Cardiovascular Risk Assessment Using Remote Photoplethysmography Signal

Abstract

Cardiovascular diseases (CVDs) are a leading health risk in the world, and accurate and convenient monitoring of cardiovascular parameters such as heart rate (HR) and oxygen saturation (SpO₂) is required. Contact-based methods such as pulse oximeters, however, are not feasible in patients with burns, cognitive impairments, or in pandemics such as COVID-19. Due to these limitations, there is a need for non-invasive contactless alternatives like remote photoplethysmography (rPPG) that estimate vital signs through light variation analysis from facial video. This study proposes a contact-free method for monitoring heart rate (HR) and blood oxygen saturation (SpO₂) using a standard ambient light camera. The author utilized the Plane-Orthogonal-to-Skin (POS) algorithm to extract rPPG signals from facial videos. Both frequency domain (via Fast Fourier Transform) and time domain analyses were applied to estimate HR, while SpO₂ was derived using red and blue channel BVP signal ratios. Signal preprocessing, including normalization and bandpass filtering, was employed to improve signal quality and reduce noise. An experimental evaluation on a dataset of 20 participants aged 20 to 50 years showed that the proposed system achieved a mean absolute error (MAE) of 3.66 beats per minute (BPM) and a root mean square error (RMSE) of 5.11 BPM in heart rate (HR) estimation. Bland-Altman analysis was employed to assess the algorithm’s performance. The mean difference between the predicted HR, SpO2, and the actual values is -2.16BPM and -5.98%, respectively. Bland-Altman analysis showed minimal bias, confirming the system’s reliability in non-contact cardiovascular monitoring.