Development of a Quantitative Visualization Technique for Gluten in Dough Using Fluorescence Fingerprint Imaging

Abstract

The distribution of constituents in food affects its end qualities such as texture, and there is a growing demand to develop a method for studying this distribution easily, accurately, and nondestructively. The objective of this study was to develop an imaging method that visualizes the precise quantity of constituents, using the fluorescence fingerprint (FF). The FF is a set of fluorescence spectra acquired at consecutive excitation wavelengths, and its pattern contains abundant information on the constituents of the sample measured. In this study, the target for visualization was the distribution of gluten in dough samples. Dough samples were prepared with different ratios of gluten, starch, and water, and fluorescence images at multiple combinations of excitation and emission wavelengths were acquired. The fluorescence intensities of a pixel at these different wavelengths constructed its FF, reflecting the constituents of the corresponding point in the sample. A partial least squares regression (PLSR) model was built using the average FFs of the samples and the corresponding gluten ratios as the explanatory and objective variables, respectively. The importance of each wavelength in the PLSR model was assessed using the selectivity ratio, and optimum wavelengths for the accurate prediction of gluten ratio were selected. Finally, the gluten ratio of each pixel was predicted with the PLSR model using the selected wavelengths, and each pixel was colored according to the predicted gluten ratio. The imaging method developed enables the distribution of constituents to be visualized with colors corresponding to their actual quantities or ratios.