Artificial intelligence auxiliary diagnosis method and system for knee joint

Abstract

The invention provides an artificial intelligence auxiliary diagnosis method and system for a knee joint, and the method comprises the steps: synchronously collecting a multi-angle dynamic X-ray image and six-degree-of-freedom contact force distribution data in an articular cavity, dynamically tracking a patella sliding track, and generating a mechanical phase track; a personalized knee joint model is constructed in combination with the biological parameters of the lower limbs of the patient, the mechanical phase trajectory and the contact force data vector are superposed to generate a dynamic mechanical fingerprint, and then the dynamic mechanical fingerprint is mapped into a biomechanical attribute map; electromyographic signal time-frequency characteristics and tibia rotation angles in the movement process are collected, an electromyographic activation mode is constructed, and the electromyographic activation mode and a biomechanical attribute map are subjected to characteristic fusion to form a joint dynamics comprehensive contour; analyzing a stress wave conduction path by using a joint anomaly dynamic propagation model to identify a compensatory injury mode; a diagnosis map is generated according to the recognition result, a preset biomechanical constraint condition is matched to correct the diagnosis map, and an individualized rehabilitation strategy is formulated. The treatment effect is improved, and the recovery speed of a patient is increased.

Description

Technical Field

[0001] The embodiments of the present application relate to the cross - technical field of biomedical engineering and artificial intelligence, and particularly to an artificial intelligence - assisted diagnosis method and system for knee joints. Background Art

[0002] With the aggravation of the global population aging and the increase in the participation in sports activities, the incidence of knee - related diseases has been rising year by year, and the demand for early accurate diagnosis and personalized treatment plans has become increasingly urgent. Especially in the athlete and elderly populations, knee injuries caused by high - intensity sports or the natural aging process not only seriously affect the quality of life of patients but may also lead to long - term health problems. To better address this challenge, the medical field requires a technical means that can integrate multi - source data (such as dynamic X - ray images, in - joint mechanical distribution, surface electromyography signals, etc.) and perform efficient comprehensive analysis. This technology should be able to deeply understand the complex biomechanical characteristics of the knee joint, provide accurate injury pattern recognition, and formulate rehabilitation strategies that meet individual differences to improve the treatment effect and quality of life.

[0003] The existing technology has tried to use advanced biomechanical modeling and computer - aided analysis methods to evaluate the functional state of the knee joint. For example, by combining high - resolution medical images with dynamic X - ray image sequences, using finite - element analysis to construct a three - dimensional knee joint model, and simulating the stress distribution under different load conditions. In addition, by collecting surface electromyography signals and gait analysis to capture muscle activities and joint movement patterns, it provides more comprehensive data support for the biomechanical evaluation of the knee joint. The application of these technologies has significantly improved the understanding of knee joint injuries and their diagnostic capabilities, making the treatment plans for specific cases more scientific and reasonable.

[0004] However, despite the above - mentioned progress, several key challenges still exist in practical applications: First, the problem of insufficient data fusion still exists, making it difficult to effectively integrate various data types such as imaging, mechanics, and electrophysiology, and unable to fully reflect the real situation of the knee joint; second, the existing models do not fully consider individual differences, with limited personalization, which affects the accuracy of diagnosis; finally, the level of automation and intelligence still needs to be improved, and most systems still require manual intervention for result interpretation and rehabilitation strategy formulation. Summary of the Invention

[0005] The embodiments of the present application provide an artificial intelligence - assisted diagnosis method and system for knee joints to solve the problems of poor treatment effect and slow recovery speed of patients in the existing technology.

[0006] In a first aspect, the embodiments of the present application provide an artificial intelligence - assisted diagnosis method for knee joints, including:

[0007] Receive a sequence of multi - angle dynamic X - ray images and six - degree - of - freedom contact force distribution data within the joint cavity, dynamically track the sequence of multi - angle dynamic X - ray images and combine with the six - degree - of - freedom contact force distribution data within the joint cavity, and encode the patellar slip trajectory to generate mechanical phase trajectory data;

[0008] Collect the lower - limb biological parameters of the patient to construct a personalized knee joint model, vectorially superimpose the mechanical phase trajectory data and the six - degree - of - freedom contact force distribution data within the joint cavity to generate a dynamic mechanical fingerprint, and map the dynamic mechanical fingerprint to a biomechanical property map through an inverse dynamics solver;

[0009] Collect the frequency characteristics of the surface electromyogram signal and the tibial rotation angle during the patient's movement, construct an electromyogram activation pattern marked with movement stages, and perform feature fusion on the biomechanical property map and the electromyogram activation pattern to generate a comprehensive joint dynamics profile;

[0010] Input the comprehensive joint dynamics profile into a pre - constructed joint abnormal dynamics propagation model, use the stress - wave conduction equation to calculate the wave propagation path in the stress - concentrated area of the meniscus and the spatio - temporal evolution trajectory of the deformation energy field of the cruciate ligament, and identify the compensatory injury pattern by analyzing the interference effect between the spatio - temporal evolution trajectory and the wave propagation path;

[0011] According to the compensatory injury pattern, combine the biomechanical property map, the electromyogram activation pattern, and the comprehensive joint dynamics profile to generate a diagnostic map, match the biomechanical constraint conditions in a preset motion feature library, and adaptively correct the diagnostic map to generate an individualized rehabilitation strategy.

[0012] Optionally, the step of inputting the comprehensive joint dynamics profile into a pre - constructed joint abnormal dynamics propagation model, using the stress - wave conduction equation to calculate the wave propagation path in the stress - concentrated area of the meniscus and the spatio - temporal evolution trajectory of the deformation energy field of the cruciate ligament, and identifying the compensatory injury pattern by analyzing the interference effect between the spatio - temporal evolution trajectory and the wave propagation path includes:

[0013] Perform hierarchical tensor decomposition on the comprehensive joint dynamics profile to obtain biomechanical tensor components and kinematic tensor components, construct a meniscus stress propagation path based on the stress - wave guide model for the biomechanical tensor components, and calculate the dynamic correlation parameters of the wave attenuation rate and the stress concentration coefficient in the meniscus stress propagation path;

[0014] Establish a ligament energy field evolution model of a visco - elastic deformation field for the kinematic tensor components, and extract the spatio - temporal gradient and the coordinates of the energy residence nodes of the ligament energy field evolution model;

[0015] By means of the dynamic correlation parameters, the spatio-temporal gradient, and the energy residence node coordinates, an interference field equation is established through the dynamic coupling of a stress waveguide model and a viscoelastic deformation field. According to the interference field equation, the wave field interference intensity distribution between the meniscus stress propagation path and the energy residence node coordinates is calculated. A compensatory damage mode vector is generated based on the three-dimensional coordinates of the abnormal resonance region and the energy overflow threshold in the wave field interference intensity distribution;

[0016] Perform modal matching between the compensatory damage mode vector and the standard motion chain parameters in the motion feature library, and perform frequency-domain weighted adjustment on the compensatory damage mode vector based on the matching result to generate a compensatory damage mode.

[0017] Optionally, the steps of establishing an interference field equation through the dynamic coupling of a stress waveguide model and a viscoelastic deformation field by means of the dynamic correlation parameters, the spatio-temporal gradient, and the energy residence node coordinates, calculating the wave field interference intensity distribution between the meniscus stress propagation path and the energy residence node coordinates according to the interference field equation, and generating a compensatory damage mode vector based on the three-dimensional coordinates of the abnormal resonance region and the energy overflow threshold in the wave field interference intensity distribution include:

[0018] Perform parametric reconstruction of the wave equation for the meniscus stress propagation path, extract the wave phase delay parameter and the amplitude attenuation rate in the meniscus stress propagation path, decompose the energy residence node coordinates into an energy density gradient field, and generate an energy density gradient vector and a residence time coefficient;

[0019] Based on the wave phase delay parameter and the energy density gradient vector, construct a dynamic coupling kernel function. Through the dynamic coupling kernel function combined with spatial convolution operation, generate a spatial grid mapping of the wave field interference intensity distribution, and mark the grid cells with energy density peaks higher than the overflow threshold in the spatial grid mapping as abnormal resonance regions;

[0020] Use the amplitude attenuation rate and the residence time coefficient to calculate the product of the energy residence time coefficient and the wave amplitude attenuation rate in the abnormal resonance region as the compensatory damage intensity factor, and extract the included angle between the three-dimensional coordinates of the abnormal resonance region and the tangent direction of the wave propagation path as the damage spatial distribution parameter. Recombine the compensatory damage intensity factor and the damage spatial distribution parameter into a tensor to generate a compensatory damage mode vector.

[0021] Optionally, the steps of constructing a dynamic coupling kernel function based on the wave phase delay parameter and the energy density gradient vector, generating a spatial grid mapping of the wave field interference intensity distribution through the dynamic coupling kernel function combined with spatial convolution operation, and marking the grid cells with energy density peaks higher than the overflow threshold in the spatial grid mapping as abnormal resonance regions include:

[0022] Perform time series normalization on the fluctuation phase delay parameter to generate a phase delay weight coefficient, and perform orthogonal basis decomposition on the energy density gradient vector to obtain a normal energy gradient component and a tangential energy diffusion rate;

[0023] Construct a two-channel coupling kernel function based on the phase delay weight coefficient and the normal energy gradient component, use the tangential energy diffusion rate as the dynamic attenuation factor of the two-channel coupling kernel function, generate a three-dimensional convolution kernel according to the dynamic attenuation factor, define the dynamic window parameter of the spatial convolution operation according to the knee joint anatomical structure, and generate an interference intensity distribution matrix through sliding integral calculation in the biomechanical spatial domain by combining the dynamic window parameter with the three-dimensional convolution kernel;

[0024] Perform multi-scale Gaussian filtering on the interference intensity distribution matrix, detect the spatial topological structure of the energy density peak in the filtered matrix, and cluster the topological units that are continuously adjacent and have a peak higher than the overflow threshold in the spatial topological structure into abnormal resonance regions.

[0025] Optionally, the constructing a two-channel coupling kernel function based on the phase delay weight coefficient and the normal energy gradient component, using the tangential energy diffusion rate as the dynamic attenuation factor of the two-channel coupling kernel function, generating a three-dimensional convolution kernel according to the dynamic attenuation factor, defining the dynamic window parameter of the spatial convolution operation according to the knee joint anatomical structure, and generating an interference intensity distribution matrix through sliding integral calculation in the biomechanical spatial domain by combining the dynamic window parameter with the three-dimensional convolution kernel includes:

[0026] Perform frequency domain transformation on the phase delay weight coefficient to generate a phase delay spectrum, extract the proportion of the main frequency band energy in the phase delay spectrum as the time modulation factor of the two-channel coupling kernel function, perform spatial Fourier transform on the normal energy gradient component to generate a normal energy wave number spectrum, and extract the high-frequency attenuation slope in the normal energy wave number spectrum as the spatial modulation factor of the two-channel coupling kernel function;

[0027] Construct the main kernel structure of the two-channel coupling kernel function based on the time modulation factor and the spatial modulation factor, introduce the logarithmic attenuation rate of the tangential energy diffusion rate as the dynamic attenuation factor of the two-channel coupling kernel function, and generate a three-dimensional convolution kernel;

[0028] Dynamically adjust the initial value of the sliding radius according to the meniscus curvature radius, define the direction vector of the convolution step size in combination with the anatomical characteristics of the ligament direction, and generate dynamic window parameters matching the knee joint biomechanical characteristics;

[0029] Perform a sliding integral calculation in the biomechanical spatial domain using the three-dimensional convolution kernel and the dynamic window parameters, and adjust the attenuation factor weight of the three-dimensional convolution kernel according to the local energy density gradient during each sliding process to generate an initial mapping of the interference intensity distribution matrix;

[0030] Perform local energy equalization processing on the initial mapping, extract the second derivative features of the energy density gradient in the equalized mapping, and adjust the sliding radius and step parameters of the three-dimensional convolution kernel through the second derivative features to generate an interference intensity distribution matrix.

[0031] Optionally, collect the frequency characteristics and tibial rotation angles of the patient's surface electromyogram signals during movement, construct an electromyogram activation pattern marked with movement phases, and perform feature fusion on the biomechanical attribute map and the electromyogram activation pattern to generate a comprehensive joint dynamics profile, including:

[0032] Perform multi-scale time-frequency decomposition on the collected surface electromyogram signals of the patient during movement, extract the energy concentration coefficient and phase synchronization index in specific frequency bands in the time-frequency domain, and generate an electromyogram time-frequency feature vector;

[0033] Collect time series data of tibial rotation angles, calculate the rotational dynamic stability coefficient of the time series data through the angle change rate, and construct an electromyogram activation pattern marked with movement phases in combination with the electromyogram time-frequency feature vector;

[0034] Extract spatial domain features from the biomechanical attribute map, obtain the cartilage contact stress distribution gradient and ligament tension change rate in the map, and generate a biomechanical feature vector;

[0035] Perform multi-modal feature alignment on the electromyogram activation pattern and the biomechanical feature vector, align the time series of movement phases through the dynamic time warping algorithm, and generate a time-synchronized multi-modal feature matrix;

[0036] Construct a comprehensive joint dynamics profile based on the multi-modal feature matrix, couple the energy concentration coefficient in the electromyogram activation pattern and the stress distribution gradient in the biomechanical feature vector through a non-linear mapping function, and generate a dynamics comprehensive profile vector.

[0037] Optionally, collect the patient's lower limb biological parameters to construct a personalized knee joint model, perform vector superposition on the mechanical phase trajectory data and the six-degree-of-freedom contact force distribution data in the joint cavity to generate a dynamic mechanical fingerprint, and map the dynamic mechanical fingerprint to a biomechanical attribute map through an inverse dynamics solver, including:

[0038] Collect the lower limb length, intercondylar width of the femur, and tibial plateau inclination angle of the patient as lower limb biomechanical parameters, and construct a parametric knee joint mesh model based on the lower limb biomechanical parameters. The parametric knee joint mesh model includes the curvature radius of the meniscus and the attachment point coordinates of the cruciate ligament that match the anatomical structure;

[0039] Perform trajectory smoothing processing on the mechanical phase trajectory data, extract the patellar slip velocity vector and acceleration change rate in the mechanical phase trajectory, and generate a mechanical phase feature vector;

[0040] Perform principal component analysis on the six-degree-of-freedom contact force distribution data in the joint cavity, extract the normal force fluctuation amplitude and tangential force coupling coefficient in the six-degree-of-freedom contact force distribution in the joint cavity, and generate a contact force feature vector;

[0041] Perform vector superposition on the mechanical phase feature vector and the contact force feature vector, and generate a dynamic mechanical fingerprint through a weighted fusion algorithm. The weighted fusion algorithm dynamically adjusts the weight ratio of the mechanical phase feature vector and the contact force feature vector according to the lower limb biomechanical parameters;

[0042] Input the dynamic mechanical fingerprint into an inverse dynamics solver, perform mechanical inverse calculation based on the parametric knee joint mesh model, solve the cartilage contact stress distribution gradient and ligament tension change rate, and generate a biomechanical property map.

[0043] In a second aspect, an artificial intelligence-assisted diagnosis system for a knee joint provided by an embodiment of the present application includes:

[0044] A receiving module, configured to receive a multi-angle dynamic X-ray image sequence and six-degree-of-freedom contact force distribution data in the joint cavity, dynamically track the multi-angle dynamic X-ray image sequence, and combine the six-degree-of-freedom contact force distribution data in the joint cavity to encode the patellar slip trajectory to generate mechanical phase trajectory data;

[0045] A collection module, configured to collect the lower limb biomechanical parameters of the patient to construct a personalized knee joint model, perform vector superposition on the mechanical phase trajectory data and the six-degree-of-freedom contact force distribution data in the joint cavity to generate a dynamic mechanical fingerprint, and map the dynamic mechanical fingerprint to a biomechanical property map through an inverse dynamics solver;

[0046] A construction module, configured to collect the frequency characteristics and tibial rotation angle of the patient's surface electromyogram signal during exercise, construct an electromyogram activation pattern marked with a motion stage, and perform feature fusion on the biomechanical property map and the electromyogram activation pattern to generate a comprehensive joint dynamics profile;

[0047] An input module, configured to input the joint dynamics comprehensive profile into a pre-constructed joint abnormal dynamics propagation model, calculate the wave propagation path of the meniscus stress concentration region and the spatio-temporal evolution trajectory of the cruciate ligament deformation energy field by using the stress wave conduction equation, and identify the compensatory injury pattern by analyzing the interference effect between the spatio-temporal evolution trajectory and the wave propagation path;

[0048] An adjustment module, configured to generate a diagnostic map according to the compensatory injury pattern, in combination with the biomechanical property map, the electromyogram activation pattern, and the joint dynamics comprehensive profile, match the biomechanical constraint conditions in a preset motion feature library, and adaptively correct the diagnostic map to generate an individualized rehabilitation strategy.

[0049] In a third aspect, an embodiment of the present application provides a computing device, including a processor and a memory, where a computer program is stored in the memory, and the processor is configured to run the computer program to execute an artificial intelligence-assisted diagnosis method for a knee joint according to any one of the first aspects.

[0050] In a fourth aspect, an embodiment of the present application provides a computer storage medium, on which computer program instructions are stored, and when the computer program instructions are executed by a processor, an artificial intelligence-assisted diagnosis method for a knee joint according to any one of the first aspects is implemented.

[0051] In the embodiment of the present application, a multi-angle dynamic X-ray image sequence and six-degree-of-freedom contact force distribution data in the joint cavity are received, the multi-angle dynamic X-ray image sequence is dynamically tracked and combined with the six-degree-of-freedom contact force distribution data in the joint cavity, and a mechanical phase trajectory data is generated by encoding the patellar slip trajectory; the lower limb biological parameters of the patient are collected to construct a personalized knee joint model, the mechanical phase trajectory data and the six-degree-of-freedom contact force distribution data in the joint cavity are vectorially superimposed to generate a dynamic mechanical fingerprint, and the dynamic mechanical fingerprint is mapped into a biomechanical property map by an inverse dynamics solver; the frequency characteristics of the surface electromyogram signal and the tibial rotation angle of the patient during exercise are collected to construct an electromyogram activation pattern marked with motion stages, and the biomechanical property map and the electromyogram activation pattern are feature-fused to generate a joint dynamics comprehensive profile; the joint dynamics comprehensive profile is input into a pre-constructed joint abnormal dynamics propagation model, the wave propagation path of the meniscus stress concentration region and the spatio-temporal evolution trajectory of the cruciate ligament deformation energy field are calculated by using the stress wave conduction equation, and the compensatory injury pattern is identified by analyzing the interference effect between the spatio-temporal evolution trajectory and the wave propagation path; according to the compensatory injury pattern, in combination with the biomechanical property map, the electromyogram activation pattern, and the joint dynamics comprehensive profile, a diagnostic map is generated, the biomechanical constraint conditions in a preset motion feature library are matched, and the diagnostic map is adaptively corrected to generate an individualized rehabilitation strategy.

[0052] The technical solution of this application has the following beneficial effects:

[0053] This application not only improves the accuracy of recognizing the damage patterns of the complex knee joint structure, but also realizes the formulation of personalized rehabilitation strategies by integrating multi-source data (such as dynamic X-ray images, contact force distribution, electromyography signals, etc.), effectively improving the treatment effect and the patient's recovery speed. At the same time, it also provides a more scientific and systematic decision-making support tool for clinical practice.

[0054] Furthermore, the embodiments of this application also perform hierarchical tensor decomposition on the comprehensive joint dynamics profile to obtain biomechanical tensor components and kinematic tensor components. Based on the stress waveguide model, a stress propagation path of the meniscus is constructed for the biomechanical tensor components, and the dynamic correlation parameters of the wave attenuation rate and the stress concentration coefficient are calculated; at the same time, for the kinematic tensor components, a ligament energy field evolution model of the viscoelastic deformation field is established, and its spatio-temporal gradient and the coordinates of the energy residence nodes are extracted. Through the above dynamic correlation parameters, spatio-temporal gradient and the coordinates of the energy residence nodes, an interference field equation is established by using the dynamic coupling of the stress waveguide model and the viscoelastic deformation field, and the wave field interference intensity distribution between the stress propagation path of the meniscus and the energy residence nodes is calculated to identify the abnormal resonance region, its three-dimensional coordinates and the energy overflow threshold, and a compensatory damage pattern vector is generated. Finally, this vector is subjected to modal matching with the standard motion chain parameters in the motion feature library, and frequency-domain weighted adjustment is performed based on the matching result to finally generate a compensatory damage pattern.

[0055] This method can accurately calculate the spatio-temporal evolution trajectory of the stress propagation path of the meniscus and the deformation energy field of the cruciate ligament by integrating multi-source data and using advanced mathematical models (such as the stress waveguide model and the viscoelastic deformation field model), and then identify the compensatory damage patterns in the knee joint. This method not only improves the accuracy and personalization level of diagnosis, but also can automatically generate rehabilitation strategies that meet the specific needs of patients, significantly improving the treatment effect and the patient's recovery speed. In addition, the accuracy of damage pattern recognition is further optimized through frequency-domain weighted adjustment, ensuring a high degree of consistency between the diagnostic atlas and the actual clinical situation, and providing strong support for the precision medicine of knee joint diseases.

[0056] These aspects or other aspects of this application will be more clearly understood in the following description of the embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the technical solutions in the embodiments of this application or in the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of this application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0058] Figure 1 It is a flowchart of an artificial intelligence-assisted diagnosis method for a knee joint provided by an embodiment of the present application;

[0059] Figure 2 It is a schematic structural diagram of an artificial intelligence-assisted diagnosis system for a knee joint provided by an embodiment of the present application;

[0060] Figure 3 It is a schematic structural diagram of a computing device provided by an embodiment of the present application. Detailed implementation manners

[0061] In order to enable those skilled in the art of the present technology to better understand the solutions of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application.

[0062] In some processes described in the specification, claims and the above-mentioned accompanying drawings of the present application, there are multiple operations that appear in a specific order. However, it should be clearly understood that these operations may not be executed in the order in which they appear in this article or may be executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish each different operation, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions such as "first" and "second" in this article are used to distinguish different messages, devices, modules, etc., and do not represent a sequence, nor do they limit that "first" and "second" are of different types.

[0063] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present application.

[0064] Figure 1 An embodiment of the present application provides a flowchart of an artificial intelligence-assisted diagnosis method for a knee joint. As Figure 1 shown, the method includes:

[0065] Step 101: Receive a multi-angle dynamic X-ray image sequence and six-degree-of-freedom contact force distribution data in the joint cavity, dynamically track the multi-angle dynamic X-ray image sequence, and combine the six-degree-of-freedom contact force distribution data in the joint cavity to encode the patellar slip trajectory to generate mechanical phase trajectory data;

[0066] In this step, the multi-angle dynamic X-ray image sequence, which is a medical imaging technique, observes the joints or other parts in a moving state by capturing continuous images of the internal structures of the human body from different angles. These images can provide information on how bones and soft tissues interact. The six-degree-of-freedom contact force distribution data in the joint cavity refers to the data on the magnitude and direction of the forces when different parts inside the joint come into contact. Six degrees of freedom means that the force can move along three spatial dimensions (front-back, left-right, up-down) and rotate around these three axes, comprehensively reflecting the complex mechanical environment inside the joint.

[0067] In actual operation, first, the system synchronously acquires the multi-angle dynamic X-ray images of the patient and the contact force distribution data in the joint cavity. Then, through an image processing algorithm, the sliding trajectory of the patella is dynamically tracked and combined with the contact force data to encode and generate phase trajectory data that can reflect the mechanical characteristics of the knee joint. This process lays the foundation for the subsequent construction of a personalized knee joint model.

[0068] For example, an athlete comes to see a doctor due to knee pain. The doctor first uses a multi-angle dynamic X-ray device to scan the athlete's knee joint and synchronously records the six-degree-of-freedom contact force distribution in the joint cavity. The system dynamically tracks the sliding trajectory of the patella based on this data and finds that there is an abnormal sliding pattern in certain movement postures. This provides key information for the next step of constructing a personalized knee joint model for this athlete.

[0069] Step 102: Collect the lower limb biological parameters of the patient to construct a personalized knee joint model, vectorially superimpose the mechanical phase trajectory data and the six-degree-of-freedom contact force distribution data in the joint cavity to generate a dynamic mechanical fingerprint, and map the dynamic mechanical fingerprint to a biomechanical property map through an inverse dynamics solver;

[0070] In this step, the lower limb biological parameters include various physiological and anatomical information related to an individual's lower limbs, such as bone length, muscle mass, joint range of motion, etc., which are used to construct a knee joint model that accurately reflects individual differences. A personalized knee joint model is a computational model established based on the specific biological parameters of the patient to simulate the function and potential problems of their knee joint. Such a model helps to understand the biomechanical characteristics of an individual in a healthy or injured state. The dynamic mechanical fingerprint is generated by vectorially superimposing the mechanical phase trajectory data and the contact force distribution data in the joint cavity, representing the mechanical behavior pattern of a specific individual's knee joint and can be used to identify abnormal or disease characteristics.

[0071] In actual operation, based on the collected lower limb biological parameters of the patient, the system constructs a knee joint model that accurately reflects individual differences. Subsequently, using the mechanical phase trajectory data and contact force distribution data, a dynamic mechanical fingerprint is generated, and through inverse dynamics analysis, it is transformed into a detailed biomechanical property map, showing information such as cartilage stress and ligament tension.

[0072] For example, continuing with the above example, the doctor further collected the lower limb biological parameters of the athlete (such as bone length, muscle strength, etc.) and input them into the system to construct a personalized knee joint model. The system used the previously obtained mechanical phase trajectory data and contact force distribution data to generate the dynamic mechanical fingerprint of the athlete, and through inverse dynamics analysis, generated a biomechanical property map, revealing the specific injury conditions of his knee joint.

[0073] Step 103: Collect the time-frequency characteristics of the surface electromyogram signal and the tibial rotation angle of the patient during exercise, construct an electromyogram activation pattern marked with the movement stage, and perform feature fusion on the biomechanical property map and the electromyogram activation pattern to generate a comprehensive joint dynamics profile;

[0074] In this step, the time-frequency characteristics of the surface electromyogram signal are used to evaluate the working state of the muscle by recording the changes in muscle electrical activity. The time-frequency characteristics refer to the characteristics of the signal changing with time and frequency, which can reflect the activation degree and coordination of the muscle during the execution of an action. The tibial rotation angle refers to the angle at which the lower leg bone (tibia) rotates relative to the thigh bone (femur). It is crucial for analyzing the stability of the knee joint during gait and movement. The electromyogram activation pattern is a model constructed based on the collected surface electromyogram signal and other relevant data, and is used to show the activity rules of each muscle group during a specific movement stage.

[0075] In actual operation, the system collects the surface electromyogram signal and tibial rotation angle data during the patient's movement, constructs an electromyogram activation pattern, and performs feature fusion on it with the previous biomechanical property map to form a comprehensive profile vector that comprehensively describes the functional state of the knee joint. This step helps to more accurately identify the complex injury patterns of the knee joint.

[0076] For example, in the next step, the doctor asked the athlete to perform several standard exercise tests while recording his surface electromyogram signal and tibial rotation angle. The system constructed the electromyogram activation pattern of the athlete based on these data, and performed feature fusion with the previous biomechanical property map to generate a detailed comprehensive joint dynamics profile, showing the specific performance of his knee joint during exercise.

[0077] Step 104: Input the joint dynamics comprehensive profile into a pre - constructed joint abnormal dynamics propagation model, calculate the wave propagation path in the stress - concentration area of the meniscus and the spatio - temporal evolution trajectory of the deformation energy field of the cruciate ligament using the stress - wave conduction equation, and identify the compensatory injury pattern by analyzing the interference effect between the spatio - temporal evolution trajectory and the wave propagation path;

[0078] In this step, the stress - wave conduction equation is a mathematical formula used to describe the propagation process of stress waves inside solid materials due to external forces. In this context, it is used to simulate the responses of the meniscus and the cruciate ligament when subjected to external forces. The wave propagation path is the direction and path of the stress wave when a force applied to the meniscus causes it to propagate within the tissue. This is very important for understanding the injury mechanism. The spatio - temporal evolution trajectory of the deformation energy field describes how the energy distribution of the deformation of the cruciate ligament changes over time under external forces, which helps analyze the development process of the injury.

[0079] In actual operation, the system inputs the joint dynamics comprehensive profile generated in step 103 into the pre - constructed model to simulate the propagation path of stress waves in the knee joint and the change of the energy field of the cruciate ligament. By analyzing this data, the system can identify the key areas and patterns that may lead to compensatory injuries.

[0080] For example, the system inputs the previously generated joint dynamics comprehensive profile into the model, simulates the propagation path of stress waves in the knee joint of this athlete, and finds that there are stress - concentration areas in the meniscus under certain motion states. In addition, the system also analyzes the change of the energy field of the cruciate ligament and determines several key positions that may cause compensatory injuries, providing a basis for further diagnosis.

[0081] Step 105: According to the compensatory injury pattern, combine the biomechanical property atlas, the electromyography activation pattern, and the joint dynamics comprehensive profile to generate a diagnostic atlas, match the biomechanical constraint conditions in a preset motion feature library, and adaptively correct the diagnostic atlas to generate an individualized rehabilitation strategy.

[0082] In this step, the compensatory injury pattern is that when a certain part of the body is injured, the adjustments made by other parts to compensate for the loss of function may lead to new injuries. This pattern helps identify these secondary injuries. The diagnostic atlas is a comprehensive report integrating all the previous data analysis results, which is used to guide treatment decisions. It can contain information such as biomechanical properties, electromyography activity patterns, and dynamic profile vectors. The individualized rehabilitation strategy is a recovery plan formulated according to the specific situation of the patient, aiming to optimize the rehabilitation effect and prevent future injuries. This strategy is customized based on the understanding of the specific biomechanical constraint conditions of the patient.

[0083] In actual operation, the system integrates all previous data and analysis results, generates a detailed diagnostic map, and adjusts it according to a preset motion feature library. Finally, the system outputs a personalized rehabilitation strategy for the patient to ensure the scientific nature and effectiveness of the treatment plan.

[0084] For example, the system generates a detailed diagnostic map based on all previous data, showing the specific injuries of the athlete's knee joint and its compensation mechanism. The system also matches the relevant constraints in the motion feature library, adaptively corrects the diagnostic map, and finally formulates a personalized rehabilitation training plan for the athlete to help him recover health faster.

[0085] This method integrates multi-source data (such as dynamic X-ray images, contact force distribution, electromyogram signals, etc.) and uses advanced mathematical models to achieve accurate identification of complex knee joint injury patterns and automatic generation of personalized rehabilitation strategies. This method not only improves the accuracy and personalization level of diagnosis, but also effectively guides clinical treatment, significantly improving the patient's recovery speed and quality of life. Each step is closely linked, gradually and deeply analyzing the biomechanical characteristics of the knee joint, providing strong support for achieving precision medicine.

[0086] To solve the problem of insufficient accuracy in identifying knee joint compensatory injury patterns in existing methods, in some embodiments, in step 104, inputting the comprehensive joint dynamics profile into a pre-constructed joint abnormal dynamics propagation model, calculating the wave propagation path of the stress concentration area of the meniscus and the spatio-temporal evolution trajectory of the deformation energy field of the cruciate ligament using the stress wave conduction equation, and identifying the compensatory injury pattern by analyzing the interference effect between the spatio-temporal evolution trajectory and the wave propagation path, includes:

[0087] Perform hierarchical tensor decomposition on the comprehensive joint dynamics profile to obtain biomechanical tensor components and kinematic tensor components. Construct a meniscus stress propagation path based on the stress waveguide model for the biomechanical tensor components, and calculate the dynamic correlation parameters of the wave attenuation rate and stress concentration coefficient in the meniscus stress propagation path. Establish an evolution model of the ligament energy field of the viscoelastic deformation field for the kinematic tensor components, and extract the spatio-temporal gradient and energy residence node coordinates of the ligament energy field evolution model. Through the dynamic correlation parameters, the spatio-temporal gradient, and the energy residence node coordinates, establish an interference field equation through the dynamic coupling of the stress waveguide model and the viscoelastic deformation field. Calculate the wave field interference intensity distribution between the meniscus stress propagation path and the energy residence node coordinates according to the interference field equation. Generate a compensatory injury mode vector based on the three-dimensional coordinates and energy overflow threshold of the abnormal resonance region in the wave field interference intensity distribution. Perform modal matching on the compensatory injury mode vector with the standard motion chain parameters in the motion feature library, and perform frequency-domain weighted adjustment on the compensatory injury mode vector based on the matching result to generate a compensatory injury mode.

[0088] In this embodiment, hierarchical tensor decomposition is a data processing technique used to decompose complex multi-dimensional data (such as the comprehensive joint dynamics profile) into multiple more easily analyzable parts. Specifically in this application, it decomposes the comprehensive joint dynamics profile into biomechanical tensor components and kinematic tensor components. The biomechanical tensor components of this part of the data mainly contain information about the mechanical properties of the internal structures of the knee joint (such as the meniscus and ligaments), such as stress distribution, deformation, etc. These data are used to construct a meniscus stress propagation path based on the stress waveguide model and calculate the dynamic correlation parameters of its wave attenuation rate and stress concentration coefficient. The kinematic tensor components of this part of the data focus on the geometric and time-varying characteristics of the knee joint during movement, such as tibial rotation angle and muscle activation pattern. These data are used to establish an evolution model of the ligament energy field of the viscoelastic deformation field and extract the spatio-temporal gradient and energy residence node coordinates of the energy field. The interference field equation is a mathematical model established based on the dynamic coupling of the stress waveguide model and the viscoelastic deformation field, used to calculate the wave field interference intensity distribution between the meniscus stress propagation path and the energy residence node. By analyzing the abnormal resonance region in this distribution and its three-dimensional coordinates and energy overflow threshold, a compensatory injury mode vector is generated.

[0089] In the embodiments of the present application, first, a hierarchical tensor decomposition is performed on the joint dynamics comprehensive profile to obtain biomechanical tensor components and kinematic tensor components. Then, the biomechanical tensor components are used to construct a meniscus stress propagation path based on the stress waveguide model, and the dynamic correlation parameters of the wave attenuation rate and the stress concentration coefficient are calculated. At the same time, the kinematic tensor components are used to establish an evolution model of the ligament energy field of the viscoelastic deformation field, and its spatio-temporal gradient and energy residence node coordinates are extracted. Then, through the above dynamic correlation parameters, spatio-temporal gradient and energy residence node coordinates, an interference field equation is established by the dynamic coupling of the stress waveguide model and the viscoelastic deformation field, and the wave field interference intensity distribution is calculated. Finally, a compensatory damage mode vector is generated according to the abnormal resonance region, its three-dimensional coordinates and the energy overflow threshold in the wave field interference intensity distribution, and it is mode-matched with the standard motion chain parameters in the motion feature library, and frequency-domain weighted adjustment is performed based on the matching result to generate the final compensatory damage mode.

[0090] The following is a specific example:

[0091] A patient came to the doctor due to repeated knee pain. The doctor first used a multi-angle dynamic X-ray device and a six-degree-of-freedom contact force sensor to collect the knee joint images and mechanical data of the patient, and combined the surface electromyogram signal and the tibial rotation angle data to generate a joint dynamics comprehensive profile. Next, the system performed a hierarchical tensor decomposition on the comprehensive profile to obtain biomechanical tensor components and kinematic tensor components. In the part of the biomechanical tensor components, the system constructed a meniscus stress propagation path based on the stress waveguide model and calculated the dynamic correlation parameters of the wave attenuation rate and the stress concentration coefficient. The results showed that there was significant stress concentration in the meniscus under certain specific movements of the patient, indicating a potential injury risk. At the same time, in the part of the kinematic tensor components, the system established an evolution model of the ligament energy field of the viscoelastic deformation field and extracted its spatio-temporal gradient and energy residence node coordinates. The analysis found that there were abnormal fluctuations in the energy field of the cruciate ligament during the patient's walking, suggesting possible early injury or dysfunction. Subsequently, the system established an interference field equation by the dynamic coupling of the stress waveguide model and the viscoelastic deformation field through the dynamic correlation parameters, spatio-temporal gradient and energy residence node coordinates, and calculated the wave field interference intensity distribution. The results showed that there was an obvious abnormal resonance region between the meniscus stress propagation path and the energy residence node, further confirming the existence of a compensatory damage mode in the patient's knee joint. Finally, the system generated a compensatory damage mode vector, mode-matched it with the standard motion chain parameters in the motion feature library, and performed frequency-domain weighted adjustment based on the matching result to generate a personalized rehabilitation strategy. This strategy not only helped the doctor accurately diagnose the patient's knee joint problem, but also provided a scientific basis for formulating an effective treatment plan. This method significantly improved the accuracy and personalization level of diagnosis, and helped to improve the patient's recovery speed and quality of life.

[0092] In order to further improve the accuracy and personalization level of knee joint compensatory injury pattern recognition, in some embodiments, in step 203, by means of the dynamic correlation parameter, the spatio-temporal gradient and the energy residence node coordinates, an interference field equation is established through the dynamic coupling of the stress wave guide model and the viscoelastic deformation field, the wave field interference intensity distribution between the meniscus stress propagation path and the energy residence node coordinates is calculated according to the interference field equation, and a compensatory injury pattern vector is generated according to the three-dimensional coordinates and the energy overflow threshold of the abnormal resonance region in the wave field interference intensity distribution, including:

[0093] Perform wave equation parameterization reconstruction on the meniscus stress propagation path, extract the wave phase delay parameter and the amplitude attenuation rate in the meniscus stress propagation path, perform energy gradient field decomposition on the energy residence node coordinates to generate an energy density gradient vector and a residence time coefficient; construct a dynamic coupling kernel function based on the wave phase delay parameter and the energy density gradient vector, generate a spatial grid mapping of the wave field interference intensity distribution through the dynamic coupling kernel function combined with spatial convolution operation, and mark the grid cells with energy density peaks higher than the overflow threshold in the spatial grid mapping as abnormal resonance regions; use the amplitude attenuation rate and the residence time coefficient to calculate the product of the energy residence time coefficient and the wave amplitude attenuation rate in the abnormal resonance region as the compensatory injury intensity factor, and extract the included angle between the three-dimensional coordinates of the abnormal resonance region and the tangent direction of the wave propagation path as the injury spatial distribution parameter, and perform tensor recombination on the compensatory injury intensity factor and the injury spatial distribution parameter to generate a compensatory injury pattern vector.

[0094] In this embodiment, the parametric reconstruction of the wave equation is a process of mathematically modeling the stress propagation path of the meniscus. By extracting the wave phase delay parameter and the amplitude attenuation rate, the propagation characteristics of stress waves in tissues are described. These parameters help to understand the propagation behavior of stress waves in different media. The energy gradient field decomposition is a technique that decomposes the energy residence node coordinates into an energy density gradient vector and a residence time coefficient. The energy density gradient vector describes the spatial variation trend of energy, while the residence time coefficient reflects the length of time that energy stays at a specific position. The dynamic coupling kernel function is a mathematical model constructed based on the wave phase delay parameter and the energy density gradient vector, and is used to describe the dynamic coupling relationship between the stress waveguide model and the viscoelastic deformation field. This kernel function, combined with spatial convolution operations, can generate a spatial grid mapping of the wave field interference intensity distribution. The compensatory damage intensity factor is calculated by the product of the energy residence time coefficient in the abnormal resonance region and the wave amplitude attenuation rate, and is used to quantify the severity of the damage. In addition, by extracting the angle between the three-dimensional coordinates of the abnormal resonance region and the tangent direction of the wave propagation path as the damage spatial distribution parameter, the specific location and morphology of the damage are further refined.

[0095] In the embodiment of the present application, first, the parametric reconstruction of the wave equation is performed on the stress propagation path of the meniscus, and the wave phase delay parameter and the amplitude attenuation rate are extracted. Then, the energy gradient field decomposition is performed on the energy residence node coordinates to generate an energy density gradient vector and a residence time coefficient. A dynamic coupling kernel function is constructed based on the wave phase delay parameter and the energy density gradient vector, and a spatial grid mapping of the wave field interference intensity distribution is generated through the combination of this kernel function and spatial convolution operations. The grid cells with energy density peaks higher than the overflow threshold are marked as abnormal resonance regions in this mapping. Then, the compensatory damage intensity factor in the abnormal resonance region is calculated using the amplitude attenuation rate and the residence time coefficient, and the angle between the three-dimensional coordinates of the abnormal resonance region and the tangent direction of the wave propagation path is extracted as the damage spatial distribution parameter. Finally, the compensatory damage intensity factor and the damage spatial distribution parameter are tensor-recombined to generate a compensatory damage mode vector.

[0096] The following is a specific embodiment:

[0097] An athlete came to the doctor due to repeated knee pain. The doctor first used a multi-angle dynamic X-ray device and a six-degree-of-freedom contact force sensor to collect the knee joint images and mechanical data of the patient, and combined the surface electromyogram signals and tibial rotation angle data to generate a comprehensive joint dynamics profile. Next, the system analyzed this comprehensive profile, especially focusing on the meniscus stress propagation path and the coordinates of the energy residence nodes. In the biomechanical tensor component part, the system performed a wave equation parametric reconstruction on the meniscus stress propagation path and extracted the wave phase delay parameter and the amplitude attenuation rate. The results showed that there was significant stress concentration in the meniscus under certain specific actions of the patient, indicating a potential injury risk. At the same time, in the kinematic tensor component part, the system performed an energy gradient field decomposition on the coordinates of the energy residence nodes and generated the energy density gradient vector and the residence time coefficient. The analysis found that there were abnormal fluctuations in the energy field of the cruciate ligament during the patient's walking, suggesting possible early injury or dysfunction. Subsequently, the system constructed a dynamic coupling kernel function based on the wave phase delay parameter and the energy density gradient vector, and through this kernel function combined with spatial convolution operations, generated a spatial grid mapping of the wave field interference intensity distribution. The grid cells with energy density peaks higher than the overflow threshold were marked as abnormal resonance regions in this mapping. The results showed that there were obvious abnormal resonance regions between the meniscus stress propagation path and the energy residence nodes, further confirming the existence of a compensatory injury pattern in the patient's knee joint. Then, the system calculated the compensatory injury intensity factor in the abnormal resonance region using the amplitude attenuation rate and the residence time coefficient, and extracted the included angle between the three-dimensional coordinates of the abnormal resonance region and the tangent direction of the wave propagation path as the injury spatial distribution parameter. Finally, the system performed a tensor recombination on the compensatory injury intensity factor and the injury spatial distribution parameter to generate a compensatory injury pattern vector. This detailed analysis not only helped the doctor accurately diagnose the patient's knee joint problem, but also provided information about the severity and specific location of the injury, providing a scientific basis for formulating an effective treatment plan. This method significantly improved the accuracy and personalization level of diagnosis, and helped to improve the patient's recovery speed and quality of life.

[0098] To further improve the accuracy of knee joint compensatory injury pattern recognition and the accuracy of spatial positioning, in some embodiments, constructing a dynamic coupling kernel function based on the wave phase delay parameter and the energy density gradient vector in step 302, and through the dynamic coupling kernel function combined with spatial convolution operations, generating a spatial grid mapping of the wave field interference intensity distribution, and marking the grid cells with energy density peaks higher than the overflow threshold as abnormal resonance regions in the spatial grid mapping, includes:

[0099] Perform time-series normalization on the fluctuating phase delay parameter to generate a phase delay weight coefficient, and perform orthogonal basis decomposition on the energy density gradient vector to obtain a normal energy gradient component and a tangential energy diffusion rate; construct a two-channel coupling kernel function based on the phase delay weight coefficient and the normal energy gradient component, use the tangential energy diffusion rate as the dynamic attenuation factor of the two-channel coupling kernel function, generate a three-dimensional convolution kernel according to the dynamic attenuation factor, define the dynamic window parameter of the spatial convolution operation according to the knee joint anatomical structure, and perform a sliding integral calculation in the biomechanical spatial domain through the dynamic window parameter combined with the three-dimensional convolution kernel to generate an interference intensity distribution matrix; perform multi-scale Gaussian filtering on the interference intensity distribution matrix, detect the spatial topological structure of the energy density peak in the filtered matrix, and cluster the topological units that are continuously adjacent and have a peak higher than the overflow threshold in the spatial topological structure into abnormal resonance regions.

[0100] In this embodiment, time-series normalization is a process of standardizing the fluctuating phase delay parameter so that it can be compared between different time points. The phase delay weight coefficient generated through this process can better reflect the propagation characteristics of stress waves at different time points. Orthogonal basis decomposition is a technique for decomposing the energy density gradient vector into multiple independent components. Specifically in this application, it decomposes the energy density gradient vector into a normal energy gradient component and a tangential energy diffusion rate, which respectively describe the change trends of energy in the directions perpendicular and parallel to the surface. The two-channel coupling kernel function is a mathematical model constructed based on the phase delay weight coefficient and the normal energy gradient component, and is used to describe the coupling relationship between the stress wave guide model and the viscoelastic deformation field. The tangential energy diffusion rate as the dynamic attenuation factor enables the kernel function to adjust its response according to the actual situation. The three-dimensional convolution kernel is a multi-dimensional filter generated based on the two-channel coupling kernel function and is used for sliding integral calculation in the biomechanical spatial domain. The dynamic window parameter of the spatial convolution operation defined in combination with the knee joint anatomical structure can more accurately capture the propagation path of stress waves in tissues. Multi-scale Gaussian filtering is an image processing technique used to smooth data and reduce noise. By detecting the spatial topological structure of the energy density peak in the filtered matrix, topological units that are continuously adjacent and have a peak higher than the overflow threshold can be identified, and these units are clustered into abnormal resonance regions.

[0101] In the embodiments of the present application, first, time series normalization processing is performed on the fluctuating phase delay parameter to generate a phase delay weight coefficient; at the same time, orthogonal basis decomposition is performed on the energy density gradient vector to obtain a normal energy gradient component and a tangential energy diffusion rate. Then, a two-channel coupling kernel function is constructed based on the phase delay weight coefficient and the normal energy gradient component, and the tangential energy diffusion rate is used as a dynamic attenuation factor to generate a three-dimensional convolution kernel. The dynamic window parameter of the spatial convolution operation is defined according to the knee joint anatomical structure, and through this parameter combined with the three-dimensional convolution kernel, sliding integral calculation is performed in the biomechanical spatial domain to generate an interference intensity distribution matrix. Finally, multi-scale Gaussian filtering processing is performed on the interference intensity distribution matrix, the spatial topological structure of the energy density peak in the filtered matrix is detected, and the topological units that are continuously adjacent and have peaks higher than the overflow threshold are clustered into abnormal resonance regions.

[0102] The following is a specific embodiment:

[0103] A patient came to the doctor due to repeated knee pain. The doctor first used a multi-angle dynamic X-ray device and a six-degree-of-freedom contact force sensor to collect the knee joint images and mechanical data of the patient, and combined the surface electromyogram signals and tibial rotation angle data to generate a comprehensive joint dynamics profile. Next, the system conducted an in-depth analysis of this comprehensive profile, especially focusing on the meniscus stress propagation path and the coordinates of the energy residence nodes. During the analysis, the system first performed time series normalization on the fluctuation phase delay parameter to generate a phase delay weight coefficient. The results showed that there was significant stress concentration in the meniscus under certain specific movements of the patient, indicating a potential risk of injury. At the same time, the system performed orthogonal basis decomposition on the energy density gradient vector to obtain the normal energy gradient component and the tangential energy diffusion rate. The analysis found that there were abnormal fluctuations in the energy field of the cruciate ligament during the patient's walking, suggesting possible early injury or dysfunction. Then, the system constructed a two-channel coupled kernel function based on the phase delay weight coefficient and the normal energy gradient component, and used the tangential energy diffusion rate as the dynamic attenuation factor of the two-channel coupled kernel function to generate a three-dimensional convolution kernel. The dynamic window parameters of the spatial convolution operation were defined according to the knee joint anatomical structure, and through this parameter combined with the three-dimensional convolution kernel, a sliding integral calculation was performed in the biomechanical spatial domain to generate an interference intensity distribution matrix. Subsequently, the system performed multi-scale Gaussian filtering on the interference intensity distribution matrix and detected the spatial topological structure of the energy density peaks in the filtered matrix. The results showed that there was an obvious abnormal resonance area between the meniscus stress propagation path and the energy residence nodes, further confirming the presence of a compensatory injury pattern in the patient's knee joint. Finally, the system clustered the topological units that were continuously adjacent and had peaks higher than the overflow threshold into abnormal resonance areas, providing detailed information on the injury location and morphology. This detailed analysis not only helped the doctor accurately diagnose the patient's knee joint problem, but also provided a scientific basis for formulating an effective treatment plan. This method significantly improved the accuracy and personalization level of diagnosis, and helped to improve the patient's recovery speed and quality of life.

[0104] In order to improve the accuracy and spatial resolution of knee joint injury pattern recognition, in some embodiments, constructing a two-channel coupled kernel function based on the phase delay weight coefficient and the normal energy gradient component in step 402, using the tangential energy diffusion rate as the dynamic attenuation factor of the two-channel coupled kernel function, generating a three-dimensional convolution kernel according to the dynamic attenuation factor, defining the dynamic window parameters of the spatial convolution operation according to the knee joint anatomical structure, and performing a sliding integral calculation in the biomechanical spatial domain through the dynamic window parameters combined with the three-dimensional convolution kernel to generate an interference intensity distribution matrix, includes:

[0105] Perform a frequency-domain transformation on the phase delay weight coefficient to generate a phase delay frequency spectrum, extract the proportion of the main frequency band energy in the phase delay frequency spectrum as the time modulation factor of the dual-channel coupling kernel function, perform a spatial Fourier transform on the normal energy gradient component to generate a normal energy wave number spectrum, and extract the high-frequency attenuation slope in the normal energy wave number spectrum as the spatial modulation factor of the dual-channel coupling kernel function; construct the main kernel structure of the dual-channel coupling kernel function based on the time modulation factor and the spatial modulation factor, introduce the logarithmic attenuation rate of the tangential energy diffusion rate as the dynamic attenuation factor of the dual-channel coupling kernel function, and generate a three-dimensional convolution kernel; dynamically adjust the initial value of the sliding radius according to the meniscus curvature radius, define the direction vector of the convolution step size in combination with the anatomical characteristics of the ligament orientation, and generate dynamic window parameters matching the biomechanical characteristics of the knee joint; perform a sliding integral calculation in the biomechanical spatial domain through the three-dimensional convolution kernel and the dynamic window parameters, and adjust the attenuation factor weight of the three-dimensional convolution kernel according to the local energy density gradient during each sliding process to generate the initial mapping of the interference intensity distribution matrix; perform local energy equalization processing on the initial mapping, extract the second-order derivative characteristics of the energy density gradient in the equalized mapping, and adjust the sliding radius and step size parameters of the three-dimensional convolution kernel through the second-order derivative characteristics to generate the interference intensity distribution matrix.

[0106] In this embodiment, the frequency-domain transformation is to perform a frequency-domain transformation on the phase delay weight coefficient to generate a phase delay frequency spectrum, and extract the proportion of the main frequency band energy therein as the time modulation factor of the dual-channel coupling kernel function. This step helps to capture the propagation characteristics of stress waves at different frequencies. The spatial Fourier transform is to perform a spatial Fourier transform on the normal energy gradient component to generate a normal energy wave number spectrum, and extract the high-frequency attenuation slope therein as the spatial modulation factor of the dual-channel coupling kernel function. This step is used to analyze the spatial distribution characteristics of energy. The dynamic attenuation factor is to introduce the logarithmic attenuation rate of the tangential energy diffusion rate as the dynamic attenuation factor of the dual-channel coupling kernel function, enabling the kernel function to be adjusted according to the actual energy diffusion situation, thereby more accurately simulating the propagation path of stress waves. The dynamic window parameters dynamically adjust the initial value of the sliding radius according to the meniscus curvature radius, and define the direction vector of the convolution step size in combination with the anatomical characteristics of the ligament orientation to generate dynamic window parameters matching the biomechanical characteristics of the knee joint. These parameters ensure that the convolution kernel can perform effective sliding integral calculations on specific anatomical structures. The local energy equalization processing is to perform local energy equalization processing on the initial mapping, extract the second-order derivative characteristics of the energy density gradient in the equalized mapping, and adjust the sliding radius and step size parameters of the three-dimensional convolution kernel through these characteristics to generate the final interference intensity distribution matrix.

[0107] In the embodiments of the present application, first, a frequency-domain transformation is performed on the phase delay weight coefficient to generate a phase delay spectrum, and the proportion of the main frequency band energy is extracted as the time modulation factor; at the same time, a spatial Fourier transform is performed on the normal energy gradient component to generate a normal energy wave number spectrum, and the high-frequency attenuation slope is extracted as the spatial modulation factor. Then, a two-channel coupling kernel function is constructed based on the time modulation factor and the spatial modulation factor, and the logarithmic attenuation rate of the tangential energy diffusion rate is introduced as the dynamic attenuation factor to generate a three-dimensional convolution kernel. The initial value of the sliding radius is dynamically adjusted according to the meniscus curvature radius, and the direction vector of the convolution step size is defined in combination with the anatomical characteristics of the ligament direction to generate the dynamic window parameter. Then, a sliding integral calculation is performed in the biomechanical spatial domain through the three-dimensional convolution kernel and the dynamic window parameter, and the attenuation factor weight of the three-dimensional convolution kernel is adjusted according to the local energy density gradient during each sliding process to generate the initial mapping of the interference intensity distribution matrix. Finally, local energy equalization processing is performed on the initial mapping, the second-order derivative characteristics of the energy density gradient in the equalized mapping are extracted, and the sliding radius and step size parameters of the three-dimensional convolution kernel are adjusted through these characteristics to generate the final interference intensity distribution matrix.

[0108] The following is a specific embodiment:

[0109] A patient came to the doctor due to repeated knee pain. The doctor used a multi-angle dynamic X-ray device and a six-degree-of-freedom contact force sensor to collect the knee joint images and mechanical data of the patient, and combined the surface electromyogram signals and tibial rotation angle data to generate a comprehensive joint dynamics profile. The system first performed a frequency-domain transformation on the fluctuation phase delay parameter to generate a phase delay spectrum, and extracted the proportion of the main frequency band energy as the time modulation factor. The results showed that there was significant stress concentration in the meniscus during certain specific movements of the patient, indicating a potential risk of injury. At the same time, the system performed a spatial Fourier transform on the normal energy gradient component to generate a normal energy wavenumber spectrum, and extracted the high-frequency attenuation slope as the spatial modulation factor. It was found through analysis that there were abnormal fluctuations in the energy field of the cruciate ligament during the patient's walking, suggesting possible early injury or dysfunction. Then, the system constructed a two-channel coupled kernel function based on the time modulation factor and the spatial modulation factor, and introduced the logarithmic decay rate of the tangential energy diffusion rate as the dynamic decay factor to generate a three-dimensional convolution kernel. The initial value of the sliding radius was dynamically adjusted according to the meniscus curvature radius, and the direction vector of the convolution step size was defined in combination with the anatomical characteristics of the ligament orientation to generate dynamic window parameters matching the biomechanical characteristics of the knee joint. A sliding integral calculation was performed in the biomechanical spatial domain through the three-dimensional convolution kernel and the dynamic window parameters, and the attenuation factor weight of the three-dimensional convolution kernel was adjusted according to the local energy density gradient during each sliding process to generate an initial mapping of the interference intensity distribution matrix. Finally, the system performed a local energy equalization process on the initial mapping, extracted the second-order derivative features of the energy density gradient in the equalized mapping, adjusted the sliding radius and step size parameters of the three-dimensional convolution kernel through these features, generated the final interference intensity distribution matrix, provided detailed injury location and morphological information, helped the doctor accurately diagnose the patient's knee joint problem, provided a scientific basis for formulating an effective treatment plan, significantly improved the accuracy and personalization level of the diagnosis, and helped improve the patient's recovery speed and quality of life.

[0110] In order to further improve the accuracy of knee joint injury pattern recognition and the effectiveness of multi-modal data fusion, in some embodiments, the frequency characteristics and tibial rotation angle during the collection of the surface electromyogram signals of the patient during movement in step 103 are used to construct an electromyogram activation pattern with movement stage markers, and the biomechanical attribute map is feature-fused with the electromyogram activation pattern to generate a comprehensive joint dynamics profile, including:

[0111] Perform multi-scale time-frequency decomposition on the surface electromyography signals collected from patients during exercise, extract the energy concentration coefficient and phase synchronization index in specific frequency bands in the time-frequency domain, and generate an electromyography time-frequency feature vector; collect the time series data of the tibial rotation angle, calculate the rotational dynamic stability coefficient of the time series data through the angle change rate, and construct an electromyography activation pattern for marking the movement stage in combination with the electromyography time-frequency feature vector; extract the spatial domain features of the biomechanical property map, obtain the cartilage contact stress distribution gradient and ligament tension change rate in the map, and generate a biomechanical feature vector; perform multi-modal feature alignment on the electromyography activation pattern and the biomechanical feature vector, align the time series of the movement stage through the dynamic time warping algorithm, and generate a time-synchronized multi-modal feature matrix; construct a comprehensive joint dynamics profile based on the multi-modal feature matrix, and couple the energy concentration coefficient in the electromyography activation pattern with the stress distribution gradient in the biomechanical feature vector through a non-linear mapping function to generate a comprehensive dynamics profile vector.

[0112] In this embodiment, multi-scale time-frequency decomposition is a signal processing technique used to extract features in different time scales and frequency ranges from surface electromyography signals. Through this method, an electromyography time-frequency feature vector can be generated, which contains the energy concentration coefficient and phase synchronization index in specific frequency bands. These features help describe the activation of muscles in different movement stages. The rotational dynamic stability coefficient is evaluated by analyzing the time series data of the collected tibial rotation angle and calculating its angle change rate. This coefficient reflects the stability and coordination of the knee joint during movement. The cartilage contact stress distribution gradient and ligament tension change rate are spatial domain features extracted from the biomechanical property map to describe the mechanical responses of the internal structures of the knee joint (such as cartilage and ligaments) in different movement states. These features provide important information about the health status of the knee joint. Multi-modal feature alignment uses the dynamic time warping algorithm (DTW) to align data in different modalities (such as electromyography signals and biomechanical features) to the same time axis, ensuring the time synchronization of data in each modality, and thus generating a time-synchronized multi-modal feature matrix. The non-linear mapping function is a mathematical model used to couple the energy concentration coefficient in the electromyography activation pattern with the stress distribution gradient in the biomechanical feature vector to generate a comprehensive dynamics profile vector. This coupling helps capture complex joint dynamics behavior.

[0113] In the embodiments of the present application, first, the surface electromyography signals of the patient collected during exercise are subjected to multi-scale time-frequency decomposition, the energy concentration coefficient and phase synchronization index in specific frequency bands in the time-frequency domain are extracted, and an electromyography time-frequency feature vector is generated. At the same time, the time series data of the tibial rotation angle are collected, the rotational dynamic stability coefficient is calculated through the angle change rate, and the electromyography activation pattern of the motion stage label is constructed in combination with the electromyography time-frequency feature vector. Then, the spatial domain features of the biomechanical property map are extracted, the cartilage contact stress distribution gradient and the ligament tension change rate are obtained, and a biomechanical feature vector is generated. Next, the electromyography activation pattern and the biomechanical feature vector are subjected to multi-modal feature alignment, and the time series of the motion stage are aligned through the dynamic time warping algorithm to generate a time-synchronized multi-modal feature matrix. Finally, a comprehensive joint dynamics profile is constructed based on the multi-modal feature matrix, and the energy concentration coefficient in the electromyography activation pattern and the stress distribution gradient in the biomechanical feature vector are coupled through a non-linear mapping function to generate a comprehensive dynamics profile.

[0114] The following is a specific embodiment:

[0115] An athlete came to the doctor due to repeated knee pain. The doctor first used a multi-angle dynamic X-ray device and a six-degree-of-freedom contact force sensor to collect the knee joint images and mechanical data of the patient, and combined the surface electromyography signal and the tibial rotation angle data to generate a comprehensive joint dynamics profile. The system began to perform multi-scale time-frequency decomposition on the collected surface electromyography signals, extracted the energy concentration coefficient and phase synchronization index in specific frequency bands, and generated an electromyography time-frequency feature vector. The results showed that the muscle activation patterns of the patient were abnormal during certain specific movements, indicating a potential risk of muscle fatigue or injury. At the same time, the system collected the time series data of the tibial rotation angle, obtained the rotational dynamic stability coefficient by calculating the angle change rate, and found that the rotational stability of the patient's knee joint was poor during walking, suggesting possible early injury or dysfunction. Then, the system extracted the spatial domain features of the biomechanical property map, obtained the cartilage contact stress distribution gradient and the ligament tension change rate, and generated a biomechanical feature vector. The analysis showed that the meniscus and cruciate ligaments of the patient were under high stress during certain sports postures, which might be one of the reasons for the pain. Subsequently, the system aligned the electromyography activation pattern and the biomechanical feature vector through multi-modal feature alignment, aligned the time series of the movement phases through the dynamic time warping algorithm, and generated a time-synchronized multi-modal feature matrix. This process ensured the precise correspondence of the electromyography signal and the biomechanical features in time, improving the accuracy of data analysis. Finally, the system constructed a comprehensive joint dynamics profile based on the multi-modal feature matrix, coupled the energy concentration coefficient in the electromyography activation pattern with the stress distribution gradient in the biomechanical feature vector through a non-linear mapping function, and generated a dynamic comprehensive profile vector. This detailed analysis not only helped the doctor accurately diagnose the patient's knee joint problem, but also provided a scientific basis for formulating an effective treatment plan. This method significantly improved the accuracy and personalization level of diagnosis, helping to improve the patient's recovery speed and quality of life. The whole process demonstrated how to use multi-modal data fusion technology to enhance the comprehensiveness and accuracy of knee joint injury identification.

[0116] To further improve the accuracy of personalized knee joint model construction and the effectiveness of biomechanical property map generation, in some embodiments, for collecting the lower limb biological parameters of the patient to construct a personalized knee joint model in step 102, vectorially superposing the mechanical phase trajectory data and the six-degree-of-freedom contact force distribution data in the joint cavity to generate a dynamic mechanical fingerprint, and mapping the dynamic mechanical fingerprint into a biomechanical property map through an inverse dynamics solver, including:

[0117] Collect the lower limb length, intercondylar width of the femur, and tibial plateau inclination angle of the patient as lower limb biomechanical parameters, and construct a parametric knee joint mesh model based on the lower limb biomechanical parameters. The parametric knee joint mesh model includes the meniscus curvature radius and the coordinates of the cruciate ligament attachment points that match the anatomical structure; perform trajectory smoothing processing on the mechanical phase trajectory data, extract the patellar slip velocity vector and the acceleration change rate in the mechanical phase trajectory, and generate a mechanical phase feature vector; perform principal component analysis on the six-degree-of-freedom contact force distribution data in the joint cavity, extract the normal force fluctuation amplitude and the tangential force coupling coefficient in the six-degree-of-freedom contact force distribution in the joint cavity, and generate a contact force feature vector; perform vector superposition on the mechanical phase feature vector and the contact force feature vector, and generate a dynamic mechanical fingerprint through a weighted fusion algorithm. The weighted fusion algorithm dynamically adjusts the weight ratio of the mechanical phase feature vector and the contact force feature vector according to the lower limb biomechanical parameters; input the dynamic mechanical fingerprint into an inverse dynamics solver, and perform mechanical inverse calculation based on the parametric knee joint mesh model to solve the cartilage contact stress distribution gradient and the ligament tension change rate, and generate a biomechanical property map.

[0118] In this embodiment, the mechanical phase feature vector is generated by performing trajectory smoothing processing on the mechanical phase trajectory data and extracting the patellar slip velocity vector and the acceleration change rate. These features describe the dynamic behavior of the patella during movement. The contact force feature vector is generated by performing principal component analysis (PCA) on the six-degree-of-freedom contact force distribution data in the joint cavity and extracting the normal force fluctuation amplitude and the tangential force coupling coefficient. These features reflect the contact force distribution between different parts inside the joint. The inverse dynamics solver is a mathematical tool that performs mechanical inverse calculation based on the parametric knee joint mesh model and the dynamic mechanical fingerprint to solve the cartilage contact stress distribution gradient and the ligament tension change rate, and generate a detailed biomechanical property map.

[0119] In the embodiments of the present application, first, lower limb biological parameters of the patient (such as lower limb length, intercondylar width of the femur, and tibial plateau inclination angle) are collected, and a parameterized knee joint mesh model is constructed based on these parameters. This model contains detailed anatomical information such as the curvature radius of the meniscus and the coordinates of the attachment points of the cruciate ligaments. Next, the mechanical phase trajectory data is smoothed, the patellar slip velocity vector and the acceleration change rate are extracted to generate a mechanical phase feature vector. At the same time, principal component analysis is performed on the six-degree-of-freedom contact force distribution data in the joint cavity, the normal force fluctuation amplitude and the tangential force coupling coefficient are extracted to generate a contact force feature vector. Then, the mechanical phase feature vector and the contact force feature vector are vectorially superimposed, and a dynamic mechanical fingerprint is generated through a weighted fusion algorithm, where the weight ratio is dynamically adjusted according to the lower limb biological parameters. Finally, the dynamic mechanical fingerprint is input into an inverse dynamics solver, and mechanical inverse calculation is performed based on the parameterized knee joint mesh model to solve the cartilage contact stress distribution gradient and the ligament tension change rate, generating a biomechanical property map.

[0120] The following is a specific embodiment:

[0121] An athlete came to the doctor due to repeated knee pain. The doctor first measured the patient's lower limb length, intercondylar width of the femur, tibial plateau inclination angle and other biological parameters, and used a multi-angle dynamic X-ray device and a six-degree-of-freedom contact force sensor to collect the patient's knee joint images and mechanical data. The system constructed a parameterized knee joint mesh model based on these lower limb biological parameters. This model included the meniscus curvature radius and the coordinates of the cruciate ligament attachment points that matched the patient's anatomical structure. Next, the system smoothed the trajectory of the mechanical phase trajectory data, extracted the patellar slip velocity vector and the acceleration change rate, and generated a mechanical phase feature vector. The results showed that the patellar slip velocity increased significantly during certain specific movements of the patient, indicating a possible potential slip abnormality or injury risk. At the same time, the system performed a principal component analysis on the six-degree-of-freedom contact force distribution data in the joint cavity, extracted the normal force fluctuation amplitude and the tangential force coupling coefficient, and generated a contact force feature vector. The analysis found that there were abnormal fluctuations in the contact force distribution inside the joint during the patient's walking, indicating possible early damage or dysfunction. Subsequently, the system superimposed the mechanical phase feature vector and the contact force feature vector vectorially, and generated a dynamic mechanical fingerprint through a weighted fusion algorithm, where the weight ratio was dynamically adjusted according to the patient's lower limb biological parameters. This process ensured the reasonable combination of different feature vectors and improved the accuracy of data analysis. Finally, the system input the dynamic mechanical fingerprint into an inverse dynamics solver, performed mechanical inverse calculation based on the parameterized knee joint mesh model, solved the cartilage contact stress distribution gradient and the ligament tension change rate, and generated a detailed biomechanical property map. This not only helped the doctor accurately diagnose the patient's knee joint problem, but also provided a scientific basis for formulating an effective treatment plan. This method significantly improved the accuracy and personalization level of diagnosis, and helped to improve the patient's recovery speed and quality of life. The whole process demonstrated how to use multi-source data fusion technology to enhance the comprehensiveness and accuracy of knee injury identification.

[0122] Figure 2 The following is a schematic structural diagram of an artificial intelligence-assisted diagnosis system for a knee joint provided by an embodiment of the present application, as Figure 2 shown. The system includes:

[0123] A receiving module 21, configured to receive a multi-angle dynamic X-ray image sequence and six-degree-of-freedom contact force distribution data in the joint cavity, dynamically track the multi-angle dynamic X-ray image sequence, and combine the six-degree-of-freedom contact force distribution data in the joint cavity to encode the patellar slip trajectory to generate mechanical phase trajectory data;

[0124] A collection module 22, configured to collect the biological parameters of the patient's lower limbs to construct a personalized knee joint model, vectorially superimpose the mechanical phase trajectory data and the six-degree-of-freedom contact force distribution data in the joint cavity to generate a dynamic mechanical fingerprint, and map the dynamic mechanical fingerprint into a biomechanical property map through an inverse dynamics solver;

[0125] A construction module 23, configured to collect the frequency characteristics and tibial rotation angles during the movement of the patient's surface electromyogram signals, construct an electromyogram activation pattern marked with movement phases, and perform feature fusion on the biomechanical property map and the electromyogram activation pattern to generate a comprehensive joint dynamics profile;

[0126] An input module 24, configured to input the comprehensive joint dynamics profile into a pre-constructed joint abnormal dynamics propagation model, calculate the wave propagation path of the stress concentration area of the meniscus and the spatio-temporal evolution trajectory of the deformation energy field of the cruciate ligament by using the stress wave conduction equation, and identify the compensatory injury pattern by analyzing the interference effect between the spatio-temporal evolution trajectory and the wave propagation path;

[0127] An adjustment module 25, configured to generate a diagnostic map according to the compensatory injury pattern, in combination with the biomechanical property map, the electromyogram activation pattern and the comprehensive joint dynamics profile, match the biomechanical constraint conditions in a preset motion feature library, and adaptively correct the diagnostic map to generate an individualized rehabilitation strategy.

[0128] Figure 2 The artificial intelligence-assisted diagnosis system of a knee joint described above can execute Figure 1 The artificial intelligence-assisted diagnosis method of a knee joint described in the embodiment shown, and its implementation principle and technical effects will not be elaborated. For the artificial intelligence-assisted diagnosis system of a knee joint in the above embodiment, the specific manners in which each module and unit perform operations have been described in detail in the embodiment related to the method, and will not be elaborated here.

[0129] In a possible design, Figure 2 The artificial intelligence-assisted diagnosis system of a knee joint in the embodiment shown can be implemented as a computing device, as Figure 3 shown, and this computing device can include a storage component 31 and a processing component 32;

[0130] The storage component 31 stores one or more computer instructions, and among them, the one or more computer instructions are called and executed by the processing component 32.

[0131] The processing component 32 is the above Figure 1 The artificial intelligence-assisted diagnosis method of a knee joint in the embodiment.

[0132] Among them, the processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above methods. Of course, the processing component may also be implemented by one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components for executing the above methods.

[0133] The storage component 31 is configured to store various types of data to support the operation of the terminal. The storage component may be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0134] Of course, the computing device may also necessarily include other components, such as input / output interfaces, display components, communication components, etc.

[0135] The input / output interface provides an interface between the processing component and the peripheral interface module, and the above peripheral interface module may be an output device, an input device, etc.

[0136] The communication component is configured to facilitate communication between the computing device and other devices in a wired or wireless manner, etc.

[0137] Among them, the computing device may be a physical device or an elastic computing host provided by a cloud computing platform, etc. At this time, the computing device may refer to a cloud server, and the above processing component, storage component, etc. may be basic server resources leased or purchased from the cloud computing platform.

[0138] The embodiment of the present application also provides a computer storage medium storing a computer program, and when the computer program is executed by a computer, it can implement the Figure 1 artificial intelligence assisted diagnosis method of a knee joint shown in the above embodiments.

[0139] Those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working processes of the above-described systems, devices, and units can refer to the corresponding processes in the foregoing method embodiments, and will not be described herein again.

[0140] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. A person of ordinary skill in the art can understand and implement it without creative effort.

[0141] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0142] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and are not intended to limit them. Although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features. And these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of each embodiment of the present application.

Claims

1. An artificial intelligence-assisted diagnosis method for knee joints, characterized in that: include: Receiving a multi-angle dynamic X-ray image sequence and six-degree-of-freedom contact force distribution data in a joint cavity, dynamically tracking the multi-angle dynamic X-ray image sequence and combining the six-degree-of-freedom contact force distribution data in the joint cavity, encoding the patellar sliding trajectory to generate mechanical phase trajectory data; Collecting biological parameters of the patient's lower limbs to construct a personalized knee joint model, performing vector superposition of the mechanical phase trajectory data and the six-degree-of-freedom contact force distribution data in the joint cavity to generate a dynamic mechanical fingerprint, and mapping the dynamic mechanical fingerprint into a biomechanical property map through an inverse dynamics solver; The frequency characteristics of the surface electromyographic signals and the tibial rotation angle of the patient during the movement are collected, the electromyographic activation pattern marked by the movement phase is constructed, and the biomechanical property map is feature-fused with the electromyographic activation pattern to generate a comprehensive profile of joint dynamics; The joint dynamics comprehensive profile is input into a pre-constructed joint abnormal dynamics propagation model, and the wave propagation path of the meniscus stress concentration area and the spatiotemporal evolution trajectory of the cruciate ligament deformation energy field are calculated using the stress wave conduction equation, and the compensatory injury pattern is identified by analyzing the interference effect between the spatiotemporal evolution trajectory and the wave propagation path; According to the compensatory injury pattern, combined with the biomechanical property map, the electromyographic activation pattern and the comprehensive profile of joint dynamics, a diagnostic map is generated to match the biomechanical constraints in the preset motion feature library, and the diagnostic map is adaptively corrected to generate an individualized rehabilitation strategy.

2. The method according to claim 1, characterized in that The joint dynamics comprehensive profile is input into a pre-constructed joint abnormal dynamics propagation model, and the wave propagation path of the meniscus stress concentration area and the spatiotemporal evolution trajectory of the cruciate ligament deformation energy field are calculated using the stress wave conduction equation, and the compensatory injury mode is identified by analyzing the interference effect between the spatiotemporal evolution trajectory and the wave propagation path, including: Performing hierarchical tensor decomposition on the joint dynamics comprehensive profile to obtain biomechanical tensor components and kinematic tensor components, constructing a meniscus stress propagation path based on a stress waveguide model for the biomechanical tensor components, and calculating dynamic correlation parameters between the fluctuation attenuation rate and the stress concentration coefficient in the meniscus stress propagation path; Establishing a ligament energy field evolution model of the viscoelastic deformation field for the kinematic tensor component, and extracting the spatiotemporal gradient and energy retention node coordinates of the ligament energy field evolution model; By means of the dynamic correlation parameters, the spatiotemporal gradient and the energy retention node coordinates, an interference field equation is established through the dynamic coupling of a stress waveguide model and a viscoelastic deformation field, and the wave field interference intensity distribution between the meniscus stress propagation path and the energy retention node coordinates is calculated according to the interference field equation, and a compensatory damage pattern vector is generated according to the three-dimensional coordinates of an abnormal resonance region in the wave field interference intensity distribution and an energy overflow threshold; The compensatory injury pattern vector is modally matched with the standard motion chain parameters in the motion feature library, and the compensatory injury pattern vector is weightedly adjusted in the frequency domain based on the matching result to generate a compensatory injury pattern.

3. The method according to claim 2, characterized in that The method includes establishing an interference field equation through the dynamic correlation parameter, the spatiotemporal gradient and the energy residence node coordinates, and dynamically coupling a stress waveguide model with a viscoelastic deformation field; calculating the wavefield interference intensity distribution between the meniscus stress propagation path and the energy residence node coordinates according to the interference field equation; and generating a compensatory damage pattern vector according to the three-dimensional coordinates of an abnormal resonance region in the wavefield interference intensity distribution and an energy overflow threshold, including: Performing wave equation parameterization reconstruction on the meniscus stress propagation path, extracting the wave phase delay parameter and amplitude attenuation rate in the meniscus stress propagation path, performing energy gradient field decomposition on the energy residence node coordinates, and generating an energy density gradient vector and residence time coefficient; A dynamic coupling kernel function is constructed based on the wave phase delay parameter and the energy density gradient vector, and a spatial grid map of wave field interference intensity distribution is generated by combining the dynamic coupling kernel function with a spatial convolution operation, and grid cells whose energy density peaks are higher than an overflow threshold are marked in the spatial grid map as abnormal resonance regions; The amplitude decay rate and the residence time coefficient are used to calculate the product of the energy residence time coefficient and the fluctuation amplitude decay rate in the abnormal resonance area as the compensatory damage intensity factor, and the angle between the three-dimensional coordinates of the abnormal resonance area and the tangent direction of the wave propagation path is extracted as the damage space distribution parameter. The compensatory damage intensity factor and the damage space distribution parameter are tensor-recombined to generate a compensatory damage pattern vector.

4. The method according to claim 3, characterized in that The method comprises: constructing a dynamic coupling kernel function based on the wave phase delay parameter and the energy density gradient vector, generating a spatial grid map of wave field interference intensity distribution by combining the dynamic coupling kernel function with a spatial convolution operation, and marking grid cells with energy density peak values ​​higher than an overflow threshold in the spatial grid map as abnormal resonance regions, including: Performing time series normalization processing on the fluctuation phase delay parameter to generate a phase delay weight coefficient, and performing orthogonal basis decomposition on the energy density gradient vector to obtain a normal energy gradient component and a tangential energy diffusion rate; A dual-channel coupling kernel function is constructed based on the phase delay weight coefficient and the normal energy gradient component, the tangential energy diffusivity is used as the dynamic attenuation factor of the dual-channel coupling kernel function, a three-dimensional convolution kernel is generated according to the dynamic attenuation factor, a dynamic window parameter of the spatial convolution operation is defined according to the anatomical structure of the knee joint, and a sliding integral calculation is performed in the biomechanical space domain in combination with the three-dimensional convolution kernel through the dynamic window parameter to generate an interference intensity distribution matrix; The interference intensity distribution matrix is ​​subjected to multi-scale Gaussian filtering, the spatial topological structure of the energy density peak in the filtered matrix is ​​detected, and the topological units in the spatial topological structure that are continuously adjacent and have peak values ​​higher than an overflow threshold are clustered as abnormal resonance regions.

5. The method according to claim 4, characterized in that The dual-channel coupling kernel function is constructed based on the phase delay weight coefficient and the normal energy gradient component, the tangential energy diffusivity is used as the dynamic attenuation factor of the dual-channel coupling kernel function, a three-dimensional convolution kernel is generated according to the dynamic attenuation factor, a dynamic window parameter of the spatial convolution operation is defined according to the anatomical structure of the knee joint, and a sliding integral calculation is performed in the biomechanical space domain in combination with the dynamic window parameter to generate an interference intensity distribution matrix, including: Performing frequency domain transformation on the phase delay weight coefficient to generate a phase delay spectrum, extracting the main frequency band energy ratio in the phase delay spectrum as the time modulation factor of the dual-channel coupling kernel function, performing spatial Fourier transformation on the normal energy gradient component to generate a normal energy wavenumber spectrum, and extracting the high-frequency attenuation slope in the normal energy wavenumber spectrum as the spatial modulation factor of the dual-channel coupling kernel function; Based on the time modulation factor and the space modulation factor, a main kernel structure of a dual-channel coupling kernel function is constructed, and the logarithmic attenuation rate of the tangential energy diffusion rate is introduced as a dynamic attenuation factor of the dual-channel coupling kernel function to generate a three-dimensional convolution kernel; The initial value of the sliding radius is dynamically adjusted according to the radius of curvature of the meniscus, and the direction vector of the convolution step is defined in combination with the anatomical characteristics of the ligament direction to generate dynamic window parameters that match the biomechanical characteristics of the knee joint. Performing sliding integral calculation in the biomechanical space domain by using the three-dimensional convolution kernel and the dynamic window parameters, adjusting the attenuation factor weight of the three-dimensional convolution kernel according to the local energy density gradient during each sliding process, and generating an initial mapping of the interference intensity distribution matrix; The initial mapping is subjected to local energy equalization processing, the second-order derivative characteristics of the energy density gradient in the equalized mapping are extracted, the sliding radius and step size parameters of the three-dimensional convolution kernel are adjusted by the second-order derivative characteristics, and an interference intensity distribution matrix is ​​generated.

6. The method according to claim 1, characterized in that The method collects the frequency characteristics of the surface electromyographic signal and the tibial rotation angle of the patient during the movement, constructs the electromyographic activation pattern marked by the movement stage, and performs feature fusion of the biomechanical property map with the electromyographic activation pattern to generate a comprehensive profile of joint dynamics, including: Perform multi-scale time-frequency decomposition on the surface electromyographic signals collected from patients during exercise, extract the energy concentration coefficient and phase synchronization index of specific frequency bands in the time-frequency domain, and generate electromyographic time-frequency feature vectors; Collecting time series data of tibial rotation angle, calculating the rotational dynamic stability coefficient of the time series data by angle change rate, and constructing the electromyographic activation pattern marked by the movement stage in combination with the electromyographic time-frequency feature vector; Performing spatial domain feature extraction on the biomechanical property atlas, obtaining the cartilage contact stress distribution gradient and ligament tension change rate in the atlas, and generating a biomechanical feature vector; Performing multimodal feature alignment on the electromyographic activation pattern and the biomechanical feature vector, aligning the time series of the movement phases through a dynamic time warping algorithm, and generating a time-synchronized multimodal feature matrix; A comprehensive joint dynamics profile is constructed based on the multimodal characteristic matrix, and the energy concentration coefficient in the myoelectric activation pattern is coupled with the stress distribution gradient in the biomechanical characteristic vector through a nonlinear mapping function to generate a comprehensive dynamics profile vector.

7. The method according to claim 1, characterized in that The method collects biological parameters of the patient's lower limbs to construct a personalized knee joint model, performs vector superposition of the mechanical phase trajectory data and the six-degree-of-freedom contact force distribution data in the joint cavity to generate a dynamic mechanical fingerprint, and maps the dynamic mechanical fingerprint into a biomechanical property map through an inverse dynamics solver, including: The patient's lower limb length, interfemoral width and tibial plateau inclination are collected as lower limb biological parameters, and a parameterized knee joint mesh model is constructed based on the lower limb biological parameters, wherein the parameterized knee joint mesh model includes a meniscus curvature radius and cruciate ligament attachment point coordinates that match the anatomical structure; Performing trajectory smoothing processing on the mechanical phase trajectory data, extracting the patellar sliding velocity vector and acceleration change rate in the mechanical phase trajectory, and generating a mechanical phase characteristic vector; Performing principal component analysis on the six-degree-of-freedom contact force distribution data in the joint cavity, extracting the normal force fluctuation amplitude and the tangential force coupling coefficient in the six-degree-of-freedom contact force distribution in the joint cavity, and generating a contact force characteristic vector; Performing vector superposition of the mechanical phase characteristic vector and the contact force characteristic vector, generating a dynamic mechanical fingerprint through a weighted fusion algorithm, wherein the weighted fusion algorithm dynamically adjusts the weight ratio of the mechanical phase characteristic vector and the contact force characteristic vector according to the biological parameters of the lower limb; The dynamic mechanical fingerprint is input into an inverse dynamics solver, and a mechanical inversion calculation is performed based on the parameterized knee joint mesh model to solve the cartilage contact stress distribution gradient and the ligament tension change rate, and generate a biomechanical property map.

8. An artificial intelligence-assisted diagnosis system for knee joints, characterized in that: include: A receiving module, used for receiving a multi-angle dynamic X-ray image sequence and six-degree-of-freedom contact force distribution data in a joint cavity, dynamically tracking the multi-angle dynamic X-ray image sequence and combining the six-degree-of-freedom contact force distribution data in the joint cavity, encoding the patellar sliding trajectory to generate mechanical phase trajectory data; A collection module, used for collecting biological parameters of the patient's lower limbs to construct a personalized knee joint model, performing vector superposition of the mechanical phase trajectory data and the six-degree-of-freedom contact force distribution data in the joint cavity to generate a dynamic mechanical fingerprint, and mapping the dynamic mechanical fingerprint into a biomechanical property map through an inverse dynamics solver; A construction module is used to collect the frequency characteristics of the surface electromyographic signal and the tibial rotation angle of the patient during the movement, construct the electromyographic activation pattern marked by the movement stage, and perform feature fusion of the biomechanical property map with the electromyographic activation pattern to generate a comprehensive profile of joint dynamics; An input module is used to input the joint dynamics comprehensive profile into a pre-constructed joint abnormal dynamics propagation model, calculate the wave propagation path of the meniscus stress concentration area and the spatiotemporal evolution trajectory of the cruciate ligament deformation energy field by using the stress wave conduction equation, and identify the compensatory injury mode by analyzing the interference effect between the spatiotemporal evolution trajectory and the wave propagation path; The adjustment module is used to generate a diagnostic map according to the compensatory injury pattern, in combination with the biomechanical property map, the myoelectric activation pattern and the comprehensive profile of joint dynamics, match the biomechanical constraints in the preset motion feature library, and adaptively modify the diagnostic map to generate an individualized rehabilitation strategy.

9. A computing device, characterized in that It comprises a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement an artificial intelligence-assisted diagnosis method for a knee joint as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that: A computer program is stored, and when the computer program is executed by a computer, an artificial intelligence-assisted diagnosis method for a knee joint as described in any one of claims 1 to 7 is implemented.

Images