Machine learning models for predicting successful vaginal birth after cesarean: a systematic review

Background: Cesarean section rates have risen globally, leading to increased interest in predicting the success of vaginal birth after cesarean (VBAC) to optimize maternal and neonatal outcomes. This systematic review aims to synthesize current evidence on the application of ML algorithms for predicting successful VBAC. Methods: Seven relevant studies published between ۲۰۲۰ and ۲۰۲۴, encompassing sample sizes from ۳۰۱ to over ۲.۵ million participants, were analyzed. The studies utilized a variety of ML techniques including gradient boosting (XGBoost, CatBoost), random forest, K-nearest neighbors (KNN), logistic regression, naïve Bayes, and decision trees. Feature sets commonly included maternal demographics, obstetric history, clinical and fetal characteristics. Data preprocessing strategies such as missing data imputation and feature selection were also assessed. Results: Tree-based ensemble methods, particularly gradient boosting algorithms (CatBoost, XGBoost), consistently showed superior predictive accuracy, with area under the curve (AUC) values ranging from ۰.۷۵ to ۰.۹۱. For instance, the CatBoost model achieved an accuracy of ۰.۹۱ and AUC of ۰.۸۹ on a Taiwanese cohort, while a large-scale U.S. cohort using XGBoost reported an AUC of ۰.۸۵ and accuracy of ۸۵.۴%. Feature selection and imputation techniques improved model performance significantly. Conclusions: Machine learning models, especially those employing gradient boosting techniques on large and diverse datasets, demonstrate promising performance for predicting successful VBAC. These tools have potential to assist personalized clinical decision-making and reduce unnecessary repeat cesarean deliveries. Future studies should focus on external validation, improving model interpretability, and clinical integration to enhance maternal and neonatal care.

کلیدواژه ها:

Machine Learning ، Vaginal Birth After Cesarean ، Prediction Models