Interpretable artificial intelligence-assisted embryo selection improved single-blastocyst transfer outcomes: a...

Area of Science:

• Reproductive Medicine
• Artificial Intelligence in Healthcare
• Embryology

Background:

• Assisted reproductive technology (ART) relies on accurate embryo selection for successful implantation.
• Traditional embryo assessment methods, like the Gardner grading system, can be subjective.
• Interpretable artificial intelligence (AI) offers a potential tool to enhance embryo selection objectivity.

Purpose of the Study:

• To evaluate the pregnancy and neonatal outcomes of an interpretable AI model for embryo selection.
• To compare the effectiveness of AI-assisted embryo selection with traditional manual methods in a prospective clinical trial.

Main Methods:

• A single-center prospective cohort study involving 250 patients undergoing fresh single-blastocyst transfer.
• Embryo selection was performed using an interpretable AI system's recommendations (AI-assisted group) or the Gardner grading system (manual group).
• Pregnancy outcomes (implantation, miscarriage, live birth, ectopic pregnancy rates) and neonatal outcomes were assessed.

Main Results:

• The AI-assisted group showed a significantly higher implantation rate (80.87%) compared to the manual group (68.15%).
• No significant differences were observed in monozygotic twin, miscarriage, live birth, or ectopic pregnancy rates.
• Neonatal outcomes, including gestational weeks, birth parameters, and malformation rates, were similar between groups.
• AI demonstrated higher consensus with embryologists for good-quality embryos (≥4BB) versus poor-quality embryos (<4BB).

Conclusions:

• The interpretable AI system effectively improved implantation rates in single-blastocyst transfer cycles.
• AI-assisted embryo selection offers a promising alternative to traditional manual evaluation, enhancing clinical outcomes.
• The AI system's reliability in selecting good-quality embryos warrants further investigation and clinical integration.