How Apple Watch Now Detects Diabetes Through Artificial Intelligence
Through a study called deep learning, an app claims that it can detect signs of diabetes with 85% accuracy.
According to Machine Learning Mastery, “Deep learning is a subfield of machine learning concerned with algorithms inspired by the structure and function of the brain called artificial neural networks.”
The Apple Watch, which has saved lives thanks to its heart rate monitor that can detect different cardiovascular diseases, can now also detect diabetes, according to a study by the mobile health data company, Cardiogram.
Until now, DeepHeart's deep learning network used data from the Apple Watch to detect arterial fibrillation, hypertension, and sleep apnea. Now, a new one is added to the list: diabetes.
The co-founder of Cardiogram, Brandon Ballinger, said that the accuracy of this app to detect signs of diabetes is 85%, a considerably high figure.
Apple is selling more watches than ever, and most of its customers use the watch for health and fitness reasons. Apple Watches have saved lives by preventing heart attacks or pulmonary embolisms. They have also helped improve people's health considerably.
However, the new Cardiogram study goes well beyond the previously recognized potential of the machinery. If the idea is to use heart rate data and extrapolate it to detect diseases that are not specific to the heart, the apps and their potential to save lives are endless.
The study, led by Cardiogram at the University of California, surveyed 14,011 people with a total of 200 million heart rate detections to train DeepHeart, and thus test the reliability and accuracy of the neural network to distinguish between people with or without diabetes. This is something that Ballinger has considered essential, since about one in four people who suffer from diabetes do not know it.
What does DeepHeart detect? It tests the variability of one's heart rate. Thus, cert n heart rate patterns in people with diabetes can be detected. Examples include, a slower recovery after physical exercise, or a high heart rate when one is not exercising. The co-founder of Cardiogram says that these patterns can detect if a person suffers from or could have diabetes.
The future development of the app will focus on confirming if one has diabetes, confirming the risk, and guiding the person through a program to prevent the development of the disease. This is precisely what Cardiogram stated in their tweet: "We are working on step 2: ways to guide people through confirmation and referral to clinically-appropriate treatment. More on this later in the year!"
Important caveat: our DeepHeart paper validates accuracy at detecting existing cases of diabetes, which is step 1.
It doesn't mean your wearable will diagnose you, or screen you, for diabetes today.
— Cardiogram (@AppCardiogram) February 7, 2018
The Apple Watch has exceeded the sales expectations of investors in 2017, reaching 18 million units sold, according to the Canalys consultancy.
Overall, the average of the estimates offered by various market analysis companies figure that the total sales of the Apple Watch are between 40 and 43 million units sold since its release in March 2015.
In another breakthrough diabetes news ExOlin, an implantable device on the liver to manage Diabetes, goes on trial.