According to the World Health Organization (WHO), diabetes kills over 1.6 people each year while high blood glucose level accounts for another 2.2 million deaths annually. Without treatment, diabetes can have serious health consequences, including kidney problems, eye conditions, heart disease, and stroke.
Currently, a doctor needs to take a blood sample to diagnose diabetes, which generally requires a trip to the clinic. For a wide range of reasons, many people do not have easy access to healthcare, so it is important to find simpler ways of detecting diabetes.
Researchers from the University of California in the US recently decided to investigate an innovative and freely available solution: a common smartphone app. Many fitness apps available on the market already use a technique called a photoplethysmography (PPG) signal to measure heart rate.
When the heart contracts to push blood it generates a pulse of pressure that moves through the body. Peripheral blood vessels swell to accommodate the incoming blood. By placing the smartphone’s flash and camera next to a finger, it is possible to observe the minute changes resulting from this expansion of the blood vessels.
With every contraction of the heart, the skin reflects an increasing amount of light. The smartphone’s camera can detect this change, and from this data, it is possible to extract information about blood flow.
To investigate how smartphone technology performed as a diagnostic tool, the researchers recruited 54,269 participants and used a popular smartphone app named Azumio Instant Heart Rate to measure heart-rates.
The scientists created a deep learning algorithm that used the app’s PPG signal to ascertain who had diabetes. When they pitted the algorithm against the database, the researchers found that it could correctly identify individuals with diabetes 72 percent of the time.
When they also incorporated information about other known risk factors, such as body mass index (BMI), age, and sex, the algorithm identified diabetes an impressive 81 percent of the time. Importantly, their algorithm also worked the other way around — it correctly identified people as not having diabetes 97 percent of the time.
The researchers said their findings could become a low-cost way to screen for diabetes at home, “because it can be derived from any optical system that has a camera and a flashlight, and most people have a smartphone. The researchers warned that the technology should not replace an official diagnosis by a doctor, but data from the tech could encourage someone to visit their doctor.