Researchers from Bringham and Women’s Hospital and the Massachusetts General Hospital created an artificial intelligence (AI) system that can help embryologists choose the healthiest and most viable embryos for implantation.
The team used thousands of embryo image examples and deep learning AI algorithms to create a system that could differentiate and identify embryos that will likely be successful. The AI system performed better than 15 experienced embryologists from five fertility centers in the U.S.
AI algorithm could help doctors choose viable embryos to be implanted.
Currently, available embryo picking tools are expensive and limited. So often embryologists are limited to relying on their observational skills and expertise.
The new assistive tool evaluates images captures with microscopes commonly found in fertility centers. The system was trained with images of embryos that were captured 113 hours after insemination. With 742 embryos, the AI system was 90% accurate when choosing the highest quality embryos.
The team assessed the system’s ability to find high-quality embryos with a normal number of human chromosomes. Its performance was compared to the performance of embryologists who analyzed the same dataset.
The system performed with an accuracy of about 75% and embryologists performed with an average accuracy of about 67%.
In the current stage, the system acts as an assistive tool for embryologists to make judgments during the embryo selection.
A paper on this technology was published in eLife.
