Are women cryopreserving enough oocytes for fertility preservation? A probabilistic model integrating AI-based oocyte morphology and age-dependent euploidy

Abstract STUDY QUESTION How many mature oocytes are biologically required to achieve at least three euploid blastocysts, by maternal age and AI-derived oocyte quality, and how does this estimate compare with real-world oocyte yields in elective fertility preservation? SUMMARY ANSWER Required oocyte numbers rise steeply with advancing maternal age and decreasing AI-derived oocyte quality, and the median number cryopreserved per fertility preservation cycle falls progressively below estimated requirements, by 12- to 30-fold in women older than 40 years. WHAT IS KNOWN ALREADY Elective oocyte cryopreservation is increasingly used to preserve future reproductive potential, but counselling typically relies on uniform age-based oocyte targets that do not account for inter-individual variability in oocyte biological competence. Although AI-based oocyte assessment has shown promise as an objective adjunct to morphological evaluation, no biologically integrated framework has been used to translate AI-derived oocyte quality into a quantitative requirement-versus-yield comparison in fertility preservation. STUDY DESIGN, SIZE, DURATION Retrospective cohort study with probabilistic modelling of blastulation and euploidy, using three datasets from a single private university-affiliated IVF centre. Model development used 13,782 mature oocytes from 1,625 autologous ICSI cycles (1,020 patients, 2021–2024) and 3,325 biopsied embryos with conclusive PGT-A results (857 patients). External validation was performed in 1,948 oocytes from 333 ovum donation cycles in 271 donors (2020–2024). Real-world application used 1,647 elective fertility preservation cycles in 1,100 patients between 2012 and 2024. PARTICIPANTS/MATERIALS, SETTING, METHODS Each oocyte was scored using a commercially available artificial-intelligence image-analysis system that assigns a 0–10 oocyte quality score. Blastulation and euploidy were modelled by logistic regression with cluster-robust standard errors. The integrated probabilistic framework computed the minimum number of mature oocytes required to obtain at least three euploid blastocysts at a probability of 90% or greater, by maternal age category and tertile of AI-derived oocyte quality score. Model-derived requirements were compared with the median number of oocytes actually cryopreserved per fertility preservation cycle. MAIN RESULTS AND THE ROLE OF CHANCE Blastulation increased with higher AI-derived oocyte quality score (OR 1.19 per point, 95% CI 1.18–1.21; P < 0.001) and decreased with maternal age (OR 0.94 per year, 95% CI 0.93–0.96; P < 0.001); the oocyte quality effect was externally validated in donors (OR 1.11 per point, 95% CI 1.08–1.14). Euploidy declined non-linearly with age, from 68.6% (<30 years) to 19.3% (>40 years). Estimated mature-oocyte requirements ranged from 9 (95% CI 8–11) at < 30 years with high AI-derived oocyte quality to 121 (104–146) at > 40 years with low AI-derived oocyte quality. The median number cryopreserved per cycle decreased from 8.3 at < 30 to 4.0 at > 40 years and fell progressively below estimated requirements; in women older than 40 years the observed yield was 12- to 30-fold below the estimated requirement. LIMITATIONS, REASONS FOR CAUTION Blastulation and euploidy probabilities were derived from autologous ICSI cycles in patients with infertility-related indications; extrapolation to elective fertility preservation candidates assumes that per-oocyte biological competence, conditional on age and AI-derived quality, is transportable across populations. The probabilistic framework assumes conditional independence between blastulation and euploidy. The model estimates oocyte requirements for euploid blastocyst formation but does not directly model implantation, clinical pregnancy, or live birth. WIDER IMPLICATIONS OF THE FINDINGS The divergence between observed yields and model-derived requirements is best interpreted not as a deficiency in retrieval practice but as an individualised benchmark. Integrating AI-based oocyte assessment with age-specific euploidy probabilities may support more personalised, biologically informed counselling, expectation-setting, and care planning, particularly for women considering oocyte cryopreservation after the mid-thirties. FUNDING This research did not receive any specific grant from public, commercial, or not-for-profit funding agencies. DISCLOSURES J.F. and P.M. are employees of Future Fertility, the developer of the artificial-intelligence oocyte-scoring system used in this study. The remaining authors declare no competing interests. TRIAL REGISTRATION NUMBER N/A