Apple Watch Detects Diabetes in Study

Screenshots from the Cardiogram app for iOSYou can use the Apple Watch to detect your heart rate during exercise. People have been relying on heart rate monitors in fitness trackers for years to help aid in cardio workouts. Keeping a sustained elevated heart rate can improve heart health, leading to a longer life and more energy. Naturally, people like to keep track of this valuable stat while exercising.

But the heart rate sensor on our wrists can do a whole lot more. Apple is using it to help detect heart rhythm abnormalities with their Heart Study app. People use it to track abnormally high or low heart rates throughout their day. Now a study has shown a new potential use for the heart rate sensor: detecting early signs of diabetes.

Apple Heart Study Screenshots

Where There’s a Will…

I previously reported that Apple was working on non-invasive means to measure blood glucose levels with the Apple Watch. The end result would be the holy grail for people with diabetes: a way to measure blood glucose in real time, without needing to draw blood. Frequent testing throughout the day can be a literal pain, and therefore, many people are interested in a solution.

There’s a Way

Apple hasn’t been the only one working on a better way to detect diabetes using the Apple Watch. Researchers used their Cardiogram app, which has already proven that the Apple Watch can detect atrial fibrillation with 97% accuracy, along with a deep neural network called DeepHeart to figure out whether or not the Apple Watch could detect diabetes. Together, their Apple Watch app and DeepHeart were able to correctly detect diabetes in people with an accuracy of 85%.

The researchers noted that normal deep learning techniques require a lot of data, which can be problematic when human lives are at risk. Therefore, they helped coax the neural net through training. This involves giving the machine learning algorithm some help before training it in differentiating the results. The algorithm was trained with both labeled and unlabeled samples, with further helped train it.

How This Helps

85% accuracy might not sound like very much. A 15% error rate is high for AI. Also, this research doesn’t get us closer to being able to directly detect glucose levels in patients with diabetes, which is still years away. How can this be helpful? Detection. Many people are reluctant to go to the doctor, myself included. It’s expensive, blows my monthly budget in one visit, and usually there’s nothing wrong. However, if that thing you’re wearing on your wrist every day is able to point out a high potential of a problem, you might be more willing to go to the doctor to get it checked out. Early detection of diabetes can help prevent other health issues. Being able to use your watch as an early warning tool for disease is something we can all benefit from.


Source: Mikey Campbell, AppleInsider

Leave a Reply

Your email address will not be published. Required fields are marked *