JBRA Assist. Reprod. 2025;29(Suppl 1):10-10
ORAL PRESENTATION
doi: 10.5935/1518-0557.20250067
1Instituto Sapientiae - Centro de Estudos e Pesquisa em Reprodução Assistida
2Science for EveryMind
3Future Fertility
4Fertility-FertGroup
Objective: Embryonic aneuploidy is a major challenge in IVF, affecting over 50% of preimplantation embryos and limiting viable pregnancies. While AI has shown considerable promise in predicting aneuploidy risk as a non-invasive alternative to genetic testing, a technology capable of assessing this at the oocyte stage would be highly valuable. The aim of this study was to evaluate whether AI-driven oocyte assessment can predict fertilization rate, embryo development and quality, and aneuploidy risk, while identifying key factors contributing to higher AI-determined oocyte quality.
Methods: In this retrospective cohort study, 14,605 oocytes from 2,306 ICSI cycles were evaluated between January 2020-May 2024. Oocyte images were captured immediately pre-ICSI and scored on a scale from 0-10, by an AI-tool (MAGENTA™) built to predict blastocyst development. Injected oocytes were incubated until day five, when 3,689 embryos were biopsied for PGT-A via nextgeneration-sequencing and received a diagnosis. Embryos were then split according to the diagnosis: euploid (n=1,274), aneuploid (n=2,219), and mosaic (n=196). To investigate the correlation of MAGENTA™ Scores (MS) to embryo aneuploidy, data were analysed using generalised linear models and Bonferroni post-hoc test, adjusting for confounders. The association between MS on oocyte fertilisation and blastocyst formation likelihood was examined. We assessed whether MS might be affected by demographic and cycle characteristics, including maternal age, body mass index (BMI), the total FSH dose, oestradiol levels, number of retrieved oocytes, and mature oocyte rate. Study power was >80%.
Results: A significant increase in MS was observed in those oocytes that developed into euploid embryos compared with aneuploid embryos; however, no significant differences in MS were found between mosaic and euploid embryos or between mosaic and aneuploid embryos (6.8±0.7, 6.7±0.1, and 6.4±0.5, p<0.01, for euploid, mosaic, and aneuploid embryos, respectively). MS were markedly lower in oocytes that failed to fertilize compared with those that successfully underwent fertilization (4.9±0.3 vs. 6.4±0.3, p<0.01). A consistent trend was observed for blastulation capacity, whereby oocytes progressing to the blastocyst stage exhibited markedly higher MS compared with those whose embryos failed to reach the blastocyst phase (6.6±0.6 vs. 5.1±0.5, p<0.01). Not only was MS predictive of blastocyst formation potential, but it also reflected overall blastocyst quality: high-quality blastocysts displayed higher MS than their counterparts (7.2±0.6 vs. 6.7±0.4, p<0.01). With respect to variables potentially impacting AIderived oocyte morphology scores, we observed a negative correlation exclusively with maternal age (OR: 0.032, CI: 0.12-0.50, p<0.01). No additional factors exerted a statistically significant effect on MS.
Discussion: In this study, we demonstrated that AI-driven evaluation of oocytes using the MAGENTA™ tool is effective in predicting fertilization success, blastocyst development, embryo quality, and aneuploidy risk. Our results reveal a significant correlation between higher MAGENTA™ scores (MS) and the development of euploid embryos, supporting the potential of AI to improve embryo selection. Moreover, MS proved to be a reliable predictor for fertilization and blastocyst formation, with higher scores associated with better-quality embryos. We also identified maternal age as a critical factor negatively affecting oocyte quality, with older women showing lower MS, consistent with the known impact of age on fertility. While our study focused exclusively on ICSI cycles, which may limit the broader application of our findings, the results remain valuable for refining embryo selection in assisted reproduction. The ability of AI to predict aneuploidy at the oocyte stage offers significant clinical advantages. It could reduce the need for invasive genetic testing, streamline clinical workflows, lower costs, and reduce patient burden. Further research is necessary to validate these findings in different patient populations and to integrate AI tools with other diagnostic techniques, such as genetic sequencing, to improve embryo selection accuracy. Thus, the use of AI-driven oocyte evaluation holds great promise for improving clinical outcomes in IVF. By offering a non-invasive, cost-effective method to enhance embryo selection, we believe that this technology could contribute to a more efficient and patient-friendly approach to fertility treatment.
Conclusion: AI-based oocyte evaluation reliably indicates chances of fertilisation, blastulation, blastocyst quality, and aneuploidy, with maternal age identified as the primary negative factor.