Upgrading IVF With the Help of Artificial Intelligence

Why you should care

Because having a baby shouldn’t be so hard.

When she started in vitro fertilization, Katie Shepard, a medical device consultant from outside St. Paul, Minnesota, knew it could take more than one round to get pregnant. So, after the grueling regimen of hormone injections, ultrasound exams, egg retrieval and transfer of embryos back into her womb, she stayed optimistic — until her second cycle. Of the 25 eggs harvested over the course of those two IVF treatments, only three developed into embryos. “It felt like someone took me out at the knees with a baseball bat,” Shepard says. Worse, the embryos didn’t take, nor did any from her third cycle. Shepard felt like a failure, and she and her husband had shelled out $30,000 for the three cycles. Since none yielded a pregnancy, they received a refund as part of a money-back program, but that didn’t include the drug costs — $6,000 per cycle.

The Shepards’ ordeal is not uncommon. An IVF cycle, at an average cost of $10,000, leads to a live birth around 20 percent to 35 percent of the time, prompting most women to undergo several cycles before getting pregnant. Embryologists often opt to select multiple embryos for transfer to the womb to improve the odds that at least one will implant and result in a pregnancy. But technology could one day alleviate the emotional and financial strain of IVF. Scientists have begun devising more precise methods of choosing the embryo with the best shot, based on characteristics like rigidity or roughness; some methods even harness artificial intelligence. The goals? Reduce IVF to a single cycle and eliminate the possibility of carrying twins or triplets, which poses risks for mothers and infants alike.

Eventually, they want to build AI into their system, using photographic data to “train” it to assess embryo quality.

After uniting egg and sperm in a petri dish, embryologists monitor the fertilized egg. At Day Five (typically), when the embryo has formed a cluster of around 100 differentiated cells known as a blastocyst, embryologists transfer the fertilized egg to the womb. When choosing embryos for transfer, they look for certain physical characteristics, such as the symmetry and number of cells — a subjective process. The push for more objective methods that could boost the odds of IVF success is largely “a sign of the times,” as more women delay childbirth for professional and other reasons, says Elpida Fragouli of Reprogenetics, part of a global network of labs providing genetic services for infertility clinics and their patients.

Tools designed to improve the accuracy of embryo selection have already cropped up in fertility clinics. Some have begun employing the recently FDA-approved Eeva Test, which uses time-lapse imaging to predict which embryo has the best chance of progressing to a blastocyst, based not only on its appearance but when it hits certain developmental milestones. There’s also preimplantation genetic screening (PGS), which involves extracting a few cells from the outer layer of the blastocyst to ensure it has the correct number of chromosomes. A blastocyst with the incorrect number of chromosomes won’t develop properly, often resulting in miscarriage, stillbirth or a child with health problems.

And yet, as many as a third of chromosomally normal embryos don’t result in a pregnancy. Fragouli and team are developing a method to identify these dud embryos by measuring the levels of DNA in the cell’s energy factories, or mitochondria, after they discovered that blastocysts with mitochondrial DNA above a certain threshold are less likely to implant in the womb. Their method predicted with perfect accuracy which of 249 chromosomally normal embryos previously selected for IVF transfer had failed to implant.

Although removing cells for PGS doesn’t pose serious risks to the embryo, “those cells were never meant to be taken out,” says Hannah Brown of the University of Adelaide. Her team, which embraces a noninvasive technology, keeps IVF embryos intact by photographing them and applying a statistical model that recognizes extremely subtle textures — invisible to the human eye — which they’ve begun linking to metabolism and DNA damage. Now they’re analyzing photos from fertility clinics to identify differences in texture between embryos that led to pregnancies and those that didn’t. Eventually, they want to build AI into their system, using photographic data to “train” it to assess embryo quality.

An embryo’s squishiness might also offer clues about its quality. Stanford researchers recently found that hour-old fertilized mouse eggs that showed resistance to a slight squeeze from a pipette had a higher chance of developing into healthy-looking blastocysts. Drawing on that data, they developed a model that correctly predicted 90 percent of the time which fertilized human eggs would progress to this stage.

At Northwestern University, scientists are studying the sparks of zinc an egg releases when fertilized to make similar predictions — finding that the brighter the spark, the more likely the egg will develop into a blastocyst. As it turns out, human eggs also radiate zinc when injected with a sperm enzyme. The goal is to develop a device that measures fluorescence from tiny probes that emit sparks when zinc latches onto them in a way that can direct embryologists to choose the most promising fertilized eggs to transfer, says Thomas O’Halloran, a chemist at Northwestern University.  

While these selection methods are “much needed … patients should be wary,” says Kate Devine, a fertility specialist and co-director of research at Shady Grove Fertility in Washington, D.C. Scientists still need to confirm they are more likely to lead to successful pregnancies than PGS. Even then, they would complement, not replace, the embryologist’s skilled eye. And, ultimately, they may prove no match for Mother Nature.

Shepard, now 33, conceived via intercourse while gearing up for a fourth round of IVF. After three failed cycles, she’s skeptical of efforts to select a single embryo. “There are just so many variables,” she says. Jennifer Palumbo, a 43-year-old infertility advocate in New York, also has doubts but remains optimistic overall. Palumbo underwent three IVF cycles before getting pregnant; she now has two sons. “Anything that helps people achieve their goals, I’m for it,” she says. 

OZYThe New + the Next

Cutting-edge trends, rising stars and big ideas.