Annotation of the results obtained in 2023
More than 650 samples of standards, model mixtures and meat products with and without packaging were produced, on the basis of which spectral data was determined, numbering almost two thousand unique records. All records were characterized by the parameters of the content of xenobiotics, water, protein, fat and carbohydrates, TMC, some were supplemented by the values of acid and peroxide numbers, the ratio of saturated and unsaturated fatty acids - determined by standard methods. The GC-MS method and the calibration curves of the chromatography-mass spectrometer for determining the fatty acid composition were refined to comply with the error of the obtained results GOST 31663-2012 - no more than 3% (rel.) of the obtained value, but no more than 1% (abs.). Tools for working with the spectral data base were developed and additions were made to its structure, ensuring the operation of several external software modules, as well as entering the characteristics of samples by type of meat product. Parameters of a machine algorithm for processing spectral data have been developed, including semantic segmentation of chilled meat, minced meat, emulsified meat products, smoked sausages, with and without packaging, based on microbial and oxidative alterations. The sensitivity of the method is sufficient for non-destructive express testing of TMC in minced meat, semi-finished meat and bone products, semi-finished minced meat products, boneless small-piece meat for compliance with the requirements of TR CU 034/2013. Training the neural network with all training data for 10 cycles showed accuracy above 80% for all models. The stability of the developed algorithm when changing the object of study has been confirmed and, thus, the possibility of its use when adding other types of meat products. A method has been developed for generating a single vector of THz parameters and hyperspectral data for training or analysis. The resulting parametric matrices were analyzed using a neural network with training. The average accuracy of semantic segmentation and detection of microbiological and oxidative changes for test samples of meat products was confirmed to be more than 90%. The possibility of detecting streptomycin and neomycin and the possibility of using the developed method and algorithm to determine quality and safety parameters for products not previously used in the training set, but belonging to the same type, for example, from a different manufacturer, were confirmed. The scientific and technical foundations for monitoring the quality and safety of meat products using hyperspectral analysis and terahertz radiation developed within the project are extremely promising from the point of view of practical implementation.

 

Publications

---

Annotation of the results obtained in 2021
The purpose of the work at the 1st stage of the project (2021) is the development of a research hyperspectral and terahertz spectrometer and the study of quality indicators, as well as the content of antibiotics and hormones in meat and meat products with its help. To develop scientific and technical methods for controlling the quality and safety of meat products using hyperspectral and terahertz radiation, confirmatory studies were carried out using known physicochemical methods (HPLC-MS on a Shimadzu LC-MS-8040 instrument, GC-MS on a Shimadzu GC-TQ-8040 instrument , HPLC on a Shimadzu LC-20 Prominence, protein content according to Kjeldahl (VELP UDK 159), fat content by Soxhlet method and others). To register the values of hyperspectral and terahertz radiation, a research spectrometer with adapted software, a radiation source, a receiver and a signal processing system was developed. For the standards of pure substances, the spectra obtained using the developed research spectrometer were selectively compared with those obtained in the IR range on a commercial Shimadzu IRA-1 IR spectrometer and a commercial instrument for terahertz time-domain spectroscopy TeraPulse 4000 (TeraView, Great Britain). The resulting hyperspectral images and terahertz spectra were compared with the data of the chemical composition, quality and safety parameters obtained by arbitration and characterized by methods. The study was consistently carried out for chemically pure standards of substances, model mixtures that simulate food products and directly for meat and meat products. Qualitative and quantitative analysis was carried out in the following types of meat products - chilled beef, minced beef, cooked and smoked sausages. For the first year of work, a study was carried out on the following quality parameters: color, compliance with the declared composition of macronutrients. Qualitative and quantitative determination was carried out for the following analytes: antibiotics of the chemical groups of aminoglycosides, beta-lactams, macrolides, quinolones, tetracyclines, cephalosporins and the hormones testosterone, zeranol, estradiol. With the help of various research objects, the influence of various matrices on the parameters of the signal of the desired object has been determined. As a result of the studies carried out, the requirements for the research hyperspectral and terahertz spectrometer are formulated, which are determined by their purpose, operating conditions and application to determine the quality and safety parameters of meat products. The choice of methods for studying experimental samples of standards, model mixtures and meat products containing antibiotics and hormones was substantiated, and the parameters of chromatography-mass spectrometry, hyperspectral and terahertz analysis were determined. Determination of the content of antibiotics and hormones was carried out by high performance liquid chromatography with mass spectrometric detection (HPLC-MS) on an LC-MS-8400 HPLC mass spectrometer (Shimadzu). The parameters of HPLC-MS and HPLC-MS-MS analysis were selected and optimized for streptomycin, neomycin, amoxicillin, erythromycin, enrofloxacin, doxycycline, cephalexin, testosterone, zeranol and estradiol. The analysis of the samples was carried out in two modes - scanning by the total ion current and in the mode of monitoring multiple reactions (MRM transition). Spectral analysis parameters include the exploration spectrometer imaging scheme, spectral ranges, and optical path options for transmission and reflection characteristics. The parameters of the radiation source, receiver and signal processing system for the hyperspectral and terahertz spectrometer are determined. To obtain hyperspectral images of the analyzed objects, the illumination system must ensure uniform irradiation of the analysis zone with radiation with a wide spectral range, ensuring analysis in the wavelength range from 450 nm to 1700 nm. For this, the choice of a halogen lamp model is justified. The developed research spectrometer in terms of the circuit for the terahertz range is implemented in the form of a coherent THz detector based on the dynamic electro-optical Pockels effect. Adapted software for calculating the dispersion of the refractive index, dielectric constant, electrical conductivity, absorption coefficient, scattering coefficient and scattering anisotropy at a specific point in the sample. In the case of analysis of opaque objects, the reflection coefficient calculated at each point of hyperspectral images can be converted into the absorption coefficient using the obtained formula. Also, the adaptation of the Spectrina software was carried out for calculating the optical properties of the objects of study in the THz frequency range. Thus, the variances of the amplitudes and phases required for the calculation are obtained from the signal by means of the Fourier transform. The result will be the dispersion of the refractive index, refractive index, penetration depth. Hyperspectra, terahertz and mass spectra of experimental samples of standards, model mixtures and meat products were obtained. Mass spectra of standard samples were obtained by HPLC-MS analysis, both in the scanning mode for the total ion current and in the MRM transition mode. In the studied samples of steamed and chilled beef, minced beef, cooked and smoked sausages (total number of samples - 63), the selected antibiotics and hormones were not detected. To search for the widest possible range of antibiotics, hormones and other xenobiotics, we screened various options for extracts from meat and meat products, without and with various options for precolumn derivatization using gas chromatography-mass spectrometry (GC-MS). In the meat samples, peaks of compounds were found that were identified by molecular weights as di-n-butyl phthalate and diisooctyl phthalate. Terahertz absorption spectra of streptomycin and neomycin standards have the same linear shape over the entire spectral range. These spectra are close to each other, since these drugs correspond to one group of aminoglycoside antibiotics. For all samples, a change in the intensity of the spectra depending on the concentration was noted. It was found that terahertz radiation is sensitive to the main macronutrients - proteins, fat and water, which were used to create model objects for meat products. This has been confirmed by spectrometric analysis of meat, minced meat and emulsified meat products with different content of water, protein and fat. When analyzing the THz spectra of prepared samples of some emulsified meat products, repeated absorption peaks of lactose were found. The change in the optical properties of the samples was established depending on the quality indicator - the appearance of the surface of meat and meat products, using the example of pigments that give the meat color: myoglobin, oxymyoglobin and metmyoglobin. Calibration curves are plotted, detection limits and method errors are found. For the used HPLC-MS methods for the determination of antibiotics and hormones, calibration curves were obtained to determine their concentration. Calibration curves and parameters of the method for investigating the content of phthalates in meat and meat products were obtained. Due to the complexity of the objects of analysis and the possibilities of setting and calibrating the receiving systems of the research spectrometer, the measurement limits and errors are determined by the parameters of the receivers of the developed spectrometer sample.

 

Publications

---

Annotation of the results obtained in 2022
In 2022, the following results were obtained at Stage 2 of the project: 1. Hyperspectra, terahertz and mass spectra of samples of meat products made with confirmed adequate content of antibiotics or hormones were determined. The spectral characteristics of the produced samples of chilled meat, minced meat, emulsified meat products and smoked sausages were studied with the addition of antibiotics and hormones at concentrations (% wt.) 5, 1, 0.5, 0.1, 0.05, 0.01, 0.001. The initial concentrations of xenobiotics before and after being added to the product, as well as after the heat treatment of the product, were confirmed using high performance chromatography-mass spectrometry (HPLC-MS). Based on the hyperspectra obtained for all types of products, it was possible to determine some of their quality indicators. Based on the obtained terahertz spectra, the presence of xenobiotics - antibiotics and hormones - in raw materials and products was determined. 2. A database of spectra of samples of standards, model mixtures and meat products has been developed. All software components that make up the web application and the database are made in a freely distributed operating environment of the modern version. The selected DBMS (database management system) PostgreSQL in the modern version is an object-oriented hierarchical DBMS. Applications can take full advantage of the object database with any extensions at the application logic level, if necessary. Additional benefits available in PostgreSQL are features such as connecting through external interfaces of other storage environments, extensive transaction support, and the ability to perform rollbacks. The structure of the database, as well as the set of objects stored in it, is completely dynamic. It makes it possible to add, delete and change any objects and their sets. The architectural solution makes it possible to use software components that support standard access protocols and are implemented in any operating and hardware environment to create a user environment. The database contains information about the experiments carried out, taken terahertz, mass and hyperspectra of standards, model mixtures and meat products. 3. Hyperspectra, terahertz and mass spectra of experimental samples of standards, model mixtures and meat products were determined to determine the content of fatty acids. In the course of the project, measurements of THz, hyperspectra and mass spectra of standard samples of rendered fats, emulsions of fats with beef protein and water, meat products - raw beef, mutton and pork fat, chilled beef, pork and mutton, minced meat and sausages were made . Hyperspectra for samples of different fats (pork, lamb, beef) have a characteristic appearance and differ from each other, which can be used to determine the falsification of fats. The refractive index of fats in the THz range is about 1.35-1.4 in the entire studied frequency range from 0.2 to 1.0 THz. The fat absorption coefficient increases from 2 to 10 cm-1 in the entire studied frequency range from 0.2 to 0.8 THz. In samples of fats of different origin and, accordingly, fractional composition, a difference was noted in the obtained spectral characteristics. 4. Calibration curves, limits of detection and method errors for new objects of study were obtained. Due to the complexity of the object of analysis and the possibilities of setting the cameras used and calibrating the acousto-optical cells of the research THz and hyperspectral spectrometer, the measurement limits and errors are determined by the parameters of the receivers of the developed research spectrometer. The established limits of mass spectrometric detection of fatty acids were 0.1–0.3 ng/input with an error of 6–11%. 5. The methodology for measuring the quality and safety parameters of meat product samples before packaging and in consumer packaging has been determined. Based on the results of a series of test measurements, sequences of actions were obtained for the methodology for measuring the quality and safety parameters of meat product samples before packaging and in consumer packaging for chilled meat, minced meat, emulsified meat products and smoked sausages. The developed sequences for determining the parameters of quality and safety for different objects differ in the choice and processing of speckle-zonal images, the set of determined parameters, and the refinement of the method of arranging samples in various types of holders in the analysis zone of the object for study. Based on the data obtained, narrower spectral ranges can be refined and selected that characterize the corresponding parameters of the quality and safety of meat products. 6. Data on the compliance of the spectral characteristics of meat products with their main quality indicators. It has been established that the greatest recorded changes in chilled meat with different storage periods occur in the red channel (660 nm), which corresponds to the relationship between quality parameters and pigment concentrations with the presented spectral characteristics. It is shown that when obtaining the corresponding spectra and compiling a database of meat and meat products samples, it is possible to conduct an optical analysis for compliance with the declared composition of water, protein and fat. When analyzing the THz spectra of prepared samples of minced meat and smoked sausages, as well as minced meat for their manufacture, repeated lactose absorption peaks were found. In minced meat samples for the production of smoked sausages, it was found that the spectral characteristics corresponded to the number of viable lactic acid bacteria or their metabolic products. A method for real-time visualization of meat products in the terahertz range is proposed, which is promising for further research and obtaining indicators of the quality and safety of meat products. 7. A self-learning software algorithm based on a neural network has been developed for processing and interpreting data obtained using hyperspectral analysis and terahertz radiation. The first version of a self-learning supervised machine learning algorithm in the Python programming language has been developed. The training and measurement samples are DataFrame format data arrays containing different data sets. The composition and method of processing data sets, as well as possible options for training and measurement categories are determined. The connection of the algorithm to the database via the API protocol has been implemented.

 

Publications

---