How Will AI Revolutionize Imaging and Cardiovascular Care? | LSI USA '25

Armen Vidian  0:04  
Armen. So first of all, I'd like to thank LSI, Henry Peck, Scott Pantel, and the entire LSI team for organizing this panel today with this fantastic group of investors and entrepreneurs. So my name is Armen Vidian, co founder and CO Managing Partner of Recode ventures. We invest at the intersection of artificial intelligence with healthcare and the life sciences. We invest there because we believe that AI is going to be fundamentally disruptive to CapEx and OpEx. We invest at the seed and the series A I'll first get started by allowing each of my fellow co panelists to do similar introduction before we get started with questions.

Jyoti Gera  0:47  
Absolutely. Thank you. Hi. My name is Jyoti. I run cardiovascular and interventional solutions for GE HealthCare. I've spent about 20 years in GE HealthCare, background in ultrasound anesthesia and now cardiology.

Adrian Lam  1:03  
Hi, thank you for having me on the panel. My name is Adrian Lam. I'm the president and CEO of cor VISTA health. I'm a recovering biomedical engineer and and I, in the second half of my career was in healthcare specialist investing. I was at a firm that actually led Dan Series C at shockwave almost nine years ago, and that was a minute ago, just a minute ago. Yeah, so very honored to be here among giants.

Daniel Hawkins  1:31  
My name is Daniel Hawkins. I am CEO of Vista. AI focused in using neural networks to enable automated scanning for MRI machines. Initial focus in cardiac as mentioned, been in this business a little bit a little bit of time, 30 plus years, not in imaging, but 30 plus years in Medtech. And one of the companies in the past that I was involved with was shockwave medical Excellent.

Armen Vidian  1:59  
So I'll get started with perhaps a bigger question, realizing that first, the core of a good product is solving a problem that's a big pain point for a customer, a patient or a physician, and it really starts there. And AI is a technology which is very interesting and has impacted a lot of corners of healthcare, but still, it kind of begs the question of, why is AI so interesting in cardiovascular medicine, and why do we believe that it has flourished so much in cardiovascular medicine? It seems first, rather than other parts of of medicine. Do we have any thoughts about that from the perspective of both GE and as entrepreneurs? Why that might be?

Jyoti Gera  2:48  
Yeah, absolutely, I can go first on that one. And you started with the unmet need. And the prior panel talked a lot about that too. Right in cardiovascular the unmet need is very clear, the number of deaths that happened, many preventable deaths that happened. So I think the pull has been there, and cardiology, to an extent, when you just think about the evolution of AI in the early days, relying a lot on data. I mean, this is rich with data right, structured, multimodal data sets. Many of them lend themselves also to pattern recognition, to automation. That's sort of where it started, right? And then you were addressing these key pain points of sonographer shortages, sonographer burnouts, because 90% of sonographers in the US actually have issues with MSK repetitive actions, right? So just really hitting upon a very clear unmet need, and I truly believe that with AI in cardiovascular it's not just about workflow or clinical decision support. It will, I think, elevate itself across the care continuum, from screening to diagnostics to interventional all the way through follow up. So the ability of a technology like AI to actually change outcomes is immense in this area. So you see a lot of that pull as well.

Armen Vidian  4:00  
I know that we often look at companies first that have really unique, interesting data sets. That's kind of the prerequisite. The AI itself is, frankly, doing the math on top of that, almost. Yeah, absolutely. Do you believe that those that come to you from the cardiovascular arena have better data sets, higher quality? More interesting is that differentiated versus other areas, or are there similar in that regard?

Jyoti Gera  4:25  
I mean, there are multiple large data sets, but I think for us, what's interesting is when it becomes multi modal, that becomes really interesting because that's now able to solve problems beyond the point algorithms that you can do, and that's where the care continuum comes into play.

Armen Vidian  4:39  
Excellent. Adrian, your thoughts on this?

Adrian Lam  4:41  
Yeah, so I take a slightly different angle to Jody here, not disagreeing, but rather than looking at the solution set and what we can sort of learn from with what we have in terms of data assets. You know, I look at the problem. It is a multi factorial problem. You. Right? And a multi factorial problem requires a multi factorial solution. It isn't really just about tech and specs. You know, as engineers and scientists and physicians, we care a lot about, say, maybe getting eking out that 5% extra gain in sensitivity and specificity a few decimal places on AUC. But really, if you look at the problems, it is about physical access, it is about compliance and adherence. It is about behavioral inertia, right? And I think that's why cardiovascular medicine is so interesting, because the problem is not just to do with that tech and specs. Is a lot to do with that behavior change and so different AI's. You know, we talk about AI as one big kind of umbrella, but there are so many different types of AI is right? And so leveraging the different forms of AI for different use cases and different problems gives the cardiovascular space a really rich opportunities. Set

Daniel Hawkins  6:03  
Daniel, so I look at it as a bit of a mix in some respects. But cardiovascular is very, very complicated, and as a result, the human mind, with lots of expertise, can manage the decision process from the onslaught of data. AI has an opportunity, if trained properly, to optimize the management of that data, right? So I'm on the Mr. Side of of the business in imaging, as I mentioned, if you think about what an MR. Needs to do, you've got to be an expert in assessing, not only the position of heart, the size of the heart. You've got to manage breathing, you've got to manage EKGs, you've got to manage different angles that tissue might present in a in a particular patient, and the disease state requirements to do all of that that's multifactorial by itself. So one can understand that level of problem, computer vision watching that is going to do better than humans. Candidly. Now, let's pop up a level and look at the broad base of cardiovascular an EKG signal. Just give me a strip. Somebody with lots of experience is probably going to pick up 99 and a half percent of what's there, but there's a half point they're going to miss. Well trained AI is not going to miss it right, right? So that com there's a level of complexity and processing that the human mind needs to do, that veering experience is going to give you different results. Properly trained AI, with the right data sets, doing the right math is going to do better than a human right? So there's an opportunity in cardiovascular because of the intensity of data, for computers to do better, and that's why, I think, to a certain extent, cardiovascular is leading the way.

Armen Vidian  7:56  
And where are the physicians with this when we see about how their questions have evolved over the past several years as AI has come along. From your perspectives, both from large companies and from new ones. How have you seen questions change? For example, from my perspective, I know when caption health, which AI guided cardiac ultrasound, which anyone can do in the first pass with the company, with customers. The big question was, will it work? And fundamentally, is this going to even be able to generate a diagnostic quality image? What questions have you seen, and how have they changed become more sophisticated or not, over the years, and what persists?

Daniel Hawkins  8:40  
You know, for the very foundation of AI, everybody started asking, so is this chat? GPT? No, it's not, right. Totally different from that. Then people started to get a little more sophisticated. Are the CNNs? Are they llms? And when they get more sophisticated, and you're working on an LLM, then they're starting asking questions about the data set, the biases that are in the data set, disease, state, geography, gender. There's a wide range of factors that can impact the ultimate result. We're upstream to that in some respects, in at Vista, because we're not doing diagnosis right. But the the the categories that are focused on diagnosis to your earlier point, data sets matter a lot, and the refreshing nature of that data set is terribly important. Now you're getting into a whole new place. How do you refresh is that part of your commercial agreement? Now you're into a place where the hospital ethics committee starts to get involved. A year ago, the notion of an AI focus group or AI committee in a hospital didn't really exist. Now. Now I'm seeing an eight, eight out of 10 commercial conversations I'm having. So there is absolutely an evolution, and maybe not at the speed of AI, but there's an evolution.

Adrian Lam  10:11  
Yeah, you know something that we've seen with in our initial couple of months, launching our product. So we are a we are very much about the core VISTA health is very much about accessible AI in the doctor's office. And the majority of the sites that we go to and we train are actually we're training them on AI for the first time. And so at first there's a excitement. You know, this is cool tech, new tech. How does it work? But then it very quickly goes to the other side of things, the trepidation, the what is the type two error? How does it get? Get things wrong, the explainability side of things. It really, really jumps into that area very, very quickly. And so I think you know, right now we see a lot of generative AI tools, as Dan mentioned, like the CNNs and all that. But you really got to differentiate. You know, are those tools right for the front line of care? You know, in the diagnostics, where you can't afford these false negatives or false positives type two errors, you know, maybe those that setting is actually more appropriate for machine learning, you know, more pure and traditional forms of machine learning, where it's really just fancy maths and fancy stats, right, versus generative AI approaches, maybe more useful for the back end. It'll probably get there eventually, right in terms of the hallucinations and, you know, all these things that we fear, but maybe not quite yet. And so it's about, again, the same thing, using the right type of AI for the right use case, I think is really, really key. And, and it's not even about the FDA, sometimes it's about giving, giving comfort to the nurses, to the doctors who are adopting your your technology. Yeah,

Jyoti Gera  12:00  
we actually, I mean, it's come a long way, to be honest. In the very early days, you would have to explain, you know, what you're trying to do with that? Okay, you could cut five steps out. You could auto measure things. You can automatically understand if there is potential LV dysfunction. You can get injection fraction out of an ultrasound, for example. So it was all interesting, but there were lots of questions, I think in the recent past, we have, I think we have passed that phase, right? They're not necessarily questioning, why? Ai? I mean, it's very clear. In fact, we get a lot of ideas back from you look at the clinician stakeholder of what else we can automate, right? Why can't you have these workflow stitch together? As an example, you know, if you go to a cat lab, they like, you have so many pre procedural data sets. Can you bring that together with what I'm doing intraoperatively? Like, why can't ai do this so we get a lot of use cases that we won't have necessarily thought about, and how we solve it is up to us, right? That's that's different AI for different use cases, but there is a lot of openness. They at the end of the day, we have to tell them how you, you know, pass the ethical considerations. How do you train where is the bias? Explainable? Ai, there are AI committees in big health systems that you have to get past, right? Which is all good, but we can ask for things like caption, health. Caption, health is like a scan guidance. AI solution that is on our handheld ultrasound that's to an extent democratizing having ultrasound in the very early stages of triaging patients, for example. Let's take aortic stenosis right in the communities. How can you try so we get asked, Why can't you do more with that? If I want to go to an ASC and ODL setup, I want to take it out of the walls of the hospital, what more can you do in that area? How can you scale that solution? So they're very, very good conversations, you know? I think we are past the point of you know how you can help us now, it's a bit more. How do you build momentum on top of that?

Armen Vidian  13:46  
Does the AI also depend on form factor? So different for handheld versus your large in house, in in or machines and so forth. Does it change the different types of requirements depending on that form factor? Is there more or less significance depending on it? I think we are

Jyoti Gera  14:05  
building models that are scalable, clearly, right? I mean, to an extent, there is a footprint issue, there is a deployment issue, and that that's certainly there. But beyond that, I'd say, look for us, the form factors are. You could be on the scanner, you could be on the edge, you could be on the cloud. It sort of doesn't matter. We are trying to create that connectivity in that ecosystem, right? That it has to be able to transfer along. It shouldn't matter where you're, for example, where you're scanning, where the procedure is being done, to an extent also who is doing it. At some point, we should be able to transcend that barrier as well, right? Who is the specialist doing it? Do you need a specialist to do something? Tell me a little

Armen Vidian  14:39  
bit about how AI has changed in its significance to GE strategy overall. How you see it in importance for customers your business strategy? Any comments about how we should think about it from GE

Jyoti Gera  14:53  
perspective? I think this changed the landscape for all of us as med techs, right? Whether you're a GE a device company, a startup. Or what have you. So I think the one thing we are learning is we are truly wanting to build that ecosystem. So we do a lot of partnerships in the area, because it's a huge problem. The unmet need is so huge, we obviously want to go at it together with multiple companies. So that's one thing we're doing. We've clearly recognized that, you know, we come traditionally from as a hardware company, and we've evolved very, very quickly. And the differentiation is truly today about solving a given problem. It's not necessarily about always the best image quality, right, as an imaging company, but it's far beyond that. Can you actually, you know, reduce redo procedures in an EP lab, and how can you use software to do that? So that differentiator is hugely important, and I think that has evolved as a company. I can tell you our early days again were many point AI solutions, like most other people in the market right and now we are at a stage where we are stitching the workflows together, redefining workflows completely and literally looking at the care continuum. If we can't stitch things across the care continuum, we won't be able to change the way outcomes are perceived by patients, right? So that's sort of the focus shift that has happened. And of course, how do you sell them? There is an entire business model shift that has happened,

Armen Vidian  16:10  
yeah. And as both of you hear that as entrepreneurs, how do you think of working together with large strategics? Do you at all? How is that important in your strategy? As you've seen, the significance of it change to companies like GE, how is that accounting for how you do business with them or your other partners?

Adrian Lam  16:33  
Yeah. I mean, because of this multi factorial problem that we said, You're not going to have all the capabilities and muscles with our own organization, and nor would you have the channels and the ears of the different stakeholders in the in the whole workflow and pathway, right? So it's important to kind of collectively, almost like a team, like Team innovators versus the versus the opponent in terms of unmet need, how do you kind of bring your strengths together to access that? And so for us, you know, we are a small company. We're trying to do some damage in a good way, and but, you know, we don't have the access that GE has, for instance, we don't have the pedigree that GE has, and maybe some of the engineering thought leaders, and so it's very, very important. I think what we what we need to do, is to really think wider about the form of collaboration. You know, we, we are a digital health company ourselves, but actually, you know, we don't, we acquire our own signals, so we don't actually leverage signals already acquired by other device companies. So maybe some of our partners are actually more of the drug companies, or where they are having difficulty in finding a certain patient pool, you know, for instance. Or maybe, you know, because our test is very safe, it doesn't put any radiation or energy into the body. We can be used potentially for, you know, drug response monitoring or drug tracking or titration, whereas you're not going to do a CT more than once a year, you know. So, so we are, our company is, you know, Cor VISTA is thinking very broadly about the type of collaborator, you know, even foundational AI models, you know, you see some of the, what is it? Bio optimist, as often Nova, you know, we can link with them, link to outcomes for potentially prognostic development, right? So it is absolutely key. And you know you're not you're in an industry where you're no longer just simply sticking two bones together or unplugging unclogging a tube. You know you need to impact so many different stakeholders, and you absolutely need partnership in order to deliver that care and Innovation

Daniel Hawkins  19:02  
at at Vista, we could not do what we do without partnerships with GE and Siemens just gonna sort of flatly. We couldn't do it. The reason is because our software actually has read write control of their MRI machines. That was developed over a 25 year period by the one of our three founders. By the way, I'm noted as a founder of Vista. I'm not. I haven't earned that designation, but one of those three founders ran the MRI Research Lab at Stanford for 25 years. It was funded by GE Siemens and for NIH grants. It was over that time period that that lab created some of what are now the de noising technologies that GM Siemens use. But one of those two companies asked that founder, could you write CNNs to control the machine? Because cardiac mr. Is too hard to do. Technologists can't do it in a repeatable fashion, and. In fact, only 2% of the MRI machines in the US are regularly used for cardiac it's the clinical gold standard, but it's literally too hard to do. So there's a default to echo for portions of a patient population that are much better served by CMR. So the cardiologist took the challenge, Dr Bob, who took that challenge? We have partnerships at a development level where we know when a new software package is going to be released by GE and Siemens. We get source code so we write based on that code to be able to control through their software, their existing install base and new machines. So we wouldn't be able to exist without it. Now, of course, that's a symbiotic relationship, but it also has its challenges, because we're an independent company, and each of them have their own efforts, right? So that's an interesting environment, and that's one of the, one of the, I'm going to say, dynamic parts of this that makes it interesting to be running Vista while at the same time, we're at the cross hairs of the challenges operationally that independent imaging centers and hospitals are dealing with, with staffing, and the clinical challenges that structural heart is dealing with because they don't have sufficient imaging to be able to do what they need to do. So it's a it's a dynamic environment for us, but truly essential to have those partnerships with industry and

Armen Vidian  21:22  
Jyioti, from your perspective,

Jyoti Gera  21:25  
I think, look, partnerships are just, it's the way we can solve the big problem, right? And I totally get it from a GE HealthCare perspective, we are truly trying to say, where can which are the startups, where is innovation coming from, and how does it fit into a platform that we can provide. I mean, our IB is huge, right? So taking this to market together makes a lot of sense commercially. It's not always straightforward, right? I mean, the business models are different. CapEx versus OpEx is different. I mean, there is no clear horse in the race that we have because we are like partnering with so many, right? So we are watching and waiting to see where, a winner might be as well. So, you know, if you just look at recently, yesterday, in fact, the day before, we announced a partnership with Nvidia, right? I mean, those, that's the type of things we want to do, to sort of bring together and say, how can we use that power as well? Go beyond everything that's happening today, physical AI, like, how can you do autonomous ultrasound? So we are really aiming big in terms of that, but we recognize fully that we can do this alone, and that's the ecosystem that I talked about.

Armen Vidian  22:28  
I know when we work with our portfolio companies, we often like to ask them to start with, where are you strong, and where is the partner going to be really strong? And it takes a little bit of self awareness and on the part of both parties to say, this is where they're going to come in to help us, and this is where we just it's not our core expertise. And I think if you start there, then you can go in with a mutually beneficial relationship.

Adrian Lam  22:54  
Yeah, I just want to add a point there, actually on, you know, being very, very open minded about the types of partnerships, right? Everything is very interdisciplinary. Now, you know, and I mentioned that we have some say, we're a device company, we're a digital health company, but actually, some of our, you know, we have a a breakthrough designation and a clearance in pulmonary hypertension. Interestingly, it wasn't internal where we were inspired to pursue that indication. We actually had a we were actually partnered with, um, we partnered with J and J actily, actually, who sell a lot of pulmonary hypertension drugs. And it was actually the CTO of the pulmonary hypertension division there at Janssen, and J and J and Jay, who actually gave us the inspiration to say, Have you thought about this rare disease called pulmonary hypertension, which this year is actually going to be big, because Marc is pushing it right with Rin Rivera, so tatersept. So, you know, it wasn't a an obvious partner for us, and actually they inspired us to actually develop that indication. They actually sponsored our clinical trial and and then actually their CTO actually ended up joining us as our CSO, actually. So, you know, very, very unconventional, and it was actually them who prompted it.

Armen Vidian  24:19  
Adrian, you brought up a couple of times now, of your pharma partnerships, is there anything that we in the Medtech community can learn from pharmas approach to AI native companies and vice versa? Anything they should be learning from us?

Adrian Lam  24:32  
That's a great question. I don't have an immediate answer for that, but what I would say is that, you know, you ultimately Diag, we are in diagnostics and and ultimately, a diagnostic is really, really only very useful if you can link to a therapeutic right for that outcome. And so, by default, there is a natural synergy. You know, that's number one. Number two is that. Yeah, I feel like diagnostics has always been a little bit more, you know, sort of mechanical and structural in terms of its diagnostic kind of scope, right. But then now, with the power of these technologies and the sensors, you know, you're going into more kind of sub grouping of diseases and a little bit of a molecular side of things, right with imaging. And I think that's where that interaction with with pharmaceuticals and pharmacological research is actually very, very interesting, you know. So we, you know, in talking to our pharma partners, they're really interested in looking at subgroup analysis in pulmonary hypertension, you know, for those of you who may or may not know about pH, there's lots of different groups. There's pulmonary arterial hypertension, which is really vessel disease, and then group two is really secondary to heart failure. So all of these are very different etiologies and and, and have different treatments, actually. And so how you interact with those different treatment types, and then how you can classify and categorize your patient pool, is probably where some of that interaction might come.

Armen Vidian  26:13  
Yeah, thanks for taking that question. And you know, I'll turn it back to Jyoti for a question about that I often like to ask is, What does aI mean for the medical device industry and how we operate, not just the impact that our products have clinically, but for how our industry is fundamentally run? Do we have any thoughts about how that may be, how that may pan out? Yeah, I think,

Jyoti Gera  26:40  
honestly, we've just crashed the surface here for the industry at large, for our providers as well, right? I think the roles and responsibilities are shifting very, very rapidly. Like, what skill do you need to get something done, right? That that is, like, it's a complete rethink, rethinking of that. What type of a device rep do you need sitting in the room, performing a procedure, helping you with the procedure, right? And I'm truly excited by what you can do in the interventional space. I feel like there has been a lot of activity on the screening space. The diagnostics are getting so much better. I think we've just scratched the surface interventionally In terms of robotics. What can that do to be completely autonomous, and for us as Medtech manufacturers, I think going to market is going to be a very, very big deal as well. How are we going to take these products out to market? Right? What types of partnerships we are going to do so there is going to be a complete redefinition of how our customers evaluate us as well. I hope there will be a day when we don't go sell a cardiac scanner for ultrasound. Rather, we actually sell a platform that essentially can tell you, how can you make sure that your heart failure patients are triaged on time, diagnosed on time, and actually can get through the pathway in the most efficient manner? Now that is a very different shift in thinking, and it's not natural to many of us who are like, Yeah, I can do the best. CT, I can do the best. Mr. But it almost doesn't matter, right? So I think that that redefinition is truly in progress. A lot of the health systems we talk to now, actually, a recent one came to us to say exactly that heart failure. Tell us what your pathway is for heart failure. How can we work together to co create that? Right? So that's a different way of thinking altogether. Yeah, it's

Armen Vidian  28:15  
that different way of thinking that I think has made AI native device companies so attractive to investors and Daniel, I'm curious, from your perspective at Vista, AI, I know that our friends at Khosla are extremely excited about the company, and we've talked several times as well. When we look at medical device companies, often, in the past, it's been kind of nerve wracking, right? Because they have large mechanical devices that require several different iterations, and often, when you do those iterations, whole new application to the FDA, often commercial traction is required, and that requires a heavy load from a sales person and a clinical person at every case, and that can be quite burdensome to invest in, and the load to get to an exit can be challenging and daunting for investors. Do you see your business in the same light in terms of its impact for investability, for the device industry, and how you look at how it is you operate fundamentally.

Daniel Hawkins  29:22  
So it is quite different. You know when, when we were doing Shockwave, we had to go through the back committee like everybody else, right? So you have to go through that process and GE, how do you get how do you break through with no reimbursement? We were charging $3,300 for a device where the reimbursement setting was about 150 bucks. So that's pretty tough, bad numbers. So how do we do that? Well, we piggybacked on tabbers, right? So without going too much into it, we became a facilitator for a procedure they desperately wanted to do that by itself. Wasn't profitable, but it became a facilitator and got us on the shelf. Then we elbow. Our way into a larger presence. You can't do that with AI. Doesn't work. Matt's wrong. The go to market strategy, as noted, completely different, refreshing in some ways, depressing in other ways, right? Because it's tough. You go from this is really interesting, to prove it to me, to AI governance committees, to four month selling cycles turn into eight month selling cycles. So while your go to market isn't body heavy, it's time heavy, right? So you're spending a lot of money, but in different ways. All of that said, once you're in your gross margins are in the 90s, right? So your payback is extraordinary, that's right. And if you can manage to be SaaS oriented, that payback starts to become economically extraordinarily compelling.

Armen Vidian  30:50  
That SaaS model for AI native companies and medical devices, I think, is revolutionary for how it is. It's investable, its impact on customers, how economic, economical it can be, and so forth, absolutely.

Daniel Hawkins  31:04  
And it becomes a category that reminds me of the very, very early, I'm going to call them cowboy days in interventional cardiology, where the devices were just popping out all over the place, and it's $600 balloons, and everybody was paying cash for them, and it was just phenomenal. It reminds me of that environment a bit, from the standpoint of the creativity, commercial creativity, and excitement around opportunity and and go to market strategy, variability that you can bring dating

Armen Vidian  31:32  
us with talk of the early days of interventional cardiology. Anyway, so I see we're out of time. So real quick, lightning round, what do we think is next for AI and cardiovascular medicine?

Daniel Hawkins  31:46  
Wow, that's an interesting one. What drew me to VISTA is what I'm going to call that what's next, and that is an ability for AI to help in in creating the images that drive therapy. The best of human eyes can recognize things that are still not quite as good as what AI could do. The reason why I decided to do one more and jump from intervention to to upstream and in imaging is to make sure that the right patient gets to the right procedure. There's an opportunity to do that in cardiovascular because it's an image heavy business, and computers can look at images better than humans.

Adrian Lam  32:29  
I think it's about I think cardiovascular disease is very has a lot of overlapping symptoms and a lot of confounding factors. So AI is going to help us be able to find those connections, those relationships that were not obvious

Jyoti Gera  32:46  
before. Yeah, no, that's those are great points. And I honestly think just managing the silos of care, even within the cardiovascular spectrum AI is going to change that completely today. There is so much fallout between every single element of how care is delivered, right episodic care. So I think it's going to completely change that by taking the complexity of it just sort of really making that easier for us to take give patients the best outcome, no matter where they are in their journey.

Armen Vidian  33:15  
That's right for us. I think it's aI integration with robotics. I mean Intuitive Surgical really coined the term robotic surgery for everybody in this room, I think for us, the next generation of that is combination AI with robotics, so that you can focus away from the grunt work of a procedure and just focus on the patient. I'm looking forward to things like robotic colonoscopies, if anyone's out there with one, and so for some of the other procedures, starting there, I think where it's not the emergency procedures, we can gain credibility and traction there and then move on more broadly from there. Well, anyway, we've run over time. I thank you all for your patience and time, and it's been a great panel. Thank you.