Measurements of pulse rate using facial video cameras in reference to ECGs from smart devices and personal computers under low-illumination conditions

ABSTRACT

Background: Effortless and contactless monitoring technologies enable the collection of vital signs in a remote fashion without the need for physical contact with the patients. Videoplethysmography (VPG) is a technique developed to measure the pulse rate of an individual using facial video recorded by the video cameras embedded into smart devices and personal computers. It is an innovative solution to the challenges associated with the new healthcare delivery system under the SARS-CoV-1/2 pandemic. Despite these advantages, there is a knowledge gap around the performance of such technologies, especially in cardiac patients using the technology in various lighting conditions. In this work, we aim to assess the performance of the VPG technology when acquiring indoor pulse rates in patients with a history of atrial fibrillation (AF). Method(s) We evaluated the VPG technology in measuring the pulse rate in AF patients at low illumination levels. We performed an experiment in which AF patients were asked to stare at the screen of multiple devices while their VPG recordings were acquired for 25 s. We tested the VPG technology when running on 3 types of devices an S10 smartphone (SP), a Tab3 tablet (TB), and a laptop with a Logitech C920 webcam (PC). We used a single-lead Holter ECG recorder to collect the ECG and measure heart rate (HR) at the time of the facial video. We evaluated the accuracy of measurements across a cohort of AF patients after collecting VPG signals from the three devices in various illumination conditions 50 (dark surrounding area) to 500 (indoor office) lux, and two types of lights (LED and incandescent). We used the Bland-Altman (BA) method to measure the level of agreement between VPG- and ECG-based measurements of HR under various conditions. Finally, we measured the rates of rejection of recordings associated with each level of illumination and source of lights. Result(s) We present the results for the cohort of AF patients enrolled in this IRB-approved study from June to September 2021. Seventeen subjects were enrolled (12 men, 68+/-11 yrs). The BA plots shown in figure below do not show any bias. Furthermore, the mean difference in heart rate from VPG and ECG was inferior to 0.7 bpm across the 3 devices with confidence intervals ranging from 3 to 12 BPM (PC device showing the largest CI). The rate of rejection of VPG signal for too low quality was 60% at 50 lx, and 23% at 500 lx. The precision of heart rate measurements was higher under LED lights than under incandescent lights measurement error+/-CI were (0.2+/-3 bpm) vs. (0.3+/-10 bpm), respectively. Conclusion(s) This VPG technology enables accurate measurements of pulse rates in patients with a history of AF when acquired in in-door conditions (low and normal illumination levels). The proposed method ensures a consistent level of accuracy even at low illumination (50 lx). This video-cardiac monitoring method represents a unique solution for physicians to measure their patient's heart rate during telemedicine visits. [Formula presented] Copyright © 2022