Maggie McKay (Host): Welcome to The Fertile Edge, a podcast from the Reproductive Science Center of the San Francisco Bay Area. I'm your host, Maggie McKay. Joining us today is infertility specialist, Dr. Eduardo Hariton, to discuss AI and embryology. Thank you so much for being here.

Eduardo Hariton, MD: Thank you so much for having me.

Host: Let's just start with what is AI in embryology, and how is it being used in IVF today?

Eduardo Hariton, MD: So, AI stands for artificial intelligence. I think five years ago, it might have been something a lot of people weren't aware of, but between ChatGPT and, you know, having it be everywhere, everybody knows about AI. When we think about the embryology lab, a lot of what we do there is take eggs, take sperm, make embryos, and grow them.

And typically, over the last 50 years of IVF, we were grading those embryos and evaluating them using our own eye through a microscope, of course, because they're quite small. However, now, we can apply artificial intelligence and look at those embryos and help us understand with a lot higher precision which embryos are more likely to lead to a baby. And then, there's other technologies that look at different parts of the process as well.

Host: That is amazing. Within that short of a time span, just a few years, we go from just looking at them to AI. How does AI-assisted embryo evaluation actually work? What is the technology analyzing?

Eduardo Hariton, MD: So, what AI does, this is a brand new technology. So, we've been studying this for years and years, but it just got FDA cleared a couple of months ago. And what AI does is that it looks at an embryo the same way a human would look at an embryo, but a human is subjective. If the embryo rotates, if you do it in the morning and then you do it again a few days later, you might give it different scores because we are inherently not perfect and accurate.

What AI does is that it looks at an image of an embryo, or there's some technologies that would look at multiple images. And it evaluates, you know, tens of thousands or hundreds of thousands or sometimes millions of little pixels to try to determine which embryo has the highest chance of success. Models are trained on hundreds or thousands of embryos that have been looked at and tied to an outcome.

So, the computer learns which embryos look like the ones that are more likely to implant, and then can look at an embryo and say, "Does this look like the one that implanted or does this look more like the one that didn't implant? Which one's better?" It looks at those features and it comes up with a ranking score of which embryo is more likely to lead to a pregnancy. So, it helps us select embryos better.

Host: Wow. And as you said, you know, we're not perfect. AI maybe isn't perfect, but has seen a lot more than the doctors, would you say? What insights can AI provide that traditional embryo grading might miss?

Eduardo Hariton, MD: So typically, embryologists train and they're the ones that are grading these embryos. The doctors are looking at those grades and making decisions. But we spend our time sitting in front of our patients and helping guiding them. So when an embryologist is training, they learn how to grade based on a scale.

But again, embryologists are subjective. AI is not subjective. AI will look at that embryo and give it a score. We're not ready to replace our embryologist. So for my embryologists out there who are listening, no one's coming for your job. If anything, this is a technology that will help make your life easier.

You know, we still want to look at the grades that our embryologist gives those embryos, but sometimes the grades between two embryos are the same. So, we are picking at random which one to transfer. Now, we have another layer based on, you know, well-evidenced technology to pick which embryo is better.

So when we're looking at embryos, the grades might be the same or they might be quite similar, but there might be something that the AI detects on the image of an embryo that tells us, "Well, we might have transferred embryo number one, but embryo number two over-layering that AI score actually looks better and has a higher chance of a pregnancy."

Our patients go through a lot of energy, pain, you know, expense to get to the process where they're ready to transfer an embryo, and we want to work as hard as we can to bring the latest technologies and help them pick the right embryo for them that has the highest chance of getting them pregnant. So, we wanted to bring this technology into our clinic to allow them to have everything we can use to pick the right embryo for them.

Host: It sounds like the best of both worlds. You have the doctor's expertise and the extra tool of AI. So, how does AI help support clinical decision-making while still relying on the expertise of embryologists, as you mentioned, and physicians?

Eduardo Hariton, MD: So, it's exactly what you said. We like to use both. We want to use what we know has worked really well to bring our pregnancy rates as, as they are today, but we also want to continue to evolve with technology. So on top of doing the same process we've done for patients, which are simulating them, culturing it under high quality conditions, trying to get as good of an embryo as we can get, and having our embryologists grade it in the way that they know how to, we now use this image and overlay that AI score for patients that want to do this, and that helps us compare embryos, you know, across that cohort between one another, and then helps us select the best one.

So, we still get to do everything we've done so far. But now, we have one more tool in our armamentarium to help that patient select the best embryo first and give them the highest chance of success with each transfer.

Host: That's so great. Because like you said, it's a journey and it's expensive and time-consuming and heartbreaking at times. So, the more help, the better. What are the current limitations of AI and what should patients understand about what it can and cannot do?

Eduardo Hariton, MD: That's a great question. I think AI is one of those evolving technologies that we have to manage carefully. There's a lot of hype around, and we want to be thoughtful about not over-promising what AI can do. You know, an AI cannot replace human judgment. An AI is built on data. So, the patients who are being taken care of have to be similar to some degree to the patients that were used to build the model and to test the model.

You know, we were part of the study where this, you know, new technology came to market. We were one of the clinics that participated, recruited patients. Our patients got to have first dibs at seeing this technology. And now, we get to use that tool. So, we know our patients and our demographics have gone through this tool and used that, so that makes us feel good. You want to make sure that technology is well-validated. So, we continue to pilot, learn, measure, and make sure that we're seeing the impact and the effect that we expect to see to make sure that technology is working well. We want to make sure that we continue to not let it run rogue. We want to make sure we pair it with what we know knows well.

So if the AI is telling us something that feels like it's not correct, we have an embryo that in our hands with our grading should be much worse than the AI, we have protocols and mechanisms to make sure that the AI is not overriding our clinical decision-making, that we're still putting the physician and the embryology judgment first as we get comfortable with the technology and continue to get more and more experience with it.

So, I think it's really important to make sure patients understand that and know that, you know, the best embryo sometimes doesn't implant either with AI or without. So, we got to continue to not just look at one score from an AI, but look at the whole clinical picture and do what's right for the patient.

Host: And as you mentioned, a good point, you don't want AI to go rogue. So, you're on it. Looking ahead, Dr. Hariton, how do you see AI shaping the future of embryology and fertility care?

Eduardo Hariton, MD: So, I think we're already seeing, at least in our clinic, AI touch some of the things that we do. So from the beginning, when I meet a patient, we have a tool that helps us use AI to help them understand their success rates. We know how old they are. We have data on how many eggs they might have, what's their diagnosis, what does the sperm look like, so we can use that to better counsel patients to pick the right treatment for them, whether that's inseminations or IVF, we can guide them through that.

We have tools that help us figure out which is the optimal day to get those eggs out and do their trigger shot to maximize the number of eggs that we get. IVF is a numbers game. So, the more eggs we can get, the better. So, we use those tools to try to help us optimize the outcomes for patients. And now, finally, we have a tool in the lab that can also help us on the embryo stage figure out which are the embryos with the highest success rates.

So, you know, we want to move these technologies into patient care really thoughtfully. We want to make sure that they work and that they're safe and that we know how to use them. But at the same time, we want to make sure that we're not sleeping on the sidelines as the whole world brings AI into what they do. We want to make sure that we continue to advance in how we take care of patients to make sure that we're giving them the best care possible, and the most advanced care that exists safely.

Host: That's quite a balancing act, I would think, you know, to make sure, like you said, you don't just rely on one or the other. But it sounds like if you need fertility help, this is a good time to be looking for it because of AI and the combination with doctors. Would you say that's accurate?

Eduardo Hariton, MD: I think that's true. I think it's a little bit of AI, a little bit of technology evolving. We've gotten our success rates to go from less than 10% to over 60% for a genetically normal embryo transfer. In a span of 50 years, we've learned how to cryopreserve embryos. We've learned how to work with very little sperm and fertilize eggs by putting one sperm inside the egg.

And this is just another evolution of that technology, where we're harnessing the power of neural networks and AI to help our patients from the moment we meet them in the clinic through the stimulation, and now at the time of embryo selection, have the best chance at a pregnancy. So, I'd say if I was trying to get pregnant, I would absolutely want to do it now versus 50 years ago.

Host: Absolutely. Well, this has been so encouraging and informative. Thank you so much for sharing your expertise. Is there anything else you'd like to add in closing?

Eduardo Hariton, MD: No, I would say for our patients listening, we're very proud to bring this technology. We're the first network and the first clinic in the whole country that is making this available for patients. So, we're happy we're in Silicon Valley, right in the Bay Area, where a lot of this technology is coming to bear, and being able to offer that to the people in our community is very special.

And for our patients who are listening to this, if this is something that you might be interested in, make sure to ask your doctor and they can walk you through whether it would be something helpful in your specific case and how do we go about turning that on for our patients.

Host: Congratulations on all the success of the clinic. That's amazing. Thank you again for your time.

Eduardo Hariton, MD: Thank you so much for having me.

Host: Again, that's Dr. Eduardo Hariton. To find out more, please visit rscbayarea.com. And if you found this podcast helpful, please share it on your socials and check out our entire podcast library for topics of interest to you. I'm Maggie McKay. Thank you for listening to The Fertile Edge from the Reproductive Science Center of the San Francisco Bay Area.