Near-infrared spectroscopy glucose measurement in aqueous solutions using single beam spectra

Quantitative glucose measurements are made directly from near-IR single-beam spectra. The 5000-4000 cm{dollar}sp{lcub}-1{rcub}{dollar} spectral region is used in this study for analysis of glucose in multi-temperature glucose-buffer sample matrices, multi-path length glucose-protein-buffer sample matrices, glucose-protein-buffer matrices and glucose-triacetin-buffer matrices.; By combining digital Fourier filtering and partial least-squares regression, single-beam spectra can be used directly to generate accurate calibration models free of interference from temperature variation and protein or triacetin concentration variations. Models based on single-beam spectra perform as well as those based on ratioed absorbance spectra.; In multi-path length studies, a path length prediction and correction algorithm is developed. Digital Fourier filtering is coupled with partial least squares (PLS) regression to generate functional glucose calibration models from path length corrected single-beam spectra. Multiplicative scatter correction is used to minimize light source variations. Performance characteristics of models generated from path length corrected multiple path length data compare favorable to those obtained with uniform path length data.; The path length prediction algorithm is evaluated as a spectra comparison tool to detect spectral outliers. Results demonstrate that this algorithm is well suited for this purpose with particular sensitivity to solution temperature and detector saturation. Predicted path length values outside the distribution of values for the whole data set indicate a likely deviation of the data collection condition.; The contribution of water displacement on PLS calibration models is investigated by using deuterium oxide to decouple water displacement and analyte concentration. This experiment proves that glucose concentration can be predicted independent of water displacement.