Intelligent generation method of dental implant microstructure based on generative adversarial network

Abstract

The present invention discloses a method for intelligently generating microstructure of dental implants based on generative adversarial networks, comprising the following steps: S1, extracting anatomical features and mechanical requirement data; S2, constructing a double-layer generative adversarial network model; S3, designing multiple macro generators and macro discriminators in a macro generation layer, wherein the macro generator generates the overall geometric structure of the dental implant according to the anatomical feature data of the patient, and multiple macro discriminators evaluate and adjust the generated overall geometric structure; S4, designing multiple micro generators and micro discriminators in a micro generation layer, wherein the micro generator generates the microstructure of the dental implant according to the overall geometric structure output by the macro generation layer; S5, introducing a non-cooperative game between the macro generator and the micro generator; S6, introducing a cooperative game between the macro discriminator and the micro discriminator; S7, repeating S3-S6. The present invention realizes multi-objective optimization of the microstructure of the implant, and significantly improves the long-term stability and clinical success rate of the implant.

Description

Technical Field

[0001] The present invention relates to the technical field of dental implants, and in particular to a method for intelligently generating a dental implant microstructure based on a generative adversarial network. Background Art

[0002] In the existing technology, the design and manufacture of dental implants mainly rely on standardized templates, and the implant geometry suitable for most patients is generated by combining traditional computer-aided design technology and manual design. Although the existing methods can provide suitable solutions for some patients, due to the large differences in the alveolar bone anatomical structure and mechanical requirements of each patient, the existing technology is difficult to provide highly personalized implant design for each individual, which leads to poor adaptability between the implant and the alveolar bone, especially at the microstructural level, which may cause poor fit between the implant and the bone tissue, affecting the bone integration effect, and then affecting the long-term stability of the implant.

[0003] In addition, existing technologies often find it difficult to balance mechanical properties and biocompatibility when designing implant microstructures. The design of the microstructure is crucial to the mechanical strength, fatigue resistance and integration of the implant with surrounding bone tissue. However, traditional design methods usually focus on the optimization of macro-geometric structures, and it is difficult to perform fine design and adjustment at the microstructural level, which may cause the implant to face problems of insufficient mechanical strength or poor biointegration during long-term use, thereby increasing the incidence of postoperative complications.

[0004] In summary, the existing technology has the following main defects: first, it is difficult to achieve personalized implant design, and it is impossible to generate fine structures according to the needs of different patients; second, the design of the microstructure is difficult to take into account both mechanical strength and biocompatibility, resulting in insufficient long-term stability and integration effect of the implant; finally, the degree of automation of the manufacturing technology is not high, and it is impossible to accurately produce complex microstructures, which limits the further improvement of implant performance. Summary of the invention

[0005] One object of the present invention is to propose a method for intelligently generating dental implant microstructures based on a generative adversarial network. The present invention achieves multi-objective optimization of the implant microstructure and significantly improves the long-term stability and clinical success rate of the implant.

[0006] According to an embodiment of the present invention, a method for intelligently generating dental implant microstructures based on a generative adversarial network comprises the following steps:

[0007] S1. Collect the patient's alveolar bone CBCT scan data for three-dimensional reconstruction, generate a three-dimensional model of the alveolar bone, and extract anatomical features and mechanical demand data;

[0008] S2. Construct a two-layer generative adversarial network model, including a macro-generation layer and a micro-generation layer;

[0009] S3, designing multiple macro generators and macro discriminators in the macro generation layer, the macro generator generates the overall geometric structure of the dental implant according to the patient's anatomical feature data, and the multiple macro discriminators evaluate and adjust the generated overall geometric structure;

[0010] S4. In the micro-generation layer, multiple micro-generators and micro-discriminators are designed. The micro-generator generates the micro-structure of the dental implant according to the overall geometric structure output by the macro-generation layer. Multiple micro-discriminators perform multi-dimensional evaluation on the generated micro-structure based on different evaluation indicators.

[0011] S5. Introducing non-cooperative games between macro-generators and micro-generators, multiple generators compete with each other, and generate more complex and diverse dental implant structures through non-cooperative games in the macro-generation layer and the micro-generation layer;

[0012] S6. Introducing the cooperative game between macro discriminators and micro discriminators, multiple discriminators evaluate the generated implant structure through cooperation, and conduct a comprehensive evaluation from different biomechanical dimensions, so that the implant structure is synergistically optimized in terms of mechanical strength, adaptability and biocompatibility;

[0013] S7, repeating S3-S6 through joint optimization of the macro-generation layer and the micro-generation layer until the overall geometric structure and microstructure of the generated dental implant meet the predetermined biomechanical and biocompatibility requirements;

[0014] S8. Based on the final generated dental implant microstructure, the dental implant is manufactured through computer-aided design and manufacturing technology.

[0015] Optionally, S1 includes the following sub-steps:

[0016] S11. Use CBCT equipment to perform a full scan of the alveolar bone area of ​​the patient to obtain an original CBCT dataset. The original CBCT dataset consists of voxels V i express;

[0017] S12, input the acquired original CBCT data set into the 3D reconstruction algorithm, and use volume rendering technology to perform 3D reconstruction to generate a 3D model M of the alveolar bone, where the 3D model M = f(V i ) through the voxel data V i The three-dimensional volume model is obtained by gradual superposition and interpolation reconstruction;

[0018] S13. Use a multi-scale filtering algorithm to extract anatomical features of the reconstructed three-dimensional model M, and identify and extract key anatomical feature points in the alveolar bone:

[0019]

[0020] Among them, T a represents the anatomical feature matrix, Ω is the 3D model area of ​​the alveolar bone, and α i is the filter weight coefficient, It is a Laplace operator, which calculates the curvature changes in the x, y, and z directions, respectively, and identifies the characteristic areas of alveolar ridge height and alveolar bone density gradient in the anatomical structure;

[0021] S14. Extract the mechanical demand data of the alveolar bone based on the three-dimensional model M. The mechanical demand data includes the bone density D b , bite force F b and bone morphology parameters S b , respectively correspond to the local hardness, stress condition and shape complexity of the alveolar bone. The mechanical demand matrix is ​​expressed as:

[0022] R m =h(D b ,F b ,S b ).

[0023] Optionally, S3 includes the following sub-steps:

[0024] S31, constructing a two-layer generative adversarial network model, the two-layer generative adversarial network comprising a macroscopic generative layer and a microscopic generative layer, the macroscopic generative layer is responsible for generating the overall geometric structure of the dental implant, and the microscopic generative layer is responsible for generating the microstructure of the dental implant;

[0025] S32. In the macro generation layer, design multiple macro generators G m,i and the macro discriminator D m,j ,i=1,2,...,N G ,j=1,2,...,N D Represent the number of macro generators and macro discriminators respectively. The macro generator G m,i Receive the patient's anatomical feature data matrix T a and the mechanical demand data matrix R m , and through the function G m,i (T a ,R m ) Generate the overall geometric structure M of the dental implant g ;

[0026] S33, multiple macro discriminators D m,j The generated overall geometry M g For evaluation, the discriminator D m,j (M g) Using the preset evaluation criteria, including geometric adaptability, anatomical feature matching and mechanical properties, the evaluation results are expressed as:

[0027]

[0028] Among them, E j is the discriminator D m,j The evaluation results, and They represent the generated geometry M g and anatomical feature data T a The Laplace operator, σ(M g ) and σ(R m ) represent the geometric structure M g and the mechanical demand matrix R m The standard deviation of is used to evaluate the consistency between the geometric structure and the mechanical requirements. β1, β2 and β3 are the weight coefficients in the evaluation criteria, which control the importance of geometric morphology, anatomical feature matching and mechanical performance evaluation respectively.

[0029] S34, Macro Generator G m,i With the macro discriminator D m,j Through zero-sum game, the confrontation optimization is carried out. The zero-sum game objective function is defined as:

[0030]

[0031] Among them, L(G m,i ,D m,j ) is the game loss function between the macro generator and the discriminator, α1 and α2 are weight parameters used to control the optimization of the overall geometric structure and mechanical properties, γ1 and γ2 are parameters related to mechanical requirements, which are used to adjust the matching degree between the generated structure and the mechanical requirements, and F g Represents the generated geometry M g The combined result of forces in different directions is and The second-order derivatives of the generated structure in the x and y directions, respectively, represent the changes in stress distribution and are used to evaluate the compressive strength and durability of the implant;

[0032] S35, repeat S32-S34 until the overall geometric structure M of the dental implant is generated. g Meet the preset anatomical characteristics and mechanical requirements.

[0033] Optionally, S4 includes the following sub-steps:

[0034] S41. In the micro-generation layer, design multiple micro-generators G w,i and the microscopic discriminator D w,j , microscopic generator Gw,i The overall geometry M that receives the output of the macro generation layer g , and generate the microstructure M of the dental implant based on the overall geometric structure w :

[0035] M w =G w,i (M g );

[0036] S42, multiple microscopic discriminators D w,j The generated microstructure M w Multi-dimensional evaluation including mechanical properties, surface roughness, porosity and biocompatibility of microstructures was performed. Each microscopic discriminator D w,j (M w ) gives evaluation results based on different evaluation criteria:

[0037]

[0038] Among them, E w,j To evaluate the results, and They represent the first-order derivatives of the microstructure and macrostructure in the x direction, respectively, and are used to evaluate the consistency between the microstructure and the overall geometric structure; γ is the parameter that controls the adjustment of structural details, S w is the surface roughness of the microstructure, P w and P target are the porosity of the generated microstructure and the target porosity, respectively, and β4, β5, and β6 are the weight coefficients in the evaluation criteria;

[0039] S43, Micro Generator G w,i and the microscopic discriminator D w,j Through dynamic game, confrontation optimization is carried out, and the dynamic game objective function is defined as:

[0040]

[0041] in, is the real distribution of the microstructure, For the distribution of macroscopic geometric structures, the microscopic generator G is optimized through dynamic game confrontation. w,i Continuously generate microstructures that meet the evaluation criteria, the microscopic discriminator D w,j Continuously optimize the evaluation accuracy of various microstructure performances;

[0042] S44, repeat S41-S43 until the generated dental implant microstructure M w Meet the preset requirements of mechanical properties, biocompatibility and surface characteristics.

[0043] Optionally, S5 includes the following sub-steps:

[0044] S51, introduce macro generator G m,i and the microscopic generator G w,i A non-cooperative game model between multiple generators, where multiple generators compete with each other and the macro generator G m,i Generate the overall geometry M of the dental implant g , microscopic generator G w,i On this basis, the microstructure M of the dental implant is generated w ;

[0045] S52. In the non-cooperative game framework, the macro generator G m,i and the microscopic generator G w,i Through optimization with different objective functions, the goal of each generator is to generate a structure that better meets the requirements while suppressing the output of other generators;

[0046] S54. In the non-cooperative game, the macro generator and the micro generator are adjusted through adversarial optimization. The goal of each generator is to gain an advantage in complexity and adaptability. The output M of the macro generator g Provide input to the micro generator, which generates more complex and mechanically compliant microstructures based on the macro generator, described by a loss function that combines the geometric complexity of the generator output structure with the performance optimization of the output structure in terms of mechanical strength, biocompatibility, and diversity:

[0047]

[0048] Among them, L total Represents the macro generator G m,i and the microscopic generator G w,i The non-cooperative game goal between Represents the geometric structure M generated by the macro generator g The loss term, ΔM g represents the overall change of the macroscopic geometric structure, γ3 is the parameter that controls the range of change of the macroscopic structure, represents the microstructure M generated by the micro-generator w The loss term, F w The generated microstructure M w The stress condition, γ4 is the weight parameter for controlling the mechanical properties, F w is the force distribution of the microstructure, α3 and α4 are the regularization parameters of the macro generator and micro generator, respectively, which control the complexity and diversity of the generated structure;

[0049] S55, repeat S51-S54 until the dental implant structure generated in the macro-generation layer and the micro-generation layer meets the preset complexity, adaptability and diversity requirements.

[0050] Optionally, S6 includes the following sub-steps:

[0051] S61. Introducing multiple macro discriminators D m,j and the microscopic discriminator D w,j The macro discriminator is responsible for evaluating the overall geometric structure M of the dental implant. g , the microscopic discriminator is responsible for evaluating the fine structure D m,j ,The generated implant structure is comprehensively evaluated from different biomechanical dimensions through cooperative game;

[0052] S62, macro discriminator D m,j The overall geometry M generated by evaluating g Mechanical strength and fit are scored based on biomechanical properties:

[0053]

[0054] Among them, E m,j is the evaluation result of the macro discriminator, β7, β8, and γ1 are weight parameters;

[0055] S63, micro discriminator D w,j The generated microstructure M w Biocompatibility and porosity assessments were performed:

[0056]

[0057] Among them, E w,j is the evaluation result of the micro discriminator, σ(M w ) and σ target represent the standard deviation and target standard deviation of the microstructure, respectively, β9 and β0 are weight parameters that control the effects of biocompatibility and porosity;

[0058] S64. Through cooperative game, macro discriminator D m,j and the microscopic discriminator D w,j The mechanical strength, adaptability and biocompatibility of the generated structure are jointly optimized, and the optimization objective function is expressed as:

[0059]

[0060] Among them, L coop is the cooperative game objective function between the macro discriminator and the micro discriminator. Through cooperative optimization, the overall structure M of the dental implant is g and microstructure M w Achieve the best balance between mechanical strength, fit and biocompatibility;

[0061] S65. Repeat S61-S64 until the overall geometric structure and microstructure of the generated dental implant meet the preset biomechanical and biocompatibility requirements.

[0062] Optionally, S8 includes the following sub-steps:

[0063] S81, the final generated dental implant overall geometric structure M g,final and microstructure M w,final Input the CAD system, model the overall geometric shape of the dental implant through the CAD system, and generate a three-dimensional model of the dental implant CAD ;

[0064] S82, according to the three-dimensional model M CAD The data is used to generate the manufacturing path of the dental implant using CAM technology. The manufacturing path is defined as:

[0065]

[0066] Among them, P CAM represents the machining path generated during the manufacturing process, Ω m is the domain of manufacturing process, is the Laplacian operator of the three-dimensional model, It is the geometric change of the model in the x direction, which is used to calculate the tool path and processing trajectory;

[0067] S83. Based on the design requirements of dental implants, materials with biocompatibility and mechanical strength should be selected for manufacturing.

[0068] The beneficial effects of the present invention are:

[0069] (1) The present invention realizes the intelligent generation of the overall geometric structure and microstructure of dental implants in the macroscopic generation layer and the microscopic generation layer respectively by constructing a two-layer generative adversarial network. Multiple generators and discriminators in the generative adversarial network are continuously optimized in the game. When zero-sum game and non-cooperative game are introduced, the generator can dynamically generate personalized implant design according to the patient's anatomical characteristics and mechanical requirements. The multi-level game mechanism generates customized implants that meet the needs of individual patients through continuous iteration and optimization, thereby improving the adaptability and clinical effect of the implant. The traditional standardized design method lacks personalization, while the present invention significantly improves the clinical adaptability of the implant by intelligently generating personalized design through the generative adversarial network.

[0070] (2) The present invention introduces a non-cooperative game model between the macro generator and the micro generator, so that the generated microstructure is significantly optimized in terms of complexity, diversity and biocompatibility. During the design process, the micro generator can not only generate microstructures that meet the mechanical strength requirements, but also optimize the biocompatibility through a dynamic game mechanism, so that the implant can maximize its bone integration effect while meeting the mechanical properties. Compared with the traditional method of compromise design between mechanical properties and biocompatibility, the present invention uses generative adversarial networks and game theory mechanisms to achieve multi-objective optimization of implant microstructures, significantly improving the long-term stability and clinical success rate of implants. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0072] Figure 1 A flowchart of a method for intelligently generating microstructures of dental implants based on a generative adversarial network proposed by the present invention;

[0073] Figure 2 A flowchart of non-cooperative game optimization between a macro generator and a micro generator in a method for intelligently generating microstructures of dental implants based on a generative adversarial network proposed by the present invention;

[0074] Figure 3 This is a structural schematic diagram of the cooperative game optimization of the macro discriminator and the micro discriminator in the intelligent generation method of dental implant microstructure based on the generative adversarial network proposed in the present invention. DETAILED DESCRIPTION

[0075] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, which only illustrate the basic structure of the present invention in a schematic manner, and therefore only show the components related to the present invention.

[0076] refer to Figure 1-3 , a method for intelligently generating microstructure of dental implants based on generative adversarial networks, comprising the following steps:

[0077] S1. Collect the patient's alveolar bone CBCT scan data for three-dimensional reconstruction, generate a three-dimensional model of the alveolar bone, and extract anatomical features and mechanical demand data;

[0078] S2. Construct a two-layer generative adversarial network model, including a macro-generation layer and a micro-generation layer;

[0079] S3, designing multiple macro generators and macro discriminators in the macro generation layer, the macro generator generates the overall geometric structure of the dental implant according to the patient's anatomical feature data, and the multiple macro discriminators evaluate and adjust the generated overall geometric structure;

[0080] S4. In the micro-generation layer, multiple micro-generators and micro-discriminators are designed. The micro-generator generates the micro-structure of the dental implant according to the overall geometric structure output by the macro-generation layer. Multiple micro-discriminators perform multi-dimensional evaluation on the generated micro-structure based on different evaluation indicators.

[0081] S5. Introducing non-cooperative games between macro-generators and micro-generators, multiple generators compete with each other, and generate more complex and diverse dental implant structures through non-cooperative games in the macro-generation layer and the micro-generation layer;

[0082] S6. Introducing the cooperative game between macro discriminators and micro discriminators, multiple discriminators evaluate the generated implant structure through cooperation, and conduct a comprehensive evaluation from different biomechanical dimensions, so that the implant structure is synergistically optimized in terms of mechanical strength, adaptability and biocompatibility;

[0083] S7, repeating S3-S6 through joint optimization of the macro-generation layer and the micro-generation layer until the overall geometric structure and microstructure of the generated dental implant meet the predetermined biomechanical and biocompatibility requirements;

[0084] S8. Based on the final generated dental implant microstructure, the dental implant is manufactured through computer-aided design and manufacturing technology.

[0085] In this implementation, S1 includes the following sub-steps:

[0086] S11. Use CBCT equipment to perform a full scan of the alveolar bone area of ​​the patient to obtain an original CBCT dataset. The original CBCT dataset consists of voxels V i express;

[0087] S12, input the acquired original CBCT data set into the 3D reconstruction algorithm, and use volume rendering technology to perform 3D reconstruction to generate a 3D model M of the alveolar bone, where the 3D model M = f(V i ) through the voxel data V i The three-dimensional volume model is obtained by gradual superposition and interpolation reconstruction;

[0088] S13. Use a multi-scale filtering algorithm to extract anatomical features of the reconstructed three-dimensional model M, and identify and extract key anatomical feature points in the alveolar bone:

[0089]

[0090] Among them, Ta represents the anatomical feature matrix, Ω is the 3D model area of ​​the alveolar bone, and α i is the filter weight coefficient, It is a Laplace operator, which calculates the curvature changes in the x, y, and z directions, respectively, and identifies the characteristic areas of alveolar ridge height and alveolar bone density gradient in the anatomical structure;

[0091] S14. Extract the mechanical demand data of the alveolar bone based on the three-dimensional model M. The mechanical demand data includes the bone density D b , bite force F b and bone morphology parameters S b , respectively correspond to the local hardness, stress condition and shape complexity of the alveolar bone. The mechanical demand matrix is ​​expressed as:

[0092] R m =h(D b ,F b ,S b ).

[0093] In this implementation, S3 includes the following sub-steps:

[0094] S31. Construct a two-layer generative adversarial network model. The two-layer generative adversarial network includes a macro-generation layer and a micro-generation layer. The macro-generation layer is responsible for generating the overall geometric structure of the dental implant, and the micro-generation layer is responsible for generating the microstructure of the dental implant.

[0095] S32. In the macro generation layer, design multiple macro generators G m,i and the macro discriminator D m,j ,i=1,2,...,N G ,j=1,2,...,N D Represent the number of macro generators and macro discriminators respectively. The macro generator G m,i Receive the patient's anatomical feature data matrix T a and the mechanical demand data matrix R m , and through the function G m,i (T a ,R m ) Generate the overall geometric structure M of the dental implant g ;

[0096] S33, multiple macro discriminators D m,j The generated overall geometry M g For evaluation, the discriminator D m,j (M g ) Using the preset evaluation criteria, including geometric adaptability, anatomical feature matching and mechanical properties, the evaluation results are expressed as:

[0097]

[0098] Among them, E j is the discriminator D m,j The evaluation results, and They represent the generated geometry M g and anatomical feature data T a The Laplace operator, σ(M g ) and σ(R m ) represent the geometric structure M g and the mechanical demand matrix R m The standard deviation of is used to evaluate the consistency between the geometric structure and the mechanical requirements. β1, β2 and β3 are the weight coefficients in the evaluation criteria, which control the importance of geometric morphology, anatomical feature matching and mechanical performance evaluation respectively.

[0099] S34, Macro Generator G m,i With the macro discriminator D m,j Through zero-sum game, the confrontation optimization is carried out. The zero-sum game objective function is defined as:

[0100]

[0101] Among them, L(G m,i ,D m,j ) is the game loss function between the macro generator and the discriminator, α1 and α2 are weight parameters used to control the optimization of the overall geometric structure and mechanical properties, γ1 and γ2 are parameters related to mechanical requirements, which are used to adjust the matching degree between the generated structure and the mechanical requirements, and F g Represents the generated geometry M g The combined result of forces in different directions is and The second-order derivatives of the generated structure in the x and y directions, respectively, represent the changes in stress distribution and are used to evaluate the compressive strength and durability of the implant;

[0102] S35, repeat S32-S34 until the overall geometric structure M of the dental implant is generated. g Meet the preset anatomical characteristics and mechanical requirements.

[0103] In this implementation, S4 includes the following sub-steps:

[0104] S41. In the micro-generation layer, design multiple micro-generators G w,i and the microscopic discriminator D w,j , microscopic generator G w,i The overall geometry M that receives the output of the macro generation layer g , and generate the microstructure M of the dental implant based on the overall geometric structure w :

[0105] M w =G w，i (M g );

[0106] S42, multiple microscopic discriminators D w,j The generated microstructure M w Multi-dimensional evaluation including mechanical properties, surface roughness, porosity and biocompatibility of microstructures was performed. Each microscopic discriminator D w,j (M w ) gives evaluation results based on different evaluation criteria:

[0107]

[0108] Among them, E w,j To evaluate the results, and They represent the first-order derivatives of the microstructure and macrostructure in the x direction, respectively, and are used to evaluate the consistency between the microstructure and the overall geometric structure; γ is the parameter that controls the adjustment of structural details, S w is the surface roughness of the microstructure, P w and P target are the porosity of the generated microstructure and the target porosity, respectively, and β4, β5, and β6 are the weight coefficients in the evaluation criteria;

[0109] S43, Micro Generator G w,i and the microscopic discriminator D w,j Through dynamic game, confrontation optimization is carried out, and the dynamic game objective function is defined as:

[0110]

[0111] in, is the real distribution of the microstructure, For the distribution of macroscopic geometric structures, the microscopic generator G is optimized through dynamic game confrontation. w,i Continuously generate microstructures that meet the evaluation criteria, the microscopic discriminator D w,j Continuously optimize the evaluation accuracy of various microstructure performances;

[0112] S44, repeat S41-S43 until the generated dental implant microstructure M w Meet the preset requirements of mechanical properties, biocompatibility and surface characteristics.

[0113] In this implementation, S5 includes the following sub-steps:

[0114] S51, introduce macro generator G m,i and the microscopic generator G w,iA non-cooperative game model between multiple generators, where multiple generators compete with each other and the macro generator G m,i Generate the overall geometry M of the dental implant g , microscopic generator G w,i On this basis, the microstructure M of the dental implant is generated w ;

[0115] S52. In the non-cooperative game framework, the macro generator G m,i and the microscopic generator G w,i Through optimization with different objective functions, the goal of each generator is to generate a structure that better meets the requirements while suppressing the output of other generators;

[0116] S54. In the non-cooperative game, the macro generator and the micro generator are adjusted through adversarial optimization. The goal of each generator is to gain an advantage in complexity and adaptability. The output M of the macro generator g Provide input to the micro generator, which generates more complex and mechanically compliant microstructures based on the macro generator, described by a loss function that combines the geometric complexity of the generator output structure with the performance optimization of the output structure in terms of mechanical strength, biocompatibility, and diversity:

[0117]

[0118] Among them, L total Represents the macro generator G m,i and the microscopic generator G w,i The non-cooperative game goal between Represents the geometric structure M generated by the macro generator g The loss term, ΔM g represents the overall change of the macroscopic geometric structure, γ3 is the parameter that controls the range of change of the macroscopic structure, represents the microstructure M generated by the micro-generator w The loss term, F w The generated microstructure M w The stress condition, γ4 is the weight parameter for controlling the mechanical properties, F w is the force distribution of the microstructure, α3 and α4 are the regularization parameters of the macro generator and micro generator, respectively, which control the complexity and diversity of the generated structure;

[0119] S55, repeat S51-S54 until the dental implant structure generated in the macro-generation layer and the micro-generation layer meets the preset complexity, adaptability and diversity requirements.

[0120] In this implementation, S6 includes the following sub-steps:

[0121] S61. Introducing multiple macro discriminators D m,j and the microscopic discriminator D w,j The macro discriminator is responsible for evaluating the overall geometric structure M of the dental implant. g , the microscopic discriminator is responsible for evaluating the fine structure D m,j ,The generated implant structure is comprehensively evaluated from different biomechanical dimensions through cooperative game;

[0122] S62, macro discriminator D m,j The overall geometry M generated by evaluating g Mechanical strength and fit are scored based on biomechanical properties:

[0123]

[0124] Among them, E m,j is the evaluation result of the macro discriminator, β7, β8, and γ1 are weight parameters;

[0125] S63, micro discriminator D w,j The generated microstructure M w Biocompatibility and porosity assessments were performed:

[0126]

[0127] Among them, E w,j is the evaluation result of the micro discriminator, σ(M w ) and σ target represent the standard deviation and target standard deviation of the microstructure, respectively, β9 and β0 are weight parameters that control the effects of biocompatibility and porosity;

[0128] S64. Through cooperative game, macro discriminator D m,j and the microscopic discriminator D w,j The mechanical strength, adaptability and biocompatibility of the generated structure are jointly optimized, and the optimization objective function is expressed as:

[0129]

[0130] Among them, L coop is the cooperative game objective function between the macro discriminator and the micro discriminator. Through cooperative optimization, the overall structure M of the dental implant is g and microstructure M w Achieve the best balance between mechanical strength, fit and biocompatibility;

[0131] S65. Repeat S61-S64 until the overall geometric structure and microstructure of the generated dental implant meet the preset biomechanical and biocompatibility requirements.

[0132] In this implementation, S8 includes the following sub-steps:

[0133] S81, the final generated dental implant overall geometric structure M g,final and microstructure M w,final Input the CAD system, model the overall geometric shape of the dental implant through the CAD system, and generate a three-dimensional model of the dental implant CAD ;

[0134] S82, according to the three-dimensional model M CAD The data is used to generate the manufacturing path of the dental implant using CAM technology. The manufacturing path is defined as:

[0135]

[0136] Among them, P CAM represents the processing path generated during the manufacturing process, Ω m is the domain of manufacturing process, is the Laplacian operator of the three-dimensional model, It is the geometric change of the model in the x direction, which is used to calculate the tool path and processing trajectory;

[0137] S83. Based on the design requirements of dental implants, materials with biocompatibility and mechanical strength should be selected for manufacturing.

[0138] Embodiment 1:

[0139] In order to verify the feasibility and effectiveness of the present invention in the intelligent generation of dental implant microstructures, the experimenters selected two public dental implant design and manufacturing-related data sets: DentalImplant3D and BoneDensityCAD. The two data sets are widely used in the design and research of dental implants, and contain rich alveolar bone CT scan data, mechanical test data, and long-term clinical follow-up data of implants. The DentalImplant3D data set covers the three-dimensional structure data of alveolar bones of 1,000 patients from 2020 to 2023, including bone density, anatomical features, and postoperative implant biocompatibility and bone integration effects in different regions. The BoneDensityCAD data set includes mechanical properties data of 700 patients collected from early 2021 to mid-2023, covering bite force, bone density distribution, and the effects of different implant designs on long-term stability.

[0140] The experimenters divided the two data sets into training set and test set in a ratio of 8:2. The training set was used to train the generative adversarial network (GANs) model, and the test set was used to verify the performance of the generated implants in terms of adaptability, mechanical properties and biocompatibility. The training data set contained bone density data, mechanical demand data and patient anatomical structure data. The anatomical data came from CBCT three-dimensional reconstruction, the bone density data came from finite element analysis, and the mechanical property data came from bite force testing.

[0141] In order to train the macro and micro generators in the generative adversarial network, the experimenters first used CBCT data to reconstruct the patient's alveolar bone in three dimensions and extract its anatomical features and mechanical requirements. These data were input into the input layer of the generative adversarial network. In order to extract visual features, the experimenters used a pre-trained 3D convolutional network with an embedding dimension of 4096 to capture key features in the three-dimensional geometric structure. For mechanical demand features (including bone density and bite force), the experimenters used traditional mechanical analysis methods to generate corresponding embedding vectors with a dimension of 1024.

[0142] In model training, the experimenters used the Adam optimizer with the learning rate set to 0.0001. The training was conducted for a total of 500 epochs and the mini-batch size was set to 32. In order to ensure the optimization of the mechanical properties and biocompatibility of the generated implants, the experimenters set the loss functions of the generator and discriminator to multi-objective optimization functions based on mechanical requirements and biocompatibility. After each round of training, the experimenters used the test set to perform mechanical simulation and biocompatibility evaluation on the generated implants, and calculated their fitness and mechanical performance indicators.

[0143] In the experiment, the experimenters compared the traditional template-based design method with the intelligent generation method based on generative adversarial networks and game theory proposed in the present invention. On the DentalImplant3D dataset, the implants generated by the method of the present invention were significantly better than the traditional method in terms of fit and mechanical properties. Table 1 lists the comparative data between the experimenter's method and the traditional design method:

[0144] Table 1 Comparison on the DentalImplant3D dataset

[0145]

[0146]

[0147] From the data in Table 1, it can be seen that the implants generated by the method of the present invention are superior to the traditional methods in terms of fit, biocompatibility and compressive strength. In terms of the generation time of the implant, the present invention greatly reduces the design time through the intelligent generative adversarial network, which is only about one-third of the traditional method. In terms of implant fit and biocompatibility, the method of the present invention reaches 96.8% and 94.5% respectively, while the traditional method has low fit and biocompatibility due to the limitations of template design, which are 82.5% and 79.3% respectively; in terms of mechanical properties, the compressive strength of the implant generated by the present invention is 315MPa, which is significantly higher than the 250MPa of the traditional method. In addition, the porosity and surface roughness of the implant are very important for bone integration. The implant generated by the present invention has a porosity of 34.6% and a surface roughness of 18.2μm, which is suitable for the attachment and growth of bone cells and effectively promotes the bone integration process. The porosity and surface roughness of the traditional method are relatively low, 20.8% and 40.5μm respectively, resulting in unsatisfactory bone integration effect.

[0148] On the BoneDensityCAD dataset, the method of the present invention performs well in optimizing mechanical properties and biocompatibility. Table 2 shows the experimental results of the experimenters on the dataset:

[0149] Table 2 Comparison on the BoneDensityCAD dataset

[0150]

[0151] It can be seen from the data in Table 2 that the method of the present invention has significant improvements in mechanical properties, bone integration success rate and manufacturing accuracy compared with traditional template design. In terms of mechanical properties, the implant generated by the present invention withstands a occlusal force of 122N, which is higher than 105N of the traditional method. In terms of bone integration success rate, the method of the present invention reaches 93.2%, while the traditional method is only 75.5%, which proves the advantage of the present invention in long-term clinical effects.

[0152] During the experiment, the experimenters used fit, compressive strength, porosity and biocompatibility as the main evaluation indicators. In order to better deal with the imbalance problem in the data set, the experimenters added weighted F1 score as a supplementary evaluation indicator to ensure the consistency and reliability of the model under different patient conditions. Table 3 shows the comparison of the results of different evaluation indicators under the two methods:

[0153] Table 3 Comparison of results of different evaluation indicators

[0154] Evaluation Metrics The method of the present invention (GANs + game theory) Traditional method (template design) Accuracy (%) 96.5 84.2 Precision (%) 94.8 80.1 Recall (%) 95.2 81.7 Weighted F1 score (%) 95.7 82.5

[0155] It can be seen from the data in Table 3 that the method of the present invention is significantly superior to the traditional design method in terms of different evaluation indicators, especially in terms of improvement in precision and recall rate, indicating that the intelligent generation method of the present invention can better handle complex personalized implant design requirements, especially in dealing with complex alveolar bone structure and mechanical requirements, showing higher reliability.

[0156] This example demonstrates the feasibility and effectiveness of the present invention in the design and manufacturing of dental implants through experimental verification of real data sets. The experimental results show that compared with traditional template design, the method of the present invention can significantly improve the fit, biocompatibility and mechanical properties of the implant, while significantly reducing the design time and manufacturing costs, providing a more efficient and accurate solution for personalized dental implant design.

[0157] The present invention constructs a two-layer generative adversarial network to realize the intelligent generation of the overall geometric structure and microstructure of the dental implant in the macroscopic generation layer and the microscopic generation layer respectively. Multiple generators and discriminators in the generative adversarial network are continuously optimized in the game. When zero-sum game and non-cooperative game are introduced, the generator can dynamically generate personalized implant design according to the patient's anatomical characteristics and mechanical requirements. The multi-level game mechanism generates customized implants that meet the needs of individual patients through continuous iteration and optimization, thereby improving the adaptability and clinical effect of the implant. The traditional standardized design method lacks personalization, while the present invention significantly improves the clinical adaptability of the implant by intelligently generating personalized design through the generative adversarial network.

[0158] The present invention introduces a non-cooperative game model between the macro generator and the micro generator, so that the generated microstructure is significantly optimized in terms of complexity, diversity and biocompatibility. During the design process, the micro generator can not only generate microstructures that meet the mechanical strength requirements, but also optimize the biocompatibility through a dynamic game mechanism, so that the implant can maximize its bone integration effect while meeting the mechanical properties. Compared with the traditional method of compromise design between mechanical properties and biocompatibility, the present invention uses generative adversarial networks and game theory mechanisms to achieve multi-objective optimization of implant microstructures, significantly improving the long-term stability and clinical success rate of the implant.

[0159] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for intelligent generation of dental implant microstructure based on generative adversarial networks, characterized in that: The steps include: S1. Collect the patient's alveolar bone CBCT scan data for three-dimensional reconstruction, generate a three-dimensional model of the alveolar bone, and extract anatomical features and mechanical demand data; S2. Construct a two-layer generative adversarial network model, including a macro-generation layer and a micro-generation layer; S3, designing multiple macro generators and macro discriminators in the macro generation layer, the macro generator generates the overall geometric structure of the dental implant according to the patient's anatomical feature data, and the multiple macro discriminators evaluate and adjust the generated overall geometric structure; S4. In the micro-generation layer, multiple micro-generators and micro-discriminators are designed. The micro-generator generates the micro-structure of the dental implant according to the overall geometric structure output by the macro-generation layer. Multiple micro-discriminators perform multi-dimensional evaluation on the generated micro-structure based on different evaluation indicators. S5. Introducing non-cooperative games between macro-generators and micro-generators, multiple generators compete with each other, and generate more complex and diverse dental implant structures through non-cooperative games in the macro-generation layer and the micro-generation layer; S6. Introducing the cooperative game between macro discriminators and micro discriminators, multiple discriminators evaluate the generated implant structure through cooperation, and conduct a comprehensive evaluation from different biomechanical dimensions, so that the implant structure is synergistically optimized in terms of mechanical strength, adaptability and biocompatibility; S7, repeating S3-S6 through joint optimization of the macro-generation layer and the micro-generation layer until the overall geometric structure and microstructure of the generated dental implant meet the predetermined biomechanical and biocompatibility requirements; S8. Based on the final generated dental implant microstructure, the dental implant is manufactured through computer-aided design and manufacturing technology.

2. The method for intelligently generating dental implant microstructure based on generative adversarial network according to claim 1, characterized in that: The S1 comprises the following sub-steps: S11. Use CBCT equipment to perform a full scan of the alveolar bone area of ​​the patient to obtain an original CBCT dataset. The original CBCT dataset consists of voxels V i express; S12, input the acquired original CBCT data set into the 3D reconstruction algorithm, and use volume rendering technology to perform 3D reconstruction to generate a 3D model M of the alveolar bone, where the 3D model M = f(V i ) through the voxel data V i The three-dimensional volume model is obtained by gradual superposition and interpolation reconstruction; S13. Use a multi-scale filtering algorithm to extract anatomical features of the reconstructed three-dimensional model M, and identify and extract key anatomical feature points in the alveolar bone: Among them, T a represents the anatomical feature matrix, Ω is the 3D model area of ​​the alveolar bone, and α i is the filter weight coefficient, It is a Laplace operator, which calculates the curvature changes in the x, y, and z directions, respectively, and identifies the characteristic areas of alveolar ridge height and alveolar bone density gradient in the anatomical structure; S14. Extract the mechanical demand data of the alveolar bone based on the three-dimensional model M. The mechanical demand data includes the bone density D b , bite force F b and bone morphology parameters S b , respectively correspond to the local hardness, stress condition and shape complexity of the alveolar bone. The mechanical demand matrix is ​​expressed as: R m =h(D b ,F b ,S b )。 3. The method for intelligently generating dental implant microstructure based on generative adversarial network according to claim 1, characterized in that: The S3 comprises the following sub-steps: S31, constructing a two-layer generative adversarial network model, the two-layer generative adversarial network comprising a macroscopic generative layer and a microscopic generative layer, the macroscopic generative layer is responsible for generating the overall geometric structure of the dental implant, and the microscopic generative layer is responsible for generating the microstructure of the dental implant; S32. In the macro generation layer, design multiple macro generators G m,i and the macro discriminator D m,j ,i=1,2,...,N G ,j=1,2,...,N D Represent the number of macro generators and macro discriminators respectively. The macro generator G m,i Receive the patient's anatomical feature data matrix T a and the mechanical demand data matrix R m , and through the function G m,i (T a ,R m ) Generate the overall geometric structure M of the dental implant g ; S33, multiple macro discriminators D m,j The generated overall geometry M g For evaluation, the discriminator D m,j (M g ) Using the preset evaluation criteria, including geometric adaptability, anatomical feature matching and mechanical properties, the evaluation results are expressed as: Among them, E j is the discriminator D m,j The evaluation results, and They represent the generated geometry M g and anatomical feature data T a The Laplace operator, σ(M g ) and σ(R m ) represent the geometric structure M g and the mechanical demand matrix R m The standard deviation of is used to evaluate the consistency between the geometric structure and the mechanical requirements. β1, β2 and β3 are the weight coefficients in the evaluation criteria, which control the importance of geometric morphology, anatomical feature matching and mechanical performance evaluation respectively. S34, Macro Generator G m,i With the macro discriminator D m,j Through zero-sum game, the confrontation optimization is carried out. The zero-sum game objective function is defined as: Among them, L(G m,i ,D m,j ) is the game loss function between the macro generator and the discriminator, α1 and α2 are weight parameters used to control the optimization of the overall geometric structure and mechanical properties, γ1 and γ2 are parameters related to mechanical requirements, which are used to adjust the matching degree between the generated structure and the mechanical requirements, and F g Represents the generated geometry M g The combined result of forces in different directions is and The second-order derivatives of the generated structure in the x and y directions, respectively, represent the changes in stress distribution and are used to evaluate the compressive strength and durability of the implant; S35, repeat S32-S34 until the overall geometric structure M of the dental implant is generated. g Meet the preset anatomical characteristics and mechanical requirements.

4. The method for intelligently generating dental implant microstructure based on generative adversarial network according to claim 1, characterized in that: The S4 comprises the following sub-steps: S41. In the micro-generation layer, design multiple micro-generators G w,i and the microscopic discriminator D w,j , microscopic generator G w,i The overall geometry M that receives the output of the macro generation layer g , and generate the microstructure M of the dental implant based on the overall geometric structure w : M w =G w,i (M g ); S42, multiple microscopic discriminators D w,j The generated microstructure M w Multi-dimensional evaluation including mechanical properties, surface roughness, porosity and biocompatibility of microstructures was performed. Each microscopic discriminator D w,j (M w ) gives evaluation results based on different evaluation criteria: Among them, E w,j To evaluate the results, and They represent the first-order derivatives of the microstructure and macrostructure in the x direction, respectively, and are used to evaluate the consistency between the microstructure and the overall geometric structure; γ is the parameter that controls the adjustment of structural details, S w is the surface roughness of the microstructure, P w and P target are the porosity of the generated microstructure and the target porosity, respectively, and β4, β5, and β6 are the weight coefficients in the evaluation criteria; S43, Micro Generator G w,i and the microscopic discriminator D w,j Through dynamic game, confrontation optimization is carried out, and the dynamic game objective function is defined as: in, is the real distribution of the microstructure, For the distribution of macroscopic geometric structures, the microscopic generator G is optimized through dynamic game confrontation. w,i Continuously generate microstructures that meet the evaluation criteria, the microscopic discriminator D w,j Continuously optimize the evaluation accuracy of various microstructure performances; S44, repeat S41-S43 until the generated dental implant microstructure M w Meet the preset requirements of mechanical properties, biocompatibility and surface characteristics.

5. The method for intelligently generating dental implant microstructure based on generative adversarial network according to claim 1, characterized in that: The S5 comprises the following sub-steps: S51, introduce macro generator G m,i and the microscopic generator G w,i A non-cooperative game model between multiple generators, where multiple generators compete with each other and the macro generator G m,i Generate the overall geometry M of the dental implant g , microscopic generator G w,i On this basis, the microstructure M of the dental implant is generated w ; S52. In the non-cooperative game framework, the macro generator G m,i and the microscopic generator G w,i Through optimization with different objective functions, the goal of each generator is to generate a structure that better meets the requirements while suppressing the output of other generators; S54. In the non-cooperative game, the macro generator and the micro generator are adjusted through adversarial optimization. The goal of each generator is to gain an advantage in complexity and adaptability. The output M of the macro generator g Provide input to the micro generator, which generates more complex and mechanically compliant microstructures based on the macro generator, described by a loss function that combines the geometric complexity of the generator output structure with the performance optimization of the output structure in terms of mechanical strength, biocompatibility, and diversity: Among them, L total Represents the macro generator G m,i and the microscopic generator G w,i The non-cooperative game goal between Represents the geometric structure M generated by the macro generator g The loss term, ΔM g represents the overall change of the macroscopic geometric structure, γ3 is the parameter that controls the range of change of the macroscopic structure, represents the microstructure M generated by the micro-generator w The loss term, F w The generated microstructure M w The stress condition, γ4 is the weight parameter for controlling the mechanical properties, F w is the force distribution of the microstructure, α3 and α4 are the regularization parameters of the macro generator and micro generator, respectively, which control the complexity and diversity of the generated structure; S55, repeat S51-S54 until the dental implant structure generated in the macro-generation layer and the micro-generation layer meets the preset complexity, adaptability and diversity requirements.

6. The method for intelligently generating dental implant microstructure based on generative adversarial network according to claim 1, characterized in that: The S6 comprises the following sub-steps: S61. Introducing multiple macro discriminators D m,j and the microscopic discriminator D w,j The macro discriminator is responsible for evaluating the overall geometric structure M of the dental implant. g , the microscopic discriminator is responsible for evaluating the fine structure D m,j ,The generated implant structure is comprehensively evaluated from different biomechanical dimensions through cooperative game; S62, macro discriminator D m,j The overall geometry M generated by evaluating g Mechanical strength and fit are scored based on biomechanical properties: Among them, E m,j is the evaluation result of the macro discriminator, β7, β8, and γ1 are weight parameters; S63, Micro Discriminator D w,j The generated microstructure M w Biocompatibility and porosity assessments were performed: Among them, E w,j is the evaluation result of the micro discriminator, σ(M w ) and σ target represent the standard deviation and target standard deviation of the microstructure, respectively, β9 and β0 are weight parameters that control the effects of biocompatibility and porosity; S64. Through cooperative game, macro discriminator D m,j and the microscopic discriminator D w,j The mechanical strength, adaptability and biocompatibility of the generated structure are jointly optimized, and the optimization objective function is expressed as: Among them, L coop is the cooperative game objective function between the macro discriminator and the micro discriminator. Through cooperative optimization, the overall structure M of the dental implant is g and microstructure M w Achieve the best balance between mechanical strength, fit and biocompatibility; S65. Repeat S61-S64 until the overall geometric structure and microstructure of the generated dental implant meet the preset biomechanical and biocompatibility requirements.

7. The method for intelligently generating dental implant microstructure based on generative adversarial network according to claim 1, characterized in that: The S8 comprises the following sub-steps: S81, the final generated dental implant overall geometric structure M g,final and microstructure M w,final Input the CAD system, model the overall geometric shape of the dental implant through the CAD system, and generate a three-dimensional model of the dental implant CAD ; S82, according to the three-dimensional model M CAD The data is used to generate the manufacturing path of the dental implant using CAM technology. The manufacturing path is defined as: Among them, P CAM represents the processing path generated during the manufacturing process, Ω m is the domain of manufacturing process, is the Laplacian operator of the three-dimensional model, It is the geometric change of the model in the x direction, which is used to calculate the tool path and processing trajectory; S83. Based on the design requirements of dental implants, materials with biocompatibility and mechanical strength should be selected for manufacturing.

Images