Comparison of the efficiency of an AI-assisted model for morphokinetic and morphological human blastocyst selection with the pre-implantation genetic testing

Introduction. Chromosomal pathologies of the embryo are the most significant ones the experts in the assisted reproductive technologies (ART) have been facing with. The existing methods for the embryo evaluation such as the assessment of morphology and NGS-based PGT-A have a number of limitations. An integration of the morphokinetic and genetic methods could increase the ART efficiency and reduce the risk of birth of newborns with the chromosomal abnormalities. Aim: to study the correlation between the results of the embryo selection based on an artificial intelligence (AI)-assisted KIDScore morphokinetic model, conventional blastocyst morphology assessment and embryo aneuploidy levels. Materials and Methods: A post-hoc analysis of the morphological and morphokinetic assessment of the development and PGT-A results involved 258 5—6-day embryos from 107 patient pairs treated for the infertility using ART methods. Results: There were 135 (52.3%) euploid and 123 (47.7%) aneuploid embryos. The probability of aneuploidy identification was higher in elder women (p<0.01). KIDScore median was higher in the euploid vs. aneuploid group (6.60 vs. 5.90, p<0.01). ROC curve for the PGT-A result relied on the KIDScore showed that the model can be used for the euploidy prediction (p<0.001). The selection of an embryo with KIDScore below 7.1 for PGT-A is highly predictive of aneuploidy. Logistic 3D model showed the age and KIDscore as significant predictors for the PGT-A result. The analysis of the dependency of blastocyst morphology on the PGT-A result revealed 1.985-fold lower chances to find a good (BB, AC, CA) blastocyst in the euploid group (p<0.05). 51.2% aneuploid embryos had low expansion of 3 vs. euploid ones (34.1%). In contrast, euploid embryos reached the expansion of 5 or 6 more frequently (17% and 14.1%) vs. aneuploid ones (7.3% and 8.1%, p<0.05). Excellent (AA, AB, BA) blastocysts had higher KIDScores in all age groups vs. good (BB, AC, CA) blastocysts, with an average value of 6.60 vs. 5.40 (p<0.001). The highest KIDScore median (8.0540) was in the embryos which had the expansion of 5 (p<0.001). Conclusions: The integration of the morphological/morphokinetic analysis with the AI-assisted PGT-A technology can improve the pregnancy outcomes and increase healthy birth numbers.