June is World Infertility Awareness Month, and a new group of technology developers in assisted reproduction technology is taking the “world” part seriously. They’re in the vanguard of a new wave of cloud-based artificial intelligence platforms intended to help clinicians everywhere improve their patients’ chances of starting a family.

The pioneers of this technology see abundant promise in improving operational efficiency, increasing the number of successful pregnancies in fewer cycles – and perhaps lowering costs – while simultaneously navigating disparate and fluid technology and healthcare privacy policies between jurisdictions. One thing, however, is clear: A community of reproductive medicine clinicians is more than willing to give this collaborative in vitro fertilization model a real shot at success.

“Those clinicians and scientists care about the science and want to be part of that,” Michelle Perugini, CEO of San Francisco, CA-based AI IVF developer Presagen said. “They want to be part of the development process. It gives them insight into how AI is built. So I sort of see our global clinical network almost as an advisory board of thousands of people all throughout the industry who have those insights, and we have to build AI in a way that’s useful for them.”

Much of the latest buzz about AI in IVF is about foundational technologies that are interpretable and transparent. But to developers like Perugini, and veteran clinician-scientists like Matthew “Tex” VerMilyea, MD, that transparency is only half the battle: the real advantage to IVF AI will come from cloud-based platforms that supply immense heterogeneous datasets paired with low operational costs.

VerMilyea, vice president of scientific advancement at Los Angeles, CA-based Ovation Fertility, has extensive experience with previous generations of IVF AI technology that utilized on-site time-lapse image analysis to predict which embryos would be most successful. While that technology did cross an acceptance hurdle among physician users, VerMilyea said it also came, literally, with a high cost.

“The acceptance of having a different device in the laboratory that helps us better understand embryo development was a big win,” he said. “The difficulty I think we come to with all that time lapse equipment is just the expense invested. That came as a real hindrance. How do you spend that kind of cap ex investment in order to bring some of that technology into the laboratory? And then it brings up the question of do you charge the patient? Do you charge a premium? Add-ons in IVF are always a hot topic.”

Enter Presagen and startups like San Francisco-based Alife Health. Both are developing AI IVF platforms that are hosted on Amazon Web Services and are banking on widespread clinician acceptance in which ease of use and versatile cross-platform compatibility will create compelling arguments for adoption (Alife CEO Paxton Maeder-York said VerMilyea, who is widely published in the field, holds a small equity stake in the company; Ovation was also an early investor in Presagen).

“A third party like Alife is really the entity that is best suited,” Maeder-York said. “A lot of these internal clinic groups have some AI here and there, but unfortunately it’s limited by the data they see within their clinical practices.”

“I always fall back on diversification of the data,” VerMilyea said. “There are multiple AI systems in play based only on internal data and it’s a flop. They are not generalizeable, they can not be used elsewhere.”

Presagen’s Perugini said that not only is it vital to construct a generalizeable AI system, but that system must also be built from the outside in. A system trained on a narrow dataset of limited demographics will likely be too cumbersome to expand easily.

“AI companies that start in one region might then say ‘OK, let’s expand to Europe,’ and they’ll try the algorithm in Europe and find it doesn’t work because it hasn’t been exposed to the right type of data from that jurisdiction. And then you have the process of re-training and rebuilding and re-validating and testing.

“We went right from day one with a global dataset that had different camera systems, different quality of images. We’ve got a very deep technology basis in our platform around managing data quality, data access, connecting data silos, and allowing us to access that global dataset form day one.”

The economic argument of IVF AI

The AI developers are placing their stakes on the improvements aggregating large amounts of data promise, in a field in which success rates are low: in a U.S. Society for Assisted Reproductive Technology report cited by Penn Medicine, the chance of having a full term, normal birth weight and singleton live birth per ART cycle using fresh embryos from non-donor eggs is 21.3 percent for women younger than 35. Those odds get even lower as maternal age climbs. The average cost per cycle, according to numerous sources, is between $12,000 and $17,000 (before medication expenses), and is dependent upon insurance coverage, patient characteristics and the treatment center.

There is already ample evidence, IVF AI advocates say, that the advanced technologies now emerging can substantially improve results over manual assessments of embryos.

“I can tell you our technology improves the accuracy over manual visual assessment of embryo quality by 25%,” Perugini said. “It reduces the time to pregnancy for patients by up to 20%, which means that patients, though they may pay a tiny incremental cost on their IVF treatment cycle, are getting the best possible chance of succeeding in the first cycle.”

“All our AI tools are built to optimize and improve the effectiveness and efficiency of an IVF cycle to the patient,” Maeder-York said. “That may slightly increase the cost to the patient on a per cycle basis, but decrease the overall cost of pregnancy, as well as reduce the heartbreak of a failed IVF cycle and having to go through the whole process again.”

Presagen’s technologies, branded LifeWhisperer, are already approved and in clinical use in 42 countries worldwide. Perugini expects FDA approval in the U.S. within the next 12 months. Alife Health’s technologies are currently in pilots and clinical trials. Maeder-York said some of its portfolio, such as its StimAssist, an adjunct decision support tool, will not need to go through a 510(k) process. Its image analysis software system, called Embryo Predict, is now being tested in trials in preparation for a 510(k) clearance.

And, though Presagen may appear to have first-mover advantage in deploying approved technology, there is no shortage of funding coming to competitors. Alife picked up $22 million in series A funding in March; Tel Aviv, Israel-based Fairtility obtained $15 million in series A funding late in May.

Policy uncertainties of IVF AI

One possible wild card IVF AI developers might have to address has nothing to do with technical performance or clinician acceptance on its merits, but rather reproductive health policy. In the United States, where the possible reversal of Roe v Wade was signaled by a leaked Supreme Court draft decision in early May, advocates for reproductive autonomy for women have expressed concern that IVF procedures might also be endangered. Such uncertainty has not unduly affected either Maeder-York or Perugini.

“Limited access to IVF in some states will drive up medical tourism, increasing costs associated with IVF, and put additional patient load on clinics in progressive states,” Maeder-York said. “This will only increase the need for tools like Alife’s that improve efficiency and efficacy of treatment.”

“There are a lot of layers, political and regulatory barriers in the U.S., but they’re everywhere,” Perugini said. “They’re in other countries as well, they’re just of a different nature. I’m not worried about it from a business perspective. I think some of that discussion will be impacting patients and that’s a shame. But from a business perspective, we are working as a technology provider to the IVF sector, and that won’t change.”