FTIR Spectroscopy and Multispectral Imaging Versus Microbiological Analysis

foods

Article Quest of Intelligent Research Tools for Rapid Evaluation of Fish Quality: FTIR Spectroscopy and Multispectral Imaging Versus Microbiological Analysis

Maria Govari 1, Paschalitsa Tryﬁnopoulou 1 , Foteini F. Parlapani 2, Ioannis S. Boziaris 2, Efstathios Z. Panagou 1 and George-John E. Nychas 1,*

1 Laboratory of Microbiology and Biotechnology of Foods, Department of Food Science and Human Nutrition, School of Food and Nutritional Sciences, Agricultural University of Athens, Iera Odos 75, 11855 Athens, Greece; [email protected] (M.G.); [email protected] (P.T.); [email protected] (E.Z.P.) 2 Laboratory of Marketing and Technology of Aquatic Products and Foods, Department of Ichthyology and Aquatic Environment, School of Agricultural Sciences, University of Thessaly, Phytoko Street, 38446 Volos, Greece; [email protected] (F.F.P.); [email protected] (I.S.B.) * Correspondence: [email protected]

Abstract: The aim of the present study was to assess the microbiological quality of farmed sea bass (Dicentrarchus labrax) fillets stored under aerobic conditions and modified atmosphere packaging ◦ (MAP) (31% CO2, 23% O2, 46% N2,) at 0, 4, 8, and 12 C using Fourier transform infrared (FTIR) spectroscopy and multispectral imaging (MSI) in tandem with data analytics, taking into account the results of conventional microbiological analysis. Fish samples were subjected to microbiological analysis (total viable counts (TVC), Pseudomonas spp., H2S producing bacteria, Brochothrix thermosphacta, lactic acid bacteria (LAB), Enterobacteriaceae, and yeasts) and sensory evaluation, together with FTIR Citation: Govari, M.; Tryfinopoulou, and MSI spectral data acquisition. Pseudomonas spp. and H2S-producing bacteria were enumerated P.; Parlapani, F.F.; Boziaris, I.S.; at higher population levels compared to other microorganisms, regardless of storage temperature Panagou, E.Z.; Nychas, G.-J.E. Quest and packaging condition. The developed partial least squares regression (PLS-R) models based on of Intelligent Research Tools for Rapid Evaluation of Fish Quality: the FTIR spectra of fish stored aerobically and under MAP exhibited satisfactory performance in the 2 FTIR Spectroscopy and Multispectral estimation of TVC, with coefficients of determination (R ) at 0.78 and 0.99, respectively. In contrast, 2 Imaging Versus Microbiological the performances of PLS-R models based on MSI spectral data were less accurate, with R values of Analysis. Foods 2021, 10, 264. 0.44 and 0.62 for fish samples stored aerobically and under MAP, respectively. FTIR spectroscopy https://doi.org/10.3390/ is a promising tool to assess the microbiological quality of sea bass fillets stored in air and under foods10020264 MAP that could be effectively employed in the future as an alternative method to conventional microbiological analysis. Academic Editor: Alexander Govaris Received: 10 December 2020 Keywords: sea bass fillets; FTIR spectroscopy; multispectral imaging; modified atmosphere packag- Accepted: 24 January 2021 ing; PLS-R Published: 28 January 2021

Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in 1. Introduction published maps and institutional affil- iations. Aquaculture is the most promising solution for the provision of animal protein for the increasing world population and the reduction of international poverty [1,2]. Nevertheless, a significant percentage of the total fish production (approximately 35%) is lost along the food supply chain due to mechanical damage, microbial growth, and/or microbial contamination, threatening food security and sustainability [3]. Of this quantity of fish Copyright: © 2021 by the authors. Licensee MDPI, Basel, Switzerland. production, 27% is wasted in post-farm gate practices, such as handling, storage, etc. [3]. This article is an open access article Microbial activity is the most important cause of fish discards in post-harvest handling, distributed under the terms and storage, processing, and distribution (post-harvest losses). Under particular conditions in conditions of the Creative Commons which fish are exposed after catch (e.g., experiencing temperature and time effects), bacteria Attribution (CC BY) license (https:// can grow and produce metabolites that result in quality loss and the sensory rejection of creativecommons.org/licenses/by/ the product [4]. Additionally, enzymatic activity and chemical oxidations can also occur, 4.0/). which are responsible for the production of spoilage-related compounds in fish tissue [4].

Foods 2021, 10, 264. https://doi.org/10.3390/foods10020264 https://www.mdpi.com/journal/foods Foods 2021, 10, 264 2 of 12

The most common preservation method used to retard the deterioration of fresh fish is refrigeration under aerobic conditions. However, the presence of oxygen at levels of 21% can favor the growth and metabolic activity of the most important microbial spoilers of fish such as Pseudomonas spp., and Psychrobacter spp. [4]. On the other hand, packaging in atmospheric conditions enriched in CO2 (modified atmosphere packaging (MAP)) has been recognized as an effective method for food preservation, including for seafood such as the European sea bass [5]. In MAP, the use of oxygen and carbon dioxide at different concentration levels can inhibit the growth of such spoilers compared to aerobic conditions, thus significantly extending (by 20–100%) the shelf life of the product [5,6]. The evaluation of spoilage status is important in determining the remaining shelf life of fish. It has been reported that fish usually spoils when its total microbial population exceeds 107 CFU/g [4,7]. Bacterial enumeration based on agar culture media is reliable, but is also laborious and time consuming. Meanwhile, the fish industry and food authorities require fast, accurate and cost-effective methods to monitor microbial population and evaluate the microbiological quality of fish. To meet stakeholders’ demands, food microbiologists must develop intelligent research-led approaches and establish rapid, reliable and easy-to-use methodologies and technologies for the evaluation of fish quality in an extremely short time. This scenario will minimize food and economic losses for the aquaculture sector, fish processors, and distributors, thereby contributing to sustainable development goals [3]. In recent years, several hyperspectral and vibrational spectroscopy techniques have been developed to evaluate chemical composition and microbial spoilage in foods [8–11]. Fourier transform infrared (FTIR) spectroscopy, when combined with partial least squares regression (PLS-R) analysis, has been reported as a promising analytical method for the detection and quantification of spoilage bacteria in meat [12,13] and, recently, in fish [14]. Multispectral imaging (MSI), a promising rapid and non-invasive technology, obtains spatial and spectral information to evaluate the food spoilage and quality characteristics of food, including fish [15–17]. European sea bass (Dicentrarchus labrax) is one of the most important fish species farmed in Mediterranean countries. In 2019, the European Union was the largest producer of sea bass in the world, providing 80% of the annual world production, while Greece produced just over one half (51%) of the total EU production of European sea bass [18]. Apart from whole or gutted fish, fresh sea bass is also distributed in the EU market as fillets, stored either aerobically or under MAP in refrigerated conditions. The aim of the present study was to investigate the potential of FTIR spectroscopy and MSI as means of evaluating the microbiological quality of sea bass fillets packaged aerobically and under MAP at different isothermal conditions.

2. Materials and Methods 2.1. Fish Fillet Samples, Storage Conditions and Sampling Farmed sea bass (Dicentrarchus labrax) fillets (about 250 g each) were obtained from Selonda Aquaculture S.A. (Athens, Greece) and transported to the laboratory on ice 48 h after harvesting. The sea bass fillets were supplied in packs in air or under modified atmosphere packaging (MAP) conditions (31% CO2, 23% O2, 46% N2). Upon arrival, the packages were stored at isothermal conditions (0, 4, 8, and 12 ◦C) in high precision (±0.5 ◦C) incubators (MIR-153, Sanyo Electric Co., Osaka, Japan). The objective was to simulate the available temperatures in retail fish outlets. The temperature profile of incubators was recorded at 15-min intervals throughout the experiment by means of electronic temperature- monitoring devices (COX TRACER, Cox Technologies Inc., Belmont, NC, USA). Sea bass fillets were analyzed upon arrival to the laboratory (t = 0) and at regular time intervals depending on storage temperature. Specifically, fish fillet samples in both packaging trials were analyzed at 24, 12, 8, and 4–6 h during storage at 0, 4, 8, and 12 ◦C, respectively. Sea bass fillets under aerobic packaging conditions were stored at 0 and 4 ◦C for 287 and 143 h, respectively, whereas other fillets were stored at 8 and 12 ◦C for 94 h. Moreover, sea bass fillets under MAP conditions were stored at 0, 4, 8, and 12 ◦C for 545, Foods 2021, 10, 264 3 of 12

377, 281, and 209 h, respectively. For each sampling point, duplicate sea bass ﬁllets from randomly selected packages under aerobic or MAP conditions at each storage temperature were subjected to: (i) microbiological analysis and pH measurement, (ii) sensory analysis, (iii) FTIR spectroscopy measurements, and (iv) MSI spectra acquisition.

2.2. Microbiological Analysis For microbiological analysis, 25 g fish fillets were transferred aseptically in a stomacher bag (Seward Medical, London, UK), containing 225 mL of sterilized peptone saline diluent (0.1%, w/v, peptone, Neogen, Lansing, MI, USA) and 0.85%, w/v, sodium chlo- ride, (Radiova, Praha, Slovakia), and homogenized in a Stomacher device (Lab Blender 400, Seward Medical, London, UK) for 60 s at room temperature. Serial 10-fold dilutions were prepared with the same diluent, and duplicate 0.1 mL portions of the appropriate dilutions were spread on the following media: Plate count agar (Biolife, Milano, Italy) for the enumeration of total viable counts (TVC), incubated at 25 ◦C for 3 d; pseudomonas agar base supplemented with Cephaloridine Fusidin Cetrimide (Neogen, Lansing, MI, USA) for Pseudomonas spp., incubated at 25 ◦C for 2 d; streptomycin sulphatethallous acetate cycloheximide (actidione) agar (Biolife, Milano, Italy) Milano, Italy for B. thermosphacta, incubated at 25 ◦C for 3 d; rose bengal chloramphenicol agar (Lab M, Lancashire, UK) for the enumeration of yeasts, incubated at 25 ◦C for 5 d. Similarly, duplicate 1.0 mL portions of the appropriate dilutions were mixed with the following media: de Man–Rogosa–Sharpe agar (Biolife, Milano, Italy) for lactic acid bacteria (LAB) incubated at 25 ◦C for 3–5 d; violet red bile glucose (Biolife, Milano, Italy) agar for the enumeration of Enterobacteriaceae, ◦ incubated at 37 C for 24 h; iron agar (IA) for the enumeration of H2S producing bacteria by counting black colonies, after incubation at 25 ◦C for 3 d, according to Gram et al. [19]. After solidification of the inoculated media, a fold of 10 mL of the corresponding melting medium was added to cover the surface. Microbial characterization was confirmed by using microscopy observation, Gram staining, oxidase and catalase results. Growth data of the different microorganisms enumerated on fish fillets were further modeled as a function of time, using the primary model of Baranyi and Roberts [20] to estimate the kinetic parameters. For data fitting, the program DMFit was used (www. combace.cc).

2.3. Gas Analysis

The gas composition (CO2 and O2) of the fish fillets packages was analyzed using a headspace gas analyzer Dansensor Chekmate 9900 (Tendringpacific, Denmark).

2.4. Sensory Analysis The sensory attributes of odor and skin color of fish fillets were assessed in duplicate. Each attribute was scored on a five-point hedonic scale. A score of one to two corresponded to high-quality fillets without any off-odors and a bright skin color (fresh). A score of three characterized fish fillets with slight-off odors or without intense skin color but of acceptable quality (intermediate freshness), whereas a score of four to five corresponded to fish fillets of unacceptable quality. Scores exceeding the value of three indicated the end of a fish fillet’s shelf life.

2.5. FTIR Spectroscopy The sea bass fillets (skin) were analyzed by FTIR spectroscopy in parallel to the microbiological analysis. A ZnSe 45◦ horizontal attenuated total reflectance (HATR) crystal (PIKE Technologies, Madison, WI, USA), an FTIR-6200 JASCO spectrometer (Jasco Corp., Tokyo, Japan) equipped with a triglycine sulphate detector, and a Ge/KBr beam splitter were used for to collect the FTIR spectral data of the sea bass fillets. The Spectra Manager Code of Federal Regulations (CFR) software version 2 (Jasco Corp.) was used, in accordance with Fengou et al. [14]. The FTIR spectra were further analyzed in the approximate wave number ranges of 3100 to 2700 and 1800 to 900 cm−1, as described previously for TVC [14]. Foods 2021, 10, 264 4 of 12

2.6. Image Acquisition Multispectral images of the sea bass ﬁllet samples (skin) were obtained using the Videometer lab device (Videometer A/S, Hørsholm, Danemark), originally developed by the Technical University of Denmark. The Videometer lab device operates in reﬂectance mode from the ultraviolet (UV, 405 nm) to shortwave near-infrared (NIR, 970 nm) regions. The operation used for image acquisition and model development has been described in detail in previous works [14,21,22].

2.7. Data Analysis Spectral Data Analysis and Correlation with Microbiological Data Partial least squares regression (PLS-R) models were developed for the quantitative analysis of the microbial population of TVC using the spectral information of FTIR and MSI as input variables and the counts of TVC as output variables. The developed PLS-R models were evaluated to provide predictions in a temperature-independent manner, as previously described [14,21]. Model calibration was based on the data derived from fish fillet samples stored at 0 and 4 ◦C(n = 144), whereas model validation was undertaken using spectral data obtained during fish fillet storage at 8 and 12 ◦C(n = 74). Multivariate data analysis was carried out using The Unscrambler ver. 9.7 software (CAMO Software AS, Oslo, Norway). The performance of the developed PLS-R models was evaluated by the calculation of the following parameters: slope, offset, root mean squared error of calibration (RMSEc), cross-validation (RMSEcv), and prediction (RMSEp), as well as the coefficient of determination (R2) for calibration, cross-validation, and prediction. The optimum number of latent variables (LVs) was assigned at the minimum prediction residual error sum of squares after cross-validation using the leave-one-out cross-validation (LOOCV) method. Prior to analysis, spectral data were pre-processed by baseline correction.

3. Results and Discussion 3.1. Microbiological Spoilage of Sea Bass Fillets The initial TVC of the sea bass fillets was 4.94 log CFU/g (Figure1), which was about 1–2 logs higher compared to whole unprocessed or gutted fish [5,6,23]. This difference could be attributed to the use of equipment, utensils and the handling of fish fillets on working surfaces that could result in contamination, even in an industrial environment with good hygiene and sanitary practices. Fish spoilage can usually occur when the TVC or Specific Spoilage Organisms (SSOs) reach around 7.0–8.0 log CFU/g. At these population levels, the activity of the microorganisms results in the production of metabolic compounds that contribute to the organoleptic rejection of the product [4,7]. Based on the sensory assessment, in aerobic storage conditions, the product was characterized as unacceptable after 191, 125, 78, and 48 h when at 0, 4, 8, and 12 ◦C, respectively, while it was characterized as unacceptable in MAP storage conditions after 305, 185, 88, and 71 h, at the same temperatures. Indeed, at the time of sensory rejection, TVC ranged between 9.0–9.8 log CFU/g for fish fillets stored aerobically and 7.0–8.0 log CFU/g for fish fillets under MAP storage conditions. Pseudomonas spp. and H2S-producing bacteria, primarily Shewanella spp. [24], were found at higher population levels compared to the other studied microorganisms (Enter- obacteriaceae, B. thermosphacta, yeasts, and LAB). Pseudomonas spp. grew faster in aerobic conditions, while its growth was inhibited under MAP [25,26]. In our study, the concentration of CO2 (31%) in combination with the concentration of O2 (23%) resulted in the outgrowth of pseudomonads and their dominance at all storage temperatures both in air and MAP conditions. This is also supported by the lower growth rates estimated in MAP compared to those of aerobic storage (Table1). Indeed, Shewanella spp. presented the highest growth rate, despite the lower initial population (3.02 and 3.68 log CFU/g), in comparison to pseudomonads under all storage conditions (Table1). Enterobacteri- aceae reached the same final population, not exceeding 6.5–7.5 and 8.0–8.7 log CFU/g at 0, 4 and 8, 12 ◦C, respectively, regardless of packaging (Figures1 and2). The MAP Foods 2021, 10, 264 5 of 12

Foods 2021, 10, x FOR PEER REVIEW 5 of 13

conditions at the higher temperatures (8 and 12 ◦C) favored the growth of B. thermosphacta,

LAB,time and of yeasts, sensory all rejection, of which TVC were ranged enumerated between 9.0– at9.8 least log CFU/g 1.8–2.0 for logfish CFU/gfillets stored higher than the correspondingaerobically and populations 7.0–8.0 log CFU/g in aerobic for fish conditions. fillets under MAP storage conditions.

Figure 1. GrowthFigure 1. Growth of microorganisms of microorganisms on on sea sea bass bass fish fish fillets fillets at different at different temperatures temperatures (0, 4, 8, and 12 (0, °C )4, in aerobic 8, and pack 12 ag-◦C) in aerobic ing conditions. Total viable counts: (◊), Pseudomonas spp.: (▪), H2S-producing bacteria: (∆), Enterobacteriaceae: (x), B. ther- packaging conditions.mosphacta: (ж), Total yeasts: viable (+), and counts: lactic acid (♦ ),bacteria:Pseudomonas (●). spp.: (), H2S-producing bacteria: (∆), Enterobacteriaceae: (x), B. thermosphacta:(ж), yeasts: (+), and lactic acid bacteria: (•). Pseudomonas spp. and H2S-producing bacteria, primarily Shewanella spp. [24], were found at higher population levels compared to the other studied microorganisms (Enter- Table 1. Estimated kinetic parameters of different microbial groups in sea bass fish fillets stored aerobically and under obacteriaceae, B. thermosphacta, yeasts, and LAB). Pseudomonas spp. grew faster in aerobic modified atmosphere packaging (MAP)conditions at different, while its temperaturegrowth was inhibited conditions. under MAP [25,26]. In our study, the concentration of CO2 (31%) in combination with the concentration of O2 (23%) resulted in the Temperature/Packaging Air MAP outgrowth of pseudomonads and their dominance at all storage temperatures both in air ◦C y0and Rate MAP conditions.±SE This yEnd is also supported Se Fit R by2 the lowery0 growth Rate rate±sSE estimated yEnd in MAP SE Fit R2 Total Viable Count 4.94compared 0.60 to those 0.05 of aerobic 9.76 storage 0.31 (Table 0.971). Indeed, 5.13 Shewanella 0.13 spp. 0.01 presented 8.60 the * high- 0.16 0.97 Pseudomonas spp. 4.68est growth 0.63 rate, 0.05 despite the 9.92 lower initial 0.32 population 0.97 5.09(3.02 and 0.14 3.68 log 0.01 CFU/g), 8.50in compar- * 0.16 0.97 Shewanella spp. 3.02ison 0.63 to pseudomonads 0.06 under 8.70 all storage 0.40 conditions 0.96 3.68(Table 1). 0.19 Enterobacteriaceae 0.01 8.03 reached 0.10 0.99 0 Enterobacteriaceae 3.40 0.25 0.02 6.37 * 0.23 0.95 3.94 0.25 0.03 7.70 * 0.21 0.96 B. thermophacta 2.50the same 0.45 final 0.01population 7.13, not exceeding 0.06 6.5 1.00–7.5 and 2.72 8.0–8.7 log 0.21 CFU/g 0.02 at 0, 4 and 7.60 8, * 12 °C, 0.36 0.96 Lactic acid bacteria 2.77respectively, 0.18 regardless 0.02 of 4.95 packaging * 0.15 (Figures 0.97 1 and3.45 2). The MAP 0.14 conditions 0.01 at 6.60 the * higher 0.21 0.97 Yeasts 3.34temperatures 0.70 (8 0.05 and 12 °C) 8.85 favored * 0.19 the growth 0.99 of B. 3.65 thermosphacta 0.10, LAB, 0.01 and yeast 6.90 *s, all of 0.13 0.97 Total Viable Count 4.94 0.91 0.04 10.02 * 0.27 0.98 5.13 0.28 0.02 9.02 0.18 0.98 Pseudomonas spp. 4.68which 0.96 were enumerated 0.04 10.05 at least * 1. 0.288–2.0 log 0.98 CFU/g 5.09higher than 0.22 the corresponding 0.01 8.60 * popu- 0.20 0.96 Shewanella spp. 3.02lations 1.02 in aerobic 0.04 conditions. 8.95 * 0.27 0.98 3.68 0.72 0.07 8.18 0.33 0.96 4 Enterobacteriaceae 3.40 0.79Lerfall et 0.09al. [27] evaluated 7.52 * the 0.22 growth 0.97changes 3.40of pseudomonads 0.32 0.03 at 4 °C in 6.73 different 0.26 0.94 B. thermophacta 2.50 0.71 0.04 6.70 * 0.16 0.99 2.72 0.66 0.06 7.34 0.24 0.99 modified atmospheres in saithe and reported a growth of 3 logs in low CO2/high O2 fish Lactic acid bacteria 2.77 0.36 0.03 4.97 * 0.13 0.97 3.45 0.33 0.03 7.07 * 0.34 0.97 Yeasts 3.34samples 1.08 (31.3% 0.03 CO2/66.0% 5.34 O2/2.7% 0.01 N2) only 1.00, whereas 3.65 in the 0.39 other gas 0.02 combinations 5.90 as- 0.06 1.00 Total Viable Count 4.94sayed 1.94, no growth 0.25 could be 9.69 detected. 0.22 Koutsoumanis 0.99 5.13et al. [25] 0.62 studied 0.07 the combined 8.96 effect 0.28 0.95 Pseudomonas spp. 4.68of storage 1.72 temperature 0.11 and 9.55 packaging 0.26 conditions 0.98 (air 5.09 and MAP) 0.53 on 0.07 red mullet 8.78 and re- 0.31 0.94 Shewanella spp. 3.02 2.17 0.22 7.71 0.35 0.96 3.68 1.01 0.11 8.28 0.29 0.96 8 Enterobacteriaceae 3.40ported 1.25 that the 0.06 main spoilage 8.00 * bacteria 0.27 in air 0.97 were Pseudomonas 3.40 0.90 spp. 0.10and Shewanella 8.60 spp., 0.35 0.96 B. thermophacta 2.50 1.24 0.10 6.45 * 0.28 0.97 2.72 0.50 0.07 8.40 * 0.61 0.94 Lactic acid bacteria 2.77 1.50 0.14 5.12 0.07 1.00 3.45 0.84 0.02 6.80 0.03 1.00 Yeasts 3.34 0.46 0.02 4.84 * 0.07 0.99 3.65 0.36 0.01 7.80 * 0.11 1.00 Total Viable Count 4.94 2.46 0.23 9.72 0.38 0.97 5.13 1.26 0.12 9.07 0.15 0.99 Pseudomonas spp. 4.68 2.37 0.20 9.71 0.34 0.97 5.09 1.26 0.17 8.87 0.19 0.98 Shewanella spp. 3.02 2.86 0.28 7.86 0.34 0.97 3.68 1.28 0.07 8.28 0.19 0.99 12 Enterobacteriaceae 3.40 2.03 0.15 8.56 0.33 0.97 3.40 1.47 0.08 8.68 0.22 0.99 B. thermophacta 2.50 3.41 0.21 6.82 * 0.13 0.99 2.72 0.69 0.06 8.50 * 0.42 0.97 Lactic acid bacteria 2.77 1.09 0.06 4.53 * 0.04 1.00 3.45 0.52 0.05 7.70 * 0.34 0.97 Yeasts 3.34 0.73 0.06 4.75 * 0.09 0.97 3.65 0.43 0.04 7.40 * 0.29 0.96 −1 y0 and yEnd: initial and final estimates (log CFU/g) of population; Rate: maximum specific growth rate (h ); SE: standard error; * indicates no estimated values, only observed values. Foods 2021, 10, x FOR PEER REVIEW 6 of 13

whereas in MAP conditions Shewanella spp. and other facultative anaerobes dominated. In sea bass fish and sea bass fillets, H2S producing bacteria were found at similarly higher Foods 2021, 10, 264 population levels compared to LAB and Br. thermosphacta under MAP storage conditions6 of 12 [5,6,28].

FigureFigure 2. 2.Growth Growth ofof microorganisms on on sea sea bass bass fish fish fillets fillets stored stored at atdifferent different temperatures temperatures (0, (0,4, 8, 4, and 8, and12 °C) 12 and◦C) modi- and fied atmosphere packaging conditions. Total viable counts: (◊), Pseudomonas spp.: (▪), H2S-producing bacteria: (∆), Entero- modified atmosphere packaging conditions. Total viable counts: (♦), Pseudomonas spp.: (), H2S-producing bacteria: (∆), Enterobacteriaceae:bacteriaceae: (x), B. (x), thermosphactaB. thermosphacta: (ж),:( yeasts:ж), yeasts: (+), and (+), lactic and lacticacid bacteri acid bacteriaa (●). (•).

Table 1. Estimated kinetic parametersLerfall etof al.different [27] evaluated microbial thegroups growth in sea changes bass fish of fillets pseudomonads stored aerobically at 4 ◦ Cand in under different modified atmosphere packaging (MAP) at different temperature conditions. modified atmospheres in saithe and reported a growth of 3 logs in low CO2/high O2 fish Temperature/Packaging samples (31.3% CO2/66.0%Air O2/2.7% N2) only, whereas in the otherMAP gas combinations o C assayed,y0 noRate growth ±SE could yEnd be detected. SE fit KoutsoumanisR2 y0 etRate al. [25 ] studied±SE yEnd the combinedSE fit effectR2 Total Viable Count of storage4.94 0.60 temperature 0.05 and9.76 packaging 0.31 conditions0.97 5.13 (air and0.13 MAP) 0.01 on red8.60 mullet * and0.16 reported 0.97 Pseudomonas spp. that4.68 the main0.63 spoilage0.05 bacteria9.92 in air0.32 were0.97Pseudomonas 5.09 0.14spp. and0.01Shewanella 8.50 * spp.,0.16 whereas 0.97 Shewanella spp. in MAP3.02 conditions0.63 0.06Shewanella 8.70 spp.0.40 and other0.96 facultative3.68 0.1 anaerobes9 0.01 dominated.8.03 0.10 In sea bass0.99 0 Enterobacteriaceae 3.40 0.25 0.02 6.37 * 0.23 0.95 3.94 0.25 0.03 7.70 * 0.21 0.96 B. thermophacta fish2.50 and sea0.45 bass 0.01 fillets, H7.132S producing0.06 bacteria1.00 were2.72 found0.21 at similarly0.02 7.60 higher * 0.36 population 0.96 Lactic acid bacteria levels2.77 compared0.18 to0.02 LAB 4.95 and * Br. thermosphacta0.15 0.97 under3.45 MAP0.14 storage0.01 conditions6.60 * [0.215,6, 28]. 0.97 Yeasts 3.34 0.70 0.05 8.85 * 0.19 0.99 3.65 0.10 0.01 6.90 * 0.13 0.97 Total Viable Count 3.2.4.94 MAP Gas0.91 Analysis0.04 10.02 * 0.27 0.98 5.13 0.28 0.02 9.02 0.18 0.98 Pseudomonas spp. 4.68The changes0.96 0.04 in gas10.05 concentration * 0.28 of MAP0.98 conditions5.09 0.22 during 0.01storage 8.60 * are presented0.20 0.96 in Shewanella spp. Figure3.02 3. The1.02 concentration0.04 8.95 of * gases0.27 remained 0.98 almost3.68 unchanged0.72 0.07 until 8.18 the time0.33 of sensory 0.96 4 Enterobacteriaceae 3.40 0.79 0.09 7.52 * 0.22 0.97 3.40 0.32 0.03 6.73 0.26 0.94 B. thermophacta rejection,2.50 regardless0.71 0.04 of storage6.70 * temperature.0.16 0.99However, 2.72 after0.66 the0.06 point of7.34 sensory 0.24 rejection, 0.99 Lactic acid bacteria CO2.772 presented 0.36 a gradual0.03 4.97 increase, * 0.13 whereas 0.97 O2 decreased3.45 0.33 until the0.03 end 7.07 of storage. * 0.34 This was0.97 ◦ Yeasts more3.34 pronounced 1.08 0.03 at the higher5.34 storage0.01 temperatures1.00 3.65 (80.39 and 120.02C). 5.90 0.06 1.00 Total Viable Count 4.94 1.94 0.25 9.69 0.22 0.99 5.13 0.62 0.07 8.96 0.28 0.95 Pseudomonas spp. 3.3.4.68 Sensory 1.7 Analysis2 0.11 9.55 0.26 0.98 5.09 0.53 0.07 8.78 0.31 0.94 Shewanella spp. 3.02Sensory 2.17 evaluation0.22 7.71 (odor and0.35 skin color)0.96 of3.68 the raw1.01 fish 0.11 fillets was8.28 also performed.0.29 0.96 Both attributes presented a decreasing trend throughout storage (Figure4). It must be underlined that sensory rejection coincided with the time when TVC ranged from 6.7 to

7.7 log CFU/g for fish fillets stored under MAP conditions and the time when TVC was greater than 9.0 log CFU/g for aerobically stored fish fillets. Kritikos et al. [6] investigated Foods 2021, 10, x FOR PEER REVIEW 7 of 13

8 Enterobacteriaceae 3.40 1.25 0.06 8.00 * 0.27 0.97 3.40 0.90 0.10 8.60 0.35 0.96 B. thermophacta 2.50 1.24 0.10 6.45 * 0.28 0.97 2.72 0.50 0.07 8.40 * 0.61 0.94

Lactic acid bacteria 2.77 1.50 0.14 5.12 0.07 1.00 3.45 0.84 0.02 6.80 0.03 1.00 Yeasts 3.34 0.46 0.02 4.84 * 0.07 0.99 3.65 0.36 0.01 7.80 * 0.11 1.00 Total Viable Count 4.94 2.46 0.23 9.72 0.38 0.97 5.13 1.26 0.12 9.07 0.15 0.99 Pseudomonas spp. 4.68 2.37 0.20 9.71 0.34 0.97 5.09 1.26 0.17 8.87 0.19 0.98 Shewanella spp. 3.02 2.86 0.28 7.86 0.34 0.97 3.68 1.28 0.07 8.28 0.19 0.99 12 Enterobacteriaceae 3.40 2.03 0.15 8.56 0.33 0.97 3.40 1.47 0.08 8.68 0.22 0.99 B. thermophacta 2.50 3.41 0.21 6.82 * 0.13 0.99 2.72 0.69 0.06 8.50 * 0.42 0.97 Lactic acid bacteria 2.77 1.09 0.06 4.53 * 0.04 1.00 3.45 0.52 0.05 7.70 * 0.34 0.97 Yeasts 3.34 0.73 0.06 4.75 * 0.09 0.97 3.65 0.43 0.04 7.40 * 0.29 0.96 Foods 2021,y100 and, 264 yEnd: initial and final estimates (log CFU/g) of population; Rate: maximum specific growth rate (h−1); 7 of 12 SE: standard error; * indicates no estimated values, only observed values.

3.2. MAP Gas Analysis Thethe changes microbial in gas spoilageconcentration of sea of MAP bass conditions fillets in during comparison storage withare presented their volatilome in and reported Figure 3. The concentration of gases remained almost unchanged until the time of sensory rejectionthat, regardless the time of storage of sensory temperature. rejection However, for cooked after the fish point coincided of sensory with rejection TVC, level, ranging from CO2 presented6.5 to 7.0 a gradual log CFU/g. increase, In whereas addition, O2 decreased Parlapani until et the al. end [5] of reported storage. This a good was association between ◦ more pronouncedTVC and sensoryat the higher rejection storage timetemperatures for whole (8 and gutted 12 °C). sea bass fish stored at 2 C.

Foods 2021, 10, x FOR PEER REVIEW 8 of 13

greater than 9.0 log CFU/g for aerobically stored fish fillets. Kritikos et al. [6] investigated the microbial spoilage of sea bass fillets in comparison with their volatilome and reported

thatFigure the 3. time Changes of sensory in headspace rejection concentration for cooked (%) fish of COcoincid2 (∙) anded with O2 (∙ TVC) during level fish, ranging fillet storage from Figure 3. Changes in headspace concentration (%) of CO2 (·) and O2 (·) during fish fillet storage under modified atmosphere 6.5under to 7.0 modified log CFU/g. atmosphere In addition, packaging Parlapani at different et al. temperatures. [5] reported Dashed a good lines association indicate thebetween time of packaging at different temperatures. Dashed lines indicate the time of sample rejection. TVCsample and rejection. sensory rejection time for whole gutted sea bass fish stored at 2 °C.

3.3. Sensory Analysis Sensory evaluation (odor and skin color) of the raw fish fillets was also performed. Both attributes presented a decreasing trend throughout storage (Figure 4). It must be underlined that sensory rejection coincided with the time when TVC ranged from 6.7 to 7.7 log CFU/g for fish fillets stored under MAP conditions and the time when TVC was

Figure 4. Sensory score for odor (a,b) and skin color (c,d) degradation of sea bass fish fillets stored Figure 4. Sensory scorein air for (a,c odor) and MAP (a,b) (b and,d) at skin 0 °C color(♦, ◊), 4 (c °C,d )(▲ degradation, ∆), 8 °C (■, □), of and sea 12 bass °C (● ﬁsh, ○). Dashed ﬁllets storedlines indi- in air (a,c) and MAP ◦ ◦ ◦ ◦ (b,d) at 0 C(, ♦), 4cateC( theN ,level∆), of 8 sampleC(, rejection.), and 12 C(•, ). Dashed lines indicate the level of sample rejection. # 3.4. Fish Spoilage Assessment Using FTIR and MSI Spectral Data Examples of typical FTIR spectra of fresh (TVC of 4.9 log CFU/g) and spoiled (TVC of 8.50 log CFU/g) sea bass fillets stored under air and MAP are shown in Figure 5. The sensory scores for spoiled sea bass fillets corresponded to those of fish samples stored at 8 °C for 144 h and 12 °C for 88 h for the air and MAP trials, respectively. The FTIR spectra in the ranges of 3100–2700 and 1800–900 cm−1 provide an overall fingerprint of the biochemical composition and freshness of the fish. The peak at 1640 cm-1 (O–H stretch) is mostly due to water and amide I. The peak at 1545 cm-1 (N–H bend, C–N stretch) is ascribed to amide II. The peaks at 1314 and 1238 cm-1(C–N stretch, N–H bend, C=O–N bend) are related to amide III. The peaks at 1162–1025 cm−1 could be due to amines (C–N stretch) [14]. Most of these peaks correspond to amides and amines that could be mainly attributed to the microbial proteolytic activity occurring during storage [14]. The growth of Pseudo- monas spp. in fish is related with the production of aldehydes, ketones, and ethyl esters (such as ethyl isovalerate or ethyl tiglate), which are considered fish spoilage components [29]. The development of ethyl esters and carbonyls in fish has been associated with peaks in the spectral regions of 2995–2860 and 1750–1705 cm-1, respectively [14]. Since Pseudomo- nas spp. was the dominant microbial group in the present study for fish fillets stored aerobically or under MAP, the aforementioned increased absorption peaks observed for the sea bass fillets (Figure 3) could be indicative of the production of spoilage metabolites, such as ethyl esters [14,28].

Foods 2021, 10, 264 8 of 12

3.4. Fish Spoilage Assessment Using FTIR and MSI Spectral Data Examples of typical FTIR spectra of fresh (TVC of 4.9 log CFU/g) and spoiled (TVC of 8.50 log CFU/g) sea bass fillets stored under air and MAP are shown in Figure5. The sensory scores for spoiled sea bass fillets corresponded to those of fish samples stored at 8 ◦C for 144 h and 12 ◦C for 88 h for the air and MAP trials, respectively. The FTIR spectra in the ranges of 3100–2700 and 1800–900 cm−1 provide an overall fingerprint of the biochemical composition and freshness of the fish. The peak at 1640 cm−1 (O–H stretch) is mostly due to water and amide I. The peak at 1545 cm−1 (N–H bend, C–N stretch) is ascribed to amide II. The peaks at 1314 and 1238 cm−1(C–N stretch, N–H bend, C=O–N bend) are related to amide III. The peaks at 1162–1025 cm−1 could be due to amines (C–N stretch) [14]. Most of these peaks correspond to amides and amines that could be mainly attributed to the microbial proteolytic activity occurring during storage [14]. The growth of Pseudomonas spp. in fish is related with the production of aldehydes, ketones, and ethyl esters (such as ethyl isovalerate or ethyl tiglate), which are considered fish spoilage components [29]. The development of ethyl esters and carbonyls in fish has been associated with peaks in the spectral regions of 2995–2860 and 1750–1705 cm−1, respectively [14]. Since Pseudomonas spp. was the dominant microbial group in the present study for fish fillets stored aerobically or under MAP, the aforementioned increased absorption peaks Foods 2021, 10, x FOR PEER REVIEW 9 of 13 observed for the sea bass fillets (Figure3) could be indicative of the production of spoilage metabolites, such as ethyl esters [14,28].

−1 FigureFigure 5. 5.RepresentativeRepresentative FTIR FTIR spectra, spectra, in in the the wavenumber wavenumber range range of of 3100 3100 to to 900 900 cm cm−, 1corresponding, corresponding to tosea sea bass bass fillets fillets stored in air (A) and MAP (B). Blue and green lines indicate fresh (total viable counts (TVC) around 4.9 log CFU/g) and stored in air (A) and MAP (B). Blue and green lines indicate fresh (total viable counts (TVC) around 4.9 log CFU/g) and spoiled (TVC around 8.5 log CFU/g) fish fillets, respectively. spoiled (TVC around 8.5 log CFU/g) fish fillets, respectively.

TheThe performance performance of ofthe the developed developed PLS PLS-R-R models models based based on the on FTIR the FTIRdata for data the for pre- the dictionprediction of TVC of TVCof sea of bass sea bassfillets fillets stored stored in air in airand and MAP MAP is shown is shown in inFigure Figure 6,6 while, while the the performanceperformance m metricsetrics are are summarized summarized in in Table Table 22. .Specifically Specifically,, for for the the sea sea bass bass fillets fillets stored stored underunder aerobic aerobic conditions, conditions, the the coefficients coefficients of determination (R(R22)) werewere 0.73, 0.73, 0.68, 0.68, and and 0.78 0.78 for formodel model calibration, calibration, cross-validation cross-validation and and prediction, prediction, respectively. respectively. Similarly, Similarly, for for the the sea sea bass bassfillets fillets stored stored under under MAP, MAP, the coefficientsthe coefficients of determination of determination (R2) were (R2) 0.99,were 0.72,0.99, and 0.72 0.99, and for 0.99model for mcalibration,odel calibration, cross-validation cross-validation and prediction, and prediction, respectively. respectively. The values The of values RMSE of of RMSEprediction of prediction were low were (0.84 low and (0.84 0.64 and for the 0.64 sea for bass the fillets sea bass stored fillets aerobically stored aerobically and under and MAP, underrespectively). MAP, respectively According). to According Chin [30], to R 2Chinvalues [30], higher R2 values than 0.67 higher indicate than a0.67 high indicate predictive a highaccuracy predictive of the accuracy PLS-R model; of the PLS ranges-R model; between ranges 0.33–0.67 between and 0. 0.19–0.3333–0.67 and indicate 0.19– moderate0.33 in- dicateand lowmoderate accuracy, and respectively, low accuracy, while respectively, values below while 0.19 values are considered below 0.19 unacceptable. are considered High unacceptable.R2 values and High low RMSER2 values values and of low calibration, RMSE values cross-validation, of calibration, and cross prediction-validation indicate, and the prediction indicate the good accuracy and precision of PLS-R models [31]. Therefore, the developed PLS-R models indicated satisfactory performance between spectral and TVC data that could provide reliable predictions of microbial populations regardless of storage temperature.

Table 2. Performance metrics of the partial least squares regression (PLS-R) models correlating TVC in sea bass fillet samples stored in air (A) and under MAP (B) on the basis of FTIR spectral data.

Storage Data Set Slope Offset R2 RMSE Calibration 0.73 2.01 0.73 0.90 Air Cross-validation * 0.68 2.45 0.59 1.14 Prediction 0.78 1.81 0.75 0.84 Calibration 0.96 0.28 0.96 0.67 MAP Cross-validation 0.72 5.41 0.99 1.05 Prediction 0.99 0.24 0.99 0.64 R2: coefficient of determination; RMSE: root mean squared error; * Leave-one-out cross-validation.

Foods 2021, 10, 264 9 of 12

good accuracy and precision of PLS-R models [31]. Therefore, the developed PLS-R models Foods 2021, 10, x FOR PEER REVIEW indicated satisfactory performance between spectral and TVC data that could provide10 of 13

reliable predictions of microbial populations regardless of storage temperature.

Figure 6. Comparison between observed and estimated counts of TVC by the PLS-R model based on FTIR spectral data for Figure 6. Comparison between observed and estimated counts of TVC by the PLS-R model based on FTIR spectral data the sea bass ﬁllets stored in air (A) and under MAP (B). Solid line indicates the line of equity (y = x); dashed lines indicate for the sea bass fillets stored in air (A) and under MAP (B). Solid line indicates the line of equity (y = x); dashed lines ±1 log unit area. indicate ±1 log unit area.

TableThere 2. Performance is a limited metrics number of the of partial publications least squares for regressionthe estimation (PLS-R) of models microbial correlating counts TVC of fishin sea directly bass fillet from samples FTIR stored spectral in air data. (A) andSpecifically, under MAP Saraiva (B) on theet al. basis [32] of investigate FTIR spectrald data.the po- tentialStorage of FTIR to predict Data the Set bacterial load Slope of salmon fillets Offset (Salmo salarR2) stored atRMSE 3, 8, and 30 °C in aerobic and MAP conditions. The developed PLS-R model for the prediction of Calibration 0.73 2.01 0.73 0.90 TVC showed R2 and RMSE prediction values of 0.81 and 0.78, respectively. The authors Air Cross-validation * 0.68 2.45 0.59 1.14 concluded that FTIR Predictioncan be used as a reliable, 0.78 accurate, 1.81 and fast 0.75method for real 0.84 time freshness evaluation of salmon fillets stored under refrigerated conditions. In a recent Calibration 0.96 0.28 0.96 0.67 study [14], FTIR spectroscopy was employed in tandem with multivariate data analysis MAP Cross-validation 0.72 5.41 0.99 1.05 [33] for the assessmentPrediction of the microbiological 0.99 quality of0.24 farmed whole 0.99 ungutted sea 0.64 bream (RSparus2: coefficient aurata of) determination; stored aerobically RMSE: root at mean0, 4, squaredand 8 °C. error; In * Leave-one-outthis study, the cross-validation. PLS-R model developed on spectral data obtained from the fish skin provided a satisfactory prediction of TVC,There as inferred is a limited by the numbervalues of of the publications coefficient of for determination the estimation (R of2, 0.72) microbial and the counts root meanof fish square directly error from (RMSE, FTIR spectral0.71). In data.a recent Specifically, work on Saraivagrass carp et al.(freshwater [32] investigated fish) [34], the a rapidpotential, online of FTIR multispectral to predict imaging the bacterial system load was of employed salmon fillets to predict (Salmo freshness salar) stored based at 3, on 8, chemicaland 30 ◦C indices in aerobic with and satisfactory MAP conditions. results. In The a different developed approach, PLS-R modelFTIR was for theused prediction to detect theof TVC adulteration showed Rbetween2 and RMSE two fish prediction species, values namely of 0.81salmon and and 0.78, salmon respectively. trout [35]. The The authors au- thorsconcluded concluded that FTIRthat FTIR can becomb usedined as with a reliable, different accurate, data analysis and fast techniques method ( forsuch real as PCA time andfreshness PLS-R) evaluation could effectively of salmon predict fillets the stored presence/absence under refrigerated of adulteration conditions. in fish In samples. a recent studyThe [14 performance], FTIR spectroscopy of the developed was employed PLS- inR tandemmodels withbased multivariate on MSI spectral data analysisdata for [the33] estimationfor the assessment of TVC of of sea the bass microbiological fillets stored under quality aerobic of farmed conditi wholeons and ungutted MAP is seapresented bream in(Sparus Figure aurata 7, while) stored the aerobicallyperformance at 0,metrics 4, and are 8 ◦C. shown In this in study, Table the 3. PLS-R model developed on spectral data obtained from the fish skin provided a satisfactory prediction of TVC, Tableas inferred 3. Performance by the values metrics of of the the coefficientPLS-R models of determinationcorrelating TVC (Rin 2sea, 0.72) bass andfillet thesamples root stored mean insquare air (A) error and under (RMSE, MAP 0.71). (B) on In the a recent basis of work multispectral on grass imaging carp (freshwater (MSI) spectral fish) data. [34 ], a rapid, online multispectral imaging system was employed to predict freshness based on chemical Storage Data Set Slope Offset R2 RMSE indices with satisfactory results. In a different approach, FTIR was used to detect the Calibration 0.58 3.10 0.58 1.12 adulteration between two fish species, namely salmon and salmon trout [35]. The authors concludedAir thatCross FTIR-validation combined * with0.48 different data3.89 analysis techniques0.40 (such as1.38 PCA and PLS-R) could effectivelyPrediction predict the0.43 presence/absence 4.23 of adulteration0.44 in fish samples.1.31 Calibration 0.65 2.79 0.65 0.77 MAP Cross-validation 0.53 3.38 0.42 0.96 Prediction 0.62 2.78 0.62 0.76 R2: coefficient of determination; RMSE: root mean squared error; * Leave-one-out cross-validation.

Foods 2021, 10, 264 10 of 12

The performance of the developed PLS-R models based on MSI spectral data for the Foods 2021, 10, x FOR PEER REVIEW estimation of TVC of sea bass ﬁllets stored under aerobic conditions and MAP is presented11 of 13 in Figure7, while the performance metrics are shown in Table3.

Figure 7. Comparison between observed and estimated counts of TVC by the PLS-R model based on MSI spectral data Figure 7. Comparison between observed and estimated counts of TVC by the PLS-R model based on MSI spectral data for for the sea bass ﬁllets stored in air (A) and under MAP (B).The solid line indicates the line of equity (y = x); dashed lines the sea bass fillets stored in air (A) and under MAP (B).The solid line indicates the line of equity (y = x); dashed lines indicate ±1 log unit area. indicate ±1 log unit area.

TableIn 3. general,Performance the metricsperformance of the PLS-Rs of the models PLS-R correlating models based TVC in on sea MSI bass spectral fillet samples data storedwere lessin air satisfactory (A) and under compared MAP (B) on to the those basis of of models multispectral based imaging on FTIR. (MSI) Specifically, spectral data. aerobically stored sea bass fillets presented lower values for the coefficient of determination (R2) Storage Data Set Slope Offset R2 RMSE (namely 0.58, 0.48, and 0.43) for model calibration, cross-validation, and prediction, respectively. Similarly,Calibration for the sea bass fillets 0.58 stored under 3.10 MAP, the 0.58 coefficients of 1.12 deter- Cross-validation * 0.48 3.89 0.40 1.38 Air 2 mination (R ) of the developedPrediction PLS-R model 0.43 were 0.65, 4.23 0.53, and 0.62 0.44 for model 1.31 calibration, cross-validation, and prediction, respectively. In addition, the values of RMSE of pre- Calibration 0.65 2.79 0.65 0.77 diction were 1.31 and 0.76 for the sea bass fillets stored under aerobic conditions and MAP Cross-validation 0.53 3.38 0.42 0.96 stored under MAP, respectively.Prediction In agreement 0.62 with the 2.78 present study, 0.62 a previously 0.76 published PLS-R model based on MSI spectral data acquired from the skin and flesh of sea R2: coefficient of determination; RMSE: root mean squared error; * Leave-one-out cross-validation. bream did not perform satisfactorily to assess the microbiological quality of fish regarding TVC Inprediction general, the[14]. performances However, Cheng of the and PLS-R Sun models[10] used based visible on MSIand near spectral-infrared data were(Vis- NIR)less satisfactory hyperspectral compared imaging to in those the ofrange models of 40 based0–1000 on nm FTIR. for Specifically, the prediction aerobically of TVC on stored the fleshsea bass of grass fillets carp presented during lower refrigerated values forstorage the coefficientat 4 °C and of reported determination that the (R 2developed) (namely PLS0.58,-R 0.48, model and based 0.43) foron the model full calibration,wavelength cross-validation, data provided a andsatisfactory prediction, performance respectively. in theSimilarly, prediction for the of seaTVC bass in filletsterms stored of RMSE under (0.57) MAP, and the R coefficients2 (0.90) values. of determination Therefore, the (R 2MSI) of spectralthe developed changes PLS-R of microbial model were growth 0.65, may 0.53, be and different 0.62 for for model various calibration, fish species. cross-validation, and prediction, respectively. In addition, the values of RMSE of prediction were 1.31 4.and Conclusions 0.76 for the sea bass fillets stored under aerobic conditions and stored under MAP, respectively.FTIR spectroscopy In agreement is a withpromising the present approach study, to apredict previously the microbiological published PLS-R quality model of seabased bass on fillets MSI spectralstored aerobic data acquiredally and from under the MAP skin when and flesh compared of sea breamwith conventional did not perform mi- crobiologicalsatisfactorily toanalysis. assess the In contrast, microbiological MSI presented quality ofa fishless regarding satisfactory TVC performance prediction [ and14]. couldHowever, not be Cheng used andeffectively Sun [10 for] used quality visible assessment and near-infrared of sea bass (Vis-NIR)fillets. Our hyperspectral findings will beimaging used to in develop the range a rap ofid, 400–1000 reliable nm, and for easy the-to prediction-use toolkit of to TVC evaluate on the the flesh microbiological of grass carp ◦ qualityduring refrigeratedof sea bass fish storage fillets at so 4 thatC and food reported microbiologists that the developed and seafood PLS-R specialists model basedcan rap- on idlythe full tackle wavelength the seafood data quality provided-related a satisfactory concerns of performance stakeholders. in Such the prediction innovations of and TVC im- in 2 provtermsements of RMSE in (0.57) food microbiologyand R (0.90) values. will minimize Therefore, food the lossesMSI spectral, contributing changes to of enhanced microbial foodgrowth security, may be welfare different, and for sustainability. various fish species.

Author Contributions: Conceptualization G.-J.E.N.; methodology, P.T.; formal analysis, M.G.; in- vestigation, M.G. and P.T.; resources, G.-J.E.N; data curation, M.G., P.T., and E.Z.P.; writing—original draft preparation, M.G., F.F.P., I.S.B., and P.T.; writing—review and editing, P.T., E.Z.P., and G.-J.E.N.; project administration, I.S.B.; funding acquisition, I.S.B and G.-J.E.N. All authors have read and agreed to the published version of the manuscript.

Foods 2021, 10, 264 11 of 12

4. Conclusions FTIR spectroscopy is a promising approach to predict the microbiological quality of sea bass fillets stored aerobically and under MAP when compared with conventional microbiological analysis. In contrast, MSI presented a less satisfactory performance and could not be used effectively for quality assessment of sea bass fillets. Our findings will be used to develop a rapid, reliable, and easy-to-use toolkit to evaluate the microbiological quality of sea bass fish fillets so that food microbiologists and seafood specialists can rapidly tackle the seafood quality-related concerns of stakeholders. Such innovations and improvements in food microbiology will minimize food losses, contributing to enhanced food security, welfare, and sustainability.

Author Contributions: Conceptualization G.-J.E.N.; methodology, P.T.; formal analysis, M.G.; inves- tigation, M.G. and P.T.; resources, G.-J.E.N.; data curation, M.G., P.T., and E.Z.P.; writing—original draft preparation, M.G., F.F.P., I.S.B., and P.T.; writing—review and editing, P.T., E.Z.P., and G.-J.E.N.; project administration, I.S.B.; funding acquisition, I.S.B. and G.-J.E.N. All authors have read and agreed to the published version of the manuscript. Funding: The project entitled “Rapid Fish FReshness Assessment Methodology” (ReFFFRAME) with MIS 5028331 have received funding from the European Union, European Maritime and Fisheries Fund, in the context of the Operational Programme “Maritime and Fisheries 2014–2020. Data Availability Statement: The data presented in this study are available on request from the corresponding author. The data are not publicly available due to privacy. Conﬂicts of Interest: The authors declare no conﬂict of interest.

References 1. Food and Agriculture Organization (FAO). FAO Framework on Rural Extreme Poverty: Towards Reaching Target 1.1 of the Sustainable Development Goals. Rome, Italy. 2019. Available online: http://www.fao.org/3/ca4811en/ca4811en.pdf (accessed on 6 December 2020). 2. Food and Agriculture Organization (FAO). The State of World Fisheries and Aquaculture 2020. Sustainability in Action. Rome, Italy. 2020. Available online: https://doi.org/10.4060/ca9229en (accessed on 6 December 2020). 3. Food and Agriculture Organization (FAO). The State of World Fisheries and Aquaculture 2018—Meeting the Sustainable Development Goals; FAO: Rome, Italy, 2018. 4. Boziaris, I.S.; Parlapani, F.F. Specific Spoilage Organisms (SSO) in Fish. In Microbiological Quality of Food: Foodborne Spoilers; Bevilacqua, A., Corbo, M.R., Sinigaglia, M., Sykes, R., Eds.; Woodhead Publishing: Cambridge, UK, 2016; pp. 60–98. 5. Parlapani, F.F.; Haroutounian, S.A.; Nychas, G.-J.E.; Boziaris, I.S. Microbiological spoilage and volatiles production of gutted European sea bass stored under air and commercial modified atmosphere package at 2 ◦C. Food Microbiol. 2015, 50, 44–53. [CrossRef][PubMed] 6. Kritikos, A.; Aska, I.; Economou, S.; Mallouchos, A.; Parlapani, F.F.; Haroutounian, S.A.; Boziaris, I.S. Volatilome of chill-stored European seabass and Atlantic salmon products under modified atmosphere packaging. Molecules 2020, 25, 1981. [CrossRef] [PubMed] 7. Gram, L.; Huss, H.H. Microbiological spoilage of fish and fish products. Int. J. Food Microbiol. 1996, 33, 121–137. [CrossRef] 8. Kamruzzaman, M.; ElMasry, G.; Sun, D.W.; Allen, P. Non-destructive prediction and visualization of chemical composition in lamb meat using NIR hyperspectral imaging and multivariate regression. Innov. Food Sci. Emerg. Technol. 2012, 16, 218–226. [CrossRef] 9. Teena, M.; Manickavasagan, A.; Mothershaw, A.; El Hadi, S.; Jayas, D.S. Potential of machine vision techniques for detecting fecal and microbial contamination of food products: A review. Food Bioprocess Technol. 2013, 6, 1621–1634. [CrossRef] 10. Cheng, J.H.; Sun, D.W. Rapid and non-invasive detection of fish microbial spoilage by visible and near infrared hyperspectral imaging and multivariate analysis. LWT Food Sci. Technol. 2015, 62, 1060–1068. [CrossRef] 11. Ropodi, A.I.; Panagou, E.Z.; Nychas, G.-J.E. Multispectral Imaging (MSI): A Promising method for the detection of minced beef adulteration with horsemeat. Food Control 2017, 73, 57–63. [CrossRef] 12. Papadopoulou, O.; Panagou, E.Z.; Tassou, C.C.; Nychas, G.-J.E. Contribution of Fourier Transform Infrared (FTIR) Spectroscopy data on the quantitative determination of minced pork meat spoilage. Food Res. Int. 2011, 44, 3264–3271. [CrossRef] 13. Argyri, A.A.; Jarvis, R.M.; Wedge, D.; Xub, Y.; Panagou, E.Z.; Goodacre, R.; Nychas, G.-J.E. A comparison of Raman and FT-IR spectroscopy for the prediction of meat spoilage. Food Contr. 2013, 29, 461–470. [CrossRef] 14. Fengou, L.C.; Lianou, A.; Tsakanikas, P.; Gkana, E.N.; Panagou, E.Z.; Nychas, G.-J.E. Evaluation of Fourier transform infrared spectroscopy and multispectral imaging as means of estimating the microbiological spoilage of farmed sea bream. Food Microbiol. 2019, 79, 27–34. [CrossRef] Foods 2021, 10, 264 12 of 12

15. Qin, J.; Chao, K.; Kima, M.S.; Lub, R.; Burks, T.F. Hyperspectral and multispectral imaging for evaluating food safety and quality. J. Food Eng. 2013, 118, 157–171. [CrossRef] 16. Dissing, B.S.; Papadopoulou, O.S.; Tassou, C.C.; Ersbøll, B.K.; Carstensen, J.M.; Panagou, E.Z.; Nychas, G.-J.E. Using multispectral imaging for spoilage detection of pork meat. Food Bioprocess Technol. 2013, 6, 2268–2279. [CrossRef] 17. Cheng, J.-H.; Sun, D.-W.; Pu, H.; Zhu, Z. Development of hyperspectral imaging coupled with chemometric analysis to monitor K value for evaluation of chemical spoilage in fish fillets. Food Chem. 2015, 185, 245–253. [CrossRef][PubMed] 18. Eurostat. Agriculture, Forestry and Fishery Statistics, 2019 Edition. Publications Office of the European Union, Luxemburg. 2019. Available online: https://appsso.eurostat.ec.europa.eu/nui/show.do?dataset=fish_aq2a&lang=en (accessed on 6 December 2020). 19. Gram, L.; Trolle, G.; Huss, H.H. Detection of specific spoilage bacteria from fish stored at low (0 ◦C) and high (20 ◦C) temperatures. Int. J. Food Microbiol. 1987, 4, 65–72. [CrossRef] 20. Baranyi, J.; Roberts, T.A. A dynamic approach to predicting bacterial growth in food. Int. J. Food Microbiol. 1994, 23, 277–294. [CrossRef] 21. Panagou, E.Z.; Papadopoulou, O.; Carstensen, J.M.; Nychas, G.-J.E. Potential of multispectral imaging technology for rapid and non-destructive determination of the microbiological quality of beef filets during aerobic storage. Int. J. Food Microbiol. 2014, 174, 1–11. [CrossRef] 22. Manthou, E.; Lago, S.L.; Dagres, E.; Lianou, A.; Tsakanikas, P.; Panagou, E.Z.; Anastasiadi, M.; Mohareb, F.; Nychas, G.-J.E. Application of spectroscopic and multispectral imaging technologies on the assessment of ready-to-eat pineapple quality: A performance evaluation study of machine learning models generated from two commercial data analytics tools. Comput. Electron. Agric. 2020, 175, 105529. [CrossRef] 23. Syropoulou, F.; Parlapani, F.F.; Bosmali, I.; Madesis, P.; Boziaris, I.S. HRM and 16S rRNA gene sequencing reveal the cultivable microbiota of the European seabass during ice storage. Int. J. Food Microbiol. 2020, 327, 108658. [CrossRef] 24. Tryfinopoulou, P.; Tsakalidou, E.; Vancanneyt, M.; Hoste, B.; Swings, J.; Nychas, G.-J.E. Diversity of Shewanella population in fish Sparus aurata harvested in the Aegean Sea. J. Appl. Microbiol. 2007, 103, 711–721. [CrossRef] 25. Koutsoumanis, K.P.; Taoukis, P.S.; Drosinos, E.; Nychas, G.-J.E. Applicability of an Arrhenius model for the combined effect of temperature and CO2 packaging on the spoilage microflora of fish. Appl. Environ. Microbiol. 2000, 66, 3528–3534. [CrossRef] 26. Tryfinopoulou, P.; Tsakalidou, E.; Nychas, G.-J.E. Characterization of Pseudomonas spp. associated with spoilage of gilthead seabream store under various conditions. Appl. Environ. Microbiol. 2002, 68, 65–72. [CrossRef][PubMed] 27. Lerfall, J.; Thomassen, G.M.B.; Jacobsen, A.N. Quality of fresh saithe (Pollachiusvirens) in modified atmosphere pachagesas affected by the gas composition. Food Packag. 2018, 18, 147–156. [CrossRef] 28. Parlapani, F.F.; Kormas, K.A.; Boziaris, I.S. Microbiological changes, shelf life and identification of initial and spoilage microbiota of sea bream fillets stored under various conditions using 16S rRNA gene analysis. J. Sci. Food Agric. 2015, 95, 2386–2394. [CrossRef][PubMed] 29. Parlapani, F.F.; Mallouchos, A.; Haroutounian, S.A.; Boziaris, I.S. Volatile organic compounds of microbial and non-microbial origin produced on model fish substrate un-inoculated and inoculated with gilt-head sea bream spoilage bacteria. LWT Food Sci. Technol. 2017, 78, 54–62. [CrossRef] 30. Chin, W.W. The Partial Least Squares Approach for Structural Equation Modeling; Marcoulides, G.A., Ed.; Modern Methods for Business Research; Lawrence Erlbaum Associates: Mahwah, NJ, USA, 1998; pp. 295–336. 31. Dijkstra, T.K.; Henseler, J. Consistent and asymptotically normal PLS estimators for linear structural equations. Comput. Stat. Data Anal. 2015, 81, 10–23. [CrossRef] 32. Saraiva, C.; Vasconcelos, H.; De Almeida, J.M.M.M. A chemometrics approach applied to Fourier transform infrared spectroscopy (FTIR) for monitoring the spoilage of fresh salmon (Salmo salar) stored under modified atmospheres. Int. J. Food Microbiol. 2017, 241, 331–339. [CrossRef] 33. Tsakanikas, P.; Karnavas, A.; Panagou, E.; Nychas, G.-J.E. A machine learning workflow for raw food spectroscopic classification in a future industry. Sci. Rep. 2020, 10, 11212. [CrossRef] 34. Cheng, J.-H.; Sun, D.-W.; Qu, J.-H.; Pu, H.-B.; Zhang, X.-C.; Song, Z.; Chen, X.; Zhang, H. Developing a multispectral imaging for simultaneous prediction of freshness indicators during chemical spoilage of grass carp fish fillet. J. Food Engin. 2016, 182, 9–17. [CrossRef] 35. Sousa, N.; Moreira, M.J.; Saraiva, C.; Almeida, J.M.M.M. Applying Fourier Transform Mid Infrared Spectroscopy to detect the adulteration of Salmo salar with Oncorhynchus mykiss. Foods 2018, 7, 55. [CrossRef]