Abstract Study question Does a free-annotation embryo scoring system (IDAScore) prediction model effective in clinical assisted embryo selection? Summary answer iDAScore could be used to predict the clinical pregnancy and live birth. What is known already iDAScore model, as a free-annotation and deep learning-based embryo scoring system, was created according to deep learning model by training 115,832 embryos and 14,644 known implantation embryos from 18 reproductive medical center. Then, every blastocyst was scored without any manipulation and intervention by embryologists. Although iDAScore's performance was validated by a few reproductive centers, it has not yet been verified in a sizable cohort study. As a result, the effectiveness of iDAScore should be assessed in more reproductive clinics. Study design, size, duration A total of 6,291 vitrified-thawed single blastocyst transfer (SBT) cycles from 2018 to 2021 at the Reproductive Medicine Center, Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, China, were analyzed a free-annotation and deep learning-based embryo scoring system (IDAScore) (EmbryoScope Plus, Vitrolife, Sweden). Participants/materials, setting, methods All blastocysts undergoing SBT were scored from 1 to 9.9 with ida system. Blastocyst scores were grouped according to quartile.The effectiveness of ida system was evaluated according to clinical outcomes, perinatal and neonatal outcomes of different groups (1.0-8.0, 8.1-8.9, 9.0-9.3, 9.4-9.9). Main results and the role of chance The iDA scores decreased with increasing maternal age. Clinical pregnancy rate and live birth rate increased with iDAScores, and there was a significant correlation (p < 0.001). The abortion rate decreased slightly with the increase of iDAScores, but there was no significant difference (p > 0.05). Ectopic pregnancy rate did not change significantly with iDAScores. No statistically significant difference were observed between single and twin live birth among different iDAScore groups (p > 0.05). For perinatal and neonatal outcomes, no significant difference was shown in four iDAScore group (p > 0.05). Among them, gestational age and premature birth rate were similar (p > 0.05). Besides, no significant difference was shown in types and rate of pregnancy complication (p > 0.05). For newborn babies, the birth weight and sex were statistically similar across all groups (p > 0.05). The birth defect outcome was also evaluated. No significant difference were shown on different types of birth defect in all groups (p > 0.05). The uni-variate and multi-variate logistic regression for live birth was also analyzed. iDAScore were significantly correlated with a positive live birth probability (OR: 1.200, 95% CI: 1.148-1.253, p < 0.05). Limitations, reasons for caution The experimental samples in this study are all blastocyst transfer cycles. There is no guiding significance for other centers using the cleavage embryo transfer cycle. Wider implications of the findings iDAScore could be used to predict the clinical pregnancy and live birth. For the blastocyst that has obtained live birth, different iDAScore did not cause significant differences in obstetric and neonatal outcomes. Therefore, iDAScore determined blastocyst implantation more than fetal development. Trial registration number not applicable
Read full abstract