16 May 2019 New AI-Directed Technology Promises to Improve Embryo Selection

By Rich Vaughn and Dr. Said Daneshmand

One of the most critical functions in the process of in vitro fertilization (IVF) is the selection and ranking of the strongest, most viable embryos for implantation. A new technique utilizing a Google deep learning algorithm, a type of artificial intelligence, promises to remove the subjectivity from embryo selection and may even improve the odds of a successful pregnancy and birth.

As a recent Wired report explained, much of the work of an IVF clinic is performed behind closed doors, where embryologists collect and fertilize eggs and cultivate embryos. One of the most taxing, time-consuming parts of their job—grading the embryos with a “quality score” based on the structure and condition of the cells. The task has also been an entirely subjective one, with embryologists reaching the same conclusion about an embryo’s condition or quality rank only about 25 percent of the time. In comparison, the new neural network, dubbed STORK, matched the majority opinion more than 95 percent of the time. Clinicians hope the new technology will improve the odds of successful outcomes for in vitro fertilization, currently estimated to be about 45 percent in the United States.

According to a report by Science Daily:

For the study, published April 4 in NPJ Digital Medicine, investigators used 12,000 photos of human embryos taken precisely 110 hours after fertilization to train an artificial intelligence algorithm to discriminate between poor and good embryo quality. To arrive at this designation, each embryo was first assigned a grade by embryologists that considered various aspects of the embryo's appearance. The investigators then performed a statistical analysis to correlate the embryo grade with the probability of going on to a successful pregnancy outcome. Embryos were considered good quality if the chances were greater than 58 percent and poor quality if the chances were below 35 percent. After training and validation, the algorithm, dubbed Stork, was able to classify the quality of a new set of images with 97 percent accuracy.

The new technology, hosted on a secure website by Weill Cornell Medicine, won’t be approved “until it can pass rigorous testing that follows implanted embryos over time, to see how well the algorithm fares in real life,” according to Wired.

One big question: Will new AI-guided embryo selection actually improve outcomes, i.e., result in pregnancy and live birth, or will it only speed up the implantation process? According to Science Daily report, “previous studies have suggested that only 80 percent of the pregnancy success rate relies on the embryo quality. Maternal age, in particular, is associated with a decreasing rate of successful embryo implantation in the uterus.”

Embryo selection is only one of the ways in which AI is expected to improve IVF procedure and outcomes.

There is clear traction and momentum in the application of Artificial Intelligence and machine learning in every aspect of IVF:  Assessment of sperm quality, optimization of IVF stimulation protocols, determination of blastocyst quality, prediction of pregnancy and live birth rates based on blastocyst assessment, to name a few.  The momentum in this burgeoning field is manifested by the number of scientific papers devoted to this area of science in the 2108 Annual Congress of the American Society of Medicine, as compared to previous years.  Sixteen different AI and ML approaches were reported in 2018, as compared to one in 2017. 

Most U.S. consumers today are using machines or processes that utilize AI in their daily lives. For example, digital assistants such as Siri and Alexa are powered by AI technology: They recognize what they hear or read and respond and make decisions based on an immense amount of data.

 What is the overarching goal of the novel application of AI in IVF?  Simply stated, to improve the efficiency and success of the IVF cycle, to increase live birth rates and decrease miscarriage rates and to apply scientifically based protocols and statistics to build machines that can process and analyze data and learn on their own, without constant human supervision. Machine-learning algorithms use statistics to find patterns in massive amounts of data. These machines can use these patterns and data to make predictions regarding IVF outcomes such as live birth. 

Physicians at San Diego Fertility Center conducted research on embryo selection parameters and predictors of endometrial receptivity over the course of the past 20 years, analyzing the cellular structure of the embryo and measuring cell diameters, levels of embryo expansion, and days to blastocyst formation. This research demonstrates certain patterns that could help with the selection of the single best quality embryo that would maximize live birth outcome.  Researchers at San Diego Fertility Center have also analyzed data regarding the hormonal environment and timing to best predict the ideal “window of implantation.”  These are but two aspects of IVF data that can be used as predictors of success. There are a myriad of other factors and data that can be best analyzed by intelligent machines to predict successful IVF outcomes.

AI and Machine Learning technologies seek to transcend the narrow focus on individual IVF variables and can uncover scientific “pearls” hidden in large data.  These pearls can have invaluable benefits in helping doctors determine the best single embryo to transfer and the timing of the transfer, based on pinpointing the “window of implantation.” 

Embryo selection is a natural starting point for the application of AI in the field of IVF.  This is because of the availability of high-quality embryo images available in most IVF Laboratories and the importance of embryo selection to IVF cycle success.  With AI and ML technologies, futile cycles with abnormal embryos can be eliminated, creating the most rapid path to a healthy delivery. 

Many challenges lie ahead as IVF providers and researches begin adopting this revolutionary technology, the most important of which is a universal adoption of electronic medical records in every IVF lab and office. In addition, machines can develop “unintentional bias” by essentially programming themselves, a phenomenon that the engineers who build these systems cannot fully explain. For this reason, human oversight and reanalysis will still be critical in honing the immense power that these machines harness.  AI should not make humans obsolete—not yet anyway!

At the end of the day, the Stork technology that is helping eliminate tedium and improving the quality of embryo selection merely scratches the surface of potential AI-powered enhancements to assisted reproductive technologies—promising a future in which biological parenthood is more reliable, accessible, affordable for more intended parents.

RICHARD VAUGHN:

Attorney Rich Vaughn is founder and principal of International Fertility Law Group, one of the world’s largest and best-known law firms focused exclusively on assisted reproductive technology, or ART, including in vitro fertilization (IVF), surrogacy, sperm donation or egg donation. Rich is co-author of the book “Developing A Successful Assisted Reproduction Technology Law Practice,” American Bar Association Publishing, 2017.

SAID DANESHMAND:

Said Daneshmand, MD, FACOG, is an internationally recognized fertility specialist with extensive experience in providing third-party reproductive services. He is internationally recognized for his expertise in egg donation and surrogacy and is one of the main providers of third-party care in the United States, with patients from all over the world. Dr. Daneshmand is known for developing and implementing cutting-edge technologies, including NGS (next generation sequencing), the latest and most precise technique for assessing embryonic genetic health.