OBJECTIVE: Remote pulse oximetry in visible light (VIS) is a relevant application of photoplethysmography (PPG). However, wavelengths penetrate at different depths and VIS-based pulse oximetry may not guarantee robustness to physiological variations of the skin properties. This paper shows how a simple manoeuver like a posture change can hamper the accuracy of a method relying on red and the less penetrating green wavelengths.
APPROACH: Stationary subjects were measured under normoxic conditions while sitting and in the supine position. For each recording, we extracted remote PPG signals from forehead video recordings of 31 healthy adults at the red and green camera channels. The resulting normalized PPG-amplitudes, and its ratio, red-over-green (RoG), were compared between postures. The observed RoG changes were translated into estimates for arterial blood oxygen saturation (SpO 2, %) errors by means of Monte Carlo simulations of the skin tissue. Simulations were also used to compare the calibratability errors of SpO 2 in VIS against the conventional red-IR wavelengths.
MAIN RESULTS: RoG differs significantly between postures (RoG: sitting, 0.100 [Formula: see text] 0.025; supine, 0.123 [Formula: see text] 0.033), mediated by PPG-amplitude changes in green. The posture interference in RoG may be mitigated by an offset correction. Without this correction, we estimated that the observed change in RoG causes SpO 2 errors >[Formula: see text]. Analogous simulations involving red-IR wavelengths indicate SpO 2 errors <[Formula: see text].
SIGNIFICANCE: Our results show that the calibrations for remote pulse oximetry in VIS require the specification of a fixed measurement position. Future work could be aimed at controlling for posture in measurements.