Ensemble Machine Learning Models for Evaluation of Sperm Quality with Respect to Success Rate of Clinical Pregnancy in IVF, ICSI, and IUI Methods

Abstract

Abstract Objective: Evaluation of the effect of sperm quality on the success rate of clinical pregnancy and the possibility of infertility. The primary objective was to determine the success rate of clinical pregnancy (CPR). The secondary objective was to evaluate the clinical pregnancy rate (FHR). Method: This retrospective study evaluated 1929 couples who were treated with In Vitro Fertilization (IVF), in Intracytoplasmic Sperm Injection (ICSI), and Intrauterine Insemination (IUI) was conducted in two infertility centers; while data from donated eggs or sperm and a surrogate uterus along with data from infertile couples with a combination of male and female factors were excluded. In this study, five ensemble machine-learning models were utilized to predict the success rate of clinical pregnancy. Results:Among the proposed ensemble models, the Random Forest (RF) model achieved the highest mean accuracy and area under the curve (AUC) and outperformed all other models in three procedures. Our results show that in cycles with 1 to 5 retrieved eggs, sperm motility and the count of sperm had a positive effect on the rate of clinical pregnancy. Furthermore, the results indicated that cut-off values of 54 (p-value=0.02, 95%-CIs (1.05, 2.13)) and 35 (p-value=0.03, 95% 95%-CIs (1.06, 2.86)) for the count parameter in IVF/ICSI, and IUI, respectively. In addition, a significant cut-off points of 30 (p-value < 0.001) was obtained for the morphology parameter in all procedures. Sperm parameters were negatively weighted in the model obtained by the RF. In addition, the acquired data illustrated that in each procedure, the morphology parameter demonstrated a significant difference in clinical pregnancy between successful and unsuccessful groups. Conclusion: The second course of IVF procedure increased success rates in clinical pregnancy in patients with lower-than-average sperm parameters, while the IUI technique was demonstrated to be more effective in patients with above-average of sperm parameters.