Predicting the Likelihood of Preterm Labor of Embryos Prior to Transfer into the Uterus using Machine Learning

Tamar Amitai 1 Yoav Kan-Tor 1 Adi Szeskin 1 Assaf Ben-Meir 3 Amnon Buxboim 1,2
1The Hebrew University of Jerusalem, Israel
2The Hebrew University of Jerusalem, Israel
3Hebrew University Hadassah Medical Center, Israel

Preterm labor is a devastating experience for any couple – especially for couples undergoing infertility treatments. On average, 10% of pregnancies lead to first trimester preterm labor. Preterm labor is associated with chromosomal aberrations in the embryo. Hence, early evaluation of the likelihood to preterm labor, high-risk embryos can be deselected for transfer into the uterus as part of in vitro fertilization (IVF) procedures. We compiled a database of 500 embryos that successfully implanted in the uterus following transfer. For each embryo, the database consists of a time-lapse video file showing its preimplantation development, clinical metadata and maternal information that we obtained from five IVF clinics across Israel. Embryos were represented based on their morphokinetic profiles and labeled for preterm labor earlier than week 14 of pregnancy. We used machine learning tools to score embryos for their likelihood to undergo preterm labor. By screening one hundred features, highly-predictive ones were selected that were associated with multiple developmental processes. Based on these features, we demonstrate a statistically significant classification of preterm labor on day-3 of embryo culture. Our research has the capacity of improving infertility treatment, thus decreasing emotional pain to the patients and shortening time to pregnancy.

Powered by Eventact EMS