AI
AI might be able to overcome this subjectivity by identifying features that point to embryo viability that human eyes cannot recognise or understand , thus making the process of evaluating and selecting embryos much more objective . medical director of IVF Australia , told Global Health Asia-Pacific .
A key challenge in performing successful IVF treatments is that poor quality embryos , once transferred into the womb , don ’ t translate into babies . In a standard IVF , embryologists aim to create multiple embryos for each woman and then observe them over the course of several days to pick the most likely to lead to a pregnancy . But this selection process is far from foolproof , as it ’ s affected by an incomplete understanding of embryonic development .
The call is made by counting cell numbers and looking at the embryonic structure to determine whether the embryos are developing normally based on past research , but embryologists don ’ t have a clear sense of which ones are viable for pregnancy , explained Professor Illingworth .
This creates a sort of guessing game where experience and professional opinion play a major role . “ There ’ s a lot of subjectivity ” when it comes to selecting the best-looking embryos to transfer , and experts can easily disagree over the best course of action , said Dr Nikica Zaninovic , an embryologist and associate professor at Weill Cornell Medicine in the US .
AI might be able to overcome this subjectivity by identifying features that point to embryo viability that human eyes cannot recognise or understand , thus making the process of evaluating and selecting embryos much more objective , he explained .
Dr Zaninovic and his research group put this idea to the test in a study where they taught an AI-powered computer to distinguish between good and bad quality embryos by sifting through about 50,000 images of embryos that were transferred into wombs and either led to a pregnancy or failed . The AI technology then used this training to predict the quality of a different set of 2,000 embryos , achieving a 97 percent accuracy rate and a huge improvement over standard methods .
Other researchers have reached similar conclusions . Scientists at IVF Australia , for instance ,
AI can make it easier for couples to achieve pregnancy through IVF developed an AI system that can pinpoint the embryos most likely to develop a foetal heart , a sign associated with a 90 percent certainty of having a live birth , in a way that surpasses the ability of well-trained embryologists , stressed Professor Illingworth .
In Malaysia , Alpha IVF and Women ’ s Specialists is one of eight fertility centres across four countries ( US , Australia , New Zealand , and Malaysia ) that collaborated in research comparing the ability of AI in selecting embryos with that of experienced embryologists . Presented at the European of Human Reproduction and Embryology Congress in July 2019 , the results showed that AI was more accurate than doctors at predicting embryos viable for pregnancy , said Alpha founder Dato ’ Dr Colin Lee , an obstetrician and gynaecologist .
The research involved the use of technology developed by Life Whisperer , a company offering fertility clinics AI analysis of embryo images . In “ every study we ’ ve done , we ’ ve been at least 15 percent more accurate than what the embryologist can do ,” stressed Dr Michelle Perugini , founder and CEO of Life Whisperer .
The findings prompted Alpha IVF to introduce the technology last December for its IVF patients who have more than one embryo to select from , and the first woman who has received an IVF treatment enhanced by AI at the clinic has fallen pregnant in March 2021 .
Life Whisperer services are currently available in several Southeast Asian countries , Australia , and New Zealand , as well as Europe and Japan .
While AI is making advances every day , it ’ s still early days in terms of its application in fertility medicine , so it ’ s not clear what impact it will have on the actual number of live births through IVF . One reason is that good embryos resulting in a pregnancy can still end up in a miscarriage because embryo quality accounts for just 80 percent of a successful pregnancy . Other factors , like the overall health of the mother and her uterus , can also influence its outcome , explained Dr Zaninovic .
Despite the unknowns , experts are confident AI will improve the efficacy of IVF , with future experiments expected to provide solid evidence to back this up .
“ We expect that AI will increase our pregnancy rates ,” said Dr Colin of Alpha . But he also cautioned that the improvements would be modest in centres where standard IVF procedures have already achieved high success rates . “ But if you ask me , between a 70 percent pregnancy rate and a 72 percent pregnancy rate , I think 72 percent is still better .”
While incremental success rates will vary , what ’ s clear is that , within the IVF field , AI is poised to go into widespread use .
“ I think that artificial intelligence will become the dominant embryo selection method within three to four years ,” said Professor Illingworth .
This doesn ’ t mean , however , that computers will replace embryologists .
36 APRIL 2021 GlobalHealthAsiaPacific . com