Decoding the Heart

Zeeshan Syed, assistant professor in the University of Michigan Department of Electrical Engineering and Computer Science, has led the discovery of subtle but potentially life-saving signals hidden in heart attack patients' EKG histories. The findings could save thousands of lives every year.

Transcript

00:00:00 I had a heart attack on that July thirty-first up in traverse city today I'm going to have a double right and left heart catheterization to see if there's any things in my heart that they can fix heart attacks myocardial infarctions are the leading cause of death in the United States and now world wide hips and shoulders there are over a million people who are admitted with a

00:00:32 heart attack a year very good Robert drink girl is one of the lucky ones he said to heart attacks in recent years and doctors have treated him aggressively he's still alive today now you're getting age groups up to twenty-five percent of heart attack survivors end up dying from complications with any year right dr. burr oak dr hi another heart attack

00:00:55 inspired a University of Michigan computer science professor to change those statistics I guess when I was an undergrad I basically just wanted to get done with my academic career as quickly as I could I raced through my undergraduate masters degrees finishing four years headed out to the west coast took up a job there and thought that would be that of that but a year into my

00:01:14 job my father had a heart attack it was a heart attack they could end up saving thousands of lives he had a silent heart attack that went undetected for a couple of days and I got very interested in this question that even in this day and age we still great miss patients who would have heart attacks Dixon couple of days to come to the hospital there must be something that we can do about people

00:01:33 like this so I went back to MIT I enrolled in the HSV program between MIT engineering and Howard medical school and I started to explore this question can we try and extract information from the kinds of data that we collect routinely in clinical practice and somehow apply sophisticated computation to movies unsophisticated but easily accessible kinds of data and extract new

00:01:54 information that might help us identify high-risk patients at early stages when they might be treated in a more cost-effective manner more successfully so yeah and his team took a closer look at the electrocardiogram the EKG or ECG it's one of the oldest tools in cardiology so the ecgs is a wave it's a somewhat period signal because the heart is a perfectly

00:02:15 periodic typically what people have done is for you notice little bumps and divots in the signal it's considered to be noise why metal noise leads might be moving something else might be going on and one of the things that we focused on is that these little imperfections about the signals those little fuzziness that you might see little small bumps they might actually tell us something useful

00:02:32 to find out what the signals were saying Syed and his colleagues partnered up with dr. Bent's karika they used complex computational techniques to analyze an entire days worth of EKG history for more than 4,500 heart attack patients at Brigham and Women's Hospital in Boston we will have algorithms that could actually data mine large volumes of data and find these patterns that had some

00:02:55 prognostic value their findings are groundbreaking we developed a method that studies whether there is consistent noise like changes that persist over long periods of time and what we found is that this noise like the ability that persists over long periods of time you alerts may startle them actually tells us how unstable the hardest because if the heart is trying to beat over and

00:03:15 over again and it's not being able to do this in a consistent manner then that dell's assets unstable there's something about the heart which is not repeatable where's hearts function should be where I think this will be a particular use is that we can use electrocardiographic data and better identify patients at risk and it's through identification of patients at higher risk that we can then

00:03:37 target therapy and hopefully reduce the risk of complications such as sudden cardiac one of the hardest over here the findings of our work showed that we can improve by almost fifty percent or fifty percent the debts that are found by cardiology target I think this there's almost a magical quality to it that we're trying to make something out of nothing it really is sophisticated

00:03:59 computational applied to unsophisticated data we're picking out these things that are routinely collected in a clinical setting and we try to extract something fundamentally new out of it something that can have real-world impact save lives affect lost at me keep track and that's almost lying around unexploited because of the wrong analytical tools you're almost discovering a treasure

00:04:19 right beneath your nose my name hi there hi how are you good everything well I just talked to dr. my feeling okay you