A recent study reveals that a new artificial intelligence (AI) system can accurately evaluate the chromosomal status of in vitro-fertilized (IVF) embryos using only time-lapse video footage and the age of the mother.
A fresh AI-based system has been developed to effectively determine the chromosomal status of in vitro-fertilized (IVF) embryos by utilizing only time-lapse videos of the embryos along with the mother’s age, according to research from Weill Cornell Medicine.
The system, known as “BELA,” was detailed in a paper released on Sept. 5 in Nature Communications. It represents the latest platform developed by the team to evaluate whether an embryo has a standard (euploid) or irregular (aneuploid) number of chromosomes, which is crucial for the success of IVF. Unlike earlier AI methods, BELA does not rely on personal judgments from embryologists, offering a more objective and universally applicable assessment. If clinical trials confirm its effectiveness, BELA could be widely adopted in embryology clinics to enhance the efficiency of IVF treatments.
“This method is completely automated and provides a more objective solution compared to previous systems. It harnesses larger sets of image data to yield improved predictive capability,” stated Dr. Iman Hajirasouliha, the study’s senior author and associate professor of physiology and biophysics at Weill Cornell Medicine’s Englander Institute for Precision Medicine.
The primary author of the study was Suraj Rajendran, a Ph.D. candidate in Dr. Hajirasouliha’s lab. The work examining the embryos was led by Dr. Nikica Zaninovic, an associate professor of embryology and director of the Embryology Laboratory at the Ronald O. Perelman and Claudia Cohen Center for Reproductive Medicine (CRM) at Weill Cornell Medicine, including NewYork-Presbyterian/Weill Cornell Medical Center. The study also included contributions from Dr. Zev Rosenwaks, the CRM’s director and chief physician, and a distinguished professor of reproductive medicine at Weill Cornell Medicine.
Typically, embryologists evaluate the quality of an IVF embryo by examining it under a microscope. If the embryo appears normal but there are concerns about potential issues, particularly in cases involving older mothers, they might conduct a more invasive test on its chromosomal status. The most reliable test is a biopsy-like procedure known as preimplantation genetic testing for aneuploidy (PGT-A). Recently, embryologists have partnered with AI experts to help automate part of this process and enhance results. In a 2022 study, Dr. Hajirasouliha and colleagues introduced STORK-A, an AI system that predicts the embryo’s ploidy status using a single microscopic image, along with maternal age and embryologist evaluations, achieving about 70 percent accuracy.
In contrast, BELA was developed to independently provide accurate predictions of ploidy without relying on embryologist assessments. The core of this system is a machine-learning model that reviews nine time-lapse video images of an embryo captured about five days post-fertilization, generating a quality score for the embryo. This score, alongside maternal age, is then utilized to predict whether the embryo is euploid or aneuploid.
For training the model, researchers utilized a de-identified dataset from Weill Cornell Medicine CRM that included image sequences of nearly 2,000 embryos and their PGT-A-confirmed ploidy status. The model was subsequently tested on new datasets from Weill Cornell Medicine CRM as well as from large IVF clinics in Florida and Spain. The results indicated that the model predicted ploidy status with improved accuracy compared to previous attempts and performed well across both internal and external datasets.
The researchers plan to further evaluate BELA’s predictive capabilities in a randomized, controlled clinical trial that is currently in the planning stages.
“BELA and similar AI models could make IVF services more accessible in areas lacking advanced IVF technologies and PGT testing, promoting equity in IVF care globally,” stated Dr. Zaninovic.
Moreover, since BELA processes extensive image data from each embryo, the researchers suggest that it could have applications beyond predicting ploidy.
“We hope that this model can also be employed for general assessments of embryo quality, prediction of developmental stages, and other applications that an embryology clinic might adapt to meet its specific needs,” Rajendran remarked.
Several physicians and scientists from Weill Cornell Medicine collaborate with external entities to encourage scientific advancements and offer expert insights. The institution publicly discloses these relationships to maintain transparency. For further details, please refer to profiles for Dr. Iman Hajirasouliha and Dr. Nikica Zaninovic.
This research was partially supported by the National Institute of General Medical Sciences, a part of the National Institutes of Health, under grant number R35GM138152.
