Skip to Content

With a Little AI, Apple Watch May Be Able to Spot a Heart Problem

The watch’s heart monitor is a long way from a medical-grade device, but smart software may be able to improve its capabilities.

The Apple Watch is not a medical device. Sure, it can keep an eye on large changes in your heart rate, making it reasonably good at telling you how long you worked out for. But anything more than that is a bit outside its capabilities.

Which is why it’s interesting that on Thursday, a study emerged suggesting that readings from an Apple Watch can indeed be used to detect atrial fibrillation—a worrisome heart condition that, if the sufferer isn’t hooked up to a heart monitor, often produces no noticeable symptoms.

The the trick behind the study, which was conducted by Greg Marcus and a team of researchers at the University of California, San Francisco, and a startup called Cardiogram, was that it used artificially intelligent software to tease information out of the Apple Watch data.

To do that, the researchers gathered data from 166 people who met several criteria: they wore an Apple Watch, were known to have atrial fibrillation, and also had an AliveCor device, an FDA-approved electrocardiogram (EKG) that hooks up to an iPhone (the company is also developing a watch band that can be worn with an Apple Watch, but that hasn’t yet received FDA approval).

Each of these users generated two streams of data simultaneously: one from the watch, and one from the EKG that was capable of accurately showing when someone’s heart was fibrillating. By training a machine-learning algorithm on the two streams, the team reasoned the algorithm could teach itself to identify what fibrillation looked like in the less precise Apple Watch data. A test on readings from about 50 people seemed to bear that out—the algorithm was 97 percent accurate in flagging fibrillation.

It’s an encouraging result, but far too small a sample size to declare success. The study also suffered from some limitations that will hamper real-world implementation. The subjects had to remain still while taking readings, for example, because Apple has said that some movements can result in inaccurate results.

Cardiogram, which has $2 million in funding from Andreessen Horowitz and others, according to BuzzFeed, is entering a field that’s been riddled with problems. Just ask the people behind Quanttus, a startup that got a hard lesson in how difficult it is to get accurate biological readings from wrist-worn devices.

It’s also a field that has competition. Alphabet’s health-tech company, Verily, is building what it calls a “Cardiac and Activity Monitor” that likely has an EKG on it, among other capabilities—part of a large-scale health study that Verily is conducting. When we spotted it last year, it appeared to be built specifically for tracking people’s health, rather than consumer use. It was plain-looking, and had an e-ink display that preserved battery life, so people could wear it for much longer stretches than Apple’s offering, which, after all, is really just a gadget.

(Read more: BuzzFeed News, “Health-Tracking Startup Fails to Deliver on Its Ambitions,” “The Apple Watch May Be About to Get a Medical-Grade Add-On,” “I Saw Alphabet’s Health Watch”)

Keep Reading

Most Popular

Russian servicemen take part in a military drills
Russian servicemen take part in a military drills

How a Russian cyberwar in Ukraine could ripple out globally

Soldiers and tanks may care about national borders. Cyber doesn't.

Death and Jeff Bezos
Death and Jeff Bezos

Meet Altos Labs, Silicon Valley’s latest wild bet on living forever

Funders of a deep-pocketed new "rejuvenation" startup are said to include Jeff Bezos and Yuri Milner.

conceptual illustration showing various women's faces being scanned
conceptual illustration showing various women's faces being scanned

A horrifying new AI app swaps women into porn videos with a click

Deepfake researchers have long feared the day this would arrive.

ai learning to multitask concept
ai learning to multitask concept

Meta’s new learning algorithm can teach AI to multi-task

The single technique for teaching neural networks multiple skills is a step towards general-purpose AI.

Stay connected

Illustration by Rose WongIllustration by Rose Wong

Get the latest updates from
MIT Technology Review

Discover special offers, top stories, upcoming events, and more.

Thank you for submitting your email!

Explore more newsletters

It looks like something went wrong.

We’re having trouble saving your preferences. Try refreshing this page and updating them one more time. If you continue to get this message, reach out to us at customer-service@technologyreview.com with a list of newsletters you’d like to receive.