Improving prenatal detection of congenital heart disease with a scalable composite analysis of six fetal cardiac ultrasound biometrics

Although screening of prenatal congenital heart disease (CHD) has improved over the last decade, the diagnosis rate can still be as low as 40%. The axial 4 chamber (A4C) is the most reliably obtained cardiac view in the fetal screening ultrasound but alone only has a maximum clinical sensitivity of 50-60%, particularly in large multicenter studies in low-risk populations. Standard biometrics, like cardiac axis (CA), cardiothoracic ratio (CTR) and cardiac chamber fractional area change (FAC), have individually been shown to be useful for CHD screening and can all be obtained from A4C alone. However, these biometrics are vastly underutilized because they are time-consuming to extract and difficult to interpret all at once. We hypothesized that using six standard biometrics in combination can improve complex CHD screening versus any one biometric alone. K-means clustering was performed to segregate the patterns of heart measurements into clusters. Sensitivity and specificity for CHD was 87% and 75%, respectively. Here, we demonstrate that a composite of six standard biometric has better sensitivity and accuracy for CHD than any one biometric alone and better than A4C visual assessment.