A monitoring device and method for the postoperative urination state of experimental animals

Abstract

The present application provides a monitoring device and method for the postoperative urination state of experimental animals, relating to the technical field of animal urination monitoring, including: a urination prediction module for collecting the dietary characteristic information of animals before the experiment and performing urination prediction; a urine volume acquisition module for performing urination monitoring after the experiment; a deviation analysis module for obtaining the first postoperative dysuria probability according to the deviation between the predicted urination information and the total urine volume information; a difficulty identification module for collecting the activity images of animals, performing dysuria identification, and combining with the first postoperative dysuria probability to calculate the postoperative dysuria probability as the monitoring result of the postoperative urination state. Through the present application, the technical problem in the prior art that the accuracy of postoperative urination monitoring is low due to the fact that urination monitoring usually relies on a single data source can be solved. By analyzing the dietary characteristic information and urine volume of experimental animals and combining with motion analysis verification, the accuracy of monitoring the postoperative urination state of experimental animals is improved.

Description

Technical Field

[0001] This application relates to the technical field of animal urination monitoring, and particularly to a monitoring device and method for the postoperative urination state of experimental animals. Background Art

[0002] With the in-depth development of biomedical research, experimental animals play an increasingly important role in scientific experiments. Monitoring the postoperative state of experimental animals, especially the urination behavior, is of great significance for evaluating the surgical effect, detecting possible complications, and improving the welfare of experimental animals. In the research on monitoring the postoperative state of experimental animals, the urination behavior is considered one of the important indicators for judging its postoperative recovery. Abnormal urination after surgery, such as difficult urination or insufficient urine volume, may indicate problems such as pain and impaired organ function in animals, and may even reflect some potential physiological or metabolic abnormalities. Many existing animal health monitoring systems mainly rely on a single data source to record the urine volume and frequency through weighing or visual observation. However, the single urine volume data is easily affected by environmental factors, diet changes, etc. For example, urine evaporation may lead to measurement errors, especially under experimental conditions with a large time span.

[0003] In summary, there is a technical problem in the prior art that since urination monitoring usually relies on a single data source and is easily affected by environmental factors, it leads to measurement errors, further affecting the accuracy of postoperative urination monitoring. Summary of the Invention

[0004] The purpose of this application is to provide a monitoring device and method for the postoperative urination state of experimental animals to solve the technical problem in the prior art that since urination monitoring usually relies on a single data source and is easily affected by environmental factors, it leads to measurement errors, further affecting the accuracy of postoperative urination monitoring.

[0005] In view of the above problems, this application provides a monitoring device and method for the postoperative urination state of experimental animals.

[0006] In a first aspect, the present application provides a monitoring device for the postoperative urination state of experimental animals. Among them, the monitoring device for the postoperative urination state of experimental animals includes: a urination prediction module, which is used to collect the dietary characteristic information of the experimental animals within the first preset time range before the experiment, perform urination prediction, and obtain predicted urination information; a urine volume acquisition module, which is used to monitor the urination of the experimental animals within the second preset time range after the experiment, and obtain the total urine volume information, where the total urine volume information is obtained through evaporation compensation; a deviation analysis module, which is used to classify and obtain the first postoperative urination difficulty probability according to the deviation between the predicted urination information and the total urine volume information; a difficulty identification module, which is used to collect the animal activity images of the experimental animals within the second preset time range, perform urination difficulty identification, obtain the second postoperative urination difficulty probability, and calculate the postoperative urination difficulty probability in combination with the first postoperative urination difficulty probability as the monitoring result of the postoperative urination state of the experimental animals.

[0007] In a second aspect, the present application further provides a method for monitoring the postoperative urination state of experimental animals. Among them, the method for monitoring the postoperative urination state of experimental animals includes: collecting the dietary characteristic information of the experimental animals within the first preset time range before the experiment, performing urination prediction, and obtaining predicted urination information; monitoring the urination of the experimental animals within the second preset time range after the experiment, and obtaining the total urine volume information, where the total urine volume information is obtained through evaporation compensation; classifying and obtaining the first postoperative urination difficulty probability according to the deviation between the predicted urination information and the total urine volume information; collecting the animal activity images of the experimental animals within the second preset time range, performing urination difficulty identification, obtaining the second postoperative urination difficulty probability, and calculating the postoperative urination difficulty probability in combination with the first postoperative urination difficulty probability as the monitoring result of the postoperative urination state of the experimental animals.

[0008] One or more technical solutions provided in the present application have at least the following technical effects or advantages:

[0009] Through a urination prediction module, the urination prediction module is used to collect dietary characteristic information of an experimental animal within a first preset time range before the experiment, perform urination prediction, and obtain predicted urination information; a urine volume acquisition module, the urine volume acquisition module is used to monitor the urination of the experimental animal within a second preset time range after the experiment to obtain total urine volume information, wherein the total urine volume information is obtained through evaporation compensation; a deviation analysis module, the deviation analysis module is used to classify and obtain the first postoperative dysuria probability according to the deviation between the predicted urination information and the total urine volume information; a difficulty identification module, the difficulty identification module is used to collect animal activity images of the experimental animal within the second preset time range, perform dysuria identification, obtain the second postoperative dysuria probability, and calculate the postoperative dysuria probability in combination with the first postoperative dysuria probability as the monitoring result of the postoperative urination state of the experimental animal. That is to say, by collecting the dietary characteristics of the animal before the experiment to predict the urination information, monitoring the urination of the animal after the experiment, and calculating the total urine volume information; comparing the predicted urination information with the total urine volume information for deviation to calculate the first postoperative dysuria probability; collecting the activity images of the animal to identify the behavioral characteristics of dysuria and calculating the second postoperative dysuria probability; combining the first postoperative dysuria probability and the second postoperative dysuria probability to obtain the postoperative dysuria probability, accurately evaluating the postoperative urination state of the experimental animal, improving the accuracy of monitoring postoperative dysuria in experimental animals, and reducing the risk of misjudgment.

[0010] The above description is only an overview of the technical solution of the present application. In order to be able to understand the technical means of the present application more clearly, it can be implemented according to the content of the specification. And in order to make the above and other purposes, features and advantages of the present application more obvious and understandable, the specific embodiments of the present application are specifically exemplified below. It should be understood that the content described in this part is not intended to identify the key or important features of the embodiments of the present application, nor is it used to limit the scope of the present application. Other features of the present application will become easily understandable through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the technical solutions in the present application or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings described below are only exemplary, and for those of ordinary skill in the art, other drawings can be obtained according to the provided drawings without creative efforts.

[0012] Figure 1 It is a schematic structural diagram of a device for monitoring the postoperative urination state of an experimental animal according to the present application;

[0013] Figure 2 It is a schematic flow diagram of a method for monitoring the postoperative urination state of an experimental animal according to the present application.

[0014] Explanation of reference numerals: Urination prediction module 11, urine volume acquisition module 12, deviation analysis module 13, difficulty recognition module 14. Specific implementation manners

[0015] This application provides a monitoring device and method for the postoperative urination state of experimental animals, which solves the technical problem in the prior art that due to the fact that urination monitoring usually relies on a single data source and is easily affected by environmental factors, resulting in measurement errors and further affecting the accuracy of postoperative urination monitoring. By collecting the dietary characteristics of animals before the experiment to predict urination information, after the experiment, the urination of the animals is monitored to calculate the total urine volume information; the predicted urination information and the total urine volume information are compared for deviation to calculate the first postoperative urination difficulty probability; the activity images of the animals are collected to identify the behavioral characteristics of urination difficulty and calculate the second postoperative urination difficulty probability; the postoperative urination difficulty probability is obtained by combining the first postoperative urination difficulty probability and the second postoperative urination difficulty probability, accurately evaluating the postoperative urination state of experimental animals, improving the accuracy of monitoring postoperative urination difficulty of experimental animals, and reducing the risk of misjudgment.

[0016] Next, the technical solutions in this application will be described clearly and completely with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all the embodiments of this application. It should be understood that this application is not limited by the example embodiments described herein. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of this application without creative efforts shall fall within the scope of protection of this application. Additionally, it should be noted that for the sake of description, only the parts related to this application are shown in the accompanying drawings rather than all of them.

[0017] Embodiment 1. Please refer to the attached Figure 1 , this application provides a monitoring device for the postoperative urination state of experimental animals. Among them, the monitoring device for the postoperative urination state of experimental animals is used to implement the steps of a monitoring method for the postoperative urination state of experimental animals. The monitoring device for the postoperative urination state of experimental animals includes:

[0018] A urination prediction module 11, which is used to collect the dietary characteristic information of experimental animals within the first preset time range before the experiment, perform urination prediction, and obtain predicted urination information.

[0019] Specifically, through preliminary experiments or historical data, collect the diet and urination monitoring data of experimental animals, including the diet characteristics of the animals and the corresponding urination data, where the diet and urination are at the same time point. Extract a series of sample diet characteristic information sets and sample urination information sets. Use the diet characteristic information set and urination information set in the sample data to train through machine learning algorithms (such as linear regression models or neural network models) to find the law between diet characteristics and urine output. After training, the urine output prediction path can predict the urine output based on the input diet characteristics.

[0020] Before the experiment, collect the diet characteristic information of the experimental animals within the first preset time range, including the daily food intake and water intake of the experimental animals. The first preset time range is a fixed time period, which is a time window set according to actual needs for data collection before the experimental animals undergo the experiment, such as 12 hours, 24 hours, 48 hours or longer. Input the collected diet characteristic information of the experimental animals into the trained urine output prediction path for urine output prediction to obtain the predicted urination information, including the predicted number of urinations and urine output of the experimental animals within the first preset time range before the experiment. By collecting the diet characteristic information of the experimental animals before the experiment, performing urine output prediction, and obtaining the predicted urination information, the urination behavior of the experimental animals can be predicted in advance while reducing unnecessary intervention.

[0021] The urine output acquisition module 12, where the urine output acquisition module 12 is used to monitor the urination of the experimental animals within the second preset time range after the experiment to obtain the total urine output information, where the total urine output information is obtained through evaporation compensation.

[0022] Specifically, within the second preset time range after the experiment, continuously monitor the experimental animals and collect relevant urination data, such as the number of urinations, time, and urine output. Through a urine output meter, a sensor installed at the bottom of the cage, or other automated devices, record the time, urine output, and other information of urination events in real time. The second preset time range is similar to the first preset time range, both are time windows for data collection set according to requirements, and can be 1 hour, 5 hours, 24 hours, etc. after the operation. During the experimental monitoring, the urine is usually exposed to the air and may evaporate due to environmental factors (such as temperature, humidity). To compensate for the error caused by evaporation, it is necessary to correct the total urine output according to the evaporation ratio. Divide the measured urine output by the value of 1 - evaporation ratio to obtain the compensated urine output. The evaporation ratio is different for different time periods, and it is necessary to perform evaporation compensation on the urine output information according to the urination time. Add up all the urine outputs after evaporation compensation to obtain the total urine output of the experimental animal within this time range. Through evaporation compensation, eliminate the influence of environmental factors (such as temperature, humidity) on urine output measurement during the experiment to ensure that the final urine output data is more accurate.

[0023] A deviation analysis module 13, which is configured to classify and obtain a first postoperative dysuria probability according to the deviation between the predicted urination information and the total urination volume information.

[0024] Specifically, calculate the deviation between the predicted urination information and the total urination volume information. By calculating the difference between the predicted urination information and the total urination volume information, and then calculating the ratio of the difference to the predicted urination information, the urination volume deviation is obtained, which quantifies the difference between the predicted value and the actual value. Collect data on the urination volume deviation of multiple samples and the actual postoperative dysuria occurrence from historical experimental data to obtain a sample urination volume deviation set and a sample first postoperative dysuria probability set, and train a dysuria classifier. Each sample contains a urination volume deviation value and the corresponding probability of postoperative dysuria. By training these data and using machine learning algorithms such as decision trees, support vector machines (SVMs), or neural networks, etc., the classifier will learn the relationship between the urination volume deviation and the postoperative dysuria probability. For example, if the deviation is less than a certain threshold, it indicates that the probability of postoperative dysuria is relatively low; if the deviation is large, the probability of postoperative dysuria is relatively high. Once the classifier is trained, input the calculated urination volume deviation value into the classifier, index the sample closest to the urination volume deviation, and output the corresponding dysuria probability, that is, the first postoperative dysuria probability. By calculating the deviation between the predicted urination information and the actual urination volume, the prediction accuracy is quantified, and the deviation value is used as a key feature and input into the classifier, so as to more accurately predict the risk of postoperative dysuria. If the dysuria probability of a certain animal is relatively high, measures can be taken in advance to avoid the aggravation of the condition and promote better postoperative recovery.

[0025] A difficulty identification module 14, which is configured to collect animal activity images of the experimental animal within the second preset time range, perform dysuria identification to obtain a second postoperative dysuria probability, and combine the first postoperative dysuria probability to calculate and obtain the postoperative dysuria probability, which is used as the monitoring result of the postoperative urination state of the experimental animal.

[0026] Specifically, when an animal has difficulty urinating, corresponding behavioral manifestations will occur. The activity images of the experimental animal within the second preset time range are collected through a video monitoring device or a camera, that is, the activity image sequence, which reflects the behavioral characteristics of the animal during the postoperative recovery process, including whether it shows behaviors of difficulty urinating, such as frequent urination attempts, abnormal urination postures, etc. Similarly, according to the detection record data of the difficulty urinating of the experimental animal, multiple sample activity image sequences, that is, different behavioral manifestations, are obtained. At the same time, the probability of the experimental animal having difficulty urinating corresponding to different sample activity image sequences is collected to obtain the second postoperative difficulty urinating probability set, which includes the probability of difficulty urinating corresponding to each activity image sequence. According to the sample activity image sequence set and the sample second postoperative difficulty urinating probability set, an activity difficulty urinating recognizer is trained using a convolutional neural network. After the collected activity image sequence is input into the activity difficulty urinating recognizer, the second postoperative difficulty urinating probability is obtained.

[0027] Obtain the urination time of the experimental animal for urination monitoring within the second preset time range. According to the ratio of the urination time to the second preset time range, the time weight is calculated. The image weight is obtained by using 1 - the time weight. The first postoperative difficulty urinating probability (the probability based on the analysis of the urine volume deviation) and the second postoperative difficulty urinating probability (the probability based on the recognition of the activity image) are weighted and combined to calculate the final postoperative difficulty urinating probability. The time weight reflects the importance of the urine volume data, while the image weight reflects the importance of the behavioral image in the prediction. The calculated postoperative difficulty urinating probability is used as the monitoring result of the postoperative urination state of the animal, providing a clear value to the experimenter for judging whether further nursing measures or interventions are needed. By combining the analysis results of the urine volume data and the activity image data, the actual difficulty urinating situation of the experimental animal is comprehensively reflected, making up for the information deficiency that may be brought by a single data source. The weighted fusion of the physical monitoring data (urine volume) and the behavioral characteristic data (activity image) can more accurately judge the probability of postoperative difficulty urinating, thus improving the reliability of the prediction.

[0028] Furthermore, the urination prediction module 11 in the above-mentioned experimental animal postoperative urination state monitoring device is further used for:

[0029] Collect the diet characteristic information of the experimental animal within the first preset time range before the experiment; according to the diet and urination monitoring data of the experimental animal, collect the sample diet characteristic information set and the sample urination information set; use the sample diet characteristic information set and the sample urination information set to train the urination prediction path; input the diet characteristic information into the urination prediction path, and predict and output to obtain the predicted urination information.

[0030] Specifically, within the first preset time range before the experiment, the dietary characteristic information of the experimental animals is recorded. The first preset time range is a preset time period, which is a time window set before the experiment on the experimental animals for data collection. For example, it could be 12 hours or 24 hours before the experiment. The dietary characteristic information refers to the dietary-related data of the experimental animals before the experiment, including daily food intake and water intake, such as the amount of food eaten, the number of eating times, the amount of water intake, the types of food, the intake time, etc. For example, troughs and water bottles equipped with weight sensors are used to record the food intake and the number of water-drinking times of mice within 12 hours before the experiment, as well as the time points of each food intake and water intake.

[0031] Based on the known dietary and urination monitoring data of the experimental animals, a certain number of sample data are selected to obtain a sample dietary characteristic information set and a sample urination information set, which respectively record the dietary characteristics of the experimental animals and the corresponding urination results. Through training with machine learning algorithms (such as linear regression models or neural network models), the relationship between dietary characteristics and urine output is found. First, the sample dietary characteristic information set and the sample urination information set are preprocessed, including steps such as data cleaning, standardization, and feature extraction. According to the task requirements, a suitable machine learning algorithm is selected, such as linear regression, decision tree, or neural network. To predict the urine output, a simple linear regression model can be used first to observe the linear relationship between dietary characteristics and urine output. If the linear relationship is weak, a more complex model may be needed, such as a random forest or a neural network.

[0032] The preprocessed sample dietary characteristics and urination information are used as the input and output data sets, and a neural network is selected for model training. The number of nodes in the input layer is equal to the number of dietary characteristics, and the number of nodes in the input layer is determined according to the collected dietary characteristics. Depending on the data complexity, 1 to 3 hidden layers are selected, and each layer contains a certain number of neurons. The output layer contains one node for outputting the predicted urine output. The input layer receives the standardized dietary characteristic data and passes the characteristic data layer by layer to the output layer to generate the predicted urine output. The neurons in each layer act through weights and biases, enabling the neural network to combine different characteristics and generate an output. The mean squared error (MSE) is used as the loss function to measure the difference between the predicted urine output and the actual urine output, obtaining the loss function. Through the backpropagation algorithm, the partial derivatives of the loss function with respect to each neuron parameter (weights and biases) are calculated to determine the change trend of the loss function, and an optimization algorithm (such as the Adam optimizer) is used to update the weights and biases, enabling the neural network to gradually reduce the prediction error. The neural network is trained through multiple rounds of iteration, and the model parameters are updated once on all training samples in each round of iteration.

[0033] Input the sample dataset (test set) that was not involved in the training into the trained neural network model to evaluate the generalization ability of the model. Calculate the MSE or mean absolute error using the test set data to measure the prediction accuracy of the model. If the prediction error is large, optimize the model by adjusting the neural network architecture (such as adding hidden layers, increasing the number of neurons), increasing the number of training times, adjusting the learning rate, etc., and finally obtain the urination prediction path. Input the collected dietary feature data into the trained urination prediction path, and the model will output the predicted urine volume. Through the training sample data, a reliable mapping path between dietary features and urine volume is established, enabling the dietary situation of the experimental animals before the experiment to be used to accurately predict the urine volume after the operation and identify the risk of abnormal urination.

[0034] Further, the urine volume obtaining module 12 in the postoperative urination state monitoring device for experimental animals is further configured to:

[0035] Within a second preset time range after the end of the experiment, monitor the urination of the experimental animal to obtain urine volume information, and collect the urination time; according to the urination time, perform evaporation compensation on the urine volume information to obtain total urine volume information.

[0036] Specifically, within a second preset time range after the end of the experiment, use a urination monitoring device to record the urine volume and urination time of the experimental animal. For example, the urine volume of each experimental animal can be automatically recorded by an electronic scale placed at the bottom of the cage, and the weighing device can accurately detect the weight of each urination and synchronously record the time of each urination. The anesthesia during the surgical experiment may cause the animal to have difficulty urinating. Within the second preset time range after the operation, such as within 1 hour, monitor the urination. Obtain the final urine volume and the time of the last urination. For example, the urine volume at 45 minutes after the operation is 20 milliliters.

[0037] Evaporation compensation is performed on the urine volume based on the urination time. Since urine will evaporate within a certain period after being discharged, the actual urine volume is less than the observed urine volume. Evaporation compensation refers to adjusting the urine volume information according to the urination time to compensate for part of the loss caused by urine evaporation. The urination time and the corresponding evaporation ratio in historical test data are collected, and an evaporation ratio classifier is trained through a machine learning model (such as a decision tree, a random forest, or a neural network). The urination time is input into the evaporation ratio classifier to map the corresponding evaporation ratio. According to the evaporation ratio, evaporation compensation is performed on the urine volume information to obtain the total urine volume information. For those with a short monitoring time, evaporation compensation can be performed based on the final urine volume. If the monitoring time is long, the evaporation amount may vary due to temperature, humidity, and the exposure time of urine. It is necessary to classify the urine volume according to time periods and perform corresponding evaporation compensation, and then sum up the urine volumes after evaporation compensation for each time to obtain the total urine volume within the second preset time range. Through the evaporation compensation method, the total urine volume excreted by the experimental animal after the experiment can be estimated more accurately, the error caused by evaporation is reduced, the experimental data is made more real and reliable, the accuracy of urine volume monitoring is improved, the underestimation caused by evaporation is avoided, and it helps to detect urination problems in time and intervene.

[0038] Further, the urine volume obtaining module 12 in the postoperative urination state monitoring device for experimental animals is further configured to:

[0039] Map and obtain the evaporation ratio according to the urination time. Among them, according to the urination monitoring data of animal experiments, a sample urination time set and a sample evaporation ratio set are collected, and an evaporation ratio classifier is constructed for mapping classification; the evaporation ratio is used to perform evaporation compensation on the urine volume information to obtain the total urine volume information.

[0040] Specifically, since urine evaporates over time after being excreted, it is necessary to calculate the evaporation ratio based on the urination time to more accurately estimate the actual urine volume. A large amount of experimental animal urination time and corresponding evaporation ratio data are collected from existing animal experiment urination monitoring data as a sample urination time set and a sample evaporation ratio set. Among them, the sample urination time set includes the urination data of experimental animals at different time points, while the sample evaporation ratio set includes the evaporation ratio of urine at the corresponding time points. The collected sample urination time set and sample evaporation ratio set are used to train an evaporation ratio classifier, such as training through algorithms like decision trees and support vector machines (SVM). The input of the classifier is the urination time, and the output is the evaporation ratio. A decision tree maps the input data to the output category through a series of rules and is a very intuitive model suitable for processing time-related data and mapping relationships. A support vector machine finds a hyperplane in a high-dimensional space to maximize the boundary between different categories for classification tasks, especially suitable for the classification of high-dimensional data and quite effective in processing time data.

[0041] Data preprocessing is performed on the sample urination time set and the sample evaporation ratio set, including formatting, standardization, or normalization to ensure that the urination time and evaporation ratio are represented in a suitable format (such as numerical or categorical). Key features are extracted from the urination time set, such as the time length after urination, environmental temperature, humidity, etc. The key features of the urination time (as input features) and the corresponding evaporation ratio (as label data) are input into the decision tree model. During training, the decision tree splits nodes according to different urination times and finds the evaporation ratio corresponding to different urination time points by recursively splitting the feature space. The goal of selecting the split point is to find the feature that can best distinguish different categories (i.e., different evaporation ratios) by splitting the feature space. The selection of the split point is usually based on measuring the information contribution of a certain feature to the target variable (such as the evaporation ratio), and common criteria include information gain and Gini index. Specifically, when starting to build a decision tree, first calculate the entropy or Gini index of the entire data set. For each feature (urination time), try different split points. For each possible split point of each feature, calculate the entropy (or Gini index) of each subset after splitting, and then calculate the information gain (or Gini gain) according to the formula. Select the feature and split point with the largest information gain or Gini gain as the current split point of the decision tree. Cross-validation is used to avoid overfitting, that is, by splitting the data set (for example, 80% for training and 20% for validation) to evaluate the performance of the model on unseen data and optimize the model parameters. After completing the training and validation of the classifier, apply it to the actual urination monitoring data.

[0042] According to the recorded urination time each time, input it into the evaporation ratio classifier to map and obtain the corresponding evaporation ratio. Use the evaporation ratio to compensate the urine volume of this urination. Assume that the urine volume at 45 minutes after the operation is 20 milliliters, and the evaporation ratio corresponding to the determined urination time point by the classifier is 8%. Then the total urine volume information after compensation is 20 / (1 - 0.08) = 21.74 milliliters. Through the intelligent mapping and compensation of the evaporation ratio of different urination time points by the evaporation ratio classifier, the accuracy of urine volume monitoring is significantly improved. The classifier is trained based on a large number of sample data, can accurately map the relationship between urination time and evaporation ratio, effectively reduce the error caused by urine evaporation, and make the experimental results closer to the actual urine volume.

[0043] Furthermore, the deviation analysis module 13 in the postoperative urination state monitoring device for experimental animals is further configured to:

[0044] Calculate the difference between the predicted urination information and the total urine volume information, and calculate the ratio of the difference to the predicted urination information to obtain the urine volume deviation; according to the urination difficulty detection record data of the experimental animal, collect the set of sample urine volume deviations, and collect the probability of the animal having urination difficulty under different sample urine volume deviations to obtain the set of sample first postoperative urination difficulty probabilities; use the set of sample urine volume deviations and the set of sample first postoperative urination difficulty probabilities to construct a urination difficulty classifier; input the urine volume deviation into the urination difficulty classifier, and index and classify to obtain the first sample postoperative urination difficulty probability corresponding to the sample urine volume deviation closest to it, to obtain the first postoperative urination difficulty probability.

[0045] Specifically, calculate the difference between the predicted urination information and the total urine volume information, and calculate the ratio of this difference to the predicted urination information to obtain the urine volume deviation, which represents the deviation degree between the predicted urine volume and the actual urine volume. The urination difficulty detection record data refers to observing the urination situation of the experimental animal in the postoperative stage and recording whether it has urination difficulty. According to the urination difficulty detection record data of the experimental animal, that is, the animal shows abnormal delays, poor urination, or pain during urination, collect multiple sample urine volume deviations and the corresponding postoperative urination difficulty record data. Postoperative urination difficulty refers to that after the animal undergoes surgery, due to physiological or drug effects and other factors, the urination function is affected. The set of urine volume deviations is a set of deviation values between the predicted urine volume and the actual urine volume of the experimental animal under different experimental conditions. The set of postoperative urination difficulty probabilities is based on sample data and records the probability of the animal having urination difficulty under different deviation values. Finally, obtain the set of sample first postoperative urination difficulty probabilities, which includes the probability of the animal having urination difficulty under different urine volume deviations.

[0046] Using the collected sample urine volume deviation set and the first postoperative dysuria probability set, a dysuria classifier is constructed. By analyzing these data, the classifier learns the relationship between urine volume deviation and dysuria, and predicts the probability of an animal having dysuria after surgery given a urine volume deviation. Supervised learning algorithms such as decision trees, support vector machines (SVMs), random forests, or neural networks are used. A model is established through a training dataset, in which the input is the urine volume deviation and the output is the probability of dysuria. Taking the decision tree as an example, the urine volume deviation set is used as the input feature, and the dysuria probability set is used as the target output to train the classifier through the training set. In the decision tree model, each node represents a feature (i.e., urine volume deviation), and each branch represents a threshold judgment. During the training process, the decision tree sequentially divides the data into different regions by selecting the best feature and split point, and finally outputs the probability of dysuria. For example, assume the decision tree structure is as follows: if the urine volume deviation ≤ 0.1, the predicted dysuria probability is 50%; if the urine volume deviation > 0.1 and ≤ 0.2, the predicted dysuria probability is 70%; if the urine volume deviation > 0.2, the predicted dysuria probability is 80%.

[0047] The calculated urine volume deviation is input into the trained dysuria classifier, and the classifier will output the dysuria probability corresponding to the closest sample urine volume deviation. For example, if the urine volume deviation is 0.18, the classifier will determine that the urine volume deviation falls between 0.1 and 0.2, and thus output a predicted dysuria probability of 70%. By establishing a dysuria classifier, it is possible to accurately predict whether an animal will have dysuria based on the urine volume deviation, improve the monitoring accuracy of the postoperative urination status of experimental animals, and avoid missing important dysuria symptoms.

[0048] Furthermore, the difficulty recognition module 14 in the postoperative urination status monitoring device for experimental animals is further configured to:

[0049] Collect the animal activity images of the experimental animal within the second preset time range to obtain an activity image sequence; according to the dysuria detection record data of the experimental animal, collect a set of sample activity image sequences, and collect the probability of the experimental animal having dysuria in different sample activity image sequences to obtain a sample second postoperative dysuria probability set; use a convolutional neural network, and use the set of sample activity image sequences and the sample second postoperative dysuria probability set as supervised training data to train an activity dysuria recognizer; input the activity image sequence into the activity dysuria recognizer to identify and obtain the second postoperative dysuria probability.

[0050] Specifically, a camera or other image capture device is used to continuously collect the activity images of the experimental animals within a second preset time range (i.e., the time range corresponding to postoperative urination monitoring) for monitoring and analyzing the behavior of the experimental animals. An activity image sequence refers to a series of images arranged in chronological order for recording and analyzing the activities of the experimental animals. When the animal has difficulty urinating, corresponding action manifestations will occur, such as frequent urination attempts, abnormal urination postures, etc. The urination difficulty detection record data refers to observing the urination situation of the experimental animals in the postoperative stage and recording whether they have difficulty urinating. According to the urination difficulty detection record data of the experimental animals, a relevant set of sample activity image sequences is collected from the activity image sequence. Different activity image sequences of the experimental animals will be labeled according to the time stamps and associated with their corresponding urination difficulty detection records. Each sample image sequence corresponds to a record of urination difficulty, labeled as difficult urination or normal urination.

[0051] At the same time, the probability of the experimental animals having difficulty urinating in different sample activity image sequences is recorded. Different action manifestations may have different probabilities of urination difficulty. Summarize all the activity image sequences of the experimental animals and their corresponding urination difficulty probabilities to form a second postoperative urination difficulty probability set, which contains multiple urination difficulty probabilities of the animals in the postoperative recovery stage, and each probability corresponds to a specific set of activity image sequences. The set of sample activity image sequences and the sample second postoperative urination difficulty probability set are used as supervised training data for training the convolutional neural network model.

[0052] A convolutional neural network (CNN) is a deep learning model that is particularly good at processing image data. It extracts features from images through convolutional operations and then gradually learns and identifies complex patterns in images through multiple layers of networks. A CNN usually consists of multiple convolutional layers, pooling layers, and fully connected layers. The convolutional neural network automatically learns the features in the active images and understands which image features (such as an animal standing, walking, urination behavior, etc.) are associated with the occurrence of dysuria. During the training process, the network gradually reduces the error in predicting the probability of dysuria by adjusting its internal weights and parameters. Using a set of sample active image sequences as the input and a set of sample second postoperative dysuria probabilities as the output, the network is optimized through the backpropagation algorithm. The input set of sample active image sequences first needs to be preprocessed, including normalization or standardization, image size adjustment, denoising, etc., to ensure that the quality of the input images is suitable for training. The convolutional layer in the CNN performs convolutional operations on the input set of sample active image sequences to extract local features (such as edges, textures, etc.). The pooling layer is used to reduce the dimensionality of the data while retaining the key features of the image. The features extracted by the convolutional layer and the pooling layer are passed to the fully connected layer for further processing, and finally, the prediction result, that is, the probability of dysuria, is output. By calculating the error between the model prediction value and the true label (usually using the mean squared error or cross-entropy loss function), that is, calculating the error between the predicted probability and the corresponding sample second postoperative dysuria probability, and updating the weights of the model through an optimization algorithm (such as Adam or SGD) to reduce the error. After training, the CNN learns to identify which image features are closely related to the occurrence of dysuria.

[0053] Once the training is completed, the active dysuria identifier (i.e., the trained CNN model) can accept a new sequence of active images as input. Based on the features extracted from the images, it can identify whether an animal has dysuria and output the corresponding second postoperative dysuria probability. When the sequence of active images of the experimental animal collected within the second preset time range is input into the active dysuria identifier, the model will predict the probability of the animal having dysuria based on the behavioral features learned previously. For example, when the sequence of active images of an animal is input, the CNN model analyzes the behavioral patterns in the images, such as frequent attempts to urinate, a long urination time, etc., and finally outputs a probability value of dysuria (such as 75%), indicating that there is a 75% possibility that the animal will have dysuria in the next period. By training the CNN model to automatically analyze the sequence of active images of animals, it can identify whether an animal has dysuria without manual intervention. Combining the sequence of active images and the probability of dysuria, the model can deeply analyze the behavioral features of animals, provide a more accurate prediction of dysuria, predict whether dysuria is likely to occur after surgery according to the behavioral patterns of different animals, and thus provide personalized care and intervention plans for each animal, which helps to improve the postoperative recovery effect of experimental animals and reduce the incidence of complications.

[0054] Furthermore, the difficulty recognition module 14 in the postoperative urination status monitoring device for experimental animals is further configured to:

[0055] Obtain the urination time of the experimental animal during urination monitoring within the second preset time range; multiply the ratio of the urination time to the second preset time range by a preset time weight to obtain a time weight; use 1 minus the time weight to obtain an image weight; use the time weight and the image weight to perform weighted calculation on the first postoperative urination difficulty probability and the second postoperative urination difficulty probability to obtain the postoperative urination difficulty probability.

[0056] Specifically, within the second preset time range, monitor the urination behavior of the animal and record the start and end times of each urination. The urination time refers to the total duration of urination of the experimental animal within the second preset time range. By monitoring the urination behavior of the experimental animal and recording the start and end times of urination, the urination time of the animal can be calculated. Calculate the ratio of the urination time to the second preset time range and multiply it by the preset time weight to obtain the time weight. The preset time weight is a pre-set value, such as 0.6, which is used to adjust the importance of the urination time during the entire monitoring period. The larger the ratio of the urination time to the second preset time range, the longer the urination time. The larger the time weight, the greater the impact of the stability and persistence of the urination behavior of the experimental animal during the postoperative recovery process on the urination difficulty probability. The image weight is a weight adjusted in a complementary manner to the time weight, calculated based on the time weight, and reflects the contribution degree of the animal activity image sequence to the prediction of the urination difficulty probability. The image weight and the time weight should add up to 1.

[0057] Based on the time weight and the image weight, a weighted calculation is performed on the first postoperative dysuria probability (based on urine volume deviation) and the second postoperative dysuria probability (based on activity images) to obtain the postoperative dysuria probability. Postoperative dysuria probability = time weight × first postoperative dysuria probability + image weight × second postoperative dysuria probability. By adjusting the time weight and the image weight, the model can flexibly respond to the urination behavior characteristics of different experimental animals. For some animals, the risk of their dysuria may be more dependent on the behavior characteristics in the activity images, while for others, the urination time may be more instructive. The dynamic adjustment of the weights makes the model more adaptable and capable of handling a wider range of animal behavior patterns. For example, assume that the first postoperative dysuria probability of animal A is 0.50, the second postoperative dysuria probability is 0.65, the time weight is 0.4, and the image weight is 0.6. Then the final postoperative dysuria probability is 0.59. By calculating the time weight and the image weight and combining the two different dysuria probabilities, researchers can more comprehensively evaluate the risk of postoperative dysuria in experimental animals, taking into account the persistence of urination time and the visual evidence of urination behavior in the activity image sequence, thereby improving the accuracy and comprehensiveness of the evaluation.

[0058] In summary, the experimental animal postoperative urination state monitoring device provided by this application has the following technical effects:

[0059] Through a urination prediction module, the urination prediction module is used to collect the dietary characteristic information of the experimental animal within the first preset time range before the experiment, conduct urination prediction, and obtain predicted urination information; a urine volume acquisition module, the urine volume acquisition module is used to monitor the urination of the experimental animal within the second preset time range after the experiment to obtain the total urine volume information, wherein the total urine volume information is obtained through evaporation compensation; a deviation analysis module, the deviation analysis module is used to classify and obtain the first postoperative dysuria probability according to the deviation between the predicted urination information and the total urine volume information; a difficulty identification module, the difficulty identification module is used to collect the animal activity images of the experimental animal within the second preset time range, conduct dysuria identification, obtain the second postoperative dysuria probability, and calculate the postoperative dysuria probability in combination with the first postoperative dysuria probability as the monitoring result of the postoperative urination state of the experimental animal. That is to say, by collecting the dietary characteristics of the animal before the experiment to predict the urination information, monitoring the urination of the animal after the experiment, and calculating the total urine volume information; comparing the deviation between the predicted urination information and the total urine volume information to calculate the first postoperative dysuria probability; collecting the activity images of the animal to identify the behavioral characteristics of dysuria and calculating the second postoperative dysuria probability; combining the first postoperative dysuria probability and the second postoperative dysuria probability to obtain the postoperative dysuria probability, accurately evaluating the postoperative urination state of the experimental animal, improving the accuracy of monitoring the postoperative dysuria of the experimental animal, and reducing the risk of misjudgment.

[0060] Embodiment 2. Based on the same inventive concept as the experimental animal postoperative urination state monitoring device in the foregoing Embodiment 1, the present application also provides an experimental animal postoperative urination state monitoring method. Please refer to the attached Figure 2 , the experimental animal postoperative urination state monitoring method includes:

[0061] Collect the dietary characteristic information of the experimental animal within the first preset time range before the experiment, conduct urination prediction, and obtain predicted urination information; within the second preset time range after the experiment, monitor the urination of the experimental animal to obtain the total urine volume information, wherein the total urine volume information is obtained through evaporation compensation; classify and obtain the first postoperative dysuria probability according to the deviation between the predicted urination information and the total urine volume information; collect the animal activity images of the experimental animal within the second preset time range, conduct dysuria identification, obtain the second postoperative dysuria probability, and calculate the postoperative dysuria probability in combination with the first postoperative dysuria probability as the monitoring result of the postoperative urination state of the experimental animal.

[0062] Further, the collection of the dietary characteristic information of the experimental animal within the first preset time range before the experiment and the conduct of urination prediction include:

[0063] Collect the dietary characteristic information of the experimental animals within the first preset time range before the experiment; according to the experimental animal diet and urination monitoring data, collect the sample dietary characteristic information set and the sample urination information set; use the sample dietary characteristic information set and the sample urination information set to train the urination prediction path; input the dietary characteristic information into the urination prediction path, and predict and output to obtain the predicted urination information.

[0064] Further, within the second preset time range after the experiment, conduct urination monitoring on the experimental animals to obtain the total urination volume information, including:

[0065] Within the second preset time range after the experiment, conduct urination monitoring on the experimental animals to obtain the urination volume information, and collect the urination time; according to the urination time, perform evaporation compensation on the urination volume information to obtain the total urination volume information.

[0066] Further, the performing evaporation compensation on the urination volume information according to the urination time includes:

[0067] Map to obtain the evaporation ratio according to the urination time, wherein, according to the experimental animal urination monitoring data, collect the sample urination time set and the sample evaporation ratio set, and construct an evaporation ratio classifier for mapping classification; use the evaporation ratio to perform evaporation compensation on the urination volume information to obtain the total urination volume information.

[0068] Further, the classifying to obtain the first postoperative dysuria probability according to the deviation between the predicted urination information and the total urination volume information includes:

[0069] Calculate the difference between the predicted urination information and the total urination volume information, and calculate the ratio of the difference to the predicted urination information to obtain the urination volume deviation; according to the dysuria detection record data of the experimental animals, collect the sample urination volume deviation set, and collect the probability of animals having dysuria under different sample urination volume deviations to obtain the sample first postoperative dysuria probability set; use the sample urination volume deviation set and the sample first postoperative dysuria probability set to construct a dysuria classifier; input the urination volume deviation into the dysuria classifier, and index and classify to obtain the first sample postoperative dysuria probability corresponding to the closest sample urination volume deviation to obtain the first postoperative dysuria probability.

[0070] Further, the collecting the animal activity images of the experimental animals within the second preset time range, performing dysuria recognition, and obtaining the second postoperative dysuria probability includes:

[0071] Collect the animal activity images of the experimental animal within the second preset time range to obtain an activity image sequence; according to the detection record data of the dysuria of the experimental animal, collect a set of sample activity image sequences, and collect the probability of the experimental animal having dysuria in different sample activity image sequences to obtain a set of sample second postoperative dysuria probabilities; use a convolutional neural network, and use the set of sample activity image sequences and the set of sample second postoperative dysuria probabilities as supervised training data to train an activity dysuria recognizer; input the activity image sequence into the activity dysuria recognizer to identify and obtain the second postoperative dysuria probability.

[0072] Further, calculating the postoperative dysuria probability by combining the first postoperative dysuria probability includes:

[0073] Obtain the urination time of the experimental animal for urination monitoring within the second preset time range; multiply the ratio of the urination time to the second preset time range by a preset time weight to obtain a time weight; use 1 minus the time weight to obtain an image weight; use the time weight and the image weight to perform weighted calculation on the first postoperative dysuria probability and the second postoperative dysuria probability to obtain the postoperative dysuria probability.

[0074] The various embodiments in this specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. The foregoing Figure 1 An experimental animal postoperative urination state monitoring device and specific examples in Embodiment 1 are equally applicable to an experimental animal postoperative urination state monitoring method in this embodiment. Through the foregoing detailed description of an experimental animal postoperative urination state monitoring device, those skilled in the art can clearly know an experimental animal postoperative urination state monitoring method in this embodiment. Therefore, for the sake of simplicity of the specification, it will not be elaborated here. For the method disclosed in the embodiment, since it corresponds to the device disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the description of the device part.

[0075] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be obvious to those skilled in the art. The general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to these embodiments shown herein, but will be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0076] Obviously, those skilled in the art can make various changes and modifications to this application without departing from the spirit and scope of this application. Thus, if these modifications and variations of this application fall within the scope of this application and its equivalent technologies, this application also intends to include these changes and modifications.

Claims

1. A device for monitoring the urination status of experimental animals after surgery, characterized in that: include: A urination prediction module, which is used to collect dietary characteristic information of the experimental animal within a first preset time range before the experiment, perform urination prediction, and obtain predicted urination information; A urine volume acquisition module, which is used to monitor the urination of the experimental animal within a second preset time range after the end of the experiment to obtain total urine volume information, wherein the total urine volume information is obtained by evaporation compensation; A deviation analysis module, the deviation analysis module is used to classify and obtain the probability of dysuria after the first operation according to the deviation between the predicted urination information and the total urination volume information; a difficulty recognition module, the difficulty recognition module is used to collect animal activity images of the experimental animal within the second preset time range, perform urination difficulty recognition, obtain a second post-operative urination difficulty probability, and calculate the post-operative urination difficulty probability in combination with the first post-operative urination difficulty probability as the post-operative urination state monitoring result of the experimental animal; Combining the first postoperative dysuria probability, calculating the postoperative dysuria probability includes: Obtaining the urination time of the experimental animal during urination monitoring within the second preset time range; According to the ratio of the urination time to the second preset time range, multiplying it by a preset time weight to obtain a time weight; Using 1-the time weight, obtain the image weight; The first probability of postoperative dysuria and the second probability of postoperative dysuria are weightedly calculated using the time weight and the image weight to obtain the probability of postoperative dysuria, wherein the probability of postoperative dysuria=time weight×first probability of postoperative dysuria+image weight×second probability of postoperative dysuria.

2. The postoperative urination status monitoring device for experimental animals according to claim 1, characterized in that: Collect the dietary characteristics of the experimental animals within the first preset time range before the experiment to predict urination, including: Collecting dietary characteristic information of the experimental animals within a first preset time range before the experiment; According to the experimental animals' diet and urination monitoring data, a sample diet characteristic information set and a sample urination information set are collected; Using the sample diet feature information set and the sample urination information set to train a urination prediction path; The dietary characteristic information is input into the urination prediction path, and the prediction output is used to obtain the predicted urination information.

3. The postoperative urination status monitoring device for experimental animals according to claim 1, characterized in that: Within a second preset time range after the end of the experiment, the experimental animal is monitored for urination to obtain total urine volume information, including: Within a second preset time range after the end of the experiment, monitoring the urination of the experimental animal to obtain urine volume information and collect urination time; According to the urination time, evaporation compensation is performed on the urine volume information to obtain total urine volume information.

4. The device for monitoring the postoperative urination status of experimental animals according to claim 3, characterized in that: According to the urination time, evaporation compensation is performed on the urine volume information, including: According to the urination time, the evaporation ratio is obtained by mapping, wherein a sample urination time set and a sample evaporation ratio set are collected according to the animal experiment urination monitoring data, and an evaporation ratio classifier is constructed to perform mapping classification; The evaporation ratio is used to perform evaporation compensation on the urine volume information to obtain total urine volume information.

5. The device for monitoring the postoperative urination status of experimental animals according to claim 3, characterized in that: According to the deviation between the predicted urination information and the total urination volume information, the probability of dysuria after the first operation is obtained by classification, including: Calculating a difference between the predicted urination information and the total urination volume information, and calculating a ratio of the difference to the predicted urination information to obtain a urination volume deviation; According to the urination difficulty detection record data of the experimental animals, the sample urine volume deviation set is collected, and the probability of urination difficulty in animals under different sample urine volume deviations is collected to obtain the sample first postoperative urination difficulty probability set; Using the sample urine volume deviation set and the sample first postoperative dysuria probability set, constructing a dysuria classifier; The urine volume deviation is input into the dysuria classifier, and the first sample postoperative dysuria probability corresponding to the closest sample urine volume deviation is obtained by index classification to obtain the first postoperative dysuria probability.

6. The device for monitoring the postoperative urination status of experimental animals according to claim 1, characterized in that: Collecting animal activity images of the experimental animal within the second preset time range, performing urination difficulty recognition, and obtaining a second postoperative urination difficulty probability, including: Collecting animal activity images of the experimental animal within the second preset time range to obtain an activity image sequence; According to the urination difficulty detection record data of the experimental animals, a set of sample activity image sequences is collected, and the probability of urination difficulty occurring in the experimental animals in different sample activity image sequences is collected to obtain a set of sample second postoperative urination difficulty probabilities; Using a convolutional neural network, the sample activity image sequence set and the sample second postoperative dysuria probability set are used as supervised training data to train an activity dysuria identifier; The active image sequence is input into the active urination difficulty identifier to identify and obtain the second postoperative urination difficulty probability.

7. A method for monitoring the urination status of experimental animals after surgery, characterized in that: The method for monitoring the postoperative urination status of an experimental animal is performed by a device for monitoring the postoperative urination status of an experimental animal according to any one of claims 1 to 6, and comprises: Collecting dietary characteristic information of the experimental animals within a first preset time range before the experiment, performing urination prediction, and obtaining predicted urination information; Within a second preset time range after the end of the experiment, monitoring the urination of the experimental animal to obtain total urine volume information, wherein the total urine volume information is obtained by evaporation compensation; According to the deviation between the predicted urination information and the total urination volume information, classify and obtain the probability of dysuria after the first operation; Collecting animal activity images of the experimental animal within the second preset time range, performing urination difficulty recognition, obtaining a second postoperative urination difficulty probability, and combining the first postoperative urination difficulty probability to calculate the postoperative urination difficulty probability as the postoperative urination state monitoring result of the experimental animal; Combining the first postoperative dysuria probability, calculating the postoperative dysuria probability includes: Obtaining the urination time of the experimental animal during urination monitoring within the second preset time range; According to the ratio of the urination time to the second preset time range, multiplying it by a preset time weight to obtain a time weight; Using 1-the time weight, obtain the image weight; The first probability of postoperative dysuria and the second probability of postoperative dysuria are weightedly calculated using the time weight and the image weight to obtain the probability of postoperative dysuria, wherein the probability of postoperative dysuria=time weight×first probability of postoperative dysuria+image weight×second probability of postoperative dysuria.

Images