Evaluating Embryo Developmental Potential based on Deep Learning of Time-lapse Imaging and Real-time Rheology

Today, in vitro fertilization – embryo transfer (IVF-ET) treatments account for 1 out of 25 newborns in Israel. After 40 years of clinical practice, there is no reliable non-invasive method for identifying the small fraction of embryos that possess the highest potential to implant and proceed to full term labor. To maintain reasonable pregnancy rates, current practice involves the transfer of multiple non-genetically diagnosed embryos into the uterus. As a result, we cannot prevent first trimester miscarriage, which is associated with chromosomal aneuploidy such as trisomy 21 (Down syndrome), and we cannot avoid the clinical complications and health risks to the newborn and the mother, which are associated with pregnancies of multiple fetuses. We designed a multiplate based device that facilitates time-lapse imaging of preimplantation development combined with real-time rheological profiling during incubation over several days in optimal culture conditions. Our deep learning algorithms outperform current state-of-the-art classifiers and offer fully automated and standardized protocol. Our preliminary data establish a link between oocyte stiffness and its capacity to mature and to undergo normal fertilization. Taken together, we present an integrated approach that combines physical characterization and visual assessment to score the developmental capacity of preimplantation embryos.