Placeholder Image

Deep Learning and Artificial intelligence to help predict the chance of pregnancy after frozen embryo transfers

Assisted reproductive technology (ART) is the most efficient treatment to address infertility, however, implantation failure remains one of the greatest challenges in ART despite reproductive advancement since its introduction 40 years ago.

Failed embryo implantation, especially repeat failure, causes psychological and financial burdens for patients, therefore, interventions that can minimize the likelihood of implantation failure are a focus of research for ART.

Mount Sinai Fertility’s Dr. Ellen Greenblatt was a driving force behind a study to help introduce “artificial intelligence (AI)” technologies in the assessment of endometrial histology (microscopic analysis) and is a promising approach to promote consistency and objectivity, and an effective way to get richer information than conventional endometrial dating methods, such as the Noyes’ criteria to analyze endometrial development. 

This study for the first time demonstrates the feasibility and excellence of Deep Learning (DL) in analyzing endometrial histology and predicting pregnancy outcomes for Frozen Embryo Transfers (FET).  The findings also suggest that important pregnancy-related DL coming from a convolutional neural network (CNN), a type of artificial neural network used primarily for image recognition and processing, due to its ability to recognize patterns in images, could be used in the future to assess biologic samples under the microscope. In such prospective trials, the correlation between this DL analysis and other meaningful clinical features can also be evaluated. Once validated, this DL-based prognostic tool may aid in deciding whether to proceed with a frozen embryo transfer (FET).in a given cycle.  A physician may, for example, proceed with embryo transfers until the live birth is achieved if the predicted chance of pregnancy is high; in contrast, physicians may delay embryo transfers and adjust endometrial preparation protocols for a more favorable intrauterine environment if the predicted chance of pregnancy is low, or suggest gestational carriers.