A knee joint three-dimensional movement analysis platform for clinical auxiliary diagnosis

Abstract

The application relates to the technical field of medical data processing, and discloses a knee joint three-dimensional movement analysis platform for clinical auxiliary diagnosis, which comprises the following steps: a data processing module adopts Magic software to collect knee joint three-dimensional movement original data, carries out coordinate correction and model processing, and generates a model group. A model analysis module imports the model group into a C3D clinical window, realizes real-time visualization of a three-dimensional movement track, calculates biomechanical parameters, and evaluates clinical pathology correlation. A man-machine interaction control module provides customized interaction logic and a graphical interface, supports parameter input, analysis process control, real-time monitoring, and automatic generation of a diagnosis report. A cloud publishing module converts the diagnosis report and the interface into a PHP dynamic webpage, deploys the webpage through HTTP, supports SSL encryption, multi-user concurrent access, and cross-platform remote calling. The application supports real-time visualization of a knee joint three-dimensional movement track, and improves data reliability.

Description

Technical Field

[0001] This invention relates to the field of medical data processing technology, and more specifically, to a three-dimensional movement analysis platform for knee joints used in clinical auxiliary diagnosis. Background Technology

[0002] Knee joint diseases have a high incidence rate in clinical practice, especially osteoarthritis, ligament injuries, and meniscus diseases, which seriously affect patients' quality of life. Traditional diagnostic methods mainly rely on static imaging examinations (such as X-rays, CT, and MRI) and physical examinations. However, these methods only provide information on the anatomical structure of the knee joint and cannot fully reflect the spatial motion characteristics and biomechanical state of the knee joint during dynamic movement, thus limiting the accuracy of clinical diagnosis and the feasibility of early intervention.

[0003] Existing knee joint motion analysis technologies suffer from several limitations. Firstly, data accuracy is limited; two-dimensional images or single-sensor data cannot fully capture the multi-degree-of-freedom three-dimensional motion information of the knee joint. These technologies are also susceptible to noise and sensor errors, leading to deviations in motion parameters and hindering accurate clinical judgment. Secondly, model analysis capabilities are insufficient. Current methods rely heavily on manual annotation or simplified models for key point identification and motion trajectory analysis, lacking automated and precise three-dimensional knee joint modeling and key anatomical point identification technologies. This makes it difficult to achieve high-precision biomechanical parameter calculations and disease-related assessments. Thirdly, interaction and visualization are limited. Traditional systems lack flexible human-computer interfaces and customized analysis workflows, failing to meet diverse clinical application needs. The lack of real-time visualization and automatic diagnostic report generation capabilities further impairs clinicians' intuitive understanding and judgment of patients' dynamic states.

[0004] Therefore, it is necessary to design a three-dimensional knee joint movement analysis platform for clinical auxiliary diagnosis to solve the problems existing in the current technology. Summary of the Invention

[0005] In view of this, the present invention proposes a three-dimensional movement analysis platform for knee joints for clinical auxiliary diagnosis, aiming to solve the problems of low data accuracy, insufficient model analysis, and limited interactive visualization.

[0006] This invention proposes a three-dimensional knee joint movement analysis platform for clinical auxiliary diagnosis, comprising:

[0007] The module includes a data processing module, a model analysis module, a human-computer interaction control module, and a cloud publishing module.

[0008] The data processing module is used to capture raw motion data of the knee joint in three dimensions using Magic software, and to perform coordinate correction and model processing operations. The coordinate correction includes noise filtering and coordinate systematization of the raw motion data. The model processing operations include simplification of the knee joint three-dimensional model, marking of key anatomical points, and accuracy optimization to generate a standardized model group.

[0009] The model analysis module is used to seamlessly import the model group into the C3D Clinical Window software. The C3D Clinical Window provides real-time visualization of the three-dimensional movement trajectory of the knee joint, calculation of biomechanical parameters, and clinical pathology correlation assessment functions.

[0010] The human-computer interaction control module is used to write customized interaction logic based on the needs of clinical diagnostic applications; it integrates clinical protocol templates and provides a graphical user interface to realize user parameter input, analysis process control, real-time result monitoring, and automatic generation of diagnostic reports.

[0011] The cloud publishing module is used to convert diagnostic reports and interactive interfaces into PHP dynamic web page format, and deploy and publish them on the cloud server via HTTP protocol, supporting SSL encrypted transmission, multi-user concurrent access, and cross-platform remote calls.

[0012] Furthermore, when the data processing module performs noise filtering and coordinate systematization on the raw motion data, it includes:

[0013] The noise filtering uses an extended Kalman filter to define a state vector containing joint angle, angular velocity, and angular acceleration. The process noise covariance parameter is adaptively adjusted according to the kinematic model, and the tolerance range is increased when the joint acceleration change exceeds the acceleration change threshold. In the measurement update stage, the data streams of optical and inertial sensors are fused, and a residual threshold monitoring process is executed to remove abnormal measurement values.

[0014] The coordinate system is initialized and calibrated based on the preoperative skeletal CT scan model using a quaternion rotation matrix algorithm. The geometric center points of the distal femoral intercondylar fossa and the tibial plateau are extracted to construct a local coordinate system. The rotation operation is represented by quaternions, and the optimal rotation parameters are solved by minimizing the point cloud registration error, thus transforming the data into a unified anatomical coordinate system.

[0015] Furthermore, when the data processing module simplifies the knee joint 3D model, marks key anatomical points, and optimizes accuracy, it includes:

[0016] The 3D model of the knee joint is simplified by using a vertex clustering polygon reduction algorithm to divide the triangular mesh into a spatial octree structure. The consistency of vertex normal vectors is analyzed within the region nodes. Vertices are merged when the curvature change is lower than the curvature change threshold, preserving the topological details of the key anatomical region of the intercondylar fossa. Key anatomical points are marked by a pre-trained U-Net convolutional neural network.

[0017] Furthermore, when the data processing module implements key anatomical point labeling using a pre-trained U-Net convolutional neural network, it includes:

[0018] The pre-trained U-Net convolutional neural network includes downsampling and upsampling layers. It takes three-dimensional voxel grid data as input and outputs a probability distribution heatmap of key points at the inferior pole of the patella and the anterior horn of the meniscus. Accuracy optimization uses the iterative nearest point algorithm for model registration. The convergence condition is set as the root mean square error being lower than the error threshold or reaching the maximum number of iterations. It is combined with Laplacian smoothing filtering to eliminate surface noise interference.

[0019] Furthermore, when the C3D clinical window provides real-time visualization of the three-dimensional movement trajectory of the knee joint, calculation of biomechanical parameters, and assessment of clinicopathological relevance, it includes:

[0020] The biomechanical parameters are calculated in real time using an inverse dynamics model combined with a musculoskeletal simulation toolkit to calculate joint torque and contact force. The clinical pathology correlation assessment is based on a support vector machine classifier. The feature vector integrates the peak torque, contact force time integral, and trajectory asymmetry index in the gait cycle, and uses a radial basis function kernel function for pattern recognition. The classifier is trained on an osteoarthritis case database and outputs a quantitative risk score.

[0021] Furthermore, when calculating the biomechanical parameters using an inverse dynamics model combined with a musculoskeletal simulation toolbox to calculate joint torques and contact forces in real time, the following steps are included:

[0022] Based on kinematic data and ground reaction force input, the equations of motion, including mass matrix, Coriolis force term and gravity term, are solved; the contact force calculation adopts Hertz contact theory model, inputting meniscus curvature radius and material elastic modulus parameters, and generating a visualization cloud map of tibiofemoral joint pressure distribution.

[0023] Furthermore, the human-computer interaction control module incorporates customized interaction logic tailored to clinical diagnostic application needs; it integrates clinical protocol templates and provides a graphical user interface to enable user parameter input, analysis process control, real-time result monitoring, and automatic generation of diagnostic reports, including:

[0024] The human-computer interaction control module integrates customized interaction logic with a visual drag-and-drop process editor, supporting the definition of analysis step sequences through a graphical node interface; data flow between nodes is driven by an event-driven mechanism to perform parameter validity verification operations; the clinical protocol template includes an osteoarthritis grading assessment template, preset anatomical index thresholds, and loads patient historical data; the graphical user interface provides real-time 3D model interaction functions, and implements touch zoom operations, frame-by-frame playback of motion trajectories, and diagnostic report generation processes based on the WebGL rendering engine.

[0025] Furthermore, when the human-computer interaction control module performs the diagnostic report generation process based on the WebGL rendering engine, it includes:

[0026] It integrates visual elements into the interface and embeds dynamic biomechanical parameter charts; the interactive logic executes user parameter input, analysis process control, and real-time result monitoring operations.

[0027] Furthermore, when the cloud-based publishing module converts the diagnostic report and interactive interface into a PHP dynamic webpage format, it includes:

[0028] The cloud-based publishing module uses a model-view-controller architecture based on the Laravel framework to implement data processing logic. The controller responds to HTTP requests and calls the database model to extract biomechanical parameters. The view layer uses a template engine to dynamically render interface content, integrates AJAX asynchronous technology to execute a data loading process without page refresh, and updates results through lightweight requests. Interactive charts are generated using a visualization library configuration, supporting the display of joint torque polar coordinate graphs and pathological risk trend lines. Webpage performance is optimized through resource compilation and compression technology to execute the page loading process.

[0029] Furthermore, the cloud publishing module is deployed and published on a cloud server via the HTTP protocol, supporting SSL encrypted transmission, concurrent access by multiple users, and cross-platform remote calls, including:

[0030] In the HTTP protocol deployment, the SSL encrypted transmission uses the TLS protocol and a high-strength key to achieve end-to-end data protection; when the client connects, the server sends a digital certificate chain and generates a secure session key through a key exchange mechanism; the multi-user concurrent access is implemented through a load balancer to distribute traffic, and a weighted round-robin algorithm is configured to dynamically adjust the connection strategy according to the server load, and a session persistence mechanism is set; the cross-platform remote call supports RESTful API interface design, defines data access endpoints, and uses JSON data format to transmit structured information.

[0031] Compared with existing technologies, the advantages of this invention are as follows: By acquiring three-dimensional motion data of the knee joint using Magic software, and combining extended Kalman filters and quaternion rotation matrix algorithms for noise filtering and coordinate systematization, high-precision capture and standardized processing of multi-degree-of-freedom knee joint movements are achieved, improving data reliability and analytical accuracy. The three-dimensional model is simplified using a vertex clustering polygon reduction algorithm, and key anatomical points are automatically labeled using a pre-trained U-Net convolutional neural network. Accuracy optimization is achieved by combining iterative nearest-point algorithms and Laplace smoothing filters, making the three-dimensional knee joint model usable for high-precision biomechanical analysis. The C3D clinical window supports real-time visualization of the three-dimensional movement trajectory of the knee joint, inverse dynamics calculation of joint torque and contact force, and clinicopathological correlation assessment based on support vector machines, enabling gait cycle analysis, peak mechanical parameter statistics, and risk score output. The human-computer interaction control module, through a graphical drag-and-drop workflow editor and a WebGL rendering engine, enables interactive parameter input, analysis workflow control, real-time operation of the three-dimensional model, and automatic generation of diagnostic reports, improving the convenience of clinical use. The cloud-based deployment module adopts the Laravel MVC architecture, converting diagnostic reports and interactive interfaces into dynamic PHP web page format, deploying them via the HTTP protocol, and supporting SSL encryption, load balancing, multi-user concurrent access, and cross-platform RESTful API calls, thus achieving secure data sharing and remote collaboration. Attached Figure Description

[0032] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0033] Figure 1 This is a functional block diagram of a three-dimensional knee joint movement analysis platform for clinical auxiliary diagnosis provided in an embodiment of the present invention. Detailed Implementation

[0034] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the disclosure to those skilled in the art. It should be noted that, unless otherwise specified, embodiments and features in the embodiments of the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0035] Current technologies primarily employ a combination of medical imaging and computer modeling for 3D reconstruction and motion analysis of the knee joint. Knee joint images are acquired through MRI or CT scans, and medical image processing software (such as Mimics and Magic) is used to segment bones and soft tissues, thereby creating a 3D anatomical model. Subsequently, motion capture systems are used to collect motion data of the patient's knee joint in static or dynamic states, and registration algorithms are used to fuse this motion data with the 3D model. Based on this, doctors can visualize the knee joint's motion trajectory and calculate some biomechanical parameters using C3D clinical analysis software or similar platforms, providing a reference for surgical planning, rehabilitation training, and disease diagnosis. However, existing technologies often suffer from errors in coordinate registration due to data from different image sources, particularly at key anatomical landmarks (such as the lateral and medial epicondyles of the femur and the edge of the tibial plateau), affecting the reliability of the 3D model. Furthermore, existing images are mostly static scans, making it difficult to achieve high-resolution, continuous motion state modeling, resulting in discrepancies between the motion analysis results and actual physiological conditions. Most systems can only provide movement trajectory display and basic parameter calculation, lacking intelligent analysis, pathological prediction and personalized diagnosis and treatment support.

[0036] For example, in the pre-operative evaluation of anterior cruciate ligament (ACL) reconstruction surgery, the surgeon used MRI to obtain data on the knee ligaments and soft tissues, and supplemented the skeletal structure with CT scans. Due to slight changes in the patient's knee posture during the two imaging processes, coordinate shifts occurred after the MRI and CT data were fused, resulting in a registration error of 2-3 mm between the femur and tibia. This error directly led to inaccurate measurements of the ACL insertion point, causing deviations when drilling the graft tunnel during surgery. Post-operatively, the patient still experienced knee instability during walking and weight-bearing activities, necessitating a second corrective surgery, prolonging the recovery period and increasing medical costs.

[0037] In view of the above problems, it is particularly necessary to design a three-dimensional movement analysis platform for clinical auxiliary diagnosis of the knee joint.

[0038] For this, please refer to Figure 1 As shown, a three-dimensional movement analysis platform for knee joints used in clinical auxiliary diagnosis includes:

[0039] The module includes a data processing module, a model analysis module, a human-computer interaction control module, and a cloud publishing module.

[0040] The data processing module is used to capture raw motion data of the knee joint in three dimensions using Magic software, and to perform coordinate correction and model processing operations. The coordinate correction includes noise filtering and coordinate systematization of the raw motion data, and the model processing operations include simplification of the knee joint three-dimensional model, marking of key anatomical points and accuracy optimization, and generation of standardized model groups.

[0041] The model analysis module is used to seamlessly import model groups into the C3D Clinical Window software. The C3D Clinical Window provides real-time visualization of the three-dimensional movement trajectory of the knee joint, calculation of biomechanical parameters, and assessment of clinical pathology correlation.

[0042] The human-computer interaction control module is used to write customized interaction logic based on the needs of clinical diagnostic applications; it integrates clinical protocol templates and provides a graphical user interface to realize user parameter input, analysis process control, real-time result monitoring, and automatic generation of diagnostic reports;

[0043] The cloud publishing module is used to convert diagnostic reports and interactive interfaces into PHP dynamic web page format, and deploy and publish them on the cloud server via HTTP protocol. It supports SSL encrypted transmission, multi-user concurrent access, and cross-platform remote calls.

[0044] Specifically, the knee joint 3D motion analysis platform consists of a data processing module, a model analysis module, a human-computer interaction control module, and a cloud publishing module, forming a complete clinical auxiliary diagnosis and information sharing system. The data processing module processes the raw data obtained from knee joint 3D motion capture using Magic software. It performs coordinate correction through noise filtering and coordinate systematization, then simplifies the 3D model, marks key anatomical points, and optimizes accuracy, generating a standardized model group suitable for clinical use. The model analysis module seamlessly imports the standardized model group into the C3D clinical window software, supporting real-time 3D visualization of knee joint motion trajectories. It can calculate joint biomechanical parameters and perform pathological correlation analysis based on clinical needs, providing quantitative evidence for disease assessment and surgical planning. The human-computer interaction control module integrates the clinical diagnostic process with interactive logic, incorporates multiple clinical protocol templates, and provides a user-friendly graphical interface, supporting physicians in parameter input, analysis process control, real-time result monitoring, and automatic generation of diagnostic reports. The cloud-based publishing module converts the generated diagnostic reports and interactive interfaces into a dynamic PHP webpage format, enabling remote deployment via the HTTP protocol. It features SSL encryption, multi-user concurrent access, and cross-platform calling capabilities, thereby facilitating data sharing and remote diagnosis between different clinical institutions.

[0045] The working principle and process of this application are as follows: The data processing module is responsible for the collection and preprocessing of raw motion data. After acquiring 3D motion capture data of the knee joint using Magic software, noise filtering is performed on the raw data to eliminate errors caused by external interference. A unified world coordinate system is established through a coordinate systematization method to ensure the consistency and accuracy of subsequent model analysis. Based on this, the 3D anatomical model of the knee joint is processed, including model simplification to reduce redundant data, marking key anatomical points for clinical identification, and accuracy optimization to improve the reliability of the model in kinetic and biomechanical analysis, ultimately generating a standardized model group. The model analysis module seamlessly imports the standardized model group into the C3D clinical window software. This module presents the trajectory changes of the knee joint under different motion states through 3D visualization, providing an intuitive display of joint dynamic kinetics. Simultaneously, it can perform real-time calculations of key motion parameters (such as joint angles, displacements, loads, and stress distribution) and combine them with a pathological information database to achieve disease correlation analysis, thereby providing doctors with scientific quantitative diagnostic evidence. The human-computer interaction control module provides clinicians with an intuitive operating interface. The module incorporates customized interactive logic and clinical protocol templates, allowing users to input parameters, control the analysis process, and monitor results through the interface. In the background, calculations and data retrieval are automatically performed based on user-defined diagnostic needs, and standardized diagnostic reports are automatically generated upon completion of the analysis, shortening the clinical diagnostic workflow. The cloud-based publishing module enables remote sharing of diagnostic results. This module converts the diagnostic report and interactive interface into dynamic PHP web pages, deploys them on a cloud server via HTTP, and uses SSL encryption to ensure data transmission security. It also supports concurrent access by multiple users and cross-platform remote access, allowing doctors to quickly retrieve results on different terminal devices, facilitating multi-center consultations and telemedicine applications. Through a closed-loop working mechanism of "data processing—model analysis—human-computer interaction—cloud publishing," the platform achieves accurate acquisition, intelligent analysis, visualization, and remote sharing of knee joint 3D motion data.

[0046] As a preferred embodiment, the specific implementation of this application is as follows: The data processing module uses Magic software to process the raw three-dimensional movement data of the patient's knee joint. Noise in the raw data is filtered out using an extended Kalman filter, and an anatomical coordinate system is established based on the skeletal structure of the patient's preoperative CT scan to systematize the coordinates of the movement data. A polygon simplification algorithm is used to simplify the structure of the three-dimensional knee joint model. A pre-trained U-Net convolutional neural network is used to mark key anatomical points such as the lateral epicondyle of the femur, the medial epicondyle of the femur, the tibial plateau, and the inferior pole of the patella. The iterative nearest point (ICP) algorithm is combined to optimize model accuracy, generating a standardized model group. The model analysis module seamlessly imports the above model group into the C3D clinical window software to achieve real-time visualization of the three-dimensional movement trajectory of the knee joint. Joint torque and contact force parameters are calculated based on the inverse dynamics model. A support vector machine classifier is used to perform pattern recognition on kinetic indicators and biomechanical features in the gait cycle, outputting an osteoarthritis-related risk score, thereby achieving a clinical pathological correlation assessment. The human-computer interaction control module pre-configures osteoarthritis grading protocol templates based on clinical diagnostic needs and uses a graphical user interface to enable parameter input, analysis process control, real-time result monitoring, and automatic generation of diagnostic reports. Doctors can select a diagnostic template and input basic patient information within the interface, automatically generating a diagnostic report containing kinematic curves, biomechanical parameter charts, and pathological risk results. The cloud-based publishing module converts the diagnostic report and interactive interface into a dynamic PHP webpage format and deploys it to a cloud server via HTTP, supporting SSL encrypted transmission, multi-user concurrent access, and cross-platform remote access, thereby enabling remote sharing and collaborative diagnosis between patients and multiple medical institutions.

[0047] Through the aforementioned technical solutions, this application employs Magic software combined with extended Kalman filtering and coordinate systematization methods to reduce noise interference. Furthermore, by simplifying the 3D model and marking key anatomical points, it achieves high-precision standardized model group construction, thereby improving the accuracy of kinetic and biomechanical analysis. The standardized model is seamlessly imported into the C3D clinical window, enabling real-time visualization of the knee joint's 3D motion trajectory. Combined with inverse dynamics calculations and pathological correlation assessment models, it can quickly output quantified biomechanical parameters and disease risk scores. The human-computer interaction control module supports customized diagnostic logic and protocol templates, providing an intuitive graphical user interface. It enables parameter input, process control, real-time monitoring, and automatic report generation, reducing manual operations for doctors and improving clinical work efficiency. The cloud publishing module converts diagnostic reports and interactive interfaces into dynamic PHP web pages and deploys them to the cloud, supporting SSL encryption, multi-user concurrent access, and cross-platform remote calls, achieving data sharing between patients and multiple hospitals.

[0048] This application further proposes a data processing module that performs noise filtering and coordinate systematization on the raw moving data, including:

[0049] Noise filtering uses an extended Kalman filter to define a state vector containing joint angle, angular velocity, and angular acceleration. The process noise covariance parameter is adaptively adjusted according to the kinetic model, and the tolerance range is increased when the joint acceleration change exceeds the acceleration change threshold. In the measurement update stage, optical and inertial sensor data streams are fused, and a residual threshold monitoring process is executed to remove abnormal measurement values.

[0050] The coordinate system is initialized and calibrated based on the preoperative skeletal CT scan model using a quaternion rotation matrix algorithm. The geometric center points of the distal femoral intercondylar fossa and the tibial plateau are extracted to construct a local coordinate system. The rotation operation is represented by quaternions, and the optimal rotation parameters are solved by minimizing the point cloud registration error, thus transforming the data into a unified anatomical coordinate system.

[0051] Specifically, the data processing module performs noise filtering and coordinate systematization on the raw data from the three-dimensional motion capture of the knee joint to ensure the accuracy and reliability of subsequent analysis. Noise filtering employs an extended Kalman filter (EKF) to model and predict the state vector containing joint angles, angular velocities, and angular accelerations. Through a kinetic model, the process noise covariance parameter is adaptively adjusted to handle data fluctuations under different movement states. When the change in joint acceleration exceeds a preset threshold, the process noise tolerance range is automatically increased to avoid the impact of abnormal deviations caused by sudden movements or sensor jitter on the filtering results. The preset threshold rules can be determined based on clinical statistical data. For example, the normal acceleration variation range during the gait cycle is 0–15 rad / s², and when the instantaneous acceleration change exceeds 15 rad / s², the tolerance increase processing is triggered; for jumping or rapid turning movements, the threshold can be dynamically adjusted to 30 rad / s² to ensure stable output from the filter under intense movement. In the measurement update phase, optical capture data and inertial measurement unit (IMU) data streams are integrated, and abnormal measurement values ​​are eliminated through a residual threshold monitoring mechanism. The residual threshold can be set to mark an anomaly when the root mean square error (RMSE) of the difference between the predicted state and the actual measured value exceeds 0.5°, and the data point is either removed or corrected by interpolation at a nearby time step. Coordinate systematization is achieved using a quaternion rotation matrix algorithm. A precise 3D skeletal model is constructed using preoperative skeletal CT scan data, and a local coordinate system is initialized at the geometric center points of the distal femoral intercondylar fossa and the tibial plateau. The original 3D motion data is represented by rotation operations using quaternions, and the optimal rotation parameters from the sensor coordinate system to the anatomical coordinate system are calculated. The optimal rotation parameters are obtained by minimizing the point cloud registration error, ensuring a high degree of matching between the joint motion trajectory and the skeletal model. All original motion data are transformed to a unified anatomical coordinate system.

[0052] As a preferred embodiment, the specific implementation of this application is as follows: A patient undergoes three-dimensional motion capture of the knee joint, and gait data is synchronously acquired using an optical motion capture system and an inertial measurement unit (IMU). In the raw acquired data, due to the patient's instantaneous shaking during rapid starts and turns, the joint acceleration fluctuation reaches 20 rad / s², significantly higher than the normal gait cycle range of 0–15 rad / s². In the data processing module, the extended Kalman filter automatically detects this abnormal acceleration change, triggering a process noise covariance parameter amplification mechanism, enabling the filter to stably output joint angle and angular velocity information even under high-dynamic movements. Simultaneously, it fuses IMU data to correct deviations caused by lost or obstructed optical data. The residual threshold is set to mark an anomaly if the RMSE exceeds 0.5°, automatically removing abnormal data points at the corresponding time step, and smoothing through interpolation from adjacent time steps. In the coordinate systematization stage, the three-dimensional models of the femur and tibia established by preoperative CT scans are used as the reference coordinate system. The geometric center points of the distal femoral intercondylar fossa and the tibial plateau are automatically extracted to construct a local anatomical coordinate system. The original captured data was rotated to a unified anatomical coordinate system using a quaternion rotation matrix algorithm. The optimal rotation parameters were determined by minimizing the point cloud registration error to ensure a high degree of match between the patient's knee joint movement trajectory and the skeletal model.

[0053] Through the above technical solutions, this application employs an extended Kalman filter to model the joint angle, angular velocity, and acceleration state vectors, and combines this with a kinetic model to adaptively adjust the process noise covariance parameter, reducing the impact of sensor noise and external interference on motion data. During the measurement update phase, optical capture and inertial measurement unit (IMU) data are fused, and anomaly measurements are eliminated through a residual threshold monitoring mechanism, ensuring data continuity and authenticity and improving the accuracy of subsequent analysis. By initializing the local coordinate system based on preoperative skeletal CT scans, using a quaternion rotation matrix algorithm, and minimizing point cloud registration errors to solve for the optimal rotation parameters, the original motion data is transformed into a unified anatomical coordinate system, ensuring high comparability and consistency of data from different acquisition environments and patients.

[0054] This application further proposes a data processing module for simplifying, marking key anatomical points, and optimizing the accuracy of the knee joint 3D model, including:

[0055] The 3D model of the knee joint is simplified by using a vertex clustering polygon reduction algorithm to divide the triangular mesh into a spatial octree structure. The consistency of vertex normal vectors is analyzed within the region nodes. Vertices are merged when the curvature change is lower than the curvature change threshold, preserving the topological details of the key anatomical region of the intercondylar fossa. Key anatomical points are marked by a pre-trained U-Net convolutional neural network.

[0056] Specifically, when simplifying, marking key anatomical points, and optimizing the accuracy of the knee joint 3D model, the data processing module divides the original triangular mesh into a spatial octree structure using a vertex clustering polygon reduction algorithm. Within each region node, the consistency of vertex normal vectors is analyzed to determine local curvature changes on the mesh surface. When the curvature change is lower than a preset curvature change threshold (e.g., set to 0.02 rad / unit length), vertices within that region are automatically merged to simplify the model. Simultaneously, topological details are preserved in the intercondylar fossa and other key anatomical regions, ensuring that the accuracy of the model's critical structures is not affected. In the key anatomical point marking stage, a pre-trained U-Net convolutional neural network processes the simplified 3D model. Inputting 3D voxel mesh data, the network outputs probability distribution heatmaps of key anatomical points such as the inferior pole of the patella and the anterior horn of the meniscus. These heatmaps, combined with a probability threshold (e.g., 0.7), determine the final marking positions, thereby achieving high-precision, automated key point identification.

[0057] As a preferred embodiment, the solution of this application is implemented as follows: In a clinical knee gait analysis, a high-resolution three-dimensional mesh model of the patient's right knee joint is acquired, containing an original triangular mesh with approximately 1.2 million vertices. To facilitate subsequent real-time analysis and storage, the data processing module first divides the triangular mesh into a spatial octree structure, statistically analyzes the vertex normal vectors within each node, and calculates the local curvature change. When the curvature change is lower than a preset threshold of 0.02 rad / unit length, the vertices in that region are clustered and merged, thereby reducing the number of model vertices to approximately 300,000. Simultaneously, the original topological details are preserved in the key anatomical regions of the intercondylar fossa and meniscus to ensure structural accuracy. The simplified three-dimensional model is input into a pre-trained U-Net convolutional neural network. The network outputs a probability distribution heatmap of the inferior pole of the patella and the anterior horn of the meniscus, and a probability threshold of 0.7 is set to determine the final key point locations. This process successfully and automatically marked key anatomical points such as the inferior pole of the patella, the anterior horns of the medial and lateral menisci, and the intercondylar fossa of the femur, providing high-precision basic data for subsequent three-dimensional motion trajectory analysis, joint biomechanical calculation, and clinical pathological evaluation.

[0058] Through the above technical solutions, this application employs a vertex clustering polygon reduction algorithm combined with a spatial octree structure to rationally merge mesh vertices, significantly reducing the number of vertices in the knee joint 3D model, alleviating storage and computational burdens while ensuring the integrity of the local topology. Original topological details are preserved in key anatomical areas such as the intercondylar fossa and meniscus, avoiding the loss of important anatomical information during model simplification. A pre-trained U-Net convolutional neural network is used to label key anatomical points on the simplified 3D model, achieving automated identification of key points such as the inferior pole of the patella, the anterior horn of the meniscus, and the intercondylar fossa of the femur, reducing manual annotation errors and improving analysis efficiency.

[0059] This application further proposes a data processing module that, when implementing key anatomical point labeling using a pre-trained U-Net convolutional neural network, includes:

[0060] The pre-trained U-Net convolutional neural network includes downsampling and upsampling layers. It takes three-dimensional voxel grid data as input and outputs a probability distribution heatmap of key points at the inferior pole of the patella and the anterior horn of the meniscus. Accuracy optimization uses the iterative nearest point algorithm for model registration. The convergence condition is set as the root mean square error being lower than the error threshold or reaching the maximum number of iterations. It also combines Laplacian smoothing filtering to eliminate surface noise interference.

[0061] Specifically, when labeling key anatomical points on a 3D model of the knee joint, the simplified and preliminarily optimized 3D model is converted into voxel mesh data and input into a pre-trained U-Net convolutional neural network. This network consists of symmetrical downsampling and upsampling layers. The downsampling layers progressively extract multi-scale features through convolution and pooling operations, while the upsampling layers progressively restore spatial resolution through deconvolution, outputting a probability distribution heatmap of key points at the inferior pole of the patella and the anterior horn of the meniscus. The value of each voxel in the probability heatmap represents the probability that the point is a target key point. A threshold of 0.7 is set; that is, when the probability is higher than 0.7, the point is identified as a key anatomical point to ensure labeling accuracy and avoid misclassification of low-confidence points. After the initial labeling of key points, the Iterative Closest Point (ICP) algorithm is further used to optimize model accuracy. This is achieved by registering the current model with a reference bone model generated from a preoperative CT scan, progressively minimizing the root mean square error (RMSE). The convergence condition of the ICP algorithm is set as follows: iteration stops when the RMSE is below 0.5 mm or the maximum number of iterations is reached (50), in order to balance computational efficiency and registration accuracy. To eliminate possible noise and outliers on the model surface, Laplacian smoothing filtering is also used to process the model surface, and local bumps and depressions are corrected by weighted averaging of neighboring vertices to improve surface continuity.

[0062] As a preferred embodiment, the specific implementation of this application is as follows: After a patient undergoes a three-dimensional motion capture scan of the knee joint, the generated three-dimensional model is input into a pre-trained U-Net convolutional neural network. After network processing, a probability distribution heatmap of key points at the inferior pole of the patella and the anterior horn of the meniscus is output. In the heatmap, the probability value of the inferior pole of the patella is 0.85, and the probability value of the anterior horn of the meniscus is 0.78, both higher than the set threshold of 0.7. These locations are marked as key anatomical points. The three-dimensional model is iteratively registered with the standard bone model generated by the preoperative CT scan. The ICP algorithm reaches a root mean square error of 0.42 mm at the 35th iteration, which is lower than the error threshold of 0.5 mm, and the registration stops. Laplace smoothing filtering is used to eliminate local unevenness on the model surface caused by scanning noise, making the model surface continuous and smooth, generating a final high-precision three-dimensional knee joint model that can be used for biomechanical analysis and clinical diagnosis.

[0063] Through the above technical solutions, the pre-trained U-Net convolutional neural network of this application can automatically identify the positions of key points such as the inferior pole of the patella and the anterior horn of the meniscus, and generate probability distribution heatmaps, thereby improving the accuracy and reliability of key point identification. During the accuracy optimization process, the iterative nearest-point algorithm ensures accurate registration between the 3D model and the standard skeletal model by setting a convergence condition where the root mean square error is below a preset threshold or the maximum number of iterations is reached, avoiding deviations caused by scanning noise or motion errors. Combined with Laplacian smoothing filtering, surface noise interference is further eliminated, resulting in a continuously smooth surface for the 3D knee joint model.

[0064] This application further proposes that when the C3D clinical window provides real-time visualization of the three-dimensional movement trajectory of the knee joint, calculation of biomechanical parameters, and assessment of clinical pathology relevance, it includes:

[0065] Biomechanical parameters are calculated in real time using an inverse dynamics model combined with a musculoskeletal simulation toolkit to calculate joint torque and contact force. Clinical pathology correlation assessment is based on a support vector machine classifier. The feature vector integrates peak torque, contact force time integral, and trajectory asymmetry index in the gait cycle, and uses radial basis kernel function for pattern recognition. The classifier is trained on an osteoarthritis case database and outputs a quantitative risk score.

[0066] Specifically, in the C3D clinical window, the real-time visualization function of the knee joint's three-dimensional movement trajectory can dynamically display the rotation angles, movement trajectories, and positional changes of each joint in the knee joint during the patient's gait cycle, allowing clinicians to intuitively observe the knee joint's movement status. Biomechanical parameter calculations employ an inverse dynamics model, combined with a musculoskeletal simulation toolkit. Based on collected kinetic data and ground reaction forces, joint torques and contact forces are calculated in real time, accurately reflecting the force distribution on the knee joint at different stages of movement. Clinical pathology correlation assessment is implemented based on a support vector machine (SVM) classifier. Key features such as peak joint torques, contact force time integrals, and movement trajectory asymmetry during the gait cycle are integrated into a multi-dimensional feature vector. A radial basis function kernel is used for pattern recognition, and by training the model on an osteoarthritis case database, a quantitative risk score is output.

[0067] As a preferred embodiment, the specific implementation of this application is as follows: For a 65-year-old patient with early-stage osteoarthritis, knee joint displacement and angular velocity data during walking are acquired through three-dimensional motion capture. The C3D clinical window renders the captured data into a three-dimensional motion trajectory in real time, and calculates the peak knee flexion-extension torque as 35 Nm and the varus / valgus contact force as 250 N using an inverse dynamics model combined with a musculoskeletal simulation toolbox. The peak torque, contact force time integral, and trajectory asymmetry index within the gait cycle are integrated into a feature vector, which is then input into a pre-trained support vector machine classifier for analysis. The classifier uses the radial basis function kernel to identify patterns and, combined with an osteoarthritis case database, generates a quantitative risk score of 0.78 (out of 1.0 indicating high risk), suggesting that the patient's knee joint load is high and there is an early degenerative trend, thus providing a diagnostic basis for doctors.

[0068] Through the above technical solutions, the C3D clinical window of this application enables clinicians to intuitively observe the dynamic changes of the joint during walking or movement by visualizing the three-dimensional movement trajectory of the knee joint in real time; the biomechanical parameter calculation, combined with the inverse dynamics model and musculoskeletal simulation toolbox, can accurately calculate the torque and contact force of the joint at different stages of movement, providing a quantitative basis for knee joint load analysis; the clinical pathology correlation assessment is based on a support vector machine classifier, which integrates the peak torque, contact force time integral and trajectory asymmetry index in the gait cycle, uses radial basis kernel function for pattern recognition, and outputs a quantitative risk score after training on an osteoarthritis case database, providing doctors with an assessment of disease progression.

[0069] This application further proposes a method for calculating biomechanical parameters using an inverse dynamics model combined with a musculoskeletal simulation toolkit to calculate joint torques and contact forces in real time, including:

[0070] Based on kinetic data and ground reaction force input, the action equations including the mass matrix, Coriolis force terms, and gravity terms are solved; the contact force calculation adopts the Hertz contact theory model, inputting the meniscus curvature radius and material elastic modulus parameters to generate a visual cloud map of the tibiofemoral joint pressure distribution.

[0071] Specifically, in the calculation of biomechanical parameters, kinetic data is collected to obtain displacement, velocity, and acceleration information of the patient's knee joint during walking or flexion-extension movements, while ground reaction force data is acquired as an external constraint input. Based on these inputs, the action equations of the joint are solved using an inverse dynamics model, including the mass matrix, Coriolis force terms, and gravity terms, to accurately calculate the joint torques at each stage of action, achieving a quantitative analysis of the internal forces on the knee joint. For the calculation of joint contact forces, the Hertz contact theory model is used, with the meniscus curvature radius and material elastic modulus parameters input into the model to calculate the pressure distribution on the femoral-tibial contact surface, and a tibial-femoral joint pressure cloud map is generated through three-dimensional visualization.

[0072] As a preferred embodiment, the solution of this application is implemented as follows: During a walking analysis of a 55-year-old patient with knee osteoarthritis, three-dimensional motion data of the knee joint and ground reaction forces are collected using a high-resolution motion capture system. The collected kinematic data is input into an inverse dynamics model, and combined with a musculoskeletal simulation toolbox, the flexion-extension torque, valgus-valgus torque, and rotational torque of the knee joint during the walking cycle are calculated in real time. The radius of curvature of the meniscus is set to 12 mm, and the elastic modulus is set to 5 MPa. The Hertz contact theory model is used to calculate the contact force distribution between the femur and tibia, and a three-dimensional pressure cloud map is generated. This cloud map can visually show that the pressure of the tibiofemoral joint is concentrated in the lateral condyle region, indicating that the patient's lateral cartilage is under excessive stress, assisting doctors in developing personalized rehabilitation training programs or surgical intervention plans, thereby achieving accurate clinical diagnosis and treatment decisions.

[0073] Through the above technical solution, this application achieves a quantitative assessment of joint load by using real-time biomechanical analysis based on an inverse dynamics model to accurately calculate the anisotropic torque and contact force distribution of the knee joint under different action states. By using kinetic data and ground reaction force input, and combining the mass matrix, Coriolis force term, and gravity term to solve the action equation, torque variation curves of the joint in flexion-extension, varus-valgus, and rotation directions can be obtained. The contact force distribution between the femur and tibia is calculated using the Hertz contact theory model, and a three-dimensional pressure cloud map is generated by combining the meniscus curvature and elastic modulus, showing the stress concentration area of ​​the joint.

[0074] This application further proposes a human-computer interaction control module that incorporates customized interaction logic based on clinical diagnostic application needs; it integrates clinical protocol templates and provides a graphical user interface to enable user parameter input, analysis process control, real-time result monitoring, and automatic generation of diagnostic reports, including:

[0075] The human-computer interaction control module integrates customized interaction logic with a visual drag-and-drop workflow editor, supporting the definition of analysis step sequences through a graphical node interface; data flow between nodes is driven by an event-driven mechanism to perform parameter validity verification operations; clinical protocol templates include osteoarthritis grading assessment templates, preset anatomical index thresholds, and load patient historical data; the graphical user interface provides real-time 3D model interaction functions, and implements touch zoom operations, frame-by-frame playback of movement trajectories, and diagnostic report generation processes based on the WebGL rendering engine.

[0076] Specifically, the human-computer interaction control module provides a comprehensive and flexible operating and analysis environment by closely integrating customized interaction logic with clinical diagnostic application needs. This module integrates a visual drag-and-drop workflow editor, allowing users to customize the order and execution conditions of analysis steps through a graphical node interface. Data flow between each node uses an event-driven mechanism for real-time parameter validity verification to ensure that input data and intermediate calculation results meet preset standards. The clinical protocol template includes an osteoarthritis grading assessment scheme, setting anatomical indicator thresholds for key anatomical points such as the distal femoral intercondylar fossa, tibial plateau, and patellar position. For example, the preset threshold for knee varus / valgus angle is ±5°, and an abnormal warning is triggered when the joint space width is less than 3 mm. Historical patient data can also be loaded for longitudinal comparison. The graphical user interface provides real-time 3D model interaction, supporting touch zoom, rotation, and frame-by-frame playback of movement trajectories based on a WebGL rendering engine, allowing users to dynamically observe the joint's movement status throughout the gait cycle. Simultaneously, a diagnostic report is automatically generated, integrating key movement parameters, abnormal indicators, and 3D visualization screenshots, outputting a standardized document format.

[0077] As a preferred embodiment, the solution of this application is implemented as follows: When performing knee gait analysis on a male patient, the clinician initiates a customized interactive logic process through the human-computer interaction control module, sets analysis steps in the graphical node interface, including data import, joint angle calculation, three-dimensional trajectory generation, and anomaly detection. Each node verifies the validity of the input data through an event-driven mechanism, such as checking whether the varus / valgus angle is within a preset threshold of ±5°. The osteoarthritis grading assessment template is invoked, and the patient's historical images and gait data are automatically loaded. The varus / valgus angle, patellofemoral joint space width, and peak joint torque are compared and analyzed. The graphical user interface displays a three-dimensional knee joint model using WebGL rendering. The doctor can touch to rotate and zoom, replaying the patient's entire gait cycle movement trajectory frame by frame. Simultaneously, a diagnostic report is automatically generated, marking abnormal indicators in red, and including a three-dimensional visualization screenshot and quantitative data, such as a patellofemoral joint space of 2.8 mm, exceeding the threshold range, indicating an early risk of osteoarthritis.

[0078] Through the aforementioned technical solution, the applicant's human-computer interaction control module integrates a visual drag-and-drop workflow editor, enabling clinicians to flexibly define the sequence of analysis steps according to the diagnostic needs of different patients. Each node has an event-driven data verification mechanism that automatically checks whether input parameters are within preset threshold ranges, such as key indicators like knee varus / valgus angle, patellofemoral joint space, and peak joint torque, and issues warnings when abnormalities occur. Using the osteoarthritis grading assessment template, the system automatically loads historical patient images and gait data, combining them with current 3D motion capture results to achieve indicator comparison and risk scoring. The graphical user interface, based on a WebGL rendering engine, supports touch rotation, zoom, and frame-by-frame playback of gait trajectories, displaying real-time joint 3D models and changes in motion parameters, while automatically generating reports containing visual screenshots, quantitative data, and diagnostic conclusions.

[0079] This application further proposes a process for generating diagnostic reports based on the WebGL rendering engine in the human-computer interaction control module, including:

[0080] It integrates visual elements into the interface and embeds dynamic biomechanical parameter charts; the interactive logic executes user parameter input, analysis process control, and real-time result monitoring operations.

[0081] Specifically, in the human-computer interaction control module, the diagnostic report generation process based on the WebGL rendering engine embeds the 3D knee joint model, gait trajectory, and biomechanical parameters such as joint torque and contact force into the interface through dynamic charts for visualization. Users can input basic patient information and analysis parameters, such as gait cycle, joint load threshold, and key anatomical point selection, through the interface. The analysis process is executed in real time, updating the 3D model's movement trajectory and related biomechanical data, and automatically annotating and prompting abnormal parameters. The module integrates real-time monitoring functions, which can dynamically generate diagnostic conclusions, biomechanical distribution cloud maps, and quantitative statistical tables during the analysis process, and finally automatically generate a comprehensive diagnostic report containing interactive charts and 3D model screenshots.

[0082] As a preferred embodiment, the solution of this application is implemented as follows: In a knee gait analysis, the doctor inputs the patient ID, gait data acquisition path, and analysis cycle parameters in the WebGL interface through the human-computer interaction control module. A three-dimensional knee joint model is automatically loaded, and real-time joint torque and contact force distribution charts are embedded in the interface. The user can rotate and scale the three-dimensional model by dragging, replaying the knee joint movement trajectory frame by frame, while observing the peak joint torque and changes in tibiofemoral contact force. When the detected torque or contact force exceeds a preset threshold, the interface automatically highlights the abnormal area. After the analysis is completed, the module integrates the movement trajectory, biomechanical cloud map, and quantitative statistical table to generate an interactive diagnostic report. The doctor can directly view the three-dimensional model, dynamic charts, and key parameters in the report, providing an intuitive basis for clinical decision-making.

[0083] Through the above technical solution, this application generates a diagnostic report based on the WebGL rendering engine through a human-computer interaction control module, which can realize real-time visualization and dynamic interaction of knee joint movement data and biomechanical parameters. Users can intuitively observe the movement trajectory, joint torque and contact force distribution of the three-dimensional model in the same interface, and monitor the analysis process and parameter changes in real time, thereby improving diagnostic efficiency and accuracy. By embedding dynamic charts and interactive controls, a structured and visualized diagnostic report can be automatically generated.

[0084] This application further proposes that when the cloud publishing module converts the diagnostic report and interactive interface into a PHP dynamic web page format, it includes:

[0085] The cloud-based publishing module uses a model-view-controller architecture based on the Laravel framework to implement data processing logic. The controller responds to HTTP requests and calls the database model to extract biomechanical parameters. The view layer uses a template engine to dynamically render interface content, integrates AJAX asynchronous technology to execute a data loading process without page refresh, and updates results through lightweight requests. Interactive charts are generated using a visualization library configuration, supporting the display of joint torque polar coordinate graphs and pathological risk trend lines. Web page performance is optimized through resource compilation and compression technology to execute the page loading process.

[0086] Specifically, by converting diagnostic reports and interactive interfaces into a dynamic PHP webpage format through a cloud-based publishing module, the platform achieves efficient separation between data processing logic and front-end presentation. Specifically, the Laravel framework's model-view-controller architecture enables the controller to accurately extract knee joint biomechanical parameters and analysis results from the database model upon receiving an HTTP request; the view layer dynamically renders interface content through a template engine and embeds AJAX asynchronous technology to achieve data loading without page refresh, allowing users to view the latest calculation results in real time while browsing the webpage without manual refresh; interactive charts utilize a visualization library to generate joint torque polar coordinate graphs and pathological risk trend lines, which can be dynamically updated according to the user-selected time point or analysis parameters.

[0087] As a preferred embodiment, the solution of this application is implemented as follows: In the knee joint three-dimensional movement analysis platform, when a doctor selects the gait analysis results of a patient on the remote diagnosis interface, the cloud publishing module immediately triggers the Laravel controller to respond to the HTTP request, calling the database model to extract the patient's knee joint torque data, contact force distribution, and pathological risk score. The view layer dynamically generates web pages through a template engine, embedding the data into interactive charts. The joint torque is displayed as a polar coordinate graph showing the peak changes of different gait cycles, and the pathological risk changes over time as a trend line. AJAX asynchronous technology allows doctors to view the updated results in real time without refreshing the page when switching between different analysis time periods or parameter settings. At the same time, the web page achieves fast loading and smooth operation through resource compression optimization.

[0088] Through the above technical solution, this application utilizes a model-view-controller architecture based on the Laravel framework, and a cloud-based publishing module to dynamically generate diagnostic reports and interactive web pages. The controller responds to HTTP requests and retrieves biomechanical parameters from the database. The view layer renders content using a template engine and integrates AJAX to achieve refresh-free data updates. Interactive charts support the display of joint torque polar coordinate graphs and pathological risk trend lines. Web page performance is optimized through resource compression, achieving fast loading and smooth operation.

[0089] This application further proposes a cloud-based publishing module deployed and published on a cloud server via the HTTP protocol, supporting SSL encrypted transmission, concurrent access by multiple users, and cross-platform remote invocation, including:

[0090] In HTTP protocol deployment, SSL encrypted transmission uses the TLS protocol and high-strength keys to achieve end-to-end data protection; when the client connects, the server sends a digital certificate chain and generates a secure session key through a key exchange mechanism; concurrent access by multiple users is handled by a load balancer to distribute traffic, and a weighted round-robin algorithm is configured to dynamically adjust the connection strategy according to the server load, and a session persistence mechanism is set; cross-platform remote calls support RESTful API interface design, define data access endpoints, and transmit structured information using JSON data format.

[0091] Specifically, the cloud-based publishing module is deployed and published on a cloud server via the HTTP protocol. It supports SSL encrypted transmission, concurrent access by multiple users, and cross-platform remote calls, ensuring data security, access stability, and compatibility. SSL encrypted transmission uses the TLS protocol and high-strength keys for end-to-end protection. Clients receive a digital certificate chain when connecting to the server and generate independent session keys through a key exchange mechanism to prevent data interception or tampering during transmission. Concurrent access by multiple users is handled by a load balancer, using a weighted round-robin algorithm to dynamically adjust server request allocation strategies. A session persistence mechanism is also configured to ensure that session state is not lost when users switch between different server nodes, improving response speed and stability. For cross-platform remote calls, the cloud-based publishing module provides a RESTful API interface, defining clearly defined data access endpoints and using JSON format for structured information transmission. This allows for easy access from mobile devices, desktops, and other third parties, enabling remote access and integrated management of diagnostic reports and biomechanical data.

[0092] As a preferred embodiment, the solution of this application is implemented as follows: A medical institution uploads the diagnostic report and interactive interface of a knee joint 3D movement analysis platform to a cloud server. Doctors and rehabilitation therapists can access patient data via web or mobile devices. SSL encrypted transmission uses the TLS 1.3 protocol and a 256-bit high-strength key to achieve end-to-end data protection. When a client initiates a connection request, the server returns a digital certificate chain. The client verifies the certificate and exchanges keys with the server to generate an independent session key, ensuring that all sensitive information (such as patient personal information, joint torque, and contact force data) is not intercepted or tampered with during transmission. For concurrent access by multiple users, the cloud deployment module uses a load balancer to distribute traffic to multiple server nodes, employing a weighted round-robin algorithm to dynamically allocate requests based on the current load of each node. A session persistence mechanism is also set up to ensure that users do not lose session state when switching between different nodes. For example, a doctor can continuously view different angles and movement trajectories without interruption while accessing a patient's 3D model. For cross-platform remote access, the platform provides a RESTful API interface that transmits structured data in JSON format. This allows hospital electronic medical record systems, mobile rehabilitation applications, or third-party analytics tools to access the interface and obtain real-time biomechanical parameters and diagnostic reports, enabling data sharing and remote analysis. For example, rehabilitation therapists can access the interface on a tablet to obtain a patient's knee joint load distribution map, directly guiding rehabilitation training programs.

[0093] Through the above technical solutions, the cloud publishing module of this application is deployed on a cloud server via the HTTP protocol, supports SSL / TLS encryption to achieve end-to-end data protection, and the client establishes a secure session through digital certificates and key exchange; concurrent access by multiple users is distributed through a load balancer and a weighted round-robin algorithm, and the session state is maintained; cross-platform remote calls provide a RESTful API interface, and structured data is transmitted in JSON format, realizing efficient and secure data access and sharing.

[0094] In summary, by acquiring 3D knee joint motion data using Magic software and combining extended Kalman filter and quaternion rotation matrix algorithm for noise filtering and coordinate systematization, high-precision capture and standardized processing of multi-degree-of-freedom knee joint movements were achieved, improving data reliability and analytical accuracy. The 3D model was simplified using a vertex clustering polygon reduction algorithm, and key anatomical points were automatically labeled using a pre-trained U-Net convolutional neural network. Accuracy optimization was achieved by combining iterative nearest-point algorithm and Laplace smoothing filter, enabling the 3D knee joint model to be used for high-precision biomechanical analysis. The C3D clinical window supports real-time visualization of the 3D knee joint movement trajectory, inverse dynamics calculation of joint torque and contact force, and clinicopathological correlation assessment based on support vector machine, realizing gait cycle analysis, peak mechanical parameter statistics, and risk score output. The human-computer interaction control module, through a graphical drag-and-drop workflow editor and WebGL rendering engine, enables interactive parameter input, analysis workflow control, real-time operation of the 3D model, and automatic generation of diagnostic reports, improving the convenience of clinical use. The cloud-based deployment module adopts the Laravel MVC architecture, converting diagnostic reports and interactive interfaces into dynamic PHP web page format, deploying them via the HTTP protocol, and supporting SSL encryption, load balancing, multi-user concurrent access, and cross-platform RESTful API calls, thus achieving secure data sharing and remote collaboration.

[0095] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program goods. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program goods embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0096] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program goods according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0097] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0098] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0099] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A three-dimensional movement analysis platform for knee joints used in clinical auxiliary diagnosis, characterized in that, include: Data processing module, model analysis module, human-computer interaction control module, and cloud publishing module; The data processing module is used to capture raw motion data of the knee joint in three dimensions using Magic software, and to perform coordinate correction and model processing operations. The coordinate correction includes noise filtering and coordinate systematization of the raw motion data. The model processing operations include simplification of the knee joint three-dimensional model, marking of key anatomical points, and accuracy optimization to generate a standardized model group. The model analysis module is used to seamlessly import the model group into the C3D Clinical Window software. The C3D Clinical Window provides real-time visualization of the three-dimensional movement trajectory of the knee joint, calculation of biomechanical parameters, and clinical pathology correlation assessment functions. The human-computer interaction control module is used to write customized interaction logic based on the needs of clinical diagnostic applications; it integrates clinical protocol templates and provides a graphical user interface to realize user parameter input, analysis process control, real-time result monitoring, and automatic generation of diagnostic reports. The cloud publishing module is used to convert diagnostic reports and interactive interfaces into PHP dynamic web page format, and deploy and publish them on the cloud server via HTTP protocol, supporting SSL encrypted transmission, multi-user concurrent access and cross-platform remote calls; When the data processing module performs noise filtering and coordinate systematization on the raw motion data, it includes: The noise filtering uses an extended Kalman filter to define a state vector containing joint angle, angular velocity, and angular acceleration. The process noise covariance parameter is adaptively adjusted according to the kinematic model, and the tolerance range is increased when the joint acceleration change exceeds the acceleration change threshold. In the measurement update stage, the data streams of optical and inertial sensors are fused, and a residual threshold monitoring process is executed to remove abnormal measurement values. The coordinate system is initialized and calibrated based on the preoperative skeletal CT scan model using a quaternion rotation matrix algorithm. The geometric center points of the distal femoral intercondylar fossa and the tibial plateau are extracted to construct a local coordinate system. The rotation operation is represented by quaternions and the optimal rotation parameters are solved by minimizing the point cloud registration error, and the data is converted to a unified anatomical coordinate system. When the data processing module simplifies the 3D model of the knee joint, marks key anatomical points, and optimizes accuracy, it includes: The 3D model of the knee joint is simplified by using a vertex clustering polygon reduction algorithm to divide the triangular mesh into a spatial octree structure. The consistency of vertex normal vectors is analyzed within the region nodes. Vertices are merged when the curvature change is lower than the curvature change threshold, preserving the topological details of the key anatomical region of the intercondylar fossa. Key anatomical points are marked by a pre-trained U-Net convolutional neural network. When the data processing module implements key anatomical point labeling using a pre-trained U-Net convolutional neural network, it includes: The pre-trained U-Net convolutional neural network includes downsampling and upsampling layers. It takes three-dimensional voxel grid data as input and outputs a probability distribution heatmap of key points at the inferior pole of the patella and the anterior horn of the meniscus. Accuracy optimization uses the iterative nearest point algorithm for model registration. The convergence condition is set as the root mean square error being lower than the error threshold or reaching the maximum number of iterations. It is combined with Laplacian smoothing filtering to eliminate surface noise interference.

2. The knee joint three-dimensional movement analysis platform for clinical auxiliary diagnosis according to claim 1, characterized in that, When the C3D clinical window provides real-time visualization of the three-dimensional movement trajectory of the knee joint, calculation of biomechanical parameters, and assessment of clinical pathology relevance, it includes: The biomechanical parameters are calculated in real time using an inverse dynamics model combined with a musculoskeletal simulation toolkit to calculate joint torque and contact force. The clinical pathology correlation assessment is based on a support vector machine classifier. The feature vector integrates the peak torque, contact force time integral, and trajectory asymmetry index in the gait cycle, and uses a radial basis function kernel function for pattern recognition. The classifier is trained on an osteoarthritis case database and outputs a quantitative risk score.

3. The knee joint three-dimensional movement analysis platform for clinical auxiliary diagnosis according to claim 2, characterized in that, The calculation of biomechanical parameters, using an inverse dynamics model combined with a musculoskeletal simulation toolbox to calculate joint torques and contact forces in real time, includes: Based on kinematic data and ground reaction force input, the equations of motion, including mass matrix, Coriolis force term and gravity term, are solved; the contact force calculation adopts Hertz contact theory model, inputting meniscus curvature radius and material elastic modulus parameters, and generating a visualization cloud map of tibiofemoral joint pressure distribution.

4. The knee joint three-dimensional movement analysis platform for clinical auxiliary diagnosis according to claim 3, characterized in that, The human-computer interaction control module incorporates customized interaction logic based on clinical diagnostic application needs; it integrates clinical protocol templates and provides a graphical user interface to enable user parameter input, analysis process control, real-time result monitoring, and automatic generation of diagnostic reports, including: The human-computer interaction control module integrates customized interaction logic with a visual drag-and-drop process editor, supporting the definition of analysis step sequences through a graphical node interface; data flow between nodes is driven by an event-driven mechanism to perform parameter validity verification operations; the clinical protocol template includes an osteoarthritis grading assessment template, preset anatomical index thresholds, and loads patient historical data; the graphical user interface provides real-time 3D model interaction functions, and implements touch zoom operations, frame-by-frame playback of motion trajectories, and diagnostic report generation processes based on the WebGL rendering engine.

5. The knee joint three-dimensional movement analysis platform for clinical auxiliary diagnosis according to claim 4, characterized in that, When the human-computer interaction control module performs the diagnostic report generation process based on the WebGL rendering engine, it includes: It integrates visual elements into the interface and embeds dynamic biomechanical parameter charts; the interactive logic executes user parameter input, analysis process control, and real-time result monitoring operations.

6. The knee joint three-dimensional movement analysis platform for clinical auxiliary diagnosis according to claim 5, characterized in that, When the cloud-based publishing module converts the diagnostic report and interactive interface into a PHP dynamic webpage format, it includes: The cloud-based publishing module uses a model-view-controller architecture based on the Laravel framework to implement data processing logic. The controller responds to HTTP requests and calls the database model to extract biomechanical parameters. The view layer uses a template engine to dynamically render interface content, integrates AJAX asynchronous technology to execute a data loading process without page refresh, and updates results through lightweight requests. Interactive charts are generated using a visualization library configuration, supporting the display of joint torque polar coordinate graphs and pathological risk trend lines. Webpage performance is optimized through resource compilation and compression technology to execute the page loading process.

7. The knee joint three-dimensional movement analysis platform for clinical auxiliary diagnosis according to claim 6, characterized in that, The cloud-based publishing module is deployed and published on a cloud server via the HTTP protocol, supporting SSL encrypted transmission, concurrent access by multiple users, and cross-platform remote calls, including: In the HTTP protocol deployment, the SSL encrypted transmission uses the TLS protocol and a high-strength key to achieve end-to-end data protection; when the client connects, the server sends a digital certificate chain and generates a secure session key through a key exchange mechanism; the multi-user concurrent access is implemented through a load balancer to distribute traffic, and a weighted round-robin algorithm is configured to dynamically adjust the connection strategy according to the server load, and a session persistence mechanism is set; the cross-platform remote call supports RESTful API interface design, defines data access endpoints, and uses JSON data format to transmit structured information.

Images