It was a huge pleasure to sit down with Steve Smith, a name synonymous with ECG expertise. Steve, renowned worldwide for his influential ECG blog, has been a pivotal figure in advancing our understanding of ECGs. Many of us have honed our ECG skills thanks to Steve’s insights. I had the opportunity to meet Steve about a decade ago at one of the SMACC conferences. Today, we delve into the fascinating world of occlusive myocardial infarction (OMI) and its comparison to STEMI (ST-elevation myocardial infarction), and explore the promising future of artificial intelligence in ECG interpretation.
Listening Time – 27:40
Exploring OMI: A Revolutionary Approach to ECG Interpretation with Steve Smith
A Legacy in Emergency Medicine and ECGs
Steve began his journey in emergency medicine 37 years ago, starting at Hennepin County Medical Centre in 1987, where he continues to work and teach residents and students. His interest in EKGs sparked in 1989 while working in the cardiac care unit. Faced with the critical decision of administering thrombolytics, Steve realized the importance of accurately interpreting EKGs to avoid false positives and potential intracranial bleeds. This led him to write “The ECG in Acute MI: An Evidence-Based Manual of Reperfusion Therapy” in 2002.
In 2008, Steve launched Dr. Smith’s ECG blog, one of the pioneering FOAMed sites, which remains free and highly regarded in the medical community. With over 1,700 posts, the blog covers a vast array of rhythm cases, but primarily focuses on OMI, which Steve considers the most crucial aspect of EKG interpretation.
Challenging the STEMI/NSTEMI Dichotomy
In 2014, during a talk at SMACC in Australia, Steve introduced the concept of the false STEMI/NSTEMI dichotomy. He argued that ST elevation is an inadequate marker for diagnosing acute coronary occlusion due to its lack of sensitivity and specificity. This led to the publication of the OMI Manifesto in 2018, advocating for the replacement of STEMI/NSTEMI with OMI/NOMI. Steve and his team have published numerous papers demonstrating that his approach to EKG interpretation is significantly more sensitive than traditional STEMI criteria.
The Data Behind Occlusive Myocardial Infarction (OMI)
Steve’s challenge to the STEMI paradigm is backed by substantial data. For instance, a study by Kojima in 2002 found that 55% of STEMI and 45% of NSTEMI patients had completely occluded arteries, both with similar mortality rates. A 2017 meta-analysis revealed that 25% of NSTEMI patients had a completely occluded artery on next-day angiogram, with these patients experiencing double the mortality compared to those with open arteries. Further studies have consistently shown higher mortality rates for patients with NSTEMI OMI compared to NSTEMI NOMI.
Identifying OMI: Beyond ST Elevation
For those familiar with the blog, identifying OMI involves looking beyond ST elevation. Subtle ST elevation, reciprocal ST depression, and other specific patterns can indicate acute coronary occlusion. Steve developed a four-variable formula, published on MDCalc, that uses the R wave amplitude in V4, QRS amplitude in V2, ST elevation in V3, and the QTc interval to diagnose LAD occlusion with high sensitivity and specificity. Although these rules are complex, they provide crucial insights into EKG interpretation.
The Role of AI in ECG Interpretation
Recognizing the potential of AI to revolutionize ECG interpretation, Steve collaborated with Powerful Medical, a company based in Bratislava, Slovakia. This partnership aimed to train AI to identify OMI as accurately as an expert. Powerful Medical’s Queen of Hearts AI, trained on thousands of EKGs, has shown remarkable accuracy in detecting OMI.
Validating AI: The Queen of Hearts
Several studies validate the efficacy of the Queen of Hearts. In one study, it demonstrated double the sensitivity of STEMI criteria. In another, it significantly reduced false positives in pre-hospital cath lab activations. At Ulst Belgium, the implementation of the Queen of Hearts reduced the median time to intervention for NSTEMI with occlusion from 16 hours to 1.4 hours.
Implementing AI in Emergency Departments
As an emergency physician in the UK, I see a significant advantage in using the Queen of Hearts for initial ECG interpretation. In a study of 51 cases of acute coronary occlusion misdiagnosed as normal by conventional algorithms, the Queen of Hearts correctly identified 38 as OMI. This demonstrates its potential to act as an initial filter, significantly reducing the burden on clinicians.
The Future of AI in ECG Machines
Integrating AI into ECG machines will require negotiation with major companies and likely FDA approval in the United States. However, once approved, the Queen of Hearts could be a game-changer in emergency departments worldwide.
Getting Started with the Queen of Hearts
For those interested in exploring this technology, the Queen of Hearts app, PM Cardio, is available on Android and Apple. The trial version allows for free use and provides a comprehensive analysis of EKGs, identifying occlusive myocardial infarction, bundle branch blocks, rhythm disturbances, and more. This tool can alleviate the stress on junior clinicians and improve patient outcomes by providing accurate and timely diagnoses.
Conclusion
The integration of AI in ECG interpretation, spearheaded by Steve Smith and Powerful Medical, marks a significant advancement in emergency medicine. By shifting from the outdated STEMI/NSTEMI paradigm to the more accurate OMI/NOMI approach, and leveraging the power of AI, we can improve diagnostic accuracy and patient care. Visit Dr. Smith’s ECG blog to learn more and try out the Queen of Hearts app to experience this revolutionary technology firsthand.
You can find out more about downloading the app here
Further Resources
OMI Literature Timeline: Key References (from Dr Smith’s ECG Blog)
2018
- Meyers & Smith: The OMI Manifesto
2019
- Meyers & Smith: “Prospective, real-world evidence showing the gap between STEMI and OMI,” International Journal of Cardiology.
2020
- Aslanger, Smith, et al.: “DIagnostic accuracy oF electrocardiogram for acute coronary OCClUsion resuLTing in myocardial infarction,” International Journal of Cardiology Heart & Vasculature.
- Meyers, Bracey, Smith, et al.: “Comparison of the STEMI vs. NSTEMI and OMI vs. NOMI paradigms of acute MI,” Journal of Emergency Medicine.
2021
- Tziakas, Chalikias, Al-Lamee, Kaski: “Total coronary occlusion in non-ST elevation myocardial infarction: Time to change our practice?” International Journal of Cardiology.
- Meyers, Bracey, Smith, et al.: “Accuracy of OMI ECG findings versus STEMI criteria for diagnosis of acute coronary occlusion myocardial infarction,” International Journal of Cardiology Heart & Vasculature.
- Aslanger, Meyers, Smith: “STEMI: A transitional fossil in MI classification?” Journal of Electrocardiography.
2022
- Xu, C et al.: “Point-of-care ultrasound may expedite diagnosis and revascularization of occult occlusive myocardial infarction,” AJEM.
- Lindow T et al.: “Low diagnostic yield of STEMI amplitude criteria in chest pain patients at the emergency department,” Scandinavian Cardiovascular Journal.
- Al-Zaiti S, Macleod R, Van Dam P, Smith SW, Birnbaum Y: “Emerging ECG Methods for Acute Coronary Syndrome Detection: Recommendations and Future Opportunities,” Journal of Electrocardiology.
- Kontos et al.: “2022 ACC Expert Consensus Decision Pathway on the Evaluation and Disposition of Acute Chest Pain in the Emergency Department,” JACC.
2023
- Sharma M et al.: “Occluded Coronary Artery among Non-ST Elevation Myocardial Infarction Patients in Department of Cardiology of a Tertiary Care Centre: A Descriptive Cross-sectional Study,” J Nepal Med Assoc.
- Zoni CR, Mukherjee D, Gulati M: “Proposed new classification for acute coronary syndrome: acute coronary syndrome requiring immediate reperfusion,” Catheterization and Cardiovascular Interventions.
- Meyers HP, Smith SW: “Hyperacute T-waves Can Be a Useful Sign of Occlusion Myocardial Infarction if Appropriately Defined,” Annals of Emergency Medicine.
- McLaren J, Meyers HP, Smith SW: “Kenichi Harumi Plenary Address at Annual Meeting of the International Society of Computers in Electrocardiology: ‘What Should ECG Deep Learning Focus on? The diagnosis of acute coronary occlusion!'” Journal of Electrocardiology.
- Spirito et al.: “Acute Coronary Occlusion in Patients With Non-ST-Segment Elevation Out-of-Hospital Cardiac Arrest,” J Am Coll Cardiol.
- Guener et al.: “Clinical implication of totally occluded infarct-related coronary artery in non-ST-segment elevation myocardial infarction: the TOTAL-NSTEMI study,” Coronary Artery Disease.
2024
- Abusharekh M, Kampf J, Dykun I, et al.: “Acute coronary occlusion with vs. without ST-elevation: impact on procedural outcomes and long-term all-cause mortality,” EHJ – Quality of Care and Clinical Outcomes.
- Smith SW, Meyers HP: “ST elevation is a poor surrogate for acute coronary occlusion. Let’s replace STEMI with Occlusion MI (OMI)!!” International Journal of Cardiology.
Podcast Transcription
It was a huge pleasure to be with Steve Smith, who many of you will know from his worldwide fame as an ECG blogger. I think most of us have taken most of our ECG knowledge from Steve over the years. I had the pleasure of meeting him about a decade ago, I think we were just saying Steve, at one of the SMACC conferences. And thank you so much for joining us. We’re going to cover a few things today. We’ll talk a bit about the blog and its background. I’d really like to hear more about occlusive myocardial infarction and what that is compared to STEMI (ST-elevation myocardial infarction). And most importantly, we’re going to talk about artificial intelligence. But first, Steve, why don’t you just introduce yourself for anybody who might not know who you are?
Sure. Well, I’ve been in emergency medicine for 37 years. I started at Hennepin County Medical Centre in 1987 and I still work there. I teach the residents and the students. I got interested in EKGs all the way back in 1989 when I was working in the cardiac care unit. We had to make the decision on whether to give thrombolytics or not. We had to make sure that the EKG was not a false positive especially because we didn’t want to cause intracranial bleeds. And then I started noticing all the subtle signs of occlusion on EKGs. So, I decided I was going to write a book on it, took me a while, but in 2002, I published “The ECG in Acute MI: An Evidence-Based Manual of Reperfusion Therapy.” In the meantime, publishing many peer-reviewed articles and in 2008, I started Dr. Smith’s ECG blog, which is one of the first FOAMed sites in the world. Actually, it’s still free. It always has been free. I’ve now got 1,700 posts on there of all kinds, a lot of rhythm cases, a lot of OMI cases and various other things. But mostly constantly on OMI because I find that to be the most important aspect of the EKG.
In 2014, I came to SMACC in Australia and talked there and I thought more about it during that and thought, you know, the STEMI/NSTEMI dichotomy is a false dichotomy. So, I started giving a talk called that, “The False STEMI/NSTEMI Dichotomy.” Later, I realized we just have to rename it. We have to come up with a new paradigm. Let’s call it the occlusion MI or non-occlusion MI because ST elevation is such a bad way of diagnosing acute coronary occlusion. It’s not sensitive, it’s not specific. And so, we published the OMI Manifesto in 2018 online on my blog. Then I started talking on, let’s replace STEMI/NSTEMI with OMI/NOMI. We’ve published many papers on this. I’ve got over about 70 peer-reviewed papers on ECG, and recently many, many, many on the OMI paradigm. We’ve shown, for instance, that my reading of an EKG is twice as sensitive as the STEMI criteria for diagnosing acute coronary occlusion.
The occlusive myocardial infarction. So, I think if you ask any cardiologist in the UK and most emergency physicians, they will describe ECGs as looking for ST elevation and sometimes they’ll say they’re looking for left bundle branch block and that’s kind of what they’re looking for. I think what you’ve challenged and said is that ST elevation is a very blunt tool for finding patient-trial occlusion and that’s what we’re interested in. The patient who needs to go to the Cath lab, who needs that occlusion and blocking and we need to try and find something different to do that. But are you getting a lot of resistance from people who just hold on to the STEMI idea?
I get resistance from a lot of people. I get acceptance from a lot of people and everybody wants more data. They want a randomized trial. Well, no one realizes that there’s never been a randomized trial of ST elevation to thrombolysis or percutaneous coronary intervention versus not doing it. They’re holding OMI to a higher standard than they ever held STEMI to. There is a lot of data. For instance, there’s a study by Kojima published all the way back in 2002. He took STEMI and NSTEMI all to the Cath lab, emergently. What do we find? 55% of STEMI had TIMI zero flow. IE, completely occluded artery. 45% of NSTEMI had a completely occluded artery. And they both had 5% mortality, very similar outcomes. In 2017, a meta-analysis of all studies of NSTEMI that they could find. These were randomized trials of various kinds. They took 40,000 NSTEMI patients. And of course, as with all NSTEMI patients, they did not get their angiogram until the next day. 10,000 of those 40,000 or 25% had a completely occluded artery on next-day angiogram. Those patients, compared to the NSTEMI patients with an open artery, had double the mortality, much higher biomarkers, worse LV function. And that all in spite of the fact that those patients with NSTEMI OMI were on average 15 years younger and had fewer comorbidities. And then since that time, there have been more studies doing this, looking at the same thing. NSTEMI OMI versus NSTEMI NOMI. And the patients with NSTEMI OMI have far higher mortality, one year, five years. Aslanger and I published a study out of Turkey where they looked at 3,000 cases, 1,000 without myocardial infarction at all, 1,000 with NSTEMI, 1,000 with STEMI. We had cardiologists who were trained in OMI recognition on the EKG to decide on the NSTEMI patients, which ones were OMI and which ones were not OMI. 28% of the NSTEMI patients were diagnosed as OMI by the EKG blindly, not knowing the outcome. And those 28% — and of course, because they were NSTEMI, they did not get — this is retrospective, they did not get emergent cath lab activation. Those 28% whom you could identify by the EKG blindly had far higher mortality over 800 days than the patients who did not by EKG have OMI. So we’ve proven that looking at the EKG and finding OMI has much higher mortality than patients without it. We have not done a randomized trial of it. Would you like to be a patient with an occluded coronary artery and be randomized to not getting your artery open? I don’t think so. On one hand, you could say it’s not ethical to do so. On the other hand, you could say, well, that’s what’s happening right now, is people with acutely occluded arteries are not going to the cath lab until the next day because they don’t have the required amount of ST elevation. I mean, it’s absurd, really, to think of it, that is happening all over the place right now, and no one’s doing anything about it. But it’s perfectly accepted. If someone has an occluded artery, but they don’t meet ST elevation criteria, it’s perfectly acceptable to let them sit there with their chest pain and have their myocardium dying. For people who haven’t read the blog, ST elevations are relatively easy concepts to understand. You learn about the lead, you learn about what you’re looking for. What should people be looking for when they’re looking for an occlusive myocardial infarction, OMI? For one thing, subtle ST elevation can also be due to acute coronary occlusion. Now, of course, some subtle ST elevation is not, so it becomes difficult to tell which ones are and which ones are not. But if you have any reciprocal ST depression, for instance, if you have a little bit of ST elevation in lead III and an ST depression in aVL of any amount, in a patient who’s got acute chest pain, it is almost certainly an acute coronary occlusion. We will prove this. One of my residents, if you want to look it up, on my blog, published that paper years ago. Another situation is left anterior descending artery occlusion can have much less than the required amount of ST elevation. How do you tell the difference between normal ST elevation and LAD occlusion? Well, I derived and validated a four-variable formula, which uses the R wave amplitude in V4 and the QRS amplitude in V2 and ST elevation at 60 milliseconds after the J point in V3 and the QTc interval and comes up with a value which with a cutoff of 18.2 is extremely sensitive and specific. Now, these rules are very hard to use. I mean, well, that has been out there for 10 years. It’s on MDCalc, that rule, but it’s still barely ever used. Then we published a paper in the Journal of the American Heart Association showing that any ST depression of any amount in V1 to V4, if it’s greater than the ST depression in V5 V6, is 97% specific for acute coronary occlusion. Do people still do that? I don’t know, but that’s a posterior MI. If V2 and V3 have any ST depression, those leads are extremely unusual. And if it’s in the setting of acute chest pain, it is an acute coronary occlusion, a posterior MI, until proven otherwise. And so we’ve published many other rules of this kind, like the Smith modified Sgarbossa criteria for left bundle branch block. We derived and validated it. And the Smith modified Sgarbossa criteria for ventricular paced rhythm. So I’ve been doing this for years looking at all different ways of codifying what I can see on an EKG. But it’s still very hard. Even the modified Sgarbossa rules are hard to apply. It takes a certain amount of skill and knowledge that is hard to learn. So I’ve always thought that the way to approach this is AI has to do it. And we’ve managed to teach AI how to do it. So I think that’s the future. And really, that’s the present now too.
I think it becomes pretty obvious that I was trying to write down some of those things you were saying as you were saying them, and I can’t keep up. And I’ve looked at your blog loads of times. And I don’t remember half of this, but the idea that having a machine that we can tell to look for, it makes a lot of sense. I would suggest to everybody that if you haven’t seen Steve’s blog, go and have a look. Firstly, it will just astound you because you’ll see ECG’s where you think, well, that’s fine. And then you’ll read the description and the outcome and think, oh, I’ve got a bit to learn here. But until we get the AI into all of our departments, we have a responsibility to make sure we’re as good as we can be. But Steve, it does feel like this development into AI makes sense. AI is, feels like it’s taking over our world a bit. It can recognize all sorts of stuff. So it makes sense. It can recognize ECGs. What brought you to that point where you were able to take the step of, I’ve got an idea, this is how I’m going to make it happen. How did you go from, I’m really keen on ECGs. I think this could be an idea to, to where you are now with a working, working model. Yeah, well, for at least 20 years, EKGs are like faces to me. I look at, I look at it and I instantly see that it’s an acute coronary occlusion or it’s not an acute coronary occlusion. And I don’t have to measure anything on it. It’s like when I see you, Ian, I know you’re Ian. I don’t have to measure your nose or measure your eyes or measure your mouth to know you’re Ian. I just know you’re Ian. And I know that computers, AI, can recognize faces just as well as I can. So if it can recognize faces like I can, it ought to be able to recognize EKG faces like I do. So for at least 15 years, I’ve been hoping to come to this point. In 2014, I started working with an AI company in EKG called Cardiologs out of Paris and I kept telling them over and over, you’ve got to focus on diagnosing OMI. And they were just obsessed with diagnosing rhythms, especially atrial fibrillation and wearables. And this is what all the AI people in EKGs have been concentrated on for the last decade. It was hard to get them to listen to me that really the biggest problem in EKG is diagnosing OMI. I finally have gotten the point where people are recognizing this, especially Powerful Medical, which had me train their system. EKG is a face. AI can recognize faces. It should be able to recognize an EKG. So how did it go from, well, how did you find Powerful Medical, which is the company that I know which, I think you have some now financial involvement with, although I would imagine it’s not buying your house quite yet. How did you get from an emergency physician in Minnesota to being a person who’s then got this company involved, and because I think they’re based in Europe aren’t they? They’re based in Bratislava, Slovakia. The reason that they have developed a system that can digitize EKGs, that’s one of the essentials of having AI do this, is being able to take like all the EKGs from my blog, make a digital file out of them so that AI can work with that digital file and no one else has been able to digitize EKGs like that. The reason they are able to is in Slovakia, they don’t record EKGs in digital format. So from the beginning of the company, they had to figure out a way because their country was behind technologically, they had to figure out a way to leapfrog that and be able to digitize the paper EKGs that they have. So they could take a, they got it so they could take a photo of a paper EKG that had been crumpled up and turn it into a digital file that AI could analyse. So that was one essential part of it. They found out about me and they contacted me and asked me, “Can you teach our system?” We put, if we put thousands of EKGs on a platform, can you code them as OMI or not OMI and then teach our system that way? As I’ve been waiting for this call for a decade, yeah, I can’t wait to do it. And my research partner, Pendle Myers, we do everything together. He and I, then on the platform, we just coded EKGs for several months. And that was all it took to train the system and we’re working on version 2 now, which is going to be a lot more complicated. But as far as the financial considerations, you know, I just have to say, I’ve been working on trying to get, trying to save people’s lives from coronary occlusion for 30 years. My blog has been free. I’ve never needed the money. I’ve never wanted the money. This is not about money. This is about saving people’s lives. And I believe we’ve found a solution to save people’s lives and save them from getting heart failure. I’ve used the app and the program quite a few times actually. And it’s been, well, revelatory in some ways. So you’ve got the app. I think Powerful Medical had made an app where they were looking at heart rhythms because that kind of was what people wanted with their wearables and whatever. And then they’ve added this bit on, which you’ve called the Queen of Hearts. The Queen of Hearts is the bit that you and Pendle have taken and educated in occlusive myocardial infarction. So you have managed to teach this artificial intelligence to recognize ECGs in the way that you do, or at least hopefully in the way you do. Has that actually been born out in trials and evidence where you’ve compared the AI generated answers to your answers to people like me, you know, ordinary people? Yeah, I’ll tell you about four studies. The first study we did was on a database of known outcomes from Ulst-Belgium, thousands of patients. The outcome of OMI was TIMI zero or one flow or TIMI two to three flow with a culprit. It’s not a perfect outcome measure, but it’s the best we could do. And we found that the Queen of Hearts had double the sensitivity at a fixed specificity. A fixed specificity of 98%. The Queen of Hearts was 68% sensitive, and the STEMI criteria were 34% sensitive. Then we did some other studies. I, at my own hospital, we looked at pre-hospital cath lab activation to see how it helped decrease false positives. In a certain time period, there were 117 cath lab activations by our medics. Now our medics, they believe in OMI already because I’m there teaching them. And so out of those 117 cath lab activations, we looked at 48 that we could classify as actual true positive OMI and 69 were false positives, or not OMI. If we applied the Queen of Hearts to these EKGs, it detected all 48 OMI and decreased false positives from 69 to 29. It could really decrease false positives. Another study we did was in Ulst Belgium, they have already implemented the Queen of Hearts in their system. It’s approved in the European Union and in England. And so they just, every patient with chest pain, they apply the Queen of Hearts. And historically, non-STEMI, which have occlusions, have had a median time to intervention of 16 hours at Ulst. In our two and a half month implementation, there were 11 non-STEMI that had an occluded artery, and they were recognized immediately by the Queen of Hearts and had a median time to intervention of 1.4 hours. So from 16 hours to 1.4 hours by implementing this, a doctor doesn’t even have to look at the EKG. There were six false positives. The false positives, one was a myocarditis, those are notoriously difficult to differentiate from OMI. There was a Takotsubo, which is also notoriously difficult to differentiate from OMI. And then there were four other cases that were cancelled by the doctor because the patient did not have any symptom that would have been worrisome for OMI. So those were the six false positives, but it detected the 11 cases that otherwise would have had delayed angiography. We have dozens of ongoing studies right now all over the world, both retrospective and prospective, and so we’re coming up with more data all the time.
So as an emergency physician in the UK, and I think this will resound with lots of people wherever they’re practicing, I can see this for one big advantage, and I’d be interested to know whether you think the Queen of Hearts would work for this. We get the ECG thrust all the time, and you probably recognize that. Everybody it seems gets an ECG in the emergency department, even if you don’t have chest pain, just because you’re tachycardic, you get an ECG, you’ve got a headache, you’ve got an ECG, everybody gets an ECG because hey, it doesn’t do any harm. We can go into whether it does or not another day, but these are then taken by somebody and they are literally thrust in front of a doctor of varying skill and experience or clinician, and they said, okay, can you sign this ECG please? And that doctor has stopped from doing their work. They then have to look at something. They ask some questions, and the whole interaction seems to take minutes and gets in the way. Is there a place whereby we could, with a decent evidence base, put this into our department for every, for want of a better term, triage ECG that’s done? And that would be safe. Yes, let me tell you about another study that I’m analysing the data right now. It’s not published yet. I have a collection on my blog of 51 cases of acute coronary occlusion that were diagnosed by the conventional algorithm, whether it be Veritas, Marquette 12SL, GE, any of those algorithms, were diagnosed as completely normal, completely normal, and yet they manifested OMI. And I could see on all 51 of these EKGs that it manifested OMI. There’s three publications so far that conclude that a doctor does not need to be shown the EKG if the conventional algorithm says it is completely normal, like I’m saying. 51 cases where the conventional algorithm says it’s completely normal, but the EKG actually was manifesting an acute coronary occlusion. So I gave those 51 EKGs to the Queen of Hearts. She diagnosed 38 of the 51 as OMI. The other 13 she said were abnormal, but did not diagnose. She did not say they were OMI. So in every case she recognized that it was an abnormal EKG, and those are the ones that need to be shown to the doctor. She did not call a single case of OMI as normal, unlike the conventional algorithms, which diagnosed 51 of them as normal. We’ve been used to sort that idea of the ECG machine gives us a little print out at the top, which I’m guessing I have no knowledge about, but I’m guessing that the intelligence behind that is utterly different from what we’re talking about here. Those algorithms that are used by the machines are not the same high level AI you’re talking about and that you’ve developed. Can you see an environment whereby in my emergency department we get hold of the app, and every ECG is scanned by the app, we can keep it on our records. And if it says there is no occlusive myocardial infarction and no abnormality or it’s normal, you would be confident in your reading of the literature, to be able to say it will not miss bad outcomes and if there’s anything abnormal that will highlight it. So that could be our first filter in the emergency department.
Yes, I mean I can’t say it will never miss something, but I can say that it will do way better than almost any physician and way better than any conventional algorithm. I remember conventional algorithms are just taught there’s code if then, if then if this is that then this is that it doesn’t, it’s not a learning algorithm like AI is. I can’t guarantee it will never miss one or never over call one. There will always be difficult EKGs that can’t be diagnosed even by the Queen of Hearts, but I can say I would feel perfectly comfortable in my emergency department having the Queen of Hearts screen every EKG without a doctor looking at it. And I mean my experience of when we’ve had a few of those missed cases, I had one at an M&M meeting not long ago. It was being discussed and a lot of my group haven’t heard of the AI type idea for this particular thing with ECGs. I put it straight in and it told me exactly what the end of the M&M presentation was which was this person had a bad outcome and it was a really visceral demonstration of how powerful that can be. How far away Steve, do you think it is from this AI going into ECG machines itself? Will that ever happen?
It will happen. I don’t know when though. That takes a lot of negotiation with the big companies that make the ECG machines. And I hear about the business side of all this every now and then, but that’s not my role in the company. And so I don’t really know how those negotiations are going. I think that they probably won’t happen until we get FDA approval in the United States. I think that will happen within 12 months, probably earlier.
Just for our UK and European listeners, what does it mean if you were to get FDA approval?
If we get FDA approval, that means we could sell the device. It’s called a device even though it’s just an algorithm. We could sell the device in the United States with the FDA approving that it’s a useful and accurate device.
So is it not available in the US at the moment?
You can use it on a trial basis. Anyone can use it whenever a post-op that shows the Queen of Hearts in action. I put a link in there. In fact, today I just have a post-op today where you can click on the link, get the Queen of Hearts for free and use it all you want. And so there are thousands of people around the world using it right now, but they’re not and they’re using it for free. It’s like a beta version.
And so the next steps, I’m completely, I mean, I’m brought into this a while ago, Steve, and we’re trying to do some work in Southampton with you to try and see how we can, there’s, I mean, there’s so many ECGs that are done in the world. It’s not like we’re looking for the cases, is it? And I’m on board with the occlusive myocardial infarction. I can get on board with that really well actually. What do you think the next steps are for people who are listening now? What would you say they should do? Download the app and have a look, see how it works? What would you say would be the next steps?
Yeah, for individuals, yeah, download the app and use it. See how you like it. I’ll tell you one story. There’s a newsletter called the Expert Witness Newsletter, and he writes cases that were malpractice cases that went to court and had decisions and had awards, you know, jury awards and multi-million dollar awards and learned about the Queen of Hearts. And he thought he wanted to go back to a case that had been a malpractice case where a woman came with chest pain, had an EKG recorded and she ended up dying, being sent home and died. And he thought, well, wonder what happens if I applied the Queen of Hearts to this EKG? And he did, and it was OMI with high confidence. Then he thought to himself, he said, you know, I had been following the OMI paradigm for a long time and I was sceptical of it because I didn’t think that physicians had the skill to read EKGs to that kind of accuracy about whether the patient has an occlusion or not. And then I heard about the Queen of Hearts AI and I thought, well, I got to test this out. So he tested it out on this case and it turned out to, it told OMI with high confidence. Then he thought, well, I wonder if it’s calling everything OMI with high confidence. So he used it on a bunch of other cases and it was giving accurate outcome of not OMI when it was not OMI. And so he became a believer in this and that’s when he sent out the expert witness newsletter last week with this case re-analysed. One’s personal experience with it is really powerful because you can use it on any EKG you want and see how it works. Now, you got to remember that critical to any test is the pre-test probability. So I would not advise using it on people with toe pain. All right. If someone’s got, you got to be thinking this patient, I’m worried about this patient having an acute coronary occlusion. You’re worried about that clinically. Then you apply the Queen of Hearts and if it says OMI with high confidence, it’s very accurate. It says not OMI with high confidence. It’s also very accurate. There are some false positives and false negatives still. So you still have to, you know, do serial EKGs and maybe echocardiogram. Sometimes you just have to do an angiogram if you’re still worried, but it is way more accurate than any human except maybe me and it is really good for your patients. So try it out. The second thing is anybody who wants to test it out formally and study it, we have protocols that we’re helping people do clinical studies on it and we’re aggregating, we want to aggregate all that data and have a very big multi-centre study showing how well it works or how well it doesn’t work either way. Please get in touch if you want to do that sort of thing. We can get you started on if you’re really serious about doing research or you can get you started on such protocol.
So for listeners who are probably sitting there with their mouths open thinking this could mean that I no longer have ECGs thrust in front of my face. There is a way forward which is, if you’ve done an ECG in the emergency department, a decent idea of first pass, put it through the Queen of Hearts which the app is called PM Cardio I think Steve isn’t it? If you look at the store. PM Cardio, there’s an app in Android and Apple just type in PM Cardio and you can get the app. Now it is not to say it does. In Europe and England, if you get the app, you get five free uses and then it costs money. If you get this trial version, which you can get from the link on my website, I believe it is free for you to just use informally. A decent start would be you’ve got an ECG in the emergency department. Have a go with the app. It’s incredibly easy. All you do is you open the app up. It says add a new page and that opens your camera on your phone. You take a picture with your phone. It asks you for some very basic demographic details which I’m guessing, Steve, add a bit to the pre-test probability of age, sex and then you ask a few symptoms as well. Have they got chest pain? Is it a routine? That sort of stuff. And then literally in seconds it will then analyse the ECG for you and give you a probability of an occlusive myocardial infarction. But it will also look at other stuff as well, won’t it? So it’ll diagnose bundle branch blocks and rhythm disturbance and all those other bits and pieces as well. Then hypertrophy, ischemia. All of the stuff that you’re really wanting in that first pass. And for me as a senior emergency physician in the UK, if I could take that stress away from our more junior clinicians who are thrown into this environment where they’re not sure what to do when they’re given an ECG. Sometimes I’m not sure what to do. If we were able to do that I think it’s a good thing. So I think it’s a good thing. I would delight to have Steve on the podcast to explain more about it. If you do want to download it go to Steve’s blog site which is Dr Smith’s ECG blog. You can find that on Google. Whatever you do go and look at the blog site. And Steve I can’t thank you enough for coming along. I think it’s a really exciting time and hopefully in a year we’ll be able to report on the FDA approval and it will be getting even bigger.
Yes, thank you so much Iain for having me.
Where to listen
You can listen to our podcast in numerous ways, ensuring you never miss an episode no matter where you are or what device you’re using. For the traditionalists, Apple Podcasts and Google Podcasts offer easy access with seamless integration across all your Apple or Android devices. Spotify and Amazon Music are perfect for those who like to mix their tunes with their talks, providing a rich listening experience. If you prefer a more curated approach, platforms like Podchaser and TuneIn specialize in personalising content to your tastes. For those on the go, Overcast and Pocket Casts offer mobile-friendly features that enhance audio quality and manage playlists effortlessly. Lastly, don’t overlook YouTube for those who appreciate a visual element with their audio content. Choose any of these platforms and enjoy our podcast in a way that suits you best!