O-158 Prediction of implantation of vitrified-warmed blastocysts using a deep learning algorithm on a single post-warming image of each embryo

Abstract Study question Can a deep learning algorithm determine the implantation likelihood of vitrified blastocysts from single time-point post-warming time-lapse images obtained EmbryoScope? Summary answer A significant positive correlation was found between score given by and rate frozen analyzed. What is known already Despite safety success vitrification technique, many best morphologically classified embryos prior to do not maintain their characteristics when returned room temperature. Since artificial intelligence (AI) has been shown be useful in assessing quality fresh at different stages, it could also used predict outcome embryo transfers. This first attempt distinguish that will implant those applying on image taken period warming transfer. design, size, duration retrospective single-centre study included 689 with data. All them were warmed using Cryotop method (Kitazato, Biopharma, Japan). Warmed assessed experienced embryologists according ASEBIR morphological scoring system degree expansion. Immediately after warming, placed EmbryoScope (Vitrolife, Denmark) incubators for 2-5 hours let re-expand until Participants/materials, setting, methods pre-transfer provided (Life Whisperer Viability), which scores 0 10. Data analyzed chi-square test or one-way ANOVA. Multivariate logistic regression analysis performed including score, oocyte origin (donated vs. autologous), age, patient handling (fresh eggs), day (5 6). The area under receiver operating characteristic (ROC) curve (AUC) calculate evaluation performance. Main results role chance There difference implanted non-implanted embryos. mean value 4.98 [95% CI 4.70-5.26]* 6.16 5.84-6.48]* When dividing quartiles, each quartile 27.7% Q1, 33.7% Q2 (Odds ratio (OR)=1.31 [0.82–2.09]), 44.8% Q3 (OR = 2.05 [1.30–3.22]*) 48.3% Q4 2.46 [1.56–3.87]*). contribution increased unit 1.16 CI: 1.09–1.23]*) statistically significant. successfully predicted an AUC 0.65 0.60–0.69]*. Moreover, better than classification, achieved 0,62 0.58–0.66] predicting implantation. *P < 0.001. Limitations, reasons caution Our clinic involved development algorithm. Furthermore, designed assess but thawed ones, so this should considered external trial. full video AI possibly provide additional information. Wider implications findings use predictive models cycles may help select transfer result failure early enough thaw another one success. It indicates can generalize assessment. Trial registration number -