An artificial intelligence algorithm developed at the Mayo Clinic could identify left ventricular dysfunction — or a weak heart pump — in most patients based on Apple Watch data, researchers shared at a conference Sunday. .
The proof-of-concept study was funded by the Mayo Clinic, based in Rochester, Minn., without technical or financial support from Apple.
Left ventricular dysfunction, which affects 2-3% of people worldwide, can be accompanied by symptoms such as shortness of breath, swelling in the legs or irregular heartbeat, but sometimes it has no symptoms, said the Dr. Paul Friedman, president of the Mayo Clinic. department of cardiovascular medicine in Rochester and a study investigator.
With improved screening for left ventricular dysfunction, providers could prescribe treatments to prevent patients from developing symptoms and reduce the likelihood of hospitalization and risk of death, he added. Diagnosis usually requires a visit to a healthcare facility for an imaging test, such as an echocardiogram, CT scan, or MRI.
The AI algorithm developed at the Mayo Clinic correctly identified 13 out of 16 patients who had a weak heart pump from Apple Watch data over a six-month study period, the researchers shared. during a presentation at the Heart Rhythm Society conference in San Francisco.
The Mayo Clinic study had two goals, according to Friedman. The first was to assess the accuracy of the AI algorithm, while determining whether researchers could successfully run a decentralized trial, in which patients were invited to participate via email and engaged with researchers via tools digital, with no in-person component in Rochester. .
The researchers recruited more than 2,400 Mayo Clinic patients from 46 states and 11 countries, each of whom already had an Apple Watch. The algorithm analyzed data from the Apple Watch’s EKG app, a feature that gained Food and Drug Administration clearance to detect atrial fibrillation in 2018.
While Apple was the first major consumer apparel company to add an FDA-approved ECG to its smartwatch, competitors like Samsung and Withings have since followed suit.
The researchers chose to use the Apple Watch for this study because Apple has app development tools that allow users to share ECG data with researchers.
Study participants downloaded a smartphone app the researchers developed with the Mayo Clinic’s Center for Digital Health, which uploaded ECGs from the Apple Watch to a data platform at the clinic.
It was up to the patients to record an ECG – data was not automatically recorded in the background – and the app prompted participants to share ECG readings every two weeks.
Of the 421 participants who had an echocardiogram on file for comparison, 16, or 3.8%, had left ventricular dysfunction. Thirteen of the 16 patients were also identified by the AI algorithm.
The sensitivity of the algorithm, or the probability of accurately identifying a positive result, was 81.2% and the specificity, or the probability of accurately identifying a negative result, was 81.3%.
The study summary released on Sunday is just a first test to show proof of concept, Friedman said. But it does suggest that “that consumer watch you’re buying has the potential to diagnose a potentially asymptomatic and life-threatening disease.”
To create the AI tool, the researchers adapted an algorithm that had already been developed at the Mayo Clinic, which detects a weak heart pump from a standard clinical ECG, which uses 12 electrodes. The Apple Watch ECG function has only one lead, so it provides less data.
The 12-lead algorithm received FDA Breakthrough Device Designation in 2019 and is licensed to Anumana, a Mayo Clinic company and software startup Nference launched last year to develop and commercialize AI algorithms for early detection of diseases. The area under the curve – a common measure of accuracy, with the most accurate tests being close to 1 – for the 12-lead algorithm was 0.93; the algorithm designed for the Apple Watch was 0.88.
The single-lead algorithm could one day be used to more easily screen cardiac patients for a weak heart pump outside of a clinical setting, the researchers wrote in the study’s abstract. They plan to continue testing the algorithm in different countries to assess whether it is accurate in all populations, what its limitations are, and how it might be used in clinical practice.
