Deep Learning Models: Their Relationship with Embryonic Euploidies and Reproductive Outcomes

Area of Science:

• Reproductive Medicine
• Artificial Intelligence in Healthcare
• Embryology

Background:

• Embryo selection in in vitro fertilization (IVF) seeks to identify embryos with the highest reproductive potential.
• Preimplantation genetic testing for aneuploidy (PGT-A) is the standard but is invasive and not always feasible.
• Deep learning (DL) models, like the iDA score, offer non-invasive embryo assessment alternatives.

Purpose of the Study:

• To critically evaluate the relationship between iDAScore (versions 1.0 and 2.0), embryo euploidy, and clinical outcomes (live birth, miscarriage).
• To assess the predictive performance of iDAScore compared to chromosomal status and reproductive results.

Main Methods:

• A narrative review of studies published from January 2020 to May 2025.
• Searches conducted on PubMed and Google Scholar using keywords: 'iDAScore,' 'deep learning,' 'euploidy,' 'live birth.'
• Inclusion of English-language, full-text studies assessing iDAScore's predictive accuracy for chromosomal status or reproductive outcomes.

Main Results:

• Six retrospective studies were included, all showing a significant association between higher iDAScore values and embryo euploidy.
• Area Under the Curve (AUC) values for euploidy prediction ranged from 0.60 to 0.68.
• iDAScore positively correlated with live birth rates and negatively with miscarriage rates, with moderate accuracy in euploid cohorts.

Conclusions:

• iDAScore shows potential as an adjunct tool for embryo assessment in IVF.
• It may assist in prioritizing embryos for transfer when PGT-A is not feasible.
• Further large-scale prospective studies are needed to confirm its clinical utility.