Experts anticipate widespread use of artificial intelligence in IVF

For several years now, artificial intelligence has been gaining traction in the clinical setting in specialties such as nephrology, cardiology, ophthalmology and primary care.

According to Eduardo Hariton, MD, MBA, a reproductive endocrinologist at the Reproductive Science Center of the Bay Area in San Ramon, California, the use of artificial intelligence (AI) in reproductive medicine is currently “still in the dial-up internet phase.” Hariton, who is also the vice president of strategic initiatives at US Fertility, estimated that there are “a small minority of early adopters” in IVF who currently use AI for clinical reasons.

However, it “will likely play an ever-increasing role in predicting clinical outcomes for patients at individual clinics,” Brent Monseur, MD, ScM, a postdoctoral medical fellow studying reproductive endocrinology and infertility at Stanford Medicine, told Healio.

Benefits of AI in IVF

In the realm of reproductive health care, the use of AI has primarily focused on optimizing IVF treatment in the form of predicting oocyte potential, assessing sperm and egg quality and determining the viability of embryos, Renjie Wang and colleagues wrote in a review published in Reproduction in 2019.

According to the authors, embryo selection is “the most critical factor for the success of IVF.” However, there is no single definitive criterion that can predict the success of an embryo. Rather, embryo selection is based on a variety of factors, making it “difficult to predict the probability of a successful pregnancy for each patient and to fully understand the cause of each failure,” they wrote. The authors suggested that AI may be able to support clinicians in filling this knowledge gap, thereby improving the success rate of IVF treatment.

For instance, recent data published in Reproductive BioMedicine Online showed that AI — specifically the Life Whisperer Viability algorithm by Presagen — may be able to reduce time to pregnancy for people undergoing IVF by 12%. In practice, the algorithm does this by “analyz[ing] static images of day 5 blastocyst-stage embryos during IVF procedures,” according to Sonya M. Diakiw, PhD, lead author of the study and Presagen’s chief medical science officer, and colleagues.

“Life Whisperer Viability scores correlate with known features of embryo quality identified by both standard morphological assessment and genetic assessment using [pre-implantation genetic testing for aneuploidy],” Diakiw said in a press release. “However, when comparing the AI with standard morphological grading, the AI performs better in evaluating embryo viability. As a result, the use of the AI could potentially reduce the time to pregnancy for couples undergoing IVF treatment.”

In a 2021 opinion article, Mark P. Trolice, MD, FACOG, FACS, FACE, and colleagues cited a reduction of time to pregnancy as one of the main benefits of using AI in IVF. Other facets of the IVF process in which AI could be used to reduce time to pregnancy include “assessing gamete quality, sperm selection for [intracytoplasmic sperm injection (ICSI)] ... recommending patient stimulation protocols, the selection of egg donors and alerting the need for maintenance of IVF equipment to name a few,” they wrote.

Similar to Presagen’s algorithm for embryo selection, AI such as those used for sperm selection and quality control of gametes relies on image analyses.

Leveraging AI in these ways is a more objective approach to IVF than the manual selection of embryos, according to Wang and colleagues.

In addition to improving clinical outcomes, AI may alleviate some of the time-consuming, monotonous, detailed work from embryologists themselves, so that they can “focus on more important tasks, like ICSI, embryo biopsy and training junior staff,” preventing “the dreaded embryologist ‘burn out,’” Trolice and colleagues wrote.

According to Monseur, “AI could streamline and optimize decision-making during IVF cycles where we collect a multitude of laboratory and sonographic data on a daily basis.”

The streamlining enabled through AI may help the IVF industry itself grow, according to Hariton.

Eduardo Hariton

“Other than improving the patient outcomes, [AI] is extremely important for our field [because] we have very severe supply side constraints,” Hariton said. “We have 1,200 practicing [reproductive endocrinology and infertility specialists] in the U.S., and we have about 50 graduating every year. The rate of growth of this industry is double digits, and the predicted size of the market, if everybody had access, is probably three to four times the size that it is now. And we are not making decisions to serve this market fast enough.”

Challenges of implementing of AI in practice

In the paper by Wang and colleagues, the authors noted that “state-of-the-art [machine-learning] algorithms such as deep learning are still in the initial stage and have not been researched adequately.” Because of this, most of these machine-learning algorithms — a type of AI that holds the most promise in IVF, they said — have “ethical and legal risks and liability issues, which may lead to [the] distrust of patients and clinicians.”

Additionally, they cautioned that machine learning models require large quantities of high-quality, unbiased data, otherwise they could lead to incorrect decisions.

Monseur noted that “the data used for predictive models will likely be specific to certain clinics and not generalizable to other clinics or a larger patient population.”

Moreover, “because much of fertility care is individualized, this tool may not work well for all patients,” he added.

In another review published in 2020, Carol Lynn Curchoe, PhD, and colleagues wrote that “seemingly similar parameters can vary between clinics. For example, if one clinic captures blastocyst images at 110 hours, another clinic might capture them just before freezing — a time that may vary based on embryo development speed and current workload in the laboratory.”

Further, Trolice and colleagues noted that there is no “gold standard” AI for IVF.

“AI requires calibration, and there is currently no agreement on how to compare performances of various AI models for optimal methods,” they wrote.

Curchoe and colleagues also noted a lack of prospective research on AI implementation in IVF. Although Diakiw and colleagues used some prospective data in their study of Presagen’s AI, more is needed.

According to Hariton, another challenge in bringing AI to more practices is the current regulatory process.

“When you are trying to show that a drug works, you go through a randomized control trial, and you prove that it works against placebo,” Hariton said. “AI is something that you need to continuously improve, you need to continuously make it better. If you go through a 2-year clinical trial ... then you have 2 more years of data to make it better. But you cannot touch the algorithm because if you touch it, you have to go through another clinical trial. It does not make any sense. A randomized control trial is not necessarily the best type of trial to prove that these technologies work.”

Future of AI in IVF treatment

Much work needs to be done to integrate AI into reproductive medicine and IVF processes specifically. As Curchoe and colleagues noted, more research is needed to assess the practical uses of AI in predicting the success of IVF pregnancies.

“In a prospective, real-world setting, how much better is AI at embryo selection than traditional approaches is a fundamental question,” they wrote. “Also, is the improvement worth the investment that will inevitably stem from licensing or purchasing AI software and changing and revalidating existing clinical workflows?”

Despite their hesitancy, Curchoe and colleagues see promise in expanding AI to IVF because of the successes seen in other specialties — including the first FDA clearance of an AI-based device in April 2021, which was designed to detect colorectal polyps.

In fact, the promise of AI has gained so much traction in reproductive medicine that there is now a meeting dedicated to it — the AI Fertility World Conference — which was hosted in Croatia in September.

“It was a meeting with palpable energy,” said Hariton, who attended the conference. “A lot of these people work for private companies that are trying to solve similar problems. You can see how that would create this competitive atmosphere. It was the complete opposite. Everybody was sharing what they were working on, how they are thinking about problems, what barriers they are reaching. The collaborative air in that conference was amazing for a field that is just so [new].”

With continued research, AI technology will improve and “may have more uses for determining egg, sperm and embryo quality in conjunction with time-lapsed imaging,” Monseur predicted.

Hariton agreed, saying he would be “surprised” if any clinics are not using AI in some capacity within the next 10 years.

References:

• Curchoe CL, et al. Fertil Steril. 2020;doi:10.1016/j.fertnstert.2020.10.040.
• Diakiw SM, et al. Reprod Biomed Online. 2022;doi:10.1016/j.rbmo.2022.07.018.
• Global study shows Artificial Intelligence can reduce time to pregnancy for IVF couples by 12%. https://www.prnewswire.com/news-releases/global-study-shows-artificial-intelligence-can-reduce-time-to-pregnancy-for-ivf-couples-by-12-301617900.html. Published Sept. 7, 2022. Accessed Sept. 15, 2022.
• Trolice MP, et al. J Assist Reprod Genet. 2021;doi:10.1007/s10815-021-02272-4.
• Wang R, et al. Reproduction. 2019;doi:10.1530/REP-18-0523.