Spectral emission profile and wavelength tolerances affect pulse oximeter performance
Discuss this preprint
Start a discussion What are Sciety discussions?Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
We investigate the effects of skin pigmentation and light source characteristics on the performance of reflective Pulse oximetry (PO) devices used in healthcare and well-being applications. We use Monte Carlo (MC) simulations to compare ideal monochromatic and realistic LED spectral emission profiles and tolerance-related wavelength shifts. The simulation covers photon transport in skin models with melanin concentrations (2.55% to 30.5%) and arterial oxygen saturations SaO 2 (70% to 100%.) Accuracy was assessed by SpO 2 error, root-mean-square error RMSE ( A rms), and percentile tail-errors (P90, P95, and P99).
Monochromatic spectral emission yielded the lowest SpO 2 error (RMSE = 1.32), while LED spectral emission profiles increased errors (RMSE = 2.10). Infrared wavelength tolerances increased SpO 2 RMSE by 1.1 ± 0.3. SpO 2 error increased with melanin concentration, from underestimation (−1.8 ± 0.1%) at 2.55% melanin concentration to overestimation (+3.9 ± 1.2%) at 30.5% for low SaO 2 (70%) and LED spectral emission profiles. At 30.5% melanin concentration, P95 and P99 exceeded FDA and DIN EN ISO 80601-2-61 thresholds, in particular at low SaO 2 (70%). Clipping SpO 2 estimates at 100% resulted in an apparent RMSE decrease of up to 3%, reflecting error masking rather than real error reduction.
In conclusion, LED spectral emission profiles and wavelength tolerances can amplify melanin-related bias in SpO 2 estimates. Monochromatic emission and tighter wavelength control can reduce SpO 2 error and should be considered in device design and regulation. Regulatory standards should discourage clipping SpO 2 estimates at 100% and mandate additional metrics as RMSE fails to reflect clinically critical percentile error thresholds, i.e. P95 and P99.
