Fetal Growth Restriction and Placental Insufficiency: Integrating Doppler Imaging with Machine Learning for Risk Stratification

Abstract

Fetal Growth Restriction (FGR) remains one of the most significant causes of perinatal morbidity and mortality worldwide, primarily linked to placental insufficiency and compromised fetoplacental circulation. Conventional diagnostic modalities based solely on estimated fetal weight and umbilical artery Doppler measurements are often insufficient for early prediction and individualized management. This study introduces an integrated diagnostic framework that combines advanced Doppler imaging parameters with machine learning–based risk stratification to enhance early detection of placental dysfunction. A retrospective dataset comprising 300 pregnancies between 24–38 weeks of gestation was analyzed, including Doppler indices such as Pulsatility Index (PI), Resistance Index (RI), and Systolic/Diastolic (S/D) ratios from the umbilical artery, middle cerebral artery, uterine artery, and ductus venosus. Machine learning models including Random Forest, XGBoost, and Support Vector Machine were trained to classify pregnancies into high-risk and low-risk FGR categories. The Random Forest model achieved the highest predictive accuracy (AUC = 0.94), outperforming conventional threshold-based clinical assessments. Statistical correlation demonstrated strong associations between altered cerebroplacental ratio (CPR) and adverse perinatal outcomes such as low birth weight and preterm delivery. The proposed integrative model establishes a scalable, data-driven approach for obstetric risk prediction, enabling clinicians to tailor interventions and surveillance for at-risk fetuses. This research underscores the potential of combining hemodynamic biomarkers and artificial intelligence to revolutionize prenatal care through precision-based prediction of fetal compromise.