Artificial neural networks in foodstuff analyses: Trends and perspectives A review

Neural network enhanced aging time measurements of diary product remaining with infrared spectroscopy

The determination of the drying degree of food residues on surfaces is an important step before efficient cleaning can be achieved. To accomplish this goal, a rapid evaluation based on a neural network and non-invasive measurement technique is introduced. Two common starch-based products and various yogurts from different manufacturers are used as example contaminants to determine the aging time of dried food residue. Near-infrared spectroscopy serves as a modern and fast measurement technique for investigating food compositions. Two analysis methods were compared for processing the measured near-infrared spectral data. The raw data were analyzed using partial least squares regression in conjunction with necessary preprocessing steps. As an alternative method, three different types of neural networks are employed. The aim of this approach is to compensate for the filtering steps before regression, which are typically necessary for multivariate regression. The challenge is to measure three different types of food and obtain a reliable prediction of moisture content in order to draw conclusions about the drying time. The experiments have shown that simple flat neural networks have similar accuracy compared to conventional regression. The use of a convolutional layer in advance demonstrates a significant improvement in prediction compared to other neural networks and even manages to surpass the accuracy of PLS regression. A network with a convolutional layer can also compensate for the sometimes strong variations between food types.

Kinetic Determination of Acetylsalicylic Acid Using a CdTe/AgInS2 Photoluminescence Probe and Different Chemometric Models

The combination of multiple quantum dots (QDs) in a multi-emitter nanoprobe can be envisaged as a promising sensing scheme, as it enables obtaining a collective response of individual emitters towards a given analyte and allows for achieving specific analyte-response profiles. The processing of these profiles using adequate chemometric methods empowers a more sensitive, reliable and selective determination of the target analyte. In this work, we developed a kinetic fluorometric method consisting of a dual CdTe/AgInS₂ quantum dots photoluminescence probe for the determination of acetylsalicylic acid (ASA). The fluorometric response was acquired as second-order time-based excitation/emission matrices that were subsequently processed using chemometric methods seeking to assure the second-order advantage. The data obtained in this work are considered second-order data as they have a three-dimensional size, I × J × K (where I represents the samples' number, J the fluorescence emission wavelength while K represents the time). In order to select the most adequate chemometric method regarding the obtained data structure, different chemometric models were tested, namely unfolded partial least squares (U-PLS), N-way partial least squares (N-PLS), multilayer feed-forward neural networks (MLF-NNs) and radial basis function neural networks (RBF-NNs).

The Characterization of Dry Fermented Sausages under the "Chorizo Zamorano" Quality Label: The Application of an Alternative Statistical Approach

The characterization of quality brand meat products, such as "Chorizo Zamorano" dry fermented sausages, involves a wide range of data which makes it necessary to use alternative statistical methodologies. In this study, the feasibility of the Categorical Principal Components Analysis as a multivariate non-linear technique for the characterization of "Chorizo Zamorano" was assessed. The data analyzed were those of eight commercial brands covered by the quality mark over an eight-year period (2013-2020) and included parameters of the physicochemical composition and organoleptic properties of the product. The results showed that "Chorizo Zamorano" has an average moisture content (28.28%), high protein (38.38%) and fat (51.05%) contents, and a very low carbohydrate concentration (1.52%). Results showed that the fat and protein content and the sensory parameters related to external and internal odor appeared to be the studied variables with the greatest influence on the classification of the products according to their quality.

Prediction of winter wheat leaf chlorophyll content based on VIS / NIR spectroscopy using ANN and PLSR

Visible-near-infrared spectroscopy is known for its rapid and nondestructive characteristics designed to predict leaf chlorophyll content (LCC) of winter wheat. It is believed that the nonlinear technique is preferable to the linear method. The canopy reflectance was applied to generate the LCC prediction model. To accomplish such an objective, artificial neural networks (ANN), along with partial least squares regression (PLSR), nonlinear, and linear evaluation methods have been employed and evaluated to predict wheat LCC. The wheat leaves reflectance spectra were initially preprocessed using Savitzky-Golay smoothing, differentiation (first derivative), SNV (Standard Normal Variate), MSC (Multiplicative Scatter Correction), and their combinations. Afterward, a model for LCC using the reflectance spectra was developed by means of the PLS and ANN. The vis/NIR spectroscopy samples at the 350-1400 nm wavelength were preprocessed using S. Golay smoothing, D₁, SNV, and MSC. The preprocessing with SNV-S.G, followed by PLS and ANN modeling, was able to achieve the most accurate prediction, with the correlation coefficient of 0.92 and 0.97, along with the root mean square error of 0.9131 and 0.7305 receptivity. The experimental findings also revealed that the suggested method utilizing the PLS and ANN model with SNV-S. G preprocessing was practically feasible to estimate the chlorophyll content of a particular winter wheat leaf area according to the visible and near-infrared spectroscopy sensors, achieving improved precision and accuracy. The nonlinear technique was proposed as a more refined technique for LCC estimating.

Diagnosis of Induced Resistance State in Tomato Using Artificial Neural Network Models Based on Supervised Self-Organizing Maps and Fluorescence Kinetics

The aim of this study was to develop three supervised self-organizing map (SOM) models for the automatic recognition of a systemic resistance state in plants after application of a resistance inducer. The pathosystem Fusarium oxysporum f. sp. radicis-lycopersici (FORL) + tomato was used. The inorganic, defense inducer, Acibenzolar-S-methyl (benzo-[1,2,3]-thiadiazole-7-carbothioic acid-S-methyl ester, ASM), reported to induce expression of defense genes in tomato, was applied to activate the defense mechanisms in the plant. A handheld fluorometer, FluorPen FP 100-MAX-LM by SCI, was used to assess the fluorescence kinetics response of the induced resistance in tomato plants. To achieve recognition of resistance induction, three models of supervised SOMs, namely SKN, XY-F, and CPANN, were used to classify fluorescence kinetics data, in order to determine the induced resistance condition in tomato plants. To achieve this, a parameterization of fluorescence kinetics curves was developed corresponding to fluorometer variables of the Kautsky Curves. SKN was the best supervised SOM, achieving 97.22% to 100% accuracy. Gene expression data were used to confirm the accuracy of the supervised SOMs.

Authentication of Herbal Medicines Based on Modern Analytical Technology Combined with Chemometrics Approach: A Review

Since ancient times, herbal medicines (HMs) have been widely popular with consumers as a "natural" drug for health care and disease treatment. With the emergence of problems, such as increasing demand for HMs and shortage of resources, it often occurs the phenomenon of shoddy exceed and mixing the false with the genuine in the market. There is an urgent need to evaluate the quality of HMs to ensure their important role in health care and disease treatment, and to reduce the possibility of threat to human health. Modern analytical technology is can be analyzed for analyzing chemical components of HMs or their preparations. Reflecting complex chemical components' characteristic curves in the analysis sample, and the comprehensive effect of active ingredients of HMs. In this review, modern analytical technology (chromatography, spectroscopy, mass spectrometry), chemometrics methods (unsupervised, supervised) and their advantages, disadvantages, and applicability were introduced and summarized. In addition, the authentication application of modern analytical technology combined with chemometrics methods in four aspects, including origin, processing methods, cultivation methods, and adulteration of HMs have also been discussed and illustrated by a few typical studies. This article offers a general workflow of analytical methods that have been applied for HMs authentication and explains that the accuracy of authentication in favor of the quality assurance of HMs. It was provided reference value for the development and application of modern HMs.

Liquid chromatographic methods coupled to chemometrics: a short review to present the key workflow for the investigation of wine phenolic composition as it is affected by environmental factors

The guarantee of wine authenticity arises great concern because of its nutritional and economic importance. Phenolic fingerprints have been used as a source of chemical information for various authentication issues, including botanical and geographical origin, as well as vintage age. The local environment affects wine production and especially its phenolic metabolites. Integrated analytical methodologies combined with chemometrics can be applied in wine fingerprinting studies for the determination and establishment of phenolic markers that contain comprehensive and standardized information about the wine profile and how it can be affected by various environmental factors. This review summarizes all the recent trends in the generation of chemometric models that have been developed for treating chromatographic data and have been used for the investigation of critical wine authenticity issues, revealing phenolic markers responsible for the botanical, geographical, and vintage age classification of wines. Overall, the current review suggests that chromatographic methodologies are promising and powerful techniques that can be used for the accurate determination of phenolic compounds in difficult matrices like wine, highlighting the advantages of the applications of supervised chemometric tools over unsupervised for the construction of prediction models that have been successfully used for the classification based on their territorial and botanical origin.

The Evolution of Data-Driven Modeling in Organic Chemistry

Organic chemistry is replete with complex relationships: for example, how a reactant's structure relates to the resulting product formed; how reaction conditions relate to yield; how a catalyst's structure relates to enantioselectivity. Questions like these are at the foundation of understanding reactivity and developing novel and improved reactions. An approach to probing these questions that is both longstanding and contemporary is data-driven modeling. Here, we provide a synopsis of the history of data-driven modeling in organic chemistry and the terms used to describe these endeavors. We include a timeline of the steps that led to its current state. The case studies included highlight how, as a community, we have advanced physical organic chemistry tools with the aid of computers and data to augment the intuition of expert chemists and to facilitate the prediction of structure-activity and structure-property relationships.

Shelf-Life Prediction of Glazed Large Yellow Croaker (Pseudosciaena crocea) during Frozen Storage Based on Arrhenius Model and Long-Short-Term Memory Neural Networks Model

In this study, the changes in centrifugal loss, TVB-N, K-value, whiteness and sensory evaluation of glazed large yellow croaker were analyzed at −10, −20, −30 and −40 °C storage. The Arrhenius prediction model and long-short-term memory neural networks (LSTM-NN) prediction model were developed to predict the shelf-life of the glazed large yellow croaker. The results showed that the quality of glazed large yellow croaker gradually decreased with the extension of frozen storage time, and the decrease in quality slowed down at lower temperatures. Both the Arrhenius model and the LSTM-NN prediction model were good tools for predicting the shelf-life of glazed large yellow croaker. However, for the relative error, the prediction accuracy of LSTM-NN (with a mean value of 7.78%) was higher than that of Arrhenius model (with a mean value of 11.90%). Moreover, the LSTM-NN model had a more intelligent, convenient and fast data processing capability, so the new LSTM-NN model provided a better choice for predicting the shelf-life of glazed large yellow croaker.

Comparison of freshness prediction method for salmon fillet during different storage temperatures

BACKGROUND

Many new forecasting models have been applied to fish freshness prediction like support vector regression (SVR) and radial basis function neural network (RBFNN). In this study, RBFNN, SVR, and Arrhenius models were established and compared for predicting and evaluating the quality of salmon fillets during storage at different temperatures, based on thiobarbituric acid (TBA), total volatile basic nitrogen (TVB-N), total viable counts (TVCs), K value, and sensory assessment (SA).

RESULTS

The TBA, TVB-N, TVC, and K values increased during storage whereas SA decreased. Residuals of the three models are random and irregular, indicating that these models were suitable for predicting the freshness of salmon fillets. The RBFNN predicted quality of salmon fillets stored at different temperatures with relative errors all within ±5% (except for the TVC value at day 6). Relative errors of the SVR model for predicting TVB-N and K value were within 10%, while the relative errors of the Arrhenius model fluctuated greatly (ranging from ±0.46 to ±38.29%) and most of it exceeded 10%. RBFNN model had the best predictive performance by comparing the residual and relative errors of the three models.

CONCLUSION

RBFNN is a promising method for predicting the freshness of salmon fillets stored at -2 to 10 °C in the cold chain. © 2021 Society of Chemical Industry.

Classification and Prediction of Bee Honey Indirect Adulteration Using Physiochemical Properties Coupled with K-Means Clustering and Simulated Annealing-Artificial Neural Networks (SA-ANNs)

The higher demand and limited availability of honey led to different forms of honey adulteration. Honey adulteration is either direct by addition of various syrups to natural honey or indirect by feeding honey bees with sugar syrups. Therefore, a need has emerged for reliable and cost-effective quality control methods to detect honey adulteration in order to ensure both safety and quality of honey. In this study, honey is adulterated by feeding honey bees with various proportions of sucrose syrup (0 to 100%). Various physiochemical properties of the adulterated honey are studied including sugar profile, pH, acidity, moisture, and color. The results showed that increasing sucrose syrup in the feed resulted in a decrease in glucose and fructose contents significantly, from 33.4 to 29.1% and 45.2 to 35.9%, respectively. Sucrose content, however, increased significantly from 0.19 to 1.8%. The pH value increased significantly from 3.04 to 4.63 with increase in sucrose feed. Acidity decreased slightly but nonsignificantly with increase in sucrose feed and varied between 7.0 and 4.00 meq/kg for 0% and 100% sucrose, respectively. Honey’s lightness (L value) also increased significantly from 59.3 to 68.84 as sucrose feed increased. Other color parameters were not significantly changed by sucrose feed. K-means clustering is used to classify the level of honey adulteration by using the above physiological properties. The classification results showed that both glucose content and total sugar content provided 100% accurate classification while pH values provided the worst results with 52% classification accuracy. To further predict the percent honey adulteration, simulated annealing coupled with artificial neural networks (SA-ANNs) was used with sugar profile as an input. RBF-ANN was found to provide the best prediction results with SSE = 0.073, RE = 0.021, and overall R2 = 0.992. It is concluded that honey sugar profile can provide an accurate and reliable tool for detecting indirect honey adulteration by sucrose solution.

Advanced multivariate techniques for the classification and pollution of marine sediments due to aquaculture

Aquaculture production has globally increased and its environmental impact is not well understood and assessed yet. Therefore, in this work nine metals and metalloids (Cu, Cd, Pb, Hg, Ni, Fe, Mn, Zn and As) and three nutrients (P, N and C) that seem to accumulate in marine sediments, were determined under the fish cages (zero distance) and about 50 and 100 m away from them, in three aquacultures in Greece. The analysis of these data is crucial due to the negative impact of the intensive aquaculture activities on fish population, human health and marine environment. This study investigated the environmental impact associated with aquaculture cages on marine sediments, using Supervised Artificial Neural Networks (ANNs) in parallel with Classification Trees (CTs). Optimised models were constructed in order to detect the significance of each variable, predict the origin of the sediment samples and successfully visualise their results. Three popular ANN architectures, as multi-layer perceptrons (MLPs), radial basis function (RBF) and counter propagation artificial neural networks (CP-ANNs) were used to assess the impact of the intensive aquaculture activities on marine sediments. In addition, more traditional multivariate chemometric techniques like CTs were applied to the same data set for comparison purposes. The modelling study showed that P, N, Cu, Cd were the most critical (and polluting) factors of those metals studied. Moreover, single-element models achieved elevated predictive percentages. The results were justified due to the usual practices used for fish feeding or cages maintenance.

Comprehensive study on applications of artificial neural network in food process modeling

Artificial neural network (ANN) is a simplified model of the biological nervous system consisting of nerve cells or neurons. The application of ANN to food process engineering is relatively novel. ANN had been employed in diverse applications like food safety and quality analyses, food image analysis, and modeling of various thermal and non-thermal food-processing operations. ANN has the ability to map nonlinear relationships without any prior knowledge and predicts responses even with incomplete information. Every neural network possesses data in the form of connection weights interconnecting lines between the input to hidden layer neurons and weights of hidden to output layer neurons, which has a significant role in predicting the output data. The applications of ANN in different unit operations in food processing were described that includes theoretical developments using intelligent characteristics for adaptability, automatic learning, classification, and prediction. The parallel architecture of ANN resulted in a fast response and low computational time making it suitable for application in real-time systems of different food process operations. The predicted responses obtained by the ANN model exhibited high accuracy due to lower relative deviation and root mean squared error and higher correlation coefficient. This paper presented the various applications of ANN for modeling nonlinear food engineering problems. The application of ANN in the modeling of the processes such as extraction, extrusion, drying, filtration, canning, fermentation, baking, dairy processing, and quality evaluation was reviewed.HIGHLIGHTS1. This paper discusses application of ANN in different emerging trends in food process.2. Application of ANN to develop non-linear multivariate modeling is illustrated.3. ANNs have been shown to be useful tool for prediction of outcomes with high accuracy.4. ANN resulted in fast response making it suitable for application in real time systems.

Insights into chemical and sensorial aspects to understand and manage beer aging using chemometrics

Beer chemical instability remains, at present, the main challenge in maintaining beer quality. Although not fully understood, after decades of research, significant progress has been made in identifying "aging compounds," their origin, and formation pathways. However, as the nature of aging relies on beer manufacturing aspects such as raw materials, process variables, and storage conditions, the chemical profile differs among beers. Current research points to the impact of nonoxidative reactions on beer quality. The effect of Maillard and Maillard intermediates on the final beer quality has become the focus of beer aging research, as prevention of oxidation can only sustain beer quality to some extent. On the other hand, few studies have focused on tracing a profile of whose compound is sensory relevant to specific types of beer. In this matter, the incorporation of "chemometrics," a class of multivariate statistic procedures, has helped brewing scientists achieve specific correlations between the sensory profile and chemical data. The use of chemometrics as exploratory data analysis, discrimination techniques, and multivariate calibration techniques has made the qualitatively and quantitatively translation of sensory perception of aging into manageable chemical and analytical parameters. However, despite their vast potential, these techniques are rarely employed in beer aging studies. This review discusses the chemical and sensorial bases of beer aging. It focuses on how chemometrics can be used to their full potential, with future perspectives and research to be incorporated in the field, enabling a deeper and more specific understanding of the beer aging picture.

A Machine Learning Method for the Detection of Brown Core in the Chinese Pear Variety Huangguan Using a MOS-Based E-Nose

The brown core is an internal disorder that significantly affects the palatability and economic value of Chinese pears. In this study, a framework that includes a back-propagation neural network (BPNN) and extreme learning machine (ELM) (BP-ELMNN) was proposed for the detection of brown core in the Chinese pear variety Huangguan. The odor data of pear were collected using a metal oxide semiconductor (MOS) electronic nose (E-nose). Principal component analysis was used to analyze the complexity of the odor emitted by pears with brown cores. The performances of several machine learning algorithms, i.e., radial basis function neural network (RBFNN), BPNN, and ELM, were compared with that of the BP-ELMNN. The experimental results showed that the proposed framework provided the best results for the test samples, with an accuracy of 0.9683, a macro-precision of 0.9688, a macro-recall of 0.9683, and a macro-F1 score of 0.9685. The results demonstrate that the use of machine learning algorithms for the analysis of E-nose data is a feasible and non-destructive method to detect brown core in pears.

Comparison of Chemometric Problems in Food Analysis Using Non-Linear Methods

Food analysis is a challenging analytical problem, often addressed using sophisticated laboratory methods that produce large data sets. Linear and non-linear multivariate methods can be used to process these types of datasets and to answer questions such as whether product origin is accurately labeled or whether a product is safe to eat. In this review, we present the application of non-linear methods such as artificial neural networks, support vector machines, self-organizing maps, and multi-layer artificial neural networks in the field of chemometrics related to food analysis. We discuss criteria to determine when non-linear methods are better suited for use instead of traditional methods. The principles of algorithms are described, and examples are presented for solving the problems of exploratory analysis, classification, and prediction.

Prediction of Degreening Velocity of Broccoli Buds Using Hyperspectral Camera Combined with Artificial Neural Networks

Developing a noninvasive technique to estimate the degreening (loss of green color) velocity of harvested broccoli was attempted. Loss of green color on a harvested broccoli head occurs heterogeneously. Therefore, hyperspectral imaging technique that stores spectral reflectance with spatial information was used in the present research. Using artificial neural networks (ANNs), we demonstrated that the reduction velocity of chlorophyll at a site on a broccoli head was related to the second derivative of spectral reflectance data at 15 wavelengths from 405 to 960 nm. The reduction velocity was predicted using the ANNs model with a correlative coefficient of 0.995 and a standard error of prediction of 5.37 × 10^-5 mg·g^-1·d^-1. The estimated reduction velocity was effective for predicting the chlorophyll concentration of broccoli buds until 7 d of storage, which was established as the maximum time for maintaining marketability. This technique may be useful for nondestructive prediction of the shelf life of broccoli heads.

Exploring the Scientific Interest for Olive Oil Origin: A Bibliometric Study from 1991 to 2018

The authenticity and traceability of olive oils have been a growing concern over the past decades, generating numerous scientific studies. This article applies the tools of bibliometric analyses to explore the evolution and strategic orientation of the research focused on olive oil geographical and varietal origins. A corpus of 732 papers published in 178 different journals between 1991 and 2018 was considered. The most productive journals, authors and countries are highlighted, as well as the most cited articles associated with specific analytical techniques. A cluster analysis on the keywords generates 8 main themes of research, each focused on different analytical techniques or compounds of interest. A network between these thematic clusters and the main authors indicates their area of expertise. The metabolomics methods are drawing increasing interest and studies focused on the relationships between the origin and the sensory or nutritional properties provided by minor compounds of olive oils appear to be future lines of research.

Alternative data mining/machine learning methods for the analytical evaluation of food quality and authenticity - A review

In recent years, the variety and volume of data acquired by modern analytical instruments in order to conduct a better authentication of food has dramatically increased. Several pattern recognition tools have been developed to deal with the large volume and complexity of available trial data. The most widely used methods are principal component analysis (PCA), partial least squares-discriminant analysis (PLS-DA), soft independent modelling by class analogy (SIMCA), k-nearest neighbours (kNN), parallel factor analysis (PARAFAC), and multivariate curve resolution-alternating least squares (MCR-ALS). Nevertheless, there are alternative data treatment methods, such as support vector machine (SVM), classification and regression tree (CART) and random forest (RF), that show a great potential and more advantages compared to conventional ones. In this paper, we explain the background of these methods and review and discuss the reported studies in which these three methods have been applied in the area of food quality and authenticity. In addition, we clarify the technical terminology used in this particular area of research.

Simultaneous sensitive determination of benzenediol isomers using multiwall carbon nanotube–ionic liquid modified carbon paste electrode by a combination of artificial neural network and fast Fourier transform admittance voltammetry

A novel electrochemical method for the simultaneous determination of catechol, hydroquinone, and resorcinol.

Cultivar classification of Apulian olive oils: Use of artificial neural networks for comparing NMR, NIR and merceological data

The development of an efficient and accurate method for extra-virgin olive oils cultivar and origin authentication is complicated by the broad range of variables (e.g., multiplicity of varieties, pedo-climatic aspects, production and storage conditions) influencing their properties. In this study, artificial neural networks (ANNs) were applied on several analytical datasets, namely standard merceological parameters, near-infra red data and ¹H nuclear magnetic resonance (NMR) fingerprints, obtained on mono-cultivar olive oils of four representative Apulian varieties (Coratina, Ogliarola, Cima di Mola, Peranzana). We analyzed 888 samples produced at a laboratory-scale during two crop years from 444 plants, whose variety was genetically ascertained, and on 17 industrially produced samples. ANN models based on NMR data showed the highest capability to classify cultivars (in some cases, accuracy>99%), independently on the olive oil production process and year; hence, the NMR data resulted to be the most informative variables about the cultivars.

Sensitivity, Prediction Uncertainty, and Detection Limit for Artificial Neural Network Calibrations

With the proliferation of multivariate calibration methods based on artificial neural networks, expressions for the estimation of figures of merit such as sensitivity, prediction uncertainty, and detection limit are urgently needed. This would bring nonlinear multivariate calibration methodologies to the same status as the linear counterparts in terms of comparability. Currently only the average prediction error or the ratio of performance to deviation for a test sample set is employed to characterize and promote neural network calibrations. It is clear that additional information is required. We report for the first time expressions that easily allow one to compute three relevant figures: (1) the sensitivity, which turns out to be sample-dependent, as expected, (2) the prediction uncertainty, and (3) the detection limit. The approach resembles that employed for linear multivariate calibration, i.e., partial least-squares regression, specifically adapted to neural network calibration scenarios. As usual, both simulated and real (near-infrared) spectral data sets serve to illustrate the proposal.