[Skip to Navigation]
Editor's Note
November 27, 2019

Diagnosing With a Camera From a Distance—Proceed Cautiously and Responsibly

Author Affiliations
  • 1Center for Digital Health, Stanford University, Stanford, California
  • 2VA Palo Alto Health Care System, Palo Alto, California
  • 3Associate Editor, JAMA Cardiology
JAMA Cardiol. 2020;5(1):107. doi:10.1001/jamacardio.2019.4572

There have been dramatic advances in diagnosing arrhythmias outside of the clinical setting from sensors widely available to the general public. The placement of a light next to the optical camera sensor on smartphones, a variant of photoplethysmography, can measure pulse rate. By measuring irregularity over longer pulse sequences like some smartwatch devices, atrial fibrillation can also be identified from camera sensors.

In this issue of JAMA Cardiology, Yan and colleagues1 further push the boundaries by asking whether video images of the human face can be used to assess pulsatile blood flow and assess for atrial fibrillation. Using a digital camera, the authors recorded 20 individuals with atrial fibrillation and 24 controls, 5 people at a time, in various sequence permutations. They extracted facial photoplethysmography signals and applied a pretrained neural network to classify atrial fibrillation. In 20% of the 64 five-person recordings, 1 of 5 individuals were misclassified. At the individual-participant level, diagnostic test characteristics showed excellent discrimination in this small sample of individuals from 1 care setting in Hong Kong.

Add or change institution