Artificial intelligence-based knee joint state identification method and system

Abstract

The embodiment of the invention relates to the technical field of medical information analysis, and discloses a knee bone joint state recognition method and system based on artificial intelligence, and the method comprises the steps: obtaining a temperature distribution diagram of a knee joint of a to-be-recognized user, and carrying out the region division of the temperature distribution diagram, so as to divide the temperature distribution diagram into a temperature region diagram; processing each temperature data point in the temperature area chart according to a set temperature adjustment logic to obtain a processed temperature standard chart; and inputting the temperature standard graph into a multi-classification recognition model for graph recognition to obtain a state classification result of the knee joint of the to-be-recognized user. According to the knee joint state recognition method based on artificial intelligence in the embodiment of the invention, the processed temperature standard diagram is input into the pre-constructed multi-classification recognition model, and the state of the knee joint can be automatically classified. According to the intelligent classification method, the recognition speed and accuracy are improved, and the accuracy of targeted protection and health care of the knee joint of the user is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of medical information processing and analysis, and in particular to a method and system for identifying the state of a knee joint based on artificial intelligence. Background Art

[0002] At present, knee osteoarthritis (KOA) is a bone and joint disease caused by cartilage degeneration and bone hyperplasia in the knee joint. With the extension of life expectancy and the progress of population aging, KOA not only greatly damages the quality of life of patients, but also brings a heavy burden to global public health. In 2020, 595 million people in the world suffer from osteoarthritis, and it is estimated that by 2050, the number of KOA cases will increase by 74.9%.

[0003] Different joint conditions will have different health care plans. For example, the two symptoms of wind-cold dampness arthralgia and damp-heat arthralgia should have differentiated health care plans. If the same health care plan is used, some users will have negative effects. Therefore, designing a solution that can accurately identify the actual state of the user's knee joint has become a technical problem that needs to be solved urgently by those skilled in the art. Summary of the invention

[0004] In view of the above-mentioned defects, an embodiment of the present invention discloses a method for identifying the state of knee joint based on artificial intelligence, which can realize accurate identification of the state of user's knee joint and provide more accurate results for user's reference.

[0005] The first aspect of the embodiment of the present invention discloses a method for identifying a knee joint state based on artificial intelligence, comprising: Acquire a temperature distribution map of a knee joint of a user to be identified, and divide the temperature distribution map into regions to divide the temperature distribution map into a temperature region map, wherein the temperature region map includes a plurality of regions to be identified, and each of the regions to be identified includes a plurality of temperature data points; Processing each temperature data point in a temperature region map having a plurality of regions to be identified according to a set temperature adjustment logic to obtain a processed temperature standard map; The temperature standard image is input into a pre-built multi-classification recognition model for image recognition to obtain a state classification result of the knee joint of the user to be identified, and the state classification result is displayed.

[0006] As an optional implementation, in the first aspect of the embodiment of the present invention, the step of obtaining a temperature distribution map of a knee joint of a user to be identified includes: The temperature distribution map of the knee joint of the user to be identified is obtained by a temperature sensor component arranged at the corresponding user's knee joint; the temperature sensor component includes a temperature sensor, a temperature processor and a wireless communication module, there are multiple temperature sensors, and the multiple temperature sensors and wireless communication modules are electrically connected to the temperature processor, the temperature sensor is used to monitor the temperature signals of different areas, and transmit the monitored temperature signals to the temperature processor, the temperature processor converts the received temperature signals into temperature data of each area after analog-to-digital conversion, signal amplification and low-pass filtering, and generates a temperature distribution map based on the temperature data, and the wireless communication module is used to transmit the temperature distribution map to the background server; Or, obtaining a first temperature distribution map of the knee joint of the user to be identified in a specific image acquisition environment by an infrared thermal imager disposed at a first position; Acquiring a second temperature distribution map of the knee joint of the user to be identified in a specific image acquisition environment by using an infrared thermal imager disposed at a second position; A third temperature distribution map of the knee joint of the user to be identified in a specific image acquisition environment is obtained by using an infrared thermal imager set at a third position; and when acquiring the temperature distribution map, the specific image acquisition environment is an independent acquisition space, the room temperature is 24°C, and the relative humidity is 50% to 60%.

[0007] As an optional implementation, in the first aspect of the embodiment of the present invention, the temperature distribution map is divided into regions to divide the temperature distribution map into a temperature region map, wherein the temperature region map includes a plurality of regions to be identified, and each of the regions to be identified includes a plurality of temperature data points, including: Inputting the temperature distribution map into a pre-built part recognition model to obtain the specific location of the patella feature, the specific location of the popliteal crease feature, and the specific location of the Zusanli feature in the temperature distribution map; Determining the corresponding patellar bottom position, patellar tip position, patellar lateral edge position and patellar medial edge position based on the specific position of the patellar feature; Determine the corresponding positions of both ends of the popliteal crease based on the specific position of the popliteal crease feature, and determine the corresponding Zusanli position based on the specific position of the Zusanli feature; Generate 4 corresponding horizontal lines according to the position above the patellar base, the patellar base, the patellar tip and the Zusanli position along the transverse direction of the temperature distribution map; Four corresponding vertical lines are generated according to the longitudinal positions of the lateral edge of the patella, the medial edge of the patella and the two ends of the popliteal transverse line along the temperature distribution diagram; and the temperature distribution diagram is divided into a temperature area diagram according to the intersection of the four corresponding horizontal lines and the four corresponding vertical lines, wherein the temperature area diagram includes nine areas to be identified, and each of the areas to be identified includes multiple temperature data points.

[0008] As an optional implementation, in the first aspect of the embodiment of the present invention, the step of processing each temperature data point in the temperature region map having multiple regions to be identified according to a set temperature adjustment logic to obtain a processed temperature standard map includes: Determine each area to be identified in the temperature area map, and obtain each temperature data point in each area to be identified; According to each temperature data point in each area to be identified, statistics are performed to obtain temperature mean data and temperature standard deviation data of each area to be identified; The temperature data points, the temperature mean data and the temperature standard deviation data in each area to be identified are calculated according to the temperature adjustment formula to obtain an updated temperature standard map; wherein the temperature adjustment formula is: ,in, are the parameters of the original temperature data points, is the temperature mean data of the corresponding area to be identified, is the temperature standard deviation data of the corresponding area to be identified.

[0009] As an optional implementation manner, in the first aspect of the embodiment of the present invention, the identification method further includes: The maximum temperature, minimum temperature and average temperature of the nine areas to be identified were determined by heat map analysis method; After inputting the temperature standard image into a pre-built multi-classification recognition model for image recognition to obtain a state classification result of the user's knee joint to be identified, the method further includes: Determine the prior probability of the corresponding knee joint state in the state classification result; Determine the highest temperature, the lowest temperature and the average temperature of the nine areas to be identified to determine the conditional probability that the corresponding user to be identified is in the corresponding knee joint state; The prior probability and conditional probability of the corresponding knee joint state are input into the Bayesian model to determine the posterior probability result of the corresponding knee joint state, and the posterior probability result is output.

[0010] As an optional implementation, in the first aspect of the embodiment of the present invention, the multi-classification recognition model is constructed by the following steps: Constructing an initial training model, the initial training model includes an input layer, a first hidden layer, a second hidden layer and an output layer, wherein the input layer includes nine input features, and the nine input features correspond to the temperature parameters of nine areas of the knee joint respectively; the first hidden layer includes 64 neurons, and the activation function is ReLU, the second hidden layer includes 32 neurons, and the activation function is ReLU, and the output layer includes 3 neurons, and the activation function is softmax; Acquire an infrared sample training set, wherein the sample training set includes an infrared sample image of a knee joint; and divide the infrared sample training set of the knee joint into a training set, a validation set, and a test set; Performing regional division on the infrared sample images of the knee joint in the training set to determine the temperature images of each region in the infrared sample images of the knee joint; Inputting the infrared sample images of the knee joint after the region division into the initial training model with completed parameter configuration for training and parameter configuration adjustment until the corresponding initial training model meets the training requirements, and the trained initial training model is used as a multi-classification recognition model; After the corresponding initial training model meets the training requirements and the trained initial training model is used as the multi-classification recognition model, the following further includes: Use the trained multi-classification recognition model to predict the infrared sample images of the knee joint in the test set, and obtain the predicted category to which each infrared sample image of the knee joint belongs; Generate a category confusion matrix based on the prediction results and true annotations of the multi-classification recognition model; the rows of the category confusion matrix represent true categories, and the columns represent predicted categories; Calculating overall performance indicators of the target detection model according to the category confusion matrix, wherein the overall performance indicators include accuracy, precision, recall rate and F1 score; The multi-classification recognition model is adjusted and optimized according to the analysis results of the category confusion matrix.

[0011] As an optional implementation, in the first aspect of the embodiment of the present invention, the step of obtaining a temperature distribution map of a knee joint of a user to be identified includes: Acquire a sequence temperature distribution map of the knee joint of the user to be identified when the leg moves from a first position to a second position; there are multiple sequence temperature distribution maps; The joint identification method further comprises: The image sequence features in the multiple sequence temperature images are extracted by using a convolutional neural network. The image sequence features are input into the LSTM model for analysis to determine the corresponding joint state results.

[0012] A second aspect of an embodiment of the present invention discloses a knee joint state recognition system based on artificial intelligence, comprising: An acquisition module is used to acquire a temperature distribution map of a knee joint of a user to be identified, and divide the temperature distribution map into regions to divide the temperature distribution map into a temperature region map, wherein the temperature region map includes a plurality of regions to be identified, and each of the regions to be identified includes a plurality of temperature data points; Temperature processing module: used to process each temperature data point in the temperature region diagram having multiple regions to be identified according to the set temperature adjustment logic to obtain a processed temperature standard diagram; Recognition module: used for inputting the temperature standard image into a pre-built multi-classification recognition model for image recognition to obtain the state classification result of the knee joint of the user to be identified, and displaying the state classification result.

[0013] The third aspect of an embodiment of the present invention discloses an electronic device, comprising: a memory storing executable program code; a processor coupled to the memory; the processor calls the executable program code stored in the memory to execute the artificial intelligence-based knee joint state recognition method disclosed in the first aspect of the embodiment of the present invention.

[0014] A fourth aspect of an embodiment of the present invention discloses a computer-readable storage medium storing a computer program, wherein the computer program enables a computer to execute the artificial intelligence-based knee joint state identification method disclosed in the first aspect of an embodiment of the present invention.

[0015] Compared with the prior art, the embodiments of the present invention have the following beneficial effects: The artificial intelligence-based knee joint state recognition method in the embodiment of the present invention inputs the processed temperature standard map into a pre-built multi-classification recognition model, and can automatically classify the state of the knee joint. This intelligent classification method not only improves the speed and accuracy of recognition, but also reduces manual intervention, and improves the accuracy of users' targeted protection and health care for the knee joint. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0017] Figure 1 It is a flowchart of a method for identifying a knee joint state based on artificial intelligence disclosed in an embodiment of the present invention; Figure 2 It is a schematic diagram of the process of dividing the temperature regions disclosed in the embodiment of the present invention; Figure 3 is a schematic diagram of the flow of temperature data processing disclosed in an embodiment of the present invention; Figure 4 It is a schematic diagram of a process of using a Bayesian model for identification disclosed in an embodiment of the present invention; Figure 5 is a comparison result diagram of the accuracy of various model algorithms disclosed in the embodiments of the present invention; Figure 6 is a schematic diagram of confusion matrix comparison disclosed in an embodiment of the present invention; Figure 7 It is an architecture diagram of a knee joint state recognition system based on artificial intelligence disclosed in an embodiment of the present invention; Figure 8 It is a structural schematic diagram of a knee joint state recognition system based on artificial intelligence provided by an embodiment of the present invention; Fig. 9 It is a structural schematic diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0018] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0019] It should be noted that the terms "first", "second", "third", "fourth", etc. in the specification and claims of the present invention are used to distinguish different objects rather than to describe a specific order. The terms "including" and "having" in the embodiments of the present invention and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device including a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0020] In recent years, artificial intelligence (AI) technology has been increasingly used in the field of medical care, especially in the classification and diagnosis of diseases based on clinical data and characteristics. By analyzing and learning large-scale data, AI technology can help doctors quickly and accurately identify complex types of symptoms, thereby improving the accuracy and efficiency of diagnosis. Traditional joint state recognition relies on the doctor's personal experience, the process is complicated and time-consuming, and often requires the assistance of large equipment, which increases the patient's medical burden. With the development of sensor technology and mobile devices, state recognition methods combined with AI have become a research hotspot. In particular, using biometric data such as temperature and pressure for state recognition can provide a non-invasive, low-cost and efficient solution. The intelligent detection system for knee joint state is still in the development stage. Existing research focuses on model training and optimization based on imaging or physical detection data, while research on knee osteoarthritis syndrome classification based on temperature data is still relatively limited. Based on this, the embodiment of the present invention discloses a knee joint state recognition method based on artificial intelligence, a system, an electronic device and a storage medium. The knee joint state recognition method based on artificial intelligence inputs the processed temperature standard map into a pre-built multi-classification recognition model, which can automatically classify the state of the knee joint. This intelligent classification method not only improves the speed and accuracy of identification, but also reduces manual intervention and improves the accuracy of users' targeted protection and health care for the knee joint.

[0021] Embodiment 1 See also Figure 1 , Figure 1 It is a flow chart of the method for identifying the state of the knee joint based on artificial intelligence disclosed in the embodiment of the present invention. Among them, the execution subject of the method described in the embodiment of the present invention is an execution subject composed of software and / or hardware, and the execution subject can receive relevant information by wired or / and wireless means, and can send certain instructions. Of course, it can also have certain processing functions and storage functions. The execution subject can control multiple devices, such as a remote physical server or cloud server and related software, or it can be a local host or server and related software that performs related operations on a device placed somewhere. In some scenarios, multiple storage devices can also be controlled, and the storage devices can be placed in the same place or different places as the devices. For example Figure 1 As shown, the knee joint state recognition method based on artificial intelligence includes the following steps: S101: Acquire a temperature distribution map of a knee joint of a user to be identified, and divide the temperature distribution map into regions to divide the temperature distribution map into a temperature region map, wherein the temperature region map includes a plurality of regions to be identified, and each of the regions to be identified includes a plurality of temperature data points; S102: processing each temperature data point in the temperature region map having multiple regions to be identified according to a set temperature adjustment logic to obtain a processed temperature standard map; S103: Inputting the temperature standard image into a pre-built multi-classification recognition model for image recognition to obtain a state classification result of the knee joint of the user to be identified, and displaying the state classification result.

[0022] The solution of the embodiment of the present invention can analyze the temperature characteristics of different parts of the knee joint in more detail by obtaining the temperature distribution map of the knee joint of the user to be identified and dividing the image into regions. This refined processing helps to capture the slight temperature changes of the knee joint, thereby improving the accuracy of knee joint identification. At the same time, the temperature data points are processed according to the set temperature adjustment logic, which can further eliminate noise and improve the reliability and recognition efficiency of the data. By displaying the results of the classification of the knee joint status, doctors and patients can intuitively understand the health status of the knee joint. This helps doctors to formulate more accurate treatment plans, and also facilitates patients to conduct self-management and rehabilitation.

[0023] Specifically, wind-cold and damp-heat have different temperature manifestations, and the targeted treatment methods required are also very different, requiring hot compress and cold compress respectively. Therefore, it is necessary to use temperature as an indicator for classification. In the embodiment of the present invention, infrared thermal imaging data of the knee joint is obtained, and the data is processed to obtain optimized temperature data; then the optimized temperature data is further identified to determine the final KOA syndrome type. When displaying the status classification results, the application can display its similarity with a certain syndrome type, and then provide the similarity and temperature distribution map to the user for reference. The solution of the present application can solve the problems of inaccurate KOA syndrome type identification, expensive equipment, and diagnosis relying on clinical experts.

[0024] More preferably, the step of obtaining a temperature distribution diagram of a knee joint of a user to be identified includes: The temperature distribution map of the knee joint of the user to be identified is obtained by a temperature sensor component arranged at the corresponding user's knee joint; the temperature sensor component includes a temperature sensor, a temperature processor and a wireless communication module, there are multiple temperature sensors, and the multiple temperature sensors and wireless communication modules are electrically connected to the temperature processor, the temperature sensor is used to monitor the temperature signals of different areas, and transmit the monitored temperature signals to the temperature processor, the temperature processor converts the received temperature signals into temperature data of each area after analog-to-digital conversion, signal amplification and low-pass filtering, and generates a temperature distribution map based on the temperature data, and the wireless communication module is used to transmit the temperature distribution map to the background server; The solution of the embodiment of the present invention can directly and accurately monitor the temperature signals of different areas of the knee joint by setting the temperature sensor component at the corresponding user's knee joint. The layout of multiple temperature sensors can cover the key parts of the knee joint to ensure the comprehensiveness and accuracy of the temperature data. The temperature processor performs analog-to-digital conversion, signal amplification and low-pass filtering on the received temperature signal, further improving the accuracy and stability of the temperature data. The temperature distribution map generated based on these precise temperature data can truly reflect the temperature distribution of the knee joint. The wireless communication module in the temperature sensor component can transmit the temperature distribution map to the background server in real time, realizing the real-time processing and remote monitoring of data. This helps doctors to obtain the patient's knee joint temperature information in a timely manner and make accurate diagnosis and treatment decisions.

[0025] Or, obtaining a first temperature distribution map of the knee joint of the user to be identified in a specific image acquisition environment by an infrared thermal imager disposed at a first position; Acquiring a second temperature distribution map of the knee joint of the user to be identified in a specific image acquisition environment by using an infrared thermal imager disposed at a second position; A third temperature distribution map of the knee joint of the user to be identified in a specific image acquisition environment is obtained by using an infrared thermal imager set at a third position; and when acquiring the temperature distribution map, the specific image acquisition environment is an independent acquisition space, the room temperature is 24°C, and the relative humidity is 50% to 60%.

[0026] The three directions here are the frontal position of the leg and the left and right lateral positions; the images in these three directions are combined to improve the accuracy of the final overall recognition. By obtaining the multi-angle temperature distribution map of the knee joint through infrared thermal imagers set in different directions, the temperature changes of the knee joint can be observed more comprehensively. This multi-angle imaging technology helps to capture the slight temperature changes of the knee joint in different directions and improve the accuracy and comprehensiveness of recognition. The acquisition of the temperature distribution map in a specific image acquisition environment ensures the consistency and comparability of image acquisition. The environmental conditions of independent acquisition space, room temperature of 24°C, and relative humidity of 50% to 60% help reduce the impact of external factors on the temperature distribution map and improve the quality and accuracy of the image. This method combines direct temperature sensing monitoring and multi-angle infrared thermal imaging technology to provide richer and more accurate knee joint temperature information. This helps doctors quickly and accurately judge the health of the knee joint and formulate effective treatment plans. When taking images, the corresponding user is also required to be in a quiet state for 15 minutes before taking the corresponding image.

[0027] More preferably, the temperature distribution map is divided into regions to divide the temperature distribution map into a temperature region map, wherein the temperature region map includes a plurality of regions to be identified, and each of the regions to be identified includes a plurality of temperature data points, including: S1011: inputting the temperature distribution map into a pre-built part recognition model to obtain the specific location of the patella feature, the specific location of the popliteal crease feature, and the specific location of the Zusanli feature in the temperature distribution map; S1012: Determine the corresponding patellar bottom position, patellar tip position, patellar lateral edge position, and patellar medial edge position based on the specific position of the patellar feature; S1013: determining the corresponding positions of both ends of the popliteal crease based on the specific position of the popliteal crease feature, and determining the corresponding Zusanli position based on the specific position of the Zusanli feature; S1014: generating four corresponding horizontal lines according to the positions of the upper part of the patellar bottom, the patellar bottom, the patellar tip, and the Zusanli position along the transverse direction of the temperature distribution map; S1015: Generate 4 corresponding vertical lines according to the longitudinal position of the lateral edge of the patella, the medial edge of the patella and the two ends of the popliteal crease along the temperature distribution diagram; and divide the temperature distribution diagram into a temperature area diagram according to the intersection of the 4 corresponding horizontal lines and the 4 corresponding vertical lines, wherein the temperature area diagram includes 9 areas to be identified, and each of the areas to be identified includes multiple temperature data points. The names of the 9 areas to be identified from top to bottom and from left to right are: upper inner area, suprapatellar bursa area, upper outer area, medial area, patellar area, lateral area, anserine bursa area, subpatellar area, and lower outer area; The scheme of the embodiment of the present invention is based on the anatomical characteristics of the knee joint and the common clinical pain sites. Four horizontal lines are made at 3 inches above the patellar base, the patellar base, the patellar tip, and Zusanli, and four vertical lines are made at the lateral edge of the patella, the medial edge of the patella, and the two ends of the popliteal transverse line. The knee joint is divided into 9 observation areas from outside to inside and from top to bottom. And there will be differences in the temperature display results in different areas. After obtaining the above temperature data, data cleaning can be performed, and the specific data cleaning is to process missing values ​​or outliers. If there is noise or abnormal points in the thermal imaging data, use interpolation methods to fill in the missing values, or use statistical methods to eliminate abnormal points.

[0028] The solution of the embodiment of the present invention can accurately identify the specific positions of the patellar features, popliteal crease features and Zusanli features in the temperature distribution map through the pre-constructed part recognition model. The accurate recognition of these feature points provides a solid foundation for the subsequent regional division. Based on the positions of these feature points, key positions such as the patellar bottom, patellar tip, lateral edge of the patella, medial edge of the patella and both ends of the popliteal crease can be accurately determined, thereby generating accurate horizontal and vertical lines. The intersection of these lines divides the temperature distribution map into multiple areas to be identified, each of which corresponds to a different part of the knee joint. Features related to the popliteal crease are extracted using image processing technology. These features may include differences in temperature distribution, texture patterns, etc. The popliteal crease may appear as a specific temperature change or texture pattern on the infrared image, which is different from the surrounding tissue; machine learning or deep learning algorithms are applied to classify and identify the extracted features. The feature pattern of the popliteal crease is learned by training the model and the recognition accuracy is improved. This step requires a large amount of labeled data to train the model to ensure that it can accurately identify the popliteal crease. Image processing technology is used to extract features related to Zusanli. These features may include differences in temperature distribution, texture patterns, etc. In infrared images, the Zusanli area shows specific temperature changes or texture patterns that are different from the surrounding tissues. Machine learning or deep learning algorithms are applied to classify and identify the extracted features. The model is trained to learn the characteristic patterns of Zusanli and improve the recognition accuracy.

[0029] The embodiment of the present invention can locate the lesion area more accurately by dividing the knee joint temperature distribution map into multiple areas. For example, in KOA patients, the temperature of certain specific areas (such as the medial edge of the patella, popliteal fossa, etc.) may rise abnormally, and the high temperature distribution of these areas can be used as an important feature for identifying KOA. Regional division helps doctors to observe the temperature conditions of different parts of the knee joint more comprehensively, thereby reducing the risk of misdiagnosis and missed diagnosis. For example, if the temperature in a certain area rises abnormally, but the doctor does not take the area into consideration, it may lead to missed diagnosis. On the contrary, if the regional division is too general, the temperature characteristics of different parts may be confused, leading to misdiagnosis. On the one hand, regional division can not only improve the final recognition accuracy, but also provide users with a basis for subsequent detailed treatment plans by providing more comprehensive data.

[0030] More preferably, the step of processing each temperature data point in the temperature region map having multiple regions to be identified according to a set temperature adjustment logic to obtain a processed temperature standard map includes: S1021: Determine each area to be identified in the temperature area map, and obtain each temperature data point in each area to be identified; S1022: performing statistics based on each temperature data point in each area to be identified to obtain temperature mean data and temperature standard deviation data of each area to be identified; S1023: Calculate each temperature data point, temperature mean data and temperature standard deviation data in each to-be-identified area according to a temperature adjustment formula to obtain an updated temperature standard map; wherein the temperature adjustment formula is: ,in, are the parameters of the original temperature data points, is the temperature mean data of the corresponding area to be identified, is the temperature standard deviation data of the corresponding area to be identified.

[0031] The embodiment of the present invention calculates the temperature mean and temperature standard deviation of each area to be identified, and uses the temperature adjustment formula to standardize the original temperature data points, thereby eliminating the temperature differences between different areas caused by individual differences, environmental differences and other factors. The temperature standard map processed in this way is more objective and unified, which is convenient for subsequent analysis and comparison. The standardized temperature data is more stable and reduces the interference of noise and outliers. This helps to improve the accuracy of KOA recognition, because certain features of KOA may be more clearly reflected in the standardized temperature data. Standardized temperature data is easier to process and analyze. For example, statistical methods, machine learning algorithms, etc. can be used to deeply mine the data to discover potential laws and characteristics, providing new ideas and methods for the diagnosis and treatment of KOA.

[0032] If the standardized temperature data is used to train the model, the generalization ability of the model can be improved. Because the standardization process eliminates the deviation and heterogeneity between the data, the model can focus more on learning the inherent laws and characteristics of the data, rather than being confused by the superficial differences. By dividing the temperature area map into multiple areas to be identified and standardizing the temperature data of each area, multi-scale and multi-dimensional analysis can be supported. For example, the temperature differences between different areas can be compared, and the trend of temperature changes over time can be analyzed, etc., providing richer and more comprehensive information for in-depth research on KOA.

[0033] More preferably, the identification method further includes: The maximum temperature, minimum temperature and average temperature of the nine areas to be identified were determined by heat map analysis method; After inputting the temperature standard image into a pre-built multi-classification recognition model for image recognition to obtain a state classification result of the user's knee joint to be identified, the method further includes: S104: Determine the prior probability of the corresponding knee joint state in the state classification result; S105: Determine the highest temperature, the lowest temperature and the average temperature of the nine areas to be identified to determine the conditional probability that the corresponding user to be identified is in the corresponding knee joint state; S106: Inputting the prior probability and conditional probability of the corresponding knee joint state into the Bayesian model to determine the posterior probability result of the corresponding knee joint state, and outputting the posterior probability result.

[0034] The embodiment of the present invention can intuitively display the temperature distribution of the nine areas to be identified, including the highest temperature, the lowest temperature and the average temperature value, through the heat map analysis method. This information provides an important basis for identifying the state of the knee joint and helps to improve the accuracy of identification.

[0035] By introducing prior probability and conditional probability and combining them with the Bayesian model for calculation, we can comprehensively consider the impact of multiple factors (such as temperature distribution, historical data, etc.) on the state of the knee joint. This method not only considers the current temperature data, but also the prior knowledge of the state of the knee joint, thereby enhancing the reliability of recognition.

[0036] By inputting the temperature standard image into the pre-built multi-classification recognition model for image recognition, the approximate classification results of the knee joint status can be quickly obtained. Then, by combining the prior probability, conditional probability and Bayesian model for further calculation, a more accurate posterior probability result can be obtained. This method improves the efficiency of diagnosis and enables doctors to make diagnoses faster. After obtaining the status classification results, it is necessary to determine the prior probability of the corresponding knee joint status in these classification results.

[0037] Prior probability refers to the possibility of a certain knee joint state in the absence of any additional information. This is usually estimated based on a large amount of historical data or expert experience. Next, it is necessary to determine the maximum temperature, minimum temperature, and average temperature values ​​of the 9 areas to be identified (these areas may be different parts or functional areas of the knee joint). These temperature values ​​will be used to calculate the conditional probability that the corresponding user to be identified is in the corresponding knee joint state. Conditional probability refers to the possibility of a certain knee joint state when certain information (such as temperature value) is known. The prior probability and conditional probability of the corresponding knee joint state are input into the Bayesian model. The Bayesian model uses this information to calculate the posterior probability, that is, the probability of a certain knee joint state after considering all available information (including image recognition results and temperature values).

[0038] In the embodiment of the present invention, a variety of models can be used to identify KOA syndromes, including: random forest algorithm, logistic regression algorithm, decision tree algorithm, XGBoost algorithm, AdaBoost algorithm and feedforward neural network.

[0039] First, random forest is a model based on ensemble learning. It constructs multiple decision trees for the temperature data set and combines their prediction results for classification. The model uses RandomForestClassifier() to train the random forest model. By sampling the temperature training set multiple times, multiple different decision trees are constructed, and the final classification result is determined by voting on the results of each tree. Random forest classifies the input temperature data by constructing a large number of decision trees, and improves the accuracy and robustness of the overall model by aggregating the prediction results of these trees. During the training process of each tree, a part of the temperature data points is randomly selected from the temperature data set, that is, different subsets of temperature data are formed through bootstrap sampling; at the same time, at each split node of each tree, not all possible features are considered, but a part of the features are randomly selected to search for the best split point.

[0040] The hyperparameters of random forest are configured as follows: n_estimators=100: the number of trees in the forest; criterion='gini': criterion for measuring split quality; max_depth=None: maximum depth of the tree; min_samples_split=2: the minimum number of samples required to split an internal node; min_samples_leaf=1: The minimum number of samples required in a leaf node.

[0041] Second, logistic regression is a linear model for binary or multi-classification. Logistic regression predicts the relationship between arthritis syndrome and knee temperature by modeling temperature data. It uses the Sigmoid function to map the output value of linear regression to between 0 and 1, thereby indicating the probability of belonging to a certain category. Among them: the model uses LogisticRegression(max_iter=1000), where max_iter=1000 ensures the maximum number of iterations before convergence; logistic regression classifies by applying a linear function to the input temperature data and using the sigmoid function to convert the result into a probability value, where: the Sigmoid (Logistic) function can compress the output value between 0 and 1, which is particularly suitable for tasks where the output is probability, recorded as: ,in, Represents the predicted probability value, which ranges from 0 to 1.

[0042] The hyperparameter configuration for logistic regression is: max_iter=1000, to ensure the maximum number of iterations before convergence; penalty='l2': regularization method, use L2 regularization (ridge regression); tol=1e-4: tolerance of the optimization algorithm; solver='lbfgs': algorithm used for optimization; max_iter=100: Maximum number of iterations for the optimization algorithm to converge.

[0043] Third, the decision tree algorithm recursively divides the temperature data into different subsets until each subset belongs to the same category or cannot be further divided. The decision tree is a supervised learning algorithm with a tree structure. It recursively divides the input temperature data set into smaller subsets, and finally forms a tree model to predict the output of new data. After inputting nine features, the decision tree will test a specific feature (such as the upper inner zone temperature) and assign the sample to an internal node based on the test result. Each node corresponds to a value of the upper inner zone temperature until the target classification result (such as wind-cold-dampness arthralgia) is reached; among them: the model uses DecisionTreeClassifier() to train the decision tree model; the model maximizes the information gain or Gini coefficient by calculating the split point of the feature to determine how to classify the data.

[0044] The hyperparameter configuration of the decision tree is: criterion='gini': criterion for measuring split quality; splitter='best': strategy used to split nodes; min_samples_split=2: the minimum number of samples required to split a node; min_samples_leaf=1: The minimum number of samples required in a leaf node.

[0045] Fourth, XGBoost is an integrated algorithm based on gradient boosting, which improves the performance of the model by gradually adding new weak learners. The model uses XGBClassifier(eval_metric='mlogloss'); mlogloss (multi-class logarithmic loss) is a common evaluation criterion for multi-classification problems, denoted as: ,in, is the sample size, is the number of categories, It is a sample Belongs to category (0 or 1), is the model prediction sample Belongs to category probability.

[0046] The XGBoost algorithm builds multiple weak learners (usually decision trees) and then weights and accumulates the prediction results of these learners to gradually optimize the loss function. , recorded as:

[0047] in, is the loss function, which means The true value of the sample With the predicted value The error between It is the regularization term of the model, which is used to control the complexity of the model and avoid overfitting; In this process, XGBoost uses gradient boosting to update the weight of each tree using first-order derivatives and second-order derivatives, recorded as:

[0048]

[0049] The update strategy of each round of XGBoost is based on gradient information to improve prediction accuracy.

[0050] The hyperparameter configuration of XGBoost is: eval_metric='mlogloss', mlogloss (multi-class logarithmic loss) is a commonly used evaluation metric for multi-classification problems; n_estimators=100: the number of base learners; learning_rate=0.3: learning rate for each iteration; max_depth=6: Maximum depth of tree construction; min_child_weight=1: determines the threshold for the minimum leaf node sample weight; objective='binary:logistic': objective function for classification.

[0051] Fifth, AdaBoost is a boosting algorithm that focuses more attention on samples that are difficult to classify by gradually adjusting the sample weights. The model uses AdaBoostClassifier() to train the model, and the default base learner is a decision tree. The Adaboost algorithm trains a weak learner 1 from the training set with initial weights, and updates the weights of the training samples based on the learning error rate of the weak learner, so that the weights of the training sample points with high learning error rates of the previous weak learner 1 become higher, so that these points with high error rates are given more attention in the subsequent weak learner 2. Then, based on the training set with adjusted weights, weak learner 2 is trained, and this is repeated until the number of weak learners reaches the pre-specified number T, and finally these T weak learners are integrated through the set strategy to obtain the final strong learner.

[0052] The hyperparameter configuration of AdaBoost is: base_estimator=None: base classifier, use DecisionTreeClassifier as the base classifier; n_estimators=50: the number of base classifiers; learning_rate=1.0: weight reduction rate of each classifier; algorithm='SAMME.R': Weight update algorithm that uses probability estimates for weighting.

[0053] Sixth, the feedforward neural network is a multi-layer perceptron type neural network, which consists of an input layer, a hidden layer and an output layer. Each layer contains a number of neurons, where: the input layer contains nine input features, corresponding to the temperatures of nine areas of the knee joint; the first hidden layer contains 64 neurons, and the activation function is ReLU, where: the output of each neuron is recorded as:

[0054] in, is the hidden layer The weighted input of each neuron, The input layer and The weight vector of hidden neurons, are input features, It is The bias of the hidden layer neurons. The superscript 1 indicates that this is the first hidden layer. This indicates that this is the first A neuron.

[0055] The second hidden layer contains 32 neurons, and the activation function is ReLU; the output layer contains 3 neurons, and the activation function is softmax, where the output of the output layer is expressed as:

[0056] in, The output layer The activation output of a neuron.

[0057] All models used were evaluated using accuracy and training loss, where: The training loss measures the performance of the model on the training set, using multi-classification cross entropy loss, denoted as:

[0058] in, is the sample size, is the number of categories, is the actual label, is the predicted probability. KOA classification results include: wind-cold-dampness arthralgia syndrome, damp-heat arthralgia syndrome and others.

[0059] After determining the optimal model, the parameters of the model are adjusted to meet the requirements of syndrome identification and classification of knee osteoarthritis. By adjusting these parameters, the accuracy and generalization ability of the model can be optimized. By comparing the final training results of each model, refer to Figure 5 and Figure 6 As shown, it can be found that the final recognition accuracy of the FNN model, that is, the feedforward neural network model and the XGboost model is higher. In the embodiment of the present invention, a detailed description of the subsequent recognition training process is performed using a feedforward neural network.

[0060] More preferably, the multi-classification recognition model is constructed by the following steps: Constructing an initial training model, the initial training model includes an input layer, a first hidden layer, a second hidden layer and an output layer, wherein the input layer includes nine input features, and the nine input features correspond to the temperature parameters of nine areas of the knee joint respectively; the first hidden layer includes 64 neurons, and the activation function is ReLU, the second hidden layer includes 32 neurons, and the activation function is ReLU, and the output layer includes 3 neurons, and the activation function is softmax; Acquire an infrared sample training set, wherein the sample training set includes an infrared sample image of a knee joint; and divide the infrared sample training set of the knee joint into a training set, a validation set, and a test set; Performing regional division on the infrared sample images of the knee joint in the training set to determine the temperature images of each region in the infrared sample images of the knee joint; Inputting the infrared sample images of the knee joint after the region division into the initial training model with completed parameter configuration for training and parameter configuration adjustment until the corresponding initial training model meets the training requirements, and the trained initial training model is used as a multi-classification recognition model; After the corresponding initial training model meets the training requirements and the trained initial training model is used as the multi-classification recognition model, the following further includes: Use the trained multi-classification recognition model to predict the infrared sample images of the knee joint in the test set, and obtain the predicted category to which each infrared sample image of the knee joint belongs; Generate a category confusion matrix based on the prediction results and true annotations of the multi-classification recognition model; the rows of the category confusion matrix represent true categories, and the columns represent predicted categories; Calculating overall performance indicators of the target detection model according to the category confusion matrix, wherein the overall performance indicators include accuracy, precision, recall rate and F1 score; The multi-classification recognition model is adjusted and optimized according to the analysis results of the category confusion matrix.

[0061] In the embodiment of the present invention, a feedforward neural network is selected to construct an AI model, and the data set is divided into a training set: validation set ratio of 8:2, and is input into the above model as input features. The neural network model consists of an input layer, a hidden layer, and an output layer, and each layer contains a number of neurons; the input layer contains nine input features, which correspond to the temperatures of nine areas of the knee joint respectively; the first hidden layer contains 64 neurons, and the activation function is ReLU, where the output of each neuron is recorded as: , in, is the hidden layer The weighted input of each neuron, The input layer and The weight vector of hidden neurons, are input features, It is The bias of the hidden layer neurons, the superscript 1 indicates that this is the first hidden layer, and the subscript This indicates that this is the first A neuron.

[0062] The second hidden layer contains 32 neurons, and the activation function is ReLU; the output layer contains 3 neurons, and the activation function is softmax, where the output of the output layer is expressed as:

[0063] in, The output layer The activation output of a neuron.

[0064] During the training process, each temperature feature input will be given a certain weight when it is input to the first hidden layer, and a certain gradient adjustment will be made after passing through the second hidden layer. Finally, all weights are weighted to obtain the corresponding knee osteoarthritis classification. After comparing with the actual classification and obtaining the accuracy, it is fed back to each hidden layer to make corresponding adjustments. After multiple iterative adjustments, a classification model with higher accuracy is obtained.

[0065] After determining the optimal model, the parameters of the model are adjusted to meet the requirements of knee osteoarthritis syndrome identification and classification. The network learning rate is 0.1, the first layer has 64 neurons, the second layer has 32 neurons, the optimizer used is the Adam optimizer, the activation function used in the hidden layer is ReLU, the activation function used in the output layer is softmax, the batch size in the training process is 8, and the number of training iterations is 50. By adjusting these parameters, the accuracy and generalization ability of the model can be optimized; the solution of the embodiment of the present invention ultimately aims to improve the accuracy of image recognition by constructing a model, thereby providing a specific basis for the subsequent accurate arthritis status.

[0066] The initial training model in the embodiment of the present invention adopts a multi-layer neural network structure, including an input layer, two hidden layers and an output layer. The number of neurons in each layer and the selection of activation functions are carefully designed to ensure that the model can fully learn the temperature characteristics in the infrared sample images of the knee joint. In particular, the use of the ReLU activation function helps to solve the gradient vanishing problem and accelerate the training process of the model.

[0067] The division of the training set, validation set, and test set ensures that the model will not be overfitted or underfitted during the training process. By dividing the infrared sample images of the knee joint in the training set into regions, the temperature images of each region are determined, providing rich feature information for model training. At the same time, by continuously adjusting the model parameters until the training requirements are met, the stability and accuracy of the model are ensured.

[0068] When implementing the model, the confusion matrix can be used to analyze the performance of the model in different categories. For example, by observing the off-diagonal elements in the confusion matrix (i.e., the number of samples with incorrect predictions), we can identify which categories are easily confused, analyze the possible reasons, and take corresponding improvement measures. According to the analysis results of the confusion matrix, the model can be adjusted and optimized. For example, for categories with more serious confusion, the number of samples can be increased or the model parameters can be adjusted to improve recognition performance.

[0069] For species with a small number of samples, you can also consider using data enhancement technology to increase the number of samples to improve the model's recognition ability for the species. You can further configure the corresponding training set for different recognition results, and optimize the specific training set through different aspects of data enhancement, thereby completing the optimization of the corresponding model.

[0070] More preferably, the step of obtaining a temperature distribution diagram of a knee joint of a user to be identified includes: Acquire a sequence temperature distribution map of the knee joint of the user to be identified when the leg moves from a first position to a second position; there are multiple sequence temperature distribution maps; The joint identification method further comprises: The image sequence features in the multiple sequence temperature images are extracted by using a convolutional neural network. The image sequence features are input into the LSTM model for analysis to determine the corresponding joint state results.

[0071] Here, "from the first position to the second position" specifically refers to the temperature image of the knee joint when the leg changes from a vertical state to a horizontal state; by obtaining the sequence temperature distribution map of the knee joint of the user to be identified when the leg changes from the first position to the second position, the temperature change of the knee joint in different positions and different motion states can be captured. Capturing this dynamic information is of great significance for understanding the motion law and pathological changes of the knee joint.

[0072] Using convolutional neural networks (CNNs) to extract features from multiple sequential temperature distribution images can automatically learn and extract key information from images, such as temperature distribution patterns, abnormal hot spots, etc. CNN's efficient feature extraction capability provides a solid foundation for subsequent joint state recognition.

[0073] Inputting the extracted image sequence features into the LSTM model for analysis can make full use of the contextual information in the time series data to make a more accurate judgment on the state of the knee joint. The LSTM model has unique advantages in processing time series data and can capture long-term dependencies in the data.

[0074] Combining the feature extraction capability of CNN and the time series analysis capability of LSTM can significantly improve the accuracy of joint state recognition. This method not only considers the temperature distribution of the knee joint at different locations, but also considers the temperature change trend over time, thereby more comprehensively evaluating the state of the knee joint.

[0075] In order to realize the sequence analysis of knee joint temperature images and then determine the arthritis syndrome, the combination of convolutional neural network (CNN) and recurrent neural network (RNN) in deep learning can be used, or long short-term memory network (LSTM) can be used directly to process sequence data. However, considering that LSTM may need to convert the image into a feature vector when processing the image sequence, which may lose some image information, it is more recommended to use a combination of CNN and LSTM, that is, ConvLSTM (convolutional long short-term memory network) or CNN and standard LSTM (that is, first use CNN to extract image features, and then use LSTM to process sequence features). Use labeled temperature image sequences (that is, image sequences with known arthritis syndromes) to train the model. During the training process, the model will learn the association between the temperature change characteristics of the knee joint in different states and the arthritis syndrome.

[0076] The solution of the embodiment of the present invention provides an easy-to-use identification method. By simplifying the operation process, ordinary users can also obtain reliable diagnosis results without professional knowledge, which promotes the popularization and application of technology. The combination of the network terminal and the mobile terminal can intelligently judge the incidence of KOA in real time, and collect relevant data for patients, doctors or caregivers to use, avoiding the uncertainty caused by relying solely on the doctor's experience, and providing a data basis for doctors to provide remote rehabilitation training guidance.

[0077] The artificial intelligence-based knee joint state recognition method in the embodiment of the present invention inputs the processed temperature standard map into a pre-built multi-classification recognition model, and can automatically classify the state of the knee joint. This intelligent classification method not only improves the speed and accuracy of recognition, but also reduces manual intervention, and improves the accuracy of users' targeted protection and health care for the knee joint.

[0078] Embodiment 2 See also Figure 8 , Figure 8 FIG. 1 is a schematic diagram of the structure of the knee joint state recognition system based on artificial intelligence disclosed in an embodiment of the present invention. Figure 8 As shown, the artificial intelligence-based knee joint state recognition system may include: Acquisition module 21: used for acquiring a temperature distribution map of a knee joint of a user to be identified, and dividing the temperature distribution map into regions to divide the temperature distribution map into a temperature region map, wherein the temperature region map includes a plurality of regions to be identified, and each of the regions to be identified includes a plurality of temperature data points; Temperature processing module 22: used for processing each temperature data point in the temperature region diagram having multiple regions to be identified according to a set temperature adjustment logic to obtain a processed temperature standard diagram; Recognition module 23: used for inputting the temperature standard image into a pre-built multi-classification recognition model for image recognition to obtain a state classification result of the knee joint of the user to be identified, and displaying the state classification result.

[0079] The artificial intelligence-based knee joint state recognition method in the embodiment of the present invention inputs the processed temperature standard map into a pre-built multi-classification recognition model to automatically classify the state of the knee joint. This intelligent classification method not only improves the speed and accuracy of recognition, but also reduces manual intervention and reduces the risk of misdiagnosis and missed diagnosis.

[0080] like Figure 7 As shown, an embodiment of the present invention provides a KOA syndrome type recognition system based on artificial intelligence. Based on the constructed KOA syndrome type judgment model, the KOA syndrome type recognition system includes a temperature acquisition module, a mobile terminal, and a network terminal connected in sequence, wherein: the knee joint temperature acquisition module is used to collect temperature data of nine areas of the knee joint of the patient to be tested, and then send it to the mobile terminal; the mobile terminal is used to store the temperature data sent by the temperature acquisition module, and then forward it to the network terminal; the network terminal is used to convert the temperature data for intelligent classification and recognition; the network terminal is provided with a KOA syndrome type judgment model; the KOA syndrome type judgment model is used to perform classification and recognition according to the temperature data transmitted by the mobile terminal, and output the knee osteoarthritis syndrome of the subject.

[0081] More preferably, the temperature acquisition module includes a temperature sensor assembly, a signal processor, and a wireless communication module; the temperature sensor assembly includes nine temperature sensors, which are respectively arranged in nine areas of the knee joint; the temperature sensor assembly is used to collect temperature signals of the knee joints of the patient to be tested; the temperature signal processor is electrically connected to the multiple temperature sensors and the wireless communication module; the temperature signal processor is used to convert the temperature signal into temperature data; The wireless communication module communicates with the mobile terminal wirelessly to send temperature data to the mobile terminal. The method is to build a model, so the collected temperature data is based on clinical practice. Clinical temperature collection is through thermal imaging, which can obtain the temperature of each pixel. More information is obtained, and the model built later is more accurate.

[0082] In addition to thermal imaging, temperature detection is also used in the specific implementation. The temperature sensing detection method can be integrated into wearable devices to expand its scope of use.

[0083] More preferably, a plurality of temperature sensors are respectively arranged at the upper inner area, suprapatellar bursa area, upper outer area, medial area, patellar area, lateral area, anserine bursa area, subpatellar area, and lower outer area of ​​the knee joint of the patient to be tested; The mobile terminal is provided with a communication transceiver module, a local storage module, and a display module; the communication transceiver module communicates wirelessly with the wireless communication module in the temperature acquisition module and is connected to the local storage module; the communication transceiver module is used to forward temperature data; the local storage module is used to store temperature data and the classification results of knee osteoarthritis syndromes in the network terminal; the display module is connected to the local storage module and is used to visualize the temperature data and the classification results of knee osteoarthritis syndromes in the network terminal.

[0084] More preferably, the network end is further provided with a data communication module, a data storage module, a calculation module, and a label module; the data communication module communicates with the communication transceiver module of the mobile end, and is used for receiving temperature data and sending the classification results of knee osteoarthritis syndrome; the calculation module is used for performing calculation processing on the temperature data to form data to be learned; the label module is used for labeling the data to be learned with the serial number of the patient to be tested, and forming labeled data to be learned; the input of the KOA syndrome judgment model is the labeled data to be learned, and the output is the syndrome of knee osteoarthritis of the patient to be tested corresponding to the label: wind-cold-dampness arthralgia syndrome, damp-heat arthralgia syndrome and others; the data storage module is respectively connected to the KOA syndrome judgment model, the data communication module, the calculation module, and the label module, and is used for storing temperature data, data to be learned, labeled data to be learned, and the classification results of the KOA syndrome judgment model.

[0085] More preferably, the temperature data is the temperature value collected by the temperature sensor; the calculation module performs the operation processing on the temperature data by standardizing the mean temperature value corresponding to each temperature sensor in turn, thereby forming the data to be learned.

[0086] More preferably, the network end is also provided with an electronic file generation module; the electronic file generation module is connected to the data storage module; the electronic file generation module is used to statistically generate electronic files based on the classification results of the KOA syndrome type judgment model, and transmit the electronic files to the data storage module for storage.

[0087] More preferably, the mobile terminal is a mobile phone and the network terminal is a server.

[0088] S1. The temperature monitoring module is used to collect temperature data of nine areas of the knee joint of patients with knee osteoarthritis. The area is divided by making four horizontal lines at 3 inches above the patellar base, the patellar base, the patellar tip, and Zusanli, and four vertical lines at the lateral edge of the patella, the medial edge, and the two ends of the popliteal crease. The knee joint is divided from outside to inside and from top to bottom into 9 observation areas: upper inner area, suprapatellar bursa area, upper outer area, medial area, patellar area, lateral area, anserine bursa area, subpatellar area, and lower outer area; a temperature sensor is bound to each of these nine areas; S1.1. When the patient to be tested sits still, the temperature sensor monitors the temperature signals of different areas. The temperature signal processor is electrically connected to each temperature sensor respectively. In this embodiment, the signal is preferably transmitted by wired connection of the signal line, and the five signal lines are wired through a connecting tube. The preferred temperature signal processor of this embodiment is specifically composed of a single-chip microcomputer and its peripheral circuits. The temperature signal processor is used to convert the received temperature signal into temperature data of each area after analog-to-digital conversion, signal amplification, and low-pass filtering. The wireless communication module is electrically connected to the temperature signal processor, and wirelessly communicates with the mobile terminal, and is used to send the temperature data generated by the temperature signal processor to the mobile terminal. The preferred specific hardware of the wireless communication module of this embodiment is a Bluetooth chip or a WIFI chip. S2, the mobile terminal is used to receive the data generated by the temperature monitoring module, store and display the data, and then send it to the network terminal, and receive electronic files from the network terminal for storage and display. The communication transceiver module is used to send and receive data and electronic files. The preferred communication transceiver module of this embodiment adopts wireless communication, and the communication modes of the wireless communication module of the temperature monitoring module and the data communication module of the network terminal correspond to each other, that is, Bluetooth or WIFI is used on one side relative to the temperature module, and 4G mobile communication is used on the other side relative to the network terminal. The communication transceiver module is electrically connected to the local storage module, and the display module is electrically connected to the local storage module. The local storage module is used to store data or electronic files. The display module is used to visualize data or electronic files. S3. The network end is used to identify the temperature characteristics of different areas, and to generate electronic files based on the accumulated identification results. S4. The data communication module, AI classification model, calculation module, label module, and electronic file generation module are respectively connected to the data storage module. S4.1. The data communication module is used to communicate with the mobile end. In this embodiment, the data communication module preferably adopts 4G mobile communication. After receiving the temperature data sent by the mobile end, the data communication module transmits the temperature data to the data storage module. The data storage module is used to store data and electronic files. The electronic file generation module is used to count the cumulative results of the temperature characteristics of different areas, and then generate an electronic file of the corresponding statistical results and transmit it to the data storage module.After the data storage module receives the corresponding electronic file, it sends the electronic file to the display module of the mobile terminal through the data communication module for visual display, so that the doctor can timely understand the patient's knee osteoarthritis syndrome data in order to make a judgment.

[0089] Embodiment 3 See also Fig. 9 , Fig. 9 Schematic diagram of the structure of an electronic device disclosed in an embodiment of the present invention. The electronic device may be a computer, a server, etc. Of course, in certain circumstances, it may also be a smart device such as a mobile phone, a tablet computer, a monitoring terminal, and an image acquisition device with processing functions. Fig. 9 As shown, the electronic device may include: A memory 510 storing executable program codes; a processor 520 coupled to the memory 510; The processor 520 calls the executable program code stored in the memory 510 to execute part or all of the steps in the artificial intelligence-based knee joint state recognition method in the first embodiment.

[0090] The above is a detailed introduction to the artificial intelligence-based knee joint state recognition method, system, electronic device and storage medium disclosed in the embodiments of the present invention. Specific examples are used in this article to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea; at the same time, for general technical personnel in this field, according to the idea of ​​the present invention, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A method for identifying knee joint status based on artificial intelligence, characterized in that: include: Acquire a temperature distribution map of a knee joint of a user to be identified, and divide the temperature distribution map into regions to divide the temperature distribution map into a temperature region map, wherein the temperature region map includes a plurality of regions to be identified, and each of the regions to be identified includes a plurality of temperature data points; Processing each temperature data point in a temperature region map having a plurality of regions to be identified according to a set temperature adjustment logic to obtain a processed temperature standard map; The temperature standard image is input into a pre-built multi-classification recognition model for image recognition to obtain a state classification result of the knee joint of the user to be identified, and the state classification result is displayed.

2. The method for identifying the state of a knee joint based on artificial intelligence as claimed in claim 1, characterized in that: The step of obtaining a temperature distribution diagram of a knee joint of a user to be identified includes: The temperature distribution map of the knee joint of the user to be identified is obtained by a temperature sensing component arranged at the corresponding user's knee joint; the temperature sensing component includes a temperature sensor, a temperature processor and a wireless communication module, there are multiple temperature sensors, and the multiple temperature sensors and the wireless communication modules are electrically connected to the temperature processor, the temperature sensor is used to monitor the temperature signals of different areas, and transmit the monitored temperature signals to the temperature processor, the temperature processor converts the received temperature signals into temperature data of each area after analog-to-digital conversion, signal amplification and low-pass filtering, and generates a temperature distribution map based on the temperature data, and the wireless communication module is used to transmit the temperature distribution map to the background server; Or, obtaining a first temperature distribution map of the knee joint of the user to be identified in a specific image acquisition environment by an infrared thermal imager disposed at a first position; Acquiring a second temperature distribution map of the knee joint of the user to be identified in a specific image acquisition environment by using an infrared thermal imager disposed at a second position; A third temperature distribution map of the knee joint of the user to be identified in a specific image acquisition environment is obtained by using an infrared thermal imager set at a third position; and when acquiring the temperature distribution map, the specific image acquisition environment is an independent acquisition space, the room temperature is 24°C, and the relative humidity is 50% to 60%.

3. The method for identifying the state of a knee joint based on artificial intelligence as claimed in claim 2, characterized in that: The temperature distribution map is divided into regions to divide the temperature distribution map into a temperature region map, wherein the temperature region map includes a plurality of regions to be identified, and each of the regions to be identified includes a plurality of temperature data points, including: Inputting the temperature distribution map into a pre-built part recognition model to obtain the specific location of the patella feature, the specific location of the popliteal crease feature, and the specific location of the Zusanli feature in the temperature distribution map; Determining the corresponding patellar bottom position, patellar tip position, patellar lateral edge position and patellar medial edge position based on the specific position of the patellar feature; Determine the corresponding positions of both ends of the popliteal crease based on the specific position of the popliteal crease feature, and determine the corresponding Zusanli position based on the specific position of the Zusanli feature; Generate 4 corresponding horizontal lines according to the position above the patellar base, the patellar base, the patellar tip and the Zusanli position along the transverse direction of the temperature distribution map; Four corresponding vertical lines are generated according to the longitudinal positions of the lateral edge of the patella, the medial edge of the patella and the two ends of the popliteal transverse line along the temperature distribution diagram; and the temperature distribution diagram is divided into a temperature area diagram according to the intersection of the four corresponding horizontal lines and the four corresponding vertical lines, wherein the temperature area diagram includes nine areas to be identified, and each of the areas to be identified includes multiple temperature data points.

4. The method for identifying the state of a knee joint based on artificial intelligence as claimed in claim 3, characterized in that: The processing of each temperature data point in the temperature region map having multiple regions to be identified according to a set temperature adjustment logic to obtain a processed temperature standard map includes: Determine each area to be identified in the temperature area map, and obtain each temperature data point in each area to be identified; According to each temperature data point in each area to be identified, statistics are performed to obtain temperature mean data and temperature standard deviation data of each area to be identified; The temperature data points, the temperature mean data and the temperature standard deviation data in each area to be identified are calculated according to the temperature adjustment formula to obtain an updated temperature standard map; wherein the temperature adjustment formula is: ,in, are the parameters of the original temperature data points, is the temperature mean data of the corresponding area to be identified, is the temperature standard deviation data of the corresponding area to be identified.

5. The method for identifying the state of a knee joint based on artificial intelligence as claimed in claim 3, characterized in that: The identification method further includes: The maximum temperature, minimum temperature and average temperature of the nine areas to be identified were determined by heat map analysis method; After inputting the temperature standard image into a pre-built multi-classification recognition model for image recognition to obtain a state classification result of the user's knee joint to be identified, the method further includes: Determine the prior probability of the corresponding knee joint state in the state classification result; Determine the highest temperature, the lowest temperature and the average temperature of the nine areas to be identified to determine the conditional probability that the corresponding user to be identified is in the corresponding knee joint state; The prior probability and conditional probability of the corresponding knee joint state are input into the Bayesian model to determine the posterior probability result of the corresponding knee joint state, and the posterior probability result is output.

6. The method for identifying the state of a knee joint based on artificial intelligence as claimed in claim 1, characterized in that: The multi-classification recognition model is constructed by the following steps: Constructing an initial training model, the initial training model includes an input layer, a first hidden layer, a second hidden layer and an output layer, wherein the input layer includes nine input features, and the nine input features correspond to the temperature parameters of nine areas of the knee joint respectively; the first hidden layer includes 64 neurons, and the activation function is ReLU, the second hidden layer includes 32 neurons, and the activation function is ReLU, and the output layer includes 3 neurons, and the activation function is softmax; Acquire an infrared sample training set, wherein the sample training set includes an infrared sample image of a knee joint; and divide the infrared sample training set of the knee joint into a training set, a validation set, and a test set; Performing regional division on the infrared sample images of the knee joint in the training set to determine the temperature images of each region in the infrared sample images of the knee joint; Inputting the infrared sample images of the knee joint after the region division into the initial training model with completed parameter configuration for training and parameter configuration adjustment until the corresponding initial training model meets the training requirements, and the trained initial training model is used as a multi-classification recognition model; After the corresponding initial training model meets the training requirements and the trained initial training model is used as the multi-classification recognition model, the following further includes: Use the trained multi-classification recognition model to predict the infrared sample images of the knee joint in the test set, and obtain the predicted category to which each infrared sample image of the knee joint belongs; Generate a category confusion matrix based on the prediction results and true annotations of the multi-classification recognition model; the rows of the category confusion matrix represent true categories, and the columns represent predicted categories; Calculating overall performance indicators of the target detection model according to the category confusion matrix, wherein the overall performance indicators include accuracy, precision, recall rate and F1 score; The multi-classification recognition model is adjusted and optimized according to the analysis results of the category confusion matrix.

7. The method for identifying the state of a knee joint based on artificial intelligence as claimed in claim 1, characterized in that: The step of obtaining a temperature distribution diagram of a knee joint of a user to be identified includes: Acquire a sequence temperature distribution map of the knee joint of the user to be identified when the leg moves from a first position to a second position; there are multiple sequence temperature distribution maps; The joint identification method further comprises: The image sequence features in the multiple sequence temperature images are extracted by using a convolutional neural network. The image sequence features are input into the LSTM model for analysis to determine the corresponding joint state results.

8. A knee joint state recognition system based on artificial intelligence, characterized in that: include: An acquisition module is used to acquire a temperature distribution map of a knee joint of a user to be identified, and divide the temperature distribution map into regions to divide the temperature distribution map into a temperature region map, wherein the temperature region map includes a plurality of regions to be identified, and each of the regions to be identified includes a plurality of temperature data points; Temperature processing module: used to process each temperature data point in the temperature region diagram having multiple regions to be identified according to the set temperature adjustment logic to obtain a processed temperature standard diagram; Recognition module: used for inputting the temperature standard image into a pre-built multi-classification recognition model for image recognition to obtain the state classification result of the knee joint of the user to be identified, and displaying the state classification result.

9. An electronic device, characterized in that: include: A memory storing executable program code; a processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the artificial intelligence-based knee joint state recognition method described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, wherein the computer program enables a computer to execute the artificial intelligence-based knee joint state identification method according to any one of claims 1 to 7.

Images