Electrocardiogram tests used to train AI to predict risk of fatal heart disease
Gerain0812/Shutterstock
Artificial intelligence systems have proven they can save lives by alerting doctors to test patients whose heart test results show they are at high risk of death. In a randomized clinical trial of approximately 16,000 patients across two hospitals, AI reduced overall deaths in high-risk patients by 31%.
“This is actually quite unusual,” he says Eric Topol Researchers at the Scripps Research Institute Translational Institute in California were not involved in the study. “Any drug, [produce] Mortality is down 31%, and it's even rarer for non-drugs, but that's just monitoring people with AI. ”
Chin Lin Researchers at Taiwan's National Defense Medical Center first trained the AI on more than 450,000 electrocardiogram (ECG) tests that measure the electrical activity of the heart, as well as survival data from ECG subjects. The AI learned to generate a percentile score representing each patient's risk of death, and considered patients at at least the 95th percentile to be at high risk.
The researchers then tested the AI alert system with 39 doctors from two different hospitals. When a nurse uploads a new patient's EKG test to a computer server, an AI system analyzes the results and immediately alerts doctors if the patient is deemed to be at high risk.
In addition to reducing the risk of death from all causes, the AI system reduced deaths in high-risk patients from heart disease by more than 90%.
Lin and his colleagues speculate that AI predictions could help focus attention on groups most at risk. Doctors usually respond to alerts with more diagnostic tests or additional treatments.
The AI alert system is already in use at 14 military hospitals in Taiwan.
“This can be implemented in every hospital in the world, and it shouldn't be expensive,” says Topol. “If it's so effective, it should become the standard of care.”
topic:
Source: www.newscientist.com
