Every year, 600 babies who should have been born, are not
written by Ruti Levi with photos by Eyal Tweig
For couples having children through IVF, fertility doctors guiding the process are the ones who receive the credit for success. But within the lab, hidden from the public eye, far lesser-known scientists are performing the complex process of in-vitro fertilization, or IVF. These are the embryologists, specialists doing the highly specialized work of locating and cleaning the eggs and sperm, injecting the sperm into the egg to fertilize it, monitoring the embryos in incubators, following their development, and eventually determining their quality for transfer.
Although the fertilization process represents the pinnacle of progress in modern medicine, those familiar with the process know that the indicators used by embryologists to determine an embryo’s quality are based on manual tests performed under a microscope. As a result, the degree of success in selecting an embryo to implant into the womb or freezing for future implantation is based on the embryologists’ experience and judgment.
Gilboa found that embryologists tend to select for implantation the first embryo they grade as good, which is not necessarily the best. This lowers the chances of becoming pregnant by 30%. “Human bias causes 600 babies a year who should have been born, not to be born,” she claims.
A computer doesn’t have to concentrate, and without any effort is able to emulate the best embryologists.
What the human eye cannot see
Today the software developed by AIVF identifies biological processes which the human eye can’t discern, such as that mitochondrial ‘energy action’ which is directly linked to the embryo’s collapse or chances of implantation.
The algorithm running the AIVF software was trained on an immense number of images and videos of embryos tagged with successful conception and leading to a normal pregnancy. Currently the program can independently identify the embryos with the highest chances of developing into a normal pregnancy.
Gilboa established the startup together with Professor Seidman, a leading expert in the field of in-vitro fertilization. Seidman, who runs a bustling clinic in Tel Aviv’s HaBarzel Street, admits that “the best fertility doctors are only as good as the best embryologists working with them. If the embryologist isn’t top class, the doctor will be known for a medium level of success.” He met Gilboa when they worked together at Assuta Ramat HaHayal.
Nonetheless, Seidman is the first to admit to the tremendous burnout and pressure under which the team operates and sees programs of the kind he’s working on as a possible solution. “A dedicated fertility doctor doesn’t have a life. Work is 7 days a week, and for 5 of them you keep going until midnight. We’re incredibly motivated. But AI would save clinician time. The more we can access data and make decisions this way, the greater effectiveness we’ll reach.”
Transparency is a significant change in a field where the concept of poor chances of success barely exists, and certainly not in a country like Israel where birth is considered so precious that the country funds treatments up to the age of 45, though with a limit of up to two children. Media, and to a great degree the physicians, shape the illusion that technology has replaced nature, and public funding signals women that everything’s possible. That is until they discover through personal and painful experience, that it is not.
In the USA, the market AIVF is eyeing has no public funding, and costs are considerable. One cycle, in which eggs are retrieved, fertilized, and attempts are made to implant the embryo, is estimated at some $20,000. This motivates clinics to introduce technology that assists in improving the chances of pregnancy. “It’s a matter of trial and error until we succeed. Some people stop when their money runs out. Others run up costs in the tens of thousands of dollars just to bring a child into the world,” Gilboa adds.
Unlike most deep learning algorithms, the AIVF platform identifies the patterns driving its actions and can explain its decisions. It will mark, for example, that it has given a low score to a specific embryo due to disruptions or disorders found during its development, due to a different behavioral pattern at the cell division stage, or due to abnormal morphological appearance such as fractures.