Process analytical technologies for fat and moisture determination in ground beef - a comparison of guided microwave spectroscopy and near infrared hyperspectral imaging. (March 2017)

• Record Type: Journal Article
• Authors: Zhao, Ming; Esquerre, Carlos; Downey, Gerard; O'Donnell, Colm P.
• Is Part Of: Food control. Volume 73:Part B(2017)
• Page Start: 1082
• Page End: 1094
• Publication Date: 2017-03
• Subjects: Guided microwave spectroscopy (GMS) -- Near infrared-hyperspectral imaging (NIR-HSI) -- Ground beef -- Partial least squares (PLS) -- Ensemble Monte Carlo variable selection (EMCVS); Food -- Quality -- Periodicals; Food -- Analysis -- Periodicals; Food handling -- Periodicals; Food industry and trade -- Quality control -- Periodicals; Aliments -- Industrie et commerce -- Qualité -- Contrôle -- Périodiques; Aliments -- Qualité -- Périodiques; Aliments -- Analyse -- Périodiques; Hygiène alimentaire -- Périodiques; Food -- Analysis; Food handling; Food -- Quality; Periodicals; Electronic journals; 664.07
• DOI: 10.1016/j.foodcont.2016.10.023
• Languages: English
• ISSNs: 0956-7135

Abstract

Abstract: This study investigated quality control of ground beef samples (n = 45) using both a guided microwave spectroscopy (GMS) system and a NIR-Hyperspectral imaging (NIR-HSI) system to mimic in-line/on-line measurement conditions. Partial least squares (PLS) regression models to predict fat and moisture content of ground beef samples were developed for both systems. The most informative spectral variables were selected by comparing the results of three different approaches (i.e. Martens' uncertainty test, genetic algorithm and an ensemble Monte Carlo variable selection (EMCVS)) to improve the consistency of the prediction results and reduce data processing time to facilitate on-line application. PLS models developed using the most informative microwave spectral variables obtained using a modified EMCVS procedure resulted in coefficient of determination in cross validation (R 2 CV) of 0.93 and coefficient of determination in prediction (R 2 P) of 0.90 for calibration and prediction of fat content respectively. Corresponding values of root mean square error of cross validation (RMSECV) and root mean square error of prediction (RMSEP) were between 2.13 and 2.18% w/w and 1.72–1.83%w/w respectively. For moisture content prediction, R 2 CV and R 2 P values were 0.89 and 0.82 with a RMSECV of 1.93% w/w and a RMSEP of 1.77% w/w respectively. Performance of PLS models developed on NIR-HSI information using EMCVS resulted in a R 2 P of 0.99 for both fat and moisture prediction with a minimum RMSEP of 0.73%w/w for fat content prediction and RMSEP of 0.64%w/w for moisture content prediction. Highlights: The potential of both GMS and NIR-HSI systems was explored for determinations of fat and moisture content in ground beef. Multivariate data analysis was used to predict fat and moisture content of ground beef samples by each technique. The most informative spectral variables were selected using EMCVS to improve the consistent accuracy of prediction.