Researchers have trained a computer algorithm to predict if a patient will have a heart attack.
Heart attacks can, in some cases, be prevented by treatment such as diet change or medication if the patient is made aware they are at risk. However, human error sometimes leads to early signs being missed.
To counter this, researchers have successfully integrated machine-learning into the process. The University of Nottingham in the UK have tested a computer algorithm which has higher accuracy than humans in cardiovascular risk prediction.
This could have wide-reaching impact. According to a report by the Heart and Stroke Foundation South Africa, premature deaths caused by heart and blood vessel diseases in people aged 35-64 years are expected to increase by 41% before 2030.
The algorithm study looked at data from nearly 400 000 patients, some of whom experienced a heart attack or stroke, in the UK, over a 10-year period (2005 - 2015).
The researchers say they based four machine-learning algorithms (random forest, logistic regression, gradient boosting machines, and neural networks) on the American College of Cardiology guidelines, which are used by doctors worldwide.
The guidelines assess patients and make a prediction based on their age, sex, race, total cholesterol, HDL cholesterol, systolic blood pressure, blood pressure lowering medication use, diabetes status, and smoking status.
First, the artificial intelligent (AI) algorithms were fed complete data from around 300 000 patients to look for patterns and create 'rules'.
The remaining records were then given to the algorithms. However, researchers held back information on whether the patients ended up having heart trouble in the next decade.
The results showed the algorithms correctly predicted risk 7.6% more often than doctors, with 1.6% less false positives. The researchers say this means 355 patients could have been told earlier they were at risk of a heart attack or stroke.
Stephen Weng, an epidemiologist at the University of Nottingham, told Science magazine that using computer algorithms allows researchers to explore counterintuitive associations in diagnosis.
For example, he says a lot of body fat could protect against heart disease in some cases.
The algorithms are not complete and did not include information on if the patient had diabetes. Weng says he hopes to include this and other factors in future tests.
The algorithm will not be incorporated into a real product yet, as it will still have to go through clinical validation, regulatory approval, and gain the trust of medical professionals.
There is currently a lot of interest in using computer science in medical diagnosis.
Last month, Google announced it was investigating how machine learning can be used to help with diagnosing breast cancer to ease pathologists' workload.
The tech giant used deep learning to help recognise patterns and images, to create an algorithm which goes through the pathology data faster than a human.
It found the algorithm was good at identifying potential cancer but also identified a lot of false positives.
Share