The application of machine learning methods to evaluate predictors of live birth in programmed thaw

system performed significantly (P<0.05) better than the embryologists in selecting two embryos for transfer among which at least one will eventually form a blastocyst. The accuracy of the CNN in selecting an embryo at 70 hpi, which developed into a high-quality blastocyst (HQB) for a single embryo transfer (SET),was 63.9% that is significantly higher (P<0.05) than the average accuracy of the embryologists (52.8%, 95%CI: 48.6-57.0%). The accuracy of the CNN in selecting an embryo at 70 hpi, which developed into HQB for a double embryo transfer (DET), was significantly higher (79.4%, P<0.05) compared to the embryologists with an average accuracy of 72.4% (95% CI: 70.7-74.0%). CONCLUSIONS: Here,we reportedanartificial intelligence-based approach for predicting the developmental fate of cleavage stage embryos. Our study shows that the developed CNN outperforms an embryologist’s morphologic assessment at 70 hpi in predicting blastocyst formation.Additionally,we demonstrated that this technology might be used to select embryos with the highest invitro developmental potential. Utilization of artificial intelligence (AI) technologies in human IVF practices may allow for more objective/standardized methods for improving embryo selection. Reference: None. SUPPORT: Financial Support:Â This work was partially supported by the Brigham Precision Medicine Developmental Award (Brigham Precision Medicine Program, Brigham and Women’s Hospital) and 1R01AI118502, R01AI138800, and R21HD092828 (National Institute of Health).