Well, thank you for joining us this early in the morning at 8. We're here to discuss the best of heart rhythm in basic science. I want to thank everyone who's been contributing manuscripts to the journal. I'm Julia Indyk from the University of Arizona. Mario Delmar from New York University. So we have four presentations this morning. I want to first bring up Dr. Jing Yao, Beyond Conduction Impairment, Unveiling the Profound Myocardial Injury in Left Bundle Branch Block. Hi, thanks for the introduction. Okay. Click on your name. Okay. And wait a few seconds. Yeah. Are you Jing? Yeah. So you can start. Okay. Click start. Just wait a few seconds. Okay. Thank you. Okay. Okay. And thank you, everyone. And I'm truly honored to be here to present our research on the left bundle branch block. As we know, LBBB is a common conduction disorder, typically asymptomatic, and often detected occasionally during routine ECG. The instance of LBBB increase with age and notably new onset, sorry, allergenic LBBB could fall in timing or more long procedure for HCM. Despite its prevalence, critical question remain poorly understood. Is LBBB a truly benign finding? If not, what pathology change does it trigger? And currently, there's limited data on the structure and the function consequence of chronic LBBB. So these knowledge gaps motivated us to explore this area further. So our study aimed to investigate the chronological progression of LBBB dysfunction in LBBB and physiological mechanism underlying LBBB-induced cardiomyopathy. To investigate this question, we designed an animal study in which LBBB was induced by operation of the main left bundle branch trunk in bear dogs. We monitored these animals over 12 months period and performing TE and ECG at baseline 30 minutes, 1 month, 3 months, 6 months, and 12 months post-operation. And at the 12 months following the SPECT-CT, we harvested the horse for detailed pathology analysis and including the staining of the purkinje fiber and histological analysis and the quantification of the connecting 43 protein expression. This slide shows how we dissect the left ventricle into segment for detailed pathology analysis. Okay, now let me highlight the main finding of our study. Here we can see the ECG. And post-operation, we can see the QI stimulation increased significantly, confirming the successful LBBB induction. Here we can see the echocardiography for chamber view of a LBBB dog at the baseline and 12 months post-operation. Here we can see the right image here. We can observe the disinclining of myocardial contraction pattern by visualization. And over 12 months period, we can observe the progressive LV dilation and the reduce of the left ventricular ejection fraction. And here we can see the global and segmental value of the strain and myocardial work. Here we can see the global longitudinal strain and global myocardial work efficiently reduced gradually. And global myocardial vestibular work increased gradually post-operation. And for the segmental value, here we use the color-coded map to represent the absolute value in each segment. Here, the dark color represent the higher value. So by this way, we can see how the segmental value change over time. For example, here we can see the segmental myocardial vestibular work. All the segments, all the segments, we can see the segmental value increased over time, especially at the basal and lateral, middle lateral and basal septum. Okay. And here we can see the progesterone loop. In echo, we can acquire the segmental or global loop with the echocardiography myocardial work so well. And this slide shows the loops of the representative of the LBB dark. Here we can see at the basal septum, the area of loop decreased with time. And from the six months, we can observe the twisting pattern here. However, at the basal lateral, the area of loop increased with time, presented with the opposite trend, right? Okay. And such alternation can be observed in clinic in LBB patients. Okay. This slide shows the segmental myocardial perfusion in SPECT-CT. And in our research, all the LBB dark presents the similar typical LBB pattern of segmental myocardial perfusion, which is lowest at apex and basal septum. And here, this slide shows the growth pathology of the LBB darks, and which is highlight of our research. As we know, the pathology change associated with LBB-induced cardiomyopathy have not well been defined in previous clinic or experimental research. And in our research, the endocardiogram of LBB dark present with the flaky or focal greasy thickening, mainly distributed in the left ventricular apex. And similar alternation here, we can see similar alternation can be observed in all of the LBB darks in our research. However, in the healthy darks, the endocardiogram appeared fresh pink with greasy appearance. In this slide, we can see the representative machioned staining in different segment of healthy dark, the first slide. And the second slide is the LBB dark. And the blue area here represent the collagen fiber. And we can see in the healthy darks, there's only a small amount of the collagen fiber can be detected in the healthy dark. However, in the LBB dark, we can see the blue area is much thicker, right? And this represent the significant increase of collagen fiber in LBB darks. And in this slide, we can see by HE and machioned staining, the pathology of the LBB, including the enlarged intercellular space, myocardial fat infiltration, and disorder arrangement of myocardial fiber, cellular swelling, fatty degeneration, and infiltration of epicardial fat into myocardium. And this slide shows the pathology of the purkinje fiber here. And the first slide is the healthy dark, and second slide is from the LBB dark. Here we can see in the healthy dark, only a small – oh, sorry. In the healthy dark, the purkinje fiber mainly located in the sub-endocardium. And occasionally, we can find it in myocardial or perivascular area. However, in the LBB darks, the purkinje fiber – the amount of purkinje fiber is reduced significantly in the sub-endocardium and increased in the myocardial and perivascular area. And this slide shows the purkinje fiber pathology change, including fibrosis here and fatty degeneration and regularization here. And in our study, we found that expression of connexin-43 decreased significantly in the LBB dark compared with the healthy darks. And as we know, connexin-43 played an important role in the conduction of the cardiac electricity and the coordination of myocardial contraction. And in the end, we conducted a correlation analysis and found the segmental myocardial work correlate positively with segmental myocardial perfusion. And endocardial collagen content correlate negatively with segmental perfusion. So, in conclusion, our study provide pioneering insight into the myocardial injury associated with LBBB and identify the chronological dynamic of LV dysfunction and uncovering a vicious cycle involving impaired myocardial perfusion and the progressive LV dysfunction by echo and pathology changes. So, this study underscored the importance of recognizing LBBB not as a benign condition, but as a potential precursor to the significant cardiac pathology. And this photo was taken after finishing the final animal. I am deeply grateful for the tremendous effort of our research team, including cardiologists and anesthesiologists and forensic pathologists and labor researchers. And a special thanks to Professor Taub for her guidance and cooperation. Thank you for your attention. I'm welcoming your questions and discussion. Thank you again. Go ahead. Samir Shaba, great presentation. Thank you. Do you have any data about any reversibility with left bundle pacing of all the changes that you saw and if there was any time beyond which you lack reversibility or you cannot reverse it anymore? So, beyond what you presented, do you have any data about pacing the left bundle and see what happens to reverse these changes? Okay, thank you. And this is what we plan to do in next project. Yeah, we are beginning to do next project just to figure out maybe CRT or left bundle pacing to such situation. Thank you. Actually, along that line, I wanted to ask, you know, when we treat patients with heart failure, in addition to biventricular pacing, but in thinking about avoiding the irreversible components that you are showing here also makes me think of the input of heart failure medications that we give. Is there any thought of maybe testing to see if concomitantly treating the animals with medications along the way? Yes. And we have next two projects. One for CRT or left bundle LBB pacing and the next project is just to try to figure out about the medication. Yes. Thank you. Yeah. Question on your histological sections. Do you see any signs of inflammation? Infiltration of other cell types in addition to the addition? Yes. We found some biomarker which showed to have some inflammation. Yeah. We got this result. But we have not do the further research on the inflammation. But yes, yes, there's some biomarker. Yeah. And finally, did you, you looked for connexin 43, but in the bundle of his and the Purkinje fibers, there is mostly connexin 40. Did you look for a connexin 40 staining in the Purkinje? I think, yes, we do the connexin 40, but connexin 43 is more dominant. Not in the Purkinje fibers. In the conduction system is mostly connexin 40. In the ventricular mass is connexin 43. So that's why I wonder. Yeah. We checked several connexin, but we found connexin 43 in our research. That's the most famous one. Yeah. Maybe we can check. Yeah. Thank you. Okay. Thank you. Okay. That was a great start. Thank you for the talk. The next presentation is by Wei-Sing Chung from China Medical University Hospital in Taiwan. And the title is ischemia-induced ventricular proarrhythmia and cardiovascular autonomic affliction after cardio neuroablation. Good morning, everyone. I'm Weixin Zhang from China Medical University in Taiwan. So my project here today is to talk about ischemia-induced ventricular arrhythmia and cardiovascular autonomic dysfunction after cardioneural ablation. As we know that, for now, almost every meeting is talking about PFA. So when you visit any medical center, the first question asked is, have you done PFA? And the second question is that, have you done cardioneural ablation, which is also known as GP ablation. But what the consequences is unknown. So at some level, a case report from a case report told us that there are some ventricular arrhythmia events after our conventional AFA ablation. Because you might incidentally damage some ganglion plexus after our PVI. So the question is that we want to know how to evaluate the long-term consequences of cardioneural ablation. But there are some questions about, what's the durability of the CNA? And what's the electrical stability of the CNA? And what's the impact of autonomic system? That is unknown. But when we want to answer this question, and when you look back all the historical literature, you will find there is no animal model telling us the consequences or the effect of cardioneural ablation. So the project started by beginning creating a model of cardioneural ablation. But there are a lot of questions, like how to find GP, how to confirm the endpoint, and which GP should be targeted. So our project is here. So we developed a spatial program to find the GPs. We used the synchronized high-frequency stimulation to detect the ganglion plexus without inducing atrial fibrillation. And you can see from here, when the blue dots means the negative spot, which means that when you stimulate area, there's no response, there's no change of the heart rate. But when you find a green area, you will see that the heart rate decrease gradually during each pulsation of the stimulation. And we also utilized the strength of animal study. We can take the heart out to see the green dots, it's here. And we do the histology examination, you can see the ganglia here, confirming that where we find the response is where the ganglia is. So confirming the accuracy of this maneuver. And then we also used the intra-jugular vein stimulation, like what we do in humans, to test the response before and after our stimulation or ablation. You can see that in the baseline, when you stimulate vagus nerve in the jugular vein, you can see the heart rate decrease, or even a systolic. For this one, it's AB block, you can see the atrial signals, but there's no blood pressure, which means AB block. After the control study, you can see the AB block remains. But when you finish the cardio-neuro ablation, you can see that the baseline, there is AB block in the baseline. But after you do the ablation, there's no AB block, no matter how strong you stimulate the vagus nerve in the intra-jugular vein. So another question is, everyone has this question about how many ganglia you have to destroy to reach that end point, it's actually unknown. But our model proves the idea that you don't have to kill all the ganglia, and in fact, you cannot. So in our calculation, usually we have to destroy 50 percent of the ganglia, then you can abolish the response of the vagus response. So what ganglia do we target? There are a lot of names and targets in the literature, it's like everywhere. So the major ganglia we target and also where we find response is mostly the right RGP, it's over here. And the second target we find that's more responsive is the LSGP, it's above the left superior pulmonary vein. So another question is that some studies point out when you ablate the GP, you will see the bradycardia response, like what we do during PVI. But actually, when we try to do that in animal model, there's no bradycardia during ablation. The only response we find is that when you really ablate the GPs, you only see the heart increase gradually by every ablation. But the response is not that profound when you ablate the left GP, but in the RGP. So almost in every animal, when you abolish the vagus effect, the heart rate increase significantly after RGP ablation. And the question is how long will the effect remain? So for the interest of time, I cannot spend 10 years waiting for the pig to grow up. So we wait for six weeks to see what's the response. So we can see that in the control group, the blue one, after six weeks, the effect gradually increase. But when you test after six weeks, you can see that in the ablation group, the response come back a little bit, but still significantly attenuated compared with the baseline. So it's the same, even you stimulate from the left side or you stimulate from the right side. Another effect of the vagus effect is also prolong the PR capduction, which means the AV node. And we can see that in the control animal, the PR prolong significantly even after six weeks. But in the ablation animals, the response remains attenuated. The most importantly, after six weeks, there's no AV block when you test, when you stimulate the vagus nerve. Another interesting finding is that we also find that when you do the GP ablation after six weeks, the sinus node activation also change. The support, the concept that the vagus, the sinus node is tightly controlled by the vagus input to the heart. And another important thing is that there's no, in human study, we cut the neck out and stimulate the vagus nerve. No one wants to do that, but we can do that in the animals. You can see that when we directly stimulate the vagus nerve in the jugular vein, in the control group, it's normal to find that the heart rate decrease and also the blood pressure decrease. But when the animal receives the cardiovascular ablation, the heart rate decrease a little bit, but that increase, and even the blood pressure increase, which means that we abolish the protective effect of the vagus nerve. So this is also true when you see all the animal does, you can see here. We increase the threshold to decrease heart rate significantly, and there's almost no heart rate change after six weeks after GP ablation, and there's no blood pressure change, and even the blood pressure will increase a little bit. It's a very important finding. And the final question of our study is to know what is the effect on sympathetic activation, because the parasympathetic is the balance between the sympathetic and parasympathetic. So what will happen when we abolish the vagus effect? You can see that immediately after ablation, the QT interval significantly increase in the ablation animal. But in the control animals, even the heart rate increase, the QT interval didn't change a lot. So this is also true when you analyze all the animals, you can see that the QT interval significantly increase after GP ablation, and this finding remains after six weeks. And when we do the sympathetic stimulation, you can see that, let me zoom in for you. So this is the baseline. When you directly stimulate from the stelic ganglia, you can see that the baseline heart rate and ECG is like this. But after you do the sympathetic stimulation, you can see the QT interval significantly increase. So this proves the concept that when you abolish the vagus effect, the sympathetic effect is under, is uncontrolled, which will make your QT interval significantly prolonged. So what is the effect on lateral stability? You can see that. In the control group, when we do the LOD ligation directly, there are minimal or some short-run non-system VTEC in the control animals, which is predictable. But when you, after GP ablation, after six weeks, you can see there are a lot of ventricular events after ligation of the LOD. We try to increase the ischemic. So this other result, we can see that the incidental VF is increased, but not significantly, because there are only, I remember, six versus six animals in our both group. And the time to the first non-system VTEC and VF is also very different. In the control group, the animal can tolerate the ischemic longer than the ablation animal, which means when the animal suffer or underwent GP ablation, they are less tolerant to the ischemic-induced ventricular arrhythmia. It's also true when you examine the non-system VTEC epso, or the polymorphic non-system VTEC epso, or the PVC, the conclusion is the same. When the animals receive GP ablation, they are less tolerant to the ischemic-induced ventricular arrhythmia. So here's my conclusion. Synchronized high-frequency stimulation, we can identify the GP and ablate the GP, and the effect of the vagus attenuation can persist more than six weeks. And the GP ablation will result in autonomic dysreflexia, and mitigating the protective effect of the parasympathetic system. And the last thing is that when we ablate the parasympathetic system, the symptom activation will increase the risk of ventricular arrhythmia following ischemic. So thank you for your attention. And this is the first pick who underwent the GP ablation. And this, thanks to all my colleagues, and my director, Dr. Olajuwimmi, and Jason Bradfield, and Maury is also here somewhere, I believe, in the anatomy section. So thank you, everyone. Thank you. Very nice study. You mentioned that you roughly have to ablate 50% of the inputs of the GPs, and beyond that you wouldn't see any more AV block, correct? Yes. So, two questions, twofold. One is how do you estimate that you're about 50%? And then second, is that only the percentage, or does it matter which GP in terms of the location of the GP with respect to AV block, per se? Okay. Thank you. Thank you for your question. So this is one of the questions that we want to ask or know in this project. So in the neuroanatomy, our GPs have more control on sinus node. In this area, you can see the B area, we call that IVC-IA-GP in some studies. It's more close to the AV node, but actually the three groups of GPs, they are connecting to each other. But RA-GP has the most powerful control of all the, you can say that it's the input of all the sinus node and AV node. So basically we're ablating these two areas, because we couldn't find any finding over this area, like what other literature say. So after ablation, I take these two areas out and do the histology estimation. I think I almost did more than 500 slides to check how many GPs they ablate. So it's like a manual calculation. So basically this area, RA-GP has the most abundant GPs over here. The AOS-GP doesn't have that much, and when I examine the RA-GP area, you can just calculate. Sometimes we're ablating the endocardial site of the ganglia, but we cannot target the ganglia. Probably in the mid-myocardium or in the fat pad, then we cannot ablate from endocardial site. But when you see the distracted, when you see this distracted ganglia, you can see, we can calculate. This, like this one, this is the destruction of the ganglia. This is the preserved ganglia. So basically we just manually calculate all the ganglia and see that we estimate 50 percent of ganglia is ablated. Then we can reach the target of abolishing all the vagal effect, which means that a lot of people are doing GP ablation now. They are not really ablating all the GPs, but they still can get it. That's why there are a lot of heterogeneous results. In all the manuals, everyone says they are effective, but why? Because you don't have to ablate everything, and you can get the attenuation of the vagal input to the heart. In the interest of time, we're going to thank you very much for your presentation. Next, Dr. Sava will come up and present Ventricular Arrhythmia Inducibility in Porcine Infarct Model after Stereotactic Body Radiation Therapy. Thank you, Dr. Indyk and Dr. Delmar, and good morning, everyone. Oh, it's good. Thank you. Go to your name and press start. Oh, thank you. All right. These are my disclosures, none relevant to this topic. I'm going to be talking about SBRT radiation to the area of myocardial infarction in terms of its ability and impact on reducing the inducibility of ventricular arrhythmias in a porcine model of myocardial infarction. I'm preaching to the choir here when I say that sudden cardiac death is the number one killer. It affects a lot of people. One of the major mechanisms is ventricular arrhythmia, be it VT or ventricular fibrillation, particularly reentrant rhythms in the context of a fibrotic myocardium for non-ischemic and ischemic cardiomyopathy. It's also fair to say in an EP meeting that the therapies that we have are not that great. So if you think about antiarrhythmic medications that we use, they're not that effective, and many of them have, and the more effective ones have more not only side effects but end organ toxicities, amyotron, lung, liver, thyroid problems. And then the ablations, the techniques have gotten better, but still it is limited by the fact that it's invasive. We have to go point by point and really find where should we ablate the critical isthmus of a ventricular arrhythmia. Many times those patients are not very stable, and these procedures take a long period of time, and sometimes it feels like we win the battle and lose the war because the patients, you know, you suppress the ventricular arrhythmia, but the patients don't do too well. So this is a recurring theme. So a few years ago, we're all aware that the Wash U group led by Phil Kukulich reported in the New England Journal a small case series of individuals with refractory ventricular arrhythmias where they borrowed the concept from radiation oncology where, you know, you radiate a certain area to treat cancer for, you know, traditionally, but it was the area of a scar in the case of these refractory ventricular arrhythmias, and that suppressed the ventricular arrhythmias over a significant period of time. Since then, there have been case series, none that have been randomized. As you probably all know, there is a randomized effort control trial happening, and that will give us a little bit more answers. But despite the value in terms of suppressing the arrhythmias, we do not know too much about the mechanism of how does it do it. There is a plausible mechanism, correct? When we talk about reentrant rhythms, there is an area of slow conduction that, you know, allows the circuit to perpetuate itself. Otherwise, the head of the, you know, the circuit of the electrical propagation meets the tail and it stops. So the thought is that, you know, the radiation will close those surviving diseased myocardial cells in the scar, in the border zone of the scar, and that would stop the arrhythmia. And by the way, this is what we try to do with the ablation point by point, try to find those areas, homogenize the scar, and suppress the abnormal rhythm. So this is what we looked for to see if that is what happens and to see the effect of that in a more controlled animal model. So we took 25 to 35 kilogram male and female pigs and we performed myocardial infarction on them, and then we randomly assigned them to getting the radiation to the scar area or not, and then followed them over time, and did invasive surgery to assess electro-anatomically what is the myocardium looking like, and then did an inducibility test to see if we can induce ventricular arrhythmias. So we allowed a good amount of time between the MI and the radiation in the pigs, in all pigs basically, in order to at least six to eight weeks to allow for maturation of the scar, and then the pigs that had the radiation had that radiation at six to eight weeks, and then four to five weeks later we did the terminal surgery, and we did a PET CT to be able to, you know, design and know how to deliver the radiation. So I'm going to go quickly over the models. This is an established model of chronic MI. We inflate an angioplasty balloon in the LAD, fairly distal in order not to kill the, you know, create a big MI, and we deliver polystyrene microspheres in that vessel, and then we see ST elevations, and on gross pathology you see the myocardial infarction. The planning with radiation oncology was tedious to be able to know how to deliver the radiation, so this was planned with a PET CT, as I mentioned. Identifying the area of the scar was a combination of hypoattenuation and thinning on the CT, as well as hypoperfusion on the 18-fluorine PET scan, and then it was a single delivery of 25 grays to the area. This was calculated by the radiation oncology folks with the intent to, you know, take into account the fact that there is no, at the delivery, there is no gating for the heart beating, but there was gating for respiration. And then the terminal surgery, we basically did the high-density anatomic mapping using the N-site system, and we used the cutoffs that we usually use in VT ablations. Less than 0.5 millivolts is, you know, dense scar, and then diseased myocardium is 0.5 to 1.5, and then normal myocardium above 1.5. We calculated the ratio of dense scar to diseased myocardium as an estimate of how much scarring do we have or how much, you know, we've obliterated the surviving myocardium in the area of the MI. Conducted analysis of conduction velocity using the isochromal bands that we can get with the N-site system, and then did program stimulation with burst spacing as well as up to three extra stimuli down to the effective refractor period, or no lower than 200 milliseconds, whichever happened first, and we considered a positive induction as any ventricular arrhythmia that sustained more than 30 seconds. So those are the results. Out of the 10 pigs that we underwent, the MI2 died after the MI before the terminal surgery, and then we were left with 4 in the MI group and 4 in the MI plus radiation group. And the mean time from the time of the MI to the terminal surgery was about 78 days, and in the 4 pigs that underwent radiation, 48 days from the MI to the radiation, and then about 32 days from the radiation to the terminal surgery. So this slide shows a sample of what we have in one of the pigs who underwent MI on the left side here, and you can see my pointer, yep, here. And then on the right, MI plus SBRT. You see it's a little bit more pinkish on the gross specimen, and more whitish, suggesting more fibrosis in the SBRT group. And then on the high-density electroanatomic mapping, you see that just to orient you to the colors, purple is normal myocardium, and all the other colors are diseased myocardium. And as you see in the MI-only group, you have different shades of colors and different spectrums, meaning that there are some voltages that are elevated, whereas you have mostly gray in the MI plus SBRT group, suggesting that there is more complete scarring of the area. And when we quantify that, although the p-values never met significance because of the small numbers, the percent of dense scar was significantly higher in the MI plus SBRT group, 33 percent versus 14 percent. In the normal myocardium, the conduction velocity, as we calculated, it was the same, north of 1.2 meters per second, but in the border zone, it was lower in the MI plus SBRT, 0.31 versus 0.56 meters per second. This is, again, a sample of the induction of monomorphic on the left and polymorphic ventricular arrhythmias on the right. We have the A-line on the top in green, intracardiac EGMs in red under it, and then six leads of EKG. The main finding is that all animals that had the MI only were inducible into sustained ventricular arrhythmia, whereas only one in the MI plus radiation sustained ventricular arrhythmia. When we look at the details, it didn't matter whether the pigs were male or female. And two out of the four ventricular arrhythmias that were induced in the MI group were monomorphic at a slower rate, but the waveform was monomorphic, and two were polymorphic. The one pig in the radiation group that was induced had only polymorphic ventricular arrhythmia. And then we looked at the histology for the fibrosis, and on the malariate trichrome, there was no difference in the dents in the middle of the scar, but at the border zone, there was significantly higher scarring in the MI plus radiation compared to the MI group only. So there are limitations to this study. We only studied the homogenization of the scar. We did not look at other mechanisms. There are some studies, as I'm sure most of you are aware, in mice primarily, looking at different doses of radiation and how radiation increases the expression of connexin 43 and NAV 1.5, the sodium current that increased basically the conduction velocity. So by increasing the conduction velocity, you can basically close that excitable gap and terminate arrhythmias. That may be a mechanism, and we didn't study that. We did not study multiple doses of radiation. We did just one fraction of 25 gray. And the big thing, we know that clinically, we don't treat monomorphic and polymorphic ventricular arrhythmias as the same, but physiologically, they're different. But for this study, we did the same. Still, if we took only the monomorphic ventricular tachycardias in the MI group only, we have two and none in the MI plus XRT pigs that were inducible. So we show that there is a difference in the scar burden in the overall MI area with the addition of radiation to the MI, and we have also much more inducibility without the XRT. So these two things taken together would suggest that the homogenization of the scar is an important, is one of the main factors that control the arrhythmia or suppress the arrhythmia, at least in this animal model. Future studies will need to look at other mechanisms. And most importantly, especially for the other mechanisms, because once you have fibrosis, you have fibrosis. I don't think it's going to be too much reversible. But for the other mechanisms of expression, looking at the durability of these changes would be important. So I want to thank the team that worked with me, EP, and again, I mean, a common theme with all these studies is that they are good studies because they involve multiple teams. Radiation oncology, we don't typically work with them. It was a great pleasure connecting with them, again, cardiac anesthesia, et cetera. And I want to thank the McCamish Foundation for their generous grant, without which this study would not have been done, as well as HeartRhythm for ushering this to publication. Thank you. Hi, it's Tim Smith from Wash U. That was great. Thank you so much. I was wondering about the inducibility tests and where you stimulated, how many different sites, if any. And I was also wondering how the histology, if you have it, correlated with the electro-anatomic mapping. Yeah. Very good. Two very good questions. One, the stimulation was from the base of the heart because particularly in the pigs that have the XRT, putting the grid, that HD grid, in the middle of the myocardial scar was not capturing. So we were in the base of the heart more so inferiorly than in other locations. In the LV? In the LV. Yeah, yeah. Everything is in the LV. We didn't touch the RV. Okay. So it's all, sorry, I didn't mention it. It's all retrograde through the aortic valve into the LV. The other question, we don't have, so the way we sampled the tissue is after the surgery, the terminal surgery, we took out the heart, looked at it grossly, took something from the middle of the scar. What we estimated was the border of the scar and then something that is far away. So we don't have that level of correlation where we've taken multiple samples, which is one of the limitations of the study, meaning that could we have, you know, with the sampling, missed the exact area of the border of the scar and maybe were a little bit more in the normal myocardium? Maybe. Great. Thanks. So we found no real difference in histology between the SBR treated pigs and the others, but we did see some differences in refractory periods depending on where we paced. And we did do stimulation in the RV as well as the LV. So thank you very much. Thank you. Thank you. Nice presentation. Did you observe any hemodynamics in those animals as well, in the control group and the treatment group? During the terminal study? So during the arrhythmia, yeah, you know, there was, I think that what you're asking is in sinus rhythm, was there a hemodynamic? I don't think so. We didn't look at that in particular, so I would have to go back to the data, but there was nothing that struck us as a difference between the two. Thank you. Sure. Did you do an echocardiogram on the heart to see if are you losing any mechanical function by increasing the size of the scar? Yeah. So we did not on a couple. So there was part of the protocol was to try to get cardiac MRIs on some of those pigs, and we ended up doing it on two pigs, and it, you know, then it was a nightmare to continue doing that. So on those, you know, one sample from each, one MI, one MI plus XRT, there was no difference. Thanks. Sure. Thank you. Thank you very much. And so we will, we will close the session with a presentation by Richard Carrick from Johns Hopkins University, and the title is Identification of High-Risk Imaging Features in Hypertrophic Cardiomyopathy Using Electrocardiography, a Deep Learning Approach. Good morning. Thank you. I feel like a little bit of an imposter in this room of great basic scientists, because this is a little bit less on the basic side. Let's see here. I have no relevant disclosures. Oh, here we go. All right. Well, thank you very much for the invitation to speak. I'm going to be talking about my project, which was Identification of High-Risk Imaging Features in Hypertrophic Cardiomyopathy Using ECG Deep Learning. So just as a little bit of background, hypertrophic cardiomyopathy is the most common genetic cardiomyopathy. It affects about 1 in 500 patients, or people, I should say. It's an important cause of sudden cardiac arrest, particularly in young adults, where if you look at patients who have had a cardiac arrest and have a cause identified, around 10 to 20% of those can be attributed to hypertrophic cardiomyopathy. And as a result, primary prevention of sudden cardiac death using ICDs is really a core component of HCM care. How do we figure out who needs an ICD? Well, the guidelines recommend consideration of an ICD in patients who have high-risk features identified for cardiac arrest. There are seven different distinct features, of which four of these, LV ejection fraction below 50%, massive left ventricular hypertrophy greater than 3 centimeters, LV apical aneurysm, and extensive LGE greater than 15%, are all based on imaging assessment of the patient. And here, cardiac MRI is either necessary for assessment or provides a more accurate assessment of these risk factors. The problem is that access to CMRs can be limited, particularly in lower-income settings. So here's an example from India, where the costs of CMR are really exaggerated relative to other testing modalities, 80 times the cost of an EKG, 10 times the cost of an echocardiogram. Availability is also dramatically lower. For example, if you compare a high-income nation to even a middle-income nation, the rate of testing is 500-fold less. So CMR is critical for sudden cardiac death risk stratification in HCM, but it's also costly, has limited availability, whereas we compare ECGs, which are inexpensive, non-invasive, and widely available. So this prompted us to ask the question of whether we could use ECG deep learning to help identify some of these sudden cardiac death risk factors and at least guide our use of cardiac MRI for their formal identification. We looked at patients who had an HCM diagnosis, had 12 lead ECG available at the time of their initial evaluation. We excluded patients who had prior septal reduction therapy, which impacts the ECG morphology. And we identified two cohorts, the first from Tufts Medical Center in Boston, a large 2,000 patients that we used for developing our deep learning model, and then a smaller cohort we used for external validation from Kochi, India, and our collaborators there, about 230 patients. And for these models, we were trying to identify each of the four imaging-based high-risk features, LV systolic dysfunction, massive LV hypertrophy, LV apical aneurysm, and extensive LGE. And within a population of HCM, the prevalences of these high-risk features are relatively low, between 2 and 4 percent. When we were designing our deep learning model, we selected an architecture of a variational autoencoder. Basically, what this does is it takes a kind of raw ECG signal, encodes it into a much smaller set of latent variables, which contain all of the information contained with that ECG, and can be used to reconstruct, through a decoder, the original ECG signal. And one of the many advantages of this type of architecture is it allows for model pre-training, which we did using the open-source Physionet data set of around 90,000 ECGs. We then used that pre-trained model and trained a subsequent high-risk feature classifier using the Boston cohort, took median beat ECG waveforms, plugged it into the pre-trained encoder portion of the variational autoencoder, and those latent variables were used as the input to a dense network classifier, along with age, sex, and heart rate, to predict each of the four high-risk features. We used C-statistics as our kind of primary method of evaluating the model performance, and the models really did quite well for systolic dysfunction. C-statistic was 0.72, massive hypertrophy, 0.83, apical aneurysm, 0.93, extensive LGE, 0.76. So all of these would be considered at least reasonably good predictors. One of the other advantages of variational autoencoders is they do have what I would consider to be intrinsic explainability, so if you take those latent space variables, you can plot those in either two-dimensional or three-dimensional space using something like principal component analysis to get an idea of the distribution of ECG characteristics that reflect patients with or without a particular diagnosis. So here I've shown in red patients who have LV apical aneurysm, and in blue patients who do not have that, and while there's overlap in the distributions here, you can see that the mean value and the standard deviations or the variance of these ECG characteristics are really quite distinct, and you can kind of formally test this statistically. And then the other thing you can do is you can use the median ECG, and you can decode that to generate a representative ECG for patients with or without this diagnosis, and we see, for example, here, again, with LV apical aneurysm, the model identifies patients who have these deep lateral precordial T-wave inversions and T-wave inversions elsewhere as well. And we can do this for each of the four different high-risk imaging features, and we see that the ECG characteristics that are highlighted are slightly different for each one, kind of reflecting the distinct natures of those diagnoses. You know, clinical implementation is really important with these models, so we tested out the use of our ECG deep learning model as part of a hypothetical clinical screening protocol. We compared this first to echo plus unselected, basically universal CMR testing, which would require all of the 1,700 patients. We just looked at the patients who had, in clinical care, actually gotten CMR, and if we use this universal selective testing, the sensitivity is very high, 100%. You get every single patient with a high-risk feature, but the testing efficiency is low. Of all the patients we tested, only about 1 in 10 will have a high-risk feature identified. Compare this to echocardiography without CMR, sensitivity is much lower, only around 50%, whereas if you use echo plus a selective ECG-guided CMR approach, where you only perform CMR in people who the ECG deep learning model recommends, a high probability of a high-risk feature, you can reduce greatly the total number of CMRs used, only about 40% compared to a universal screening approach, while maintaining a very high sensitivity to detect those high-risk features of around 97%. And then the testing efficiency, as you might imagine, increases quite a bit as well, so 1 in 4 patients using this approach who had a CMR had a high-risk feature identified. We then externally validated this using the India cohort, slightly different cohort than the Boston cohort. Patients tended to have less outflow obstruction, less kind of advanced heart failure symptoms. CMR, as you might imagine, was performed in a much smaller percentage of the total cohort, only around 40%, and the rate of these high-risk features was much higher, about 4 to 5-fold increased with the exception of LD apical aneurysm, where there's only 1 in this cohort. And really the model held up for most of these high-risk imaging features, C-statistics greater than 0.7 for systolic dysfunction, massive hypertrophy, and apical aneurysm. The one exception here is that the extensive LGE model did drop off below 0.7. We attribute that to the very low rate of CMR testing within this cohort. So just to wrap up, so we put together a novel set of computational tools that can accurately identify HCM high-risk features using ECG, namely systolic dysfunction, massive hypertrophy, apical aneurysm, and extensive LGE. I think there's some potential clinical utility to this, specifically for improving the allocation of CMRs and identifying patients with high-risk disease phenotypes. And then one thing I personally feel strongly about is making these models publicly available. So for those of you who are interested, you can access the code to this model directly on GitHub at the link provided here. So I wanted to thank the Heart Rhythm Journal for publishing our paper this past spring, as well as to the many co-authors on our project. Thank you very much. Questions? Very, very nice work. How contemporary were the EKGs and the imaging that correlated with them? And I think you alluded to it, I don't know, I didn't, you know, get to do the numbers on the fly, but some patients may have had, you know, the septal thickness and the aneurysm or vice versa. So how did the model perform with these things? There was, so to answer the first question, we tried to limit it to six months within evaluation. And for the second question, there was definitely some overlaps, you know, almost all the patients who had LV apical aneurysm are going to have some amount of LGE, and likewise for the massive hypertrophy. We looked at that in the manuscript, I don't have the numbers off the top of my head, but I would probably say, you know, 20% of the patients had more than one in the holdout testing cohorts. Thank you. I was wondering if you did any failure mode analysis on this, on like looking at are there any common features of cases where it did fail, anything you can kind of glean or extract from that? It's a good question. We didn't do that. Yeah, it's a good question, but we have not looked at that previously. Okay. Thank you. Yeah. Good morning. Thank you. Pat Nufke with Catholic Precision. Did you happen to, do you have any data or any way to look at the appropriateness of this screening, you know, rather than just looking at the state of the art now, but could you see if there's like appropriate shock was delivered, and like if you can train your model to look for, you know, that in the future? Yeah. We didn't include this as part of the manuscript, but we actually did during our analysis look at outcomes, and you can risk stratify using these, the ECG prediction of high-risk features. It's less strong, as you might imagine, since it's a less perfect test for the high-risk features than the high-risk features themselves. So we wouldn't advocate using the model prediction directly for risk stratification necessarily. Okay. Thanks. Good question. Some of the differences of this, just blinking over the ECGs, seem to be changes in the voltage amplitude of the QRS. Yeah. Absolutely. I wonder if you would correct for body mass, or if some of the failures were obese patients. Yeah. It's a good question, and, you know, definitely more of those high-risk features are associated with people who have more severe disease, as you might imagine, and that often shows up with an increased QRS amplitude. We did not include body mass index, but that is a, it's totally an appropriate thing to think about. Yeah. Makes sense. Thank you. Thank you very much. If there are no further questions, this closes the session. I want to thank all of the speakers and the attendees, and I want to tell you to please keep submitting your papers to Heart Rhythm. Next year you may be here. So thank you all very much.

Heart Rhythm conference

cardiac conditions

left bundle branch block

cardioneuro ablation

ventricular arrhythmias

radiation therapy

ECG deep learning

hypertrophic cardiomyopathy

myocardial injuries