Machine learning demonstrates normal fetal Doppler velocimetry associated with reduced risk of necrotizing enterocolitis among preterm infants with growth restriction

Background Necrotizing enterocolitis (NEC) is an intestinal ischemic disease that affects preterm infants with fetal growth restriction (FGR). The role of fetal Dopplers in stratifying risk for developing NEC is unclear but the innovative use of machine learning technology may aid in identifying their contribution. Methods This is a single center retrospective cohort of 164 infants born before 33 weeks’ gestation with FGR from 2016 to 2019. We used machine learning to classify NEC and to evaluate the predictive values of gestational age (GA), birth weight (BW), and the presence of abnormal umbilical artery (UA) Dopplers before delivery. Results Early GA and lower BW strongly predicted NEC. The presence of normal fetal UA Dopplers was heavily weighted in classifying infants unlikely to develop NEC. Fetal UA Dopplers had a 95% specificity (15% sensitivity) for NEC. Conclusions If validated, normal fetal UA Doppler studies may identify infants with FGR at low risk for NEC who may avoid conservative NEC-prevention strategies.