An AI model can now detect chromosomal abnormalities in IVF . embryos
An aneuploidy, or an abnormal number of chromosomes, is the main reason embryos created through in vitro fertilization (IVF) fail to implant or lead to a healthy pregnancy. One of the current methods for diagnosing aneuploidy includes biopsy-like sampling and cytogenetic testing from embryos—a method that increases the cost of in vitro fertilization and embryo invasion. The novel STORK-A algorithm, presented in a paper published December 19, 2022 in The Lancet Digital Health, can help detect aneuploidy without the need for a biopsy. It works by evaluating microscopic images of embryos and integrating information about the mother’s age and the IVF clinic’s classification for embryo appearance.
“We hope that we will eventually be able to predict heterochromia in a completely non-invasive way, using artificial intelligence and computer vision techniques,” said the study’s lead author, Dr. Iman Hajirasouliha, associate professor of computational genomics and physiology and biophysics at Weill Cornell Medicine and a member of the Englander Institute of Precision Medicine.
According to the Centers for Disease Control and Prevention, more than 300,000 cycles of IVF will be performed in the United States by 2020, resulting in about 80,000 live births. IVF specialists are constantly looking for ways to improve success rates and produce more successful pregnancies with fewer embryo transfers—which requires creating better embryos to identify viable embryos. Microscopy is currently being used by fertility clinic staff to screen embryos for large-scale defects that correspond to poor viability. To get information about chromosomes, clinic staff may also use a biopsy called preimplantation genetic testing for aneuploidy (PGT-A), mainly in women over 37 years old.
Harnessing artificial intelligence technology to select IVF embryos
Investigators from the Center for Reproductive Medicine collaborated with colleagues from the Englander Institute to create a computer-based approach to embryo assessment that leverages the use of advanced time-lapse photography. of the Laboratory of Embryology.
In a 2019 study, researchers built an artificial intelligence (AI) program, STORK, that can assess embryo quality just like IVF clinic professionals. For the latest study, they developed STORK-A as a potential alternative to PGT-A or as a more selective means of deciding which embryos should be tested for PGT-A.
The new STORK-A algorithm uses microscopic images of embryos taken 5 days after fertilization, clinic staff’s assessment of embryo quality, maternal age, and other information collected regularly during pregnancy. throughout the IVF process. Because it uses artificial intelligence, the system automatically ‘learns’ how to associate specific aspects of the data, often too subtle for the human eye, with the risk of aberration. The scientists trained STORK-A on a dataset of 10,378 blastocysts, whose spongy state was known. They estimated the algorithm’s accuracy in predicting aneuploid ‘euploid’ embryos against normal chromosomes to be around 70% (69.3%) based on its performance. STORK-A had an accuracy of 77.6% in predicting aneuploidies involving more than one chromosome (complex aneuploidy) compared with allografts.
The work serves as a proof of concept for a currently experimental strategy. Standardizing the use of STORK-A in the clinic will require clinical studies comparing it to PGT-A as well as FDA approval—all of which will take years. However, the new method represents a step forward in the selection of IVF embryos that are less dangerous, subjective, less expensive and more accurate.
“This is another great example of how AI has the potential to transform medicine. The algorithm turns tens of thousands of embryo images into AI models that can ultimately be used to help improve IVF and further democratize access by reducing costs,” said co-author Dr. .Olivier Elemento, director of the Englander Institute of Precision Medicine and professor of physiology, biophysics, and computational genomics in computational biomedicine at Weill Cornell Medicine.
“We believe that ultimately by using this technology we can reduce the number of embryos that need to be biopsied, reduce costs, and provide a very good tool to advise patients when they need to decide whether to should perform PGT-A or not.” Dr. Zaninovic said.
The team now plans to build on this success using algorithms trained on videos of embryonic development. “Using video classification, we can leverage both temporal and spatial information about embryonic development, and hopefully that will enable the detection of developmental trends that help with embryo development,” says Barnes. distinguish aneuploidy and aneuploidy with higher accuracy.
Co-author Dr Zev Rosenwaks, director and physician, said: “This technology is being optimized in the hope that at some point its accuracy will be close to that of genetic testing, here is the gold standard and is more than 90% accurate. —director of the Ronald O. Perelman and Claudia Cohen Center for Reproductive Medicine at NewYork-Presbyterian/Weill Cornell Medical Center and Weill Cornell Medicine, and Revlon Distinguished Professor of Reproductive Medicine in Obstetrics and Gynecology at Weill Cornell Medicine. “But we recognize that this goal is aspirational.”
- A non-invasive artificial intelligence approach to predicting human ploidy blastocysts: a validation study and retrospective model development – (https:www.thelancet.com/journals/landig/article/PIIS2589-7500(22)00213-8/)