Evaluating the concordance between AI-based and conventional embryo selection: implications for clinical...

Area of Science:

• Reproductive medicine
• Artificial intelligence in healthcare
• Embryology

Background:

• Standardized embryo scoring is crucial for improving in vitro fertilization (IVF) success rates.
• Current methods rely on subjective interpretation by embryologists, leading to variability.
• Identifying embryos with the highest potential for pregnancy and live birth remains a challenge.

Purpose of the Study:

• To evaluate the accuracy of an artificial intelligence (AI) embryo selection algorithm (ESA) in identifying viable embryos.
• To compare the ESA's performance against human embryologists in embryo scoring.
• To determine if AI can standardize embryo assessment and improve pregnancy prediction.

Main Methods:

• A multicentre, retrospective analysis involving six centers across five countries.
• An ESA and 20 embryologists independently assessed 1681 embryo pairs.
• Accuracy was calculated for the ESA and each embryologist, with differences analyzed using McNemar's test.

Main Results:

• The ESA achieved an accuracy of 70.1% in selecting the best embryo.
• Individual embryologists' accuracy ranged from 64.2% to 68.9%, with a mean of 67.7%.
• The ESA showed a statistically significant advantage over 14 out of 20 embryologists and the mean embryologist accuracy.

Conclusions:

• The ESA demonstrated superior accuracy compared to the majority of individual embryologists and the average performance.
• AI has the potential to serve as a standardized tool to augment expert embryologist judgment.
• Further prospective, multicenter trials are necessary to confirm the ESA's effectiveness and generalizability in clinical practice.