Joint replacement operation simulation system based on pressure sensing feedback

Abstract

The invention discloses a joint replacement operation simulation system based on pressure sensing feedback. The system comprises a data acquisition and preprocessing module, a model construction and parameter initialization module, a multi-modal coupling modeling module and a dynamic response optimization module. The data acquisition and preprocessing module is used for acquiring and preprocessing original data related to a joint replacement operation, and the original data comprises intraoperative three-dimensional bone tissue image data, instrument-bone tissue interaction real-time mechanical data and operation related anatomical parameters. According to the joint replacement surgery simulation system based on pressure sensing feedback, the mechanical state and the structural change in the key operation process can be visually displayed, quantitative feedback and risk early warning are provided for an operator, the scientificity of surgery planning and the stability of intraoperative operation can be improved, and the remarkable teaching and clinical application value is achieved.

Description

Technical Field

[0001] This invention relates to the field of medical simulation, and more specifically, to a joint replacement surgery simulation system based on pressure sensing feedback. Background Technology

[0002] Joint replacement surgery is one of the most widely used and complex surgical procedures in modern orthopedic clinics. It is characterized by its delicate operation and high risk, demanding extremely high levels of spatial positioning, force control, and anatomical understanding from the surgeon. Currently, mainstream surgical training methods include ex vivo specimen manipulation, virtual simulation training systems, and robot-assisted surgical platforms. However, these traditional methods generally suffer from problems such as limited data sources, delayed feedback, and insufficient model realism. For example, virtual simulation systems often rely on static image models, lacking the acquisition and feedback of real-time intraoperative biomechanical interaction data, making it difficult to reproduce the actual stress and deformation characteristics during key procedures such as prosthesis implantation and bone surface reshaping. Furthermore, due to the failure to effectively integrate imaging data, biomechanical data, and individual anatomical parameters, the differences between patients are often overlooked, resulting in a lack of specificity and accuracy in simulation training.

[0003] Furthermore, existing systems generally lack adaptive learning and dynamic optimization mechanisms, making it impossible to update model parameters in real time based on surgeon errors and hindering closed-loop training involving perception, response, and correction. Especially when simulating high-precision procedures such as grinding, implantation, and compaction, the model's static response cannot accurately reflect the nonlinear mechanical changes during surgery, limiting the realism and educational value of the training. Therefore, there is an urgent need for a joint replacement surgery simulation system that integrates multimodal information, possesses pressure sensing feedback, and self-learning optimization capabilities to achieve highly realistic dynamic simulation and individualized training support throughout the entire surgical process. Summary of the Invention

[0004] The purpose of this invention is to provide a joint replacement surgery simulation system based on pressure sensing feedback. This system solves the problems of existing methods, such as single data sources, feedback delays, and insufficient model realism. It fails to effectively integrate imaging data, mechanical data, and individual anatomical parameters, often ignoring the differences between different patients. This results in a lack of targeted and accurate simulation training, a lack of adaptive learning and dynamic optimization mechanisms, and an inability to update model parameters in real time based on the surgeon's operational errors, thus failing to meet the needs of users.

[0005] This invention achieves the above objective through the following technical solution: a joint replacement surgery simulation system based on pressure sensing feedback, the system comprising: The module includes data acquisition and preprocessing, model building and parameter initialization, multimodal coupling modeling, and dynamic response optimization. The data acquisition and preprocessing module is used to acquire and preprocess raw data related to joint replacement surgery. The raw data includes intraoperative three-dimensional bone tissue imaging data, real-time mechanical data of instrument-bone tissue interaction, and surgical-related anatomical parameters. The model building and parameter initialization module is used to build a multimodal fusion adaptive feedback model and determine the initial parameters of the model. The multimodal coupling modeling module is used to input preprocessed data into the adaptive feedback model, and realize adaptive coupling modeling through multimodal fusion algorithm to dynamically adjust simulation parameters; The dynamic response optimization module is used to continuously optimize the model through a cyclic optimization mechanism to achieve a highly realistic simulation of the entire joint replacement surgery process.

[0006] Furthermore, the intraoperative three-dimensional bone tissue image data acquired by the data acquisition and preprocessing module covers the core bone tissue region of joint replacement, which includes: At least one of the following: acetabulum, proximal femur, and tibial plateau; The real-time mechanical data is collected by pressure sensors distributed on the surface of the prosthesis, the tip of the grinding instrument and the contact surface of the compaction tool, including at least one of grinding cutting force, prosthesis implantation pressure and interface compaction reaction force. The anatomical parameters include at least one of the following key parameters: femoral neck-shaft angle, acetabular abduction angle, prosthesis size, cortical bone thickness, and cancellous bone porosity.

[0007] Furthermore, the data acquisition and preprocessing module includes the following steps: The three-dimensional bone tissue image data were normalized, the real-time mechanical data were filtered and denoised, and the anatomical parameters were standardized. The filtering and denoising process employs algorithms including Gaussian filtering, while the normalization and standardization processes optimize data consistency by eliminating differences in data units.

[0008] Furthermore, the model building and parameter initialization module constructs an adaptive feedback model based on a multi-resolution convolutional-graph convolutional hybrid network, wherein the hybrid network includes a multi-resolution convolutional module, a graph convolutional module, and a feature fusion module; The multi-resolution convolution module is used to extract multi-scale features from three-dimensional bone tissue image data; The graph convolution module is used to fuse real-time mechanical data and anatomical parameters into a multi-dimensional feature vector, and to construct a graph structure based on the bone tissue-prosthesis interaction topology. The feature fusion module employs an attention mechanism to adaptively weight and fuse multi-resolution image feature maps and fused features.

[0009] Furthermore, the initial parameters determined by the model construction and parameter initialization module include: Initial values ​​for attention weights and error threshold; The initial value of the attention weight is determined based on the information entropy ratio of the multimodal data in the training set; The error threshold is set based on the statistical characteristics of a real clinical joint replacement surgery interactive dataset and is dynamically corrected in combination with the accuracy requirements of key surgical operations, including at least one of grinding, prosthesis implantation, and interface compaction.

[0010] Furthermore, the multimodal coupling modeling module obtains bone tissue stiffness distribution parameters and prosthesis fitting parameters through a multimodal fusion algorithm; The bone tissue stiffness distribution parameters are set according to the anatomical structure by dividing the bone tissue into zones, and the zones include at least one of the acetabular roof, femoral distance, and medial tibial plateau. The prosthesis fitting parameters include at least one of the following dimensional parameters: prosthesis implantation depth, tilt angle, offset, and soft tissue tension compensation coefficient.

[0011] Furthermore, the multimodal coupling modeling module establishes a mapping relationship model between stiffness distribution and bone density and anatomical parameters, determines the bone density level based on the gray values ​​of different regions in the three-dimensional bone tissue image, and dynamically adjusts the stiffness coefficient of the corresponding region in combination with cortical thickness and cancellous bone porosity parameters. The mapping relationship model achieves dynamic stiffness adaptation by associating the baseline stiffness coefficient, normalized bone mineral density value, stiffness adjustment coefficient, and anatomical parameters.

[0012] Furthermore, the multimodal coupling modeling module establishes a correlation model between prosthesis fitting parameters and anatomical parameters and mechanical data, determines the initial fitting parameters of the prosthesis based on the anatomical parameters, and adjusts the fitting parameters through real-time mechanical data feedback to ensure uniform force distribution and interface fit during prosthesis implantation. The correlation model achieves dynamic optimization through initial adaptation parameters, parameter adjustment coefficients, and the difference between the real-time interactive average force and the reference force value.

[0013] Furthermore, the dynamic response optimization module continuously optimizes the adaptive feedback model through a perception-response-correction loop, the loop including: The three phases of perception, response, and correction. The perception phase acquires interactive state data of the current surgical operation, including at least one of bone tissue cutting amount, prosthesis implantation progress, and interface pressure distribution. The response phase simulates bone tissue deformation, prosthesis placement process, and interface stress transmission during surgical operations, generating dynamic response output. The correction phase calculates the error between the dynamic response output and the preset real surgical interaction data, and updates the model parameters based on the error feedback.

[0014] Furthermore, the model parameters updated by the dynamic response optimization module include: Attention weights, prosthesis adaptation parameters, weight parameters of multi-resolution convolutional-graph convolutional hybrid networks, and adjustment coefficients of stiffness mapping relationship models; The attention weights are adaptively adjusted based on error feedback and surgical operation type, and the prosthesis fitting parameters are adjusted based on error feedback and anatomical parameter constraints. The optimization process repeats the sensing-response-correction cycle until the simulation error is less than a preset threshold, thereby achieving a highly realistic dynamic response for key operations in joint replacement surgery.

[0015] The beneficial effects of this invention are as follows: 1. By placing pressure sensors on the surface of the prosthesis, the tip of the grinding instrument, and the contact surface of the compaction tool, mechanical information such as grinding cutting force, implantation pressure, and interface reaction force is collected in real time. This information is then combined with three-dimensional image data and anatomical parameters to achieve dynamic coupling and high-precision modeling of multimodal data during surgery.

[0016] 2. The system uses a multi-resolution convolutional-graph convolutional hybrid network to construct an adaptive feedback model, and combines an attention mechanism to achieve weighted fusion of multimodal features. It can automatically adjust the simulation response according to the bone density, cortical thickness and prosthesis parameters of different patients, which significantly improves the individualization of the simulation.

[0017] 3. Through the perception-response-correction loop mechanism, the model parameters are continuously self-learned and error corrected, so that the simulation process can reflect the deformation of bone tissue, the placement of the prosthesis and the changes in interface stress during the operation in real time, and ensure that the dynamic simulation results are highly consistent with the clinical operation.

[0018] 4. This invention can intuitively display the mechanical state and structural changes during key operations, providing surgeons with quantitative feedback and risk warnings, which helps to improve the scientific nature of surgical planning and the stability of intraoperative operations, and has significant teaching and clinical application value. Attached Figure Description

[0019] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments of this application and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a flowchart illustrating the overall architecture of the present invention; Figure 2 This is a flowchart of the multimodal coupling modeling process of the present invention; Figure 3This is a diagram of the dynamic response optimization mechanism of the present invention. Detailed Implementation

[0020] The present application will now be described in further detail with reference to the accompanying drawings. It should be noted that the following specific embodiments are only used to further illustrate the present application and should not be construed as limiting the scope of protection of the present application. Those skilled in the art can make some non-essential improvements and adjustments to the present application based on the above application content.

[0021] Example 1: Please see Figure 1-3 This invention provides a technical solution: a joint replacement surgery simulation system based on pressure sensing feedback, the system comprising: The module includes data acquisition and preprocessing, model building and parameter initialization, multimodal coupling modeling, and dynamic response optimization. The data acquisition and preprocessing module is used to acquire raw data related to joint replacement surgery, including intraoperative three-dimensional bone tissue imaging data, real-time mechanical data of instrument-bone tissue interaction collected by pressure sensors, and anatomical parameters related to joint replacement surgery, and to preprocess the raw data. Among them, intraoperative 3D bone tissue imaging data is image data that reflects the three-dimensional morphology and structure of bone tissue, acquired through medical imaging equipment such as CT and MRI during joint replacement surgery. This data can clearly present the three-dimensional features of bone tissue, providing a morphological basis for subsequent simulation. Real-time mechanical data of instrument-bone tissue interaction is obtained by pressure sensors installed on surgical instruments during surgery, measuring the mechanical signals generated when the instruments interact with bone tissue in real time, including the magnitude and direction of force, reflecting the mechanical interaction during the surgical operation. Anatomical parameters related to joint replacement surgery include various parameters of the anatomical structure of the human joint, such as joint size, bone shape, and articular surface angles. These parameters are crucial for accurately simulating the surgical process and its effects. Preprocessing involves a series of processing operations on the acquired raw data to remove noise, correct data deviations, and standardize data formats, making the raw data more suitable for subsequent model construction and analysis, thus improving data quality. The model building and parameter initialization module is used to build an adaptive feedback model for multimodal force perception-image fusion based on a multi-resolution convolutional-graph convolutional hybrid network, and to determine the initial values ​​of attention weights and error thresholds in the model. Among them, the multi-resolution convolutional-graph convolutional hybrid network is a network structure that combines the characteristics of multi-resolution convolutional neural networks and graph convolutional neural networks. Multi-resolution convolutional neural networks can extract features from data at different resolution levels, while graph convolutional neural networks are good at processing graph-structured data. This hybrid network can make full use of the advantages of the two networks to better process multimodal data. The multimodal force-image fusion adaptive feedback model is a model built on the multi-resolution convolutional-graph convolutional hybrid network. This model can fuse force data, such as the real-time mechanical data mentioned above, and image data, such as three-dimensional bone tissue image data, and has an adaptive feedback mechanism that dynamically adjusts the model output according to the input data to more accurately simulate the joint replacement surgery process. The initial value of attention weight is the initial setting value of the weight parameters related to the attention mechanism in the model. The attention mechanism allows the model to focus on important information when processing data. The initial value of attention weight is the starting state of attention allocation when the model starts running. The error threshold is a pre-set value used to measure the error range between the model simulation result and the actual situation or the expected result. When the simulation error is less than this preset threshold, the model is considered to have reached a certain accuracy requirement. The multimodal coupling modeling module is used to input preprocessed real-time mechanical data, three-dimensional bone tissue image data and anatomical parameters into the adaptive feedback model. Adaptive coupling modeling is performed through multimodal fusion algorithm to dynamically adjust the bone tissue stiffness distribution and prosthesis fitting parameters in the simulation, so that the force feedback in the virtual surgery matches the different bone density areas in the image, the prosthesis implantation angle and the prosthesis-bone interface fit. Among them, the multimodal fusion algorithm is used to fuse different types of data, such as real-time mechanical data, three-dimensional bone tissue imaging data, and anatomical parameters. Through this algorithm, multiple modalities of data can be organically combined to explore their intrinsic relationships and provide support for adaptive coupling modeling. Adaptive coupling modeling involves inputting pre-processed data into an adaptive feedback model, which then dynamically adjusts its internal parameters and relationships according to the multimodal fusion algorithm to achieve adaptive matching and coupling between different data, thereby more realistically simulating various phenomena during the surgical process. Bone tissue stiffness distribution refers to the distribution of bone tissue's ability to resist deformation in different parts of the body. Bone tissue stiffness varies in different bone density regions. Accurately simulating bone tissue stiffness distribution in surgical simulation is crucial for realistically reflecting the surgical experience and results. Prosthesis adaptation parameters are various parameters that describe the degree of adaptation between the prosthesis and bone tissue, such as the prosthesis implantation angle and the degree of fit between the prosthesis and bone interface. These parameters directly affect the stability and function of the prosthesis in the bone tissue, and these parameters need to be accurately adjusted in surgical simulation to achieve highly realistic simulation. The dynamic response optimization module is used to continuously optimize the adaptive feedback model through a perception-response-correction loop, dynamically update the attention weight and prosthesis adaptation parameters until the simulation error is less than the preset threshold, and achieve highly realistic dynamic response for key operations such as instrument polishing, prosthesis implantation and interface compaction in joint replacement surgery, thus completing the simulation of the entire joint replacement surgery process. The perception-response-correction loop is a mechanism for continuous model optimization. The model first perceives the input data and the current simulation state (perception phase), then makes corresponding adjustments based on the perception results (response phase). Next, it compares the model with preset standards or real-world conditions to identify and correct errors (correction phase). This process is repeated continuously to optimize model performance. Dynamically updating attention weights and prosthesis adaptation parameters involves adjusting these parameters in real-time based on the results of each iteration of the perception-response-correction loop. This allows the model to better adapt to different surgical situations and data changes, improving the accuracy and realism of the simulation. High-fidelity dynamic response for key operations addresses critical operations in joint replacement surgery, such as instrument polishing, prosthesis implantation, and interface compaction. Through the aforementioned optimization process, the model can highly realistically simulate the mechanical feedback and imaging changes during these operations, providing a reliable simulation environment for surgical training and planning. The full-process simulation of joint replacement surgery, achieved through the collaborative work of the preceding modules, simulates the entire process from start to finish, including preoperative planning, intraoperative procedures, and postoperative evaluation, providing medical personnel with a comprehensive surgical simulation experience and training platform.

[0022] It should be noted that during use, the data acquisition and preprocessing module comprehensively collects and processes various types of raw data, providing rich and accurate basic information for the simulation. The model building and parameter initialization module utilizes a multi-resolution convolutional-graph convolutional hybrid network to construct an adaptive feedback model, which can better integrate multimodal data and reasonably set initial parameters to ensure the scientific initial state of the model. The multimodal coupling modeling module can dynamically adjust key parameters, enabling precise matching of force feedback and images, thereby improving the realism of the simulation. The dynamic response optimization module continuously optimizes through a perception-response-correction loop, constantly updating parameters until the error reaches the target, achieving highly realistic dynamic responses for key operations and completing the entire simulation process. All modules of the entire system work together, from data to model to optimization, comprehensively ensuring the accuracy, realism, and comprehensiveness of the simulation, providing strong support for joint replacement surgery training and planning.

[0023] In one embodiment, raw data related to joint replacement surgery is acquired, and the raw data is preprocessed, including: Acquiring intraoperative three-dimensional bone tissue imaging data:

[0024] in , , These represent the height, width, and depth dimensions of the image, respectively. This indicates the number of frames in the image sequence, and the image data covers the core bone tissue areas of joint replacement, such as the acetabulum, proximal femur, and tibial plateau. Acquire real-time mechanical data from pressure sensors:

[0025] in, This indicates the number of pressure sensing channels, distributed across the prosthesis surface, the tip of the grinding instrument, and the contact surface of the compaction tool. This indicates the sampling time step. The mechanical data includes grinding cutting force, prosthesis implantation pressure, and interface compaction reaction force. Obtaining anatomical parameters for joint replacement surgery This includes the femoral neck-shaft angle, acetabular abduction angle, prosthesis size and model, cortical bone thickness, and cancellous bone porosity. Key parameters; The three-dimensional bone tissue image data is normalized using the following expression:

[0026] in, , , , , and These represent the minimum and maximum values ​​of the three-dimensional bone tissue image data, respectively. The real-time mechanical data is filtered and denoised using a Gaussian filtering algorithm, the expression of which is:

[0027] in, , , This represents the Gaussian filter kernel function. Indicates the half-width of the filtering window; The anatomical parameters are standardized, and the expression is:

[0028] in, , Indicates the first The mean of the anatomical parameters, Indicates the first The standard deviation of the anatomical parameters.

[0029] This design, through methods for acquiring and preprocessing raw data from joint replacement surgery, normalizes 3D bone tissue imaging data to eliminate the influence of dimensions; real-time mechanical data filtering and denoising improves data quality; and standardized processing of anatomical parameters ensures data comparability. Comprehensive and precise data acquisition and preprocessing provide a high-quality, standardized data foundation for subsequent model construction. Different types of data, after targeted processing, can more accurately reflect the actual surgical situation, avoiding model deviations due to data differences. This helps improve the model's simulation accuracy of the surgical process, making the simulation results closer to real surgical scenarios and providing a reliable basis for surgical training and planning.

[0030] In one embodiment, an adaptive feedback model for multimodal force perception-image fusion is constructed based on a multi-resolution convolutional-graph convolutional hybrid network, and the initial values ​​of attention weights and error thresholds in the model are determined, including: A multi-resolution convolutional module was constructed, and 3D convolutional kernels of different sizes were used to extract features at multiple scales from the preprocessed 3D bone tissue image data. The module focused on capturing features of key areas such as the acetabular cup coverage area, the femoral prosthesis stem implantation channel, and the tibial plateau osteotomy surface, generating multi-resolution image feature maps. ,in Indicates the resolution level, the first Layer feature map , , , and They represent the first The height, width, depth, and number of channels of the layer feature map; A graph convolution module is constructed to fuse preprocessed real-time mechanical data and anatomical parameters into a multi-dimensional feature vector. Based on the bone tissue-prosthesis interaction topology in joint replacement surgery, a graph structure is built, embedding the feature vector into graph nodes, including bone tissue partition nodes, prosthesis key site nodes, and instrument action point nodes. Graph convolution operations are used to capture the spatial dependencies between mechanical data, anatomical parameters, and surgical procedures, generating fused features.

[0031] in, Indicates the number of nodes in the graph. Indicates the dimension of the fused features; A feature fusion module is constructed, employing an attention mechanism to combine multi-resolution image feature maps with fused features. Perform adaptive weighted fusion, initial values ​​for attention weights The determination rule is as follows: it is calculated based on the information entropy ratio of multimodal data in the training set, and the expression is:

[0032] in, This represents the information entropy of three-dimensional bone tissue imaging data. This represents the information entropy after fusing real-time mechanical data with anatomical parameters. The expression for calculating information entropy is:

[0033] Represents the probability distribution of the data; Error threshold The rules for determining it are as follows: Based on the statistical characteristics of a real clinical joint replacement surgery interactive dataset, one-third of the dataset's standard deviation was used as the initial threshold. This threshold was then dynamically adjusted to reflect the precision requirements of key surgical procedures. The threshold coefficients were set as follows: 0.8 for grinding operations, 0.9 for prosthesis implantation operations, and 1.0 for interface compaction operations. The adjusted expression is as follows:

[0034] in, The standard deviation of the real surgical interaction dataset is represented by the standard deviation of the dataset. This represents the threshold coefficient corresponding to the operation type.

[0035] This design constructs a multimodal force perception-image fusion adaptive feedback model and a method for determining initial parameters. The multi-resolution convolution module captures key image features, the graph convolution module mines the spatial relationship between mechanical and anatomical parameters, and the attention mechanism fuses features. Initial weights are determined based on information entropy, and error thresholds are set in conjunction with clinical data. The multi-resolution and graph convolution modules extract features from different angles to comprehensively reflect surgical information. The attention mechanism achieves adaptive fusion, highlights important features, and reasonably determines initial parameters and error thresholds, making the model's initial state scientific and reasonable, enabling faster convergence, improving model training efficiency and simulation accuracy, and better simulating the complex interactive processes during surgery.

[0036] In one embodiment, preprocessed real-time mechanical data, three-dimensional bone tissue imaging data, and anatomical parameters are input into an adaptive feedback model. Adaptive coupling modeling is performed using a multimodal fusion algorithm to dynamically adjust the bone tissue stiffness distribution and prosthesis fitting parameters in the simulation, including: Based on multimodal fusion features Bone tissue stiffness distribution parameters are obtained through mapping of fully connected layers. Fitting parameters for implants ,in This indicates the number of bone tissue zones, which are divided according to anatomical structure into regions such as the acetabular roof, femoral tenaculum, and medial tibial plateau. Indicates the first Stiffness coefficient of each partition, This indicates the dimensions of prosthesis fitting parameters, including prosthesis implantation depth, anteversion angle, offset, soft tissue tension compensation coefficient, etc. A mapping model was established between stiffness distribution and bone mineral density and anatomical parameters. Bone mineral density grades were determined based on the grayscale values ​​of different regions in three-dimensional bone tissue images. The stiffness coefficients of corresponding regions were dynamically adjusted in conjunction with parameters such as cortical thickness and cancellous bone porosity. The mapping expression is as follows:

[0037] in, Indicates the reference stiffness coefficient. Indicates the first Normalized bone mineral density values ​​for each region, This represents the stiffness adjustment coefficient, corresponding to cortical bone. Cancellous bone corresponds to , Indicates the first Cortical thickness in each zone, Indicates average cortical thickness. Indicates the first Porosity of cancellous bone in each zone This represents the average porosity of cancellous bone. A correlation model was established between prosthesis fitting parameters and anatomical parameters and biomechanical data. Initial prosthesis fitting parameters were determined based on anatomical parameters such as the femoral neck-shaft angle and acetabular abduction angle. Adjustments were made using real-time biomechanical data feedback to ensure uniform force distribution and interface fit during prosthesis implantation. The correlation model expression is as follows:

[0038] in, Indicates the first The initial values ​​of each adaptation parameter are obtained by looking up a table based on anatomical parameters. Indicates the first The adjustment coefficient of each parameter, This represents the average force between the device and bone tissue at the current moment. Indicates the first The reference force values ​​corresponding to each parameter; It receives mechanical data updates from pressure sensors and image data frame updates in real time, and adaptively adjusts stiffness distribution parameters and prosthesis fitting parameters based on a multimodal fusion algorithm to achieve dynamic matching of virtual surgical force feedback with bone density region, prosthesis implantation status, and soft tissue tension.

[0039] This design adaptively couples preprocessed data into the model, dynamically adjusting bone stiffness distribution and prosthesis fitting parameters. By establishing a mapping and correlation model, parameters are determined based on imaging, anatomical, and mechanical data, and are updated in real time. This allows for dynamic adjustment of key parameters according to the actual surgical situation, ensuring that force feedback in virtual surgery matches bone density, prosthesis status, and other parameters in real time. This dynamic adjustment mechanism makes the simulation more closely resemble real surgical changes, improving the realism and interactivity of the simulation. It provides medical personnel with a more realistic surgical experience, helping to improve surgical skills and the ability to handle complex situations.

[0040] In one embodiment, the adaptive feedback model is continuously optimized through a perception-response-correction loop, dynamically updating the attention weights and prosthesis adaptation parameters until the simulation error is less than a preset threshold, including: Sensing phase: The pressure sensor senses the mechanical signals of the instrument-bone / prosthesis interaction during key operations of joint replacement surgery in real time. Combined with the real-time update of intraoperative three-dimensional bone tissue images and dynamic correction of anatomical parameters, the interaction status data of the current surgical operation is obtained, including bone tissue cutting amount, prosthesis implantation progress, interface pressure distribution, etc. Key operations include: grinding, implantation, and compaction; Response phase: Based on the current parameters of the adaptive feedback model, simulate bone tissue deformation, prosthesis placement process, and interface stress transmission during joint replacement surgery, and generate dynamic response outputs such as instrument force response, prosthesis displacement trajectory, and tissue deformation degree; Calibration phase: Calculate the error between the dynamic response output and the preset real surgical interaction data. The weighted error calculation method focuses on the errors in prosthesis implantation angle, interface pressure distribution, and bone tissue cutting amount. The expression is as follows: in, , , These are the weighting coefficients for angle error, pressure error, and cutting amount error, respectively, with a total value of 1. The default values ​​are 0.3, 0.4, and 0.3, respectively. , , These represent the number of validation samples corresponding to the error. , , These represent the actual surgical angle, pressure, and cutting amount data, respectively. , , These represent the corresponding data of the simulated response; The dynamic update rule for attention weights is: adaptive adjustment based on error feedback and surgical procedure type, expressed as:

[0041] in, This represents the updated attention weights. Indicates the current attention weight. Represents the learning rate (within a range of values). ), This represents the partial derivative of the error with respect to the current attention weights. Indicates the surgical procedure type coefficient, and the grinding procedure. Implantation procedure compaction operation ; The dynamic update rule for prosthesis fitting parameters is: based on error feedback and anatomical parameter constraints, the expression is:

[0042] in, Indicates the updated number One adaptation parameter, Indicates the current number One adaptation parameter, This represents the learning rate, a parameter for adaptation, and its value ranges from [value missing]. , This represents the partial derivative of the error with respect to the current fitting parameters. , They represent the first The anatomically reasonable range threshold of each fitting parameter; Based on error The gradient descent algorithm is used to update other parameters of the adaptive feedback model, including the weight parameters of the multi-resolution convolutional-graph convolutional hybrid network and the adjustment coefficients of the stiffness mapping relationship model, to complete one perception-response-correction loop; Repeat the above cyclical process until the error... Less than the preset threshold This enables continuous optimization of the model and outputs highly realistic dynamic responses throughout the entire joint replacement surgery process.

[0043] This design employs a perception-response-correction cyclic optimization method for the adaptive feedback model. The perception phase acquires surgical interaction state data, the response phase generates dynamic response output, and the correction phase calculates errors and updates attention weights and prosthesis fitting parameters until the error falls below a threshold. This continuous optimization mechanism constantly improves the model and enhances simulation accuracy. Through real-time perception and response, model parameters can be adjusted promptly to adapt to surgical changes. Focusing on key errors, the simulation results are ensured to be highly consistent with real surgery in important aspects. Ultimately, this achieves a highly realistic dynamic response throughout the entire joint replacement surgery process, providing a high-quality simulation environment for surgical training and planning.

[0044] Example 2 Data acquisition equipment: 3D image acquisition equipment: A 128-slice spiral CT scanner with a slice thickness of 0.625mm was used. The scanning range covered the pelvis, proximal femur, and soft tissues around the hip joint to acquire 3D bone tissue image data. The resolution was set to 512×512×256 (H×W×D). The number of image sequence frames was M=30, with one frame acquired every 10 seconds for 5 minutes.

[0045] Pressure sensing devices: A fiber optic pressure sensor array with a total of K=16 sensing channels is used. Four channels are distributed on the surface of the acetabular prosthesis, evenly distributed on the dome and anterior and posterior walls of the prosthesis; six channels are distributed on the surface of the femoral stem prosthesis, evenly arranged along the long axis of the prosthesis stem; three channels are distributed on the tip of the acetabular grinding instrument; and three channels are distributed on the contact surface of the interface compaction tool. The sensor range is 0-500N, the sampling frequency is 100Hz, and the sampling time step is T=0.01s.

[0046] Anatomical parameter measuring equipment: Anatomical parameters, including femoral neck-shaft angle, acetabular abduction angle, acetabular depth, proximal femoral medullary canal diameter, cortical thickness, and cancellous bone porosity (S=8 key parameters), are automatically extracted using a medical imaging workstation. At the same time, parameters such as prosthesis size and model are obtained by combining them with the prosthesis specification manual.

[0047] Analog interactive devices: A force feedback robotic arm with a force feedback resolution of 0.01N and a position resolution of 0.03mm is used to simulate the operation of surgical instruments and force feedback interaction.

[0048] Raw data acquisition: CT imaging data of 50 patients who underwent total hip arthroplasty were selected (approved by ethics review). The imaging data were in DICOM format and covered patients of different ages (45-75 years) and bone mineral density grades (normal, osteopenia, osteoporosis) to ensure data diversity.

[0049] Mechanical data during the simulated surgery were collected using pressure sensing devices, including the cutting force during acetabular grinding (range 5-80N), the axial pressure during femoral stem prosthesis implantation (range 50-300N), and the reaction force after compaction of the prosthesis-bone interface (range 30-150N).

[0050] Anatomical parameters were extracted using a medical imaging workstation. Examples of parameters are as follows: femoral neck-shaft angle 125-135°, acetabular abduction angle 35-45°, cortical thickness 2-5mm, cancellous bone porosity 30%-60%, and prosthesis size model with diameter 44-56mm (acetabular prosthesis) and stem diameter 12-16mm (femoral stem prosthesis).

[0051] Data preprocessing: 3D bone tissue image data normalization: grayscale values ​​of each frame of DICOM image data are normalized and mapped to [0,1], where min(Y)=0 (air region) and max(Y)=255 (cortical bone region). The processed image data is stored in NIfTI format.

[0052] Real-time mechanical data filtering: A Gaussian filtering algorithm is used, with the filter window half-width n=3, and the Gaussian filter kernel function... ,in =1.0, filtering sensor noise and high-frequency interference to ensure smooth mechanical data.

[0053] Anatomical parameter standardization: Mean values ​​were calculated based on anatomical parameters from 50 patients. and standard deviation For example, the femoral neck-shaft angle μ=130°, σ=3°, and the acetabular abduction angle μ=40°, σ=2°. All parameters are standardized to the range of [-1,1] to eliminate dimensional differences.

[0054] Model building and parameter initialization: Construction of multi-resolution convolutional-graph convolutional hybrid networks: Multi-resolution convolutional module: Employs 3 layers of 3D convolutional layers with kernel sizes of 3×3×3, 5×5×5, and 7×7×7, all with a stride of 2 and same padding. ReLU activation function is used, and each convolutional layer is followed by a BatchNorm layer and a Dropout layer (dropout rate = 0.2). Outputs L = 3 resolution level image feature maps, where the first layer (high resolution) feature map size is 256×256×128×64, the second layer (medium resolution) is 128×128×64×128, and the third layer (low resolution) is 64×64×32×256.

[0055] The graph convolution module constructs a graph structure containing N=128 nodes, including 32 bone tissue partitioning nodes (divided according to the anatomical structure of the acetabulum and proximal femur into regions such as the acetabular roof, anterior acetabular wall, posterior acetabular wall, femoral distance, and proximal femoral medullary canal), 64 prosthesis key site nodes (32 sites for the acetabular prosthesis and 32 sites for the femoral stem prosthesis), and 32 instrument action point nodes (8 for grinding instruments, 12 for implantation tools, and 12 for compaction tools). The preprocessed mechanical data (16 channels × T time steps) and anatomical parameters (8 items) are fused into a 24-dimensional feature vector and embedded into the corresponding graph nodes. A two-layer graph convolutional layer is used, with the first layer outputting the feature dimension. The second layer outputs feature dimensions. The activation function used is GELU.

[0056] Feature fusion module: Employs a multi-head attention mechanism (8 heads) to perform weighted fusion of multi-resolution image feature maps (aggregated dimension 256) and graph convolutional fusion features (dimensional 256). Initial values ​​for attention weights are... Based on information entropy calculation, where H_img = 4.2 (image data information entropy) and H_force-ana = 3.8 (mechanical-anatomical fusion data information entropy), then... .

[0057] Error threshold determination: Based on an interactive dataset of 50 real clinical surgeries (including surgical biomechanics data, prosthesis implantation angle, interface pressure distribution, etc.), the dataset standard deviation std(Y_real) was calculated to be 5.0 (normalized error value). Therefore, the initial threshold was 1 / 3 × 5.0 ≈ 1.67. The operational threshold was then refined by adjusting for surgical operation type. Prosthesis implantation procedure Interface compaction operation .

[0058] Multimodal coupling modeling: Feature fusion and parameter mapping: Multi-resolution image feature maps are aggregated into a 256-dimensional vector using global average pooling, and then weighted and fused with graph convolutional fusion features (256-dimensional) through an attention mechanism to obtain multimodal fusion features. (Wei); will Input is fed into 3 fully connected layers (hidden layer dimensions 512, 256, and 128), outputting bone tissue stiffness distribution parameters. (Stiffness coefficient of each partition) and prosthesis fitting parameters The following parameters are used for fitting: prosthesis implantation depth, acetabular prosthesis anteversion angle, acetabular prosthesis abduction angle, femoral stem prosthesis anteversion angle, femoral stem offset, and soft tissue tension compensation coefficient.

[0059] Dynamic adjustment of stiffness distribution: Reference stiffness coefficient (Baseline value for cancellous bone), based on normalized bone mineral density value Image grayscale mapping, cortical bone , cancellous bone Osteoporosis area Stiffness adjustment coefficient λ (cortical bone λ=1.4, cancellous bone λ=0.7), and cortical thickness ratio. and the proportion of cancellous bone porosity Calculate the stiffness coefficients for each zone; for example, the acetabulum, predominantly cortical bone.

[0060] Femoral talus cancellous bone region

[0061] Dynamic adjustment of implant fitting parameters: Initial fitting parameters for implants Obtained by looking up anatomical parameters, such as the initial value of the acetabular prosthesis anteversion angle. (Based on a table referenced at a hip abduction angle of 40°), initial value of femoral stem eccentricity. (Based on a proximal femoral medullary canal diameter of 14mm, refer to the table); adjustment coefficient. The settings are as follows: implantation depth δ=0.5mm, anteversion angle δ=1°, abduction angle δ=1°, eccentricity δ=0.5mm, soft tissue tension compensation coefficient δ=0.05; reference force value. Based on clinical data, such as the mean force reference value for prosthesis implantation. When the current average force At that time, the depth of implantation This ensures uniform implantation force.

[0062] Dynamic response optimization and surgical simulation: The perception-response-correction cycle is executed: Perception phase: The force feedback robotic arm captures the doctor's operating actions (such as the rotation angle and advance speed during acetabular grinding, and the axial thrust and rotation angle during prosthesis implantation). The pressure sensor collects interactive mechanical signals in real time (sampling frequency 100Hz). The CT image is updated every 10 seconds, and the status data such as bone tissue cutting amount (calculated through image registration), prosthesis implantation progress (current implantation depth / target depth), and interface pressure distribution (sensor array data) are acquired simultaneously.

[0063] Response phase: Based on the current parameters of the MR-CGCN model, simulate bone tissue deformation (using a simplified finite element model and combining stiffness distribution parameters to calculate the degree of deformation), prosthesis placement process (simulating prosthesis displacement and rotation based on adaptation parameters), and interface stress transmission (outputting real-time force feedback through a force feedback robotic arm); for example, when grinding the cortical bone area, the robotic arm outputs greater resistance (corresponding to a high stiffness coefficient), while the resistance decreases when grinding the cancellous bone area, simulating the feel of cutting real tissue.

[0064] Correction phase: Calculate the weighted error E, where the angle error weights are... (Pay attention to the errors in the prosthesis's anteversion and abduction angles), and the weighting of the pressure error. (Focus on the uniformity of interface pressure distribution) and the weight of cutting quantity error. (Focusing on the precision of bone tissue cutting); If the current surgical procedure involves prosthesis implantation, the calculated root mean square error of angle is 1.2°, the root mean square error of pressure is 8N, and the root mean square error of cutting amount is 0.2mm. After normalization, the error E = 0.3 × 0.8 + 0.4 × 0.6 + 0.3 × 0.4 = 0.24 + 0.24 + 0.12 = 0.60, which is less than the threshold. No significant adjustments are needed; if the error E=1.80 (greater than the threshold), then parameter updates will be initiated.

[0065] Parameter updates and model optimization: Attention weight update: Set the learning rate η = 0.005, the surgical procedure type coefficient ζ = 1.0 (prosthesis implantation), and calculate the partial derivative of the error with respect to the attention weights. ,but Increase the weight of image features to improve the accuracy of prosthesis positioning.

[0066] Implant fitting parameters updated: Set the adaptation parameter learning rate γ=0.01, for example, the current value of the acetabular prosthesis anteversion angle. , Error partial derivative Reasonable range of anatomical lean angle , ,but Fine-tune the lean angle to reduce angle error.

[0067] Other model parameters updated: The Adam optimizer (learning rate = 0.001, β1 = 0.9, β2 = 0.999) is used to update the weight parameters of the convolutional and graph convolutional layers of MR-CGCN and the adjustment coefficients of the stiffness mapping relationship model based on the error E, completing one loop (loop period = 0.01s, consistent with the sampling time step).

[0068] Example 3 Full surgical procedure simulation: Grinding stage: The simulated acetabular grinding instrument is advanced from the initial position to the acetabular cartilage surface. The grinding resistance is dynamically adjusted based on the stiffness distribution parameters. When grinding reaches the cortical bone area, the force feedback robotic arm outputs increased resistance. At the same time, the progress of bone tissue cutting is displayed through image updates to ensure that the acetabular cup coverage area is ≥80% after grinding.

[0069] Prosthesis implantation stage: First, the acetabular prosthesis is implanted. The implantation angle and depth are adjusted based on the fitting parameters. The pressure distribution at the prosthesis-bone interface is displayed in real time (visualized through sensor data) to ensure uniform pressure (maximum pressure difference ≤20%). Then, the femoral stem prosthesis is implanted to simulate the compression fitting process between the prosthesis stem and the medullary cavity. The force feedback gradually increases with the implantation depth to avoid over-implantation that could lead to bone tissue damage.

[0070] Interface compaction stage: The prosthesis-bone interface is compacted using a compaction tool to simulate the interface pressure adjustment before bone cement hardening. The force feedback robotic arm outputs the compaction reaction force to ensure that the interface fit is ≥95%. The entire surgical process is simulated, with a total simulation time of ≈30 minutes, which is consistent with the clinical surgical time.

[0071] Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0072] The above embodiments provide a detailed description of the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A joint replacement surgery simulation system based on pressure sensing feedback, characterized in that, The system includes: The module includes data acquisition and preprocessing, model building and parameter initialization, multimodal coupling modeling, and dynamic response optimization. The data acquisition and preprocessing module is used to acquire and preprocess raw data related to joint replacement surgery. The raw data includes intraoperative three-dimensional bone tissue imaging data, real-time mechanical data of instrument-bone tissue interaction, and surgical-related anatomical parameters. The model building and parameter initialization module is used to build a multimodal fusion adaptive feedback model and determine the initial parameters of the model. The multimodal coupling modeling module is used to input preprocessed data into the adaptive feedback model, and realize adaptive coupling modeling through multimodal fusion algorithm to dynamically adjust simulation parameters; The dynamic response optimization module is used to continuously optimize the model through a cyclic optimization mechanism to achieve a highly realistic simulation of the entire joint replacement surgery process.

2. The joint replacement surgery simulation system based on pressure sensing feedback according to claim 1, characterized in that, The intraoperative three-dimensional bone tissue imaging data acquired by the data acquisition and preprocessing module covers the core bone tissue region of joint replacement, which includes: At least one of the following: acetabulum, proximal femur, and tibial plateau; The real-time mechanical data is collected by pressure sensors distributed on the surface of the prosthesis, the tip of the grinding instrument and the contact surface of the compaction tool, including at least one of grinding cutting force, prosthesis implantation pressure and interface compaction reaction force. The anatomical parameters include at least one of the following key parameters: femoral neck-shaft angle, acetabular abduction angle, prosthesis size, cortical bone thickness, and cancellous bone porosity.

3. The joint replacement surgery simulation system based on pressure sensing feedback according to claim 1, characterized in that, The data acquisition and preprocessing module includes the following steps: The three-dimensional bone tissue image data were normalized, the real-time mechanical data were filtered and denoised, and the anatomical parameters were standardized. The filtering and denoising process employs algorithms including Gaussian filtering, while the normalization and standardization processes optimize data consistency by eliminating differences in data units.

4. The joint replacement surgery simulation system based on pressure sensing feedback according to claim 1, characterized in that: The model building and parameter initialization module constructs an adaptive feedback model based on a multi-resolution convolutional-graph convolutional hybrid network, which includes a multi-resolution convolutional module, a graph convolutional module, and a feature fusion module. The multi-resolution convolution module is used to extract multi-scale features from three-dimensional bone tissue image data; The graph convolution module is used to fuse real-time mechanical data and anatomical parameters into a multi-dimensional feature vector, and to construct a graph structure based on the bone tissue-prosthesis interaction topology. The feature fusion module employs an attention mechanism to adaptively weight and fuse multi-resolution image feature maps and fused features.

5. The joint replacement surgery simulation system based on pressure sensing feedback according to claim 4, characterized in that, The initial parameters determined by the model building and parameter initialization module include: Initial values ​​for attention weights and error threshold; The initial value of the attention weight is determined based on the information entropy ratio of the multimodal data in the training set; The error threshold is set based on the statistical characteristics of a real clinical joint replacement surgery interactive dataset and is dynamically corrected in combination with the accuracy requirements of key surgical operations, including at least one of grinding, prosthesis implantation, and interface compaction.

6. The joint replacement surgery simulation system based on pressure sensing feedback according to claim 1, characterized in that: The multimodal coupling modeling module obtains bone tissue stiffness distribution parameters and prosthesis fitting parameters through a multimodal fusion algorithm; The bone tissue stiffness distribution parameters are set according to the anatomical structure by dividing the bone tissue into zones, and the zones include at least one of the acetabular roof, femoral distance, and medial tibial plateau. The prosthesis fitting parameters include at least one of the following dimensional parameters: prosthesis implantation depth, tilt angle, offset, and soft tissue tension compensation coefficient.

7. The joint replacement surgery simulation system based on pressure sensing feedback according to claim 6, characterized in that: The multimodal coupling modeling module establishes a mapping relationship model between stiffness distribution and bone density and anatomical parameters. It determines the bone density level based on the gray values ​​of different regions in the three-dimensional bone tissue image and dynamically adjusts the stiffness coefficient of the corresponding region in combination with cortical thickness and cancellous bone porosity parameters. The mapping relationship model achieves dynamic stiffness adaptation by associating the baseline stiffness coefficient, normalized bone mineral density value, stiffness adjustment coefficient, and anatomical parameters.

8. The joint replacement surgery simulation system based on pressure sensing feedback according to claim 6, characterized in that: The multimodal coupling modeling module establishes a correlation model between prosthesis fitting parameters and anatomical parameters and mechanical data. Based on the anatomical parameters, it determines the initial fitting parameters of the prosthesis and adjusts the fitting parameters through real-time mechanical data feedback to ensure uniform force distribution and interface fit during the prosthesis implantation process. The correlation model achieves dynamic optimization through initial adaptation parameters, parameter adjustment coefficients, and the difference between the real-time interactive average force and the reference force value.

9. The joint replacement surgery simulation system based on pressure sensing feedback according to claim 1, characterized in that, The dynamic response optimization module continuously optimizes the adaptive feedback model through a perception-response-correction loop, which includes: The three phases of perception, response, and correction. The perception phase acquires interactive state data of the current surgical operation, including at least one of bone tissue cutting amount, prosthesis implantation progress, and interface pressure distribution. The response phase simulates bone tissue deformation, prosthesis placement process, and interface stress transmission during surgical operations, generating dynamic response output. The correction phase calculates the error between the dynamic response output and the preset real surgical interaction data, and updates the model parameters based on the error feedback.

10. The joint replacement surgery simulation system based on pressure sensing feedback according to claim 9, characterized in that, The model parameters updated by the dynamic response optimization module include: Attention weights, prosthesis adaptation parameters, weight parameters of multi-resolution convolutional-graph convolutional hybrid networks, and adjustment coefficients of stiffness mapping relationship models; The attention weights are adaptively adjusted based on error feedback and surgical operation type, and the prosthesis fitting parameters are adjusted based on error feedback and anatomical parameter constraints. The optimization process repeats the sensing-response-correction cycle until the simulation error is less than a preset threshold, thereby achieving a highly realistic dynamic response for key operations in joint replacement surgery.

Images