Simulation study of in vitro glucose measurement by NIR spectroscopy and a method of error reduction

The effects of some important factors on the blood glucose measurements by NIR spectroscopy are investigated by numerical simulation, and a method is proposed to significantly reduce the prediction errors induced by these effects. The changes in the absorbance spectra with the changes in the glucose concentration, temperature and scattering characteristics of background tissue are obtained by a Monte Carlo simulation of light propagation for the wavelength range from 1200 nm to 1800 nm. The glucose concentration is predicted by applying a multivariate analysis to the numerically simulated spectra. This process estimates the errors in the prediction of the glucose concentration induced by the temperature and scattering changes. It has been found that only 1degreesC change in the temperature or only 1% change in the scattering coefficient induces about 500 mg dl(-1) or 300 mg dl(-1) errors, respectively, in the prediction of the glucose concentration. These errors can be significantly reduced to less than 20 mg dl(-1) of the glucose concentration by incorporating the effects of the temperature and scattering characteristics on the spectra to the multivariate analysis.