Method and device for identifying contaminated beef based on dynamic metabolic stimulation and odor fingerprint

Abstract

The application discloses a kind of based on dynamic metabolic excitation and smell fingerprint pollution beef identification method and device.The method includes the following steps: S1 sample preparation and inoculation;S2 dynamic metabolic excitation and signal acquisition: the sample after inoculation is placed in the environment of program temperature control, and the high-resolution time series data aligned with temperature-time curve is obtained by executing the preset dynamic temperature cycle program;S3 dynamic feature extraction and processing: the time series data collected is preprocessed and feature extracted;High-dimensional dynamic feature vector is constructed;S4 pollution identification: the high-dimensional dynamic feature vector is input into the machine learning classification model trained in advance, and the microbial contamination type identification result of beef sample is output by the model.The method and device of the application can actively excite, capture and analyze the new method and device of specific metabolic signal in complex microbial contamination system, to realize the rapid and accurate identification of various foodborne pathogens in beef.

Description

Technical Field

[0001] This invention relates to the field of food safety testing, and in particular to a method and apparatus for identifying contaminated beef based on dynamic metabolic stimulation and odor fingerprinting. Background Technology

[0002] Meat products, especially beef, are at risk of microbial contamination throughout the entire supply chain, from farming to consumption, which is one of the main causes of foodborne illnesses. In actual distribution and storage, beef contamination often results from the coexistence of multiple bacteria, such as Salmonella, Escherichia coli O157:H7, and Listeria monocytogenes. These different bacterial communities engage in complex ecological competition and metabolic interactions, leading to overlapping and interference in the characteristic volatile organic compound (VOC) profiles they release. This makes it difficult for traditional detection methods to accurately distinguish the specific type and extent of contamination, posing a significant challenge to food safety early warning and control.

[0003] Currently, mainstream technologies for microbial testing in meat have significant limitations. Traditional plate culture methods, while reliable, are time-consuming (usually several days) and cumbersome, failing to meet the needs of rapid on-site screening. Molecular detection methods based on nucleic acids (such as PCR) or immunological principles (such as ELISA), while highly sensitive, require complex sample pretreatment and specialized operations, resulting in high costs and difficulty in reflecting the actual metabolic activity and dynamic risks of microorganisms in the food matrix. More importantly, these methods typically target a single target; in the presence of multiple microorganisms, synergistic or antagonistic interactions can lead to false negative or false positive results.

[0004] In recent years, detection technologies based on odor fingerprinting (VOCs spectra) have attracted attention due to their potential for speed and non-destructive operation. However, most existing technologies collect VOCs under static or isothermal conditions. In such steady-state environments, the metabolic activities of microorganisms tend to be stable, resulting in high similarity and insufficient specificity in the spectral characteristics of released VOCs. Especially in complex systems with multiple coexisting bacteria, dominant bacteria may mask the signals of inferior bacteria, leading to detection results that fail to accurately reflect the complete composition of pollution. Existing technologies lack both a clear definition of the signal thresholds for each bacterial species in mixed pollution and an effective strategy for distinguishing different pollution proportions, thus limiting their reliability and accuracy in practical applications. Summary of the Invention

[0005] To address the problem in existing technologies for microbial detection in meat where dominant bacteria in complex systems with multiple coexisting microorganisms may mask the signals of inferior bacteria, resulting in test results that fail to accurately reflect the complete composition of contamination, and to solve the limitations in reliability and accuracy of existing technologies in practical applications due to the lack of clear definitions of signal thresholds for each microbial species in mixed contamination and the absence of effective strategies to distinguish different contamination proportions, this invention proposes a novel method and apparatus capable of actively stimulating, capturing, and analyzing specific metabolic signals in complex microbial contamination systems to achieve rapid and accurate identification of multiple foodborne pathogens in beef.

[0006] To achieve this goal, the present invention adopts the following technical solution: A method for identifying contaminated beef based on dynamic metabolic excitation and odor fingerprinting includes the following steps: S1 Sample preparation and inoculation: Cut the beef sample into small pieces, and under aseptic conditions, inoculate the sample surface with a bacterial suspension of one or more foodborne pathogens. Then place the sample in a sealed or semi-sealed container; this step is the basis for building a controlled contamination model. S2 Dynamic Metabolic Excitation and Signal Acquisition: The inoculated sample is placed in a temperature-controlled environment and a preset dynamic temperature cycle program is executed. During the temperature cycle, the dynamic response signal of the sample headspace gas is continuously acquired using a sensor array to obtain high-resolution time-series data aligned with the temperature-time curve. S3 Dynamic Feature Extraction and Processing: Preprocessing and feature extraction are performed on the collected time-series data to extract multi-dimensional features, including steady-state features, transient dynamic features, and cycle consistency features, and construct a high-dimensional dynamic feature vector; S4 Contamination Identification: The high-dimensional dynamic feature vector is input into a pre-trained machine learning classification model, and the model outputs the identification result of the microbial contamination type of the beef sample.

[0007] Furthermore, in the method for identifying contaminated beef based on dynamic metabolic stimulation and odor fingerprinting of the present invention, in step S1, the foodborne pathogens include one or more of Salmonella, Escherichia coli O157:H7, and Listeria monocytogenes. These bacteria are key foodborne pathogens in beef and have important detection significance.

[0008] Furthermore, in the method for identifying contaminated beef based on dynamic metabolic stimulation and odor fingerprinting of the present invention, in step S1, when inoculating with one or more foodborne pathogens, the inoculation amount of each pathogen is not less than 50% of the minimum load threshold that can produce a detectable odor signal when it is infected alone. This ratio is an empirical value determined based on preliminary experiments, which can balance the relationship between detection sensitivity and real contamination simulation.

[0009] In addition, in the contaminated beef identification device based on dynamic metabolic stimulation and odor fingerprint of the present invention, in step S2, the preset dynamic temperature cycle program includes the following steps: a dual-cycle or multi-cycle mode between 5°C and 15°C, including a cycle unit of maintaining 5°C for a first duration, raising the temperature to 15°C and maintaining it for a second duration, and then cooling it back to 5°C and maintaining it for a third duration.

[0010] Furthermore, in the contaminated beef identification device based on dynamic metabolic stimulation and odor fingerprinting of the present invention, the dynamic temperature cycling program includes the following steps: first, maintaining the temperature at 5°C for 30 minutes, then raising the temperature to 15°C and maintaining it for 20 minutes, then lowering the temperature back to 5°C and maintaining it for 30 minutes, repeating this cycle at least twice. This program design fully utilizes the metabolic inhibition characteristics of the target bacterial strain at low temperatures and its metabolic activation characteristics in the mesophilic region, amplifying its metabolic differences through cyclic perturbation.

[0011] Furthermore, in the contaminated beef identification device based on dynamic metabolic excitation and odor fingerprinting of the present invention, in step S3, the steady-state characteristics include one or more selected from peak value, integral value, average differential value, energy value, and steady-state standard deviation; the transient dynamic characteristics include one or more selected from response delay time, maximum rate of change, time required to reach a predetermined percentage peak value, area under the transient curve, rise / fall time constant, and transient waveform fitting parameters; the cyclic consistency characteristics include one or more selected from the peak ratio of corresponding stages between different cycles, delay drift, and steady-state offset. These three types of characteristics together constitute fingerprint information that comprehensively describes the dynamic metabolic behavior of microorganisms.

[0012] In addition, in the device for identifying contaminated beef based on dynamic metabolic excitation and odor fingerprint of the present invention, step S3 further includes processing the high-dimensional dynamic feature vector by a dimensionality reduction method after feature extraction. The dimensionality reduction method includes principal component analysis, linear discriminant analysis, independent component analysis or isometric mapping.

[0013] Furthermore, in the contaminated beef identification device based on dynamic metabolic excitation and odor fingerprinting of the present invention, in step S4, the pre-trained machine learning classification model is obtained through the following method: acquiring a dynamic feature vector dataset containing beef samples of various known contamination types; dividing the dataset into a training set and a test set; using the training set, employing a Bayesian optimization algorithm to optimize the hyperparameters of candidate machine learning classifiers, and evaluating the model performance through cross-validation to obtain the optimal model; the candidate machine learning classifiers include support vector machines, random forests, backpropagation neural networks, K-nearest neighbors, or Naive Bayes. This model construction process ensures the accuracy, stability, and generalization ability of the final identification model.

[0014] The present invention also includes a device for identifying contaminated beef based on dynamic metabolic excitation and odor fingerprinting for implementing the above methods, the device comprising the following modules: An open detection chamber module includes a chamber body and an openable pull plate. The bottom of the chamber body is provided with an annular membrane-breaking blade and a sealing ring for sealing connection with a culture container. The pull plate is configured to open in the detection position to expose the sensor array inside the chamber, and to close in the cleaning position to seal the sensor array and form an independent cavity. The sensor array module, fixed inside the gas chamber, contains multiple sensors sensitive to different volatile organic compounds, used to detect the dynamic changes in the composition of the sample headspace gas in real time. The gas cleaning module includes an air pump and an airflow channel. When the pull plate is in the cleaning position, the air pump can pump clean gas into the independent cavity to clean the sensor array. The signal acquisition and processing module is electrically connected to the sensor array module and is used to acquire and process sensor signals and coordinate the timing control of each module.

[0015] In addition, in the device for identifying contaminated beef based on dynamic metabolic excitation and odor fingerprinting of the present invention, the sensor array module includes a metal oxide semiconductor sensor, which has a specific response to volatile organic compounds such as ammonia, sulfides, alcohols, aldehydes and ketones.

[0016] The purpose of this invention is to overcome the shortcomings of existing technologies and provide a method and apparatus for identifying multi-microbial contamination in beef based on dynamic metabolic stimulation and odor fingerprinting. This method disturbs the microbial metabolic homeostasis through controlled dynamic temperature cycling, stimulating a more discriminative transient metabolic response. Combined with high-dimensional odor fingerprint analysis and machine learning methods, it achieves rapid and accurate identification of contaminated systems with multiple coexisting pathogens.

[0017] Therefore, the method and apparatus for identifying contaminated beef based on dynamic metabolic stimulation and odor fingerprinting of the present invention have the following technical effects.

[0018] This invention overcomes the limitations of static detection by actively stimulating the specific metabolic response of microorganisms through dynamic temperature cycling. Combined with high-dimensional dynamic feature extraction and intelligent methods, it significantly improves the accuracy and reliability of identifying mixed contamination of multiple microorganisms. This method is rapid, non-destructive, and the device is highly automated, showing promising prospects for field applications. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the overall structure of the contaminated beef identification device based on dynamic metabolic stimulation and odor fingerprint according to a specific embodiment of the present invention.

[0020] Figure 2This is a schematic diagram of the overall structure of the contaminated beef identification device based on dynamic metabolic stimulation and odor fingerprint according to a specific embodiment of the present invention.

[0021] Figure 3 This is a schematic diagram of the overall process of the method for identifying contaminated beef based on dynamic metabolic stimulation and odor fingerprinting according to a specific embodiment of the present invention.

[0022] Figure 4 This is a schematic diagram of a dynamic temperature cycling program (5℃-15℃ double cycle) in a method for identifying contaminated beef based on dynamic metabolic stimulation and odor fingerprinting according to a specific embodiment of the present invention.

[0023] Figure 5 This is a schematic diagram of the typical response timing signal of the sensor array under dynamic temperature cycling in the method for identifying contaminated beef based on dynamic metabolic excitation and odor fingerprinting according to a specific embodiment of the present invention.

[0024] Figure 6 This is a schematic diagram of the dynamic feature extraction framework in the method for identifying contaminated beef based on dynamic metabolic excitation and odor fingerprinting according to a specific embodiment of the present invention, which shows three-dimensional features of steady state, transient state, and cyclic state.

[0025] Figure 7 This is a flowchart illustrating the training and identification process of a machine learning model in the method for identifying contaminated beef based on dynamic metabolic stimulation and odor fingerprinting, as described in a specific embodiment of the present invention.

[0026] Figure 8 This is a visualization of the distribution of samples with different contamination types after feature extraction and dimensionality reduction (LDA projection map) in the method for identifying contaminated beef based on dynamic metabolic excitation and odor fingerprint according to a specific embodiment of the present invention.

[0027] Figure 9 This is a PCA analysis diagram of the odor fingerprint characteristics of beef samples with different types of contamination in the method for identifying contaminated beef based on dynamic metabolic excitation and odor fingerprint according to a specific embodiment of the present invention.

[0028] Figure 10 The image shows radar diagrams of odor fingerprint responses of beef samples with different contamination types in the method for identifying contaminated beef based on dynamic metabolic excitation and odor fingerprinting according to a specific embodiment of the present invention. Detailed Implementation

[0029] The present invention will now be described in detail with reference to the accompanying drawings.

[0030] The following detailed exemplary embodiments are disclosed. However, the specific structural and functional details disclosed herein are merely for the purpose of describing exemplary embodiments.

[0031] However, it should be understood that the present invention is not limited to the specific exemplary embodiments disclosed, but covers all modifications, equivalents, and substitutions falling within the scope of this disclosure. Throughout the description of the drawings, the same reference numerals denote the same elements.

[0032] Referring to the accompanying drawings, the structures, proportions, sizes, etc., depicted in the drawings are merely for illustrative purposes to aid those skilled in the art in understanding and reading the content disclosed herein. They are not intended to limit the conditions under which the invention can be implemented and therefore have no substantial technical significance. Any modifications to the structure, changes in proportions, or adjustments to the size, without affecting the effects and objectives achieved by the invention, should still fall within the scope of the technical content disclosed herein. Furthermore, the positional limitations used in this specification are merely for clarity of description and are not intended to limit the scope of the invention. Changes or adjustments to their relative relationships, without substantially altering the technical content, should also be considered within the scope of the invention's implementation.

[0033] It should also be understood that the term “and / or” as used herein includes any and all combinations of one or more of the related listed items. Furthermore, it should be understood that when a component or unit is referred to as “connected” or “coupled” to another component or unit, it may be directly connected or coupled to the other component or unit, or there may be intermediate components or units. In addition, other words used to describe the relationship between components or units should be understood in the same manner (e.g., “between” versus “directly between,” “adjacent” versus “directly adjacent,” etc.).

[0034] Figure 1 and Figure 2 This is a schematic diagram of the overall structure of the contaminated beef identification device based on dynamic metabolic excitation and odor fingerprinting according to a specific embodiment of the present invention. As shown in the figure, the open detection chamber module of the contaminated beef identification device based on dynamic metabolic excitation and odor fingerprinting according to a specific embodiment of the present invention includes a chamber body 5 and an openable pull plate 4. The bottom of the chamber body 5 is provided with an annular membrane-breaking blade 6 and a sealing ring for sealing connection with a culture dish 7, exposing the sensor mounting position 2. The pull plate 4 is configured to open in the detection position, exposing the sensor 8 array inside the chamber, and close in the cleaning position, sealing the sensor 8 array to form an independent cavity. The sensor array module, fixed inside the gas chamber, contains multiple sensors sensitive to different volatile organic compounds, used to detect the dynamic changes in the composition of the sample headspace gas in real time. The gas cleaning module includes an air pump and an airflow channel. When the pull plate is in the cleaning position, the air pump can pump clean gas into the independent cavity to clean the sensor array. The signal acquisition and processing module is electrically connected to the sensor array module and is used to acquire and process sensor signals and coordinate the timing control of each module.

[0035] Specifically, the sensor array module, fixed to the top plate inside the gas chamber, includes multiple metal oxide semiconductor sensors sensitive to different VOCs, used for real-time detection of changes in headspace gas composition. Sensor types may include, but are not limited to, sensors with specific responses to ammonia, sulfides, alcohols, aldehydes, ketones, etc.

[0036] The gas cleaning module includes a miniature air pump, an air inlet 1, and an air outlet 3. When the pull plate 4 is in the cleaning position, the air pump pumps clean air into the independent cavity through the air inlet 1 to clean the sensor 8, and the exhaust gas is discharged through the air outlet 3, ensuring that the response baseline of the sensor 8 is stable and avoiding cross-contamination between samples.

[0037] The signal acquisition and processing module, including the main control unit, is used to acquire real-time electrical signals from the sensor array and to perform filtering, feature extraction, and digital processing on the signals. This module is also responsible for coordinating the timing control of each module.

[0038] Figure 3 This is a schematic diagram of the overall process of a method for identifying contaminated beef based on dynamic metabolic stimulation and odor fingerprinting according to a specific embodiment of the present invention. As shown in the figure, the specific embodiment of the present invention includes a method for identifying contaminated beef based on dynamic metabolic stimulation and odor fingerprinting, comprising the following steps: S1 Sample preparation and inoculation: Cut the beef sample into small pieces, and under aseptic conditions, inoculate the sample surface with a bacterial suspension of one or more foodborne pathogens. Then place the sample in a sealed or semi-sealed container; this step is the basis for building a controlled contamination model. S2 Dynamic Metabolic Excitation and Signal Acquisition: The inoculated sample is placed in a temperature-controlled environment and a preset dynamic temperature cycle program is executed. During the temperature cycle, the dynamic response signal of the sample headspace gas is continuously acquired using a sensor array to obtain high-resolution time-series data aligned with the temperature-time curve. S3 Dynamic Feature Extraction and Processing: Preprocessing and feature extraction are performed on the collected time-series data to extract multi-dimensional features, including steady-state features, transient dynamic features, and cycle consistency features, and construct a high-dimensional dynamic feature vector; S4 Contamination Identification: The high-dimensional dynamic feature vector is input into a pre-trained machine learning classification model, and the model outputs the identification result of the microbial contamination type of the beef sample.

[0039] Furthermore, in the method for identifying contaminated beef based on dynamic metabolic stimulation and odor fingerprinting of the present invention, in step S1, the foodborne pathogens include one or more of Salmonella, Escherichia coli O157:H7, and Listeria monocytogenes. These bacteria are key foodborne pathogens in beef and have important detection significance.

[0040] Furthermore, in the method for identifying contaminated beef based on dynamic metabolic stimulation and odor fingerprinting of the present invention, in step S1, when inoculating with one or more foodborne pathogens, the inoculation amount of each pathogen is not less than 50% of the minimum load threshold that can produce a detectable odor signal when it is infected alone. This ratio is an empirical value determined based on preliminary experiments, which can balance the relationship between detection sensitivity and real contamination simulation.

[0041] In addition, in the contaminated beef identification device based on dynamic metabolic stimulation and odor fingerprint of the present invention, in step S2, the preset dynamic temperature cycle program includes the following steps: a dual-cycle or multi-cycle mode between 5°C and 15°C, including a cycle unit of maintaining 5°C for a first duration, raising the temperature to 15°C and maintaining it for a second duration, and then cooling it back to 5°C and maintaining it for a third duration.

[0042] Furthermore, in the contaminated beef identification device based on dynamic metabolic stimulation and odor fingerprinting of the present invention, the dynamic temperature cycling program includes the following steps: first, maintaining the temperature at 5°C for 30 minutes, then raising the temperature to 15°C and maintaining it for 20 minutes, then lowering the temperature back to 5°C and maintaining it for 30 minutes, repeating this cycle at least twice. This program design fully utilizes the metabolic inhibition characteristics of the target bacterial strain at low temperatures and its metabolic activation characteristics in the mesophilic region, amplifying its metabolic differences through cyclic perturbation.

[0043] Furthermore, in the contaminated beef identification device based on dynamic metabolic excitation and odor fingerprinting of the present invention, in step S3, the steady-state characteristics include one or more selected from peak value, integral value, average differential value, energy value, and steady-state standard deviation; the transient dynamic characteristics include one or more selected from response delay time, maximum rate of change, time required to reach a predetermined percentage peak value, area under the transient curve, rise / fall time constant, and transient waveform fitting parameters; the cyclic consistency characteristics include one or more selected from the peak ratio of corresponding stages between different cycles, delay drift, and steady-state offset. These three types of characteristics together constitute fingerprint information that comprehensively describes the dynamic metabolic behavior of microorganisms.

[0044] In addition, in the device for identifying contaminated beef based on dynamic metabolic excitation and odor fingerprint of the present invention, step S3 further includes processing the high-dimensional dynamic feature vector by a dimensionality reduction method after feature extraction. The dimensionality reduction method includes principal component analysis, linear discriminant analysis, independent component analysis or isometric mapping.

[0045] Furthermore, in the contaminated beef identification device based on dynamic metabolic excitation and odor fingerprinting of the present invention, in step S4, the pre-trained machine learning classification model is obtained through the following method: acquiring a dynamic feature vector dataset containing beef samples of various known contamination types; dividing the dataset into a training set and a test set; using the training set, employing a Bayesian optimization algorithm to optimize the hyperparameters of candidate machine learning classifiers, and evaluating the model performance through cross-validation to obtain the optimal model; the candidate machine learning classifiers include support vector machines, random forests, backpropagation neural networks, K-nearest neighbors, or Naive Bayes. This model construction process ensures the accuracy, stability, and generalization ability of the final identification model.

[0046] The following are some more specific examples to illustrate specific embodiments of the present invention.

[0047] Specific Example 1: Detailed Steps and Parameter Settings of the Identification Method This example uses the identification of single and mixed contamination of Salmonella Enteritidis ATCC 13076, Escherichia coli O157:H7 ATCC 43895, and Listeria monocytogenes ATCC 19115 as examples to detail the methodological process (e.g., Figure 3-8 (As shown).

[0048] 1. Sample preparation and inoculation: Take fresh beef tenderloin and cut it into small pieces of 2 cm × 2 cm × 1 cm using a sterile knife in a biosafety cabinet.

[0049] Three target strains were activated and purified, and bacterial suspensions with concentrations of 10^8 CFU / mL were prepared using physiological saline. The concentrations were then accurately determined using the plate count method.

[0050] Signal threshold determination: Through preliminary experiments, single-bacterial suspensions of different concentrations were inoculated onto beef chunks and cultured statically at 5°C for 24 hours before detection. The minimum bacterial load required to produce a stable signal distinct from the control group was determined and recorded as the signal threshold for each bacterium: Th_Sal (Salmonella), Th_EC (Escherichia coli), and Th_Lis (Listeria). For example, Th_Sal ≈ 10^3 CFU / sample.

[0051] Experimental Groups: Control group: Inoculated with 100 μL of sterile saline.

[0052] Single-strain infection group: 100 μL of a single bacterial suspension was inoculated, with the bacterial count being twice the threshold (e.g., 2×Th).

[0053] Cross-infection group: Inoculate 100 μL of bacterial suspensions mixed in different ratios. Ensure that the final inoculation amount of each bacterium in the mixture is not less than 50% of its own Th. Ratios such as Salmonella:Escherichia coli:Listeria = 1:1:1, 10:1:1, 1:10:1, 1:1:10, etc. Five replicates are prepared for each ratio.

[0054] After inoculation, the suspension was evenly spread over the entire surface of the meat piece using a sterile spreader. The meat piece was then placed in a sterile sample bag with a 0.2 μm hydrophobic filter membrane and lightly sealed.

[0055] 2. Dynamic metabolic activation and signal acquisition: Place the sample bag into a programmable thermostatic incubator.

[0056] Figure 4 This diagram illustrates the dynamic temperature cycling program (5℃-15℃ dual cycle) in the method for identifying contaminated beef based on dynamic metabolic stimulation and odor fingerprinting according to a specific embodiment of the present invention. In this example, the following is executed: Figure 4 The dynamic temperature cycling program shown is as follows: Initial temperature 5°C, held for 30 minutes; then increase the temperature to 15°C at a rate of 2°C / min, held for 20 minutes; then decrease the temperature to 5°C at the same rate, held for 30 minutes. This constitutes one cycle, totaling 100 minutes. This embodiment executes two complete cycles consecutively.

[0057] At the start of the second cycle, the detection device is activated. The petri dish (simulating the sample bag environment) is placed under the device, the air chamber is closed, and the membrane is punctured. The pull plate is opened to the detection position. The sensor array begins continuous data acquisition at a sampling frequency of 1 Hz. The acquired signals are time-series data strictly aligned with the temperature profile. Figure 5 This is a schematic diagram of the typical response timing signal of the sensor array under dynamic temperature cycling in the method for identifying contaminated beef based on dynamic metabolic excitation and odor fingerprinting according to a specific embodiment of the present invention. Figure 5 As shown, at the inflection point of temperature rise and fall, different sensors exhibited different transient response peaks.

[0058] 3. Dynamic Feature Extraction and Processing: Preprocessing: The raw signal is smoothed and baseline corrected by applying a Savitzky-Golay filter (window size 7, polynomial order 3).

[0059] Feature extraction (e.g.) Figure 6 As shown in the framework, Figure 6This is a schematic diagram of the dynamic feature extraction framework in the method for identifying contaminated beef based on dynamic metabolic excitation and odor fingerprinting according to a specific embodiment of the present invention, showing steady-state, transient, and cyclic three-dimensional features. Steady-state characteristics: extracted from the last 5-minute data window of each temperature holding phase. For example, from the end of the first 5°C holding phase: the average value (steady-state value) and standard deviation of the signals from each of the 8 sensors; from the end of the 15°C holding phase: the peak value, the integral value of the signal in that phase (approximate total VOCs released), and the average value of the absolute value of the first derivative (average rate of change).

[0060] Transient dynamic characteristics: For each heating (5°C→15°C) and cooling (15°C→5°C) process. Taking the heating process as an example, for the response curve of each sensor, calculate: the delay time from the start of temperature rise to the start of a significant signal rise; the maximum slope (maximum rate of change) of the signal rise segment; the time required from the start of the rise to reach 90% of the peak value of the heating response; and perform exponential function fitting on the transient peak to obtain the rise time constant and attenuation coefficient.

[0061] Cyclic consistency characteristics: Compare the responses at the same temperature stage (e.g., two 15°C hold periods) in two cycles. Calculations: The ratio of the peak value of the second cycle to the peak value of the first cycle; the difference in the same transient characteristic (e.g., delay time) between the two cycles.

[0062] Organized by “sensor number × feature type × cycle stage”, this embodiment extracts a total of 8 sensors × (5 steady state + 4 transient state) × 3 stages = 216-dimensional initial feature vectors.

[0063] Feature optimization: Outlier features are removed using the local outlier factor algorithm. Then, principal component analysis and linear discriminant analysis (LDA) are attempted for dimensionality reduction. By comparing the classification performance of the dimensionality-reduced features on the validation set, this embodiment selects LDA, reducing the features to 5-10 dimensions, which best separates different contamination categories. Figure 8 This is a visualization of the distribution of samples with different contamination types after feature extraction and dimensionality reduction (LDA projection map) in the method for identifying contaminated beef based on dynamic metabolic excitation and odor fingerprinting according to a specific embodiment of the present invention. Figure 9 This is a PCA analysis diagram of the odor fingerprint characteristics of beef samples with different types of contamination in the method for identifying contaminated beef based on dynamic metabolic excitation and odor fingerprint according to a specific embodiment of the present invention. Figure 10 This is a radar image showing the odor fingerprint response of beef samples with different contamination types in a method for identifying contaminated beef based on dynamic metabolic excitation and odor fingerprinting, according to a specific embodiment of the present invention. Figures 8-10It can be seen that the method and apparatus for identifying contaminated beef based on dynamic metabolic stimulation and odor fingerprinting according to specific embodiments of the present invention can overcome the limitations of static detection, and significantly improve the identification accuracy and reliability of mixed contamination of multiple microorganisms by combining high-dimensional dynamic feature extraction and intelligent methods.

[0064] 4. Model training and contamination identification (e.g.) Figure 7 As shown in the process, Figure 7 This is a flowchart illustrating the training and identification process using a machine learning model in a method for identifying contaminated beef based on dynamic metabolic excitation and odor fingerprinting, according to a specific embodiment of the present invention. The final feature vectors and labels of all samples (including control group, single bacteria group, and mixed groups with different proportions at 24h, 48h, and 72h) were collected to form the total dataset.

[0065] The training set and the independent test set are randomly divided in a 7:3 ratio.

[0066] On the training set: Bayesian optimization tools (such as MATLAB's bayesopt) are used to automatically optimize the hyperparameters of support vector machines, random forests, and backpropagation neural networks, with the optimization objective being the average accuracy of 5-fold cross-validation.

[0067] The evaluation found that the support vector machine based on the radial basis function kernel function performed best on this dataset, with a cross-validation accuracy of 98.5%.

[0068] The final discrimination model is obtained by retraining on the entire training set using the optimal hyperparameters.

[0069] Identification Application: For a new beef sample to be tested, the same procedures are strictly followed for preparation, dynamic excitation, signal acquisition, and feature extraction to obtain its feature vector. This vector is then input into a trained SVM model, and the model output is the predicted contamination type (e.g., "Listeria monocytogenes single contamination" or "Salmonella and Escherichia coli mixed contamination, ratio approximately 10:1").

[0070] Example 2: Overall Operation Procedure of the Method and Apparatus This embodiment outlines the complete operation process from sample to result, and further clarifies the collaborative working process of the device.

[0071] The operator first cuts and inoculates the beef sample with pathogens on a sterile table, then places the sample into a dedicated petri dish. The petri dish is then placed in an incubator with a pre-set dynamic temperature program. When the testing time point is reached (e.g., 24 hours after inoculation), the following procedures are performed: 1. Remove the target culture dish from the incubator and quickly place it in the positioning slot below the detection device.

[0072] 2. Initiate the "Detection" command via the host computer software. The main control unit of the device controls the gas chamber to descend, the annular membrane-breaking blade punctures the sealing membrane of the culture dish, and the sealing ring is tightened.

[0073] 3. The main control unit drives the pull plate to move to the "detection position", exposing the sensor array.

[0074] 4. The sensor begins acquiring data, while the incubator executes a pre-set, shortened dynamic temperature excitation cycle (e.g., only the second cycle described in Example 1 is executed). The data is uploaded to the host computer in real time.

[0075] 5. After the excitation cycle and data acquisition are completed, the host computer software automatically extracts features and calls the loaded machine learning model for real-time identification. The results are displayed on the software interface.

[0076] 6. The main control unit drives the pull plate to move to the "cleaning position" and starts the micro air pump to purge the sensor cavity with clean air for 90 seconds.

[0077] 7. The air chamber rises, and the petri dish is removed. One test is completed, the apparatus is ready, and the next sample can be tested.

[0078] Example 3: Optimization of Cycling Parameters at Different Temperatures This embodiment aims to illustrate the adjustability of the dynamic temperature cycling program and its impact on discrimination performance, and is a moderate extension of the method of the present invention. Its core lies in perturbing microbial metabolism through programmed temperature changes. Those skilled in the art will understand that, for different detection targets, the cycling parameters can be optimized to obtain a more discriminative dynamic response signal.

[0079] In addition to the 5-15℃ dual-cycle method used in Example 1, other temperature ranges and cycle modes can be explored. For example: Option A: To improve the detection sensitivity for low bacterial count contamination, a high-frequency cycling mode with small temperature variations (such as three cycles or more) can be used, or a small temperature fluctuation intervention can be performed in the range close to the minimum growth temperature of the target bacteria (such as 4℃-12℃) to achieve effective accumulation and amplification of metabolic response signals.

[0080] Option B: To better distinguish between bacterial species with similar metabolic behaviors, the rate of temperature change could be adjusted. For example, using a slower heating rate (e.g., 1°C / min) might help capture richer transient dynamic details, thus providing more differentiated characterization.

[0081] Option C: To further amplify the differences in temperature adaptability among different bacterial species, short-term, minute temperature fluctuations can be introduced into the main temperature plateau (such as inserting a brief cooling pulse during the 15℃ plateau period) to observe the differences in their metabolic recovery capabilities.

[0082] The above is merely an illustrative description of the parameter optimization direction, intended to illustrate the flexibility and adjustability of the method of the present invention. In practical applications, the optimal dynamic temperature cycling program can be determined through limited conventional experiments based on the characteristics of the target pathogen.

[0083] Example 4: Targeting other foodborne pathogens It should be noted that the core technical idea of ​​this invention—that is, to disturb the metabolism of microorganisms through a preset dynamic temperature cycling program and to collect and analyze their dynamic odor fingerprints—is not limited to identifying Salmonella, Escherichia coli O157:H7 and Listeria monocytogenes in Example 1.

[0084] This method can, in principle, be applied to other foodborne pathogens that are common in beef, have metabolic activity, and produce characteristic volatile organic compounds, such as Staphylococcus aureus or Clostridium perfringens.

[0085] For example: For Staphylococcus aureus, cycling in a relatively high temperature range (e.g., 10℃-25℃) can be considered to stimulate its metabolism. For the strictly anaerobic Clostridium perfringens, a cycling program can be designed under anaerobic conditions, including a range from low temperature (inhibition) to medium temperature (activation), and a brief heat shock pulse can be added to observe its (spore) response.

[0086] Subsequently, following the same process as described in this invention: determining the signal threshold under the corresponding excitation procedure, constructing contaminated samples, collecting dynamic response signals, extracting multi-dimensional dynamic features (steady-state, transient, and cyclic consistency features), and training a new machine learning model, a method for identifying new target bacterial species can be established.

[0087] For those skilled in the art, once the type of target microorganism is identified, its suitable dynamic excitation parameters can be determined through limited conventional experiments, thus extending the application of the method of the present invention to new detection targets.

[0088] The foregoing description illustrates and describes several preferred embodiments of the present invention. However, as mentioned above, it should be understood that the present invention is not limited to the forms disclosed in this specification and should not be construed as excluding other embodiments. It can be used in various other combinations, modifications, and environments, and can be altered within the scope of the inventive concept described in this specification through the foregoing teachings or techniques or knowledge in related fields. Any modifications and variations made by those skilled in the art that do not depart from the spirit and scope of the present invention should be within the protection scope of the appended claims.

Claims

1. A method for identifying contaminated beef based on dynamic metabolic excitation and odor fingerprinting, characterized in that, Includes the following steps: S1 Sample preparation and inoculation: Cut the beef sample into small pieces, and under aseptic conditions, inoculate the sample surface with a bacterial suspension of one or more foodborne pathogens. Then place the sample in a sealed or semi-sealed container. S2 Dynamic Metabolic Excitation and Signal Acquisition: The inoculated sample is placed in a temperature-controlled environment and a preset dynamic temperature cycle program is executed. During the temperature cycle, the dynamic response signal of the sample headspace gas is continuously acquired using a sensor array to obtain high-resolution time-series data aligned with the temperature-time curve. S3 Dynamic Feature Extraction and Processing: Preprocessing and feature extraction are performed on the collected time-series data to extract multi-dimensional features, including steady-state features, transient dynamic features, and cycle consistency features, and construct a high-dimensional dynamic feature vector; S4 Contamination Identification: The high-dimensional dynamic feature vector is input into a pre-trained machine learning classification model, and the model outputs the identification result of the microbial contamination type of the beef sample.

2. The method for identifying contaminated beef based on dynamic metabolic stimulation and odor fingerprinting as described in claim 1, characterized in that, In step S1, the foodborne pathogens include one or more of Salmonella, Escherichia coli O157:H7, and Listeria monocytogenes.

3. The method for identifying contaminated beef based on dynamic metabolic stimulation and odor fingerprinting as described in claim 1, characterized in that, In step S1, when inoculating one or more foodborne pathogens, the inoculation amount of each pathogen shall not be less than 50% of the minimum load threshold that can produce a detectable odor signal when it is infected alone.

4. The method for identifying contaminated beef based on dynamic metabolic stimulation and odor fingerprinting as described in claim 1, characterized in that, In step S2, the preset dynamic temperature cycling program includes the following steps: a dual-cycle or multi-cycle mode between 5°C and 15°C, including a cycle unit that maintains 5°C for a first duration, raises the temperature to 15°C and maintains it for a second duration, and then cools it back to 5°C and maintains it for a third duration.

5. The method for identifying contaminated beef based on dynamic metabolic stimulation and odor fingerprinting as described in claim 4, characterized in that, The dynamic temperature cycling program includes the following steps: first, maintain the temperature at 5°C for 30 minutes, then raise the temperature to 15°C and maintain it for 20 minutes, then lower the temperature back to 5°C and maintain it for 30 minutes, and repeat this cycle at least twice.

6. The method for identifying contaminated beef based on dynamic metabolic stimulation and odor fingerprinting as described in claim 1, characterized in that, In step S3, the steady-state characteristics include one or more selected from peak value, integral value, average derivative value, energy value, and steady-state standard deviation; the transient dynamic characteristics include one or more selected from response delay time, maximum rate of change, time required to reach a predetermined percentage peak value, area under the transient curve, rise / fall time constant, and transient waveform fitting parameters; the cyclic consistency characteristics include one or more selected from peak ratio of corresponding stages between different cycles, delay drift, and steady-state offset.

7. The method for identifying contaminated beef based on dynamic metabolic stimulation and odor fingerprinting as described in claim 1, characterized in that, Step S3 further includes processing the high-dimensional dynamic feature vector using a dimensionality reduction method after feature extraction. The dimensionality reduction method includes principal component analysis, linear discriminant analysis, independent component analysis, or isometric mapping.

8. The method for identifying contaminated beef based on dynamic metabolic stimulation and odor fingerprinting as described in claim 1, characterized in that, In step S4, the pre-trained machine learning classification model is obtained by acquiring a dynamic feature vector dataset containing beef samples with multiple known contamination types. The dataset is divided into a training set and a test set. Using the training set, the Bayesian optimization algorithm is used to optimize the hyperparameters of the candidate machine learning classifiers, and the model performance is evaluated through cross-validation to obtain the optimal model. The candidate machine learning classifiers include support vector machines, random forests, backpropagation neural networks, K-nearest neighbors, or Naive Bayes.

9. A device for identifying contaminated beef based on dynamic metabolic excitation and odor fingerprinting for implementing the method of any one of claims 1-8, characterized in that, Includes the following modules: An open detection chamber module includes a chamber body and an openable pull plate. The bottom of the chamber body is provided with an annular membrane-breaking blade and a sealing ring for sealing connection with a culture container. The pull plate is configured to open in the detection position to expose the sensor array inside the chamber, and to close in the cleaning position to seal the sensor array and form an independent cavity. The sensor array module, fixed inside the gas chamber, contains multiple sensors sensitive to different volatile organic compounds, used to detect the dynamic changes in the composition of the sample headspace gas in real time. The gas cleaning module includes an air pump and an airflow channel. When the pull plate is in the cleaning position, the air pump can pump clean gas into the independent cavity to clean the sensor array. The signal acquisition and processing module is electrically connected to the sensor array module and is used to acquire and process sensor signals and coordinate the timing control of each module.

10. The device for identifying contaminated beef based on dynamic metabolic excitation and odor fingerprinting as described in claim 9, characterized in that, The sensor array module includes a metal oxide semiconductor sensor, which has a specific response to volatile organic compounds such as ammonia, sulfides, alcohols, aldehydes, and ketones.

Images