Never a lost cause: can artificial intelligence help embryologists when only poor-quality embryos are available?

Area of Science:

• Reproductive Medicine
• Artificial Intelligence in Healthcare
• Embryology

Background:

• Embryo assessment is crucial for successful in vitro fertilization (IVF).
• Limited availability of high-quality embryos presents a challenge in IVF.
• Objective embryo assessment tools are needed to support clinical decision-making.

Purpose of the Study:

• To evaluate the efficacy of EMBRYOLY, an AI system, in assisting embryologists with embryo assessment.
• To determine if EMBRYOLY can help identify viable poor-quality embryos for transfer when superior options are absent.
• To assess the impact of EMBRYOLY on clinical pregnancy and live birth rates.

Main Methods:

• Utilized data from 15,767 embryos across 15 clinics (four countries) from 2019-2024.
• Employed EMBRYOLY's transformer-based model to analyze poor-quality embryos and compare with embryologist assessments.
• Evaluated agreement, pregnancy outcomes, and time to pregnancy, comparing EMBRYOLY-recommended poor-quality embryos with non-poor embryos.

Main Results:

• Embryologists encountered only poor-quality embryos in 29% of cases.
• EMBRYOLY's top embryo choice agreed with embryologists in 66% of cases.
• EMBRYOLY's scoring was significantly associated with clinical pregnancies and live births (P < 0.001).
• Adjunct EMBRYOLY use could reduce cycles to pregnancy by 19% and increase first cycle pregnancy rate by 65%.
• EMBRYOLY-recommended poor-quality embryos showed comparable pregnancy potential to higher-quality embryos.

Conclusions:

• Objectively trained AI, like EMBRYOLY, can effectively aid embryologists in selecting viable poor-quality embryos.
• This AI assistance is particularly valuable when high-quality embryos are unavailable for transfer.
• EMBRYOLY demonstrates potential to improve IVF efficiency and success rates.