Embryo Ranking Intelligent Classification Algorithm (ERICA): artificial intelligence clinical assistant predicting...

Area of Science:

• Reproductive medicine
• Artificial intelligence in healthcare
• Embryology

Background:

• Assisted reproductive technologies rely on accurate embryo selection for successful implantation.
• Traditional methods for embryo assessment can be subjective and time-consuming.
• Identifying euploid embryos is crucial for improving IVF outcomes.

Purpose of the Study:

• To evaluate the performance of a deep machine learning algorithm (ERICA) in predicting ploidy and implantation potential from static blastocyst images.
• To compare ERICA's predictive accuracy against chance and experienced human embryologists.
• To assess ERICA's ability to rank euploid embryos effectively.

Main Methods:

• A dataset of 1231 blastocyst images with known outcomes was utilized.
• The ERICA algorithm was trained to classify embryos based on ploidy and implantation potential.
• ERICA's predictions were compared against random assignments and senior embryologists' assessments using metrics like accuracy and normalized cumulative gain.

Main Results:

• ERICA achieved an accuracy of 0.70 for predicting euploidy, with a positive predictive value of 0.79.
• The algorithm demonstrated significantly better embryo ranking performance than random selection and two senior embryologists (P < 0.05).
• ERICA successfully ranked an euploid blastocyst first in 78.9% of cases and within the top two in 94.7% of cases, with an average ranking time under 25 seconds.

Conclusions:

• Artificial intelligence, exemplified by ERICA, shows strong potential for image pattern recognition in embryology.
• ERICA can effectively rank embryos based on ploidy and implantation potential using single static images, assisting embryologists in embryo selection.
• This AI tool offers a non-invasive, time-saving alternative to current methods, without requiring time-lapse imaging or biopsy.