OC07.01: Prediction of small‐ and large‐for‐gestational‐age neonates: artificial intelligence model

Accurate individualised assessment of fetal size enables the identification of patients most likely to benefit from early delivery, appropriate pregnancy care pathways and surveillance. The aim of this study was to evaluate the performance of artificial intelligence models for the prediction of small (SGA) and large for gestational age (LGA) neonates. This was a cohort study of low-risk singleton pregnancies undergoing routine ultrasound between 35+0 and 37+6 weeks. Maternal age, body mass index, abdominal circumference centile, estimated fetal weight centile, Umbilical Artery Pulsatility Index (PI) Multiples of the Median (MoM), Middle Cerebral Artery PI MoM, Uterine Artery (UtA) PI MoM, Cerebroplacental ratio CPR) MOM as numerical variables, nulliparity, ethnicity, previous stillbirth, smoking status, alcohol consumption, previous CS, gestational diabetes mellitus, pre-eclampsia, fertility treatment, oligohydroamnios and polyhydroamnios were classified as categorical variables and used in the algorithm. Two separate Elastic-net regression algorithms were created for the SGA and LGA prediction models. The Sklearn library and Python software were used to implement the algorithms. A total of 14,161 pregnancies were included in the study; 80% of the dataset was used for machine learning and 20% as a test group. The sensitivity and specificity of the prediction model for SGA prediction are 0.36 and 0.99. The positive predictive value and the negative predictive value are 0.70 and 0.95 respectively. The sensitivity and specificity of the predictive model for LGA are 0.01 and 1.00. The positive predictive value and the negative predictive value are 0.50 and 0.93, respectively. In the ROC analysis for these two models, the AUC values were 0.90 and 0.88 for SGA and LGA, respectively. Routinely available patient characteristics and ultrasound variables at 35-37 weeks can reliably predict both SGA and LGA with a high degree of accuracy.