AI can predict death or heart attack better than humans, according to a new study presented at the International Conference on Nuclear Cardiology and Cardiac CT (ICNC) in Lisbon.
A machine learning algorithm was able to predict heart attack or death by identifying patterns correlating variables to both outcomes with more than 90% accuracy. By repeatedly analyzing 85 variables in 950 patients with known six-year outcomes, the algorithm learn how imaging data interacts and outpaced humans in predicting death or heart attack.
Current methods to make treatment decisions are based on the use of risk scores, but doctors aren’t able to take into account as many variables as machine learning, resulting in less accuracy in these scores.
“Our study shows that very high dimensional patterns are more useful than single dimensional patterns to predict outcomes in individuals and for that we need machine learning,” study author Luis Eduardo Juarez-Orozco, MD, PhD, of the Turku PET Centre, Finland, said in a statement.
The study looked at 950 patients with chest pains who underwent usual protocol by the Turku PET Centre in Finland to search for coronary artery disease. Fifty-eight pieces of data on presence of coronary plaque, vessel narrowing and calcification were found in a coronary computed tomography and angiography (CCTA). Patients with scans that suggested the disease underwent further testing, including a positron emission tomography, producing 17 blood flow variables, and researchers pulled out another 10 variables, such as sex, age, smoking and diabetes, from medical records.
Using just the 10 variables from the medical records produced modest results in predictive performance, with an area under the curve (AUC) of 0.65. When the PET scan data was added, AUC improved to 0.69, with the predictive performance increasing significantly with CCTA added to an AUC of 0.82 and more than 90% accuracy.
Over an average six-year follow-up with the patients, there were 24 heart attacks and 49 deaths.
“Doctors already collect a lot of information about patients – for example those with chest pain,” Juarez-Orozco said. “We found that machine learning can integrate these data and accurately predict individual risk. This should allow us to personalize treatment and ultimately lead to better outcomes for patients.”