Orthopedic surgery risk assessment method and system

Abstract

The invention relates to the field of orthopedic surgery risk assessment, in particular to an orthopedic surgery risk assessment method and system. The method comprises the following steps: acquiring preoperative data and intra-operative data of an orthopedic surgery of a patient, performing data preprocessing on the preoperative data and the intra-operative data, and extracting bone microstructure spatial features in initial orthopedic surgery data by using a cascade feature extraction network; calculating an association weight of the image data and the biological signal through a cross-modal attention mechanism, and constructing a patient specific skeleton model based on finite element analysis to perform real-time scene simulation on the characteristic orthopedic surgery data; the influence of joint contact stress distribution under different osteotomy angles, a bone implant interface micro-motion threshold after implantation of an internal fixation instrument and hemodynamic changes on a bone healing microenvironment is simulated in real time, complications caused by misoperation are remarkably reduced, and the success rate of an operation is increased.

Description

Technical Field

[0001] The present invention relates to the field of medical insurance technology, and in particular to an orthopedic surgery risk assessment method and system. Background Art

[0002] Traditional orthopedic surgical risk assessment mainly relies on static clinical indicators and two-dimensional imaging features, which have significant limitations. Existing technologies are mostly based on isolated data sources, lack effective integration of multimodal data, and cannot fully reflect individual differences among patients. Traditional assessment models are only run once before surgery and cannot perceive the mechanical changes of bone-instrument interactions and fluctuations of vital signs during surgery in real time, resulting in delayed response to sudden risks. Existing finite element simulations mostly use group average parameters, ignoring key features such as patient-specific bone density distribution and microstructural anisotropy. In addition, the prediction of postoperative complications and long-term follow-up evaluation are independent of each other, lacking a full-cycle risk management framework covering "intraoperative-perioperative-long-term". Summary of the Invention

[0003] The purpose of the present invention is to solve the above problems and to design a method and system for orthopedic surgery risk assessment.

[0004] To achieve the above-mentioned purpose, the technical solution of the present invention is as follows: in an orthopedic surgery risk assessment method, the orthopedic surgery risk assessment method comprises the following steps:

[0005] Acquiring preoperative data and intraoperative data of a patient's orthopedic surgery, and performing data preprocessing on the preoperative data and the intraoperative data to obtain initial orthopedic surgery data;

[0006] A cascaded feature extraction network is used to extract the spatial features of bone microstructure, the time series signals of instrument stress, and the topological relationship of the biomechanical conduction path in the initial orthopedic surgery data; and a cross-modal attention mechanism is used to calculate the correlation weight between the image data and the biological signal to obtain characteristic orthopedic surgery data.

[0007] Constructing a patient-specific bone model based on finite element analysis to perform real-time scenario simulation on the characteristic orthopedic surgery data;

[0008] Real-time simulation of the joint contact stress distribution at different osteotomy angles, the micro-motion threshold of the bone-implant interface after internal fixation device implantation, and the impact of hemodynamic changes on the bone healing microenvironment to obtain the risk status of orthopedic surgery;

[0009] Furthermore, in the above-mentioned orthopedic surgery risk assessment method, the step of obtaining preoperative data and intraoperative data of the patient's orthopedic surgery and performing data preprocessing on the preoperative data and intraoperative data to obtain initial orthopedic surgery data includes:

[0010] Obtaining preoperative and intraoperative data from orthopedic surgery, wherein the preoperative data includes at least: high-resolution CT images, MRI images, gait analysis data, serum bone metabolism markers, and susceptibility gene SNP sites; and the intraoperative data includes at least: bone density impedance signals, internal fixation device stress distribution, and patient vital sign time series;

[0011] Aligning the space and time of the CT images and MRI images in the preoperative data with the intraoperative data using an affine transformation matrix, detecting outliers in the time series data of the intraoperative data using a Grubbs test, deleting the detected outliers, and processing the genetic data using hierarchical normalization to obtain first orthopedic surgery data;

[0012] After enhancing the image data in the first orthopedic surgery data, the gene data and phenotypic data are corrected using KL divergence distribution to obtain initial orthopedic surgery data.

[0013] Furthermore, in the above-mentioned orthopedic surgery risk assessment method, the use of a cascaded feature extraction network to extract the topological relationship between the bone microstructure spatial features, instrument stress timing signals, and biomechanical conduction paths in the initial orthopedic surgery data includes:

[0014] Define the cascade feature extraction network as a triplet Its output characteristics satisfy the following formula:

[0015]

[0016] Among them, Φ fuse represents the cross-modal fusion function, X represents the image data, S represents the stress time series signal, and A represents the biomechanical topological adjacency matrix; Represents a 3D convolutional neural network; Represents an LSTM network; Represents GNN graph neural network, F total represents the output features;

[0017] Extracting bone microstructure spatial features from the initial orthopedic surgery data using the 3D convolutional neural network; extracting instrument stress temporal features from the initial orthopedic surgery data using the LSTM network; and obtaining spatiotemporal orthopedic surgery data;

[0018] Based on the GNN graph neural network, the topological relationship of the biomechanical conduction path is established and the biomechanical conduction map is defined. node represents the bone substructure segmented from the CT image, and the edge e ij ∈ε represents the force conduction path determined based on finite element simulation, and the edge weight ω ij =exp(-||fi -f j || 2 ), f i and f j Represents the mechanical eigenvectors of nodes i and j;

[0019] Using RGCN relational graph convolution network for anisotropic graph convolution:

[0020]

[0021] in, represents the hidden state of node i in layer l+1, l represents the number of hidden layers, R represents the set of predefined biomechanical relationship types, represents the set of neighbor nodes of node i under relationship r, Indicates the size of the neighbor node set, used for normalization; Represents the weight matrix of relation r at layer l, used to transform the hidden state of neighboring nodes ReLU represents the activation function, which performs a nonlinear transformation on the weighted summation result.

[0022] Furthermore, in the above-mentioned orthopedic surgery risk assessment method, the correlation weights between the image data and the biological signals are calculated by the cross-modal attention mechanism to obtain the characteristic orthopedic surgery data, including:

[0023] defining a joint embedding mapping of the initial orthopedic surgery data between the imaging modality and the biological signal modality in a cascaded feature extraction network feature extraction;

[0024] A bidirectional query-key-value interaction structure is established, an inter-modality weight matrix is generated through a gating network, Kullback-Leibler divergence loss is added to promote attention sparsity, and the dynamic association weights of the cascaded feature extraction network are optimized;

[0025] The correlation strength of the cascade feature extraction network in extracting initial orthopedic surgery data is adjusted by the real-time biological signal change rate to obtain characteristic orthopedic surgery data.

[0026] Furthermore, in the above-mentioned orthopedic surgery risk assessment method, the patient-specific bone model is constructed based on finite element analysis to perform real-time scenario simulation on the characteristic orthopedic surgery data, including:

[0027] A patient-specific bone model is constructed based on finite element analysis, wherein the patient-specific bone model includes a spatially variable material model based on the nonlinear relationship between CT value and bone density:

[0028] E(x)=α·ρ(x) β +γ·softplus(GSNP );

[0029] Where E(x) represents the material strength at voxel x, ρ(x) = a·HU(x) + b represents the apparent density at voxel x, a and b represent the correlation coefficients, and HU(x) represents the CT value of voxel x; G SNP represents the correction term of gene characteristics for material strength, α=2045, β=1.55, γ=0.3 represents clinical calibration parameters, which are used to quantify the relationship between material strength and other variables, and softplus represents a nonlinear function, which is used to correct G SNP Perform transformations;

[0030] The patient-specific bone model also includes conformal tetrahedral mesh generation using a modified Marching Cubes algorithm:

[0031]

[0032] in, Represents the generated conformal tetrahedral mesh area, which is a set of three-dimensional points that meet specific conditions; N represents the number of elements in the set, V i Represents a point in three-dimensional space, constituting the basic element of the grid, φ represents the signed distance function of the CT image, τ represents the threshold parameter, and δ represents the adaptive offset based on the direction of the trabeculae; represents the signed distance function, which is used to adjust the threshold according to the gradient direction of Φ;

[0033] Based on GPU acceleration of the preconditioned conjugate gradient method, a block diagonal preprocessing matrix of the sparse stiffness matrix is constructed, and parallel optimization of matrix-vector multiplication is performed under the CUDA architecture to obtain a finite element solver for the specific bone model.

[0034] Furthermore, in the above-mentioned orthopedic surgery risk assessment method, the real-time simulation of the joint contact stress distribution at different osteotomy angles, the micro-motion threshold of the bone-implant interface after internal fixation device implantation, and the impact of hemodynamic changes on the bone healing microenvironment to obtain the orthopedic surgery risk status includes:

[0035] The real-time simulation of the joint contact stress distribution at different osteotomy angles includes calculating the angle change range with a resolution of 0.1° based on the characteristic orthopedic surgery data, determining the joint contact stress peak value and distribution pattern at each angle, and obtaining a risk result.

[0036] The micro-motion threshold of the bone implant interface after the internal fixation device is implanted includes the strain signal of the piezoelectric film array on the object surface in the characteristic orthopedic surgery data, the change in the propagation characteristics of the bone acoustic guided wave, and the interface gap data of the intraoperative OCT imaging. A three-dimensional feature vector of the micro-motion amplitude, frequency, and spatial distribution is established to determine whether the current micro-motion pattern belongs to the high-risk type, and obtain a Class II risk result.

[0037] The impact of hemodynamic changes on the bone healing microenvironment includes generating a four-dimensional bone healing progress map based on blood flow simulation results, quantifying local repair capacity using the MA I metabolic activity index, determining the regional MA I warning line for 5 consecutive minutes, and obtaining three types of risk results;

[0038] The three types of risk assessment results are dynamically weighted to obtain the risk status of orthopedic surgery.

[0039] To achieve the above-mentioned purpose, the technical solution of the present invention is, further, in an orthopedic surgery risk assessment system, the orthopedic surgery risk assessment system comprises:

[0040] A surgical data acquisition module is used to acquire preoperative data and intraoperative data of a patient's orthopedic surgery, and perform data preprocessing on the preoperative data and intraoperative data to obtain initial orthopedic surgery data;

[0041] a data feature fusion module for extracting the spatial features of bone microstructures, instrument stress time series signals, and topological relationships of biomechanical conduction pathways from the initial orthopedic surgery data using a cascaded feature extraction network; and calculating the correlation weights between image data and biological signals using a cross-modal attention mechanism to obtain characteristic orthopedic surgery data;

[0042] A real-time scene simulation module, used for constructing a patient-specific bone model based on finite element analysis to perform real-time scene simulation on the characteristic orthopedic surgery data;

[0043] The risk status assessment module is used to simulate in real time the joint contact stress distribution at different osteotomy angles, the micro-motion threshold of the bone-implant interface after internal fixation device implantation, and the impact of hemodynamic changes on the bone healing microenvironment to determine the risk status of orthopedic surgery;

[0044] Furthermore, in the above-mentioned orthopedic surgery risk assessment system, the surgery data acquisition module includes the following submodules:

[0045] An acquisition submodule is used to acquire preoperative and intraoperative data of patients undergoing orthopedic surgery. The preoperative data includes at least: high-resolution CT images, MRI images, gait analysis data, serum bone metabolism markers, and susceptibility gene SNP sites; the intraoperative data includes at least: bone density impedance signals, internal fixation device stress distribution, and patient vital sign time series;

[0046] an alignment submodule, configured to align the space and time of the CT images and MRI images in the preoperative data with the intraoperative data using an affine transformation matrix, detect outliers in the time series data in the intraoperative data using a Grubbs test, delete the detected outliers, and process the genetic data using hierarchical normalization to obtain first orthopedic surgery data;

[0047] A submodule is obtained, which is used to enhance the image data in the first orthopedic surgery data, and then correct the gene data and phenotypic data through KL divergence distribution to obtain initial orthopedic surgery data.

[0048] Furthermore, in the above-mentioned orthopedic surgery risk assessment system, it is characterized in that the data feature fusion module includes the following submodules:

[0049] A definition submodule is used to define a joint embedding mapping of the image modality and the biological signal modality of the initial orthopedic surgery data in the cascade feature extraction network feature extraction;

[0050] Establishing submodules for establishing a bidirectional query-key-value interaction structure, generating an inter-modality weight matrix through a gating network, adding Kullback-Leibler divergence loss to promote attention sparsity, and optimizing the dynamic association weights of the cascaded feature extraction network;

[0051] The extraction submodule is used to adjust the association strength of the cascade feature extraction network in extracting initial orthopedic surgery data through the real-time biological signal change rate to obtain characteristic orthopedic surgery data.

[0052] Its beneficial effect lies in that, by fusing preoperative imaging data with intraoperative biomechanical signals, it effectively integrates fragmented data, eliminates noise interference, and forms structured initial orthopedic surgical data, solving the problems of isolated data and inconsistent formats in traditional surgery. The cascaded feature extraction network breaks through the limitations of single-modality data; at the same time, the cross-modal attention mechanism strengthens the semantic association between different modalities by calculating the correlation weights between imaging data and biological signals, achieving in-depth analysis of bone mechanical behavior and instrument-bone interface interactions in surgical scenarios, and avoiding the problem of feature omission caused by modal splitting in traditional methods. The patient-specific bone model constructed based on finite element analysis can accurately reflect the geometric morphology, material properties and biomechanical characteristics of individual bones, solving the problem of disconnection between traditional general models and the actual patient situation. By simulating in real time the effects of different osteotomy angles and internal fixation implantation schemes on joint contact stress, interface micro-motion threshold and hemodynamics, it can provide surgeons with a preliminary assessment of surgical results and support dynamic adjustment of surgical plans during surgery. By quantitatively analyzing the impact of surgical operations on the biomechanical environment of bones, empirical surgical decisions can be transformed into data-driven precise regulation, significantly reducing complications caused by improper operations and improving surgical success rates. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Various other advantages and benefits will become apparent to those skilled in the art by reading the following detailed description of the preferred embodiment.The accompanying drawings are only for the purpose of illustrating the preferred embodiment and are not to be considered as limiting the present invention.

[0054] Figure 1 This is a schematic diagram of a first embodiment of a method for risk assessment of orthopedic surgery according to an embodiment of the present invention;

[0055] Figure 2 This is a schematic diagram of a second embodiment of a method for risk assessment of orthopedic surgery according to an embodiment of the present invention;

[0056] Figure 3 This is a schematic diagram of a third embodiment of a method for risk assessment of orthopedic surgery according to an embodiment of the present invention;

[0057] Figure 4 This is a schematic diagram of a first embodiment of an orthopedic surgery risk assessment system in an embodiment of the present invention. DETAILED DESCRIPTION

[0058] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0059] Those skilled in the art will appreciate that, unless otherwise stated, the singular forms "a," "an," "said," and "said" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present invention refers to the presence of the stated features, integers, steps, operations, elements, and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0060] The present invention will be described in detail below with reference to the accompanying drawings. Figure 1 As shown, a method for risk assessment of orthopedic surgery includes the following steps:

[0061] Step 101: obtaining preoperative data and intraoperative data of a patient's orthopedic surgery, and performing data preprocessing on the preoperative data and the intraoperative data to obtain initial orthopedic surgery data;

[0062] Specifically, in this embodiment, preoperative and intraoperative data of a patient undergoing orthopedic surgery are obtained. The preoperative data include at least: high-resolution CT images, MRI images, gait analysis data, serum bone metabolism markers, and susceptibility gene SNP sites; the intraoperative data include at least: bone density impedance signals, internal fixation device stress distribution, and patient vital sign time series.

[0063] The affine transformation matrix was used to align the space and time of the CT and MRI images in the preoperative data with the intraoperative data. The Grubbs test was used to detect outliers in the time series data of the intraoperative data, and the detected outliers were deleted. The genetic data were processed using stratified normalization to obtain the first orthopedic surgery data.

[0064] After enhancing the image data in the first orthopedic surgery data, the gene data and phenotypic data are corrected through KL divergence distribution to obtain the initial orthopedic surgery data.

[0065] Step 102: Using a cascaded feature extraction network to extract the spatial features of bone microstructure, instrument stress time series signals, and topological relationships of biomechanical conduction pathways from the initial orthopedic surgery data; calculating the correlation weights between the image data and the biosignals through a cross-modal attention mechanism to obtain feature orthopedic surgery data;

[0066] Specifically, in this embodiment,

[0067] Define the cascade feature extraction network as a triplet Its output characteristics satisfy the following formula:

[0068]

[0069] Among them, Φ fuserepresents the cross-modal fusion function, X represents the image data, S represents the stress time series signal, and A represents the biomechanical topological adjacency matrix; Represents a 3D convolutional neural network; Represents an LSTM network; Represents GNN graph neural network, F total represents the output features;

[0070] The spatial features of bone microstructure in the initial orthopedic surgery data are extracted using a 3D convolutional neural network. The temporal features of instrument stress in the initial orthopedic surgery data are extracted using an LSTM network to obtain spatiotemporal orthopedic surgery data.

[0071] Based on the GNN graph neural network, the topological relationship of the biomechanical conduction path is established and the biomechanical conduction map is defined. node represents the bone substructure segmented from the CT image, and the edge e ij ∈ε represents the force conduction path determined based on finite element simulation, and the edge weight ω ij =exp(-||f i -f j || 2 ), f i and f j Represents the mechanical eigenvectors of nodes i and j;

[0072] Using RGCN relational graph convolution network for anisotropic graph convolution:

[0073]

[0074] in, represents the hidden state of node i in layer l+1, l represents the number of hidden layers, R represents the set of predefined biomechanical relationship types, represents the set of neighbor nodes of node i under relationship r, Indicates the size of the neighbor node set, used for normalization; Represents the weight matrix of relation r at layer l, used to transform the hidden state of neighboring nodes ReLU represents the activation function, which performs a nonlinear transformation on the weighted summation result.

[0075] Define the joint embedding mapping of the initial orthopedic surgery data between the imaging modality and the biological signal modality in the cascade feature extraction network feature extraction;

[0076] We establish a bidirectional query-key-value interaction structure, generate an inter-modal weight matrix through a gating network, add Kullback-Leibler divergence loss to promote attention sparsity, and optimize the dynamic association weights of the cascaded feature extraction network.

[0077] The correlation strength of the cascade feature extraction network in extracting initial orthopedic surgery data is adjusted by the real-time biological signal change rate to obtain characteristic orthopedic surgery data.

[0078] Step 103: construct a patient-specific bone model based on finite element analysis to perform real-time scenario simulation on the characteristic orthopedic surgery data;

[0079] Specifically, in this embodiment, a patient-specific bone model is constructed based on finite element analysis. The patient-specific bone model includes a spatially variable material model based on the nonlinear relationship between CT value and bone density:

[0080] E(x)=α·ρ(x) β +γ·softplus(G SNP );

[0081] Where E(x) represents the material strength at voxel x, ρ(x) = a·HU(x) + b represents the apparent density at voxel x, a and b represent the correlation coefficients, and HU(x) represents the CT value of voxel x; G SNP represents the correction term of gene characteristics for material strength, α=2045, β=1.55, γ=0.3 represents clinical calibration parameters, which are used to quantify the relationship between material strength and other variables, and softplus represents a nonlinear function, which is used to correct G SNP Perform transformations;

[0082] The patient-specific bone model also includes conformal tetrahedral mesh generation using a modified Marching Cubes algorithm:

[0083]

[0084] in, Represents the generated conformal tetrahedral mesh area, which is a set of three-dimensional points that meet specific conditions; N represents the number of elements in the set, V i Represents a point in three-dimensional space, constituting the basic element of the grid, Φ represents the signed distance function of the CT image, τ represents the threshold parameter, and δ represents the adaptive offset based on the direction of the trabeculae; represents the signed distance function, which is used to adjust the threshold according to the gradient direction of Φ;

[0085] Based on GPU acceleration of the preconditioned conjugate gradient method, a finite element solver for the specific bone model is obtained by constructing a block diagonal preprocessing matrix of the sparse stiffness matrix and performing parallel optimization of matrix-vector multiplication under the CUDA architecture.

[0086] Step 104: Real-time simulation of the joint contact stress distribution at different osteotomy angles, the micro-motion threshold of the bone-implant interface after internal fixation device implantation, and the impact of hemodynamic changes on the bone healing microenvironment to obtain the risk status of orthopedic surgery;

[0087] Specifically, in this embodiment, real-time simulation of the joint contact stress distribution at different osteotomy angles includes calculating the angle variation interval with a resolution of 0.1° based on characteristic orthopedic surgery data, determining the joint contact stress peak value and distribution pattern at each angle, and obtaining a risk result.

[0088] The micro-motion threshold of the bone-implant interface after internal fixation device implantation includes the strain signal of the piezoelectric film array on the object surface in the implant characteristic orthopedic surgery data, the changes in the propagation characteristics of the bone acoustic guided wave, and the interface gap data of intraoperative OCT imaging. A three-dimensional feature vector of micro-motion amplitude, frequency, and spatial distribution is established to determine whether the current micro-motion pattern belongs to the high-risk type, and obtain a Class II risk result.

[0089] The impact of hemodynamic changes on the bone healing microenvironment includes generating a four-dimensional bone healing progress map based on blood flow simulation results, quantifying local repair capacity using the MA I metabolic activity index, determining the regional MA I warning line for 5 consecutive minutes, and obtaining three types of risk results;

[0090] The three types of risk assessment results are dynamically weighted to obtain the risk status of orthopedic surgery.

[0091] A three-level surgical risk warning mechanism should be established based on the risk status of orthopedic surgery, including at least intraoperative warning, short-term complications and long-term failure.

[0092] Specifically, in this embodiment, intraoperative warnings include combining intraoperative LSC I laser speckle imaging with simulation predictions. When local blood flow drops by 30% compared to baseline and persists for more than two minutes, the hemostatic device's standby mode is activated. Based on the patient's anatomical variation data, a 3mm dynamic buffer zone is generated around blood vessels and nerve pathways. When the device touches the buffer zone, an optical flashing warning is triggered.

[0093] Short-term complications include combining lower limb venous ultrasound elastography data to trigger gradient compression stockings wearing reminders and anticoagulant dosage adjustment recommendations when peak blood flow velocity drops by 40% and D-dimer is >5 mg / L;

[0094] Long-term failure includes building a patient digital twin based on dynamic monitoring data one year after surgery to predict the amount of prosthesis wear, the extent of osteolysis, and the optimal time window for revision surgery 5-10 years later.

[0095] Its beneficial effect lies in that, by fusing preoperative imaging data with intraoperative biomechanical signals, it effectively integrates fragmented data, eliminates noise interference, and forms structured initial orthopedic surgical data, solving the problems of isolated data and inconsistent formats in traditional surgery. The cascaded feature extraction network breaks through the limitations of single-modality data; at the same time, the cross-modal attention mechanism strengthens the semantic association between different modalities by calculating the correlation weights between imaging data and biological signals, achieving in-depth analysis of bone mechanical behavior and instrument-bone interface interactions in surgical scenarios, and avoiding the problem of feature omission caused by modal splitting in traditional methods. The patient-specific bone model constructed based on finite element analysis can accurately reflect the geometric morphology, material properties and biomechanical characteristics of individual bones, solving the problem of disconnection between traditional general models and the actual patient situation. By simulating in real time the effects of different osteotomy angles and internal fixation implantation schemes on joint contact stress, interface micro-motion threshold and hemodynamics, it can provide surgeons with a preliminary assessment of surgical results and support dynamic adjustment of surgical plans during surgery. By quantitatively analyzing the impact of surgical operations on the biomechanical environment of bones, empirical surgical decisions can be transformed into data-driven precise regulation, significantly reducing complications caused by improper operations and improving surgical success rates.

[0096] In this embodiment, please refer to Figure 2 A second embodiment of an orthopedic surgery risk assessment method according to an embodiment of the present invention obtains preoperative data and intraoperative data of a patient's orthopedic surgery, performs data preprocessing on the preoperative data and the intraoperative data, and obtains initial orthopedic surgery data, including the following steps:

[0097] Step 201: Obtain preoperative and intraoperative data of a patient undergoing orthopedic surgery. The preoperative data includes at least: high-resolution CT images, MRI images, gait analysis data, serum bone metabolism markers, and susceptibility gene SNP sites; the intraoperative data includes at least: bone density impedance signals, internal fixation device stress distribution, and patient vital sign time series.

[0098] Step 202: align the space and time of the CT images and MRI images in the preoperative data with the intraoperative data using an affine transformation matrix, detect outliers in the time series data of the intraoperative data using a Grubbs test, delete the detected outliers, and process the genetic data using hierarchical normalization to obtain the first orthopedic surgery data;

[0099] Step 203: After enhancing the image data in the first orthopedic surgery data, the gene data and phenotype data are corrected by KL divergence distribution to obtain initial orthopedic surgery data.

[0100] Its beneficial effect lies in the fusion of preoperative imaging data (CT / MRI) with intraoperative biomechanical signals (instrument stress and hemodynamic parameters), followed by standardized preprocessing. This effectively integrates fragmented data, eliminates noise interference, and forms structured initial orthopedic surgical data. This process solves the problems of isolated and inconsistent data formats in traditional surgery, providing high-quality multi-dimensional input for subsequent analysis and ensuring comprehensive and reliable feature extraction.

[0101] In this embodiment, please refer to Figure 3 In a third embodiment of an orthopedic surgery risk assessment method and system according to an embodiment of the present invention, calculating the correlation weight between image data and biological signals through a cross-modal attention mechanism to obtain characteristic orthopedic surgery data includes the following steps:

[0102] Step 301: defining a joint embedding mapping of the image modality and the biological signal modality in the cascade feature extraction network for the initial orthopedic surgery data;

[0103] Step 302: Establish a bidirectional query-key-value interaction structure, generate an inter-modality weight matrix through a gating network, add Kullback-Leibler divergence loss to promote attention sparsity, and optimize the dynamic association weights of the cascaded feature extraction network;

[0104] Step 303: Adjust the correlation strength of the cascade feature extraction network in extracting the initial orthopedic surgery data through the real-time biosignal change rate to obtain feature orthopedic surgery data.

[0105] Its beneficial effect is that the cross-modal attention mechanism strengthens the semantic association between different modalities by calculating the correlation weights between imaging data and biological signals (dynamically coupling the bone morphology in X-ray images with intraoperative stress sensor data), achieving in-depth analysis of bone mechanical behavior and instrument-bone interface interactions in surgical scenarios, and avoiding the problem of feature omission caused by modal splitting in traditional methods.

[0106] The above describes an orthopedic surgery risk assessment method provided by an embodiment of the present invention. The following describes an orthopedic surgery risk assessment system according to an embodiment of the present invention. Figure 4 In one embodiment of the orthopedic surgery risk assessment system of the present invention, the system includes:

[0107] The surgical data acquisition module is used to obtain the preoperative data and intraoperative data of the patient's orthopedic surgery, perform data preprocessing on the preoperative data and intraoperative data, and obtain the initial orthopedic surgery data;

[0108] The data feature fusion module is used to extract the spatial characteristics of bone microstructure, instrument stress time series signals, and topological relationships of biomechanical conduction pathways from the initial orthopedic surgery data using a cascaded feature extraction network. The module also calculates the correlation weights between image data and biological signals through a cross-modal attention mechanism to obtain characteristic orthopedic surgery data.

[0109] A real-time scenario simulation module is used to construct a patient-specific bone model based on finite element analysis to perform real-time scenario simulation on characteristic orthopedic surgery data;

[0110] The risk status assessment module is used to simulate in real time the joint contact stress distribution at different osteotomy angles, the micro-motion threshold of the bone-implant interface after internal fixation device implantation, and the impact of hemodynamic changes on the bone healing microenvironment to determine the risk status of orthopedic surgery;

[0111] A three-level surgical risk warning mechanism should be established based on the risk status of orthopedic surgery, including at least intraoperative warning, short-term complications and long-term failure.

[0112] The basic principles, main features, and advantages of the present invention are shown and described above. Those skilled in the art should understand that the present invention is not limited to the above-described embodiments. The above-described embodiments and descriptions are merely preferred examples of the present invention and are not intended to limit the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention, and such changes and modifications fall within the scope of the invention claimed.

Claims

1. A method for risk assessment of orthopedic surgery, characterized in that: The orthopedic surgery risk assessment method comprises the following steps: Acquiring preoperative data and intraoperative data of a patient's orthopedic surgery, and performing data preprocessing on the preoperative data and the intraoperative data to obtain initial orthopedic surgery data; A cascaded feature extraction network is used to extract the spatial features of bone microstructure, the time series signals of instrument stress, and the topological relationship of the biomechanical conduction path in the initial orthopedic surgery data; and a cross-modal attention mechanism is used to calculate the correlation weight between the image data and the biological signal to obtain characteristic orthopedic surgery data. Constructing a patient-specific bone model based on finite element analysis to perform real-time scenario simulation on the characteristic orthopedic surgery data; Real-time simulation of the joint contact stress distribution at different osteotomy angles, the micro-motion threshold of the bone-implant interface after internal fixation device implantation, and the impact of hemodynamic changes on the bone healing microenvironment can be used to obtain the risk status of orthopedic surgery.

2. The orthopedic surgery risk assessment method according to claim 1, wherein: The step of obtaining preoperative data and intraoperative data of an orthopedic surgery of a patient and performing data preprocessing on the preoperative data and the intraoperative data to obtain initial orthopedic surgery data includes: Obtaining preoperative and intraoperative data of orthopedic surgery on patients, wherein the preoperative data includes at least: high-resolution CT images, MRI images, gait analysis data, serum bone metabolism markers, and susceptibility gene SNP sites; the intraoperative data includes at least: bone density impedance signals, internal fixation device stress distribution, and patient vital sign time series; Aligning the space and time of the CT images and MRI images in the preoperative data with the intraoperative data using an affine transformation matrix, detecting outliers in the time series data of the intraoperative data using a Grubbs test, deleting the detected outliers, and processing the genetic data using hierarchical normalization to obtain first orthopedic surgery data; After enhancing the image data in the first orthopedic surgery data, the gene data and phenotypic data are corrected using KL divergence distribution to obtain initial orthopedic surgery data.

3. The orthopedic surgery risk assessment method according to claim 1, wherein: The method of extracting the topological relationship between the bone microstructure spatial features, the instrument stress time series signal, and the biomechanical conduction path in the initial orthopedic surgery data using a cascade feature extraction network includes: Define the cascade feature extraction network as a triplet Its output characteristics satisfy the following formula: Among them, Φ fuse represents the cross-modal fusion function, X represents the image data, S represents the stress time series signal, and A represents the biomechanical topological adjacency matrix; Represents a 3D convolutional neural network; Represents an LSTM network; represents GNN graph neural network, F total represents the output features; Extracting bone microstructure spatial features from the initial orthopedic surgery data using the 3D convolutional neural network; extracting instrument stress temporal features from the initial orthopedic surgery data using the LSTM network; and obtaining spatiotemporal orthopedic surgery data; Based on the GNN graph neural network, the topological relationship of the biomechanical conduction path is established and the biomechanical conduction map is defined. node represents the bone substructure segmented from the CT image, and the edge e ij ∈ε represents the force conduction path determined based on finite element simulation, and the edge weight ω ij =exp(-||f i -f j || 2 ), f i and f j Represents the mechanical eigenvectors of nodes i and j; Using RGCN relational graph convolution network for anisotropic graph convolution: in, represents the hidden state of node i in layer l+1, l represents the number of hidden layers, R represents the set of predefined biomechanical relationship types, represents the set of neighbor nodes of node i under relationship r, Indicates the size of the neighbor node set, used for normalization; W r ( l ) Represents the weight matrix of relation r at layer l, used to transform the hidden state of neighboring nodes ReLU represents the activation function, which performs a nonlinear transformation on the weighted summation result.

4. The orthopedic surgery risk assessment method according to claim 3, wherein: The cross-modal attention mechanism is used to calculate the association weights between image data and biological signals to obtain characteristic orthopedic surgery data, including: Defining a joint embedding mapping of the initial orthopedic surgery data between the imaging modality and the biological signal modality in the cascade feature extraction network feature extraction; A bidirectional query-key-value interaction structure is established, an inter-modality weight matrix is generated through a gating network, Kullback-Leibler divergence loss is added to promote attention sparsity, and the dynamic association weights of the cascaded feature extraction network are optimized; The correlation strength of the cascade feature extraction network in extracting initial orthopedic surgery data is adjusted by the real-time biological signal change rate to obtain characteristic orthopedic surgery data.

5. The orthopedic surgery risk assessment method according to claim 1, wherein: The method of constructing a patient-specific bone model based on finite element analysis to perform real-time scene simulation on the characteristic orthopedic surgery data includes: A patient-specific bone model is constructed based on finite element analysis, wherein the patient-specific bone model includes a spatially variable material model based on the nonlinear relationship between CT value and bone density: E(x)=α·ρ(x) β +γ·softplus(G SNP ); Where E(x) represents the material strength at voxel x, ρ(x) = a·HU(x) + b represents the apparent density at voxel x, a and b represent the correlation coefficients, and HU(x) represents the CT value of voxel x; G SNP represents the correction term of gene characteristics for material strength, α=2045, β=1.55, γ=0.3 represents clinical calibration parameters, which are used to quantify the relationship between material strength and other variables, and softplus represents a nonlinear function, which is used to correct G SNP Perform transformations; The patient-specific bone model also includes conformal tetrahedral mesh generation using a modified Marching Cubes algorithm: in, Represents the generated conformal tetrahedral mesh area, which is a set of three-dimensional points that meet specific conditions; N represents the number of elements in the set, V i Represents a point in three-dimensional space, constituting the basic element of the grid, Φ represents the signed distance function of the CT image, τ represents the threshold parameter, and δ represents the adaptive offset based on the direction of the trabeculae; represents the signed distance function, which is used to adjust the threshold according to the gradient direction of Φ; Based on GPU acceleration of the preconditioned conjugate gradient method, a block diagonal preprocessing matrix of the sparse stiffness matrix is constructed, and parallel optimization of matrix-vector multiplication is performed under the CUDA architecture to obtain a finite element solver for the specific bone model.

6. The orthopedic surgery risk assessment method according to claim 1, wherein: The real-time simulation of the joint contact stress distribution at different osteotomy angles, the micro-motion threshold of the bone-implant interface after internal fixation device implantation, and the impact of hemodynamic changes on the bone healing microenvironment, and the risk status of orthopedic surgery are obtained, including: The real-time simulation of the joint contact stress distribution at different osteotomy angles includes calculating the angle change range with a resolution of 0.1° based on the characteristic orthopedic surgery data, determining the joint contact stress peak value and distribution pattern at each angle, and obtaining a risk result. The micro-motion threshold of the bone implant interface after the internal fixation device is implanted includes the strain signal of the piezoelectric film array on the object surface in the characteristic orthopedic surgery data, the change in the propagation characteristics of the bone acoustic guided wave, and the interface gap data of the intraoperative OCT imaging. A three-dimensional feature vector of the micro-motion amplitude, frequency, and spatial distribution is established to determine whether the current micro-motion pattern belongs to the high-risk type, and obtain a Class II risk result. The impact of hemodynamic changes on the bone healing microenvironment includes generating a four-dimensional bone healing progress map based on blood flow simulation results, quantifying local repair capacity using the MA I metabolic activity index, determining the regional MA I warning line for 5 consecutive minutes, and obtaining three types of risk results; The three types of risk assessment results are dynamically weighted to obtain the risk status of orthopedic surgery.

7. An orthopedic surgery risk assessment system, characterized in that: The orthopedic surgery risk assessment system includes the following modules: A surgical data acquisition module is used to acquire preoperative data and intraoperative data of a patient's orthopedic surgery, and perform data preprocessing on the preoperative data and intraoperative data to obtain initial orthopedic surgery data; a data feature fusion module for extracting the spatial features of bone microstructures, instrument stress time series signals, and topological relationships of biomechanical conduction pathways from the initial orthopedic surgery data using a cascaded feature extraction network; and calculating the correlation weights between image data and biological signals using a cross-modal attention mechanism to obtain characteristic orthopedic surgery data; A real-time scene simulation module, used for constructing a patient-specific bone model based on finite element analysis to perform real-time scene simulation on the characteristic orthopedic surgery data; The risk status assessment module is used to simulate in real time the joint contact stress distribution under different osteotomy angles, the micro-motion threshold of the bone-implant interface after internal fixation device implantation, and the impact of hemodynamic changes on the bone healing microenvironment to obtain the risk status of orthopedic surgery.

8. The orthopedic surgery risk assessment system according to claim 7, wherein: The surgical data acquisition module includes the following submodules: An acquisition submodule is used to acquire preoperative and intraoperative data of patients undergoing orthopedic surgery. The preoperative data includes at least: high-resolution CT images, MRI images, gait analysis data, serum bone metabolism markers, and susceptibility gene SNP sites; the intraoperative data includes at least: bone density impedance signals, internal fixation device stress distribution, and patient vital sign time series; an alignment submodule, configured to align the space and time of the CT images and MRI images in the preoperative data with the intraoperative data using an affine transformation matrix, detect outliers in the time series data in the intraoperative data using a Grubbs test, delete the detected outliers, and process the genetic data using hierarchical normalization to obtain first orthopedic surgery data; A submodule is obtained, which is used to enhance the image data in the first orthopedic surgery data, and then correct the gene data and phenotypic data through KL divergence distribution to obtain initial orthopedic surgery data.

9. The orthopedic surgery risk assessment system according to claim 7, wherein: The data feature fusion module includes the following submodules: A definition submodule is used to define a joint embedding mapping of the image modality and the biological signal modality of the initial orthopedic surgery data in the cascade feature extraction network feature extraction; Establishing submodules for establishing a bidirectional query-key-value interaction structure, generating an inter-modality weight matrix through a gating network, adding Kullback-Leibler divergence loss to promote attention sparsity, and optimizing the dynamic association weights of the cascaded feature extraction network; The extraction submodule is used to adjust the association strength of the cascade feature extraction network in extracting initial orthopedic surgery data through the real-time biological signal change rate to obtain characteristic orthopedic surgery data.

Images