Gerard Downey

Prediction of beef eating quality from colour, marbling and wavelet texture features.

Beef longissimus dorsi colour, marbling fat and surface texture are long established properties that are used in some countries by expert graders to classify beef carcasses, with subjective and inconsistent decision. As a computer vision system can deliver objective and consistent decisions rapidly and is capable of handling a greater variety of image features, attempts have been made to develop computerised predictions of eating quality based on these and other properties but have failed to adequately model the variation in eating quality. Therefore, in this study, examination of the ribeye at high magnification and consideration of a broad range of colour and marbling fat features was used to attempt to provide better information on beef eating quality. Wavelets were used to describe the image texture of the beef surface at high magnification rather than classical methods such as run lengths, difference histograms and co-occurrence matrices. Sensory panel and Instron analyses were performed on duplicate steaks to measure the quality of the beef. Using the classical statistical method of partial least squares regression (PLSR) it was possible to model a very high proportion of the variation in eating quality (r(2)=0.88 for sensory overall acceptability and r(2)=0.85 for 7-day WBS). Addition of non-linear texture terms to the models gave some improvements.

Review: Authentication of food and food ingredients by near infrared spectroscopy

Food authenticity is an issue of concern to food processors, retailers, regulatory authorities and consumers alike. Near infrared (NIR) spectroscopy has many potential advantages as an authenticity testing tool and has already been applied to a number of authentication problems using a range of sample presentation and chemometric techniques. This review outlines the principles of the statistical procedures used so far, and summarises the work reported to-date on a range of foods and food ingredients.

Food and food ingredient authentication by mid-infrared spectroscopy and chemometrics

Abstract Authenticity is an important food quality criterion. Rapid methods for confirming authenticity or detecting adulteration are increasingly demanded by food processors, consumers and regulatory bodies. Mid-infrared spectroscopy has recently been applied to this problem and this short review reports on some of the experimental and statistical approaches used and the results obtained. Implications for future applications are considered.

Non-destructive prediction of selected quality attributes of beef by near-infrared reflectance spectroscopy between 750 and 1098 nm

Heifers (n = 70) were slaughtered and hung conventionally in an industrial meat plant. Near infrared (NIR) spectroscopy was studied for its ability to predict selected meat quality attributes, i.e. Warner-Bratzler shear force (WBSF), sensory tenderness, texture, flavour and acceptability. Freshly cut steaks (2.5 cm thick) were taken from the longissimus dorsi (LD) muscle at 1, 2, 7 and 14 days post mortem for NIR analysis. Other samples (also 2.5 cm thick) were taken at 2, 7 and 14 days post mortem, vacuum-packaged in plastic bags and stored at -20 °C for WBSF measurement and sensory analysis. Heifers were slaughtered in two groups; between slaughterings, replacement of the spectrophotometer lamp and lamp assembly was necessitated by a bulb failure. Using principal component regression (PCR), correlation coefficients of 0.82 and 0.73 were obtained for the prediction of WBSF in sample sets 1 and 2, respectively. On merging both sample sets, this value was lowered slightly (r = 0.61). Correlation coefficients obtained for the prediction of tenderness, texture, flavour and acceptability were 0.67, 0.53, 0.51 and 0.46 respectively (set 1); 0.72, 0.71, 0.45 and 0.67 (set 2); 0.53, 0.54, 0.24 and 0.42 (combined sets).

Geographic Classification of Extra Virgin Olive Oils from the Eastern Mediterranean by Chemometric Analysis of Visible and Near-Infrared Spectroscopic Data

Visible and near-infrared reflectance spectra have been examined for their ability to classify extra virgin olive oils from the eastern Mediterranean on the basis of their geographic origin. Classification strategies investigated were partial least-squares regression, factorial discriminant analysis, and k-nearest neighbors analysis. Discriminant models were developed and evaluated using spectral data in the visible (400–750 nm), near-infrared (1100–2498 nm), and combined (400–2498 nm) wavelength ranges. A variety of data pretreatments was applied. Best results were obtained using factorial discriminant analysis on raw spectral data over the combined wavelength range; a correct classification rate of 93.9% was obtained on a prediction sample set. Though the overall sample set was limited in numbers, these results demonstrate the potential of near-infrared spectroscopy to classify extra virgin olive oils on the basis of their geographic origin.

A review of near infrared spectroscopy in muscle food analysis: 2005–2010

Muscle foods (meat and fish) are very important from the perspective of human nutrition and economic activity, both nationally and internationally. At a research and development level, major efforts continue to be focussed on improving the quantity and quality of raw and processed muscle food types available on the market and also to monitor their compliance with compositional, safety and, increasingly, provenance issues. Publications dealing with the development of near infrared (NIR) applications for the analysis of muscle foods (meat and fish) over the period 2005–2010 have been assembled and reviewed. Well-described advantages of NIR spectroscopy suit the food processing industry in terms of operating speed and possible implementation of in-line, on-line or at-line process monitoring; it also has the ability to meet consumer expectations in terms of product quality and safety assurance. These advantages allow food processors to easily monitor and manipulate processing conditions to avoid the production and release of defective products, thereby guaranteeing product quality and enhancing the possibility of repeat purchasing by customers. For public regulatory organisations which have responsibilities to both food producers and consumers, NIR technology may be able to contribute efficiently to these aims. Interrogation of NIR datasets by increasingly powerful and sophisticated chemometric techniques continues to improve calibration robustness and accuracy while the appearance of extensive suites of algorithms in commercially-available software packages helps in their deployment. The aim of this review is to provide an update on work in these areas which has been published in the period from 2005 to 2010. While targeted chiefly at researchers active in the field, it should also be of relevance to technical personnel in the meat and fish industries and to regulatory personnel.

Authentication of Coffee Bean Variety by Near-infrared Reflectance Spectroscopy of Dried Extract

The potential of near-infrared (NIR) reflectance spectroscopy for discriminating between coffee beverage prepared from pure Arabica, pure Robusta and blends of these two varieties was investigated. Dried beverages were produced by both lyophilisation and air-drying under vacuum on glass-fibre filter paper. Spectral collections were treated by principal component and factorial discriminant analyses. Using the wavelength range 1100–2498 nm, only three of 65 test samples were misclassified when the filter paper approach was used. When freeze-dried coffee beverages were analysed, nine of the 65 test samples were misclassified. The basis for this discrimination appears to involve caffeine and/or other alkaloids.

Preventing over-fitting in PLS calibration models of near-infrared (NIR) spectroscopy data using regression coefficients

Selection of the number of latent variables (LVs) to include in a partial least squares (PLS) model is an important step in the data analysis. Inclusion of too few or too many LVs may lead to, respectively, under or over‐fitting of the data and subsequently result in poor future model performance. One well‐known sign of over‐fitting is the appearance of noise in regression coefficients; this often takes the form of a reduction in apparent structure and the presence of sharp peaks with a high degree of directional oscillation, features which are usually estimated subjectively. In this work, a simple method for quantifying the shape and size of a regression coefficient is presented. This measure can be combined with an indicator of model bias (e.g. root mean square error) to aid in estimation of the appropriate number of LVs to include in a PLS model. The performance of the proposed method is evaluated on simulated and and real NIR spectroscopy datasets sets and compared with several existing methods. Copyright

Classification of Commercial Skim Milk Powders According to Heat Treatment Using Factorial Discriminant Analysis of Near-Infrared Reflectance Spectra

Near-infrared reflectance (NIR) spectroscopy is well established as a rapid and nondestructive analytical technique in many agro-food industries. Most published applications have been concerned with the use of NIR for quantitative analyses of technologically important chemical constituents such as water and protein in grain, alcohol in wine, oil in mayonnaise, etc., although successes have been reported with the prediction of less precisely defined but nonetheless functionally important sample attributes (e.g., wheat hardness). A number of mathematical techniques have been used to develop accurate and stable prediction equations, among which may be mentioned stepwise multiple linear regression, principal component regression, partial least-squares regression, and derivative quotient mathematics.

Detection of honey adulteration by addition of fructose and glucose using near infrared transflectance spectroscopy

Samples of artisanal honey produced in Ireland over a number of harvests have been obtained directly from producers. Adulterant solutions containing both fructose and glucose at ratios of 0.7: 1, 1.2: 1 and 2.3: 1 w/w were prepared. Honeys and adulterants were adjusted to a constant solids content. Visible and near infrared (400–2498 nm) transflectance spectra of the honeys were collected before and after adulteration at levels of 7, 14 and 21% w/w of each of the fructose-plus-glucose adulterant solutions. Chemometric analysis of the spectral collection by discriminant partial least squares regression (PLS1), k-nearest neighbours (k-NN) and soft independent modelling of class analogy (SIMCA) have been performed with a view to discriminating between the unadulterated and adulterated honey samples. Discriminant partial least squares regression proved to be the most accurate of these three methods.