All right, good afternoon, ladies and gentlemen. My name is D.J. Lakharedi. I'm a cardiac electrophysiologist from the Kansas City Heart Atom Institute, Kansas City. And I'm joined here by Dr. Roy John, my co-chair from Stanford University. So this is an exciting session that really talks about some of the innovations in the hemodynamic aspects of cardiac devices. Again, when you really look at the cardiac device space, innovation kind of stopped about 10, 15 years ago, and there has really been not a whole lot going on here. Everything that we are inhaling and drinking and breathing, everything happens to be PFA. For a change, we're going to talk about some exciting innovations in the space of cardiac devices. So with that, I would introduce our first speaker, Dr. Chandra Boma, from the Advent Health Medical Group, Cardiology and Cardiac Electrophysiology, who's going to talk about clinical outcomes after ICD therapies. This is the mouse. So go to start. Good afternoon, everyone. I thank HRS for giving me opportunity to discuss this topic. I got about eight minutes. I'm going to try to do my best. And the topic, obviously, I've seen, clinical outcomes after ICD therapies, and ICD is basically implantable cardioverter defibrillator, but standard AICDs are the ones which we know most about. And in the recent eight to nine years, we have sub-QICDs and EVICDs as recently in the last couple of years. And ICD therapy starts from implantation of the device, and followed by that, obviously, the main therapies would be anti-tachycardia pacing, followed by appropriate, inappropriate shock, they're all included in the therapies. And the other part I wanted to discuss was pacing, which is CRT, which also is a device therapy. And the outcomes of IICD therapies usually are survival benefits and mortality and activity level and quality of life after therapies, and psychosocial outcomes, such as anxiety about the therapies in terms of shocks. And improvement in CHF, especially with CRT, which is a therapy, again, device therapies, and adherence to the treatment plans post-therapy. So a few studies I looked at to discuss this topic were here as follows. MEDIC-II basically published that 80% of the survival rate at one year after an appropriate therapy for most of the people who received ICDs in MEDIC-II, and most of these patients, after the appropriate therapy rescues them, have non-sudden cardiac death comorbidities which have to be addressed and dealt with. And those are basically coming up from appropriate therapies, could be due to substrate issues, and they need to complete the ischemic workup and optimize the treatment, guideline-mediated treatment, and device reprogramming, and such and so on and so forth. See, the other review article which we reviewed was basically showing most of the implantable cardioverter defibrillator therapies reduce the risk of sudden cardiac death and prolonged life in selected populations, but both appropriate and inappropriate ICD shocks are common and associated with adverse effect on health outcomes and quality of life and mortality. The pain-free trial we published shows that device-measured physical activity level significantly decreased after ICD shocks. Anxiety level remained high after ICD shocks, did not show any effect on mobility and self-care. So conclusion would be ICD shocks have both negative, objective, and subjective effects on patients and must be addressed. And survival after implantable cardioverter defibrillator shock data published that, I think they looked at five landmark ICD trials which shows underlying arrhythmic substrate is most important than the ICD therapy itself to look at the mortality in the future. The first appropriate shock was associated with increased mortality, irrespective of concomitant, appropriate, or inappropriate shocks followed by that. Prognostic importance of defibrillator appropriate shocks and anti-tachycardia pacing in patients with mild heart failure. This data shows that mortality is associated more with fast ventricular arrhythmias and the fast ventricular arrhythmias increase mortality irrespective of type of therapy they received to convert patient back into sinus rhythm. The incidence of risk factors for first and recurrent ICD shock therapy in patients with implantable cardioverter defibrillators. This study shows that more than one-third of the patients received appropriate shocks and most of these patients who received appropriate shocks would have a second shock or future shocks. So the take-home message was once a patient has an AICD shock, especially an appropriate shock, at least one-third of these people would have follow-up shocks and that needs to be addressed and treated accordingly. This altitude study shows that the risk of death after first shock compared to etiologies of the shocks, the reason for the shock, looking at monomorphic PT and polymorphic PT, et cetera, have a highest risk of death after the first shock and they're coming to as simple as lead malfunctioning, lead fractures, things would not put you at a higher risk for sudden cardiac death or should not put you at a higher risk for mortality. This graph shows that patients with VTVF, no VTVF, have a better survival and patients with VTVF treated with ATP have a better survival, but patients with VTVF treated with shocks have a lower survival. So ICD shocks are associated with increased mortality. It is unclear whether the shocks are merely a marker of a more severe disease or directly contribute to mortality. Shock in the absence of spontaneous arrhythmias have a neutral effect on mortality. This can be debated because inappropriate shocks also would increase the risk of mortality. Treating shocks by ICD reprogramming reduces mortality and regardless of prognostic implications of shock, they are painful, psychologically detrimental and should be avoided as much as possible. Coming to inappropriate shocks, most of the inappropriate shocks have been studied in this clinical trial, inappropriate implantable cardiopulmonary defibrillator shocks and it shows that 13% of the people have inappropriate shocks and the reasons for inappropriate shocks are various types including atrial fibrillation and lead malfunctioning and SVTs, et cetera. So inappropriate shocks and results, what could inappropriate shock result in for patient long-term traumatic experience, possibly as much as PTSD, reduced quality of life, higher mortality rate but less than appropriate shocks. Inappropriate shocks will have definitely higher mortality rate but compared to appropriate shocks, it would be a little less. The more number of inappropriate shocks, the more increased mortality. With inappropriate shocks, hospital cost will go up and admission and diagnostic testing cost will go up and obviously atrial fibrillation will put a patient at a higher risk for mortality than with compared to lead fractures. And what are the strategies we can use to avoid inappropriate shocks or basically if it's SVT or atrial fibrillation, we can consider ablation and medical therapies and with T-wave over-sensing which is one of the reasons for inappropriate shocks, we can device reprogramming and there are a lot of other things we can use device reprogramming as a remedy. And lead noise and fractures obviously we have to replace the leads and EMI is another reason why you could get inappropriate shocks and basically removal of EMI interference and device malfunctioning would need a device change out. And CRT and remote monitoring, I put a couple of studies here. Survival after upgrade to resynchronization therapy is same as or better in terms of survival benefit when compared to de novo CRT devices. So CRT always help to improve survival and decrease the CHF parameters. And CRT implantation and remote follow-up shows that remote follow-up has better outcomes than in-office follow-ups. I have put up a couple of studies here. Compared with standard follow-up through office visits, remote follow-up monitoring, remote monitoring is associated with reduced death and cardiovascular hospitalization in patients with ICD in clinical practice. This one more study about octogenarians who are 80 years or older, we have to be careful in deciding ICD implantation in this population due to the fact that their mortality is high at one year. So we have to go case by case on these patients. In summary, implantable cardio defibrillators reduce the risk of sudden cardiac death and prolong life in selected patients. And both appropriate and inappropriate ICD shocks are common and are associated with adverse effects on health outcomes and quality of life. Underlying arrhythmic substrate rather than the ICD therapy is more important determinant of mortality in ICD recipients. And the rate of ventricular arrhythmia has a direct correlation to higher mortality. And once the patient has an appropriate shock, he or she will have more appropriate shocks in the future. Survival improves with de novo CRT or an upgrade to CRT. Remote monitoring reduces mortality and hospitalization in AICD patients. When deciding on AICD and octogenarians, you have to be extra careful and look at all the risks and benefits. And I had non ischemic cardiomyopathy, rare cardiomyopathies like hypertrophic cardiomyopathy, brugada, chagas, and other diseases also benefit from AICD therapies. But they do get inappropriate shocks and that should be taken into consideration when you recommend and implant ICD. So in last two statements, it is always important to have a detailed evaluation of the patient and overall health status and quality of life and discuss appropriate and inappropriate shocks especially with patients and procedure related complications and choice of ICD such as transvenous ICD versus sub-QICD before ICD is implanted. This will minimize the negative outcomes in the long run. Treat all ICD therapy patients proactively to avoid future ICD shocks and worsening of the pump such as worsening of the pump. Use all the modalities we have such as starting with ischemic workup and optimizing guideline mediated therapy and device reprogramming and optimizing electrolyte imbalances and consider ablations and medical management and dietary compliance. Thank you. Thank you, Dr. Varma. Yeah, we should move on because we're going to save questions for towards the end of the session and type them into your phone so you won't forget. Dr. Niraj Varma is next. He's from Cleveland Clinic and he's going to be talking to us about performance and shortcomings of heart failure diagnostics in the current generation of CIADs. Niraj, you have two minutes. No, no. Keep you all on time. Thank you, Dr. John, Dr. Lekharadi. Good afternoon, ladies and gentlemen. So switching gears now, and this is regarding heart failure diagnostics. And as we've heard earlier, we have remote monitoring that's been established for many years. Three sets of guidelines over more than a decade have issued a class 1A recommendation for the use of remote monitoring, ready to access device-based diagnostics, class 1A recommendation. So that's very strong. And the reason for that is we have continuous monitoring in this era of digital medicine, continuous monitoring that can relay data as and when needed. And that means we have the ability for early detection. So here from more than 10 years ago, we have from the randomized trial that early detection within 48 hours of significant events occurring, whether they are symptomatic or asymptomatic. So we have the ability to collect a lot of data from implantable devices. And this can relate to device conditions, lead fracture, lead function here, device function. And ultimately, as we've heard earlier, inappropriate shocks, for instance, were reduced with remote monitoring, so direct effect on patient outcomes here. However, this is all related to device-related function. Can we do the same and switch to alert-based care, so see patients when they need to be seen? Can we do the same with heart failure? Because that might influence prognosis. So does heart failure carry the same set of parameters that we can find out about via remote monitoring that will influence outcomes? So the answer should be yes, because before each decompensation, we know up to two to three weeks beforehand, there's a change in hemodynamics. Several weeks before the patient actually increases weight and develops symptoms. So can we intervene at a preclinical stage? This really is the holy grail of remote monitoring for disease intervention, particularly for heart failure, which is so important to us. So the answer to that question is yes. I mean, the early data, which directly measured hemodynamics, and here we have the laptop trial data using this left atrial sensor screwed into the septum here, providing data directly to the patient. And we see her left atrial pressure, high pressure, and with treatment low stable pressure, we see the patient profile, high pressures, stabilizing with medications, destabilizing when GDMT was withdrawn, restabilizing again. So yes, device-based data can influence heart failure prognosis. But that is a significantly invasive device. So maybe we can use a less invasive device, and this is a CardioMEMS device placed in the pulmonary artery, so not so invasive. And again, inpatients receiving CRT, so a best device-based therapy for heart failure patients, the addition of CardioMEMS-based data and action taken in response to pulmonary artery pressure changes reduced event rates, reduced event rates significantly within a short period of time, treatment versus control. So yes, direct hemodynamically measured data will influence prognosis, but they require extra hardware. Can we do the same without adding additional hardware to the heart? Can we just use the diagnostics that are available in an implantable device? And we have a large number of these. Can we use those to predict heart failure and maybe preempt heart failure? And here we see a patient who was admitted with ADHF, so heart failure decompensation symptomatic really at this end of the spectrum. But if we look at his data, and these are standard CIED data, we see that events started perhaps three months ahead of time with atrial fibrillation, ventricular rates, loss of BIV pacing, physical activity, and thoracic impedance metrics. So they started way ahead of time. Can we intervene early? So there was a lot of interest in thoracic impedance. Most of the trials that were done at that stage in the last decade were regarding thoracic impedance, particularly Optival. So could this metric predict heart failure and prevent decompensation? The answer is no. Multiple trials, and I show you one here, OptiLink HF, large randomized trial, primary endpoint, death and cardiovascular hospitalization, absolutely no change between standard of care and Optival-related remote monitoring. So disappointing. But maybe one parameter is not enough. I began with device diagnostics for device function, and lead fracture is binary. It's there or it's not there, so that's easy. But heart failure is more complicated, it's more nuanced. Maybe we should use all the parameters. So multi-parametric analysis. So this was a randomized trial using remote monitoring without thoracic impedance, and this showed a reduction in all-cause mortality in patients randomized to remote monitoring and looking at all the diagnostic data that were available. This was manually done. It's a lot of work. It's not feasible in real world, but under trial conditions, it was enabled. But there was a significant change in mortality. So this is a hard endpoint. Heart failure hospitalizations were not changed, however. So can we automate this? So there are several multi-parametric algorithms that have been used, and these are manufacturer-specific. I'll go with the one that's probably the most investigated, is HeartLogic. Five parameters out of the list that I showed you, including in first and third heart sounds. When a threshold is reached, it triggers an alert, and this may occur six weeks prior to a heart failure event. It is patient-dependent, so this is personalized care. This is personalized care, it's patient-dependent. And the algorithm learns. So daily measurements are compared to a reference calculated over the previous three-month window, and this is rolling. So personalized and adjusted dynamically. So for predicting heart failure, not bad. Sensitivity, 70%. Specificity, 86%. So this yielded a false positive rate of 1.56 per patient-year alerts and unexplained alerts of 1.47. Doesn't sound like very much, but it is, because these alerts may trigger interventions, at least at clinical contacts. So this increases workload, and this is false positive information. So this is not desirable. This is a problem. Other manufacturers have come out with their own algorithms here. Seven parameters here. Sensitivity, 66% for predicting heart failure. Relatively modest. Again, a lower, but still significant, false alert rate. Less information on triage HF. It's weighted by the optovalve. Specificity is the lowest. So perhaps these multi-parametric algorithms don't deliver in the current era. Perhaps we need more. And especially for patients who don't have ICDs or CRTs. So what am I referring to here? We have patients who have pacemakers, or ILRs, and don't have heart failure reduced DF. That's the indication for ICD and CRT, but may nevertheless be vulnerable to heart failure. But these algorithms are not enabled in those patients. So this is the HFPAF population. And HFPAF accounts for 50% of heart failure admissions. And the survival after HFPAF is as poor as for HFREF. Can we do anything for this population? So perhaps AI is a solution here. And this is one FDA-approved algorithm, SignalHF, for predicting heart failure. It's embedded not in the device, but in a third-party platform. It includes pacemakers, so suddenly we have the ability to predict heart failure in patients with HFPAF who have pacemakers. It's manufacturer-agnostic. You could use it in any device. And it collects information over a series of device-based diagnostics and assigns a weight to them. And when that composite AI-generated metric crosses a certain threshold, it triggers an alert. So then we can see it's the patient. So sensitivity is high, specificity is high. And this is being tested now in a prospective study. What about insertable cardiac monitors? So not a therapeutic device like a pacemaker or an ICD or a CRT, but simply a monitor. Well, it turns out that the information collected, and this is surprising, this is in a subcutaneous pocket, the information that this device can retrieve from that space is significant. Heart rate, HRV, AF burden. This is from the ECG, which it's designed to collect, but also activity, fluid from impedance measurements and respiratory rate. And that composite score, at least in this first study, does predate a heart failure hospitalization in patients who are vulnerable to heart failure events. So perhaps an ICM could do the same. But nevertheless, in this first iteration, 20% of monthly evaluations that were adjudicated as high risk, only 7% were followed by heart failure hospitalization or diuretic in the next 30 days. So they're not very specific. So this, at the moment, generates a high alert burden and a high percentage of false positives. So it's really not there yet. So in summary, we know that direct hemodynamic measurement is very useful, the laptop, the pulmonary artery pressure sensor. But when using surrogate parameters of hemodynamic compromise, the predictive ability of heart failure events amongst patients is modest. Indeterminate alerts or false positive alerts generate increased patient contact, possibly anxiety, more clinic admissions, and work burden for the clinical workforce. This is very undesirable. The prediction of heart failure events in patients without CRTs or ICDs, that is the HEF-PEF population, not the HEF-REF population, is really just beginning. But there's promise there. Treatment at a preclinical stage by that I mean before a patient is actually decompensated is still rudimentary. I mean that takes a frame shift in the way we treat our patients. And alert-based management, as opposed to remote monitoring for device function, for disease management and heart failure management, alert-based management ability remains very limited. So as a result, as opposed to class 1A recommendation for device management, for heart failure management, this consensus statement, I gave a 2A recommendation, which seems generous in the way that I presented the data. And in fact, Europe assigns a 2B at best for this. So I think there's a ways to go when using CID-based parameters for predicting heart failure. But there is promise, particularly with AI in the current era. Thank you. Thank you. So the next speaker is Dr. Mark Kroll from the University of Minnesota. He's gonna talk to us about the current diagnostic algorithms and their real-life performance in transvenous, subcutaneous, and extravascular ICD therapies. Dr. Kroll is a preeminent name in this space. And I know, Mark, it's hard to cover all of this in eight minutes, but you can do this. If I can figure out how to click a slide. Let's see. Is there an engineer in the house? Okay, thank you. Well, I thought I would give you the perspective from the technology side from someone that's never touched a patient. And we're gonna go to one more slide here. And we're not going to have time to talk about subcutaneous ICDs or extravascular. So we're just going to focus on conventional transvenous ICDs. Okay. I'm going to go here. So here's a brief history that suggests that maybe we're going to go around in a circle. In 1970, Mirowski suggested a hemodynamic sensor, specifically a pressure sensor in the right ventricle to detect VF. And by 1980, he realized that was impractical. And in his patent application, he went with the morphology algorithm. And here we see this. Let's see if we can figure out the mouse here. And we can see the probability distribution function of the voltage because of VF. The voltages tend to be pretty close to zero versus the QRS complex where you've got big deflections way out. That didn't work out. So then he settled on ventricular rate. And then we added interval analysis. Later came dual chamber interval analysis. And we're going to go back. Let's see. We're going to go back one here, I think. There we go. And then in 1998, St. Jude introduced their morphology algorithm. In 2002, Medtronic introduced their morphology algorithm. And you're all familiar with that. And maybe now we're looking at circling back to hemodynamic sensor again. The challenge that we faced originally was just sensitive detection of VF. Then we started dealing with the problems of inappropriate shocks from supraventricular tachycardias. And so we added rate stability, sudden onset, and then the morphology discriminators. The next challenge faced was over-sensing from lead noise, T waves, and also P waves. And I'm going to suggest now that the biggest challenge we have is HRS guidelines. Because these fail to appreciate the differences in the manufacturing algorithms. What do you expect? Naturally, the technology guy is going to blame the audience here for the problems, right? But hear me out. There might be some truth in it. Let's take something pretty simple. I just want to set the cutoff rate at 200 beats per minute. And look what that means with the Boston Scientific Device. Initially, I need 8 out of 10 fast beats to get in the game. But after that, regardless of how long you set the duration, I only need 6 out of 10 beats to be fast. So I could have AF at 330 milliseconds and fulfill a faster rate on average because I want to pick up 6 of those 10 being fast, so I have reduced specificity. If you set a Medtronic device at 200 beats per minute, you have consecutive interval counting. A single long interval resets the counter to zero. So one miss and you're out. One miss has a veto. Imagine a patient on AMU with polymorphic VT. You get this irregular amplitude, so you're going to miss some of the peaks, and you're going to get reduced activation sensing, reduced detection, and reduced sensitivity. So different algorithms are very, I'm sorry, the different, in fact, are very different algorithms. Let's look at something that's a little more sophisticated, the morphology algorithms. Because this kind of stuff I like. I always like mathematics. It beat real work. So the Abbott was the first. They sample the complex at 8 points, evenly spaced points, and they nail these voltages down. And then they compare those just like you would do a correlation to a template. The Boston morphology algorithm also samples 8 points, but they pick out landmarks, the positive peak, the negative peak, and so on, and then they compare those. Medtronic algorithm is more sophisticated. It uses a wavelet transform, very similar to what JPEG does to compress your pictures on your phone. So you have a digital image here, just binary, plus 1, minus 1, showing that I've got a biphasic complex. And then I've got a more narrow one showing the differences in voltage here, and then it gets more narrow and more narrow. And I just have to line up these wavelets. Very sophisticated algorithm. Problem is the sophisticated algorithm does not work that well. The sensitivity and specificity with the wavelet algorithm is relatively poor. Depending upon where you set the match threshold, you might get lucky to get a 70% sensitivity and specificity. With these simpler 8-point matching algorithms work reasonably well. You can get sensitivity and specificities up in the 80s and 90s. So that's one of the challenges that you face. But here's a bigger challenge, HRS consensus. So where did this come from? Took all of these studies that use manufacturer-specific programming, such as MATED RRT, PROVIDE, ADVANCE III, et cetera. They were stirred in this pot to arrive at an expert consensus. That goes in the programming, and then it's used to program all of the five brands. Let's look at the impact on clinical practice. Here is a small case series out of Denmark. They looked at 10 patients that had failure to timely deliver VF therapy. They blamed eight of them on consensus programming. Of those, seven, because the consensus detection rate was faster than what was tested in the clinical trial, and in five of those patients, VF was never detected. Here's some data out of Holland and Belgium. They considered 50 episodes of failed VT-VF detection in 24 patients. Twelve were considered life-threatening events. Twelve had VF. Five have deaths. Seven were able to be externally resuscitated. They blamed nine on polymorphic VT-VF being misclassified as SVT and said that in all nine, it was due to rate stability being used for 222 to 250 beats per minute. And here we have a simple cartoon of that situation, not an actual tracing, of course. Another nine patients had monomorphic VT misclassified as SVT, and they blamed most of these on sudden onset criterion. Bob Houser and Chuck Swerdlow dug into the MAUD database covering five-year period from 2019 to 2023, 858 reports, including 96 deaths, 173 malfunctions. The most common cause of failure to treat in a non-malfunctioning device, because we know we have plenty of those, 55% misclassified as SVT-AF. Most common cause of death, 66% under-sensing. Another 16% included programming errors. So I took their data for failure to treat, divided that by the number of implants that were in, number of existing implants in that year, and did a simple correlation. In 2019, we went from five failures to treat per 100,000 implants to 11 failures to treat 100,000 implants, so that doubled in a five-year period. So maybe we are losing ground. I mean, obviously, you don't want to shock people with AF. This has devastating psychological consequences. You know it typically at a median of five inappropriate shocks. These people have anxiety, clinical levels of anxiety and depression. But maybe the pendulum swung too far. But here's another way to look at the question. Are the engineers failing to deliver the magic algorithm to you practitioners? Could they make a more sophisticated algorithm? The answer is yes. That's what they love to do. Engineers love making complicated algorithms. But already, the algorithms are so complicated that HRS wants to simplify them with consensus guidelines. And you run into the unintended consequences of complexity. You make a system complicated enough to intertwine, it has effects that are unpredictable. Do programming practices contribute to the problem? It's hard to believe, because I know we're dealing with the smartest physicians in any specialty, but even you folks do have some problems once in a while, appreciating the differences in these algorithms and programming them. So the fair question is, have we reached the fundamental limits of information in the electrical signals? It's kind of an epistemological question. Is there just not enough knowledge there? I would say no, because how do we know that these were inappropriately diagnosed? An EP looked at the complete printout, granted with the benefit of hindsight, and then compared to that what the machine did and said, with all the brilliance of hindsight, here's the information. This should have been programmed, but the algorithm should have been this. So the information is really theoretically there, but you don't want the algorithms more complicated. So what's the right answer? What's the buzzword of the day? Artificial intelligence. But guess what? You can't put a bunch of NVIDIA chips inside a device. There's not enough real estate. There's not enough power. You can do it offline, but we don't have all day for a diagnosis of heart failure. So this isn't going to happen. It looks like an imitation to hemodynamic sensing. Thank you very much. So hemodynamic sensing it is, and the next speaker, Dr. Ji Chang is going to be talking to us about that. He's from Texas Heart Institute, and his lecture is titled, The Role of Novel Embedded Hemodynamic Sensor Technology in Tachycardia Therapy Misdiagnostics and Mitigation of Inappropriate Shocks. This is my disclosure. So I'm going to present some data in our group looking at hemodynamic sensing in implantable device. In device therapy. We know that ICD saves lives, but there is a bit, but on the previous talks indicating that we have inappropriate shocks, sometimes we miss the appropriately and miss some of the ventricular arrhythmia could be fatal, although it's rare, and more importantly, for patient with a heart failure, most people who has a device, we're not taking full advantage of that, but my talk is going to focus on inappropriate shocks, and we know the bulk of the inappropriate shock is caused by atrial fibrillation, and inappropriate shocks is not benign, and it's fairly common. As a clinician, we deal with not daily, but at least weekly or monthly. With patient, we come in with inappropriate shock that lead to emotional, psychological trauma. Some patients just tell me, Dr. Chen, you need to take it out, let me die. It's not exaggeration. And there is reports that suggest that shocks were appropriately associated with increased mortality, and the cost of healthcare nowadays, everybody is needing to be cost conscious, and suddenly, we see the significant change, increase in healthcare resource utility, and hence, the cost. After decades of effort, we try to improve the technology to reduce in-patient shocks, but it still exists, why? I think one answer lies in the fact that our information is limited. Just think about it, we're sitting across the table with the patient, whatever rhythm has, if patient's still talking, I'm not gonna shock the patient. We do not have enough information. And then looking back the history that we've spent so much effort and looking at electrograms, our electrograms, analyzing to improve the technology, now the time they really need to think outside the box. It's the box that EGM only. And then we look at all the common feature of in-operative shocks, they share one common feature. Hemodynamically, they are stable. So theoretically, if we have way of telling hemodynamics, we can totally eliminate in-operative shocks. So we come across a technology that use optic sensors, which is very tiny, it's close, the diameter is only 50 to 100 microns, it's thinner than air. It can be easily embedded in any implantable lead without changing any physical property of the lead at all. One of the important advantage is the signal pathway, the optical pathway is totally embedded in the lead. So it does not have any interaction with tissue, scars, anything, it does not affect that. And does not require optical window that most optic signal system require to interact with tissue, it does not. It's totally embedded within the lead. This is the signal that the sensor produced from the right ventricle and from CS, the left ventricle. It's a string signal, basically displacement from diastole to systole. The peak is systole and the bottom here is the, and the systole and here is end diastole. So we hypothesized and also validated that that difference is delta string then can be, is significantly associated with stroke volume. Here's the data. In 100 to 100 cardiac cycles, we look at the string pattern changes during cardiac cycle and we use the pressure loop, pressure volume loop that's pretty much a gold standard for hemodynamic measurement. Stroke volume and start volume. Use that signal to decipher the signal, string signals recorded from those optic sensors that are embedded in the RV and the CS lead. And the high, very high correlation, the p-value is, that's actually the computer printout, it's all zero. I had to put a one there to make it more realistic. So we, our data demonstrate that the delta string during cardiac cycle correspond to, or reflect the change in stroke volume. So as the first application, we set to test a hypothesis where that hemodynamic data achieved from a miniaturized sensor can be utilized to reduce input shock. We induced or simulated atrial fibrillation and ventricular tachyarrhythmias in pigs. We defined the effective heartbeat that correspond to the cardiac cycle length, which produced at least 45% of the delta string during sinus rhythm. That is, for a given heart cycle, if the delta string, i.e. the cardiac output, is greater or equal to 45% of baseline sinus rhythm, we call effective heartbeat. Meaning that that beat is expected to produce adequate post-pressure. And also we define a new parameter, an index we call a hemodynamic stability index, which is simply a percentage of effective heartbeats over the total heartbeat during the detection period. We set up seven second. Okay, but I think we, here is some example here. During sinus rhythm, we have a delta string and also arterial post-pressure. That's the far field, the EGM, and the I-electrogram, RV-electrogram. And during atrial fibrillation, not all the beats produce effective cardiac output. So that some of them may drop off, that it does not generate adequate pulse, and substantial of them did. So in this case, the hemodynamic stability index is about 70%. Versus VT, you do have the occasional beats that generate enough pulse pressure, but most of them fall off. Okay, and of course, during V-fib, you do not have much pulse pressure, and the delta string is almost flat lined. And just simply looking at the hemodynamic stability index, in the simulated tachyarrhythmias, they're consistent of 460 H-fib, 135 VT, and 110 VF. It produced 100% accuracy, weighs the additional far field EGM with additional criteria. A combination of EGM width and this hemodynamic stability index produce 100% accuracy, 100% sensitivity, and 100% specificity, okay? And panel C is simply another graph showing that a cutoff value of hemodynamic stability index of 0.6 would separate those hemodynamic stable beats, tachycardia, versus hemodynamic unstable tachycardia, whether it's A-fib or VT, regards underlying rhythm issue at all. And of course, A-fib rarely can be hemodynamic unstable. Those are the result we get from based on the conventional feature expressions. It's a very tedious work. We are using artificial intelligence in this case, like, okay, we are now gonna be limited by the small number of data animal tested, and also some artificially derived thresholds. We simply put the data in AR model, end-to-end, and see what's the result. Again, that's the structure model in the 70 layers. And here's the result. The same set of data, okay? 135 episode VT, 110 episode VF, and 469 episode A-fib with RVR, okay? We put through, those are the recorded electrograms. We put through the RCD, the dual-chamber RCD, and it results 100% accuracy for ventricular tachyarrhythmia, as we expected. But it failed 41% in a purpose-diagnosed A-fib episode that misdiagnosed the VT or VF. And as I showed before, if we use conventional feature extraction algorithm, zero percent error. We put through the AI model, it reduced the error rate from 41% to 8.3%. And we have to keep in mind, this model is by no means perfect. That is our initial attempt. And that model can be further improved. Importantly, those are the calculation of the AI model, took 38 millisecond to complete. So our model can be practical. It does not, and also we tested with CPU on a cell phone, which is very low power, and require 2.7, 2.4 second to complete the calculation to derive the result. So make it a very practical alternative. And this is a slide showing that we did it with electrical noise, about 260 magnetic noise, and 287 lead factor simulated. We put through the same process. Again, the current technology, this is what in the market, I mark the manufacturer names, so it won't be biased. They won't be mad at me, but lead to about 50% error rate. Whereas with the consideration of hemodynamic data error, the accuracy is 100% with zero error. In conclusion, I think monitoring hemodynamics with a miniaturized sensor that embed in the lead, it does not affect any physical property of the lead at all, and can lead to a disruptive paradigm shift. And so from a tech heart detection, current technology is tech heart detection, and you go rhythm discrimination, which has problems other way understand. We go direct to hemodynamics. If a patient's hemodynamics stable, we're not gonna shock the patient. And that can, we're based on a hemodynamic, and we can, in combination with rhythm discrimination, really lead to 100% accuracy in therapy decisions, and totally eliminate the inoperable shocks. And thank you, and as you can tell, this is a huge project, and a lot of, we have a team that work on it, and we appreciate it, and want to say thanks to them. Thank you. So with that, with that we go into the last talk, which basically explores the future applications of such hemodynamic data in heart failure diagnostics and management in patients who have cardiac devices. By Dr. Nilesh Mathuria from Houston Methodist. Thank you, Dr. Lacaretti, thank you, Dr. Jung for the introduction, and let's see here. And I know we're, I think, running a little behind, so I'll be as efficient as possible. And let's see here, okay, great. So I've been tasked to discuss future applications of hemodynamic data in heart failure diagnosis and management in patients with CIDs. Before we talk about the future, the present, I think Dr. Verma, Dr. Kroll, others have really gone over this extensively as far as our current tools we have with regard to hemodynamic assessment, but at the same time, there are limitations. And even microelectromechanical sensors such as the CardioMEMS and the Endotronics device, which is also approved here in the U.S., although they provide incredible amount of data, can have been shown to help reduce heart failure hospitalizations, there are two fundamental issues. Number one, although the risks are low, there's still another procedure, another device implant for the patient. And then number two, more often than not, at least in my clinical practice, the patients that receive these devices have already had difficulties with heart failure management, et cetera, and so this is now an adjunctive therapy that's being considered in addition to what is already being used. So clearly, we're already behind the eight ball, if you will, at least in that patient population. So what is the future, or what could be the future? Let's put it that way, and I'll discuss three general topics, laser Doppler sensors, admittance, cardiac output, and then of course, and then work that I've been fortunate enough to be a part of with the Optical Strain Sensor Project. So laser Doppler, so Doppler, like any other Doppler, we're looking at movement or shift, if you will. So if you apply light to moving red blood cells, there will be a degree of shift, or a Doppler shift in the light that's reflected back, and that degree of shift can correlate to pulsatile flow. And so this is a study that was published in Jack EP about five, six years ago where they looked at 50 patients, so this is not even preclinical, it's a clinical study with patients with ICD where VF was actually induced, and then using this Doppler algorithm, can they generate or can they assess for hemodynamic instability? So the setup here was actually in patients with an existing ICD, two various optical sensors on the skin, and then that would be converted into a digital signal and correlated, or synced, if you will, sorry, gated with regard to electrograms. At least it was shown, at least in proof of concept data that perhaps this could be something that could be used with regard to understanding at least acute hemodynamic status, given that some of these sensors could be perhaps incorporated into ICD leads. As you can imagine, just with the term Doppler, there are gonna be issues with regard to limitations with motion artifact, and over time, capsule or just fibrosis would that mitigate or alter some of these sensors' capabilities, if you will. And then moving forward, the concept of admittance-derived cardiac output has been published, now it's been about roughly 10 years. This is also work actually done in Texas, and there was a startup technology called, or a company called Admittance Technologies, which actually, or this was a preclinical study in a canine model, were actually using existing biventricular ICDs, i.e. you really need an RV coil with a ring and tip electrode. If one delivers a current from that RV ring over to the LV ring, and then the return voltage can actually have a phase shift, and this company had some proprietary technology where they could, in theory, separate the blood and myocardial component of this. And interestingly, the blood conductance would actually correlate with stroke volume based on some of their preclinical data. At least, again, further advancing this concept, can we get real-time hemodynamic assessment in patients with implantable devices? Fast forward a few years, the same group, and this time with the late, great David Haynes, actually assessed this same concept in patients. So 22 patients with ICDs underwent a stroke volume assessment with comparison to 3D echocardiography at the time of generator change. And as you see here, at least with this approach, there was a very reasonable correlation with regard to, at least, derived stroke volume, both through the technology, admittance technology, as well as, at least in comparison with 3D echo. Now, numerous limitations. Number one, namely, all of these patients had to have biventricular ICDs, so namely an RV coil, LV lead, as Dr. Verma already mentioned. That does not fit all our patients. There are patients with, quote-unquote, pacemakers or other forms of devices. And then also, 3D echo itself had some limitations. There was significant variability with regard to volume assessment, with regard to this. But again, it created further proof of concept that there can be this sort of more, further information gleaned from these patients with regard to their devices. And then now, transitioning to the work that I've been fortunate to be a part of down in Houston with the previous speaker, Dr. Chang, and colleagues at Texas Heart Institute, we've been looking at optical strain sensors, as was described previously. Specifically, optical fibers with so-called fiber-bragged gratings within the fiber core. And when light is delivered through these FPG aspects of the optical fiber, some light is reflected back. And when there is strain within these gratings, there can be a change, or there are changes in the distances between those gratings, which can cause a shift, or wavelength shift in that reflected light. And then that wavelength shift can be converted into a digital signal, which then can be analyzed, if you will. As was mentioned in the prior talk, these sensors can be quite small, and ranging in the micrometer range, so it could be applicable to current CID leads, if you will. And as we all know from our imaging colleagues, strain in general, myocardial strain, has shown to have significant sort of prognostic value in A, diagnostic purposes, and various infiltrative cardiomyopathies, restrictive cardiomyopathies, and also with regard to predicting or anticipating hemodynamic alterations. So somewhat of a busy slide here, but at least based on that hypothesis, at least in a porcine model, we implanted biventricular ICDs, where we put these optical sensors at the tip of both the RV and the LV lead, and then had a PV loop inserted for comparison as sort of a control or a ground truth. And as was mentioned by Dr. Cheng in the previous talk, when assessing and looking at comparison with pulmonary vein, sorry, with the PV loop data, the so-called delta strain or difference between the systolic and end-diastolic strain correlated with stroke volume, and here's sort of a blown-up picture of this, where you have your arterial waveform and LV pressure. The blue there is your pressure volume loop, and then the green here is our strain data. And here, that difference, if you will, is the so-called delta strain between the diastole and systole, if you will, correlating with the stroke volume. And then, of course, our electrogram that we're familiar with from the EP standpoint at the bottom. But beyond even the so-called delta strain and how it correlated with various hemodynamic alterations, both with, at least in this, as I'm showing here on the top right, qualitatively with dobutamine, as well as saline infusion, it's just simply end-diastolic strain also correlated with end-diastolic volume, which is really an important marker that we know that we use clinically, that we lean on our echo colleagues day in, day out, with our patients in and out of the hospital to assess do they have some component of heart failure, preserved ejection, or even patients with a reduced ejection fraction, you know, where is their hemodynamic status, perhaps, and does it correlate with any potential clinical symptoms, if you will. And even that parameter correlated not with the various sort of hemodynamic alterations. So here, again, somewhat of a busy slide, I've highlighted with the red circle there, end-diastolic volume and end-diastolic strain were correlated both in this preclinical model when we have dobutamine infusion, where they both decreased, and then conversely with saline bolus, the end-diastolic volume increased, and that correlated with also increased strain on the end-diastolic strain. So again, perhaps providing a marker, aside from simply the concept of preventing or reducing inappropriate shocks, could this be another form or another way to assess hemodynamic status in patients and perhaps predict issues, or predict even, perhaps even provide a diagnostic tool, because we have many patients on the EP side with CIDs implanted for various reasons, sometimes even VT for unclear radiology, and could this provide some clue, perhaps, to either restrictive cardiomyopathies, also monitoring patients if they're getting cardiotoxins, i.e. chemotoxicity, et cetera. So again, it could provide another avenue for research, and again, another diagnostic tool. This is just more further data, even in rapid pacing in the same model, in a subset of this Porcine model, again, with these hemodynamic alterations. Again, somewhat of a busy slide, but in this case, in the rapid pacing, essentially every other beat is perfused, and if one can see here on the strain data, which is the third, sort of third from the top, if you will, or the middle, one can see that every other beat, there is this evidence of sort of lack of perfusion, and in turn, lack of strain, if you will. And also, just another showing that both in the LV and the RV, that these assessments, end-diastolic volume, and in particular, in this case, stroke volume correlated with regard to when they're normalized for body surface area, et cetera, and correlated with our PV loop data. And again, even in this condition, both with dobutamine infusion, as well as saline bolus, the changes in strain correlated with end-diastolic measurements, as well, with our PV loop data. So again, this could provide another avenue of assessment of patients with regard to this. So one of the other challenges, if you will, with regard to incorporating strain data, or this sort of optical sensor data, especially in the medical realm, specifically the CID, CID standpoint has been, as I mentioned, all of this data ultimately comes from reflected light, and converting this into some sort of digital signal. And historically, this has been fairly large and bulky, and which is why we have not seen incorporation of this sort of approach in our space, if you will. But working alongside, and really, more importantly, learning, I'm the lowly clinical EP, but learning alongside the incredible engineers at the group, the group has been able to actually create, essentially, a miniaturized version, if you will, of this sort of converter, if you will, to help convert this optical wavelength shift to electrical signals, and this has been validated, even in different angles with different leads, and so it can really provide further promise that perhaps we can use this sort of technology, optical sensor technology, in current CID patients, if you will, without disrupting device size, et cetera, if you will. So with that, obviously, this is just initial proof of concept data, certainly this needs to be addressed even pre-clinically in other models, cardiomyopathy, post-MI type models, RV versus LV sensors, location of sensors, also fatigue testing of these optical sensors, although there's preliminary data to suggest that this should be sufficient for a patient life cycle, but it provides, obviously, an incredible avenue of research that hopefully can be, perhaps, paradigm shifting, and as already mentioned by Dr. Kroll, all of this data ultimately would have to be driven by some sort of machine learning as far as how it would perhaps impact patient care. So with that, I'll conclude, so hemodynamic data in patients with CIDs offer unique opportunities for early diagnosis and or intervention. As I mentioned, minimization of optical sensors and fiber optic cables may provide new avenues for real-time hemodynamic assessment, and future device optimization may be refined with such continuous hemodynamic monitoring. So hopefully, I got us caught up, if you will, with this last talk, and thank you again. Thank you. Yeah, unfortunately, we don't have much time left to ask questions, but if there's a burning question, please come up to the microphone, and it doesn't look like this much. So all that's left for me to do is thank the speakers for a wonderful session, and all of you for attending it. Thank you.

cardiac electrophysiologist

hemodynamic innovations

implantable cardioverter defibrillator

ICD therapies

remote monitoring

heart failure diagnostics

ICD detection algorithms

miniaturized hemodynamic sensors

optical strain sensors

cardiac device functionality