Dr. Zev Williams, Chief of the Division of Reproductive Endocrinology and Fertility at NewYork-Presbyterian and Columbia, explains how his team built an artificial intelligence system for sperm identification and collection, opening doors for countless couples struggling to conceive due to male infertility factors.
On this episode of Advances in Care, host Erin Welsh speaks with Dr. Zev Williams, Chief of the Division of Reproductive Endocrinology and Fertility at NewYork-Presbyterian and Columbia and Director of the Columbia University Fertility Center. Recently, Dr. Williams and a team of researchers and clinicians used artificial intelligence to develop a system called STAR, or Sperm Track and Recovery, which combines advanced imaging with innovations in microfluidic chip technology to more accurately identify and capture sperm in samples from patients with azoospermia – a condition that often leaves men with untraceable numbers of sperm in their semen.
Dr. Williams explains that some azoospermia patients might have two or three sperm cells as opposed to the typical two or three million and having human researchers looking for those cells under a microscope is painstaking and rarely leads to success. Inspired by the AI-powered technology that astrophysicists use to find stars, Dr. Williams and his colleagues set out to build a tool that could help embryologists not only find those few sperm in a field of cell debris, but also collect them gently for future fertilization in an expedited manner.
The effort took five years of research and development, along with a collaborative bench-to-bedside research approach that Dr. Williams says is unique to the Columbia University Fertility Center. But the work paid off, resulting in a successful pregnancy and a promising example of how AI will continue to transform reproductive medicine.
***
Dr. Zev Williams is the Wendy D. Havens Associate Professor of Women's Health at Columbia and the Chief of the Division of Reproductive Endocrinology and Infertility at NewYork-Presbyterian Columbia University Irving Medical Center. As a physician scientist, Dr. Williams' focus has been on helping those suffering from recurrent pregnancy loss and infertility and developing novel technologies and treatments to improve patient success.
For more information visit nyp.org/Advances
Dr. Zev Williams: Over the last few decades, we've seen fertility medicine really being able to incorporate novel technologies to allow it to help those who previously have not had a real hope of success.
Erin: This is Dr. Zev Williams, a reproductive endocrinologist at NewYork-Presbyterian and Columbia.
Dr. Zev Williams: We've gone from having success rates of under 1% to now in good prognosis patients well over 90%. And we've also seen this field expand from being solely helping women with tubal factor infertility to now male factor.
Erin: In his practice, Dr. Williams often treats women who struggle with fertility issues like tubal factors, repeated miscarriages, diminished egg quantity, and more. But male factor infertility accounts for about 40% of fertility issues, and one of those issues is azoospermia, or the absence of sperm in a semen sample. Azoospermia has long perplexed physicians because the patient is often healthy and their semen sample looks completely normal. But when you look at it under a microscope –
Dr. Zev Williams: Instead of seeing a field of sperm swimming around, you just see fields of debris. So cells, cell fragments, but no sperm. What's interesting though is when we say there's no sperm present, we're oftentimes almost correct.
Erin: It looks like there’s no sperm, because the few sperm that do exist are extraordinarily difficult to locate. A normal semen sample contains two or three million sperm, but a sample from someone with azoospermia might contain two or three. Finding them is a painstaking, labor intensive process for technicians in a lab.
Dr. Zev Williams: It's something that is akin to trying to find the needle hidden within a thousand haystacks, but doing it in a short time period so that once you find that sperm, you're able to take that sperm and still use it to fertilize an egg successfully and have it result in an ongoing pregnancy. That's the kind of task that for a human to do is really a monumental and nearly impossible task.
Erin: But where some might see a roadblock, Dr. Williams saw an opportunity to look beyond the limits of human ability, and to leverage the power of artificial intelligence. He and a team of researchers spent the last five years developing a system that has made the nearly impossible possible – an AI-fueled advancement that is helping couples who never thought they'd be able to have a biological child of their own.
Dr. Zev Williams: You know, fertility is such an ancient part of the human experience, and to think that you have literally the most advanced technologies trying to solve some of the most ancient challenges that we've faced, I think there's something incredibly profound about that.
Erin: I'm Erin Welsh and this is Advances in Care, a podcast about groundbreaking developments in modern medicine.
Today, my conversation with Dr. Zev Williams, chief of the Division of Reproductive Endocrinology and Infertility at NewYork-Prebyterian and Columbia. He shares the story behind his latest achievement — STAR, which stands for Sperm Track and Recovery – an innovative technology to address azoospermia that’s part of a new era for reproductive medicine.
Erin: Dr. Williams, thank you so much for taking the time to chat with me today. It is great to meet you.
Dr. Zev Williams: Thank you, Erin. It's a pleasure to be here.
Erin: I'm curious how you found yourself drawn to this field. Was there a particular moment or experience that made you want to pursue this as a career?
Dr. Zev Williams: For me, when I was seeing those struggling with infertility who were really otherwise healthy, and then suddenly this thing that they wanted so badly that was so important and so meaningful and wasn't able to happen to them, I really felt that challenge. And helping those who had infertility and recurrent pregnancy loss.
But it was also interesting because at an intellectual level, this was such a black box field. There's so many areas where we just don't fully understand how it's happening. And it's one of those areas from a research point of view where you can make, within the scope of a career, actual discoveries that lead to changes in patients' lives. That, I think, is enormously motivating.
Erin: Yeah, and what makes research in reproductive medicine different from others? That shorter turnaround from research to impact – how is that more achievable than in other fields?
Dr. Zev Williams: We often speak in medicine about this transition from bench to bedside. To be able to do research and eventually translate it into clinically meaningful changes. And fertility medicine is unique in that the bench is literally at the bedside.
These are not two separate areas. And you can actually stand and see on your left side the procedure room where patients have their eggs retrieved. And on the other side, the IVF lab where the eggs are fertilized and grown. And so the ability then to be able to innovate and to introduce changes can happen at such a quick pace that you can really see multiple discoveries go from concept, initial design, development, validation and ultimately seeing patients having success from it.
And we measure success in two ways. One is the major goal of having a child when previously wouldn't have been possible. But the other is developing new technologies to make the fertility journey easier.
Erin: Right. And I wanna touch on one of these areas of technology, broadly speaking, and that is AI. I'd love to hear your thoughts on the potential benefits that AI might hold for medicine overall. Not just fertility medicine, but just medicine.
Dr. Zev Williams: Yeah, it's so interesting and I feel like we collectively are so fortunate to live in this major transition time in human history. We're all, you know, trying to envision and figure out how to, how it will enhance what we do, how it could augment what we do. And for us, one of the most exciting things too is not just how could it enhance or augment, but how can it do things that otherwise wouldn't have been possible? And whether some of the image analysis that's being done now by AI, some of the ability to look at large data sets and extract things that we might have been totally blind to.
Erin: Right, I mean, AI holds so much promise but it’s really only useful if you can get it from concept to practice, which is why that bench-to-bedside research approach is so crucial. So, tell me more about the setup that you have at your labs in this respect. I know that at many institutions, the basic science and clinical research teams are often siloed with really few opportunities for interaction. What have you done differently at NewYork-Presbyterian and Columbia in order to innovate?
Dr. Zev Williams: The traditional model for medicine is exactly as you described it. There's usually these basic science labs located at the medical center and then the clinical spaces. And they're two separate areas. The first thing that we did was to try to merge the two. So we created a whole new lab space and brought our NIH funded research lab from the medical center and embedded it within Columbia University Fertility Center.
What that meant was that the clinical scientists who are processing sperm samples are working literally within inches of the people developing the technologies to better analyze those samples. When we have this gift of having such talented and incredible clinicians and scientists all working together, it's not just the opportunity to be able to advance medicine, I think there's a real responsibility and obligation on us to do that because globally there's few places as well positioned to advance medicine to help give treatments and cures that are better positioned than NYP and Columbia to be able to do all of that together.
Erin: So let’s get into the details of how this setup at NewYork-Presbyterian and Columbia is helping to advance medicine. You mentioned that researchers are analyzing sperm samples – can you walk me through why azoospermia seemed like a promising area to apply this research?
Dr. Zev Williams: We started with the problem. And the problem was, what is one of the most devastating, catastrophic diagnoses that exist in fertility? What's a population of patients that currently do not have an opportunity for success? And it was quite clear that azoospermia was an example of that. When you tell somebody that they don't have sperm, their hopes, their dreams of becoming a biological parent are just vanquished.
And it's such a truly heartbreaking thing to see. And so we said, okay, here is a problem that if we can solve it would make such a tremendous difference. We can literally help bring life and joy to the world by solving this problem. And then we said, okay, well how do we do that? How do we actually find these rare sperm?
How do we do it in a way that's gentle, so the sperm could be used? Because of the advances that preceded us in IVF medicine, you only need one or two sperm. Like, the technology to take a single sperm and place it into an egg has been very well developed. The challenge then is how do you find that one or two sperm? And that's where we said, okay, let's try to come up with a solution to do that.
Erin: That is a complicated challenge that I’m sure took some complex technology to try to solve. And I know that there’s a unique backstory to this development. Can you talk about the inspiration behind this solution that you and your team built?
Dr. Zev Williams: We start off by thinking of the problem that astrophysicists have where they're sort of scanning through the sky. They'll see billions of stars and they'll wanna be able to detect, when is there a new star born?
What are some of the ways that they go about solving that problem? Because we're looking through fields with billions of fragments and we're trying to find that one special one. Here's a challenge where as humans it's just very, very difficult bordering on impossible for us to do. For men who have these samples where there's almost – at least when you do a standard semen analysis, even a more intense one, where you'll spin down the sample, you can't find any sperm – to look through this over 4, 5, 6, 7 hours or to do a surgery to try to remove sperm from the testes, which is often, you know, which isn't always successful, those aren't great options. So can we harness AI to be able to help us in this process?
Erin: Well, I mean, it sounds like that's exactly what you have done in creating this STAR system for sperm track and recovery. So I’m curious – what types of technology and expertise did you bring together to both find the sperm and then collect them for fertilization?
Dr. Zev Williams: So I think a good way to put together the expertise that are represented is to understand the elements that had to come together to make it happen. And you needed three separate technologies to converge to allow this to happen. Without any one of these three it couldn't happen.
So the first is to be able to make advances in microfluidics so that you can have a sperm sample that you can both detect sperm in and sort the sperm within one microfluidic chip. Up till now, there have been some chips that are very good for looking and some microfluidic chips that are very good for sorting. We had to develop a new chip, what's called the fusion DTX chip, which stands for diagnosis and treatment, that allows us to both look and sort sperm. So you need that advanced microfluidics.
You also needed high speed imaging technology. So the ability to image – we can go up to 8 million images per hour to be able to look through the sample without needing any lasers or dyes or harsh treatments.
And the last component is the AI component. The ability to look through fields of tons of debris and cell fragments, and to be able to pick out literally that one or two sperm samples out of millions of other images. And the whole thing had to happen so quickly that it could be done in real time.
Literally milliseconds from being able to identify, capture the image, identify that there's a sperm, and sort out that sperm. So to make that happen, we needed experts in microfluidics, experts in AI, experts in being able to take those really rare cells and still be able to capture them and use them to fertilize the egg. It's this entire team approach that's required.
Erin: Wow! So your team created this DTX chip, which is a completely new development in microfluidics, and you’re using the AI system to make the sperm detection process happen exponentially faster. Can you tell me more about how these components come together and what happens after you find the sperm.
Dr. Zev Williams: Identifying the sperm was actually the easy part. The real challenging part was then being able to isolate and recover that one sperm. So you start off with a semen sample that looks like a regular, healthy, normal semen sample, except there's just cell fragments and debris in it. You flow it through the special microfluidics chips. So at the end of the process from, let's say, two milliliters of semen sample, you'll end up with a droplet of about 10 microliters with the sperm cells in it.
Now, normally in reproductive medicine, you have hundreds of millions of sperm and let's say 10 eggs, just a few eggs. The rate limiting step is the number of eggs. In these cases, it's reversed. So we'll often get those 10 eggs, but might only have two or three sperm. So in practice what we do is, men who have this diagnosis where we're recovering sperm from their samples, will give multiple ejaculate samples in advance.
We're able to pick up those individual sperm and freeze them so we'll have more sperm than eggs. And then on the day of the actual egg retrieval, we'll remove the eggs and collect a fresh sample and preferentially use that fresh sample that didn't have to go through freezing. But if we have even more eggs available, we can then thaw the eggs and use those as well.
Erin: And so just to get a comparison or a really a perspective on how the STAR system has completely changed the game when it comes to treating or managing azoospermia, how long might that have taken and what would the success rate have been using these manual or traditional approaches compared to this new system?
Dr. Zev Williams: One of the first patients to have gone through this was somebody who had had a surgical procedure done and a sample was extracted and the lab team looked through the sample for two days. So really highly skilled, some of the most talented people looking through the sample for two days, couldn't find any sperm, went through the STAR system and within one hour found forty-four sperm.
Erin: Forty, forty-four! So after two days of looking versus one hour.
Dr. Zev Williams: Yeah.
Erin: That is incredible.
Dr. Zev Williams: And these are really great people. Like there's not a better group of embryologists who can do this, but it is just such a monumental task for any human to be able to do.
And it's amazing because, you know, we have now successfully created embryos because of this technology, have successfully enabled pregnancy to occur and there'll be children being born and hopefully thousands of children being born because of this same technology.
Erin: It must feel incredibly fulfilling to have created that first mousetrap, to lead the development of a technology that will truly change people's lives. And it already has, as I understand it. So the STAR system has recently helped a couple struggling to conceive. What can you tell me about their story?
Dr. Zev Williams: So the first couple to have successfully gone through the STAR system had been struggling with infertility for 19 years, almost two decades, but never gave up hope. The man gave a sperm sample and whereas a normal semen analysis and even careful looking would find no sperm, we found some, we were able to freeze them. And on the day of the egg retrieval, he gave another sperm sample. We were able to find seven sperm, some of which were actually moving, which is incredible. We didn't need to use any frozen sperm. The fresh sperm could be used. They fertilized the eggs, successfully created an embryo, transferred the embryo, and the patient’s now about four months pregnant with a healthy, ongoing pregnancy.
Erin: Wow, that is just a beautiful outcome to the story of this technology. I mean, would you have thought five years ago that this is what you’d be able to accomplish with AI?
Dr. Zev Williams: That was certainly the hope and the dream from the beginning. And what's interesting is how technology works, where we'll take something that was just a concept, design it, develop it, validate it, and implement it, and how quickly it goes from being, like, just a concept to like, well, yeah, this is just routine.
We've had a lot of fertility clinics from around the world reach out and ask to be able to offer this. And that's where we're working to try to make a lot of the user interface on the computer side and the microfluidics simpler to use so that a regular laboratory technician could run it.
So the bar, like, as soon as we achieve something, suddenly it becomes, okay, well now how do we elevate it more? That's just, I think that's part of the natural human drive. I think it's why we keep progressing. So of course, we developed the ability to now identify the sperm. The next thing was like, yeah, okay, but how do we now use it to get the healthiest sperm? As soon as you reach one bar, it’s, okay, how do we raise that again?
Erin: Well, I look forward to seeing how this kind of research will allow you to keep raising that bar in the years ahead. Dr. Williams, thank you so much for sharing all of this with us today. It is really so exciting to get these insights into how artificial intelligence is helping humans create life.
Dr. Zev Williams: Well, Erin, I really appreciate you taking the time to speak with me and explore this topic because I think this is one of those areas where all of us are living through this transformational period in human history, and especially biomedical history.
Erin: Many thanks to Dr. Zev Williams for sharing the story behind the STAR “sperm track and recovery” system, and its success – the first pregnancy, thanks to AI.
I’m Erin Welsh.
Advances in Care is a production of NewYork-Presbyterian Hospital. As a reminder, the views shared on this podcast solely reflect the expertise and experience of our guests. To listen to more episodes of Advances in Care, be sure to follow and subscribe on Apple Podcasts, Spotify, or wherever you get your podcasts. And to learn more about the latest medical innovations from the pioneering physicians at NewYork-Presbyterian, go to nyp.org/advances.