Artificial intellegence and wearable technologies: Mayo Clinic Radio



welcome back to Mayo Clinic radio I'm dr. Tom Shyam's and I'm Tracy McRae Tracy some five million people in the US have congestive heart failure their heart is just too weak to pump enough blood now that usually doesn't happen all of a sudden and there are ways to predict that it will or that it may happen but they're expensive and they're not always readily available but now by combining artificial intelligence with a simple EKG Mayo Clinic doctors have found a quick inexpensive way to do it a cheaper way to discover that there is a problem with the pump the pump being your arm coming up we'll hear from two members of the research team about the new technique and later on in this segment we'll talk with a woman from Scottsdale Arizona who discovered that her heart was an atrial fibrillation when she tried on her husband's new Apple watch a great story joining us in studio to tell us about combining AI with EKG to detect early heart failure is dr. Paul Friedman chair of the Midwest department of cardiovascular medicine and Mayo Clinic cardiologist dr. Peter Noseworthy gentlemen welcome both of you to the program thank you great to be here now before we start I want to know what that is hanging around your neck obviously can't see it but it looks like you've got a speaker on the end of your stethoscope well this is a digital stethoscope and it has two metal plates on either end of the listening piece so in addition to digitally recording the sounds that I would listen to it also records an ECG simultaneously so by putting the stethoscope on the chest we get ECG as well as heart sounds the beauty of that is they're digital so that when we can apply artificial intelligence to that so any practitioner of any specialty can benefit from these algorithms and essentially have an expert cardiologists in his or her pocket to say hey there may be a problem here let's check Tracy's heart so that we can better understand this this new technique if wannabe would first explain congestive heart failure and also what is meant by left ventricular dysfunction great well congestive heart failure is a weakening of the heart muscle and it typically results in shortness of breath or inability to exercise many people are very bothered by a congestive heart failure but a small portion of the population may have weakening of the heart muscle that they're not yet aware of and if we could pick that up early then we can intervene with medications or devices to prevent it from progressing and to try to prevent people from having some of the adverse effects that are associated with heart failure so that's our goal finding the people who may not even know yet that they have a weak heart muscle why is it hard to diagnose it would seem that it would be easy to figure out well it's easy when it's obvious but it's harder when when you're doing okay and we use typically use something called an echocardiogram similar to ultrasound that's used to look at a baby for instance and there you can look at the actual muscle of the heart in action and see how it's working but you have to have a reason to do that test and you have to have about an hour and it costs a fair amount of money and it's not a practical screening test at the level of the general population so you have come up with something new something different and tell us about that so what we did was we used an neural network to empower the ECG to make it a more powerful test so that the ubiquitous inexpensive widely available ten-second non-invasive ECG can screen to see whether or not a weak heart pump is present and that you may then benefit for an echocardiogram and the way it worked is as follows a neural network requires a big amount of data to be trained on and then once it's trained it can run on a smartphone or on a stethoscope or something small and at Mayo Clinic fortunately we have big amounts of digital data so we took roughly fifty thousand data points it is ECGs and people who also had echocardiogram and we would feed the ECG into the network and say this person's ejection fraction which is a measure of heart pump strength is 50% which is normal this one is 35 percent after you do that about 50 thousand times and then which you know for the computer made may take an hour to the network learns it learns from the data the data actually train the mathematical equations in this nonlinear way so now you have a network that you can give it any ECG and say is a weak heart pump president yes or no so much the way a child learns you hold up a fruit and say this is an apple they'll say ok after a time they learned it's firm has a stem hold up another one this is an orange pitted different color sort of the same concept but with 50,000 ECGs you can pick up a lot of patterns subtle things that humans may not appreciate it's hidden in plain sight and so let me tell you how it performed we measure how good a test is by the area under the curve a perfect test would be a 1 if it says you've got the condition you have it if it says you don't you don't if you flip a coin it's a 50-50 right so it's a point 5 for medical tests such as a pap smear it's about a point 7 a treadmill test is a point 8 5 this tests the ability of the AI algorithm reading an ECG to tell you feel weak heart pump is 0.93 a very powerful test and to give you a sense of just how powerful it is we then said maybe if we tell the computer if the person whose test it is is a man or a woman it'll work even better because we know that gender and age affect heart disease we tried it it made no difference and then we said maybe the computer already knows if you're a man or woman from the ECG and guess what it performed phenomenally area under the curve for determining gender is 0.97 in other words a computer and reading an ECG is better determining someone's gender then you or I would be just walking down the street looking at somebody statistically speaking a female heart or a male heart that's right interest poetic isn't it hold on what's the difference I mean what's the difference between a female heart in a meal heart in a medical sense Phyllis as opposed to a philosophical or poetic sense we don't know and I say we don't know because the tool is a black box it learns but what it's looking at is not known and that's true of all neural networks so what everybody who has an EKG now at the Mayo Clinic would get this additional artificial intelligence with it and would tell them whether or not their heart is beginning to weaken so Peter runs our ECG lab maybe you want to talk about how we're implementing it yeah we're just now starting to figure out how to apply this to practice and we are building the infrastructure to allow to run the algorithm on all the ECGs that we obtain in a given year we did $250,000 now no cardiologist or doctor or internist has ordered an ECG to look for a low ejection fraction just like they wouldn't order it to tell if their patient is a man or a woman so it requires a bit of a change in the way we approach this test which is not intended as a screen for a low ejection fraction but we're going to be rolling it out in a randomized trial essentially within our primary care practice starting to give the results to the doctors and see how they make sense of it whether they order follow-up echoes with the diagnostic yield is and hopefully we start to prove that we're able to detect a disease earlier and make a difference for some of our patients so low ejection fraction means that the heart is not pumping as well as it could correct and you it used to be that in order to figure that out you needed an echocardiogram which required an hour fairly expensive then you can do the same thing now by using artificial intelligence and combining that with an EKG right and of course we would want a confirmatory test if we ran this we took off dr. Friedmann stethoscope and handed it to you and it was a positive screen we wouldn't leave it at that we would want to take a good look at your heart with an echo and figure out exactly what's going on but as an initial first pass it's really exciting for us and why is it important to know that well the key point is that 7 million Americans have a weak heart pump and don't know about it meaning and the people where it's more overt where it's already progressed to symptoms they know but if you have this condition and don't know about it you have an increased risk of developing symptoms an increased risk of dying and there are more than 5 professional Society publications saying here are medications that we know lower the risk of getting symptoms strengthen your heart pump make you live longer so the goal is to prevent bad things from happening people by detecting them early so rather than the old paradigm I feel sick I go to the doctor tests are ordered we can do a screening test that can be run ultimately when it's done even at home you know from a smartphone based ECG a watch base ECG something else now we haven't extensively tested it that way but we are in the process of doing that but the goal is detect disease early stop it from being bad and and help people in the process who should have this des to everyone or anyone who's just at risk of congestive heart failure so I would say that any our goal is to vet it because that in an actual practice setting as Peter was mentioning it and then in that context we would envision rolling it out so that ultimately anyone getting a medical order 12-lead ECG would have the test as a starting point and in the future then we'd have to see what groups it may make sense in absolutely fantastic I think these guys are pretty smart I changed my mind I do want to have a new technique to detect and predict a problem with a heart that is now widely available and inexpensive our guests cardiologist dr. Paul Friedman and dr. Peter knows worth its time for a short break and when we come back we'll talk with a woman from Scottsdale Arizona who discovered that she had a heart arrhythmia when she tried on her husband's new Apple watch and ask our experts what wearable technologies can and cannot do welcome back to Mayo Clinic radio I'm dr. Tom chimes and I'm Traci McCray Traci when I was in Scottsdale recently I was talking with a friend and she is a patient at Mayo Clinic Arizona and she told me that recently her husband got a new Apple watch mm-hmm and she said I tried that on it and she did and as you may know the model has an EKG on it and she said well I'm gonna check that out and she did an EKG on herself and it said she had atrial fibrillation so she ended up going to the emergency room and she's on the phone with us very shown back thanks so much for joining us thank you doctor so nice to be with you tell us the story I got the first part right didn't I you you you sure did yes you did I I did where my husband's Apple watch he recently had gotten a series for and the reason was I I was feeling a little off I guess great medical term I know but I didn't have any pain I had no shortness of breath no flu symptoms but something was telling me something was just a little different with my body and and I did have a few heart palpitations in hindsight I I realized I had that but I attributed that to I exercise most days so I thought well I'm getting older and and it's probably exercising and that but when I did it was just a quick reprogramming of my husband's watch for I believe its height and weight so it knew my information and to my surprise that said afib and I'll be honest I really didn't know what a and then this coincided with my annual exam and getting blood work done and that showed that some of mine my numbers were off on my blood work and that led to a phone call with my Mayo PA and nurse and to my surprise they did say you need to get to the ER I attribute the series for Apple watch to bringing to my attention what my body was telling me and and also being able to communicate that I honestly with with Mayo I honestly don't think I would have mentioned anything about my heart or you know feeling off is really not to medical so I am very thankful that and I now have a series forever watch myself I have a feeling but I have I have told a story just because I I really wasn't educated in this and not feeling sick is was really the key for me I didn't think anything was really wrong other than just minor symptoms yeah if you show up in the emergency room saying I feel a little off you're gonna sit there awhile right well the first thing they did was an EKG so that that did say yep you you are I had the rapid heart rate I had all the signs so I'm I'm very thankful that it all played out as I did and obviously I'm a fan of the series for Apple watch well they ultimately figured out the cause of your peb relation didn't they they did and it all is attributed to my thyroid she always is hyperthyroidism which I had no idea I had either and part of that was when the blood when my blood work came back my lab results it it showed that so Mayo was very quick and saying okay the the afib plus my lab work they said you go to the a you go to the ER you may be in thyroid storm which I didn't realize either so it it all played out very well for me and I'm very thankful so they got the thyroid problem fixed and you're now in normal sinus rhythm yes I check that once a while well as well as my I guess is the bpm heart rate would there's really two features on that Apple watch that we check periodically yeah pretty interesting story so dr. Friedman dr. Noseworthy I think there are some six million people in this country who have atrial fibrillation why is it a concern and is it important that people know whether or not they have it exactly it's a common condition and it increases your risk of stroke which is the major issue that we're trying to prevent with atrial fibrillation so if we can diagnose atrial fibrillation we can treat it usually with a blood thinner and anybody who has any other risk factors for stroke and it's not uncommon that a first presentation of atrial fibrillation is at the time of a stroke or many times people have a stroke you have no idea there was no harbinger whatsoever so detecting atrial fibrillation early is key does everybody who you detect and that has atrial fibrillation that you know about do you put them on everybody on a blood thinner not necessarily everybody but it's a conversation you should have with your doctor it ends up being about 85% of people who have atrial fibrillation because atrial fibrillation itself tracks with other risk factors but there are people who are young and otherwise healthy who may have atrial fibrillation or like your caller who had it related to her thyroid she may not require a long long term anticoagulation once that I I said it always is the thyroid because it always seems like the thyroid is the last thing that people end up checking but how many people end up with atrial fibrillation because of a thyroid is that common it's one of the presenting symptoms of hyperthyroidism not a terribly common one but we always look for it if you have atrial fibrillation your doctor will check to definitely check your thyroid function at least once but Mary's story I think brings out a few points in Mary I want to thank you for sharing it with everyone because first of all while heart disease sometimes will knock you off your feet sometimes it can be subtle and you said you know I don't feel right that's not a medical term but in reality you know that that's your body was adjusting well but picking that up early allows treatment for the underlying disorder if one is present in your case the high thyroid and it looks for screening if there's another potential disorder and it can allow for the use of medication shorter long term to prevent stroke so I think it's it's an important point and I think as we go forward there'll be more and more of these opportunities to because it's far better to detect it take medication than something else the second point Mary made which was a little more subtle but worth mentioning is that there's two things that the watch and most sensors are giving her one is the heart rate the heart rate is measured by using a light on the back it's called a PPG and that can sometimes give you warnings but it can often also give false alarms and so a lot of people are calling the doctors that said my heart rate was 200 and it may or may not be but the newer devices that actually record an ECG that we can look at and review that's powerful because then we can it can be reviewed by a physician who can look at it and say oh yes this actually is atrial fibrillation in each treatment what's the future hold I mean it this is pretty an amazing breakthrough to be able to put something on your on your wrist and it can tell you what your heart rate is and what and measure your EKGs run at EKG but but what's next I mean this is we're just at the tip of the iceberg aren't we we are there a long list of conditions that I think will be amenable because our bodies are giving off these invisible signals all the time and the more we can pick them up I mean just think about this before someone has a heart attack they've had atherosclerosis or narrowing of the arteries for 10 to 20 years and so the more we can pick up the bodies invisible signals and intervene sooner the more we can prevent bad things from happening and we're just on the beginning cusp of this I think the science and the technology is going to lurch well ahead of the regulatory issues around these that will be actually require significant addressing because there are no state boundaries when it relates to health but there are in terms of practice licenses and things like that where the future is nothing less than exciting is it all right an EKG with artificial intelligence can now detect previously hidden heart disease and holds great promise for saving lives and improving health and we talked about wearable technology showing real promise in likely one day we'll be able to prevent disease before it strikes our thanks to Mary Sean back from Scottsdale Arizona and for sharing her story and de Mayo Clinic heart specialists dr. Paul Friedman dr. Peter Noseworthy thank you very much thank you thank you thank you

One Comment

  1. vikram kashyap said:

    Sir .. as much as I understand that if we will make a device with help of neural network…the device will work by itself on basis of feeding data..

    Its means that device can also tell patient that your heart vessels are narrow or you have blockages in your heart vessels..by data it could advice patient about medicme too ?

    May 23, 2019
    Reply

Leave a Reply

Your email address will not be published. Required fields are marked *