An experimental animal behavior monitoring system for a feeding room

Abstract

The application discloses an experimental animal behavior monitoring system of a feeding room, and relates to the technical field of animal behavior research.The system comprises a behavior information collection module, which uses a monitoring device to collect original data of experimental animal behavior information of the feeding room in real time, and transmits the original data to a data processing and classification module after collection; the data processing and classification module receives the original data transmitted by the behavior information collection module, calculates, classifies, counts and stores the original data by using a processing and classification algorithm, and forms animal behavior characteristic data after processing the original data, and respectively transmits the animal behavior characteristic data to a correlation analysis module and a comparative analysis module; the data processing and classification module, the correlation analysis module and the comparative analysis module are adopted, original data of experimental animal behavior information collected by the behavior information collection module in real time and comprehensively is accurately processed, and the correlation law and the behavior mode between experimental animal behaviors in a large amount of animal behavior data are found.

Description

Technical Field

[0001] This invention relates to the field of animal behavior research technology, and more specifically to a laboratory animal behavior monitoring system for a breeding room. Background Technology

[0002] With the continuous exploration of medical scientific research experiments and the ongoing development of teaching and talent training, small animals kept in breeding rooms have played a significant role in experimental research. By monitoring the animals' behavior, physiological and psychological states, they provide more accurate and objective experimental data, helping researchers obtain reliable research results and better understand the effects and impacts of animal experiments. The experimental animal behavior monitoring system in the breeding room typically monitors multiple indicators of the experimental animals in real time, including their environment, movement behavior, eating behavior, sleep behavior, and social behavior. This comprehensive assessment of animal behavioral characteristics helps to capture and identify potential abnormal behavioral features, and further optimizes experimental design and breeding conditions by regulating influencing factors.

[0003] For example, an experimental animal behavior monitoring system disclosed in application number CN201910897386.3 continuously monitors animals and calculates their autonomous activity levels for observation and recording without the need for frequent alerts from staff. However, it suffers from drawbacks such as uncontrollable monitoring of abnormal animal behavior signals, lack of state assessment and comparative analysis of pre- and post-test data, leading to deviations in experimental research direction, wasted manpower and resources in multiple experiments, low efficiency, and impact on the judgment of behavioral evidence. Furthermore, monitoring only the activity level of mice lacks behavioral analysis, resulting in mismatched control strategies and affecting the behavioral defects of other mice. An animal behavior monitoring system disclosed in application number CN201810545588.7 uses a monitoring platform, image acquisition device, optogenetic regulation device, EEG fiber optic coupling acquisition device, and biofeedback device to implement a strategy of choosing between competitive high-reward and low-reward food rewards with peers. However, it suffers from the drawback of animals preferring refined feed and reducing their demand for coarse feed, increasing the cost of animal husbandry.

[0004] The existing technology has the following shortcomings: When the monitoring accuracy of the experimental animal behavior characteristics in the experimental animal behavior monitoring system in the breeding room decreases, or when the monitored factors are not comprehensive or when no correlation analysis is established between the factor indicators, the accuracy of animal behavior detection will be reduced. If individual differences in animals, subtle changes in posture or body shape, or faint changes in sound are not monitored in time, it will pose a great threat to the life safety of the experimental animals and affect the validity and accuracy of the experimental results.

[0005] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] The purpose of this invention is to provide a monitoring system for the behavior of laboratory animals in a breeding room. This invention employs a data processing classification module, a correlation analysis module, and a comparative analysis module, which enables the behavior information acquisition module to collect and monitor the behavioral activity information of laboratory animals in real time and accurately process it. This allows for the discovery of correlation patterns and behavioral patterns among a large amount of animal behavior data, thereby solving the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a laboratory animal behavior monitoring system for a breeding room, comprising a behavior information acquisition module: using monitoring equipment to collect raw data of laboratory animal behavior information in the breeding room in real time, and after collection, transmitting the raw data to a data processing and classification module;

[0008] Data processing and classification module: Receives raw data transmitted from the behavior information collection module, performs calculations, classifications, statistics, and storage using processing and classification algorithms, and processes the raw data to form animal behavior characteristic data, which is then transmitted to the association analysis module and the comparison analysis module respectively.

[0009] Association Analysis Module: By establishing a K-means clustering algorithm model and then processing the animal behavior characteristic data classified by the data processing and classification module, association influence coefficients are generated to discover the association patterns between animal behaviors and to reveal non-intuitive behavioral patterns.

[0010] Comparative analysis module: Collects animal behavioral characteristic data, establishes neural network models for comparative calculations, and evaluates the degree of influence of environmental, behavioral and physiological information on experimental animal behavior, obtains the influence concealment coefficient, and is used to identify and judge behavioral changes caused by environmental, behavioral and physiological information;

[0011] Abnormal Behavior Diagnosis Module: Based on predefined abnormal behavior parameters and learned behavior patterns, the module analyzes the received concealment coefficients to identify and diagnose any abnormal behavior and reports the diagnosis results to the comprehensive monitoring and control module.

[0012] Integrated monitoring and control module: Integrates the outputs of various modules within the experimental animal behavior monitoring system, presents the data in a visual form, and automatically adjusts monitoring parameters and issues alarms based on the standards set in the breeding room and the results of data analysis, providing researchers in animal experiments with complete and real-time animal behavior monitoring results.

[0013] Preferably, the raw data includes environmental information of the animal living in the laboratory, labeled EI; behavioral information during movement, labeled BI; and physiological information of vital signs, labeled PI. The environmental information EI includes temperature, humidity, gas concentration, light intensity, and noise, and is labeled Eit, Eih, Eig, Eil, and Ein, respectively. The behavioral information BI includes activity level, social status, feeding status, sleep pattern, and cleaning and maintenance behavior, and is labeled Bia, Bis, Bie, Bip, and Bic, respectively. The physiological information PI includes weight, body temperature, respiratory rate, heart rate, and neuropsychiatric mood, and is labeled Piw, Pit, Pir, Pih, and Pin, respectively.

[0014] Preferably, the logical steps for obtaining the classification algorithm are as follows:

[0015] Supervised learning algorithms were used to train the experimental animals in the breeding room on raw data to obtain animal behavioral characteristic data;

[0016] For classification, a nonlinear support vector machine is used, and a kernel function is introduced to map the sample parameters in the original data to a high-dimensional eigenspace for linear separability. The calculation formula is as follows: In the formula, ω T The direction is represented by the hyperplane. Represented as the function value of the original data in the eigenspace. denoted as the feature vector after mapping the original data, b represents the bias constant;

[0017] The formula for minimizing the objective function is: In the formula, These represent class labels for environmental information, behavioral information, and physiological information, respectively. Eit, Eih, Eig, Eig, Ein, Bia, Bis, Bie, Bip, Bic, Piw, Pit, Pir, Pih, and Pin represent training samples.

[0018] The algorithm is introduced to learn and solve the optimal classification function, and the calculation formula is as follows: In the formula, Represented as the inner product of the sample and label in the feature space, the solution is obtained as follows: In the formula, The optimal classification value for extracting features from the original data is denoted as BC = {EI′, BI′, PI′}, which is used to obtain accurate animal behavior feature data. In this formula, BC represents the set of animal behavior feature data.

[0019] Preferably, the animal behavioral characteristic data is obtained by performing data denoising and cleaning, data normalization, feature extraction, and data classification on the original data, wherein the calculation formula for data normalization is as follows: In the formula, EI′, BI′, and PI′ represent the normalized data of environmental information EI, behavioral information BI, and physiological information PI, respectively. min(EI) and max(EI) represent the minimum and maximum values ​​of environmental information EI, min(BI) and max(BI) represent the minimum and maximum values ​​of behavioral information BI, and min(PI) and max(PI) represent the minimum and maximum values ​​of physiological information PI, respectively.

[0020] Preferably, the steps for establishing the K-means clustering algorithm model are as follows:

[0021] Three data points, EI0, ​​BI0, and PI0, representing environmental, behavioral, and physiological information, were selected as the initial center points.

[0022] Using the Euclidean distance formula, for each point sample in the animal behavioral characteristic data, the sum of the distances to each center point EI0, ​​BI0, and PI0 is calculated. The formula for calculating EI for environmental information is as follows: And x i = (Eit, Eih, Eig, Eil, Ein), where i represents the number of feature points of the environmental information, x i Let d1(x) be the i-th feature value of data point x. i (EI0) 2 Represented as x i Assign it to the cluster nearest to the environmental information EI;

[0023] The formula for calculating behavioral information (BI) is as follows: And x k = (Bia, Bis, Bie, Bip, Bic), where k represents the number of feature points of behavioral information, x k Let d2(x) be the k-th feature value of data point x. k ,BI0) 2 Represented as x k Assign it to the cluster closest to the behavioral information BI;

[0024] The formula for calculating the physiological information PI is as follows: And x m = (Piw, Pit, Pir, Pih, Pin), where m represents the number of feature points of physiological information, x m Let d3(x) be the m-th feature value of data point x. m ,PI0) 2 Represented as x m Assign it to the cluster closest to the physiological information PI.

[0025] Preferably, for each cluster, the average value algorithm is used to recalculate and update the center point, and the calculation formula is as follows: In the formula, EI1, BI1, and PI1 represent the updated center points, respectively;

[0026] After iteratively calculating the operation of updating the centroids several times, when the cluster assignments no longer change, the minimum sum of squared errors of all clusters is obtained, denoted as SSE, and the formula is: In the formula, SSE=d(x i,k,m EI1, BI1, and PI1 represent data points x, respectively. i x k x m The distances to the corresponding update center points EI1, BI1, and PI1.

[0027] Using the sum of squared errors for each cluster in the K-means clustering algorithm, association rule mining is performed to generate association influence coefficients, which are then denoted as Cic. The calculation formula is as follows: In the formula, ρ represents the strength of the correlation among environmental information, behavioral information, and physiological information, and d1(x i ,EI0),d2(x k ,BI0), d3(x m , PI0) represent data points x respectively i x k x m The distances to each corresponding center point EI0, ​​BI0, PI0.

[0028] Preferably, the processing steps of the neural network model are as follows:

[0029] A recurrent neural network (RNN) model is constructed by dividing the collected animal behavior feature data into training and testing sets and inputting them into the RNN input layer, so that the number of input nodes matches the number of selected animal behavior features.

[0030] The predicted output of the neural network is calculated through forward propagation, and a loss function based on mean squared error is defined to measure the difference between the predicted output and the actual label. In the forward propagation calculation, an initial weight matrix w is first set. EI w BI w PI , w z and initialize the hidden state h0;

[0031] For each time step t, the input feature EI of the current time step is... t BI t PI t The hidden state h of the previous time stept-1 As input, to compute the hidden state h at the current time step t. t ,and In the formula, G represents the activation function, and · represents matrix multiplication;

[0032] Then calculate the predicted output z based on the hidden state. t The formula for calculating the prediction result is z. t =G(w z ×h t ).

[0033] Preferably, the loss function calculation formula for the mean squared error (MSE) is defined as follows: In the formula, 15 represents the total number of samples in the original data, z (EI、BI、PI) Let y represent the predicted output for each sample in the environmental, behavioral, and physiological information categories, respectively. (EI、BI、PI) These are respectively represented as the actual labels corresponding to environmental information, behavioral information, and physiological information;

[0034] Based on the predicted output z t The gradient is calculated using the backpropagation algorithm, and the network parameters are updated using the stochastic gradient descent optimization algorithm to minimize the loss function. This is achieved by introducing a learning rate η that controls the step size of the parameter updates, and the network parameters w to be updated, where w = {w...} EI w BI w PI w ht w z gradient of the loss function Perform backpropagation, calculate the effect of the gradient on each parameter, and use the gradient descent formula. In the formula, w′ represents the updated network parameters, and the network parameters are updated.

[0035] Preferably, the calculation logic for the influencing concealment coefficient is as follows:

[0036] The test set is predicted using a trained recurrent neural network (RNN) model, and the prediction result z is obtained. t Based on feature importance analysis, the weight of each feature in the RNN model prediction is evaluated to calculate the influence concealment coefficient of each feature. This coefficient is calculated by subtracting all combinations BC′ that form a subset of the target feature from each feature data BC.

[0037] For each combination of target feature subsets, calculate the predicted difference value ΔY of the corresponding output of the RNN model, i.e., ΔY = z t (BC)-z t (BC′);

[0038] For each target feature subset, calculate the total number of combinations ΔX of target feature subsets of different sizes, i.e., ΔX = m! (15! - m! - 1), where 15 represents the total number of feature data, m represents the size of the target feature subset, and ! represents the factorial;

[0039] Calculate the concealment coefficient φ for each feature. i ,Right now and 1≤i≤15, where BC i Let be the i-th feature in the target feature subset, and BC i ∈BC 1-15 BC 1-15 = {Eit, Eih, Eig, Eig, Ein, Bia, Bis, Bie, Bip, Bic, Piw, Pit, Pir, Pih, Pin}, where S represents the feature subset and i represents the number of target features, used to assess the specific impact of each environmental, behavioral, and physiological information on the behavior of experimental animals.

[0040] Preferably, the monitoring method comprises the following steps:

[0041] S1. Install cameras and smart sensor devices to capture the behavior and activities of laboratory animals and obtain raw data on the behavior of laboratory animals.

[0042] S2. Perform preprocessing and normalization on the raw data to generate animal behavioral characteristic data that is easy to process and calculate later.

[0043] S3. Use the K-means clustering algorithm to perform association analysis on animal behavioral characteristic data to discover the association patterns between different behaviors;

[0044] S4. Use neural network comparative analysis to identify and judge behavioral changes caused by environmental information, behavioral information and physiological information, so as to detect changes and abnormalities in animal behavior;

[0045] S5. Diagnose abnormal behavioral changes through monitoring methods, reversely assess the causes of abnormal behavior, and adjust environmental parameters and conduct human intervention on experimental animals based on the results of data visualization and report generation from the comprehensive monitoring and control module.

[0046] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0047] This invention employs data processing and classification modules, correlation analysis modules, and comparative analysis modules. These modules enable the real-time and comprehensive collection and accurate processing of behavioral activity information from laboratory animals. This allows for the discovery of correlations and patterns among large amounts of animal behavior data. Furthermore, analysis based on received concealment coefficients identifies and diagnoses any abnormal behaviors in the animals, facilitating timely detection and resolution of potential problems and ensuring the health of the animals and the reliability of the research. Finally, the comprehensive analysis and visualization provided by the integrated monitoring and control module allow caretakers to understand the animals' behavioral status in real time, detect abnormalities early, and take timely measures, improving the efficiency and accuracy of husbandry and management, while simultaneously supporting the reliability and accuracy of scientific experiments and results. Attached Figure Description

[0048] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0049] Figure 1 This is a block diagram of the experimental animal behavior monitoring system in the breeding room of the present invention.

[0050] Figure 2 This is a flowchart of the experimental animal behavior monitoring method in the breeding room of the present invention. Detailed Implementation

[0051] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that the description of this disclosure will be more complete and fully convey the concept of the exemplary embodiments to those skilled in the art.

[0052] like Figure 1 As shown, the present invention provides a laboratory animal behavior monitoring system for a breeding room, including a behavior information acquisition module: using monitoring equipment to collect raw data of laboratory animal behavior information in the breeding room in real time, and after collection, transmitting the raw data to a data processing and classification module;

[0053] It should be noted that the monitoring equipment includes environmental sensors, visual cameras, and vital sign detectors, which are used to capture and collect environmental information about the animals' daily life in the laboratory, behavioral information about the animals' movement, eating, socializing and sleeping, as well as physiological information such as the animals' heart rate and body temperature.

[0054] The raw data includes environmental information of the animals living in the laboratory, labeled EI; behavioral information during movement, labeled BI; and physiological information of vital signs, labeled PI. Among them, environmental information EI includes temperature, humidity, gas concentration, light intensity, and noise, and is labeled Eit, Eih, Eig, Eil, and Ein, respectively; behavioral information BI includes activity level, social status, feeding status, sleep pattern, and cleaning and maintenance behavior, and is labeled Bia, Bis, Bie, Bip, and Bic, respectively; and physiological information PI includes weight, body temperature, respiratory rate, heart rate, and neuroemotional state, and is labeled Piw, Pit, Pir, Pih, and Pin, respectively.

[0055] Data processing and classification module: Receives raw data transmitted from the behavior information collection module, performs calculations, classifications, statistics, and storage using processing and classification algorithms, and processes the raw data to form animal behavior characteristic data, which is then transmitted to the association analysis module and the comparison analysis module respectively.

[0056] It should be noted that the logical steps for obtaining the classification algorithm are as follows:

[0057] Supervised learning algorithms were used to train the experimental animals in the breeding room on raw data to obtain animal behavioral characteristic data;

[0058] For classification, a nonlinear support vector machine is used, introducing a kernel function to map the sample parameters in the original data to a high-dimensional eigenspace for linear separability. The calculation formula is as follows: In the formula, ω T The direction is represented by the hyperplane. Represented as the function value of the original data in the eigenspace. denoted as the feature vector after mapping the original data, b represents the bias constant;

[0059] The formula for minimizing the objective function is: In the formula, These represent class labels for environmental information, behavioral information, and physiological information, respectively. Eit, Eih, Eig, Eig, Ein, Bia, Bis, Bie, Bip, Bic, Piw, Pit, Pir, Pih, and Pin represent training samples.

[0060] The algorithm is introduced to learn and solve the optimal classification function, and the calculation formula is as follows: In the formula, Represented as the inner product of the sample and label in the feature space, the solution is obtained as follows: In the formula, The optimal classification value for extracting features from the original data is denoted as BC = {EI′, BI′, PI′}, which is used to obtain accurate animal behavior feature data. In this formula, BC represents the set of animal behavior feature data.

[0061] Animal behavioral characteristic data involves denoising and cleaning the raw data, normalizing the data, extracting features, and classifying the data. Data classification is achieved using labeled data—environmental, behavioral, and physiological information—collected from experimental animals in predefined enclosures. This classification is accomplished through dynamic learning using a supervised learning algorithm. The formula for data normalization is as follows: In the formula, EI′, BI′, and PI′ represent the normalized data of environmental information EI, behavioral information BI, and physiological information PI, respectively. min(EI) and max(EI) represent the minimum and maximum values ​​of environmental information EI, min(BI) and max(BI) represent the minimum and maximum values ​​of behavioral information BI, and min(PI) and max(PI) represent the minimum and maximum values ​​of physiological information PI, respectively.

[0062] Association Analysis Module: By establishing a K-means clustering algorithm model and then processing the animal behavioral characteristic data classified by the data processing and classification module, association influence coefficients are generated to discover the association patterns between animal behaviors, revealing non-intuitive behavioral patterns and helping to understand the complex behavioral states of animals.

[0063] It should be noted that the steps for establishing the K-means clustering algorithm model are as follows:

[0064] Three data points, EI0, ​​BI0, and PI0, representing environmental, behavioral, and physiological information, were selected as the initial center points.

[0065] Using the Euclidean distance formula, for each point sample in the animal behavioral characteristic data, the sum of the distances to each center point EI0, ​​BI0, and PI0 is calculated. The formula for calculating EI for environmental information is as follows: And x i = (Eit, Eih, Eig, Eil, Ein), where i represents the number of feature points of the environmental information, x i Let d1(x) be the i-th feature value of data point x. i (EI0) 2 Represented as x i Assign it to the cluster nearest to the environmental information EI;

[0066] The formula for calculating behavioral information (BI) is as follows: And x k= (Bia, Bis, Bie, Bip, Bic), where k represents the number of feature points of behavioral information, x k Let d2(x) be the k-th feature value of data point x. k ,BI0) 2 Represented as x k Assign it to the cluster closest to the behavioral information BI;

[0067] The formula for calculating the physiological information PI is as follows: And x m = (Piw, Pit, Pir, Pih, Pin), where m represents the number of feature points of physiological information, x m Let d3(x) be the m-th feature value of data point x. m ,PI0) 2 Represented as x m Assign it to the cluster closest to the physiological information PI;

[0068] For each cluster, the centroid is recalculated and updated using the average value algorithm. The calculation formula is as follows: In the formula, EI1, BI1, and PI1 represent the updated center points, respectively;

[0069] After iteratively calculating the operation of updating the centroids several times, when the cluster assignments no longer change, the minimum sum of squared errors of all clusters is obtained, denoted as SSE, and the formula is: In the formula, SSE=d(x i,k,m EI1, BI1, and PI1 represent data points x, respectively. i x k x m The distances to the corresponding update center points EI1, BI1, and PI1.

[0070] Using the sum of squared errors for each cluster in the K-means clustering algorithm, association rule mining is performed to generate association influence coefficients, which are then denoted as Cic. The calculation formula is as follows: In the formula, ρ represents the strength of the correlation among environmental information, behavioral information, and physiological information, and d1(x i ,EI0),d2(x k ,BI0), d3(x m , PI0) represent data points x respectively i x k x m The distances to each corresponding center point EI0, ​​BI0, PI0.

[0071] Comparative analysis module: Collects animal behavioral characteristic data, establishes neural network models for comparative calculations, and evaluates the degree of influence of environmental, behavioral and physiological information on experimental animal behavior, obtains the influence concealment coefficient, and is used to identify and judge behavioral changes caused by environmental, behavioral and physiological information;

[0072] It should be noted that the processing steps of the neural network model are as follows:

[0073] A recurrent neural network (RNN) model is constructed by dividing the collected animal behavior feature data into training and testing sets and inputting them into the RNN input layer, so that the number of input nodes matches the number of selected animal behavior features.

[0074] The predicted output of the neural network is calculated through forward propagation, and a loss function based on mean squared error is defined to measure the difference between the predicted output and the actual label. In the forward propagation calculation, an initial weight matrix w is first set. EI w BI w PI , w z and initialize the hidden state h0;

[0075] For each time step t, the input feature EI of the current time step is... t BI t PI t The hidden state h of the previous time step t-1 As input, to compute the hidden state h at the current time step t. t ,and In the formula, G represents the activation function, and · represents matrix multiplication;

[0076] Then calculate the predicted output z based on the hidden state. t The formula for calculating the prediction result is z. t =G(w z ×h t );

[0077] The loss function for calculating the mean squared error (MSE) is defined as follows: In the formula, 15 represents the total number of samples in the original data, z (EI、BI、PI) Let y represent the predicted output for each sample in the environmental, behavioral, and physiological information categories, respectively. (EI、BI、PI) These are respectively represented as the actual labels corresponding to environmental information, behavioral information, and physiological information;

[0078] Based on the predicted output z tThe gradient is calculated using the backpropagation algorithm, and the network parameters are updated using the stochastic gradient descent optimization algorithm to minimize the loss function. This is achieved by introducing a learning rate η that controls the step size of the parameter updates, and the network parameters w to be updated. gradient of loss function Perform backpropagation, calculate the effect of the gradient on each parameter, and use the gradient descent formula. In the formula, w′ represents the updated network parameters, and the network parameters are updated.

[0079] It should be noted that the calculation logic affecting the concealment coefficient is as follows:

[0080] The test set is predicted using a trained recurrent neural network (RNN) model, and the prediction result z is obtained. t Based on feature importance analysis, the weight of each feature in the RNN model prediction is evaluated to calculate the influence concealment coefficient of each feature. This coefficient is calculated by subtracting all combinations BC′ that form a subset of the target feature from each feature data BC.

[0081] For each combination of target feature subsets, calculate the predicted difference value ΔY of the corresponding output of the RNN model, i.e., ΔY = z t (BC)-z t (BC′);

[0082] For each target feature subset, calculate the total number of combinations ΔX of target feature subsets of different sizes, i.e., ΔX = m! (15! - m! - 1), where 15 represents the total number of feature data, m represents the size of the target feature subset, and ! represents the factorial;

[0083] Calculate the concealment coefficient φ for each feature. i ,Right now and 1≤i≤15, where BC i Let be the i-th feature in the target feature subset, and BC i ∈BC 1-15 BC 1-15 = {Eit, Eih, Eig, Eig, Ein, Bia, Bis, Bie, Bip, Bic, Piw, Pit, Pir, Pih, Pin}, where S represents the feature subset and i represents the number of target features, used to assess the specific impact of each environmental, behavioral, and physiological information on the behavior of experimental animals.

[0084] Abnormal Behavior Diagnosis Module: Through predefined abnormal behavior parameters and learned behavior patterns, it analyzes the received concealment coefficient to identify and diagnose any abnormal behavior, and reports the diagnosis results to the comprehensive monitoring and control module to warn clinical researchers of any potential health problems in laboratory animals and behavioral changes that require extra attention.

[0085] It should be noted that the abnormal behavior diagnosis module includes a monitoring model. Analysis is performed using this model to identify any abnormal behavior. The monitoring steps of the model are as follows:

[0086] Based on the specific laboratory animal housing setting and animal behavioral characteristics, multiple parameters of abnormal behavior are defined in advance, including unusual movement patterns or specific sound frequencies;

[0087] Based on the defined abnormal behavior parameters, relevant features are extracted from the data processed by the correlation analysis module and the comparative analysis module.

[0088] The extracted relevant features are compared and matched with predefined abnormal behavior parameters. If the predefined rules for abnormal behavior are met, the behavior is determined to be abnormal.

[0089] Integrated monitoring and control module: Integrates the outputs of various modules within the experimental animal behavior monitoring system, presents the data in a visual form, and automatically adjusts monitoring parameters and issues alarms based on the standards set in the breeding room and the results of data analysis, providing researchers in animal experiments with complete and real-time animal behavior monitoring results.

[0090] like Figure 2 As shown, this invention provides a laboratory animal behavior monitoring system for a breeding room, and the monitoring method includes the following steps:

[0091] S1. Install environmental sensors, visual cameras, and vital sign detectors to capture the behavior and activities of laboratory animals and obtain raw data on the behavior of laboratory animals.

[0092] S2. Perform preprocessing and normalization on the raw data to generate animal behavioral characteristic data that is easy to process and calculate later.

[0093] S3. Use the K-means clustering algorithm to perform association analysis on animal behavioral characteristic data to discover the association patterns between different behaviors;

[0094] S4. Use neural network comparative analysis to identify and judge behavioral changes caused by environmental information, behavioral information and physiological information, so as to detect changes and abnormalities in animal behavior;

[0095] S5. Diagnose abnormal behavioral changes through monitoring methods, reversely assess the causes of abnormal behavior, and adjust environmental parameters and conduct human intervention on experimental animals based on the results of data visualization and report generation from the comprehensive monitoring and control module.

[0096] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0097] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0098] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0099] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0100] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A monitoring system for the behavior of laboratory animals in a feeding room, characterized in that, Includes a behavior information collection module: using monitoring equipment to collect raw data on the behavior of experimental animals in the feeding room in real time, and then transmitting the raw data to the data processing and classification module; Data processing and classification module: Receives raw data transmitted from the behavior information collection module, performs calculations, classifications, statistics, and storage using processing and classification algorithms, and processes the raw data to form animal behavior characteristic data, which is then transmitted to the association analysis module and the comparison analysis module respectively. Association Analysis Module: By establishing a K-means clustering algorithm model and then processing the animal behavior characteristic data classified by the data processing and classification module, association influence coefficients are generated to discover the association patterns between animal behaviors and to reveal non-intuitive behavioral patterns. The comparative analysis module collects animal behavioral characteristic data, establishes a neural network model for comparative calculations, and evaluates the degree of influence of environmental, behavioral, and physiological information on experimental animal behavior, obtaining an influence concealment coefficient to identify and determine behavioral changes caused by environmental, behavioral, and physiological information. The processing steps of the neural network model are as follows: A recurrent neural network (RNN) model was constructed, and the collected animal behavioral feature data was divided into a training set and a test set. The animal behavioral feature data included environmental information. Behavioral information and physiological information The input is fed into the RNN input layer of the recurrent neural network, so that the number of input nodes matches the number of selected animal behavioral features. The predicted output of the neural network is calculated through forward propagation, and a loss function based on mean squared error is defined to measure the difference between the predicted output and the actual label. In the forward propagation calculation, an initial weight matrix is ​​first set. and initializing hidden state ; For each time step The input features at the current time step The hidden state of the previous time step As input, to calculate the current time step Hidden state ,and In the formula, Represented as an activation function, Represented as matrix multiplication; Then calculate the predicted output based on the hidden state. The formula for calculating the prediction result is: ; Abnormal Behavior Diagnosis Module: Based on predefined abnormal behavior parameters and learned behavior patterns, the module analyzes the received concealment coefficients to identify and diagnose any abnormal behavior and reports the diagnosis results to the comprehensive monitoring and control module. Integrated monitoring and control module: Integrates the outputs of various modules within the experimental animal behavior monitoring system, presents the data in a visual form, and automatically adjusts monitoring parameters and issues alarms based on the standards set in the breeding room and the results of data analysis, providing researchers in animal experiments with complete and real-time animal behavior monitoring results.

2. The experimental animal behavior monitoring system for a feeding room according to claim 1, characterized in that, The raw data includes environmental information about the animals' living conditions in the laboratory, calibrated as... Behavioral information and labeling during exercise And physiological information of vital signs, denoted as Among them, environmental information This includes temperature, humidity, gas concentration, light intensity, and noise, and each is calibrated as follows: , , , and Behavioral information This includes activity level, social status, eating habits, sleep patterns, and cleaning and maintenance behaviors, each labeled as follows: , , , and Physiological information This includes weight, body temperature, respiratory rate, heart rate, and neuropsychological state, each labeled as follows: , , , and .

3. The experimental animal behavior monitoring system in a feeding room according to claim 2, characterized in that, The steps for obtaining the classification algorithm are as follows: Supervised learning algorithms were used to train the experimental animals in the breeding room on raw data to obtain animal behavioral characteristic data; For classification, a nonlinear support vector machine is used, introducing a kernel function to map the sample parameters in the original data to a high-dimensional eigenspace for linear separability. The calculation formula is as follows: In the formula, The direction is represented by the hyperplane. Represented as the function value of the original data in the eigenspace. Represented as the feature vector after mapping the original data. Represented as a deviation constant; The formula for minimizing the objective function is: In the formula, These are respectively represented by class labels for environmental information, behavioral information, and physiological information. Represented as training samples; The algorithm is introduced to learn and solve the optimal classification function, and the calculation formula is as follows: In the formula, Represented as the inner product of the sample and label in the feature space, the solution is obtained as follows: In the formula, This represents the optimal classification value of the extracted features from the original data, in order to obtain accurate animal behavioral feature data. In the formula, It represents a collection of animal behavioral characteristic data.

4. The experimental animal behavior monitoring system for a feeding room according to claim 3, characterized in that, The animal behavioral characteristic data is obtained by performing data denoising and cleaning, data normalization, feature extraction, and data classification on the raw data. The calculation formula for data normalization is as follows: In the formula, These are respectively represented as environmental information. Behavioral information and physiological information The data after normalization processing These are respectively represented as environmental information. The minimum and maximum values, These are respectively represented as behavioral information. The minimum and maximum values, Represented as physiological information The minimum and maximum values.

5. The experimental animal behavior monitoring system for a feeding room according to claim 1, characterized in that, The steps for establishing the K-means clustering algorithm model are as follows: Three data points were selected: environmental information, behavioral information, and physiological information. As the initial center point; Using the Euclidean distance formula, for each point sample in the animal behavioral characteristic data, the distance to each center point is calculated separately. The distance and sum, then for environmental information The calculation formula is as follows: ,and In the formula, This represents the number of feature points in the environmental information. Represented as data points The 1 eigenvalue, Indicated as Assigned to environmental information In the nearest cluster; For behavioral information The calculation formula is as follows: ,and In the formula, This is represented by the number of feature points for behavioral information. Represented as data points The 1 eigenvalue, Indicated as Assigned to behavioral information In the nearest cluster; For physiological information The calculation formula is as follows: ,and In the formula, The number of feature points representing physiological information. Represented as data points The 1 eigenvalue, Indicated as Assigned to physiological information In the nearest cluster.

6. The experimental animal behavior monitoring system for a feeding room according to claim 5, characterized in that, For each cluster, the centroid is recalculated and updated using the average value algorithm. The calculation formula is as follows: , , In the formula, These are respectively represented as the updated center points; After iteratively calculating several times the operation of updating the centroids, when the cluster assignments no longer change, the result is the minimized sum of squared errors for all clusters, calibrated as follows: The formula is In the formula, They are represented as data points. and corresponding update center point The distance; Using the sum of squared errors for each cluster in the K-means clustering algorithm, association rule mining is performed to generate association influence coefficients, which are then calibrated as... The calculation formula is: In the formula, This represents the strength of the correlation between environmental information, behavioral information, and physiological information. , , They are represented as data points. and corresponding to each center point The distance.

7. The experimental animal behavior monitoring system for a feeding room according to claim 1, characterized in that, The loss function for calculating the mean squared error (MSE) is defined as follows: In the formula, 15 represents the total number of samples in the original data. These represent the predicted outputs for each sample in the environmental, behavioral, and physiological information categories, respectively. These are respectively represented as the actual labels corresponding to environmental information, behavioral information, and physiological information; Based on the predicted output The gradient is calculated using the backpropagation algorithm, and the network parameters are updated using the stochastic gradient descent optimization algorithm to minimize the loss function. The learning rate is then introduced to control the step size of the parameter updates. and network parameters to be updated ,and The gradient of the loss function Perform backpropagation, calculate the effect of the gradient on each parameter, and use the gradient descent formula. In the formula, This represents the updated network parameters, and the network parameters are updated accordingly.

8. The experimental animal behavior monitoring system for a feeding room according to claim 1, characterized in that, The calculation logic for the influencing concealment coefficient is as follows: The test set is predicted using a trained recurrent neural network (RNN) model, and the prediction results are obtained. Based on feature importance analysis, the weight of each feature in the RNN model prediction is evaluated to calculate the influence concealment coefficient of each feature, i.e., for each feature data... Subtract all combinations that form a subset of target features from themselves. ,and ; For each combination of target feature subsets, calculate the predicted difference value of the corresponding output of the RNN model. ,Right now ; For each subset of target features, calculate the total number of combinations of target feature subsets of different sizes. ,Right now In the formula, 15 represents the total number of feature data. Represented as the size of the target feature subset. Represented as factorial; Calculate the concealment coefficient for each feature. ,Right now ,and In the formula, Represented as the first in the target feature subset One feature, and , , Represented as a feature subset, This is represented by the number of target features, used to assess the specific impact of each environmental, behavioral, and physiological feature on the behavior of experimental animals.

9. The experimental animal behavior monitoring system for a feeding room according to claim 1, characterized in that, The steps of the monitoring method are as follows: S1. Install cameras and smart sensor devices to capture the behavior and activities of laboratory animals and obtain raw data on the behavior of laboratory animals. S2. Perform preprocessing and normalization on the raw data to generate animal behavioral characteristic data that is easy to process and calculate later. S3. Use the K-means clustering algorithm to perform association analysis on animal behavioral characteristic data to discover the association patterns between different behaviors; S4. Use neural network comparative analysis to identify and judge behavioral changes caused by environmental information, behavioral information and physiological information, so as to detect changes and abnormalities in animal behavior; S5. Diagnose abnormal behavioral changes through monitoring methods, reversely assess the causes of abnormal behavior, and adjust environmental parameters and conduct human intervention on experimental animals based on the results of data visualization and report generation from the comprehensive monitoring and control module.

Images