Deep machine learning applied to support clinical decision-making in the treatment of infertility using assisted reproductive technologies

Abstract

Introduction. Machine learning (ML) applied to data analysis allows to more accurately and targetedly determine the most significant correctable and non-correctable predictors of onset of pregnancy in assisted reproductive technology (ART) programs in patients of different age groups. Analysis of data using various techniques and comparison of results obtained via two models will determine the most significant factors for onset of pregnancy in the ART program.Aim. To determine the most significant clinical and embryological predictors of onset of pregnancy using standard regression analysis and a decision tree algorithm to predict pregnancy in the ART program.Materials and methods. A total of 1,021 married couples were included in the retrospective study. The study analysed clinical and laboratory test findings and stimulated cycle parameters depending on the effectiveness of the ART program. A regression analysis was carried out and a decision tree algorithm was built using the Gini criterion to determine the most significant factors.Results. We identified “general” signs that require further validation on other models, including ML: the presence/absence of a history of pregnancies, stimulated cycle parameters (oocyte cumulus complex, number of metaphase II (MII) oocytes, number of zygotes), spermogram indicators on the day of puncture, number of high and good quality embryos, as well as the embryo grading.Conclusion. rFSH (follitropin-alpha, Gonal-f) gives a significant result in two of the five available age groups, follitropin-beta, corifollitropin alfa – in one of the five groups only. Building a model that includes not only the couple’s medical history data, but also molecular markers using machine learning methods will not only allow us to most accurately determine the most promising groups of patients for in vitro fertilization (IVF) programs, but also increase the efficiency of ART programs by selecting the highest quality embryo to be transferred.