Enhancing clinical utility: deep learning-based embryo scoring model for non-invasive aneuploidy prediction

Area of Science:

• Reproductive biology
• Artificial intelligence in medicine
• Genetics

Background:

• Preimplantation genetic testing for aneuploidies (PGT-A) is effective but resource-intensive.
• There is a need for more accessible, non-invasive methods for assessing embryo ploidy.
• AI-driven image analysis offers potential for objective embryo assessment.

Purpose of the Study:

• To evaluate the clinical applicability of an AI-based deep learning algorithm for predicting embryo ploidy.
• To assess the performance of the intelligent data analysis (iDA) Score in identifying euploid embryos.

Main Methods:

• Retrospective analysis of 3448 biopsied blastocysts from 979 Time-lapse (TL)-PGT cycles.
• Application of the iDA Score, a deep learning algorithm, to assign scores (1.0-9.9) to blastocysts.
• Multivariate logistic regression analysis incorporating clinical and embryonic characteristics.

Main Results:

• Significant differences in iDA Scores were observed between blastocysts with varying ploidy.
• Higher iDA Scores strongly correlated with euploidy (p < 0.001).
• The iDA Score alone achieved an Area Under the Curve (AUC) of 0.612 for predicting euploidy, increasing to 0.688 when combined with other factors.

Conclusions:

• The iDA Score demonstrates potential as a non-invasive, cost-effective tool for embryo ploidy assessment.
• This AI-driven approach could benefit patients unable to undergo or afford traditional PGT-A.
• Accuracy of embryo ploidy prediction remains contingent on confirmatory next-generation sequencing (NGS) analysis.