Early Prediction of SGA-LGA Fetus at the First Trimester Ending Through Weighted Voting Ensemble Learning Approach

Small for Gestational Age (SGA) and Large for Gestational Age (LGA) describe fetuses whose weights deviate below or above the expected average for their gestational age (24 to 42 weeks). SGA fetuses typically fall below the 10th percentile in weight, indicating intrauterine growth restriction, while LGA fetuses surpass the 90th percentile, signifying excessive growth. Accurate classification of SGA and LGA is crucial for evaluating fetal health and guiding prenatal care decisions. This study proposes a Weighted Soft Voting Ensemble Classifier (WSVEC) capable of predicting these classifications by the end of the first trimester. Data were collected from 7,943 pregnant women, including 424 SGA, 928 LGA, and 6,591 AGA (Appropriate for Gestational Age) cases, at the Third Affiliated Hospital of Sun Yat-sen University in Guangzhou, China, from 2015 to 2021. Machine learning algorithms were developed to classify fetuses into SGA, LGA, and AGA groups based on maternal history, morbidity, and biochemical parameters around 13 weeks of gestation. An algorithm was also designed to compute feature importance and identify factors associated with SGA and LGA. The proposed WSVEC classifier achieved an average accuracy of 90.9%, the highest and narrowest negative mean squared error among the tested models, outperforming five state-of-the-art algorithms. Our algorithms also identified key maternal attributes directly correlated with SGA-LGA fetuses, including weight change, pre-pregnancy weight, height, age. Factors that increased the risk of SGA-LGA fetuses, including fetal growth restriction, previous LGA births. Biomarkers such as OGTT-1h, TG, HDL, OGTT-2h, OGTT-0h, TC, FPG, LDL reflected the current state of SGA-LGA fetuses, underscoring the importance of early nutritional and prenatal interventions. These innovative solutions provide valuable support to medical practitioners, enabling early interventions and enhancing strategies for managing fetal and maternal well-being.