Transcript
Speaker: Welcome to another episode of Crossroads by Alantra, where we explore the cutting edge of digital health innovation. In this session, we have the privilege of hosting Elad Wallach, the CEO and founder of ADOC, a trailblazer in clinical AI applications.
Speaker: Elad will walk us through the story of how he and his three co-founders set out to tackle one of healthcare's most persistent challenges, reducing medical misdiagnoses, which remained, unfortunately, alarmingly common to date.
Speaker: Allad will also shed light on why many health systems struggle to integrate multiple best-of-breed vendors, challenge that inspired ADOC to develop comprehensive platform capable of seamlessly incorporating both their and third-party algorithms.
Speaker: We will close out this session by diving through the role the technology titans are playing in the clinical AI ecosystem. We hope you find this interview as enlightening and insightful as we did. Welcome to Crossroads. Welcome all to another episode of Crossroads by Alantra.
Speaker: As part of our coverage of the radiology technology segment, we have the great pleasure of hosting Elad Uvalak, CEO of ADOC. Elad and three other members of the Israeli Air Force Program, Talpia, co-founded ADOC in 2016.
Speaker: ADOC has developed the medical imaging diagnostic platform, which has received 17 FDA clearances to date, the most in the industry by far. It is used for radiology, cardiology, neurology, and vascular applications.
Speaker: Joining me today on this episode is Deris Kudo. Deris is an associate in our healthcare care practice and is jointly covering the radiology software sector with me. Thank you both for taking the time today.
Speaker: I really appreciate it. Likewise. Excited to be here. Thank you for inviting me. Coming out of the Israeli Defense Force, what was the genesis to create a radiology AI vendor? And tell us about the early days, please.
Speaker: Yeah, well, first, maybe I'll start by, I think we're well beyond a radiology eye vendor at this point, I would say. So we view ourselves more as a clinical eye vendor, but we can get more to that later.
Speaker: My dad was actually was a scientist and he worked at IBM Watson. He was one of the originators that pushed IBM Watson to go into health. I'd been once on health and he was pushing for it.
Speaker: When went back home after like a week and in the service, then he would share with me some of the things he saw in the care gaps. I always had this bug to go into healthcare. care And I was very fortunate to meet my co-founders. so we were in the same program, as you mentioned, in the Ministry of Defense.
Speaker: Immediately upon finishing your service, which was around the same time after nine years, we all had this shared passion of like really doing something together, opening a startup, and going into healthcare.
Speaker: But as you can imagine, having nine years, I've served in the Air Force, headed AI department there. ah you know, really nothing about business, nothing about healthcare, care and definitely nothing about the US healthcare, I would say.
Speaker: One could ask, why did we even go? Or how do we go on this journey? And basically the first year upon finishing the show, we spent a bunch of time in hospitals. And a lot of it it was in Chiba Hospital, which is the biggest hospital in in Israel, but also great US hospitals, really to trying and see the problems they had.
Speaker: it sounds like it runs the family. So your dad was at IBM Watson. and That is very, very interesting. That's a fun fact. I truly believe we have amazing people in healthcare. care We have some of the smartest, most passionate people that are trying to do good.
Speaker: But unfortunately, we lack the processes and tools to handle the current realities of the u US healthcare. read this Hopkins study from last year that they've estimated that 10 to 20% of all care decisions are wrong.
Speaker: or delayed. And their estimate was that in the US today, there are about 400,000 deaths every year due to misdiagnosis.
Speaker: It's shocking. 400,000. That is the third largest cause of death. Number one, cardiovascular disease. Number two, cancer. But number three is this, not by a wide margin.
Speaker: And guess what? This is actually preventable. So that was a eureka moment for you? i think we've realized we have to do something about the care gap that we have. We have to develop better tools for physicians.
Speaker: And that was a eureka moment where we said, look, we have to help health system utilize your data better and kind of started us a journey for the clinical AI platform. Yeah. Early days, my understanding is you really focus on radiologists and the caseloads.
Speaker: They were just overwhelmed by the caseload. Was radiology kind of your initial foray or from the get-go, you said, I need to fix this more comprehensive issue, which is pretty much ah across specialties, right? It's not only about radiologists.
Speaker: If you look at the vast majority of how the AI market is built, because it's so complex to build AI solutions, it is really built disease by disease. Health systems are very slow moving enterprises. They cannot engage with 50 vendors to solve 50 diseases.
Speaker: So always the ADOC thesis was, hey, let's build this comprehensive platform. Let's give health system a way to scale and not just solve one disease, but actually do big strategic impact across their whole portfolio.
Speaker: We certainly see the same thing across a lot of the folks in radiology, AI, IT, where you know all the demand drives are there, but health systems are slow moving. There's a lot of mistrust in AI by providers.
Speaker: think the value of AI goes beyond just algorithms or any platform. In the past, you've said that it also involves changes in process. I would argue that a lot of mistrust by providers comes from bad prior experiences and process changes, not any sort of model performance or performance of a platform.
Speaker: How are you addressing these kinds of barriers in your product development, commercialization, etc.? The way we we look at the problem space, we basically divide a good product into three layers.
Speaker: The first layer is what a lot of people talk about. That's the algorithmic layer. It has to be accurate. Otherwise, it's not good. But let's say even if you the best algorithm in the world, doesn't mean anything, right? So you have to do the step two, which is workflow integration, which is what do we call the you build from an algorithm, you build a product.
Speaker: If it's that first layer of the algorithm, we measure sensitivity specificity, and that second layer, we measure engagement. Like if you give it to physicians, would they actually use it and engage with it on a daily basis?
Speaker: But even that is not enough because that doesn't mean it provides value. It could be the cool thing in the world, but is it really moving the needle on care? And that's why we believe in that third layer, which is what I would call moving from a product to a solution.
Speaker: That you have a clear identification of a problem. You can change the workflow to solve the problem, and then you can measure the outcomes on that outcome. And that, I think, is where a lot of companies don't look at the full picture.
Speaker: But for example, one of the things we found is that the change management for the hospital across the workflow is very meaningful. And I'll give an example. We had a stroke implementation at Ochsner.
Speaker: We ran with them. They've had another stroke vendor. Really what they wanted to do is reduce time to needle. So they had a clear problem state. But for years, they were stuck at about 100 minutes door to needle, even with this AI.
Speaker: And then we came in and they published on it. Like the door to needle time dropped by 40 minutes after implementation of ADOT. That's a lot and massively improves care. So honestly, I've asked my team,
Speaker: What's going on? Like, how did we do that? And what they told me is, look, Alad, we actually think, yes, the product is great and a great engagement. But the main impact was around the change management. It's because we went in there with the amazing Oshner team and we've mapped every step on the workflow.
Speaker: And we've meticulously mapped what needs to happen differently and then how the AI drives a workflow change. It's not just a product you stick on top of an existing workflow, but you actually drive a different workflow.
Speaker: That drove the needle much more meaningfully. That is an example of how it's really not about you could have the same product, but by driving change, ah you can make a much bigger impact.
Speaker: Maybe one follow up to that. Health systems are known to be very under-resourced today. They have a limited set of things they can do as it relates to any sort of technical implementation process changes.
Speaker: How much learning, hand-holding do you have to do for them We have dedicated teams, hold multiple teams that are dedicated to change management that we want to support our partners with.
Speaker: We definitely need to invest a lot of human capital in it because of what you mentioned. I think health systems are really stretched thin. The good news is that despite me saying that change management is so important,
Speaker: The amount of work needed for change management with AI is not as significant. To some extent, AI is the help, is the enabler for much easier change management because they don't need to remember all the guidelines.
Speaker: You don't need to rely on 10 physicians remembering what's the best practice. You have this AI that drives the workflow in a much more meaningful way. And the outcomes are pretty amazing for that. That is why we think it's so transformative because it allows health system to do the care they want to do, but just enables that change.
Speaker: Our understanding is payer reimbursement in both the US and Europe, public-private payer environments, they've had their ups and downs in terms of reimbursement. Only a few algorithms are covered today, is our understanding.
Speaker: How will you make your case to payers? In the US, I'll say to the US that you need to adopt this. This needs to be ah ubiquitous. I will tell you, most successful AI companies we're seeing today have built solutions that do not rely on reimbursement.
Speaker: that have direct ah ROI to the provider system. It slow moving are a lot of difficulties in creating reimbursement for AI. Maybe I'll share one story for me that was really telling about the US healthcare ecosystem.
Speaker: So very early in our journey, I've met with one of the executives, one of the big payers. And I remember I told them we haven't developed such a product, but I told them, hey, we could essentially develop a product that like finds more lung cancer and helps with the follow up of those patients and effectively can find a lot more cancer a lot earlier.
Speaker: Would something like that be interesting to a payer? He was saying that, look, a lot, I will have to pay more for the workup if I find cancer. Yes, but we all know that you find cancer early, right? You save costs later you reduce total cost of care.
Speaker: That's amazing. But then he told me something that scared me and like really made my eyes open on the US healthcare. He told me, Alad, you have to understand, I only own the patients for two to three years.
Speaker: I do not have any financial in incentive to what happens to them after that. That's crazy. Think about what this means. It means that there is no entity, apart from potentially CMS, but that's a whole nother discussion, that optimizes for the full patient care.
Speaker: Or lifetime. Or lifetime. That means that me as a company has to develop things the market will buy. And therefore, I will have to do things that have at least some sort of ah ROI in a fee-for-service environment direct to provider.
Speaker: So I think there is immense potential in reimbursement because we can then actually have people optimizing for the full lifetime of a patient. But, there's a big but, it's going to take time.
Speaker: So staying on the topic of you finding more disease earlier, acting on it earlier... One of the prerequisites for this is more ubiquitous adoption of AI across therapeutic areas, disease states, not just individual point solutions.
Speaker: From an R&D perspective, one of the more impactful trends that we've seen surround the development of foundation image models, similar to what OpenAI has done in the NLP space. You just announced a landmark partnership with AWS on your CARE Foundation model, which you recently unveiled late in 2024.
Speaker: Even without any transfer learning, specificity and sensitivity over your algorithms rival what you already have with this more generalist kind of model. And of course, you can take this and rapidly speed up development of higher accuracy, you new point solutions.
Speaker: What does your work with Foundation models, and now that includes Amazon, mean for the sector? I will tell you, I think it's one of the biggest inflection points I've seen for the whole industry.
Speaker: There are thousands of diseases. And the question is, how do we increase the coverage? And the step limiting factor is that developing AI is slow and expensive.
Speaker: Even for us, and we're probably the fastest in the industry, it takes a year to a year and a half to get one FDA clearance. we had to think differently about the whole problem, which is why the foundation model is so transformative. The foundation model is a piece of technology that can identify many, many diseases all at once.
Speaker: And therefore, if you want to develop more use cases, now instead of taking a year and a half to develop them, you can do it in a week. And this is not an exaggeration. Like we actually have. And one of the things right now, we submitted to the FDA product that was practically built in a couple of weeks, obviously based on a foundation model and our validation, et etc e cetera, et cetera.
Speaker: It is very transformative because finally we have line of sight to really much broader coverage of disease states. And I think the impact on patient care is going to be immense because it can literally touch like 100% of what a health system does in a few years.
Speaker: So as you said earlier, this industry is very fragmented. There's over 500 imaging applications out there, almost as many vendors. Many of them have to rely on marketplaces for commercialization. They can't offer the same platform approach as you because ultimately the customer wants a full solution suite, as you said, and more than any of these individual vendors can offer on their own.
Speaker: You have a marketplace today. Now you have your foundation model, which could eventually speed up development of a lot more disease states. Do you think there's going to be long-term value in these marketplaces? Or will the space eventually be consolidated by those that can offer that kind broad single vendor platform?
Speaker: I firmly believe in an ecosystem. I don't believe that one player can do it all. And I think there is going to be a lot. There is a lot of that. We see it today and we're committed to this ecosystem vision.
Speaker: The problem is, i agree, is consolidation. Like, how do I have my singular layer? But in the market, you might see two different approaches. One of them is what you'd call the marketplace. But the marketplace as it stands today is a typically very shallow integration.
Speaker: The big value of the marketplace is... the ability to have basically one contracting vehicle, right? That is not what we do. If you onboard to the AIOS, which has advantages and disadvantages, it's a full stack thing. It's the monitoring, it's the analytics, it's the workflow integration.
Speaker: Practically, you don't care if it's ADOC or not ADOC. It's ah the same unified experience from IT perspective, from a physician perspective. That is very difficult to build. We spent over $100 million dollars in building the AIOS.
Speaker: We spend millions of dollars per vendor for onboarding. It's very expensive because we want to create the simplest experience. But that means we have a vetted ecosystem, right?
Speaker: It's high value solutions. It's trusted partners. There are people that are vetted. So it's a slightly different offering. i obviously believe in the power of a platform to really scale it.
Speaker: And I'm guessing the added benefit is to be on your platform, the monitoring, especially in the face of heightened regulation, ah would make it easier on the health system to comply with regulatory burden, and right?
Speaker: That's definitely a benefit. so Two more questions, more on the the overall landscape, and especially around the big tech, the Microsofts, the Googles, the Amazons. Of course, we just mentioned your partnership. Congrats, by the way, for the achievement.
Speaker: But we feel at Elantra that they should be lured by the market potential. Also, the untapped nature of clinical AI, all of the above. What we've noticed is they were first to market in terms of pattern recognition, deep learning, you know, Microsoft bought Nuance.
Speaker: They've partnered with Viz. Amazon, we just mentioned it, partnered with you. Google appears to be doing some research in the sector. you feel this is the start of a more concerted push by the big guys in terms of being more present in clinical AI? Yeah.
Speaker: I sure do hope so. Look, healthcare data, I've recently read up research saying that in a few years is going to be 30% of all global data, all data in the world, right? It's pretty spectacular. So there is a lot of potential there. And I think health systems is definitely an untapped market and that has a lot of potential.
Speaker: And I do really believe in the hyperscalers potential. Our partnership with Amazon has been fantastic in our ability to scale. If I go to a health system today and I want to grow them from one use case to like 30%,
Speaker: Guess what? There is no on-prem infrastructure that can do that. It's like not a thing. You have to find these ways to really scale. I think we'll see more and more of cloud adoption as well as broader partnerships with some of these hyperscalers for sure.
Speaker: Outside of the big tech, you also have the the big OEMs, like the big manufacturers, the incumbent in terms of healthcare IT vendors, the Epic, the Oracle Health slash Cerner, right? And you got the pharma biotech.
Speaker: We have not seen them as much creating diagnostic algorithms or clinical AI. They would stand to gain from clinical AI. and Why do you think they've struggled or been reluctant to invest in diagnostic AI? Yeah.
Speaker: I think developing clinical AI is is really, really, really tough. There is a reason why vast majority of the market is built on point solutions. You need like literally a whole company to identify and build a workflow of one disease.
Speaker: The data is not normalized. The workflow is not integrated. Developing an algorithm is super tough. Getting to regulation is super tough. Even if you've done all of this, you need clinical studies about the outcomes. You need the ability to then distribute and then integrate into the workflow.
Speaker: It's not for the faint of heart at all. And I think it's just such a dedicated new core competency that it's very hard for people. We have people, just to give you a sense, we have people that are experts at monitoring, that are experts at validation.
Speaker: We people we call experiment babysitters that monitor experiments, AI. You have all these new professions that you haven't even thought about when you want to develop AI. The market evidence is there. And I think that's why you might be saying that some people choose to partner because it's a way more, you know, both cost efficient.
Speaker: Honestly, even Microsoft and OpenAI, right? I think it's ah one of the best examples. Microsoft is an amazing company with some of the smartest people on earth. They chose to partner with OpenAI because they understood that you need to move it different core competence. You need to move differently.
Speaker: Not a bad comparison. Elad, again, thank you so much on behalf of Darius and myself and the listeners. This was great. it was a really great discussion. Maybe parting words from you, Darius?
Speaker: So I appreciate the time, Elad. is a space that we're passionate about, you're passionate about, and it's good to see that you're really one of the trailblazers unlocking this industry and ultimately improving patients' care.
Speaker: Thank you. Thanks again for tuning into another episode of Crossroads by L'Entre, an initiative that strives at bringing together our team of sectorial experts and healthcare care leaders to discuss innovative technologies and themes that are reshaping our industry.
Speaker: If you're seeking to explore your strategic options, our team is uniquely positioned to assist you. Our sectorial expertise and global network is simply unparalleled in the mid-market. To learn more, please feel free to visit our website or contact one of our team members directly.
Speaker: you
