So, before we dive into the study details, I would like to take a step back and have a look at the broader picture. So, ischemic stroke remains one of the leading causes of morbidity and mortality worldwide and its burden is not shared equally. Minorities especially face higher mortality and experience stroke at younger ages and among those, American Indians face the highest stroke incidence in the U.S. These disparities highlight an urgent need to identify modifiable risk factors and improve prevention especially for those who need it most. So, for a long time, atrial fibrillation has been considered the main culprit in stroke risk prediction and it is the foundation of our most widely used risk tool, the CHA2D2-VA But evidence showing a temporal dissociation between AF episodes and stroke events complicated that picture and led to a shift in thinking. Maybe atrial fibrillation isn't the direct cause of stroke but rather a marker of an underlying atrial disease and that's where atrial myopathy comes in which is defined as any structural or functional atrial abnormality and it may turn out to be a better predictor of stroke compared to AF alone. So our research question builds on this evolving concept of atrial myopathy and we ask, in American Indians with no history of atrial fibrillation and no history of stroke, markers of atrial myopathy, whether markers of atrial myopathy both from imaging and from ECG are independently associated with ischemic stroke even after adjusting for traditional stroke risk factors and competing risks and we do so over 25 years of follow-up. The data for our analysis comes from the Strong Heart Study which is the largest population-based longitudinal cohort of American Indians and it enrolls participants from Oklahoma, Arizona and North and South Dakota and it investigates cardiovascular disease and our study enrolled patients from Phase II between 1993 to 1995 and we followed them up until 2019 and these are patients with us participants with no history of a fibro stroke and available ECG data. So, to assess stroke risk we consider two sets of exposures, the traditional risk factors and the markers of atrial myopathy which included left atrial diameter indexed to body surface area and from the ECG PR interval, interatrial block, P-wave dispersion, mean P-wave duration and an oval marker, P-wave coefficient of variation which is defined as the standard deviation of P-wave duration across all 12 leads divided by the mean and it is a measure that attempts to capture the variability of P-wave duration. Our primary outcome was definite incident stroke either fatal or non-fatal and stroke events were identified through annual morbidity and mortality surveillance, they were adjudicated by independent committee and classified and we defined ischemic stroke as or cardio-embolic events and all other thrombotic events excluding lacunar strokes. We used Cox proportional hazard model and fine grade competing risk analysis for our data and we handled missing data with multiple imputations. So our analysis included 2,000 participants with a mean age of 60 and just above 60% were female, the prevalence of hypertension and diabetes was above 40% and around 20% of participants had CKD. So after 25 years of follow-up, 148 patients experienced ischemic stroke and participants who developed stroke were more likely to have been smokers at any time in life, to have a history of cardiovascular disease and increased LAD index and systolic blood pressure. This is a cumulative incidence curve for ischemic stroke and death over our follow-up and it highlights that the stroke risk accumulates gradually and the non-stroke mortality is substantial with around 60% of our patients dying from other causes prior to stroke occurrence. Here we show the cumulative curves for stroke incidence stratified by some key risk factors and all these are accounting for the competing risk of non-stroke death and we saw significantly higher stroke incidence in participants with history of cardiovascular disease and smoking while hypertension showed a trend towards increased incidence without reaching significance and diabetes was not associated. Among atrial myopathy markers, the largest tertile of left atrial diameter index was associated with an increased incidence of stroke and here we see a beautiful dose-response relationship while the P-wave variability variable, we observed again a trend towards a higher increase in the highest tertile but the global T-test was not significant. So in a pooled across the imputations, univariate Cox regression, we saw that several traditional stroke risk factors remained associated with increased hazard of ischemic stroke and this included age, smoking, cardiovascular disease, hypertension, HbA1c and among the atrial myopathy markers, left atrial diameter index, interatrial block and P-wave variability were also significantly associated with stroke while the other markers were not predicted. So in a univariate Cox analysis, which are the red lines, the final model retained seven variables that were independent predictors of stroke and these were age, smoking, cardiovascular disease, HbA1c, systolic blood pressure and LAD index and P-wave variability but after we accounted for the competing risks using a fine gray model, which are the blue lines, we see that age, HbA1c, P-wave variability and cardiovascular disease lost their significance. However, smoking, LAD index and systolic blood pressure remained significant predictors and this means that these factors are specifically consistently associated with stroke incidence even after adjusting for the high competing risk of non-stroke death. And findings that support the connection of atrial myopathy and stroke have made anticoagulation an attractive idea but trials in post-cryptogenic stroke patients and even those focusing on patients with markers of atrial myopathy, they have failed to improve the outcomes. So we are thinking should this data stop us from further investigating this relationship and we think not. Instead we believe that this data suggests that there is a lack of a reliable and specific marker for atrial myopathy that help us better identify which patterns of atrial disease actually drive thrombosis. So in conclusion, our study strengthens the link between atrial myopathy and ischemic stroke. We show that dose-response relationship between LAD index and stroke and all this was confirmed in a high-risk population of American Indians with strong competing mortality and a follow-up study spanning 25 years adding real-world strength to these findings. And our data support the idea that atrial disease, even in the absence of atrial fibrillation, is a meaningful and modifiable risk factor but only if we can better define it and measure it appropriately. And future efforts should aim to identify reliable markers and personalized intervention because the failures of the recent trials, they show us that in atrial myopathy, one size does not fit all. And at this point, I would like to acknowledge that this work was done at Dr. Stavrakis Lab at the University of Oklahoma. I would like to thank all the members of the lab, the collaborators, and our funding and thank you all for your time. Any questions from the group, go ahead. So the question is, was atrial myopathy associated with cardioembolic strokes? Thank you for your question. It is very important that initially we aimed for that outcome, however, this is a really long study, 25 years, and the events that we could get were classified as ischemic. So that's why we decided to include all because when we further categorized them, due to the years, some were misclassified under the ischemic stroke umbrella. So we wanted an outcome that we would be sure of its robustness during the adjudication process. That's why. Thank you. All right, so our next presentation is First in Man Assessment of a Novel Atrial Fibrillation Activation Mapping System. Here is Ollie Jones of Imperial College of London. Thank you all for the opportunity to speak to you all today. My name is Oliver Jones. I'm a PHF-funded clinical research fellow undertaking a PhD at the National Heart and Lung Institute at Imperial College London under the supervision of Professor Prapa Kanagratnam. Today I present the First in Man Assessment of our Novel Atrial Fibrillation Activation Mapping System. So activation mapping has proven instrumental in our understanding and management of complex, complex tachycardias. It's also been applied to atrial fibrillation, often in the epicardial setting of epicardial surgical mapping studies, and it's demonstrated different types of activation waves. And whilst these have been productive, these approaches all remain incredibly labor-intensive, requiring manual annotation of activations and of isochrones as well. And that's because atrial fibrillation represents or presents a unique technical obstacle that makes activation mapping exceptionally challenging. First of all, most clinical activation mapping is predicated upon a stable cycle length, and that allows a fixed reference electrogram and a fixed window of interest throughout the entire mapping process. And secondly, standard peak detection algorithms used to detect annivation, such as voltage threshold, DVDT, or template matching, are vulnerable to failure because of the way that electrogram morphology changes in atrial fibrillation. But especially, rapidly varying activation patterns in AF mean that activation can't simply be represented with a single static isochronal image. So instead, it's necessary to select and visualize only the key data. And my predecessors in the group have spent many years addressing these challenges and published several methodological papers outlining our progress on this front. And this work has now been adapted into a commercial system, and that's what we're presenting today. This system is called Tau20 RetroMapping, and it's a 48-channel recording and stimulation system with an associated novel activation mapping system. The Tau20 RetroMapping system is an investigational medical device. It's an atrial fibrillation mapping system, and it's compatible with and it runs alongside existing commercial 3D navigation systems, and in fact, any system with a fixed electrode spacing catheter. You can see here on the bottom right there, or the bottom left, sorry, the junction box and the console next to it. And then the big image on the right is the software interface that we use in the cath lab. And you have the electrograms visible in the top right there of the software interface. You can see what we call the wave watcher, and below that, you can see an activation map. And I'll go on to these in more detail in the next slide. So here are some examples from our recent validation work of this system over the past year. And at the end of the presentation, you'll see, as in the picture, that the system's being used with an AFocus 2 catheter. That's a 20-electrode double-loop catheter. So first, we validated an atrial tachycardia. And in the top right of the user interface, you can see the wave watcher. That's the spiral of yellow dots. And the shape of that reflects the 20 electrodes of the AFocus 2 catheter. And when those light up, that confirms good signal quality, and that's a proxy for contact and healthy tissue, and all that's required for activation mapping with retro mapping. And then below this, you can see the activation map, and that shows activations on the face of the catheter itself. So again, it mirrors the shape of the AFocus catheter in this example, and it updates automatically. Red areas are early, and blue areas are late activations, just like conventional. So the activation map, that's generated immediately, and it updates in real time with each wave event as soon as the operator places the catheter on the atrial wall. So here we are. So here it is, activation mapping, and you can see the atrial tachycardia being mapped there using the system. Moving on to something a bit different, we validated with sinus rhythm with ectopy. And the first wave here is sinus rhythm, as you can see, and it's activating earliest at electrode one. So that shows up as red on the right of the activation map. So here you can see the sinus beat. And now next, we have following this an ectopic beat, and you can see the activation propagates in the opposite direction. So now you have red on the left of the activation map. And these are more ectopic beats coming through. And then after this, we're going to see yet another sinus beat, and you'll see the activation map reverts to red on the right, and finally, an ectopic beat again. Okay? So while that's relatively easy to follow for us all here, one, it's a simple pattern because you simply have two different directions of activation, and two, we've slowed it down fivefold. And what we found, actually, is when you run this at full speed and in more complex arrhythmias, it's not really easy at all for operators to follow that at full speed while they're doing everything else in the cath lab. So that's why we introduced retro mapping, and this is the key idea in the whole system. And the retro map, which you can now see overlaid where we did have the activation map in the same position, it displays a smooth, dynamic average of all organized activation patterns in the given area. So looking at this recording again, you can see the sinus beats on the right, and they've been dominating, but now you'll see the ectopic beats, as they accumulate, they start to emerge and compete with those sinus beats. And you can see both of those things being represented at the same time. And we have red on the right and red on the left representing simultaneously the two earliest points of activation for these competing patterns. Okay, so now actually returning back to that atrial tachycardia that we saw early, what you likely didn't realize is that wasn't a standard activation map, that was actually a retro map too. And you can tell it's a retro map because it's dynamically updating. You can see it's shifting around. And although it's relatively static, that's obviously because in atrial tachycardia, as you'd expect, most of the activation is originating from the same direction. So let's move on to something more complex. This is atrial fibrillation now, induced AF. And you can see when I play the activation map here, you can see there's a shifting pattern from the top right down to the bottom right. So electrode 4 to electrode 12. As the pattern of activation continuously shifts, you see the retro map updating live and following that activation around without any need for any window of interest or anything else like that. And if you look at the end here, the AF actually breaks into sinus rhythm. We captured this in a real case. And you'll see now immediately the retro map updates and shows the red at the bottom of the activation map of the retro map. So we sought to formally validate both the system and investigate further these uniform waves, which have been described in AF before. A uniform wave is essentially a transient organized activation within the AF milieu, where it propagates uniformly across the catheter and you get a straight wave front. And we studied the frequency and the orientation of these waves, and most especially with respect to PVI, because they've been proposed as pointing towards focal drivers of atrial fibrillation. And so to this end, we undertook an observational study of patients undergoing a clinically indicated ablation for persistent atrial fibrillation at Hammersmith Hospital in London. And for each patient, we made multiple 30-second recordings within the left atrium using an AFocus II catheter. And then we returned after PVI to exactly the same sites to make repeat recordings. We studied nine patients and the demographics were typical of our center. We made 60 recordings in total. That was 32 minutes of atrial fibrillation data. And that yielded just over 4,000 wave events identified by the Tau 20 retro mapping system. And now in the image here, you can see that manual validation confirmed reliable algorithmic detection of uniform waves. The specificity was 94.1%. And you can see those green highlights showing that this is what the operator sees in the cath lab. And you can see that they can review the waves one by one. And it's highly selective. Actually, it's highly specific for uniform waves to the point of emitting borderline waves, which we found to be the best way to set the system up for the highest fidelity and accuracy. So looking at these uniform waves in a little more detail, we found they occurred with statistically similar frequency before and after PVI. That is 1.1 and 1.4 waves per second respectively before and after PVI. And in just over half of the recordings, we saw a clear dominant uniform wave orientation. By clear dominant uniform wave orientation, what I mean is that all the uniform waves that we captured were all propagating in exactly the same direction. So this can be seen in the image here, which actually shows a very interesting example of what we saw when we saw a clear dominant uniform wave both before and after PVI in the same site. We saw that this changed direction. So here on the left, you see three electrograms with red timestamps from before PVI and a red arrow showing the dominant wave activation. And then you can see the green arrow shows that it changed direction after PVI. So to finish, I'm just going to show a very quick example of using the system to track down focal drivers outside the pulmonary veins. So you can see here it's shifting, but it's mostly centered at the bottom around electrode 4. So then because of that, we highlight that point on the map, and in fact we move the catheter to overlie that red point. And then we get this characteristic bullseye pattern, which shows that the waves are propagating in all directions. This fits our model for a focal driver atrial fibrillation. And then to conclude, AF represents a unique challenge to activation mapping, but it's also an opportunity to understand this complex and heterogeneous disease on an individual patient level. The Tau20 retro mapping system allows us to do this, and we actually have everything in place to start a randomized controlled trial to study these uniform waves and track them back and ablate non-PV drivers of AF. Thank you. An excellent presentation. Questions from the audience? So the question involves, you know, looking at the activation mapping AFib, where are we targeting? Are we targeting the center of the bullseye? So a little more, I think, clarification. Yeah, excellent question. So the red point shows the earliest activation. Now it can be that you track back that and it keeps shifting back, if it was a macro re-entrant loop, for example. But what we look for is when we see the red point pointing in different directions. And so as you travel further, you see it pointing back from where you came from, and then you know that both red points are pointing towards the same focus. So the question involves, after PVI, the vector change in direction, what does that mean as far as what PVI changed? So strictly what it means is that the dominant activation direction, which is to say the trend of uniform waves, is pointing in the opposite direction. It would imply that the driver from within the pulmonary veins is no longer visible from that point, and now all we're seeing is the opposite, the non-PV driver, which is now dominating. All right. In the interest of time, we'll need to go on to the next presentation. So the title of the next presentation is Augmented Wide Area Circumferential Catheter Ablation for Reduction of Atrial Fibrillation. Recurrence Randomized Controlled Trial of the AWARE RCT Extended Follow-up Results. Dr. A.J. Sharma of the University of Ottawa Heart Institute. So hello everyone. My name is A.J. Sharma. I'm an AP fellow from the Ottawa Heart Institute, and I'm grateful today to be able to present results of extended follow-up from the AWARE RCT trial focusing on patients enrolled from the Ottawa Heart Institute. I would like to thank my staff colleagues at the Ottawa Heart Institute, and particular thanks to Dr. Nair for his support through this project. So a bit of background, as we know, atrial fibrillation ablation strategies have progressed over the years with improved one procedure success rates with pulmonary vein isolation. However, atrial arrhythmia recurrence has remained an issue. Various studies have looked at improving durable PVI and adjunctive ablation on top of that to improve PV isolation, although long-term results are seldom reported. This is just a handful of studies over the past 10 years that have looked with various means of ablation technique, RF, cryo, and pulse field ablation, all of which demonstrated comparable one-year success rates with regards to freedom of AA recurrence, all using different monitoring tools. But, you know, again, probably 25 to 30 percent chance of AA recurrence. The original AWARE trial, which was published in JAMA Cardiology in 2023, it was a randomized trial with 10 centers in Canada, 398 patients enrolled with one-to-one randomization to traditional wide area circumferential ablation versus double sequential AF ablation. And these patients were enrolled from March 2015 to May 2017. This is a diagram from the paper which demonstrates the double sequential wide area circumferential ablation that was performed. You see voltage maps on the left and then just the FAM on the right. And important to note at this time of this trial, randomization took place after the first set of WACAs were performed on the left and right pulmonary veins, minimizing any chance of, you know, operator maybe thought that they may be doing a second ablation. So, you know, they were still very focused at the first go around. All these patients were done under general anesthetic with low tidal volumes. There was very meticulous ablation with standardized protocols with regards to AI, you know, on the anterior and posterior walls, as well as, you know, using pacing from the CS for catheter stability. The results of that trial demonstrated that there was no statistical difference between both arms with regards to AA recurrence after a year, which is something that we've seen alongside other trials, you know, such as STAR-AF trial as well. However, there was, you know, equivalent and good one-year success in regards of freedom of AA regardless. With regards to recurrence of AA, we know, you know, from papers such as these that often early recurrence is understood to be related to pulmonary vein reconnection. However, mechanisms for late recurrence are less well understood. There have been some, obviously, retrospective analyses looking at long-term outcomes of catheter ablation with single procedure for paroxysmal atrial fibrillation. The study on the top, you know, at three years demonstrated 62.3% freedom from AA recurrence. And that paper also looked at mechanisms of recurrence in three of the studies, which found that in nearly 97% of patients that underwent for peak catheter ablation, there was at least one pulmonary vein reconnected. So, obviously, that certainly is a problem. The paper at the bottom looked over a longer duration between 2003 and 2021. Obviously, different eras of catheter ablation tools, you know, contact force versus irrigated catheters. And it noted, again, you know, at five years, the recurrence was around 60%. Sorry, the freedom was 60% at five years. So, still 40% AA recurrence. And not surprisingly, when compared with between different ablation sort of tools, the contact force catheters had the best results. So, it speaks to importance of, you know, meticulous ablation from the get-go. So, what we looked at was the extended follow-up from patients enrolled from the Ottawa Heart Institute in their WEAR trial. Patients were compared from freedom of AA and also the need for repeat catheter ablation. Again, as long-term studies are seldom reported. And how we did this is we used our electronic medical records. At University of Ottawa, we use EPIC, as well as Connecting Ontario, which gives us information from other hospitals in the province that are signed up to it. And we looked at all clinic assessment halters, ECGs, hospital presentations, any electrophysiology reports that were available for 242 patients. This is just a quick slide. I appreciate a busy slide, but just to say that there was no clinical significance between both arms. There were 122 people from the standard arm, and 120 in the augmented arm. Now, the results of this analysis demonstrated, again, there was no statistical difference in the overall AA reoccurrence between both arms, and no statistical difference between durable PVI on repeat ablation, for those of which underwent repeat ablation. But what we did note was that overall long-term survival, without the need for repeat ablation, was statistically significantly lower in the augmented wide area circumferential arm. As part of the full manuscript, we're going to be looking at patterns of reconnection, using these types of tokata diagrams, just to identify the patients that underwent repeat ablation, and where, if so, pulmonary vein reconnection took place. This is an example of a diagram from that, for a patient who underwent augmented ablation, so the double WACAs, and we look to see where the reconnection was, and again, identify that in the tokata diagram, so you could see there was some reconnection there on the right side, and ablation performed in that area. So overall, we saw that there was low recurrence rates of atrial arrhythmia over the extended follow-up in both arms. The lack of statistical difference may reflect the meticulous ablation that was performed with strict protocols from the get-go, but clinically, the double WACA may allow for some meaningful improvement in patient symptoms, even in those having recurrence. As we know, 31 seconds of recurrence of AF for a patient, you know, technically may count as a failure in the trial, but doesn't necessarily reflect something that's clinically significant for the patient, and when offered, and as we do offer patients repeat ablation if there's reoccurrence, you know, they may decline, so if it feels it's not impacting their quality of life. Another factor is when you do double sequential WACAs, you're covering a larger area of electrical isolation, and that may address some of the extra pulmonary triggers, you know, that was sort of talked about in some of the previous presentations even today. Limitations, obviously any, you know, study has limitations. This was an observational study of a completed RCT from a single center from Canadian institutions, obviously there's caveats with that, and AA recurrence assessments were limited to what are available results. All the patients as part of the WARE trial underwent at least three 14-day holters as part of the WARE trial. Obviously that's not necessarily the same as an ILR, but certainly comparable with regards to picking up meaningful symptomatic reoccurrence. So in conclusion, we reported the extended follow-up phase of the WARE RCT for patients enrolled at the Ottawa Heart Institute over a mean of 6.3 years follow-up, and there was demonstrated statistically significant reduction in repeat catheter ablation. This may suggest a role for adjunctive ablation alongside PVI in patients undergoing first time catheter ablation for paroxysmal AF. Where next from here? So, as mentioned, despite augmented ablation strategies, results are limited by pulmonary vein reconnection. Depending on the monitoring strategy, as is even in our study here, it's reported and quoted anywhere between 25 to 43%. And data from our center and other studies show a lack of durable PVI, resulting in pulmonary vein connection remains the major reason for recurrent AF. Further ablation as an adjunct to PVI is obviously balanced with a risk of significant injury. As we know, esophageal fistulas, phrenic nerve injury. So, it raises the concept or the thought of, is there sort of a ceiling effect from pulmonary vein isolation alone? And how much of a role is there for extra-pulmonary triggers? And the literature suggests that there is a percentage of patients, somewhere in the realm of 10 to 15% of what I see sort of published, that there can be a role of extra-pulmonary PACs, as I talked about previously, AVNRT. For some patients, CTI flutter that degenerates into AF. So, in patients that have chronic PVI isolation on a repeat catheter ablation, the thought is to perhaps try to tackle some of these other possible triggers that may be present. So, at the Ottawa Heart Institute, we just recently got a RAB approval for the WARE-2 trial, where we will be randomizing one-to-one patients for PVI ablation alone, using pulse-fetal ablation versus a PVI-plus approach. Basically, using RF ablation to do the double WACAs. And then, looking for any extra-pulmonary triggers at the index procedure. And this will be using isoprenolin as part of our standardized protocol, a full EP study to see if there's any other triggers that are tackled, and then, again, looking at the freedom of AA recurrence and need for repeat catheter ablation. Thank you. So I'll pose a question then, you know, this is a very interesting, you know, sequential WAC approach. Tell me about, say, tell me about, were there any evidence of coronary stenosis, because you're doing two levels of band-like fibrosis, so comment on that for me. Thank you for that. So from the original AWARE trial, they looked at safety outcomes, I actually sort of have a slide on that, and actually there was no statistical difference between both arms with regards to significant adverse outcomes such as esophageal fistula or, you know, or esophageal ulceration. There was one patient, for example, in the augmented arm that had an esophageal ulceration, but that particular patient was immunocompromised. With regards to pulmonary vein stenosis, you know, we haven't seen any signal of significant increase in that in the augmented arm, probably because we're taking that line sort of further out, if anything, the second line is further out from the initial traditional location for the WACO. Okay, well thank you very much. Thank you. Okay, our next presentation is Artificial Intelligence Enhanced Electrocardiography Quantifies the Systemic Disease Associated with Atrial Fibrillation. This is Dr. Sal of the Imperial College in London. Thank you very much. Thank you. So, I'm from Imperial. My name's Aaron, one of the EP Fellows and Clinical Lecturers. I'm here to talk to you about AI-ECG for quantifying systemic disease associated with AF, and it leads on nicely from the first talk, which was considering part of the elements about this. So, as we all know, AF is a growing epidemic around the world, and it's growing in frequency drastically, but the mechanisms and pathways that are associated with the increased incidence of AF are somewhat complex. They include things like metabolic syndrome, obesity, and associated sleep disorders, breathing, coronary artery disease, hypertension, and just generally an ageing population, and all these pathways together lead to inflammation, vascular disease, increased atrial ectopy, atrial myopathy, and atrial fibrillation. Now, a concept that was alluded to in the first talk was this principle of AF as a risk factor versus as a risk marker, particularly in the concept of AF-related stroke. One argument is that AF is a risk factor. AF, as we know, results in reduced left atrial appendage function, which can predispose individuals to left atrial appendage thrombus, which can embolize causing cardiometabolic strokes. AF is also a risk marker. The systemic disease that I've described, there's a vascular disease associated with obesity, metabolic syndrome, diabetes, hypertension, leads to an atrial and vascular disease, so both an atrial myopathy and vascular disease syndrome. That can predispose to thrombus, which can cause thrombotic stroke, and AF-related cardiomebolic stroke. And in this context, AF is a lower part of the pathway and may be a marker of risk rather than necessarily causing risk of stroke. Now, the concept of atrial cardiomyopathy is an evolving one, and this recent statement from this consensus group described the concept in more detail and showed how we can identify this through 3D mapping and voltage during catheter ablation, through MRI, and also through histology. And also demonstrated that there is an advanced, there is a complex interaction between the systemic symptoms and syndromes that are associated with AF, and also the atrial cardiomyopathy element which is more specific to the atrium. Now artificial intelligence ECG has been used for a wide range of tasks, pioneered by the Mayo Clinic initially for the detection of AF from sinus rhythm ECGs, as demonstrated by these two papers and many, many more subsequently. We at Imperial have applied AI ECG for a wide range of tasks, including for the prediction of mortality using the discrete time survival loss function approach, using a training data set from the Beth Israel Deaconess Medical Center of about 200,000 patients. We then have moved on to applying this to predict a range of future diseases, including cardiovascular disease, heart failure, ventricular arrhythmia, complete heart block, and atrial fibrillation. And I want to focus today on the models relating to atrial fibrillation. We call this platform the AI ECG Risk Estimation Platform, or AIR. AIR-AF, in particular, can predict the risk of someone developing atrial fibrillation in the future. Two cohorts here are displayed, a US secondary care cohort on the left, and a volunteer cohort from the UK on the right, UK Biobank, and we found that a high AIR-AF score can indicate a greatly increased risk of development of atrial fibrillation in the high-risk group in red here, compared to the low-risk group in green, in both the US-based derivation cohort and the UK-based volunteer external validation cohort. Now one element that we're very keen on is to use AIR-ECG for mechanistic insights, to derive insights into both the AIR-ECG mechanisms, and also potentially mechanisms of disease. These papers here describe our approach, which includes genetics, pheno-wide association, and imaging-wide association studies. So for the purpose of this current study, I wanted to use these techniques to see if we could elucidate anything about AIR-AF and disease mechanisms using this approach. So firstly, to try and understand what is the AIR-ECG model looking at, we use this variational autoencoder-based approach, and identify that AIR-AF is looking at multiple components of the ECG, which includes both the P waves, as demonstrated by the middle panel, heart rate on the right panel, but also the QRS morphology and conduction system, which indicates that we're not just looking at the atrium, but also the ventricles and about the cardiovascular disease as a whole. I therefore hypothesized that AIR-AF, as a score, was quantifying the systemic disease element of atrial fibrillation, and that therefore this could be used as a marker, as alluded to in the first talk, that may be at least as good as, if not better than, current markers of the systemic disease. One element that supports this is the fact that we see, in an aggregate fashion, AIR-AF scores increase gradually over many years prior to a diagnosis of AF, before a diagnosis of AF is made, and then plateaus, indicating over time there is an increase in the vascular syndrome and atrial myopathy that relate to AF, before finally AF is diagnosed and develops, after which there is generally a plateau and stabilization of this. We also found that, regardless of whether or not someone had AF, AIR-AF scores were associated with the diseases that are known to be associated with AF. This example here shows the blue panel, patients without the disease, and the red panel, patients with the disease, here looking at stroke, and we find that patients with a history of stroke have a greatly increased AIR-AF score compared to those without. And this is true for the other diseases associated with AF, heart failure, hypertension, diabetes, obesity, CKD, sleep apnea, and dementia. We also found that AIR-AF is associated with atrial myopathy, measured on the left in terms of cardiac MRI indices, so left atrial ejection fractions and volumes, and on the right in terms of echo, different LA volumes, LA sizes, indicating that AIR-AF is a measure of atrial myopathy. AIR-AF also strongly correlates with diastolic function. We find that when looking at echocardiographic diastolic function scores, which are derived from the AHA parameters, that a higher degree of diastolic dysfunction is associated with a higher AIR-AF score. And finally, we found, using a genome-wide association study in the UK Biobank, that AIR-AF scores are associated with everything that we know to be associated with the vascular syndrome of atrial fibrillation, that is, genes associated with metabolic syndrome, heart failure, cardiomyopathies, and diastolic dysfunction. So what can this tell us about disease and disease risks and AF? All of this put together makes us think that AIR-AF is a mechanism of quantifying the systemic syndrome associated with AF. So these plots show the hazard ratios for the presence of the disease of atrial fibrillation and how that impacts on outcomes. We can see here that the presence of atrial fibrillation increases the risk of cardiovascular death by approximately fourfold and the risk of all-cause death by about twofold in the whole cohort. But there is a differential effect depending on the phenotype of the patient. So if someone has a low AIR-AF score, i.e. they have a relatively low vascular disease syndrome, they have a much greater impact from developing AF than if you have a high vascular disease AIR-AF score. So basically saying, if you are a relatively healthy person and you develop AF, your incremental risk from developing AF is very high, whereas if you already have a very advanced disease phenotype in terms of vascular myopathy, atrial myopathy, developing AF actually confers a relatively low incremental increase and may be even slightly protective in this example. We then expanded that to multiple other endpoints, so looking at stroke, diabetes, sleep apnea, CKD, hypertension, heart failure, these are all things that we know to be associated with AF. We found a very similar pattern in that those patients who have a low disease phenotype derive a much greater increase in risk when they develop AF compared to those with an advanced disease phenotype. So to summate this, we think that our findings are suggestive of a dual, of both hypotheses being true, that AF is both a risk marker and a risk factor for stroke and other AF-related diseases. In subjects that have a disease phenotype that is quite advanced, AF may be more of a risk marker, and the development of atrial fibrillation here may not be adding that much more compared to their existing risks. But in people who have low AF, younger patients, or with those that have a relatively early disease phenotype, the development of AF is predominantly a risk factor for stroke in the traditional left atrial appendage-related thrombus. So in conclusion, air AF can identify subjects at risk of AF from sinus rhythm ECGs. A high air AF score may identify subjects with an advanced systemic disease phenotype. This potentially could be used as a marker for future clinical trials for score-derived anticoagulation, ablation strategies, or rhythm control strategies. And we identified that AF may be both a risk factor and a risk marker that is dependent on patient phenotype. Thank you very much for the whole team at Imperial College London and our funders. I'm happy to take any questions. Thank you. Questions from the group? So I guess I'll ask a question. So mechanistically, we're intervening downstream from all of your air AF metrics. So, you know, performing ablations, performing left atrial appendage occlusion, you know, I guess in the context of your scoring system, how should we interpret our interventions? And were there any study patients that had actually undergone occlusion? Thank you. So unfortunately, the numbers of people having appendage occlusion are quite small and not enough to make any meaningful statistical analyses. In terms of how our metrics might affect this, we did some early exploratory analysis looking at, for example, AF ablation, PVI. There's not enough to make any definitive conclusions, but there's a suggestion that intervention with PVI may stabilize progression, not reversing, but maybe stabilize progression in those that retain sinus rhythm. And that may be a signal for future research. Thank you very much. Thank you. All right. All right, so we'll move on to our final presentation, kind in urine pathway metabolites, predictive biomarkers for atrial fibrillation recurrence and metabolic benefits of pulsed field ablation. This is Dr. Li of Yunnan Affiliated Kunming Medical University. And we have Dr. Li here. It appears our final – well, he did not upload his slides either, actually, I was checking earlier. So it appears maybe our final presenter wasn't able to join us. So we'll give you all the gift of time. All right. All right, well, thank you all for coming, and congratulations again to the presenters.

atrial myopathy

ischemic stroke

American Indians

Strong Heart Study

atrial fibrillation

ECG indicators

left atrial diameter

stroke risk factors