Using Web Cams to Detect Arrhythmias

Researchers from upstate New York have developed software that can detect atrial fibrillation by analyzing video from ordinary webcams.

Researchers from upstate New York have developed software that can detect atrial fibrillation (AF) by analyzing video from ordinary webcams.

Results from one small trial, which were published in Heart Rhythm, suggest the new technology’s error rate is about 20%. The corresponding error rates for automated electrocardiogram (ECG) measurements ranged from 17% to 29%.

The software scans video of people looking into the webcam, searching for subtle beat-to-beat variations in skin color that indicate cardiac pulsatile signal. It then times those variations to infer the patient’s beat-to-beat pulse rate.

The study team tested the system on 8 men and 3 women who had been referred for electrical cardioversion. The video plethysmographic signal (VPG) and ECG systems both detected the significant differences in pulses per minute (ppm) that distinguish periods of AF from periods of sinus rhythm: 72±9 vs. 57±7ppm (p<0.0001) for the VPG, and 80±17 vs. 56±7bpm (p<0.0001) for the ECGs.

The study team synchronized ECG and VPG signals, divided them into 407 15-second epochs and used the results to test the accuracy of the experimental detection method. The video system, which uses a novel quantifier of pulse variability that its creators call “Pulse Harmonic Strength (PHS),” performed about as well as an ECG.

“This technology holds the potential to identify and diagnosis cardiac disease using contactless video monitoring,” said the study’s lead author, Jean-Philippe Couderc, PhD, in a news release that accompanied publication of the study results. Couderc is the assistant director of the University of Rochester's Heart Research Follow-up Program. “This is a very simple concept, but one that could enable more people with AF to get the care the care they need,” he said.

Hemoglobin in the blood absorbs green light, so each time the heart pumps blood through the face, the webcam registers a little dip in the amount of green light hitting its sensor. The changes are invisible to the human eye, but they are easy for a machine to detect because the skin on the face is unusually thin and the blood vessels are unusually close to the surface.

If further trials confirm the accuracy of the new test, it offers several potential advantages over current methods for diagnosing AF. Webcams obviously cost far less than ECGs, so the new test would be cheaper. It would also be considerably faster. Patients simply look into a camera for 15 seconds and the system produces results a few seconds after that.

Ease-of-use may actually be the biggest advantage of the new system, which was developed through a partnership between Xerox Corp. and the University of Rochester’s School of Medicine and Dentistry. AF, in its earliest stages, tends to come and go, often without producing any symptoms that send patients to the doctor. Even if patients feel bad enough to seek treatment, their heart rates have often returned to normal by then, so their condition goes undiscovered.

Overall, studies estimate, about 30% of all the people with AF at any given time have yet to be diagnosed.

Patients who have been taught to monitor their own pulse can diagnose themselves at home, but few ever master the technique. An easy-but-accurate home test, one that might even run automatically in the background while people read email, could thus speed the detection and treatment of AF and improve patient outcomes.