Researchers have created an artificial intelligence program that can predict when patients will die from a heart disorder. It is hoped that the software could be used by doctors to make informed decisions about how to treat patients before a condition deteriorates.
The software works by analyzing MRI scans taken of patients with pulmonary hypertension. It looks at over 30,000 points in the heart as it contracts on each heartbeat and then builds a virtual 3D heart of each patient. Coupled with eight years’ worth of health records, the artificial intelligence was able to assess what aspects predicted when a patient was likely to die up to five years into the future. The work has been published in the journal Radiology.
“The computer performs the analysis in seconds and simultaneously interprets data from imaging, blood tests and other investigations without any human intervention,” says co-author Dr Tim Dawes, who developed the algorithms that inform the program, in a statement. “It could help doctors to give the right treatments to the right patients, at the right time.”
The software analyzes over 30,000 points of the heart on every heartbeat to create 3D recreations. MRC London Institute of Medical Sciences LMS/YouTube
This is an improvement on current techniques, making faster and more accurate predictions. The newly created software was able to correctly predict which of 256 patients tested would be alive after one year with an 80 percent success rate, compared to 60 percent for doctors using the older method. This could therefore provide valuable insight into when to give patients certain treatments and medicines to better improve their chances.
The software was only trialed on those suffering from pulmonary hypertension. This is when there is high pressure in the blood vessels that lead to the lungs, which in turn puts unwanted strain on parts of the heart and can lead to organ failure. Without any treatment, patients can die within a couple of years, but if caught in time, drugs, injections, or even lung transplants can help.
It is hoped that eventually the software could be used on multiple different heart conditions, for example assessing those who might need a pacemaker. But for now, the researchers plan on testing the software at other hospitals in the UK to further gauge its accuracy and verify their findings.