EUVIMEDEUVIMED entdecken
Crossref

L26/P-194 Artificial intelligence redefines oocyte donation programs: from random to individual, probability-based oocyte allocation to optimize blastocyst yield

L. Murria, L. Carrión-Sisternas, A. Gamero-Gómez, C. Micheletti, L. Bori, A. Cobo, M. Meseguer

Human Reproduction · 01.07.2026 · 41 (Supplement_1)

DOI: 10.1093/humrep/deag083.530 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Human Reproduction, online erschienen am 08.07.2026. Verfasst von L. Murria und 6 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Study question Can artificial intelligence (AI)-based oocyte assessment optimize donor oocyte allocation in IVF by reducing the number of oocytes assigned per recipient without compromising blastocyst outcomes? Summary answer AI-based oocyte assessment generated oocyte groups identifying which and the minimum number of oocytes required for predefined probabilities of achieving ≥1 or ≥ 2 blastocysts. What is known already Oocyte assessment in clinical practice remains largely morphology-based and subjective, with high observer variability and limited ability to predict developmental competence. Consequently, oocyte donation programs and oocyte banks lack tools to ensure predefined outcomes or determine which and how many oocytes are required for treatment success. In the absence of objective criteria, oocyte allocation has traditionally relied on random or empirical strategies, often involving fixed large groups of 8-12 oocytes per recipient. Recently, AI-based approaches have emerged to estimate oocyte developmental potential. ROSETM is an AI-based tool designed for allocating oocytes into groups achieving predefined reproductive outcomes. Study design, size, duration This retrospective single-centre study included oocyte images from donor cycles performed between 2015 and 2019, restricted to cycles in which embryos were cultured in time-lapse incubators. A total of 1,226 donation cycles comprising 14,544 oocytes were analysed. ROSE generated individual oocyte competence predictions and automatically allocated oocytes into groups targeting a minimum predicted probability of 85% of achieving at least one or at least two blastocysts. Model performance was evaluated at the group level. Participants/materials, setting, methods Oocyte images were obtained immediately after intracytoplasmic sperm injection (ICSI) using time-lapse incubators (EmbryoScope). ROSE sequentially allocated oocytes from each cohort into groups until the predefined probability threshold was reached for achieving at least one or at least two blastocysts. Remaining oocytes not meeting the threshold formed lower-probability residual groups. Allocated groups were compared with observed blastocyst outcomes using contingency tables. Group-level precision and the number of blastocysts obtained per group were assessed. Main results and the role of chance The AI-based oocyte assessment algorithm generated oocyte groups with a predicted probability ≥85% of obtaining at least one blastocyst. These groups had a mean size of 2.5 oocytes and were stratified by the most frequent group sizes: 3 (n = 3,067), 4 (n = 964), and >4 oocytes (5-7 oocytes; n = 107). In these groups, the observed clinical rate of obtaining at least one blastocyst was 89% for 3-oocyte groups, 88% for 4-oocyte groups, and 90% on average for groups >4 oocytes. Prediction precision for the ≥1 blastocyst endpoint, based on contingency table analysis, was consistent across group sizes (89.62–89.84%). Among correctly predicted positive groups, the observed proportion achieving more than one blastocyst increased with group size (67%, 72%, and 77%, for 3, 4, and >4 oocytes, respectively). For the ≥2 blastocyst endpoint, a mean group size of seven oocytes was needed to reach the ≥85% predicted probability according to the AI model. Seven-oocyte groups showed an observed ≥2 blastocyst rate of 87% (precision 91.94%), while eight-oocyte groups achieved an observed rate of 94% (precision 96.81%). Among correctly classified groups, more than two blastocysts were obtained in 91.23% of seven-oocyte groups and 95.31% of eight-oocyte groups. Limitations, reasons for caution This study is limited by its retrospective and single-centre design. Prospective studies are required to determine whether this tool improves oocyte allocation compared with current random or empirical strategies. Wider implications of the findings This data-driven, AI-based approach enables competence-based donor oocyte allocation to meet predefined blastocyst probability targets, reducing the number of oocytes per recipient compared with current empirical fix-group allocation (9–12 oocytes). This approach improves outcome predictability and resource efficiency, increases the number of recipients who benefit, and reduces embryo surplus. Trial registration number No

Zitieren

APA

Murria, L., Carrión-Sisternas, L., Gamero-Gómez, A., Micheletti, C., Bori, L., Cobo, A., & Meseguer, M. (2026). L26/P-194 Artificial intelligence redefines oocyte donation programs: from random to individual, probability-based oocyte allocation to optimize blastocyst yield. Human Reproduction, 41(Supplement_1). https://doi.org/10.1093/humrep/deag083.530

Vancouver

Murria L, Carrión-Sisternas L, Gamero-Gómez A, Micheletti C, Bori L, Cobo A, et al. L26/P-194 Artificial intelligence redefines oocyte donation programs: from random to individual, probability-based oocyte allocation to optimize blastocyst yield. Human Reproduction. 2026;41(Supplement_1). doi:10.1093/humrep/deag083.530

BibTeX
@article{Murria2026,
  author = {Murria, L. and Carrión-Sisternas, L. and Gamero-Gómez, A. and Micheletti, C. and Bori, L. and Cobo, A. and Meseguer, M.},
  title = {L26/P-194 Artificial intelligence redefines oocyte donation programs: from random to individual, probability-based oocyte allocation to optimize blastocyst yield},
  journal = {Human Reproduction},
  year = {2026},
  volume = {41},
  number = {Supplement_1},
  doi = {10.1093/humrep/deag083.530}
}
Kostenlos starten →