New digital tools are seen as a way to screen for conditions such as heart disease before they become an emergency. But the accuracy of wearables and other devices often remain an issue, particularly with more complex health problems.
One potential solution, however, involves training artificial intelligence tools on “noisy” electrocardiograms that simulate the kind of messy data readings a wearable would generate, a new study shows.
The investigation, conducted by a team of researchers at Yale, suggests that by mimicking the challenges of wearable-derived data right away, such as poor contact with the skin, AI can better detect heart problems.
A “noise adapted” AI model designed to detect heart failure performed much better than a standard version, STAT reported this week.
Heart disease overall is the leading cause of death among older adults.
Currently, the research team is putting together a follow-up study with actual patients. Future goals are to clarify how the AI accounts for different demographics and whether the noise-tested AI can be “device agnostic” or only works with specific wearables.
The ability for wearables to accurately detect heart disease would particularly benefit people who live in “low resource settings” with “limited access to hospital grade equipment,” the researchers note.
The main disease the study tested for, left ventricular systolic dysfunction, is a common concern among older adults. But although the condition is underdiagnosed, LVSD could be detected early in roughly 90% of patients, previous studies have shown.
Wearable heart monitors are increasingly popular, with tools such as portable EKGs, smartwatches and patches in development or on the market.
Beyond bulkier wearables, one company, iRhythm, has developed a wearable patch that could monitor heart conditions, one of several examples of miniaturized EKG tech, the McKnight’s Tech Daily reported.
