Heart attacks can be hard to anticipate and those who experience it for the first time are rather clueless about its symptoms. Now, scientists have devised a mechanism where artificial intelligence (AI) can teach itself to predict heart attacks even better than doctors.
While doctors are known to follow certain risk factors like age, cholesterol level, blood pressure and more to determine heart attacks, the new machine learning mechanism teaches itself to pick up other disease and lifestyle factors that may contribute to the attack. This is particularly important as the average age of those suffering heart attacks has come down in the past few decades owing to significant lifestyle changes.
An estimated 20 million people die every year from the effects of cardiovascular disease that include heart attacks, strokes, blocked arteries, and other circulatory system malfunctions. To predict these doctors use guidelines given by the American College of Cardiology/American Heart Association (ACC/AHA) that are based on eight major risk factors.
But these factors alone are too simplistic to predict heart attacks accurately.
"There's a lot of interaction in biological systems," says Stephen Weng, an epidemiologist at the University of Nottingham and member of the research team. "What computer science allows us to do is to explore those associations," he says referring to the new AI algorithm.
Weng and his team compared the use of the regular ACC/AHA guidelines with four machine-learning algorithms — random forest, logistic regression, gradient boosting, and neural networks —all of which have picked up the various risk factors involved in heart diseases. The results of the comparison showed, all four AI techniques analysed patient data well and that too without any human instruction.
Many of the risk factors the AI identified as strong predictors were not even included in the ACC/AHA guidelines, like mental illness and taking oral steroids. On the other hand, the AI failed to consider diabetes as a risk factor, which is high on the ACC/AHA list.
Weng says it will be easier to include more medical, lifestyle and genetic factors in the algorithms to further improve its accuracy. If implemented successfully the new method could save thousands of lives by alerting patients to get to hospital on time.