P-274 Clinical limitations of manual oocyte quality scoring by embryologists are overcome by an artificial intelligence (AI) oocyte image analysis tool

Abstract Study question Do manual embryologist annotations (individual/combined) of dysmorphisms within mature denuded oocytes display better correlation with blastocyst development compared to an AI oocyte assessment tool? Summary answer Manual scoring cannot reliably determine quality in comparison assessments by image analysis tool (MAGENTA), which significantly correlates formation. What is known already Despite the importance on development, there no validated visual employed clinical practice. While individual can be identified, none have been shown consistently correlate laboratory and reproductive outcomes. Even when systems aggregate multiple dysmorphisms, correlations fall short. An (MAGENTA) has trained parameters oocyte, including those that may imperceptible human eye, outcomes, producing more objective, accurate time-saving assessments. design, size, duration A retrospective dataset consisting 1009 from 166 IVF-ICSI patients attending Bahçeci Umut ART Centre (Istanbul, Turkey) during 2016-2019 was assessed. 879 were included final data analysis, after removing missing morphological assessments, lab outcomes (fertilization development), or whose static images could not linked The resulting correspond 149 ages 22-45 (mean age 33.9 years old). Participants/materials, setting, methods Two embryologists, blinded retrospectively annotated for 7 dysmorphism categories: shape/size, cytoplasmic granularity, polar body fragmentation, zona pellucida thickness, perivitelline space smooth endoplasmic reticulum (SER) presence, vacuole presence. Annotations assigned points combined into overall score ranging 0-10 based a recent review (Bartolacci 2022). Each, previously unseen, analyzed MAGENTA, assigning (0-10), where increasing scores greater development. Main results role chance between each observed categories outcome assessed individually using Chi-squared test. Only presence SER found correlated whether became (p < 0.05), however sample size small (27 out 879; 3% dataset). MAGENTA Welch’s two-sample t-test. mean whole 9.2. No differences developed blastocysts (9.1) did (9.2). system assigns 10 unless are noted, defaulting assumption high noted otherwise. This reflected skewed distribution most receiving due lack visible dysmorphisms. Conversely, 4.2. Oocytes had higher (4.7) (4.2) 0.05). evenly distributed across entire spectrum (0-10). Limitations, reasons caution To avoid reducing male factor influence removed. performed retrospectively, consequently allowing embryologists time assess than typically available. number fragmentation too few appropriately Wider implications findings Because rare, subjective, don’t annotating them clinically nonbeneficial inefficient use embryologist’s time. immediate, easily incorporated workflow improved insights decision-making. Trial registration Not applicable.