Medical

AI algorithm detects asymptomatic heart disease from Apple Watch ECG data

AI algorithm detects asymptomatic heart disease from Apple Watch ECG data
A research team developed an algorithm that can detect if a subject has left ventricular dysfunction (aka weak heart pump) using Apple Watch ECG data
A research team developed an algorithm that can detect if a subject has left ventricular dysfunction (aka weak heart pump) using Apple Watch ECG data
View 1 Image
A research team developed an algorithm that can detect if a subject has left ventricular dysfunction (aka weak heart pump) using Apple Watch ECG data
1/1
A research team developed an algorithm that can detect if a subject has left ventricular dysfunction (aka weak heart pump) using Apple Watch ECG data

A new algorithm developed by researchers at the Mayo Clinic can effectively detect patients with a dangerous heart dysfunction using data gathered by an Apple Watch. A large trial is now underway looking to test the clinical utility of the algorithm in one million people.

Although Apple introduced electrocardiogram (ECG) measurement capabilities into its smartwatch back in 2018, it’s only recently that the technology has begun to deliver meaningful health reports. While the ECG feature on an Apple Watch can currently detect a type of irregular heart rhythm called atrial fibrillation, it is still mostly recommended as a tool to help detect cardiac abnormalities in conjunction with a doctor.

A traditional ECG test involves up to 12 electrodes attached to various parts of the body. These electrodes allow doctors to record the heart’s electrical activity and subsequently detect a number of different cardiac abnormalities.

The Apple Watch ECG only gathers data from one point on the wearer’s wrist, so of course it can never be as accurate a diagnostic tool as a 12-lead ECG test in a clinic. To transform that Apple Watch ECG data into something clinically useful researchers must turn to AI algorithms, designed to identify tiny signals in the data that can correspond with heart problems.

A couple of years ago a team from the Mayo Clinic developed a novel algorithm that could automatically detect a heart condition known as left ventricular dysfunction using traditional 12-lead ECG data. The condition, informally referred to as a weak heart pump, is often asymptomatic and is estimated to afflict nearly seven million Americans.

"Left ventricular dysfunction – a weak heart pump – afflicts 2 percent to 3 percent of people globally and up to 9 percent of people over age 60,” explained Paul Friedman, a Mayo Clinic researcher working on the project. “It may have no symptoms, or be associated with shortness of breath, leg swelling or racing heart beats. What is important is that once we know a weak heart pump is present, there are many lifesaving and symptom-preventing treatments available.”

The new research, yet to be published in the peer-reviewed journal, was presented recently at the Heart Rhythm Society conference. Led by Itzhak Zachi Attia, the research team adapted the previously developed algorithm to effectively interpret the single-lead Apple Watch ECG data as if it were from a 12-lead machine.

A small preliminary six-month study looked at data from a number of people participating in an ongoing Mayo Clinic study. The participants shared a large volume of ECG data alongside Apple Watch data to test how effective the new algorithm was at detecting a weak heart pump.

"Approximately 420 patients had a watch ECG recorded within 30 days of a clinically ordered echocardiogram, or ultrasound of the heart, a standard test to measure pump strength,” said Attia. “We took advantage of those data to see whether we could identify a weak heart pump with AI analysis of the watch ECG. While our data are early, the test had an area under the curve of 0.88, meaning it is as good as or slightly better than a medical treadmill test.”

In early April the Mayo Clinic launched a massive study looking to test a range of algorithms designed to predict heart disease from Apple Watch ECG data. The study hopes to enroll one million subjects who will be followed for one year, supplying Apple Watch ECG data to an anonymous database. This data will be retrospectively compared with each participant’s medical records to assess the quality of the AI algorithm’s health predictions.

So at best, it is probably one or two years before an Apple Watch starts diagnosing people with this particular heart condition. Nevertheless, Friedman is optimistic this kind of AI technology will transform medicine in the future, as cheap health-tracking wearables will increasingly be able to catch serious diseases at their earliest stages without needing patients to come into hospital and perform expensive or onerous tests.

“It is absolutely remarkable that AI transforms a consumer watch ECG signal into a detector of this condition, which would normally require an expensive, sophisticated imaging test, such as an echocardiogram, CT scan or MRI," added Friedman. “This is what the transformation of medicine looks like: inexpensively diagnosing serious disease from your sofa.”

Source: Mayo Clinic

3 comments
3 comments
FB36
Smartwatches definitely need to have more health sensors, for measuring blood pressure/sugar/alcohol/THC & even for detecting pregnancy (early)!
BlueOak
… and that $300+ price for the Apple Watch Series 7 starts to look reasonable…
P51d007
I held off getting a smartwatch until mid 2018. I was just going to get one, play with it a few days, send it back. Seemed kind of useless.
But after the first day of not having to dig out my phone to see who was calling/texting/emailing, I kept it.
Noticed it had sleep tracking. So, started wearing it at night. Noticed my heart rate at night would sometimes jump into the 80-90 bpm.
Next time I was at a routine checkup with my doctor, mentioned it, showed them the chart of my night HR. Doctor suggested a sleep
study. Came back with a "mild to moderate" case of sleep apnea. First night I used the CPAP, slept through the night, which never happened
before.
Amazing "gadgets".