A personalized orthopedic insole design and rapid mechanical simulation method

Abstract

This invention belongs to the fields of digital medicine and medical image processing technology, and discloses a method for personalized orthotic insole design and rapid mechanical simulation. It constructs a statistical shape model of the foot epidermis and a statistical shape model of the orthotic insole using generalized polymorphic analysis and principal component analysis. A conditional Gaussian distribution is used to correlate the corresponding shape coefficient matrices in the two statistical shape models. By matching the foot epidermis model of the applicant to the statistical shape model of the foot epidermis, the shape coefficients of the foot epidermis model are calculated. Then, the shape coefficients of the corresponding orthotic insole are calculated using the conditional Gaussian model, generating the corresponding orthotic insole model from the statistical shape model of the orthotic insole. By adjusting the arch and forefoot regions in the orthotic insole model, multiple mechanical simulation results are generated, and a mechanical simulation library is constructed. This invention achieves real-time rapid mechanical simulation and allows for further adjustments to the insole model based on the mechanical simulation results until the optimal result is generated.

Description

Technical Field

[0001] This invention relates to the fields of digital medicine and medical image processing technology, specifically to a personalized orthopedic insole design and rapid mechanical simulation method. Background Technology

[0002] The arch of the foot is a unique structure that evolved in the human body. It cushions the powerful impact generated during upright walking and running, helping humans adapt to various terrains such as rocks and sand, playing a vital role in shock absorption and protection. However, with industrialization, human living conditions have greatly improved, and average human weight has increased significantly, while the corresponding amount of physical activity has decreased drastically. This has resulted in insufficient exercise for the muscles and tendons of the arch, while the load on the arch has continuously increased. Simultaneously, humans have transitioned from wilderness survival to urban living, with the ground environment shifting to flat surfaces such as asphalt roads, floorboards, and brick pavements. The lack of exercise for the arch directly impacts its development.

[0003] For the reasons mentioned above, the incidence of flat feet among Chinese adolescents has been steadily increasing in recent years. It often manifests as collapse of the medial longitudinal arch, abduction of the talus joint, and eversion of the subtalar joint. The arch disappears under weight-bearing conditions, leading to foot fatigue and even pain after standing or walking. In its early stages, flat feet can cause gait abnormalities. If left untreated for a long time, it can lead to serious problems such as pelvic tilt, scoliosis, lower back pain, and asymmetrical musculoskeletal development, causing irreversible damage to the body. Therefore, early detection and treatment of flat feet are crucial.

[0004] Treatment for flat feet primarily involves orthotic insoles. These insoles provide support to the forefoot, midfoot, and posterior foot, altering the foot's biomechanics and changing the distribution of plantar pressure to reduce stress and restore arch height. Currently, most orthotic insoles on the market are prefabricated. These are mass-produced using factory-designed molds, resulting in standardized, template-like shapes that often fail to meet the individual's needs. Each person's foot shape is unique, requiring different supports, sometimes even with the opposite effect. Current hospital-made custom orthotic insoles are based on the patient's external foot shape and plantar pressure tests. They are generated using foreign insole design software and then modified based on the doctor's personal experience. This process is lengthy, has a significant shortage of experienced doctors, and demands a high level of expertise from physicians. Summary of the Invention

[0005] To address the current problems of lengthy design time, low production efficiency, and a large shortage of doctors required for orthopedic insole design, this invention provides a personalized orthopedic insole design and rapid mechanical simulation method. This method can conveniently and quickly provide patients with matching orthopedic insole models for treatment, significantly improving the efficiency of orthopedic insole design, reducing the workload of orthotists, and enabling personalized design of matching orthopedic insole models for patients and rapid mechanical simulation.

[0006] The technical solution of this invention is as follows: A method for designing personalized orthotic insoles and rapid mechanical simulation, comprising the following steps:

[0007] S1. Align the foot epidermal model and the orthotic insole model in spatial position; the two should conform to the state when worn normally in spatial position;

[0008] S2. Select any foot skin model from the foot skin model sample dataset as the foot skin template model, and normalize the remaining foot skin models; extract the point cloud data from the foot skin template model and the remaining foot skin models separately as foot template point cloud models and foot sample point cloud models; match the foot skin template point cloud model with the remaining foot skin sample point cloud models to obtain the three-dimensional coordinate set of all points matched from the template point cloud model to each sample point cloud model; use generalized Protodyakonov analysis to construct a shape vector for each sample point cloud model, use principal component analysis to extract the main shape vectors from the shape vectors, and construct a statistical shape model of the foot skin using the main shape vectors;

[0009] From the orthotic insole model sample dataset, any one orthotic insole model is selected as the orthotic insole template model, and the remaining orthotic insole models are normalized. The point cloud data of the orthotic insole template model and the remaining orthotic insole models are extracted separately as orthotic insole template point cloud models and orthotic insole sample point cloud models. The orthotic insole template point cloud model and the remaining orthotic insole sample point cloud models are matched to obtain the three-dimensional coordinate set of all points matched from the template point cloud model to each sample point cloud model. Generalized Protodyakonov analysis is used to construct a shape vector for each sample point cloud model, and principal component analysis is used to extract the main shape vectors from the shape vectors. The orthotic insole statistical shape model is constructed through the main shape vectors.

[0010] S3. Match the statistical shape model of the foot skin to the foot skin model of the candidate, and set the number of deformation parameters in the statistical shape model of the foot skin; based on the shape feature matrix in the statistical shape model of the foot skin and the shape vector of the deformation of the statistical shape model of the foot skin obtained when matching the statistical shape model of the foot skin to the foot skin model of the candidate, use the least squares method to perform linear regression fitting to obtain the shape coefficient of the foot skin model of the candidate in the shape feature matrix;

[0011] S4. Construct a conditional Gaussian model for the statistical shape model of the foot skin and the statistical shape model of the orthotic insole; input the shape coefficients of the foot skin model of the candidate to be designed in the statistical shape feature matrix into the calculation formula of the conditional Gaussian model to obtain the shape coefficients of the orthotic insole model of the candidate to be designed in the statistical shape model of the orthotic insole that match the foot skin model of the candidate to be designed; multiply the shape coefficients of the orthotic insole model of the candidate to be designed in the statistical shape model of the orthotic insole with the shape feature matrix of the statistical shape model of the orthotic insole to obtain the point cloud model of the orthotic insole of the candidate to be designed;

[0012] S5. Generate a morphological template library of foot + insole based on the statistical shape model of foot epidermis and the statistical shape model of orthopedic insole; for each foot + insole instance, adjust the arch and forefoot parts of the insole, and perform mechanical simulation using the finite element method for each combination of different foot shapes and insole protrusions; the stress and strain results contained in each set of results constitute a mechanical simulation library containing different foot shapes and different insole protrusions;

[0013] S6. Adjust the arch and forefoot regions of the orthopedic insole point cloud model obtained in step S4; in the mechanical simulation library obtained in step S5, perform interpolation calculations based on the shape coefficient of the foot epidermal model of the candidate in step S4 in the statistical shape feature matrix and the degree of protrusion of the orthopedic insole at the arch and forefoot regions to obtain the stress and strain values ​​of each surface vertex on the foot epidermal model, and obtain the final mechanical simulation results.

[0014] The normalization process for the remaining foot skin models involves aligning the centroid of the foot skin template model with the centroids of the remaining foot skin models; the normalization process for the remaining orthotic insole models involves aligning the centroid of the orthotic insole template model with the centroids of the remaining orthotic insole models.

[0015] Specifically, obtaining the set of three-dimensional coordinates of all points matched from the template point cloud model to each sample point cloud model involves matching the template point cloud model with the remaining sample point cloud models using an annealing algorithm. The stopping condition for the annealing algorithm is set to the Euclidean distance between the two models being less than a set threshold. The deformation coefficients matched from the template point cloud model to the remaining sample point cloud models are calculated, namely the affine transformation parameter d, the non-affine transformation parameter w, and the Euclidean distance PHI between the two models. The template point cloud model is then matched one by one to all sample point cloud models to obtain the set of three-dimensional coordinates of all points matched from the template point cloud model to each sample point cloud model.

[0016] V x =V0·d+PHI·w

[0017] Where V0 is the template point cloud model, V x These are the point cloud models of the remaining samples.

[0018] The specific steps for constructing the statistical shape model of the foot epidermis are as follows: For each sample point cloud model, construct a shape vector, let s... i =[p i ,l i ]∈R 3N Let p be the shape vector of the sample point cloud model i, where p i ∈R 3N l is the set of 3D coordinates of all points obtained after matching the template point cloud model with the sample point cloud model i. i ∈R 3K This is the set of three-dimensional coordinates of all anatomical calibration points in the sample point cloud model; using the shape vector s i Before performing statistical modeling, the shape vectors s of all sample point cloud models were analyzed using the generalized Protodyakonov analysis method. i Perform alignment processing;

[0019] For all aligned shape vectors s i Principal component analysis was used to model the shape changes, resulting in the statistical shape model of the foot epidermis, as shown below:

[0020]

[0021] in, Φ represents the average shape of the sample point cloud model dataset. s =R )(N+K)×M This is the feature vector matrix obtained after performing principal component analysis on the sample point cloud model dataset, where the M feature vectors represent the M deformation patterns learned by the statistical shape model of the foot epidermis, and M is the number of sample point cloud models; b s ∈R M b is a shape parameter s The M elements in the data represent Φ respectively. s The weights of the M deformation patterns superimposed on the average shape; s∈R 3(N+K) The shape of the foot epidermis is represented by the statistical shape model, s is affected by b s Control, by adjusting b s The value controls the deformation of the foot statistical shape model.

[0022] The specific steps of matching the statistical shape model of the foot skin to the foot skin model of the candidate are as follows: Point cloud data from the foot skin model of the candidate is extracted as the foot morphology point cloud model of the candidate. An annealing algorithm is used to match the statistical shape model of the foot skin and the foot morphology point cloud model of the candidate. The stopping condition for the annealing algorithm is that the Euclidean distance between the two is less than a set threshold. The deformation coefficients from the statistical shape model of the foot skin to the foot skin point cloud model of the candidate are calculated, namely the affine transformation parameter d, the non-affine transformation parameter w, and the Euclidean distance PHI. The statistical shape model of the foot skin is then matched to the foot skin point cloud model of the candidate, obtaining the three-dimensional coordinate set of all points from the statistical shape model of the foot skin to the foot skin point cloud model of the candidate.

[0023] The number of deformation parameters in the statistical shape model of the foot skin is set. Based on the shape feature matrix extracted from the statistical shape model of the foot skin and the shape vector of the deformation of the statistical shape model of the foot skin after matching the statistical shape model of the foot skin to the foot skin model of the designer, the shape parameters of the foot skin model are obtained by linear regression fitting using the least squares method.

[0024] The shape parameter b s The input formula for the conditional Gaussian model is as follows:

[0025]

[0026]

[0027] Where μ S μ is the mean of the shape factor S of the orthopedic insole model. F Let F be the mean of the shape coefficient F of the foot epidermal model, ∑ S ∑ is the variance of the shape factor S of the orthopedic insole model; F Let F be the variance of the shape coefficient F of the foot epidermal model, ∑ SF and ∑ FS It is the covariance matrix of the orthopedic insole model shape coefficient S and the foot epidermal model shape coefficient F. s To match the statistical shape model of the foot skin to the shape parameters calculated from the foot skin model of the designer.

[0028] The orthopedic insole point cloud model of the patient obtained in adjustment step S4 is implemented by designing smooth raised portions for the arch and forefoot areas; the height of the raised portions is a function of the two-dimensional coordinates of the vertex of the insole surface, described by a two-dimensional Gaussian function, as shown below:

[0029]

[0030] Where x and y are the horizontal and vertical coordinates of the vertices on the surface of the orthopedic insole point cloud model, σ x σ y The influence range of the protrusion is controlled, which is represented as the width in the x or y direction in the point cloud model of the orthopedic insole. x0 and y0 represent the coordinates of the center point of the protrusion, and the coefficient η controls the amplitude of the protrusion. The corresponding points are subjected to two-dimensional Gaussian operation to change the shape of the part and make detailed position adjustments, which are used to generate multiple results in the later mechanical simulation.

[0031] The beneficial effects of this invention are as follows: By utilizing existing models of the patient's foot skin and orthotic insoles, statistical shape models of both models are constructed. The patient's foot skin model is then matched to these statistical shape models to generate the corresponding orthotic insole model. Based on the generated orthotic insole model, multiple mechanical simulation results are generated, and a mechanical simulation library is constructed. The model with the best treatment effect and wearing comfort is selected as the final result. This invention achieves automatic, efficient, and personalized orthotic insole design solutions for patients, significantly reducing the design time for orthotists and improving the design efficiency of orthotic insoles. Attached Figure Description

[0032] Figure 1 This is a flowchart illustrating the overall workflow of a personalized orthotic insole design and rapid mechanical simulation method.

[0033] Figure 2 These are detailed adjustment display images;

[0034] Figure 3(a) shows the results;

[0035] Figure 3(b) is a top-view representation of the corresponding instance of Figure 3(a). Detailed Implementation

[0036] To address the current problems of lengthy design time, low production efficiency, and a large shortage of doctors required for orthopedic insole design, this invention provides a personalized orthopedic insole design and rapid mechanical simulation method. This method enables an automated, efficient, and personalized orthopedic insole design process, and can conveniently and quickly provide matching orthopedic insoles for patients to receive treatment, greatly improving treatment efficiency.

[0037] This invention proposes a personalized orthotic insole design and rapid mechanical simulation method, which enables an automated, efficient, and personalized orthotic insole design process. Based on a large number of patient foot models and corresponding orthotic insole models, statistical shape models of the foot epidermis and orthotic insoles are constructed. A conditional Gaussian model is applied to the two statistical shape models, and the patient's foot epidermis model is matched to the statistical shape model of the foot epidermis to obtain the orthotic insole model. Details of the generated orthotic insole model are adjusted, generating multiple mechanical simulation results and constructing a mechanical simulation library. Real-time rapid mechanical simulation is achieved, and the orthotist selects the optimal simulation result.

[0038] First, a model of the patient's foot skin and a matching orthotic insole model are collected. Next, generalized Probabilistic Analysis (GPA) and Principal Component Analysis (PCA) are used to construct statistical shape models of the foot skin and orthotic insoles, respectively. A conditional Gaussian distribution is used to correlate the corresponding shape coefficient matrices in the two statistical shape models. By matching the patient's foot skin model to the statistical shape model, the shape coefficients of the foot skin model are calculated. Then, the shape coefficients of the corresponding orthotic insole are calculated using the conditional Gaussian model, generating the corresponding orthotic insole model from the statistical shape model. Based on the generated orthotic insole model, multiple mechanical simulation results are generated by adjusting the arch support and the forefoot support, constructing a mechanical simulation library. Orthotists can adjust the insole model, call the mechanical simulation library, and perform real-time, rapid mechanical simulations. They can then further adjust the insole model based on the simulation results until the optimal result is achieved. The specific steps are as follows:

[0039] S11. Data Preprocessing. Align the foot epidermal model and the corresponding orthotic insole model in the sample data in spatial position, so that their spatial positions conform to the state when worn normally.

[0040] S12. Construct a statistical shape model of the foot epidermis;

[0041] S121. Foot skin model normalization. Any data point from the foot skin model sample data is used as the foot skin template model, and then the centroid of the template model is calculated. Then, the centroids of the remaining sample data in the foot skin model sample data are calculated. Based on the centroids of the remaining sample data models and the centroid of the foot skin template model, the remaining sample data models are translated so that their centroids are aligned with the centroid of the foot skin template model.

[0042] S122. Foot Skin Model Matching. Point cloud data from the foot skin model is extracted to form the foot skin point cloud model. An annealing algorithm is used to match the template point cloud model with the remaining sample point cloud models. The stopping condition for the annealing algorithm is set to the Euclidean distance between the two models being less than a set threshold. The deformation coefficients from the template point cloud model to the remaining sample point cloud models are calculated, representing the affine transformation parameter d, the non-affine transformation parameter w, and the Euclidean distance PHI. This is achieved using the function shown below:

[0043] V x =V0·d+PHI·w

[0044] Where V0 is the template point cloud model, V x These are the remaining sample point cloud models. By matching the foot skin template point cloud model one by one to all the sample point cloud models, we can obtain the set of three-dimensional coordinates of all points matched from the foot skin template point cloud model to each sample point cloud model.

[0045] S123, Generalized Protodyakonov Analysis. Construct a shape vector for each sample point cloud model, let s... i =[p i ,l i ]∈R 3N Let p be the shape vector of the sample point cloud model i, where p i ∈R 3N The set of three-dimensional coordinates of all points obtained after matching the foot skin template point cloud model with the sample point cloud model i. i ∈R 3K This is the set of three-dimensional coordinates of all anatomical calibration points in the sample point cloud model. When using the shape vector s... i Before performing statistical modeling, the shape vectors s of all sample point cloud models were analyzed using the generalized Protodyakonov analysis method. i Perform alignment processing.

[0046] S124. Principal Component Analysis. For all aligned shape vectors s... i Principal component analysis was used to model the shape changes, and the resulting statistical shape model is shown below:

[0047]

[0048] in, Φ represents the average shape of the sample model dataset. s =R 3(N+K)×M This is the feature vector matrix (subscript s denotes shape) obtained after principal component analysis of the sample model dataset, where M feature vectors represent the M deformation patterns learned by the statistical shape model, and M is the number of sample models; b s ∈R M b is a shape parameter sThe M elements in the data represent Φ respectively. s The weights of the M deformation patterns superimposed on the average shape. s∈R 3(N+K) The shape of the statistical shape model is represented by s, which is subject to b. s Control, by adjusting b s The value is used to control the deformation of the statistical shape model.

[0049] S13. Construct a statistical shape model for orthopedic insoles;

[0050] S131. Orthopedic Insole Model Normalization. Any data point in the orthopedic insole model sample data is used as the orthopedic insole template model. Then, the centroid of the orthopedic insole template model is calculated. Next, the centroids of the remaining sample data in the orthopedic insole model sample data are calculated. Based on the centroids of the remaining sample data and the centroid of the orthopedic insole template model, the remaining sample data are translated so that the centroids of the remaining sample data models are aligned with the centroids of the orthopedic insole template model.

[0051] S132. Orthopedic Insole Model Matching. Point cloud data from the orthopedic insole model is extracted as the orthopedic insole point cloud model. An annealing algorithm is used to match the template point cloud model with the remaining sample point cloud models. The stopping condition for the annealing algorithm is set to the Euclidean distance between the two models being less than a set threshold. The deformation coefficients from the template point cloud model to the remaining sample point cloud models are calculated, representing the affine transformation parameter d, the non-affine transformation parameter w, and the Euclidean distance PHI. This is achieved using the function shown below:

[0052] V x =V0·d+PHI·w

[0053] Where V0 is the template point cloud model, V x These are the remaining sample point cloud models. By matching the template point cloud model to each of the sample point cloud models one by one, we can obtain the set of three-dimensional coordinates of all points matched from the template point cloud model to each sample point cloud model.

[0054] S133, Generalized Protodyakonov Analysis. Construct a shape vector for each sample point cloud model, let s... i =[pi,l i ]∈R 3N Let p be the shape vector of the sample point cloud model i, where p i ∈R 3N l is the set of 3D coordinates of all points obtained after matching the template point cloud model with the sample point cloud model i. i ∈R 3K This is the set of three-dimensional coordinates of all anatomical calibration points in the sample point cloud model. When using the shape vector s... i Before performing statistical modeling, the shape vectors s of all sample point cloud models were analyzed using the generalized Protodyakonov analysis method. iPerform alignment processing.

[0055] S134, Principal Component Analysis. For all aligned shape vectors s... i Principal component analysis was used to model the shape changes, and the resulting statistical shape model is shown below:

[0056]

[0057] in, Φ represents the average shape of the sample model dataset. S =R 3(N+K)×M This is the feature vector matrix (subscript s denotes shape) obtained after principal component analysis of the sample model dataset, where M feature vectors represent the M deformation patterns learned by the statistical shape model, and M is the number of sample models; b s ∈R M b is a shape parameter s The M elements in the data represent Φ respectively. s The weights of the M deformation patterns superimposed on the average shape. s∈R 3(N+K) The shape of the statistical shape model is represented by s, which is subject to b. s Control, by adjusting b s The value is used to control the deformation of the statistical shape model.

[0058] S14. Foot Skin Statistical Shape Model Matching. The statistical shape model of the foot skin is matched to the foot skin model of the candidate to obtain the shape coefficients of the candidate's foot skin model in the statistical shape feature matrix. The specific method is as follows:

[0059] S141. Matching the Foot Skin Model of the Subject to be Designed. The point cloud data from the foot skin model of the subject to be designed is extracted as the foot skin point cloud model. An annealing algorithm is used to match the average shape point cloud model and the foot skin point cloud model of the subject to be designed. The stopping condition for the annealing algorithm is set to the Euclidean distance between the two models being less than a set threshold. The deformation coefficients from the average shape point cloud model to the foot skin point cloud model of the subject to be designed are calculated, namely the affine transformation parameter d, the non-affine transformation parameter w, and the Euclidean distance PHI. The following function is used:

[0060] V x =V0·d+PHI·w

[0061] Where V0 is the average shape point cloud model, V x This is the point cloud model of the foot skin of the person to be designed. By matching the average shape point cloud model to the point cloud model of the foot skin of the person to be designed, the set of three-dimensional coordinates of all points matched from the average shape point cloud model to the point cloud model of the foot skin of the person to be designed can be obtained.

[0062] S142. Set the deformation parameter b in the statistical shape model of the foot epidermis. s The number of [elements]. The deformation parameter b can be defined directly. s The number of values ​​can also be selected, or deformation parameter b can be chosen to retain more than 95% of the data information. s The number depends on the deformation parameter b. s The number of elements is obtained by extracting the corresponding statistical shape feature matrix Φ from the statistical shape model. s .

[0063] S143. Calculate the personalized shape parameter b s Based on the shape feature matrix Φ extracted from the statistical shape model. s And the shape vector of the hindfoot epidermal model deformation was matched, and linear regression fitting was performed using the least squares method to calculate the linear equation system Φ. s b s =s p The shape parameters b of the foot epidermal model are obtained. s , where s p Φ represents the displacement deviation between the shape vector of the foot skin of the person being designed and the average shape vector. s This represents the shape feature matrix contained in the shape model within the statistical model.

[0064] S15. Constructing a conditional Gaussian model for the foot epidermis model and orthotic insole model;

[0065] S151. Establish a correlation calculation formula for the shape feature matrix of the statistical shape model of the foot epidermis and the statistical shape model of the orthotic insole, using a conditional Gaussian model, as shown in the following formula:

[0066] P(S|F)=N(μ,∑)

[0067] Where S and F represent the shape feature matrices of the orthopedic insole model and the foot epidermal model statistical shape model, respectively, μ is the conditional mean of the Gaussian model, and ∑ is the conditional covariance of the Gaussian model.

[0068] S152, The shape parameter b calculated by matching the statistical shape model of the foot skin to the foot skin model of the designer. s The formula for calculating the Gaussian model with applied conditions.

[0069]

[0070]

[0071] Where μ S μ is the mean of the shape factor S of the orthopedic insole model. F Let F be the mean of the shape coefficient F of the foot epidermal model, ∑ S∑ is the variance of the shape factor S of the orthopedic insole model; F Let F be the variance of the shape coefficient F of the foot epidermal model, ∑ SF and ∑ FS It is the covariance matrix of the orthopedic insole model shape coefficient S and the foot epidermal model shape coefficient F. s To match the statistical shape model of the foot skin to the shape parameters calculated from the foot skin model of the designer.

[0072] The calculated μ and ∑ are the Gaussian model coefficients of the orthopedic insole model of the candidate, which are matched with the candidate's foot epidermal model. Substituting them into the Gaussian model, the point cloud model of the candidate's orthopedic insole is obtained.

[0073] S16. Detail Adjustments. A smooth raised section is added to the corresponding position of the forefoot depression in the orthotic insole model to distribute pressure on the forefoot and reduce foot burden. A height design is added to the arch area of ​​the orthotic insole model to support the arch and restrain the foot. The height of the raised section is a function of the two-dimensional coordinates of the insole surface vertex, described using a two-dimensional Gaussian function, as shown below:

[0074]

[0075] Where x and y are the horizontal and vertical coordinates of the vertices of the orthopedic insole model surface, and σ x σ y The influence range of the protrusion is controlled, represented by its width in the x or y direction in the orthopedic insole model. x0 and y0 represent the coordinates of the center point of the protrusion, and the coefficient η controls the amplitude of the protrusion. Two-dimensional Gaussian operations are performed on the corresponding points to change the shape of the part and adjust its detailed position, which is used to generate multiple results in subsequent mechanical simulations. (Appendix) Figure 2 The shapes of the protrusions at the forefoot and arch are shown. Since there are two protrusions, η is a two-dimensional vector. To control the degree of protrusion, each dimension of η can be adjusted within the range [0,1] to achieve different degrees of local protrusion adjustment in the insole.

[0076] S17. Construct a mechanics simulation library.

[0077] S171. Generate a morphological template library for the foot and insole based on the statistical shape model of the foot epidermis and the statistical shape model of orthotic insoles. The shape coefficients b of the two statistical shape models... s Perform uniform sampling: Assume there are k shape coefficients, then b s Let there be k-dimensional vectors, spanning a k-dimensional shape space. For this shape space, sample n times at equal intervals along each dimension within the shape coefficient interval [-2, 2]. Then the entire space can be sampled n times. k Using a sample and substituting it into the statistical shape model formula, generate n... kExamples of foot and insole models with different foot shapes.

[0078] For each model instance, fine-tune the local protrusions of the insole. Adjust the degree of protrusion at the center of the forefoot and the medial arch. The degree of these two protrusions is uniformly sampled in the range [0,1]. Each coefficient is sampled t times, resulting in a total sampling of 2 times for the two protrusion coefficients. t One sample.

[0079] There are n combinations of different foot shapes and insole protrusion levels. k 2 t For these n k 2 t Each combination was mechanically simulated using the finite element method to obtain n. k 2 t The mechanical simulation results are collected, and the stress and strain data contained in each set of results are saved to form a mechanical simulation library that includes different foot shapes and different insole protrusions.

[0080] S18. Based on the foot surface scan data of the candidate, calculate the shape coefficient b that conforms to the shape of the foot epidermis of the candidate. s And an orthopedic insole model, then apply 2 to the insole t The sampling value of the convexity coefficient η is used to obtain the 2 corresponding to the designer. t Examples of insole models. For each foot + insole model instance, its shape is represented by a k+2 dimensional shape coefficient [b]. s The shape coefficient is used to describe the stress and strain values ​​of each surface vertex on the designed foot skin and orthopedic insole model. The interpolation is performed in the k+2 space where the mechanical simulation library is located, and the interpolation object is the stress and strain values ​​of each surface vertex on the designed foot skin and orthopedic insole model.

Claims

1. A personalized orthotic insole design and rapid mechanical simulation method, characterized in that, The steps include the following: S1. Align the foot epidermal model and the orthotic insole model in spatial position; the two should conform to the state when worn normally in spatial position; S2. Select any foot skin model from the foot skin model sample dataset as the foot skin template model, and normalize the remaining foot skin models; extract the point cloud data from the foot skin template model and the remaining foot skin models separately as foot template point cloud models and foot sample point cloud models; match the foot skin template point cloud model with the remaining foot skin sample point cloud models to obtain the three-dimensional coordinate set of all points matched from the template point cloud model to each sample point cloud model; use generalized Protodyakonov analysis to construct a shape vector for each sample point cloud model, use principal component analysis to extract the main shape vectors from the shape vectors, and construct a statistical shape model of the foot skin using the main shape vectors; From the orthotic insole model sample dataset, any one orthotic insole model is selected as the orthotic insole template model, and the remaining orthotic insole models are normalized. The point cloud data of the orthotic insole template model and the remaining orthotic insole models are extracted separately as orthotic insole template point cloud models and orthotic insole sample point cloud models. The orthotic insole template point cloud models and the remaining orthotic insole sample point cloud models are matched to obtain the three-dimensional coordinate set of all points matched from the template point cloud model to each sample point cloud model. Generalized Protodyakonov analysis is used to construct a shape vector for each sample point cloud model, and principal component analysis is used to extract the main shape vectors from the shape vectors. The orthotic insole statistical shape model is constructed through the main shape vectors. S3. Match the statistical shape model of the foot skin to the foot skin model of the designer, and set the number of deformation parameters in the statistical shape model of the foot skin; Based on the shape feature matrix in the statistical shape model of the foot skin and the shape vector of the deformed statistical shape model of the foot skin obtained when matching the statistical shape model of the foot skin to the foot skin model of the candidate, the shape coefficient of the foot skin model of the candidate in the shape feature matrix is ​​obtained by linear regression fitting using the least squares method. S4. Construct a conditional Gaussian model for the statistical shape model of the foot epidermis and the statistical shape model of the orthotic insole; The shape coefficients of the foot epidermal model of the candidate to be designed in the statistical shape feature matrix are input into the conditional Gaussian model calculation formula to obtain the shape coefficients of the orthopedic insole model of the candidate to be designed in the statistical shape model of the orthopedic insole that match the foot epidermal model of the candidate to be designed. The shape coefficients of the orthopedic insole model of the candidate to be designed in the statistical shape model of the orthopedic insole are multiplied with the shape feature matrix of the statistical shape model of the orthopedic insole to obtain the point cloud model of the orthopedic insole of the candidate to be designed. S5. Generate a morphological template library of foot + insole based on the statistical shape model of foot epidermis and the statistical shape model of orthopedic insole; for each foot + insole instance, adjust the arch and forefoot parts of the insole, and perform mechanical simulation using the finite element method for each combination of different foot shapes and insole protrusions; the stress and strain results contained in each set of results constitute a mechanical simulation library containing different foot shapes and different insole protrusions; S6. Adjust the arch and forefoot regions of the orthopedic insole point cloud model obtained in step S4; in the mechanical simulation library obtained in step S5, perform interpolation calculations based on the shape coefficient of the foot epidermal model of the candidate in step S4 in the statistical shape feature matrix and the degree of protrusion of the orthopedic insole model of the candidate in the arch and forefoot regions to obtain the stress and strain values ​​of each surface vertex on the foot epidermal model, and obtain the final mechanical simulation results.

2. The personalized orthotic insole design and rapid mechanical simulation method according to claim 1, characterized in that, The normalization process for the remaining foot skin models involves aligning the centroid of the foot skin template model with the centroids of the remaining foot skin models; the normalization process for the remaining orthotic insole models involves aligning the centroid of the orthotic insole template model with the centroids of the remaining orthotic insole models.

3. The personalized orthotic insole design and rapid mechanical simulation method according to claim 2, characterized in that, Specifically, obtaining the set of three-dimensional coordinates of all points matched from the template point cloud model to each sample point cloud model involves matching the template point cloud model with the remaining sample point cloud models using an annealing algorithm. The stopping condition for the annealing algorithm is set to the Euclidean distance between the two being less than a set threshold. The deformation coefficients matched from the template point cloud model to the remaining sample point cloud models are calculated, namely the affine transformation parameter d, the non-affine transformation parameter w, and the Euclidean distance PHI between the two. The template point cloud model is then matched one by one to all sample point cloud models to obtain the set of three-dimensional coordinates of all points matched from the template point cloud model to each sample point cloud model. V x =V0·d+PHI·w Where V0 is the template point cloud model, V x These are the point cloud models of the remaining samples.

4. The personalized orthotic insole design and rapid mechanical simulation method according to claim 3, characterized in that, The specific steps for constructing the statistical shape model of the foot epidermis are as follows: For each sample point cloud model, construct a shape vector, let s... i =[p i , l i ]∈R 3N Let p be the shape vector of the sample point cloud model i, where p i ∈R 3N l is the set of 3D coordinates of all points obtained after matching the template point cloud model with the sample point cloud model i. i ∈R 3K This is the set of three-dimensional coordinates of all anatomical calibration points in the sample point cloud model; using the shape vector s i Before performing statistical modeling, the shape vectors s of all sample point cloud models were analyzed using the generalized Protodyakonov analysis method. i Perform alignment processing; For all aligned shape vectors s i Principal component analysis was used to model the shape changes, resulting in the statistical shape model of the foot epidermis, as shown below: in, Φ represents the average shape of the sample point cloud model dataset. s =R 3(N+K)×M This is the feature vector matrix obtained after performing principal component analysis on the sample point cloud model dataset, where the M feature vectors represent the M deformation patterns learned by the statistical shape model of the foot epidermis, and M is the number of sample point cloud models; b s ∈R M b is a shape parameter s The M elements in the data represent Φ respectively. s The weights of the M deformation patterns superimposed on the average shape; s∈R 3(N+K) The shape of the foot epidermis is represented by the statistical shape model, s is affected by b s Control, by adjusting b s The value controls the deformation of the foot statistical shape model.

5. The personalized orthotic insole design and rapid mechanical simulation method according to claim 4, characterized in that, The specific steps of matching the statistical shape model of the foot skin to the foot skin model of the candidate are as follows: Point cloud data from the foot skin model of the candidate is extracted as the foot morphology point cloud model of the candidate. An annealing algorithm is used to match the statistical shape model of the foot skin and the foot morphology point cloud model of the candidate. The stopping condition for the annealing algorithm is that the Euclidean distance between the two is less than a set threshold. The deformation coefficients from the statistical shape model of the foot skin to the foot skin point cloud model of the candidate are calculated, namely the affine transformation parameter d, the non-affine transformation parameter w, and the Euclidean distance PHI. The statistical shape model of the foot skin is then matched to the foot skin point cloud model of the candidate, obtaining the three-dimensional coordinate set of all points from the statistical shape model of the foot skin to the foot skin point cloud model of the candidate.

6. The personalized orthotic insole design and rapid mechanical simulation method according to claim 5, characterized in that, The number of deformation parameters in the statistical shape model of the foot skin is set. Based on the shape feature matrix extracted from the statistical shape model of the foot skin and the shape vector of the deformation of the statistical shape model of the foot skin after matching the statistical shape model of the foot skin to the foot skin model of the designer, the shape parameters of the foot skin model are obtained by linear regression fitting using the least squares method.

7. The personalized orthotic insole design and rapid mechanical simulation method according to claim 6, characterized in that, The shape parameter b s The input formula for the conditional Gaussian model is as follows: Where μ S μ is the mean of the shape factor S of the orthopedic insole model. F Σ represents the mean of the shape coefficient F of the foot epidermal model. S Σ represents the variance of the shape factor S of the orthopedic insole model; F Σ represents the variance of the shape coefficient F of the foot epidermal model. SF and Σ FS It is the covariance matrix of the orthopedic insole model shape coefficient S and the foot epidermal model shape coefficient F. s To match the statistical shape model of the foot skin to the shape parameters calculated from the foot skin model of the designer.

8. The personalized orthotic insole design and rapid mechanical simulation method according to claim 7, characterized in that, The orthopedic insole point cloud model of the patient obtained in adjustment step S4 is implemented by designing smooth raised portions for the arch and forefoot areas; the height of the raised portions is a function of the two-dimensional coordinates of the vertex of the insole surface, described by a two-dimensional Gaussian function, as shown below: Where x and y are the horizontal and vertical coordinates of the vertices on the surface of the orthopedic insole point cloud model, σ x σ y The influence range of the protrusion is controlled, which is represented as the width in the x or y direction in the point cloud model of the orthopedic insole. x0 and y0 represent the coordinates of the center point of the protrusion, and the coefficient η controls the amplitude of the protrusion. The corresponding points are subjected to two-dimensional Gaussian operation to change the shape of the part and make detailed position adjustments, which are used to generate multiple results in the later mechanical simulation.

Images