Shaping effect simulation display method and system based on 3D simulation technology

Abstract

The invention discloses a plastic effect simulation display method and system based on a 3D simulation technology, and relates to the technical field of medical information, and the method comprises the steps: collecting the facial data of a patient, and constructing a multi-level facial structure initial model; constructing a hybrid deformation model; individual characteristic parameters of a patient are extracted, and an individual biomechanical parameter set is generated through the biomechanical parameter prediction network; applying the personalized biomechanical parameter set to the mixed deformation model; operation input parameters of plastic surgery are received, and the deformation result of each layer of structure is calculated and displayed in real time based on a GPU parallel computing framework; constructing a tissue healing and recovery kinetic model, and generating appearance prediction data at different postoperative time points in combination with the deformation result; and performing three-dimensional visualization processing on the appearance prediction data. According to the method, CT / MRI, 3D structured light scanning and expression dynamic video sequences are creatively integrated, the limitation of traditional single-mode data modeling is broken through, and the soft tissue boundary recognition precision is remarkably improved.

Description

Technical Field

[0001] The present invention relates to the field of medical information technology, and in particular to a method and system for simulating and displaying plastic surgery effects based on 3D simulation technology. Background Art

[0002] With the rapid development of medical information technology, 3D simulation technology is increasingly used in the prediction of plastic surgery results. However, existing technologies still face significant challenges in practical applications: First, the deformation simulation accuracy is insufficient. Traditional methods mostly rely on a single mechanical model (such as finite element or mass spring model), which makes it difficult to balance computational efficiency and the nonlinear deformation characteristics of multi-layer soft tissues, resulting in simulation results that deviate from the true biomechanical behavior. Second, personalized adaptation is insufficient. Existing models mostly use general biomechanical parameters, ignoring individual differences among patients and affecting prediction accuracy. In addition, complex deformation calculations rely on CPU serial processing, making it difficult to achieve real-time feedback for dynamic adjustment of surgical parameters, limiting clinical applicability.

[0003] At the same time, existing technologies lack modeling of dynamic recovery after surgery. Traditional methods mostly focus on intraoperative instant deformation simulation, without considering long-term morphological evolution such as tissue healing and scar contraction, and it is difficult to show the appearance changes at different recovery stages after surgery. Although some studies have attempted multi-model fusion or parameter optimization, they have not effectively integrated multi-level modeling, hybrid computing architecture and dynamic healing prediction, resulting in difficulty in balancing the precision, personalization and real-time nature of plastic surgery simulation, which restricts its clinical application value. Summary of the invention

[0004] In view of the problems in existing plastic surgery simulations, such as insufficient simulation accuracy of soft tissue deformation, insufficient consideration of individual differences, and low real-time interactive performance, the present invention is proposed.

[0005] Therefore, the problem to be solved by the present invention is how to achieve accurate, personalized and real-time simulation of plastic surgery effects through multi-level soft tissue modeling, personalized parameter optimization and hybrid computing architecture.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: In the first aspect, an embodiment of the present invention provides a method for simulating and displaying plastic surgery effects based on 3D simulation technology, which includes collecting patient facial data and constructing a multi-level facial structure initial model; constructing a hybrid deformation model, which uses a combination of finite element method and mass spring model to describe the mechanical properties of each layer of structure; extracting patient individualized characteristic parameters, constructing a biomechanical parameter prediction network, and converting the individualized characteristic parameters into corresponding personalized biomechanical parameter sets; applying the personalized biomechanical parameter set to the hybrid deformation model to optimize the deformation response characteristics of each layer of structure; receiving the operation input parameters of the plastic surgery, and calculating and displaying the deformation results of each layer of structure in real time based on the GPU parallel computing framework; constructing a tissue healing and recovery dynamics model, and generating appearance prediction data at different time points after surgery according to the deformation results and the tissue healing and recovery dynamics model; performing three-dimensional visualization processing on the appearance prediction data to form a plastic surgery effect comparison and display interface.

[0007] As a preferred solution of the plastic surgery effect simulation display method based on 3D simulation technology described in the present invention, the construction of a multi-level facial structure initial model includes the following steps: collecting patient facial CT / MRI data and 3D structured light scanning data, and recording the patient's facial expression dynamic video sequence; applying a deep learning segmentation algorithm to process the CT / MRI data to generate a preliminary segmentation mask of the five-layer structure of skin, fat, muscle, fascia, and bone; spatially aligning the 3D structured light scanning data with the CT / MRI data, and correcting the outer contour of the skin layer segmentation mask; at the same time, optimizing the segmentation mask of the muscle layer and fascia layer by constructing a facial muscle movement pattern database; performing boundary smoothing processing on the segmentation mask of each layer structure to generate three-dimensional boundary data; measuring the deformation characteristics of each layer structure based on the patient's facial expression dynamic video sequence, and calculating material property parameters; fusing the three-dimensional boundary data and material property parameters to generate a multi-level facial structure initial model.

[0008] As a preferred solution of the plastic surgery effect simulation display method based on 3D simulation technology described in the present invention, the construction of a mixed deformation model includes the following steps: constructing a finite element model of the bone layer and the fascia layer, setting the grid nodes of the bone layer as rigid fixed boundary conditions, and defining the connection constraint relationship between the fascia layer and the bone layer; constructing a composite deformation model of the muscle layer, applying a nonlinear finite element method to the muscle fiber direction area, applying a particle spring model to the muscle connection area, and defining the sliding constraint relationship between the muscle layer and the fascia layer; constructing a particle spring model of the fat layer, setting spring parameters according to the viscoelastic properties of the fat tissue, and defining the adhesion constraint relationship between the fat layer and the muscle layer; constructing a nonlinear finite element model of the skin layer, introducing a hyperelastic material constitutive equation to describe the skin deformation characteristics, and defining the connection constraint relationship between the skin layer and the fat layer; integrating the deformation models of each layer structure to form a mixed deformation model, applying material property parameters to the mixed deformation model, and configuring corresponding initial model parameters according to the material properties of each layer structure.

[0009] As a preferred solution of the plastic surgery effect simulation display method based on 3D simulation technology described in the present invention, wherein: constructing a biomechanical parameter prediction network includes the following steps: collecting basic information of the patient to form an individualized feature parameter set of the patient, and extracting the facial structure feature vectors of the multi-level facial structure initial model, integrating and generating comprehensive feature data; at the same time, constructing a reference database of biomechanical properties of facial tissue; constructing a multi-branch neural network architecture, and establishing independent prediction channels for the skin layer, fat layer, muscle layer, fascia layer, and bone layer respectively; using a multi-branch neural network architecture, pre-training a general model based on the reference database of biomechanical properties of facial tissue, and training a biomechanical parameter prediction network through a transfer learning method combined with the comprehensive feature data of historical patients; inputting the comprehensive feature data of the current patient into the biomechanical parameter prediction network, and optimizing it in combination with the patient's facial tissue aging degree evaluation index and similar group data in the historical surgical case database to generate a personalized biomechanical parameter set for the patient.

[0010] As a preferred solution of the plastic surgery effect simulation display method based on 3D simulation technology described in the present invention, the personalized biomechanical parameter set is applied to the hybrid deformation model, including the following steps: updating the material property parameters of the hybrid deformation model according to the personalized biomechanical parameter set, and mapping the updated parameters to the grid nodes and unit properties of the hybrid deformation model; performing a comprehensive deformation test, including a loading-unloading cycle test to verify the deformation trajectory and recovery characteristics of each layer structure, and a local deformation test to calculate the displacement field and stress distribution of the facial area, and record the deformation response characteristic curve; comparing the deformation response characteristic curve with the standard curve of a similar patient group in the historical data of the historical surgical case database, calculating the deviation value, and generating a parameter fine-tuning guide; optimizing the biomechanical parameters of each layer structure according to the parameter fine-tuning guide, repeatedly performing parameter mapping and comprehensive deformation testing until the deformation response accuracy reaches a preset accuracy threshold, and outputting the optimized hybrid deformation model.

[0011] As a preferred solution of the plastic surgery effect simulation display method based on 3D simulation technology described in the present invention, the following steps are included: real-time calculation and display of the deformation results of each layer structure based on the GPU parallel computing framework: receiving the plastic surgery parameters input by the doctor, and converting the plastic surgery parameters into the mechanical constraints and load distribution parameters of the hybrid deformation model, and building a digital mapping model of the surgical operation; configuring the GPU parallel computing framework, dividing the calculation subdomain according to the spatial distribution of the surgical area, and applying asynchronous adaptive grid refinement technology to perform grid encryption on the direct surgical action area; building a heterogeneous parallel solver, assigning the sparse matrix solution task of the finite element model to the GPU tensor core, and the particle calculation of the particle spring model to the CUDA multi-threaded block, and synchronizing the inter-layer force data through shared memory; performing real-time calculation of tissue deformation, and outputting the displacement field, stress field and strain field; asynchronously transmitting the displacement field data to the visualization module through the GPU-CPU heterogeneous pipeline to generate the deformation result data of each layer structure.

[0012] As a preferred solution of the plastic surgery effect simulation display method based on 3D simulation technology described in the present invention, generating appearance prediction data at different time points after surgery includes the following steps: constructing a tissue healing and recovery dynamics model, including an acute phase sub-model, an edema regression sub-model, a tissue fibrosis sub-model, and a scar formation sub-model, and defining the state variable transfer rules between the sub-models; importing the deformation result data as the initial state after surgery to construct a time series model; sequentially executing the acute phase, edema regression, tissue fibrosis, and scar formation stages in the time series model, each stage uses the output state of the previous stage as input, and generates facial tissue state prediction data at the corresponding time point through the corresponding sub-model; integrating the output data of each stage through the time series model to generate full-cycle appearance prediction data for the postoperative recovery process.

[0013] In the second aspect, an embodiment of the present invention provides a plastic surgery effect simulation display system based on 3D simulation technology, which includes a facial structure initialization module, which is used to collect patient facial data, build a multi-level facial structure initial model, and obtain three-dimensional boundary data and material property parameters of each layer structure; a hybrid deformation modeling module, which is used to build a hybrid deformation model, which uses a finite element method and a mass spring model to describe the mechanical properties of each layer structure; a parameter personalization module, which is used to extract the patient's individualized characteristic parameters, build a biomechanical parameter prediction network, and convert the individualized characteristic parameters into corresponding personalized biomechanical parameter sets; a model parameter optimization module, which is used to apply the personalized biomechanical parameter set to the hybrid deformation model to optimize the deformation response characteristics of each layer structure; a real-time deformation calculation module, which is used to receive the operation input parameters of the plastic surgery, and calculate and display the deformation results of each layer structure in real time based on the GPU parallel computing framework; a postoperative recovery prediction module, which is used to build a tissue healing and recovery dynamics model, and generate appearance prediction data at different time points after surgery according to the deformation results and the tissue healing and recovery dynamics model; a visualization display module, which is used to perform three-dimensional visualization processing on the appearance prediction data to form a plastic surgery effect comparison display interface.

[0014] The beneficial effects of the present invention are as follows: the present invention creatively integrates CT / MRI, 3D structured light scanning and expression dynamic video sequences, breaks through the limitations of traditional single-modal data modeling, and significantly improves the accuracy of soft tissue boundary recognition; it pioneers the "finite element + particle spring" hybrid deformation model architecture, and performs differentiated modeling for different tissue characteristics, while innovatively designing the interlayer sliding and adhesion constraint relationship, which not only improves the computational efficiency but also maintains a high degree of biomechanical fidelity; it realizes personalized biomechanical parameter prediction through a multi-branch neural network architecture, and converts the individual characteristics of patients into precise simulation parameters, realizing truly personalized and precise modeling; it constructs a complete tissue healing and recovery dynamics model from the acute phase to scar formation, and for the first time realizes the prediction of appearance changes at different time points after surgery, providing doctors and patients with a comprehensive postoperative recovery assessment tool. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0016] Figure 1 This is a framework flow chart of the plastic surgery effect simulation display method based on 3D simulation technology.

[0017] Figure 2A flow chart is constructed for the hybrid deformation model of the plastic surgery effect simulation display method based on 3D simulation technology.

[0018] Figure 3 A flow chart for obtaining a patient-specific biomechanical parameter set for a simulation display method of plastic surgery effects based on 3D simulation technology. DETAILED DESCRIPTION

[0019] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0020] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0021] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor is it a separate or selective embodiment that is mutually exclusive with other embodiments.

[0022] Example 1, reference Figure 1 to Figure 3 , which is the first embodiment of the present invention, and provides a method for simulating and displaying plastic surgery effects based on 3D simulation technology. The framework flow chart is as follows Figure 1 As shown, including, S1: Collect the patient's facial data, build an initial model of the multi-layer facial structure, and obtain the three-dimensional boundary data and material property parameters of each layer structure.

[0023] Specifically, collect the patient's facial CT / MRI data, 3D structured light scanning data, and the patient's facial expression dynamic video sequence (i.e., the video sequence of performing preset facial expressions); apply the deep learning segmentation algorithm to process the CT / MRI data to generate a preliminary segmentation mask of the five-layer structure of skin, fat, muscle, fascia, and bone; spatially align the 3D structured light scanning data with the CT / MRI data to correct the outer contour of the skin layer segmentation mask; extract the motion trajectory of the feature points in the facial expression dynamic video sequence, build a facial muscle movement pattern database, and optimize the segmentation mask of the muscle layer and fascia layer based on the facial muscle movement pattern database. It should be noted that when the deep learning segmentation algorithm processes CT / MRI data, if the segmentation mask confidence is lower than the preset confidence threshold (such as <90%), the interactive correction module is started to allow doctors to manually mark key anatomical landmarks and iteratively optimize the segmentation mask through semi-supervised learning.

[0024] Furthermore, boundary smoothing is performed on the segmentation mask of each layer of the structure to generate three-dimensional boundary data; the image feature values ​​of each layer of the structure are extracted, and a mapping relationship between the image feature values ​​and material property parameters is established; at the same time, the deformation characteristics of each layer of the structure are measured based on the facial expression dynamic video sequence, and the material property parameters are calculated; the three-dimensional boundary data and material property parameters are fused to generate a multi-level facial structure initial model.

[0025] It should be noted that the multi-layered initial facial structure model is divided into five layers of skin, fat, muscle, fascia, and bone. The preference is that it not only conforms to the actual composition of the facial anatomical structure, but also accurately reflects the unique biomechanical properties of each layer (such as the elasticity of the skin, the contractility of the muscle, and the rigid support of the bone). At the same time, these five layers are the main target tissues for most facial plastic surgery, and can accurately simulate the inter-layer deformation transfer relationship and surgical impact.

[0026] Preferably, the present invention creatively integrates CT / MRI, 3D structured light scanning and expression dynamic video sequences, breaking through the limitations of traditional single-modal data modeling; through deep learning segmentation and semi-supervised correction mechanism (confidence threshold triggering), the anatomical accuracy of medical images and the biomechanical characteristics of dynamic videos are organically combined, which significantly improves the accuracy of soft tissue boundary recognition, realizes accurate modeling of the five-layer structure of skin, fat, muscle, fascia and bone, and greatly reduces the need for manual intervention in the modeling process.

[0027] S2: A hybrid deformation model is constructed based on three-dimensional boundary data and material property parameters. The hybrid deformation model uses a combination of finite element method and mass-spring model to describe the mechanical properties of each layer structure.

[0028] Specifically, the hybrid deformation model construction flow chart is as follows: Figure 2 As shown, the following steps are included: S2.1: Import three-dimensional boundary data, perform tetrahedral meshing on the boundaries of each layer structure, and generate a discrete grid structure with topological connection relationship.

[0029] S2.2: Construct a finite element model of the bone layer. Based on the rigid characteristics of the bones, set the bone layer mesh nodes as rigid fixed boundary conditions to simulate its supporting role in facial deformation.

[0030] Specifically, the discrete grid structure of the bone layer is imported, the bone boundary is identified based on the bone layer segmentation mask, and a high-precision bone surface mesh is generated; the tetrahedral meshing technology is applied to segment the bone surface to ensure the mesh quality and computational efficiency; the material property parameters of the bone layer are defined, and the local elastic modulus and Poisson's ratio parameters are set according to the bone structure characteristics to reflect the mechanical properties of the bone; fixed boundary conditions are set, and key anatomical structure points are identified and used as fixed constraint points to provide a stable deformation reference benchmark; the stiffness matrix is ​​assembled to establish the overall mechanical properties of the bone layer as a supporting structure for the deformation of other tissue layers.

[0031] Furthermore, special boundary conditions are established in the bone connection area, and degrees of freedom are set at active joints such as the temporomandibular joint to allow relative movement in a specific direction. The bone force and deformation information is transmitted to the connection layer, and the displacement and force boundary conditions of the bone-fascia interface are calculated through an interpolation algorithm as input conditions for the deformation of the fascia layer.

[0032] S2.3: Construct a finite element model of the fascia layer, define the connection constraint relationship between the fascia layer and the bone layer, and establish the association equation between the fascia deformation and the bone fixed point.

[0033] Specifically, the discrete grid structure of the fascia layer is imported, and the fascia layer segmentation mask is used to generate a grid representing the anatomical characteristics of the fascia; the nonlinear elastic material method is applied to define the constitutive equation of the fascia material to reflect its mechanical properties under tension and compression; the fascia-bone connection constraints are constructed, the attachment points of the fascia and the bones are identified based on their respective segmentation masks, the node correspondence is established, and the displacement constraints are defined; a variable stiffness connection scheme is designed, different connection stiffnesses are set for different anatomical regions, and the tight and loose connections between fascia and bones are simulated; the stiffness matrix is ​​assembled based on the tetrahedral grid, the connection constraints are integrated, and the overall mechanical analysis framework is established.

[0034] Furthermore, the load transfer mechanism is defined, the transmission path of the fascia force and deformation information to the muscle layer is established, and the displacement and force boundary conditions of the fascia-muscle interface are calculated; the implicit integration format is used to calculate the deformation response of the fascia under the action of external force to capture its mechanical behavior characteristics.

[0035] S2.4: Construct a composite deformation model of the muscle layer, apply the nonlinear finite element method to the muscle fiber direction area, apply the mass-spring model to the muscle connection area, and define the sliding constraint relationship between the muscle layer and the fascia layer.

[0036] Specifically, the muscle area is divided, the direction of muscle fibers is identified based on CT / MRI data, and the muscle tissue is divided into fiber direction area and connection area; a finite element model of the fiber direction area is constructed, and the anisotropic nonlinear finite element method is used to define the active-passive coupling constitutive equation, and the Hill model is introduced to describe the dynamic mechanical properties of muscle contraction, and the activation factor parameters are set to simulate the muscle response under neural drive; a particle spring model of the connection area is constructed, and a dual-stiffness spring is designed to distinguish the mechanical behavior of the tendon under different deformation states; a transition zone is constructed, and a weighted function is applied at the junction of the regions to smoothly merge the mechanical responses of the two models to ensure deformation continuity.

[0037] Furthermore, the sliding constraint relationship between the muscle layer and the fascia layer is defined, the normal impenetrability condition and the tangential Coulomb friction are imposed, and the constraint force is calculated to simulate the physiological sliding characteristics; the mixed deformation model is solved, the mixed stiffness matrix is ​​assembled, the dynamic equations are solved using the implicit integration format, and the deformation response of the muscle layer under different activation states is calculated.

[0038] Preferably, through multi-scale modeling and physical drive parameter calibration, the problem of balancing accuracy and efficiency in traditional muscle modeling is solved, providing innovative technical support for the precise planning and effect prediction of plastic surgery.

[0039] S2.5: Construct a mass spring model of the fat layer, set the spring parameters according to the viscoelastic properties of the fat tissue, and define the adhesion constraint relationship between the fat layer and the muscle layer.

[0040] Specifically, the discrete grid structure of the fat layer is imported, the fat distribution area is identified based on the fat layer segmentation mask, and the surface fat distribution is corrected in combination with the 3D structured light scanning data to generate a grid that expresses the distribution characteristics of adipose tissue; a multi-resolution particle system is constructed, and particle distributions of different densities are set according to the requirements of local anatomical details to balance the calculation accuracy and efficiency; a viscoelastic spring network is designed to establish the connection relationship between particles, and the spring constant and damping coefficient are defined according to the viscoelastic properties of adipose tissue; the Maxwell-Kelvin structure is applied to set the spring viscoelastic parameters to simulate the hysteresis response characteristics of adipose tissue during loading and unloading; the fat density change area is defined, and differentiated particle density and spring parameters are set corresponding to different facial areas to reflect the local distribution characteristics of adipose tissue.

[0041] Furthermore, a fat-muscle adhesion constraint is constructed, the contact area between the fat layer and the muscle layer is identified based on their respective segmentation masks, the node correspondence is established, and the adhesion force function is defined to simulate the interaction between the two layers of tissue; a multi-layer sliding mechanism is designed to allow the fat layer to slide limitedly on the surface of the muscle layer while maintaining overall connectivity to reflect the actual physiological characteristics; the Verlet integration method is used to calculate the evolution of the particle position with time to simulate the deformation response of the fat layer under the action of external force.

[0042] S2.6: Construct a nonlinear finite element model of the skin layer, introduce the hyperelastic material constitutive equation to describe the skin deformation characteristics, and define the connection constraint relationship between the skin layer and the fat layer.

[0043] Specifically, a hyperelastic constitutive equation is defined, a hyperelastic material model is used to describe the nonlinear mechanical properties of the skin, and the material property parameters are initialized based on the image eigenvalues; a finite element model of the skin layer is constructed, the stiffness matrix is ​​assembled based on the tetrahedral mesh, and boundary conditions are applied to constrain the deformation behavior; the connection constraint relationship between the skin layer and the fat layer is defined, contact node pairs are constructed, normal adhesion force and tangential friction force are applied to simulate the mechanical interaction between the two layers; the nonlinear dynamic equations are solved, and the deformation response of the skin layer is calculated.

[0044] S2.7: Integrate the deformation models of each layer structure, establish the interlayer force transfer equations, and form a mixed deformation model.

[0045] S2.8: Apply material property parameters to the mixed deformation model and configure corresponding initial model parameters according to the material properties of each layer structure.

[0046] The material property parameters include but are not limited to elastic modulus, Poisson's ratio, density, viscosity parameter, hyperelastic parameter and anisotropy parameter. In addition, the material property parameters are initialized through the mapping relationship between the image feature value and the material property parameter.

[0047] S2.9: Perform preset deformation tests to verify the accuracy of the deformation response of each layer of the mixed deformation model under different forces, adjust the interlayer sliding constraint parameters and adhesion constraint parameters, and optimize the interlayer sliding and adhesion effects.

[0048] Preferably, the present invention has pioneered the design of a hybrid model framework of "finite element + mass spring" to perform differentiated modeling for different tissue characteristics. Compared with the traditional single physical model, this multi-scale hybrid modeling method improves computational efficiency while maintaining high precision, and balances the problem of accuracy and efficiency in traditional methods. At the same time, the innovative introduction of sliding constraints and adhesion constraints to simulate the interaction between tissue layers solves the deformation and distortion problem caused by traditional fixed constraints, which is more in line with actual physiological characteristics.

[0049] S3: Extract the patient's individual characteristic parameters, construct a biomechanical parameter prediction network, and convert the individual characteristic parameters into the corresponding personalized biomechanical parameter set.

[0050] Specifically, the flowchart for obtaining the patient's personalized biomechanical parameter set is as follows: Figure 3 As shown, the following steps are included: S3.1: Collect basic patient information and form a set of individualized characteristic parameters for the patient.

[0051] Among them, the patient's basic information includes but is not limited to the patient's age, gender, race, BMI index, etc.

[0052] In addition, the collection of basic patient information requires the patient's consent or authorization.

[0053] S3.2: Based on clinical trial data and literature research, construct a reference database of biomechanical properties of facial tissues of different ages, genders, and ethnicities.

[0054] S3.3: Design a feature extraction neural network to automatically identify the locations of key anatomical landmarks on the multi-level facial structure initial model and extract the facial structure feature vector.

[0055] Specifically, a feature extraction architecture based on a convolutional neural network is designed, which includes an input layer, a feature extraction layer, a landmark detection layer, and a feature vector generation layer. Through training, the network can automatically identify key anatomical landmarks of the face and generate vectors that represent facial structural features.

[0056] S3.4: Integrate the patient's individualized feature parameter set with the facial structure feature vector to generate comprehensive feature data.

[0057] S3.5: Construct a multi-branch neural network architecture and establish independent prediction channels for the skin layer, fat layer, muscle layer, fascia layer, and bone layer respectively.

[0058] Specifically, a shared feature extraction backbone network of a multi-branch neural network is defined, and the facial tissue cross-sectional feature map is extracted through cascaded convolutional layers and spatial pyramid pooling modules to generate a feature vector encoding the patient's anatomical structure; a skin layer prediction channel is constructed, and a fully connected layer sequence and a parameter regularization constraint layer are used to output a material property parameter set suitable for the skin layer finite element model; a fat layer prediction channel is constructed, and a temporal memory unit and a residual connection structure are combined to produce a material property parameter set suitable for the fat layer particle spring model; a muscle layer prediction channel is constructed, and a parallel dual-channel network is designed, in which one channel processes the muscle fiber direction tensor field and the other channel processes the connection area topology. The intention-force mechanism is integrated to output a material property parameter set suitable for the mixed deformation model of the muscle layer; a fascia layer prediction channel is constructed, and the graph convolution unit is used to process the fascia-bone connection point distribution data. The cascaded fully connected layer processes the fascia material properties, and outputs a material property parameter set suitable for the finite element model of the fascia layer; a bone layer prediction channel is constructed, and the bone structure is converted into a three-dimensional feature body through voxelization processing. The density distribution characteristics are extracted through the three-dimensional convolution layer, and the parameter set suitable for the boundary conditions of the bone layer is obtained; an inter-layer force transfer learning module is constructed, and an inter-layer connection matrix is ​​established based on the anatomical position relationship. The inter-layer force transfer rules are learned through the graph neural network to ensure the consistency between the prediction parameters of each layer.

[0059] S3.6: A multi-branch neural network architecture is used to pre-train a general model based on the reference database of facial tissue biomechanical properties. The biomechanical parameter prediction network is trained by combining the comprehensive feature data of historical patients with the transfer learning method.

[0060] Specifically, data samples were extracted from the reference database of biomechanical properties of facial tissues, and the data sample set was divided into training subset, validation subset and test subset according to a preset ratio (7:2:1). The training subset data was normalized and enhanced. A phased training strategy was adopted. First, the dedicated prediction channels were fixed, and only the shared feature extraction backbone network was trained to optimize the feature representation capability. Each dedicated prediction channel was unfrozen, and the multi-branch neural network was trained using the small batch gradient descent method. A weighted loss function was used to balance the prediction accuracy of each layer. Based on the performance evaluation of the validation subset, a learning rate decay strategy and an early stopping mechanism were adopted to prevent model overfitting and obtain a pre-trained general model.

[0061] Furthermore, the comprehensive feature data of historical patients were imported, the underlying parameters of the shared feature extraction backbone network were frozen, and the transfer learning technology was used to fine-tune the high-level network parameters. A loss function that adapts to domain transfer was designed to balance the retention of general knowledge and the adaptation to the characteristics of specific patient groups, and a biomechanical parameter prediction network for the target patient group was trained. The test subset was used to evaluate the performance of the biomechanical parameter prediction network, the prediction errors of the material property parameters of each layer were calculated, and a model performance report was generated.

[0062] S3.7: Input the comprehensive characteristic data of the current patient into the biomechanical parameter prediction network to generate a patient-personalized biomechanical parameter set.

[0063] S3.8: Based on the patient's facial tissue aging assessment indicators, adjust the elasticity parameters and viscosity parameters in the patient's individualized feature parameter set.

[0064] Among them, the aging degree assessment indicators include skin elasticity measurement values, facial wrinkle depth data, and facial bone density measurement data; the comprehensive aging coefficient is calculated based on these measurement data, and according to the preset biomechanical rules, the reduction ratio of the elastic parameters and the increase ratio of the viscosity parameters are adjusted accordingly to accurately simulate the aging characteristics of tissues.

[0065] Among them, the comprehensive aging coefficient is calculated by weighted summation, the adjustment rule of elastic parameter E and viscosity parameter The adjustment rules are: , ,in is the comprehensive aging coefficient, and are adjustment coefficients (respectively controlling the change ratio of elastic parameters and viscosity parameters with aging degree), are the elastic and viscosity parameters before adjustment, are the adjusted elastic and viscosity parameters.

[0066] S3.9: Optimize deformation response parameters based on similar group data in the historical surgical case database to generate the final patient-personalized biomechanical parameter set.

[0067] Among them, the similar group data matching criteria include age, gender, race and BMI index. The final patient-personalized biomechanical parameter set includes the elastic parameters, viscosity parameters, hyperelastic parameters, anisotropic parameters, and interlayer coupling parameters corresponding to each layer structure.

[0068] It should be noted that the deformation response parameters are intermediate optimization parameters obtained through the analysis of historical surgical case data, and the optimization results are used to adjust the elastic parameters, viscosity parameters, hyperelastic parameters, anisotropic parameters, and interlayer coupling parameters in the final parameter set to improve the accuracy of facial soft tissue deformation prediction. In addition, the optimization methods of deformation response parameters include but are not limited to regression analysis, machine learning models (such as random forests, support vector machines) or deep learning models (such as neural networks), which are fitted and predicted based on the biomechanical response characteristics of similar group data.

[0069] S4: Apply personalized biomechanical parameter sets to the hybrid deformation model to optimize the deformation response characteristics of each layer structure.

[0070] Specifically, the personalized biomechanical parameter set is parsed to extract the elastic parameters, viscosity parameters, hyperelastic parameters, anisotropic parameters, and interlayer coupling parameters corresponding to each layer of the structure; the material property parameters of the hybrid deformation model are updated according to the personalized biomechanical parameter set, including updating the elastic parameters of each layer of the structure, configuring the viscosity parameters of each layer of the structure in the hybrid deformation model, modifying the hyperelastic parameters of the skin layer, adjusting the anisotropic parameters of the muscle layer, and updating the interlayer coupling parameters between adjacent tissue layers in the hybrid deformation model; model parameter mapping is performed to map the updated material property parameters to the mesh nodes and unit properties of the hybrid deformation model.

[0071] Furthermore, a loading-unloading cycle test is performed to verify the deformation trajectory and recovery characteristics of each layer of the structure during the stress and unloading process, and to detect the deformation response accuracy of the mixed deformation model after the parameters are applied; a local deformation test is performed to calculate the displacement field and stress distribution of at least two facial areas under the preset force, and record the deformation response characteristic curve; the deformation response characteristic curve is compared with the standard curve of a similar patient group in the historical data of the historical surgical case database, the deviation value is calculated, and a parameter fine-tuning guide is generated; the biomechanical parameters of each layer of the structure are optimized according to the parameter fine-tuning guide, and the model parameter mapping, loading-unloading cycle test and local deformation test are repeatedly performed until the deformation response accuracy reaches the preset accuracy threshold; an optimized mixed deformation model is generated, which includes the updated material property parameters and interlayer interaction parameters of each layer of the structure.

[0072] It should be noted that the deviation value is calculated by the mean square error, and the parameter fine-tuning guidance includes the proportion of increase / decrease of the elastic parameters; the preset accuracy threshold is a relative error of less than 5%; the model parameter mapping adopts the weighted average method, taking into account the grid density and unit type.

[0073] Preferably, traditional plastic surgery simulation relies on universal parameters, cannot reflect individual differences of patients, and only focuses on static deformation characteristics, resulting in large deviations between simulation results and actual surgical effects. The present invention realizes individualized and precise modeling by applying personalized biomechanical parameter sets to hybrid deformation models, significantly improving simulation accuracy; designs independent prediction channels for different tissue layers, and combines graph neural networks to process interlayer force transmission, breaking through the limitations of traditional universal parameter models; combines loading-unloading cycle tests with local deformation tests to comprehensively evaluate the dynamic response and local deformation characteristics of the model, making the simulation results more consistent with the mechanical behavior in actual surgery, and providing reliable support for the precise planning and effect prediction of plastic surgery.

[0074] S5: receiving the plastic surgery operation input parameters, and calculating and displaying the deformation results of each layer structure in real time based on the GPU parallel computing framework.

[0075] Specifically, the plastic surgery parameters input by the doctor are received, including the type of surgery, the area of ​​action, the operating force, the implant properties and the tissue resection range; the plastic surgery parameters are converted into the mechanical constraint conditions and load distribution parameters of the hybrid deformation model, and a digital mapping model of the surgical operation is constructed, including the following steps: parsing the input surgery type parameters, matching the preset biomechanical constraint rule library, and activating the corresponding tissue deformation calculation mode; generating the dynamic mechanical load distribution acting on the target area according to the operating force parameters and the geometric characteristics of the surgical tools; constructing the initial boundary conditions of the implant-tissue contact surface and the constraint equations of the tissue defect area based on the implant parameters and the tissue resection range; integrating the spatial range of action and the biomechanical characteristics of the surgical area to generate the global distribution matrix of the hybrid deformation model; performing the validity check and dynamic adjustment of the plastic surgery parameters to ensure the numerical stability of the boundary conditions and the distribution matrix.

[0076] Furthermore, a GPU parallel computing framework is configured to divide the computing subdomains according to the spatial distribution of the surgical area, pre-allocate the GPU memory pool and set a dynamic expansion strategy, where the dynamic expansion strategy refers to the use of CUDA's unified memory management or memory pool technology to automatically allocate additional memory blocks when more memory is detected to be needed, and release them when not needed; asynchronous adaptive mesh refinement technology is applied to perform mesh encryption on the surgical direct action area in an independent CUDA stream, and synchronize the encrypted mesh to the main computing thread in real time; a heterogeneous parallel solver is constructed to assign the sparse matrix solution task of the finite element model to the GPU tensor core, and the mass point calculation of the mass spring model to the CUDA multi-thread block, and synchronize the inter-layer force data through shared memory; real-time calculation of tissue deformation is performed, GPU utilization and calculation delay are monitored, the mesh density of non-critical areas is dynamically adjusted, and displacement fields, stress fields and strain fields are output, where non-critical areas refer to areas that are not directly affected by surgery and whose biomechanical influence is negligible; the displacement field data is asynchronously transmitted to the rendering module through the GPU-CPU heterogeneous pipeline to generate deformation result data of each layer structure.

[0077] Preferably, the present invention realizes the digital mapping of surgical operations by automatically converting plastic surgery parameters into mechanical constraints and load distribution parameters of the hybrid deformation model; designs a GPU computing architecture specifically for soft tissue deformation, especially realizes adaptive mesh refinement of the surgical area and dynamic mesh simplification of non-critical areas, and balances accuracy and performance; at the same time, based on the GPU parallel computing framework, adopts dynamic memory management, asynchronous adaptive mesh refinement and heterogeneous parallel solver, which significantly improves computing efficiency; in addition, by dynamically adjusting the mesh density of non-critical areas and the GPU-CPU heterogeneous pipeline, it optimizes the allocation of computing resources and provides doctors with intuitive preoperative planning support.

[0078] S6: Construct a tissue healing and recovery dynamics model, and generate appearance prediction data at different time points after surgery based on the deformation results and the tissue healing and recovery dynamics model.

[0079] Specifically, a tissue healing and recovery dynamics model is constructed, including an acute phase sub-model, an edema regression sub-model, a tissue fibrosis sub-model, and a scar formation sub-model, and the state variable transfer rules between the sub-models are defined, including the following steps: constructing an acute phase sub-model, using the nonlinear elastic material constitutive equation to describe the edema tissue characteristics, using the inflammatory factor concentration gradient equation to simulate edema diffusion, and establishing a mathematical mapping relationship between surgical trauma parameters and inflammatory response intensity; constructing an edema regression sub-model, using an exponential decay function to characterize the body fluid reabsorption process, combining the tissue permeability matrix to calculate the liquid flow rate, and integrating the stress-strain recovery equation to simulate tissue morphological recovery; constructing a tissue fibrosis sub-model, using the collagen fiber directional deposition algorithm to describe the fibrosis process, using the anisotropic material stiffness tensor to characterize the tissue hardness change, and using the local stress field as a fibrosis process regulating factor; constructing a scar formation sub-model, using the contraction-stress coupling equation to describe the evolution of scar tissue, simulating tissue plastic deformation through the von Mises yield criterion, and integrating the optical property change matrix to generate scar appearance characteristics.

[0080] Furthermore, the state variable transfer rules between sub-models are defined, the transfer matrix of displacement field, stress field, tissue density and material parameters is established, the data format conversion protocol is formulated, and the method of maintaining the consistency of state parameters between different models is designed; the time series state transition algorithm is integrated, and the piecewise continuous function is used to describe the stage characteristics of the recovery process, and the parameterized correspondence between the type of surgery and the duration of the recovery stage is established to construct a complete tissue healing and recovery dynamics model.

[0081] Furthermore, the deformation result data is imported as the initial state after surgery to construct a time series model, in which the time series model performs the following operations: setting the postoperative recovery timeline and dividing the effective time interval of each sub-model, including the acute phase, edema subsidence phase, tissue fibrosis phase and scar formation phase; establishing the phase switching trigger condition; defining the data interface between the sub-models, and using the output state of the previous phase as the input of the next phase.

[0082] Furthermore, when the time series model advances to the acute phase, the initial postoperative state is input into the acute phase sub-model to generate the predicted data of the facial tissue state at the first time point after surgery, including the spatial distribution of edema volume, tissue displacement field, and surface short-term deformation; when the time series model advances to the edema regression phase, the output state of the acute phase sub-model is input into the edema regression sub-model to generate the predicted data of the facial tissue state at the second time point after surgery, including the residual edema area positioning, tissue retraction, and surface smoothness; when the time series model advances to the tissue fibrosis phase, the output state of the edema regression sub-model is input into the tissue fibrosis sub-model to generate the predicted data of the facial tissue state at the third time point after surgery, including the deep elastic modulus distribution, morphological offset, and dynamic stability parameters; when the time series model advances to the scar formation phase, the output state of the tissue fibrosis sub-model is input into the scar formation sub-model to generate the predicted data of the facial tissue state at the fourth time point after surgery, including scar contraction, final displacement field, and morphological confidence interval; the output data of each stage are integrated through the time series model to generate the full-cycle appearance prediction data of the postoperative recovery process.

[0083] In addition, the first time point, the second time point, the third time point, and the fourth time point are 7 days, 30 days, 180 days, and 365 days, respectively. The preference of these time points is reflected in the turning period of key biological processes: 7 days corresponds to the peak period of acute inflammation (maximum diffusion of edema), 30 days covers the middle period of fluid index attenuation (edema subsides by 70%-80%), 180 days matches the stable period of collagen remodeling (fibrosis hardness tends to be stable), and 365 days corresponds to the end point of scar maturation (shrinkage rate <5%). Its design is consistent with the biological time constants of tissue healing (such as the macrophage activity cycle of 28 days and the collagen half-life of 90-120 days), and it also fits the clinical follow-up nodes, balancing the observation needs of short-term drastic changes and long-term gradual processes through exponential time spans.

[0084] Preferably, traditional plastic surgery simulation usually only focuses on the immediate effect of the surgery, ignoring the important impact of the postoperative tissue healing and recovery process on the final appearance, resulting in patients lacking accurate expectations of changes at different stages after surgery. The present invention achieves full-cycle simulation from the acute postoperative period to scar formation by constructing a complete tissue healing and recovery dynamics model, providing doctors with a comprehensive postoperative recovery assessment tool, and enabling patients to form reasonable expectations for postoperative appearance changes, significantly improving the clinical value of plastic surgery effect prediction and patient satisfaction.

[0085] S7: Perform three-dimensional visualization on the appearance prediction data to form a multi-viewing and multi-time point plastic surgery effect comparison display interface.

[0086] Furthermore, the present embodiment also provides a plastic surgery effect simulation display system based on 3D simulation technology, including a facial structure initialization module, which is used to collect patient facial data, build a multi-level facial structure initial model, and obtain three-dimensional boundary data and material property parameters of each layer structure; a hybrid deformation modeling module, which is used to build a hybrid deformation model, and the hybrid deformation model uses a finite element method and a mass spring model to describe the mechanical properties of each layer structure; a parameter personalization module, which is used to extract the patient's individualized characteristic parameters, build a biomechanical parameter prediction network, and convert the individualized characteristic parameters into corresponding personalized biomechanical parameter sets; a model parameter optimization module, which is used to apply the personalized biomechanical parameter set to the hybrid deformation model to optimize the deformation response characteristics of each layer structure; a real-time deformation calculation module, which is used to receive the operation input parameters of the plastic surgery, and calculate and display the deformation results of each layer structure in real time based on the GPU parallel computing framework; a postoperative recovery prediction module, which is used to build a tissue healing and recovery dynamics model, and generate appearance prediction data at different time points after surgery according to the deformation results and the tissue healing and recovery dynamics model; a visualization display module, which is used to perform three-dimensional visualization processing on the appearance prediction data to form a plastic surgery effect comparison display interface.

[0087] In summary, the present invention creatively integrates CT / MRI, 3D structured light scanning and expression dynamic video sequences, breaks through the limitations of traditional single-modal data modeling, and significantly improves the accuracy of soft tissue boundary recognition; it pioneers the "finite element + particle spring" hybrid deformation model architecture, and performs differentiated modeling for different tissue characteristics. At the same time, it innovatively designs the interlayer sliding and adhesion constraint relationship, which not only improves the computational efficiency but also maintains a high degree of biomechanical fidelity; it realizes personalized biomechanical parameter prediction through a multi-branch neural network architecture, and converts the individual characteristics of patients into precise simulation parameters, realizing truly personalized and precise modeling; it constructs a complete tissue healing and recovery dynamics model from the acute phase to scar formation, and for the first time realizes the prediction of appearance changes at different time points after surgery, providing doctors and patients with a comprehensive postoperative recovery assessment tool.

[0088] Example 2, reference Figure 1 to Figure 3 , which is the second embodiment of the present invention, provides a method for simulating and displaying plastic surgery effects based on 3D simulation technology. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0089] In real-life application, this plastic surgery simulation display method based on 3D simulation technology first collected facial data from a 55-year-old female patient. The patient faced the problem of facial sagging and hoped to improve her appearance through face lift surgery. The acquisition team used Siemens SOMATOM CT scanner to obtain the patient's facial CT scan data with a resolution of 0.5mm; the facial surface morphology data was collected through Artec Eva 3D structured light scanner with an accuracy of 0.1mm; and high-definition video sequences of the patient performing 7 preset facial expressions such as smiling and frowning were recorded at the same time. The nnU-Net deep learning segmentation framework was used to process the CT data, and the initial segmentation mask confidence was 87.3%, which was lower than the 90% threshold, and then the interactive correction module was activated. Professional physicians annotated 42 key anatomical landmarks, and after 10 iterations of semi-supervised learning, the segmentation mask confidence increased to 95.7%. After boundary smoothing and material property parameter extraction, an accurate initial facial model consisting of five layers of structure, namely skin, fat, muscle, fascia and bone, was successfully constructed. Compared with traditional single-modal data modeling, the accuracy of anatomical boundary recognition was improved by 28.3% and the manual intervention time was reduced by 64.5%.

[0090] Based on the acquired three-dimensional boundary data and material property parameters, the research team constructed a hybrid deformation model. First, tetrahedral meshing was performed on each layer of the structure to generate a discrete structure containing 283,564 mesh units. The bone layer was set as a rigid fixed boundary condition with an elastic modulus of 13,700 MPa; a connection constraint relationship was established between the fascia layer and the bone layer, with an elastic modulus of 80 MPa; the muscle layer was modeled in a hybrid way, with a nonlinear finite element method applied to the fiber direction region, a mass spring model applied to the connection region, and the active contraction region described by the Hill model, with an elastic modulus set to 32.8 MPa and an activation factor range of 0-1.0; a sliding constraint relationship was defined between the muscle layer and the fascia layer, with the normal impenetrable condition and the tangential Coulomb friction coefficient set to 0.28; the fat layer used a mass spring model, with a spring stiffness set to 4.3 kPa and a damping coefficient of 0.32; the skin layer used the Ogden hyperelastic model, with a shear modulus of 23.1 kPa and a bulk modulus of 0.4 MPa. In order to optimize the interlayer interaction, 10 sets of preset deformation tests were performed, and the interlayer sliding constraint parameters and adhesion constraint parameters were adjusted, so that the deformation response error of each layer structure under different loads was controlled within 4.2%. The computational efficiency was improved by 65.3% compared with the traditional single physical model, while maintaining an accuracy level of no less than 94%.

[0091] According to the personalized biomechanical characteristics of patients, the system collected basic information of patients (55 years old, female, Asian, BMI index of 24.2) as an individualized feature parameter set, and the collection of basic information of patients required the consent or authorization of patients. Combined with the biomechanical parameter data of 2,376 facial tissues in the clinical trial database, a multi-branch neural network architecture was constructed. The network contains 5 independent prediction channels, targeting the skin layer, fat layer, muscle layer, fascia layer and bone layer respectively. The shared feature extraction backbone network extracts features through the spatial pyramid pooling module. The skin layer prediction channel uses 3 fully connected layers, and parameter regularization constraints are set at the end; the fat layer prediction channel combines LSTM units and residual connections; the muscle layer prediction channel designs a parallel dual-channel structure to process muscle fiber direction and connection area respectively; the fascia layer prediction channel uses a graph convolution unit to process the connection point distribution; the bone layer prediction channel extracts density distribution features through 3D convolution. The multi-branch network was trained based on a 75% / 15% / 10% training / validation / test set split, using the Adam optimizer with an initial learning rate of 0.0002 and a decay of 10% every 10 epochs. The final model prediction error was less than 7.3%, and the generated patient-specific biomechanical parameter set included elasticity, viscosity, hyperelasticity, and anisotropy parameters for each layer of tissue. Based on the measured skin elasticity (0.62 mm / N), the average facial wrinkle depth (1.8 mm), and bone density assessment, the calculated comprehensive aging coefficient was 0.78, corresponding to the adjustment of the elastic parameter reduction ratio (-22.3%) and the viscosity parameter increase ratio (+35.1%).

[0092] After applying the personalized biomechanical parameter set to the hybrid deformation model, the system performed a face lift simulation. The surgical parameters entered by the doctor included: SMAS face lift, zygomatic area traction, and an operating force of 2.8 N / cm 2, no implants, and a temporal incision width of 4.2 cm. These parameters are converted into load distribution through a biomechanical constraint rule library, and the mesh density of the surgical area is increased to three times the original density, reaching about 850,000 units. The NVIDIA RTX 3090 GPU parallel computing framework is used for real-time calculations, and the computational domain is divided into 32 subdomains, each of which is assigned to a separate CUDA thread block. The system applies asynchronous adaptive mesh refinement technology to perform mesh encryption on the direct surgical area, and dynamically adjusts the non-critical area to 65% of the original density. The sparse matrix solution task of the finite element model is assigned to the GPU tensor core, and the mass calculation of the mass spring model is assigned to the CUDA multi-thread block, and the inter-layer force data is synchronized through shared memory. The final computing performance achieves a real-time rendering rate of 24 frames per second, and the simulation calculation delay is less than 42 milliseconds, which is 28.4 times faster than traditional CPU calculations, while maintaining the deformation accuracy at 95.2%. The surgical deformation results showed that the maximum displacement of the SMAS layer reached 7.2 mm, the stress concentration area was located at the edge of the temporal incision, and the maximum stress value was 0.58 MPa.

[0093] Based on the results of surgical deformation, the system applied the tissue healing and recovery dynamics model to predict the appearance changes at different time points after surgery. The acute phase submodel (7 days after surgery) predicted the distribution of facial edema volume, with the maximum edema thickness of 5.3 mm, concentrated in the zygomatic area and jawline; the edema submodel (30 days after surgery) showed that the residual edema dropped to 18.2% of the initial value, and the tissue retraction was 1.3 mm; the deep elastic modulus calculated by the tissue fibrosis submodel (180 days after surgery) increased by 27.4%, and the morphological offset was 17.5% of the original traction; the scar formation submodel (365 days after surgery) predicted a scar contraction of 0.8 mm, and the final displacement field retracted by 10.3% compared with the immediate effect of surgery. The system generated a three-dimensional visualization interface containing each time point, showing the complete appearance change process of the patient from preoperative, postoperative acute phase, edema subsidence, fibrosis phase to the final effect. By calculation and comparison, the final appearance prediction 365 days after surgery was compared with the actual follow-up 3D scanning data, and the average deviation was only 1.8mm, and the position matching rate reached 93.5%, which was much better than the traditional simulation method. As shown in Table 1, the present invention has significant advantages over the prior art.

[0094] Table 1 Performance comparison between the present invention and the prior art As can be seen from Table 1, the plastic surgery effect simulation display method based on 3D simulation technology proposed in the present invention is superior to the existing technology in many performance indicators. This method constructs a high-precision facial structure model through multimodal data fusion, uses a mixed deformation model to balance calculation accuracy and efficiency, introduces a personalized biomechanical parameter prediction network to achieve patient-specific modeling, uses a GPU parallel computing framework to achieve real-time simulation calculations, and constructs a tissue healing and recovery dynamics model to predict the appearance changes throughout the postoperative cycle. These innovations together constitute a complete plastic surgery effect prediction system, which provides clinicians with accurate surgical planning tools, while helping patients form reasonable postoperative recovery expectations, significantly improving the safety of plastic surgery and patient satisfaction.

[0095] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for simulating and displaying plastic surgery effects based on 3D simulation technology, characterized in that: include, Collect patient facial data and build a multi-level initial facial structure model; Constructing a hybrid deformation model, wherein the hybrid deformation model uses a finite element method and a mass-spring model to describe the mechanical properties of each layer structure; Extracting individualized characteristic parameters of patients, constructing a biomechanical parameter prediction network, and converting the individualized characteristic parameters into corresponding personalized biomechanical parameter sets; Applying the personalized biomechanical parameter set to the hybrid deformation model to optimize the deformation response characteristics of each layer structure; Receive the input parameters of plastic surgery operations, calculate and display the deformation results of each layer structure in real time based on the GPU parallel computing framework; Constructing a tissue healing and recovery kinetic model, and generating appearance prediction data at different time points after surgery according to the deformation results and the tissue healing and recovery kinetic model; The appearance prediction data is processed into three-dimensional visualization to form a plastic surgery effect comparison display interface.

2. The method for simulating and displaying plastic surgery effects based on 3D simulation technology according to claim 1, characterized in that: The construction of the multi-level facial structure initial model comprises the following steps: Collect the patient's facial CT / MRI data and 3D structured light scanning data, and record the patient's facial expression dynamic video sequence; Applying a deep learning segmentation algorithm to process the CT / MRI data to generate a preliminary segmentation mask of the five-layer structure of skin, fat, muscle, fascia, and bone; The 3D structured light scanning data and the CT / MRI data are spatially registered to correct the outer contour of the skin layer segmentation mask; at the same time, the segmentation masks of the muscle layer and the fascia layer are optimized by constructing a facial muscle movement pattern database; Perform boundary smoothing processing on the segmentation mask of each layer structure to generate three-dimensional boundary data; Based on the patient's facial expression dynamic video sequence, the deformation characteristics of each layer structure are measured to calculate material property parameters; The three-dimensional boundary data and material property parameters are integrated to generate a multi-level facial structure initial model.

3. The method for simulating and displaying plastic surgery effects based on 3D simulation technology according to claim 1, characterized in that: The construction of the blended deformation model comprises the following steps: Construct finite element models of the bone layer and fascia layer, set the mesh nodes of the bone layer as rigid fixed boundary conditions, and define the connection constraint relationship between the fascia layer and the bone layer; Construct a composite deformation model of the muscle layer, apply the nonlinear finite element method to the muscle fiber direction area, apply the mass-spring model to the muscle connection area, and define the sliding constraint relationship between the muscle layer and the fascia layer; Construct a mass spring model of the fat layer, set the spring parameters according to the viscoelastic properties of the fat tissue, and define the adhesion constraint relationship between the fat layer and the muscle layer; A nonlinear finite element model of the skin layer is constructed, the hyperelastic material constitutive equation is introduced to describe the skin deformation characteristics, and the connection constraint relationship between the skin layer and the fat layer is defined; The deformation models of each layer structure are integrated to form a mixed deformation model, material property parameters are applied to the mixed deformation model, and corresponding initial model parameters are configured according to the material properties of each layer structure.

4. The method for simulating and displaying plastic surgery effects based on 3D simulation technology according to claim 1, characterized in that: The construction of the biomechanical parameter prediction network comprises the following steps: Collect the basic information of the patient to form the patient's individualized feature parameter set, extract the facial structure feature vector of the multi-level facial structure initial model, and integrate it to generate comprehensive feature data; at the same time, build a reference database of facial tissue biomechanical properties; Construct a multi-branch neural network architecture and establish independent prediction channels for the skin layer, fat layer, muscle layer, fascia layer, and bone layer; The multi-branch neural network architecture is adopted to obtain a general model based on the reference database of facial tissue biomechanical properties through pre-training, and a biomechanical parameter prediction network is obtained through training by combining the comprehensive feature data of historical patients through a transfer learning method; The comprehensive characteristic data of the current patient is input into the biomechanical parameter prediction network, and is optimized by combining the patient's facial tissue aging assessment indicators and similar group data in the historical surgical case database to generate a personalized biomechanical parameter set for the patient.

5. The method for simulating and displaying plastic surgery effects based on 3D simulation technology according to claim 1, characterized in that: Applying the personalized biomechanical parameter set to the blended deformation model comprises the following steps: updating material property parameters of the hybrid deformation model according to the personalized biomechanical parameter set, and mapping the updated parameters to mesh node and unit properties of the hybrid deformation model; Perform comprehensive deformation tests, including loading-unloading cycle tests to verify the deformation trajectory and recovery characteristics of each layer structure, and local deformation tests to calculate the displacement field and stress distribution in the facial area and record the deformation response characteristic curve; Compare the deformation response characteristic curve with the standard curve of similar patient groups in the historical data of the historical surgical case database, calculate the deviation value, and generate parameter fine-tuning guidance; The biomechanical parameters of each layer structure are optimized according to the parameter fine-tuning guidance, and the parameter mapping and comprehensive deformation test are repeated until the deformation response accuracy reaches the preset accuracy threshold, and the optimized hybrid deformation model is output.

6. The method for simulating and displaying plastic surgery effects based on 3D simulation technology according to claim 1, characterized in that: The real-time calculation and display of the deformation results of each layer structure based on the GPU parallel computing framework includes the following steps: Receive the plastic surgery parameters input by the doctor, and convert the plastic surgery parameters into mechanical constraints and load distribution parameters of the hybrid deformation model to build a digital mapping model of the surgical operation; Configure the GPU parallel computing framework, divide the computing subdomains according to the spatial distribution of the surgical area, and apply asynchronous adaptive mesh refinement technology to perform mesh encryption on the direct surgical area; Build a heterogeneous parallel solver to assign the sparse matrix solving task of the finite element model to the GPU tensor core, assign the mass calculation of the mass-spring model to the CUDA multi-threaded block, and synchronize the inter-layer force data through shared memory; Perform real-time calculation of tissue deformation and output displacement field, stress field and strain field; The displacement field data is asynchronously transmitted to the visualization module through the GPU-CPU heterogeneous pipeline to generate the deformation result data of each layer structure.

7. The method for simulating and displaying plastic surgery effects based on 3D simulation technology according to claim 1, characterized in that: The generating of appearance prediction data at different time points after surgery comprises the following steps: Construct a tissue healing and recovery dynamics model, including an acute phase sub-model, an edema resolution sub-model, a tissue fibrosis sub-model, and a scar formation sub-model, and define the state variable transfer rules between the sub-models; Import the deformation result data as the initial state after surgery and construct a time series model; The acute phase, edema regression, tissue fibrosis, and scar formation phases in the time series model are executed in sequence. Each phase uses the output state of the previous phase as input and generates the predicted data of the facial tissue state at the corresponding time point through the corresponding sub-model. The output data of each stage are integrated through the time series model to generate the appearance prediction data of the whole cycle of the postoperative recovery process.

8. A plastic surgery effect simulation display system based on 3D simulation technology, based on the plastic surgery effect simulation display method based on 3D simulation technology according to any one of claims 1 to 7, characterized in that: Also includes, The facial structure initialization module is used to collect patient facial data and build a multi-level facial structure initial model; A hybrid deformation modeling module is used to construct a hybrid deformation model, wherein the hybrid deformation model uses a finite element method and a mass spring model to describe the mechanical characteristics of each layer structure; A parameter personalization module is used to extract individualized characteristic parameters of patients, construct a biomechanical parameter prediction network, and convert the individualized characteristic parameters into corresponding personalized biomechanical parameter sets; A model parameter optimization module, used for applying the personalized biomechanical parameter set to the hybrid deformation model to optimize the deformation response characteristics of each layer structure; The real-time deformation calculation module is used to receive the operation input parameters of the plastic surgery, and calculate and display the deformation results of each layer structure in real time based on the GPU parallel computing framework; A postoperative recovery prediction module is used to construct a tissue healing and recovery dynamics model, and generate appearance prediction data at different time points after surgery based on the deformation results and the tissue healing and recovery dynamics model; The visualization display module is used to perform three-dimensional visualization processing on the appearance prediction data to form a plastic surgery effect comparison display interface.

Images