EUVIMEDEUVIMED entdecken
Crossref

L26/P-137 Who predicts live birth best? A comparison of artificial inteligence, KIDScore™ and embryologist-based embryo selection

A. Rodriguez Isern, L. Peinado, P. Ruiz, F. Moffa, M. De Rocco, P. Andrade, A. Marta, S. Rovira, S. Novo

Human Reproduction · 01.07.2026 · 41 (Supplement_1)

DOI: 10.1093/humrep/deag083.473 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Human Reproduction, online erschienen am 08.07.2026. Verfasst von A. Rodriguez Isern und 8 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Study question Do morphological, morphokinetic and artificial intelligence–based strategies differ in how they prioritise blastocysts for live birth after single embryo transfer? Summary answer AI-based ranking showed a non-significant monotonic trend in prioritising embryos for live birth, providing greater internal differentiation than morphology- and morphokinetic-based approaches. What is known already Embryo prioritisation is a critical step in IVF, particularly in single embryo transfer strategies. Conventional embryologist-based selection relies on blastocyst morphology, while morphokinetic algorithms and artificial intelligence–based systems have been developed to support embryo selection. However, direct comparisons between these prioritisation approaches and their association with real reproductive outcomes remain limited, especially in cohorts where all embryos have been transferred and final outcomes are fully known. Study design, size, duration Retrospective cohort study including 104 oocyte recipient cycles performed between 2020 and 2024. All cycles exhausted all embryos available for transfer, allowing complete follow-up of reproductive outcomes at the cycle and embryo-transfer level. Only cycles with at least two embryos available for prioritisation and single blastocyst cryotransfer were included. Participants/materials, setting, methods 985 fresh donor oocytes generated 694 embryos, 357 usable blastocysts were transferred. Embryos were retrospectively prioritised using three strategies: embryologist ranking based on Gardner morphology, KIDScore D5 v3.1 morphokinetic ranking (Vitrolife), and the Chloe EQ v3.10 artificial intelligence system. Agreement between rankings was assessed using Spearman correlation and weighted Cohen’s kappa. Prioritisation performance was evaluated by comparing outcome gradients across ranking categories. Discriminative ability for clinical pregnancy and live birth was assessed using ROC curves. Main results and the role of chance Agreement between embryo prioritisation strategies was high, particularly between embryologist-based morphology and KIDScore (Spearman ρ = 0.915; weighted Cohen’s κ = 0.925). Concordance between Chloe EQ system and the other two was moderate (Spearman ρ = 0.688 with KIDScore and 0.677 with embryologist ranking; weighted κ ≈ 0.72). Discriminative ability for clinical pregnancy was modest across strategies, with AUCs of 0.543 (95% CI 0.485–0.597) for embryologist-based morphology, 0.527 (0.471–0.584) for KIDScore D5 v3.1, and 0.533 (0.474–0.594) for the Chloe EQ v3.10 system. Discrimination for live birth was close to chance level for all approaches, with AUCs of 0.519 (95% CI 0.463–0.574) for morphology, 0.508 (0.451–0.564) for KIDScore, and 0.523 (0.463–0.583) for Chloe EQ, indicating no meaningful predictive capacity. When live birth rates were analysed across ordered ranking categories, morphology-based prioritisation showed substantial overlap between positions (43.3%, 44.0% and 44.4% for ranks 1–3; p = 0.99), as did KIDScore (43.3%, 42.2% and 47.6%; p = 0.78). The Chloe EQ system showed a monotonic decrease in live birth rates from rank 1 (46.6%) to rank 2 (41.7%) and rank 3 (35.9%), although this trend did not reach statistical significance (p = 0.39), indicating a clearer internal separation among top-ranked embryos. Limitations, reasons for caution The retrospective design and restriction to oocyte recipient cycles may limit generalisability. Among 694 embryos cultured, the Chloe EQ system ranked 93 embryos within the top three positions; 61 were classified by embryologists as poor-quality blastocysts and not vitrified, and their reproductive potential if transferred remains unknown. Wider implications of the findings While none of the evaluated strategies improved global prediction of live birth, AI-based prioritisation provided greater granularity among top-quality blastocysts. This may support clinical decision-making when multiple embryos share similar morphological scores, suggesting a complementary rather than substitutive role for AI in embryo selection. Trial registration number No

Zitieren

APA

Isern, A. R., Peinado, L., Ruiz, P., Moffa, F., Rocco, M. D., Andrade, P., Marta, A., Rovira, S., & Novo, S. (2026). L26/P-137 Who predicts live birth best? A comparison of artificial inteligence, KIDScore™ and embryologist-based embryo selection. Human Reproduction, 41(Supplement_1). https://doi.org/10.1093/humrep/deag083.473

Vancouver

Isern AR, Peinado L, Ruiz P, Moffa F, Rocco MD, Andrade P, et al. L26/P-137 Who predicts live birth best? A comparison of artificial inteligence, KIDScore™ and embryologist-based embryo selection. Human Reproduction. 2026;41(Supplement_1). doi:10.1093/humrep/deag083.473

BibTeX
@article{Isern2026,
  author = {Isern, A. Rodriguez and Peinado, L. and Ruiz, P. and Moffa, F. and Rocco, M. De and Andrade, P. and Marta, A. and Rovira, S. and Novo, S.},
  title = {L26/P-137 Who predicts live birth best? A comparison of artificial inteligence, KIDScore™ and embryologist-based embryo selection},
  journal = {Human Reproduction},
  year = {2026},
  volume = {41},
  number = {Supplement_1},
  doi = {10.1093/humrep/deag083.473}
}
Kostenlos starten →