#42: Medical breakthrough for bad hearts

This is going to be a different one. Today we're studying how AI is changing, how cardiologists can study the heart, and it might save your life. So who are you?

I'm Hugh McNaught, CEO of VenturePoint Diagnostics. I'm here with Jeff Sobel, who's the CEO of Ascent Cardiovascular, and Christina Almeida from VenturePoint.

Yeah. And tell me what you, what you all are working on, because it's really exciting.

Yeah. So we are applying artificial intelligence to cardiac ultrasound. So taking a ubiquitous technology, 2D ultrasound is everywhere. The installed base is thousands and thousands of systems. We're improving the accuracy of that for hard imaging, the levels equivalent to mri. So faster, cheaper, far more accessible.

So I have three kids, and every time my wife was pregnant, they would do ultrasound and see the baby inside of her. Is that the same kind of ultrasound sensor that you're talking about here?

Similar ones, absolutely. We get into the game a little bit later than pregnancy, but our systems have been imaged to have been used to image babies young as 30 minutes old. So maybe Christine could touch on that. As a practicing sonographer.

You are right. That is the same equipment, the same ultrasound machines that we would use when your, your wife was pregnant. But like Hugh says, we tackle cardiac hearts once the baby is actually born. We can do hearts anywhere as small as 3ml. So very small pediatric premature babies all the way up to adults in, in, you know, their 80s and 90s. So we handle all patients and regardless of, as well, the condition that their heart is in. So, yeah, it's faster.

And, and what, what is the breakthrough here? Because, you know, I've, I've seen a few of those ultrasound scans. You know, you, you can see what inside the body looks like on a monitor. What does the AI. And, and what is the breakthrough here? What does it let you do? It let you see or let you understand about the human heart that you couldn't before?

Yeah. So we're very focused on what we call volumetric measurements of all four chambers of the heart. So the volumes of each of those chambers, plus what we call the ejection fractions, which essentially provide the cardiologist with a measure of the efficiency of how the heart is working. So uniquely, we have the ability to do this for all four chambers, including the right ventricle, with an accuracy that is equivalent to mri. So that is unique.

So you don't have to stick the baby in an mri. That's a big machine and you can't have the human touch.

Right, right.

Or a Claustrophobic adult or people with implants. You know, Jeff might be well positioned to kind of speak to this.

Yeah, yeah. Hi there. Thanks for having me.

Yeah.

MRI exams on kids are kind of tough because they don't follow instructions. They're upset, you know, to have something done. They can't hold their breath for you the way adults will during an mri. So often. Actually, an mri, you have to give general anesthesia to a child and that's a big undertaking. It has certain risks. And if you can get the same data through a much simpler, non invasive mechanism, that's highly preferable.

That's cool. Do you guys have a demo or, or what it actually looks like for people to see? We can do that later as well and then overlay it on the video. But if you have one and you can share your screen, that would be awesome. This is really interesting. It seems like this is a real breakthrough for your, for your industry and how, how doctors can understand the heart, understand defects, and make a plan for proceeding much better and much cheaper too.

Yeah, absolutely. Yeah. This has been another 10 year overnight success, but we have recently launched what we call version 4 in a number of countries and we're expecting clearance by FDA on that very soon. So I think the form factor now is about right for mass adoption. And certainly this is part of our partnership with JustGroup looking at how to integrate this into the clinical environment.

The 3D scan that it gives you, does that really help a surgeon then know what, what he or she is up against if they need to do some operation to fix, fix a defective heart or something like that?

Well, I think it's sort of much broader than that and probably Jeff is best positioned to speak to it. But I, you know, I think the areas we're focusing on as a company, at least in terms of what we perceive to be near term opportunities for us, are congenital heart defects. So there's two and a half million people in America with heart defects, you know, one million children, but also one and a half million adults, you know, cancer patients. So a number of chemotherapies are actually toxic to the heart. So these patients need to be carefully monitored. Pulmonary hypertension, and we're just beginning to do a dive into valve procedures. But, but Jeff can speak more knowledgeably about each of these sort of indications.

Yeah, there's really a very wide range of heart disease that this is applicable to. And, you know, one way to think about this is standard echocardiography or heart ultrasound or, you know, even what you might be familiar with when A woman is pregnant and has a fetal ultrasound. It's kind of like a slice through the heart. That's a two dimensional image, kind of like a pane of glass. You know, you're only really seeing a two dimensional slice. And obviously the heart's a three dimensional structure, really often referred to as kind of a four dimensional structure because it has a time dimension, it's moving with time. And so the beauty of this technology is being able to take those, a series of two dimensional slices and using AI, map that information onto models of the heart that reconstruct all that information in three or four dimensions and give you an output that's fairly equivalent to the four dimensional data you would get from a cardiac mri. Really, ultrasound is ubiquitous in our world in cardiology. Ultrasound is really the workflow modality. It's used day in and day out. It allows you to study people sequentially over time and see what changes are going on. So even though we have other ways of looking at the heart, like nuclear cardiology, kind of stress test type scans and CT scans and MRIs, and even though we do procedures where we put catheters up in the heart and do all kinds of fancy measurements and interventions, really ultrasound is the workhorse modality. And when you add this kind of layer of capability, you make it just that much more powerful and that much more applicable to a group of patients, particularly the pediatric patients, who are much more limited in how you can treat them.

Yeah, my friend made a digital stethoscope, it seems, which would see the heart in a different, different, different way than the old doctor who had put a stethoscope on your chest and listened to your heart working. It seems like you're getting some fantastic new tools to understand the heart, partly because of your efforts. Right. What's happening? You know, man, I, I get so excited about what I call the holodeck because we're, we're moving from 2D computing to 3D computing. I mean, I have an augmented reality hat on from a company in Ger for that reason. This seems like it's building a new kind of slam map of the heart so you can really spin it around and zoom into it and understand how it works in a whole new way.

Yeah, I would say, you know, that's one of the focus areas of the sun. Cardiovascular is how you pull information together from all these different areas. And, you know, the way I would kind of explain it is each way we look at the heart has kind of a different information domain. So if you put a digital stethoscope on somebody, you're obviously getting a digital audio signal and you can feed that into AI and get all kinds of interesting results from that. You could obviously do a plain chest X ray or an electrocardiogram. You can kind of go up the information space, so to speak, into ultrasound, which are sound waves that bounce off of structures in the heart and recreate a picture. You can Connecticut and Mr. Work differently in terms of how they look at radiation pictures or magnetic fields that the heart is in. So they're all just kind of biophysical information that you're getting out of the heart using different modalities. And cardiology has become very sophisticated over time in kind of putting all that information together in the best possible way to help treat patients. We're always looking for. Well, AI is obviously revolutionary because it can be applied to any of these biophysical outputs. Really, people are working on every kind of way to squeeze more lemon from the juice, so to speak, from these information streams. But VenturePoint has been unique in that they've taken the approach of actually taking two dimensional reconstructions and using AI to turn them into the equivalent of three dimensional data sets, which is really something that hasn't been possible to do in the ultrasound field before this.

And this all happens real fast. I mean, if you have a baby that's a few seconds old and the baby is having problems and you're a doctor trying to figure out what's going on, seconds matter at that point.

Right. Cardiology is a very real time, you know, domain. Things happen real quick. And time can definitely be of the essence when we're caring for patients who.

Are critically ill. Just so I can add to that, just to give an idea of how this program and how our software actually works. If you did have a baby in the hospital that needed an and echoes, like we were saying, are very prevalent. It's a go to modality within the hospital because you can do it quickly. You can see real time action of the heart. So when we come with our system as well, which is in conjunction with the ultrasound machine, is we have two sensors. So one that'll go on the baby or beside the baby, and then one that goes on the actual probe. So when we're scanning and taking Those pictures, those 2D slices that Jeff was just talking about, we're now gathering information in the spatial XYZ planes. Right. Every time I angle, every time I move that probe and I take a picture, it's now in correlation with the sensor, so it knows where I am in space. Once we take those 2D pictures, and they're classic 2D pictures that everybody has been trained to do. We put anatomical points on them. We label them. This is an apex. This is a valve. We give that X, Y, Z shape, since we know where the probe was in space. And then from those points, we create a 3D model which represents that baby's heart that you just scanned. And then we have those cardiac output values that you got just bedside. You know, it was on top of the echo that was already scheduled to be done, and it only took an additional five minutes more on the pictures that you already had to take anyways. And now you have this cardiac output values that are equal to MRI that were done just bedside. So now you're able to, like we said, monitor the baby as you go. And it's not to say that MRI is to be replaced, because it is the gold standard, but it's a different mode and a faster mode to get information so you can give better bedside care to your patients in a much faster way. And then when you see some major changes, maybe that's when you're like, now it's worth going for an mri, you know, or maybe it's about changing your protocols. And we're talking in the instance right now of being inside a hospital. But this also pertains to globally as well. What about those children in cities that don't have access to mri, or those adults that have pacemakers that don't have access to MRIs, but now they can be followed in their hometown by their local physician. And you can create protocols that say, okay, once we see 10 or 15% change within the lab, then we'll send you out to get additional, more accurate MRIs. But this is to help feed and create better patient care around everyone in. Regardless of what space you. You're in or where you live? Because ultrasound is that available.

It is a lot.

These devices are not very expensive compared to an mri. Right. Mri, you need a whole room. It needs to be magnetically shielded, and you don't want anything metal in that room. Right. It'll get attracted to the mri.

Absolutely.

And it also takes a long time. I had a full body MRI scan. I was in that thing for an hour. Right. And so it's is just boom, boom, boom, boom. You're going for a certain, you know, you can see a lot. This real exciting. You know, I just. As a layman, listening to you all talk about how this is changing the world, that's real cool.

Can you.

Can you all tell me about The AI that's doing the stitching of these images together and making the 3D model of the heart and sort of what's going on underneath the hood.

So I can. Oh, sorry, go ahead.

No, go ahead, Christina.

So when they created this algorithm and this AI, they actually joined forces with a bunch of different hospitals here, both in North America, in Canada, in the US they actually had multiple thousands of cardiac MRIs donated to the hospital. And they had those physicians at those hospitals actually mark up those MRIs. So it was all done by physicians on physician quality MRIs from hospitals, and they created the algorithm from those MRIs, which is why we can say, and we can make a claim, that we're within 10% of MRI because we use them as the grounding foundation to create the algorithm. Now, I can't tell you how the other one was. An algorithm was created.

A lot of Nvidia cards, right?

That's right. I can't share that part. But it was, it's founded on actual MRIs. We have a catalog. And it's not normal, it's all sorts of different pathologies. Because that MRI catalog is essentially a shape catalog. It's what the AI was built off of. And it's not, not a learning AI. We don't have to continually feed it. It's one that's born from these MRIs. Anything else that's out there at the moment makes no claims about the values that they have and how accurate or good they are. They just say, we've created a value, but we can actually compare it to a gold standard because we used it as our foundation.

That's interesting. So you have a large training set of MRI scans that you were able to feed to the model, to make the model. Wow. It's just AI is changing everything in the world and it's just so exciting to hear, hear how it's changing this for, for you all. It, it means you can see, you can see your patient in a whole new way and much quicker and much cheaper.

Right.

I think in terms of, you know, healthcare systems are looking at things at the macro level. You know, cardiovascular disease is the leading cause of death worldwide.

Yeah.

Better at treating it, whether it's through drugs or surgical interventions or devices. So these patients are living longer, but then need to be carefully managed. People are living longer. Generally, a number of countries are looking at increasing standards of living and with that comes unhealthy lifestyles, cardiac disease. So you've got this tsunami of patients.

I don't eat a lot of ice cream at Night.

Yeah. The question is really, how do we effectively manage all of these people and putting them all into MRIs? You know, clearly it's not a scalable approach. So we think this is more practical, you know, much more efficient and certainly for those patients that need to be regularly monitored, you know, I think it's just a better tool.

Yeah, no, this is. Yeah, I, I totally get that. What, whatever. What are you all learning by doing this? You know, you know, building an AI system and that, that models the heart in a new way. What, what are you taking away from this after how long have you all been working on this project? It's probably been a few years. Right.

The company's been in existence since 2007. Christina and I have been a little more recent additions, but, you know, we've been at it for a few years anyways. So, yeah, I think the learnings are constant. Actually, this morning we had a very exciting team discussion just around the AI and how that's evolving and just the improving accuracy of it analogy. This morning we're sort of at the stage of stay with cars, of having driver assistance, and we're looking forward to sort of at some point creating an autonomous driving vehicle.

That's interesting because I worked with a company called Mediview, which put scans over the human body for the surgeons who were going to be cutting a tumor out of you or something like that. And it's almost like playing a video game. They, they can see when the cutting tool is in the right spot and it turns green and tells them, okay, you're in the right spot for cutting. Does this help, help that kind of thing to go down? I mean, a surgeon someday is going to be wearing a, a pair of augmented reality glasses and be inside your model of the heart and be able to fix a problem or something like that.

Well, not quite there yet, but one thing we're looking at right now is with heart valves, so, you know, that's a very dynamic area, going from repairs to replacements and actually, you know, replacements that are introduced through catheter rather than cracking the chest open. One thing that's really interesting for us is to be able to assess the, you know, characteristics of these valves, you know, through the chest wall, just like they were counting your wife when she was pregnant, rather than going the route with greedy echo where they would have to anesthetize the patient, put the probe down their throat. So it's invasive, time consuming, it's expensive. I think that's the next step for us. But Jeff has been living and breathing this the last 20 years, maybe he has more insight.

Yeah, well, so a few things I've observed over the course of my career and you know, computers and cardiology. One, cardiologists are very tech savvy. I mean, we love technology, so we're kind of like radiologists and surgeons who also kind of take care of patients at the same time and talk to people. And so our, our kind of technology adoption bandwidth is quite high and. But yet we work within systems. I mean, one of the important things to understand about AI and the healthcare space is the cardiovascular enterprise is kind of a system of its own within a larger hospital and healthcare system. And so AI has to kind of plug into the system and be used efficiently by people in the system, the providers, the practitioners. So that's really the value of our partnership with VenturePoint. They've created a very unique AI technology. It's kind of like a new tool in the toolkit, but the cardiologist needs all of those tools and they need to have those tools available to them in their workflow and in their environment, kind of at the right time for the right patient. So that's a lot of what Ascend does, is bring together tools we've created and tools that our partners have created that are best of breed and cutting edge, some of which are AI tools, some of which are other kinds of imaging or information assets and trying to make that work within the cardiovascular enterprise and within the overall healthcare system. And it's probably true of many other domains, whether you're talking about finance or manufacturing or whatever. AI has to solve a problem. It has to be relevant. It has to either create a better outcome, which in our case means a better patient outcome, or it has to make the system more efficient or ineffective. It has to have one of those goals. But it can't kind of live in a silo. It can't just be in its own little bubble. It has to be part of and transform the system that becomes part of.

How is this going to change over the next few years, you know, you know, five years from now, what does this system look like? What is it going to be able to do that you can't do today?

Yeah, I think going back to my earlier comments about form factor and you know, how we're ready to really steer things, steer our company into the phase, you know, we're going to have increasing levels of software automation. So right now we're working with the cardiologist to guide them or to make suggestions regarding certain, certain program functions or features. Will be increasingly automated and done in a way that they have much more confidence. So at this point it's not about opposing this automation on them, but you know, getting them really to use and trust the system and you know, we'll build on it from there. Again, right now it's, you know, providing, you know, driver assistance in new cars versus, you know, having self driving cars at some point down the road.

That's cool. I hear we have a demo ready to go. Could we see it?

Yes, absolutely. So I'm just going to share my screen or if someone will allow me to share my screen. Oh, here we go.

Share.

Do you guys see my screen?

I see a white box. Hopefully it comes up.

Let's see, let's see. Do you see like a bunch of scans on my screen or not?

I can't, no.

Okay, let's try again.

And what are the hospitals this is being used at?

We've got a number of placements. So in the US the Mayo Clinic, Duke University Hospital, which is one of the top cardiac centers in America. Seattle Children's, Phoenix Children's, Sick Children's Hospital or Hospital for Sick Children, that's one of the top two pediatric sites globally. King's College in London, Erasmus in Rotterdam, we've got them in Berlin, Vienna. So you know, the early adopters really are the, I would suggest the thought leading centers of cardiology.

Very cool. So what are we seeing?

Okay, so this is what the program would look like after you gathered your images. Okay, so I didn't mention there was two sensors and you use a regular ultrasound machine like normal. And then we have. So let me click on. Oh, there we go. Okay, so as you can see to the right hand side here we have seven images. And these are the images that would guide the sonographer that need to be taken. And all these images that you see here on the side are very regular classic images that all sonographer schools kind of teach as basic images. So there's nothing sort of out of the normal here. And if I go to the very top, you can see we do all four chambers of the heart. Now depending on which chamber you want to do, those images will change. But again they're all still very standard type images. So when we talk about our program, we do talk about the RV the most because it is the ventricle that is not seen as well and doesn't have as many measurements available to it. But a quick understanding of how it works is for each of the images there that are on the side we're going to Put these points down, which are different colors and associated with different anatomical landmarks. So we do that for sisyl and diastole, and then it'll give us a 3D reconstruction. So you can see here, my blue points here are for the RV endocardium, and then my green ones are for the septum. So we put those both down for system for sisly and diastole.

And then you're scanning. You're. You're tagging these.

Yeah, it's a post processing event.

Okay.

So after you've scanned, you would send these to your computer. Or you could do it on our actual laptop. So I'll show you what it looks like. This is what our actual ventry point system looks like. It's essentially a laptop on wheels. So you could do the post processing on this right beside the patient bedside, or you could send the images back to like a workstation and do it on a workstation. But after you've put your points down, you get your 3D reconstruction. So this is a representation of a 3D RV and how you would look at it. So we have sisal and diastole together. So we have our one inflow. This is tricuspid valve inflow. And then the outflow flow for the blood tract.

So, Christine, if I could just jump in for a second, we're using a little bit of technical language here.

Yeah, yeah. Dr. Nerdy on me.

Sorry.

No, that's cool.

So, yeah, what she's referring to is the heart when it's relaxed and compressed. So your blood pressure, if it was 120 over 80, that's systole over diastole. So that's what she's referring to. And the image here is showing both states. Yep. Did I get that right? Jeff, I'm not going to question you on it.

So then we have this rv. This is a representation of the heart that should be on the table that you just scanned. So you look at areas that are dilated or stenotic. It's a representation of how well stenotic. How well, like how big things are versus how small or undersized things are versus oversized. And then you can see here on the right hand side, we have all of the values that would be important to cardiologists. We have the volume of blood when the heart is relaxed versus the volume once it's squeezing your output values. So these is. These are the values that we say are comparable to MRI within 10%.

Very cool.

And then we just have. Yeah, some fancy features. This shows how all of the Dots are connected so you can see how you've scanned the heart. We have something called a scan plane so that you can see how you cut the heart with your 2D image. You can do more than one at a time. So it's a great learning tool. Like you said, it's really hard for some people to think in 2D, 2D to 3D, because it is a three dimensional heart. But we're scanning in 2D. So this just gives people an opportunity to learn a little bit more and wrap their mind around it.

Wow, that's cool.

Yeah. And this can all be done with just. With just a few minutes added onto a regular echo?

Yeah, that's crazy. You're turning hard into a holodeck, I call it.

So that's just a very quick version.

No, that's cool. Would it help you to be in like an Apple Vision Pro or something like this when you're looking at the hardware, or is it just good enough to see it on a 2D screen?

Well, that's one I can probably take, actually, in the sense that our team spent a lot of time with smart glasses and some of the VR headsets and all. Part of what we're very interested in is not just the image analysis and how to improve that, but also the information delivery to the providers. Because at the end of the day, for the foreseeable future at least, cardiologists are going to stay in the loop in patient care and other cardiology providers. So when you think about the VR headsets, there are some people who are working on visualizing the heart, mostly for electrophysiology procedures to understand not just the three dimensional structure of the heart, but the electrical circuits through the heart and how they can modify some of those electrical circuits in three dimensions. They're called electromechanical maps. And so there are some very sophisticated cardiac displays that might benefit from holographic or VR type displays. But. But that's not the only useful thing. I mean, it's really interesting with kind of augmented and mixed reality and smart glasses. There's such a spectrum of technologies now that we can marry with AI to improve the delivery of information. Point of care, particularly in cardiology, where we're often mobile, we're in the icu, we're going to different labs. I think it's one of the most interesting areas for innovation actually, is the intersection of AI and AR VR. Kind of a smart glass revolution.

I'm so excited. This opens up a whole new way to do healthcare. I just want to thank you all because you're inspiring. And thanks for helping people, you know, have better lives and fix their hearts, you know, keep, keep track of their hearts. Is there anything I should be asking because I'm a layman, I, you know, I don't live in your world very, very much so. Is there, is there a question I should be asking?

Christina?

I think you've covered it pretty well, I would say. You know, again, one of the challenges we're having in healthcare is, you know, across the entire economy is people getting used to the idea that these AI driven innovations are going to be, you know, part of their everyday experience. And so I think podcasts like yours and ways of helping educate, you know, the population, everybody's going to be is a patient someday. And also, the more we all understand that this is going to hopefully help us all, the better off we'll all be.

Do you have a success story about how this really changed somebody's life or really brought a better outcome?

So just from, like, traveling around and, and seeing different cases in different hospitals, I have had, like, lots of different feedback. We had one family in particular at a children's hospital who we did this on, and they were so grateful because they didn't want to have to. He was already a little bit delayed and had multiple, multiple hospital appointments. And to avoid the anesthesia that comes with having to put your child through an MRI and then not being able to actually be beside your child as well, because they're not allowed to be in room. The room. So they actually expressed how once we went through all of this and did it on their child, how grateful they were that they did not have to put their child through that one more time yet again, you know, and they were able to, even if, like I said, it was delaying it for two years and just do this until that happens, then just how easy it was on family and their care and, and all of the stuff that the child has already gone through. And then, I mean, that was one example that you really felt it from the parents and how much they really appreciated it. But then I've gone to an adult place as well when we were demoing this product and we did a patient that had a 3D echo, done an ultrasound with the regular ultrasound machine, and then we did our equipment as well. And he was just happy to see because we went through the results with him and his physician to show that our results were closer than the actual 3D of the Echo, so he wouldn't have had to do that MRI for the next. So he was very happy to hear that he was able to combine it into one that it was going to be more cost effective for the hospital. And then he wouldn't be disturbed to have to come back again to get results or to do the MRI again, because he was able to do it in his one appointment. So from both sides of the spectrum, whether it was PEDs or adults, I've seen some really positive feedback from patients and families who've had an opportunity to use the system recall.

Well, thank you so much for spending a little bit of time to me with me to show you. Show me the system and explain its importance and why it's changing how we understand the heart.

Well, it's a very exciting area. Thank you for allowing us to share this with you. Yeah.

Thank you so much.