Smartphone-Based Screening Tool Detects Prevalent Diabetes Using Photoplethysmography

Smartphone App Cardiovascular Health
Smartphone App Cardiovascular Health
Researchers harnessed the power of deep learning and smartphone technology to develop a smartphone-based diabetes screening tool.

Through the combination of deep learning and smartphone technology, researchers from the University of California, San Francisco Medical Center have developed an app-based screening tool to test patients for diabetes. Research was presented at the American College of Cardiology 2019 Scientific Sessions, held March 16-18, 2019 in New Orleans, Louisiana. 1

Robert Avram, MD, a post-doctoral fellow at UCSF, and colleagues harnessed the power of the smartphone camera, combining it with photoplethysmography (PPG) signals, which are capable of “[picking] up vascular changes based on changes in blood flow.” Dr Avram explained the process in an ACC press release that accompanied the research presented.2

“[V]ariations in blood volume that occur with every heartbeat can be captured by shining a smartphone flashlight on a fingertip. With every heart contraction, blood pressure…causes [the blood vessels] to expand, which then increases the amount of light reflected by the skin to the optical sensor of the phone’s camera.” That input is then converted to a waveform “representing the volumetric change of the blood volume in a vessel.”

The study2 conducted by Dr Avram and colleagues included 22,298 participants (mean age: 47±14 years; 69% male, 6% with self-reported diabetes) enrolled in the online Health eHeart Study. A deep learning algorithm was applied to the smartphone-captured PPG signal recordings.

Researchers found that the model correctly identified people with diabetes in more than 72% of cases using PPG signal; a strong predictive value of 97% was also noted.

“The potential to transition screening that’s normally done by physicians or nurses to the patient themselves through a smartphone app is a very novel concept and gives us a glimpse into how health care might work in the future,” Dr Avram concluded.2

Related Articles

“We are hopeful this technology will assist with early diabetes detection.”


  1. Avram R, Tison G, Kuhar P, et al. Predicting diabetes from photolethysmography using deep learning. Presented at: American College of Cardiology 2019 Scientific Sessions; New Orleans, LA; March 16-18, 2019.
  2. App uses smartphone camera, flashlight to detect diabetes [news release]. Washington, DC: American College of Cardiology. Published March 6, 2019. Accessed March 6, 2019.