Smartphone camera and AI to detect diabetes, developed by UCSF researchers

Accuracy of over 80%, enabling early detection.

Engadget JP (Translation)
Engadget JP (Translation) , @Engadget_MT
2020年08月19日, 午前 11:15 in egmt
0シェア
FacebookTwitter
UCSF
UCSF

This article is based on an article from the Japanese edition of Engadget and was created using the translation tool Deepl.


University of California, San Francisco (UCSF) research team announced that they have developed a "digital biomarker" that uses a smartphone's built-in camera to detect type 2 diabetes.

Type 2 diabetes is a major lifestyle disease that needs no explanation at this time, as it is caused by a person's physical constitution (genetics), excessive consumption of a high-calorie, high-fat diet, and lack of exercise, which reduces the amount of insulin secretion and the effectiveness of insulin, making it difficult for blood sugar levels to drop. Since almost no symptoms appear in the early stages of the disease, you may believe you are healthy, but before you know it, the disease has progressed to a very advanced stage.

The method developed by UCSF applies a technique called photoplethysmography (PPG), in which light is shone on tissue to detect changes in blood volume. PPG technology is also used in hospitals to measure blood oxygen levels, for example, in the form of a pair of clothespins that are used to pinch your little finger.

The team used PPG records from 2.6 million people, including 53,870 diabetics, to train a 39-layer convolutional deep neural network (DNN) to be able to detect diabetes from PPG data acquired by a smartphone's built-in camera. The algorithm reportedly was able to correctly identified the presence of diabetes in up to 81 percent of patients in two separate datasets.

The researchers said that with this level of predictive performance, the algorithm could serve a similar role to other widespread disease screening tools to reach a much broader group of people, followed by a physician’s confirmation of the diabetes diagnosis and a treatment plan, and and that smartphone-based tool like this could encourage confirmatory testing of suspected diabetics at a low cost.

The next step we can immediately think of is to create a smartphone app for diabetes (suspected) diagnosis with this algorithm. However, the researchers recommend conducting further research to determine the effectiveness of this technology for specific clinical applications, such as diabetes diagnosis and treatment monitoring.

Either way, if the signs of diabetes can be detected more easily and early than ever before through familiar means like smartphone apps, it should help halt the progression of the disease before it becomes irreversible, and help people live healthier lives.

Incidentally, as diabetes progresses, various complications appear. Among them, diabetic retinopathy is the most common one that affects your daily life. When high blood sugar levels persist, the capillaries in the eye break down, preventing oxygen and nutrients from reaching the tissues, and in the worst cases, one day you can suddenly become blind. A method of diagnosing this is also being studied using a smartphone camera.

Source: UCSF (1), (2), Nature Medicine


This article is based on an article from the Japanese edition of Engadget and was created using the translation tool Deepl. The Japanese edition of Engadget does not guarantee the accuracy or reliability of this article.

 
新型コロナウイルス 関連アップデート[TechCrunch]

 

関連キーワード: egmt, healthcare, smartphone, Camera, deep learning, Medicine, news, gear
0シェア
FacebookTwitter