We have an amazing audience that woke up early. I guess they didn't really Stay up too late last night I'm so delighted to have this distinguished panel with me and we're going to be talking about something that's top of mind for a lot of us here PPG PPG signal is something that's been around a long time, but recently because of the development in wearables and smartphones and other Wearable patches it's become a ubiquitous, you know clinical practice so I'm excited to kind of you know, have the experts in the field and Pick their brain about what PPG is and how it could be used. So Without further ado. I want them to go around and introduce themselves And good morning, thank you all for joining us, my name is Dan Canelon I'm an EP cardiologist by background was at the Cleveland Clinic for 16 years before I joined Massimo We are a medical device company based in Irvine, California that specializes in non-invasive monitoring and one of the largest Manufacturers globally for pulse oximetry and PPG type sensors My name is Joseph Safia Harp, I'm the founder and CEO of happy tech We detect atrial fibrillation using only the smartphone with no additional hardware we're conducting large screening studies in the Netherlands over 160,000 patients and The beauty of the technology is that it's accessible and scalable because you need nothing else Good morning. My name is Ben Green. I am the senior vice president of services at a live core I'm a family physician by background and I lead a live cores clinical services, which includes cardiology overreads cardiology Services and then that support our devices which are ECG based devices Have a lot of experience using PPGs over the years as a house call provider for many years so we can we can dive into that later Hi, my name is Chris Lee. I'm a serial medical device entrepreneur currently the CEO and chairman Huxley medical we commercialize a multi-diagnostic platform that includes multiple sensors including both PPG and ECG on the chest and really the idea is to use the redundancy of multiple signals to come up with multiple diagnosis to streamline care Good morning, everyone. My name is Lars I'm one of the co-founders and CEO of FibriCheck. FibriCheck delivers a regulated AI platform that converts consumer devices into medical devices Main focus today is utilizing PPG for cardiac arrhythmias and that's where we want to contribute to this panel Well, thank you so much. So let's start with You know educating us a little bit about what PPG is and How actually PPG signal varies depending on the acquisition of the PPG signal? So we'll start with that Maybe you could tell us a little bit about how you think about PPG and and how it varies from you know different hardwares that you actually produce so sort of the basics of PPG is that you are looking at the the arterial waveform Passing through the tissue of focus You're using a light emitting diode LED. You've got a couple different ways You can get the the signal you can do transmission where you're pushing from an emitter and you have a detector You can do reflection where you have a source and then some peripheral receivers in which the signals derived And and I think for cardiologists that maybe are a little bit less Familiar with sort of optical sensors and optical science you can think of it as sort of the same principles apply to Ultrasound to to you know, Doppler echocardiograms, you know, the the wave mechanics are very similar And so there's a lot of different ways that you can use The the technology not just things like arterial oxygen saturation But, you know, for example, my company has a technology where we can non invasively assay the hemoglobin molecule by pushing light and understanding the the The absorptive characteristics of hemoglobin we can assay that non-invasively continuously within one gram per deciliter Specification so, you know light and optical sensors are very powerful tools indeed I Can comment on the smartphone based PPG and what makes it really interesting is the variety of the sensors so you can imagine every different phone has a different camera and That leads to a wide variety of different signals that you can obtain and there's a couple of components in generating the PPG signal One is the user themselves so they can have blood blood perfusion or they can have a very cold fingers and And it's the combination usually of the user's finger and the camera lens that will generate the PPG signal which also is one of the challenges of smartphone based PPG and making sure that You're able to generate a good quality signal every time Whereas for example massimo's devices are made for this purpose the smartphones initially were not So this is some of the things we're working on every day to try to figure out how to generate that kind of signal Because with a good quality signal you can also lead to a good Ben maybe you can tell us a little bit about how PPG varies in quality but also in quantity and when you compare it to ECG how how should we think about ECG and PPG together? Are they complementary? Are they competing against each other? Yeah, I certainly think they're complementary I think you know, the the acquisition of PPG is important I think the acquisition of PPG is very important. I mean, you know, it's it's it's a it's it's it's it's The acquisition of PPG is important I think to call out the beauty of PPG is the the ability to detect You know results in a variety of sites, you know as we mentioned there's there's fingertip applications There's smartphone applications. There's video applications And really almost any part of the body can can can generate a PPG signals using various detectors You know as we look at the difference between a PPG and an ECG I think you know today as we know For a fib diagnosis and for for other arrhythmia diagnosis ECG remains remains the gold standard remains the diagnostic tool of choice, however PPG is a great way to detect detect a rhythm is to detect in a potential a fib So so it is a continuum and I like to look at as a continuum where PPG can can detect abnormalities ECGs and now the ability to easily capture 30-second ECGs You know, it does prevent a present a complementary relationship between between the two That's that's great Chris Maybe you could tell us a little bit about how you know acquisition site from you know going from fingertip to smartphone and you know a patch can actually like change the quality of the signal and how should we think about That signal that comes from a different place from your patch Yeah, no, that's a good question so we've collaborated with a lot of academic and clinical advisors to kind of think through this so as You mentioned site has a big part and you know, there's a lot of variables when you think about physiological site There's obviously skin tone. I think that's been a hot topic for the FDA and just regulatory in general Over the past few months few years, but probably even more importantly is the tissue composition So when we look at perfusion when we look at density of the capillaries and the fascicular bed that's underneath It's gonna vary quite a bit not only by site, but also by age by gender And also by ethnicity So I think when we think through this It's tough to use one approach or one technology to be able to overcome these different challenges and the noises that are That you're gonna see with this level of variability So really when we designed our device and when we think through these technologies We try to use a multiple approach whether that's sensor innovation. That's form factor innovation. That's algorithm innovation We kind of combine all three of these to overcome the variability that you have from different sites Okay What are your thoughts about the differences on acquisition sites and and the quality of the signals that you're getting The The PPG signal is a very difficult signal to interpret Because the morphology of the signal is quite low in frequency spectrum compared to ECG, which has clear signatures That means that noise Confounds the interpretation and noise can come from different regions So if you're not able to identify the noise from the potential rhythm problem, you have a lot of interpretation issues So the way that you need to look at those signals and where data comes from and needs an older device Acquisition capabilities needs to be managed in one way or another that you harmonize what it comes in may be junk but what comes out needs to be clear and That is a key asset to understanding which PPG you can use or cannot use and that that is a crucial component of quality model And PPG will make the difference in acceptance compared to ECG, which is a different methodology of getting insights and information That was fantastic kind of segues into my next Topic that I want to explore with this panel You know as you heard acquisition of PPG is different using different technologies So that leads to you know, you can monitor PPG continuously or you could do a spot check So, can you maybe talk a little bit about what that means? Particularly for detection of arrhythmias and you know the one that we care, you know We really think about a lot of this is a sharp revelation. How should we think about that? You know on you know spot check versus continuous and what are the implications? That's a great question And I just want to come back to the previous comments that were really excellent in terms of summarizing some of the limitations and challenges There's so much Signal processing that's needed for the PPG waveform in order to get it, right you have to remove the corrupting signal from noise and some of the biggest challenges that exist with with PPG which gets back to your question are under those states of low perfusion, right because You're measuring the ratio of arterial blood to venous blood both of which is you know being in the path of The path link of the light that's passing through the tissue. And so when we talk about what that means For a wearable device, you know, like a like a watch or something that is being worn continuously It has to be able to pick up that signal, you know during motion During states of low perfusion, you know Or when the cardiac output is low and as was mentioned earlier across a plurality of skin tones and so That's where you know when you're talking about continuous sensors like the ones that we use in in hospitals and in neonates You know You can't ask the baby to hold still so you can measure the SpO2 the baby's screaming and waving their arms and legs around And you have to have sophisticated Components of the system. So for example in our technology to to to address the issue of Skin pigment we have an auto amplification of our light source and so we can control the light Amplification on the emitter, but also we have an automatic gain control on the detector. So it helps us, you know accurately Measure that signal earring even during those states of low perfusion where there's more absorption, you know with darker skin pigments And so as we think about what that means for continuous versus spot check You have to understand that that's going to have implications if you're using irregularity in the PPG waveform in order to Detect a potential arrhythmia. And so I agree really what what you said earlier Ben that you know PPG and ECG really are complementary rather than competitive and as we know the most recent Guidelines that were published last year from the professional societies indicate that you do need to have that a Visual inspection of the ECG waveform to confirm the diagnosis, but PPG can become a really useful tool to screen potential Arrhythmias based on the irregularity of the pulsation But you know what I've seen is that it's very difficult Extremely difficult to look at the signature for PPG of sinus rhythm with very frequent PACs or runs of you know bursts of atrial tachycardia Versus just you know straight out a fib especially when you know as many cases the atrial fibrillation is very regularized and so that's where I think It becomes extremely difficult to get a very high precision on a fib with just PPG alone And then what you mentioned You know that has implications for over saturating the signal when you're increasing light intensity and battery depletion, too So there's a lot of things you have to think about on the engineering side By the way, I could listen to this for an hour We should be more PPG sessions Thanks for the insight. So maybe just a comment on spot versus Continuous if somebody has a watch then they don't need a spot check because it it kind of solves the needs of the patient of monitoring themselves But if they don't and they are in a rural area four hours away from hospital or on another continent Then a spot check solution based on a smartphone would be a good solution. So it really depends on what is available and Of course that continues is always the ideal solution and you can also detect a fib burden and more information but if it's not available, then You can in the same way that ECG and PPG are complementary a spot check and a complement continuous Monitoring are also complementary to each other depending on what is available But Ben, how are you how are you thinking about this as you know We starting to think maybe that the burden of atrial fibrillation is probably important as a target of therapy But also prognostically how are you thinking about PPG and ECG? Particularly when you think about ECGs that are longer than 30 seconds Yeah, and I think the the continuous versus spot check Element of PPG is is similar in the in the world of ECG and we we we look at that as well You know with with PPG and with the ability to do continuous monitoring while that certainly expands the the Potential for capturing ECG or capturing a fib rather You know, we know there's there's a lot of false positives and we see that with you know with ECG as well And we're also catching low burden Low burden AF. So the longer you monitor, you know, we saw that with the loop loop study That we're detecting lower burden AF that may not need early intervention And I think that's that's controversial right now and that that is why the AF screening guidelines are not yet Universal and I think I think we're seeing the same thing with PPG. We're able now to detect abnormalities With comfort confirmatory ECGs rapidly, which is great But are we are we finding the AFib that that needs to be treated urgently? And I think that's that's the big debate here just because we can can detect it early It's our responsibility to have the right pathways in place for you for follow-up but for management of those patients What do you Think about acquiring signal simultaneously and ECG maybe Combined with the PPG signal it do you think that there is value in something like that or particularly when it comes to like detecting? Other arrhythmias or you know ruling out false positives. What are your thoughts on? Yeah, I think anytime you have another modality That's not dependent on each other. It just provides a robustness, you know detection and diagnosis of certain Arrhythmias and other types of disorders that may be associated with it when we think about You know kind of in the spirit of this conference HRX in the spirit of you know, not where are we now? But where will where will we be five ten years from now? I think we know that algorithm development sensor development is going to rapidly improve over the next five ten years So it's not for me to say what we're able to do with these technologies now But I think when we look at what's not going to change five to ten years from now I think we know for sure that the signals that we're getting from PPG or just it's a very robust signal, right? You're getting not only really valuable Cardiovascular information, but you're getting great respiration In other signals and other data that's really rich in that and I think there's still a lot of development Involved to figure out and really deconvolute what's in that but I think the other thing that's not going to change Is that it's still a very peripheral signal. It's not a direct measurement Essentially on the heart so it's going to be a lot more susceptible to noise It's gonna be a little bit trickier to analyze, but I think with the right innovative approach Yes, you can use both signals simultaneously to come up with not only better detection, but also better management And maybe provide context around what you're actually seeing right atrial fibrillation Combined with level of activity for example or quality of sleep I think I just can you know, really you can understand the clinical situation a lot better and make better clinical decisions and problem Lars, how are you thinking about a spot check versus a continuous PPG monitoring? Do you think that that should really be clinician driven or is it kept? Yeah Well, it depends on the patient population that you're trying to deploy the technology in I Would first like to take one step back on addressing that question because we we all use PPG But we all use it differently PPG is a technology It is not a methodology to determine if you have an arrhythmia like ECG for instance is very standardized you have standardization how to acquire an ECG and Independently of the acquisition you can develop algorithms to actually extract the information and a clinician is trained how to read the ECG When you look at PPG today We know it in the form factor of something is being measured and a number gets out of there or there is an alert being raised and that is how PPG is perceived today and that exact exactly is the challenge that the accuracy of the detection of AF and differentiation between other arrhythmias versus AF is therefore overlooked in the capability of actually looking deeper at the signal capability to Identify those other problems. So addressing the question. Do you need a spot check or seat of continuous for me? The question is then how would you actually utilize this and what would be the main use case? Are you approaching PPG as a way to say? Oh Something is possibly wrong send them to a diagnostic or I am actually creating a halter concept using PPG and making it Interpretable to assess based on the PPG alone And I'm a real fan of the latter one because I believe there's much more than meets the eye That is actually not yet understood fully or not yet Perceived in practice today yet And I think there's a lot of untapped potential in that area that PPG can bring much further than just a fib detection Which is actually the easy part to start with actually then the journey starts to get you to get deeper And you would double click on that a little bit more for me Like what do you mean? It's a different arrhythmias or am I gonna be able to drive additional information like, you know with ECG for example I can tell if somebody has heart failure or maybe aortic stenosis with you know with AI Are you is that what you're alluding to? I think it's a vertical and horizontal question So PPG has the potential to dive deeper in the vertical arrhythmia space not only to say hey, these are ectopies But we're able to identify which type of ectopies they are They're not able to localize it like an ECG can do in the ventricles around the atrium on a specific place But it's maybe one step deeper than actually is perceived today next to this as my colleague just Educated is that the cardiovascular biomarkers are encoded in a signal We cannot see them with a naked eye, but the AI can definitely identify those things So when you start utilizing PPG, for instance to determine Age of a person you're actually able to extract the age of a person out of a single PPG reading So that means that that information cardiovascular wise is encoded there and there's much more in there that can be used and combined to Evolve further, but that's that's a different road and really addressing the problem of arrhythmias in helping patients today I love that. I do agree with you that there's a lot more to the PPG signal than just arrhythmia detection So I want to go back and think about the you know, patient selection a little bit and the comments that Dan made can you Kind of you know educate us a little bit about how should we think about the point this technology? Particularly if you could you know, maybe dive a little bit deeper into the skin tone question We know that there's definitely issues with a pulse proximity and and You know, the readings can be sometimes not as accurate. How should we think about that when it comes to AF detection So, how are you thinking about that does that matter where the acquisition of the signal is coming from or you know Are there other things that play here? You know you you alluded to this earlier when you were talking about some of the pitfalls of amplifying the light source and you know Adjusting, you know controls and the emitter and the detector. It has to work a little bit like the automatic transmission your car You know, these are very sophisticated sensors. These are very sophisticated instruments and we have validated You know the performance of our of our sensors Across a plurality of skin tones to prove that they are quite accurate even during circumstances of low perfusion. And we have ongoing clinical studies at Medical University of South Carolina and critically ill adults. And there's an FDA-funded study that's ongoing at Stanford in pediatric patients, which hopefully we should see some results out of that early next year. But to get back to your question about the patient selection piece, I think where all of the signal that's embedded in the PPG waveform that can be extracted with sophisticated algorithms is amplified exponentially when you think about the possibility of blending your sensors. Because these are complementary modalities. You have electrical sensors like ECG. You have optical sensors that do PPG. You have acoustic sensors. You have gyrocellarometers. You have thermistors. And so the way we think about that at Massimo is that, look, for example, in our next generation inpatient hospital wearable system, when we're trying to detect something like tachypnea, which is the most sensitive marker of clinical deterioration, whether you're talking about heart failure, DVT, PE, opiate-induced respiratory depression, just relying on the pleth waveform to give you the respiratory rate can be a pitfall. So what we do with our multi-parameter solution is that we measure the respiratory rate from the pleth waveform on the finger, from the impedance on the chest with the ECG electrodes, with the gyrocellarometer in the puck so we can look at the excursion of the chest wall, and also with an optional acoustic sensor that we plug in. And so what you're doing is by going multi-modality, it allows you to report to the clinician, OK, what is the one true respiratory rate? Because you can relatively easily, without using a lot of sophisticated AI or advanced transformer models, you can actually relatively easily report an accurate respiratory rate in real time that the clinician can use. And it overcomes some of the inherent limitations that you're going to see with motion artifact on the finger or some of the other things which we know affect impedance measurements on the chest, for example. Yeah, I kind of agree more with Dan on that comment. I think that's what we're talking about in five years, right? We're not talking about either ECG or PPG. We're talking about devices that do all of it. And because they do all of it, you're just going to get a level. And I think we kind of danced around this, but then say explicitly level of specificity that you're just not getting currently. Because you have different redundancy in these different optical and electrical type measurements that you're just able to eliminate noise, and you're just able to get a lot more specificity. And what you're trying to measure, and also a lot more context in what you're trying to measure as well. Yosef, what do you think? PPG is being used probably only people on this panel and people that are here are used to looking at it. Do you see a future where PPG could be used easily? And clinicians, regular clinicians like electrophysiologists can look at it and say, you know, this is atrial fibrillation or other arrhythmias? Or do you think that's never going to happen? That's going to be ECG, it has to be there to confirm it? I'll have Ben respond to that. It's a great question, Hamid, and I'd like to add on what great things Lars said. I feel right now PPG is like French and ECG is English. Everybody can speak English, but French is still in a very small group of people. And what we've seen, we're deployed across, I think, about 14 hospitals in the Netherlands, is that once, let's say, the Holter technician team start to learn looking at PPG, they're able to reduce the number of false positives. They are able to see other kinds of arrhythmias by looking at the PPG graph. So I think learning to have, let's say, more information available on the types of PPG signals and how they look like, but also presenting them in a way that they are used to in a tachogram or a Poincaré plot so that information is available. And then maybe the second point to that is really choosing your patient population as correctly as possible based on the device characteristics and, of course, the guidelines that are there. So I think if there's more educational on PPG signals and how you can look at them and how you can interpret it, and if the devices also become more accurate, there's more evidence. Right now, the evidence on PPG, ECG, there was a multi-analysis that showed the studies were still on a small scale. That could really help drive PPG having a closer seat to the table next to ECG. But what do you think? Do you think there's going to be a future where I'm just going to look at a PPG and be like, no, I don't need anything to reduce? Well, I think potentially. I think, though it also depends on the use case, it depends why we're deploying the PPG in the first place. Is it for diagnosing new a-fib? Is it for managing someone with existing a-fib? And I think, as Joseph mentioned, the clinical validation studies need to be done, more of those to show that PPG can be accurately used for diagnosing arrhythmias, a-fib in particular. For monitoring someone with established a-fib, however, or established arrhythmias, I think there's more near-term application of PPG, where you have an individual who has a-fib, and they're able to now use a PPG to detect recurrences. And I think that's where physicians and other providers are able to use these technologies and these wearables to help manage an ongoing condition like that arrhythmia. Actually, I mentioned earlier I used to do house calls. I remember I had a patient with a-fib, and I used to carry around a pulse oximeter with me. It may have been a Massimo device. We didn't have ECG, we didn't have PPG that did any arrhythmia detection, but I left a pulse oximeter with her, and in consultation with her cardiologist, we were able to keep her at home. She would go to the ER very frequently for anxiety and for mild symptoms, but because she had that heart rate monitor, it kept her at home. So I think things like that, where we're able to use technology, use things like PPG to help a patient better understand their condition and help the provider better manage that patient, I mean, that's what I'm really excited about. I'm impressed that you were used to do house calls. That's, we have to talk about that again. We'll talk about that later. Lars, do you see an opportunity for standardizing PPG signal? And if you do, who should lead this? Should it be HRS, should it be industry, or should we work together? How should we think about this? I would love to be on the committee that helps standardize PPG for arrhythmias. I think it's a combination of the different stakeholders to get this done, together with an IEEE kind of organization behind it to make sure that this is really getting out there. Because what you see is that once the technology is being used in routine clinical practice and you have patient population selection, then they're even talking about avoiding performing ECGs, avoiding taking holders, avoiding prescribing a patch. So the acceptance is there once they start to get the trust and the feeling how the technology works. But the challenge will be is that everybody's coming with a different solution, and therefore you always have this acceptance barrier to get over there. So that standardization will be key, and that needs to be done with the different stakeholders. And industry is definitely one part, clinicians, and then international organization to streamline this like HRS or an IEEE kind of institute. Perhaps a large library of ECG and PPG signals and educational efforts that can educate people much earlier on so they become comfortable with it. That would be great. Well, we can't have a panel without talking about AI. So what do you think role of AI is in, well, acquisition of PPG, cleaning up the signal, and improving the quality of it, but also in interpretation? So get your thoughts on it, Dan. Yeah, so I think really two things. I just wanna come back very briefly because the optical sensors, I would encourage everyone to think beyond just PPG because I think with different wavelengths of light, there's a lot of sophistication to optical sensors beyond just the tachograms and that signal. For example, I'm talking about existing technology, not futuristic stuff. We have spot check devices that go all over the world that are robust, that can go to, with EMS first responders to burn victims, and we can measure carbon monoxide levels in the blood non-invasively, spot check accurately in the field, and that can be used to triage people. So this, it's a lot bigger than AFib when you think about the implications of being able to screen people for anemia. For example, a blood donation centers all over the world using our technology that assays the hemoglobin molecule. So I think there's a lot of digital biomarkers, and I wanted to call that out because I didn't wanna gloss over that completely before we move on to AI. Because look, I think the capacity, the ceiling is quite high, right? Obviously for what you can do when you build those algorithms and you layer them on top of your signal, but again, I think you have to keep coming back to the importance of the hardware and the importance of not just the signal acquisition, but the signal processing. Because I think if you don't really get that part right, I would argue respectfully that it becomes an algorithm that is a garbage in, garbage out. You're building AI on top of a poor foundation, and so I think that's where we hang our hat as a company in our accuracy through motion, through low perfusion, through oxygen desaturation, across a plurality of skin tones, both continuously and SpioCheck, because those are quite different. The finger, the anatomy of the human figure was like it mostly was built for photoplath, right? Because just the conical shape and the symmetry and the fact that you can align an emitter and a detector, it doesn't get any better than that. Whereas when you look at even among the best companies in the world that manufacture optical sensors, you're gonna lose some accuracy when you go from the finger to the forehead or to the earlobe. And so we just have to keep in mind these big categorical differences between transmission and reflection and the types of measurements, because these are the details that matter when you're building algorithms, because you can't overlook those things if you're trying to capture a certain signal of interest. I wanna have Chris respond to that, because you're obviously acquiring differently from the chest, so how are you thinking about the quality of the signal from where you're getting it? Yeah, we completely agree. It's when you acquire it from the chest it's an order of magnitude less, maybe even two orders of magnitude less profuse than what you get at the finger. The fact that you're using a reflectance mode versus transmissive obviously leads to additional noise. So I think we think of it exactly the same way. So how do you overcome that? How do you use different configurations of the sensors? How do you use different configurations of how that sensor transmits and absorbs light? How do you then process that signal? It's all about the fundamental, how do we improve that initial signal to noise ratio before you apply any other algorithms to analyze that signal? So I couldn't agree more with that comment. Lars, what about you? You're getting it from the smartphone. It's obviously not better. Very fascinating question. And the smartphone is not made to acquire PPG, but still it does actually an excellent job at doing it for what we need it. That's important. CO detection or carbon monoxide detection, that is something that will not be possible to smartphone and flashlight, but for the space in heart rhythm monitoring, what we've seen, we're compatible. I have now 600,000 users, 17 million recordings that are annotated that compiles data from 6,500 different smartphones. And there's no difference in diagnostic accuracy across those phones, but there is a difference in quality across those phones. So the question is then, how do you manage an imperfect sensor to get to a clinical grade result? And again, that quality assessment, there the AI plays a crucial role in making that possible to make the selection, which data is good and which data is not good. So I'm a fan of using sensors out there and actually using the AI to leverage them to a higher level that they were not capable before, rather than developing the perfect sensor for the use case that we're currently in. But that's use case driven for me. So I'm very interested in that viewpoint. That's awesome. Yeah, so how, okay, let's agree that, where you get the signal and the input is important. But how do we, should we think about applying AI once we have our signal? Obviously it matters where you get it from, but how should we think about that? So I think as AI here also as a tool, like a recipe where you can apply it to the different components of the algorithm, whether it's the acquisition, the cleaning of the signal, the interpretation. So you can use various AI models to do that. And what the panel just mentioned also is the quality input in and input out that it needs to be good. And so AI models are really, really important, but making sure that the user is able to use the device properly is also a very, very important component, especially when we're dealing with very vulnerable patients that are 75 plus, 80 plus, and designing your solution in a way that they are able to interact with it properly. And if you do it in that way, you increase the chances of starting with a good quality signal. And then the options are endless in terms of the different AI models that you can apply to analyze the signal. So there's like thousands of combinations and it's kind of like a recipe where you try, let me try this, let's try that, and then find what's most optimal. But that we can go for an hour on that topic as well. We can, you know, we just go over. You at AliveCorp have been leading in with AI models, particularly when it comes to ECG interpretation. So how are you thinking about, you know, actually having the human out of the loop when you apply AI models, when it comes to, you know, rendering a diagnosis? Do you think that's a good idea? I mean, obviously, I won't answer for you, but do you think it's a good idea? Well, you know, we use humans in our solutions today. I think there's absolutely a role for humans in whether it's AF detection, you know, as Dan mentioned, there's a lot of other opportunities with PPG and ECG for that matter, and using advanced AI to pull out more information, with PPG specifically, I think one of the hopes and promises is we're able to really distinguish different AFib and AFib mimics, you know, from the signal. But having a human in the loop to really shepherd in this new era, just like we've done with ECG, and especially, you know, consumer-acquired ECGs, you know, having a human to verify that ECG, verify that AI determination, I think is critically important. We, at AliveCorps, we offer that service to our consumers, so when an individual gets an AFib determination, and they want a human to validate that, we offer that, and I think that's the same thing with PPG. I think we need to better educate providers across the continuum on what these signals are providing, because there's rich data, but there's also the worry that we're over-promising value, we're over-promising that we can do a lot of things, but, you know, I'm excited about what's possible. You know, with obstructive sleep apnea, for instance, you know, PPG is quite readily being accelerated, so there's a lot of opportunity, but, again, it's how do we, you know, cautiously and responsibly, you know, interpret that data for patients. Well, you know, this sounds amazing. It sounds like the PPG is what I should be using all the time for arrhythmia detection, so why are we using it all the time? What's stopping us? So maybe I can go around and kind of talk about, like, what are the barriers to adoption of this technology worldwide right now? It's those pesky cardiac arrhythmias. You know, I mean, think about, like, all the challenges, you know, and a huge kudos to Ben and his colleagues at AliveCorps for really the work that they've done, you know, with the automated detection algorithms, and I agree with everything you said about, you know, the importance of the human overread, and I don't think that's going away anytime soon, but there's just the, you know, and Dave posts a lot, you know, on social media about these really interesting tracings, very, very, you know, there's so much, you know, we've been talking a lot about all of the parameters that go with, you know, the engineering of optical sciences, but, you know, we've almost completely neglected the pathophysiology of the disease, right, which is quite heterogeneous and difficult, and so I think, you know, to answer your question, it's because it's very hard. It's very hard to get the type of, you know, diagnostic accuracy that you're gonna need for a standalone arrhythmia detection with PPG by itself, because, again, it's those very challenging arrhythmias, you know, the differentiation between sinus rhythm with, you know, more ectopic P waves on the screen than there are native sinus beats, and then you've got other situations where the baseline is very isoelectric with AFib, and it's quite regular, and so those are, when you're talking about the plough waveform, that becomes extremely challenging. I'm not saying it's impossible to get it to, you know, as they say in engineering, to the nines, but it's very difficult, and so I think we see, and I think a lot of people see, blended sensors as the way past that barrier, because, as I made the analogy earlier with, you know, the different ways we can acquire respiratory rate and then you can layer algorithms on top of that and say, okay, well, you know, the plough waveform right now is giving me noise, so I'm gonna look and see what's happening with the impedance on the chest. Okay, well, the impedance is off. Let's go to the acoustic sensor, and so what you can do is you can use that to leverage the strengths against the weaknesses of each of these different diagnostic modalities and sensors that I mentioned, so it's not to say that it's impossible that you're gonna have a highly reliable system, you know, for AF or arrhythmia detection using PPG only. It's just that it's gonna be a lot easier if you give yourself some flexibility to look at other, you know, sensor modalities in parallel. Great, we have two minutes, so maybe I can go around and get final thoughts on what do you think is stopping us from adopting it? There is a saying in cardiology, right? In God we trust, all others must show data, so we still have limited data on PPG clinical validation studies. That's one, and defining the right use cases where you show that clinical data. The good news is that the European Society of Cardiology, I think two years ago or one year ago, they put out a practical guide to managing arrhythmias using digital devices. I highly recommend you read that paper. Talk about when to use PPG, when to use ECG, I think it's a great article. It would be amazing if HRS would do something like that as well. And we've seen that at least in Europe, the number of PPG sessions increasing during the European Society of Cardiology or European Heart Rhythm Association, so having publications, presentations on PPG will help aware and I'll stop here. Great, thank you. We have one minute, so go ahead. Yeah, I absolutely echo what Josef just said. That paper's a fantastic one. ESC just came out with their latest recommendations on AFib management. And still, they're not yet ready to say PPG can be used for diagnosing AF. And why is that? Well, I think even though the advancements in all the hardware and the software are accelerating rapidly, we're still unfortunately seeing too many false positives and motion artifact that's not yet hitting that data bar and that data validation bar. And not to say that that's not gonna happen. It's just at this time, we're not ready to do that. But as I mentioned before, it's a great tool for helping manage patients with diagnosed conditions today. Chris? It's back to what Ben said. I think you have a lot of noise in terms of physiology, in terms of variability of how do you get that signal, including other arrhythmias, in terms of affecting your ability to get AFib. And then also variability in the measurement technique, like the PPG on your watch, on your O-rings, very different from what you're gonna get from, you know, a dance company. Lars, where are you? I just wanna end with a question. What study is missing to demonstrate that this is possible? Because data tells a story, but there's nobody who can tell me this is the study that we need to do to show this. So what study is missing? That's my final comment. Yeah, so that's terrific. Well, this was a fantastic panel. Thank you so much for taking your time. I definitely learned a ton, and please stay back and ask questions. Thank you.

Photoplethysmography

PPG technology

arrhythmias

atrial fibrillation

wearables

AI

clinical validation

electrocardiogram