Periodic Reporting for period 1 - IVFMakeSense (An embryo transfer decision-support tool optimizes live-birth rates in IVF treatments)
Période du rapport: 2021-05-01 au 2023-04-30
There is a growing need for in vitro fertilization - embryo transfer (IVF-ET) treatments owing to socio-demographic changes worldwide, including reproductive ageing. The overall goal of this project was to develop computational classifiers that can asses the developmental potential of the embryos that are available for transfer using machine learning. The algorithms that we generated are compatible with current clinical needs and can be integrated to improve fertility treatments by overcoming existing impediments as specified below.
Even if an embryo has successfully implanted in the uterus and a pregnancy was generated, there are too many incidents that live birth was not the outcome due to embryo miscarriage. Amitai et al. 2022 reports the very first classifier that evaluates the risk of a given embryo to undergo first trimester miscarriage (the most prevalent miscarriage event) during preimplantation stages. Hence, using this classifier, the clinicians can select embryos for transfer not only based on their implantation and pregnancy potential but also based on their risk of miscarriage. This is a significant step forward.
The evaluation of the developmental potential of embryos is often based on morphokinetic annotation, namely the time points from fertilization in which specific events occur. This includes pronuclei appearance and fading, blastomere cleavage, morula compaction and blastulation. Despite the established utility morphokinetic evaluation of embryo potential, these models are rarely utilized in the IVF lab due to the overwhelmingly busy environment where dedicating time and effort by the medical staff is often not permitted. Moreover, manual annotation may introduce inter-observer and intra-observer variations that compromise the validity of this approach. Hence, we set out to generate an algorithm that will perform automatic morphokinetic annotation in real time. Zabari et al. reports the development of an automatic annotation classifier that we further used to generate profiles of ten of thousands embryos. In this manner we managed to classify human embryo into several clusters based on the morphokinetic profiles. By correlating the specific features of these clusters with their developmental potential to be selected for transfer and implant in the uterus, we elucidated a specific event that differentiates between poor quality and high quality embryos (quality related to embryo developmental terms).
We are currently seeking mechanisms for implementing the decision support tools that we developed in IVF clinics to improve infertility treatments by increasing cumulative pregnancy and live birth rates, decreasing medical risks and shortening time to pregnancy.
Even if an embryo has successfully implanted in the uterus and a pregnancy was generated, there are too many incidents that live birth was not the outcome due to embryo miscarriage. Amitai et al. 2022 reports the very first classifier that evaluates the risk of a given embryo to undergo first trimester miscarriage (the most prevalent miscarriage event) during preimplantation stages. Hence, using this classifier, the clinicians can select embryos for transfer not only based on their implantation and pregnancy potential but also based on their risk of miscarriage. This is a significant step forward.
The evaluation of the developmental potential of embryos is often based on morphokinetic annotation, namely the time points from fertilization in which specific events occur. This includes pronuclei appearance and fading, blastomere cleavage, morula compaction and blastulation. Despite the established utility morphokinetic evaluation of embryo potential, these models are rarely utilized in the IVF lab due to the overwhelmingly busy environment where dedicating time and effort by the medical staff is often not permitted. Moreover, manual annotation may introduce inter-observer and intra-observer variations that compromise the validity of this approach. Hence, we set out to generate an algorithm that will perform automatic morphokinetic annotation in real time. Zabari et al. reports the development of an automatic annotation classifier that we further used to generate profiles of ten of thousands embryos. In this manner we managed to classify human embryo into several clusters based on the morphokinetic profiles. By correlating the specific features of these clusters with their developmental potential to be selected for transfer and implant in the uterus, we elucidated a specific event that differentiates between poor quality and high quality embryos (quality related to embryo developmental terms).
We are currently seeking mechanisms for implementing the decision support tools that we developed in IVF clinics to improve infertility treatments by increasing cumulative pregnancy and live birth rates, decreasing medical risks and shortening time to pregnancy.