Near-infrared (NIR) Prediction of trans-Fatty Acids in Ground Cereal Foods

Abstract

Near-infrared (NIR) reflectance spectroscopy was evaluated as a rapid method for prediction of trans-fatty acid content in ground cereal products without the need for oil extraction. NIR spectra (400-2498 nm) of ground cereal products were obtained with a dispersive NIR spectrometer and correlated to trans- and cis-fatty acid content determined by a modification of AOAC Method 996.01. Partial least-squares regression and Marten's uncertainty test were applied to calculate models for prediction of trans-fatty acids using spectral regions affected by lipid absorption. The best model (n=84) for trans-fat prediction used the 700-2498 nm region and second-derivative processing of spectra. When used to predict test samples (n=27) the model had an RPD of 4.8 with a standard error of performance of 0.70% (range of 0.05-11.74%) and r(2) of 0.97. Optimum models for cis-fatty acids were developed with the 1100-2498 and 700-2498 nm ranges and had an RPD of 4.0. Regression coefficients indicated that useful absorbance for prediction of trans- and cis-fatty acids was in the overtone and combination regions for lipid absorption.