Prediction of twin preterm birth based on quadratic discriminant analysis and electrohysterogram

Preterm birth is a significant cause of adverse maternal and fetal outcomes. Clinical evidence demonstrates a significantly higher incidence of preterm birth in twin pregnancies compared to singletons. This study evaluates the feasibility of using electrohysterogram in combination with quadratic discriminant analysis models to predict preterm birth in twin pregnancies. Electrohysterogram signals were obtained from 35 twin-pregnant women. Robust electrohysterogram features were selected to construct quadratic discriminant analysis models for preterm birth prediction. A decision-level fusion method based on weighted majority voting rules was implemented to combine all single-channel classification results, aiming to enhance classification performance. The constructed model exhibited a specificity of 96.33% and good generalizability. The classification results showed that the electrohysterogram achieved high accuracy (93.33%), sensitivity (93.56%), and strong binary classification performance (AUC = 0.99) in distinguishing preterm from term birth in twin pregnancies. Integrating electrohysterogram signals with machine learning algorithms enables accurate prediction of preterm birth in twin pregnancies. These findings carry substantial clinical significance, offering opportunities for timely interventions in abnormal pregnancies and potential improvements in perinatal outcomes.