Artificial Intelligence Assisted Blastocyst Assessment Predicts Implantation Success and Demonstrates Concordance with PGT-A Results in Frozen Embryo Transfer Cycles
AI-assisted blastocyst assessment may improve embryo selection in frozen embryo transfer cycles, but it should not replace preimplantation genetic testing for genetic screening.
Where it sits
this study against the rest of the hcg (human chorionic gonadotropin) corpusSummary and findings
This study evaluated the predictive value of an AI-assisted blastocyst assessment system for implantation outcomes in frozen embryo transfer cycles. A total of 340 FET cycles were analyzed, with 69.7% achieving successful implantation. The study compared AI-assisted scoring with conventional grading and preimplantation genetic testing.
Abstract
<title>Abstract</title> <p>Embryo selection remains one of the most critical challenges in assisted reproductive technology (ART), as morphologically normal embryos frequently fail to implant despite technically successful transfer. While conventional morphological grading using the Gardner and Schoolcraft system has been the established method for embryo assessment, it is inherently subjective and highly dependent on embryologist expertise, leading to significant variability in evaluation between different laboratories and practitioners. To address these limitations, this study evaluated the predictive value and clinical utility of Life Whisperer™, an artificial intelligence (AI)-assisted blastocyst assessment system, for predicting implantation outcomes in frozen embryo transfer (FET) cycles and compared its performance with both conventional Gardner morphological grading and preimplantation genetic testing for aneuploidy (PGT-A). A retrospective cohort study was conducted analyzing 340 FET cycles from Indira IVF Fertility Centre, Delhi, India, between January 2022 and December 2024. Day 5 blastocysts were assessed using both conventional Gardner scoring and Life Whisperer™ AI-based viability scoring, with implantation success determined by serum β-hCG positivity. Predictive performance was evaluated using receiver operating characteristic (ROC) analysis, and concordance between AI-derived genetic prediction and PGT-A results was assessed using Cohen's kappa coefficient. Results demonstrated that 237 of 340 embryos (69.7%) achieved successful implantation, with higher Gardner morphology grades (4AA and 3AB) associated with superior implantation rates of 80.7% and 81.0%, respectively. AI-assisted viability scores demonstrated strong and independent predictive performance, with implantation rates increasing from 53.5% (low viability) to 74.2% (high viability), and ROC analysis revealed excellent discriminatory ability with an area under the curve of 0.91. However, AI-derived genetic prediction showed only fair concordance with PGT-A (Cohen's κ = 0.257). These findings demonstrate that AI-assisted blastocyst assessment offers a promising, non-invasive, and objective tool to complement conventional morphological evaluation for optimizing embryo selection and improving implantation outcomes in FET cycles, though it cannot currently substitute for PGT-A in genetic screening.</p>
Background
This paper addresses the challenge of embryo selection in assisted reproductive technology, where conventional methods often yield variable results. Prior research has indicated that morphological grading can be subjective and dependent on the expertise of embryologists. The introduction of AI in embryo assessment aims to provide a more objective and reliable method for predicting implantation success, which is crucial for improving outcomes in frozen embryo transfer cycles.
Methods
A retrospective cohort study was conducted analyzing 340 frozen embryo transfer cycles from Indira IVF Fertility Centre, Delhi, India, between January 2022 and December 2024. Day 5 blastocysts were assessed using both conventional Gardner scoring and Life Whisperer™ AI-based viability scoring. The primary outcome measure was implantation success, determined by serum β-hCG positivity.
Results
The primary endpoint showed that 69.7% of embryos achieved successful implantation, n=340. Higher Gardner morphology grades (4AA and 3AB) were associated with implantation rates of 80.7% and 81.0%, respectively. AI-assisted viability scores indicated an increase in implantation rates from 53.5% for low viability to 74.2% for high viability. ROC analysis revealed an area under the curve of 0.91, indicating excellent discriminatory ability.
Interpretation
The findings suggest that AI-assisted blastocyst assessment can enhance embryo selection by providing a more objective evaluation compared to conventional methods. However, while the statistical significance of the results is notable, the clinical significance of the AI-derived genetic predictions is limited, as indicated by the fair concordance with PGT-A. The study's retrospective design and single-center nature may also confound the generalizability of the results, suggesting that further research is needed to validate these findings across diverse populations.
Key findings
- 69.7% of embryos achieved successful implantation, n=340.
- Implantation rates for Gardner grades 4AA and 3AB were 80.7% and 81.0%, respectively.
- AI-assisted viability scores increased implantation rates from 53.5% (low viability) to 74.2% (high viability).
- ROC analysis showed an area under the curve of 0.91 for AI-assisted scoring.
- Cohen's κ for AI-derived genetic prediction and PGT-A was 0.257.
Limitations
- retrospective study design
- single-center analysis
- fair concordance with PGT-A (Cohen's κ=0.257)
- subjective nature of conventional grading
- potential variability in embryologist expertise