Search for a command to run...
Abstract Study question Can an embryo score assigned by an AI model predict the likelihood of an embryo leading to clinical pregnancy and live birth with varying prognosis? Summary answer An AI embryo quality score can predict whether an embryo is able to lead to a clinical pregnancy and live birth. What is known already Single embryo transfers are the standard practice in most IVF procedures and have been accepted worldwide. However, patients with low quality embryos or multiple replacement failures might benefit from double embryo transfers. In many instances, patients with multiple failures have embryos that are of lesser quality. Ability to select which embryo to transfer in these patients may be difficult due to the inherent subjectivity already involved in embryo selection. The potential of an AI (Chloe EQ) to automatically register morphokinetic events and other developmental parameters, may assist embryologists in the selection of embryos in good and poor prognosis patients’ population. Study design, size, duration 360 embryos from a total of 287 patients were transferred in a single private clinic. 214 patients underwent single embryos transfer (SET), and 73 patients went through double embryo transfer (DET). Only patients with known pregnancy and live birth outcomes were included. In cases of DETs, only patients with twin live births or failure to implant were included as to avoid the uncertainty of not knowing which replaced embryo resulted in a positive outcome. Participants/materials, setting, methods All transferred embryos received an AI EQ Score between 0 and 10 from the AI model. EQ Score prediction of Clinical Pregnancy and Live Birth was assessed with Binary Logistic Regression. Multiple 2-sample t-tests were performed to assess the differences between the mean EQ Score for embryos that achieved a clinical pregnancy/live birth as compared to those that did not, as well as to assess the mean developmental time for the different morphokinetic parameters. Main results and the role of chance The AI scoring model was determined to be predictive of clinical pregnancy and live birth with an AUC of 0.66 for both outcomes. Embryos that achieved a clinical pregnancy had significantly higher (p < 0.001) mean EQ score than those that failed to develop a clinical pregnancy. Mean EQ score was 8.09 for embryos that achieved a clinical pregnancy as compared to 6.39 for those that did not. Similarly, embryos that resulted in a live birth had a significantly higher (p < 0.001) mean EQ scores than those that failed to end in a live birth. Mean EQ score was 8.11 for embryos that achieved a live birth versus 6.45 for those that did not. When comparing morphokinetics of embryos that achieved a live birth with embryos that did not, the following events showed significant differences: t6 (p = 0.022), t7 (p = 0.011), t8 (p = 0.013), tM (p = 0.005), tSB (p = 0.000), tB (p = 0.000), tEB (p = 0.002). Specifically, events related to blastulation were the most significantly related to live birth: tSB (96.82 ± 7.13 vs 101.20 ± 9.81), tB (103.92 ± 7.34 vs 108.7 ± 10.9) and tEB (109.39 ± 8.51 vs 113.6 ± 0.87). Other morphokinetic events (tPNa, tPNf, t2, t3, t4, t5, t9) were not found to be significant. Limitations, reasons for caution This retrospective data has been collected from only a single centre. Patients with double embryo transfers were only included in this study if both embryos failed to implant or subsequently both embryos resulted in a live birth. Wider implications of the findings AI could be utilized to predict and select embryos for patients with good prognosis and for those patients that have failed more than one replacement or have poor quality embryos. AI has the potential to reduce the inherent subjectivity in assessing patients with poor prognoses. Trial registration number No