Predictive model integrating two-dimensional ultrasound indicators and hemodynamic features for fetal coarctation of the aorta.

backgroundPrenatal coarctation of the aorta (CoA) diagnosis still has high false-positive and false-negative rates. This study aimed to develop and validate a new model combining two-dimensional(2D) ultrasound parameters and hemodynamic indicators and construct a CoA nomogram to enhance diagnostic accuracy. METHODS AND

resultsBetween January 2013 and June 2024, 222 fetuses with suspected CoA on prenatal ultrasound were collected retrospectively and divided into CoA (n = 69) and Non-CoA (n = 153) groups based on postnatal follow-up. Univariate analysis found that gestational age (GA) at examination, the pulmonary artery-to-aorta diameter ratio (PA/AO), aortic isthmus (AOI) Z-score, increased diastolic flow at the AOI, flow reversal at the AOI, and ventricular septal defect (VSD) were CoA predictors (P < 0.05). PA/AO, diastolic flow increase, absence of flow reversal, and VSD were identified as the final predictors after multivariate logistic regression analysis. The combined model including the final predictors above mentioned achieved the best performance with a mean AUC of 0.862216 (standard deviation = 0.0439) in the 5-fold cross-validation. Bootstrap validation indicated this combination was the top model in 26% of iterations and had the highest mean out-of-bag AUC of 0. 8524.This combined model achieved an AUC of 0.85, with 83.33% sensitivity and 72.09% specificity in validation set.

conclusionsThis study developed and validated a prenatal CoA prediction model integrating multiple parameters. Combining 2D ultrasound anatomical indicators (PA/AO, VSD) and AOI hemodynamic features (diastolic flow increase, absence of flow reversal), the model demonstrated good predictive performance.