Artificial intelligence-based knee joint X-ray image classification method
Through the artificial intelligence-based knee X-ray image classification method, the characteristic parameters of knee X-ray image are extracted and analyzed, and the problem of insufficient subjectivity and accuracy in traditional evaluation methods is solved, higher diagnostic accuracy and consistency are achieved, and the scope of analysis applications is expanded.
Patent Information
- Application Number
- CN202411699582.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-26
- Publication Date
- 2025-03-11
AI Technical Summary
Traditional knee X-ray evaluation relies on the subjective evaluation of the clinician, and the accuracy is influenced by the physician's experience and may be low, resulting in insufficient consistency and accuracy of the diagnosis.
A method of knee X-ray image classification based on artificial intelligence is proposed. By acquiring and preprocessing knee X-ray images, multiple feature parameters are extracted, differentiated matrix of feature parameters are calculated, main parameters and their weights are determined, and the neural network is input for convolutional operation, and target parameters are output for image classification.
Through automated and objective evaluation methods, the diagnostic accuracy and consistency of knee X-ray images are improved, the error of subjective judgment is reduced, the status of the knee joint can be reflected more accurately, and the application range of knee X-ray image analysis is broadened.
Smart Images
Figure CN119600356A_ABST
Abstract
Description
Technical Field
[0001] The invention proposes a knee joint X-ray image classification method based on artificial intelligence, and relates to the technical field of image classification. Background Art
[0002] The knee joint is composed of bones, cartilage, ligaments, meniscus and other tissues. It is one of the largest and most complex joints in the human body and plays an extremely important role in maintaining normal human activities. With the continuous improvement of living standards, people's demand for sports and entertainment is also increasing, and the incidence of knee sports injuries is also increasing. Knee sports injuries are mostly concentrated in soft tissues such as cartilage, ligaments, and meniscus. If diagnosis and treatment are not timely, it will cause the stability of the knee joint to decrease and accelerate the occurrence of knee osteoarthritis.
[0003] With the advancement of medical image acquisition technology, image scanning equipment can collect data at a faster speed and higher resolution. However, the image interpretation process, especially the interpretation of medical images, is mostly performed by doctors, which is limited by doctors' subjectivity, cognitive differences and fatigue.
[0004] Application of deep learning in medical imaging: Deep learning, especially convolutional neural network (CNN), has become the leading machine learning tool in the field of computer vision. It can automatically obtain intermediate and high-level abstract features from raw data (images), which is very effective for target recognition and positioning. Deep learning technology has shown promising results in interpreting X-rays, which can improve the accuracy of diagnosis and generate uniformly classified fracture records for studying treatment strategies for different types of fractures.
[0005] However, traditional knee X-ray assessment relies on the clinician's subjective assessment, and the accuracy varies greatly depending on the clinician's experience and can be particularly low. Therefore, the development of automated and objective assessment methods is crucial to improve the consistency and accuracy of diagnosis. Summary of the invention
[0006] In order to solve the above technical problems, the present invention proposes a knee joint X-ray image classification method based on artificial intelligence, comprising the following steps:
[0007] Step 1: Acquire a knee joint X-ray image and pre-process the acquired knee joint X-ray image;
[0008] Step 2: extracting multiple characteristic parameters of the knee joint X-ray image;
[0009] Step 3: Calculate the difference values between the multiple feature parameters to form a differential matrix of the multiple feature parameters of the knee joint X-ray image;
[0010] Step 4: Determine multiple feature parameters with difference values greater than zero in the differentiation matrix as main parameters, and calculate the weight of each main parameter;
[0011] Step 5: Input the multiple main parameters and corresponding weights calculated in step 4 into the neural network for convolution operation, and output the target parameters;
[0012] Step 6: Classify multiple knee joint X-ray images based on the target parameters.
[0013] In a preferred embodiment, in step 3, for the n characteristic parameters of the knee joint X-ray image: X1, X2, ..., X i ...,X j ...,X n ; Among them, X i ,X j For the i-th, j-th feature parameter of the knee joint X-ray image, calculate the difference matrix of the feature parameter:
[0014] C ij =E[(X i -E[X i ])-(X j -E[X j ])];
[0015] Where E represents the standard value of the variable in [], and the element C in the differentiation matrix ij Represents the characteristic parameter X i With X j The difference between .
[0016] In the preferred embodiment, in step 4, multiple feature parameters with difference values greater than zero in the differentiation matrix are determined as main parameters, and the weight of the kth main parameter is set to w k , by calculating the change in weight of each main parameter Δw k to update the weights, and then adjust the weights according to the changed values.
[0017] In a preferred embodiment, the step 4 comprises the following steps:
[0018] S41. Let the weight of the kth main parameter be w k , all weights w k Initialized to 0.5;
[0019] S42. For each weight w k , calculate the change value Δw of the weight k ;
[0020] S43. Update each weight using the following formula. The updated weight is w k0 :
[0021] w k0 =max(0,min(w k +Δw k ,1))
[0022] Among them, Δw k is the weight w k The change value of
[0023] S44, iteration: repeat steps S42 and S43 until the maximum number of iterations is reached.
[0024] In a preferred embodiment, the multiple main parameters and corresponding weights calculated in step 4 are input into the neural network for convolution, and the target parameter is output. The predicted value y output by the convolution layer of the neural network is t As shown below:
[0025] y t =x t *w t +b
[0026] The weight w corresponding to the tth main parameter t As the convolution kernel, with the tth main parameter x t Perform convolution, b is the correction term.
[0027] In a preferred embodiment, the predicted value y t Input to the matching layer of the neural network, the matching layer calculates the predicted value y t Matching value L(t) with reference value Z:
[0028]
[0029] The matching layer outputs the main parameter corresponding to the predicted value with the smallest matching value L(t) as the target parameter for classifying multiple knee joint X-ray images.
[0030] In a preferred embodiment, the step six comprises the following steps:
[0031] S61, calculating the variance of the target parameters of any two knee joint X-ray images to form a target parameter sequence;
[0032] S62, finding two knee joint X-ray images with the smallest variance, and merging them into an initial cluster;
[0033] S63, calculating the mean of the target parameters of the initial clustered knee joint X-ray images, and updating the target parameter sequence;
[0034] S64, calculating the variance of the target parameters of other knee joint X-ray images that have not been clustered and the mean of the target parameters of the initial cluster, finding the unclustered knee joint X-ray image with the smallest variance, and adding it to the initial cluster to form a new cluster;
[0035] S65, continuously repeating steps S64 and S65, each time replacing the previous target parameter with the mean value of the target parameter of the new cluster, until the number of knee joint X-ray images in the cluster reaches a preset number.
[0036] In a preferred embodiment, the plurality of characteristic parameters include: knee joint angle, cartilage thickness, anterior cruciate ligament length, and posterior cruciate ligament length.
[0037] Compared with the prior art, the present invention has the following beneficial technical effects:
[0038] 1: Classification method of knee joint X-ray images based on artificial intelligence. First, obtain the knee joint X-ray image and pre-process the acquired image; extract multiple feature parameters of the knee joint X-ray image. By extracting multiple feature parameters, the state of the knee joint can be described from different angles. Compared with a single feature, it can more comprehensively reflect the actual situation of the knee joint and avoid the errors that may be caused by relying solely on human observation and subjective judgment.
[0039] 2: The present invention calculates the differentiation matrix of multiple characteristic parameters of knee joint X-ray images, calculates the difference values between multiple characteristic parameters; determines multiple characteristic parameters with difference values greater than zero in the differentiation matrix as main parameters, calculates the weight of each main parameter, inputs the calculated multiple main parameters and corresponding weights into the neural network for convolution, and outputs the target parameters; and classifies multiple knee joint X-ray images based on the target parameters. By calculating the differentiation matrix of multiple characteristic parameters and determining the characteristic parameters with difference values greater than zero as the main parameters, the screening of many features is achieved, which helps to focus on those features that are truly distinguishable and important in the knee joint X-ray images, avoiding the increase in analysis complexity and possible interference caused by too many irrelevant or redundant features, so that subsequent processing can be more accurately carried out around the key factors that have a substantial impact on the state of the knee joint.
[0040] 3: Calculate weights for each main parameter, further refining the consideration of the importance of different main features. Different main parameters may contribute differently to accurately describing the condition of the knee joint and performing operations such as classification. Assigning weights can more reasonably reflect this difference, so that in subsequent neural network processing and other links, each main parameter can play a role based on its actual importance, thereby more accurately characterizing the various states of the knee joint.
[0041] 4: The whole set of processes described in this invention, including calculating characteristic parameters, determining main parameters and weights, and using neural network classification, provides a standardized method for knee joint X-ray image analysis. In addition to direct applications in the traditional field of medical diagnosis, it can also be expanded to other related scenarios, such as tracking and monitoring the recovery of the knee joint during rehabilitation treatment, and evaluating the effects of different treatment schemes on knee joint images in drug development, further broadening the application scope of knee joint X-ray image analysis and providing more extensive support for medical research and practice related to the knee joint. BRIEF DESCRIPTION OF THE DRAWINGS
[0042] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.
[0043] Figure 1 A schematic diagram of the process of the knee joint X-ray image classification method based on artificial intelligence of the present invention;
[0044] Figure 2 A processed knee joint X-ray image saved by the present invention;
[0045] Figure 3 A schematic diagram of the steps of calculating the weight of each main parameter in step 4 of the present invention;
[0046] Figure 4 A schematic diagram of the process of classifying multiple knee joint X-ray images based on target parameters according to the present invention. DETAILED DESCRIPTION
[0047] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.
[0048] In the drawings of the specific embodiments of the present invention, in order to better and more clearly describe the working principles of the various components in the system, the connection relationship of the various parts in the device is shown, which only clearly distinguishes the relative position relationship between the various components, and cannot constitute a limitation on the signal transmission direction, connection sequence and structural size, size and shape of each part within the component or structure.
[0049] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.
[0050] like Figure 1 FIG. 1 is a flow chart of a knee joint X-ray image classification method based on artificial intelligence according to the present invention. The knee joint X-ray image classification method comprises the following steps:
[0051] Step 1: Obtain knee joint X-ray images and preprocess the acquired images.
[0052] 1. Image acquisition:
[0053] X-ray images of the knee joint are taken by medical imaging equipment (such as an X-ray machine) to obtain clear images.
[0054] 2. Image transmission:
[0055] The acquired knee joint X-ray images are transferred from the medical imaging device to the computer system for further processing and analysis.
[0056] 3. Knee joint X-ray image preprocessing:
[0057] Denoising: Use a filter to remove noise in the knee joint X-ray image. Preferably, a filter such as a Gaussian filter or a median filter is used.
[0058] Contrast enhancement: Improve the contrast of the knee joint X-ray image to make the knee joint structure more obvious, which is convenient for subsequent feature extraction and analysis. Preferably, a contrast enhancement method such as histogram equalization and linear stretching is used.
[0059] Resizing: Remove irrelevant parts of the knee joint X-ray image, retain the main area of the knee joint, and uniformly scale all knee joint X-ray images to the same resolution to reduce the amount of calculation.
[0060] Normalization: Adjusting the pixel value range to a fixed interval (such as between 0 and 1) helps model training and prediction.
[0061] 4. Preservation and Recording:
[0062] After completing the preprocessing, save the processed knee joint X-ray image, such as Figure 2 as shown, and record all preprocessing steps taken.
[0063] Step 2: Extract various feature parameters of knee joint X-ray images.
[0064] Obtain characteristic parameters of knee joint X-ray images, including:
[0065] Knee Angle: The change in knee angle is critical to the normal function of the knee. The biomechanical properties of the knee joint change in different flexion states.
[0066] Cartilage thickness: The thickest part of the patellar articular cartilage can reach 5mm. Cartilage thickness is very important for joint cushioning and reducing friction.
[0067] Ligament length:
[0068] Anterior cruciate ligament (ACL): originates from the front of the intercondylar eminence of the tibia, and attaches to the medial surface of the lateral femoral condyle obliquely posteriorly and superiorly; posterior cruciate ligament (PCL): originates from the back of the intercondylar eminence of the tibia, and attaches to the lateral surface of the medial femoral condyle obliquely and anteriorly and superiorly; medial collateral ligament (MCL): located on the medial side of the knee joint, originates from the medial epicondyle of the femur, and ends downward at the medial surface of the medial tibial condyle; lateral collateral ligament (LCL): located on the outside of the knee joint, originates from the lateral epicondyle of the femur, and ends at the head of the fibula.
[0069] Muscle strength: Knee flexion and extension are primarily produced by the hamstrings and quadriceps. Muscles such as the quadriceps, sartorius, popliteus, gracilis, biceps femoris, semitendinosus, and semimembranosus are essential for maintaining knee stability.
[0070] Since different characteristic parameters have different dimensions and magnitudes, the data needs to be standardized to ensure that each characteristic parameter has the same importance in the analysis.
[0071] Step 3: Calculate the difference matrix of various feature parameters of the knee joint X-ray image, and calculate the difference values between the various feature parameters.
[0072] Multiple feature parameters of each knee joint X-ray image are extracted, and the difference matrix of the multiple feature parameters is calculated to count the differences between the multiple feature parameters.
[0073] The differentiation matrix is a matrix used to represent the relationship between multidimensional feature parameters. The values on the diagonal of the differentiation matrix represent the difference distance of the same feature parameters, that is, 0, and the values on the off-diagonal represent the difference distance between different feature parameters.
[0074] For a set of feature parameters of a knee joint X-ray image:
[0075] X1,X2,...,X i ...,X j ...,X n , X i ,X jis the i-th, j-th characteristic parameter of the knee joint X-ray image. For example, X1 is the knee joint angle; X2 is the cartilage thickness; X3 is the anterior cruciate ligament length, etc.
[0076] Let the differentiation matrix C be an n×n matrix, where each element C ij Represents the characteristic parameter X i With X j The difference between .
[0077] C ij =E[(X i -E[X i ])-(X j -E[X j ])];
[0078] Where E represents the standard value of the variable in [].
[0079] It should be explained that the standard value is a measure of the variables within [].
[0080] Specifically, the standard value can be an expected value. The expected value is the average value that a variable may take under a large number of repeated experiments. The expected value reflects the average expected results under various possibilities.
[0081] The standard value can also be the mean. The mean is the arithmetic average of a set of data, which is obtained by adding all the data values and then dividing by the number of data.
[0082] The expected value or mean is a standard value that quantitatively describes the characteristic significance of a variable under statistical data, which can better understand and analyze the relationship between variables.
[0083] In the differentiation matrix C, C ij =C ji .
[0084] If C ij >0, then the characteristic parameter X j With X j The correlation is large, that is, X j The increase in X is often accompanied by j increase.
[0085] If C ij <0, then the characteristic parameter X j With X j The correlation is small, that is, X j An increase in one variable is often accompanied by a decrease in another variable.
[0086] Step 4: Determine multiple feature parameters whose difference values in the differentiation matrix are greater than zero as main parameters, and calculate the weight of each main parameter.
[0087] Determine multiple feature parameters whose difference values in the differentiation matrix are greater than zero as the main parameters, and set the weight of the kth main parameter to be w k The weight update method calculates the change Δw of the loss function with respect to the weight of each main parameter k to update the weights, and then adjust the weights according to the changed values.
[0088] Weight is a key parameter. Weight can reflect the importance of different feature parameters. Specifically, Figure 3 As shown, adjusting the weight specifically includes the following steps:
[0089] S41. Initialize weight: Let the weight of the kth main parameter be w k , all weights w k is initialized to 0.5. This initialization helps ensure that the change value is not too small, thereby avoiding slow learning.
[0090] S42. Calculate the weight change value: For each weight w k , calculate the change value of the weight Δw k The change value indicates the rate of change of the weight.
[0091] S43. Update weights: Use the following formula to update each weight. The updated weight is w k0
[0092] w k0 =max(0,min(w k +Δw k ,1))
[0093] Among them, Δw k is the weight w k This formula ensures that the weights remain in the range of [0,1] after each update. Limiting the weights to the range of [0,1] helps maintain the stability of the weights. If the weights exceed this range, outliers will appear in subsequent calculations, which will affect the performance of the entire image classification.
[0094] S44, iteration: repeat steps S42 and S43 until a stop condition is met, such as the change value is small enough or the maximum number of iterations is reached.
[0095] Step 5: Input the multiple main parameters and corresponding weights calculated in step 4 into the neural network for convolution and output the target parameters.
[0096] The multiple main parameters and corresponding weights calculated in step 4 are input into the convolution layer of the neural network for convolution, and the target parameters are output. The model of the neural network is as follows:
[0097] yt =x t *w t +b
[0098] Wherein, t=1...T, T is the total number of the multiple main parameters calculated in step 4; x t is the tth main parameter, and the weight corresponding to the tth main parameter is used as the convolution kernel, which is combined with the tth main parameter x t Perform convolution, b is the correction term.
[0099] For the predicted value y output by the convolutional layer of the neural network t , the predicted value y t Input to the matching layer of the neural network, the matching layer calculates the predicted value y t Matching value L(t) with reference value Z:
[0100]
[0101] The main parameter corresponding to the predicted value with the smallest output matching value L(t) is used as the target parameter for subsequent classification of multiple knee joint X-ray images.
[0102] Step 6: Classify multiple knee joint X-ray images based on target parameters.
[0103] Classification calculation is performed on multiple knee joint X-ray images based on target parameters, such as Figure 4 The specific classification steps are as follows:
[0104] S61. Calculate the variance of the target parameters of any two knee joint X-ray images.
[0105] The target parameters of all knee joint X-ray images are extracted to form a target parameter sequence. For any two knee joint X-ray images, the variance of their target parameters is calculated respectively.
[0106] S62. Initial clustering.
[0107] Among the variance calculation results of the target parameters of all pairwise knee joint X-ray images, find the two knee joint X-ray images with the smallest variance and merge them into an initial cluster, which means that the two knee joint X-ray images are most similar in the target parameters.
[0108] S63, calculating the mean of the target parameters of the initial clustered knee joint X-ray images, and updating the target parameter sequence.
[0109] The mean of the target parameters of the two knee joint X-ray images in the initial cluster is calculated, and the mean is used to replace the target parameters of the original two knee joint X-ray images in the target parameter sequence.
[0110] S64, extended clustering.
[0111] The variance of the target parameters of other unclustered knee X-ray images and the mean of the target parameters of the initial cluster is calculated, and the unclustered knee X-ray image with the smallest variance is found and added to the initial cluster to form a new cluster.
[0112] S65, loop operation to form the final clustering.
[0113] The above steps S64 and S65 are continuously repeated, each time replacing the target parameters in the previous target parameter sequence with the mean of the target parameters of the new cluster, and then continuing to look for the unclustered knee joint X-ray image with the smallest variance with the mean of the new target parameters to add to the current cluster.
[0114] The operation steps are continuously repeated until the number of knee joint X-ray images in the cluster reaches a preset number. At this time, the knee joint X-ray images in the clustering results have similar target parameters, that is, the features reflected by these knee joints in the X-ray images are similar and can be considered to be the same category of knee joint conditions.
[0115] In one embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above method embodiments when executing the computer program.
[0116] In one embodiment, a computer-readable storage medium is provided, storing a computer program, which implements the steps in the above method embodiments when executed by a processor.
[0117] In one embodiment, a computer program product or computer program is provided, the computer program product or computer program includes computer instructions, the computer instructions are stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device performs the steps in the above-mentioned method embodiments.
[0118] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.
[0119] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0120] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.
Claims
1. A knee joint X-ray image classification method based on artificial intelligence, characterized in that: The steps include: Step 1: Acquire a knee joint X-ray image and pre-process the acquired knee joint X-ray image; Step 2: extracting multiple characteristic parameters of the knee joint X-ray image; Step 3: Calculate the difference values between the multiple feature parameters to form a differential matrix of the multiple feature parameters of the knee joint X-ray image; Step 4: Determine multiple feature parameters with difference values greater than zero in the differentiation matrix as main parameters, and calculate the weight of each main parameter; Step 5: Input the multiple main parameters and corresponding weights calculated in step 4 into the neural network for convolution operation, and output the target parameters; Step 6: Classify multiple knee joint X-ray images based on the target parameters.
2. The knee joint X-ray image classification method according to claim 1, characterized in that: In step 3, for the n characteristic parameters of the knee joint X-ray image: X1, X2, ..., X i ...,X j ...,X n ; Among them, X i ,X j For the i-th, j-th feature parameter of the knee joint X-ray image, calculate the difference matrix of the feature parameter: C ij =E[(X i -E[X i ])-(X j -E[X j ])]; Where E represents the standard value of the variable in [], and the element C in the differentiation matrix ij Represents the characteristic parameter X i With X j The difference between .
3. The knee joint X-ray image classification method according to claim 2, characterized in that: In step 4, multiple feature parameters with difference values greater than zero in the differentiation matrix are determined as main parameters, and the weight of the kth main parameter is set to w k , by calculating the change in weight of each main parameter Δw k to update the weights, and then adjust the weights according to the changed values.
4. The knee joint X-ray image classification method according to claim 3, characterized in that: The step 4 comprises the following steps: S41. Let the weight of the kth main parameter be w k , all weights w k Initialized to 0.5; S42. For each weight w k , calculate the change value Δw of the weight k ; S43. Update each weight using the following formula. The updated weight is w k0 : In k0 =max(0,min(in k +Δw k ,1)) Among them, Δw k is the weight w k The change value of S44, iteration: repeat steps S42 and S43 until the maximum number of iterations is reached.
5. The knee joint X-ray image classification method according to claim 4, characterized in that: The multiple main parameters and corresponding weights calculated in step 4 are input into the convolution layer of the neural network for convolution. The predicted value y output by the convolution layer of the neural network is t As shown below: y t =x t *w t +b The weight w corresponding to the tth main parameter t As the convolution kernel, with the tth main parameter x t Perform convolution, b is the correction term.
6. The knee joint X-ray image classification method according to claim 5, characterized in that: The predicted value y t Input to the matching layer of the neural network, the matching layer calculates the predicted value y t Matching value L(t) with reference value Z: The matching layer outputs the main parameter corresponding to the predicted value with the smallest matching value L(t) as the target parameter for classifying multiple knee joint X-ray images.
7. The knee joint X-ray image classification method according to claim 4, characterized in that: The step six comprises the following steps: S61, calculating the variance of the target parameters of any two knee joint X-ray images to form a target parameter sequence; S62, finding two knee joint X-ray images with the smallest variance, and merging them into an initial cluster; S63, calculating the mean of the target parameters of the initial clustered knee joint X-ray images, and updating the target parameter sequence; S64, calculating the variance of the target parameters of other knee joint X-ray images that have not been clustered and the mean of the target parameters of the initial cluster, finding the unclustered knee joint X-ray image with the smallest variance, and adding it to the initial cluster to form a new cluster; S65, continuously repeating steps S64 and S65, each time replacing the previous target parameter with the mean value of the target parameter of the new cluster, until the number of knee joint X-ray images in the cluster reaches a preset number.
8. The knee joint X-ray image classification method according to claim 1, characterized in that: The multiple characteristic parameters include: knee joint angle, cartilage thickness, anterior cruciate ligament length, and posterior cruciate ligament length.