A method for fabricating and driving a knee orthotic brace based on 4D printing, and the brace structure.

Abstract

This invention relates to a method for fabricating and driving a knee orthotic brace based on 4D printing, as well as the brace structure. The fabrication and driving methods combine three-dimensional scanning surface data of the knee joint with medical images to establish a biomechanical model with knee joint deformity characteristics, providing a model basis for printing the brace and predictive parameters for driving the brace, thus improving its practicality. The brace structure includes a first orthotic plate and a second orthotic plate, connected by a first connecting plate and a second connecting plate, facilitating assembly and ensuring that it does not affect walking after being worn. The driving method utilizes the biomechanical model with knee joint deformity characteristics, and then combines it with a deep learning algorithm to construct a knee joint orthotic target angle prediction model, providing a reference for various morphological changes of the orthotic brace. Furthermore, the model can be adjusted and reused according to specific stages, improving efficiency and resource utilization.

Description

Technical Field

[0001] This invention belongs to the field of orthotic brace technology, specifically a method for preparing and driving a knee orthotic brace based on 4D printing, the brace structure, and the driving method. Background Technology

[0002] Traditional knee orthotic braces are typically standardized designs, making it difficult to perfectly match individual patient differences. This results in limited corrective effects and may even cause discomfort or complications. Furthermore, traditional braces lack intelligent and dynamic adaptability during manufacturing, failing to adjust to biomechanical changes during patient rehabilitation. With the continuous development of 3D printing technology, especially the emergence of 4D printing, new possibilities have been provided for personalized customization and dynamic adaptation of orthotic braces. 4D printing technology combines three-dimensional spatial printing with shape changes over time, enabling more complex structures and functions, bringing revolutionary changes to the design and manufacturing of orthotic braces. Summary of the Invention

[0003] The purpose of this section is to outline some aspects of embodiments of the present invention and to briefly describe some preferred embodiments. Simplifications or omissions may be made in this section, as well as in the abstract and title of this application, to avoid obscuring the purpose of these documents; however, such simplifications or omissions should not be construed as limiting the scope of the invention.

[0004] Given the following technical problems with the existing technology: the existing braces are used by changing different braces according to the patient's condition at each stage, which is too costly; and the production steps of the braces at each stage are also cumbersome.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a method for preparing a knee orthotic brace based on 4D printing, comprising,

[0006] S1, collect patient data, including three-dimensional surface data of the knee joint and medical imaging CT or MRI data;

[0007] S2, a biomechanical model of knee joint deformity characteristics obtained by three-dimensional reconstruction of medical imaging data;

[0008] S3, the finite element analysis results of the initial orthotic brace structural model, and the 4D printing model is obtained after designing the structure of the orthotic brace based on the analysis results;

[0009] S4 utilizes shape memory polymer smart materials to 4D print the obtained model using melt direct molding technology, and then assembles it into orthopedic braces.

[0010] As a preferred technical solution for the fabrication and driving method of knee orthotic braces based on 4D printing, the biomechanical model for obtaining knee joint deformity characteristics includes,

[0011] Import medical imaging data;

[0012] Threshold segmentation preserves the target knee joint bone while removing most soft tissue and noisy data.

[0013] 3D reconstruction generates skeletal model B;

[0014] Noise removal and model integrity restoration were performed on skeletal model B to obtain the optimized 3D skeletal model B. * ;

[0015] A biomechanical model M with knee joint deformity characteristics was established by combining the patient's known deformity feature parameters. bio , represented as:

[0016] M bio = (B * ,P,Λ)

[0017] Among them: B * The above-mentioned optimized skeletal geometric model; P represents the physical and mechanical properties of the bones and soft tissues; Λ represents the set of deformity angles, joint lines, or other deformity characteristic parameters.

[0018] As a preferred technical solution for the fabrication and driving method of knee orthotic braces based on 4D printing, the threshold segmentation includes setting an initial threshold range [I] based on the difference in grayscale values ​​(or signal intensity) between bone tissue and soft tissue in CT / MRI scans. min I max ];

[0019] For image voxel position x, if I min ≤I(x)≤I max If the voxel is identifiable as belonging to skeletal tissue, its mask label M(x) = 1; otherwise, M(x) = 0; as follows:

[0020]

[0021] Based on whether the masked 3D preview image fully displays the patient's lower limb bones and joint structures, if there is a significant amount of soft tissue or residual noise, adjust [I]. min I max The range of thresholds; if there are breaks or gaps in the bone structure, the upper and lower limits of the thresholds will be modified accordingly.

[0022] As a preferred technical solution for the fabrication and driving method of knee orthotic braces based on 4D printing, the initial orthotic brace structural model undergoes finite element analysis preprocessing, including dimensional checks and necessary adjustments to the model:

[0023] Based on the actual use of the brace and biomechanical analysis, corresponding loads and boundary conditions are applied to the outer surface of the brace to simulate the actual stress conditions of the brace under gravity, compression, and corrective torque.

[0024] Based on the finite element analysis results, the mechanical strength and stability of the model are determined, the displacement distribution of the model is determined by the displacement cloud map, the location and magnitude of the maximum displacement are read, and the impact of the degree of deformation on the functionality is evaluated.

[0025] To address issues such as stress concentration areas and areas of excessive displacement, the structure of the orthotic brace is optimized by adjusting its shape, position, size, or changing the overall structural layout. After the design is completed, a 4D printing model is obtained, which is used for subsequent 4D printing of knee orthotic braces.

[0026] As a preferred technical solution for the fabrication and driving method of knee orthotic braces based on 4D printing, the printing material used for the orthotic brace is a shape memory polymer smart material.

[0027] As a preferred technical solution for the fabrication and driving method of knee orthotic braces based on 4D printing,

[0028] In step S4,

[0029] Set the nozzle temperature in the printing parameters to within 5°C of the melting point of the selected smart polymer material to ensure that the material can melt fully without excessive degradation, thus preserving the optimal molecular chain state for subsequent shape memory properties.

[0030] Set the substrate temperature in the printing parameters to within the range of -15℃ to -20℃ of the glass transition temperature of the selected smart polymer material. Printing at this temperature, slightly below the glass transition temperature, will allow the activity of the smart polymer material molecules to be reasonably regulated.

[0031] As a preferred technical solution for the preparation and driving method of knee orthotic brace based on 4D printing, the method includes: calculating the varus / valgus angle based on the knee joint biomechanical model, constructing a knee joint orthotic target angle prediction model using a deep learning algorithm, and obtaining the orthotic brace structural model by compensating for the orthotic angle.

[0032] By driving the deformation and stiffness adjustment of the orthotic brace through external stimuli, and changing the brace's correction angle by combining the orthotic brace structural model, an orthotic brace with orthotic capabilities can be obtained.

[0033] During the re-correction, the deformity data is modified, and a new correction angle is obtained to obtain the structural model of the orthotic brace for the next stage.

[0034] As a preferred technical solution for the fabrication and driving method of knee orthotic braces based on 4D printing, the step of constructing a knee orthotic target angle prediction model using a deep learning algorithm includes:

[0035] Acquire and preprocess 3D data of the patient's knee joint, and divide the dataset;

[0036] Define network models and network structures;

[0037] Model training and loss function construction;

[0038] Cross-validate and optimize the model;

[0039] Model output.

[0040] As a preferred technical solution for the fabrication and driving method of knee orthotic braces based on 4D printing, the acquisition and preprocessing of the patient's three-dimensional knee joint data, and the division of the dataset, include:

[0041] Based on the obtained three-dimensional data of the patient's knee joint and the clinically known target varus / valgus angles after orthopedics, training datasets and validation datasets are formed.

[0042] The patient's original knee joint deformity angle and the final clinically confirmed corrected angle were used as training labels, and the corrected angle was denoted as Y. i ;

[0043] Multimodal data is formatted in a unified manner, and relevant mechanical parameters are structured and encoded for input into the network.

[0044] As a preferred technical solution for the fabrication and driving method of knee orthotic braces based on 4D printing, the input layer is defined as:

[0045] Let the preprocessed data be denoted as X, where:

[0046] X = {x1, x2, ..., x} N}

[0047] Each training data x i The patient's knee joint's three-dimensional morphological features and mechanical parameters are input into the network in the form of vectors or tensors.

[0048] Define the output layer as:

[0049] That is, output the predicted value of the patient's target varus / valgus angle. Used for regression prediction.

[0050] As a preferred technical solution for the fabrication and driving method of knee orthotic braces based on 4D printing,

[0051] The loss function L is defined as follows:

[0052]

[0053] in: Y represents the predicted angle of the model for the i-th sample; i Let be the true corrected angle (label value) of the i-th sample; Ω(W) is the regularization term, where W represents all trainable parameters in the network; α is the regularization coefficient, used to balance data fitting error and model complexity.

[0054] The orthotic brace is then repositioned to a different angle by external stimulation to induce deformation and stiffness control. Specifically, one or more of the following methods can be used: thermal stimulation, electrical stimulation, optical stimulation, magnetic stimulation, and microwave stimulation. This aims to drive the orthotic brace to achieve the desired correction angle.

[0055] A brace structure prepared according to the aforementioned 4D-printed knee orthotic brace preparation and driving method includes,

[0056] A first orthotic plate and a second orthotic plate, wherein a first connecting plate is provided at one end of the first orthotic plate and a second connecting plate is provided on the second orthotic plate, and the first connecting plate and the second connecting plate are connected by a clamp.

[0057] The first connecting plate is provided with a first deformation section, and the second connecting plate is provided with a second deformation section;

[0058] The first connecting plate has a first gear tooth at its end and the second connecting plate has a second gear tooth at its end. The first gear tooth and the second gear tooth mesh with each other. The first connecting plate has a first through hole and the second connecting plate has a second through hole. A first pin passes through the first through hole and is connected to the clamping plate. A second pin passes through the second through hole and is connected to the clamping plate.

[0059] The first orthotic plate is provided with a first strap hole, and the second orthotic plate is provided with a second strap hole.

[0060] The beneficial effects of this invention are:

[0061] By collecting detailed data and reconstructing 3D data from patients, a biomechanical model with specific knee joint deformity characteristics is generated, thereby designing an orthotic brace that is highly matched to the individual patient's condition. A knee joint orthotic target angle prediction model is constructed using deep learning algorithms, which not only improves the prediction accuracy of the orthotic angle, but also compensates for the initial orthotic angle based on the prediction results, ensuring that the initial design of the orthotic brace structure is close to the optimal state. Furthermore, the 4D-printed orthotic brace can be repeatedly adjusted to adapt to the orthotic stage without the need for remanufacturing, reducing expenses. Attached Figure Description

[0062] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein:

[0063] Figure 1 This is a schematic diagram of the overall method flow structure in this invention;

[0064] Figure 2 This is a schematic diagram of the software interface after the data import is completed in this invention;

[0065] Figure 3 This is a schematic diagram of the threshold range adjustment interface in this invention;

[0066] Figure 4 This is a schematic diagram of the 3D mask model after threshold range adjustment according to the present invention;

[0067] Figure 5 This is a schematic diagram of the export parameter setting interface in this invention;

[0068] Figure 6 This is a schematic diagram of the interface for model optimization and knee joint varus / valgus angle adjustment in this invention;

[0069] Figure 7 This is a schematic diagram of the overall structure of the orthopedic brace;

[0070] Figure 8 This is a schematic diagram of the connection point of the orthotic brace.

[0071] Reference numerals: First gear 102, Second gear 201a, First connecting plate 101, Second connecting plate 201, First through hole 101b, First pin 101c, Second through hole 201b, Second pin 201c, Clamping plate 300, First orthotic plate 100, First strap hole 101d, Second orthotic plate 200, Second strap hole 201d. Detailed Implementation

[0072] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0073] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0074] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0075] Secondly, the present invention is described in detail with reference to the schematic diagrams. When detailing the embodiments of the present invention, for ease of explanation, the cross-sectional views illustrating the device structure may be partially enlarged, not according to the usual scale. Furthermore, the schematic diagrams are merely examples and should not limit the scope of protection of the present invention. In addition, actual fabrication should include three-dimensional spatial dimensions of length, width, and depth.

[0076] Example 1

[0077] Reference Figures 1 to 8 This embodiment provides a preferred technical solution for the fabrication and driving method of a knee orthotic brace based on 4D printing, including:

[0078] Step 1: Collect patient data, including 3D surface data of the knee joint and medical imaging CT or MRI data;

[0079] Step 2: Perform three-dimensional reconstruction of medical imaging data to obtain a biomechanical model of knee joint deformity characteristics; calculate the varus / valgus angle based on the knee joint biomechanical model, and compensate for the initial orthopedic angle to obtain the initial orthopedic brace structure model;

[0080] Step 3: Perform finite element analysis on the initial orthotic brace structural model, and design the structure of the orthotic brace based on the analysis results to obtain the 4D printed model;

[0081] Step 4: Using shape memory polymer smart materials, the obtained model is 4D printed using melt direct molding technology, and the orthopedic brace is assembled, including a hinge structure, angle indicator, and length adjustment device.

[0082] Step 5: Drive the orthotic brace to deform and adjust its stiffness through external stimuli, thereby changing the brace's correction angle to meet the patient's ever-changing biomechanical needs during the rehabilitation process.

[0083] Furthermore, steps 1 and 2 specifically include:

[0084] Collect medical imaging data of the patient's areas requiring correction. The medical imaging data is CT or MRI scan data conforming to the DICOM image file standard format. Reconstruct the patient's areas requiring correction, optimize the model by denoising, and obtain a biomechanical model with knee joint deformity characteristics.

[0085] The angle measurement function was used to measure the varus / valgus angles of the knee joint in a biomechanical model with knee joint deformity characteristics, and the initial varus / valgus angles of the patient were obtained.

[0086] A knee orthopedic target angle prediction model is constructed using deep learning algorithms, such as convolutional neural networks (CNNs) or recurrent neural networks (RNNs). Through learning from training data, features related to the orthopedic target angle are automatically extracted, and a complex mapping relationship between input parameters and the target angle is established. Cross-validation and regularization techniques are used to optimize the model, outputting the target varus / valgus angle after orthopedic correction for the patient. The initial orthopedic brace structural model is then compensated. The initial orthopedic brace structural model is obtained.

[0087] The biomechanical models obtained with knee joint deformity characteristics include:

[0088] (1) Data reading and import

[0089] Medical imaging data of the patient's knee joint (the area requiring orthopedic surgery) is obtained through CT or MRI scans and stored as a file conforming to the DICOM standard format.

[0090] For each patient, the medical imaging data can be denoted as D, where D = {d1, d2, ..., dn}. n}, d i Let be the i-th computed tomographic image; n is the total number of CT / MRI images in this group.

[0091] In the Mimics software main interface, select "File" → "New projects", select the DICOM data folder in the pop-up file browser, and then click "Next".

[0092] The software automatically reads and displays basic information about the DICOM sequence image, including the number of image layers, layer thickness, pixel spacing, etc.; if multiple sets of DICOM data need to be added, repeat the above operation.

[0093] Click "Open" and wait for the software to finish processing. Mimics will then present a preview of the base image after the CT or MRI data reconstruction is complete. You can use this preview interface to verify the integrity and accuracy of the data.

[0094] (2) Threshold segmentation and mask creation

[0095] In the Mimics interface, in the project management menu on the right, click the mask option, right-click in the blank area and select "New Mask" to open the "Thresholding" threshold settings interface.

[0096] Based on the difference in grayscale values ​​(or signal intensity) between bone tissue and soft tissue in CT / MRI scans, an initial threshold range [I] is set. min I max ].

[0097] The corresponding mathematical expression can be defined as: for image voxel position x, if I min ≤I(x)≤I max If the voxel is 1, then it is considered to belong to the skeletal tissue, and its mask label M(x) = 1; otherwise, M(x) = 0, as follows:

[0098]

[0099] After clicking the confirmation button on the threshold setting interface, the software will display the preliminary 3D model of the skeleton obtained based on the threshold range in real time in the 3D preview area on the right.

[0100] Based on whether the masked 3D preview image fully displays the patient's lower limb bones and joint structures, if there is a significant amount of soft tissue or residual noise, adjustments can be made accordingly. min I max The range of thresholds; if there are breaks or gaps in the bone structure, the upper and lower limits of the thresholds will be modified accordingly.

[0101] Repeat the above adjustment process until the target knee joint bones can be fully and clearly displayed in the preview model, and as much soft tissue and noise data as possible is removed.

[0102] (3) Generating a skeletal model through 3D reconstruction

[0103] After achieving satisfactory threshold segmentation results, you can select either "Calculate 3D" or "3D Preview / Update" in the mask management interface. The software will then automatically generate a 3D surface model based on the established mask M(x). This model is stored as a triangular patch or a polygonal mesh, denoted as B.

[0104] B is a three-dimensional geometric representation of the patient's lower limb bones (especially the knee joint area), which includes information on the location and morphology of bony structures.

[0105] Export the 3D skeletal model B in commonly used formats such as STL, OBJ, or PLY for subsequent denoising and optimization.

[0106] In subsequent algorithms or software modules, B needs to be associated with the patient's deformity features (such as known deformity angles or force line offsets) to obtain a biomechanical model with knee joint deformity features.

[0107] (4) Noise reduction and other optimizations

[0108] The 3D model B is denoised or filtered to remove isolated floating meshes, spikes, or non-bone fine structures.

[0109] Common methods include:

[0110] Mesh smoothing: Using algorithms such as Laplacian smoothing or bidirectional filtering, vertex normals or coordinates are iteratively updated to remove local burrs on the model surface;

[0111] Small Component Removal: Through connected component analysis, if the number of faces or voxels in a certain connected mesh is less than a preset threshold δ, it is considered noise or an artifact and removed.

[0112] For holes or tortuosities on the model surface caused by threshold segmentation or scanning defects, topological repair (HoleFilling) or interpolation compensation (Interpolation) can be performed; or key areas (such as the distal femur, proximal tibia, etc.) can be cut or selected according to clinical needs.

[0113] The final optimized 3D skeletal model is denoted as B. * It has been improved in terms of geometric continuity, surface smoothness and deformity feature representation, and can be used for subsequent mechanical simulation and correction parameter calculation.

[0114] Next, add deformity feature information and build a biomechanical model: Model B... * Combined with the patient's known deformity characteristics (such as the original varus / valgus angle, knee joint force line offset distance, etc.), and data annotation or attribute assignment is performed according to mechanical boundary conditions (such as the mechanical properties of muscles, ligaments or cartilage).

[0115] In Mimics or subsequent mechanical analysis software (such as ANSYS, Abaqus, etc.), a biomechanical model M with knee joint deformity characteristics is established based on bone material properties, joint contact surfaces, soft tissue support, and force line direction. bio It can be represented as: M bio = (B * ,P,Λ)

[0116] in:

[0117] B * The above is the optimized skeletal geometry model;

[0118] P represents the physical and mechanical properties of bones and soft tissues (such as elastic modulus, Poisson's ratio, tendon tensile strength, etc.);

[0119] Λ is a set of deformity angles, joint force lines, or other deformity characteristic parameters.

[0120] Through the specific implementation of step S1 above, the following technical effects can be achieved:

[0121] Efficient data import: Utilizing multiple CT / MRI tomographic images conforming to the DICOM standard format ensures the integrity and accuracy of the patient's original knee joint data;

[0122] Threshold segmentation and mask management: Using threshold strategies in Mimics software to accurately segment bone tissue and quickly generate preliminary 3D models;

[0123] Denoising and model optimization: Through operations such as smoothing, component removal, and hole repair, the geometric continuity and reliability of the model are improved;

[0124] Biomechanical information integration: Combining the patient's deformity angle and mechanical boundary conditions, an accurate biomechanical basis is laid for subsequent orthopedic parameter calculations and 4D printed brace design.

[0125] In summary, step S1, through reading, thresholding, masking, and post-denoising optimization of the patient's knee CT / MRI images, not only obtains a complete and clear three-dimensional skeletal model, but also constructs a biomechanical model M by combining deformity feature parameters. bio This provides scientific and precise data support for the subsequent steps (steps S2-S5) of orthotic brace design and 4D printing process. The process is operable, accurate, and innovative, and can be applied to the technical solution of the "4D Printed Knee Orthotic Brace Preparation and Driving Method" involved in this invention. Specific Implementation

[0127] I. Data Import

[0128] (I) Preparing Medical Imaging Data

[0129] Collect medical imaging data (CT or MRI scan data) of the areas of the patient requiring correction, ensuring that the data format is the standard DICOM image file format, stored in the same folder, and that the sequence is complete, the images are clear, and there are no artifacts or motion interference. Data quality directly affects the accuracy and effect of reconstruction.

[0130] (II) Launching Mimics software and importing data

[0131] After launching Mimics software, select "File" - "New projects" on the main interface. In the pop-up file browser, select the DICOM data folder and click "Next". The software will automatically read the data and display the image sequence in the image browsing interface. You can view basic image information (such as number of slices, slice thickness, pixel spacing), verify the integrity and accuracy of the data, and add multiple sets of DICOM data files. Finally, click "Open" and wait for the software to complete the processing. The software will then display the reconstructed image from the CT or MRI scan data.

[0132] Reference Figure 2 This is a schematic diagram of the software interface after the data import is complete.

[0133] II. Image Processing

[0134] (I) Threshold Adjustment

[0135] In the Project Management menu on the right, click the Mask option, right-click in the blank area to open the menu bar, select New Mask, initially adjust the "Thresholding" threshold range, and click OK.

[0136] Figure 3 This is a schematic diagram of the threshold range adjustment interface.

[0137] At this point, the lower right view area is still blank. Click the 3D mask preview icon on the right, and a 3D mask model will appear in the blank lower right view area. Continuously adjust the threshold range based on the 3D mask preview. Since the grayscale value of bone tissue is higher than that of surrounding soft tissue, use the difference in tissue grayscale value to initially filter bone and joint tissues. Move the slider to set the threshold range until the 3D model clearly and completely displays the patient's lower limb bones and joints, while minimizing the appearance of other soft tissues. The desired 3D model of the patient's skeleton is obtained, but further optimization is needed due to image noise and partial volume effects.

[0138] Figure 4 The bottom right corner shows a schematic diagram of the masked 3D model after the threshold range has been adjusted.

[0139] (II) Exporting STL files

[0140] On the main interface, select "File" - "Export" - "STL". For easier reading and processing in 3D modeling software later, choose either ASCII STL or binary STL format. There are differences between ASCII and binary STL files. ASCII is text-based, readable and editable, with an identifier in the header. Face data is recorded line by line, resulting in a neat but redundant file. Binary, on the other hand, uses compact binary encoding, cannot be read directly, and stores data sequentially after reserving header information and face counts. This results in smaller files and faster read / write speeds. Each format is suitable for specific 3D model processing needs; choose based on your computer configuration and processing requirements. After selecting the file format, select the masked 3D model to export, click Next to set the export parameters, and then click Finish. The software will convert the 3D model to an STL file and save it to the specified path.

[0141] Figure 5 A diagram of the export parameter settings interface.

[0142] III. Model Angle Calculation

[0143] (I) Model Optimization

[0144] Open the STL file of the masked 3D model using ZBrush software, optimize the model by deleting unnecessary background and soft tissue, and use the software's editing tools for further optimization. The "Smoothing" tool smooths the model surface according to the set number of iterations and parameters, reducing jagged edges; the "Remeshing" tool adjusts the model's mesh layout and density to improve topology quality to meet the needs of subsequent analysis and manufacturing.

[0145] Figure 6 Schematic diagram of the model optimization and knee valgus / varus angle adjustment interface

[0146] The knee joint orthopedic target angle prediction model constructed using deep learning algorithms includes:

[0147] 1. Data Acquisition and Preprocessing

[0148] (1) Obtain training data:

[0149] Based on the three-dimensional data of the patient's knee joint obtained in step S1 (including the knee joint biomechanical model information obtained from body surface scan data and CT / MRI medical image reconstruction) and the clinically known target varus / valgus angle after correction (or the actual correction angle value verified by surgery and rehabilitation program), a training dataset and a validation dataset are formed.

[0150] (2) Data labeling:

[0151] The patient's original knee joint deformity angle (e.g., varus / valgus angle) and the final clinically confirmed corrected angle are used as training labels. For supervised learning in the deep learning model, this corrected angle needs to be denoted as Y. i .

[0152] (3) Data preprocessing:

[0153] Multimodal data (such as 3D surface meshes, volumetric data reconstructed from CT / MRI, and patient physiological and mechanical parameters) undergo uniform formatting processing, including size normalization, coordinate alignment, and noise filtering, to ensure that the input data meets the requirements of subsequent deep learning models. Simultaneously, relevant mechanical parameters (such as force line distribution and soft tissue data) are structured and encoded for input into the network.

[0154] 2. Model Building and Network Structure Design

[0155] (1) Types of deep learning networks:

[0156] Depending on the actual data type and feature extraction requirements, choose deep network structures such as Convolutional Neural Networks (CNN) or Recurrent Neural Networks (RNN). For situations with a large amount of 3D morphology and image data, CNN structures can be used first to extract spatial or image features.

[0157] (2) Input layer definition:

[0158] Let the preprocessed data be denoted as X, where:

[0159] X = {x1, x2, ..., x} N}

[0160] Each training data x i The patient's knee joint's three-dimensional morphological features (such as point cloud or voxel data) and mechanical parameters (such as joint space, force line distribution, bone density information, etc.) are input into the network in the form of vectors or tensors.

[0161] (3) Network main structure:

[0162] For CNNs, they can consist of multiple convolutional layers, pooling layers, and fully connected layers; for RNNs, they can consist of multiple recurrent units (LSTM or GRU) and fully connected layers. The main function of the network is to automatically learn deep features that are highly correlated with the correction angle in multimodal knee joint data.

[0163] (4) Output layer definition:

[0164] The output layer is used for regression prediction, that is, it outputs the predicted value of the patient's target varus / valgus angle. In this technical solution, Record It means "the predicted target angle (varus / valgus direction) of the patient's knee joint after orthopedic surgery".

[0165] 3. Model Training and Loss Function Construction

[0166] (1) Training dataset and validation dataset:

[0167] The dataset is divided into training and validation sets according to a certain ratio for model training and performance validation.

[0168] (2) Loss function:

[0169] The mean squared error (MSE) in regression form is used as the basic loss function, combined with a regularization term to suppress overfitting. Assuming there are N samples in the training set, the loss function L can be defined as:

[0170] in:

[0171] Let be the predicted angle of the model for the i-th sample;

[0172] Y i The true post-correction angle (label value) for the i-th sample;

[0173] Ω(W) is the regularization term, where W represents all trainable parameters in the network;

[0174] α is the regularization coefficient, used to balance data fitting error and model complexity.

[0175] (3) Parameter update and optimization:

[0176] The network weights W and bias b are updated by minimizing the loss function L using stochastic gradient descent (SGD), Adam, or other optimization algorithms. The update process is as follows:

[0177]

[0178] Where η is the learning rate.

[0179] Furthermore, K-fold cross-validation or leave-one-out cross-validation (LOOCV) is used to evaluate the training process. The training dataset is divided into K folds, and one fold is used as the validation set in each iteration, while the remaining K-1 folds are used as the training set. The training is repeated cyclically, and the validation error is calculated. Finally, the stability and generalization ability of the model are evaluated by combining the results of K validation iterations.

[0180] (2) Regularization and hyperparameter selection:

[0181] Based on the validation error and the size of the regularization term, select appropriate regularization coefficient α and network structure hyperparameters (such as convolutional kernel size, number of network layers, number of hidden units, learning rate η, etc.) to ensure that the model avoids overfitting or underfitting while maintaining prediction accuracy.

[0182] (3) Early stop strategy:

[0183] During training, monitor the validation set loss. When the validation set loss no longer decreases or shows an increasing trend after several consecutive iterations, terminate training and retain the model parameter combination W that performs best on the validation set. * ,b * .

[0184] After training, the final "knee joint orthopedic target angle prediction model" can be represented as:

[0185]

[0186] in:

[0187] f(·) is the nonlinear mapping function represented by the deep neural network;

[0188] W * ,b * These are the network parameters after training and optimization;

[0189] X represents the multimodal features (knee joint biomechanical model, 3D data, etc.) corresponding to the current patient (or a newly input sample).

[0190] Model output The predicted optimal varus / valgus angle for the patient takes into account the patient's original deformity characteristics obtained in step S1, skeletal geometry, and soft tissue biomechanical factors.

[0191] Will This serves as the target correction angle (including compensation for the initial correction angle) required for subsequent orthotic brace design. In steps S3 to S5, the 4D-printed brace structure can be designed, optimized, and dynamically adjusted based on this prediction result to meet the individualized needs of patients during their rehabilitation process.

[0192] Through the specific implementation of step S2 above, the following technical effects can be achieved:

[0193] Automatic feature extraction: Effective features are extracted from knee joint 3D data and its biomechanical information using deep learning networks such as CNN or RNN;

[0194] Personalized complex mapping: Mapping the patient's multimodal data to personalized orthopedic target angles to achieve more accurate prediction of varus / valgus angles;

[0195] Model generalization ability: By using techniques such as cross-validation and regularization, overfitting of the model is prevented, and the robustness and accuracy of predictions are improved.

[0196] It can be linked with subsequent steps: The output of this prediction model can be directly used in subsequent steps S3 to S5 to guide the design of orthopedic braces and the 4D printing process.

[0197] By implementing the aforementioned deep learning method, this invention enables 4D-printed knee orthotic braces to be more personalized and accurate in angle setting, which can fully meet the ever-changing biomechanical needs of patients during the rehabilitation process, demonstrating significant innovation and feasibility.

[0198] Open the STL file of the masked 3D model using 3D modeling software. Measure the knee joint's valgus and varus angles using the angle measurement function. The patient's initial valgus angle is X1, which is adjusted to X2 according to the clinician's treatment plan. The skeletal model and surface data at this point serve as the basis for brace design. Based on the recalculated skeletal model and surface data, an initial orthotic brace structural model is obtained.

[0199] Specifically, step S3 includes:

[0200] The initial orthotic brace structural model undergoes finite element analysis preprocessing, including dimensional checks and necessary adjustments.

[0201] Based on the actual use of the brace and biomechanical analysis, corresponding loads and boundary conditions are applied to the outer surface of the brace to simulate the actual stress conditions of the brace under gravity, compression, and corrective torque.

[0202] Based on the finite element analysis results, the mechanical strength and stability of the model are determined, the displacement distribution of the model is determined by the displacement cloud map, the location and magnitude of the maximum displacement are read, and the impact of the degree of deformation on the functionality is evaluated.

[0203] To address issues such as stress concentration areas and areas of excessive displacement, the structure of the orthotic brace is optimized by adjusting its shape, position, size, or changing the overall structural layout. After the design is completed, a 4D printing model is obtained, which is used for subsequent 4D printing of knee orthotic braces. Specific Implementation

[0205] The acquired 3D models of fracture and soft tissue injury sites are preprocessed, including model size design, cutout design, and file format conversion.

[0206] In the finite element analysis software, the model is meshed. Meshing should, in principle, use appropriate mesh type and size to ensure a balance between computational accuracy and efficiency. If no problems are found after inspection, proceed to the next step; if problems are found, the mesh is re-generated.

[0207] After importing the model, create a new material and assign it an elastic modulus, Poisson's ratio, and density. These are 2000 MPa, 0.34, and 1.25 ePa, respectively. -9 T / mm 3 These values ​​represent the basic properties of PLA (polylactic acid) materials.

[0208] Create a "shell" section of type "mean". Select the area to be assigned to the interface. In the viewport, select the entire model with the mouse. Click "Finish". Select the shell section created in the previous step for the entire model's section and fill in the thickness value according to the actual size of the model.

[0209] Based on the actual usage of the brace and biomechanical analysis, corresponding loads and boundary conditions are applied to the outer surface of the brace (such as simulating compression and corrective torque).

[0210] After applying loads and setting boundary conditions on the model, create and configure a static analysis step for simulation. Finally, create a new job, submit it, and wait for the software to run and output the results.

[0211] After the calculations are completed, the results interface allows analysis of the obtained data, primarily focusing on the stress and displacement contour maps. The stress contour map determines the stress distribution within the model, identifying the location and magnitude of the maximum stress. This stress is compared to the material's yield strength to assess the risk of yielding or failure, determining the model's mechanical strength and stability. The displacement contour map reveals the model's displacement distribution and displays the location and magnitude of the maximum displacement. Based on the maximum displacement, the degree of model deformation is evaluated, assessing the impact of this deformation on functionality.

[0212] Specifically, the orthotic braces use shape memory polymer smart materials for printing, including polyurethane shape memory polymers, polynorbornene shape memory polymers, polylactic acid shape memory polymers, epoxy resin shape memory polymers, styrene shape memory polymers, polycaprolactone (PCL) based shape memory polymers, polymethyl methacrylate (PMMA) based shape memory polymers, polyetheretherketone (PEEK) based shape memory polymers, ethylene-vinyl acetate copolymer (EVA) based shape memory polymers, liquid crystal elastomer shape memory polymers, polycarbonate (PC) based shape memory polymers, polyvinylidene fluoride (PVDF) based shape memory polymers, silicone rubber based shape memory polymers, and polypropylene (PP) based shape memory polymers. The orthotic shape memory polymer comprises one or more of the following: nylon (polyamide, PA) based shape memory polymers, furan resin based shape memory polymers, polyethylene terephthalate-1,4-cyclohexanediethanol ester (PETG) based shape memory polymers, polyimide (PI) based shape memory polymers, polyvinyl chloride (PVC) based shape memory polymers, natural rubber based shape memory polymers, phenolic resin based shape memory polymers, polyester elastomer (TPEE) based shape memory polymers, polyarylene ether nitrile (PEN) based shape memory polymers, polybutadiene based shape memory polymers, cellulose based shape memory polymers, polyurethane acrylate (PUA) based shape memory polymers, polystyrene (PS) based shape memory polymers, and polylactic acid (PLA) based shape memory polymers. External stimulation employs one or more of the following: thermal excitation, electrical excitation, optical excitation, magnetic excitation, and microwave excitation. When the orthotic brace undergoes deformation and stiffness adjustment driven by external stimulation conditions...

[0213] The specific implementation method is as follows.

[0214] A quantitative database of relationships between external stimuli and changes in the deformation, stiffness, and correction angle of orthotic braces was established. For thermal excitation, the thermal performance of shape memory polymer smart materials was tested, obtaining deformation curves and stiffness change data under different heating rates and temperature holding times. The orthotic brace was placed in a temperature-controlled device, and the temperature was controlled by a program according to the needs of different stages of the patient's rehabilitation process. Throughout the rehabilitation process, based on the patient's regular medical examination data, changes in knee varus / valgus angles, bone growth, etc., were used to adjust the external stimulation parameters in real time, dynamically changing the brace correction angle to meet the patient's constantly changing biomechanical needs during rehabilitation.

[0215] Specifically, refer to Figure 7 and Figure 8The knee orthotic brace structure based on 4D printing in this invention includes a first orthotic plate 100 and a second orthotic plate 200. A first connecting plate 101 is provided at one end of the first orthotic plate 100, and a second connecting plate 201 is provided on the second orthotic plate 200. The first connecting plate 101 and the second connecting plate 201 are connected by a clamping plate 300. A first deformable segment 102 is provided on the first connecting plate 101, and a second deformable segment 202 is provided on the second connecting plate 201.

[0216] The first deformable segment 102 and the second deformable segment 202 are printed using 4D memory material.

[0217] The first connecting plate 101 has a first gear tooth 101a at one end, and the second connecting plate 201 has a second gear tooth 201a at one end. The first gear tooth 101a and the second gear tooth 201a mesh with each other. The first connecting plate 101 has a first through hole 101b, and the second connecting plate 201 has a second through hole 201b. A first pin 101c passes through the first through hole 101b and is connected to the clamping plate 300. A second pin 201c passes through the second through hole 201b and is connected to the clamping plate 300.

[0218] The first orthotic plate 100 is provided with a first strap hole 101d, and the second orthotic plate 200 is provided with a second strap hole 201d.

[0219] It should be noted that the clamping plate 300 is provided with two pairs of pin holes for the first pin 101c and the second pin 201c to pass through and connect.

[0220] Specifically, both the first orthotic plate 100 and the second orthotic plate 200 have an arc-shaped structure and are used to fit the patient's thigh and calf respectively. They are equipped with straps for fixation. The first orthotic plate 100 and the second orthotic plate 200 are connected to each other by the first connecting plate 101 and the second connecting plate 201. It should be noted that the meshing of the first gear 101a and the second gear 201a makes it easier for the first connecting plate 101 and the second connecting plate 201 to rotate relative to each other, which facilitates the relative movement between the first connecting plate 101 and the second connecting plate 201 during walking.

[0221] In step S4,

[0222] Set the nozzle temperature in the printing parameters to within 5°C of the melting point of the selected smart polymer material to ensure that the material can melt fully without excessive degradation, thus preserving the optimal molecular chain state for subsequent shape memory properties.

[0223] Setting the substrate temperature in the printing parameters to within the range of -15℃ to -20℃ of the selected smart polymer material's glass transition temperature, and printing at this temperature slightly below the glass transition temperature, allows for reasonable regulation of the activity of the smart polymer material molecules. The intermolecular forces remain at a relatively stable level, meeting the stringent requirements of knee orthotic braces for material stability and shape accuracy.

[0224] It should be understood that numerous specific implementation decisions can be made during the development of any practical implementation, such as in any engineering or design project. Such development efforts may be complex and time-consuming, but for those skilled in the art who benefit from this disclosure, the development effort will be a routine work of design, manufacturing, and production without requiring much experimentation.

[0225] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for fabricating and driving a knee orthotic brace based on 4D printing, characterized in that: include, S1, collecting patient data, including three-dimensional surface data of the knee joint and medical imaging CT or MRI data; S2, a biomechanical model of knee joint deformity characteristics obtained by three-dimensional reconstruction of medical imaging data; S3, Finite element analysis results of the initial orthotic brace structural model; 4D printing model obtained after designing the orthotic brace structure based on the analysis results; S4 utilizes shape memory polymer smart materials to 4D print the obtained model using melt direct molding technology, and then assembles the orthopedic brace. S5 calculates the varus / valgus angle based on the knee joint biomechanical model, constructs a knee joint orthopedic target angle prediction model using a deep learning algorithm, and obtains the orthopedic brace structure model by compensating for the orthopedic angle; S6, through external stimulation conditions, drives the orthotic brace to deform and adjust its stiffness, changing the brace's correction angle to meet the patient's constantly changing biomechanical needs during the rehabilitation process; The biomechanical model for obtaining knee joint deformity characteristics includes, Import medical imaging data; Threshold segmentation preserves the target knee joint bone while removing most soft tissue and noisy data. 3D reconstruction generates skeletal models ; For skeletal models Noise removal and model integrity restoration were performed to obtain an optimized 3D skeletal model. ; A biomechanical model with knee joint deformity characteristics was established by combining the patient's known deformity feature parameters. , represented as: in: The above is the optimized skeletal geometric model; P represents the physical and mechanical properties of the bones and soft tissues. It is a set of deformity angles, joint lines, or other deformity characteristic parameters; The method of constructing a knee joint orthopedic target angle prediction model using deep learning algorithms includes: Acquire and preprocess 3D data of the patient's knee joint, and divide the dataset; Define network models and network structures; Model training and loss function construction; Cross-validate and optimize the model; Model output; Define the input layer as: Let the preprocessed data be denoted as X, where: Each training data The patient's knee joint's three-dimensional morphological features and mechanical parameters are input into the network in the form of vectors or tensors. Define the output layer as: That is, output the predicted value of the patient's target varus / valgus angle. , used for regression prediction; The loss function L is defined as follows: in: Let be the predicted angle of the model for the i-th sample; The actual post-correction angle of the i-th sample; For regularization terms, This represents all trainable parameters in the network; α is the regularization coefficient, used to balance data fitting error and model complexity.

2. The method for fabricating and driving a 4D-printed knee orthotic brace according to claim 1, characterized in that: The threshold segmentation includes setting an initial threshold range based on the difference in grayscale values ​​between bone tissue and soft tissue in CT / MRI scans. ; For image voxel locations ,like If the voxel belongs to skeletal tissue, its mask label is recorded. ;otherwise ;as follows: Based on whether the masked 3D preview image fully displays the patient's lower limb bones and joint structures, adjustments are made if there is significant soft tissue or residual noise. Scope; If there are breaks or gaps in the bone structure, the upper and lower limits of the threshold will be adjusted accordingly.

3. The method for fabricating and driving a 4D-printed knee orthotic brace according to claim 2, characterized in that: The initial orthotic brace structural model undergoes finite element analysis preprocessing, including dimensional checks and necessary adjustments. Based on the actual use of the brace and biomechanical analysis, corresponding loads and boundary conditions are applied to the outer surface of the brace to simulate the actual stress conditions of the brace under gravity, compression and correction torque. Based on the finite element analysis results, the mechanical strength and stability of the model are determined, the displacement distribution of the model is determined by the displacement cloud map, the location and magnitude of the maximum displacement are read, and the impact of the degree of deformation on the functionality is evaluated. To address issues such as stress concentration areas and areas of excessive displacement, the structure of the orthotic brace is optimized by adjusting its shape, position, and size, or by changing the overall structural layout of the brace. After the design is completed, a 4D printing model is obtained, which is used for subsequent 4D printing of knee orthotic braces.

4. The method for fabricating and driving a 4D-printed knee orthotic brace according to claim 3, characterized in that: The orthopedic brace is printed using shape memory polymer smart material.

5. The method for fabricating and driving a 4D-printed knee orthotic brace according to claim 4, characterized in that: In step S4, Set the nozzle temperature in the printing parameters to be within 5°C of the melting point of the selected smart polymer material to ensure that the material can melt fully without excessive degradation, thus preserving the optimal molecular chain state for subsequent shape memory properties. Set the substrate temperature in the printing parameters to within the range of -15℃ to -20℃ of the glass transition temperature of the selected smart polymer material. Printing at this temperature below the glass transition temperature will allow the activity of the smart polymer material molecules to be reasonably regulated.

6. The method for fabricating and driving a 4D-printed knee orthotic brace according to claim 5, characterized in that: This includes calculating the varus / valgus angle based on a knee joint biomechanical model, constructing a knee joint orthopedic target angle prediction model using a deep learning algorithm, and obtaining an orthopedic brace structure model by compensating for the orthopedic angle; during the second correction, by changing the deformity data, compensating for the orthopedic angle again to obtain the orthopedic brace structure model for the next stage, and changing the orthopedic angle again.

7. The method for fabricating and driving a 4D-printed knee orthotic brace according to claim 6, characterized in that: The process of acquiring and preprocessing the patient's three-dimensional knee joint data, and dividing the dataset, includes: Based on the obtained three-dimensional data of the patient's knee joint and the clinically known target varus / valgus angles after orthopedics, training datasets and validation datasets are formed. The patient's original knee joint deformity angle and the final clinically confirmed corrected angle were used as training labels, and the corrected angle was recorded as... ; Multimodal data is formatted in a unified manner, and relevant mechanical parameters are structured and encoded for input into the network.

8. The method for fabricating and driving a 4D-printed knee orthotic brace according to claim 7, characterized in that: By driving the deformation and stiffness adjustment of the orthotic brace through external stimulation conditions, and changing the correction angle of the brace in combination with the orthotic brace structural model, an orthotic brace with orthotic ability can be obtained; specifically, the orthotic brace is driven to change to the target angle by using one or more of thermal excitation, electrical excitation, optical excitation, magnetic excitation and microwave excitation.

9. A brace structure based on the 4D printing-based knee orthotic brace fabrication and driving method according to any one of claims 1 to 8, characterized in that: include, A first orthotic plate (100) and a second orthotic plate (200) are provided. A first connecting plate (101) is provided at one end of the first orthotic plate (100), and a second connecting plate (201) is provided on the second orthotic plate (200). The first connecting plate (101) and the second connecting plate (201) are connected by a clamping plate (300). The first connecting plate (101) is provided with a first deformable section (102), and the second connecting plate (201) is provided with a second deformable section (202). The first connecting plate (101) has a first gear tooth (101a) at one end, and the second connecting plate (201) has a second gear tooth (201a) at one end. The first gear tooth (101a) and the second gear tooth (201a) mesh with each other. The first connecting plate (101) has a first through hole (101b), and the second connecting plate (201) has a second through hole (201b). A first pin (101c) passes through the first through hole (101b) and connects to the clamping plate (300). A second pin (201c) passes through the second through hole (201b) and connects to the clamping plate (300). The first orthotic plate (100) is provided with a first strap hole (101d), and the second orthotic plate (200) is provided with a second strap hole (201d).

Images