Cheese is a dairy product derived from milk and is produced in a variety of flavors and textures. It comprises proteins and fat from milk and forms by coagulation of the milk fat casein. The milk is often acidified and the addition of the enzyme rennet causes the coagulation. Most cheeses are made from whole cow milk, with worldwide production of 18.7 million metric tons in 2014. The United States accounts for approximately 29% of this production, with various European countries accounting for most of the rest. Other types of milk used for cheese are skimmed cow, goat, sheep, and buffalo milk. There are approximately five hundred different varieties of cheese recognized by the International Dairy Federation and strict standards are often applied to these different varieties based on many factors. While cheese manufacturing methods can vary greatly for different types of cheese, the constituents and parameters of interest during the manufacturing process are usually similar. Parameters such as fat, protein, moisture, dry matter, and acidity are critical in nearly all types of food manufacturing and these are very important in cheese as well. Cheesemaking is a highly selective skill and the cheesemaker is often reliant on sensory skills and chemical analysis to determine if the cheese being made is going to meet quality standards. After the manufacturing process, aging is also critical for many types of cheese as many physical parameters and flavor are finalized with age. In the case of fresh cheese, proper transport, temperature, and handling are critical as the shelf-life for such cheeses is often very short. Adulteration and authenticity are issues with cheese as well, especially in cheese manufactured in specific regions according to strict standards. Current methods for testing these parameters are expensive, laborious, and time-consuming, especially when implemented in a process setting. The timing aspect is especially critical for cheese because the time of steps as the process progresses can be quite short and there is often little time to make adjustments if needed. There is a need for fast, cost-effective, and real-time monitoring of parameters at all stages of the cheese manufacturing process. One such method that has been examined is NIR spectroscopy.
- Dry Matter (DM)
- Crude Protein (CP)
- Total Solids
- Sodium Chloride (Salt)
- Total Nitrogen
- Shelf Life
- Adulteration and Authenticity
- Rind %
- Months of Ripening
Summary of Published Papers, Articles, and Reference Materials
Measurement of chemical parameters in major constituents of cheese for quality control purposes has been studied using NIR spectroscopy. The results of most studies have been promising. One study examined monitoring the following important parameters of processed cheese manufacturing: dry matter (DM), fat, crude protein (CP), pH, and the rheological property penetration. Results were excellent for DM, fat, and CP and are considered good enough for real-time analysis. pH and penetration results are considered good enough for screening purposes. Fat, protein, and moisture are essential components in food manufacturing and cheese is no exception. These parameters were examined in ricotta cheese for NIR spectroscopic analysis, showing excellent results for fat and protein. Results for moisture were not as good but this is likely due to all samples having a high moisture content and a wider range in the sample set will improve results, as moisture is well-known to be a measurable constituent using NIR spectroscopy because of high absorption from water. Curd formation and cutting is an important step in cheese manufacturing. NIR spectroscopy was examined to compare results between regression models for total solids and protein in both non-homogenized and homogenized samples of cheese curds. The results were excellent and comparable for both models, indicating the potential to use NIR spectroscopy as a method for real-time measurements of these parameters in an industrial setting. Another comparative study was conducted to measure fat, protein, and sodium chloride (NaCl) in processed cheese that was both unwrapped and wrapped in polyethylene (PE) film. As was the case with the cheese curd study, the results showed little difference in performing these measurements between the wrapped and unwrapped samples and proved the feasibility of measuring these constituents in processed cheese wrapped in PE film. Aging of cheese is important and many of the main parameters that are measured during cheese manufacturing do not change discernably during the aging process. Total nitrogen and amino acids are two parameters that change with aging and NIR spectroscopy was examined to measure total nitrogen, tyrosine, and tryptophan. Results were good for total nitrogen but a lot of variability was shown during validation analysis for the amino acids. This likely occurred due to reference error and low concentration of the constituents, but there is potential to use NIR spectroscopy as a screening tool for measuring amino acids. Shelf-life of fresh cheese is critical and samples of Crescenza cheese were examined using both FT-NIR and FT-IR spectroscopy to determine the feasibility of classifying this cheese based on shelf-life. Classification analysis showed that both sets of spectral data can determine whether a sample is fresh, aged, or old. Adulteration is a major problem in the food industry and grated cheese samples were examined to determine authenticity based on standards for Parmigiano-Reggiano (PR) cheese in Italy. Results showed that NIR spectra could be used to classify both compliance and non-compliance based on standards and could also distinguish between PR cheese and competitor brands of cheese. Rind % and months of ripening were quantified with reasonable accuracy as well from the NIR spectra and calibration models.
Scientific References and Statistics
NIR Spectroscopy: A Useful Tool for Rapid Monitoring of Processed Cheeses Manufacture – Curda, Kukackova, Journal of Food Engineering 61 (2004) 557-560
Rapid monitoring of processed cheese manufacture is essential to obtain a high-quality product with minimal cost. NIR spectroscopy was examined as a method for assessing dry matter (DM), fat, crude protein (CP), pH, and the rheological property penetration in processed cheese samples. Fifty processed cheese samples from fourteen different Czech producers were procured for the study. Samples were left at room temperature for a minimum of twelve hours before being scanned. NIR spectra were collected using an FT-NIR spectrometer and a fiber optic probe from 900 nm to 2500 nm. Three separate spectra were collected from three different points on each sample and averaged into one spectrum. Standard reference methods were performed on the samples and the NIR spectra and reference values were used to create Partial Least Squares (PLS) regression models to correlate the spectra to the parameters of interest. Various pre-processing methods and selective wavelength ranges were used to optimize the calibration models.
|DM||R2= 0.998||RMSEP= 0.429%|
|Fat||R2= 0.995||RMSEP= 0.997%|
|CP||R2= 0.996||RMSEP= 0.303%|
|pH||R2= 0.945||RMSEP= 0.062|
|Penetration||R2= 0.925||RMSEP= 1.330mm|
Both the DM and fat models used the pre-processing method Multiplicative Scatter Correction (MSC) and selective wavelength ranges (1200 nm to 2200 nm for DM and 1000 nm to 2200 nm for fat) for model optimization. CP used no pre-processing but a selective wavelength range from 900 nm to 2400 nm. pH and penetration used the full wavelength range and no pre-processing. Results were excellent for DM, fat, and CP and were good enough to use these models for real-time analysis of manufactured cheese. Results were still good for pH and penetration but are better suited to estimate the values. Cross validation testing indicates that the lower precision for these two parameters is likely due to a small range of values for pH and error in the reference method for penetration.
Determination of Fat, Protein, and Moisture in Ricotta Cheese By Near Infrared Spectroscopy and Multivariate Calibration – Madalozzo, Sauer, Nagata, Journal of Food Science & Technology, March 2015 52 (3): 1649-1655
NIR spectroscopy was examined as a method for determining fat, protein, and moisture in ricotta cheese without any complex sample preparation. Nineteen samples of ricotta from different manufacturers in the southern region of Brazil were procured for the study. Sample varieties were fresh, pressed, conventional, and low-fat. Each sample was cut to provide a flat surface representative of the interior. NIR spectra were collected in diffuse reflectance mode from 1100 nm to 2500 nm using 1 nm resolution. Two different portions of each sample were scanned for thirty-eight total spectra. Thirty-three spectra were used for a calibration set and the remaining five were used for a validation set. Standard reference methods were performed on the samples to obtain fat, protein, and moisture values. Reference values and the NIR spectra were used to create Partial Least Squares (PLS) calibration models correlating the spectral data to the parameters of interest.
|Fat||R2= 0.968||RMSEP= 1.3%|
|Protein||R2= 0.968||RMSEP= 0.7%|
|Moisture||R2= 0.851||RMSEP= 2.7%|
There was a large amount of variability observed in the reference tests for the samples among the same varieties of samples, especially for fat and protein. While this variability is good for creating regression models, it does lend concern to commercial samples because ricotta is often used by people with dietary restrictions. Correlation was excellent for fat and protein and the validation set predictions proved the validity of the models. In the case of moisture, the variability in samples was lower but all samples surpassed the high moisture threshold of 55% for classification by Brazilian standards. A sample set with a wider range of values should improve the moisture model. Overall, this study proved the feasibility of using NIR spectra and regression models to determine fat, protein, and moisture in ricotta without using wet chemistry methods and requiring minimal sample preparation.
Using Near Infrared Spectroscopy for the Determination of Total Solids and Protein Content in Cheese Curd – Sultaneh, Rohm, International Journal of Dairy Technology, Volume 60, No.4, November 2007
Curd formation followed by curd cutting is considered to be one of the most essential steps in cheese manufacturing. Continuous information about the composition of cheese curd during processing (such as during drainage on conveyor belts between the cheese vat and pre-pressing unit) could help control the cheesemaking process, leading to improvements in quality. One previous study did measure total solids and protein in cheese curd but required excessive sample preparation that would not be applicable to on-line or in-line analysis. NIR spectroscopy was examined as a method for measuring total solids and protein in native unhomogenized cheese curd. A total of two hundred forty-two cheese curd samples were prepared in the laboratory for the study. Variation in total solids and protein was created by changing the intensity and duration of curd cutting as well as variation in scalding temperature and scalding time. After sample preparation and draining of the curd, NIR spectra were collected from 12000 cm-1 to 2000 cm-1 in diffuse reflectance mode. Thirty-two scans were averaged into one spectrum and collected at 8 cm-1 spectral resolution. Three individual spectra were collected in this manner on different portions of each sample. In order to compare the effect of homogenizing the samples, each sample was mixed together for one minute and the spectra collection process was repeated to create two different data sets. Standard reference methods were performed on the samples to obtain reference values for total solids and protein. Two sets of regression models were created for both total solids and protein using the NIR spectra and reference values.
|Total Solids||R2= 0.994||RMSEP= 0.502%|
|Protein||R2= 0.985||RMSEP= 0.548%|
|Total Solids||R2= 0.997||RMSEP= 0.388%|
|Protein||R2= 0.992||RMSEP= 0.381%|
While the results are slightly improved for the homogenized samples, the improvement is minimal and high correlation was obtained for both total solids and protein for both data sets. Predictions from a validation set proved the feasibility of the calibration models as a method for determining total solids and protein in cheese curd. Because of the minimal sample preparation, this study shows that NIR spectra and regression models could be used for industrial, real-time, on-line measurement of parameters of interest in cheese curd during the cheese manufacturing process.
Non-Destructive Determination of Components in Processed Cheese Slice Wrapped with a Polyethylene Film Using Near-Infrared Spectroscopy and Chemometrics – Pi, Shinzawa, Ozaki, Han, International Dairy Journal 19 (2009) 624-629
NIR spectroscopy was examined as a method for determining fat, protein, and sodium chloride in processed cheese slices that are wrapped in polyethylene (PE) film. Processed cheese is often covered in plastic wrapping to preserve quality and prolong shelf life. Fifty-one batches of processed cheese slices were obtained from commercial markets for the study. Each batch contained ten slices and all slices were wrapped in 25 µm thick PE film. NIR spectra of the wrapped samples were collected using an FT-NIR spectrometer in diffuse reflectance mode from 1000 nm to 2500 nm at 2 nm intervals. Thirty-two scans were averaged per spectrum. Three random points were chosen on each sample for this process and the three spectra for each sample were averaged into one spectrum as well. For comparative purposes, each sample was unwrapped from the PE film and this process was repeated. Reference values were obtained for fat, protein, and salt using traditional methods. A second derivative pre-processing was performed on both sets of spectra before chemometric modeling. Second derivative processing helps remove scattering due to non-homogenous distribution of particles, enhances peak separation, and can remove undesirable effects from baseline drift and slope. Forty-one spectra for each data set were used to build calibration models and the remaining ten spectra were used for a validation set.
With PE Film:
|Fat||R2= 0.984||RMSEP= 0.625%|
|Protein||R2= 0.999||RMSEP= 0.355%|
|Salt||R2= 0.991||RMSEP= 0.105%|
Without PE Film:
|Fat||R2= 0.990||RMSEP= 0.598%|
|Protein||R2= 0.998||RMSEP= 0.308%|
|Salt||R2= 0.998||RMSEP= 0.092%|
Visual comparison of the NIR spectra of the wrapped and unwrapped samples did show some contribution from the PE film in the wrapped samples, especially in the C-H vibrational absorbing areas of the NIR wavelength range. However, the effects of scattering were removed by the second derivative processing. Prediction values on the validation sets proved the feasibility of both models for measuring the parameters of interest. It must be noted that NaCl does not directly absorb in the NIR wavelength range. However, the high correlation indicates that the measurement is valid. Most likely, the salt has an effect on water molecules which is readily measurable using NIR spectra. Analysis of regression coefficients showing the relevant wavelength ranges for the calibration indicates this is the case. An indirect measurement of a component from NIR spectra is acceptable but must be carefully examined and validated. The results indicate that not only parameters in cheese can be measured when wrapped in PE film, but the potential exists for measuring other products wrapped in film as well by using NIR spectra with second derivative processing and reference values for the constituents of interest.
Application of Near Infrared Spectroscopy to Estimate Selected Free Amino Acids and Soluble Nitrogen During Cheese Ripening – Mlcek, Rop, Dohnal, Sustova, ACTA VET. BRNO 2011, 80: 293-297
The composition of young cheese is a key to aging but the traditional main components are relatively stable during the aging process and thus not a good indicator of the course of ripening as cheese ages. Cheese acquires its typical taste, smell, consistency, and appearance over the course of aging due to fermentation processes. One important barometer for this process is the type and quantity of free amino acids, which influence taste and provide information about the state and progression of aging. NIR spectroscopy was examined as a method for determining soluble nitrogen and two important amino acids (tyrosine and tryptophan) in Edam cheese. Samples were obtained from two different dairy factories, each providing four different types of Edam cheese with different values for dry matter and different starter cultures. Three portions of each sample were used for a total of two hundred eighty-eight separate portions for the analysis. The samples were aged for six months and at monthly intervals, both NIR spectra and reference tests for the parameters of interest were obtained. NIR spectra were collected from 12500 cm-1 to 2000 cm-1 in reflectance mode. Eighty scans were averaged per spectrum and spectral resolution was 4 cm-1. For reference values, samples were prepared for UV spectrophotometric analysis and values were obtained by the traditional method for each parameter of interest. After all data was collected over six months, Partial Least Squares (PLS) regression models were created using the NIR spectra and reference values.
|Total Nitrogen||R2= 0.911||RMSEP= 1.33%|
|Tryptophan||R2= 0.929||RMSEP= 0.00292 mg/100g|
|Tyrosine||R2= 0.959||RMSEP= 0.00705 mg/100g|
While the models did show good correlation, predictions obtained during cross-validation analysis showed very high variability, indicating that the models may not be suitable for real-time analysis. One possible reason for this is error in the UV reference method. It must be noted that the concentrations of tryptophan and tyrosine are below the threshold of measurement using NIR spectroscopy. Most likely, the models are correlating to another parameter which may or may not be affected by a change in the two amino acids. Before these models can be considered for real-time use, they must be further validated and carefully examined to determine the validity of the measurement. Despite this, the results offer promise for using NIR spectroscopy as a tool for estimating the degree of ripeness in Edam cheese as well as selecting a raw optimum material for making processed cheeses.
Application of FT-NIR and FT-IR Spectroscopy to Study the Shelf-Life of Crescenza Cheese – Catteneo, Giardina, Sinelli, et al., International Dairy Journal 15 (2005) 693-700
Crescenza cheese covers more than 40% of the Italian fresh cheese market and is produced only from whole cow milk. It is a fresh cheese and the components that are associated with its freshness are low acidity, limited proteolysis, and no bitter taste. Both FT-NIR and FT-IR spectroscopy were examined as methods for evaluating the shelf-life period for freshness in Crescenza cheese. Two different types of Crescenza made in the same industrial plant but using different technologies were procured for the study. The two types differed in their fat composition and starter components used during manufacturing. Samples were analyzed at different times over twenty days. FT-NIR spectra were collected from 12000 cm–1 to 4000 cm-1 averaging sixteen scans per spectrum at 16 cm-1 resolution. An optical probe was used to collect the spectra and measurements were collected in replicates varying from four to eight. FT-IR spectra were collected in duplicate for each sample from 4000 cm-1 to 600 cm-1 using an ATR crystal as a background. Sixteen scans were averaged per spectrum and 4 cm-1 spectral resolution was used. After spectra were collected, chemical and physiochemical analyses were performed to determine pH, Dry Matter (DM), Titratable Acidity (TA), and Hue (a color measurement) for each sample to determine thresholds for when sample quality has become poor. Based on this analysis, the samples were found to maintain freshness for six days, an significant decrease in freshness for the next few days, and an unacceptable freshness level for human consumption after eight to nine days. Various pre-treatments were performed on both the FT-NIR and FT-IR spectra and Principle Component Analysis (PCA) showed a clear grouping between these three thresholds. Samples could be clearly grouped from both sets of data into “Fresh”, “Aged”, and “Old” groups from the PCA scores plot. The results here indicate that both FT-NIR and FT-IR spectra can be used as a classification tool to determine the shelf-life of Crescenza cheese.
Screening of Grated Cheese Authenticity by NIR Spectroscopy – Cevoli, Fabbri, Gori, et al., Journal of Agricultural Engineering 2013; volume XLIV(s2):e53
Parmigiano-Reggiano (PR) cheese is one of the oldest traditional cheeses produced in Europe and has a Protected Designation of Origin in Italy. It is manufactured exclusively from whole PR wheels that correspond to the production standard. Grated PR cheese must be matured for a period of twelve months and characterized by a rind content of less than 18%. NIR spectroscopy was examined as a method for determining the authenticity of PR grated cheese. Four hundred samples were procured for the study with the following classifications: Compliance PR, Non-Compliance PR, PR with Rind Content > 18%, and Competitors (various commercial brands of grated cheeses obtained from local markets). NIR spectra were collected using an FT-NIR spectrometer in diffuse reflectance mode from 12500 cm-1 to 4000 cm-1. Thirty-two scans were averaged per spectrum and 8 cm-1 spectral resolution was used. Three replicate spectra were collected per sample. Various pre-processing treatments were performed on the NIR spectra. Principle Component Analysis (PCA) was first performed as an exploratory tool to define discrimination between Compliance and Non-Compliance PR samples or Competitors. Artificial Neural Network (ANN) models were created using software to test the feasibility of predicting each specific class from the spectral data. Reference values of Rind % and Months of Ripening were used with the NIR spectra to create Partial Least Squares (PLS) regression models for predicting these values from the spectra.
|Compliance PR Classification||100% for Training Set||100% for Validation Set|
|Competitors Classification||100% for Training Set||95.5% for Validation Set|
|Non-Compliance PR Classification||100% for Training Set||100% for Validation Set|
|Rind Content > 18% Classification||100% for Training Set||100% for Validation Set|
|Rind %||R2= 0.982||RMSEP= 1.473%|
|Months of Ripening||R2= 0.986||RMSEP= 0.805|
The results obtained in this study for both types of models were excellent and confirmed the ability of NIR spectroscopy to be used as a screening tool for determining grated cheese authenticity. The ANN model was able to 100% predict compliance or non-compliance in PR samples and detect competitor grated cheese at with an accuracy at above 95%. More competitor samples in the model will likely improve these results. In the case of PLS, the results were especially good considering there was some question in the reference values for rind %. A regression model can only predict within the error of the reference method and these results should improve as well with more accurate reference testing. Ripening can be predicted to within an accuracy of less than one month. NIR spectroscopy can be used as a fast, non-destructive screening tool for determining the authenticity of grated cheese.