Research Finds Artificial Intelligence Enhances Embryo Selection Accuracy in IVF

How transparent and reliable are AI scoring models in embryo selection—and what do they actually measure? As artificial intelligence (AI) becomes woven into reproductive medicine, clinicians are right to question how AI-derived embryo scores relate to established morphokinetic parameters and whether they genuinely improve IVF outcomes.

Study Design: Retrospective Cohort Evaluates IDAScore Interpretation

This single-center retrospective study included 312 elective single embryo transfer cycles—102 fresh transfers of cleavage-stage embryos, all graded 8C1, and 210 frozen-thawed blastocyst transfers, all 4AA quality—at a large reproductive medicine center in China (Aug 2023–Mar 2025). Embryos were scored automatically using IDAScore V2.0, an AI model that analyzes continuous time-lapse (TL) imaging data. Patients and embryos were grouped by IDAScore for comparison of morphokinetics and clinical outcomes.

Results: Higher IDAScore Reflects Faster Embryo Development & Better Outcomes

Cleavage-stage embryos with higher IDAScore developed more rapidly to the 3-, 4-, 5-, and 8-cell stages, with shorter intervals between cell divisions (notably, the cc2 interval).

Blastocysts with higher IDAScore also reached key stages (2-, 3-, 4-, and 5-cell; blastocyst expansion) more quickly and scored better for inner cell mass and trophectoderm quality.

Clinical pregnancy rates, rates of high-quality embryos, blastocyst formation, and usable blastocysts were all significantly higher in groups with elevated IDAScore. Miscarriage rates were lower in high-score blastocyst groups.

Logistic regression identified cc2 as an independent predictor of cleavage-stage IDAScore, and time to blastocyst expansion (tB) plus trophectoderm quality as key predictors for blastocyst scoring.

Why It Matters: Moving Toward Interpretable AI in IVF

The key clinical implication is that IDAScore does not function as a mysterious “black box”—it is tightly correlated with established, observable embryo morphokinetics and morphology. Faster-developing embryos, as measured by TL parameters, consistently earn higher AI scores and show better implantation potential. This transparency builds trust in AI-driven embryo assessment and supports its use in patient counseling and embryo selection.

Technical Terms, Briefly Explained

IDAScore: An AI-based, fully automated scoring system that evaluates embryos via continuous TL imaging, generating real-time, second-level scores.

Morphokinetics: The timing and sequence of key embryonic developmental milestones, such as cell divisions and blastocyst expansion.

cc2, tB: Specific time intervals measuring cell division synchrony and timing to full blastocyst expansion, respectively.

Study Limitations

Single-center, retrospective design; all embryos were first-cycle transfers, possibly limiting generalizability.

No stratification by infertility etiology or ovarian stimulation protocol.

Results support further multicenter, larger-sample studies for external validation.

Conclusion

IDAScore aligns closely with embryo morphokinetics and reliably predicts clinical outcomes. These findings provide much-needed clarity for clinicians using AI-based embryo selection, especially in non-Western populations, and mark a step forward for transparent, reproducible, and effective IVF care.

Key Points

High IDAScore embryos develop faster and are more likely to result in pregnancy.

AI-derived scores correlate strongly with established morphokinetic markers.

cc2 interval and blastocyst expansion timing independently predict IDAScore.

Using IDAScore can enhance confidence in embryo selection and patient counseling.

Further large, multicenter studies are warranted for wider adoption.

Citation:

He L., Ke Q., Li H. et al. Correlation between artificial intelligence-based IDAScore and embryo morphokinetic parameters: a retrospective cohort study. BMC Pregnancy Childbirth. 2026; [in press]. https://doi.org/10.1186/s12884-026-09662-5