Orthopedic implantation instrument design method based on structure-material-tissue regeneration relation

Abstract

The invention provides an orthopedic implantation instrument design method based on a structure-material-tissue regeneration relationship, which comprises the following steps: S1, acquiring anatomical structure, bone characteristics and mechanical property data of an implantation part of a patient, obtaining structure-material and tissue regeneration quantitative data in combination with an animal experiment, and constructing a data set; s2, based on the data set, establishing a quantitative mapping model among the structure, the material parameters, the biomechanical property and the tissue regeneration effect by using deep learning, and improving the model precision through historical data training and cross validation; and S3, decomposing functional requirements and design parameters by applying axiom design, generating a structure and material scheme which is matched with anatomical features and meets functional requirements through Z-shaped mapping and deep learning reverse optimization, and obtaining a final design of the personalized implant instrument after iterative verification. The method is based on a deep learning data driving model and an axiom design theory, it is ensured that design parameters achieve expected performance, and design accuracy and reliability are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of orthopedic medical device technology, specifically to a design method for orthopedic implantable devices based on the structure-material-tissue regeneration relationship. Technical Background

[0002] Orthopedic implants are widely used in clinical surgeries such as fracture repair, bone defect reconstruction, joint replacement, and orthopedics. Traditional implants typically employ standardized designs, making it difficult to meet the individualized bone tissue repair needs of different patients. This often leads to problems such as poor biomechanical fit, unsatisfactory tissue repair effects, and even complications. With the development of technologies such as digital medicine, computer-aided design, and artificial intelligence, implant design methods based on individual patient data have gradually emerged. However, these methods fail to fully consider the complex relationship between the structure and materials of the implant and bone tissue regeneration, making it difficult to accurately meet individual differences and easily resulting in poor biomechanical fit between orthopedic implants and bone tissue, and unsatisfactory tissue repair effects.

[0003] The bone tissue regeneration process is influenced by the structural features and material properties of implanted devices. Currently, the design of orthopedic implanted devices relies heavily on experience, and the design of structural parameters and material components is often carried out independently without fully considering their coupling effect on the bone tissue regeneration process. There is a lack of a quantitative model that systematically describes the relationship between "structure-material-tissue regeneration", and it is not yet possible to achieve precise, efficient and intelligent design.

[0004] Furthermore, the macro / microstructural parameters and material composition of orthopedic implants often exhibit inconsistent or even contradictory effects on their mechanical properties and tissue regeneration, requiring careful consideration during the design process. Moreover, the regulatory mechanisms governing the mechanical properties and tissue regeneration of implants based on their macro / microstructural parameters and material composition are complex and often difficult to express directly using explicit functions, significantly increasing the complexity of multi-objective design optimization. This results in existing design methods being cumbersome, lacking flexibility, and inefficient. Summary of the Invention

[0005] This invention addresses the problem that existing orthopedic implant design methods lack a systematic quantitative model describing the "structure-material-tissue regeneration" relationship, which easily leads to poor biomechanical compatibility between orthopedic implants and bone tissue, unsatisfactory tissue repair effects, and technical problems such as cumbersome design process, insufficient flexibility, and low efficiency. Therefore, this invention provides an orthopedic implant design method based on the structure-material-tissue regeneration relationship.

[0006] The technical solution adopted by the present invention to solve the above-mentioned technical problems is as follows:

[0007] A design method for orthopedic implantable devices based on the structure-material-tissue regeneration relationship includes the following steps:

[0008] S1: Using medical imaging equipment, acquire anatomical structure data and bone characteristic data of the implantation site in the patient, establish a mechanical property model of bone tissue to obtain mechanical property data; combine with animal experiments to obtain quantitative results data on the device structure-material combination and tissue regeneration effect; the mechanical property data and quantitative results data together constitute a dataset;

[0009] S2: Based on the dataset obtained in step S1, a quantitative mapping model between implantable device structural parameters, material synthesis parameters, biomechanical properties, and tissue regeneration effects is constructed using deep learning algorithms. Historical clinical cases and experimental data are used for training and validation of the model.

[0010] S3: Intelligent design of implantable devices using axiomatic design theory, specifically including:

[0011] S3.1: Define the functional requirements of the implantable device to form a functional domain, and use the corresponding structural and material design as top-level design variables to form a physical domain; at the same time, the macroscopic shape of the implantable device must match the anatomical structure data of the patient's site to be repaired obtained in step S1 as a design constraint.

[0012] S3.2: Perform a Z-shaped mapping transformation between the functional domain and the physical domain to decompose the functional requirements and corresponding design variables layer by layer. Express the mapping relationship between the underlying functional indicators and design parameters through the underlying design matrix. Use a deep learning network for intelligent reverse design. Through objective function optimization, quickly generate structural and material design parameters in the physical domain that can realize the corresponding underlying functional requirements to form a preliminary design scheme.

[0013] S3.3: Verify whether the preliminary design scheme can achieve the predetermined functional requirements and meet the design constraints. Based on the verification results, repeatedly optimize the design parameters to achieve the final implantable device design scheme that meets the individual needs of patients.

[0014] In step S2, cross-validation is used to improve the model's generalization ability and accuracy.

[0015] In step S1, CT or MRI scans are used to obtain three-dimensional image data of the implantation site, a three-dimensional geometric model is established, and anatomical structure data is extracted.

[0016] The bone characteristic data in step S1 includes bone density data.

[0017] In step S2, the specific method for improving the generalization ability and accuracy of the model through cross-validation is as follows: during the training process, cross-validation and training / validation set splitting are used to evaluate the model's prediction accuracy. The loss function adopts the mean squared error between the predicted performance and the target performance to ensure that the prediction error of the output on the biomechanical and osseointegration performance of the implanted device is below the acceptable range. By improving the generalization ability and accuracy of the model through cross-validation, the model can quickly predict the biomechanical and osseointegration performance of the new design scheme, providing model preparation for subsequent reverse engineering.

[0018] Ensure that the prediction error of the output for the biomechanical and osseointegration performance of the implanted device is less than 5%.

[0019] In step S3.2, a design matrix is ​​constructed to quantitatively express the mapping relationship between the same functional domain and the physical domain within the functional domain. Based on the quantitative mapping model in step S2, the elements in the design matrix are quantified. Based on the independence principle of axiomatic design, the sensitivity of each design variable to each functional index is calculated, and the functional domain is decoupled to achieve a one-to-one correspondence or weak coupling between functional index and design variable. The above steps are repeated until the functional domain and physical domain are decomposed to the point that they cannot be further subdivided, so as to realize the mapping relationship between the underlying functional index and design parameter expressed by the underlying design matrix.

[0020] In S3.3, numerical simulation, biomechanical experiments, materials science experiments and biocompatibility tests are used to verify the preliminary design scheme generated in S3.2, to verify whether it can achieve the predetermined functional requirements and meet the design constraints. Based on the verification results, the design parameters are repeatedly optimized to achieve the final design scheme that meets the individual needs of patients.

[0021] The orthopedic implant design method based on the structure-material-tissue regeneration relationship in this invention has the following advantages:

[0022] (1) The orthopedic implant design method based on the structure-material-tissue regeneration relationship in this invention combines deep learning algorithms with axiomatic design theory to construct an efficient intelligent reverse design process. Through deep learning, a mapping model between implant structure parameters and performance indicators is established, enabling the ability to deduce design parameters from target functional requirements. This invention introduces axiomatic design principles into the design process, ensuring that the generated schemes follow strict design axioms and scientific criteria, guaranteeing that various functional requirements are independent and non-conflicting, reducing iterative processes of repeated trial and error, and efficiently achieving multi-objective trade-off design. Through this intelligent reverse design process, this invention can automatically generate implant design schemes that meet the requirements simply by providing the desired mechanical properties or tissue regeneration goals, significantly improving design efficiency.

[0023] (2) The orthopedic implant device design method based on the structure-material-tissue regeneration relationship in this invention is based on the anatomical features and bone condition of the specific implantation site of the patient, the mechanical environment of the actual service process and the tissue repair function requirements. This allows the structural features and mechanical properties of the implant device to be customized for different patients, greatly improving the fit and functional adaptability of the implant device to the surrounding tissues.

[0024] (3) The orthopedic implant design method based on the structure-material-tissue regeneration relationship in this invention involves cross-scale collaborative design of the macroscopic and microscopic structural parameters of the implant. Macroscopic structures, such as the overall shape, size, and internal support layout of the implant, determine the initial mechanical stability of the implant and its compatibility with the anatomical features of the patient's implantation site; microscopic structures, such as pore size, porosity, and surface morphology, affect regeneration processes such as cell adhesion, growth, and vascularization of new tissue. By carrying out cross-scale collaborative design of macroscopic and microscopic structural parameters, this invention enables the implant to possess both sufficient mechanical strength and stability, while maximizing cell proliferation and tissue regeneration.

[0025] (4) The orthopedic implant design method based on the structure-material-tissue regeneration relationship in this invention significantly improves the scientific rigor, accuracy, and clinical adaptability of the design compared to traditional implant design methods. The design method proposed in this invention is guided by a data-driven model based on deep learning and axiomatic design theory, providing a solid scientific basis for design decisions. It no longer relies solely on the experience and judgment of engineers, ensuring that design parameters achieve the expected performance and significantly improving the accuracy and reliability of the design. Furthermore, this invention supports personalized customization and cross-scale collaborative design of macro and micro structures, adapting to different patients' bone defect conditions and physiological environments, providing effective support for precise and personalized clinical treatment.

[0026] To make the technical solution of the orthopedic implant device design method based on the structure-material-tissue regeneration relationship of the present invention clearer, the present invention will be further described below in conjunction with specific embodiments. Detailed Implementation

[0027] Example 1

[0028] This embodiment uses an interbody fusion device as an example to describe in detail the specific implementation process of the orthopedic implant device design method based on the structure-material-tissue regeneration relationship described in this invention, which includes the following steps:

[0029] S1: The three-dimensional anatomical structure of the patient's spinal lesion segment is obtained through CT scan. The shape of the vertebral body and intervertebral space is extracted and three-dimensional reconstruction is performed using image segmentation technology to obtain anatomical structure data. At the same time, the mechanical performance model of the patient's vertebral tissue is established by mapping the CT gray value to bone density data and obtaining mechanical performance data, including the bone elastic modulus and bone compressive strength of the vertebral tissue.

[0030] Mapping bone mineral density data using CT grayscale values ​​involves converting the grayscale values ​​(HU) of CT images into apparent bone mineral density (BMD, g / cm³). Then, based on the "bone mineral density-mechanical property" relationship, the mechanical property parameters of bone tissue are predicted. The specific steps are as follows:

[0031] (1) CT image acquisition and segmentation:

[0032] Quantitative CT (QCT) is used to scan the target area. To ensure accurate density calibration, a density calibration phantom containing a known concentration of hydroxyapatite or potassium phosphate solution is scanned simultaneously during the scan. This allows for precise segmentation of bone tissue from the surrounding soft tissue in the CT images.

[0033] (2) Conversion between grayscale value HU and apparent bone mineral density:

[0034] Using data from the calibration phantom, a correlation was established between CT grayscale value HU and equivalent hydroxyapatite (mgHA·cm⁻¹). -3 A linear relationship between them.

[0035] ρ HA =a·HU+b

[0036] Where ρ HA Hydroxyapatite equivalent (mgHA·cm) -3 The slope a and intercept b are determined by the calibration phantom.

[0037] (3) Density system conversion: convert ρ HA Converted to engineering apparent density ρ app or ash density ρ ash This facilitates integration with empirical formulas for the mechanical properties of bone tissue.

[0038] ρ app =c1ρ HA +c0

[0039] ρ ash =0.6ρ app

[0040] The empirical coefficient c1 ranges from 0.001 to 0.0015 (g·cm³). -3 / mgHA·cm -3 c0 takes values ​​between 0 and 0.05.

[0041] (4) Bone mineral density-mechanical property conversion:

[0042] Elastic modulus of cancellous bone (MPa)

[0043] Where α ranges from 1000 to 3700, and β ranges from 1.8 to 2.2.

[0044] Compressive strength of cancellous bone (MPa)

[0045] Where k takes values ​​from 5 to 120, and n takes values ​​from 1.6 to 2.0.

[0046] Cortical bone elastic modulus (MPa).

[0047] A spinal fusion model was obtained through animal experiments on rabbits or sheep. Quantitative data on the osseointegration effect of different fusion device "structure-material" combinations were acquired. The quantitative results of the osseointegration effect included changes in trabecular thickness, bone volume fraction, trabecular number, trabecular separation, and bone density. The collected experimental data underwent denoising and standardization processing. The mechanical property data and the quantitative data on osseointegration effect together constituted the dataset.

[0048] S2: Quantitative Modeling of the "Structure-Material-Tissue Regeneration" Relationship: Based on the aforementioned dataset, using the macro / microscopic structural design parameters and material synthesis parameters of the interbody fusion device as input features, and the mechanical performance data and quantitative index data of osseointegration effect from the aforementioned dataset as output variables, a quantitative mapping model is constructed between the macro / microscopic structure of the interbody fusion device, material synthesis parameters, and biomechanical and osseointegration performance. The training process uses cross-validation and training / validation set splitting to evaluate the model's prediction accuracy. The loss function uses the mean squared error between the predicted performance and the target performance to ensure that the prediction error of the output for the biomechanical and osseointegration performance of the interbody fusion device is below an acceptable range, preferably with a target parameter error of <5%. Cross-validation is used to improve the model's generalization ability and accuracy, ensuring that the model can quickly predict the biomechanical and osseointegration performance of new design schemes, providing model preparation for subsequent reverse engineering.

[0049] S3: Intelligent design of implantable devices is completed by combining axiomatic design theory, specifically including the following steps:

[0050] S3.1: Define the functional requirements of the implantable device to form a functional domain, and use the corresponding structural and material designs as top-level design variables to form a physical domain; simultaneously, ensure that the macroscopic shape of the implantable device matches the anatomical features of the patient's site to be repaired, as a design constraint; the specific method is as follows:

[0051] a. For the intelligent design of intervertebral fusion devices, it is first necessary to clarify the overall goal of implanting an intervertebral fusion device into the degenerated segment during spinal fusion surgery: to restore the normal intervertebral height and normal physiological curvature of the spine in the degenerated segment, and to provide stable mechanical support to promote bony fusion between adjacent vertebrae. Based on the overall goal, the top-level functional requirements FR0 of the intervertebral fusion device are formed: to restore the normal intervertebral height and normal physiological curvature of the spine in the degenerated segment, and to provide stable mechanical support to promote bony fusion between adjacent vertebrae, thus forming a functional domain. At the same time, the macroscopic shape of the intervertebral fusion device needs to match the anatomical structure data of the patient's spinal lesion segment obtained in step S1, serving as a design constraint and forming a design space.

[0052] b. Establishment of design variable DP: Based on the established FR0, the top-level design variable DP0 corresponding to the realization of FR0 is: the overall design of the interbody fusion device, forming the physical domain DPs.

[0053] S3.2: A Z-shaped mapping transformation is performed between the functional domain and the physical domain to decompose functional requirements and corresponding design variables layer by layer. The mapping relationship between underlying functional indicators and design parameters is expressed through the underlying design matrix. Intelligent reverse design is performed using a deep learning network. Through objective function optimization, structural and material design parameters that can realize the corresponding underlying functional requirements are quickly generated in the physical domain, forming a preliminary design scheme. Specifically, this includes:

[0054] After determining the FR0 and DP0 of the interbody fusion cage, the functional requirements FRs of the interbody fusion cage are decomposed within the functional domain. Each FR is analyzed... i The design variables of the interbody fusion device that may have an impact are decomposed into DP0 in the physical domain to form corresponding control parameters for each FR. i Design Variables DP i This process will create a Z-shaped mapping between the functional domain and the physical domain.

[0055] For example, the FR0 of an interbody fusion cage can be specifically broken down as follows:

[0056] FR 11 Geometric fit and structural support performance. The fusion cage shape must conform to the patient's vertebral endplate morphology, bear the physiological loads of the spine, maintain structural stability under axial compression and torsional loads, maintain intervertebral height, and prevent post-implantation displacement.

[0057] FR 11 It can be further broken down into postoperative segmental flexion-extension range of motion (FR). 111 lateral bending mobility FR 112 , Axial rotational mobility FR 113 Postoperative intervertebral height change rate (FR) 114 Postoperative recovery degree of spinal physiological curvature FR115 Specific quantitative indicators, etc.

[0058] FR 12 Mechanical performance matching. The mechanical properties of the interbody fusion device should be well matched with the surrounding bone tissue to avoid stress shielding. If it is a biodegradable interbody fusion device, its degradation rate needs to match the bone regeneration rate so that it can be gradually replaced after new bone formation, while avoiding premature mechanical failure due to excessively rapid degradation.

[0059] FR 12 This can be further decomposed into the difference between the overall stiffness of the fusion cage and the Young's modulus of the surrounding bone tissue (FR). 121 , Fusion yield strength FR 122 FR of the fusion unit 123 For biodegradable fusion devices, the degradation rate FR is also included. 122 Rate of decrease in mechanical strength, degradation uniformity (FR) 124 Specific quantitative indicators, etc.

[0060] FR 13 Stress transfer performance. That is, the fusion device should reasonably transmit the physiological load of the spine, avoid stress concentration and fracture risk in the surgical segment, and reduce local pressure on the endplate.

[0061] FR 13 It can be further decomposed into the rate of change of intradiscal pressure (FR) between adjacent segments. 131 FR (Face Joint Cartilage Stress Distribution Change Rate) 132 FR uniformity of stress distribution in the final plate 133 Specific quantitative indicators, etc.

[0062] FR 14 Osteofusion performance. The surface and internal structure of the fusion device should promote osteoblast adhesion and bone ingrowth to achieve osteofusion between vertebral bodies.

[0063] FR 14 It can be further decomposed into bone volume fraction FR 141 Bone mineral density FR 142 FR (Bone Trabecular Thickness) 143 FR (Bone Trabecular Separation) 144 Specific quantitative indicators, etc.

[0064] FR 15 Biocompatibility. The fusion cage material and its potential degradation products are non-toxic, do not trigger immune rejection, and ensure postoperative imaging visualization, facilitating monitoring of the intervertebral bone integration process.

[0065] FR 15 It can be further decomposed into material cell compatibility FR 151 FR material blood compatibility 152For fusion devices that are prone to degradation, the cellular compatibility of the degradation products should also be considered. 153 Degradation products blood compatibility FR 154 Specific indicators, etc.

[0066] Accordingly, the DP0 of the interbody fusion device can be specifically broken down as follows:

[0067] DP 11 Macro-structural design

[0068] DP 11 It can be further decomposed into the macroscopic shape DP of the fusion unit. 111 Geometric dimensions DP 112 and tilt angle DP 113 This determines its contact pattern and stability with the vertebral endplate.

[0069] DP 12 Microstructure design

[0070] DP 12 It can be further decomposed into a porous unit topology DP 121 DP aperture 122 , rib width DP 123 DP wall thickness 124 Porosity DP 125 Inter-unit connection method DP 126 These factors affect the macroscopic mechanical properties, permeability, and osseointegration performance of the fusion device.

[0071] DP 13 Material synthesis design

[0072] DP 13 It can be further decomposed into material components DP 131 Group allocation ratio DP 132 Coating component DP 133 Surface roughness DP 134 These factors affect the macroscopic mechanical properties, biocompatibility, and osseointegration performance of the fusion device.

[0073] Design Matrix Construction and Decoupling: A design matrix is ​​constructed to quantitatively express the mapping relationship between functional variables (FRs) and design variables (DPs) at the same level within a functional domain. The elements in the design matrix are quantified based on the quantitative mapping model obtained in step S2. According to axiomatic design theory, during the decomposition and Z-shaped mapping of FRs and design variables (DPs), each FR in the functional domain should be controlled by only one DP at the same level in the physical domain to satisfy the principle of independence. By calculating the sensitivity of each DP to each FR, the coupling degree between DPs is evaluated. Based on this, the design matrix is ​​optimized to tend towards diagonalization, thereby decoupling the DPs and achieving a one-to-one correspondence or weak coupling between FRs and DPs. The above steps are repeated to decompose FRs and DPs until they cannot be further subdivided, obtaining the bottom-level FRs and DPs. The mapping relationship between the bottom-level FRs and DPs is expressed through the bottom-level design matrix.

[0074] Deep Learning Reverse Design: Based on the established underlying FRs-DPs mapping relationship, a deep neural network is used for reverse design. According to the three-dimensional anatomical features and bone condition of the patient's spinal lesion segment, specific target values ​​for each underlying FR are determined. In this embodiment, the specific target value for each underlying FR is that the mechanical properties of the interbody fusion cage should match the bone quality of the lesion segment; specifically, the deviation range between the bone elastic modulus and bone compressive strength of the vertebral tissue and the corresponding performance indicators of the lesion segment should be ≤10%. These target values ​​are used as input to the deep learning network. Through the combination of the deep learning neural network and advanced optimization algorithms, the specific values ​​of each underlying DP of the fusion cage are generated and output within the design space according to the given underlying FR targets, forming a preliminary design scheme. The deep learning neural network used in this embodiment includes, but is not limited to, convolutional neural networks, conditional generative networks, and backpropagation neural networks. The advanced optimization algorithms include, but are not limited to, topology optimization and genetic algorithms.

[0075] S3.3: Design Verification and Optimization. The preliminary design scheme generated in step S3.2 is verified by comprehensively applying numerical simulation, biomechanical experiments, materials science experiments, and biocompatibility tests to verify whether it can achieve the predetermined functional requirements and meet the design constraints. If the preliminary design scheme cannot achieve the predetermined functional requirements or meet the design constraints, the design parameters are further optimized based on the verification results until the verification is passed, so as to achieve the final interbody fusion device design scheme that meets the individual needs of patients.

[0076] Further optimization of the design parameters can be achieved through repeated deep learning reverse engineering, adjusting the key adjustable structures and hyperparameters of the neural network during the optimization process. The key adjustable structures and hyperparameters of the neural network include at least one of the following: depth (number of layers in the network), width (number of neurons per layer), activation function type, loss function type, network weights, learning rate (the step size controlling weight updates), and training cycle (the total number of times the entire training dataset passes through the network). By adjusting these key adjustable structures and hyperparameters, an optimized design scheme can be formed.

[0077] In addition to the aforementioned methods of deep learning reverse design, further optimization of design parameters can employ one or more of the following methods: macroscopic structural matching optimization, topology optimization, porous structure optimization, and multi-scale optimization. Macroscopic structural matching optimization, based on patient imaging data, improves the fit with the patient's anatomy by optimizing the implant's shape and macroscopic dimensions, aiming to reduce additional cutting during surgery to adapt to bone and improve initial stability. Topology optimization, under given design space, load conditions, and constraints (such as maximum stress, target weight, or stiffness), uses finite element analysis iterative calculations to remove material from low-stress areas and retain material along high-stress paths, aiming to achieve lightweighting while ensuring mechanical strength and precisely controlling stiffness to obtain the optimal structure. Porous structure optimization balances mechanical strength and bone ingrowth space, optimizing porosity, pore size, and basic unit shape, aiming to obtain the optimal porous structure that simultaneously meets the requirements for mechanical strength and bone ingrowth promotion performance. Multi-scale optimization, while performing macroscopic topology optimization, allocates optimal microscopic material to each point in the structure, aiming to generate a smooth transition structure from macroscopic to microscopic.

[0078] Example 2

[0079] This embodiment uses a bone repair scaffold for large long bone defects or nonunion as an example to illustrate the specific implementation process of the intelligent design method for orthopedic implants based on the "structure-material-tissue regeneration" relationship described in this invention. The specific steps are as follows:

[0080] S1: Data Acquisition and Preprocessing: Three-dimensional image data of the patient's bone defect area are acquired through CT or MRI scans. The defect shape and microscopic parameters of the surrounding bone tissue are extracted. The image data are segmented and registered to establish a three-dimensional geometric model of the defect area and adjacent normal bone tissue. Key anatomical parameters such as defect length and bone shaft diameter are extracted to obtain anatomical structure data. The bone density distribution of the residual bone is assessed using CT grayscale mapping. A mechanical property model of the bone tissue surrounding the patient's bone defect area is established to obtain mechanical property data, including the bone elastic modulus and bone compressive strength. The CT grayscale values ​​are mapped to the bone density data to convert the grayscale values ​​HU of the CT images into apparent bone density (BMD, g / cm³). Then, the mechanical property parameters of the bone tissue are predicted based on the "bone density-mechanical property" relationship. The specific steps are the same as in Example 1.

[0081] Quantitative indicators of the bone regeneration effect of different bone repair scaffold "structure-material" combinations were obtained by collecting data from tissue engineering experiments and animal experiments. These indicators specifically included changes in trabecular bone thickness, bone volume fraction, trabecular bone number, trabecular bone separation, and bone density. The collected experimental data underwent noise reduction and standardization. The mechanical property data and the quantitative indicators of bone integration effect together constituted the dataset.

[0082] S2: Quantitative Modeling of the "Structure-Material-Tissue Regeneration" Relationship: Based on extensive clinical data, animal experiments, and in vitro simulation data, a dataset is constructed using macro / microscopic structural design parameters and material synthesis parameters of the bone repair scaffold as input features, and the mechanical performance data and quantitative indicators of bone integration effect as output variables. Multi-scale machine learning methods are used for supervised learning of the input and output to construct a quantitative mapping model between the "macro / microscopic structure of the bone repair scaffold, material synthesis parameters, and the efficiency of new bone tissue regeneration and mechanical performance reconstruction." The training process uses cross-validation and training / validation set splitting to evaluate the model's prediction accuracy. The loss function uses the mean squared error between the predicted performance and the target performance to ensure that the prediction error of the output on the efficiency of new bone tissue regeneration and mechanical performance reconstruction is below an acceptable range, preferably with a target parameter error of <5%. Cross-validation is used to improve the model's generalization ability and accuracy, ensuring that the model can quickly predict the efficiency of new bone tissue regeneration and mechanical performance reconstruction in new design schemes, providing model preparation for subsequent reverse engineering.

[0083] S3: Intelligent design of implantable devices is completed by combining axiomatic design theory, specifically including the following steps:

[0084] S3.1: Define the functional requirements of the implantable device to form a functional domain, and use the corresponding structural and material designs as top-level design variables to form a physical domain; simultaneously, ensure that the macroscopic shape of the implantable device matches the anatomical features of the patient's site to be repaired, as a design constraint; the specific method is as follows:

[0085] a. For the intelligent design of bone repair scaffolds, it is first necessary to clarify the overall goal of implanting a bone repair scaffold at the bone defect site during bone repair surgery: to induce new bone regeneration at the defect site to achieve bone end bridging and realize the dual reconstruction of morphology and mechanical function of the bone tissue at the defect site. Based on the overall goal, the top-level functional requirements FR0 of the bone repair scaffold are formed: to induce new bone regeneration at the defect site to achieve bone end bridging and realize the dual reconstruction of morphology and mechanical function of the bone tissue at the defect site, thus forming a functional domain. At the same time, the macroscopic shape of the bone repair scaffold needs to match the anatomical structure data of the patient's bone defect site obtained in S1, serving as a design constraint and forming a design space.

[0086] b. Based on the established FR0, the top-level design variable DP0 corresponding to the realization of FR0 is: the overall design of the bone repair scaffold, forming the physical domain DPs.

[0087] S3.2: A Z-shaped mapping transformation is performed between the functional domain and the physical domain to decompose functional requirements and corresponding design variables layer by layer. The mapping relationship between underlying functional indicators and design parameters is expressed through the underlying design matrix. Intelligent reverse design is performed using a deep learning network. Through objective function optimization, structural and material design parameters that can realize the corresponding underlying functional requirements are quickly generated in the physical domain, forming a preliminary design scheme. Specifically, this includes:

[0088] After determining the FR0 and DP0 of the bone repair scaffold, the functional requirements FRs of the bone repair scaffold are decomposed within the functional domain. Each FR is analyzed... i The design variables of the bone repair scaffold that may have an impact are decomposed into DP0 in the physical domain to form corresponding controls for each FR. i Design Variables DP i This process will create a Z-shaped mapping between the functional domain and the physical domain.

[0089] For example, the FR0 of a bone repair scaffold can be specifically broken down as follows:

[0090] FR 11 Geometric fit and structural support performance. This means the shape of the bone repair scaffold must conform to the anatomical shape of the patient's bone defect, temporarily replacing the mechanical support function of the defect, bearing physiological loads, and creating a stable mechanical environment for the growth and functional reconstruction of new bone tissue.

[0091] FR 11 It can be further decomposed into the yield strength FR of the stent. 111 Fatigue limit FR of the stent 112 FR (Fitness) between the scaffold and the defective tissue 113 Specific quantitative indicators, etc.

[0092] FR12 Mechanical property compatibility. The mechanical properties of the bone repair scaffold should be well-matched with the surrounding bone tissue to avoid stress shielding. If it is a biodegradable scaffold, its degradation rate must match the bone regeneration rate so that it can be gradually replaced after new bone formation, while avoiding premature mechanical failure due to excessively rapid degradation.

[0093] FR 12 This can be further decomposed into the difference between the overall stiffness of the bone repair scaffold and the Young's modulus of the surrounding bone tissue (FR). 121 The biodegradable scaffold also includes a degradation rate FR 122 The rate of decrease in mechanical strength FR 123 Degradation uniformity FR 124 Specific quantitative indicators, etc.

[0094] FR 13 : Induced regeneration performance. Bone repair scaffolds should effectively promote the adhesion, proliferation, and differentiation of cells involved in osteogenic processes, thereby improving the efficiency of bone tissue regeneration and functional reconstruction.

[0095] FR 13 It can be further decomposed into stent-permeable FR 131 FR of the support 132 vascularization efficiency FR 133 FR (Frequency Rate of New Bone Tissue) 134 FR (Reconstruction Rate of Mechanical Properties of New Bone Tissue) 135 Specific quantitative indicators, etc.

[0096] FR 14 Biocompatibility. The bone repair scaffold material and its degradation products (if any) are non-toxic, do not trigger immune rejection, and ensure postoperative imaging visualization to facilitate monitoring of the bone regeneration process.

[0097] FR 14 It can be further decomposed into material cell compatibility FR 141 FR material blood compatibility 142 Degradable scaffolds should also include cell compatibility features for degradation products. 143 Degradation products blood compatibility FR 144 Specific indicators, etc.

[0098] Accordingly, the DP0 of the bone repair scaffold can be specifically broken down as follows:

[0099] DP 11 Macro-structural design

[0100] DP 11 It can be further broken down into the macroscopic shape DP of the stent. 111 Geometric dimensions DP 112 and tilt angle DP 113This determines its contact with and mechanical stability of the bone tissue surrounding the defect area.

[0101] DP 12 Microstructure design

[0102] DP 12 It can be further decomposed into a porous unit topology DP 121 DP aperture 122 , rib width DP 123 DP wall thickness 124 Porosity DP 125 DP (Aperture Variation Gradient) 126 These factors affect the macroscopic mechanical properties and induced regeneration performance of the stent.

[0103] DP 13 Material synthesis design

[0104] DP 13 It can be further decomposed into material components DP 131 Group allocation ratio DP 132 Coating component DP 133 Surface roughness DP 134 These factors affect the macroscopic mechanical properties, biocompatibility, and induced regeneration performance of the stent.

[0105] Design Matrix Construction and Decoupling: A design matrix is ​​constructed to quantitatively express the mapping relationship between functional variables (FRs) and design variables (DPs) at the same level within a functional domain. The elements in the design matrix are quantified based on the quantitative mapping model obtained in step S2. According to axiomatic design theory, during the decomposition and Z-shaped mapping of FRs and design variables (DPs), each FR in the functional domain should be controlled by only one DP at the same level in the physical domain to satisfy the principle of independence. By calculating the sensitivity of each DP to each FR, the coupling degree between DPs is evaluated. Based on this, the design matrix is ​​optimized to tend towards diagonalization, thereby decoupling the DPs and achieving a one-to-one correspondence or weak coupling between FRs and DPs. The above steps are repeated to decompose FRs and DPs until they cannot be further subdivided, obtaining the bottom-level FRs and DPs. The mapping relationship between the bottom-level FRs and DPs is expressed through the bottom-level design matrix.

[0106] Deep learning reverse design: Based on the established underlying FRs-DPs mapping relationship, a deep neural network is used for reverse design. According to the three-dimensional anatomical features and bone quality of the patient's bone defect area, specific target values ​​for each underlying FR are determined. In this embodiment, the specific target value for each underlying FR is that the mechanical properties of the bone repair scaffold should match the bone quality of the diseased segment. Specifically, the deviation range between the bone elastic modulus and bone compressive strength of the vertebral tissue and the corresponding performance indicators of the diseased segment should be ≤10%.

[0107] These target values ​​are used as input to a deep learning network. By combining deep learning neural networks (including but not limited to convolutional neural networks, conditional generative networks, backpropagation neural networks, etc.) with advanced optimization algorithms (including but not limited to topology optimization, genetic algorithms, etc.), the specific values ​​of various underlying DPs of the bone repair scaffold are generated and output in reverse within the design space according to the given underlying FRs target.

[0108] S3.3: Verify the preliminary design scheme generated in step S3.2 by comprehensively applying numerical simulation, biomechanical experiments, materials science experiments and biocompatibility tests to verify whether it can achieve the predetermined functional requirements and meet the design constraints. If the preliminary design scheme cannot achieve the predetermined functional requirements or meet the design constraints, further optimize the design parameters based on the verification results until the verification is passed, and finally achieve the bone repair scaffold design scheme that meets the patient's personalized needs.

[0109] Further optimization of the design parameters can be achieved through repeated deep learning reverse engineering, adjusting the key adjustable structures and hyperparameters of the neural network during the optimization process. The key adjustable structures and hyperparameters of the neural network include at least one of the following: depth (number of layers in the network), width (number of neurons per layer), activation function type, loss function type, network weights, learning rate (the step size controlling weight updates), and training cycle (the total number of times the entire training dataset passes through the network). By adjusting these key adjustable structures and hyperparameters, an optimized design scheme can be formed.

[0110] In addition to the above-mentioned methods of repeated deep learning reverse design, one or more of the following methods can be used to further optimize the design parameters: macroscopic structure matching degree optimization method, topology optimization method, porous structure optimization method, and multi-scale optimization method.

[0111] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the claims.

Claims

1. A design method for orthopedic implantable devices based on the structure-material-tissue regeneration relationship, characterized in that, Includes the following steps: S1: Obtain anatomical structure data and bone characteristic data of the implantation site in the patient, establish a mechanical property model of bone tissue to obtain mechanical property data; combine animal experiments to obtain quantitative results data on the device structure-material combination and tissue regeneration effect; The mechanical performance data and the quantification results data together constitute the dataset; S2: Based on the dataset, a quantitative mapping model between implantable device structural parameters, material synthesis parameters, biomechanical properties, and tissue regeneration effects is constructed using deep learning algorithms. Historical clinical cases and experimental data are used for model training and validation. S3: Intelligent design of implantable devices using axiomatic design theory, specifically including: S3.1: Define the functional requirements of implantable devices to form a functional domain, and use the corresponding structural and material designs as top-level design variables to form a physical domain; The macroscopic shape of the implantable device must match the anatomical structure data of the patient's implantation site, which serves as a design constraint. S3.2: Perform a Z-shaped mapping transformation between the functional domain and the physical domain to decompose the functional requirements and corresponding design variables layer by layer. Express the mapping relationship between the underlying functional indicators and design parameters through the underlying design matrix. Use a deep learning network for intelligent reverse design. Through objective function optimization, generate structural and material design parameters in the physical domain that can realize the corresponding underlying functional requirements to form a preliminary design scheme. S3.3: Verify whether the preliminary design scheme can achieve the predetermined functional requirements and meet the design constraints. If it cannot achieve the predetermined functional requirements or meet the design constraints, further optimize the design parameters based on the verification results until the verification is passed, so as to achieve an implantable device design scheme that meets the individual needs of patients.

2. The orthopedic implant design method according to claim 1 or 2, characterized in that, In step S1, CT or MRI scans are used to obtain three-dimensional image data of the implantation site, a three-dimensional geometric model is established, and anatomical structure data is extracted.

3. The orthopedic implant design method according to claim 3, characterized in that, The bone characteristic data in step S1 includes bone density data.

4. The orthopedic implant design method according to claim 3, characterized in that, In step S2, cross-validation is used to improve the model's generalization ability and accuracy.

5. The orthopedic implant design method according to claim 4, characterized in that, In step S2, the specific method for improving the generalization ability and accuracy of the model through cross-validation is as follows: during the training process, cross-validation and training / validation set splitting are used to evaluate the model's prediction accuracy. The loss function adopts the mean squared error between the predicted performance and the target performance to ensure that the prediction error of the output on the biomechanical and osseointegration performance of the implanted device is below the acceptable range. By improving the generalization ability and accuracy of the model through cross-validation, the model can quickly predict the biomechanical and osseointegration performance of the new design scheme, providing model preparation for subsequent reverse engineering.

6. The orthopedic implant design method according to claim 5, characterized in that, Ensure that the prediction error of the output for the biomechanical and osseointegration performance of the implanted device is less than 5%.

7. The orthopedic implant design method according to claim 6, characterized in that, In step S3.2, a design matrix is ​​constructed to quantitatively express the mapping relationship between the same functional domain and the physical domain within the functional domain. Based on the quantitative mapping model in step S2, the elements in the design matrix are quantified. Based on the independence principle of axiomatic design, the sensitivity of each design variable to each functional index is calculated, and the functional domain is decoupled to achieve a one-to-one correspondence or weak coupling between functional index and design variable. The above steps are repeated until the functional domain and physical domain are decomposed to the point that they cannot be further subdivided, so as to realize the mapping relationship between the underlying functional index and design parameter expressed by the underlying design matrix.

8. The orthopedic implant design method according to claim 7, characterized in that, In S3.3, numerical simulation, biomechanical experiments, materials science experiments and biocompatibility tests are used to verify the preliminary design scheme generated in S3.2, to verify whether it can achieve the predetermined functional requirements and meet the design constraints. Based on the verification results, the design parameters are repeatedly optimized to achieve the final design scheme that meets the individual needs of patients.