Near-infrared spectroscopy: a non-invasive tool for quality evaluation of seafood

The present-day food supply chain is globalized, and this has led to an increase in awareness among consumers regarding seafood quality measures. For ensuring the quality parameters, although suitable preservation techniques are used, industries still face quality issues during production, storage, and distribution. However, traditional methods like sensory evaluation, biochemical and microbiological analysis are cumbersome, susceptible to variations in results, and time-consuming paving the way to explore alternative tools for quality assessment. Thus, non-destructive optical techniques involving visible and near-infrared wavelengths are now being applied for evaluating the quality of seafood on a real-time basis enabling online monitoring of all product samples. The infrared radiation of the electromagnetic spectrum is the invisible band between the visible and microwave region, having a wavelength range of 0.76 to 350 μm emitted out of substances whose temperature exceeds absolute zero, from the sun to electric heaters and gas-fired heaters. NIR techniques utilize the concept that molecules tend to absorb specific frequencies of light characterizing the corresponding structure of the molecules, enable rapid data acquisition, saves time, and determines multiple parameters. In Near-Infrared spectroscopy, the substance to be analyzed is illuminated with a broad-spectrum nearinfrared source, by absorption, transmittance, reflectance, or scattering. This article reviews the application of Near-Infrared Spectroscopy for the seafood quality assessment, different modes of spectra measurement, and various instruments used in NearInfrared Spectroscopy. Near-Infrared Spectroscopy coupled with chemometrics is a propitious tool useful in the prediction of several fish and seafood’s quality attributes and authentication of various fish-and-fishery products.


Introduction
Seafood is an important food commodity providing a balanced diet to human with several health benefits. These include shellfish and fish, with a high content of protein, long-chain omega-3 polyunsaturated fatty acids, vitamins, and minerals (Weichselbaum et al., 2013). It is highly perishable in nature because of the low content of connective tissue, the presence of autolytic enzymes, high water activity (aw), and neutral pH (Hassoun and Karoui, 2017). This necessitates the adoption of suitable preservation techniques to sustain the quality attributes and ensure its safety for human consumption. Despite the best techniques adopted for seafood preservation, the industry often faces quality issues during production, storage and distribution.
The globalization of the food supply chain and the shift in consumer behavior and eating habits have led to growing awareness regarding seafood quality and the preservation methods adopted (Karoui et al., 2010). Therefore, seafood quality and it's authentication is a major concern worldwide; and the processors and regulatory agencies today are required to assure seafood quality based on traceability of the product (geographical origin), labeling (species), production method (farmed or wild), processing conditions, substitutions or additives used, etc. (Hassoun and Karoui, 2017). An important aspect of seafood quality is an analysis of quality indices using several methods. Traditionally, sensory evaluation, chemical and microbiological analysis of the seafood is done to determine the quality. In the seafood industries and retail stores, freshness evaluation of fish products is measured by evaluating through sensory means which is either descriptive or discriminative in nature (Sant'Ana et al., 2011). Chemical analysis of volatile compounds (Castro et al., 2006), measurement of lipid oxidation (Li et al., 2011), an assay of nucleotide and amine degradation products are done to determine the degree of freshness of seafood (Li et al., 2011 andOnal et al., 2013). Additionally, texture measurements using texture analyzers and electrical properties using Torrymeter and Fischtester are applied for seafood quality monitoring (Coppes-Petricorena, 2010;Oehlenschlager, 2005;Oehlenschlager, 2014).
However, traditional methods have several disadvantages like sensory evaluation is not considered ideal for working with a large amount of sample, chemical analysis is cumbersome, slow, may produce toxic waste and destructive to the product causing irreversible damage to it. The microbiological analysis identifies biological organisms and no other contents. These traditional techniques are also susceptible to variations in results and time-consuming paving the way to explore alternative tools for quality assessment. This led to the development of newer technologies that are non-destructive. One such approach is an optical technique involving the use of visual and near-infrared wavelengths. These optical methods determine results on a real-time basis that helps in online monitoring of all product samples (Alandar et al., 2013). Spectroscopy in the visible (VIS) and infrared (IR) regions of the electromagnetic spectrum as an analytical technique has gained the attention of the researchers and processors for measuring the quality of seafood. Spectroscopic techniques are an optical method that enable rapid data acquisition, save time, and simultaneously determine multiple parameters (Hassoun and Karoui, 2017).

History of near-infrared spectroscopic analysis
Near-Infrared energy as defined by the International Union of Pure and Applied Chemistry (IUPAC), extends from 780-2500nm. It was discovered by Fredrick William Herschel in 1800 in an experiment wherein he projected a rainbow using a prism and observed that heating effect increased from the blue to the red zone and found maximum heating effect into the black area beyond the end of his spectrum i.e., the invisible part of the spectrum (Davies, 1998). The present-day applications of NIR spectroscopy are done depending on the earlier researches by Karl Norris who was awarded "First Fellow of Near-Infrared Spectroscopy" by the International NIR community during the 8 th conference of The International Committee for NIR Spectroscopy, in the year 1997, in Germany. Norris is credited with the award due to his research on the use of Near-Infrared spectroscopy in the determination of analytes in agricultural commodities especially the protein content of wheat (Davies, 1998). Ever since Norris pioneered the use of NIR spectroscopy, a lot of research was done on its application in agriculture and food. Norris and Hart published the works on the direct spectrophotometric determination of moisture content of grain and seeds in "Proceedings of the International Symposium on Humidity and Moisture" of the US Department of Agriculture (USDA) (Norris and Hart, 1996). Later, Phil Williams from the Canadian Grain Commission established an analysis of wheat protein all-over Canada in1975. During the early days, the application of NIR was only used for quality evaluation of products with lowmoisture content such as cereal grains like wheat, corn, and soybean. This is because of water absorption in the region of 1900-2500 nm wavelengths. Later on, silicon sensors were developed to enable data acquisition even in the shorter wavelengths of 700-1100 nm. Shorter wavelengths have better penetration which allowed non-destructive evaluation of thick samples. This enabled analysis of high moisture samples which are thick as well, like intact kernels, fruit, meat, and fish (Alandaret al., 2013). NIR spectroscopy measures the chemical properties based on the changes in spectra. Therefore, the technique needs to be calibrated based on a reference method for analysis of the target parameter (Osborne, 2006). Over the years, data processing and mining are used to calibrate the NIR spectroscopy techniques based on chemometric tools (Hassoun and Karoui, 2017). Alandar et al. (2013) defined chemometrics as the optical measurement results correlated with the chemical analysis and statistics. Presently, the NIR spectroscopy is used for the prediction of various quality parameters and to authenticate seafood. The research in the NIR spectroscopy is mainly aimed at the assessment of freshness, chemical, and microbiological parameters of seafood (Liu et al., 2013).

Near-Infrared waves and Near-Infrared spectroscopy
The infrared radiation of the electromagnetic spectrum is the band between the visible region and microwaves having a wavelength range of 0.76 to 350 µm. This is the invisible part of the spectrum (Fig 1.). It constitutes 66% of solar radiation and causes heat. Infrared energy is emitted from any substances whose temperature exceeds absolute zero, from the sun to electric heaters and gas-fired heaters. Almost all sources of light and heat generate some energy in the infrared region (Aboud et al., 2019). Infrared radiation is categorized into three spectrums: 1. Near-IR (NIR) with wavelength ranging from 0.75 to 1.4 µm. 2. Mid-IR (MIR) with a wavelength between 1.4 and 3 µm. 3. Far-IR radiation (FIR) with a wavelength between 3 and 1000 µm. The heating caused by infrared radiation depends on several factors: a. the emission spectrum of the source b. the temperature of the source c. the direction of the emission falling on the surface d. absorption and scattering of the medium The process of scattering and absorption of the medium is important in heating. Absorption is the process of converting radiation to any other forms of energy and scattering causes the radiated energy to be directed to another destination from the original direction of propagation through reflection, refraction, and deviation ( Figure 2). These result in the weakening of electromagnetic radiation (Aboud et al., 2019). The Near-Infrared, i.e., the wavelength range 780nm to 2500nm of the spectrum and the technique based on the absorption within this range of electromagnetic radiation is referred to as NIR spectroscopy (Osborne, 2006). Williams and Norris, (1987) noted that in the NIR region, the fundamental molecular vibrations of chemical bonds C-H, O-H, N-H, C = O, and other functional groups are detected and concentrations of food components like protein, fat, water, and carbohydrate having these bonds, can be determined using classical absorption spectroscopy. Although the method requires to be calibrated against a standard method as reference using multivariate mathematics (chemometrics) (Osborne, 2006). Food substances contain various functional groups which when irradiated with NIR light, absorbs the light with frequencies that match with the vibrations of a particular functional group. The rest of the frequencies is either transmitted or reflected (Foley et al., 1998). The light undergoing reflection, transmission, or absorption into was detected and the biochemical constituents of food were determined based on the calibration using chemometrics (André and Lawler, 2003).

Modes of Measurement of Near-Infrared Spectra
There are four standard modes of measurement of NIR spectra of samples such as transmission, reflection, transfection, and interaction (interactance) (Alandar et al., 2013). The choice of the model depends on several factors, of which the type of sample to be tested is of importance.

Transmission mode
In the transmission mode of measurement, the light incident on the sample passes through it, and part of the light that exits the sample is measured at a point exactly opposite to the source of light. In solid samples, the light source is placed at a right angle to the detector (Fig. 3)

Reflection mode
In this mode, the source of light is positioned on the same side as that of the sample. The detectors are usually aligned at 45 o angle to the sample plane for avoiding surface reflection. An integrating sphere is used in certain cases for concentrating the reflected light so that the light can be captured and integrated from different directions into the detector (Fig. 4)

Transflection mode
This is a combination of transmission and reflection wherein opposite to the light source a reflector is positioned and the transmitted light passing through the sample is reflected. This reflected light then once again travels through the sample before reaching the detector. The alignment of the detector and source remains the same as in reflection measurement mode. In transflection mode, the physical path length is double the thickness of the sample as the transmitted light passes back within the sample after being reflected (Fig. 5)

Interactance mode
This is also a combination of transmission and reflectance with certain modifications to make it suitable for solid samples. The illumination of the sample is achieved using fiber optic cables and it remains in direct contact with the sample surface or holder. Such direct contact avoids the chances of surface reflection and maximizes the penetration depth. The light passing through the sample is either reflected, transmitted, and/or absorbed. Source and the fibers are aligned on the same side and the transmitted and reflected light return to the detector (Fig. 6).

Liquid samples
Liquid samples such as vegetable oil, filtrated fruit juice are clear in nature and cause low scattering. Hence, such samples can be placed in a quartz cuvette, disposable glass vial, or a transflectance sample holder. The measurement model used for analysis is transmission or transflection. The liquid samples also provide an additional benefit of adjusting the thickness of the sample based on the absorptivity of major constituents.

Ground and relatively small solid samples
This category includes samples like the ground and intact cereal grains, soybean meals, powder milk, and crushed processed foods. These food samples exhibit a strong scattering characteristic which necessitates that the NIR spectra be measured using reflectance. The samples to be tested are generally positioned in glass window sample holders or quartz sample holders.

Relatively large samples that require non-destructive or noninvasive measurement
Samples such as meat, fish, fruit, and nut kernels which are sold intact fall under this category. For such samples, the composition of parameters such as fat, protein, and moisture are not distributed evenly throughout and variations exist in different parts of the sample like surface or sample core.
Hence, for these samples the interactence and transmission modes of measurements are applied. Silicon detectors are used in data acquisition which generally is in the range of 700-1100 nm (Alandar et al., 2013). The food commodities and the parameters measured using NIR spectroscopy is given in table1

Soluble solids contents and pH in orange juice
Chemometrics and Vis−NIRS Cen et al. (2006) Texture and colour of dry-cured ham Visible and near infrared spectroscopy using a fiber optic probe Garcia-Rey et al.
Dry matter, fat, pH, vitamins, minerals, carotenoids, total antioxidant capacity, and color in fresh and freeze-dried cheeses Visible-near-infrared reflectance spectroscopy Lucas et al. (2008) .

Sample presentation techniques in NIR spectroscopy
The food samples can be in different forms like liquid, slurry, powdered, or solid, and therefore, suitable sample presentation techniques must be available for each type. The different types of sample presentations are

Diffuse Transmittance
Light passing through a food sample can be reflected, transmitted or absorbed. If scattering is absent as in clear transparent liquids, Beer's law is applied for defining a proportionality between transmittance and concentration of the absorbing food sample and path length. However, in food samples such as meat and cheese, scattering of the light occur and Beer's law is not applicable. This occurs as the scattering causes a change in the length of the path and the length becomes undefined. Diffuse transmittance is applied for such samples and performed in a range of 800-1100 nm wavelength (Fig. 8). This is near-infrared transmittance (NIT) and it is ideal for analyzing whole grains. For whole grains, the sample is presented in a receptacle termed as hopper, from where grains in predetermined quantities are dispensed into the measurement chamber. After the analysis is completed, the grains are discharged into a collection tray. This sample presentation is also used for on-line measurements. For meat and cheese, 1-2 cm thickness is usually maintained to enable proper data acquisition.

Diffuse reflectance
Food samples, in which the maximum portion of the radiation incident on the surface is reflected by regular or specular reflection, the amount of absorption is nil and transmittance is negligible. In the spectra range of 1100 to 2500 nm, most of the incident radiation is scattered and this situation is termed diffuse reflectance (Nilsen and Heia, 2009). Diffuse reflectance is applicable for samples like flours and powdered milk and the NIR spectroscopy for such samples is done in reflectance mode. In reflectance mode, the radiation reflected is measured by a set of detectors set at 45° to the incident beam (Fig. 9). Alternatively, an integrating sphere may be used for focussing the diffused reflection on a single detector. The samples are placed in a 1 cm deep sample cell, then enclosed with a quartz window. The cell is then put inside the instrument for NIR spectroscopy measurement. This system of sample presentation can also be modified for on-line measurements using a powder sample. Liquid samples that are otherwise measured by transmittance can also be measured by diffuse reflectance by modifying the sample presentation. Liquid samples are placed in a ceramic tile beneath the sample and the radiation after being transmitted through the sample will be reflected from the ceramic tile and transmitted back finally reaching the detector. This is called transflectance as discussed earlier. Further, fiber-optic probes are used for presenting the samples for evaluation by using interactance mode. Such an arrangement is particularly useful for large samples such as fruit.

On line samplers
NIR instruments are relatively low cost, capable of measuring the target analytes rapidly and in a non-destructive manner, which makes them a suitable instrument for online analysis.

Fig. 9 Diffuse reflectance analysis for powdered sample (Osborne, 2006)
The NIR used in the online analysis is classified into three categories

Remote Sensor
This system is characterized by sensors that are set remote from the sample and has application in the food industry for the determination of different constituents like protein, fat, and moisture in whole biscuits, flour, and potato crisps. It is also used for sorting fruits and monitoring bread dough development.

By-pass Sampler
This sample presentation is used for measuring protein content in flour and is used in online systems to control the addition of gluten to flour. The sample presentation is achieved by taking direct samples from the flour stream into a sample cup using a suitable mechanism for NIR measurement.

Fibre-optic prob
This sample presentation is applied for the production of extruded snack foods, online analysis of dairy products, meat, fruit, and beer. In extrusion cooking, NIR is used to monitor the changes in starch structure. Transflectance measurement mode is generally used and the probe is installed in the die of the extruder (Osborne, 2006).

Instruments
Near-Infrared spectroscopy is applied depending on the type of instrument chosen. It is mainly of three types (Osborne, 2006)

Monochromators
Grating monochromators are versatile instruments used for measurements in both visible and NIR spectrum. The measurement modes can be either transmittance or reflectance. There are three types of detectors: silicon (400-1100 nm), indium gallium arsenide (800-1700 nm) and lead sulphide (1100-2500 nm) detector.
A combination of silicon and lead sulphide detector may also be integrated in a system to get a wide range of spectrum inbetween 400-2500 nm. Acousto-optically tuneable filter (AOTF) is another type of dispersive monochromator comprising of TeO2 crystal, through which acoustic wave is produced at 90 O angle to the incident light beam. The crystal, thus, behave like longitudinal diffraction grating with a periodicity equal to the wavelength of sound across the material.

Diode array spectrometers
These spectrometers use an array of IR emitting diodes, that function as the light source as well as wavelength selection system. The instrument works fast and non-invasive in nature and operates within the range 400-1700 nm. Diode array spectrometers are useful for ultra-rapid online measurements.

Filter Instruments
In these instruments, limited number of interference filters is added, each representing the absorption spectrum of the required applications such as protein, moisture and oil in agricultural samples. These equipments are made for routine analysis in the laboratory or for online system.

Application of Near-Infrared spectroscopy in seafood
Near-infrared spectroscopy in combination with the chemometric tools is considered a promising analytical technique for application in seafood industry. The technique is shown to be useful in prediction of several fish quality attributes and to authenticate fish and other seafoods. The seafood research involving application of NIR spectroscopy is focussed around evaluating freshness, chemical and microbiological parameters (Liu et al., 2013). The different applications of the technique in seafood industry are given in table 2. The trends in use of NIR spectroscopy in seafood industry (table 2) indicates the following broad areas of application suggesting that NIR spectroscopy is a promising tool that has a widespread application in seafood industry. i.
Evaluation of seafood freshness ii.
Authenticity of seafood and detection of fraud/mislabeling iii.
Detection of microbiological behaviour in seafood iv.
Determination of spoilage and estimation of shelf life of seafood v.
Prediction of chemical composition of seafood and its quality vi.
Detection of contaminants/adulterants in seafood

Live, shucked, and freeze-dried Abalone
Glycogen concentrations in the foot muscle of cultured abalone NIR spectroscopy, reflectance, 350-2500 nm, PLSR NIR is a rapid method to monitor the factors (glycogen content) associated with abalone quality Fluckiger et al. (2011)

Red Crayfish
Inorganic arsenic content VIS/NIR spectroscopy NIR can be used to screen inorganic arsenic content in red crayfish Font et al.

Crab meat
Authenticity and adulteration VIS/NIR spectroscopy NIR is applied for species authenticity and detection of adulteration in crab meat Gayo and Hale, (2007) Conclusion NIR spectroscopy is a valuable and useful analytical tool that could be applied for the rapid assessment of fish quality and authenticity. It can provide extensive information on changes of various chemical components (proteins, fat, and water) during handling, processing, and storage. The disadvantage of this technique is the need to develop calibration models based on the use of chemometrics to predict unknown samples.
However, features like low energy costs, portable nature of the instrument, accuracy level, possibility to align it for online determination of parameters and rapid testing ability makes this technique a highly promising alternative for assessment of seafood quality and authenticity.