Using neural network technologies to predict premature birth

Objective. The study to develop a model for the early prediction of preterm birth based on neural network technology. Material and Methods. A prospective, continuous study of 1.417 pregnant women in early pregnancy was conducted. 131 women delivered preterm, while 1.286 delivered at term. To identify risk factors for preterm birth, 36 parameters obtained from medical records and questionnaires were analyzed. A mathematical model for predicting preterm birth was constructed using neural network approximation, and cross-validation was used to assess the stability and generalizability of the model. Results. The developed model for predicting preterm birth demonstrated an average accuracy of 0.7573 with a sensitivity of 0.7571. The ROC-AUC integral characteristic was 0.8492. Conclusion. The application of the preterm birth prediction model makes it possible to identify patients with a high risk of this pathology at an early stage of gestation. This allows for a personalized approach and timely implementation of preventive measures, which justifies the implementation of this model in clinical practice.