Artificial intelligence on guard of reproductive health

Abstract: International large-scale studies have shown that infertility affects about 186 million people worldwide. At the same time, there is a negative trend: the number of infertile couples increases every year. With the clinical introduction of assisted reproductive technologies (ARTs), mainly artificial insemination, there are chances to solve this problem. However, despite the high level of development of modern reproductive medicine, only about one third of interventions are successful. To make the diagnosis of infertility more accurate and treatment more effective, it is advisable for the medical community to use software products based on artificial intelligence technologies in everyday practice. This will be able to timely identify and study potential relationships in large datasets and to create reliable predictive models using machine learning techniques. The most advanced areas of artificial intelligence research in reproductology are to improve the quality of biomaterial assessment for in vitro fertilization and to predict the outcome of artificial insemination, by taking into account the data of married couples. The main machine learning methods used to solve problems in reproductive medicine are the support vector method, the random forest algorithm, the decision tree algorithm, Bayesian classifiers, and artificial neural networks. The major goal of recent research in this field is to achieve maximum accuracy of software algorithms and to obtain results that can subsequently provide reliable prediction, diagnosis, and treatment of diseases.

Conclusion. This review considers the main ways of applying the machine learning algorithms in reproductive medicine, the stages of creating training models, some limitations and prospects for introducing these methods inclinical practice.