Correlation between artificial intelligence-based IDAScore and embryo morphokinetic parameters: a retrospective...

The rapid advancement of artificial intelligence (AI) has brought new opportunities for embryo evaluation. The integration of AI with time-lapse (TL) imaging technology has significantly enhanced the pregnancy rates in assisted reproductive technologies (ART). However, the fully automated AI-based embryo screening model, IDAScore lacks transparency and specific justifications in assessing the internal processes of embryos, and its interpretability remains limited. The aim of this study is to explore the correlation between IDAScore and embryonic morphological kinetics, providing a convincing explanation for AI-driven embryo assessment and selection.

A retrospective analysis was performed on clinical data from 102 first fresh embryo transfer cycles involving cleavage-stage embryos, all morphologically graded as 8-cell, grade 1 (8C1), and 210 first frozen-thawed blastocyst transfer cycles with embryos uniformly scored as 4AA, conducted at the Reproductive Medicine Center of the Affiliated Hospital of Guangdong Medical University between August 2023 and March 2025.All embryos were cultured using a time-lapse imaging system and evaluated by the fully automated AI-based scoring model, IDAScore V2.0.Cleavage-stage embryos were stratified into two groups based on their IDAScore: Group A [8cells, IDA:1.0-5.7]n=27and Group B [8cells, IDA:5.8-8.0]n=75. Blastocysts were similarly divided into Group C [IDA:1.0-7.7]n=98 and Group D [IDA:7.8-9.9]n=112. Clinical baseline characteristics, ovarian stimulation parameters, embryonic kinetic data, and pregnancy outcomes were compared across the four groups.

There were no significant differences in baseline clinical characteristics-including patient age, anti-Müllerian hormone (AMH) levels, body mass index (BMI), and basal hormone profiles-among the four groups. In cleavage-stage embryos, Group B (higher IDAScore) showed significantly faster developmental timings compared to Group A (low IDAScore), including time to 3 cells (t3), 4 cells (t4), 5 cells (t5), and 8 cells (t8), as well as the duration from 2-cell to 3-cell (cc2) and from 3-cell to 5-cell (cc3) stages (P < 0.05). Moreover, Group B demonstrated significantly higher rates of high-quality embryos, blastocyst formation, usable blastocysts, and clinical pregnancy compared to Group A (P < 0.05).Logistic regression analysis revealed that the cc2 interval was significantly associated with IDAScore on Day 3 (OR = 0.410; 95% CI: 0.253-0.666; P < 0.001).In blastocyst-stage embryos, Group D (higher IDAScore) exhibited significantly faster development to the 2-cell (t2), 3-cell (t3), 4-cell (t4), and 5-cell (t5) stages, as well as earlier time to full blastocyst expansion (tB), compared to Group C (low IDAScore) (P < 0.05). Additionally, significant differences were observed between Groups C and D in both inner cell mass (ICM) and trophectoderm (TE) grading (P < 0.05). The blastocyst formation rate in Group D was significantly higher, and the miscarriage rate was significantly lower than that in Group C, with both differences being statistically significant. Stepwise logistic regression analysis showed that the time to full blastocyst expansion (tB) (OR=0.780;95%CI: 0.721-0.845;P < 0.001) and trophectoderm (TE) grading (OR=4.224;95%CI:1.633-10.926;P=0.003) remained significant factors influencing blastocyst scoring.

There is a significant correlation between IDAScore and both embryonic morphology and kinetic parameters. Faster-developing embryos are generally associated with higher IDAScore values and demonstrate greater implantation potential. This finding provides a scientific basis and theoretical explanation for the use of IDAScore in embryo selection.