EUVIMEDEUVIMED entdecken
Crossref

L26/P-245 Real-time 3D morphokinetic analysis for non-invasive embryo ploidy prediction using a time-lapse neural network

G.L. Dantas Trajano, G. Guzzon, H. Pelanda

Human Reproduction · 01.07.2026 · 41 (Supplement_1)

DOI: 10.1093/humrep/deag083.581 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Human Reproduction, online erschienen am 08.07.2026. Verfasst von G.L. Dantas Trajano und 2 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Study question Can a three-dimensional time-lapse deep learning model non-invasively predict embryo ploidy with accuracy similar to preimplantation genetic testing for aneuploidy (PGT-A)? Summary answer A novel 3D morphokinetic neural network accurately predicted embryo ploidy, supporting its value as a non-invasive embryo selection tool. What is known already PGT-A is widely used to improve embryo selection but requires an invasive trophectoderm biopsy, increasing treatment cost and complexity. Time-lapse imaging has enabled morphokinetic assessment, and recent artificial intelligence models using two-dimensional image analysis have demonstrated associations with implantation and ploidy status. However, embryos are three-dimensional dynamic structures, and conventional 2D approaches might be unable to capture subtle spatial-temporal developmental features associated with chromosomal competence. Study design, size, duration This retrospective multicentre cohort study analysed time-lapse recordings of blastocyst-stage embryos cultured over multiple years. Embryo development data were linked to corresponding PGT-A results, which served as the reference standard. Model development, training, and evaluation were conducted using temporally separated datasets to minimise bias and reduce the risk of data leakage. Participants/materials, setting, methods Time-lapse videos from standard incubator systems were reconstructed into volumetric representations of embryos. A custom 3D convolutional neural network analysed continuous morphokinetic development from fertilisation to the blastocyst stage. The model was trained to classify embryos as euploid or aneuploid. Performance was assessed using the area under the receiver operating characteristic curve (AUC), sensitivity, specificity, and calibration metrics. Main results and the role of chance The three-dimensional neural network demonstrated strong discriminatory ability for embryo ploidy classification in the independent test dataset. Performance was consistently superior to baseline morphokinetic approaches and remained stable across relevant clinical subgroups. The model identified reproducible spatial–temporal developmental patterns associated with euploid embryos. Internal validation employing resampling and randomization analyses demonstrated that model effectiveness was significantly above chance and robust to sampling variability. When applied as an embryo-ranking tool, the model prioritised embryos that were later confirmed as euploid. Limitations, reasons for caution The retrospective design and reliance on PGT-A as an imperfect reference standard limit causal inference. Prospective validation with implantation and live-birth outcomes is required. Wider implications of the findings Three-dimensional AI-based embryo assessment may lessen reliance on invasive biopsy, lower costs, and facilitate accessibility to advanced embryo selection strategies in IVF. Trial registration number No

Zitieren

APA

Trajano, G. D., Guzzon, G., & Pelanda, H. (2026). L26/P-245 Real-time 3D morphokinetic analysis for non-invasive embryo ploidy prediction using a time-lapse neural network. Human Reproduction, 41(Supplement_1). https://doi.org/10.1093/humrep/deag083.581

Vancouver

Trajano GD, Guzzon G, Pelanda H. L26/P-245 Real-time 3D morphokinetic analysis for non-invasive embryo ploidy prediction using a time-lapse neural network. Human Reproduction. 2026;41(Supplement_1). doi:10.1093/humrep/deag083.581

BibTeX
@article{Trajano2026,
  author = {Trajano, G.L. Dantas and Guzzon, G. and Pelanda, H.},
  title = {L26/P-245 Real-time 3D morphokinetic analysis for non-invasive embryo ploidy prediction using a time-lapse neural network},
  journal = {Human Reproduction},
  year = {2026},
  volume = {41},
  number = {Supplement_1},
  doi = {10.1093/humrep/deag083.581}
}
Kostenlos starten →