Machine learning may have the potential to provide a cardiovascular disease prognosis that is just as accurate as a professional’s prediction. Adoption of this artificial intelligence (AI) technology could not only facilitate the management of cardiovascular disease patients but would significantly reduce costs for the NHS as well.
This research recently emerged from a group of scientists at Cardiff University who produced evidence that portrays how well machine learning can assess patients. Published in PLOS One, the study showed how diagnostic AI can match traditional methods of providing reliable prognoses for cardiac conditions. Requiring no professional training or human interaction, this method has the potential to greatly alleviate the healthcare provider’s workload.
The team used a technique known as genetic programming in this study, a method that was inspired by natural evolutionary processes. The technique involves training computer programs with a set of genes that undergo modification or evolution. Simply put, evolutionary algorithms such as genetic programming allow humans to discover solutions to problems we are not able to solve.
Genetic programming has utility over traditional AI algorithms created by researchers in that it reduces potential human error and bias, while simultaneously leaving room for environmental changes to be integrated into its system. This system also allows complex interactions revealed by AI to be easily seen by healthcare professionals, making the technology simple to integrate into their work.
The study used genetic programming to analyze risks of future cardiovascular events including fatality, and non-fatal strokes/myocardial infarctions. Over 3,800 patients aged 19-83 were involved in this study, conducted over a 10-year period.
25 predictors from patient data were used by the machine-learning algorithms, including sex, age, alcohol and tobacco use, BMI, and blood pressure. The researchers found that the algorithms were able to predict risks associated with individual patients at a level comparable to traditional methods.
Statisticians and clinicians have traditionally went about using numbers to estimate disease risks by developing their own arithmetic equations. The benefit of using an AI method like genetic programming in this instance is that it can reveal complex statistical associations in the data that would be challenging to see by other means.
“If we can refine these methods, they will allow us to determine much earlier those people who require preventative measures,” said study co-author Craig Currie, Professor at Cardiff University’s School of Medicine. “This will extend people’s lives and conserve NHS resources.”
“Although we already have reliable methods of forecasting people according to their degree of risk of serious heart events, artificial intelligence promises new ways of interrogating data and the likelihood of more reliable classification of risk,” Professor Currie concluded.
— Cardiff University (@cardiffuni) January 29, 2019