Open access peer-reviewed chapter

Applications of Near-Infrared Spectroscopy (NIRS) in Fish Value Chain

Written By

Sonia Nieto-Ortega, Rebeca Lara, Giuseppe Foti, Ángela Melado-Herreros and Idoia Olabarrieta

Submitted: 10 May 2022 Reviewed: 07 June 2022 Published: 16 July 2022

DOI: 10.5772/intechopen.105736

From the Edited Volume

Infrared Spectroscopy - Perspectives and Applications

Edited by Marwa El-Azazy, Khalid Al-Saad and Ahmed S. El-Shafie

Chapter metrics overview

172 Chapter Downloads

View Full Metrics

Abstract

Near-infrared spectroscopy (NIRS) has undergone a significant evolution in the last years due to the numerous scientific studies that revealed its potential for industrial applications, attracting a growing interest in the food sector. Furthermore, new advances have allowed the reduction in size and cost of the NIR devices, making them appropriate for on-site determinations. The complex structure of the fish value chain, combined to its high market value, makes this sector particularly vulnerable to fraud and adulteration practices. Also, the perishable nature of fish and fish products, as well as the lack of traceability, arises the urgent need for a fast, reliable and portable tool capable of precisely characterizing the quality and authenticity of the product while also ensuring its safety. In this chapter, the capabilities of NIRS combined to several chemometric techniques for quality, authenticity and safety applications are presented through an extensive review of the most recent research works.

Keywords

  • NIR applications
  • food testing and analysis
  • spectral analysis
  • industrial applications
  • fish chain
  • chemometrics

1. Introduction

1.1 Principles of NIR spectroscopy and chemometrics

Near-infrared spectroscopy (NIRS) is a technique that measures the absorption of the electromagnetic region from 750 nm to 2500 nm (Figure 1a), between the visible and the mid-infrared (MIR) region. Like MIR spectroscopy and some part of the far-infrared (FIR) spectroscopy, this technique detects molecular vibrations, which gives information about the different chemical bonds within the molecule and its environment [1, 2].

Figure 1.

a) Electromagnetic spectrum of UV-vis-IR. b) Vibrational transitions involved.

However, there are some differences between these spectroscopic techniques, based on the nature of the vibrational transitions. While in the MIR/FIR region (2500 nm–50,000 nm), almost all fundamental transitions (v0 → v1) can be observed, resulting in sharp and intense spectra, in the near-infrared (NIR) region (750 nm–2500 nm) only the much less intense overtones (v0 → vn) of those transitions appear. NIR absorptions are based on overtones and combination bands and, due to their lower transition probabilities, intensities usually decrease by a factor of 10 or 100 from the fundamental overtone to the next one (Figure 1b) [1, 3].

It could be logical to think that it would be better to use MIR bands (fundamental transitions) instead of NIR. However, NIR instruments are simpler and more affordable. Furthermore, NIR has a deeper penetration in skin and organic tissues (optical window) than MIR, so that a larger area or volume can be measured [4].

While MIR chemical information can be directly obtained from peak intensities, shape and energy, NIR spectra need special treatment to extract useful information that involves multivariate analysis of data, usually called chemometrics [5]. Chemometrics is the science that applies mathematics and statistical methods to process chemical data, in order to obtain useful information about the sample [6, 7]. These two sciences are closely related: NIRS would not have evolved without chemometrics, and chemometric algorithms usually use NIRS examples to illustrate their power. This process needs computational resources that only since the last few decades have been available. In NIR applications, chemometrics covers a wide range of multivariate methods, which involve preprocessing techniques, as well as qualitative and quantitative analysis [8].

NIRS data usually need to be preprocessed before the model construction, because the spectra usually contain plenty of noise and background information that must be filtered out to find the desired spectral signatures. Some of the most used preprocessing techniques are divided in: i) scatter correction methods and ii) spectral derivatives. The scatter correction methods include techniques such as multiplicative scatter correction (MSC), detrending, normalization, or standard normal variate (SNV). Regarding the spectral derivatives, the Savitzky-Golay (SG) derivative after smoothing is one of the most used [9]. Chemometric methods can be divided into unsupervised, when groups or patterns inside data are not known, and supervised (classification and regression) when categories or analytes are already quantified, and the objective is to predict them for new samples [10].

The rapid development of chemometric methods during the last decades has allowed the application of NIRS to different areas of food science [11], such as food fraud and authenticity [12, 13] and food quality [14, 15].

1.2 Fish value chain

The demand for fish consumption has increased during the last years as a consequence of the better dietary habits of the population. Fish has become one of the most valuable foods, due to its higher protein content and the healthier fat profile that it has compared with meat [16]. However, fish and fish products present complex problematics in terms of food quality, authenticity and safety. In addition, a greater awareness of food quality and sustainability has led consumers to demand increasingly reliable information about the products purchased in the market, which has stimulated the food value chain to strengthen efforts in the quest for a better and more trusty traceability [17].

Fisheries and aquaculture production are very heterogeneous in terms of species and fish products and, due to its high economic value, are one of the most vulnerable targets to adulteration and fraud. These practices are favoured by the complex structure of the supply chain (Figure 2), the lack of transparency and the high percentage of the production destined to the elaboration of processed fish products (around the 50%) such as fillets, portions or elaborated products, in which the morphological characteristics are absent, making these frauds particularly elusive [18]. Furthermore, fish is a perishable product that may suffer a fast enzymatic decomposition and microbial spoilage compared with other foods [19]. This means that small changes in the preservation conditions along the value chain, such as temperature or salt level variations, can have a high impact on the fish quality and safety, resulting in a fast degradation, higher food waste and the risk of intoxications.

Figure 2.

Flow chart of seafood value chain from catch to consumer, indicating the different points where the quality of the fish should be monitored. ) Monitoring of chain conditions and quality attributes; ) inspection technologies and quality control.

On the market side, consumers are demanding more transparency in the labelling of seafood. To ensure the prevention and prompt detection of illegal activities and the access to all the information about the nature, origin and characteristics of the fish, monitoring of all the processes and quality analysis must be done thoroughly through the whole production chain and market, until it arrives to the consumers [20, 21]. The implementation of reliable methods to ensure a complete and correct traceability of the product and to avoid fraud is of primary interest for both industry and consumers.

1.3 NIRS in fish analysis

Nowadays, the quality, authenticity and safety control procedures in fish industry are carried out using traditional laboratory techniques, making virtually impossible any real-time application. However, some rapid techniques, such as NIRS, have shown the capacity of obtaining valuable information in a rapid way, which is important for the fish chain analysis [22].

NIRS is a tool already well developed in agriculture [23] and other products [24, 25]. The capacity of this technique to detect vibrations of molecules with polar bonds, such as those with hydrogen as C-H (aliphatic compounds, fat, oils, proteins), N-H (proteins) or O-H (water, alcohols, acids), is especially useful to determine the composition of organic matter. This makes this technique suitable for the quantification of the chemical composition of samples with a high percentage of fat, protein or water content and allows the determination of other parameters that depend on many chemical and physical properties such as the species [26], freshness, storage conditions (unfrozen or frozen-thawed fish) [27] or even non-polar analytes in low concentrations through indirect ways.

For all the previously mentioned reasons, NIRS, coupled with chemometrics, is starting to be considered as essential as other analytical methods for fish analysis. Indeed, the number of Web of Science indexed articles in which NIRS techniques have been applied to determine fish quality, safety or authenticity have increased rapidly during the last 20 years. While this technique is mostly used for fish quality control and authenticity, applications such as fish safety have begun to be explored in the last 10 years (Figure 3).

Figure 3.

Number of publications since 2001 regarding NIR and fish. Data extracted from Web of Science in April 2022. The keywords used in topic, common for all the sections, were ‘fish’ AND ‘NIR’ OR ‘near infrared spectroscopy’. For quality control, two independent searches were made: i) ‘composition’ OR ‘fat protein’ AND ‘quality’ ii) ‘freshness’ AND ‘storage’ NOT ‘safety’. For authenticity, four independent searches were made: i) ‘specie*’ AND ‘identification’ OR ‘substitution’ OR ‘authentication’ ii) ‘origin’ iii) ‘farm* wild’ OR ‘production method*’ iv) ‘fresh thawed’ OR ‘unfrozen thawed’ OR ‘fresh frozen’ OR ‘unfrozen frozen’. For safety, two independent searches were made: i) ‘spoilage’ OR ‘toxic’ OR ‘*amine*’ AND ‘safety’ ii) ‘micro*’ OR ‘bacteria’ AND ‘safety’.

Advertisement

2. Applications of NIRS in fish chain

While writing this chapter, the authors reviewed the use of NIRS in the fish chain regarding three different application fields: fish quality control, fish authenticity and fish safety (Figure 4). The supplementary material, which shows a summary of the works that will be reviewed in this chapter, is available in the Appendix.

Figure 4.

Classification of the problematics considered and reviewed under this chapter.

2.1 Fish quality control

Assuring the quality of products is one of the objectives that any industry must consider, and it is of major importance in the fish chain, due to the problems that are usually associated with its evaluation at an industrial level [28]. Unlike other foodstuff, quality of fish and fish products is more difficult to control, due to several factors, such as the variations in species, sex, age or habitats [29]. In this sense, fish quality control refers specially to the determination of the proximate composition of the fish samples and the evaluation of the freshness of the products. From one side, the measurement of the proximate composition of fishery species is very important, since it varies from one species and one individual to another. Moreover, it also gives an idea of the sexual stage of the fish [30], and it changes depending on many factors (diet, origin, rearing systems). The proximate composition determines the organoleptic quality of a specific fish product [31]. At the same time, the determination of the freshness of samples is another indicator of the quality of the product. Freshness is an ambiguous term, which can be interpreted in different ways, and it should not be used unless it is properly defined. Through this chapter, the concept of freshness is going to be a quality descriptor of fish samples, and it will refer to the time past after the catch of the fish. If that time is short, the sample will be considered fresh, as it will retain its original characteristics. As time passes, samples will lose quality, due to the action of biochemical, chemical, physical and microbial processes, and therefore losing freshness [32]. Nowadays, NIRS is becoming an alternative as a quality control method, due to its advantages over traditional analysis [33].

2.1.1 Proximate composition

Fish is mainly composed of water, protein, fat and other minoritarian compounds [28]. As fish proximate composition is a quantitative analysis, regression techniques (usually partial least square regression or PLSR) are used to determine the composition of fish by NIRS, as an alternative to laboratory determinations.

Many authors have determined the proximate composition of different fish species using NIRS. In one of the first studies, NIRS was used to determine moisture, fat and protein content in Atlantic salmon (Salmo salar) fillets and in minced samples [34]. Principal component regressions (PCRs) were used to perform the calibrations, which were evaluated using a full cross-validation. This study illustrated the potentiality of NIRS for predicting the proximate composition of minced fish, obtaining with the best models a root mean square error of cross-validation (RMSECV) of 6.6, 3.8 and 3.0 g/Kg in fat, moisture and protein analysis, respectively. However, they failed in the analysis of the whole fish, with RMSECV higher than 13.2 g/Kg in all the models. Years later, another study performed PLSR directly in fish fillets of Chinese export tilapia [31]. The regressions showed good results for the determination of moisture and lipid content, with r2 = 0.96 and 0.97 and standard error of cross-validation or SECV = 0.87 and 0.71, respectively. Even though the regression made on the protein content had a lower coefficient of determination in the cross-validation (r2 = 0.31), it showed good results regarding the error (SECV = 0.55). This difference was due to the small protein differences that tilapia fillets have between them, and it could be solved adding samples with more variability to the calibration set. It is important to highlight that they performed an external validation and not only a cross-validation, for moisture and protein determination, using new model-outside samples, which presented variations that authors considered acceptable between predicted and measured values (a maximum relative error of 1.67% for moisture and 6.48% for lipid). Similar studies regarding the analysis of the proximate composition in fish samples are shown in Table 1.

Fish speciesSample presentationMeasured parametersAlgorithmResultsReference
Atlantic salmon (Salmo salar)WholeFat and moisturePLSRr2 = 0.87, 0.86; RMSECV = 1.12, 0.98%, respectively[35]
Sea bass (Dicentrarchus labrax)Fillet portions, whole fresh minced fillet, freeze-dried minced filletWater, ether extract and crude proteinPLSRSECV <2.81% in all the cases[36]
Sea bass (Dicentrarchus labrax)Fresh and freeze-dried minced filletsWater, ether extract and crude proteinPLSRSECV ≤0.39% in all the cases[37]
Skipjack tuna (Katsuwonus pelamis) and yellow fin (Thunnus albacares)Fish fleshMoisture, protein, free and total fatPLS-2r2 > 0.90 in all, small deviations between reference and predicted values in the cross-validation[38]
Nile tilapia’s (Oreochromis niloticus)FilletsFat content and FAsPLSRcorrelation coefficient = 0.70, 0.71; (RMSEP) = 2.39, 4.76 mg FA/g of total lipids, respectively[39]
Saithe (Pollachius virens) and hoki (Macruronus novaezelandiae)Thawed fish piecesTotal lipid content, phospholipids, (PUFAs)* and (MUFAs)PLSRr2 ≥ 0.73, 0.96. The highest RMSEP = 5.31, 4.10%, respectively.[33]
Cured salmon roe (Oncorhynchus keta), cured Atlantic salmon (Salmo salar) and cold smoked Atlantic salmon (S. salar).Roe and salmon piecesMoisture and sodium chloridePCR, MLS, PLSR and different neural network algorithmsr2 ≥ 0.71 and standard error ≤ 1.87%, r2 ≥ 0.72 and standard error ≤ 2.04%, r2 ≥ 0.64 and standard error ≤ 4.10% in the cross-validation for both components, respectively[40, 41, 42]

Table 1.

Use of NIRS to analyse proximate composition of fish.

PUFAs: polyunsaturated fatty acids; MUFAs: monounsaturated fatty acids, FAs: fatty acids; MLR multiple linear regression; RMSEP: root mean squared error of prediction.


2.1.2 Freshness

In this chapter, freshness has been considered as a quality characteristic, and it will be defined as the time passed after the catch. It is also influenced by the temperature during the storage [32]. Freshness is usually evaluated using sensory analyses, which consider several parameters such as flesh elasticity, pupil colour and skin odour, among others [43].

Although principal component analysis (PCA) is not a classificatory nor a discriminant method (since it is a non-supervised technique, which only takes into account the spectral information), and it should not be used with that purpose, some authors have applied it in order to see the differences between fresh fish and fish samples that have been stored in the fridge for days. In this sense, some authors used short-wavelength near infrared spectroscopy (SW-NIRS) to differentiate fresh from non-fresh rainbow trout (Oncorhynchus mykiss) samples, in fillets and minced [44]. Fish samples were stored at 4°C for 8 days, or at room temperature (21°C) for 24 hours. PCA results showed a clear separation between control samples (1 day stored) and fish kept 4 days or longer in refrigeration. Regarding fish at room temperature, a segregation was found for samples stored 10 hours or longer in comparison to control samples (stored 0 hours). Other studies performed PCA to discriminate between fresh Atlantic salmon (Salmo salar) and salmon stored at 4°C for 9 days, finding a clear separation between fresh and non-fresh fillets [45]. A similar experiment used PCA to discriminate golden pompano (Trachinotus ovatus) samples regarding the storage time using visible/near-infrared spectroscopy (VIS/NIRS) [46]. However, they concluded that samples stored for 0, 2 and 3 days could not be well discriminated using this analysis. They also created models to predict the storage time, but the model which used VIS/NIRS spectroscopy was not validated (only a model combining VIS/NIRS data and electronic nose information was validated, obtaining an accuracy of 93.3%). Another study analysed the spectra of rohu fish (Labeo rohita) sliced into pieces, using SW-NIRS and PCA to distinguish between fresh fish (after being collected) and fish stored in the fridge at 4°C during 4, 7 and 12 days [47]. In the PCA analysis, samples were well grouped by their freshness, so that four clusters were created (one per each measurement day).

Different studies aim not only to separate fresh samples from non-fresh ones, but also to make models to correlate spectral data with the storage time. One of them explored the suitability of SW-NIRS to estimate the freshness of the fillets of two fish species: cod (Gadus morhua) and salmon (Salmo salar) [32]. PLSR models were created to predict the number of days that the fish has been stored and validated using a full cross-validation. Cod gave the best correlation with the visible wavelength range (r2 = 0.97, RMSE = 1.04 days) and salmon with the NIR range (correlation of 0.98 and RMSE = 1.20 days). A similar study predicted the freshness in cod (Gadus morhua) fillets with a handheld interactance probe, using VIS/NIRS. Fillets were stored in a cold room (between 2 and 4°C) for 13 days, making measurements at days 0, 2, 4, 7, 10 and 13 [48]. PLSR models were validated using a leave-one-out cross-validation. The best results were obtained selecting some wavelengths of the visible range (448, 487, 606 and 646 nm) and using SNV, giving a correlation coefficient of 0.93 and an RMSECV = 1.66 days. Other researchers used VIS/NIRS to perform a similar study on Atlantic salmon (Salmo salar L.) fillets [49]. In those, PLSR models using the wavelength region between 605 and 965 nm and the pre-treatment with SNV gave the lowest error in the leave-one-out cross-validation (RMSECV = 1.91 days). In another work the freshness of cod (Gadus morhua) was predicted using VIS/NIRS [50]. Authors built PLSR models to predict the quality index method (QIM), a sensory analysis made to determine the freshness of cod fillets. A full cross-validation was used to test the performance of the models, giving the analysis made with the visible part of the spectra the best results (correlation coefficients = 0.96 and RMSE = 2.6 points).

2.2 Fish authenticity

The globalization and expansion of the fish and aquaculture sector, in addition to the increasing public concern about food quality, have caused a growing interest in several issues related to fish authenticity. According to the European Regulation (EU) n. 1379/2013 [51], fishery and aquaculture products must be labelled with the commercial designation, proper scientific name of the species, production method (e.g. caught, farmed), fishing gear (e.g. hook, trap, trawl), catch or production area and storage method (unfrozen or frozen-thawed). Other claims, such as environmental and production techniques, can also be reported on the label, and all this information must be verified by effective methods. Special mention is the differentiation of farmed from wild-produced seafood, which is becoming a concern among consumers, since the quality and the price may be affected by the production method. The presence of errors in the label information about fish origin and production process is increasing, so that nowadays fish is the second most vulnerable category of food to fraud [52]. In order to protect consumers and with the aim to avoid it, the assessment of seafood origin is increasing as a security measure. Regulatory interventions are trying to stop the mislabelling or the substitution of wild with farmed fish, which is mitigating risks for the consumer’s confidence and health [53].

Both the geographical origin and the production method, among others, can strongly affect the characteristics of the two types of products, whose discriminating properties are usually difficult to determine. Several analytical techniques have been traditionally used to assess fish authenticity. However, even though they are well established, there is still a necessity for faster, easier and more affordable methods [52].

2.2.1 Species identification

The substitution of valuable species with cheaper ones at any point of the supply chain is one of the most common frauds, sometimes in an unintentional way due to the similarities between species or the use of different names for the same species. However, the differences in the economic value, the exploitation of endangered species, the replacement with poisonous fish and their difficult identification, make the substitution of fish species an extended problem, especially severe after processing, at the retailers and supermarkets [54]. The mentioned frauds particularly affect fish fillets and ready-to-eat products, such as fish patties, which cannot be recognized through the traditional morphological analysis.

The use of NIRS with the objective of discriminating between species has been explored during the last years [52]. For instance, in a preliminary study under industrial conditions, some authors used NIRS coupled with partial least square discriminant analysis (PLS-DA) and linear discriminant analysis (LDA) on PCA scores to discriminate between fishmeal made with different fish species: salmon (Salmo salar), blue whiting (Micromesistius poutassau), mackerel (Scomber scombrus) or herring (Clupea harengus) [55]. They obtained a correct prediction percentage > 80%, tested using cross-validation.

In other studies, authors measured the skin and meat of two different species of cod (winter cod and cod), mullet (red mullet and Atlantic mullet) and trout (samlet and salmon trout) with a handheld NIR spectrometer [56]. The data were analysed using PCA and soft independent model of class analogies (SIMCA), obtaining promising results with the last method. Other authors used Fourier transform NIR (FT-NIR) to distinguish fillets of red mullet (Mullus surmuletus) and plaice (Pleuronectes platessa), fish with higher economic value, from Atlantic mullet (Pseudupeneus prayensis) and flounder (Platichthys flesus flesus), species which are cheaper [57]. LDA and SIMCA were used to classify them, obtaining a sensitivity higher than 70% and a specificity of 100% for FT-NIR and SIMCA, and a 100% correct classification with LDA for both techniques.

The differentiation between different carp species: silver (Hypophthalmichthys molitrix) black (Mylopharyngodon piceus), bighead (Aristichthys nobilis), grass (Ctenopharyngodon idellus) and common carp (Cyprinus carpio) and also crucian (Carassius auratus) and bream (Parabramis pekinensis) has also been studied [58]. Only nine selected wavelengths in the 1000–1799 nm region after the pre-treatment were used. In this case, two models using PCA-LDA and the fast Fourier transformation pre-processing (FFT) before LDA analysis were built. They showed a 100% accuracy, specificity, sensitivity and precision. Fish surimi made from white croaker (Argyrosomus argentatus), hairtail (Trichiurus haumela) and redcoat (Nemipterus virgatus) was successfully classified by using LDA and multiplicative scatter correction (MSC) as pre-treatment. The results of the prediction of the calibration and the validation data sets were 98.5% and 100%, respectively [59].

More recently, an attempt to distinguish Atlantic cod (Gadus morhua) from haddock (Melanogrammus aeglefinus) was made. Raw fillets and patties were measured with a handheld and a benchtop instrument, with the spectra pre-processed using SNV, MSC or SG, and analysed with LDA and SIMCA. LDA with MSC pre-treatment obtained a 100% of correct classifications for both fillets and patties for the external validation set, regardless of the instrument. The SIMCA method also obtained a 100% of successful classification of fillets using SNV and SG (first and second derivative) for the benchtop instrument and using MSC for the portable NIR device. With SIMCA, 100% correct predictions of patties were obtained for all instruments and pre-processing techniques [60].

2.2.2 Farmed vs. wild fish and types of farming

In the last years, there has been an increasing interest from farming industries, retailers and even from consumers in investigating the differences between cultured and wild fish [61]. Most of the studies which differentiate farm from wild fish are centred in fish species such as European sea bass (Dicentrarchus labrax L.), which is one of the most economically important fish species in the whole Mediterranean area [62]. The intense competition among producing countries in the Mediterranean Sea (Turkey and Greece are the largest producing countries and Italy, Spain and France are the world’s leading importers) and the lowering of the market prices are requiring ways to differentiate sea bass quality. The differences in the production methods (farm and wild specimens) and, among farmed individuals, in the rearing systems and the feeding regimens used, may affect the flesh quality. Usually, farmed fish presents lighter fillet colour, softer texture and milder flavour than wild fish [36, 61]. Previous studies aimed to differentiate between farmed and wild sea bass were based on discriminating the two classes depending on their protein and amino acids composition, metallic ions and fatty acids groups, being usually the lipid profile the differential factor. However, the extraction of lipids and the transformation of the triglycerides into their corresponding fatty acid methyl esters (FAMEs) plus the analysis by chromatography with a flame ionization detector (GC-FID) are a tedious and time-consuming methodology, which not only generates toxic waste but can also provoke the oxidation of the polyunsaturated groups [63].

NIRS has shown its potentiality in some studies as a rapid method to classify sea bass according to rearing systems [53]. Some authors used an NIR spectrometer to discriminate between four types of sea bass rearing systems: extensive, semi-intensive, intensive and sea-cage in fresh minced fish and freeze-dried minced samples [36]. They performed a four-cluster SIMCA analysis, which was evaluated using a full cross-validation. The best results were obtained with the freeze-dried minced fillets (83%, 80%, 74% and 83% of samples well classified for extensive, semi-intensive, intensive and sea-cage methods, respectively). Similar results were obtained by [37] in the discrimination of organic or semi-intensive sea bass. Some years later, NIRS was also used to discriminate between farm and wild seabass fillets, which were minced before the analysis [53]. They developed three different approaches: an authentication using measured chemical variables (PLS-DA_mc), an authentication using estimated chemical variables (PLS-DA_ec), an authentication with direct use of spectral data (PLS-DA_NIR) and the WPTER algorithm. The evaluation of the validation data set gave very similar results for all the techniques in the classification of declared wild samples. Regarding the declared farming samples, PLS-DA_mc classified correctly all the samples, while PLS-DA_ec misclassified 2, and PLS-DA_NIR and WPTER misclassified 3 (in all the cases out of 34). All the methods indicated that the most relevant spectral regions in the models were the ones related to the absorbance of the CH, CH2, CH3 and H2O groups, which corresponded to fat, fatty acids, and water content, highlighting that the difference between farm and wild sea bass is very related with the fat/water percentage and the lipid profile. Other authors also studied the differences between sea bass, according to the production method (wild or farmed) and the rearing system used (extensive, semi-intensive or intensive) [62]. In this study, an NIR spectrometer was used to analyse minced and homogenized fillets. In this case, cross-validation and an external validation were made. PCA was used as exploratory analysis and two different orthogonal partial least square-discriminant analysis (OPLS-DA) models were performed. The evaluation of the external validation showed an overall classification rate of a 100% for the production method and 94.44% for the farming system. Looking at the categories, all the farm and wild samples were correctly classified, while regarding the farming system, wild samples presented a classification rate of 100%, extensive samples 66.67%, semi-intensive samples 80.00% and intensive samples 100%. Although most of the studies regarding discrimination between farm and wild fish are performed in sea bass, other species have also been analysed. Segregation between wild and farm sole (Solea solea) and turbot (Psetta maxima) has also been performed [64]. Here, several chemometric methods were tested, obtaining the best results with logistic regression in sole (90% of precision, recall and F1-score) and with support vector machine in turbot (99% of precision, recall and F1-score) in the evaluation of the test set.

2.2.3 Geographical origin

NIRS is one of the most used techniques to determine the origin of a lot of types of foods (oil, coffee, wine, etc). However, especially in fish and fish products, its use for such purpose is a challenge due to the difficulty in proving the geographical origin of a particular fish. A very large number of samples must be measured for each location due to a high heterogeneity among them, depending on the season or different years, the available food, if the fish has been captured before or after the spawning season, etc. With the aim of determining the viability of NIRS regarding the geographical origin, it often requires multi-disciplinary methods that could consider all the environmental and genetic factors that determine fish characteristics. Even so, there are some studies that successfully applied NIRS to determine the geographical origin of fish products.

Some authors used NIRS to classify tilapia fillets in four different Chinese regions (Guangdong, Hainan, Guanxi and Fujian provinces) [31]. The classification models were carried out with SIMCA, achieving more than 80% correct classification for the Guangdong, Hainan and Fujian and 75% of the Guangxi samples by the cross-validation method. None of them were assigned to more than one category, less than 20% to none of the classes and only less than 2% were incorrectly classified. The differentiation between minced fish fillets made using European seabass (Dicentrarchus labrax L.) caught in three areas of the Mediterranean Sea has been also explored [62]. In this case, NIRS was used in combination with PLS-DA method. The external validation with a test set resulted in 100% of the eastern, 88% central and 85% western Mediterranean samples were correctly classified. In another study, NIRS was used successfully to assess the geographical origin (Morocco, Spain, Tunisia or Croatia) of minced semi-finished and finished salted anchovies (Engraulis encrasicholus) [65]. The combination of models based on orthogonal PLS-DA gave the correct identification of the origin of both types of anchovies with >98% sensitivity, > 99% specificity and > 99% accuracy in an external validation using a test set with the 20% of the samples. Moroccan samples were distinguished for the proteins and degradation compounds absorption bands, Tunisian anchovies for the unsaturated lipids, while Croatian and Spanish samples were classified for their differences on both types of absorption bands.

2.2.4 Unfrozen vs. frozen-thawed fish

Freezing, due to its advantages, is intensively used for preserving fish. It maintains the product safety, the sensory qualities and also the nutritional characteristics. However, freezing the products, as well as the storage and the thawing process, changes the physical and biochemical properties of the fish muscle [66]. Recent technological advances in freezing and the posterior storage have allowed to minimize the damage caused by the thawing process, while maintaining the product quality. The impact the freezing process does on the product depends on several factors inherent to fish characteristics (genetics, muscle type, husbandry) and the post-mortem treatments (freezing rate, storage time, storage pressure and temperature, thawing method) [67]. Slow freezing leads to the formation of large extracellular ice crystals which damage the cell membranes and muscle proteins [68, 69, 70]. Also, in general, freezing-thawing cycles are related to the degradation of textural properties due to physical damages and protein denaturation and oxidation [66].

As it has been reported before, most of the analytical techniques used for determining whether the fish has been frozen or not, despite their accuracy, are destructive, time-consuming and expensive, making them not suitable for commercial applications [71]. As an alternative, NIRS, in combination with qualitative chemometric methods, has been adopted as an efficient solution for this determination.

Several authors have used NIRS and chemometrics to discriminate between unfrozen and frozen-thawed fish. FT-NIRS has been used, coupled with two different classification algorithms (LDA and SIMCA), to discriminate between unfrozen and thawed Atlantic mullet (Pseudupeneus prayensis) fillets [57]. LDA provided the most accurate results in the evaluation of the external validation set, with correct classification rates of 97.2% for unfrozen and 100% for thawed samples. The same classification algorithms were used in [72] to distinguish whether red sea bream (Pagrus major) samples were unfrozen or frozen-thawed using an NIR spectrophotometer in interactance mode. They observed that the change in total reflectance, the major effect of the freeze-thawing process, arises from changes in the scattered light at the surface of the fish due to alterations in its structure. The classification accuracy when using absorbance spectra without any pre-processing (100%) was much higher than for spectra pre-treated with the multiplicative scatter correction (MSC) (81%), confirming that light scattering contains most of the information. VIS/NIRS techniques were as well used to distinguish unfrozen from thawed swordfish (Xiphias gladius L) cutlets stored at different temperatures (−10 and − 18°C) for several days [73]. VIS/NIR spectra were collected using a portable spectrophotometer and a NIR monochromator. The correct classification percentage for VIS/NIR spectra was ≥96.7%, while for the NIR spectra was ≥90.0%, with no effect on the classification due to the temperature or days that the frozen samples were stored. In another reference, the authors distinguished unfrozen from frozen tuna fillets (Thunnus thynnus) which were frozen and, after 5, 21 and 35 days, thawed and then re-scanned [74]. In this study, the PLS-DA algorithm with double cross-validation was used to classify unfrozen from frozen-thawed fillets with a 92% and 82% of correctly classified samples, respectively. Some researchers have tried to discriminate between tilapia fillets (Oreochromis) that have been subjected to different freezing-thawing cycles, from one to seven [75]. In this project, NIR spectra were collected at each cycle and the test data analysed with PCA and discrimination analysis, obtaining a correct classification of 86% for once frozen-thawed fillets and 93% for samples subjected to several freezing cycles. Finally, in other recent study, unfrozen and frozen-thawed bigeye tuna (Thunnus obesus) fillets were also measured with NIRS for their discrimination [76]. To include variability, samples were injected with different commercially used concentrated solutions (NaCl, polyphosphates and protein hydrolysate). Three different models, based on PLS-DA, were developed giving NIRS the best results with an accuracy of 0.91 and an error rate of 0.10 for the prediction of the validation set.

2.3 Fish safety

Fish and fish products are very perishable, being subject to a fast deterioration process and becoming easily unsuitable for consumption and, therefore, a possible threat to public health. The biological composition and characteristics of fish, i.e. low percentage of connective tissue, autolytic enzymes, neutral pH, high water activity [27] and the growth of spoilage bacteria, contribute to the short shelf-life of these products [45]. It is estimated that 25% of the fish losses are initiated by microbial and chemical deterioration [77].

Even though most studies regarding NIRS in the fish industry have been conducted to determine their quality and authenticity, the use of this technique in food safety evaluation is still relatively new but constantly increasing [25]. During the last years, many studies have been conducted in the field of fish safety, in order to determine the presence of spoilage compounds and microorganism in fish products.

2.3.1 Detection of spoilage compounds

Some toxic compounds produced during fish spoilage can be measured by NIRS to evaluate fish safety. An important group is formed by biogenic amines, with histamine being one of the most critical, as it is related with the most common foodborne illness associated with fish consumption [78]. They are a group of nitrogenous compounds, usually measured to evaluate the safety of aquatic products. Among them, the most common are histamine, putrescine, cadaverine and tyramine. Biogenic amines are formed by the decarboxylation of amino acids or by the amination and transamination of aldehydes and ketones. The consumption of fish products with excessive levels of biogenic amines may lead to food intoxication, so they have become an important index to evaluate fish safety [79].

NIRS has been used as a tool to determine the cadaverine content in homogenized and ultrafiltered solutions of Chub Mackerel (Scomber japonicus) stored at different temperatures (5 and 25°C) [80]. The r2 obtained in the regression analysis was 0.98 and 0.99 for the different dilutions made. However, authors mentioned that NIRS measurements of cadaverine content made directly in the fish were not successful. Histamine was also determined by this spectroscopic technique. It is the only biogenic amine for which international regulatory levels have been established [78]. The presence of this amine is due to the bacterial decarboxylation of histidine, an amino acid present in high levels in fish muscle, specially from Scomberiscida and Scombridae families (which include tuna). Studies have tried to predict this compound in order to assure fish safety. Histamine content has been measured in different dry extracts of skipjack tuna (Katsuwonus pelmis) using NIRS [81]. The standard error of prediction or SEP in all the cases (using different solvents and different pre-processing techniques) was between 2.94 and 3.47 in the validation set, with r2 between 0.63 and 0.79. Later on, in a similar experiment, histamine in raw (Thunnus albacares) and in processed tuna (Katsuwonus pelamis and Thunnus albacares) was detected, which were previously minced and homogenized [78]. Both models showed r2 values higher than 0.97 and RMSEP lower than 10 mg kg−1, with the obtained results being better for processed tuna. Considering the mentioned studies, promising results have been seen in all the works regarding the detection of biogenic amines in fish by NIRS. However, more efforts are still needed to determine these compounds directly in the fish fillet, without the necessity of destroying the sample.

Biogenic amines are not the only compounds associated with the spoilage of fish. Trimethylamine concentration can also be used as an indicator of fish spoilage, since it is related with the putrefaction process and the fishy odour. This compound has been estimated in fresh fish (Hypophthalmichthys molitrix) using FT-NIRS and different chemometric algorithms [77]. In this case, measurements were made directly in the fish, which was sliced in small cubes. Regression analysis showed RMSEP between 5.10 and 5.75 mg N/100 g and correlation coefficients of prediction between 0.94 and 0.98. K-value is another indicator of fish deterioration that is related to the degradation of ATP in the fish flesh. Likewise, this index was also estimated in fresh silver carps (Hypophthalmichthys molitrix) using FT-NIRS to make the measurements directly in the fish. Several chemometric techniques were compared, obtaining as best result an RMSEP of 0.036 and a correlation coefficient of prediction of 93.74% [82]. Another way to measure the spoilage of saithe (Pollachius virens) and hoki (Macruronus novaezelandiae) fillets has been also explored [33]. Here the authors developed PLSR models to predict free fatty acids (FFA), fluorescent interaction compounds (OFR) and thiobarbituric acid reactive substances (TBARS), in order to measure the lipid oxidation, one of the major deteriorative reactions in fish muscle. Results showed acceptable results for the test set in both species (r2 between 0.63 and 0.98), although some high errors were found (RMSEP of 2.90 and 3.43 g FFA/100 g lipids, 0.97 and 0.55 μmol malonaldehyde diethylacetal/kg of samples for TBARS and 5.23 and 1.20 for OFR in hoki and saithe, respectively). Similar results were found regarding FFA [83] and on finding lipid oxidation products in vegetable oils [84]. Other similar analysis were performed using NIRS to predict the pH, TBARS, total volatile basic nitrogen (TVB-N) and K-value of minced bighead carp (Aristichthys nobilis) [85]. PLSR together with competitive adaptive reweighted sampling (CARS) algorithm and pre-processing methods were used to make regressions, obtaining satisfactory results for the prediction of the test set (coefficients of prediction of 0.945, 0.932, 0.954 and 0.807 and RMSEP of 0.081, 2.099, 0.107 and 6.509 for pH, TVB-N, TBA and K value, respectively).

2.3.2 Microorganisms

Fish is one of the most vulnerable aquatic products, as it serves as a growth medium for microorganisms, which can be either pathogenic or cause fish spoilage. In this sense, spectroscopic techniques have shown potential for spoilage monitoring to confirm if fish is safe for consumption or not, since conventional methods to determine microorganisms are really time-consuming and tedious [86]. However, although some works have been done regarding the detection of microorganisms by NIRS and the potentiality of the technique is big, this is an application that must be studied carefully, and it must be validated using techniques with limits of detection more sensitive than NIR.

Some studies tried to use SW-NIRS to quantify microbial loads (expressed as total viable count) in rainbow trout (Oncorhynchus mykiss) [44]. Regression models were built based on spectra taken in the flesh and skin of fish fillets, stored at 4°C for 8 days, and also in minced samples stored at room temperature 24 hours. All the models gave goods results, the model made in the flesh being slightly better (r2 = 0.97 and SEP = 0.38 log CFU/g) compared with the models of the skin and the minced samples (r2 = 0.94 and 0.82 and SEP = 0.53 and 0.82 log CFU/g, respectively). Similar experiments have used FT-NIRS to detect and predict microbial spoilage on Atlantic salmon fillets (Salmo salar) [45]. The results showed that the regression analysis was able to predict the numbers of bacteria (expressed as total aerobic plate counts) after 9 days of storage at 4°C, with an r2 of 0.64 and a RMSEP of 0.32 log cfu/g in the validation set. A portable spectrometer was also used to determine the total bacteria content in flounder fillets (Paralicthys olivaceus), expressed as the aerobic plate count of total bacteria, after 8 days of storage at 4°C [87]. Different models based on different chemometrics techniques were developed, obtaining the best results using a combination of genetic algorithm and artificial neural networks, with an r2 of 0.966 and an RMSEP of 0.083 in the test set.

Advertisement

3. Conclusions

NIRS has been proved as an effective analysis tool for the fish value chain. Although the use of this technology in fish and fish products routine inspection has recently increased, more and more studies are available each year, some applications are more developed than others. The monitoring of the quality and authenticity of the fish chain by NIRS is well known, with applications that are nowadays implemented in the food industry. However, more efforts are still needed regarding the use of this technology for safety issues, where the limit of detection still represents a big challenge. Even if some studies have been carried out successfully during the last decade, these applications should be studied carefully, and more investigations must be done before trying to implement NIRS as an alternative to the already established destructive methods.

Advertisement

Acknowledgments

The authors greatly acknowledge the Basque Government—Department of Economic Development, Sustainability and Environment—Vice. Dept. of Agriculture, Fishing and Food Policy, Directorate of Quality and Food Industries for the funding of the project ELIKatea 4.0 and for the scholarship of Sonia Nieto-Ortega.

Advertisement

ApplicationSpecies (sample preparation)Measured propertyInstrumentChemometric methodStatistical resultsReference
Proximate composition (Quality)Tilapia (fillets)Moisture, protein and lipidReflectance NIRPLSRr2cv = 0.959, SECV = 0.872 g/100 g for moisture r2cv = 0.306, SECV = 0.550 g/100 g for protein r2cv = 0.971, SECV = 0.706 g/100 g for lipid[31]
Geographical origin (Authenticity)Chinese region originSIMCA> 75% Correct classification
Freshness (Quality)Cod Gadus morhua Salmon Salmo salar (fillets)Storage timeTransflection Vis–NIRPLSRr2p = 0.97, RMSEP = 1.04 days for cod in visible region r2p = 0.98, RMSEP = 1.20 days for salmon in NIR region[32]
Proximate composition (Quality)Saithe Pollachius virens Hoki Macruronus novaezelandiae (thawed portions)Total lipid content, phospholipids, PUFAs and monounsaturated fatty acids (MUFAs)Reflectance FT-NIRPLSRr2v ≥ 0.96 hoki and r2v ≥ 0.73 saithe for lipid composition RMSEP <4.10 hoki and RMSEP <5.31% saithe for lipid composition[33]
Spoilage compounds (Safety)Lipid oxidation components: free fatty acids (FFA), fluorescent interaction compounds (OFR) and thiobarbituric acid reactive substances (TBARS)PLSRr2v ≥ 0.63 hoki and r2v ≥ 0.67 saithe for lipid oxidation RMSEP <5.23 hoki and RMSEP <3.43% saithe for lipid oxidation
Proximate composition (Quality)Atlantic salmon S. salar (fillets and minced)Moisture, fat and proteinOn-line NIR fibre-optic probe and off-line reflectance NIRPCRRMSECV = 6.6 g/Kg for fat in minced sample RMSECV = 3.8 g/Kg for moisture in minced sample RMSECV = 2.0 g/Kg for protein in minced sample RMSECV = 12.4 g/Kg for fat in whole fillet RMSECV = 11.3 g/Kg for moisture in whole fillet[34]
Proximate composition (Quality)Atlantic salmon salmo salar (whole fish)Fat and moisture contentInteractance NIRPLSRr2cv = 0.87, RMSECV = 1.12% for fat r2cv = 0.86, RMSECV = 0.98% for moisture[35]
Proximate composition (Quality)European sea bass Dicentrarchus labrax (fillet, fresh minced and freeze-dried minced)Water, ether extract and crude proteinAbsorbance NIR + monocromatorPLSRr2cv > 0.47, SECV <2.76% for water in fillets r2cv = 0.95, SECV = 0.87% for water in minced r2cv = 0.97, SECV = 0.70% for water in freeze-dried
r2cv > 0.48, SECV <2.81% for ether in fillets r2cv = 0.97, SECV = 0.70% for ether in minced r2cv = 0.97, SECV = 0.62% for ether in freeze-dried
r2cv = 0.30, SECV = 0.55% for protein in minced r2cv = 0.68, SECV = 0.35% for protein in freeze-dried
[36]
Rearing system (Authenticity)Extensive, semi-extensive, intensive or sea-cagesSIMCA> 37% correct classification for minced fillets
> 74% correct classification for freeze-dried minced fillets
Proximate composition (Quality)European sea bass Dicentrarchus labrax (fresh and freeze-dried minced)Water, ether extract and crude proteinAbsorbance NIR + monocromatorPLSRr2cv = 0.935, SECV = 0.387% for water in minced r2cv = 0.953, SECV = 0.328% for water in freeze-dried r2cv = 0.984, SECV = 0.154% for ether in minced r2cv = 0.982, SECV = 0.175% for ether in freeze-dried r2cv = 0.587, SECV = 0.290% for protein in minced r2cv = 0.640, SECV = 0.274% for protein in freeze-dried[37]
Rearing system (Authenticity)Organic or semi-intensiveSIMCA> 20% correct classification for minced samples >65% correct classification for freeze-dried samples
Proximate composition (Quality)Skipjack tuna Katsuwonus pelamis and yellow fin tuna Thunnus albacares (cutlets)Moisture, protein, free and total fatNIR + fibre optic probePLS-2r2p = 0.98 for moisture r2p = 0.99 for protein r2p = 0.95 for total fat r2p = 0.96 for free fat[38]
Proximate composition (Quality)Tilapia Oreochromis niloticus (fillets)Omega-3 and omega-6Reflectance NIRPLSRr2p = 0.70, RMSEP = 2.39 mg FA/g for omega-3 r2p = 0.71, RMSEP = 4.76 mg FA/g for omega-6[39]
Proximate composition (Quality)Salmon roe Oncorhynchus keta (cured)Salt and moistureSW-NIR + fibre optic probePLSRr2p > 0.711, SEP < 1.13% for salt r2p > 0.609, SEP < 1.87% for moisture[40]
Proximate composition (Quality)Atlantic salmon Salmo salar (cold smoked)Salt, water and water phase salt (WPS)SW-NIR + fibre optic probePCR, MLR, PLSR, BPNNr2cv > 0.639, RMSCV <0.77 for salt r2cv > 0.854, RMSCV <4.10 for water r2cv > 0.717, RMSCV <1.55 for WPS[41]
Proximate composition (Quality)Atlantic salmon Salmo salar (cured portions)Salt and moistureSW-NIR + fibre optic probePLSR, BPNNr2cv > 0.701, RMSCV <1.43 for salt r2cv > 0.784, RMSCV <2.08 for moisture[42]
Freshness (Quality)Rainbow trout Oncorhynchus mykiss (flesh and skin of fillets and minced)Storage time (2 temperatures)SW-NIRSPCASegregation between fresh and non-fresh fish[44]
Microorganisms (Safety)Microbial loadsPLSRr2p = 0.97, SEP = 0.38 log CFU/g in flesh side of fillet r2p = 0.94, SEP = 0.53 log CFU/g in skin side of fillet r2p = 0.82, SEP = 0.82 log CFU/g in minced samples
Freshness (Quality)Atlantic salmon Salmo salar (fillets)Storage timeFT-NIRPCAClear separation between fresh and non-fresh fillets[45]
Microorganisms (Safety)Numbers of bacteriaPLSRr2cv = 0.64, RMSECV = 0.32 log cfu/g of bacteria
Freshness (Quality)Golden pompano Trachinotus ovatus (portions)Storage timeVis–NIRPCASamples stored for 0, 2 and 3 days could not be well discriminated[46]
Freshness (Quality)Rohu Labeo rohita (sliced pieces)Storage timeAbsorption NIRPCASamples were well grouped by their freshness[47]
Freshness (Quality)Atlantic cod Gadus morhua (fillets)Storage timeInteractance Vis–NIRPLSRr2cv = 0.93, RMSECV = 1.66 days[48]
Unfrozen vs. Thawed (Authenticity)Unfrozen vs. ThawedKNN> 96% correct classification
Freshness (Quality)Atlantic salmon Salmo salar L (fillets)Storage timeInteractance Vis–NIRPLSRr2cv = 0.95, RMSECV = 1.91 days[49]
Unfrozen vs. Thawed (Authenticity)Unfrozen vs. ThawedKNN100% correct classification
Freshness (Quality)Cod Gadus morhua (fillets)Quality index method (QIM)Transflection Vis–NIRPLSRr2p > 0.93, RMSEP <3.1 pints[50]
Wild vs. Farmed (Authenticity)Seabass Dicentrarchus labrax (minced)Wild vs. FarmedReflectance NIRPLS-DA with measured chemistry, PLS-DA with estimated chemistry, PLS-DA and WPTER100% correct classification with PLS-DA_mc
94% correct classification with PLS-DA_ec
91% correct classification with PLS-DA_NIR
91% correct classification with WPTER
[53]
Species identification (Authenticity)Salmon Salmo salar, blue whiting Micromesistius poutassau, mackerel Scomber scombrus and herring Clupea harengus (Fishmeal)Species identificationReflectance NIRDPLS and LDA>80% correct classification with DPLS
> 70% correct classification with LDA
[55]
Species identification (Authenticity)Winter cod and cod, red mullet and Atlantic mullet, and samlet and salmon trout (skin and meat of fillets)Species identificationHandheld reflectance NIRSIMCA100% correct classification of cod and winter cod
100% correct classification of red mullet and Atlantic mullet 100% correct classification of salmon trout and samlet
[56]
Species identification (Authenticity)Red mullet Mullus surmuletus and plaice Pleuronectes platessa, Atlantic mullet Pseudupeneus prayensis and flounder Platichthys flesus flesus (fillets)Higher from lower economical value speciesFT-NIRLDA and SIMCA100% correct classification of red mullet and Atlantic mullet >67% correct classification of plaice and flounder[57]
Unfrozen vs. Thawed (Authenticity)Unfrozen vs. Thawed Atlantic mulletFT-NIRLDA and SIMCA> 82% correct classification of unfrozen and thawed Atlantic mullet
Species identification (Authenticity)Silver carp Hypophthalmichthys molitrix, black carp Mylopharyngodon piceus, bighead carp Aristichthys nobilis, grass carp Ctenopharyngodon idellus, common carp Cyprinus carpio, crucian Carassius auratus and bream Parabramis pekinensis (minced)Species identificationReflectance NIRPCA-LDA, PLS-LDA, CARS-LDA and FFT - LDA> 96% correct classification for PCA-LDA
> 89% correct classification for PLS-LDA
> 88% correct classification for CARS-LDA
> 96% correct classification for FFT-LDA
[58]
Species identification (Authenticity)White croaker Argyrosomus argentatus, hairtail Trichiurus haumela, and redcoat Nemipterus virgatus (surimi)Species identificationReflectance NIRPC-DA100% correct classification for the validation set[59]
Species identification (Authenticity)Atlantic cod G. morhua and haddock Melanogrammus aeglefinus (raw fillets and patties)Species identificationReflectance handheld NIR and FT-NIRLDA and SIMCA> 76% correct classification for portable NIR and LDA
100% correct classification for FT-NIR and LDA
> 60% correct classification for portable NIR and SIMCA
> 91% correct classification for FT-NIR and SIMCA
[60]
Wild vs. Farmed (Authenticity)Sea bass Dicentrarchus labrax (minced)Wild or farmed and rearing systemReflectance NIROPLS-DA100% correct classification for production method
> 67% correct classification for farming system
[62]
Geographical origin (Authenticity)Areas of the Mediterranean SeaOPLS-DA> 85% correct classification
Proximate composition (Quality)Alaskan pollock Gadus chalcogrammu, Atlantic cod Gadus morhua, European plaice Pleuronectes platessa, common sole Solea solea, and turbot Psetta maxima (fillets)Protein, lipids, and moistureReflectance NIRPLSRr2p = 0.80, RMSEP = 0.15 g/100 g for lipids
r2p = 0.80, RMSEP = 0.95% for protein
r2p = 0.79, RMSEP = 1.01% for moisture
[64]
Wild vs. Farmed (Authenticity)Wild vs. farmedLR and SVM90% correct classification for sole with LR
99% correct classification for turbot with SVM
Unfrozen vs. Thawed (Authenticity)Unfrozen vs. ThawedXGB, LDA and RF88% correct classification for cod with XGB
100% correct classification for plaice with LDA
90% correct classification for sole with RF
Geographical origin (Authenticity)Anchovies Engraulis encrasicholus (minced semi-finished and finished salted)Mediterranean countryReflectance FT-NIROPLS-DA> 95% correct classification[65]
Unfrozen vs. Thawed (Authenticity)Red sea bream Pagrus major (whole fish)Unfrozen vs. ThawedInteractance NIRLDA and SIMCA> 63% correct classification for SIMCA
> 79% correct classification for LDA
[72]
Unfrozen vs. Thawed (Authenticity)Swordfish Xiphias gladius L (cutlets)Unfrozen vs. ThawedVis–NIR and NIR + monocromatorPLS-DA> 96.7% correct classification for VIS/NIR
> 90.0% correct classification for NIR
[73]
Unfrozen vs. Thawed (Authenticity)Tuna Thunnus thynnus (fillets)Unfrozen vs. ThawedReflectance Vis–NIRPLS-DA92% correct classification for unfrozen
82% correct classification for frozen–thawed
[74]
Unfrozen vs. Thawed (Authenticity)Tilapia (Oreochromis) (fillets)Freeze-thawing cyclesFT-NIRMahalanobis distance-DA> 60.0% correct classification for once frozen–thawed fillets
> 60.0% correct classification for samples subjected to several freezing cycles
[75]
Unfrozen vs. Thawed (Authenticity)Bigeye tuna Thunnus obesus (fillets)Unfrozen vs. ThawedReflectance handheld NIRPLS-DA91% of correct classification[76]
Spoilage compounds (Safety)Silver carp Hypophthalmichthys molitrix (sliced portions)Trimethylamine contentFT-NIRPLS, Si-PLS and GA-PLSr2p = 0.94, RMSEP = 5.75 mg N/100 g with PLS
r2p = 0.95, RMSEP = 5.45 mg N/100 g with Si-PLS
r2p = 0.98, RMSEP = 5.10 mg N/100 g with GA-PLS
[77]
Spoilage compounds (Safety)Tuna Katsuwonus pelamis and Thunnus albacares (minced raw and canned)Histamine contentReflectance FT-NIROPLSRr2 p > 0.97, RMSEP <10 mg kg−1[78]
Spoilage compounds (Safety)Chub Mackerel Scomber japonicus (homogenized and ultrafiltered)Cadaverine contentNIRPLSRr2p > 0.98[80]
Spoilage compounds (Safety)Skipjack tuna Katsuwonus pelmis (dry extracts)Histamine contentReflectance FT-NIRPLSRr2c > 0.76, SEP < 3.47 ppm[81]
Spoilage compounds (Safety)Silver carp Hypophthal michthys molitrix (whole fish, measured in the eye)K-valueReflectance FT-NIRPLS, i-PLS, Si-PLS, SVMR and Si-SVMRRp = 93.7%, RMSEP = 0.036 for Si-SVMR[82]
Spoilage compounds (Safety)Mackerel oil (extracted from minced samples)Free fatty acidsTransmittance NIRMLR and PLSRp > 0.862%, SEP < 0.154% for MLR
Rp > 0.441, 0.862%, SEP < 0.23, 0.131% for PLS
[83]
Spoilage compounds (Safety)Bighead carp Aristichthys nobilis (minced)pH, TBARS, total volatile basic nitrogen (TVB-N) and K-valueReflectance NIRCARS-PLSRRp = 0.945, RMSEP = 0.081 for pH
Rp = 0.932, RMSEP = 2.099 for TVB-N
Rp = 0.954, RMSEP = 0.107 for TBARS
Rp = 0.807, RMSEP = 6.509 for K-value
[85]
Microorganisms (Safety)Flounder (fillets)Total bacteria contentPortable NIRPLSR and genetic algorithm with BP-ANNr2p = 0.916, RMSEP = 0.40 with PLS
r2p = 0.966, RMSEP = 0.083 with BP-ANN
[87]

References

  1. 1. Siesler HW. Basic principles of near-infrared spectroscopy. In: Burns DA, Ciurczac EW, editors. Handbook of Near-infrared Analysis. 3rd ed. Boca Raton, FL: CRC press; 2008. pp. 7-20
  2. 2. Ozaki Y, Morisawa Y. Principles and characteristics of NIR spectroscopy. In: Ozaki Y, Huck C, Tsuchikawa S, Engelsen SB, editors. Near-Infrared Spectroscopy: Theory, Spectral Analysis, Instrumentation, and Applications. Singapore: Springer Singapore; 2021. pp. 11-35. DOI: 10.1007/978-981-15-8648-4_2
  3. 3. Ozaki Y. Infrared spectroscopy—mid-infrared, near-infrared, and far-infrared/terahertz spectroscopy. Analytical Sciences. 2021;37:1193-1212. DOI: 10.2116/analsci.20R008
  4. 4. Beć KB, Grabska J, Huck CW. Near-infrared spectroscopy in bio-applications. Molecules. 2020;25:2948. DOI: 10.3390/molecules25122948
  5. 5. El-Azazy M. Introductory Chapter: Infrared spectroscopy - a synopsis of the fundamentals and applications. In: El-Azazy M, editor. Infrared Spectroscopy - Principles, Advances, and Applications. London: IntechOpen; 2018. DOI: 10.5772/intechopen.82210
  6. 6. Brereton RG, Jansen J, Lopes J, Marini F, Pomerantsev A, Rodionova O, et al. Chemometrics in analytical chemistry—part I: History, experimental design and data analysis tools. Analytical and Bioanalytical Chemistry. 2017;409:5891-5899. DOI: 10.1007/s00216-017-0517-1
  7. 7. Brereton RG, Jansen J, Lopes J, Marini F, Pomerantsev A, Rodionova O, et al. Chemometrics in analytical chemistry—part II: Modeling, validation, and applications. Analytical and Bioanalytical Chemistry. 2018;410:6691-6704. DOI: 10.1007/s00216-018-1283-4
  8. 8. Blanco M, Villarroya I. NIR spectroscopy: A rapid-response analytical tool. TrAC Trends in Analytical Chemistry. 2002;21:240-250. DOI: 10.1016/S0165-9936(02)00404-1
  9. 9. Rinnan Å, Fvd B, Engelsen SB. Review of the most common pre-processing techniques for near-infrared spectra. TrAC Trends in Analytical Chemistry. 2009;28:1201-1222. DOI: 10.1016/j.trac.2009.07.007
  10. 10. de Lima MD, Barbosa R. Methods of authentication of food grown in organic and conventional systems using chemometrics and data mining algorithms: A review. Food Analytical Methods. 2019;12:887-901. DOI: 10.1007/s12161-018-01413-3
  11. 11. Yu H, Guo L, Kharbach M, Han W. Multi-way analysis coupled with near-infrared spectroscopy in food industry: Models and applications. Food. 2021;10:802. DOI: 10.3390/foods10040802
  12. 12. Nobari-Moghaddam H, Tamiji Z, Akbari-Lakeh M, Khoshayand MR, Haji-Mahmoodi M. Multivariate analysis of food fraud: A review of NIR based instruments in tandem with chemometrics. Journal of Food Composition and Analysis. 2022;107:104343. DOI: 10.1016/j.jfca.2021.104343
  13. 13. McVey C, McGrath TF, Haughey SA, Elliott CT. A rapid food chain approach for authenticity screening: The development, validation and transferability of a chemometric model using two handheld near infrared spectroscopy (NIRS) devices. Talanta. 2021;222:121533. DOI: 10.1016/j.talanta.2020.121533
  14. 14. Melado-Herreros A, Nieto-Ortega S, Olabarrieta I, Gutiérrez M, Villar A, Zufía J, et al. Postharvest ripeness assessment of ‘Hass’ avocado based on development of a new ripening index and Vis-NIR spectroscopy. Postharvest Biology and Technology. 2021;181:111683. DOI: 10.1016/j.postharvbio.2021.111683
  15. 15. Ncama K, Magwaza LS, Mditshwa A, Tesfay SZ. Application of visible to near-infrared spectroscopy for non-destructive assessment of quality parameters of fruit. In: El-Azazy M, editor. Infrared Spectroscopy - Principles, Advances, and Applications. London: IntechOpen; 2018. DOI: 10.5772/intechopen.80069
  16. 16. Coppola D, Lauritano C, Palma Esposito F, Riccio G, Rizzo C, de Pascale D. Fish waste: From problem to valuable resource. Marine Drugs. 2021;19:116. DOI: 10.3390/md19020116
  17. 17. Ruiz-Salmón I, Laso J, Margallo M, Villanueva-Rey P, Rodríguez E, Quinteiro P, et al. Life cycle assessment of fish and seafood processed products – A review of methodologies and new challenges. Science of The Total Environment. 2021;761:144094. DOI: 10.1016/j.scitotenv.2020.144094
  18. 18. Fox M, Mitchell M, Dean M, Elliott C, Campbell K. The seafood supply chain from a fraudulent perspective. Food Security. 2018;10:939-963. DOI: 10.1007/s12571-018-0826-z
  19. 19. Sheng L, Wang L. The microbial safety of fish and fish products: Recent advances in understanding its significance, contamination sources, and control strategies. Comprehensive Reviews in Food Science and Food Safety. 2021;20:738-786. DOI: 10.1111/1541-4337.12671
  20. 20. Ye B, Chen J, Fu L, Wang Y. Application of nondestructive evaluation (NDE) technologies throughout cold chain logistics of seafood: Classification, innovations and research trends. LWT- Food Science and Technology. 2022;158:113127. DOI: 10.1016/j.lwt.2022.113127
  21. 21. Ndraha N, Hsiao H-I, Vlajic J, Yang M-F, Lin H-TV. Time-temperature abuse in the food cold chain: Review of issues, challenges, and recommendations. Food Control. 2018;89:12-21. DOI: 10.1016/j.foodcont.2018.01.027
  22. 22. Yu W. Applications of near infrared spectroscopy for fish and fish products quality: A review. IOP Conference Series: Earth and Environmental Science. 2021;657:012115. DOI: 10.1088/1755-1315/657/1/012115
  23. 23. Shenk JS, Workman JJ, Westerhaus MO. Application of NIR spectroscopy to agricultural products. In: Burns DA, Ciurczac EW, editors. Handbook of Near-infrared Analysis. 3rd ed. Boca Raton, FL: CRC press; 2008. pp. 347-386
  24. 24. Mendez J, Mendoza L, Cruz-Tirado JP, Quevedo R, Siche R. Trends in application of NIR and hyperspectral imaging for food authentication. Scientia Agropecuaria. 2019;10:143-161. DOI: 10.17268/sci.agropecu.2018.01.16
  25. 25. Qu J-H, Liu D, Cheng J-H, Sun D-W, Ma J, Pu H, et al. Applications of near-infrared spectroscopy in food safety evaluation and control: A review of recent research advances. Critical Reviews in Food Science and Nutrition. 2015;55:1939-1954. DOI: 10.1080/10408398.2013.871693
  26. 26. Power A, Cozzolino D. How fishy is your fish? Authentication, provenance and traceability in fish and seafood by means of vibrational spectroscopy. Applied Sciences. 2020;10:4150. DOI: 10.3390/app10124150
  27. 27. Franceschelli L, Berardinelli A, Dabbou S, Ragni L, Tartagni M. Sensing technology for fish freshness and safety: A review. Sensors. 2021;21:1373. DOI: 10.3390/s21041373
  28. 28. Liu D, Zeng X-A, Sun D-W. NIR spectroscopy and imaging techniques for evaluation of fish quality—A review. Applied Spectroscopy Reviews. 2013;48:609-628. DOI: 10.1080/05704928.2013.775579
  29. 29. Venugopal V. Biosensors in fish production and quality control. Biosensors and Bioelectronics. 2002;17:147-157. DOI: 10.1016/S0956-5663(01)00180-4
  30. 30. Yeannes MI, Almandos ME. Estimation of fish proximate composition starting from water content. Journal of Food Composition and Analysis. 2003;16:81-92. DOI: 10.1016/S0889-1575(02)00168-0
  31. 31. Liu Y, Ma D-H, Wang X-C, Liu L-P, Fan Y-X, Cao J-X. Prediction of chemical composition and geographical origin traceability of Chinese export tilapia fillets products by near infrared reflectance spectroscopy. LWT - Food Science and Technology. 2015;60:1214-1218. DOI: 10.1016/j.lwt.2014.09.009
  32. 32. Nilsen H, Esaiassen M, Heia K, Sigernes F. Visible/near-infrared spectroscopy: A new tool for the evaluation of fish freshness? Journal of Food Science. 2002;67:1821-1826. DOI: 10.1111/j.1365-2621.2002.tb08729.x
  33. 33. Karlsdottir MG, Arason S, Kristinsson HG, Sveinsdottir K. The application of near infrared spectroscopy to study lipid characteristics and deterioration of frozen lean fish muscles. Food Chemistry. 2014;159:420-427. DOI: 10.1016/j.foodchem.2014.03.050
  34. 34. Isaksson T, Tøgersen G, Iversen A, Hildrum KI. Non-destructive determination of fat, moisture and protein in salmon fillets by use of near-infrared diffuse spectroscopy. Journal of the Science of Food and Agriculture. 1995;69:95-100. DOI: 10.1002/jsfa.2740690115
  35. 35. Wold JP, Isaksson T. Non-destructive determination of fat and moisture in whole atlantic salmon by near-infrared diffuse spectroscopy. Journal of Food Science. 1997;62:734-736. DOI: 10.1111/j.1365-2621.1997.tb15446.x
  36. 36. Xiccato G, Trocino A, Tulli F, Tibaldi E. Prediction of chemical composition and origin identification of european sea bass (Dicentrarchus labrax L.) by near infrared reflectance spectroscopy (NIRS). Food Chemistry. 2004;86:275-281. DOI: 10.1016/j.foodchem.2003.09.026
  37. 37. Trocino A, Xiccato G, Majolini D, Tazzoli M, Bertotto D, Pascoli F, et al. Assessing the quality of organic and conventionally-farmed European sea bass (Dicentrarchus labrax). Food Chemistry. 2012;131:427-433. DOI: 10.1016/j.foodchem.2011.08.082
  38. 38. Khodabux K, L’Omelette MSS, Jhaumeer-Laulloo S, Ramasami P, Rondeau P. Chemical and near-infrared determination of moisture, fat and protein in tuna fishes. Food Chemistry. 2007;102:669-675. DOI: 10.1016/j.foodchem.2006.05.057
  39. 39. dos Santos DA, Coqueiro A, Gonçalves TR, Carvalho JC, Bezerra JS Jr, Matsushita M, et al. Omega-3 and omega-6 determination in nile Tilapia’s fillet based on MicroNIR spectroscopy and multivariate calibration. Journal of the Brazilian Chemical Society. 2020;31:1883-1890. DOI: 10.21577/0103-5053.20200082
  40. 40. Huang Y, Rogers TM, Wenz MA, Cavinato AG, Mayes DM, Bledsoe GE, et al. Detection of sodium chloride in cured salmon roe by SW−NIR spectroscopy. Journal of Agricultural and Food Chemistry. 2001;49:4161-4167. DOI: 10.1021/jf001177f
  41. 41. Huang Y, Cavinato AG, Mayes DM, Bledsoe GE, Rasco BA. Nondestructive prediction of moisture and sodium chloride in cold smoked atlantic salmon (Salmo salar). Journal of Food Science. 2002;67:2543-2547. DOI: 10.1111/j.1365-2621.2002.tb08773.x
  42. 42. Huang Y, Cavinato AG, Mayes DM, Kangas LJ, Bledsoe GE, Rasco BA. Nondestructive determination of moisture and sodium chloride in cured atlantic salmon (Salmo salar) (Teijin) using short-wavelength near-infrared spectroscopy (SW-NIR). Journal of Food Science. 2003;68:482-486. DOI: 10.1111/j.1365-2621.2003.tb05698.x
  43. 43. Chun H-N, Kim B, Shin H-S. Evaluation of a freshness indicator for quality of fish products during storage. Food Science and Biotechnology. 2014;23:1719-1725. DOI: 10.1007/s10068-014-0235-9
  44. 44. Lin M, Mousavi M, Al-Holy M, Cavinato AG, Rasco BA. Rapid near infrared spectroscopic method for the detection of spoilage in rainbow trout (Oncorhynchus mykiss) fillet. Journal of Food Science. 2006;71:S18-S23. DOI: 10.1111/j.1365-2621.2006.tb12400.x
  45. 45. Tito NB, Rodemann T, Powell SM. Use of near infrared spectroscopy to predict microbial numbers on Atlantic salmon. Food Microbiology. 2012;32:431-436. DOI: 10.1016/j.fm.2012.07.009
  46. 46. Zhang X, Zhou H, Chang L, Lou X, Li J, Hui G, et al. Study of golden pompano (Trachinotus ovatus) freshness forecasting method by utilising Vis/NIR spectroscopy combined with electronic nose. International Journal of Food Properties. 2018;21:1257-1269. DOI: 10.1080/10942912.2018.1440239
  47. 47. Dei J, Naskar H, Dasgupta S, Nag S, Behera BK, Tudu B, et al. Fish freshness assessment using NIR spectroscopy. In: Proceedings of the 2020 International Conference on Emerging Frontiers in Electrical and Electronic Technologies (ICEFEET); 10-11 July 2020; Patna, India. 2020. pp. 1-3. DOI: 10.1109/ICEFEET49149.2020.9187016
  48. 48. Sivertsen AH, Kimiya T, Heia K. Automatic freshness assessment of cod (Gadus morhua) fillets by Vis/Nir spectroscopy. Journal of Food Engineering. 2011;103:317-323. DOI: 10.1016/j.jfoodeng.2010.10.030
  49. 49. Kimiya T, Sivertsen AH, Heia K. VIS/NIR spectroscopy for non-destructive freshness assessment of Atlantic salmon (Salmo salar L.) fillets. Journal of Food Engineering. 2013;116:758-764. DOI: 10.1016/j.jfoodeng.2013.01.008
  50. 50. Nilsen H, Esaiassen M. Predicting sensory score of cod (Gadus morhua) from visible spectroscopy. LWT - Food Science and Technology. 2005;38:95-99. DOI: 10.1016/j.lwt.2004.05.001
  51. 51. European Regulation (EU) n. 1379/2013. Available from: https://eur-lex.europa.eu/legal-content/EN/TXT/HTML/?uri=CELEX:32013R1379&from=EN [Accessed: 2022-05-09]
  52. 52. Ghidini S, Varrà MO, Zanardi E. Approaching authenticity issues in fish and seafood products by qualitative spectroscopy and chemometrics. Molecules. 2019;24:1812. DOI: 10.3390/molecules24091812
  53. 53. Ottavian M, Facco P, Fasolato L, Novelli E, Mirisola M, Perini M, et al. Use of near-infrared spectroscopy for fast fraud detection in seafood: Application to the authentication of wild european sea bass (Dicentrarchus labrax). Journal of Agricultural and Food Chemistry. 2012;60:639-648. DOI: 10.1021/jf203385e
  54. 54. Blanco-Fernandez C, Ardura A, Masiá P, Rodriguez N, Voces L, Fernandez-Raigoso M, et al. Fraud in highly appreciated fish detected from DNA in Europe may undermine the Development Goal of sustainable fishing in Africa. Scientific Reports. 2021;11:11423. DOI: 10.1038/s41598-021-91020-w
  55. 55. Cozzolino D, Chree A, Scaife JR, Murray I. Usefulness of near-infrared reflectance (NIR) spectroscopy and chemometrics to discriminate fishmeal batches made with different fish species. Journal of Agricultural and Food Chemistry. 2005;53:4459-4463. DOI: 10.1021/jf050303i
  56. 56. O'Brien N, Hulse CA, Pfeifer F, Siesler HW. Near infrared spectroscopic authentication of seafood. Journal of Near Infrared Spectroscopy. 2013;21:299-305. DOI: 10.1255/jnirs.1063
  57. 57. Alamprese C, Casiraghi E. Application of FT-NIR and FT-IR spectroscopy to fish fillet authentication. LWT - Food Science and Technology. 2015;63:720-725. DOI: 10.1016/j.lwt.2015.03.021
  58. 58. Lv H, Xu W, You J, Xiong S. Classification of freshwater fish species by linear discriminant analysis based on near infrared reflectance spectroscopy. Journal of Near Infrared Spectroscopy. 2017;25:54-62. DOI: 10.1177/0967033516678801
  59. 59. Zhang X-Y, Hu W, Teng J, Peng H-H, Gan J-H, Wang X-C, et al. Rapid recognition of marine fish surimi by one-step discriminant analysis based on near-infrared diffuse reflectance spectroscopy. International Journal of Food Properties. 2017;20:2932-2943. DOI: 10.1080/10942912.2016.1261153
  60. 60. Grassi S, Casiraghi E, Alamprese C. Handheld NIR device: A non-targeted approach to assess authenticity of fish fillets and patties. Food Chemistry. 2018;243:382-388. DOI: 10.1016/j.foodchem.2017.09.145
  61. 61. Costa C, D’Andrea S, Russo R, Antonucci F, Pallottino F, Menesatti P. Application of non-invasive techniques to differentiate sea bass (Dicentrarchus labrax, L. 1758) quality cultured under different conditions. Aquaculture International. 2011;19:765-778. DOI: 10.1007/s10499-010-9393-9
  62. 62. Ghidini S, Varrà MO, Dall'Asta C, Badiani A, Ianieri A, Zanardi E. Rapid authentication of European sea bass (Dicentrarchus labrax L.) according to production method, farming system, and geographical origin by near infrared spectroscopy coupled with chemometrics. Food Chemistry. 2019;280:321-327. DOI: 10.1016/j.foodchem.2018.12.075
  63. 63. Vidal NP, Goicoechea E, Manzanos MJ, Guillén MD. Fourier transform infrared spectroscopy as a tool to study farmed and wild sea bass lipid composition. Journal of the Science of Food and Agriculture. 2014;94:1340-1348. DOI: 10.1002/jsfa.6417
  64. 64. Gonçalves DB, Santos CSP, Pinho T, Queirós R, Vaz PD, Bloore M, et al. Near infrared reflectance spectroscopy coupled to chemometrics as a cost-effective, rapid, and non-destructive tool for fish fraud control: Monitoring source, condition, and nutritional value of five common whitefish species. Journal of AOAC International. 2020;104:53-60. DOI: 10.1093/jaoacint/qsaa114
  65. 65. Varrà MO, Ghidini S, Ianieri A, Zanardi E. Near infrared spectral fingerprinting: A tool against origin-related fraud in the sector of processed anchovies. Food Control. 2021;123:107778. DOI: 10.1016/j.foodcont.2020.107778
  66. 66. Hassoun A, Shumilina E, Di Donato F, Foschi M, Simal-Gandara J, Biancolillo A. Emerging techniques for differentiation of fresh and frozen-thawed seafoods: Highlighting the potential of spectroscopic techniques. Molecules. 2020;25:4472. DOI: 10.3390/molecules25194472
  67. 67. Hong H, Luo Y, Zhou Z, Bao Y, Lu H, Shen H. Effects of different freezing treatments on the biogenic amine and quality changes of bighead carp (Aristichthys nobilis) heads during ice storage. Food Chemistry. 2013;138:1476-1482. DOI: 10.1016/j.foodchem.2012.11.031
  68. 68. Leygonie C, Britz TJ, Hoffman LC. Impact of freezing and thawing on the quality of meat: Review. Meat Science. 2012;91:93-98. DOI: 10.1016/j.meatsci.2012.01.013
  69. 69. Nakazawa N, Okazaki E. Recent research on factors influencing the quality of frozen seafood. Fisheries Science. 2020;86:231-244. DOI: 10.1007/s12562-020-01402-8
  70. 70. Tolstorebrov I, Eikevik TM, Bantle M. Effect of low and ultra-low temperature applications during freezing and frozen storage on quality parameters for fish. International Journal of Refrigeration. 2016;63:37-47. DOI: 10.1016/j.ijrefrig.2015.11.003
  71. 71. Bozzetta E, Pezzolato M, Cencetti E, Varello K, Abramo F, Mutinelli F, et al. Histology as a valid and reliable tool to differentiate fresh from frozen-thawed fish. Journal of Food Protection. 2012;75:1536-1541. DOI: 10.4315/0362-028x.jfp-12-035
  72. 72. Uddin M, Okazaki E, Turza S, Yumiko Y, Tanaka M, Fukuda Y. Non-destructive visible/NIR spectroscopy for differentiation of fresh and frozen-thawed fish. Journal of Food Science. 2005;70:c506-c510. DOI: 10.1111/j.1365-2621.2005.tb11509.x
  73. 73. Fasolato L, Balzan S, Riovanto R, Berzaghi P, Mirisola M, Ferlito JC, et al. Comparison of visible and near-infrared reflectance spectroscopy to authenticate fresh and frozen-thawed swordfish (Xiphias gladius L). Journal of Aquatic Food Product Technology. 2012;21:493-507. DOI: 10.1080/10498850.2011.615103
  74. 74. Reis MM, Martínez E, Saitua E, Rodríguez R, Pérez I, Olabarrieta I. Non-invasive differentiation between fresh and frozen/thawed tuna fillets using near infrared spectroscopy (Vis-NIRS). LWT – Food Science and Technology. 2017;78:129-137. DOI: 10.1016/j.lwt.2016.12.014
  75. 75. Wang W-L, Chen W-H, Tian H-Y, Liu Y. Detection of frozen-thawed cycles for frozen tilapia (Oreochromis) fillets using near infrared spectroscopy. Journal of Aquatic Food Product Technology. 2018;27:609-618. DOI: 10.1080/10498850.2018.1461156
  76. 76. Nieto-Ortega S, Melado-Herreros Á, Foti G, Olabarrieta I, Ramilo-Fernández G, Gonzalez Sotelo C, et al. Rapid differentiation of unfrozen and frozen-thawed tuna with non-destructive methods and classification models: Bioelectrical impedance analysis (BIA), near-infrared spectroscopy (NIR) and time domain reflectometry (TDR). Food. 2022;11:55. DOI: 10.3390/foods11010055
  77. 77. Agyekum AA, Kutsanedzie FYH, Mintah BK, Annavaram V, Zareef M, Hassan MM, et al. Rapid and nondestructive quantification of trimethylamine by FT-NIR coupled with chemometric techniques. Food Analytical Methods. 2019;12:2035-2044. DOI: 10.1007/s12161-019-01537-0
  78. 78. Ghidini S, Chiesa LM, Panseri S, Varrà MO, Ianieri A, Pessina D, et al. Histamine control in raw and processed tuna: A rapid tool based on NIR spectroscopy. Food. 2021;10:885. DOI: 10.3390/foods10040885
  79. 79. Liu Y, Wang C, Xia Z, Chen J. Nondestructive evaluation of biogenic amines in crayfish (Prokaryophyllus clarkii) by near infrared spectroscopy. Journal of Near Infrared Spectroscopy. 2021;29:330-338. DOI: 10.1177/09670335211054298
  80. 80. Shim K, Jeong Y. Freshness evaluation in chub mackerel (Scomber japonicus) using near-infrared spectroscopy determination of the cadaverine content. Journal of Food Protection. 2019;82:768-774. DOI: 10.4315/0362-028x.jfp-18-529
  81. 81. Pochanagone S, Rittiron R. Preliminary study on the determination of ppm-level concentration of histamine in tuna fish using a dry extract system for infrared coupled with near-infrared spectroscopy. ACS Omega. 2019;4:19164-19171. DOI: 10.1021/acsomega.9b02438
  82. 82. Ding R, Huang X, Han F, Dai H, Teye E, Xu F. Rapid and nondestructive evaluation of fish freshness by near infrared reflectance spectroscopy combined with chemometrics analysis. Analytical Methods. 2014;6:9675-9683. DOI: 10.1039/c4ay01839g
  83. 83. Zhang H-Z, Lee T-C. Rapid near-infrared spectroscopic method for the determination of free fatty acid in fish and its application in fish quality assessment. Journal of Agricultural Food Chemistry. 1997;45:3515-3521. DOI: 10.1021/jf960643r
  84. 84. Yildiz G, Wehling RL, Cuppett SL. Method for determining oxidation of vegetable oils by near-infrared spectroscopy. Journal of the American Oil Chemists' Society. 2001;78:495-502. DOI: 10.1007/s11746-001-0292-1
  85. 85. Zhou JJ, Wu XY, Chen Z, You J, Xiong SB. Evaluation of freshness in freshwater fish based on near infrared reflectance spectroscopy and chemometrics. LWT-Food Science and Technology. 2019;106:145-150. DOI: 10.1016/j.lwt.2019.01.056
  86. 86. Cheng J-H, Dai Q, Sun D-W, Zeng X-A, Liu D, Pu H-B. Applications of non-destructive spectroscopic techniques for fish quality and safety evaluation and inspection. Trends in Food Science & Technology. 2013;34:18-31. DOI: 10.1016/j.tifs.2013.08.005
  87. 87. Duan C, Chen C, Khan MN, Liu Y, Zhang R, Lin H, et al. Non-destructive determination of the total bacteria in flounder fillet by portable near infrared spectrometer. Food Control. 2014;42:18-22. DOI: 10.1016/j.foodcont.2014.01.023

Written By

Sonia Nieto-Ortega, Rebeca Lara, Giuseppe Foti, Ángela Melado-Herreros and Idoia Olabarrieta

Submitted: 10 May 2022 Reviewed: 07 June 2022 Published: 16 July 2022