Objective To develop and evaluate the performance of a novel method for predicting neonatal respiratory morbidity based on quantitative analysis of the fetal lung by ultrasound. Methods More than 13¿000 non-clinical images and 900 fetal lung images were used to develop a computerized method based on texture analysis and machine learning algorithms, trained to predict neonatal respiratory morbidity risk on fetal lung ultrasound images. The method, termed ‘quantitative ultrasound fetal lung maturity analysis’ (quantusFLM™), was then validated blindly in 144 neonates, delivered at 28¿+¿0 to 39¿+¿0¿weeks' gestation. Lung ultrasound images in DICOM format were obtained within 48¿h of delivery and the ability of the software to predict neonatal respiratory morbidity, defined as either respiratory distress syndrome or transient tachypnea of the newborn, was determined. Results Mean (SD) gestational age at delivery was 36¿+¿1 (3¿+¿3) weeks. Among the 144 neonates, there were 29 (20.1%) cases of neonatal respiratory morbidity. Quantitative texture analysis predicted neonatal respiratory morbidity with a sensitivity, specificity, positive predictive value and negative predictive value of 86.2%, 87.0%, 62.5% and 96.2%, respectively. Conclusions Quantitative ultrasound fetal lung maturity analysis predicted neonatal respiratory morbidity with an accuracy comparable to that of current tests using amniotic fluid.

Peer Reviewed