JBRA Assist. Reprod. 2025;29(Suppl 1):30-30
POSTER PRESENTATION
doi: 10.5935/1518-0557.20250084
P-016. Correlation Between the Euploidy Classification by an Artificial Intelligence Software (EMA) and Genetic Results of Blastocysts Assessed by PGT-A
1Reproferty - Center for Human Reproduction. São José dos Campos - SP, Brazil
Objective: To evaluate the correlation between the probability of euploidy classification provided by the artificial intelligence software EMA and the genetic results obtained from embryo biopsy.
Methods: This retrospective study analyzed 294 blastocysts evaluated by the EMA software prior to biopsy for genetic analysis through Preimplantation Genetic Testing for Aneuploidies (PGT-A). The EMA software automatically assigns each embryo a probability score for euploidy, categorized into four groups: A (highest probability of euploidy), B, C, and D (lowest probability). Embryos were grouped according to this classification, and PGT-A results were used to determine whether embryos classified as A indeed exhibited a higher euploidy rate compared to the other groups.
Results: Table 1 shows the total number of embryos classified by the EMA software into groups A through D, along with the proportion of euploid and aneuploid embryos in each category. Embryos classified as “A” showed the highest euploidy rate (76.9%), while those classified as “D” showed the lowest (11.6%), demonstrating a decreasing trend in euploidy rate as the EMA classification worsened. A statistical analysis was performed to assess the linear trend between EMA classification (A to D) and euploidy rates. The Cochran-Armitage trend test showed a statistically significant decrease in euploidy rates as classification worsened (Z = -4.56; p<0.00001).These findings are in line with recent studies that explore the application of artificial intelligence and non-invasive methods to predict the genetic potential of embryos. Cai et al. (2022) demonstrated that the majority of transferred embryos resulted in healthy live births using EMA-based morphology assessment combined with noninvasive PGT-A from cell-free DNA in blastocyst fluid. Similarly, Souza (2022) developed an artificial neural network capable of predicting embryo ploidy based on morphological, morphokinetic, and patient-related variables, reinforcing AI as a valuable tool in embryo selection.
Conclusion: The data suggest a positive correlation between the euploidy probability classification assigned by the EMA software and the genetic results obtained through embryo biopsy. Embryos classified as “A” had the highest euploidy rate, while those classified as “D” had the lowest, indicating that artificial intelligence may be a promising non-invasive tool to support embryo selection.
Table 1.
References