Research results that diabetes can be detected with a surprise precision of 85% if there is a wearable terminal that can measure heart rate

AppleWe launched a secret research team within the company to bring innovation to diabetes treatment with Apple WatchAlthough it was reported as being reported, the latest research shows that if there is a wearable terminal that can measure heart rate, we can detect diabetes with high accuracy of 85%.

Ordinary wearables can flag signs of diabetes, according to new Cardiogram study

Apple Watch detected diabetes with 85% accuracy | Cult of Mac

In the United States, it is said that more than 100 million adults are diabetes and diabetes reserve forces, but a quarter of diabetic patients are not aware that they are diabetic themselves, and as a diabetes reserve army It seems that 88.4% is not aware of it. In the medical community there is no need to insert instruments in the skin for decades so that many people can judge whether it is diabetes more quicklyNoninvasiveWe are trying to develop a method to measure glucose concentration in the blood in a similar way.

At the time of writing, there is no established method to measure glucose concentration in the blood in a non-invasive way and to check whether it is diabetes, but in a newly published study, heart rate like Apple Watch By combining a wearable terminal with a deep neural network, it has been shown that wearers can investigate the early signs of diabetes with surprising precision.

A team of the University of California San Francisco school succeeded in detecting diabetes by combining a wearable terminal and a neural network. The research team used samples for heart rate recording applications provided for Apple Watch and Android Wear "Cardiogram"Data of 10,4011 people for users. Also, in order to let the deep neural network "DeepHeart" learn what health data of diabetes patient is like, health data of 33,628 persons is used separately from Cardiogram data.

When using a neural network such as DeepHeart, it is necessary to learn "abnormal heart rhythm of a person who causes a heart attack" using millions of data, but in this case, what kind of symptoms Learning with both unsupervised learning and supervised learning by using 'Health data for 33,228 people' who knows whether they are suffering from the disease and 'unknown' data for 10,411 people I heard that a method was taken.

Ultimately DeepHeart succeeded in detecting that 462 of the data providers are diabetes, and its detection accuracy was surprisingly high accuracy of 84.51%. Since DeepHeart can be used if heart rate data is available, wearable terminals that can measure heart rate, such as Android Wear and Fitbits, will be able to detect diabetes with considerable precision even if it is not Apple Watch .

in Mobile,   Hardware,   Science, Posted by logu_ii