Objective: Photoplethysmography (PPG) is a noninvasive technique to measure the blood-volume pulse and derive various vital signs. Camera-based PPG imaging was recently proposed for clinical microvascular assessment, but motion robustness is still an issue for this technique. Our study aims to quantify cardiac-related, i.e., ballistocardiographic (BCG), motion as a source of artifacts in PPG imaging.
Methods: In this paper, using the human head as a relevant region of interest, the amplitude of BCG-artifacts was modeled for a Lambertian surface illuminated by a light source. To derive peak-to-peak head displacements for the model, we recorded, on 54 subjects, PPG and inertial sensor data at the pulse and cranial vertex. We simulated the effect of light source location at a mesh representation of a human face and conducted additional experiments on a real subject.
Results: Under nonorthogonal illumination, the relative strength of the BCG artifacts is strong enough, compared to the amplitude of PPG signals, to compromise PPG imaging in realistic scenarios. Particularly affected are the signals obtained in the nongreen part of the spectrum and/or when the incident angle at the skin surface exceeds 45 (°).
Conclusion: From the model and an additional experiment conducted on real skin, we were able to prove that homogenous and orthogonal illumination is a means to minimize the problem.
Significance: Our illumination recommendation provides a simple and effective means to improve the validity of remote PPG-imagers. We hope that it helps to prevent mistakes currently seen in many publications on remote PPG.