Good morning, and welcome again to the 2026 Jefferies Healthcare Services Conference. I am Brian Tanquilut, Healthcare Services analyst here at Jefferies. Our next fireside chat is with RadNet, and with us today is the company's Chief Strategy Officer, Greg Sorensen. Greg and I have known each other for about 6.5 years, since RadNet bought his company called DeepHealth. Maybe Greg, I will pass it to you. Just share with us what that journey has looked like over the last 6+ years, and what RadNet is today in your mind.
Yeah. Thank you, Brian. It is great to be here with you all. Thanks for the attention. It is exciting to be at RadNet. As Brian said, I started DeepHealth, and RadNet acquired it in the summer of 2020, June of 2020. That was back when DeepHealth was just focused on mammography. As you probably know, RadNet does about as many mammograms every year as the country of England, so it is a big part of our business at RadNet. Since then, AI has exploded in impact and RadNet has grown substantially, and that combination is why I am still here. I was just whispering to Brian that my wife is like, "Greg, why aren't you retiring? You sold your company. Let's move to Paris." I just cannot see a better place to have high impact in medicine than RadNet right now.
Radiology is the most digital of all of the medical specialties. We were digital in the 1990s for lots of reasons, and it is deep in the DNA of our specialty, more than pathology, more than anything else. The opportunity for AI to transform medicine is going to come right through our specialty. When you couple that with the impact imaging can have in population health, in urgent care management, in inpatient management, it just touches everybody's lives. As you all know, 90%, 95% of everybody who goes into an ER comes through the radiology department. Lots of outpatient imaging that is growing. Lots that is not being done by radiology, in OB and in med, and all of these things say to me that the opportunities to improve healthcare and have impact on patient health is, and coupled with the innovations in AI, is going to happen in radiology.
The scale that RadNet brings that allows us to get data and therefore train the AI, deploy it, figure out where it is working and where it is not working, and iterate, that sort of flywheel of innovation cycle, I just do not see any better place to be. That is why I am still here. It has been a great six years, and we are a much different company than we were six years ago with a huge vision, and there is lots more to come.
So maybe I'll hit on that last point you made. What does RadNet look like today? If you can walk us through what the digital health side looks like and what is the core business today?
Yes. RadNet, as you all know, or as many of you know, reports two segments today. We have an imaging center business that's over $2 billion in revenue, and then we have this digital health business that's taken the DeepHealth name, but really does more than mammography. It does all kinds of AI and other IT solutions for outpatient and inpatient radiology. That's, on this year, we've guided to $135 million-$145 million in revenue. So over $2 billion and $145 million, clearly very different sizes. The reason we break out the digital health business is because of its higher growth rate. We're definitely, I think, in Q2, we grew at 90%. Also because of the investments that are needed in AI. We wanted to make it clear how those investments are returning to those investors.
We also have this services business, which is just doing great. As you saw in our last quarter, we reported 25% revenue growth, and that was on top of an earlier quarter where we reported 22% revenue growth. Of that, I think 10%+ was same-store MRI growth. We had 8% or 9% same-store PET and CT growth. So the core imaging business is doing fantastic and well. The AI business, if you will, or the digital health business, has actually been the enabler for a lot of that. The reason that we had 10% same-store growth in MR wasn't because the demand hadn't been there before. It was because the digital health tools allowed us to meet the surging demand for radiology by opening up slots that our techs could now scan, scheduling more efficiently, speeding up the exam time, speeding up the reporting.
Essentially, that demand is still out there for even more. It's can we hire the people? Can we execute in order to meet that demand?
Maybe, Greg, one of the questions we get asked a lot, I mean, this is a healthcare services conference first and foremost, is what does the demand or the volume or utilization environment look like today? I think there are lots of concerns that volumes have slowed down, but you're not seeing that in imaging.
No.
Curious, what do you think is driving all that and the sustainability there?
Yeah. It's a great question, and you're right. When I look at my old employer on the OEM side or other companies, I don't see them in double-digit revenue growth, and yet we've guided to that for this year, and we don't see any reason to slow down. There are a couple of things. There is a lot of new technology in imaging, whether it's cardiac that's growing with things like HeartFlow, whether it's new PET tracers like PyLarify, where we just didn't have a way to monitor prostate cancer before, and now we do. Whether it's the increased recognition of screening as a way to lower healthcare costs, whether with lung cancer screening that's really shown stage shift in the ability to improve not just health, but the cost of that healthcare by diagnosing cancer earlier.
Or just the demographics, that patients are getting older, and as they get older, they get sicker, and they need more imaging. All of those tailwinds are really pushing to boost the demand for imaging broadly. When you add to that the lower cost, the cost pressures that say, "Let's get out of the high-cost locations, inpatient or HOPDs, to pure outpatient settings." As I think you've seen us talk about, Brian, we've now got three mammography locations in Walmart, trying to get the costs even lower. The desire to get to low-cost settings is universal, and that's a big tailwind for us. None of those seem to be slowing down. I think they're very sustainable. The Trump administration is talking about site-neutral payments.
I don't know that they'll be able to get that through, but whether they do, which would be a huge tailwind for us, or whether they don't, and it's just the payers who keep pushing to go to lower cost settings. The reason I'm very bullish on RadNet specifically is we already are in the lowest reimbursement states today. That's our base in California and some of the East Coast. We have not the richest mix in terms of advanced procedures. We're still under 30% advanced procedures, so we can only get better. We can do more advanced procedures than we're doing. We could move into states with better reimbursement than we're getting. And that's, I think, how margin expansion, even if we just keep doing what we're doing, we're just in different geographies, I think we are going to have an opportunity for margin expansion.
Greg, just to that last point you made about the remaining runway, is that what gives you guys confidence that this growth rate that's in the high single to low double digits, kind of organic-
Yeah
Is sustainable for the next several years?
Yes.
Is that the right way to think about that?
It is. I think it's very sustainable. For sure, same store growth we see, we're building de novos because now that we've got the quality up with these digital tools, the demand is very high. Another key to the demand driver is the JVs with the hospitals. It helps them reduce leakage. It helps us ensure that demand is consistent. I do think that these numbers are not a flash in the pan. They're quite sustainable.
All right. Let's shift to digital health, which is
Great. Yeah. Something I actually know something about.
I was going to say something that you're very passionate about.
Yeah.
All right, so maybe let's start with, maybe if we can describe to the audience, because we get this question a lot. Can you break down what the digital health segment looks like? Or what are RadNet's digital health or AI assets? Because there's a lot of them.
Yeah. There are. I think one easy way to think about them is kind of in a two by two. There's clinical AI, which is sort of the pixel stuff that I used to do, and then there's more operational or kind of RIS/PACS, general purpose healthcare IT. And right now, those are about 50/50 in terms of the amount of revenue. The other is internal and external, inside RadNet and outside RadNet. And today it's about 40/60 or maybe 35 RadNet, 65 external for both of those. So you can think about these assets. If you're just talking about the assets, are you talking about things that look at the pixels? That's mostly the clinical AI, or are you talking about moving images or scheduling patients or doing revenue cycle management or kind of reporting? That's more the, I would say kind of the operational radiology software.
Both of these are growing rapidly. We're investing in both of them. Very simply put, the generative AI stuff is probably helping the reporting and the non-pixel AI even more than it's helping the pixel AI stuff. The hyperscalers are not really investing so much of their benchmarking on pixels, on pictures. Although they do with some of the image generation stuff. More of it is on words, and so all of the scheduling and all the tools that it takes to generate a report, to summarize old exams and prior history so that the radiologist can quickly know what to say, that's all on the gen AI side, and our tools are leveraging that dramatically. So on the pixel side, there's probably, I think now 27, 28 FDA cleared products because if you analyze images and provide useful information to a doctor, the FDA asserts regulatory jurisdiction over that.
If you're just manipulating words, generally that's not FDA cleared or it's a class one device, and so the regulatory path there is a lot lower and easier. The exception there is so-called draft reporting, which is very hot these days, which is a combination. You pour the images in like a CT scan, and out comes a full-blown report. And of course, the challenge is that shouldn't be a hallucination. It should be actual, real data. And that's where the regulatory rub is, that there is no yet full draft reporting product by anybody on the market. We've, I think, made one step on there that we're about to put an article out, but it's public.
We got an updated FDA clearance for one of our chest X-ray products, and the claims are very similar, but the engine is a so-called VLM, a Visual Language Model, as opposed to just the simple sort of standard convolutional neural network or CNN that these tools were based on before. That is, I hope, kind of a little bit of an overview of what the assets are. If I were to now to link those back to RadNet, we use almost all of those tools inside RadNet, and that is what has allowed the speed up of the productivity and the same store growth beyond and as to meet that demand, and happy to double click on a couple examples if that would be helpful.
Yeah, I would love to double click first on the clinical side.
Sure.
What are the products that you have both FDA approved and are pending?
Yeah.
And what is reimbursable or how do you monetize these products?
Great questions. One of the things, I used to be an academic, and when I was an academic, we survived off grants, and it was just about being clever. Now that I'm at RadNet, you have to be clever in a different way. You have to be sustainable, and that means you have to somehow be paid for what you're doing, or you can't keep doing it. There has to be a revenue source. We've really focused our clinical AI on what's sustainable. How do we get paid? There are now, I would say, two big categories of ways we're getting paid, maybe three. For sure, the easiest to track is new revenue. There are now Category III CPT codes that allow us to bill, and older Category I codes that also allow us to bill for using these software tools.
Probably a single shining example that's great to talk about is our thyroid ultrasound product, and now just most recently, we've got the breast ultrasound product. Both of these are tools that take what used to be a very complex, manually heavy process where a technologist would fill out a worksheet with diagrams and numbers, and then the radiologist would take that worksheet and dictate it into a report, just explaining what all the figures were. Now AI does all of that. It creates the worksheet, it moves that worksheet into the report for the doctor to review, and it does that by looking at those pixels, and old exams, and then comparing, and then creating a draft report. Thyroid ultrasound used to be the bane of our existence, I'm a radiologist, as we radiologists, because we had to do all of that really boring drudge work.
It was slow, and the RVU per hour rate was not great when it was slow. With AI, that's completely inverted. Our radiologists accept those reports 94% or 5% of the time, and there is a Category III code that lets us get paid about $60 and about, well, a little more than half of our payers are willing to pay us that. The cherry on top is that because the techs no longer have to fill out all this worksheet, we can schedule them every 20 minutes exams now instead of every 30 minutes. So for the same tech labor and the same ultrasound box and the room rent and all that, we're getting an improvement in productivity. That AI tool really pays for itself through all three ways.
We get new revenue, we get a physician productivity win, and we get a technologist and room productivity win. In other cases, we have self-pay. We've talked before about the enhanced breast cancer detection. This year, we've gotten breast arterial calcification off mammograms cleared. We've submitted the risk product, for determining what the likelihood of breast cancer is, not today, but in three or five years off of a mammogram. I expect clearance of that by the end of the year. We think that will boost the number of women who pay the $40 from 50% to probably 60% or 70%, as well as be a really good public health good by figuring out who actually needs more. There's not a Category III code for that yet, but I know that's in the works. People are working on a Category III code for breast arterial calcification and for risk.
We see ways to get paid for that, and so that means our customers, of DeepHealth's customers, can get paid as well, and so then they can buy the AI. We see a real growth opportunity in the clinical AI side of things. I think we've disclosed publicly that the mammography self-pay is over $15 million, I think approaching $20 million a year of added revenue. The T code stuff, we do 250,000 thyroid ultrasounds. We're now getting paid on many of them. We do over 600,000, maybe it's close to 1 million this year, breast ultrasounds. Now with that, we'll be able to get the T code on many of them. So that's added revenue on top of the work we're already doing for doing the work more efficiently.
I think it's the kind of thing that our colleagues in other academic medical centers and other outpatient centers are saying, "Yeah, why wouldn't I want to do this? We'll buy your AI from you.
No, that's great, Greg. So maybe shifting to the other side, or the whole suite.
Yes
You've really built up that team within the digital health segment. If I'm not mistaken, you have what, close to 600-
Yes
Employees on that side of the business. What does that look like in terms of translating those investments into revenues and also driving external sales?
Yes.
Because right now you said it's 65/35.
Yeah.
What's the expectation there?
I think that external growth is going to continue. We've got 3,000 customers outside that like these products, love these products, and most of them don't have the full suite. The classic SaaS model of land and expand, I think is still possible for DeepHealth. That was why we did the Gleamer acquisition. It wasn't just to get their pixel AI around X-rays, but also, all of those are customers that we can now sell these other reporting tools in. There's a whole suite of non-pixel operational software tools, and those range from everything from how to schedule patients, to checking them in. Now we're building tools to allow people to check in on their phone, so you have fewer front office staff members, to scanning.
We've talked before about TechLive, this tool we've built that allows a technologist to do an MRI scan while they're in Arizona, but the patient is in New York, and even run three scanners in New York while they're sitting in Arizona in their pajamas. They love it. We can hire better techs. They'll work weekends, they'll work nights. Where we couldn't get labor and couldn't keep the scan rooms open on weekends, that's part of that 10% MRI same-store growth is the MRI was there, we just couldn't get the labor, in New York. But now we can hire them in Florida or wherever. That's a tool that a lot of our external customers are salivating over. Then there's the full reporting suite for the radiologist.
The viewer, as everybody in the radiology world knows, PowerScribe end-of-lifed as of this month, and so you have to replace it with something. Better speech recognition software and AI tools that allow the doc to just dictate part and then have the AI fill in the rest, these are very powerful and are helping the docs become more efficient. We've developed a suite of those tools all the way, and then at the end of revenue cycle management and billing. That whole full sort of OS, if you will, the radiology operating system, is now being built and being deployed across RadNet and is being offered for sale and is being purchased.
I think when we announced the Trinity partnership in Idaho, we also announced that they were one of the groups that has decided to move over to our operating system for their full operations. We see a lot of synergy between the pixel AI and the OS, if you will. In fact, really, going back to where I started six years ago, when we brought in the mammo AI, we realized that unless we had the operating system, we couldn't really recognize the full value of what the pixel AI was doing. Just to give one last specific example there, the real value of the pixel AI on the mammo front isn't finding the cancers, it's bringing a second doctor in on the trickiest cases. That's all workflow. All that workflow solution means when do I call the second doctor?
How do I make sure that if the patient didn't pay, a second doctor isn't called? How does the first doctor and the second doctor interact? Does the report get paused while the two doctors confer? How do they resolve what they're going to do? All of that has nothing to do with the pixels, but has everything to do with the workflow AI that was enabled by smart pixel AI. We needed that team, and I didn't bring that. I had a team of 12 really smart scientists back six years ago. Now we're 600 because we need all of that really clever workflow. That's what's transforming medicine. It's not just the AI, it's novel ways of leveraging that AI to work more efficiently and work more intelligently. I think that's going to happen in other specialties, too.
Greg, when I think about the differentiating factor for RadNet on the digital health side
Yes
I think it's that, at least from where I sit, I see a unified offering that no one else has.
Correct.
Maybe if you can just walk us through what is that pitch like to a health system or a clinic?
Yeah. The pitch, you always have to meet your customer where they are. We don't want to overwhelm them by saying we have this big monolithic thing. We say, "What's interesting to you? Let's get that started." If that's draft reporting or faster reporting or faster viewing, great, we'll start there. But then we try to open the kimono and let them see the bigger vision, and see that you can buy a piece of the operating system and use it. But in fact, there's this unified data stream and unified operating system that allow you to have efficiencies and cost savings and productivity and quality improvements, all that work together, that you can expand into at your own pace. We're seeing that.
Cleveland Clinic is a customer of ours for the mammo AI, but now they're looking and saying, "Oh, well, look, we could actually get our workflow better if we had this tool or that tool, or we could scan our patients remotely." It is back to that kind of land and expand model. We do not want to land with the full armada. We want to say, let's see what's of interest to you. Where does your budget allow you to start? Then you can see that just like RadNet needed all of these pieces to move itself into the digital era and becoming the modern AI tech-enabled services company, you could become a tech-enabled services company too, by picking up whichever of these pieces work well for you.
That has meant we've had to do more interfacing, which is, I think, a little annoying sometimes, but Epic is Epic. You have to figure out how to work with them, right? That's life. Again, we kind of try to meet people where they are and interface with what they've got, but then once they see the vision, they tend to say, "Ah, okay, more and more of this makes sense for me.
Maybe just to double-click on this conversation. What you described sounds more like health system focused than your legacy-
True
Focus on-
Yeah
Outpatient clients, right?
Just pure outpatient stuff. Yeah.
Is that sales process different? Where do you stand in terms of trying to gain more clients on the hospital side?
It's an excellent question, I think there's some synergy there. RadNet, Dr. Berger has said over and over, we're at 30%+ JV partnerships for our 442 imaging centers as of the end of the quarter. 36%, I think, was the number. He'd like it to be 50%. He'd like it to be more. What that's a signal, you're right, that's a little bit of a change. I think maybe the older Dr. Berger or the earlier Dr. Berger was a little more independent, I think now he's seeing partnership is the right model. These hospital systems can be very good partners. They are for us, there's a win-win here to be had. Let's work in a way where we can focus all of this energy and excitement around AI and imaging, let you, hospital system, focus on what's important to you.
It's probably not imaging. Or you have so many things, why don't you work with a best-in-class partner on imaging, then you can focus on whatever it is you want to focus on, your inpatient products. That is a shift. It's a shift for how RadNet approaches its JV partners, and I think it is driven by the digital health story. The digital health story is definitely going straight to hospital partners and saying, "Look, you're the majority of the market. We know we need to solve your problems." We've solved some of them for RadNet, but they're in an outpatient setting, but you can see that there are inpatient versions of your problem, and some of them are the same. The remote tech stuff is the same. The thyroid ultrasound is the same. Some of it's not the same.
RadNet doesn't do much in the way of acute care, and that's why Gleamer's products for fracture detection, if you've got a fracture concern, you're in an ER, you're not at a RadNet center. But that's why we've started moving more into these areas, these domains that are important to inpatient. I don't know that I've seen good stats on this, Brian. I think if you look at all of imaging, it's roughly 50/50 inpatient, outpatient. I think that's kind of the number I carry around in my head. But even the acuity is different, too. I think we want to impact patients wherever they are across the spectrum, and that means we can't just stay outpatient-only focused, especially on the digital health time. I don't think we'll build all the workflow tools and revenue cycle for inpatient. That's not our expertise.
But there's enough commonality between how a radiology department runs in a hospital and out of a hospital that I think much of what we're building really resonates with those clients.
Even, Greg, just to think through that, right? There are providers out there when I think of ProMedicus-
Yeah
I think of Sectra, right, that are in the RIS or, sorry, in the PACS side.
The viewer business. Yeah, the PACS.
Of your business.
Yes, exactly.
If I'm a hospital looking at your product versus theirs, say I'm walking the RSNA booth
Yeah
What would pull me to DeepHealth's OS product versus, say, a Sectra or a
Yeah
ProMedicus-
A Pro Medicus or something. Yeah. I think the Visage product, look, it's in my old hospital where I used to work. It's a well-accepted, well-liked product. In that sense, we could bring a viewer and compete with them, and I think we do have a competitive product there. But that's not what I think people come to the DeepHealth story, if they've already got Pro Medicus on board for. It's that full OS approach. It's not just one thing, it's kind of the whole thing, if you will. Now, what's of interest to individuals when they walk the RSNA booth, they're usually looking for a couple of things. They're not looking for everything. I think that means we need to be flexible enough to offer what they want, where they want it, and to be best-in-class in multiple things.
I think the good news is we're doing that. We've got the pressure. We think our radiologists who have moved from PowerScribe to the new reporting tool say it's better than anything they've ever used and better than anything they're hearing from their friends. Our viewer is as fast as Visage. Is it going to be as integrated in 3D and installed? No, it's not installed in those places, so we can't go to a customer and say, "You should rip out what you've just spent and have a seven-year contract left for." But you can put ours right alongside it, and if that's the reporting tool or that's the clinical AI tool or that is a scheduling thing that helps you run your facilities better, or it's TechLive because now you can. Fine.
I think that's the fun of being at a full service sort of OS provider is we really have a unified vision, but you don't have to buy the vision in order to love the individual products.
That makes sense. Maybe in the 2.5 minutes we have left, I've got a couple more. Number one, capital allocation.
Yes.
You have a lot of cash on your balance sheet. You obviously have looked and done a lot of acquisitions
Yes
On the digital health side and also on the clinic side. How should investors think about capital allocation and M&A going forward?
Active. RadNet's traditionally done a dozen plus a year, and we see a lot of opportunity. I do think discipline, though. You've seen Dr. B. and others, they really hate overpaying for assets. I do think that we're going to do more acquisitions in both sides, but disciplined, valuable ones. I get calls literally every day from people who want us to acquire them, and we just have to find the right ones that we can convince ourselves and our shareholders and our stakeholders that we can have 1 + 1 = 3. We can pay them what they want, but that we can turn it into something even more valuable now that RadNet owns it. We do see those. You'll hear more from us on that front this year. We see new geographies that we can consider. We see actually adjacencies to what we're doing.
Radiology is a very dynamic area on both the services side and on the digital health side. That's part of what's fun. When I first started following RadNet, we had a leverage ratio of, like, six. And now, with a leverage ratio under two, and a stock that's a valuable asset to do acquisitions with, I think we're going to be very active there. It's one of the things I've actually admired about Dr. Berger. He's a really smart guy when it comes to deal-making and acquisitions, and has deployed the capital well.
All right. Since you're a radiologist, one minute left. What would radiology look like five years from today?
Oh, my wife and I argue about this all the time. I'm convinced that we radiologists will still have jobs. She's like, "How could that be?" Given not just what Geoffrey Hinton said in 2016, but just the amazing pace of AI. I think it comes down to Jevons paradox. I think that the cost of radiology is going to drop, but the value of radiology is going to continue and expand, and as a result, the demand for radiology is going to explode. Every doctor I know wants more imaging. They want it in the OR, they want it in the ER, they want it in their office. If you could get reports instantly, they would be ordering more imaging. 75% of admissions go through the ER. They're all too slow. Outpatient scheduling, it all needs imaging.
I think radiology as a specialty is fine, but I do think that less and less will be done by us radiologists and more will be done automated.
Awesome, Greg. Thank you so much.
Exciting to be here.
Really appreciate your time today.
Thanks so much.
Thank you.
Thanks for the time.
Thank you so much.
Yep. As always.
Nice to see you as always.
Yeah. No, you bet.