Performance of Noncontact Video-Based Detection of Pulse Rate and Atrial Fibrillation on Personal Computers Utilizing a Webcam.

We developed a video-based monitoring technology to detect atrial fibrillation and pulse rate to support telemedicine applications. In the past, we reported results and performance of the technology implemented on Android based devices (tablets and smartphones). Recently, the technology was implemented on the Windows operating system. In this work, we present performance results of the recent implementation over a Windows 10 laptop connected to a commercially available webcam. Our analysis is based on a clinical validation study involving 22 patients diagnosed with paroxysmal and persistent Atrial Fibrillation. Results demonstrate very high accuracy in heart rate estimation. Namely, 95% of estimated pulse rate values have no more than 5 BPM deviation from reference heart rate. In addition, specificity and sensitivity of 0.86 and 0.85 respectively are obtained when detecting atrial fibrillation. These results match those of traditionally accepted medical grade devices.