Fetal Congenital Heart Disease Detection

Congenital heart disease (CHD) is a common birth defect that is challenging to detect with current fetal screening methods, which often have low sensitivity. Using 107,823 images from 1,326 fetal echocardiograms, researchers trained an ensemble of neural networks to identify key cardiac views and differentiate between normal and complex CHD. The model achieved high performance with an AUC of 0.99, 95% sensitivity, and 96% specificity, and demonstrated potential to significantly enhance the detection of fetal CHD when applied to guideline-recommended imaging.