O-240 Does embryo categorisation by existing artificial intelligence, morphokinetic, or morphological embryo selection models correlate with blastocyst euploidy rates?

Abstract

Abstract Study question Does embryo categorization by existing artificial intelligence, morphokinetic, or morphological embryo selection models correlate with blastocyst euploidy? Summary answer Our results show that existing blastocyst scoring models correlate with ploidy status. What is known already A previous study suggested that morphokinetic parameters should not be used yet as a surrogate for PGT-A to determine embryo ploidy in vitro. As for predicting pregnancy likelihood, AI models have been proposed for predicting ploidy status. Discrimination performance in terms of AUC has been reported of up to 0.80 when including metadata such as maternal age, and 0.63 when AI predictions were made based on time-lapse images alone. However, as the studies only looked at models trained for ploidy prediction, it is unclear how existing blastocyst scoring systems would perform at discriminating euploids and aneuploids. Study design, size, duration A total of 834 patients, 3,573 blastocysts were retrospectively analyzed. The cycles were stratified into five maternal age groups according to the Society for Assisted Reproductive Technology (SART) age groups (< 35, 35–37, 38–40, 41–42, and >42 years). The quality and scoring of embryos were assessed by iDAScore v1.0 (iDA, Vitrolife, Sweden), KIDScoreTM D5 v3 (KS; Vitrolife), and Gardner grading (GG). Participants/materials, setting, methods Embryos were cultured in the EmbryoScope+ and EmbryoScopeFlex (Vitrolife). iDA was automatically calculated using the iDAScore model running on the EmbryoViewer (Vitrolife). KS was calculated in EmbryoViewer after annotation of the required parameters. ICM and TE were annotated according to the Gardner grading. The degree of expansion in all blastocysts was Grade 4 due to our freezing policy. Furthermore, Gardner's scores were stratified into four grades (A: AA, B: AB BA, C: BB, D: others). Main results and the role of chance At first, the correlation between all assessment methods and euploid rates were analyzed using a trend-test. Euploidy rates significantly correlated with iDA in each age group (p = 0.035 for age <35 years, p < 0.001 for all other age groups) and decreased progressively with increasing maternal age in each score group. Furthermore, euploidy rates were significantly correlated with KS (P < 0.0001), except in the youngest age group (p = 0.07). They decreased progressively with increasing maternal age in each score group. Similarly, euploidy rates were significantly correlated with GG (P < 0.0001), except in the youngest age group (p = 0.06). They decreased progressively with increasing maternal age in each score group. Secondly, the performance of the euploidy prediction for each embryo scoring model was compared using the area under the curve (AUC) of the receiver operating characteristic curve. The AUCs for euploid prediction of iDA, KS, and GG for all ages were 0.666, 0.655, and 0.642, respectively. iDA and KS were significantly higher than Gardner grading (iDA vs. GG: p = 0.004; KS vs. GG: p = 0.02). Additionally, there was no significant difference between iDA and KS. Limitations, reasons for caution It was based on minimal stimulation and the use of natural cycle for IVF treatment, which involved only insemination by ICSI. Furthermore, the blastocyst assessment models that we tested were initially developed for pregnancy prediction and not for the prediction of the chromosomal status of an embryo. Wider implications of the findings Existing blastocyst assessment models that reflect blastocyst viability in view of implantation potential to some extent correspond with ploidy status. Therefore, existing blastocyst scoring models can be used to support the decision for embryo biopsy and/or to reduce the patient’s cost by deciding on which embryo(s) to biopsy. Trial registration number not applicable