on images from a tertiary center, to fetal ultrasound images during the second-trimester anomaly scan in a low-risk population. The DL model outperforms initial clinical assessment, achieving higher sensitivity in detecting congenital heart defects (CHD). It performs well on community-acquired images, even on lesions it had not encountered before.