Non-destructive Near Infrared Spectroscopy for the labelling of frozen Iberian pork loins.

Iberian pigs fed on acorns and pasture were slaughtered from January until March of 2018 and 2019. The meat from those Iberian pigs is a seasonal food that only can be found fresh, at the marketplace, during a limit period of the year. Selling frozen-thawed meat is a legal practice, but consumers must be informed about it on the product label. However, to declare as fresh meat, meat previously frozen, is one of the most frequent meat frauds. The present study compares the performance of two rather different Near Infrared Spectroscopy instruments, based on Fourier Transform and Linear Variable Filter technologies, for the in-situ detection of fresh and frozen-thawed acorns-fed Iberian pig loins using Partial Least Discriminant Analysis (PLS-DA). The performance of the models developed for both instruments offered a very high discriminant ability. Furthermore, the models showed consistent results and interpretation when were evaluated with several scalars and graphical methods.

Short Communication: The potential of portable near infrared spectroscopy for assuring quality and authenticity in the food chain, using Iberian hams as an example.

This communication assesses the use of a portable near infrared (NIR) instrument to measure quantitative (fatty acid profile) properties and qualitative ('Premium' and 'Non-premium') categories of individual Iberian pork carcasses at the slaughterhouse. Acorn-fed Iberian pigs have more unsaturated fats than pigs fed conventional compound feed. Recent advances in miniaturisation have led to a number of handheld NIR devices being developed, allowing processing decisions to be made earlier, significantly reducing time and costs. The most common methods used for assessing quality and authenticity of Iberian hams are analysis of the fatty acid composition of subcutaneous fat using gas chromatography and DNA analysis. In this study, NIR calibrations for fatty acids and classification as premium or non-premium ham, based on carcass fat measured in situ, were developed using a portable NIR spectrometer. The accuracy of the quantitative equations was evaluated through the standard error of cross validation or standard error of prediction of 0.84 for palmitic acid (C16:0), 0.94 for stearic acid (C18:0), 1.47 for oleic acid (C18:1) and 0.58 for linoleic acid (C18:2). Qualitative calibrations provided acceptable results, with up to 98% of samples (n = 234) correctly classified with probabilities ⩾0.9. Results indicated a portable NIR instrument has the potential to be used to measure quality and authenticity of Iberian pork carcasses.

Fourier transform near-infrared spectroscopy coupled to a long fibre optic head for the quality control of IBERIAN pork loins: Intact versus minced.

Conventional chemical analyses of meat products are time-consuming, expensive and destructive. The advantages of NIR spectroscopy are its speed, portability, suitability for both at-line and on-line analysis, low cost and the possibility of simultaneously measuring many different parameters in a large number of samples. The purpose of this study was to develop and validate calibrations for the prediction of moisture, protein and fat in Iberian pig pork loins using an FT-NIR instrument coupled to a 5-m fibre optic sensor head. The best equations obtained for intact loin in both modes of analysis (full and optimal spectral range) displayed Standard Error of Cross-Validation (RMSECV) of 1.06% and 1.09% and Determination Coefficient of Cross-Validation (RCV2) of 0.69 and 0.77 for fat: RMSECV of 0.87% and 0.77% and RCV2 of 0.67 and 0.73 for moisture; while for protein, the RMSECV values were 0.51% and 0.49% and the RCV2 values were 0.66 and 0.70.

Prediction of fatty acids content in pig adipose tissue by near infrared spectroscopy: at-line versus in-situ analysis.

A handheld micro-electro-mechanical system (MEMS) based spectrometer working in the near infrared region (NIR) (1600-2400nm) was evaluated for in-situ and non-destructive prediction of main fatty acids in Iberian pig (IP) carcasses. 110 IP carcasses were measured. Performance of the instrument was compared with at-line high-resolution NIRS monochromators working in two analysis modes: melted fat samples (transflectance cups) and intact adipose tissues (interactance fiber optic). Standard Error of Prediction (SEP) values obtained on the MEMS-NIRS device were: 0.68% (stearic), 1.30% (oleic), 0.55% (linoleic) and 1% (palmitic), explaining a variability of 83%, 84%, 81% and 78%, respectively. As expected, this represented a loss of predictive capability in comparison to at-line models, even with the same spectral characteristics as on the handheld device. However, the estimated total errors were at the same level for gas chromatography and NIRS analysis. This indicates that the MEMS-NIRS in-situ analysis of each individual carcass provides a cost-effective and real-time quality control system with suitable accuracy.

Data pre-processing to improve the mining of large feed databases.

The information stored in animal feed databases is highly variable, in terms of both provenance and quality; therefore, data pre-processing is essential to ensure reliable results. Yet, pre-processing at best tends to be unsystematic; at worst, it may even be wholly ignored. This paper sought to develop a systematic approach to the various stages involved in pre-processing to improve feed database outputs. The database used contained analytical and nutritional data on roughly 20 000 alfalfa samples. A range of techniques were examined for integrating data from different sources, for detecting duplicates and, particularly, for detecting outliers. Special attention was paid to the comparison of univariate and multivariate solutions. Major issues relating to the heterogeneous nature of data contained in this database were explored, the observed outliers were characterized and ad hoc routines were designed for error control. Finally, a heuristic diagram was designed to systematize the various aspects involved in the detection and management of outliers and errors.

Handling of missing data to improve the mining of large feed databases.

Feed databases often have missing data. Despite their potentially major effect on data analysis (e.g., as a source of biased results and loss of statistical power), database managers and nutrition researchers have paid little attention to missing data. This study evaluated various methods of handling missing data using mining outputs from a database containing data on chemical composition and nutritive value for 18,864 alfalfa samples. A complete reference dataset was obtained comprising the 2,303 cases with no missing data for the attributes CP, crude fiber (CF), NDF, ADF and ADL. This dataset was used to simulate 2 types of missing data (at random and not at random), each with 2 loss intensities (33 and 66%), thus yielding a total of 4 incomplete datasets. Missing data from these datasets were handled using 2 deletion methods and 4 imputation methods, and outputs in terms of the identification and typing of alfalfa (using ANOVA and descriptive statistics) and of correlations between attributes (using regressions) were compared with outputs from the complete dataset. Imputation methods, particularly model-based versions, were found to perform better than deletion methods in terms of maximizing information use and minimizing bias although the extent of differences between methods depended on the type of missing data. The best approximation to the uncertainty value was provided by multiple imputation methods. It was concluded that the choice of the most suitable method for handling missing data depended both on the type of missing data and on the purpose of data analysis.

Authentication of organic feed by near-infrared spectroscopy combined with chemometrics: a feasibility study.

Organic products tend to retail at a higher price than their conventional counterparts, which makes them susceptible to fraud. In this study we evaluate the application of near-infrared spectroscopy (NIRS) as a rapid, cost-effective method to verify the organic identity of feed for laying hens. For this purpose a total of 36 organic and 60 conventional feed samples from The Netherlands were measured by NIRS. A binary classification model (organic vs conventional feed) was developed using partial least squares discriminant analysis. Models were developed using five different data preprocessing techniques, which were externally validated by a stratified random resampling strategy using 1000 realizations. Spectral regions related to the protein and fat content were among the most important ones for the classification model. The models based on data preprocessed using direct orthogonal signal correction (DOSC), standard normal variate (SNV), and first and second derivatives provided the most successful results in terms of median sensitivity (0.91 in external validation) and median specificity (1.00 for external validation of SNV models and 0.94 for DOSC and first and second derivative models). A previously developed model, which was based on fatty acid fingerprinting of the same set of feed samples, provided a higher sensitivity (1.00). This shows that the NIRS-based approach provides a rapid and low-cost screening tool, whereas the fatty acid fingerprinting model can be used for further confirmation of the organic identity of feed samples for laying hens. These methods provide additional assurance to the administrative controls currently conducted in the organic feed sector.

In-situ Iberian pig carcass classification using a micro-electro-mechanical system (MEMS)-based near infrared (NIR) spectrometer.

Iberian pig (IP) products are gourmet foods highly appreciated at international markets, reaching high prices, because of its exquisite flavors. At present, there aren't practical and affordable analytical methods which can authenticate every single piece put on the market. This paper reports on the performance of a handheld micro-electro-mechanical system (MEMS)-based spectrometer (1600-2400nm) for authentication-classification of individual IP carcasses into different commercial categories. Performance (accuracy and instrumental design) of the instrument was compared with that of high-resolution NIRS monochromators (400-2500nm). A total of 300 carcasses of IPs raised under different feeding regimes ("Acorn", "Recebo" and "Feed") were analyzed in three modes (intact fat in the carcass, skin-free subcutaneous fat samples and melted fat samples). The best classification results for the MEMS instrument were: 93.9% "Acorn" carcasses correctly classified, 96.4% "Feed" and 60.6% "Recebo", respectively. Evaluation of model performance confirmed the suitability of the handheld device for individual, fast, non-destructive, low-cost analysis of IP carcasses on the slaughterhouse line.

Building a metadata framework for sharing feed information in Spain.

Information about the nutritional aspects and uses of feed is of widespread interest, hence systematic efforts of laboratories to obtain it. The way this information is currently being handled leaves something to be desired, underscoring the need to use computerized systems and statistical techniques that allow the management of large volumes of heterogeneous information. This project seeks to develop a structure that will facilitate the exchange and exploitation of information on feeds produced in Spain. To this end, metadata and data mining techniques have been adopted by the Feed Information Service at the University of Cordoba. The structure has been designed to work on the basis of a server-client architecture, in which information is stored on local software (Califa) by its own creators so that it can subsequently be incorporated into a database server where it can be accessed online. Various aspects of the structure are described in this paper: organization (participants and data shared), format (physical features), logistics (data description), quality (reliability of information), legality (correct use of data), and financing (revenue and expenditure). An indication is given of the amount of information accumulated to date, now exceeding 200,000 numerical data and associated metadata, arranged in several thematic databases. The activities carried out highlight the heterogeneous nature of the information produced, as well as the large number of errors and ambiguities that slip through the normal filters and reach the end-user of the data.

Multivariate near-infrared reflection spectroscopy strategies for ensuring correct labeling at feed bagging in the animal feed industry.

A key concern in animal feed factories is guaranteeing the correct labeling of compound feeds. Therefore, due to incorrect labeling, there is an urgent need for new control methods on the claims that can be made. In this study, this question has been tackled with different multivariate classification algorithms based on the near-infrared spectral fingerprint obtained from a given compound feed analyzed in its original physical market presentation form (i.e., cubes, coarse meals, pellets). The objective of this paper is the evaluation of different methods for establishing a separation among 24 feed types. Two linear methods, soft independent modeling of class analogy (SIMCA) and partial least squares (PLS) with two approaches to classification (PLSD and PLS-LDA); and one nonlinear method, support vector machines (SVM), were studied. The database used had the following structure: a first division was made between granules and meals; within these two groups, there was a second division according to three animal species to which the feed was marketed (bovine, ovine, and porcine); within each species there was a third division according to the age or physiological status of the animal (i.e., lactating dairy cattle, starters, etc.). Given the database structure, all the methods were evaluated following two strategies: (1) development of a model composed of the nine classification models corresponding to the structure of the data; and (2) development of a unique model that discriminates among the 24 classes of different feeds. With both strategies the lowest percentage of misclassified samples was achieved with the SVM method (3.96% with strategy 1 and 2.31% with strategy 2). Among the linear methods evaluated, SIMCA yielded the best results, with a percentage of 8.47% misclassified samples with strategy 1 and 4.05% misclassified samples with strategy 2. The results in this study show the ability of near-infrared spectroscopy to make acceptable classifications of feed types based only on spectral information, with differences in performance depending on the multivariate algorithm used.

A feasibility study on the use of near-infrared spectroscopy for prediction of the fatty acid profile in live Iberian pigs and carcasses.

This work reports on the development and optimisation of NIRS technology for fat characterization both in live pigs and in carcasses in the slaughterhouse; use of this technology would enable implementation of a real-time traceability and control system based on non-destructive sensors. A total of 52 Iberian pig fat samples were analysed using a LabSpec(®)Pro A108310 spectrophotometer (Analytical Spectral Device Inc.), with a high-intensity fiber-optic contact probe. Spectra were collected in five analysis modes: from the live animal, from the carcass in the slaughterhouse, from a subcutaneous fat sample with skin, from a skin-free subcutaneous fat sample and from a transverse section. Calibrations were developed for the prediction of the four main fatty acids in Iberian pig fat, obtaining for palmitic acid SECV values of 1.24% for in vivo analysis and 0.82% for carcass analysis, for stearic acid 0.67% and 0.94%, for oleic acid 1.42% and 1.48% and for linoleic 0.36% and 0.55%, respectively. The calibrations accounted for between 60% and 74% of the variation recorded in live animals, and between 31% and 87% of variation in carcasses. These results confirm the feasibility of NIRS technology for the on-site inspection and control of Iberian pig, both in the field and in the slaughterhouse.

Advanced nonlinear approaches for predicting the ingredient composition in compound feedingstuffs by near-infrared reflection spectroscopy.

For quantitative applications, the most common usage of near-infrared reflection spectroscopy (NIRS) technology, calibration involves establishing a mathematical relationship between spectral data and data provided by the reference. This model may be fairly complex, since the near-infrared spectrum is highly variable and contains physical/chemical information for the sample that may be redundant, and multivariate calibration is usually required. When the relationship to be modeled is nonlinear, classical regression methods are inadequate, and more complex strategies and algorithms must be sought in order to model this nonlinearity. The development of NIRS calibrations to predict the ingredient composition, i.e., the inclusion percentage of each ingredient, in compound feeds is a complex task, due to the nature of the parameters to be predicted and to the heterogeneous nature of the matrices/formulas in which each ingredient participates. The present paper evaluates the use of least squares support vector machines (LSSVM) and two local calibration methods, CARNAC and locally biased regression, for developing NIRS models to predict two of the most representative ingredients in compound feed formulations, wheat and sunflower meal, using a large spectral library of 7523 commercial compound feed samples. For both ingredients, the best results were obtained using CARNAC, with standard errors of prediction (SEP) of 1.7% and 0.60% for wheat and sunflower meal, respectively, and even better results when the algorithm was allowed to refuse to predict 10% of the unknowns. Meanwhile, LSSVM performed less well on wheat (SEP 2.6%) but comparably on sunflower meal (SEP 0.60%), giving results very similar to those reported previously for artificial neural networks. Locally biased regression was the least successful of the three methods, with SEPs of 3.3% for wheat and 0.72% for sunflower meal. All the nonlinear methods improved on the standard approach using partial least squares (PLS), which gave SEPs of 5.3% for wheat and 0.81% for sunflower meal.

Nutritive evaluation of olive tree leaves by near-infrared spectroscopy: effect of soil contamination and correction with spectral pretreatments.

Olive leaves obtained as a byproduct in the Mediterranean region could play an important role in the nutrition of extensive ruminant systems. However, the reported variation in their nutritive value, among other reasons due to discrepancies in mineral content, is considered an important obstacle for their common use. Near-infrared spectroscopy (NIRS) could fulfill the requirements of these productive systems, providing analytical information in a rapid and economic way. In this work, the effect of soil contamination on NIR spectra has been studied, as well as its correction with some of the most commonly used spectral pretreatments (derivatives, multiplicative scatter correction, auto scaling, detrending, and a combination of the last two transforms). Effects were evaluated by visual inspection of the transformed spectra and comparison of the calibration statistics obtained to estimate acid insoluble ash and total ash contents and in vitro pepsin cellulase digestibility of organic and dry matter. The incidence of spectral curvature effects caused by soil contamination that can be conveniently corrected with pretreatments such as derivatives was confirmed.

Non-linear regression methods in NIRS quantitative analysis.

Due to its speed and precision, near-infrared reflectance spectroscopy (NIRS) has become a widely used analytical technique in many industries. It offers, moreover, a number of other advantages which make it ideal for meeting current demands in terms of control and traceability: low cost per sample analysed; little or no need for sample preparation; ability to analyse a wide range of products and parameters; a high degree of reproducibility and repeatability. NIRS can be built into in-line processes, and - since no reagents are required - produces no waste. However, the major drawback to the use of NIRS for its most traditional application (the generation of prediction equations) is that it is a secondary method, and as such needs to be calibrated using a conventional reference method. For quantitative applications, calibration involves ascertaining the optimum mathematical relationship between spectral data and data provided by the reference method. The model may be fairly complex, since the NIRS spectrum is highly variable and contains physical/chemical information for the sample which may be redundant. As a result, multivariate calibration is required, based on a set of absorption values from several wavelengths. Since the relationship to be modelled is often non-linear, classical regression methods are unsuitable, and more complex strategies and algorithms must be sought in order to model this non-linearity. This overview addresses the most widely used non-linear algorithms in the management of NIRS data.

Use of artificial neural networks in near-infrared reflectance spectroscopy calibrations for predicting the inclusion percentages of wheat and sunflower meal in compound feedingstuffs.

The use of near-infrared reflectance spectroscopy (NIRS) calibrations to predict the ingredient composition in compound feeds (i.e., inclusion percentage of each ingredient) is a complex task, regarding both the nature of the parameters to be predicted, since they are not well-defined chemical entities, and the heterogeneousness of the matrices/formulas in which each ingredient participates. The present paper evaluates the use of nonlinear regression methods, such as artificial neural networks (ANN), for developing NIRS calibrations to predict these parameters. Two of the most representative ingredients in the Spanish compound feed formulations (wheat and sunflower meal) were selected for evaluating ANN possibilities, using a large spectral library comprising a total of 7523 commercial compound feed samples; 7423 were used as training set and 100 as validation set. Three general models of networks were studied: multilayer perceptron with back-propagation training (BP), multilayer perceptron with Levenberg-Maquartd training (LM), and radial basis function nets (RBF); moreover, in accordance with a factorial design, more complex architectures were evaluated gradually, changing the number of hidden layers and hidden neurons, for the determination of the optimal network topology. For both ingredients, the best results were obtained using ANN with BP training, showing prediction error values (SEP) of 2.72% and 0.66% for wheat and sunflower meal, respectively. These SEP values showed a significant improvement (19%-49% for sunflower meal and wheat, respectively) in comparison with those obtained using calibrations developed with linear methods.

Evaluation of pretreatment strategies for near-infrared spectroscopy calibration development of unground and ground compound feedingstuffs.

Chemometric procedures are usually applied to near-infrared (NIR) spectra in order to obtain prediction models. These procedures include the application of different combinations of spectral mathematical pretreatments for the improvement of calibrations and the selection of the best model on the basis of validation results. In this work, we used an automatic routine to obtain calibrations for unground and ground compound feedingstuffs (N=354 samples), including 49 combinations of pretreatments (first and second derivatives, an auto scaling procedure, detrending and two versions of multiplicative scatter correction). Calibrations for crude fiber and crude protein were developed without elimination of outliers and with 2 or 9 maximum passes of elimination of outliers. Validation statistics were highly influenced by the pretreatments used, as a combined result of their ability to improve the detection of outliers and the model adjustment. The standard error of prediction (SEP) values ranged from 0.61 to 1.27 for crude protein (CP) and from 0.74 to 1.33 for crude fiber (CF). In spite of the fact that validation statistics did not show a clear distribution pattern, some combinations of pretreatments provided consistently better results.

Optimization of discriminant partial least squares regression models for the detection of animal by-product meals in compound feedingstuffs by near-infrared spectroscopy.

This paper evaluates two multivariate strategies for classifying near-infrared (NIR) spectroscopic data for the detection of animal by-product meals (henceforth generically termed AbP) as an ingredient in compound feedingstuffs. Classification models were developed to discriminate between the presence and absence of animal-origin meals in compound feeds using two forms of discriminant partial least squares (PLS) regression: the algorithms PLS1 and PLS2. The training set comprised 433 commercial feeds, of which 148 contained AbP and the other 285 were stated to be AbP-free. Since the initial set contained unequal numbers of each class, the effect of this imbalance was analyzed by applying the same algorithms to a training set containing equal numbers of AbP-free and AbP-containing samples. The best classification model (97.42% of samples correctly classified), obtained with PLS2, that showed less sensitivity to the use of class-unbalanced sets, was externally validated using a set of 18 samples (10 AbP-containing and 8 AbP-free); all samples were correctly classified, except for one AbP-free sample that was classified as containing AbP (false positive). The results suggest that the application of PLS discriminant analysis to NIR spectroscopic data enables detection of AbP, a feed ingredient banned since the bovine spongiform encephalopathy (BSE) crisis; this confirms the value of NIRS qualitative analysis for product authentication purposes.

Implementation of LOCAL algorithm with near-infrared spectroscopy for compliance assurance in compound feedingstuffs.

Seven thousand four hundred and twenty-three compound feed samples were used to develop near-infrared (NIR) calibrations for predicting the percentage of each ingredient used in the manufacture of a given compound feedingstuff. Spectra were collected at 2 nm increments using a FOSS NIRSystems 5000 monochromator. The reference data used for each ingredient percentage were those declared in the formula for each feedingstuff. Two chemometric tools for developing NIRS prediction models were compared: the so-called GLOBAL MPLS (modified partial least squares), traditionally used in developing NIRS applications, and the more recently developed calibration strategy known as LOCAL. The LOCAL procedure is designed to select, from a large database, samples with spectra resembling the sample being analyzed. Selected samples are used as calibration sets to develop specific MPLS equations for predicting each unknown sample. For all predicted ingredients, LOCAL calibrations resulted in a significant improvement in both standard error of prediction (SEP) and bias values compared with GLOBAL calibrations. Determination coefficient values (r(2)) also improved using the LOCAL strategy, exceeding 0.90 for most ingredients. Use of the LOCAL algorithm for calibration thus proved valuable in minimizing the errors in NIRS calibration equations for predicting a parameter as complex as the percentage of each ingredient in compound feedingstuffs.