A pulse oximeter measures arterial oxygen saturation. It relies on
The device measures the ratio of the pulsatile absorbances (each normalized by the nonpulsatile part to account for different LED intensities):
$$R = \frac{AC_{660} / DC_{660}}{AC_{940} / DC_{940}}$$
And, by applying Beer-Lambert twice, we can see
$$R = \frac{\Delta L}{\Delta L} \frac{(\epsilon_{\text{oxy}} C_{\text{oxy}} + \epsilon_{\text{deoxy}} C_{\text{deoxy}})_{660}}{(\epsilon_{\text{oxy}} C_{\text{oxy}} + \epsilon_{\text{deoxy}} C_{\text{deoxy}})_{940}}$$
$$R = \frac{(\epsilon_{\text{oxy}} C_{\text{oxy}} + \epsilon_{\text{deoxy}} C_{\text{deoxy}})_{660}}{(\epsilon_{\text{oxy}} C_{\text{oxy}} + \epsilon_{\text{deoxy}} C_{\text{deoxy}})_{940}}$$
If we then assume that there is only OxyHb and DeoxyHb i.e. \(F_{oxy} + F_{deoxy} = 1\) then
$$R = \frac{(\epsilon_{oxy} F_{oxy} [Hb] + \epsilon_{deoxy} (1-F_{oxy}) [Hb])_{660}}{(\epsilon_{oxy} F_{oxy} [Hb] + \epsilon_{deoxy} (1-F_{oxy}) [Hb])_{940}}$$
$$R = \frac{(\epsilon_{oxy} F_{oxy} + \epsilon_{deoxy} (1-F_{oxy}))_{660}}{(\epsilon_{oxy} F_{oxy} + \epsilon_{deoxy} (1-F_{oxy}))_{940}}$$
We could then solve for \(F_{oxy}\) directly; unfortunately, there is a lot of error (due to differential scattering of the different wavelengths, such that \(\Delta L_{660} \neq \Delta L_{940}\)
In practice, R is regressed against SpO2 of healthy subjects breathing gas of varying hypoxic FiO2. R of 1 \(\approx\) SO2 of 95%
Issues due to R's calibration
Issues due to abnormal Hb species
Issues due to pulsatility