Medical beauty implant material intelligent optimization method and system based on artificial intelligence

Abstract

The invention relates to a medical beauty implant material intelligent optimization method and system based on artificial intelligence, and the method comprises the following steps: S1, obtaining the performance parameters of different medical beauty implant materials, capturing the anatomical features of a patient through 3D scanning and MRI images, and carrying out the statistics of the personalized biological data of the patient; s2, constructing a multi-task learning model and training based on a multi-target supervised learning data set, wherein the target is to predict the performance of different implant material formulas in a medical beauty implant environment; s3, performing multi-objective optimization by adopting an NSGA-III algorithm to obtain an optimal implant material formula; s4, reconstructing a three-dimensional model of the patient by using a reverse modeling technology based on anatomical features of the patient; and S5, based on the three-dimensional model, according to the optimal implant material formula and the personalized biological data, simulating the biomechanical characteristics of the patient by utilizing finite element analysis, and outputting biocompatibility scores and suitability suggestions of the material in different implant environments. According to the method, the clinical requirements on customized design and application of the medical beauty implant material are effectively met.

Description

Technical Field

[0001] This invention relates to the field of intelligent optimization simulation, and in particular to an intelligent optimization method and system for medical aesthetic implant materials based on artificial intelligence. Background Technology

[0002] In the field of modern medical aesthetics, implant materials are widely used in various applications such as plastic surgery filling, contouring, and functional repair. Commonly used medical aesthetic implant materials include silicone, polylactic acid (PLA), and hyaluronic acid. These materials need to meet high requirements in multiple performance indicators such as shape stability, hydrophilicity, and biocompatibility. However, the biological characteristics (such as skin thickness, tissue activity, and metabolic rate) and specific needs of different patients vary greatly, making it impossible for existing general-purpose materials to fully adapt to personalized implantation needs. At the same time, material development has traditionally relied on a large number of laboratory tests and experience accumulation, which is time-consuming, inefficient, and costly, making it difficult to meet rapidly changing medical needs. Summary of the Invention

[0003] To address the aforementioned issues, the present invention aims to provide an intelligent optimization method and system for medical aesthetic implant materials based on artificial intelligence. This method not only improves the performance prediction and design accuracy of implant materials but also accelerates product development through intelligent optimization and simulation verification closed-loop.

[0004] To achieve the above objectives, the present invention adopts the following technical solution:

[0005] An AI-based intelligent optimization method for medical aesthetic implant materials includes the following steps:

[0006] S1: Obtain performance parameters of different medical aesthetic implant materials, capture patient anatomical features using 3D scanning and MRI images, and collect personalized biodata of patients;

[0007] S2: Construct a multi-objective supervised learning dataset, build a multi-task learning model and train it based on the multi-objective supervised learning dataset, with the goal of predicting the performance of different implant material formulations in the medical aesthetic implantation environment;

[0008] S3: Based on the predicted performance of different implant material formulations in the medical aesthetic implantation environment, the NSGA-III algorithm is used for multi-objective optimization to obtain the optimal implant material formulation.

[0009] S4: Based on the patient's anatomical features, reverse modeling technology is used to reconstruct a three-dimensional model of the patient's skin, soft tissue, and bones;

[0010] S5: Based on a 3D model, using finite element analysis to simulate the biomechanical characteristics of patients according to the optimal implant material formulation and personalized biological data, it outputs biocompatibility scores and fit recommendations for materials in different implantation environments.

[0011] Furthermore, performance parameters of different cosmetic implant materials were obtained through experimental data, literature data, and clinical data, as detailed below:

[0012] Obtain experimental test data for the materials, including tensile strength, elongation at break, and degradation rate;

[0013] The literature data was collected from the database using web crawling technology, focusing on papers related to medical aesthetic materials. A pre-trained NLP model was then used to extract performance data from the literature and obtain relevant performance indicators.

[0014] Clinical data is collected by studying historical application cases of implantable materials, including material types and their corresponding clinical feedback data.

[0015] The acquired performance data is annotated, with standardized units and ranges to form a structured data format. A material performance database is then constructed, categorized by material, performance, and parameter source.

[0016] Furthermore, 3D scanning and MRI images are used to capture the patient's anatomical features and collect personalized biodata, specifically as follows: using 3D scanning equipment, the geometric features of the patient's skin surface are captured, including thickness and contour characteristics; based on MRI images, the anatomical structures of the skin, soft tissue, and bones are acquired; allergic reactions and immune system characteristics are analyzed through the patient's medical history, combined with the patient's age, metabolic behavior, etc., to form an individualized characteristic description.

[0017] Furthermore, S2 specifically refers to:

[0018] The extracted input features include material performance data, patient-specific data, and target performance parameters; the material performance data includes the material composition ratio and preparation process parameters; the patient-specific data includes anatomical and biological characteristics; and the target performance parameters include elastic modulus E, biocompatibility score B, and durability D.

[0019] The input features are normalized to eliminate differences in units, and a multi-objective supervised learning dataset is constructed.

[0020] Combine material data and personalized features in the form of feature vectors:

[0021] X i =[C PLA C additive ,T heat ,P mold ,h skin ,ρ bone [IR,MR];

[0022] Among them, C PLAPLA content; C additive The proportion of additives; T heat h is the heat treatment temperature. skin For thickness; ρ bone Bone density; IR is the immune score; P mold Molding pressure; MR represents metabolic score;

[0023] The target value is stored as:

[0024] Y i =[E,B,D];

[0025] Where E is the elastic modulus, B is the biocompatibility score, and D is the durability.

[0026] The complete dataset is:

[0027] ;

[0028] The multi-task learning model employs a multi-task learning framework, completing the prediction task by sharing a feature extraction layer and a task-specific output layer.

[0029] The shared bottom layer extracts a general representation through a shared hidden layer, captures the correlation between different tasks, and improves data utilization efficiency. The input sample features are mapped and converted into shared features H by a nonlinear function.

[0030] The task branching layer uses independent neural network branches to handle task-specific characteristics and objectives, making predictions for each task separately.

[0031] Furthermore, the multi-task learning model includes an input layer, a shared hidden layer, and a task branching layer, as detailed below:

[0032] The input layer receives the feature vector X. i ;

[0033] The shared hidden layer extracts shared features through a fully connected layer:

[0034] H=ReLU(W s X i +b s );

[0035] Among them W s Let b be the weight matrix. s Here, m represents the input feature dimension, and d represents the hidden layer dimension;

[0036] The task branching layer, based on the shared feature H, allows different tasks to predict target results through independent branches; the task branching layer includes an elastic modulus prediction branch, a biocompatibility prediction branch, and a durability prediction branch:

[0037]

[0038] in, , and These are the weighting coefficients; , and For bias; , and These are the predicted values ​​for elastic modulus, biocompatibility, and durability, respectively.

[0039] The loss function is ;

[0040] in, , , , Task weights are dynamically adjusted.

[0041] Update the weights of the shared layer and task branch using gradient descent:

[0042] ;

[0043] Where η is the learning rate and t is the number of iterations.

[0044] Furthermore, based on the predicted performance of different implant material formulations in the medical aesthetic implantation environment, the NSGA-III algorithm was used for multi-objective optimization to obtain the optimal implant material formulation, as follows:

[0045] Optimization objectives include high elastic modulus f1 ( ), high biocompatibility f2 ( ) and high fatigue resistance f3 ( ):

[0046]

[0047] Constraints include material strength, degradation rate, and price.

[0048] Fracture strength limit g1 ( ):

[0049] ;

[0050] Where σ min Lower limit of fracture strength; fracture strength of implanted materials;

[0051] Degradation rate limit g2 ( ):

[0052] ;

[0053] Where, r degradation Indicates the degradation rate of the material, in the range r min ≤rd egradation ( )≤r max ;

[0054] Cost range constraint g3 ( ):

[0055] ;

[0056] Where: c( ) represents the unit price of the material, c max This represents the upper limit of the budgeted cost.

[0057] The final constraint problem is modeled as follows:

[0058] Find to minimize F( )=[f1( ),f2( ),f3( subject to G( )≤0;

[0059] Among them, G( ) including g1( g2 ) and g3( The optimal implant material formulation was obtained by using the NSGA-III algorithm for multi-objective optimization.

[0060] First, an initial population P of N candidate solutions is generated, and each sample... ∈P is randomly initialized;

[0061] For each candidate solution Calculate the aforementioned objective function values ​​f1, f2, f3, and check whether they satisfy the actual constraints. If the solution violates the constraints, define a constraint penalty function P( ):

[0062] ;

[0063] in, Let j be the constraint function;

[0064] Perform a non-dominated sort on all solutions in the population and divide them into distinct Pareto fronts: if solutions Better than solution ,but Dominant Solution And count the dominance of each solution:

[0065] Among the candidate solutions of the Pareto front, the diversity of solutions is maintained by selecting reference points from a predefined distribution, with the number of reference points being L:

[0066] ;

[0067] Where H represents the level of detail in the reference point division; M represents the number of targets;

[0068] When the algorithm iterates to the maximum number of algebras G or the solution converges, it outputs the Pareto optimal set, with each solution corresponding to an implant material formulation.

[0069] Furthermore, based on the patient's anatomical characteristics, reverse modeling techniques were used to reconstruct three-dimensional models of the patient's skin, soft tissue, and bones, as follows:

[0070] Based on the patient's anatomical characteristics, the U-Net deep learning framework was used to extract key anatomical information of the skin, soft tissue, and bone regions.

[0071] Based on key anatomical information, the two-dimensional segmented image data is converted into a three-dimensional geometric surface model using the reverse modeling algorithm Marching Cubes, and the generated surface mesh structure is optimized to obtain the final three-dimensional model.

[0072] Furthermore, based on the patient's anatomical characteristics, the U-Net deep learning framework was used to extract key anatomical information from the skin, soft tissue, and bone regions, as follows:

[0073] The system captures the surface morphology (contour characteristics, thickness, etc.) of the patient's skin using 3D scanning equipment, generates point cloud data of the skin, and converts it into a triangular mesh model.

[0074] Bone, soft tissue, and skin data were captured using T1 / T2 weighted MRI images, and the MRI images were then denoised and normalized in intensity.

[0075] Using the U-Net deep learning framework, skin, soft tissue, and bone are segmented from MRI images, the anatomical information of the segmented parts is extracted, and key region masks are generated.

[0076] The triangular mesh model and key region masks are registered to integrate the anatomical structures of soft tissue, bone, and skin, thereby obtaining key anatomical information for each region.

[0077] Furthermore, S5 specifically refers to:

[0078] S51: Constructing the finite element model:

[0079] Based on the patient's three-dimensional anatomical model obtained from S4, and the optimal implant material formula obtained from S3, the geometric model of the implant is obtained.

[0080] Boolean operations are performed between the implant geometry model and the patient's 3D anatomical model to define the contact interface and generate a finite element model. :

[0081] ;

[0082] in, To obtain the geometric model of the implant; These are three-dimensional anatomical models of bones, soft tissues, and skin, respectively.

[0083] S52: Mapping of Material Properties and Boundary Conditions

[0084] Based on the optimal formulation parameters, its mechanical properties are predicted through a multi-task learning model.

[0085] Biological tissue properties, with bones using an orthogonal anisotropic elastic model and soft tissues using a hyperelastic model;

[0086] Apply mechanical loads according to the patient's activity scenario and define the contact relationship between the implant and surrounding tissues;

[0087] S53: Solving biomechanical equations:

[0088] The equilibrium equations for finite element analysis are:

[0089] Ku=F;

[0090] Where K is the global stiffness matrix; u is the nodal displacement vector; and F is the external load vector.

[0091] The Newton-Raphson method is used to iteratively solve the nonlinear constitutive relations of implants or soft tissues:

[0092] ;

[0093] in, No. The tangent stiffness matrix of the next iteration; For displacement increment; This is the internal force vector; This is the external load vector;

[0094] Output stress / strain distribution, von Mises stress on implant surface:

[0095] ;

[0096] S54: Weighted calculation of biocompatibility score B:

[0097] ;

[0098] ;

[0099] ;

[0100] ;

[0101] Where w1, w2, and w3 are weighting coefficients; For stress shielding scoring; For wear rating, Assess implant displacement. For bone stress; The stress on bones in a healthy state; k is the wear attenuation coefficient; d slip u is the sliding distance of the contact surface; max The maximum displacement of the implant; u threshold Maximum permissible displacement limit.

[0102] An AI-based intelligent optimization system for medical aesthetic implant materials includes a processor, a memory, and a computer program stored in the memory. When the processor executes the computer program, it specifically performs the steps of the AI-based intelligent optimization method for medical aesthetic implant materials described above.

[0103] The present invention has the following beneficial effects:

[0104] 1. This invention combines 3D scanning, MRI images and patient-specific biological data, and reconstructs the patient's anatomical structure model through reverse modeling technology. This enables personalized implant material design for different patients throughout the entire process, and can accurately simulate the interaction between the implant and biological tissues, ensuring that the designed implant material better meets individual needs.

[0105] 2. This invention employs a multi-task learning model and a multi-objective optimization algorithm (NSGA-III) to predict the performance of different material formulations in the medical aesthetic environment under the drive of big data, and to find the optimal solution for implanted materials. Compared with the trial and error process of traditional material development, this method efficiently explores the performance space of different formulations, accelerates the material research and development process, and can balance multiple objectives such as mechanical properties, biocompatibility, and degradation rate.

[0106] 3. This invention comprehensively evaluates the interaction between implanted materials and the patient's biological environment through three-dimensional biomechanical simulation, outputs biocompatibility scores and fit suggestions, and forms a feedback closed loop optimization of material design based on the score results. This not only ensures the mechanical stability and long-term fit of the implanted materials, but also significantly improves the safety and clinical translation rate of the method. Attached Figure Description

[0107] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0108] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments:

[0109] refer to Figure 1 In this embodiment, an intelligent optimization method for medical aesthetic implant materials based on artificial intelligence is provided, including the following steps:

[0110] S1: Obtain the physical, chemical, and biological performance parameters of different medical aesthetic implant materials (such as elastic modulus, biodegradation rate, and corrosion resistance); data sources include experimental studies, literature archives, and existing clinical test records; use 3D scanning and MRI images to capture the patient's anatomical characteristics (such as skin thickness, bone structure, etc.) and collect the patient's personalized biological data (such as immune response, allergies, and relevant medical history).

[0111] S2: Construct a multi-objective supervised learning dataset, build a multi-task learning model and train it based on the multi-objective supervised learning dataset, with the goal of predicting the performance of different implant material formulations in the medical aesthetic implantation environment;

[0112] S3: Based on the predicted performance of different implant material formulations in the medical aesthetic implantation environment, the NSGA-III algorithm is used for multi-objective optimization to obtain the optimal implant material formulation.

[0113] S4: Based on the patient's anatomical features, reverse modeling technology is used to reconstruct a three-dimensional model of the patient's skin, soft tissue, and bones;

[0114] S5: Based on a 3D model, using finite element analysis to simulate the biomechanical characteristics of patients according to the optimal implant material formulation and personalized biological data, it outputs biocompatibility scores and fit recommendations for materials in different implantation environments.

[0115] In this embodiment, the performance parameters of different cosmetic implant materials were obtained through experimental data, literature data, and clinical data, as detailed below:

[0116] Obtain experimental test data for the materials, including tensile strength, elongation at break, and degradation rate;

[0117] The literature data was collected from databases (such as PubMed, Springer, and Elsevier) using web crawling techniques (such as Scrapy and Beautiful Soup) to collect papers related to medical aesthetic materials; and performance data from the papers was extracted using pre-trained NLP models (such as BERT or SciSpacy), and keyword searches were configured, such as "tensile strength", "degradation rate", and "elastic modulus", to obtain relevant performance indicators.

[0118] Clinical data is collected by studying historical application cases of implantable materials, including material types and their corresponding clinical feedback data.

[0119] The acquired physical, chemical, and biological performance data are annotated, with standardized units and ranges to form a structured data format. A materials performance database is then constructed, categorized by material, performance, and parameter source.

[0120] In this embodiment, 3D scanning and MRI images are used to capture the patient's anatomical features and collect the patient's personalized biodata, as follows: the geometric features of the patient's skin surface, including thickness and contour characteristics, are captured using a 3D scanning device; the anatomical structures of the skin, soft tissue, and bones are acquired based on MRI images; and allergic reactions and immune system characteristics are analyzed through the patient's medical history, combined with the patient's age, metabolic behavior, etc., to form an individualized feature description.

[0121] In this embodiment, S2 specifically refers to:

[0122] The extracted input features include material performance data, patient-specific data, and target performance parameters; the material performance data includes the composition ratio of the material (e.g., PLA content C). PLA Additive ratio C additive ), preparation process parameters (such as heat treatment temperature T) heat Patient-specific data includes anatomical features (skin thickness h). skin Bone density ρ bone ) and biological characteristics (immune score IR, metabolic score MR); target performance parameters include elastic modulus E, biocompatibility score B, and durability D;

[0123] The input features are normalized to eliminate differences in units, and a multi-objective supervised learning dataset is constructed.

[0124] Combine material data and personalized features in the form of feature vectors:

[0125] X i =[C PLA Cadditive ,T heat ,P mold ,h skin ,ρ bone [IR,MR];

[0126] Among them, C PLA PLA content; C additive The proportion of additives; T heat h is the heat treatment temperature. skin For thickness; ρ bone Bone density; IR is the immune score; P mold Molding pressure; MR represents metabolic score;

[0127] The target value is stored as:

[0128] Y i =[E,B,D];

[0129] Where E is the elastic modulus, B is the biocompatibility score, and D is the durability.

[0130] The complete dataset is:

[0131] ;

[0132] The multi-task learning model employs a multi-task learning framework, completing the prediction task by sharing a feature extraction layer and a task-specific output layer.

[0133] The shared bottom layer extracts a general representation through a shared hidden layer, captures the correlation between different tasks, and improves data utilization efficiency. The input sample features are mapped and converted into shared features H by a nonlinear function.

[0134] The task branching layer uses independent neural network branches to handle task-specific characteristics and objectives, making predictions for each task separately.

[0135] In this embodiment, the multi-task learning model includes an input layer, a shared hidden layer, and a task branching layer, as detailed below:

[0136] The input layer receives the feature vector X. i ;

[0137] The shared hidden layer extracts shared features through a fully connected layer:

[0138] H=ReLU(W s X i +b s );

[0139] Among them W s Let b be the weight matrix. sHere, m represents the input feature dimension, and d represents the hidden layer dimension;

[0140] The task branching layer, based on the shared feature H, allows different tasks to predict target results through independent branches; the task branching layer includes an elastic modulus prediction branch, a biocompatibility prediction branch, and a durability prediction branch:

[0141]

[0142] in, , and These are the weighting coefficients; , and For bias; , and These are the predicted values ​​for elastic modulus, biocompatibility, and durability, respectively.

[0143] The loss function is ;

[0144] in, , , , Task weights are dynamically adjusted.

[0145] Update the weights of the shared layer and task branch using gradient descent:

[0146] ;

[0147] Where η is the learning rate and t is the number of iterations.

[0148] In this embodiment, based on the predicted performance of different implant material formulations in the medical aesthetic implantation environment, the NSGA-III algorithm is used for multi-objective optimization to obtain the optimal implant material formulation, as detailed below:

[0149] Optimization objectives include high elastic modulus f1 ( ), high biocompatibility f2 ( ) and high fatigue resistance f3 ( ):

[0150]

[0151] Constraints include material strength, degradation rate, and price.

[0152] Fracture strength limit g1 ( ):

[0153] ;

[0154] Where σ min Lower limit of fracture strength; fracture strength of implanted materials;

[0155] Degradation rate limit g2 ( ):

[0156] ;

[0157] Where, r degradation Indicates the degradation rate of the material, in the range r min ≤rd egradation ( )≤r max ;

[0158] Cost range constraint g3 ( ):

[0159] ;

[0160] Where: c( ) represents the unit price of the material, c max This represents the upper limit of the budgeted cost.

[0161] The final constraint problem is modeled as follows:

[0162] Find to minimize F( )=[f1( ),f2( ),f3( subject to G( )≤0;

[0163] Among them, G( ) including g1( g2 ) and g3( The optimal implant material formulation was obtained by using the NSGA-III algorithm for multi-objective optimization.

[0164] First, an initial population P of N candidate solutions is generated, and each sample... ∈P is randomly initialized;

[0165] For each candidate solution Calculate the aforementioned objective function values ​​f1, f2, f3, and check whether they satisfy the actual constraints. If the solution violates the constraints, define a constraint penalty function P( ):

[0166] ;

[0167] in, Let j be the constraint function;

[0168] Perform a non-dominated sort on all solutions in the population and divide them into distinct Pareto fronts: if solutions Better than solution ,but Dominant Solution And count the dominance of each solution:

[0169] The solution of level 1 is the first-level Pareto front; the remaining solutions are sorted again to obtain the second-level Pareto front.

[0170] In the candidate solutions of the Pareto front, the diversity of solutions is maintained by selecting reference points with a predefined distribution. The number of reference points (uniformly distributed in the solution space) is L:

[0171] ;

[0172] Where H represents the level of detail in the reference point division; M represents the number of targets;

[0173] When the algorithm iterates to the maximum number of algebras G or the solution converges, it outputs the Pareto optimal set, with each solution corresponding to an implant material formulation.

[0174] In this embodiment, based on the patient's anatomical features, reverse modeling technology is used to reconstruct a three-dimensional model of the patient's skin, soft tissue, and bones, as detailed below:

[0175] Based on the patient's anatomical characteristics, the U-Net deep learning framework was used to extract key anatomical information of the skin, soft tissue, and bone regions.

[0176] Based on key anatomical information, the two-dimensional segmented image data is converted into a three-dimensional geometric surface model using the reverse modeling algorithm Marching Cubes, and the generated surface mesh structure is optimized to obtain the final three-dimensional model.

[0177] In this embodiment, based on the patient's anatomical features, the U-Net deep learning framework is used to extract key anatomical information of the skin, soft tissue, and bone regions, as detailed below:

[0178] The system captures the surface morphology (contour characteristics, thickness, etc.) of the patient's skin using 3D scanning equipment, generates point cloud data of the skin, and converts it into a triangular mesh model.

[0179] Bone, soft tissue, and skin data were captured using T1 / T2 weighted MRI images, and the MRI images were then denoised and normalized in intensity.

[0180] Using the U-Net deep learning framework, skin, soft tissue, and bone are segmented from MRI images, the anatomical information of the segmented parts is extracted, and key region masks are generated.

[0181] The triangular mesh model and key region masks are registered to integrate the anatomical structures of soft tissue, bone, and skin, thereby obtaining key anatomical information for each region.

[0182] In this embodiment, S5 specifically refers to:

[0183] S51: Constructing the finite element model:

[0184] Based on the patient's three-dimensional anatomical model (bone, soft tissue, skin) obtained from S4, and the optimal implant material formula obtained from S3, the geometric model of the implant is obtained.

[0185] Boolean operations are performed between the implant geometry model and the patient's 3D anatomical model to define the contact interface and generate a finite element model. :

[0186] ;

[0187] in, To obtain the geometric model of the implant; These are three-dimensional anatomical models of bones, soft tissues, and skin, respectively.

[0188] S52: Mapping of Material Properties and Boundary Conditions

[0189] Based on the optimal formulation parameters, its mechanical properties are predicted through a multi-task learning model.

[0190] Biological tissue properties, with bone using an orthotropic elastic model (elastic modulus E) bone =10 GPa); soft tissue was modeled using a hyperelastic model (Mooney-Rivlin material parameters C10, C01);

[0191] Apply mechanical loads according to the patient's activity scenarios (such as chewing and walking), and define the contact relationship between the implant and surrounding tissues;

[0192] S53: Solving biomechanical equations:

[0193] The equilibrium equations for finite element analysis are:

[0194] Ku=F;

[0195] Where K is the global stiffness matrix (determined by material properties and mesh topology); u is the nodal displacement vector; and F is the external load vector.

[0196] The Newton-Raphson method is used to iteratively solve the nonlinear constitutive relations of implants or soft tissues:

[0197] ;

[0198] in, No. The tangent stiffness matrix of the next iteration; For displacement increment; This is the internal force vector; This is the external load vector;

[0199] Output stress / strain distribution, von Mises stress on implant surface:

[0200] ;

[0201] S54: Weighted calculation of biocompatibility score B:

[0202] Stress shielding effect: risk of bone resorption around the implant.

[0203] Inflammatory response: Concentration of inflammatory factors induced by interfacial fretting wear.

[0204] Mechanical stability: Whether the displacement of the implant exceeds the threshold.

[0205] ;

[0206] ;

[0207] ;

[0208] ;

[0209] Where w1, w2, and w3 are weighting coefficients; For stress shielding scoring; For wear rating, Assess implant displacement. For bone stress; The stress on bones in a healthy state; k is the wear attenuation coefficient; d slip u is the sliding distance of the contact surface; max The maximum displacement of the implant; u threshold Maximum permissible displacement limit.

[0210] An AI-based intelligent optimization system for medical aesthetic implant materials includes a processor, a memory, and a computer program stored in the memory. When the processor executes the computer program, it specifically performs the steps in the AI-based intelligent optimization method for medical aesthetic implant materials as described above.

[0211] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0212] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0213] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0214] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0215] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. An artificial intelligence-based intelligent optimization method for medical beauty implant materials, characterized in that, The method comprises the following steps: S1: Obtain the performance parameters of different medical implant materials, capture the patient's anatomical features using 3D scanning and MRI images, and count the patient's personalized biological data; S2: Construct a multi-objective supervised learning data set, construct a multi-task learning model and train based on the multi-objective supervised learning data set, the goal being to predict the performance of different implant material formulations in a medical implant environment; S3: Based on the predicted performance of different implant material formulations in a medical implant environment, use the NSGA-III algorithm for multi-objective optimization to obtain the optimal implant material formulation; S4: Based on the patient's anatomical features, use reverse modeling technology to reconstruct the three-dimensional model of the patient's skin, soft tissue, and bone; S5: Based on the three-dimensional model, according to the optimal implant material formulation and personalized biological data, use finite element analysis to simulate the patient's biomechanical properties, output the biocompatibility score and adaptability suggestion of the material in different implant environments.

2. The method of claim 1, wherein the method is based on artificial intelligence. The performance parameters of different medical implant materials are obtained through experimental data, literature data, and clinical data, as follows: Obtain experimental test data of the material, including tensile strength, elongation at break, and degradation rate; The literature data is collected from the database using web crawler technology related to medical materials; And use a pre-trained NLP model to extract performance data from the literature to obtain relevant performance indicators; Clinical data, by collecting historical implant material application cases, including material type and corresponding clinical feedback data; The obtained performance data is labeled and processed, with uniform units and ranges, forming a structured data format, and a material performance database is constructed, classified by material, performance, and parameter source. 3.The method of claim 1, wherein the method is characterized by, The patient's anatomical features are captured using 3D scanning and MRI images, and the patient's personalized biological data is counted, as follows: Capture the geometric features of the patient's skin surface, including thickness and contour characteristics, using a 3D scanning device; Collect the skin, soft tissue, and bone anatomical structure based on MRI images; Analyze the patient's medical history to determine the allergic reaction and immune system characteristics, and combine the patient's age, metabolism, and other individualized characteristics.

4. The method of claim 1, wherein the method is based on artificial intelligence. S2 is specifically: Extract input features including material performance data, patient personalized data, and target performance parameters; the material performance data includes material composition ratio and preparation process parameters; the patient's personalized data includes anatomical features and biological characteristics; The target performance parameters include elastic modulus E, biocompatibility score B, and durability D; Normalize the input features to eliminate dimensional differences and construct a multi-objective supervised learning data set; Combine the material data and personalized features in the form of a feature vector: X i =[C PLA ,C additive ,T heat ,P mold ,h skin ,ρ bone ,IR,MR] wherein, C PLA is the PLA content; C additive is the additive ratio; T heat is the heat treatment temperature; h skin is the thickness; p bone is the bone density; IR is the immune score; P mold is the molding pressure; MR is the metabolic score; The target value is stored as: Y i = [E, B, D]; Where E is the elastic modulus, B is the biocompatibility score, and D is the durability; The complete data set is: ； The multi-task learning model uses a multi-task learning framework to complete the prediction task through shared feature extraction layers and task-specific output layers, The shared bottom layer extracts a general representation through a shared hidden layer, captures the correlation between different tasks, improves data utilization efficiency, and converts input sample features into shared features H through a nonlinear function mapping; The task branch layer is used for processing task-specific characteristics and targets through independent neural network branches, and predicts each task separately.

5. The artificial intelligence-based medical beauty implant material intelligent optimization method according to claim 4, characterized in that, The multi-task learning model comprises an input layer, a shared hidden layer and a task branch layer, and specifically comprises the following: The input layer receives the feature vector X i ; The shared hidden layer extracts shared features through a fully connected layer: H = ReLU(W s X i +b s ); where W s is a weight matrix, b s is a bias term, m is the input feature dimension, and d is the hidden layer dimension; The task branch layer independently predicts target results through independent branches based on the shared features H; the task branch layer comprises an elastic modulus prediction branch, a biocompatibility prediction branch and a durability prediction branch: wherein, , and are weight coefficients; , and are biases; , and are the elastic modulus prediction, biocompatibility prediction, and durability prediction values, respectively. The loss function is ; wherein, , , , is a dynamically adjusted task weight; The weights of the shared layer and the task branch are updated through gradient descent: ； wherein, η is a learning rate, and t is an iteration number.

6. The artificial intelligence-based medical beauty implant material intelligent optimization method according to claim 5, characterized in that, According to the predicted performance of different implant material formulations in the medical beauty implant environment, a NSGA-III algorithm is used for multi-objective optimization to obtain an optimal implant material formulation, and specifically comprises the following: The optimization objectives include high elastic modulus f1 , high biocompatibility f2 , and high fatigue resistance f3 : Constraints include strength of material, degradation rate, and price constraints fracture strength limit g1 ): ； where σ min is the lower limit of the breaking strength; the breaking strength of the implant material; Degradation rate limiting g2 ): ； where r degradation represents the rate of degradation of the material, ranging from r min ≤ rd egradation ( )≤r max ; Cost range constraint g3 ): ； where: c represents the unit price of the material, c max is the upper limit of the budget cost; The final constraint problem is modeled as: Find to minimize F( )=[f1( ),f2( ),f3( )] subject to G( )≤0; Wherein, G( ) includes g1( ), g2( ) and g3( );NSGA-III algorithm is used for multi-objective optimization, and the optimal implant material formula is obtained: An initial population P of N candidate solutions is first generated, each sample ∈ P is randomly initialized; For each candidate solution , the preceding objective function values f1, f2, f3 are computed and it is checked whether the actual constraints are met, and if the solution violates the constraints, a constraint penalty function P( ) is defined: ； wherein is the jth constraint function; Non-dominated sorting is performed for all solutions in the population, dividing the different Pareto fronts: if solution is better than solution , then dominates solution , and the domination of each solution is counted: Among the candidate solutions on the Pareto front, the diversity of solutions is maintained by selecting reference points according to a predefined distribution, and the number of reference points is L: ； wherein, H is a refinement degree of the reference point division; and M is a target number; When the algorithm is iterated to a maximum number G or the solution converges, a Pareto optimal set is output, and each solution corresponds to an implant material formulation.

7. The method of claim 1, wherein the method is based on artificial intelligence. Based on the patient's anatomical features, a three-dimensional model of the patient's skin, soft tissue and bone is reconstructed using reverse modeling technology, and specifically comprises the following: Based on the patient's anatomical features, a U-Net deep learning framework is used to extract anatomical key information of the skin, soft tissue and bone regions; According to the anatomical key information, a two-dimensional segmentation image data is converted into a three-dimensional geometric surface model by a reverse modeling algorithm Marching Cubes, and the generated surface grid structure is optimized to obtain a final three-dimensional model.

8. The artificial intelligence-based medical beauty implant material intelligent optimization method according to claim 7, characterized in that, Based on the patient's anatomical features, a U-Net deep learning framework is used to extract anatomical key information of the skin, soft tissue and bone regions, and specifically comprises the following: According to the patient's skin surface morphology (contour characteristics, thickness, etc.) captured by the 3D scanning device, point cloud data of the skin is generated, and is converted into a triangular mesh model; T1 / T2 weighted MRI images are used to capture bone, soft tissue and skin data, and the MRI images are denoised and intensity normalized; A U-Net deep learning framework is used to segment the skin, soft tissue and bone in the MRI images, extract the segmented anatomical information, and generate key region masks; The triangular mesh model and the key region mask are registered, and the anatomical structures of the soft tissue, bone and skin are integrated to obtain the anatomical key information of each region.

9. The intelligent optimization method for medical aesthetic implant materials based on artificial intelligence according to claim 1, characterized in that, S5 specifically comprises: S51: constructing a finite element model: According to the patient's three-dimensional anatomical model obtained in S4, the optimal implant material formulation obtained in S3 is used to obtain an implant body geometric model; The implant geometry model is subjected to a Boolean operation with the three-dimensional anatomic model of the patient to define the contact interface and generate the finite element model : ； wherein, is a geometrical model of the implant; are three-dimensional anatomical models of bone, soft tissue, skin, respectively. S52: material attribute and boundary condition mapping: Based on the optimal formulation parameters, the mechanical properties are predicted through the multi-task learning model; wherein, η is a learning rate, and t is an iteration number. According to the predicted performance of different implant material formulations in the medical beauty implant environment, a NSGA-III algorithm is used for multi-objective optimization to obtain an optimal implant material formulation, and specifically comprises the following: The final constraint problem is modeled as: Among the candidate solutions on the Pareto front, the diversity of solutions is maintained by selecting reference points according to a predefined distribution, and the number of reference points is L: wherein, H is a refinement degree of the reference point division; and M is a target number; When the algorithm is iterated to a maximum number G or the solution converges, a Pareto optimal set is output, and each solution corresponds to an implant material formulation. Based on the patient's anatomical features, a three-dimensional model of the patient's skin, soft tissue and bone is reconstructed using reverse modeling technology, and specifically comprises the following: Based on the patient's anatomical features, a U-Net deep learning framework is used to extract anatomical key information of the skin, soft tissue and bone regions; According to the anatomical key information, a two-dimensional segmentation image data is converted into a three-dimensional geometric surface model by a reverse modeling algorithm Marching Cubes, and the generated surface grid structure is optimized to obtain a final three-dimensional model. Based on the patient's anatomical features, a U-Net deep learning framework is used to extract anatomical key information of the skin, soft tissue and bone regions, and specifically comprises the following: According to the patient's skin surface morphology (contour characteristics, thickness, etc.) captured by the 3D scanning device, point cloud data of the skin is generated, and is converted into a triangular mesh model; T1 / T2 weighted MRI images are used to capture bone, soft tissue and skin data, and the MRI images are denoised and intensity normalized; A U-Net deep learning framework is used to segment the skin, soft tissue and bone in the MRI images, extract the segmented anatomical information, and generate key region masks; The triangular mesh model and the key region mask are registered, and the anatomical structures of the soft tissue, bone and skin are integrated to obtain the anatomical key information of each region. S5 specifically comprises: S51: constructing a finite element model: According to the patient's three-dimensional anatomical model obtained in S4, the optimal implant material formulation obtained in S3 is used to obtain an implant body geometric model; S52: material attribute and boundary condition mapping: Based on the optimal formulation parameters, the mechanical properties are predicted through the multi-task learning model; Biological tissue properties, in which the bone adopts an orthotropic elastic model; the soft tissue adopts a hyperelastic model; Apply mechanical load according to patient activity scene, and define the contact relationship between the implant and the surrounding tissue; S53: Solve the biomechanics equation: The balance equation of finite element analysis is: Ku=F; Wherein, K is the global stiffness matrix; u is the node displacement vector; F is the external load vector; The Newton-Raphson method is used to iteratively solve the nonlinear constitutive relation of the implant or soft tissue: ； wherein, the first tangent stiffness matrix of the iteration; is the displacement increment; is the internal force vector; is the external load vector; Output stress / strain distribution, von Mises stress on the surface of the implant: ； S54: Weighted calculation of biocompatibility score B: ； ； ； ； wherein w1, w2, w3 are weight coefficients; stress shield score; wear score, implant displacement score; skeletal stress; skeletal stress in healthy state; k is wear attenuation coefficient; d slip contact surface slip distance; u max maximum displacement of the implant; u threshold allowed maximum displacement limit.

10. An artificial intelligence-based intelligent optimization system for medical beauty implant materials, characterized in that, The computer program product comprises a processor, a memory, and a computer program stored on the memory, and when the processor executes the computer program, the steps of the artificial intelligence-based medical and cosmetic implant material intelligent optimization method according to any one of claims 1-9 are specifically executed.

Images