AI for Early Detection: A New Chapter in Breast Cancer Prevention

Is 2025 a pivotal inflection point for AI in health care? Dr. Connie Lehman, Co-Founder of Clairity, thinks so, and she has strong cause for optimism. Her organization’s software-as-a-medical-device product, Clairity Breast, recently received authorization from the Food and Drug Administration as the first AI platform that predicts a woman’s five-year risk of developing breast cancer.

On this episode, Dr. Lehman shares her journey with Clairity, from the paper she read as a medical student that sparked the idea, to her experience navigating the new domain of image-based risk assessment with the FDA. Her current focus is on implementing the technology through education and advocacy.

Dr. Lehman is passionate about advancing medicine toward risk assessment and disease prevention. Understanding risk empowers patients and their health care providers to choose the best path. Dr. Lehman envisions a future where image-based risk information is accessible and available to improve health outcomes for all.

Creators and Guests

Host

Alex Maiersperger

SAS

Guest

Connie Lehman MD, PhD

Co-Founder, Clairity

What is The Health Pulse?

How can data, AI and advanced analytics accelerate health innovation? Which new technologies hold the most promise? What are the biggest roadblocks to progress? How can we solve endemic problems?

Join us for The Health Pulse podcast series as we explore fresh perspectives on digital transformation in health care and life sciences. With a special guest expert on each episode*, we’ll tackle the most pressing issues affecting the delivery of health care and therapies worldwide.

All presentations represent the opinions of the presenter and do not represent the position or the opinion of SAS.

CONNIE LEHMAN: I think we're going to look back at 2025 and say that's really the year that it all started to really change. That's when we really were able to start to put into practice the incredible scientific discoveries in AI from the past, literally, 20 to 30 years.

ALEX MAIERSPERGER: Wouldn't it be amazing if we could take these patterns and have a computer identify women destined to be diagnosed with breast cancer. As a student, our guest today read this idea in a medical journal, and years later created the first FDA approved AI-powered prediction tool, changing the way diagnostics will be done forever. Today, we welcome to the podcast, Dr. Connie Lehman, founder of Clarity.

Dr. Lehman, most people see an X-ray or an image of the body and they think, tell me what's wrong right now. Your life work has been to look at those images very differently. Was there an aha moment, or has this been a gradual progression of how you see these images to get to this point?

CONNIE LEHMAN: I would really say it was a series of aha moments. So the first was way back when I was in training as a resident, and I was on my mammography rotation. And I was reading papers in the Journal of Radiology. And John Wolfe, in 1976, published this paper that I thought was amazing.

And he did a very careful study and said it turns out that women that develop breast cancer seem to have different patterns of breast tissue than his patients that did not develop breast cancer. And he's like, wouldn't it be amazing if we could take these patterns and have a computer analyze them and then help identify the women destined to become diagnosed with breast cancer.

And so that was in-- around 1990. I talked with my mentors and faculty. And they said oh, that was crazy. Everyone thinks he's nuts. It's too bad because he did a lot of important things in the field, but that was just nutty. And so I set that aside.

Then I kept going along, decided I really wanted to work in breast cancer. I was so excited about women's health and the power of radiology in that domain. And I had the good fortune to work with a whole group of colleagues with the Breast Cancer Surveillance Consortium. And they really opened my eyes to the importance of population health and looking at epidemiology and trends over time and ways of not just diagnosing breast cancer, but identifying the tools that would really have the sensitivity and the specificity that we needed.

A lot of our focus in radiology is on finding the cancers. And they really opened my eyes to the importance of having that balance. We could have 100% sensitivity if we told everyone that they needed a biopsy for breast cancer and found all those.

So anyway, that continued. My research continued to evolve. When I landed in Boston, I had the very good fortune to meet Regina Barzilay. She had gone through her own personal journey of breast cancer. And we decided we could combine our interests and our backgrounds to apply the latest in computer vision.

And a lot had happened between 2010 and 2015 in computer vision with computer scientists, apply that to identifying those women back in the John Wolfe story destined to develop breast cancer. So that's when it started. And really, just sort of opened up my eyes to how much more images of the human body can offer us and just to detect and diagnose present visible disease.

ALEX MAIERSPERGER: We're going to talk about turning those dreams into reality and the incredible achievement of being the first of its kind AI-powered prediction tool to receive FDA authorization. But first, let's talk about breast cancer specifically a little bit more in depth.

I have a wife, two daughters, a mom, sister, nieces. Conventional wisdom says to someone like me, breast cancer is extremely rare in our family history. It's not something I worry about, or I don't need to worry about it. Is that what you see?

CONNIE LEHMAN: I'm so glad you brought this up because we, radiologists, when we do breast biopsies, and we talk to our patients to explain to them the results of the biopsy, it is one of our more common stories that we tell of the woman being diagnosed with breast cancer says, that's just impossible. No one in my family has ever had breast cancer, Are you sure? Can you double-check?

Many of us remember the Katie Couric story, where she wrote a beautiful article that she started off with a title, "Why Me?" How could this be? No one in my family has had breast cancer. I don't have a genetic mutation. Otherwise, I'm healthy. It doesn't make any sense.

But that's actually the story of the vast majority of patients with breast cancer. The majority have no family history. In fact, 85% of patients diagnosed with breast cancer cannot identify a single member of the family.

So this tells us that breast cancer is in a domain of what we refer to as sporadic. So this isn't following strong family lines. And we really need to focus equally on these sporadic cancers as we have. We've made so many great gains in understanding the breast cancers that are related to familial patterns and are related to genetic mutations that are inherited.

So that work is so important. I would say that's probably 15% of patients diagnosed with breast cancer. Can't stress the importance of that enough. But we've left 85% behind. And that's what our products, that's what our approach, that's what we're tackling.

ALEX MAIERSPERGER: That was not the answer I wanted to hear. Thank you for giving me the facts. But definitely sounds like it's something we all need to be prepared for and thinking about and maybe worried about somewhere in the back of our minds as we get to the ages and stages where we need the screenings.

CONNIE LEHMAN: It's interesting because the age of breast cancer onset is earlier and earlier. It's the most rapidly growing domain of women being diagnosed with breast cancer or young women. And so we're needing to think differently. We just need to not be shackled by data that is now 20, 30, 40 years old. We really need to be following these current trends and adjusting so we can get ahead of this.

We all know someone diagnosed with breast cancer. The more common story that I'm hearing from my friends, from my family, from my patients, is another woman diagnosed with breast cancer in her 30s, in her early 40s. It wasn't until pretty recently that we even were all agreed that we're going to start screening at the age of 40. And so we want to get ahead of this. That's what we're trying to do.

ALEX MAIERSPERGER: Is this the food we're eating? Is it some way about modern society that this is happening? We see these-- like you said, we all know someone. And it does feel-- even from a societal level, it just feels like it's happening earlier. Oh, my gosh, this came out of nowhere, those type of things. Is that something about how we've designed our lives currently?

CONNIE LEHMAN: Certainly, we know that the environment really matters. Actually, I was watching and listening to your podcast with Otis Brawley, which it was so great to listen to, and how much he emphasized just the data.

When we look at the data and we see how much environment matters-- the zip code that one lives in, the part of the country that one lives in. So some of it is access to health care, but there's also elements about the environment itself.

So when we see that women living in my state, so particularly in the area of race and environment, so Black women living in Massachusetts are significantly more likely to survive breast cancer than Black women living in certain states in the South. And that tells us a lot about what patterns can we look for in the environment.

Now, what I'm excited about is those elements of the environment that everyone's been studying-- the influence of diet on breast cancer risk, of environmental toxins and exposures, of hormones in dairy products or in our meat products. Hormones definitely have a strong role in breast cancer, and we need to study that better.

What I'm excited about is until now, we haven't had a way to actually measure those impacts on the individual woman. So I could live in the same neighborhood, have the same exposures than my neighbor. My body responds to that differently, the inflammatory processes that are set up. My weight, my body mass index-- they can be different even though we're in the same environment.

And now we have tools that can extract that impact of the environment on the human body from images of the human body. So it's a new domain we're opening up. And we're pretty excited about it. We've needed a better way to measure the impact of the environment on the human body. And we think we've opened up a big window to see into this area that before, we've really been pretty restrained in how we can study it.

ALEX MAIERSPERGER: And you've helped get us into this new world. So congratulations are in order. The product from the company you founded, Clarity Breast, received de novo authorization from the FDA to predict a woman's five-year risk of developing breast cancer. So it's a big deal. It's a new thing. What's different about your platform, your approach? Why did the FDA give the green light for this one?

CONNIE LEHMAN: So I learned a lot. This was the first time I had founded a company. It was the first time I had been through the FDA process. I'm not a particularly patient person. So I thought that it would be frustrating trying to get through the FDA. Actually, there were some days that were frustrating.

I have to say, I came out of it with such a deep appreciation for the importance of the FDA regulatory process. They worked with us with such focus and such curiosity and such a willingness to figure out together, how can we move into this new domain? And they are very organized.

So the first question was, does this need FDA regulatory oversight? Because there's a lot of products we use in health that don't require that. They were very clear that this does. It's something-- it's a domain of software as a medical device. And whether we have software as a medical device using human tissue or images of the human body, the FDA is going to be involved. And I think that's really important.

So it was clear that, yes, we were going to need to submit to the FDA in order to have this available to women in the US. The second point was, well, is there a predicate? Because if someone else has done something similar to this, we could follow that pathway. And they determined in many meetings with us for the approach we were taking and how we were training and testing our model that there was no predicate.

Then they had to really think carefully and have a lot of sessions with us on, well, what are the potential risks, so they could decide what class is this. And it was a class II. And that's when it led right to, well, then you will do a de novo application. There is no predicate. It does need to be regulated. It's a class II. So this will be a de novo application.

And then it got really interesting and really exciting because the questions were so much around the questions that we're all facing as we bring AI technology into health care. What will be the role of the physician, of the health care provider, of the radiologist in this?

And we needed to be really clear that this application of AI and computer vision is doing a task that I, as a human, as a well-trained, specialized breast imager, I can't do. I cannot look at a mammogram and give a estimate of the woman's five-year future risk of breast cancer.

So this isn't similar to the CAD products that have been out there since 1998, cleared by the FDA, where it assists the radiologist to do the radiologist's job hopefully a little better, to identify those lesions that might be cancerous on the mammogram. But this was truly something that will be autonomous.

The computer can do this. Our AI deep learning algorithms can do this, and I cannot. No matter how specialized, no matter how good I am, as a human, I cannot see these signals that the deep learning computer vision pathways were seeing and identifying.

So with that, then we had to look into, well, what's the comparison? And the FDA ended up feeling, well, there's no good comparison because the traditional risk models have such low performance. And they're not regulated at all by the FDA. Breast density is being used somewhat to try to triage women into high and low risk. But that performance is so weak. They didn't think that was viable either.

So we co-created, developed very rigorous methods for two important pieces of a risk tool. One is how well does it discriminate the high- and medium- and low-risk patients? And we were doing this on a 0% to 100% scale. So it was going to be very precise.

But the second very important piece was, is it calibrated? Most women are going to be in the 0.5% to 5% five-year risk. And so we needed to have the calibration really good so actually it's meaningful if a woman is 1.3% five-year risk versus 1.7% five-year risk. Doesn't sound like a big difference. But the clinical care pathway that women go on based on these different risk scores really matters.

So I know that was a lengthy answer. It was sort of a lengthy process. But I'm telling you, it's a new domain with the FDA. For the CAD products, for tools to assist the radiologist to read a mammogram better, those studies were 240 heavily enriched selected mammograms read by 12 to 20 radiologists.

We had over 77,000 consecutive mammograms from five distinct geographies across the US. We wanted to show how well this generalized, how it was meant for all women, and how it was really a high bar for performance. And I think it's a new era, where we will see what is being authorized by the FDA in this domain of image-based risk assessment is going to perform very strongly across the full diversity of patients we serve.

ALEX MAIERSPERGER: You talked about the incredible power of the machine beyond what a human is capable of, in some respects. The ability right now-- you talked about just that human element of the physician in the room, understanding that score and having the belief in it and the confidence and trust that that score is accurate.

Telling a woman you're at risk of breast cancer in these next few years based on something I don't see on the scans, but based on what the computer is telling me, how's that-- are we navigating that right now as a society? Do you give instructions to these physicians? Is that a conversation they have to have of, your risk score is this. I don't see it now, but we're probably going to see it in the future?

CONNIE LEHMAN: I'm really glad that you asked this question, because now that we've accomplished the FDA authorization, the whole world has opened up for us to now get this out to women. And part of getting it out to women and their health care providers is the educational piece.

We all, as a society, are learning every day about what AI can do, what it can't do, the benefits, the risks, the dangers, the excitement, the enthusiasm. So as a company, we really want to lead in that.

How can we focus on all of these elements? How will our patients understand these risk scores? How will they leverage the power of these risk scores to improve their outcomes? Because that's really the link we want to make. And how can we work with providers?

We've had fantastic partners since we first founded the company, who were all in this vision that we had. And we're continuing to work with them as now we're implementing clinically to make sure that we continue to co-create together. We learn together. And there's just-- there are going to be a lot of pieces to it. And society's learning more and more.

I like those expressions people use-- we focused a lot on literacy. How do we write our materials for our patients so that they can understand them? There's an area also, the numeracy. How can we help people understand what percentage risk means? You can have the same data and have it look completely different.

So I could tell a woman that she has a 3% risk of breast cancer in the next five years. And she could think, well, that sounds great. That's like getting an A-plus. 97% chance I'm not going to have breast cancer. Or you could say, you're at the highest risk category for future breast cancer. You're three times more likely to develop breast cancer than a woman with similar patterns and similar risk factors that has this lower average risk.

So we want to think a lot about that. We really want to focus on, this is information that is knowledge. It's empowering. By knowing your risk, you can actually change your path. You can choose the best path for you with your health care provider. And in the event that you are diagnosed with breast cancer, we can detect it early when it can be cured.

And even more importantly, so much work has been done to prevent cancer. And that's the domain I'm most excited about. Let's move from late-stage disease treatment, late-stage diagnosis, the work being done for earlier detection.

Let's keep pushing that envelope. Let's push it all the way back to risk assessment. We're actually preventing the disease from occurring. And now that we have a dynamic risk factor, we can really get into this game. And I can't wait for what the next five years are going to bring.

ALEX MAIERSPERGER: It is so incredibly exciting and so incredibly meaningful. And again, congratulations. You probably felt like you were a little bit back in the FDA hot seat of answering questions and having to get in depth about societal challenges and things. So we'll give you a little bit of fun here in the middle with a speed round meant to be quick-paced questions, whatever comes to your mind first. What's one place on your travel bucket list?

CONNIE LEHMAN: So my son Sam just got back from this fantastic trip where he went to one of those destination weddings in Italy and then stopped in the Azores on the way home. He said it was like the Hawaii of Europe. I guess that's the tagline. I've never been there. And all my kids are fantastic travelers, so I like learning from them and then sort of following in their jet stream.

ALEX MAIERSPERGER: That's awesome. Hawaii of Europe-- I have heard that. I think that's added to mine as well. What's your favorite app on your phone?

CONNIE LEHMAN: Do you know I have a love-hate relationship with my phone. I try to turn it off a lot. Favorite app-- I do-- I actually honestly do love the podcast app because I can go there, put in people or topics or things.

One of the ones I've been doing recently is Wiser than Me with Julia Louis-Dreyfus. Love that. Love the stories these women tell of their discoveries. So I would say that. And then that thing at night you can turn on to make white noise or burbling streams or whatever. That's a nice one, too.

ALEX MAIERSPERGER: So you mentioned nighttime. Are you a morning or night person?

CONNIE LEHMAN: Depends on how you define it. So nighttime, 5:00 to 8:00 PM, I'm all in. If we're talking 9:00 to midnight, I'm squarely in morning people.

ALEX MAIERSPERGER: All right. Wait, what time of morning people? I think that's also defined.

CONNIE LEHMAN: Right. It really depends. Some of my best work actually is in the morning, but it's not-- I have some friends that are up at 4:30 and 5:00. But I'm going to be more of a 7:00 AM kind of person.

ALEX MAIERSPERGER: All right. Splitting the difference. You're a prime hours person. I like it. Books or movies?

CONNIE LEHMAN: Love them both so much. Love stories, stories of the human condition. But I'd have to say if I had to pick, it would be books.

ALEX MAIERSPERGER: Mountain or beach?

CONNIE LEHMAN: Ooh. Mountain.

ALEX MAIERSPERGER: Favorite-- this is my personal most important question of what's your favorite ice cream flavor?

CONNIE LEHMAN: Oh, my gosh. My favorite ice cream flavor is going to be some kind of fresh fruit. Like, whatever is fresh. If it's a homemade fresh peach or fresh strawberry or whatever. My husband Paul, it is cookie dough-- what is it called again? Cookie dough ice cream? Cookie dough--

ALEX MAIERSPERGER: Anything cookie dough qualifies me and Paul as friends in this category.

CONNIE LEHMAN: He actually thinks he invented it. He's pretty certain he invented it because he was getting those rounds of the uncooked cookies, breaking them up, and then putting them into vanilla ice cream. But for getting ahead of Ben and Jerry, he could have retired early.

ALEX MAIERSPERGER: All right. Your contributions to society are great, and we thank you. But we thank your husband Paul almost equally for the invention of cookie dough ice cream. Tell him his check's in the mail.

CONNIE LEHMAN: [LAUGHS]

ALEX MAIERSPERGER: What's-- this follows the ice cream question every time of what's your go-to workout or exercise?

CONNIE LEHMAN: OK, so Paul also-- I definitely ride in his jet stream. He is a fantastic athlete. He's into paddle and all these great sports. So we go to Equinox's barre classes together. We've been doing it for about a year. And I think they're fantastic.

They are just-- they're like muscles you didn't know you had. It's a lot about balance. I think as Paul and I are getting older, we really want to make sure we've got balance. We've got muscle strength. And so the bar classes are really fun. He's usually-- there's a class of 40, and it's Paul and maybe one or two other men. And he's great. He's all in, and we have fun doing it together.

ALEX MAIERSPERGER: He's leading the way on ice cream and exercise. I like this trailblazing. A job you'd be doing if it wasn't saving the world from breast cancer?

CONNIE LEHMAN: My other life-- being a writer. I would have just loved that. I love-- and maybe it's because I want to be like these women I admire so much that tell a fabulous story and just know how to write about the human condition. But I think that.

ALEX MAIERSPERGER: Love that. All right. We're putting you back on this FDA hot seat and getting into the hard-pressing questions. But thank you for giving us some insight into your life and the adventures that you have and will have.

Speaking of writing, Clarity's tagline is AI-powered human health. What does all of this mean that we've discussed for the future of health and healthcare? I think we see some of the headlines. There's three camps of people.

There's, we're going to live forever. There's another of brain and computer interface. This is the future. And then there's another headline of, we're going to cure all diseases in the next 10 years, so we don't need to worry about some of this stuff. Are any of these headlines anywhere close or on the right track? Where does medicine go from here?

CONNIE LEHMAN: I think the, we're going to live forever and we're going to cure all diseases, I just-- I don't think that way. I don't think human history has shown us that that is a reasonable way to think.

It certainly aspirational. It's exciting to think could we just find ways that we just eradicate all disease? But how great would it be if during our lifetimes, whether right now we're seeing our lifetime as another 10 years, another 50 years, whatever that is, if in our lifetime we kept pushing towards a healthier society.

And there's elements to that that are-- we could enact within the next three years with the right focus, having health care be more of a right of all people rather than a privilege. That would be incredible. That would just have such an impact.

I do think we're seeing more and more of groups trying to, again, pull back from late-stage disease treatment and pull back towards early risk identification and preventing the disease. And we see this in cardiovascular disease, what we're trying to do to tackle the obesity epidemic, really understanding more how does the tissue in the body-- how does it become inflamed? And what is that inflammation? What are those inflammatory processes? What are they doing, and how do we reduce them?

The work being done to make our food healthier, to remove the toxins from our food, from our environment, from our drinking water-- I think we could become so much healthier, feel better with knowledge that we have now implementing that and implementing that across the full diversity of humans globally.

I think that's what the real challenge is and one that I think with AI, with the power of AI, we have an opportunity more so in health than any other domain, to use AI for good, to have some of our biggest challenges in having all humans lead healthier lives be addressed and addressed on a much faster scale than we could imagine before the AI revolution.

ALEX MAIERSPERGER: You don't have to lay out your whole product roadmap, but specific to the imaging world that you live in, is there a rush to go through recent images? And so you've found that this prediction tool works for breast images. Is there a last five-year window of outreach? So some people that got images five years ago that didn't have this technology at the time, are you rushing to get through that? How quickly does it become the standard?

And then does it expand to other things? I have three active kids, and so we've had our fair share of X-rays and ultrasounds and images. Is there something we can learn from those of vitamin deficiency or any perspective on other imaging that's coming up?

CONNIE LEHMAN: Yes. Yes to all those questions. So first, I think that if I close my eyes and I imagine that we had blood from people all over the world stored and available and ready for testing-- we can just-- it's already been collected. We pull that out, and we can run tests on it to inform each individual patient so much more about the best sort of designed precision health care pathway for them.

And we're seeing more and more people are doing that. Send in your blood, and we can run all these different tests and then give you very specific recommendations about what you should be doing with your diet, with your exercise, with medications.

Now, we open our eyes, and we close them again, and we imagine all those images stored all over the world sitting there collecting digital dust. I mean, they're not being used at all. And now we can.

Now we can extract those because those are data points. Those are every bit as important as a patient's blood. And we haven't been using them for their full potential because we've been limiting them to detection and diagnosis of current present disease. So we're going to open that all up. And, as a company, we're so excited to think about how you can leverage the mammogram to predict diseases outside of breast cancer.

And again, I go back in history. It's like the John Wolfe story. There are also stories of radiologists that said, we see these arterial calcifications in the breast tissue. Wonder if that has anything relevant to cardiac disease? Because we see these arteries in the heart, the coronary arteries, we see calcifications there that make people at higher risk for having a myocardial infarction. So could we put those together?

And they started to-- it was like they heard this faint signal, or they saw this really hazy, hazy signal. And we believe just in the way that AI and computer vision has really realized John Wolfe's, that faint signal he heard or saw into a reality, we think we can do the same thing with AI and the mammogram, predicting a woman's cardiovascular disease risk. And that is so important because while breast cancer is critical, it's a leading cause of death in women, cardiovascular disease is the lead killer for women. And most women aren't aware of their risk.

Men are much more likely when indicated to go on statins compared to women. They just don't see it as a disease they need to worry about. They worry about it in their brothers and their fathers and their husbands, but they don't worry about it in themselves. And so we're really excited to move forward in that domain.

There was another part of your question-- there was so much good information there I wasn't sure where to start. There's another part, which is, well, what about all of these images from the prior years? And one part is like, well, that's exciting because they're there, and we can pull them out and then assess them.

So if a woman-- even just starting a very practical level, if a woman says, well, I want this risk score. I had my mammogram four months ago, we can pull that and process that and provide the risk score. So that's one thing.

But the other part that's really exciting is that these now give us the opportunity for the first time ever to have what we refer to as a dynamic risk score. So if you think about it, you test a woman for the BRCA mutation or other genetic mutations, and it's either positive or negative, and it's not going to change.

And the traditional risk model is largely based on family history and age. The family history doesn't change a lot. Every now and then they'll be like, oh, actually, my aunt was diagnosed with breast cancer last year, so that adjusted a bit. But usually they're pretty static.

Every time a woman gets a mammogram, there are subtle differences in her breast tissue. We know that with AI, we can determine a woman's body mass index from her mammogram, and we can see how that changes over time. And even my colleagues and I that just with our human eyes look at mammograms, we can see changes over time in women as they age, as they gain or lose weight, as they undergo hormone treatment, or they come off their hormone treatment.

And so we did a study, and we took a group of women that had been diagnosed with breast cancer. And then the controls were those that had not been diagnosed with breast cancer. And we took all their mammograms going back six years.

So we knew that the women diagnosed with breast cancer, as far back as six years, were going to have higher scores than the cohort that did not develop breast cancer. But what was really exciting for us to see was that it was the slope of change over time. And that change started six years before the diagnosis.

So it's not just having that score, but also comparing the score to the prior year, a little bit like what we do with PSA values. Like, this is your score. Gosh, you seem to be steady there for 10 years. Or this is your score, but you've really-- you've jumped up 10%. And that's a jump up that we're going to pay more attention to.

And also when you have a dynamic score, we're going to be better at testing the impact of our interventions to reduce risk on the individual woman. So some women are recommended to go on chemoprevention, such as tamoxifen.

Most women don't stay on those chemoprevention therapies. We're hopeful that if they can see the impact with our dynamic risk score, it's like going on a medication to reduce your cholesterol and then seeing with that medication the impact, or for your high blood pressure and seeing the impact.

And I think that that helps keep the patient engaged in that intervention to reduce that risk, to tackle that problem. So this is going to be a whole new domain. And the serial mammograms over time, because of the way that we screen-- in most regions here it's annually. In Europe, it can be every two or three years. But most women have prior exams. And we can look at changes over time and inform them even more with those dynamic risk scores.

ALEX MAIERSPERGER: This is so incredibly exciting to be living in this moment and getting to learn this and hear about the potential that we have in the future of all of this can be applied to. You've talked a lot about history, and certainly, John Wolfe and other physicians that you've come in contact with that have heard those sort of signals in the noise, heard that little bit, and then acted on it over time and how it takes.

How is history going to look back at the year 2025 in medicine? There's also so much that can go wrong in the future with algorithmic takeovers. What makes you optimistic that we're going to get things right in the future?

CONNIE LEHMAN: I think it's exactly your point. Who are the leaders? Who are the people? And what visions do they have?

So there's an incredible, incredible woman, Fei-Fei Li. If you haven't read The Worlds I See, highly recommend it. It has so many layers of the human story. So Fei-Fei Li, her parents figure it out, as they were living in China, she is so smart. Let's get to the US, and let's have formal education for her.

Her father came over as an immigrant, did everything he could to figure out then how to bring his wife and daughter over. And Fei-Fei arrived when she was 16 in New Jersey. And she has become the world's leading expert in this domain of computer vision and AI.

Fei-Fei was the one who developed ImageNet. And that was what-- the huge jump in our ability to leverage deep learning and computer vision was because of her incredible hard work and diligence and brilliance in designing and developing ImageNet and having it available for others to use.

She is someone in her personal journey who also was always looking for how can you use this technology for good? And particularly in healthcare, how do we use this for good? So when we have people like that who are so resilient, who are so brilliant, who are so visionary, others want to join them. And that becomes the inflection point, that becomes where we all then want to follow, and where we also want to lean in and lead.

So I always come back to-- and I was in DC recently talking to a lot of people about our technology and how can we have it more accessible to more people because there's quite a gap between FDA authorization and then payment. And I was a little nervous, honestly, spending all that time on the Hill and in human health services.

I came back feeling so positive, which was a surprise to me. Because you hear so much about the discord and the problems and nothing's getting done. There are incredible people who are smart and focused, and they're in those domains because they want the world to be better. Maybe they want it to be better in their state, in their district, in their country, but they want things to be better.

And so I think we're going to look back at 2025 and say, that's really the year that it all started to really change, that we really started to see that we were able to move in this direction for better health care outcomes, for the full diversity of patients that we serve in the health care community, regardless of age, regardless of rural or urban or race or ethnicity. And I think we'll look back as that's when we really were able to start to put into practice the incredible scientific discoveries in AI from the past literally 20 to 30 years.

ALEX MAIERSPERGER: It's probably hard as you go through the day-to-day challenges of building a company and going down all of the paths of where to focus and who to help next and who to partner with, all of those things. But I hope you can appreciate and sense the sense of accomplishment and excitement and your place in history. As the way you talk about John and the way you talk about Fei-Fei, there's so many who are saying, Dr. Connie Lehman is that person to them now and will be for the years and decades to come. So, Dr. Lehman, thank you so much. This was such a joy.

CONNIE LEHMAN: That was incredibly kind. Thank you so much, and thanks for all the work you do. I am a new fan of your podcast. I love going for long walks and listening and just learning so much. And you do it in a spirit that really is just-- I'm very grateful for that.

ALEX MAIERSPERGER: Thank you so kindly. Thank you for joining us today. If you'd like to add another vote for cookie dough ice cream, or if you'd like to join us as a guest or leave a comment, please email us, [e-mail]. See you next time.

[MUSIC PLAYING]

Wouldn't it be-- wouldn't it-- wouldn't it-- wouldn't it. We can do-- we'll do two takes of it. Do I need to say The Health Pulse Podcast, or just today we welcome? Sweet. Two and done.