CHAPTER 12
Recent Advances for Rapid Detection of Quality and Safety of Fish by Hyperspectral Imaging Analysis
Chao-Hui Feng
The University of Tokyo
Chengdu University
Sichuan Agricultural University
Yoshio Makino and Masatoshi Yoshimura
The University of Tokyo
Francisco J. Rodríguez-Pulido
University of Seville
Contents
12.1 Introduction: Background and Driving Forces
12.2 Freshness
12.3 Physical Properties
12.4 Chemical Compositions
12.5 Nematodes Inspection
12.6 Microbial Spoilage Inspection
12.7 Conclusions
Acknowledgments
References
12.1 Introduction: Background and Driving Forces
Fish has been one of the most important components of several and nutritious diets in the world. Its contribution to human health is well documented, being an essential topic for researchers. The reason of its importance lays in its nutritional composition. Fish is an important source of protein, vitamins, trace elements, and other nutritional components, with the most significant being, without doubt, the polyunsaturated fatty acids which are related to the long-chain ω-3 eicosapentaenoic acid and docosahexaenoic acid (DHA) (Hernández-Martínez et al., 2013). ω-3 fatty acids are extremely important for the neural development in the infant in utero and the first few years after birth (Uauy et al., 2000). Fish and seafood are the major elements for many dietary guidelines, such as Mediterranean and Japanese diets. Humans lack the enzymes necessary to produce ω-3 fatty acids that have to be obtained from the diet or produced in vivo from diet-derived ω-3 fatty acid precursors such as α-linolenic acid. Similarly, DHA, which is a primary structural component of the human brain, cerebral cortex, skin, and retina, is an essential nutrition for infants and can be greatly obtained from fish or other types of marine-derived food (Montaño et al., 2001). Fishes such as salmon, tuna, mackerel, and sardines are particularly rich in polyunsaturated fatty acids (PUFAs). Nowadays, high quality of the fish is highly demanded from both consumers and fishery industries. Significant efforts have been made by the industries to enhance the quality and safety of the aquatic and seafood products by using new technologies such as novel cooling (Feng et al., 2013a, b; 2014a, b, c; 2016; Feng and Sun, 2014; Feng and Li, 2015), freezing (Crane et al., 2016; Sánchez-Alonso et al., 2018), and packaging (Kuswandi et al., 2012; Heide and Olsen, 2017). Meanwhile, manufactures, consumers, and government regulators increasingly and imperatively require reliable, rapid, and practical analytical methods and techniques for safety detection and assessment of the aquatic products (Kamruzzaman et al., 2015a).
However, high-quality and safe fish products are also closely related to human health and dietary benefits. For example, the nematode, as a common parasite in fish muscle, exerts a negative impact on consumers’ purchase decision (Heia et al., 2007). Anisakis, as a genus of parasitic nematodes that have lifecycles involving fish and marine mammals, can cause anisakiasis or an allergic reaction (Berger and Marr, 2006). This parasite poses a high risk to human health via intestinal infection with worms from the eating of uncooked or underprocessed fish. Within a few hours of ingestion of fish with Anisakis, the parasitic worm tries to burrow from the intestinal wall to escape. As it cannot penetrate and gets stuck and eventually dies in the human body, it arouses an immune response from our body and forms a ball-like structure surrounding the dead worms, leading to severe abdominal pain and vomiting due to the block of the digestive system caused by the ball-like structure. If the worm (larvae) passes into the bowel or large intestine, it may cause a severe eosinophilic granulomatous response and symptoms mimicking Crohn’s disease.
Herein, a noninvasive and rapid analysis and evaluation of fish and other seafood quality is highly demanded to ensure the humans health. Hyperspectral imaging (HSI), as an emerging and innovative tool, has been intensively exploited and investigated for nondestructively detecting the food and food products during the past few years (Kamruzzaman et al., 2015a, b, c; 2016a, b, c, d; Feng et al., 2018). In this chapter, the detection of fish freshness, evaluation of physical properties and chemical composition and inspection of microbial spoilage in fish by HSI are discussed.
TABLE 12.1 Application of Hyperspectral Imaging for the Detection of Fish Properties
12.2 Freshness
Fish freshness is always regarded as one of the most important integrated quality attributes for evaluating the quality of the fish. This critical index affects safety, nutritional quality, and edibility of fish, which may be caused by physical, chemical, microbiological, and biochemical processes (Cheng et al., 2013). Table 12.1 summarizes the recent applications to freshness detection in fish. Although the conventional well-established methods and analytical techniques such as sensory analysis, high performance liquid chromatographic, spectrophotometric, and electrochemical approaches are able to detect the fish freshness, the aforementioned methods are generally time-consuming, tedious, and relatively expensive (He and Sun, 2015a). Moreover, a skilled personnel is required to perform the experiments, which is difficult to be used in on-field applications (Cheng et al., 2014). Compared with traditional methods, HSI is a environmental friendly, toxic-free, noninvasive, time-saving technique (Kamruzzaman et al., 2016a; Feng et al., 2018). The spectroscopic changes in fresh salmon stored under different atmospheres (air, modified atmosphere packaging, and 90% vacuum) were studied using HSI (400–1,100 nm) (Sone et al., 2012). The main spectral changes occurred at wavelengths of 606 and 636 nm, which could be used to classify fresh salmon fillets with different types of packaging used for the storage (Sone et al., 2012). The authors attributed the spectroscopic changes to the different oxidation states of the heme proteins in the salmon (Sone et al., 2012). Similarly, HSI with the short wave near-infrared SW-NIR range was employed to evaluate the shelf life of the vacuum-packed chilled smoked salmon (Ivorra et al., 2016). The partial least square (PLS) regression model using fat tissue presented a better cross validation outcome (Rcv2 = 0.80). A support vector machine model successfully classified the samples stored for 0, 10, 20, 40, and 60 days, and the prediction accuracy being 87.2% (Ivorra et al., 2016).
Fillets of cod under different programs of freezing, thawing, and storage were investigated by HSI (Washburn et al., 2017). The results showed that samples frozen to −40°C and −20°C can be accurately predicted using the region of 450–600 nm, with accuracy rate of 98% and 99%, respectively. The reason for differentiating with fresh, once thawed, and twice thawed samples is because of the increased light scattering with each freeze-thaw cycle. It was proposed that the increased light scattering is as the result of the denaturation of proteins during freezing and thawing (Love, 1962).
K value, as an important freshness index, is widely used for the assessment of chemical spoilage and nucleotide degradation (Cheng et al., 2015b). As the increase in K value indicates the endogenous enzyme actions and bacterial spoilage (Lowe et al., 1993), the K value of 20% is regarded as the optimal freshness limit, while the K value of 60% being the rejection point (Ehira, 1976). The K values in grass carp and silver carp fillets were investigated by using HSI (400–1,000 nm). Seven optimal wavelengths, i.e., 432, 455, 588, 635, 750, 840, and 970 nm, were selected in the study by Cheng et al. (2015b). The simplified PLS model using the optimal wavelengths shows excellent performance ability (Rp2 = 0.935; Root Mean Square Error of Prediction (RMSEP) = 5.17%), with comparable accuracy to the original models using the whole wavelengths. According to the distribution map of fish fillet (Figure 12.1), the storage for 2 days at 4°C ± 1°C showed the limit freshness for the fish and it started to spoil from the bottom of the fish sample (Cheng et al., 2015b).
FIGURE 12.1 Distribution maps of K value in fish fillets at four different K values (freshness): (a) K = 24.2%, storage for 0 days; (b) K = 45.6%, storage for 2 days; (c) K = 78.1%, storage for 4 days; (d) K = 89.8%, storage for 6 days (cited from Cheng et al., 2015b).
Quality index method (QIM) is a standard to evaluate and describe the fish freshness. Although it is the one of the most wholesome and straightforward ways to describe the fish freshness, it is subjective, time-consuming, and not of practical use for the large-scale commerce. Cheng and Sun (2015a) predicted sensory quality index scores (QIS) of grass carp fish fillet by using HSI in tandem with data fusion technique. Whitworth et al. (2010) also measured QIM by HSI. They acquired images from exterior surface and fillets of cod and measured the average spectra from areas having demonstrated the greatest potential for discrimination of changes along time. After applying a segmentation criterion for selecting the region of interest, they divided the fillets into head, middle, and tail pieces of equal size for considering the average spectra. Then, the authors used PLS regression with full cross validation for predicting some reference measurements of freshness, such as storage time and Torry scores for raw and cooked cod. They proved that HSI improves the results obtained by visible/NIR spectroscopy. In fact, the best results were obtained for the head end of the fillet (R2 = 0.92 and SECV = 1.66 days). Rigor mortis is also related to freshness.
12.3 Physical Properties
Color is an important criterion for evaluating the quality of the foodstuffs as consumers use color as the first visual attribute to judge the quality of the foods, resulting in purchasing intention. For example, redder salmon is preferable for consumers as it is associated with better quality, higher flavor, and freshness (Troy and Kerry, 2010). Wu et al. (2012) investigated the color of salmon from Scotland, Ireland, and Norway using HSI with the spectral range of 897–1,753 nm. Three local absorption maxima were reported to appear at 980, 1,220, and 1,450 nm, which may due to the presence of water, fat, and protein in the sample, respectively. The PLSR model using full spectral range achieved Rp2 of 0.864, 0.736, and 0.798 for L*, a*, and b*, respectively. For the reduced wavelengths selection, the final prediction models were considered as the multiple linear regression (MLR) model, with Rp2 of 0.869, 0.729, and 0.788 for L*, a*, and b*, respectively. Approximate 98.5%, 98.5%, and 96.5% of wavelengths were reduced for L*, a*, and b* during wavelengths selection, respectively, which will significantly reduce the high dimensionality with redundancy and collinearity among spectral wavelengths and facilitate the designing of a multispectral imaging system (Wu et al., 2012).
Apart from the color that influences the consumer’s preference, texture is also an important index to evaluate food or processed food products to control different processing operations (Ma et al., 2017). The textural properties (Warner–Bratzler shear force [WFSF], hardness, gumminess, and chewiness) of fish fillets for vacuum freeze-drying at different durations (3, 6, 12, 18, 24, 30, and 36 h) were investigated using HSI (400–1,000 nm). The authors stated that choosing optimal wavelengths did not significantly reduce the accuracy of the model for predicting WBSF in comparison with the full wavelengths. The Rp2 of model using the selected wavelengths even improved a little bit (0.81%) as compared to the full wavelengths. The predicting ability in the study by Ma et al. (2017) was reported to be better th...
