Machine learning helps researchers predict cardiovascular disease

deepmind nhs records legal dispute doctor working at office desk
stokkete/123RF
Researchers at Nottingham University have demonstrated that machine learning algorithms could be better at predicting cardiovascular risk than the medical models that are currently in place. Four algorithms were put through their paces during the study; random forest, logistic regression, gradient boosting, and neural networks.

A team of primary care researchers and computer scientists compared these algorithms with the standard guidelines for cardiovascular disease risk assessment offered by the American College of Cardiology. A data set comprising 378,256 patients from almost 700 medical practices in the United Kingdom was used to facilitate the investigation.

All four algorithms were found to improve overall prediction accuracy compared to established risk prediction methodology based on a metric known as the ‘Area Under the Receiver Operating Characteristic curve,’ according to a report from Phys.org. The level of improvement varied from 1.7 to 3.6 percent.

Neural networks was found to be the highest achieving algorithm in the study, correctly predicting 7.6 percent more patients who would eventually develop cardiovascular disease.

“Cardiovascular disease is the leading cause of illness and death worldwide,” said Dr. Stephen Weng, of Nottingham University’s National Institute for Health Research School for Primary Care Research. “Our study shows that artificial intelligence could significantly help in the fight against it by improving the number of patients accurately identified as being at high risk and allowing for early intervention by doctors to prevent serious events like cardiac arrest and stroke.”

Based on the results of the study, the team is confident that artificial intelligence and machine learning techniques have a key role to play in fine-tuning risk management strategies for individual patients.

The researchers say that there’s more to be learned about the potential predictive accuracy of machine learning techniques, with other large clinical datasets, different population groups, and different diseases all providing avenues to expand upon the study.

Editors' Recommendations