Application of Random Forest Algorithm in Pregnancy Prediction after Fallopian Tube Recanalization

Background Protecting female fertility stands as a central goal and vision in a fertility-friendly society, and fallopian tube recanalization offers the possibility of pregnancy for patients with tubal infertility. Objective This study aims to accurately identify the influencing factors affecting successful pregnancy after fallopian tube recanalization and explore the application of the random forest algorithm in screening and predicting pregnancy influencing factors in such patients. Methods The study collected and analyzed data from 170 patients who underwent laparoscopic combined with hysteroscopic fallopian tube recanalization at Capital Medical University Affiliated Beijing Shijitan Hospital between 2016 and 2018. Based on whether the patients achieved successful natural pregnancy within 2 years after the surgery, they were divided into the pregnancy and non-pregnancy groups. Using the R software, a random forest model for predicting pregnancy risk after tube recanalization was established on the training data set (108 cases, 63.2% of cases, extracted via Bootstrap method), and its prediction accuracy, sensitivity, specificity, positive predictive value, negative predictive value, and area under the curve (AUC) were evaluated on the verification data set. Results The study comprised 82 cases in the pregnancy group and 88 in the non-pregnancy group, with a spontaneous pregnancy rate of 48.2% post-surgery. The random forest algorithm, trained on the training set, demonstrated robust predictive capability upon validation, with an accuracy of 87.1%, sensitivity of 93.1%, specificity of 81.8%, positive predictive value of 81.8%, negative predictive value of 93.1%, and an AUC of 0.921. The random forest algorithm was employed to rank the importance of factors influencing pregnancy following fallopian tube recanalization by using variable importance scores. The analysis identified the top three significant predictor variables: duration of infertility, history of previous pregnancies, and patient age. Conclusion The random forest algorithm emerges as a viable tool for predicting factors influencing pregnancy after fallopian tube recanalization. The predictive model, predicated on infertility duration, history of prior pregnancies, and age, exhibits notable discrimination and accuracy. Early identification of key factors post-recanalization allows for timely and effective interventions. We recommend that patients presenting risk factors consider utilizing assisted reproductive technology to improve pregnancy rates.