Chemometric classification of Brazilian artisanal cheeses from different regions according to major and trace elements by ICP-OES

Enhanced food authenticity control using machine learning-assisted elemental analysis

With the increasing attention being paid to the authenticity of food, efficient and accurate techniques that can solve relevant problems are crucial for improving public trust in food. This review explains two main aspects of food authenticity, namely food traceability and food quality control. More explicitly, they are the traceability of food origin and organic food, detection of food adulteration and heavy metals. It also points out the limitations of the commonly used morphology and organic compound detection methods, and highlights the advantages of combining the elements in food as detection indicators using machine learning technology to solve the problem of food authenticity. Taking elements as detection objects has the significant advantages of stability, machine learning technology can combine large data samples, ensuring both the accuracy and efficiency. In addition, the most suitable algorithm can be found by comparing their accuracy.

Enterotoxigenic Staphylococcus aureus in Brazilian artisanal cheeses: Occurrence, counts, phenotypic and genotypic profiles

The present study aimed to assess the occurrence and counts of Staphylococcus aureus in Brazilian artisanal cheeses (BAC) produced in five regions of Brazil: Coalho and Manteiga (Northeast region); Colonial and Serrano (South); Caipira (Central-West); Marajó (North); and Minas Artisanal cheeses, from Araxá, Campos das Vertentes, Cerrado, Serro and Canastra microregions (Southeast). The resistance to chlorine-based sanitizers, ability to attach to stainless steel surfaces, and antibiogram profile of a large set of S. aureus strains (n = 585) were assessed. Further, a total of 42 isolates were evaluated for the presence of enterotoxigenic genes (sea, seb, sec, sed, see, seg, sei, sej, and ser) and submitted to typing using pulsed-field gel electrophoresis (PFGE). BAC presented high counts of S. aureus (3.4-6.4 log CFU/g), varying from 25 to 62.5%. From the S. aureus strains (n = 585) assessed, 16% could resist 200 ppm of sodium hypochlorite, whereas 87.6% produced strong ability to attach to stainless steel surfaces, corroborating with S. aureus ability to persist and spread in the environment. Furthermore, the relatively high frequency (80.5%) of multidrug-resistant S. aureus and the presence of enterotoxin genes in 92.6% of the strains is of utmost attention. It reveals the lurking threat of SFP that can survive when conditions are favorable. The presence of enterotoxigenic and antimicrobial-resistant strains of S. aureus in cheese constitutes a potential risk to public health. This result calls for better control of cheese contamination sources, and taking hygienic measures is necessary for food safety. More attention should be paid to animal welfare and hygiene practices in some dairy farms during manufacturing to enhance the microbiological quality of traditional cheese products.

Comparison of hyperspectral imaging and spectrometers for prediction of cheeses composition

Artisanal cheeses are part of the heritage and identity of different countries or regions. In this work, we investigated the spectral variability of a wide range of traditional Brazilian cheeses and compared the performance of different spectrometers to discriminate cheese types and predict compositional parameters. Spectra in the visible (vis) and near infrared (NIR) region were collected, using imaging (vis/NIR-HSI and NIR-HSI) and conventional (NIRS) spectrometers, and it was determined the chemical composition of seven types of cheeses produced in Brazil. Principal component analysis (PCA) showed that spectral variability in the vis/NIR spectrum is related to differences in color (yellowness index) and fat content, while in NIR there is a greater influence of productive steps and fat content. Partial least squares discriminant analysis (PLSDA) models based on spectral information showed greater accuracy than the model based on chemical composition to discriminate types of traditional Brazilian cheeses. Partial least squares (PLS) regression models based on vis/NIR-HSI, NIRS, NIR-HSI data and HSI spectroscopic data fusion (vis/NIR + NIR) demonstrated excellent performance to predict moisture content (RPD > 2.5), good ability to predict fat content (2.0 < RPD < 2.5) and can be used to discriminate between high and low protein values (∼1.5 < RPD < 2.0). The results obtained for imaging and conventional equipment are comparable and sufficiently accurate, so that both can be adapted to predict the chemical composition of the Brazilian traditional cheeses used in this study according to the needs of the industry.

Elemental Fingerprinting of Pecorino Romano and Pecorino Sardo PDO: Characterization, Authentication and Nutritional Value

Sardinia, located in Italy, is a significant producer of Protected Designation of Origin (PDO) sheep cheeses. In response to the growing demand for high-quality, safe, and traceable food products, the elemental fingerprints of Pecorino Romano PDO and Pecorino Sardo PDO were determined on 200 samples of cheese using validated, inductively coupled plasma methods. The aim of this study was to collect data for food authentication studies, evaluate nutritional and safety aspects, and verify the influence of cheesemaking technology and seasonality on elemental fingerprints. According to European regulations, one 100 g serving of both cheeses provides over 30% of the recommended dietary allowance for calcium, sodium, zinc, selenium, and phosphorus, and over 15% of the recommended dietary intake for copper and magnesium. Toxic elements, such as Cd, As, Hg, and Pb, were frequently not quantified or measured at concentrations of toxicological interest. Linear discriminant analysis was used to discriminate between the two types of pecorino cheese with an accuracy of over 95%. The cheese-making process affects the elemental fingerprint, which can be used for authentication purposes. Seasonal variations in several elements have been observed and discussed.

Heavy metals and health risk assessment of Brazilian artisanal cheeses

Dairy products stand out as a food matrix susceptible to the contamination of heavy metals via cattle feed and environmental or processing conditions. Specifically, in the case of cheese, the concentrations can be further increased depending on the production process. The artisanal cheese market has been standing out, especially in Brazil, due to cultural and gastronomic reasons. Eight types of Brazilian artisanal cheese were analyzed for metal concentrations (chromium, copper, cadmium, lead, arsenic, and mercury, n = 80, 10 samples of each cheese) using inductively coupled plasma mass spectrometry. Based on the results, a health risk assessment was carried out, based on the determination of estimated daily intake, target hazard quotient (THQ), and hazard index (HI). Variable concentrations were observed between the types of cheese, but in all cases the THQ and HI values were less than 1, indicating an absence of potential risk in the consumption of artisanal cheeses in relation to the intake of heavy metals.

Determination of Minerals in Soft and Hard Cheese Varieties by ICP-OES: A Comparison of Digestion Methods

For sample preparation prior to mineral analysis, microwave digestion (~2 h) is quicker and requires lower acid volume as compared to dry (6-8 h) and wet digestion (4-5 h). However, microwave digestion had not yet been compared systematically with dry and wet digestion for different cheese matrices. In this work, the three digestion methods were compared for measuring major (Ca, K, Mg, Na and P) and trace minerals (Cu, Fe, Mn and Zn) in cheese samples using inductively coupled plasma optical emission spectrometry (ICP-OES). The study involved nine different cheese samples with moisture content varying from 32 to 81% and a standard reference material (skim milk powder). For the standard reference material, the relative standard deviation was lowest for microwave digestion (0.2-3.7%) followed by dry (0.2-6.7%) and wet digestion (0.4-7.6%). Overall, for major minerals in cheese, strong correlation was observed between the microwave and the dry and wet digestion methods (R² = 0.971-0.999), and Bland-Altman plots showed best method agreement (lowest bias), indicating the comparability of all three digestion methods. A lower correlation coefficient, higher limits of agreement and higher bias of minor minerals indicate possibilities of measurement error.

Intelligent Discrete Deep Learning Based Classification Methodology in Chemometrics

In recent years, deep learning models have attracted much attention for classification purposes in chemometrics. The popularity of deep learning models in this field comes from their unique features like universal approximation capability with the desired accuracy. Deep learning classifiers use several intelligent processing layers to model mixed, complex, and nonlinear patterns in the underlying data sets, which is why the development of deep learning based models has never been stopped in the chemometrics literature. Despite the variety of deep learning classification models used in this field, they all use a continuous distance-based cost function in their learning processes. Although using a continuous cost function for learning deep classifiers is a common approach, it conflicts with the discrete nature of the classification problem. In fact, applying a continuous cost function for inherently discrete classification problems can reduce the performance of the classification. In this research, a novel discrete learning based classification approach is proposed and implemented on a deep feed-forward neural network as one of the most commonly used deep learning models to develop a different learning process for deep classification models. The basis of the proposed learning approach is maximizing a discrete matching function of the actual and fitted values instead of minimizing a continuous distance-based cost function. The proposed classification approach is evaluated on five benchmark data sets in the chemistry field. The empirical results indicated the superiority of the proposed discrete deep learning approach over its classic continuous form. The results of this study demonstrate the important effect of discrete learning processes on the performances of deep learning classification models. Therefore, the proposed methodology can be a powerful alternative to common classification approaches to analyze chemical data in the chemometrics field.