A personalized design method of knee brace based on 3D printing

Abstract

The application relates to the technical field of knee joint brace design, and discloses a knee joint brace individual design method based on 3D printing. The method comprises bone structure, soft tissue contour and joint motion range information; based on stereoscopic scanning data, in combination with a preset knee joint biomechanics knowledge base, the individual anatomical feature set of the knee joint is determined; according to the individual anatomical feature set and at least one corresponding design rule file, the stereoscopic scanning data is divided into a macroscopic structure region and a microscopic detail region; the macroscopic structure region outputs bone frame data, and the data is transmitted to the microscopic detail region to guide soft tissue matching analysis; based on the division result, a three-dimensional model of the knee joint brace is constructed; the three-dimensional model is converted into a 3D printing instruction file, which is used for driving a 3D printing device to produce an individual knee joint brace. The method realizes individual manufacturing of the knee joint brace through multi-dimensional data fusion and structure collaborative design.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of knee brace design, in particular to a knee brace personalized design method based on 3D printing. BACKGROUND

[0002] As a complex load-bearing joint of the human body, the knee joint often needs to be fixed, corrected or assisted in movement with the help of a brace when it is damaged or dysfunctional. The design and production of traditional knee braces rely on manual measurement of limb size by medical personnel, and the basic parameters of the brace are determined by experience. Such a method is limited by the subjectivity of manual operation and is prone to measurement errors, resulting in deviations between the brace and the anatomical structure of the individual's knee joint area.

[0003] Standardized knee braces produced in batches can meet some general needs, but they are difficult to adapt to the bone morphology, soft tissue thickness and joint movement characteristics of different patients. Some patients have significant differences in limb morphology and standard models due to knee joint deformity and anatomical variation after wound healing, and the standardized brace is prone to local compression, looseness and other problems when worn, affecting the daily use experience.

[0004] With the development of digital technology, 3D scanning technology has been gradually applied to brace design, which can obtain three-dimensional morphological data of the knee joint area. However, existing design methods based on 3D scanning often treat bone structure and soft tissue contour as independent data for processing, and lack consideration of their correlation. In the model building process, either the macroscopic bone framework is overemphasized and the dynamic changes of soft tissue are ignored, or the micro details are piled up, resulting in insufficient stability of the overall structure of the brace.

[0005] Biomechanical properties are an important consideration for knee braces to function. In traditional design, the analysis of knee biomechanics relies on general databases and fails to incorporate individualized information such as joint range of motion and muscle distribution, making it difficult for the brace to coordinate with individual biomechanical characteristics when assisting joint movement, which may cause new limb discomfort over time.

[0006] 3D printing technology makes personalized brace production possible, but in existing technology, the conversion of three-dimensional models to printing instructions often results in conflicts between the bone support part and the soft tissue fitting part of the printed brace due to unreasonable model structure division, failing to fully utilize the advantages of 3D printing in personalized manufacturing. SUMMARY

[0007] The purpose of the present application is to provide a knee brace personalized design method based on 3D printing to solve the problems raised in the background art.

[0008] To achieve the above object, the application provides a knee brace personalized design method based on 3D printing, which comprises the following steps:

[0009] Collecting stereoscopic scanning data of the knee joint region of a patient by a three-dimensional scanning device, wherein the stereoscopic scanning data contains bone structure, soft tissue contour and joint motion range information;

[0010] Based on the stereoscopic scanning data, combining a preset knee biomechanics knowledge base, the personalized anatomical feature set of the knee joint is determined;

[0011] According to the personalized anatomical feature set and at least one design rule file corresponding to the feature set, the stereoscopic scanning data is divided into a macro-structure region and a micro-detail region;

[0012] Based on the division results of the macro-structure region and the micro-detail region, the stereoscopic scanning data is subjected to three-dimensional model construction of the knee brace;

[0013] The constructed three-dimensional model is converted into a 3D printing instruction file for driving a 3D printing device to produce a personalized knee brace;

[0014] In the division process of the macro-structure region and the micro-detail region, the macro-structure region outputs bone framework data, which is transmitted to the micro-detail region for guiding soft tissue matching analysis.

[0015] Preferably, the process of determining the personalized anatomical feature set of the knee joint comprises:

[0016] The stereoscopic scanning data is input into a multi-feature recognition model, which outputs key anatomical landmark points in the scanning data;

[0017] The key anatomical landmark points are input into a preset knee biomechanics knowledge base, and a biomechanics property set including all the landmark points is recognized according to the associated mapping relationship in the knowledge base;

[0018] Based on the biomechanics property set, the joint activity angle range and the load distribution parameter are calculated, which are used for data interaction in subsequent three-dimensional model construction.

[0019] Preferably, the division process of the macro-structure region and the micro-detail region comprises:

[0020] According to the personalized anatomical feature set, the design rule file corresponding to the feature set is called, and the design rule file is subjected to semantic analysis by an intelligent analysis engine, and a plurality of design constraint conditions are output;

[0021] Based on the design constraints, the stereoscopic scanning data is divided into macro-structure region and micro-detail region, the macro-structure region generates the bone topology data, and the micro-detail region generates the soft tissue deformation data, the bone topology data is directly input into the soft tissue deformation analysis module to optimize the segmentation result.

[0022] Preferably, the process of constructing the three-dimensional model of the knee brace based on the stereoscopic scanning data comprises:

[0023] According to the divided macro-structure region and micro-detail region, the pre-processed stereoscopic scanning data is split into macro-structure sub-data set and micro-detail sub-data set;

[0024] The macro-structure sub-data set and the associated biomechanical parameters are input into the structure optimization model, and the micro-detail sub-data set is input into the detail generation model;

[0025] The output results of the structure optimization model and the detail generation model are combined to generate a complete three-dimensional model of the knee brace, and the intermediate data of each model construction is recorded.

[0026] Preferably, the execution process of the structure optimization model and the detail generation model comprises:

[0027] The structure optimization model is established based on a geometric constraint algorithm, the structure optimization model first performs contour fitting on the macro-structure sub-data set to obtain brace support frame information, and determines whether the mechanical stability requirement is met by comparing the matching degree of the support frame information and the biomechanical parameters;

[0028] The detail generation model is used for surface texture analysis to realize the adaptive smoothing function of the fine surface in the micro-detail sub-data set, and the support frame information is input into the surface texture analysis to enhance the data consistency.

[0029] Preferably, when the macro-structure sub-data set is subjected to contour fitting, a micro-detail enhanced region is further extracted from the initial macro-structure sub-data set, which belongs to an iterative optimization process, and the specific steps of the iterative optimization process comprise:

[0030] The macro-structure sub-data set is obtained, the curvature change rate and the density distribution index of each local unit of the macro-structure sub-data set are calculated, and when the curvature change rate or the density distribution index of any local unit exceeds a preset critical value, it is determined that the local unit has geometric defects that need to be finely reconstructed;

[0031] Based on the determined geometric defect local unit, the corresponding micro-detail enhanced data set is separated from the macro-structure sub-data set;

[0032] According to the geometric defect characteristics of different knee joint types, data extraction rules associated with the defect characteristics are established, and the geometric defect determination results are directly input to the data extraction rule module to drive the separation operation.

[0033] Preferably, the specific steps of the iterative optimization further include:

[0034] After separating the micro-detail enhancement dataset from the macro-structure sub-dataset, a data synchronization mechanism is established between the structure optimization model and the detail generation model;

[0035] The separated region is evaluated using the structure optimization model, and the coordinate range, size parameters and evaluation results of the separated region are transmitted to the detail generation model;

[0036] After receiving the information transmitted by the structure optimization model, the detail generation model reconstructs the surface of the micro-detail enhancement dataset, detects and generates specific surface optimization information, including texture coordinates, curvature change trend and integration state with the support frame information, and the surface optimization information is fed back to the frame evaluation module to complete the closed-loop optimization.

[0037] Preferably, the specific steps of the iterative optimization further include:

[0038] The generated surface optimization information is fed back to the structure optimization model in real time;

[0039] During the multiple data separation and model optimization processes, the optimization threshold and model configuration parameters are adjusted according to the deviation of each optimization result from the preset standard;

[0040] The final generated complete three-dimensional model information, key parameters and optimization log are output to the designer end, and the full data record of each optimization process is stored, and the model configuration parameter adjustment result is transmitted to the subsequent three-dimensional model construction step to update the processing logic.

[0041] Preferably, the specific process of separating the corresponding micro-detail enhancement dataset from the macro-structure sub-dataset includes:

[0042] Determine the local unit with geometric defects in the macro-structure sub-dataset, which is composed of boundary voxels defined by a set of three-dimensional coordinate points;

[0043] Based on the geometric defect local unit, the parameters required for separating the micro-detail enhancement dataset are calculated, including the starting coordinate point, voxel width and height dimension;

[0044] Using the calculated separation parameters, the corresponding micro-detail enhancement dataset is separated from the macro-structure sub-dataset through data cropping operation;

[0045] The separated micro-detail enhancement data set is verified whether it completely covers the defect area and excludes redundant data interference, and the separated parameters are output to a verification module for an automatic checking process.

[0046] Preferably, the combination structure optimization model and the detail generation model output results, adopt a fusion weight distribution mechanism, different weight coefficients are allocated according to the confidence scores of the structure optimization model and the detail generation model, the confidence scores are obtained by training historical data and updated in real time, and the weight coefficients are output to the model output integration module to coordinate the generation of the three-dimensional model.

[0047] Compared with the prior art, the present application has the following advantages:

[0048] The three-dimensional scanning device collects the stereoscopic scanning data of the knee joint area of the patient, covering the bone structure, soft tissue contour and joint movement range, so that the obtained information is more comprehensive and can reflect the anatomical features and movement state of the individual knee joint in detail. Based on these stereoscopic scanning data, combined with the pre-set knee joint biomechanics knowledge base, the personalized anatomical feature set is determined, so that the design process takes into account the individual biomechanical characteristics, and the brace design is more suitable for the body function and movement needs of the patient.

[0049] According to the personalized anatomical feature set and the corresponding design rule file, the stereoscopic scanning data is divided into macro-structure region and micro-detail region, so that different levels of structure information can be processed. The bone framework data output by the macro-structure region is transmitted to the micro-detail region for guiding soft tissue matching analysis, so that the structure information of the bone and soft tissue is coordinated in the model construction, avoiding the disconnection of the two in the design, so that the overall structure of the brace not only meets the support needs of the bone, but also adapts to the shape and dynamic changes of the soft tissue.

[0050] Based on the division results of the macro-structure region and the micro-detail region, the three-dimensional model of the knee brace is constructed, so that the model can take into account the macro-structure stability and the micro-precision at the same time, and present a structure shape that is more suitable for the individual knee joint shape and movement characteristics. The constructed three-dimensional model is converted into a 3D printing instruction file to drive the 3D printing device to produce a personalized knee brace, so that the personalized features in the design can be directly converted into the properties of the physical brace, and the brace can be highly matched with the patient's knee joint area in shape and structure, making the contact with the body more natural when worn, reducing the local pressure or looseness caused by improper fitting, and improving the comfort and fit during use.

[0051] The entire design method forms a coherent personalized process from data acquisition to final production, and each link is developed around individual characteristics, so that the final knee brace can better adapt to the anatomical differences, biomechanical characteristics and movement needs of different patients, and form a closer match with the individual in shape and function. Attached Figure Description

[0052] Figure 1 This is a schematic diagram illustrating the working principle of the 3D-printed personalized knee brace design method described in this invention.

[0053] Figure 2 A flowchart for determining a personalized anatomical feature set for the knee joint;

[0054] Figure 3 Flowchart for constructing a 3D model of a knee brace;

[0055] Figure 4 A flowchart for data synchronization and closed-loop optimization. Detailed Implementation

[0056] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0057] Please see Figure 1 This invention provides a method for personalized design of knee braces based on 3D printing, the method comprising:

[0058] A 3D scanning device is used to acquire stereoscopic scan data of the patient's knee joint region. This data includes information on skeletal structure, soft tissue contours, and joint range of motion. Based on this data and a pre-defined knee joint biomechanical knowledge base, a personalized anatomical feature set for the knee joint is determined. According to this personalized anatomical feature set and at least one corresponding design rule file, the stereoscopic scan data is divided into macroscopic structural regions and microscopic detail regions. Based on the division of these regions, a 3D model of a knee brace is constructed. This 3D model is then converted into a 3D printing instruction file to drive the 3D printing equipment to produce the personalized knee brace. During the division of the macroscopic and microscopic detail regions, the macroscopic structural region outputs skeletal framework data, which is then transferred to the microscopic detail region to guide soft tissue matching analysis.

[0059] Example 1: See Figure 2In the process of determining the personalized anatomical feature set of the knee joint, the stereoscopic scanning data is collected by a three-dimensional scanning device, containing bone structure, soft tissue contour and joint motion range information. These data are stored in the form of point cloud or mesh, covering the geometric and dynamic characteristics of the knee joint region. After the collection is completed, the data are input into the multi-feature recognition model for processing. The multi-feature recognition model adopts a deep learning architecture, analyzes the scanning data through a convolutional neural network and a feature extraction algorithm, and recognizes the spatial distribution of anatomical landmark points. The key anatomical landmark points output by the model include femoral condyle, tibial plateau, patellar edge and other feature positions with clear biomechanical significance.

[0060] The key anatomical landmark points are then input into a preset knee biomechanics knowledge base. The knowledge base adopts a graph database structure for storage, containing the topological relationship between anatomical landmark points, the mechanical transmission path and the stress distribution rule under typical motion modes. The knowledge base correlates the input landmark points with predefined biomechanical characteristics through a correlation mapping mechanism. For example, the radius of curvature of the femoral condyle is associated with the joint contact pressure distribution, and the inclination angle of the tibial plateau is associated with the load transmission efficiency. The system generates a biomechanical property set based on the mapping relationship, which is stored in the form of a structured data table, containing indicators such as stiffness coefficient, active freedom range, ligament attachment point mechanical parameters, etc.

[0061] Based on the biomechanical property set, the system derives the joint activity angle range and load distribution parameters through numerical calculation methods. The activity angle range is determined by simulating the limit positions of the knee joint flexion, rotation and other movements, and the elastic deformation of soft tissue and the sliding friction of bone contact surface are considered in the calculation process. The load distribution parameters are estimated by finite element analysis, which quantifies the pressure distribution of the knee joint in static standing and dynamic gait cycle into regionalized numerical values. These parameters are stored in the form of a matrix, which is used to guide the structure optimization and material distribution in the subsequent three-dimensional model construction.

[0062] In the process of dividing the macro-structure region and the micro-detail region, the system calls the design rule file matched with the personalized anatomical feature set. The design rule file is written in a semantic description language, containing geometric constraints, material distribution gradient requirements and mechanical performance thresholds for region division. The intelligent analysis engine parses the rule file, extracts key instructions and converts them into executable calculation tasks. During the parsing process, the engine uses natural language processing technology to identify the logical relationships in the rules, such as the conditional statements "if the curvature of the femoral condyle is greater than the threshold, then increase the thickness of the lateral support", which are converted into executable program branches.

[0063] Based on the design constraints obtained by analysis, the stereoscopic scanning data is divided into macro-structure region and micro-detail region. The division of macro-structure region is based on the mechanical bearing requirements of the skeleton, and the high-density region is identified by voxel clustering algorithm to generate the skeleton topology data. The data is represented by non-uniform rational B-spline surface, which accurately restores the bearing framework of the skeleton. The division of micro-detail region focuses on the morphological changes of soft tissue and skin surface, and the high gradient change region is identified by curvature manifold analysis to generate soft tissue deformation data. The data contains fine parameters such as epidermis texture characteristics and subcutaneous fat thickness distribution.

[0064] The skeleton topology data is transmitted in real time to the soft tissue deformation analysis module through the data pipeline. The module uses elastic deformation simulation algorithm to quantify the soft tissue displacement caused by bone movement as deformation field data. During the analysis, the module dynamically adjusts the vertex weight of the soft tissue deformation data, so that the generated micro-detail region and macro-skeleton movement maintain biomechanical consistency. For example, the skin stretching effect caused by the change of patella trajectory is quantified as a vertex displacement vector and fed back to the surface reconstruction process of the micro-detail region.

[0065] The system improves the accuracy of region division through iterative optimization mechanism. In each iteration, the skeleton topology data of macro-structure region and the soft tissue deformation data of micro-detail region are compared to detect the continuity error of the boundary region. When the density or curvature mutation between adjacent voxels is detected, the system automatically adjusts the segmentation threshold and recalculates the region division result. The optimization process continues until the transition gradient between macro and micro regions meets the preset smoothness standard, and finally outputs the knee joint digital model with hierarchical structure.

[0066] During data interaction, the system uses timestamp synchronization mechanism to ensure the timeliness of each module processing. The coordinate update of key anatomical landmarks triggers real-time query of the biomechanics knowledge base, and the change of bearing distribution parameters directly feeds back to the dynamic loading of design rule file. This tightly coupled data flow design makes the determination of personalized anatomical feature set and the region division process form a closed loop system, effectively capturing the correlation characteristics of knee joint morphology and function.

[0067] In terms of implementation details, the multi-feature recognition model uses transfer learning strategy to adapt to the anatomical variations of different patients through pre-trained weights. The knowledge base introduces fuzzy logic processing in the association mapping, which is compatible with the noise and partial missing in the scanning data. The intelligent analysis engine is equipped with rule version management function, which supports clinical experts to visually edit and simulate the effect of design rules. The whole implementation process runs in a distributed computing framework, which uses parallel computing to accelerate large-scale data processing, and uses checkpoint mechanism to ensure data recoverability after operation interruption.

[0068] The stereoscopic scanning data is processed by coordinate system normalization in the preprocessing stage, and all subsequent calculations are based on the standardized anatomical reference system. The intermediate results generated during data division are stored incrementally, only the differences are retained to reduce memory usage. The system interface provides three-dimensional visualization tools, allowing the operator to review the regional division effect from any perspective, and manually correct the automatic division result through the annotation tool. The correction operation is recorded as a new design rule supplement to the knowledge base, forming a self-improving mechanism for the system.

[0069] The technical scheme of this embodiment gradually converts the original scanning information into a structured feature set that can guide the design of the brace through a hierarchical data processing flow. Automatic identification of anatomical landmarks reduces subjective bias in manual annotation, and the application of biomechanics knowledge base converts clinical experience into quantifiable design parameters. The double-layer structure of regional division takes into account the overall mechanical performance and local fitting accuracy, providing differentiated data processing basis for subsequent three-dimensional model construction. The automation of the entire process significantly shortens the period from data acquisition to brace design, while standardized processing ensures the comparability of design results between different cases.

[0070] The log data generated during system operation includes processing time, parameter adjustment records, and exception event reports for each module. These logs are stored through a time series database for analyzing process bottlenecks and optimizing computing resource allocation. The interactive events of the operation interface are abstracted as a state machine model to ensure the traceability of user operations. Role-based permission management is used for data security to ensure the encryption protection of patient privacy information during transmission and storage. The network communication uses a checksum mechanism to prevent bit errors during data transmission from affecting calculation accuracy.

[0071] The technical scheme of this embodiment realizes the automatic conversion from knee joint scanning data to design feature set through the collaborative work of multiple modules. The digital expression of anatomical features provides accurate input conditions for the design of personalized medical devices, and the hierarchical processing strategy balances the demand for calculation efficiency and model accuracy.

[0072] Example 2: see Figure 3 In the three-dimensional model construction process of the knee brace, the system first performs regionalization processing on the preprocessed stereoscopic scanning data. According to the macro-structure regions and micro-detail regions divided earlier, the scanning data is systematically divided into two logically independent but data-related subsets. The macro-structure sub-data set mainly contains the geometric information of the skeletal framework, stored in the form of high-precision triangular mesh, and each vertex coordinate is attached with local curvature and thickness parameters. The micro-detail sub-data set records the texture features and dynamic deformation characteristics of the soft tissue surface, expressed in the form of parameterized surface combined with displacement field.

[0073] The macro-structure sub-dataset is inputted into the structure optimization model together with the associated biomechanical parameters. The biomechanical parameters are stored in JSON format, including key indicators such as joint movement angle limits, pressure distribution maps, and material stiffness coefficients. The structure optimization model uses a multi-objective optimization algorithm, taking into account both the mechanical support performance of the brace and the comfort requirements of the wearer. The model first performs a topology analysis on the input skeletal framework, identifying areas that need to be strengthened and non-critical parts that can be appropriately lightened. Then, based on the finite element method, it simulates the stress distribution under different load conditions and automatically adjusts the thickness distribution and reinforcement layout of the brace framework.

[0074] The operation process of the structure optimization model adopts a phased strategy. In the initial stage, a coarse-grained model that meets the basic mechanical requirements is quickly generated, which serves as the basis for subsequent refinement optimization. In the middle stage, a constraint condition processing algorithm is introduced to ensure that the brace structure matches the patient's anatomical form with sufficient tolerance. In the final stage, the high-stress concentration areas are strengthened by locally encrypting the grid, and the support framework information that meets the biomechanical requirements is output.

[0075] The micro-detail sub-dataset is inputted into the detail generation model in parallel. This model focuses on processing the properties of the inner surface of the brace that comes into contact with the patient's skin. The model first performs curvature analysis and contact pressure prediction on the soft tissue contour, identifying sensitive areas that require special treatment. Then, using an adaptive subdivision algorithm, it dynamically adjusts the grid density in high-pressure areas while maintaining the overall surface continuity. The detail generation model also integrates tribological parameters to automatically generate differentiated surface texture patterns based on the skin characteristics of different body parts.

[0076] The workflow of the detail generation model includes two dimensions: geometric processing and physical simulation. In terms of geometric processing, the model preserves the natural morphological features of the anatomical structure through feature line detection and surface parameterization techniques. In terms of physical simulation, the model calculates the interaction between the brace and the soft tissue, predicting changes in pressure distribution that may occur due to long-term wear.

[0077] The output results of the structure optimization model and the detail generation model are integrated through a fusion weight allocation mechanism. The system assigns dynamic weight coefficients to each model, which are automatically adjusted based on the confidence score of the model. The confidence score is calculated by analyzing the historical performance data of the model and the matching degree of the current input, and is updated in real time using a sliding window algorithm. The weight allocation process considers multiple dimensions of influencing factors, including geometric accuracy, mechanical performance, and manufacturing feasibility.

[0078] In the model fusion stage, the system first performs a topological consistency check on the mesh structures output by both models. When mismatched areas are found, automatic triggering of local recalculation or interpolation smoothing processing. The fusion algorithm preserves the main framework of the macro-structure model while accurately mapping the surface characteristics of the micro-detail model to the corresponding areas. For transition areas with weight disputes, the system uses a gradual mixing strategy to ensure smooth transition of geometric features. The complete three-dimensional model after fusion is processed by lightweight, deleting redundant vertices and invisible surfaces, optimizing data storage efficiency.

[0079] The system records the intermediate data of each model construction, including the mesh model of each optimization stage, parameter adjustment record and performance evaluation index. These data are managed in a version control manner, supporting designers to trace back to any step of the intermediate state. The data storage format takes into account processing efficiency and storage space, and key parameters use differential encoding to reduce redundancy.

[0080] The quality verification of three-dimensional model adopts a multi-level checking mechanism. The geometric level checks the closure, normal consistency and mesh quality of the model; the functional level simulates the mechanical behavior of the brace in different motion states; the manufacturing level analyzes the printability of the model and the support structure requirement. Problems found in the verification process are automatically fed back to the corresponding processing module, triggering local recalculation or parameter adjustment. The verified model is converted to a lightweight intermediate format, ready for subsequent print instruction generation.

[0081] In terms of system implementation, the model construction module uses a microservice architecture, with structure optimization and detail generation as independent services. Asynchronous communication between services through message queues ensures the elastic expansion of computing resources. The task scheduler monitors the load state of each service and dynamically allocates computing tasks. The user interface provides real-time visualization, displaying model construction progress and key parameter changes. Interface operations are decoupled from background calculations to avoid blocking user interactions with complex calculations.

[0082] The data processing pipeline uses fault-tolerant design, with checkpoints set at key computing steps. After an abnormal interruption, it can be recovered from the most recent stable state. The system resource monitoring module tracks CPU, memory and GPU usage, and automatically reduces calculation precision or suspends secondary tasks when resources are insufficient. Network communication uses compression transmission technology to reduce latency when exchanging large amounts of data. Database query optimizes index structure to speed up historical model retrieval and analysis.

[0083] The model construction process supports manual intervention mechanisms. Designers can pause automatic calculations at any stage, manually adjust parameters or directly modify model geometry. Manual modification records are stored separately from automatic calculation logs, facilitating analysis of automation process improvement space. The system provides a variety of editing tools, including local surface adjustment, feature line editing and mechanical parameter overlay functions, with all modifications visualized in real time.

[0084] In terms of system performance, a multi-level parallel computing strategy is adopted. Macro-structure optimization utilizes multi-core CPUs for task decomposition of finite element analysis, while micro-detail generation leverages GPUs to accelerate surface processing algorithms. Intelligent caching mechanisms are employed for memory management, with high-frequency access data retained in fast storage areas. Disk IO is optimized through asynchronous writing to avoid blocking the computing process with storage operations.

[0085] In terms of security measures, patient data is encrypted during transmission and storage. Access control is based on a role-based permission model, ensuring that sensitive operations require authorized authentication. System logs record complete data flow trajectories, supporting security audits and compliance checks. Network communication employs two-way authentication to prevent man-in-the-middle attacks.

[0086] The technical solution of this embodiment achieves complete conversion from anatomical data to printable models through the synergistic work of structure optimization and detail generation. The dual-model architecture balances overall performance and local adaptation, and the weight distribution mechanism balances the priority of different design objectives. Complete intermediate data records support the traceability of the design process, and multi-level verification ensures the quality and reliability of the output model. The system implements modern software engineering practices to ensure processing efficiency while maintaining sufficient flexibility. The entire model construction process strikes a balance between automation and manual control, providing technical support for the accurate manufacturing of personalized braces.

[0087] Example 3: The construction of the structure optimization model is based on a geometric constraint algorithm system. This algorithm uses a hierarchical processing strategy when processing the macro-structure sub-data set. In the initial stage, the input data is denoised and smoothed to eliminate surface irregularities introduced during the scanning process. The processed data is converted to a voxel representation for spatial topology analysis. The voxelization process uses an adaptive resolution strategy to automatically increase the sampling density in high-curvature areas. The structure optimization model identifies the mechanical key points of the skeletal structure through a feature extraction algorithm, which forms the basis for positioning reference for the brace support frame.

[0088] The contour fitting process uses an iterative optimization method. The system first determines the initial support frame topology based on biomechanical parameters. The frame generation algorithm automatically plans the orientation of the reinforcing ribs and the position of the connection nodes based on the curvature distribution and thickness variation of the bone surface. In the mechanical stability evaluation stage, the model matches the generated support frame information with the input biomechanical parameters. The matching degree calculation considers multiple dimensions, including stiffness matching coefficient, stress dispersion efficiency, and weight distribution balance. When the main indicators do not meet the preset threshold, the model triggers the frame reconstruction mechanism, adjusts the key parameters, and recalculates.

[0089] The mechanical stability evaluation is quantitatively analyzed using the following formula:

[0090] ;

[0091] wherein, represents the frame matching degree deviation value, is the number of key evaluation points, is the measured stiffness value of the i-th point, is the theoretical stiffness value of the corresponding point, is the weight coefficient of the i-th point, is the stress distribution balance factor, is the stress unevenness. The calculation result of the formula is used to judge whether the frame structure needs to be adjusted, and the optimization cycle is triggered when the deviation value exceeds the threshold value.

[0092] When processing the micro detail sub-data set, the model adopts a multi-scale analysis method. The model first parameterizes the soft tissue surface and establishes a unified UV coordinate system. The surface texture analysis algorithm identifies sensitive areas that need special processing by calculating the local curvature change rate and the normal direction distribution. Adaptive smoothing processing uses a physics-based diffusion equation to eliminate irregularities at the micro scale while preserving the main features. During the processing, the model receives real-time support frame information from the structure optimization model and converts this information into geometric constraints to ensure that the detail processing is coordinated with the overall structure.

[0093] The iterative optimization process of the macro structure sub-data set adopts a dynamic data separation strategy. When the system performs geometric quality evaluation on each local unit, it calculates two key indicators: the curvature change rate reflects the degree of change in surface smoothness, and the density distribution index describes the spatial distribution uniformity of voxels. The evaluation algorithm establishes a moving analysis window in three-dimensional space, and the window size is dynamically adjusted according to the anatomical structure characteristics. When the statistical value of any indicator in the window exceeds the preset critical value, mark the area as a geometric defect area that needs fine reconstruction. The critical value setting considers the normal physiological variation range of different knee joint parts to avoid invalid reconstruction caused by excessive sensitivity.

[0094] The separation operation of the geometric defect area adopts an accurate boundary recognition technology. The system first performs morphological dilation processing on the defect area to ensure complete coverage of the problem area. Then, an edge detection algorithm is applied to determine the separation boundary, and the boundary curve meets the second-order geometric continuity requirement. The data cropping module extracts the micro detail enhancement data set from the macro structure sub-data set according to the boundary information. The extraction process preserves the topological connection relationship of the original data to ensure seamless integration in subsequent processing. The separation parameter calculation considers the anisotropy characteristics of three-dimensional space and sets higher separation accuracy in the mechanical principal direction.

[0095] The data extraction rule module establishes a classification processing mechanism according to the knee joint type. The system maintains a feature rule library containing typical geometric defect patterns under different pathological conditions. The rule matching process uses fuzzy reasoning methods to handle common variations in clinical practice. After the geometric defect judgment result is input into the rule module, the corresponding data processing flow is triggered. For example, the cartilage wear area of an osteoarthritis patient is processed using edge enhancement, while the abnormality of the ligament attachment point caused by sports injury focuses on surface smooth transition. The rule engine supports dynamic updating, and clinical experts can add new processing strategies or adjust existing parameters.

[0096] The data interaction between the structure optimization model and the detail generation model uses an event-driven mechanism. When the macro-structure sub-dataset completes a specific processing stage, a standardized event message is generated, containing processing area coordinates, feature parameters, and intermediate results. The message bus routes this information to subscribed detail generation model instances, triggering corresponding micro-processing flows. Reverse communication also follows this pattern, ensuring that the processing progress of the two models remains synchronized. The data synchronization process uses optimistic concurrency control, and when conflicts occur, version merging is based on timestamps.

[0097] The framework evaluation module uses a multi-criteria decision-making method, with evaluation indicators including geometric matching degree, mechanical performance indicators, and production feasibility. The evaluation process generates a detailed report, marking key areas that require special attention. Report data is converted into visual markers and superimposed on the three-dimensional model, making it easy for designers to quickly locate problems. The evaluation results are input into the detail generation model, guiding the focus of surface reconstruction processing. For example, high-stress areas require more detailed surface subdivision, while major load-bearing parts require stricter dimensional tolerances.

[0098] The generation of surface optimization information uses a parameterization method, with the system recording the optimization type, intensity parameters, and influence range of each processing area. Texture coordinate processing takes into account the characteristics of subsequent manufacturing processes to ensure that the generated surface patterns can be accurately reproduced during printing. Curvature trend analysis establishes a mathematical model to predict wear patterns during long-term use of the brace. Integration state checks verify whether the micro-processing results maintain mechanical continuity with the overall structure, triggering local recalculation when inconsistencies are found.

[0099] The geometric defect judgment algorithm uses machine learning methods to improve accuracy. The system collects processing records from historical cases to train classification models to identify common defect patterns. Model reasoning combines low-level feature analysis and high-level semantic understanding to reduce misjudgment rates. The judgment result is accompanied by a confidence score, and low-confidence cases are automatically transferred to the manual review process. The algorithm is retrained periodically with new data to gradually adapt to new variations encountered in clinical practice.

[0100] The processing of the micro-detail enhancement dataset adopts a hierarchical refinement strategy. The base layer processing guarantees the basic functionality of the geometry, the intermediate layer optimizes the parameters related to wearing comfort, and the surface layer processing focuses on the aesthetic and tactile properties. Each processing layer has independent quality check standards, and only after passing the current layer verification can it enter the next stage. This progressive processing method effectively controls the computational complexity and avoids unnecessary precision waste.

[0101] The data cropping operation performs a complete integrity check before implementation, verifying the topological consistency and geometric validity of the input data. The cropping process adopts a conservative strategy, retaining redundant data in the transition area for flexible adjustment in subsequent processing stages. The operation log records the cropping parameters and the affected range in detail, supporting operation rollback and data reconstruction. The verification module uses a combination of sampling inspection and global analysis to ensure that the separated dataset meets the requirements of subsequent processing.

[0102] Example 4: see Figure 4 In the personalized design process of the knee brace, after separating the micro-detail enhancement dataset from the macro-structure sub-dataset, the system establishes a data synchronization mechanism between the structure optimization model and the detail generation model. This process can be clearly demonstrated through a specific clinical case: a 45-year-old male patient with right knee instability due to sports injury, and three-dimensional scanning of the knee joint area was performed, with a scanning resolution of 0.2mm, covering the distal femur, proximal tibia and surrounding soft tissue.

[0103] The data synchronization mechanism is first reflected in the processing of the structure optimization model for the separated area. Taking the patient's medial condyle of the femur as an example, the system identifies 3 geometric defects in this area that require fine processing, with sizes of 8.7mm x 6.3mm, 5.2mm x 4.1mm and 12.5mm x 9.8mm. When the structure optimization model performs framework evaluation on these areas, it generates an evaluation report containing the following key parameters:

[0104]

[0105] The evaluation report is transmitted to the detail generation model through a standardized data interface, triggering the corresponding surface reconstruction process. In this case, after receiving the coordinate range information of area 3 (X: 124.5-137.0mm, Y: 56.7-66.5mm, Z: 89.2-99.0mm), the detail generation model first performs data preprocessing on this area. The preprocessing steps include point cloud data resampling, adjusting the original scanning point density from 36 points per square millimeter to 25 points, and eliminating noise points generated during the scanning process.

[0106] The surface reconstruction algorithm adopts an adaptive subdivision strategy, adjusting the processing intensity based on the received evaluation results. For the high stress concentration areas of the patient's femoral medial condyle (region 2 and region 3), the algorithm automatically increases the grid subdivision level, reducing the average size of the triangular patches from the default 1.2mm to 0.8mm. When processing areas with excessive thickness variation rates, the system activates the thickness compensation function, generating a gradual transition structure on the inside surface of the brace, converting the abrupt thickness changes into smooth gradients.

[0107] The generation process of surface optimization information integrates multi-source data. Taking the processing of region 1 as an example, the system combines the curvature deviation data in the frame evaluation results and the normal direction information of the original scanning points to generate surface textures with specific orientations. The texture units are elliptical, with the long axis direction consistent with the principal stress direction, and the unit spacing is dynamically adjusted according to the curvature change gradient. This texture design ensures the fit of the brace inner surface with the patient's anatomical structure and optimizes the pressure distribution during long-term wear.

[0108] The closed-loop optimization mechanism is reflected in the real-time feedback of surface optimization information to the frame evaluation module. In the above case, after the detail generation model completes the surface reconstruction of region 2, it packages the processing results into a feedback data packet, including the following key information: the reconstructed surface curvature distribution map, the vertex displacement vector field, and the texture unit density parameters. After receiving these data, the frame evaluation module recalculates the mechanical parameters of this region and finds that the stress concentration coefficient decreases from the initial 2.3 to 1.9, close to the safety threshold.

[0109] The technical implementation of the data synchronization mechanism relies on a distributed message system. Clinical cases show that when processing a medium complexity case containing 15 defect areas, the system exchanges an average of 83 messages per minute between the structure optimization model and the detail generation model, with a peak of 120 messages per minute. The message content is binary coded, with an average message size of 4.7KB, containing coordinate transformation matrices, attribute bitmaps, and operation instructions. The message queue has a priority mechanism to ensure that synchronization requests for high stress areas are given priority.

[0110] During multiple iterations of optimization, the system records complete logs of each data exchange. Taking the treatment data of the 45-year-old patient as an example, the system optimizes the medial condyle area of the femur for 4 rounds of iteration, and the intermediate data generated in each iteration includes: the frame adjustment scheme output by the structure optimization model, the surface correction parameters generated by the detail generation model, the synchronization timestamp between the two models, and the manual intervention record (if any). These data are stored in time series form, supporting designers to trace back to the state of any optimization stage.

[0111] The implementation of inter-model data synchronization is also reflected in the exception handling mechanism. In clinical practice, when two models have significant differences in treatment recommendations for the same area (such as opposite curvature adjustment directions), the system automatically triggers a negotiation protocol. During the treatment of this patient, such a situation occurred in the proximal tibial anterior region, where the structural optimization model recommended increasing support stiffness, while the detail generation model recommended reducing contact pressure based on soft tissue adaptability considerations. The system resolves the disagreement by pausing the automatic optimization process, displaying three-dimensional renderings of both treatment options simultaneously in the visualization interface, prompting the clinician to make a judgment, and ultimately adopting a hybrid solution that maintains the support structure while adding a cushioning layer design.

[0112] The integration status check of surface optimization information uses a hierarchical verification method. Taking the final optimization result of the above case as an example, the system performs three levels of checks: geometric level verifies surface continuity and boundary fit, physical level simulates contact pressure distribution under different flexion angles, and manufacturing level analyzes the feasibility of 3D printing. Each check level generates a detailed pass / fail report, and the items that fail the check automatically generate correction suggestions for the next round of optimization.

[0113] The time characteristics of data synchronization show significant value in clinical applications. System monitoring shows that from the completion of the structural optimization model to the return of the surface optimization information from the detail generation model, the average delay is 37 seconds, and 90% of requests are completed within 1 minute. This near real-time interaction capability enables the optimization of complex areas to be completed in a single clinic visit, avoiding the cumbersome process of multiple scans and tryouts in traditional methods.

[0114] In terms of system implementation, inter-model communication uses a dedicated data channel based on TCP / IP protocol, and transmission layer encryption ensures patient privacy and security. Network exception handling strategies include: automatic retry mechanism (up to 3 times), local cache of unsent data, and degradation processing (switching to batch processing mode when consecutive failures occur). In terms of hardware configuration, it is recommended to allocate independent computing nodes for each model, interconnected through a high-speed local area network. Actual test data shows that dual-node configuration improves performance by 62% compared to single-node configuration.

[0115] The clinician's operation interface provides real-time visual monitoring of data synchronization. Interface elements include: inter-model data transmission status indicator, three-dimensional positioning map of the current processing area, synchronization progress bar, and exception alert area. The physician can pause the automatic synchronization process at any time and manually adjust parameters or switch processing modes. All operation records are stored together with the automatically generated synchronization log to form a complete design process document.

[0116] The application data of this embodiment in 20 clinical cases shows that through the collaborative work of the structure optimization model and the detail generation model, the average number of final brace design scheme modifications is reduced from 6.2 times of the traditional method to 2.8 times. The initial adaptation period after the patient wears it is shortened, which benefits from the improvement of design accuracy brought by the fine data synchronization between models.

[0117] In the personalized design process of the knee brace, the real-time feedback mechanism of surface optimization information constitutes the core link of system self-improvement. After the detail generation model completes the reconstruction of the micro surface, the generated surface optimization information is transmitted to the structure optimization model through a special data channel. This feedback is not a simple data transmission, but a composite information package containing multiple levels of processing results and correction suggestions. During the transmission process, the system intelligently compresses and prioritizes the data package to ensure that the feedback information of critical areas is processed first, such as optimization data of high stress concentration areas or frequently contacted parts, which are marked as emergency level and automatically sorted to the front in the message queue.

[0118] After receiving the feedback information, the structure optimization model starts a multi-round evaluation and verification process. The model first analyzes the geometric correction parameters in the feedback data package and compares them with the current frame structure. The comparison process uses a spatial mapping algorithm to convert the adjustment amount of the micro surface to the equivalent correction amount of the macro frame. For example, when the feedback information shows that a contact area needs to increase a 0.5 millimeter buffer layer, the structure optimization model will adjust the thickness distribution and reinforcement layout of the support structure in that area accordingly. This conversion is not a simple linear correspondence, but a nonlinear mapping that considers the material mechanics characteristics and the influence of the overall structure.

[0119] During multiple data separation and model optimization processes, the system establishes a dynamic threshold adjustment mechanism. After each optimization cycle, the system automatically compares the deviation of the current result from the preset standard, and intelligently adjusts the decision threshold and convergence conditions for the next round of optimization based on the distribution characteristics and trends of the deviation. Deviation analysis not only focuses on numerical differences, but also examines the topological consistency of geometric features and the coordination of mechanical parameters. When the optimization effect of a particular area is consistently unsatisfactory, the system gradually relaxes the convergence conditions in that area while enhancing the constraint strength in adjacent areas, achieving a balance in overall performance through this compensation strategy.

[0120] The adjustment of model configuration parameters follows the principle of gradual optimization. The system maintains a parameter influence coefficient matrix, which records the influence weight of each configuration parameter on the final design quality. After each optimization iteration, the system analyzes the correlation between parameter adjustment and effect improvement and updates this influence coefficient matrix. Based on the updated matrix, the parameter adjustment scheme for the next round of optimization is generated. This data-driven parameter optimization method enables the system to gradually adapt to variations in anatomical features and differences in individual needs of different patients.

[0121] The full data record of the optimization process adopts a tiered storage strategy. Raw scan data, intermediate processing results, and final design models are stored in different levels of databases, linked by a unified timestamp and case number. Each optimization iteration generates a data package containing a complete processing log, recording the basis parameters and execution steps of model decisions. Log data is stored in a structured format, facilitating both machine automatic parsing and manual review and analysis. The system establishes an independent data version branch for each design task, supporting backtracking to any historical state at any time.

[0122] The output information on the designer's end is intelligently integrated and visualized. The system automatically extracts key parameter change curves and important decision nodes in the optimization process, generating a concise and concise design report. The report content adopts a hierarchical display method, the first level presents the overall design quality and main performance indicators, the second level shows the optimization details of each functional area, and the third level provides query interfaces for raw data and analysis methods. The visual interface supports interactive browsing of three-dimensional models, and designers can view the structural characteristics and parameter distribution of any cross section through gesture operations.

[0123] The storage system is designed to consider the long-term data management needs. The full data generated by each optimization process is archived in a standardized format, along with the generation of metadata description files, recording the data generation background and processing history. Archival data uses differential storage technology, saving only the incremental changes from the previous version, significantly reducing storage space occupancy. The data retrieval system supports multi-condition combination queries, allowing clinicians to quickly locate target data according to time range, anatomical region, design parameters, and other dimensions.

[0124] The adjustment results of model configuration parameters are transmitted to subsequent processing modules through a standardized interface. Interface data includes parameter change explanation, expected impact assessment, and implementation suggestions. After receiving these data, the subsequent modules first perform compatibility checks to confirm that parameter adjustments will not cause processing flow conflicts or contradictions. After passing the check, each module updates its internal processing logic according to the new parameter configuration. This update can be immediate or delayed until the next processing task begins, with the specific strategy determined dynamically based on the importance and impact of the parameters.

[0125] The system exception handling mechanism runs throughout the entire optimization process. When data anomalies or processing timeouts are detected, the system automatically starts the diagnosis process, analyzes the causes of the anomaly and attempts to self-repair. Diagnosis information is displayed to the designer in real time, along with multiple processing suggestions. For repeated occurrences of the same type of anomaly, the system automatically adjusts related parameter thresholds or optimization algorithm configurations to reduce the probability of future occurrences. In the event of a serious anomaly that causes processing to be interrupted, the system can automatically save the current state and continue execution from the breakpoint after recovery.

[0126] Clinical decision support functions are embedded in various stages of the optimization process. The system flags critical design points in the optimization results that require human confirmation, such as parameter adjustments beyond the normal range or innovative structural solutions. The flagging information includes detailed background explanations and reference cases to help the designer quickly understand the essence of the problem. For optimization choices with multiple processing options, the system presents a parallel comparison of the pros and cons of each option, but does not automatically make a final decision, leaving the decision-making power to the clinician.

[0127] The hardware resource management module dynamically regulates the computational load of the optimization process. The system monitors the use of CPU, memory, and video memory in real time, and automatically reduces the processing accuracy of non-critical tasks or suspends background analysis processes when resources are tight. For optimization iterations that take a long time, the system supports distributed computing and breakpoint resume functions, allowing tasks to be executed in parallel on multiple computing nodes or to be resumed from the middle after interruption. Resource allocation strategies are automatically adjusted according to different optimization stages, such as focusing on processing speed in the initial coarse optimization stage and focusing on computational accuracy in the later fine adjustment stage.

[0128] The security audit function records all modifications to system configuration parameters. Whether it is an automatic optimization triggered parameter adjustment or a designer manually input parameter change, the information such as operator identity, modification time, original parameter value and new parameter value will be recorded in detail. The audit log is stored using tamper-proof technology, supporting regular compliance checks and security vulnerability analysis. The parameter modification impact tracking function can show the chain effect of a specific adjustment on the subsequent design process, helping to understand the reasons for changes in system behavior.

[0129] The user interface design emphasizes the smoothness of human-machine collaboration. Designers can intervene in the automatic optimization process at any time, view the current state, adjust the processing direction or modify specific parameters. The interface operation and background calculation use asynchronous interaction mode, and the user's operation instructions enter the priority queue and do not block the ongoing optimization calculation. The system provides multiple view modes, both global overview of optimization progress and in-depth processing process of details in a certain anatomical area. The visual prompt system highlights the design changes and abnormal situations that need attention, but avoids excessive alarms that cause interference.

[0130] It is to be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting; it is not intended to exclude myriad other embodiments of the present application that other inventors can develop based on the same general inventive concepts embodied by the described embodiments. That is, although the present application is described in terms of particular embodiments and implementations, it is to be understood that the terminology used is for the purpose of descriptive clarity and that it should be taken in its broadest possible sense. For example, the terms "a", "an", and "the" include both singular and plural referents unless the context clearly dictates otherwise. The terms "comprises", "comprising", "includes", "including" and the like can be used in conjunction with the term "consisting of to include the elements or steps listed after such conjunctive language, but not to the exclusion of other elements or steps. The singular forms "a", "an" and "the" include plural referents unless the context clearly dictates otherwise.

[0131] While the embodiments of the application have been shown and described herein, it is to be understood that the application is not limited to these embodiments. Rather, many modifications, changes and substitutions are intended to fall within the scope of the present application, which is limited only by the scope of the following claims and their equivalents.

Claims

1. A method for personalized design of knee braces based on 3D printing, characterized in that, include: Three-dimensional scanning data of the patient's knee joint area is collected using a three-dimensional scanning device. The three-dimensional scanning data includes information on bone structure, soft tissue contours, and joint range of motion. Based on stereoscopic scan data and combined with a pre-set knee joint biomechanical knowledge base, a personalized anatomical feature set of the knee joint is determined. Based on a personalized anatomical feature set and at least one design rule file corresponding to that feature set, the stereoscopic scan data is divided into macroscopic structural regions and microscopic detail regions. Based on the division results of macroscopic structural regions and microscopic detail regions, a three-dimensional model of a knee brace is constructed from the stereoscopic scan data; The constructed 3D model is converted into a 3D printing instruction file, which is used to drive 3D printing equipment to produce personalized knee braces; In the process of dividing the macroscopic structural region and the microscopic detail region, the macroscopic structural region outputs skeletal framework data, which is then transmitted to the microscopic detail region to guide soft tissue matching analysis. The process of constructing a three-dimensional model of a knee brace from stereoscopic scan data includes: Based on the division of macroscopic structural regions and microscopic detail regions, the preprocessed stereoscopic scan data is split into macroscopic structural subsets and microscopic detail subsets. The macroscopic structural subset and associated biomechanical parameters are input into the structural optimization model, and the microscopic detail subset is input into the detail generation model. By combining the outputs of the structural optimization model and the detail generation model, a complete 3D model of the knee brace is generated, and intermediate data for each model construction is recorded. The execution process of the structure optimization model and the detail generation model includes: The structural optimization model is established based on the geometric constraint algorithm. The structural optimization model first performs contour fitting on the macroscopic structural subset to obtain the brace support frame information. By comparing the matching degree between the support frame information and biomechanical parameters, it is determined whether the mechanical stability requirements are met. The detail generation model is used for surface texture analysis, enabling adaptive smoothing of subtle surfaces in the micro-detail subset, and supporting framework information as input to the surface texture analysis to enhance data consistency. When performing contour fitting on the macroscopic structure subset, the process involves extracting microscopic detail enhancement regions from the initial macroscopic structure subset state. This is an iterative optimization process, and the specific steps of the iterative optimization include: Obtain a macroscopic structural subset of data. For each local unit in the macroscopic structural subset of data, calculate its rate of curvature change and density distribution index. When the rate of curvature change or density distribution index of any local unit exceeds a preset threshold, it is determined that the local unit has a geometric defect that needs to be finely reconstructed. Based on the identified local units of geometric defects, the corresponding micro-detail enhancement dataset is separated from the macro-structure subset dataset; Based on the geometric defect characteristics of different knee joint types, data extraction rules associated with the defect characteristics are established, and the geometric defect judgment results are directly input into the data extraction rule module to drive the separation operation.

2. The method for personalized design of knee braces based on 3D printing as described in claim 1, characterized in that, The process of determining the personalized anatomical feature set of the knee joint includes: The stereoscopic scan data is input into a multi-feature recognition model, which outputs key anatomical landmarks in the scan data. Key anatomical landmarks are input into a pre-defined knee joint biomechanical knowledge base. Based on the association mapping relationship in the knowledge base, the set of biomechanical properties including all landmarks is identified. Based on the set of biomechanical characteristics, the range of joint movement angles and load distribution parameters are calculated, which are used for data interaction in subsequent 3D model construction.

3. The method for personalized design of knee braces based on 3D printing as described in claim 1, characterized in that, The process of dividing the macroscopic structural region and the microscopic detail region includes: Based on the personalized anatomical feature set, the design rule file corresponding to the feature set is called, and the intelligent analysis engine performs semantic parsing on the design rule file to output multiple design constraints. Based on design constraints, the stereoscopic scan data is segmented into macroscopic structural regions and microscopic detail regions. The macroscopic structural regions generate skeletal topology data, while the microscopic detail regions generate soft tissue deformation data. The skeletal topology data is directly input into the soft tissue deformation analysis module to optimize the segmentation results.

4. The method for personalized design of knee braces based on 3D printing as described in claim 1, characterized in that, The specific steps of the iterative optimization also include: After separating the micro-detail enhancement dataset from the macro-structure subset, a data synchronization mechanism is established between the structure optimization model and the detail generation model. The structural optimization model is used to perform a frame evaluation on the separated region, and the coordinate range, size parameters and evaluation results of the separated region are passed to the detail generation model. After receiving the information from the structural optimization model, the detail generation model performs surface reconstruction on the micro-detail enhancement dataset, detects and generates specific surface optimization information, including texture coordinates, curvature change trends, and integration status with the supporting frame information. The surface optimization information is fed back to the frame evaluation module to complete closed-loop optimization.

5. The method for personalized design of knee braces based on 3D printing as described in claim 4, characterized in that, The specific steps of the iterative optimization also include: The generated surface optimization information is fed back to the structural optimization model in real time; During multiple data separation and model optimization processes, the optimization threshold and model configuration parameters are adjusted based on the deviation between each optimization result and the preset standard. The final generated complete 3D model information, key parameters, and optimization logs are output to the designer's end, and the full data record of each optimization process is stored. The model configuration parameter adjustment results are passed to the subsequent 3D model construction steps to update the processing logic.

6. The method for personalized design of knee braces based on 3D printing as described in claim 1, characterized in that, The specific process of separating the corresponding micro-detail augmentation dataset from the macro-structure subset includes: Identify local units with geometric defects in a macroscopic structural subset of data. These local units with geometric defects are composed of boundary voxels defined by a set of three-dimensional coordinate points. Based on the local unit of geometric defects, the parameters required to separate the micro-detail enhancement dataset are calculated, including the starting coordinate point, voxel width, and height dimension. Using the calculated separation parameters, the corresponding micro-detail enhancement dataset is separated from the macro-structure subset dataset through data pruning operations; Verify whether the separated micro-detail enhancement dataset completely covers the defect area and eliminate redundant data interference. The separation parameters are output to the verification module for automatic verification.

7. The method for personalized design of knee braces based on 3D printing as described in claim 1, characterized in that, When combining the outputs of the structural optimization model and the detail generation model, a fusion weight allocation mechanism is adopted. Different weight coefficients are assigned according to the confidence scores of the structural optimization model and the detail generation model. The confidence scores are obtained through training with historical data and updated in real time. The weight coefficients are output to the model output integration module to coordinate the generation of the 3D model.

Images