Using different sample holders in determining protein and moisture content in whole wheat flour by means of Fourier transform near infrared (FT-NIR) spectroscopy

Abstract

Fourier transform near infrared (FT-NIR) spectroscopy was used to record the spectra of 92 whole wheat flour samples presented in three different sample holders. The conventional sample cup with a sapphire-glass base (provided with the spectrophotometer), borosilicate-glass vials and soda-glass vials were used. Calibrations were derived for protein and moisture content, respectively, of whole wheat flour by performing partial least square (PLS) regression on multiplicative scatter corrected (MSC) spectra and tested using independent validation procedures. Best results were obtained with the sample set analysed in the borosilicate-glass vials, although not significantly different (P > 0.05). The standard error of prediction (SEP), root mean standard error of prediction (RMSEP) and the correlation coefficient (r) were, respectively, 0.5%, 1.16% and 0.92 for the protein calibration model, and 0.15%, 0.38% and 0.94 for the moisture calibration model. The differences in spectral data due to the use of the different sample holders were investigated using correlation coefficients (r) and maximum distances (d). The significance of differences in prediction results was determined by performing an analysis of variance (ANOVA) on predicted protein and moisture values. Differences within each type of sample holder were also determined. All differences were found to be statistically insignificant (P > 0.05). The conventional sample cup can therefore be replaced by one of the more convenient vial types where each sample can be stored and presented to the spectrophotometer individually. © 2001 Taylor & Francis Group, LLC.