The team behind a new algorithm claims that it can predict someone’s chance of having a heart attack up to four hours before doctors see the early warning signs.
Heart attacks are hard to forecast, but doctors rely on a scorecard system called the Modified Early Warning Score which looks at indications like a patient’s heart rate, blood pressure, and body temperature to estimate their chances of being affected.
As New Scientist reports, researchers from Carnegie Mellon University in Pittsburgh led by Sriram Somanchi wanted to find out if a computer could do a better job of calculating the odds of a cardiac emergency.
So they studied data from 133,000 people who visited four Chicago hospitals from 2006- 2011. Then they chose 72 variables, ranging from heart rate to age to blood sugar levels, and taught an algorithm to predict which patients would go on to have a heart attack. Their computer was even able to guess correctly from data taken four hours before the attack; much sooner than the Modified Early Warning Score would alert a doctor to the problem.
Which is fantastic, but the problem is it’s only accurate about two-thirds of the time, and 20% of results are false positives. But it’s better to err on the side of caution, and Somanchi and his team are now using larger groupings of data to train the system to become increasingly precise.
They’ll present their work so far at this year’s Knowledge Discovery and Data Mining conference, which takes place in New York from 24-27 August.
Image via NEC Corporation of America’s Flickr.