Study: Cardiogram's DeepHeart detects diabetes through smartwatches

By Laura Lovett
11:52 am

A new study from Cardiogram suggests that ordinary wearables like Fitbit and Apple watches can detect diabetes and other medical conditions when integrated with Cardiogram’s DeepHeart app. 

“Twenty-four percent of people with diabetes, and 88.4 percent with pre-diabetes, don't realize they have it,” Brandon Ballinger, CEO of Cardiogram, told MobiHealthNews in an email. “We don't want to turn people into patients, but by catching diabetes early, we can guide people to convenient treatments they can perform in their everyday lives, like diabetes prevention programs. Ultimately, this means healthier patients and cheaper healthcare for all of us — for example, Medicare recently certified $2,650 in cost savings for every person with pre-diabetes that enrolls in a diabetes prevention program.”

The study, which was funded by Cardiogram and conducted in partnership with the University of California San Francisco, found that the DeepHeart app was 85 percent accurate in distinguishing between people with and without diabetes. The app was also able to detect high blood pressure with 80 percent accuracy, and sleep apnea with 83 percent accuracy.

Cardiogram’s DeepHeart uses deep neural network technology and machine learning to analyze large sets of data. 

The study, which was presented at this week's AAAI Conference on Artificial Intelligence, had 14,011 participants from all over the world, and collected 200 million unlabeled sensor measurements. Participants completed a medical history, which included previous diagnoses and medications. They were also given a mobile app, which integrated with HealthKit and continuously stored and processed the participants' heart rate steps and daily activity, according to the study. 

Researchers then compared two semi-supervised training methods and found that both significantly better than baseline analyses. The authors claim that these methods also outperformed hand-engineered biomarkers for detecting health conditions that have been reported in past medical literature.

“We believe our work suggests a new approach to patient risk stratification based on cardiovascular risk scores derived from popular wearables such as Fitbit, Apple Watch, or Android Wear,” the authors of the study wrote. 

In November another study released by Cardiogram and the UC San Fransisco Health eHeart Study found that DeepHeart could accurately detect hypertension and sleep apnea. However, diabetes was not measured in that investigation. 

This is just one of a number of studies that is set to validate the product and, according to the company, pave the way for use as an intervention. 

“Phase 1 of Cardiogram was validation: proving that wearables, when combined with artificial intelligence, could accurately detect health conditions like diabetes, hypertension, sleep apnea, and atrial fibrillation,” Ballinger wrote. “To that end, over the last year or so, we've published five validation studies in academic medical or artificial intelligence conferences. Phase 2 of Cardiogram is intervention: actually helping our users become healthier in real-world scenarios. Over the next few months, you'll see us launch new features that guide you through the process of screening, confirmation, and referral to clinically-appropriate treatments for diabetes, pre-diabetes, and more.”

But the product isn’t just designed to help diabetes. Ballinger explains that it is set up with all heart related conditions in mind. 

“Since your heart is connected to your pancreas, blood vessels, brain, stomach, intestines, and more through the autonomic nervous system, we think heart rate is not only a reflection of your heart health, but also a powerful vantage point into the rest of your body,” said Ballinger. 


The latest news in digital health delivered daily to your inbox.

Thank you for subscribing!
Error! Something went wrong!