Artificial Intelligence Tool May Improve IVF Embryo Selection

Researchers at Weill Cornell Medicine have developed an artificial intelligence (AI) tool that can non-invasively identify in vitro fertilized embryos with aneuploidy, or an abnormal number of chromosomes, with about 70 percent accuracy.

The press release states that aneuploidy is a major factor contributing to whether an embryo derived from in vitro fertilization (IVF) will fail to implant or result in a healthy pregnancy. Current methods for detecting the condition involve genetic testing and biopsy-like sampling of cells from an embryo, which is both invasive and costly for those undergoing IVF.

To address these issues, the researchers sought to develop an AI algorithm that could help predict aneuploidy without the drawbacks of traditional testing methods. Their tool, known as STORK-A, achieves this by analyzing microscope images of the embryo and information about maternal age and the IVF clinic’s scoring of the embryo’s appearance, according to a study published this week in The Lancet Digital Health.

“Our hope is that we’ll ultimately be able to predict aneuploidy in a completely non-invasive way, using artificial intelligence and computer vision techniques,” said study senior author Iman Hajirasouliha, PhD, associate professor of computational genomics and physiology and biophysics at Weill Cornell Medicine and a member of the Englander Institute for Precision Medicine, in the press release.

The new algorithm builds on the researchers' existing work in this area. In 2019, the research team developed STORK, an AI approach that could assess embryo quality on par with IVF clinic staff.

STORK-A expands on this, using microscope images of embryos taken five days past fertilization, clinic staff’s scoring of embryo quality, maternal age, and other information that is usually gathered as part of the IVF process to enhance the use of preimplantation genetic testing for aneuploidy (PGT-A), a biopsy method used to obtain information about an embryo’s chromosomes.

The press release indicates that incorporating these additional data and AI may make STORK-A a potential method to decide which embryos should have PGT-A testing or whether to replace PGT-A altogether.

The algorithm was trained on a dataset of 10,378 blastocysts for which ploidy status was already known. It was then evaluated on its ability to accurately identify aneuploid versus normal-chromosome 'euploid' embryos. For this task, STORK-A was found to be 69.3 percent accurate.

The researchers also found that the tool was 77.6 percent accurate in predicting aneuploidy involving more than one chromosome — complex aneuploidy — versus euploidy.

The study provides a proof of concept for the experimental approach, the press release states.

“This is another great example of how AI can potentially transform medicine. The algorithm turns tens of thousands of embryo images into AI models that may ultimately be used to help improve IVF efficacy and further democratize access by reducing costs,” said study co-author Olivier Elemento, PhD, director of the Englander Institute for Precision Medicine and a professor of physiology and biophysics and computational genomics in computational biomedicine at Weill Cornell Medicine, in the press release.

Moving forward, the researchers plan to build on this research by developing algorithms trained on videos of embryo development, which could help identify aneuploidy with even higher accuracy by leveraging both spatial and temporal information.

“We believe that ultimately by using this technology, we can reduce the number of embryos to be biopsied, reduce the costs, and provide a very good tool for consultation with the patient when they need to make [a] decision whether to do PGT-A or not,” said Nikica Zaninovic, PhD, associate professor of embryology in clinical obstetrics and gynecology and director of the Embryology Laboratory at the Ronald O. Perelman and Claudia Cohen Center for Reproductive Medicine at Weill Cornell Medicine and NewYork-Presbyterian/Weill Cornell Medical Center, in the press release.

This is the latest effort to leverage AI to improve IVF. In 2020, researchers showed that a deep-learning system was capable of choosing the most high-quality embryos for IVF with 90 percent accuracy in a dataset of images captured from 742 embryos taken 113 hours post-insemination.

The tool’s performance was also compared to that of 15 embryologists from five fertility centers across the US. During the assessment, the model performed with an accuracy of approximately 75 percent, while the embryologists performed with an average accuracy of 67 percent.