Next Article in Journal
Review of Hydrogen-Driven Power-to-X Technology and Application Status in China
Previous Article in Journal
Prediction of Short-Term Winter Photovoltaic Power Generation Output of Henan Province Using Genetic Algorithm–Backpropagation Neural Network
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Processes
Volume 12
Issue 7
10.3390/pr12071517
processes-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

A Preliminary Study on Determining Seasonal Variations in Halloumi Cheese Using Near-Infrared Spectroscopy and Chemometrics

by
Maria Tarapoulouzi
1,*,
José-Antonio Entrenas
2,
Dolores Pérez-Marín
2,
Ioannis Pashalidis
1 and
Charis R. Theocharis
1
1
Department of Chemistry, University of Cyprus, P.O. Box 20537, 1678 Nicosia, Cyprus
2
Department of Animal Production, University of Cordoba, Campus Rabanales, 14071 Córdoba, Spain
*
Author to whom correspondence should be addressed.
Processes 2024, 12(7), 1517; https://doi.org/10.3390/pr12071517
Submission received: 19 June 2024 / Revised: 15 July 2024 / Accepted: 18 July 2024 / Published: 19 July 2024
(This article belongs to the Special Issue Research Advances and Applications of Isotopic and Spectroscopic Analysis in Foods and Beverages)
Browse Figures
Versions Notes

Abstract

:
Cheese quality is affected by seasonal variations. These variations can influence several aspects of cheese, including its flavor, texture, nutritional content, and overall sensory qualities. The aim of this study was to assess the performance of near-infrared (NIR) instrumentation in terms of its ability to detect seasonal variations in Halloumi cheese samples when applying limited sample preparation compared to traditional protocols. Therefore, the use of NIR spectroscopy was examined for the determination of seasonal variations in Halloumi cheese samples from Cyprus in combination with chemometrics. Partial Least Squares Discriminant Analysis (PLS-DA) was applied. We found that NIR and chemometrics successfully discriminated the Halloumi cheese samples based on different climate conditions, the four seasons in the year when the milk collection took place. To externally validate the model, the dataset was divided into training and test sets. The innovation of this study is that Halloumi cheese was studied regarding seasonal variations by applying NIR for the first time. The outcome of this preliminary study is positive in terms of the capability of NIR to distinguish seasonal variations in Halloumi cheese, especially those due to differences in fatty acid molecules throughout the year. Future studies will include more samples to increase the current database.

1. Introduction

Nowadays, cheesemaking is one of the most significant businesses in terms of economic growth and one of the most well-studied industries in terms of quality and authenticity [1,2,3,4,5,6]. Cheese quality studies deal with the season (time of year) of collection of milk, which is a very important parameter influencing cheese quality, similar to geographical and species origin, milk treatment, type and amount of starter added, ripening time, etc. [7,8]. The time of year affects milk composition due to its relationship with lactation and the diet of animals, as well as various environmental factors [9].
Various chemical parameters like fat composition, fatty acid profile, and the total concentration of fat and proteins are mainly affected by seasonality in cheese. In addition, differences in proteolysis (free amino acid content) are affected by seasonality, as well as the occurrence of terpenes in cheese, which are transferred in milk during feeding with summer plants [10,11,12]. Lypolysis and microbiological quality as well as the variations in milk fat and proteins are influenced by season, as reported by Muñoz et al. [11]. Moreover, the concentration of microorganisms in milk is higher during summer; however, during winter, the microorganism count is significantly lower, posing a challenge for cheese fermentation [13]. Furthermore, Hickey et al. [9] stated that different seasonal milk collections gave considerable biochemical and compositional differences in Cheddar cheese, which change its sensory and volatile characteristics.
NIR spectroscopy is a technique that exploits the absorption of light in the near-infrared region (typically 780–2500 nm) by molecular overtone and combination vibrations. Using a diffusion reflection mode, NIR spectroscopy represents a rapid method for the determination of cheese composition in terms of fat, total nitrogen, water-soluble nitrogen, and non-protein nitrogen [8]. In general, this technique is characterized as non-destructive, cheap, and automatable for industrial purposes; additionally, no sample pre-treatment is needed. NIR instruments come in various forms, including online, handheld, and benchtop devices. Online NIR instruments are designed for real-time monitoring and control of processes in food production lines. Handheld NIR instruments offer portability and flexibility for on-site measurements in various environments. Benchtop NIR instruments are designed for laboratory-based analysis, offering higher precision and accuracy compared to handheld devices [14,15]. Based on these advantages, NIR spectroscopy is very popular in authentication and characterization studies of food products [16,17,18,19,20]; however, it has never been applied to study Halloumi cheese, except in the study of Tarapoulouzi et al. [21].
Chemometrics is a very important science as it deals with the interpretation of large and complex datasets coming from spectroscopy [7,17,19,20,21,22,23,24,25,26]. The chemometric method PLS-DA has been widely applied with great success in studies dealing with spectroscopical data. PLS-DA is particularly useful in food analysis as it can handle high-dimensional data (e.g., spectra with thousands of variables) and is robust in the presence of collinearities and noise. It is widely used for tasks such as authentication, classification of food products based on geographical origin or quality parameters, detection of adulteration, and monitoring of food processes [17,27,28,29,30]. More particularly, when combined, NIR spectroscopy and PLS-DA offer a comprehensive approach to food quality analysis. As already mentioned, NIR spectroscopy provides rich chemical information, while PLS-DA enables efficient classification and prediction based on this information. This combination is particularly valuable in industries where rapid and non-destructive analysis of large numbers of samples is required, such as food production and quality control [28,29].
Regarding Halloumi cheese, it is a well-known fact that consumer preferences tend to fluctuate seasonally, with the popularity of Halloumi cheese peaking during specific times of the year. Halloumi cheese which is produced between May and June is the most tasteful, and demanding consumers tend to purchase Halloumi in this specific period. In this context, the objective of this study was to assess the performance of NIR instrumentation in terms of its ability to detect seasonal variations in Halloumi cheese samples when applying limited sample preparation. It is noteworthy that this study represents the second examination of Halloumi cheese using NIR spectroscopy, and publications focusing solely on PDO-certified Cypriot cheese remain limited overall. As cheese quality is affected by the season of collection of milk, this method may be used as a screening test in the future as it is fast and non-destructive. Moreover, the outcome of this preliminary study will form the basis of more investigations on Halloumi cheese with NIR instrumentations, such as online, handheld and benchtop NIR instruments, to compare their performance in cheese analysis.

2. Materials and Methods

2.1. Selection and Pre-Treatment of Samples

The Halloumi cheese samples were purchased from two local, artisanal producers in Nicosia, Cyprus. Τhe animals were kept at nearby farms within 15 km; therefore, there were no plant differences due to locality. In particular, 34 samples with known dates of production were used in this study. In 2019, the seasons were defined as follows: winter (December–February), spring (March–May), summer (June–August), and autumn (September–November). During each of these seasons, the primary ingredient of the samples, milk, was obtained. Specifically, 5, 5, 10, and 14 samples were purchased in winter, spring, summer, and autumn, respectively. Because it is well-known that several factors influence the characteristics of cheese, i.e., milk species origin, animal breed, milk treatment, and ripening time (Karoui et al., 2006 [8]), these parameters were kept constant for all the samples in this study. This ensured that any observed differences in the cheese could be attributed specifically to seasonal effects. This means that all the samples in this study were subjected to consistent conditions regarding milk species origin (1:1 goat/sheep milk ratio), animal breed (Damascus goat and Cyprus fat-tailed sheep), milk treatment (during cooking time of Halloumi cheese), and ripening time (1–2 days after production), thereby isolating the influence of season on the characteristics of cheese. The feeding regime was not taken into account in this study. Freeze-drying of the samples was performed by applying the same conditions used elsewhere [24]. After freeze-drying, the samples were all grated for homogenization with a drum grater (Ghizzoni mod. GS electric, Retsch, Haan, Germany). The ground samples were kept in sealed plastic containers at room temperature.

2.2. NIR Measurements

The samples were measured using a MicroNIRTM 1700 Spectrometer (VIAVI Solutions, Inc., San Jose, CA, USA). This instrument is based on linear variable filter technology (LVF) and is designed for analysis in situ. This portable miniature spectrophotometer is extremely light (only 64 g, excluding the 150 g handle and the acquisition/data processing device). The optical window is about 227 mm2, and it has a 910 to 1676 nm spectral range, with a constant interval of 6.2 nm. The sensor integration time was 12.2 ms, and each spectrum was the mean of 12 scans. The instrument’s performance was checked every 10 min. A white reference measurement was obtained using a NIR reflectance standard (Spectralon™) with a 99% diffuse reflectance, while a dark reference was obtained from a fixed point in the room. The instrument was used in reflectance mode (log1/R). All the samples were measured in triplicate; thus, an average spectrum from the triplicates of each sample was calculated and used to produce the chemometric models.
Measurements were taken also by using a QUANT (Q-Interline A/S, Tølløse, Denmark) Spectrometer. It is a high-performance Fourier Transform NIR (FT-NIR) instrument prepared for at-line analysis, in the laboratory. The samples of this study were placed in petri dishes before taking measurements. It has an analysis window of approximately 4.15 cm2 and is capable of rotating the sample to obtain a more representative spectrum, although in this work, due to the small amounts of the samples, the analysis was performed in static mode without rotation. The range of this device is 671.82–2702.70 nm and it uses the HORIZON MBTM spectrum acquisition software (version 3.6.1.0). The selected spectral resolution was 30 cm−1, and the number of scans was set at 64. The reference was measured every 30 min with a standard material with 99% diffuse reflectance. Consequently, the samples were analyzed using both NIR instruments since they were both accessible, marking the initial evaluation of Halloumi cheese samples in a specific laboratory. Thus, it was crucial to assess their performance within this particular food matrix.

2.3. Pre-Processing of Spectral Data

No noise was detected; hence, no sections of the spectrum were excluded. Consequently, the complete spectra (940–1676 nm) recorded were utilized in the data analysis, following the approach outlined in Turgut et al. [19]. Different spectral pre-processing methods were applied, like multiplicative scatter correction (MSC), Savitzky–Golay (SG), standard normal variate (SNV), first derivative (D1), and second derivative (D2). The SG method proved to be the most appropriate for smoothing, and additionally, D1 was employed. These two data preprocessing methods were essential to enhance the performance of subsequent models by eliminating scattered noise, correcting wavelength-dependent variations, and smoothing the spectral data [19].

2.4. Chemometric Analysis for Model Establishment

The chemometric software SIMCA (version 17, Umetrics, Umeå, Sweden) was used to interpret the NIR measurements. PLS-DA was the applied chemometric method. Even though the sample number is small, external validation took place by random division of the dataset into training and test sets, which contained 75% and 25% of the total number of samples, respectively, as in Tarapoulouzi et al. [24] and Antonova et al. [31]. Based on these percentages, 25 samples were kept for the training set and 9 for the test set. The level of significance was determined at p < 0.05. In addition, Hotelling’s ellipse in score plots provided a visual representation of the confidence region for the scores in multivariate analysis, aiding in the interpretation of model results, identification of outliers, and visualization of the data structure and relationships.
Misclassification tables were produced to check the correct classification of each model, as well as each group. The misclassification table provides a summary of the effectiveness of the chosen models in categorizing observations into their respective known classes. The default setting involves assigning each observation exclusively to the nearest class. Fisher’s probability of the misclassification table refers to the likelihood of the observed table arising randomly. This condition is met when the p-value is less than 0.05, indicating a 95% confidence level.
The selection of components or factors is an important number; thus, the root mean square error for the cross-validation (RMSECV) plot was used to evaluate the statistical significance of the model. In the 2D score scatter plots in PLS-DA, the terms t[1] and t[2] refer to the scores on the first and second latent variables or components, respectively. In PLS-DA, latent variables are new variables that are constructed as linear combinations of the original variables (spectroscopic data). These latent variables, t[1] and t[2], capture the maximum covariance between the predictors (X) and the response (Y), which is typically a class label in PLS-DA. Moreover, t[1] is the most important component in terms of explaining the variation in the data related to the classification, while t[2] is orthogonal to the first component, and still considers the response variable [17].
In addition, the R2 parameter, which shows the degree of adjustment to good fitting, as well as the Q2 parameter, which explains the robustness and predictive capacity of the model, were assessed; R2 and Q2 should be higher than 0.5 and their difference less than 0.3. A variable importance in the projection (VIP) plot was extracted for the general model to check which variables (wavelengths) had a VIP > 1 as this is an indicator of the most important ones influencing the model. The VIP value indicates the importance of each variable in predicting the response variables. Higher VIP values suggest greater importance, while lower values suggest less importance. The calculation of this parameter takes into consideration other factors like the number of response variables, the number of latent variables/components in the model, the weight of the variable for the latent variable and the score of the latent variable. Moreover, the Kruskal–Wallis test was performed to check whether the important wavenumbers of PLS-DA modeling were statistically significant (p < 0.05) across groups at the univariate analysis level, as suggested in Chen et al. [32]. In more detail, a null hypothesis states that there are no significant differences in scores between the groups. Conversely, an alternative hypothesis means that there are significant differences in scores between the groups. Rejection of the null hypothesis (due to p < 0.05) indicates that there are significant differences in scores between the groups and that the model is significant.

3. Results

3.1. Characterization of Halloumi Cheese via NIR Measurements

Evaluation of the spectra taken by both NIR instruments showed similar spectral characteristics as well as chemometric models; thus, only one of the two datasets, the dataset recorded with the MicroNIR™ 1700 Spectrometer, is presented in the next sections. The raw spectra and their first-order derivatives are shown in Figure 1a,b, respectively.
The NIR spectrum revealed several characteristics through the absorption bands. The C-H bond (a basic constituent of fatty acid molecules) resulted in absorbance at wavelengths close to 1150–1200 nm. In more detail, the C-H bond is a characteristic feature of organic molecules, particularly fatty acids, which contain long hydrocarbon chains. This absorption band at wavelengths close to 1150–1200 nm likely corresponds to the presence of fatty acids in the samples. The absorption at 1210 nm may be a consequence of the second overtone of the CH2 bond. In addition, the particular absorption indicates the presence of methylene groups (-CH2-) in the sample. Methylene groups are commonly found in fatty acids, which consist of a hydrocarbon chain with a carboxyl (-COOH) group at one end. The spectra curves have an increasing global trend with two strong local peaks at wavelengths around 1210 and 1450 nm, as observed also by Curto et al. [33] and Revilla et al. [34]. This subregion indicates the overall presence of organic compounds with varying molecular structures. The strong local peaks at wavelengths around 1210 nm and 1450 nm suggest specific functional groups or molecular vibrations. For instance, the peak at 1450 nm may correspond to the presence of C-H bonds in the methyl (CH3) groups, which are also common in fatty acid molecules.

3.2. Data Analysis

The components plot presents the eight components based on R2Y(cum) and Q2(cum), which were important and kept for the models’ construction, as seen in Figure 2. Therefore, eight components were used in data analysis.
Training (a) and test (b) sets, and their misclassification tables, (c) and (d), respectively, are presented in Figure 3. The misclassification tables show a 100% correct classification of each set, based on the 100% correct classification of each group.
Regarding both Figure 3 and Figure 4, it can be noted that while t[1] and t[2] provide the primary insights, higher-order components (t[3], t[4], etc.) have been examined to ensure that no critical information is overlooked. Overall, all 34 samples were combined, and based on the PLS-DA method, a general model was extracted, which is presented in Figure 4. R2X(cum) = 0.919, R2Y(cum) = 0.816 and Q2(cum) = 0.888 were obtained for the overall PLS-DA model, as well as a 100% correct classification of the overall model based on the 100% correct classification of each group. Furthermore, as can be observed, each season influences milk differently, and that is why four correctly classified groups occurred in the model.
A VIP plot was extracted, as shown in Figure 5, to check which variables (wavelengths) had a VIP > 1, as this is an indicator of the most influential ones in the classification. Based on the standard error computed for the model, the results from all the cross-validation rounds were 1100–1112, 1360, 1521, 1533 and 1639 nm, which are related to the fat and protein contents, as well as the occurrence of bound water [2,35]. These wavelengths had VIP values in the range 1.07–1.47.
In addition, based on the Kruskal–Wallis test, a p-value (0.02689) less than the significance level (0.05) was obtained, so the null hypothesis was rejected. This indicated that there were significant differences in the scores between the groups and that the model is important.

4. Discussion

The classification was expected to group the samples into two categories: winter + spring and summer + autumn. This approach creates one class for cold conditions and one for hot conditions. Cyprus has a typical Mediterranean climate, characterized by long, dry summers from mid-May to mid-October and mild winters from December to February. Short autumn and spring seasons are between these two main climatic periods. Nevertheless, based on the results of this study (i.e., four well-separated classes were obtained), it seems that each of the four seasons (winter: December–February, spring: March–May, summer: June–August and autumn: September–November) is different in Cyprus, and each one influences milk differently. This is a starting point for more research in this field which would initially involve adding more samples to the model.
The outcome of this study is characterized by the presence of fatty acid molecules throughout the year. The study of González-Martín et al. [7] is in agreement with our study and other authors who stated that seasonal changes in milk composition influence cheese yield and quality, the evolution of volatile components, free amino acid and free fatty acid contents, and sensory characteristics. More intense fruity and sweet overtones may be present in spring rather than in winter cheese. The above authors referred to the highest concentrations of free fatty acid contents reached in Manchego, La Serena, and Zamorano cheeses made in spring. Summer milk has lower levels of saturated fatty acids as well as higher unsaturated fatty acid and conjugated linoleic acid concentrations compared to the milk of other seasons. In addition, summer cheese is different due to deeper proteolysis taking place compared to cheese made in other seasons. Furthermore, summer cheese has important levels of terpenes in summer, but they do not appear in winter and spring samples. Lastly, the concentrations of volatile compounds derived from the catabolism of the branched-chain amino acids valine and leucine, such as 2-methypropan-1-ol, 3-methylbutan-1-ol, 2-methylpropanoic acid, and 3-methylbutanoic acid, are increased in cheese produced during summer [7].
Since this is a preliminary study conducted in one year, the climatic conditions parameter was not studied. In order to control this parameter, samples are usually collected from at least two consecutive years. In future studies, samples may be collected for another whole year to take into consideration this parameter. Feeding regime is also an important factor since it is influenced by the weather conditions [36,37,38]. However, to our knowledge, it is not a well-studied factor in the literature concerning Halloumi cheese. Changes in milk characteristics are influenced by seasonal variations in feeding regimes, and it would be very interesting to extend this study in order to study the seasonal variation in Halloumi cheese using NIR based on well-organized feeding regimes.
The aim of this study and the innovation regarding distinguishing among cheeses produced in different periods of the year are very important for several reasons. Due to the fact that the diet of dairy animals can vary depending on the time of year, the flavor, texture, and composition of the milk they produce may vary significantly. For instance, differences in the taste and quality of the milk are possible, which influences the cheese characteristics. Changes in temperature, humidity, and other environmental factors can influence the cheesemaking process, which can result in variations in flavor, aroma, and texture between cheeses made at different times of the year. For instance, cheeses produced in warmer months might be creamier or more flavorful due to a higher fat content in the milk, and due to sunlight, more Vitamin D is present in milk, which makes cheese more tasteful. Complementary fatty acid and Vitamin D analyses could be the objective of future research using the current Halloumi samples. In addition, in some regions, cheesemaking follows seasonal rhythms that have been established over generations based on traditional practices. These traditions may dictate when certain types of cheese are produced and are often tied to factors like the availability of milk or cultural festivities. Furthermore, market demand and consumer preferences may also vary seasonally, with some cheeses being more popular during certain times of the year. For instance, due to milk quality, richer-tasting Halloumi cheese might be preferred in May to June by demanding Halloumi lovers. In terms of quality, by monitoring and distinguishing cheeses produced in different seasons, producers can ensure quality control and consistency in their products. Understanding how seasonal variations affect cheese production allows producers to adapt their methods to maintain desired standards and meet consumers’ preferences.

5. Conclusions

This was the first time that seasonal variations in Halloumi cheese were studied by using NIR and chemometrics. Since the performance of the two NIR instruments was the same regarding Halloumi cheese measurements, the study proceeded successfully to multivariate data analysis of the spectra. PLS-DA provided 100% correct classification of the Halloumi cheese samples based on the season of milk production. Therefore, NIR spectroscopy coupled with PLS-DA resulted in an efficient methodology for discriminating Halloumi cheese based on seasonal variations. The method presented in this study eliminates the need for more detailed and complementary chemical analyses and could help screening samples that require further analysis in a low-cost manner.
Future studies should consider feeding regimes that are influenced by the weather conditions, as the occurrence of seasonal plants can affect milk characteristics, in combination with fatty acids measurements. Moreover, another future target is to develop the extracted model by increasing the number of samples. After that, the protocol could be applied to monitor Halloumi cheese batches quickly and easily throughout the production line to ensure the best possible product quality. By understanding and adapting to seasonal variations, cheese producers can optimize their processes, ensure consistent quality, and create unique marketing opportunities.

Author Contributions

Conceptualization, M.T., I.P. and C.R.T.; methodology, M.T. and J.-A.E.; software, M.T., J.-A.E. and D.P.-M.; validation, M.T., I.P. and C.R.T.; investigation, M.T.; resources, D.P.-M. and C.R.T.; data curation, M.T.; writing—original draft preparation, M.T.; writing—review and editing, D.P.-M., I.P. and C.R.T.; visualization, M.T.; supervision, J.-A.E., D.P.-M. and C.R.T.; project administration, C.R.T.; funding acquisition, D.P.-M. and C.R.T. All authors have read and agreed to the published version of the manuscript.

Funding

The researcher/author Dr. Maria Tarapoulouzi was financially supported by the “Short Term Scientific Mission” grant from “European Cooperation in Science & Technology (COST)” [Action Information: “CA19145—European Network for assuring food integrity using non-destructive spectral sensors”] regarding travelling and subsistence in Córdoba, Spain where this research took place.

Data Availability Statement

Data are contained within the article.

Conflicts of Interest

The authors declare no conflicts of interest.

References

  1. Ayvaz, H.; Mortas, M.; Dogan, M.A.; Atan, M.; Yildiz Tiryaki, G.; Karagul Yuceer, Y. Near-and mid-infrared determination of some quality parameters of cheese manufactured from the mixture of different milk species. J. Food Sci. Technol. 2021, 58, 3981–3992. [Google Scholar] [CrossRef] [PubMed]
  2. Calvini, R.; Michelini, S.; Pizzamiglio, V.; Foca, G.; Ulrici, A. Evaluation of the effect of factors related to preparation and composition of grated Parmigiano Reggiano cheese using NIR hyperspectral imaging. Food Control 2022, 131, 108412. [Google Scholar] [CrossRef]
  3. de Andrade, B.M.; Margalho, L.P.; Batista, D.B.; Lucena, I.O.; Kamimura, B.A.; Balthazar, C.F.; Brexo, R.P.; Pia, A.K.R.; Costa, R.A.S.; Cruz, A.G.; et al. Chemometric classification of Brazilian artisanal cheeses from different regions according to major and trace elements by ICP-OES. J. Food Compos. Anal. 2022, 109, 104519. [Google Scholar] [CrossRef]
  4. Malegori, C.; Oliveri, P.; Mustorgi, E.; Boggiani, M.A.; Pastorini, G.; Casale, M. An in-depth study of cheese ripening by means of NIR hyperspectral imaging: Spatial mapping of dehydration, proteolysis and lipolysis. Food Chem. 2021, 343, 128547. [Google Scholar] [CrossRef] [PubMed]
  5. Ozturk, M.; Dogan, M.A.; Menevseoglu, A.; Ayvaz, H. Infrared spectroscopy combined with chemometrics as a convenient method to detect adulterations in cooking/stretching process in commercial cheese. Int. Dairy J. 2022, 128, 105312. [Google Scholar] [CrossRef]
  6. Tarapoulouzi, M.; Theocharis, C.R. Discrimination of Cheddar, Kefalotyri, and Halloumi cheese samples by the chemometric analysis of Fourier transform infrared spectroscopy and proton nuclear magnetic resonance spectra. J. Food Process Eng. 2022, 45, e13933. [Google Scholar] [CrossRef]
  7. González-Martín, I.; Hernández-Hierro, J.M.; Salvador-Esteban, J.; González-Pérez, C.; Revilla, I.; Vivar-Quintana, A. Discrimination of seasonality in cheeses by near-infrared technology. J. Sci. Food Agric. 2011, 91, 1064–1069. [Google Scholar] [CrossRef] [PubMed]
  8. Karoui, R.; Mouazen, A.M.; Dufour, É.; Pillonel, L.; Schaller, E.; De Baerdemaeker, J.; Bosset, J. Chemical characterisation of European Emmental cheeses by near infrared spectroscopy using chemometric tools. Int. Dairy J. 2006, 16, 1211–1217. [Google Scholar] [CrossRef]
  9. Hickey, D.K.; Kilcawley, K.N.; Beresford, T.P.; Sheehan, E.M.; Wilkinson, M.G. The influence of a seasonal milk supply on the biochemical and sensory properties of Cheddar cheese. Int. Dairy J. 2006, 16, 679–690. [Google Scholar] [CrossRef]
  10. Collomb, M.; Bisig, W.; Bütikofer, U.; Sieber, R.; Bregy, M.; Etter, L. Seasonal variation in the fatty acid composition of milk supplied to dairies in the mountain regions of Switzerland. Dairy Sci. Technol. 2008, 88, 631–647. [Google Scholar] [CrossRef]
  11. Muñoz, N.; Ortigosa, M.; Torre, P.; Izco, J.M. Free amino acids and volatile compounds in an ewe’s milk cheese as affected by seasonal and cheese-making plant variations. Food Chem. 2003, 83, 329–338. [Google Scholar] [CrossRef]
  12. Vivar Quintana, A.M.; Lopez, B.; Revilla, I.; González Martín, I.; Hernández Hierro, J.M.; González Pérez, C. Seasonal evolution of hydrophilic and hydrophobic peptide contents in cheeses made from ewe′s goat′s or cow’s milk. Czech J. Food Sci. 2009, 27, S106–S108. [Google Scholar]
  13. Mendia, C.; Ibañez, F.C.; Torre, P.; Barcina, Y. Influence of the season on proteolysis and sensory characteristics of Idiazabal cheese. J. Dairy Sci. 2000, 83, 1899–1904. [Google Scholar] [CrossRef] [PubMed]
  14. Beć, K.B.; Grabska, J.; Huck, C.W. Principles and applications of miniaturized near-infrared (NIR) spectrometers. Chem.–A Eur. J. 2021, 27, 1514–1532. [Google Scholar] [CrossRef] [PubMed]
  15. Pu, Y.; Pérez-Marín, D.; O’Shea, N.; Garrido-Varo, A. Recent advances in portable and handheld NIR spectrometers and applications in milk, cheese and dairy powders. Foods 2021, 10, 2377. [Google Scholar] [CrossRef] [PubMed]
  16. Alinovi, M.; Mucchetti, G.; Tidona, F. Application of NIR spectroscopy and image analysis for the characterisation of grated Parmigiano-Reggiano cheese. Int. Dairy J. 2019, 92, 50–58. [Google Scholar] [CrossRef]
  17. Firmani, P.; La Piscopia, G.; Bucci, R.; Marini, F.; Biancolillo, A. Authentication of P.G.I. Gragnano pasta by near infrared (NIR) spectroscopy and chemometrics. Microchem. J. 2020, 152, 104339. [Google Scholar] [CrossRef]
  18. Parastar, H.; van Kollenburg, G.; Weesepoel, Y.; van den Doel, A.; Buydens, L.; Jansen, J. Integration of handheld NIR and machine learning to “Measure & Monitor” chicken meat authenticity. Food Control 2020, 112, 107149. [Google Scholar]
  19. Turgut, S.S.; Entrenas, J.A.; Taşkın, E.; Garrido-Varo, A.; Pérez-Marín, D. Estimation of the sensory properties of black tea samples using non-destructive near-infrared spectroscopy sensors. Food Control 2022, 142, 109260. [Google Scholar] [CrossRef]
  20. Visconti, L.G.; Rodriguez, M.S.; Di Anibal, C.V. Determination of grated hard cheeses adulteration by near infrared spectroscopy (NIR) and multivariate analysis. Int. Dairy J. 2020, 104, 104647. [Google Scholar] [CrossRef]
  21. Tarapoulouzi, M.; Logan, N.; Hardy, M.; Montgomery, H.; Haughey, S.A.; Elliott, C.T.; Theocharis, C.R. A Pre-Trial Study to Identify Species of Origin in Halloumi Cheese Utilising Chemometrics with Near-Infrared and Hyperspectral Imaging Technologies. Analytica 2024, 5, 17–27. [Google Scholar] [CrossRef]
  22. Carames, E.T.d.S.; Piacentini, K.C.; Alves, L.T.; Pallone, J.A.L.; Rocha, L.d.O. NIR spectroscopy and chemometric tools to identify high content of deoxynivalenol in barley. Food Addit. Contam. Part A 2020, 37, 1542–1552. [Google Scholar] [CrossRef] [PubMed]
  23. Segato, S.; Merlanti, R.; Bisutti, V.; Montanucci, L.; Serva, L.; Lucatello, L.; Mirisola, M.; Contiero, B.; Conficoni, D.; Balzan, S. Multivariate and machine learning models to assess the heat effects on honey physicochemical, colour and NIR data. Eur. Food Res. Technol. 2019, 245, 2269–2278. [Google Scholar] [CrossRef]
  24. Tarapoulouzi, M.; Kokkinofta, R.; Theocharis, C.R. Chemometric analysis combined with FTIR spectroscopy of milk and Halloumi cheese samples according to species’ origin. Food Sci. Nutr. 2020, 8, 3262–3273. [Google Scholar] [CrossRef] [PubMed]
  25. Tyska, D.; Mallmann, A.O.; Vidal, J.K.; Almeida, C.A.A.d.; Gressler, L.T.; Mallmann, C.A. Multivariate method for prediction of fumonisins B1 and B2 and zearalenone in Brazilian maize using Near Infrared Spectroscopy (NIR). PLoS ONE 2021, 16, e0244957. [Google Scholar] [CrossRef] [PubMed]
  26. Achata, E.M.; Oliveira, M.; Esquerre, C.A.; Tiwari, B.K.; O′Donnell, C.P. Visible and NIR hyperspectral imaging and chemometrics for prediction of microbial quality of beef Longissimus dorsi muscle under simulated normal and abuse storage conditions. LWT 2020, 128, 109463. [Google Scholar] [CrossRef]
  27. Bassbasi, M.; De Luca, M.; Ioele, G.; Oussama, A.; Ragno, G. Prediction of the geographical origin of butters by partial least square discriminant analysis (PLS-DA) applied to infrared spectroscopy (FTIR) data. J. Food Compos. Anal. 2014, 33, 210–215. [Google Scholar] [CrossRef]
  28. Bian, X.; Zhang, C.; Liu, P.; Wei, J.; Tan, X.; Lin, L.; Chang, N.; Guo, Y. Rapid identification of milk samples by high and low frequency unfolded partial least squares discriminant analysis combined with near-infrared spectroscopy. Chemom. Intellig. Lab. Syst. 2017, 170, 96–101. [Google Scholar] [CrossRef]
  29. de Oliveira, V.M.A.T.; Baqueta, M.R.; Março, P.H.; Valderrama, P. Authentication of organic sugars by NIR spectroscopy and partial least squares with discriminant analysis. Anal. Methods 2020, 12, 701–705. [Google Scholar] [CrossRef]
  30. Yazgan, N.N.; Genis, H.E.; Bulat, T.; Topcu, A.; Durna, S.; Yetisemiyen, A.; Boyaci, I.H. Discrimination of milk species using Raman spectroscopy coupled with partial least squares discriminant analysis in raw and pasteurized milk. J. Sci. Food Agric. 2020, 100, 4756–4765. [Google Scholar] [CrossRef] [PubMed]
  31. Antonova, O.; Calvo, J.; Seifert, A. Rapid Detection of Thermal Treatment of Honey by Chemometrics-Assisted FTIR Spectroscopy. Foods 2021, 10, 2892. [Google Scholar] [CrossRef]
  32. Chen, X.; Zeng, M.; Cai, D.; Zhou, H.; Wang, Y.; Liu, Z. Correlation between APOE4 gene and gut microbiota in Alzheimer’s disease. Benef. Microbes 2023, 14, 349–360. [Google Scholar] [CrossRef] [PubMed]
  33. Curto, B.; Moreno, V.; García-Esteban, J.A.; Blanco, F.J.; González, I.; Vivar, A.; Revilla, I. Accurate prediction of sensory attributes of cheese using near-infrared spectroscopy based on artificial neural network. Sensors 2020, 20, 3566. [Google Scholar] [CrossRef] [PubMed]
  34. Revilla, I.; Escuredo, O.; González-Martín, M.I.; Palacios, C. Fatty acids and fat-soluble vitamins in ewe’s milk predicted by near infrared reflectance spectroscopy. Determination of seasonality. Food Chem. 2017, 214, 468–477. [Google Scholar] [CrossRef] [PubMed]
  35. Darnay, L.; Králik, F.; Oros, G.; Koncz, Á.; Firtha, F. Monitoring the effect of transglutaminase in semi-hard cheese during ripening by hyperspectral imaging. J. Food Eng. 2017, 196, 123–129. [Google Scholar] [CrossRef]
  36. Nudda, A.; Battacone, G.; Boaventura Neto, O.; Cannas, A.; Francesconi, A.H.D.; Atzori, A.S.; Pulina, G. Feeding strategies to design the fatty acid profile of sheep milk and cheese. Rev. Bras. Zootec. 2014, 43, 445–456. [Google Scholar] [CrossRef]
  37. Serrapica, F.; Masucci, F.; Di Francia, A.; Napolitano, F.; Braghieri, A.; Esposito, G.; Romano, R. Seasonal variation of chemical composition, fatty acid profile, and sensory properties of a mountain pecorino cheese. Foods 2020, 9, 1091. [Google Scholar] [CrossRef]
  38. Osorio, M.T.; Koidis, A.; Papademas, P. Major and trace elements in milk and Halloumi cheese as markers for authentication of goat feeding regimes and geographical origin. Int. J. Dairy Technol. 2015, 68, 573–581. [Google Scholar] [CrossRef]
Processes 12 01517 g001
Figure 1. (a) Raw NIR data; and (b) NIR data with SG and D1 for Halloumi cheese samples recorded with MicroNIR™ 1700 Spectrometer.
Figure 1. (a) Raw NIR data; and (b) NIR data with SG and D1 for Halloumi cheese samples recorded with MicroNIR™ 1700 Spectrometer.
Processes 12 01517 g001
Processes 12 01517 g002
Figure 2. Components plot based on R2Y(cum) and Q2(cum), where Comp[number] stands for Component[number of Component].
Figure 2. Components plot based on R2Y(cum) and Q2(cum), where Comp[number] stands for Component[number of Component].
Processes 12 01517 g002
Processes 12 01517 g003
Figure 3. PLS-DA score scatter plots (the x-axis—labelled t[1]—shows the scores of the samples on the first latent variable, and the y-axis—labelled t[2]—shows the scores of the samples on the second latent variable) of the (a) training set—25 samples; and (b) test set—9 samples (yellow: winter, blue: spring, green: autumn, and red: summer milk collection). (c) Misclassification table of the training set; and (d) misclassification table of the test set.
Figure 3. PLS-DA score scatter plots (the x-axis—labelled t[1]—shows the scores of the samples on the first latent variable, and the y-axis—labelled t[2]—shows the scores of the samples on the second latent variable) of the (a) training set—25 samples; and (b) test set—9 samples (yellow: winter, blue: spring, green: autumn, and red: summer milk collection). (c) Misclassification table of the training set; and (d) misclassification table of the test set.
Processes 12 01517 g003
Processes 12 01517 g004
Figure 4. (a) PLS-DA score scatter plot (the x-axis—labelled t[1]—shows the scores of the samples on the first latent variable, and the y-axis—labelled t[2]—shows the scores of the samples on the second latent variable) of the overall model (R2X(cum) = 0.919, R2Y(cum) = 0.816 and Q2(cum) = 0.888), (yellow: winter, blue: spring, green: autumn, and red: summer milk collection). (b) The misclassification table of this model, with a Fisher probability of 3.1 × 10−17.
Figure 4. (a) PLS-DA score scatter plot (the x-axis—labelled t[1]—shows the scores of the samples on the first latent variable, and the y-axis—labelled t[2]—shows the scores of the samples on the second latent variable) of the overall model (R2X(cum) = 0.919, R2Y(cum) = 0.816 and Q2(cum) = 0.888), (yellow: winter, blue: spring, green: autumn, and red: summer milk collection). (b) The misclassification table of this model, with a Fisher probability of 3.1 × 10−17.
Processes 12 01517 g004
Processes 12 01517 g005
Figure 5. (a) VIP plot of the PLS-DA model; and (b) the most influencing spectral wavelengths. * Based on the 8 selected components. ** Jack-knife standard error computed for model results from all the cross-validation rounds.
Figure 5. (a) VIP plot of the PLS-DA model; and (b) the most influencing spectral wavelengths. * Based on the 8 selected components. ** Jack-knife standard error computed for model results from all the cross-validation rounds.
Processes 12 01517 g005
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Tarapoulouzi, M.; Entrenas, J.-A.; Pérez-Marín, D.; Pashalidis, I.; Theocharis, C.R. A Preliminary Study on Determining Seasonal Variations in Halloumi Cheese Using Near-Infrared Spectroscopy and Chemometrics. Processes 2024, 12, 1517. https://doi.org/10.3390/pr12071517

AMA Style

Tarapoulouzi M, Entrenas J-A, Pérez-Marín D, Pashalidis I, Theocharis CR. A Preliminary Study on Determining Seasonal Variations in Halloumi Cheese Using Near-Infrared Spectroscopy and Chemometrics. Processes. 2024; 12(7):1517. https://doi.org/10.3390/pr12071517

Chicago/Turabian Style

Tarapoulouzi, Maria, José-Antonio Entrenas, Dolores Pérez-Marín, Ioannis Pashalidis, and Charis R. Theocharis. 2024. "A Preliminary Study on Determining Seasonal Variations in Halloumi Cheese Using Near-Infrared Spectroscopy and Chemometrics" Processes 12, no. 7: 1517. https://doi.org/10.3390/pr12071517

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop