Machine Learning Approaches to predict Intra-Uterine Insemination Success Rate- Application of Artificial Intelligence in Infertility

Introduction: Assisted Reproductive Technology (ART) has been widely utilized for infertility management. Despite its low success rate, Intra-Uterine Insemination (IUI) is one of the first alternatives and most important approaches regarding many cases of infertility treatment. Given the numerous influencing factors and limitations associated with time and resources, the development of a reliable model to predict the success rate of ART methods can significantly contribute to decision-making processes. Materials and methods: We reviewed the demographic, clinical, and laboratory data regarding 157 IUI treatment cycles among 124 women using their partner’s sperm from May2017 to June2019. Primary outcome measures were clinical pregnancy and live birth. Some prediction models were constructed and compared to the logistic regression analysis. Results: Woman’s mean age was 30.1 ± 5.2 years and the infertility had a female cause in 24.3% of the cases, male cause in 32.6% of cases, and combined causes in 32.6% of the cases. Concerning the first IUI cycle, the clinical pregnancy rate per cycle was 16.9% (N= 21). Data were prepared according to cross-industry standard process for data mining (CRISP-DM) methodology, and the following models were fitted to the data: J48 Decision Tree, Perceptron Multilayer (MLP) Neural Network, Support Vector Machine (SVM) with radial basis function (RBF) kernel, K-Nearest Neighbors (KNN) with one neighborhood, and Bayesian Network. J48 Decision Tree, with a sensitivity of 95% and specificity of 98%, had the most optimal performance, and the KNN model was the weakest one. Conclusion: To predict the results of IUI as a simple and less invasive therapy for infertile couples, some models were applied based on artificial intelligence and J48 Decision Tree was recommended.