Multispectral Analysis for Quantitative Measurements of Myoglobin Oxygen Fractional Saturation in the Presence of Hemoglobin Interference

Abstract

Quantitative values for myoglobin oxygen fractional saturation were extracted from visible absorption spectra of myoglobin and hemoglobin solutions by analysis with three algorithms: classical least-squares, partial least-squares, and stagewise multiple linear regression. In an effort to mimic in vivo conditions, oxygen tensions and concentrations of myoglobin and hemoglobin solutions in separate cuvettes were varied independently. Transmission measurements were made through both cuvettes so that spectra contained contributions from both myoglobin and hemoglobin. Oxygen tensions in the myoglobin solutions spanned the rapidly varying region of the myoglobin oxygen saturation curve with pO2ranging from 0 to 4.79 Torr, corresponding to fractional saturation values between 0 and 0.903. A range of hemoglobin oxygenations from fully oxygenated to fully deoxygenated was used. Estimation of myoglobin fractional saturation by the classical least-squares algorithm had a standard error (SEest) of 0.094, while the partial least-squares method resulted in an SEestof 0.070. Partial least-squares estimations resulted in an SEestof 0.041 when a limited wavelength range was used. The stagewise multiple linear regression method had an SEestof 0.052. Results indicate that stagewise regression and partial least-squares yielded estimates of myoglobin fractional saturation that were more accurate than those obtained from classical least-squares.