I was asked by Jen to moderate this session on what would be called the future of AI and machine learning in pediatrics and congenital heart disease. And I want to just pull up, we obviously won't need the Q&A stuff here. This is a small group. It's meant to be an open discussion, but we have a number of people who have been similarly prepped, and I just want to make sure that I identify everybody. The experts at the table include Charlie Brule, who is from Children's National, Artie Dallal from Vanderbilt, and to my left, I have Kate Spivak, who I just met, from CHOP, Nikon Bergen from Madison now, right? And then Andy Blaufachs from Long Island, and Lisa Rowell, I think I may have met you once before, but welcome. I'm from WashU in St. Louis. WashU. Right, exactly. So I just want to read to get things started. I'm going to toss it out to all the five or six of us I just introduced, but you guys are not only free and welcome, but encouraged to jump in and make this truly, make this rectangular table more of a round table, and I just will give you the opening of the prompts that Jen Silva sent to us, which were in response to, I think, a couple people in this group who more or less said, what the hell am I supposed to talk about? There's a significant interest from our innovation partners, and by that, in the context of HRX, Jen means industrial development, to understand what is the pediatric market. Charlie, you don't want the blinking purple. You want the solid purple. So keep going. Here are some questions I hear, Jen meaning I, hear from innovation partners that you may wish to consider for conversation at your round table, and the first one or two I want to read. How important is it for us to have indications for use for products in AI in pediatrics? What does regulatory clearance, what's the significance of that to pediatrics, and what do we wish as people who are in the pediatric and congenital heart space, what do we wish that we had with regards to AI and digital health tools? Is what you see as necessary and important something that's already there and we just haven't figured out how to utilize and implement it yet, or are there things that are outside the bounds of what's available? So I'm going to throw those questions open to our panelists. I would like, as you get started, to identify not only who you are again, but also why you think Jen Silva may have picked you for this panel. For instance, I was just talking to Kate, and I'll start with you, who explained to me that she started a company, oh sorry, Kate next to me over here, that she started a company based on a product that she developed and has become a successful product. So she is here not so much from the AI space perhaps, but from the entrepreneurial space and the regulatory space. So Kate, I'll toss it to you first. I think that I have learned so much, like creating a product and bringing it to market, some things that I have never learned, and you know, I obviously have just a medical background, and I was never in business, and so thankfully I had a partner who was, who was also my husband. And just- Now Kate, tell us what your product is before, we're curious. Sure, yeah. So except being a PA in EP, I'm also an IBCLC, which is an International Board Certified Lactation Consultant, and I created a product which helps sustain breastfeeding, and really what it helps you do is supplement at the breast when you're feeding a baby to basically eliminate bottle use and keep the baby at the breast and helps you promote milk supply. So we did that for a few years together, and I think, so like going back to the question that was originally asked, I think it's definitely important because going through this process we had to learn about like FDA and what it means to be, like to get all the clearances, and going through like what it actually took to get there, I understand the importance of it. And so yeah, I think it definitely would be important for us. I've got a lot of reverb in mine. Do you guys have reverb in yours too? Okay. So I'm Nick Vonberg, and so I'm a pediatric cardiac, it was better for a second, pediatric cardiac electrophysiologist from Madison. I also am on this panel partly because I invented what's called the Atrium, hopefully some of your, hopefully all of your centers are using the Atrium and hopefully use it a little bit more. It's a device that allows continuous real-time monitoring of the atrial electrogram after heart surgery by hooking up to the atrial wires and then putting that on the bedside monitor. And so, so as I think about the question of what about indications for use, in my experience with things like the Atrium, I think the indications for use seems to be one of the things that we can grab onto as kind of the beachhead, here's where we can use this product. But then once you find that something is especially useful, then it seems like we move from there. So as we think about innovation and electrophysiology, I think indications for use are beneficial at the outset, and then we start to figure out how could it be used even a little bit more or a little bit broader. I will wait on the question of what do I want for the future, but from the least indication for use, I think yes, I think it's important initially and then less important as you move forward. I'll have to hold it down, I guess. Andy Blaufox from Cohen Children's in New York. I'm not really sure why Jen picked me other than I'm on the Digital Health Committee and I've known her for a long time. But anyway, to me, I try to be somewhat innovative but not as innovative as these people who spoke previously. But I think the question about FDA and regulation to me is incredibly important as a pediatrician knowing that the market is not where we're at. And so I think having the FDA involved and having pediatric indications and having some sort of drive for industry to make sure that there are pediatric indications and or data to support the use of devices in the pediatric population is critical. You know, taking, for instance, like the Apple Watch, every time I talk to a patient about using the Apple Watch, you have to log out and you have to now lie and say you're over 18 and so they can use it. And clearly, the algorithms aren't developed for patients who maybe have as much sinus arrhythmia as my patients or have faster heart rates and all these other things. So I think getting the FDA involved and having tighter regulations and making sure that the products are directed to a pediatric market is important for me. Hi, I'm Lisa Raleigh. I am from Wash U in St. Louis. I actually work with Jen. But in our work, we do a lot of research using digital devices and digital health tools with our patients. So we're trying to help push that research into our community by trying out various things. We've tried, we've used different stethoscopes we've sent out to patients in toolkits. We use various ECG recording devices and compare the readings on those. I think it's with movement. The reverb comes back with movement. And I think that that is huge in pediatrics that we need to continue to do that type of research. And to piggyback on Andy, what we're finding too is that patients also want to know that it works for their kid, that we're not getting all of the false information or false reads. And the way that we can avoid some of that is when we have pediatric algorithms. I think we see that too in our remote monitoring of devices that even when people are, we have tech support who overread some of our devices like ILRs, they're reading based on adult criteria. And so we're overreading everything that they're reading because it's not technically correct for our patients. Yeah, so I agree with both what Andy and Lisa have said. So again, Arti Dalal, I'm in Nashville at Vanderbilt. I'm a pediatric electrophysiologist. And I think one of the things that's really challenging, we in the pediatric community or even in the adult congenital space have gotten very used to sort of using what is there and adapting it to our patient population. And that's important when you're talking about, as Lisa was alluding to, you can sort of use the same values and we can't use the same values. So I think having specific pediatric use in our patient population is important because again, like Andy said, I had a patient that uses the Apple Watch and we sort of talked about not using it necessarily. And I will sort of look at the information as it comes through. But the Apple Watch, the ECG sort of app stopped working. So they went to the Apple store to say, hey, this wasn't working. Said they were using it in their 14-year-old kid. It's not made for that. They made the app inaccessible for that family. And so you can't use the product if you can't troubleshoot it. And if you're using the app off-label, then it makes it very challenging. So I think making sure that if we're going to make use of this technology that exists, we have to make sure that there is pediatric indications because then our patients can then actually get the support they need when things go wrong. Wow, that's a good story. I'm Charlie Burrell. I'm impressed with they need to go to a different Genius Bar, I think, find other geniuses. I'm at Children's National in Washington, D.C., and also spun off a company called Pericor, which is involved with miniaturization of devices and minimally invasive ways to get to the heart. And Contego works with me at Nationwide and several others around the country. So I'm not so much involved with AI and digital health, but involved more with miniaturization of device development. We all use AI and digital health in various ways, whether we realize it or not. I'm just going to take another stance on this, though. Just from having a device, I wish, like I definitely think we need the pediatric indications when we talk to our patients. But I think as clinicians, we should be more open minded to hearing other ideas, because there were many times where we went to hospitals and they would say, like, are you clear? And we're like, no, not yet. But like, listen to our idea. And like, we really think it would work. And really what it took is just some people who believed in us and we were able to prove our case. And now we have everything that we need. But I think it's important to listen to, like, have an open mind to innovation. And I'm going to round this out just by my own experience in this. Over the last couple of years, there are two main areas that I've been involved with. One is a group working with Ann Dube and Charlie and others, forging a collaborative environment. Actually, some people are calling them collaboratories between clinicians, industry, and the FDA, recognizing that if we are going to use products in an on-label manner in pediatrics, then that allows us to collaborate very directly with industry in recognizing the shortcomings, the opportunities, and the specific factors that are necessary to address in order to make the products better over time in collaboration with industry partners. The other thing I've been involved with, which I hope is why Jen put me on this table, is we've been very engaged in the more research development of artificial intelligence as a tool for interpretation and diagnostic prediction, specifically in pediatrics and congenital heart disease. And I've been working with a fellow who's also a doctor of biomedical engineering in that space named Josh Mayorian. And so I'll be bringing that to the table today. So I'm going to send it around once more. And I'd like to, I mean, I noticed that all the faculty are on one side and all the visitors, if you will, on the other. It probably would have been better if we were all mixed up. But I'm going to invite everyone to participate. And the next particular question, which is kind of like a free-form thought exercise, which is if there was an opportunity for you to put forward a wish list of things in this particular space that you thought might be useful in your practice, I mean, it doesn't have to be comprehensive, but what kind of things might that include? And Nick, I know you were about to talk about something, so I'll toss that out to you and you can say whatever you want. Yeah, well, you know, I think as we think about the last question, it was what about the FDA guidelines? I think we have to think in the big, big picture here. The big picture is if we want innovation. So I read something that said if you're just inventing, you're just inventing. If you want actually innovation, you need to actually have something get to the marketplace. And indications are something that allows you to have the financial kind of, you know, cojones to say we can move forward. Like if we have indications and you can possibly get reimbursement, that can have a business model. Otherwise, innovation will not flourish. As we think about the wish list, you know, I'll take a side note here a little bit. I was going to Mexico to do some medical mission work and I was doing electrophysiology studies down there and, you know, spending a whole lot of money moving stuff down there to do eight cases. And after I did that, the question is how many immunizations could that have bought? It could have bought a ton of immunizations. But instead, I took care of eight patients with, well, more than that. But either way. But getting back to the question, you know, I think as we think about what is our wish list, we should also think about what can we accomplish. And I think if we think about our particular pain points in our particular areas of expertise, that's where we should focus. You know, a lactation consultant should think about that pain point. And, you know, for me, postoperative arrhythmias with inability to always know what the rhythm was. It just seemed to be crazy that we didn't have that. So what is my wish list? It's always to push things a little bit forward at the particular pain points that I feel, ideally with a high-risk, high-reward benefit for intervention. And I think digital health allows us to go in so many different areas. I find that kind of exciting. Okay. One's not on. I'm Brynn. I'm from Ann Arbor University of Michigan. One of the things that I think about often is who are the kids that we're seeing that come in with Apple Watch? They're usually well-to-do kids or well-to-do families, where the mom is giving the kid their Apple Watch. So what is a way that we can use what we have? What is a way that we can use what we have? What is a way that we can use what we have? Or well-to-do families, where the mom is giving the kid their Apple Watch. So what is a way that we can use what we have, these digital products, and make them cheaper or make them more available for families who may live in inner-city Detroit, in my world, or way up in the Upper Peninsula, where it's like Alaska and there's just not a lot of resources? So how can we get the products that we know work for our kids? We diagnose SPT and lots of things using Apple Watches or whatever digital product. So how can we get this more on the market? I mean, everyone was just saying, we have to tell the kids you have to be over 18, so we have to lie, you know, blah, blah, blah. So how can we get this out there for, you know, all of our patients and get them into the inner cities that might need this type of diagnosis, where they don't have the transportation to get to the hospital for a visit, but we can use these products to really kind of expand, you know, DEI and, you know, things like that. So that's would be my wish list moving forward, is to make these products more available to everyone. You know, the topic of health equity in digital medicine, I think I'm so glad that you brought that up. Any comments about that? I think that there's an enormous opportunity to democratize expertise and expand the access to our services. Yeah, I was gonna say, as preparing for this panel, I did a little research and came across, Lisa, some of the work you did on loaning out devices, which I found fascinating, because I know in our population, if we loaned out something, we probably never got it back. And so I was curious what your return rate was. At least in your paper, it was pretty high, and it seemed like it worked pretty well. That seems to be a lower cost way of arrhythmia monitoring than doing a ZO patch, for example. Sure, so we had multiple devices in those kits. We had two stethoscopes, three ECG devices, and some pulse oximetry devices. So the total cost of those kits was around $2,000, and we gave them a cell phone that had the apps to all of those things connected. The way that we had the 100% return rate is we had them, in the consent, agree that if they didn't send it back, they were responsible for the cost of the entire kit. So I think that that was why we got 100% back. Nobody wanted to pay for something in the research study. I know that my colleague, Nate Miller, also did a study with using the CARDIA ECGs post-ablation, and I think he had a pretty good return rate as well. There was only one lost. I was going to say two. So there was only one that was not returned, and I think he had a similar thing, saying that if you don't return this, this is up to you to pay for. So I think one of the big things in that is having the patient or the patient family know the cost of it, but. Maybe people are more honest than the Midwest. I think on the East Coast, before the days of disposable Holter monitors and those kinds, when they were actual recorders, the patient signed that if they didn't return it, they're responsible for the device, $2,500 a piece, and we lost many of those devices. We continue to lose them, right? We still. I wonder if that has to do with who they think is paying for it. So if it's not, we presented it as it was part of the research study. We upfronted the cost for this, but the hospital pays for it, right? Hospitals have endless amounts of money to a lot of our patients. They think that that's just all that there is that comes out of it. There's no business understanding of what's in and out of the hospital. So for them to lose property of the hospital, I think is no big deal. But it goes back to, right, I think what Bryn's point was. So again, it goes back to money, right? And who owns the product? The thing about loaning somebody these devices, they still have to return them, right? The benefit of owning an Apple Watch or an equivalent is to say you have the ability to record this data at any time, whenever you may have a symptom, right? So I guess it goes back to, maybe the ask is how do we make this more accessible, more cost efficient? Because you also don't need the Apple Watch, right? You need a Fitbit that has an EKG monitor, right? And so how do we, but then is also able to, and I don't know, I don't have kids in school, but now my understanding is kids are not allowed to have smartphones in school because of all these rules. So what, can we create a product or is there a product that allows us to sort of get that information? Do they get special dispensation to have the, like if they, we know that they are a congenital kid that's having chronic arrhythmias. Like how do you, we have to sort of work through that as well now, right? You're fighting with the place where they exist, which is in school. I wonder if some of the accessibility can be mitigated with the IFU discussion, the indications for use discussion from earlier, because if we got proper indications for use, then perhaps insurance, especially like Medicare, would have to pay for it. And so that would probably increase the level of accessibility for lower socioeconomic folks who wouldn't otherwise be able to pay out of pocket. Yeah, I agree with that. I think having a low cost device, and I don't have any brand preference, but a couple of them back there, Cardia and some of the ones for $79 instead of $500 for an Apple Watch is gonna be a lot more cost effective to just prescribe to the patient and be covered by insurance, as well as developing codes for interpretation so that there's physician buy-in that, because if you read a Xeo, you get paid to read the Xeo. If you read a Barty, you get paid to read the Barty. If you read a Cardia device, you get nothing for an Apple Watch. I mean, there are CPT codes, at least for having a discussion with the patient afterwards about the results and any management, but there's nothing just for interpretation. I thought there was a code that you could use for reading the AliveCore. I think there are these two codes that are maybe being used? I think they're technically for discussion, up 20 minutes and then 20 more minutes. Oh, okay, okay. I just read it on the plane today. No, I think he's right. David Slatweiner talked about it, reimbursement of digital data, probably about, maybe it was 2023 HRS, and he did present some codes that you could use, but they primarily were for discussion, billing for MyChart and that kind of thing. As a model for this, how many people at this table take care of patients with devices who live at a distance more than, say, 50 or 100 miles from your center? And do we generally, I think it would be interesting, perhaps, to talk about the model of device follow-up, the finances that underpin that and the opportunities to deploy these types of tools in a similar way in monitoring devices. One thing that I would like to mention also, just in terms of something to think about, is that if we start charging the patients for interpretation, sometimes that decreases the use. We see a lot in our population in Columbus, so we're a Midwestern state, we don't have as good healthcare packages. Patients get a lot of coinsurance payments and they have to pay a fraction of things. A lot of our patients with ICDs and pacemakers just unplug their CareLinks or Latitudes. They don't send in transmissions because they don't want to get a fraction of that $400 interpretation payment every three months, and then they will come see us in clinic, we'll check their device, but they leave their home devices unplugged because they don't want that extra charge. So I think that there's sort of two sides to that coin, just something to think about. And then the other thing I was gonna say was my wishlist, I think, dovetails into the idea of the IFUs and we want to be able to prescribe a cardia that could get paid for by insurance, but the company itself needs to sort of seek out that IFU in the first place. Like all of the companies that make devices, and we've been dealing with this for years in the miniaturization of devices space, is that nothing is really designed for pediatrics, it's all designed for adults and we adapt it, right? And so what can incentivize companies to design something for kids on the first go around instead of having to wait 10, 15 years down the line for them to then make that adjustment later on? I mean, are there similar, you have exclusivity for pharmacological products, are there similar things for INDs that allow patent to go on? Not really. No, we've talked about that in the FDA panels as the carrot for device companies, but there's nothing out there like there are for brand name drugs. I was gonna bring up the, someone asked the question of, can, I think John, you alluded to it, can you do 100% remote monitoring? You implant the device in the hospital and after that, as long as everything's fine, why do they need to come back for their in-person visit? If they don't unplug their CareLink or Latitude or Merlin, we've showed during the pandemic that it can be done. And if they're fine, there's not much more we do in person. Yeah, sometimes I wonder what my death skill is for. I agree. One of the things that we've been talking about is along those same points, both of these points, is going to, trying to go to an alert-based care. We don't need every three month downloads. We don't need them. You get tons of information, but what we need to know is if the lead breaks or the battery goes or they have a lot of AFib or atrial tachycardia or whatever. So Dave Bradley and I are working on programming all of our alerts very carefully to go to an alert-based, because we have the same problem. They just unplug their monitors. So the thing is, though, if we go to that, they have to have them plugged in. So we have to have some sort of understanding that if we go to this alert-based care, we don't charge you every three months, but you have to have this plugged in all the time and we have to kind of keep track of it. But that will do two things. One, that's going to decrease our reimbursement because right now we're getting reimbursed every three months to do this. So we would lose out on some of that revenue. But at the same token, some of our patients aren't following through because they don't want the charge. So we're going to see what happens. That's our kind of wishlist over the next year is to move some of our patients that are not dependent on a pacemaker, just we have a criteria to more of an alert-based care. We're going to start with the sub-Qs. What is the point of checking a sub-Q in clinic? I mean, there just isn't. So we're going to start doing some of those types of things coming up too. You raise an interesting point and I think it comes down to a little bit how much do we trust our algorithms and our devices to do things which historically have been done by nurses and physicians and other APCs. I will say that I was doing a talk a few years ago on pacemaker follow-up and dug into the sort of archives of the literature to find out where the every three month criteria came for pacemaker follow-up and it turns out it was just made up. And has become enshrined in practice and importantly enshrined in billing practice and financial workflows. So I'd like to toss out to people, if there was something that you think we need in this pediatric and congenital space right now, does the first thing that floats into your mind, is it some sort of physical device that we need or is it something in the way of a smart thinking machine that is going to replace expertise at the locus that currently resides which is usually in our field at a large urban medical center? I don't know that you have to choose one or the other. No, but I'm making you. I would choose device because I think... Okay, what device? So I think straight off the bat, if you look at our patient population, I love a device that would allow me to monitor rate and rhythm and adjust it for the patient, whether if it's a congenital that's gonna have a lot of flutter versus an eight-year-old that's having SVT. But I wanna be able to monitor what is... Like when your Apple Watch goes off because you have a resting heart rate above 120, I wanna be able to modify that rate. And I don't want the irregularity of an AFib, I wanna sort of be able to modify what triggers the algorithm. And so I want to be able to manipulate it based on the patient I'm placing it on. That's what I would love, a device that lets me do that. Because I think, to Andy's point, I don't think it can be an algorithm or something that sort of makes decisions for us because our patient population, maybe I'm just sort of, and I'm being naive in this, but I think our patient population is too nuanced for it to sort of fit the general AFib algorithm that exists right now. And so, yes, AFib is the most common arrhythmia, but our patients have such variation that I still don't trust what's out there to make decisions for me. But I do want it to give me good data. Yeah, and I'm not sure if AI will ever be able to do that because a patient population is so nuanced and so heterogeneous, right? Like AFib's the most common and there's like hundreds of thousands of them, and how many TGAs do we have? So let me summarize. Highly programmable event monitoring. Well, you also need the other and the non-technical end of the spectrum, right? You need smaller watch bands, right? To fit around an infant wrist, for example. Charlie says baby-sized watch bands. Yeah, I mean, that's the non, you know, it's the non-digital part. I can have them made by the truckload in Southern China for you and get them to. Exactly, dollar a piece. Charlie likes everything small. Make it smaller, yeah. Lisa Raleigh's been waiting patiently with her finger poised over the button. Yes, thank you. For me, it would be something where we can easily obtain the data and then put it into the electronic medical record. And I think that that is a need for a lot of people as an APC who deals with a lot of the remote monitoring data. This is something that I think is missing because we have a lot of stuff for CIEDs. We do not have stuff for wearable devices. Interoperability. Now, I'm looking at Stephanie, Tina, Lindsay, and Caitlin. We haven't heard anything from that corner of the table. Does one of you or more want to step in with your ideas about what type of device or algorithm you'd like to see? Caitlin. So I should say I'm a journalist, I'm not an MD, but I've been thinking about this as I've been listening because I have a 10 and a 12-year-old who love devices. I think what'd be useful in designing these is to have that user interface for the patient side of things at a level which the kids could participate and understand. And I assume there's nothing like that out there, but I think it'd be interesting to see what things would get the kids to be more willing to participate and also not scare them. The alerts probably would be different than they would be for adults. So maybe gamifying the interface for kids a little bit. Anything else over here? When I was thinking, sorry. I mean, I think we have something for babies, right? We have the Owlet, which babies wear, monitors their SVT, very successful. And now we can write a prescription for it and patients can get it. The patient population that I actually struggle the most with are like the three to 10-year-olds who are way too small for the Apple Watch and their heart rates are way too fast. They can't wear the Owlet. And so we have this, and some of them are too young to say what's going on. So verbally, we need a way to monitor them more effectively. And I think that's the age of population that's really, really challenging for me right now. I think, so I had a conversation with our techs at CHOP and I was like, what would you guys want? And one of the things they mentioned is when they're putting on monitors, people are really ready to get out of there. They just had their appointment, they waited for a long time, they know the plan, but now they want to get out of there. And they mentioned having education for parents and then also for, like you said, for a kid at the kids' level so that when they go home, they're not confused about everything that just happened and they don't know who to reach out to. So, like short videos and content that is engaging and easy to understand. And I think from a kid's standpoint, I wish our devices were just more kid-friendly. They looked more kid-friendly, like Tamagotchi-like, something that they are interested and not afraid of. And so, I wish they didn't have a cell phone they need to carry around with them because a four-year-old doesn't need to carry a cell phone. I wish they had a transmitter that was a watch or something like a Whoop maybe that went on your wrist or on your foot, on your ankle, that if a parent had some kind of concern, it's right there. They can notify and something also that looked cute and friendly to the child. Pikachu. Pikachu. So, what do we perceive? I think it's very interesting that one of the topics that Caitlin's brought up and Kate just amplified on was child-friendly devices should frankly be a super simple hurdle. What are the more general market barriers that we individually in this group perceive to be in the way of adoption, development, adoption, and success of products that might be designed for pediatric and congenital populations? Can I just push back one second? I actually don't. I mean, these products, I find that the kids, even at three, they're really quick at working through the technology. I actually don't. I mean, if we're trying to market to children so they like wearing it, that's a different story. If we're talking about usability and getting the data from these kids, I think it's really just a size thing, right? But I will tell you, I give like a two-year-old an iPad or an Apple Watch, their little fingers are going through that little thing and they're doing it better than we are. Yeah, but they're not logging into Epic. So again, the conversation, the question is, what do we need to create it so we can get good data, right? That's one thing. But then it's about, yes, then the question becomes, how is that data downloaded easily into a workable format? Now, that's a completely different conversation. But it is part of the same need, I agree with you. But the purpose of creating a kid-friendly product with Pikachu on it, I mean, then we're saying, yes, we want the kid to feel good about wearing this. Do we want every kid to have a monitor that's allowing us to get real... That's going to be a crazy amount of data, guys. It's like putting a cartoon camel on a box of cigarettes. The question of repackaging what we do now versus doing something different, I think some of this is repackaging what we do now just for kids, but how do we actually do it better or different, I think is what we should always focus on in moving forward. What would my wish list be? I have interest in the post-operative space. I would love to go from the EKGs, which terrifies nurses and some physicians, and move it into a more of a risk base. Is it sinus? Is it not? Is it jet? Is it not? Is it something else? Is it not? Can you even abstract the information to what we want as opposed to necessarily just the EKG so that we have one less layer of an understanding of what we want to know? Moving from there, can we go from, we have this rhythm, let's add CVP in there. Can we actually increase accuracy? What about moving now into the understanding of physiology with a pulse ox? Do we have too much pulmonary over-circulation or not? We have ways that we can go with adequate data moving in that I think would be incredible if we think about the big picture of improving more detail than I would say in ICU digital platforms so that we can see exactly what's going on with the patient's physiology, understand it better, and then potentially offer treatments. If we can get that into an automated space, I think that would be incredible for standardization of care, especially at hospitals that don't have the sheer numbers of other places. That would be my wish list, one of the things that I hope collaboration works together to move towards. I think there's movement looking at pre-arrest variables. That's pretty hot in the ICU, and so some of the work that you're doing can be integrated with that as well to help make those prediction models better. I'd say if you want to stay in the hospital, this is a pet peeve maybe, is maybe to figure out how to, hopefully it won't offend anyone here, how to get ERs to be able to read EKGs a little bit better so they're not calling my fellows at 3 o'clock in the morning about things that don't need to be called about, or just to help frontline providers deal with these tools that maybe they don't use as frequently as we do. Then I'll answer to your question, how do we overcome the bias. Andy, for the ER thing, I was going to say, I think this is a space where maybe AI could actually be very helpful. The first thing I teach the fellows is, does it look normal or abnormal, like sinus or not sinus? Just having an AI algorithm to maybe not fully interpret the EKGs, but at least potentially prioritize them. We can wait until 6 AM. It's very interesting. This is exactly what Josh and I have been working on. The topics that we might engage in here, obviously we've talked a lot about the bread and butter of pediatric arrhythmia, which is a kid with symptoms that we need to diagnose. We've talked a little bit about congenital heart disease and management of patients in the perioperative period. There's also interesting adult congenital issues over chronically ill people. But screening is a third one. There are only a certain number of pediatric cardiologists and EPs in the world. There are an infinitely larger number of emergency rooms where people report, and a very small percentage of them have potentially lethal conditions that can show up on simple screening measures like an ECG. What we have found is that the approach that Rohan has just proposed is really a smart way to go in algorithm development as well. Not try to duplicate reading of an EKG, but instead say, is this EKG normal or not? Does it have an easily recognizable pattern in it like WPW or long QT syndrome? And what we find is that as far as you can obviously tune these systems, but what we do is we get a very high negative predictive value set. And with that, we can increase the positive predictive value by much less. If the incidence of the problem is 1 in 1,000, we might be able to increase it to 2 in 100. So there's still a lot of false positives, but you can safely eliminate 70% or 80% of the studies that are brought in front of you. And that may be of great assistance out in Oshkosh before they send their ECG in to be read by a fellow at Milwaukee in the middle of the night. Question in Queens, New York. Come and speak into the mic. Hi. I'm Sunduk. I'm from Halifax, north of you, Canada. Yeah, so I totally agree with Nick and John's point and Rohan's point. I think that's where it makes difference because of the scarcity of our expertise. In Canada, we are only 13 pediatric EP in whole Canada. And we cover up to Yukon, even up to the border of Russia. So for our patient, they have to take two days to travel to our center. So we started with telehealth 25 years back with a transmitted echo, person doing echo there, and we are seeing in Halifax. Embarking on that, we also see a lot of patients who travel just because of the murmur. So we find that the murmur, patients with the murmur, 80% are normal, even more than 80%, probably 90% normal. But there are sort of a lot of resources used for traveling those patients to our center. So coming to the simple solution, rather than teaching them different murmurs, we came up that whether it's making murmur, no murmur. Innocent murmur versus pathological murmur. And actually, the data so far we have, we find that just doing simple, that solution, even parents can do it. So we built a product, and we founded a company based on that. And we are using that, and we are just in training and testing data. I'm not an AI expert, but my colleague is here. He's an AI expert. And we can reduce 80% of those referrals just based on a simple algorithm. Rather than going into, oh, this is a solid murmur. This is a grade 3, 4, 5. Rather than teaching that, just building simple, murmur, no murmur, innocent or pathological murmur. And those patients with pathological murmur, even we have high sensitivity. Even specificity is a bit low. That's okay. But still, we can reduce 80% of those referrals. And we also have a collaboration with another company. Instead of having digital stethoscope, we can use the phone. So phone can record heart sounds and send it to. And in a second, they can know whether it's the same with the ECG. Also, ECG can be used in addition to that. That's great. I think democratization of the expertise is a really important aspect of this. Now, we're all going to be experts in artificial intelligence soon. My partner and friend of many years, Charlie Berul, can attest that I've artificially been intelligent for years. We're coming up to the top of the hour. And I'm going to actually ask Charlie if you wouldn't mind kind of summarizing some of the topics that you heard discussed today. And I would say myself, this has been far exceeded my expectations. Perfect size group. Very interactive. And a great topic. So, Charlie? Thanks, John. He always makes me do his work. John and I were like a married couple for 15 years. Shared an office and shared an assistant. And, yeah, we're now missing each other, not working together the past 15. So I think the main issues and challenges that we talked about today are digital health and how we integrate that into the electronic health record, things that aren't in there now. Specifically, things like consumer-offered ECG products, Cardia, Apple Watch, et cetera. That's one challenge. And the burden on the clinical staff of interpreting those, the challenges of lack of reimbursement for that, we talked about a little briefly and was a little off topic. I think that was one of the big issues. We talked about how we can make these devices more appropriate for pediatrics, including the product size and features, making it more child-friendly was brought up. And most importantly, but also probably most challenging, is making the algorithms more pediatric-specific, which I think is the biggest challenge for artificial intelligence in that not every child is the same and not every child's arrhythmia is the same. And so that's going to be the hard challenge, whereas AFib, for the most part, is AFib. The ventricular rate may vary, but AFib looks like AFib. So I think that's going to be the biggest challenge for artificial intelligence and bringing it down to pediatrics, as well as the common challenge of pediatrics, which is market size is small. And so from a business standpoint, these companies are going to be less incentivized to develop pediatric-specific products. That's my summary. Great. Any final comments that people want to offer? So we talked a lot about rhythm, which I think makes sense because we're electrophysiologists. But when we think about the pediatric population and just thinking about what Nick had sort of talked about, can we use technology to sort of look at other things? Another patient population, this using AI might be super helpful, would be our high-risk single ventricle population, right? The things like the WHOOP, there was a study that was published that could predict when women were pregnant, before they were pregnant, right, based on their heart rates. Like, I feel like this very high-risk single ventricle population is something that we should probably think about. Can we use all this technology and AI to identify before they get sick so they don't have to, you know, spend their interstage in the hospital and actually, right, so they're not traveling hundreds of miles. Or almost a hospital at home. Nick, I detected that you were about to reach for the mic as well. I made just something that I find very exciting for the future. A colleague of mine mentioned that I thought I would put out there so you guys think about it too. You know, how much can we get from the EKG? I think this isn't something that we know. And one of the things he found, one of my colleagues found, is that he could even tell with vaping. Like, could we use EKG screening for things that we don't necessarily think of as cardiac? Like, a screening for someone that's vaping. Instead of a depression index, do you actually say, well, let's look for a vaping index on the EKG? Like, these unique areas that have these implications that we haven't even thought of. And I'm really excited about the potential for stuff like that. Go back to measuring the pulse. I'm also really excited to see if we can gather all this information that we're getting. So, like, serial holters and come up with a summary. Like, use AI to make a summary for these patients. And these congenital patients who are now moving on to adult care, can we have a summary of their first 18 years of holters to, like, provide to the next physician so that they have all the relevant information they need? On a credit card. I think that's a beautiful idea. I mean, and it would not be incredibly hard to do. It's, you know, for the congenital heart population. The practice at our hospital for many years, and I think amongst many pediatric cardiologists, is there's a section in every letter that you send from clinic, which is kind of a growing ladder of every test, every finding, every intervention that they've had, which sometimes can become so long that it's never read. And the opportunity, I mean, this is the next application for CHAT-GPT. The opportunity to synthesize that into a coherent and digestible story would be something that would be very easy to do, and it could be managed around waveforms and electrocardiographic patterns as well. There's a sudden hubbub in the room. I think all of us are free to stand around here and chat more as we want, but if you have an urgent need to get up and wander around or do the next thing, thank you all for attending today.

AI in pediatrics

machine learning

congenital heart disease

regulatory clearance

pediatric-specific algorithms

affordable monitoring devices

digital health tools

interoperable devices

diagnostics