at the University of Washington in Seattle. Here chairing the session with Dr. Zhao, who is from the University of Auckland Institute of Bioengineering. And we have a very exciting session this afternoon that's entitled The Many Faces of the AFib Substrate. We're all introduced to AFib as an electrical abnormality, but most of us in this room realize that it's not just that and our speakers today were carefully chosen to cover various aspects of that substrate. So what we're going to do is have our speakers deliver their talks. We're going to go for about 10 minutes and then we're going to have a couple of minutes of questions for each speaker after. And obviously, if we have time at the end, we'll also take some more questions. I would encourage everyone to come up to the microphones and ask their questions. There's also an option to type them in. And we have this iPad here that we'll use to track any questions that you guys send in through the app. So with that, we'll get started. Our first speaker is Dr. Yogi Swaran from the University of Washington. And she is going to cover some of her work covering the functional parameters of atrial myopathy and what she's learned from cardiac MR. So hi, everyone. I'm one of the ECHO fellows at the University of Washington. And as Dr. Akum mentioned, I'll be talking about functional parameters of atrial myopathy. My main objectives are to highlight some key CMR studies that advance our current understanding of atrial myopathy and present some new insights from our work in the UK Biobank. So before we begin, what is atrial myopathy? So the definition is broad. And oftentimes, you'll hear this interchangeably with atrial cardiomyopathy and atrial cardiopathy. But the current consensus guidelines defines this as any complex of structural, architectural, contractile, or electrophysiologic changes that can affect the cardiac atria with the potential to produce clinically relevant manifestations. Now, several large epi and clinical studies have shown us that imaging measures and EKG measures of atrial myopathy are independently associated with an increased risk of not only just atrial fibrillation, but also its downstream complications, like ischemic stroke and heart failure. However, a major limitation of this prior work is they have mostly relied on insensitive measures, such as LA diameter and measures that are challenging to reproduce, like echocardiographic LA strain. CMR helps us overcome this limitation and is a gold standard for atrial volume and function assessment with its high spatial and temporal resolution. In head-to-head studies comparing CMR and echo, echo has been found to underestimate LA size by almost 15 milliliters. Additionally, as many of you have seen clinically, CMR is the added benefit of tissue and scar characterization. Previously, in the multi-ethnic study of atherosclerosis, or MESA, a large cohort study based out of the United States, researchers observed that CMR-derived atrial parameters were independently associated with atrial fibrillation. At baseline, individuals who went on to develop AFib had a 17% larger LA size. And two studies observed that RA volumes, LA volumes, LA ejection fraction, and LA strain were independently associated with the longitudinal risk of AFib after adjustment for clinical risk factors. Now, these early studies have been limited. Some of them had low event numbers, and others had low sample sizes, which has limited our ability to really understand atrial myopathy as a generalizable risk measure. The UK Biobank imaging study has allowed us to overcome this. By allowing for a large-scale platform for assessing atrial myopathy. The UK Biobank is a large longitudinal cohort based out of the United Kingdom. They began recruiting in 2006, and has recruited over 600,000 participants to date. The imaging cohort sub-study began in 2014, and has over 65,000 cardiac MRIs. And there's a repeat study currently ongoing. Several existing UK Biobank analyses over the last few years have shed some more insight into this concept of myopathy. For example, the original UK Biobank investigators identified that CMR measures of LA structure and function are independently associated with incident CVD and prevalent CVD after adjustment for vascular risk factors. Auberg and colleagues identified 18 novel loci associated with LA myopathy. And Pericelli's group identified loci associated with LA myopathy and RA myopathy. Algorithms have enabled our group to also do the same. By and colleagues from the UK Biobank published a fully convolutional network that takes the CMR images as inputs, learns the features through a series of convolutions, and then finally predicts the image segmentation, which you can use to get the automated measurements. One of our AI research scientists, Jennifer Brody, applied this published algorithm to the 65,000 available CMRs, which has allowed us to access this data. So in our study, after I quality controlled those measures and excluded individuals with baseline IFIB and those without follow-up data, we ended up with a cohort of 51,693 participants with high-quality CMR measures. So our mean age was 65, 48% were male. It was predominantly of European descent, 97%, and the cohort was relatively healthy. Less than 1% had a history of heart failure and only 2% had a history of prior MI. For the results of our primary analysis, during a median follow-up time of four years, we observed that both left and right atrial structure and function measures were independently associated with not only an increased risk of atrial fibrillation, but also ischemic stroke and heart failure. So all of these hazard ratios are reported per standard deviation increase in each unit of the atrial measure. So for example, each increase in left atrial index minimum volume was associated with a 55% increased risk in atrial fibrillation over time. Notably, we also observed a novel association between left atrial index minimum volume and the risk of nuance at dementia. Now these results remained relatively unchanged when we adjusted for secondary adjustment parameters, which included EKG measures of atrial myopathy, biomarkers, and the Townsend Index. Additionally, the ischemic stroke, heart failure, and dementia analyses remained robust after we adjusted for AFib during the follow-up period, also called intercurrent AFib. We also evaluated for effect modification and observed important sex-specific differences in risk. So for example, each standard deviation increase in left atrial index minimum volume was associated with a 17% higher risk of AFib and an 18% higher risk of heart failure in women compared to men. Now similar interactions were observed with left and right atrial maximum volume measures. Next, we looked at categorical associations by evaluating these measures as quintiles. So here, participants in the highest LA quintiles had more than a three times increased risk of atrial fibrillation compared to those in the reference quintile. For ischemic stroke, the highest LA measures, again, had more than a 1.5 times increase in risk. And for heart failure, the highest LA quintiles had more than a 2.5 times increase in risk compared to the reference quintile. And these right-sided associations were observed to be not as impactful as the left-sided associations, which is also demonstrated in our AFib and ischemic stroke analyses. And for dementia, we did not see any significant associations when we observed these measures as quintiles. Our group also performed genetic analyses to further evaluate the associations between left atrial myopathy and these longitudinal outcomes. And we identified 51 loci that were associated with left and right atrial structure and function measures, 27 of which were novel. So 25 of these were associated with LA myopathy and 26 with RA myopathy. And in our cohort, we did not find any significant overlap in the measures. Additionally, when we did genetic correlation analyses, we observed that right and left atrial measures had relatively low levels of correlation as shown in the figure on the right. For example, LA volumes and RA volumes had a correlation coefficient of less than 0.3. Now these results are not unexpected and align well with our epi results. We also compared our genetic analyses to the largest existing AFib GWAS and the previously reported atrial GWASes and observed that several of our loci do not overlap within 500 kb of either. Some of these loci, when evaluated individually, are near regions like TTN, AMZ1, and PITX2. These are regions that have been mapped to cardiomyopathy, AFib, and AFib risk factors. So TTN is TITAN, which may be familiar to some of you, and is associated with cardiomyopathy and several studies have published an association with early-onset AFib. AMZ1 stands for arculase and family metallopeptidase 1 and is reported in a HEIT GWAS, which is a known risk factor for atrial fibrillation. And PITX2 is paired like homeodomain transcription factor 2 and has been well-published for its association in the pathogenesis behind AFib. So taken together, what we've learned is that biatrial myopathy is independently associated with an increased risk of atrial fibrillation and atrial fibrillation-related outcomes, and these associations were consistently stronger in women, highlighting the need for personalized risk assessment that takes into account for sex-specific findings. Our genetic findings reveal distinct loci associated with left and right atrial traits, which might be future therapeutic targets and warrant further study. And clinically, perhaps most important of them all, the study highlights that high-quality measures of atrial structure and function might be a novel screening tool that we can use to identify those who are at highest risk for adverse cardiovascular events, especially in women who are at highest risk for AFib-related outcomes. So thank you to my main mentors, Dr. James Floyd and Dr. Nazim Akoum, my co-first author, Jennifer Brody, the University of Washington Cardiovascular Health Research Unit, my many collaborators, mentors, and for the Heart Rhythm Society Postdoctoral Grant for funding my work this year. Thank you. Any questions? Thank you. Vid, very, very impressive work. I have a question for you regarding the right atrium. We sometimes get too focused on the left side, but what you're showing us here is that we should look at both sides. Were there any specific features? You mentioned some sex differences, but was there anything that should make us look rightward instead of leftward from your analysis? So I think one thing that was interesting to me is I feel like most of the literature to date really talks about the left atrium, and I feel like our left-sided associations, especially for AFib and heart failure, were not that different than our right-sided, or that our right-side associations were not that different than our left-sided associations. And there's actually that one study from MESA that shows that right atrial measures, independent of left atrial measures, are also associated with the risk of atrial fibrillation. And although I didn't present that here, when we adjusted for left atrial measures in our right-sided associations, we found that these measures were still independently associated with longitudinal outcomes. So I think it's a topic that warrants further study, and it's something I'm excited about. Go ahead. The microphone is right behind you. Hi, Gunjan Shukla from New Jersey. Great, fascinating work. Just one question is, if we find a loci, and we see that those people are at high risk for earlier atrial fibrillation, what intervention we should do? Is there any work towards that to prevent, or does it guide you to treat AFib ablation differently in those regions? I don't think I can comment on the AFib ablation question, but regarding what we do with those loci, I think now the public research platforms have been made relatively available to most of us. So there's one, FUMA, where if you plug in the different loci, you can find genes that are mapped close by. So in our analysis of the 51 loci, 26 of which were right atrial associations, we observed that there were maybe 27 different drugs that were currently approved, and 20, 25 or so that were currently investigational, and then several more experimental. So I think we're still several steps out, but it does feel like the current research platforms are making this more available to us, so we can say, hey, maybe these medications can be repurposed and studied in that way. I have one question to follow up. Another question is, you look at the left and right volume independent associated with atrial fibrillation. Have you looked at combining them together, how much increase left plus right volume? Oh, like combining the? Yeah, yes, yes. I didn't combine them in a risk measure, but I think that's an interesting, oh, yeah, sorry, I haven't combined them in a risk measurement tool yet, but I think that would be an interesting thing to evaluate. I think there was one small study that did find there was additive benefit of combining both. Great, there's one question from the audience. Can you comment on the relative importance of absolute versus indexed atrial size? Absolute versus? Indexed atrial size. I think as imagers, we tend not to really think about absolute measurements just because if somebody has a BMI of 40, their atrial size is gonna be different than someone with a BMI of 18. So I think for the most part, we're more interested in index measurements. All right. If you can make it quick so we can stay on time, that would be fantastic. Thank you. Arjuna, hi. One question here, which is the palmiven antrum ratio to the LA volume. Did you look at that in terms of correlation between AF and the size? Thank you. Did you say the pulmonary vein? Yeah, PV antrums to the size of the left atrial. I think that would be an interesting future study to look at. I think from the published algorithms to study it on this large of a scale, we only have those volume and function measures. Great. We're gonna move on to our next speaker. Is Dr. Kalman here? Not yet. So we'll skip and hopefully he will join us later. We'll go to Dr. Heichel next. And he is gonna tell us about metabolic shifts in the atrium that are associated with atrial fibrillation and what he discovered using PET imaging. Hello everyone. My name is Romano Saikal. I'm a research fellow at the University of Washington, and today I'll be speaking about metabolic shifts in AF, evidence from PET imaging. So as we all know, the cardiomyopathic atrium is marked by multiple remodeling processes, including structural, functional, and electrical remodeling processes. But one substrate must not be overlooked. It's the metabolic substrate. The metabolic substrate involves, but is not limited to, glucose, fatty acids, and keto metabolism. In glucose metabolism, there is a shift from a glucose oxidation to glycolysis. And to put that in numbers, under physiologic condition, 10 to 20 percent of the ATP supply comes from glucose metabolism. This number increases by 10 to 20 folds in atrial fibrillation. And why am I focusing on glucose metabolism? Because our topic today is on FDG PET imaging. So FDG is being used in cardiac sarcoidosis. It has its own methods, and it yields metabolism images that, in conjunction with perfusion, could aid in diagnosing cardiac sarcoidosis, as well as quantifying the severity of inflammation in the ventricles. But does it actually have a role in the atrium? Does it have a role in atrial fibrillation? Our team recently published a meta-analysis of six studies looking at the association between FDG uptake in the atrium and atrial fibrillation. And it found that the left atrium was associated with an odds ratio of 14.5, while, interestingly, the right atrium had an odds ratio of 52. Now, some of those studies even went further and segregated patients into paroxysmal versus persistent AF patients. And they noted that persistent AF patients had a higher left and right atrial wall uptake. And this is evident in the quantitative analysis showing higher SUV max, as well as target-to-background ratio in the persistent AF population. This is a catheter ablation study looking into data from catheter ablation. The first image, the top right-hand side image, is a voltage map. And this was compared to PET scan imaging quantifying atrial uptake. And it found that voltage scores was negatively correlated with atrial uptake. And as you can see in the bottom right-hand side, TBR was negatively correlated with the voltage score. Interestingly, AF termination in persistent AF patients was found to be associated with the presence of an abnormal right atrial uptake with an odds ratio of 10. However, none of the other FDG parameters were predictors of atrial fibrillation recurrence. This is a study of 36 persistent AF patients undergoing catheter ablation. Prior to the ablation procedures, they undergo PET scans. And if they achieve sinus rhythm two months later, then they would repeat the PET scans. And we can see here that in those population who achieved sinus rhythm, the left atrial wall uptake decreases significantly from pre-ablation. And there was actually no difference in either SUV or target-to-background ratio between them and healthy volunteers with no known atrial fibrillation. Now, one of the outcomes, famous outcomes, looked at in atrial fibrillation is stroke. From a PET perspective, atrial uptake was associated with an increased prevalence of stroke. And the higher CHADS2-VASc score correlates with a higher right atrial wall SUVmax. And I am focusing on the right atrial wall because it is repeated in each and every study, the predominance of the right atrial wall. And if we want to take a deeper look into this table, we could actually see here that both qualitative and quantitative variables were significantly elevated in the stroke group. Now, this is a study of 118 cardiac sarcoid patients that were followed over a period of six years. And it was found that in patients who had PET positive evidence of atrial uptake, the cumulative incidence of atrial fibrillation was higher than those without any atrial uptake. And they went a step further to show that atrial uptake was a better predictive of incident AF than atrial dilation measured by MRI scan. Now, our work at the University of Washington involves investigating FDG affinities and linking it to an AF diagnosis and adverse cardiovascular outcomes. So what we're doing is we are evaluating those PET scans and looking at atrial uptake. We are using something called a 3D region of interest across the highlighted area, as you can see, and this would yield an SUV mean and an SUV max. And we normalize this value to the blood pool to be able to compare them between multiple patients according to this formula. We are also reporting on qualitative assessment. We are describing pattern as focal, like in the top right-hand side image, diffuse, like in the top left-hand side, and focal undiffused, like in the bottom two images. We are also comparing the intensity of those to the liver, yielding an avidity score of 0 to 3. Now, our first aim, linking atrial avidity to incident AF, we excluded patients who have evidence of incomplete suppression on PET scans, as well as a previous documented history of AF. And out of the 206 patients, 50 of them developed AF. And what was also reiterated in our study is the predominance of the right atrial uptake. As you can see, the right atrial uptake out of 35 patients was present in 25 of them. Now, this is a Kaplan-Meier curve showing the AF resurvival with a follow-up period of six years. And we could see that the atrial avidity was associated with a new onset atrial fibrillation with a hazard ratio of 3.19, even after adjusting for cardiac sarcoidosis. Now, as for the adverse cardiovascular outcome, out of the 313 PET scans that we read, 154 patients experienced major cardiovascular outcomes. And it was found that the ATBR mean and max in the MACE group was significantly elevated than the patients who did not experience any adverse cardiovascular outcome. Now, this clustered bar graph on the left-hand side shows the individual and the cumulative outcomes. And we can see that the significance, the percentage of patients with avidity is significantly elevated in the heart failure hospitalizations and the mortality as individual outcomes. However, we see similar trends in stroke and acute coronary syndrome, but we didn't achieve any significance yet. PET scan imaging. A new lens on an old problem. As we all know, atrial fibrillation is more than an electrical disorder. It involves metabolic and inflammatory changes. And I think, although PET scans are relatively young, they are highly capable of identifying active remodeling, identifying high-risk patients, and identifying potential upstream targets for management. Thank you so much. I want to thank my team. And I'm open for any questions. I have a question. Go ahead. Yeah, Larry Stearns from Victoria. Very interesting stuff. I'm just wondering about the right atrial predominance. I mean, seeing so much of this, was there any correlation made with clinical characteristics? And I'm thinking about obesity or sleep apnea, or the difference between atrial fibrillation and atrial flutter? In other words, did a lot of them have more right atrial arrhythmias, like flutter, than fibrillation? Well, some of those studies, for the second question, some of those studies separated atrial fibrillation and flutter. They only included atrial fibrillation patients. And the other studies, like our ones, they include both atrial fibrillation and flutter. And you are seeing the same thing with regards to the right atrium. There is no difference. There is a predominance in the right atrium. So that's the question. As for the first question, we are seeing some trends in obesity and sleep apnea. Patients with obesity and sleep apnea have an increased avidity, not necessarily right atrial avidity, maybe left or biatrial. And that's it. Thank you. Thank you. Fantastic talk. Congratulations. Very important message you deliver. And the question, I have multiple questions, but first question, what exactly do you see when you say about metabolic? Do you actually highlight atrial, myocardial tissue? Who is actually producing, for example, in right atria, this high loading intensity? Oh, it's a fat or it's any inflammatory cells? Well, mechanistically, who is providing the signal? Yeah, I understand that. It actually depends. We try to separate the epicardial adipose tissue from the atrial wall, from the blood pool. But again, PET scans have its own limitation. It is subjected to partial volume effects. So sometimes they might overlap. We try to use the Hounsfield unit to be able to separate those. For epicardial adipose tissue, it usually ranges between minus 50, minus 200 and minus 50. But we try to separate them. It's not always possible. We exclude cases that we think where the PET scan has a very low resolution. We are not able to like clearly distinguish. But I think overall we are capturing the atrial wall. Actual atrial cell, myocardial cells? Yeah, atrial myocardial cells. And another question regarding any electrophysiology data, like from this any patient, do they show any low voltage or any abnormal conduction? Any abnormal sinus rhythm? Again, it was taken, these scans, during a FIP or during sinus rhythm? Yeah, so not all studies distinguish the rhythm during the PET scan. So some studies actually report if the patients were in PET scan, like this stroke one I showed. There is like the top column that actually says how much patients were in atrial fibrillation during the PET scans and not. Now that's for the first part. For the second part, the catheter ablation data that we have, the voltage mapping, when we compare those to the PET scan and showed atrial uptake, the voltage score was actually the propensity of low voltage area segregated across six different areas in the atrium. So they separated the atrium into six different areas and they checked for the propensity of low voltage zone in each one of them. And if, for example, low voltage zones were increased in all six areas, the voltage score would be six. And this voltage score was negatively correlated with target to background ratio. So the higher the uptake, the lower the voltage score. Okay, you do have this correlation. Yeah. In left atria or in right too? In the left atrium only. Left atrium. Yeah, we don't have in the right atrium. Any correlation, my last question, any correlation with just first speaker present about size, different function, left, right atria? Yeah. Did they provide some like correlation with these parameters? Yeah, there is actually one study. It combined FDG uptake and function using MRI PET scans. And it showed that when patients return to sinus rhythm, the function increases. It becomes much better. And the atrial uptake decreases. Decreases. Yeah. In sinus rhythm. Thank you so much. You're most welcome. Very exciting work. Thank you. I'm going to have to move on. Yeah, we're going to move on. The next speaker is Dr. John Carman from Royal Melbourne Hospital. So he will teach us about the epipose and fibrosis. Thank you very much. And my apologies for my delayed arrival, which was unavoidable. Okay. So somewhat of a clinical bent. And just to make sure the slides are moving. So it's very well recognised that population studies relate obesity to incident atrial fibrillation. A number of studies over the years. But what's also clear is if you look at these studies, patients can change their AF trajectory depending on changes in their risk factors and in particular their BMI. And in this particular study, the women's health study from Usha Tedro and Christine Albert, patients who had high, rather patients who increased to the greater than 30 BMI from normal BMI had an increment in risk, whereas those who reduced from above 30 had a change in risk. And this was first subjected to a randomised trial over a decade ago by the group in Adelaide, Pratt Sanders group, and showed that if you address risk factors and you lose weight and have significant reductions in BMI, that you see significant reductions in AF burden. Now what is going on at more of a functional level? And this is very well-known work of obesity in sheep and weight loss and looking at the pathological and electrophysiological changes. And obese sheep develop inflammation, they develop adipocyte infiltration, they develop fibrosis and they have reductions in connexin levels. And all of those changes with progressive weight loss reverse really completely, at least again emphasising a relatively short-term model. The changes do correlate with electrophysiologic changes. So here's baseline stimulation over here, normal propagation. In the obese sheep, you see this bunching of isochrones on the epicardial plaques and changes in wavefront direction that signify conduction slowing, conduction heterogeneity increases. And with that, induced AF duration increases. And again emphasising the relatively short-term nature of the model, weight loss reverses this. In addition, as previously indicated, in areas where there was regional areas of adiposity, there was infiltration of the fat in between myocytes. And this did translate here into reduction in voltage. And so there was both site-specific and more global regional conduction changes in relation to this. Now, we've long known about this association between body mass index and atrial fibrillation, the increased risk and waist circumference and other measures. But it turns out that the strength of the association between AF and obesity is greatest if you look specifically at epicardial fat, and that there's a greater dissociation. Increasing epicardial fat associates with higher relative risk of AF, which overall was really significantly greater than BMI alone. So there's something very particular happening with the epicardial fat. This is a human study of electroanatomic mapping, Rajiv Mahajan, patients with BMI over 30 versus patients under 25. And again, in humans, as in the animal models, you see conduction slowing in green, increment in abnormal signals, the fractionated signals, increment in regions of low voltage, percent low voltage, as indicated by the red regions on these electroanatomic maps. And further, on the MRI, the strongest correlation was in the regions between fat and electrical changes were in the regions where there was high-density epicardial adiposity. So here's the correlation with fractionation and conduction with BMI, relatively low, but improved inverse relationship to conduction velocity, direct relationship with increasing fractionation, when you looked at posterior left atrial adipose tissue and compared this to fractionation in the posterior left atrium. Krishan Mallya in our group looked at this in a mapping study in the operating room. And so patients had a pre-surgical theta CT scan and were divided according to or stratified according to the amount of epicardial adipose tissue on the CT scan. We were looking particularly in the regions that we could map in the OR, so anterior right atrium. Here are the maps comparing low anterior fat with high anterior fat. These are our epicardial plaques. And you can see, again, the bunching of isochrones in those patients with high epicardial fat and changes in direction of activation, so conduction heterogeneity. And between plaque activation time and anterior RA adiposity, conduction velocity, complex points, there was quite a reasonable correlation, a correlation that did not hold up when you just looked at more global biatrial adipose volume, again, suggesting there's something going on regionally. And indeed, the same held true for fibrosis. If you look at the regions of high anterior adiposity and underlying these, you will see a clear correlation between adipose volume and the amount of fibrosis, the extent of percent fibrosis seen in the tissue sections. And further, as we talked about previously, the infiltration of fat in between the myocytes, this also causes conduction change. So you can see in these regions of high adiposity that sort of infiltration of fat into the region. And with that increment in conduction heterogeneity, an effect that was independent of the fibrosis. And further affecting conduction are changes in connexin with clear lateralisation of connexin in regions of high anterior adiposity. Now, colleagues at the University of Melbourne performed a series of studies culturing atrial myocytes with either mouse pericardial or inguinal adiposity and showed that it was just the pericardial fat that caused prolonged activation and slowing of conduction. And also demonstrated that the epicardial adipose tissue has a secretome, which has a distinct proteomic profile when compared with inguinal adiposity. And multiple factors raised as a potential in causing all of these different effects that we're seeing, including the fibrosis. So the question that has sort of been interesting to address is whether this is reversible. And these are unpublished data. This is a randomised substrate study. Again, my colleagues, Dr. Parfait and Dr. Sanders, randomised patients with obesity and atrial fibrillation to either standard care or aggressive risk factor management. And over a follow-up period of 12 months, there was really significant improvements in bipolar voltage and in conduction velocity in the risk factor management group, not seen in the control arm. And with this, when looking on MRI, atrial epicardial adipose volume significantly decreased, left atrial volume significantly decreased. So both electrical and structural remodelling in concert with the reduction in atrial epicardial tissue. And the clinical correlates, which we touched on earlier, multiple studies looking at this. This is a nice study from the Cleveland Group showing that patients who are morbidly obese who undergo bariatric surgery have the same sorts of AF ablation outcomes as those who are non-obese, whereas the morbidly obese without surgery have very poor outcomes with obviously epicardial adiposity driving progression of remodelling as the hypothesis for that. And with the reduction in weight, all of these beneficial changes, blood pressure reduces. The phenotype of the AF changes more from a persistent to a paroxysmal phenotype. And that was shown also in the reverse AF study. Melissa Middeldorp showed that in patients over five years who retain more than 10% body weight loss, the progression from paroxysmal to persistent AF in orange down here, 3%, is rare. And further, that reversal occurs. So the phenotype in a significant proportion of those patients who could maintain, obese patients with atrial fibrillation, persistent AF who could maintain weight loss, frequently transitioned back to a paroxysmal phenotype. So significant reversibility. Now, I don't know if I've got a minute or so left. I just wanted to touch on a further aspect of imaging that's coming to the fore. And again, work from my colleagues. We know of the relationship between epicardial adipose volume and risk of atrial fibrillation in red. But there's this increasing understanding that increasing attenuation of fat on the CT scan reflects inflammation and a higher probability of atrial fibrillation. And in this study, the less negative or increased attenuation of the fat around the pulmonary veins did actually correlate with increasing AF risk. The attenuation is most marked adjacent to the veins. As you move away, the fat attenuation becomes less. And on multivariable prediction, both volume and attenuation of peripulmonary venous fat were predictors of atrial fibrillation. So inflammation, part of the story. And others have suggested that fat that's more attenuated is associated with higher risk of recurrence. And this is just, I think, an emerging space that can help us maybe predict who's more likely to have adverse or poorer outcomes from their AF management. So in conclusion, there's a clear relationship between what's happening locally with adiposity and arrhythmogenesis. And there are multiple mechanisms. Fibrosis, fat infiltration, inflammation, lateralization, and reduction of connections. And all of this slows conduction and changes refractoriness and creates the substrate for AF. And we've seen that, to a significant extent, the impact on AF risk is reversible. Thank you very much. Two questions. One regarding, when it's transition happening from persistent to paroxysmal, is it without any ablation, treatment, it just happening in this patient? They have persistent effipens and going, so you just record it, loop recorder, is it based on that? It's based on monitoring, intermittent monitoring. It is a study over five years and looking at the dominant phenotypes. Independent from treatment, like anti-arrhythmic treatments? Yeah, there are a lot of variables that are hard to control for in, you know, it's an observational study and it's an observational study over five years, but it's quite striking in the relatively smaller percentage of the entire group that could maintain that level of weight loss over that period of time, saw this, you know, reversal of the phenotype. Yeah, that's great. And another, like, mechanistical question. Reading all your studies and paper regarding fat infiltration and then reversible, does it mean that just kind of mechanistic fat cells, adipocytes, can actually diffuse inside of the tissue? Or are they still there and just increasing in terms of the volume significantly? Because they have huge potential. It's fine, like, I understand that we have kind of the same number of adipocytes in adult heart, or it's actually really diffusive process? You know, the answer is I don't know. We didn't, in the animal studies, I don't think that's been looked at, and obviously in the humans we just have a snapshot in time and a sort of comparison with other patients where we see no fat in between the myocardial cells. But, you know, are there sort of adipocytes that are much smaller and we're not seeing? I guess that's an interesting possibility, but I don't know the answer. Interesting question. Just to connect to the previous speaker, it would be very interesting, is obesity and this fat, epicardial fat layer, associated with FDG, like load. Do you see any combination? So a beast, body mass index, high body mass index, associated with epicardial fat as well as FDG uptake. There are some preliminary analysis that shows that atrial wall uptake could also relate to epicardial adipostitial uptake, and it could be like through an inflammatory process we're hypothesizing, but it's still a preliminary analysis. Cool. Thank you so much. This is great to have that much interaction. I just want to make sure that we're all okay staying beyond our hour. I want to keep this going. This is great. But if we want to stay within the hour, do we need to give the room up after? Go ahead. Sorry, hopefully a quick question. I work in a jurisdiction where if the patient's BMI is over 35, they're not funded for an atrial fibrillation ablation at all. And if it's 30 to 35, you have to go through a twin track approach of lifestyle modification plus the AF ablation. With regards to the over 35, from your work and your reading of the other work, do you think that's a justifiable position to take? Yeah, that's a great question. I think it's a health care system question. I think it's a difficult one because you've got to look at that data and say what percentage of patients can achieve that amount of weight loss. Are we making the new inhibitor drugs such as Azempic widely available enough for patients to be able to lose that amount of weight? We know that weight loss is really, really challenging. In practice, in clinical practice, I just tell patients we're in it together and they're symptomatic and they're bothered by their AF and I don't send them away for a year to lose weight. I do the ablation and tell them, look, if you don't lose weight, the likelihood of recurrence and address your risk factors is going to be a lot higher. Another quick question. Is there a strong correlation between obesity and epicardial fat, i.e., do you get obese people who don't have epicardial fat and vice versa? There's clearly a degree of disconnect between your BMI and your epicardial adipose levels. Broadly speaking, if I can use that term, obese people will have more adipose tissue, but the variation is significant and also regional. Just to piggyback on all that, has someone looked at the GLP-1s in terms of risk reduction for atrial fibrillation and making that part of your practice in terms of management for atrial fibrillation, your non-ablative or procedural treatment? Also a really interesting question. It's been hampered until now by the lack of availability of the drug to do a clinical trial. Certainly in many jurisdictions, in ours, available widely up until relatively recently only for patients with diabetes. I know that those studies are underway. There are a number of those studies that are underway. I think hopefully we'll see some interesting results from randomized data, observational data suggesting impact. So the weight loss associated with those medications would probably also correlate to adipose reduction, the cardiac adipose reduction? That's the hope. That would be my guess. Thank you so much, Dr. Campbell. So our AV team has told me that we don't have to give the room up, so we do have time. Hopefully everyone gets to stay. We want to be fair to our speakers, but so many great questions and great discussions. It's my pleasure to introduce Dr. Barnard as our next speaker. He's going to cover the myocardial metabolomics aspect of atrial fibrillation. There we go, great. Great, thanks for coming. I'm John Barnard, I'm a translational molecular epidemiologist at the Cleveland Clinic. This work is sponsored in part by funding from the NIH as well as from the American Heart Association and before I go into the talk, I want to thank my many colleagues at the Cleveland Clinic, particularly Julie Renneson and Mina Chung. Don't have any disclosures. We know from a long history of work that, as previous speakers have mentioned, that AF has a metabolic component, both potentially as a cause and consequence. There's a lot of metabolic changes. There haven't been a lot of data in humans. A lot of animal studies, a limited amount of data in humans, particularly directly measuring metabolites and what I'm gonna show today is directly actually measuring enzymes in the myocardium of humans. So the goals here today are to look at, as you go across AF state, and I'll clarify what I mean there, how do the metabolite and the enzymes controlling these metabolite processes change as you increase AF state? So there's a beautiful review paper late last year that sort of outlines a lot of the knowledge about metabolism in AF that I highly recommend. So we published a paper a couple years ago, Julie Renneson was the first author, along with my colleague David Van Wagner, where we had a very small human study, 12 samples. They were carefully selected to be matched valve surgery patients from the Cleveland Clinic where they either had very early forms of AF, so they were in sinus rhythm, either paroxysmals early in the process, or they had been persistent at some point and moved back to a paroxysmal state, versus long-standing permanent AFs. And then in this small study, we did untargeted proteomics. We saw 3,000 proteins, about 800 of them were enzymes. And we saw a lot of changes in AF, which isn't surprising. So here's a volcano plot of that. We saw 115 proteins in general differ between AF rhythm and sinus rhythm in this particular design. But the top one was a metabolic enzyme. Anilose-3 is a known metabolite involved in glycolysis. And so here on the right are some pathways that are changing in this small analysis. Glycolysis is one of the strongest down-regulated processes as you go from AF rhythm to sinus rhythm, where oxidative phosphorylation is increasing in this advanced AF state versus a very early AF state. Also, we saw electron transport proteins, metabolites greatly increase in the advanced forms of AF. And this is a beautiful picture that Julie Renison put together, showing particularly glycolysis enzymes going down, while complex I and some of the fatty acids, which is known in AF, are increasing with long-term AF, again, in this very small study. So in terms of, I wanna show some new data here today. We have a much bigger cohort. We have 222 cardiac surgery patients from the Cleveland Clinic. And here we've then divided them up into sort of what we're calling AF state combinations of rhythms. So they're all surgery patients. So the no AFs are predominantly valve patients, some cabbage-only patients. We have paroxysmals, persistent sinus rhythms. So at one point they were persistent. They were either cardioverted or they had reverted back into a sinus rhythm state. Then in sort of an early form of AF rhythm patients. So they haven't been diagnosed with AF that long. And then finally, sort of now called long-standing, what we previously called permanent, they've had AF a long time in their AF rhythm. So sort of trying to think of, even though this is cross-sectional, sort of progression of AF from an early state, from no state to an early form of AF and beyond. And so we've got a fairly large cohort here. Interestingly, we've got some lone AFs in there where they don't have anything particularly wrong with them, and they actually were going for MACE procedures at that time. So it's a large cohort. It's diverse, a lot of different factors. Obesity, hypertension, valve disease, coronary artery disease. So it's a broad spectrum of folks. I wanted to highlight a couple of things here in terms of age at surgery. A little bit older in the no AFs, but generally fairly consistent across the age range. It's slightly older in the AF rhythm groups. And as I mentioned earlier, the paroxysmals tend to not have the diseases long. The longstanding ones here over in the, you can't really see the mouse, but the yellow one all the way to the left, you can see the permanents have had AF a long time, and they're in AF rhythm. So we then did a newer proteomic method called data independent acquisition that allows you to do untargeted proteomics from one sample, one run per person. So we were able to scale it up to a much larger cohort. We saw about 2,500 proteins, which is a little disappointing. You know, there's obviously a lot more proteins in the heart than that. Of those, we found 371 metabolic enzymes. And so that's what I'm gonna focus on here and show you we did some attempts at adjustment. This is a very heterogeneous cohort, sort of like the previous speaker said. So we had a fairly sophisticated adjustment model, including use of what's called surrogate variables that attempts to sort of implicitly adjust for the major vectors beyond what you're interested in. So I think this is the most interesting plot I'm gonna show here today. So this is adjusting for the sort of known risk factors of the COVID rate model. This is called a multiple dimension scaling plot. So essentially, it's trying to preserve distances among these 371 metabolic enzymes. So here, as the colors go across from no AF, paroxysmal, persistent sinus rhythm, AF rhythm, and finally permanence. So as you go across, with the idea that the closer these two samples are, the more similar metabolic enzyme profile is, this global profile. And as you progress sort of through the quote AF states, you see that the triangles here are the average of those states. You see how the metabolic profile shifts, the enzyme shift away from a sort of no AF state, early forms of AF all the way to the yellow, very strong metabolic shifts as you go across that. You can see that as well here with, this is a little busy heat map, but here we've clustered the enzymes. I won't go too much into what the elements of the clusters are. But more interestingly, as you go across the AF states, you see some enzymes that are high. This is high early, they decrease as you increase AF state. You see some that are sort of much higher later in AF. So AF rhythm, and they even get stronger and permanent. So again, you're seeing heterogeneity in the patterns. It's a complex set of progression in these metabolite profiles. Going into the specifics of individual metabolites, individual enzymes, excuse me, that are involved in metabolic processes, we do see some strong players. And so here there's two different comparisons. And this, the top one's including everyone's in sinus rhythm versus AF rhythm, adjusting for all the things I mentioned. The bottom one is only looking at those with AF. Again, sinus rhythm versus AF rhythm. And what you see here, without going into much of the details here on the right, are fatty acid enzymes. Here in the green are mainly glycolysis enzymes. So we're seeing a decrease in glycolysis metabolites, an uptake in fatty acid and ketone driving metabolites, keeping metabolite enzymes, excuse me. Again, we're seeing this NLOS3 is the strongest association in this much larger cohort. So the original 12 was valve surgery patients, sort of well-designed, very small. In this much, much larger one, it's still the number one enzyme involved in metabolism. This is an approach of trying to ask which enzymes vary as you increase state. And so again, ENO3 increases sort of almost linearly as you increase state. I think that'll be more clear here in the picture. So here's the metabolite, here's the enzymatic levels of the ENO3 across the states. So you can see here in no AF, it's massively higher than it is in permanent, a longstanding AF. And you sort of see this nice beautiful decrease as you go through the states. Continues to go down. Versus acetyl-Co, acetyltransferase, consistently goes up as you go through AF states. But other ones, as you can sort of see from the heat map, are varied. So glutathione reductase only goes up as you go from no AF to AF, then it's sort of flat. And then similarly, you've got this lactate dehydrogenase only really seems to change when you're in AF rhythm. So having AF itself doesn't seem to have really elevated that. So just to sort of finish up here, in terms of pathways, sort of major ideas, I've sort of touched on this already. If you do a pathway analysis of the enzymes and ask what's changing, what we're seeing is actually the enzymes involved in glycolysis, although glucogenesis is actually reversing the glucose, back, reversing it back into glucose. All the sort of glucose-related stuff is down as you go through, as you increase AF state. While in ENOS activation, interestingly, is up, although that goes down when you're in permanent AF. So it's an interesting, it's sort of, at some point it gets activated and it seems to drop. So there's a lot here, it's obviously a complex story. Just to sort of wrap up, we're seeing metabolic shifts in enzymes that control those metabolites pretty clearly as you go through AF, but it's very heterogeneous. Even with the adjustment models, you can see that some folks are not behaving the same way at all. But there is sort of a very strong, typical type of shift. We also have RNA-seq data on a lot of these folks. We see the same sort of signals in the RNA-seq. Proteins and RNA are not the same thing, obviously, and we have seen in many genes where they don't match. They generally match here in the same folks, so which is encouraging. But the thing that was a little surprising to us was this idea that the glycolysis enzymes were consistently going down as you advance through AF. That was a little bit surprising. Not the fatty acid utilization part wasn't too surprising, but that enzyme shift wasn't quite what we expected. Limitations, enzyme levels and enzyme activity, that's clearly a major limitation. The other part is the proteome coverage we have on these human samples is fairly poor. We know from our RNA-seq data, basically on the same patients, a little bit larger, we can see almost 1,800 metabolic enzymes. This is bulk, of course, tissue, so it's mostly myocardial, since the cardiomyocytes make up most of the volume. But we're only measuring 380 of them, so we're missing a lot of the story. And clearly, this is a cross-sectional. We know this isn't really progression. This is really state change. And of course, this is incredibly heterogeneous human population. These adjustment models, residual confounding, the usual limitations of observational studies. Next steps, we have a lot of work to do. We've got really cool data where we actually have real metabolite levels in these folks as well. So we're trying to integrate metabolite levels, the RNA-seq data, and the protein to sort of get a better prediction of what's the actual metabolic state, and how did that metabolic state, at least in the bulk level, change as AF gets worse. Thank you. Thank you. One question from the audience here. Any changes in de novo lipogenesis pathways? We didn't see that, at least at the pathway level. Doesn't mean there aren't individual members of that in terms of lipogenesis. We did see one that I didn't mention that was a ketone enzyme. Not exactly, I mean, there are a bunch of fatty acid enzymes in there. But we didn't, surprising us, maybe just because there wasn't enough coverage. Fatty acid metabolism itself didn't show up as being significantly, statistically different as AF state progressed. But individually, there are some members of those. Thank you. Appreciate it. Thanks. We have last speaker of the section, Dr. Kassab from University of Washington. So his talk title is Oxygen Consumption and Symptoms of Atrial Fibrillation. So good afternoon everyone. Thank you for the kind introduction. My name is Ahmed Kassar. I'm a research fellow at the University of Washington Medical Center And today we're going to have a small discussion about a really interesting topic, which is oxygen consumption and symptoms of atrial fibrillation. So I do not have any pertinent disclosures. So I'd like to start my discussion with a brief introduction of two concepts which complement each other. I would like to start with the right side, which is the classic view where we tie usually symptoms of atrial fibrillation to a mechanistic point of view where the loss of atrial contraction, which ultimately leads to a decrease in LV filling and the manifestation of symptoms due to a reduced cardiac output. Now at the University of Washington, we're looking at this aspect from a complementary way from another point of view, which is a metabolic point of view. So there has been multiple reports which tied increased oxygen extraction seen using cardiopulmonary exercise testing in patients with atrial fibrillation. And this is why we wanted to look at depth in regards to this metabolic mismatch. So this is a small diagram, which I will not really dwell onto, which shows that the decreased cardiac output feeds into the symptoms of atrial fibrillation and this sheds light on the mechanistic approach when we talk about symptoms and AF. So some of the symptoms that are reported are palpitations, dyspnea, fatigue, exercise intolerance, and ultimately this all lead to a decrease in the functional status of the patients. Now the symptoms did not show up overnight. These have been reported in patients with atrial fibrillation ever since the 1999 by a paper by Levi et al, which was published in Circulation. And even after separating the patients into the different groups of atrial fibrillation, paroxysmal persistent, we noticed in each and every group palpitations, dyspnea, and fatigue were the most being reported. So what did you want to do at the University of Washington? We had three main aims. First of all, we wanted to look at the difference in myocardial oxygen extraction based on the underlying rhythm. So whether it was in an AFib rhythm or in a sinus rhythm. We also wanted to look at the components that affect this myocardial oxygen extraction. And in the final analysis, we wanted to look at AF related symptoms and try to link that to the oxygen extraction across the myocardium. And we came up with two hypotheses that we would like to address. The first hypothesis was that reduced cardiac output and decreased myocardial blood flow leads to a higher myocardial oxygen extraction. And patient reported symptoms are associated with higher oxygen extraction versus those with no reported symptoms. And to answer these questions, we relied on cardiac magnetic resonance imaging and specifically phase contrast in order to be able to contour the coronary sinus across the entire cardiac cycle. And this allowed us to generate a biphasic flow curve, coronary sinus flow curve, which is shown below. Now, we didn't leave out the mechanistic approach. We wanted also to look at cardiac power. And in order to derive cardiac power, it was the product of mean arterial pressure by cardiac output divided by 450. Now, as for myocardial mass, so we contoured. We also resorted to cardiac magnetic resonance imaging, two chamber views and four chamber view in order to derive myocardial mass by contouring the endocardium and epicardium in green and red of the left ventricle in order to be able to derive the myocardial mass. And this allowed us to adjust cardiac power along with coronary sinus blood flow to myocardial mass. Now, to go a little bit back to the cardiac power calculation, there are a lot of things that feed into this equation. So first of all, we would like to highlight systolic blood pressure and diastolic blood pressure, which feed into mean arterial pressure. And we cannot forget that cardiac output is the product of heart rate and stroke volume. Now, how did we calculate cardiac output? So we considered this as a closed circuit and we relied on Fick's law in order to derive cardiac output. So this was a sample of 45 patients undergoing an AF ablation at the University of Washington Medical Center. We standardized the blood draws before even the ablation starts. We had left atrial, which is arterial blood and coronary sinus venous oxygen blood. And we noted down the saturation of blood in all samples. We also had pre-procedural hemoglobin levels. We went then thereafter and calculated the arterial venous oxygen difference. And this allowed us to calculate oxygen extraction in milliliters of oxygen extracted per deciliter of blood. And I would like to share with you a couple of results that we have. So first of all, as I said, these were 45 patients. These are atrial fibrillation patients. We classified them based on the presenting rhythm through which we drew the blood. And we had 27 sinus rhythm patients, 18 atrial fibrillation patients. Most patients were in paroxysmal AF in both groups. And we noted also down modified European heart rhythm association scores, which touches base on how these symptoms affect the daily living of the patients. And you could notice that in the AF group, nearly 78% of the patients were belong to a Mehra score greater than one. Now as for the heart rate, so you could notice that patients in atrial fibrillation had a higher heart rate at 80.6 versus 64 in patients in sinus rhythm with a significant P value. And as for LA oxygen saturation, so there were no difference between the two groups. However, we noticed a significant drop in coronary sinus oxygen saturation in the AF group at 44% versus 50.4% with a significant P value. Now moving on to the main variable, which is myocardial oxygen extraction. We can obviously see that patients in an atrial fibrillation rhythm, they had more myocardial oxygen extraction at 10.8 milliliters of oxygen per deciliter of blood versus those in sinus rhythm at 8.9, which was also statistically significant. Now what about the two other valuable variables that we hypothesized that they affect this myocardial oxygen extraction? So you could see that cardiac power in patients in atrial fibrillation was lower when compared to those in sinus rhythm at 1.07 with a significant P value. As for coronary sinus blood flow per 100 grams of myocardial mass, you notice that there's a trend towards a decreased coronary sinus blood flow in the AF group when compared to the sinus rhythm group. However, it did not attain statistical significance. Now for the completion of our results, we ran a multivariable and univariable linear regression models. So model one is without the cardiac power and without coronary sinus blood flow. And what we noticed is that even after adjusting for cardiac power and coronary sinus net flow, you notice that you still have a beta of 3.93 on the multivariable regression, indicating that patients in an underlying atrial fibrillation rhythm had more myocardial oxygen extraction when compared to those in sinus rhythm. Now in the final analysis, we wanted to stratify these patients according to two levels. We did talk about initially that we had patients in an AF and an SR rhythm, and now we wanted to stratify them on another level and look at the symptoms of atrial fibrillation. So you could see that the grouping over here in orange, if the patients reported the symptoms, they were highlighted in orange. However, if they did not report the symptoms, they were highlighted in green. And what we could see here is that patients in the atrial fibrillation group who reported dyspnea, exercise intolerance, and palpitations had more myocardial oxygen extraction when compared to those who did not report any of these symptoms. So to discuss these findings a bit further, so faster heart rates, lower cardiac output, and we have the manifestations of the symptoms. Also with the loss of the atrial kick, on top of that having irregularity of the rhythm, you have symptoms of atrial fibrillation, which we have discussed. Now what we did is we tied these metabolic factors, the oxygen extraction by the heart, to these symptoms and showed that patients in an atrial fibrillation rhythm have more myocardial oxygen extraction. Is it really due to reduced coronary sinus flow, which leads to a more transit time across the coronary sinus and more oxygen extraction? Or is it due to a reduced cardiac power and more stagnation of blood within the ventricular myocardium and ultimately more myocardial oxygen extraction? So it could be a bit of both. But ultimately, we do know that we have metabolic inefficiency that's leading to fatigue, dyspnea, and reduced reserve even after controlling for heart rate. So are we saying that we should start targeting myocardial metabolism in atrial fibrillation management? Well, it's too early to say. And I would like to share with you this recent paper that we have published a couple of days ago in the American Heart Journal, and you could scan your QR code in order to access it where we dwell more on this topic. Now I'd like to touch base on a couple of limitations. So we had a relatively small sample size with an N of 45. And hopefully later in the future, we will be able to increase the sample size to integrate more clinically relevant factors such as AF burden as another important limitation, which is the patients were under general anesthesia and mechanical ventilation. So this might not reflect ambulatory physiology. Now for the future clinical implications, so our findings might explain why some AF patients feel worse than others. So in the future, could metabolic profiling guide therapy in AF management? And do therapies that reduce myocardial oxygen related symptoms do so beyond rhythm and rate control? I'd like to thank you all for your time. This is a wonderful picture of the team at the University of Washington who has put all of this work together. And I'd be more than happy to answer any of your questions. Yeah, I'll kick it off. Firstly, congratulations. Really original and interesting work. And it is a massive question, isn't it? Why some of our patients with AF are so debilitated and others seem to have a similar clinical phenotype and are completely asymptomatic. I guess I draw you out a little bit because when you looked at the, well I guess a couple of related questions. Firstly, these are all patients coming for ablation. So they would have a level of symptoms, you know, to have qualified for their ablation. They wouldn't be completely asymptomatic. Did you look, when you looked at it, you know, did you look at the entire group, those who were in sinus at the beginning of the procedure and those who were in AF? And were you able to take a more granular look maybe at their AF symptoms with a, you know, AFECT or AFSS and see whether there's a, you know, a broader correlation rather than a more binary analysis that you showed us? I realize the original, you know, and very difficult study to do. So, but yeah, great, great work. Thank you for your question. So we wanted to binarize it initially because we wanted, this was a sub-analysis of what we were planning to do, where we looked at the AF-related symptoms and we resorted to the most recent clinic visit in order to be able to derive the modified rapine heart rhythm association along with the binarized presence or absence of symptoms. But this is one of the aspects that we would like to highlight hopefully in future research. One last question. Thank you. Ilhan Gokhan from Yale University. Thank you for the really interesting talk. I wanted to ask about this extra oxygen consumption at the myocyte and the mitochondria level. Do you think some of it is getting shunted towards reactive oxygen species? And is that something that you could potentially measure? Because the role of ROS in the mitochondria might be contributing to AF. Thank you. Thank you for your question. We are not specifically looking at reactive oxygen species, but one analysis that we're doing in regards to arteriovenous gradients in patients with atrial fibrillation is that we're looking at lipidomics and metabolomics of atrial fibrillation and seeing how that ties to oxygen consumption later on. But we didn't specifically look at reactive oxygen species. Great. That concludes our session. Thank you everyone for joining us. Great discussions and great presentations. Thank you.

atrial fibrillation

cardiac magnetic resonance

atrial myopathy

UK Biobank

metabolic shifts

PET imaging

epicardial fat

obesity

myocardial metabolomics

oxygen consumption

personalized management