Predicting sensory evaluation indices of Cheddar cheese texture by fluorescence fingerprint measurement coupled with an optical fibre

Abstract

A fluorescence fingerprint (FF) was used to develop a quick and non-contact practical method for predicting the sensory evaluation index of cheese texture (cheese body measurement). A partial least-squares (PLS) model was constructed from FF data and cheese body measurements of Cheddar cheeses. The cheese body measurement was successfully predicted by the PLS model with a coefficient of determination for calibration of 0.800. Notably, the reproducibility of the prediction value and the model accuracy were comparable with those of a conventional FF model despite the non-contact measurement. By exploring the variable importance in projection (VIP) and selectivity ratio (SR) of the PLS model, the fluorescence likely corresponding to oxidised lipids, Maillard reaction products, and compounds of proteins and amino acids with oxidised lipids was found to increase in intensity with the progress of ripening. This suggests that the fluorescence of these compounds contributes to the accuracy of the PLS model.