Partial least-squares (PLS) and second-derivative preprocessing were used to obtain estimates of myoglobin oxygen fractional saturation from diffuse reflectance spectra of solutions containing myoglobin, hemoglobin, and a scatterer. A computer model and solutions in vitro were used to simulate several physiological situations. The maximum standard error (SE) was 0.082 for these trials; myoglobin fractional saturation varies between 0 and 1. These results show that a statistical approach can differentiate two highly overlapping absorbance peaks in the presence of diffuse scatter. A robust PLS model was created by using a calibration set with a range of scattering coefficients and concentrations of hemoglobin. Second derivatives of the spectra were less affected by changes in scattering coefficients than were the original spectra. A linear scaling of PLS estimates produced accurate myoglobin saturations from in vitro prediction set spectra that had scattering and absorption coefficients both within and beyond the range represented by the calibration set. Preliminary estimates of myoglobin fractional saturation from spectra acquired from the rat hind limb suggest that this calibration set is appropriate for use in vivo.
Lorilee S. L. Arakaki, Martin J. Kushmerick, and David H. Burns, "Myoglobin Oxygen Saturation Measured Independently of Hemoglobin in Scattering Media by Optical Reflectance Spectroscopy," Appl. Spectrosc. 50, 697-707 (1996)