Measurement of Glucose in Water with First-Overtone Near-Infrared Spectra

Partial least-squares (PLS) regression analysis was used to build calibration models for three unique spectral data sets of glucose in water. Spectra in the first data set were collected with a 2.0 mm optical pathlength. For these data, the measured root-mean-square (rms) noise of 100% lines over the 5975-5850 cm⁻¹ spectral range was 4.5 micro-absorbance units (μAU). Spectrometer upgrades permitted a 5.2 mm optical pathlength for the second data set, and the resulting spectra had an rms noise of 5.9 μAU. Further spectrometer adjustments allowed the use of a 10.0 mm optical pathlength for the third data set, and the resulting spectral rms noise was 8.4 μAU. In each case, the instrumentation was modified individually in order to provide high radiant powers at the detector while avoiding detector saturation. Poor calibration models for the first data set indicate that a 2.0 mm optical pathlength is insufficient for adequate glucose measurements at clinically relevant concentrations. Calibration and prediction errors for the data collected at 5.2 and 10.0 mm pathlengths ranged from 0.40–0.50 and 0.35–0.40 mM, respectively. Digital Fourier filtering significantly improved model performance by reducing the required number of latent variables (factors) in the PLS models and by reducing the wavelength dependency of these models. For the best calibration model, spectra in the data set corresponding to a 10.0 mm pathlength were Fourier filtered with a Gaussian-shaped filter defined in digital frequency units (f) by a mean position of 0.0206 f and a standard deviation width of 0.0031 f. These filtered spectra were then submitted to a one-factor PLS model that is limited to the 5975-5850 cm⁻¹ spectral range. Consideration of different spectral ranges and an analysis of spectral loading vectors indicate that the 5920 cm⁻¹ absorption band for glucose is critical for useful analytical measurements.