Search for a command to run...
Abstract Study question Can artificial intelligence (AI) predict utilisable blastocyst conversion rates from day 3 (D3) and day 5 (D5) embryos? Summary answer Higher D3 and D5 iDAScore values correlate with better morphology, and iDAScore can strongly predict conversion to utilisable blastocysts. What is known already iDAScore is an AI model developed using deep learning and a neural network, trained to analyse time lapse images of developing embryos. Integrated directly with the time lapse platform; a score is automatically generated for each embryo, correlating with implantation, live birth rates and euploidy. A significant improvement in the predictive performance of iDAScore in clinical outcomes has been demonstrated from cleavage to blastocyst stages, as available developmental information increases. While iDAScores values have been shown to correlate with morphological and morphokinetic events, currently there is limited published data on the ability of AI to successfully predict blastocyst conversion. Study design, size, duration A prospective single centre study, including 2393 embryos from 336 patients undergoing fresh IVF treatment cycles, conducted between February and June 2024. There are two types of iDAScore values; those which are temporary and continuously change based on developmental information provided to the system (Continuous) and a final saved value generated at the end of the culture period (Finalised). Continuous iDAScore values were recorded during routine morphological grading on D3 and D5. Participants/materials, setting, methods Embryos were individually cultured using GTL culture medium (Vitrolife), in the EmbryoscopeTM + time lapse incubator, under 6% CO2 and 5% O2 at 37 °C, for a maximum of 6 days. The ability of Continuous iDAScore v2.0 values on D3 and D5 to discriminate the embryos likely to develop into utilisable blastocysts, was assessed using logistic regression, area under the curve (AUC) and t-tests on STRATA 16 and GraphPad Prisim statistics platforms. Main results and the role of chance A significant positive association was demonstrated between D3 Continuous iDAScores and D3 and D5 Morphology (p < 0.001). The AUC for D3 Continuous iDAScore was 0.8583 for D3 Morphology and 0.7770 for D5 Morphology, suggesting that D3 Continuous iDAScores are a strong predictor of D3 and D5 morphology outcomes. The odds of blastocyst formation increased 2.03 fold for each one-unit increase in Day 3 IDA score. A significant positive correlation was demonstrated between D5 Continuous iDAScores and day 6 (D6) Morphology (p < 0.0001, AUC 0.8413). Focusing specifically on morula and early blastocyst D5 groupings, the AUCs for D5 Continuous iDAScores in Morula, and early blastocysts for different expansions (1 to 3) were 0.8118, 0.8183, 0.7997 and 0.6753 respectively. This suggests that D5 Continuous iDAScore is a strong predictor of the likelihood of conversion to good quality utilisable D6 blastocysts, from slower developing embryos on D5. Limitations, reasons for caution The study was carried out in a single centre; therefore, the results may not be universally applicable. Application of the model is only available using the Embryoscope Viewer platform. Only the finalised iDAScore values are stored on the system, therefore contemporaneous documentation of continuous iDAScore values is essential. Wider implications of the findings iDAScore may improve our ability to predict conversion to utilisable blastocysts, especially when used in combination with morphology and other parameters. It could also help to provide personalised prediction for individual patients undergoing treatment. Trial registration number No