A real scene interactive anesthesia simulation training management method and system based on VR technology

Abstract

The application discloses a real scene interactive anesthesia simulation training management method and system based on VR technology, and relates to the technical field of intelligent medical simulation, which comprises the following steps: generating a virtual three-dimensional anatomical view by using a generative adversarial network based on MRI / CT images and medical records of a patient, configuring a linear mapping function by using a linear regression method, generating virtual patient physiological parameters, and embedding the virtual patient physiological parameters into the virtual three-dimensional anatomical view to obtain a virtual patient model; calculating the interaction force of a puncture needle in the virtual patient model in real time by macroscopic finite element analysis and microscopic molecular dynamics; and calculating the interaction force deviation value and the anesthetic deviation value based on the interaction force and the pharmacodynamic curve, and giving real-time prompts to a trainee through a VR head-mounted display and a tactile glove. By generating the virtual patient model, the application realizes accurate modeling of individualized anatomical structures and physiological parameters, and significantly improves the authenticity and individualization of the patient simulation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent medical simulation, in particular to a real scene interactive anesthesia simulation training management method and system based on VR technology. BACKGROUND

[0002] In recent years, with the rapid development of virtual reality (VR) technology and its in-depth application in the field of medical education, VR-based medical simulation training systems have gradually become an important complementary means of medical practice, especially in the field of anesthesiology. Traditional clinical training is limited by the scarcity of real patient resources, high operation risk and ethical issues, prompting researchers to explore new models for anesthesia skill training using high-fidelity simulation environments.

[0003] Currently, related technologies mainly include medical image three-dimensional reconstruction, immersive interaction, haptic feedback devices, and dynamic simulation of physiological parameters, for example, medical image processing technology has been able to convert MRI or CT scan results of patients into three-dimensional visual models to support surgical planning. At the same time, the development of haptic feedback devices also allows users to experience close-to-real tactile sensations in virtual environments.

[0004] However, in terms of virtual patient modeling, although existing technologies can generate three-dimensional anatomical models based on medical images, these models often lack personalized features and cannot fully reflect the differences between individuals. SUMMARY

[0005] In view of the above existing problems, the present application is proposed.

[0006] Therefore, the present application provides a real scene interactive anesthesia simulation training management method based on VR technology to solve the problem of lack of personalization in virtual patient models and difficulty in reflecting individual differences.

[0007] To solve the above technical problems, the present application provides the following technical solutions:

[0008] In a first aspect, the present application provides a real scene interactive anesthesia simulation training management method based on VR technology, which includes,

[0009] Based on patient MRI / CT images and medical records, a virtual three-dimensional anatomical view is generated using a generative adversarial network, a linear regression method is used to configure a linear mapping function, and a virtual patient physiological parameter is generated. The virtual patient physiological parameter is embedded in the virtual three-dimensional anatomical view to obtain a virtual patient model;

[0010] The interaction force and tissue node displacement of the puncture needle in the virtual patient model are calculated in real time by macroscopic finite element analysis and microscopic molecular dynamics, and the interaction force includes macroscopic resistance and microscopic resistance;

[0011] The trainer performs anesthetic puncture operation through a VR head-mounted display and a tactile glove, uses a quantum algorithm to solve a pharmacokinetic nonlinear differential equation, and generates a pharmacokinetic curve;

[0012] Based on the pharmacokinetic curve, a linear mapping method is used to calculate the changes of physiological parameters of the patient, and the virtual patient model is updated;

[0013] Based on the interaction force and the pharmacokinetic curve, the interaction force deviation value and the anesthetic agent deviation value are calculated, and the trainer is prompted in real time through the VR head-mounted display and the tactile glove.

[0014] As a preferred scheme of the real scene interactive anesthesia simulation training management method based on the VR technology, wherein: the virtual three-dimensional anatomical view is generated by using a generative adversarial network, and the specific steps are as follows,

[0015] The patient MRI / CT image is preprocessed by using a DICOM standard analysis tool to obtain a three-dimensional image tensor;

[0016] Based on 3D CycleGAN, a generative adversarial network architecture is configured, which includes a generator, an inverse generator and a discriminator;

[0017] The three-dimensional image tensor is input as a source domain, and the virtual three-dimensional anatomical view is generated by the generator;

[0018] The virtual three-dimensional anatomical view generated by the generator is converted back to the three-dimensional image tensor by the inverse generator;

[0019] The discriminator is used to compare the virtual three-dimensional anatomical view with the real three-dimensional anatomical view, and output the authenticity probability of the virtual three-dimensional anatomical view as the real three-dimensional anatomical view

[0020] The Cycle consistency loss value of the generator and the inverse generator, and the adversarial loss value of the discriminator are calculated;

[0021] Based on the Cycle consistency loss value and the adversarial loss value, a weighted sum method is used to calculate the total loss value of the generative adversarial network, and the parameters of the generative adversarial network are updated;

[0022] Based on the convergence criterion of the generative adversarial network, a total loss value threshold is set, when the total loss value of the generative adversarial network is less than the total loss value threshold, the training of the generative adversarial network is completed, and the virtual three-dimensional anatomical view is output.

[0023] As a preferred scheme of the real scene interactive anesthesia simulation training management method based on the VR technology, wherein: the virtual patient physiological parameters are generated, and the specific steps are as follows,

[0024] The medical record is cleaned and normalized to generate a three-dimensional conditional vector;

[0025] Based on the three-dimensional condition vector, a linear mapping function is configured by using a linear regression method to generate a virtual patient physiological parameter.

[0026] As a preferred scheme of the real scene interactive anesthesia simulation training management method based on the VR technology, the virtual patient physiological parameter is embedded into the virtual three-dimensional anatomical view to obtain a virtual patient model, and the specific steps are as follows,

[0027] The virtual three-dimensional anatomical view is converted into a three-dimensional grid model by using a Marching Cubes algorithm.

[0028] The virtual patient physiological parameter is converted into a key-value pair by using a JSON serialization tool to create a physiological parameter attribute table.

[0029] The physiological parameter attribute table is embedded into the three-dimensional grid model by using a vertex attribute editing function of Unreal Engine 5 to obtain the virtual patient model.

[0030] As a preferred scheme of the real scene interactive anesthesia simulation training management method based on the VR technology, the interactive force of the puncture needle in the virtual patient model is calculated, and the specific steps are as follows,

[0031] According to the Young's modulus of soft tissue in biomechanics, the elastic modulus of the tissue in the anesthesia puncture region is set.

[0032] Based on the elastic modulus of the tissue in the anesthesia puncture region, the macroscopic resistance is calculated by using Hooke's law through finite element analysis.

[0033] Based on the collagen fiber molecular structure in the protein database, a collagen fiber molecular model is constructed by using GROMACS software.

[0034] The atomic pair in the collagen fiber molecular model is traversed, the interatomic force is calculated by using a Lennard-Jones potential function, and the interatomic force of all atomic pairs is accumulated to obtain the microscopic resistance.

[0035] Based on the macroscopic resistance and the microscopic resistance, the interactive force is calculated by using a weighted summation method.

[0036] As a preferred scheme of the real scene interactive anesthesia simulation training management method based on the VR technology, the specific steps of generating the pharmacodynamic curve are as follows,

[0037] The injection dose and injection rate of the anesthetic agent of the training person in the training process are recorded in real time.

[0038] In the Qiskit framework, a variational quantum algorithm is configured, and a pharmacokinetic nonlinear differential equation is converted into a Hamiltonian.

[0039] Based on the VQE algorithm, the central chamber anesthetic concentration and the peripheral chamber anesthetic concentration of each step are calculated using a discretization method, and the concentration time series of the central chamber and the peripheral chamber are obtained;

[0040] The anesthetic concentration time series of the central chamber and the peripheral chamber are integrated using a JSON serialization tool to generate a pharmacodynamic curve.

[0041] As a preferred scheme of the real scene interactive anesthesia simulation training management method based on VR technology, the specific steps of updating the virtual patient model are as follows,

[0042] From the pharmacodynamic curve, the central chamber anesthetic concentration is extracted, and a pharmacodynamic-physiological response mapping function is constructed based on a linear mapping method to update the virtual patient physiological parameters;

[0043] The updated virtual patient physiological parameters are embedded in the three-dimensional grid model to update the virtual patient model.

[0044] In a second aspect, the present application provides a real scene interactive anesthesia simulation training management system based on VR technology, which comprises a patient virtual module, a calibration module, an interactive force module, a pharmacodynamic curve module, a model updating module and a prompt module,

[0045] The patient virtual module is used to generate a virtual three-dimensional anatomical view based on patient MRI / CT images and medical records using a generative adversarial network, configure a linear mapping function using a linear regression method, generate virtual patient physiological parameters, and embed the virtual patient physiological parameters in the virtual three-dimensional anatomical view to obtain a virtual patient model;

[0046] The interactive force module is used to calculate the interactive force and tissue node displacement of the puncture needle in the virtual patient model in real time through macro finite element analysis and micro molecular dynamics, and the interactive force includes macro resistance and micro resistance;

[0047] The pharmacodynamic curve module is used for anesthesia puncture operation by the trainee through the VR head-mounted display and the haptic glove, and a pharmacodynamic curve is generated by solving a pharmacokinetic nonlinear differential equation using a quantum algorithm;

[0048] The model updating module is used to calculate the change of the patient physiological parameters based on the pharmacodynamic curve using a linear mapping method, and update the virtual patient model;

[0049] The prompt module is used to calculate the interactive force deviation value and the anesthetic deviation value based on the interactive force and the pharmacodynamic curve, and to provide real-time prompts to the trainee through the VR head-mounted display and the haptic glove.

[0050] In a third aspect, the present application provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and wherein the computer program, when executed by the processor, implements any step of the VR technology-based real scene interactive anesthesia simulation training management method according to the first aspect of the present application.

[0051] In a fourth aspect, the present application provides a computer readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements any step of the VR technology-based real scene interactive anesthesia simulation training management method according to the first aspect of the present application.

[0052] The present application has the following beneficial effects: the present application generates a virtual patient model by using a generative adversarial network, realizes accurate modeling of individualized anatomical structure and physiological parameters, and significantly improves the authenticity and individualization of patient simulation, adopts macro finite element analysis combined with micro molecular dynamics to calculate the interaction force and tissue deformation in the puncture process, not only captures the macro mechanical behavior of soft tissue, but also considers the nonlinear response of micro structures such as collagen fibers, and provides a more realistic haptic feedback experience, uses a quantum algorithm to efficiently solve pharmacokinetic differential equations, generates a pharmacokinetic curve in real time and updates the physiological state change of the virtual patient, and ensures the dynamic simulation of drug action to be accurate and correct. BRIEF DESCRIPTION OF DRAWINGS

[0053] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0054] Fig. 1 The flowchart of the VR technology-based real scene interactive anesthesia simulation training management method.

[0055] Fig. 2 The module diagram of the VR technology-based real scene interactive anesthesia simulation training management system.

[0056] Fig. 3 The flowchart of generating a virtual three-dimensional anatomical view by using a generative adversarial network.

[0057] Fig. 4 The flowchart of embedding a virtual patient physiological parameter into a virtual three-dimensional anatomical view to obtain a virtual patient model. DETAILED DESCRIPTION

[0058] In order to make the above-mentioned objects, features and advantages of the present application more apparent and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0059] In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present application. However, it will be apparent to one skilled in the art that the present application can be practiced without the specific details set forth in this description. In other instances, well-known methods, procedures, components, and circuits have not been described in detail so as not to unnecessarily obscure aspects of the present application.

[0060] Secondly, the "one embodiment" or "an embodiment" referred to herein means a specific feature, structure, or characteristic under discussion. Thus, "one embodiment" does not mean a single embodiment nor is it to be taken individually or selectively from other embodiments.

[0061] Reference is made to Figs. 1-4 For one embodiment of the present application, the embodiment provides a VR technology-based real scene interactive anesthesia simulation training management method, comprising the following steps:

[0062] S1, based on the MRI / CT image and medical record of the patient, a virtual three-dimensional anatomical view is generated by using a generative adversarial network, a linear regression method is used to configure a linear mapping function, a virtual patient physiological parameter is generated, and the virtual patient physiological parameter is embedded in the virtual three-dimensional anatomical view to obtain a virtual patient model.

[0063] The MRI / CT image and medical record of the patient are extracted from the local server of the hospital;

[0064] The patient's MRI / CT image is preprocessed by using a DICOM standard analysis tool, such as an ITK library;

[0065] Specifically, the DICOM header information is verified, which includes the DICOM header information containing the patient ID, the scanning date, and the modal type. The patient ID is replaced with a random UUID string to remove the patient's identity information, and an MRI / CT image matrix is obtained. The pixel value of the MRI / CT image matrix is linearly mapped to the range of 0-1 by using the maximum and minimum minimization method to generate a normalized MRI / CT image matrix. A three-dimensional Gaussian filtering algorithm is applied to perform convolution operation on the normalized MRI / CT image matrix to generate a smooth filtered MRI / CT image matrix. The preprocessed patient MRI / CT image is converted into a three-dimensional tensor format by using the image tensor function of the ITK library to obtain a three-dimensional image tensor;

[0066] The medical record is cleaned and normalized to generate a three-dimensional condition vector;

[0067] Specifically, the data in the medical record is normalized by maximum minimization method, for example, the body weight is scaled to 0-1, the gender is coded as binary value, such as male is 1 and female is 0, the allergy history is coded as binary value, such as 1 for having and 0 for not having, a three-dimensional conditional vector is generated, the missing values in the medical record are processed, for example, for the missing fields in the medical record, such as the body weight or the allergy history is empty, the default value filling method is used, the default value of the body weight is set to 70 kg based on the average body weight of adults, the default value of the gender is 0, and the default value of the allergy history is 0, the data in the medical record is unit standardized, the unit of the body weight is converted to kilogram, for the records with non-kilogram unit, the weight unit conversion formula is used for conversion, the JSON format integrity of the medical record is verified to ensure the field name;

[0068] Based on 3D CycleGAN, the generative adversarial network architecture is configured;

[0069] Specifically, the generative adversarial network includes a generator, an inverse generator and a discriminator, the generator network is a U-Net architecture with 32 layers, including 16 down-sampling layers and 16 up-sampling layers, the convolution kernel size of each layer is 3x3x3, the stride is 1, the padding is 1, and the activation function is ReLU, the structure of the inverse generator is the same as that of the generator, the discriminator network is a PatchGAN architecture with 5 convolution layers, the convolution kernel size of each layer is 4x4x4, the stride is 2, the receptive field is 70x70x70, and the activation function is LeakyReLU, and the weights of the generative adversarial network are initialized by Xavier;

[0070] The three-dimensional image tensor is input as the source domain, forward propagation calculation is performed, the features of the MRI / CT scan data are extracted through the 16 down-sampling layers of the generator, for example, the resolution is reduced by 2 times each time, the virtual three-dimensional anatomical view is generated through the 16 up-sampling layers, for example, the resolution is enlarged by 2 times each time, for example, the size is 512x512x256, the pixel value is normalized to 0-1, and the virtual three-dimensional anatomical view contains skin, fat, muscle and blood vessel distribution, the virtual three-dimensional anatomical view generated by the generator is converted back to the three-dimensional image tensor through the inverse generator, the cycle consistency error of the generator and the inverse generator is calculated, that is, the sum of the absolute values of the pixel differences between the generated data, for example, the generation results of the generator and the inverse generator, for example, the virtual three-dimensional anatomical view and the three-dimensional image tensor, and the original data, for example, the inputs of the generator and the inverse generator, for example, the three-dimensional image tensor and the virtual three-dimensional anatomical view, is taken as the generation error, and the calculation formula is as follows:

[0071] ;

[0072] Wherein, represents the Cycle consistency loss value, represents the three-dimensional image tensor, represents the inverse generator, represents a virtual three-dimensional anatomical view generated by a generator, represents a virtual three-dimensional anatomical view generated by a generator, represents a virtual three-dimensional anatomical view generated by a generator, is not a corresponding relationship, is generated by a generator based on different three-dimensional image tensors (non ) for unsupervised learning, and represent expected values of the three-dimensional image tensor and the virtual three-dimensional anatomical view, respectively, and represent batch pixel average values of and , respectively;

[0073] Further, represents that the three-dimensional image tensor is first converted into a virtual three-dimensional anatomical view (non ) by a generator, and then the virtual three-dimensional anatomical view is converted back into the three-dimensional image tensor by an inverse generator, Similarly, since the traditional generative adversarial network requires paired data for supervised learning, but the virtual three-dimensional anatomical view generated in this scheme lacks paired training data, that is, the direct mapping between the three-dimensional image tensor and the corresponding virtual three-dimensional anatomical view, therefore, through the inverse generator, bidirectional mapping is performed to realize unsupervised learning, and the generation quality and diversity of the virtual three-dimensional anatomical view are improved;

[0074] In the discriminator, the virtual three-dimensional anatomical view is compared with the real three-dimensional anatomical view, and the authenticity probability of the virtual three-dimensional anatomical view as the real three-dimensional anatomical view is output, and the adversarial loss value is calculated, and the calculation formula is as follows:

[0075] ;

[0076] wherein, represents the adversarial loss value, represents is the authenticity probability of the virtual three-dimensional anatomical view (generated by the generator based on different three-dimensional image tensors (non ) by the discriminator, represents the authenticity probability of by the discriminator;

[0077] The total loss value of the generative adversarial network is calculated by weighted summation method, and the calculation formula is as follows:

[0078] ;

[0079] wherein, is the total loss value, ​is a Cycle consistency loss value weight, and the value range is [0, 1], is an adversarial loss weight, the value range is [0, 1], and ;

[0080] Based on the total loss value, the Adam optimizer is used to calculate the generative adversarial network weight increment, and the generative adversarial network weight is updated, and the calculation formula is as follows:

[0081] ;

[0082] wherein, represents the generative adversarial network weight increment, is a learning rate, is an initial generative adversarial network weight;

[0083] Based on the convergence criteria of the generative adversarial network, such as Nash equilibrium and early stopping method, a total loss value threshold is set, when the total loss value of the generative adversarial network is less than the total loss value threshold, the training of the generative adversarial network is completed, and the virtual three-dimensional anatomical view is output;

[0084] Based on the three-dimensional conditional vector, a linear regression method is used to configure a linear mapping function to generate a virtual patient physiological parameter, wherein the virtual patient physiological parameter includes heart rate, blood pressure (systolic pressure) and respiratory rate, and the calculation formula is as follows:

[0085] ;

[0086] wherein, represents the th virtual patient physiological parameter, represents an index variable of the virtual patient physiological parameter, specifically, when , represents the heart rate, when , represents the blood pressure, when , represents the respiratory rate, represents the respiratory rate, represents the respiratory rate, represents the respiratory rate, represents the weight coefficient of the th virtual patient physiological parameter, is a normalized weight value, represents the gender coding coefficient of the th virtual patient physiological parameter, is a gender code, represents the allergy history coding coefficient of the th virtual patient physiological parameter, is an allergy history code, is a constant term of the th virtual patient physiological parameter, , and The value range is [0,5]. The value range is [0,100];

[0087] Using the Marching Cubes algorithm, the virtual 3D anatomical view is converted into a 3D mesh model;

[0088] Specifically, the tensor reading function of the VTK library is used to convert the virtual three-dimensional anatomical view into a voxel grid format that can be processed by the Marching Cubes algorithm. The isosurface threshold is set, for example, fixed to 0.5 in the range of [0, 1], representing the boundary of the anatomical structure, such as the skin-fat interface. The vertex accuracy of the triangular mesh is set, and the lookup table of the Marching Cubes algorithm is initialized. The lookup table contains 256 voxel configurations, each of which defines 0-4 triangles. The Marching Cubes algorithm is used to traverse each voxel of the virtual three-dimensional anatomical view in the voxel grid format, and the 8 vertex pixel values ​​of each voxel are compared with the isosurface threshold of 0.5 to determine the configuration index of the voxel. Triangles are generated according to the lookup table, where each triangle contains 3 vertices. The vertex coordinates are calculated by linear interpolation. All triangles are merged to output a three-dimensional mesh model.

[0089] The three-dimensional mesh model is smoothed using the Laplace smoothing method;

[0090] Using Unreal Engine 5's texture generation tool, initialize the texture map, set the background color of the texture map to a neutral gray to avoid subsequent color deviation, and assign different areas to the skin, fat, muscle, and blood vessels of the 3D mesh model;

[0091] According to the pixels of each voxel in the three-dimensional grid model, the voxel labeling method is used to assign a label value to the pixels of each voxel. For example, the skin label value is [0.1, 0.3], the fat label value is [0.3, 0.5], the muscle label value is [0.5, 0.7], and the blood vessel label value is [0.7, 0.9].

[0092] Through the UV mapping method, the pixel label value is projected to the texture map pixel corresponding to the UV coordinate to generate the texture area division. The RGB color is set according to the pixel label value. The skin area is set to semi-transparent, for example, the RGB value is 200, 180, 150, and the transparency is 0.5. The fat area is set to yellow, for example, the RGB value is 240, 220, 100, and the transparency is 1.0. The blood vessel area is set to red, for example, the RGB value is 200, 50, 50, and the transparency is 1.0. Linear interpolation is used to smooth the color transition of adjacent texture areas, and the texture map after color assignment is output;

[0093] According to the material binding function of Unreal Engine 5, a material instance is created, the skin, fat, muscle and blood vessel regions of the three-dimensional grid model are assigned materials, and material parameters are set, for example, the skin material enables subsurface scattering, the scattering coefficient is 0.2, and the range is 0-1; the fat material enables diffuse reflection, the reflectivity is 0.4; the muscle material enables metal luster, the metal degree is 0.1; the blood vessel material enables self-luminous, the luminous intensity is 0.3, and the texture map after color distribution is mapped to the corresponding region of the three-dimensional grid model through UV mapping;

[0094] The virtual patient physiological parameters are converted into key-value pairs by using a JSON serialization tool to create a physiological parameter attribute table, and the physiological parameter attribute table is embedded into the three-dimensional grid model by using the vertex attribute editing function of Unreal Engine 5 to obtain a virtual patient model;

[0095] It should be noted that the above information is obtained with the consent of the user and for legal purposes.

[0096] S2, the interaction force and tissue node displacement of the puncture needle in the virtual patient model are calculated in real time by macroscopic finite element analysis and microscopic molecular dynamics, and the interaction force includes macroscopic resistance and microscopic resistance.

[0097] The anesthetic puncture region in the virtual patient model is divided, and the anesthetic puncture position is determined;

[0098] According to the Young's modulus of soft tissue in biomechanics, the elastic modulus of the skin, fat and muscle of the anesthetic puncture region is set, for example, the skin elastic modulus is [1, 5] MPa, the fat elastic modulus is [0.05, 0.2] MPa, and the muscle elastic modulus is [0.3, 1] MPa, based on Hooke's law, the macroscopic resistance and tissue node displacement are calculated by finite element analysis;

[0099] Based on a protein database, such as Protein Data Bank, PDB, human skin collagen is screened, and entries containing complete triple helix structures are selected to obtain collagen fiber molecular structures;

[0100] The collagen fiber molecular structure is preprocessed by using visual molecular dynamics software such as VMD (Visual Molecular Dynamics), and the collagen fiber molecular structure is converted into a topology file recognizable by GROMACS software, such as a.top format, wherein the topology file includes atom types, bond connections and potential parameters;

[0101] Specifically, the pre-processing collagen fiber molecular structure includes repairing missing atoms in the initial collagen fiber molecular structure data, such as by the Auto PSF plug-in of VMD, supplementing hydrogen atoms, the target number of hydrogen atoms is about 7000, ensuring the stability of the triple helix structure by adjusting the molecular conformation, the bond length error is less than 0.01 nanometers, setting the molecular center coordinates, such as moving to the origin x=0, y=0, z=0, the accuracy is 0.001 nanometers, and the like.

[0102] The collagen fiber molecular structure in the topology file is placed in the simulation water tank by using the genbox tool of the GROMACS software, and sodium ions and chlorine ions are added to neutralize the charge, and a solvated topology file is generated.

[0103] The steepest descent method is used to minimize the collagen fiber molecular energy, and the lowest energy collagen fiber molecular configuration is generated, and the gmx check tool of the GROMACS software is used to check the atomic coordinate integrity, chemical bond stability and molecular configuration rationality of the collagen fiber molecular structure, and a collagen fiber molecular model is generated.

[0104] The tissue node displacement is used as a displacement constraint, and the displacement constraint is applied to the boundary atoms of the collagen fiber molecular model.

[0105] The atomic pairs in the collagen fiber molecular model are traversed, the Lennard-Jones potential function is used to calculate the interatomic force, and the interatomic force of all atomic pairs is accumulated to obtain the micro resistance.

[0106] Based on the macro resistance and the micro resistance, the weighted sum method is used to calculate the interaction force.

[0107] The vertex buffer editing function of Unreal Engine 5 is used to traverse the triangular mesh vertices of the anesthesia puncture area, and the triangular mesh vertex coordinates of the anesthesia puncture area are updated according to the tissue node displacement of the anesthesia puncture area in the virtual patient model.

[0108] The texture map color of the anesthesia puncture area is redefined, for example, light red, RGB value 220, 150, 150, transparency 0.8, range [0, 1], the closer to the anesthesia puncture position, the lower the transparency.

[0109] Real-time ray tracing is used to dynamically render the virtual patient model according to the change of the tissue node displacement of the anesthesia puncture area, and the tissue deformation view is obtained.

[0110] By generating the tissue deformation view, the tissue changes of the puncture operation are reflected in real time, the visual effect of the deformation area is highlighted, the realism of the skin depression is enhanced, the trainer can identify the puncture influence, the interactive intuitiveness is improved, the trainer is provided with ultra-high precision tactile and visual feedback, and the real tissue interaction is simulated, such as skin elasticity, muscle viscosity and fiber rupture.

[0111] S3, the training person performs anesthetic puncture operation through the VR head-mounted display and the tactile glove, solves a pharmacokinetic nonlinear differential equation by using a quantum algorithm, and generates a pharmacodynamic curve.

[0112] The virtual patient model is imported into the VR device, the training person confirms an anesthetic puncture point through the VR head-mounted display, performs an anesthetic puncture operation by wearing the tactile glove, and records an anesthetic injection dose and an injection rate in real time;

[0113] The pharmacokinetic nonlinear differential equation is calculated, and the calculation formula is as follows:

[0114] ;

[0115] ;

[0116] wherein, represents a central compartment anesthetic concentration, and represents a blood anesthetic concentration, is a time step, is a metabolic rate, the value range is [0.05, 0.2], and represents an anesthetic clearance rate from the central compartment, is a distribution rate, the value range is [0.02, 0.1], and represents an anesthetic transfer rate from the central compartment to the peripheral compartment, is a return rate, the value range is [0.01, 0.05], and represents a drug return rate from the peripheral compartment to the central compartment, is a peripheral compartment anesthetic concentration, and represents a tissue anesthetic concentration, is an anesthetic injection dose, is a bolus rate factor, the value range is [0.5, 2], and represents an anesthetic injection rate, is a central compartment volume;

[0117] In the Qiskit framework, a variational quantum algorithm such as a VQE algorithm is configured, a quantum circuit is initialized, a time step of 0.01 minutes is set, the pharmacokinetic nonlinear differential equation is converted into a Hamiltonian, the central compartment anesthetic concentration and the peripheral compartment anesthetic concentration at each step are calculated based on the VQE algorithm by using a discretization method such as the forward Euler method, and the concentration time series of the central compartment and the peripheral compartment are obtained;

[0118] The residual value of the VQE solution and the Runge-Kutta solution is calculated by solving the pharmacokinetic nonlinear differential equation by using the Runge-Kutta method, the accuracy of the VQE solution is verified, and the calculation formula is as follows:

[0119] ;

[0120] wherein, is a time step The residual value when and The time steps The concentration of anesthetic agent in the central and peripheral ventricles of the VQE solution is and The time steps The concentration of anesthetic agent in the central and peripheral chambers of the Runge-Kutta solution at 37°C;

[0121] According to the clinical requirement of concentration accuracy based on the pharmacokinetics of anesthetic agents, the error is required to be <1% of the peak concentration, approximately 0.01 mg / L, and the residual threshold is set;

[0122] If the residual value is greater than the residual threshold, the VQE optimizer learning rate is adjusted, for example, by increasing it by 10% each time, from [0.01, 0.011], and the central chamber anesthetic agent concentration and the peripheral chamber anesthetic agent concentration are recalculated until the residual value is no greater than the residual threshold;

[0123] The JSON serialization tool was used to integrate the time series of anesthetic agent concentrations in the central and peripheral chambers to generate a pharmacodynamic curve with time as the horizontal axis and anesthetic agent concentration as the vertical axis.

[0124] S4. Based on the drug efficacy curve, the linear mapping method is used to calculate the changes in the virtual patient's physiological parameters and update the virtual patient model.

[0125] The concentration of the central anesthetic agent is extracted from the pharmacodynamic curve. Based on the linear mapping method, a pharmacodynamic-physiological response mapping function is constructed to update the physiological parameters of the virtual patient and further update the virtual patient model. The calculation formula is as follows:

[0126] ;

[0127] in, After the update Virtual patient physiological parameters, For the The effect coefficient of the physiological parameters of a virtual patient, with a value range of [10,30];

[0128] The updated physiological parameters of the virtual patient are embedded into the three-dimensional mesh model to update the virtual patient model.

[0129] S5. Based on the interaction force and drug efficacy curve, the interaction force deviation value and the anesthetic drug deviation value are calculated, and real-time prompts are given to the trainee through the VR headset and tactile gloves.

[0130] The segmented deviation analysis method is used to divide the anesthesia puncture process into three segments of skin layer penetration, fat layer penetration and muscle layer penetration, and based on the elastic modulus and thickness of the tissue corresponding to each segment of the anesthesia puncture process, the target interaction force value of each segment of the anesthesia puncture process is calculated according to the Hooke's law;

[0131] The absolute value of the difference between the actual interaction force value and the target interaction force value in each segment of the anesthesia puncture process is calculated as the interaction force deviation value of each segment of the anesthesia puncture process.

[0132] The real-time central chamber anesthetic concentration is extracted from the pharmacodynamic curve, and the real-time target central chamber anesthetic concentration is calculated according to the physiological parameters in the virtual patient model and the patient medical record, and the calculation formula is as follows:

[0133]

[0134]

[0135] wherein, is the time step target central chamber anesthetic concentration, is the clinically recommended dose, is the unit weight dose, is the patient's body weight;

[0136] Based on the real-time central chamber anesthetic concentration and the real-time target central chamber anesthetic concentration, the anesthesia deviation value is calculated by the root mean square error method.

[0137] Using Unreal Engine 5, the puncture path is generated according to the coordinate change trajectory of the needle tip of the puncture needle, wherein the width of the puncture path is the same as the cross-sectional size of the puncture needle, and the puncture path is mapped and rendered in the same way as the rendering and mapping of the tissues in the virtual patient model, and is highlighted, wherein highlighting the puncture path does not mean increasing the display brightness of the puncture path, but selecting a color with high contrast with the rendering color of the tissues around the puncture path for rendering, such as red, RGB value 255, 0, 0, and the transparency is set to be lower than the transparency of the surrounding tissues, and the path is highlighted, such as setting to 0.8, and triggering the vibration of the HaptX haptic glove, the larger the interaction force deviation value, the higher the vibration frequency of the HaptX haptic glove, and for the anesthesia deviation value, the pharmacodynamic curve and physiological abnormality warning are displayed in real time through the VR head-mounted display, such as giving a text prompt "drug effect exceeds standard".

[0138] ​​The embodiment also provides a VR technology-based real scene interactive anesthesia simulation training management system, which comprises a patient virtual module, an interactive force module, a drug effect curve module, a model updating module and a prompt module, the patient virtual module is used for generating a virtual three-dimensional anatomical view by using a generative adversarial network based on patient MRI / CT images and medical records, adopting a linear regression method, configuring a linear mapping function, generating virtual patient physiological parameters, and embedding the virtual patient physiological parameters into the virtual three-dimensional anatomical view to obtain a virtual patient model; the interactive force module is used for calculating the interactive force and tissue node displacement of a puncture needle in the virtual patient model in real time through macro finite element analysis and micro molecular dynamics, wherein the interactive force comprises macro resistance and micro resistance; the drug effect curve module is used for training a person to perform anesthesia puncture operation through a VR head-mounted display and a haptic glove, solving a pharmacokinetic nonlinear differential equation by using a quantum algorithm, and generating a drug effect curve; the model updating module is used for calculating patient physiological parameter changes based on the drug effect curve by using a linear mapping method, and updating the virtual patient model; and the prompt module is used for calculating interactive force deviation values and anesthesia agent deviation values based on the interactive force and the drug effect curve, and performing real-time prompting on the training person through the VR head-mounted display and the haptic glove.

[0139] The embodiment also provides a computer device suitable for the VR technology-based real scene interactive anesthesia simulation training management method, which comprises a memory and a processor; the memory is used for storing computer executable instructions, and the processor is used for executing the computer executable instructions to realize the VR technology-based real scene interactive anesthesia simulation training management method proposed in the above embodiment.

[0140] The computer device can be a terminal, and the computer device comprises a processor, a memory, a communication interface, a display screen and an input device which are connected through a system bus. The processor of the computer device is used for providing computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used for wired or wireless communication with external terminals. The wireless communication can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or can be a key, a trackball or a touchpad arranged on the shell of the computer device, or can be an external keyboard, a touchpad or a mouse, etc.

[0141] The embodiment also provides a storage medium on which a computer program is stored, the program being executed by a processor to implement the method for implementing real scene interactive anesthesia simulation training management based on VR technology as proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0142] To sum up, the application generates a virtual patient model by using a generative adversarial network, realizes accurate modeling of individualized anatomical structure and physiological parameters, significantly improves the authenticity and individualization of patient simulation, adopts macro finite element analysis combined with micro molecular dynamics to calculate the interaction force and tissue deformation in the puncture process, not only captures the macro mechanical behavior of soft tissue, but also considers the nonlinear response of micro structures such as collagen fibers, and provides a more realistic haptic feedback experience, solves pharmacokinetic differential equations by using a quantum algorithm, generates a pharmacokinetic curve in real time and updates the physiological state change of the virtual patient, and ensures the dynamic simulation of drug action to be accurate and correct.

[0143] It should be noted that the above embodiments are only used to illustrate the technical solutions of the application but not limit the application, although the application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the application can be modified or replaced equivalently without departing from the spirit and scope of the technical solutions of the application, and all should be covered in the scope of the claims of the application.

Claims

1. A real-life interactive anesthesia simulation training management method based on VR technology, characterized by: The method comprises the steps of: Based on the patient's MRI / CT image and medical record, a virtual three-dimensional anatomical view is generated by using a generative adversarial network, a linear regression method is used to configure a linear mapping function, a virtual patient physiological parameter is generated, and the virtual patient physiological parameter is embedded in the virtual three-dimensional anatomical view to obtain a virtual patient model; The interaction force and tissue node displacement of the puncture needle in the virtual patient model are calculated in real time by macroscopic finite element analysis and microscopic molecular dynamics, and the interaction force includes macroscopic resistance and microscopic resistance; The specific steps of calculating the interaction force of the puncture needle in the virtual patient model are as follows, According to the Young's modulus of soft tissue in biomechanics, the elastic modulus of the tissue in the anesthesia puncture area is set; Based on the elastic modulus of the tissue in the anesthesia puncture area, the macroscopic resistance is calculated by finite element analysis using Hooke's law; Based on the collagen fiber molecular structure in the protein database, a collagen fiber molecular model is constructed by using GROMACS software; The atomic pairs in the collagen fiber molecular model are traversed, the interatomic force is calculated by using the Lennard-Jones potential function, and the interatomic force of all atomic pairs is accumulated to obtain the microscopic resistance; Based on the macroscopic resistance and the microscopic resistance, the interaction force is calculated by using the weighted summation method; The trainee performs anesthesia puncture operation through the VR head-mounted display and the haptic glove, and uses a quantum algorithm to solve a pharmacokinetic nonlinear differential equation to generate a pharmacodynamic curve; Based on the pharmacodynamic curve, the change of the patient's physiological parameter is calculated by using the linear mapping method, and the virtual patient model is updated; Based on the interaction force and the pharmacodynamic curve, the interaction force deviation value and the anesthetic agent deviation value are calculated, and the trainee is prompted in real time through the VR head-mounted display and the haptic glove.

2. The VR technology-based live interaction anesthesiology simulation training management method of claim 1, wherein: The specific steps of generating the virtual three-dimensional anatomical view by using the generative adversarial network are as follows, The patient's MRI / CT image is preprocessed by using a DICOM standard analysis tool to obtain a three-dimensional image tensor; Based on the 3D CycleGAN, a generative adversarial network architecture is configured, which includes a generator, an inverse generator and a discriminator; The three-dimensional image tensor is input as a source domain, and the virtual three-dimensional anatomical view is generated by the generator; The virtual three-dimensional anatomical view generated by the generator is converted back to the three-dimensional image tensor by the inverse generator; The discriminator is used to compare the virtual three-dimensional anatomical view with the real three-dimensional anatomical view, and output the authenticity probability of the virtual three-dimensional anatomical view as the real three-dimensional anatomical view; The Cycle consistency loss value of the generator and the inverse generator, and the adversarial loss value of the discriminator are calculated; Based on the Cycle consistency loss value and the adversarial loss value, the total loss value of the generative adversarial network is calculated by using the weighted summation method, and the parameters of the generative adversarial network are updated; Based on the convergence criterion of the generative adversarial network, a total loss value threshold is set, and when the total loss value of the generative adversarial network is less than the total loss value threshold, the training of the generative adversarial network is completed, and the virtual three-dimensional anatomical view is output. 3.The VR technology-based live interaction anesthesiology simulation training management method of claim 2, wherein: The specific steps of generating the virtual patient physiological parameter are as follows, The medical record is cleaned and normalized to generate a three-dimensional conditional vector; Based on the three-dimensional conditional vector, a linear regression method is used to configure a linear mapping function to generate a virtual patient physiological parameter.

4. The VR technology-based live interaction anesthesiology simulation training management method of claim 3, wherein: The virtual patient physiological parameter is embedded into the virtual three-dimensional anatomical view to obtain a virtual patient model, and the specific steps are as follows, The virtual three-dimensional anatomical view is converted into a three-dimensional grid model by using a Marching Cubes algorithm; The virtual patient physiological parameter is converted into a key-value pair by using a JSON serialization tool to create a physiological parameter attribute table; The physiological parameter attribute table is embedded into the three-dimensional grid model by using a vertex attribute editing function of Unreal Engine 5 to obtain the virtual patient model. 5.The VR technology-based live interaction anesthesiology simulation training management method of claim 1, wherein: The specific steps of generating the pharmacokinetic curve are as follows, The training person records the anesthetic injection dose and injection rate in real time during the training process; In the Qiskit framework, the pharmacokinetic nonlinear differential equation is converted into a Hamiltonian by configuring a variational quantum algorithm; Based on the VQE algorithm, the central compartment anesthetic concentration and the peripheral compartment anesthetic concentration at each step are calculated using a discretization method to obtain the concentration time series of the central compartment and the peripheral compartment; The anesthetic concentration time series of the central compartment and the peripheral compartment are integrated by using a JSON serialization tool to generate the pharmacokinetic curve. 6.The VR technology-based live interaction anesthesiology simulation training management method of claim 1, wherein: The specific steps of updating the virtual patient model are as follows, The central compartment anesthetic concentration is extracted from the pharmacokinetic curve, and a pharmacodynamic-physiological response mapping function is constructed based on a linear mapping method to update the virtual patient physiological parameter; The updated virtual patient physiological parameter is embedded into the three-dimensional grid model to update the virtual patient model.

7. A VR technology-based real scene interactive anesthesia simulation training management system based on the VR technology-based real scene interactive anesthesia simulation training management method of any one of claims 1-6. It includes a patient virtual module, an interactive force module, a pharmacokinetic curve module, a model updating module, and a prompt module, The patient virtual module is used to generate a virtual three-dimensional anatomical view based on patient MRI / CT images and medical records by using a generative adversarial network, configure a linear mapping function by using a linear regression method, generate a virtual patient physiological parameter, and embed the virtual patient physiological parameter into the virtual three-dimensional anatomical view to obtain a virtual patient model; The interactive force module is used to calculate the interactive force and tissue node displacement of the puncture needle in the virtual patient model in real time by using macroscopic finite element analysis and microscopic molecular dynamics, and the interactive force includes macroscopic resistance and microscopic resistance; The pharmacokinetic curve module is used for the training person to perform anesthetic puncture operation through a VR headset and a haptic glove, and a pharmacokinetic nonlinear differential equation is solved by using a quantum algorithm to generate a pharmacokinetic curve; The model updating module is used to calculate the change of the patient physiological parameter based on the pharmacokinetic curve by using a linear mapping method to update the virtual patient model; The prompt module is used to calculate the interactive force deviation value and the anesthetic deviation value based on the interactive force and the pharmacokinetic curve, and to provide real-time prompts to the training person through the VR headset and the haptic glove.

8. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that: The processor executes the computer program to realize the steps of the VR technology-based real scene interactive anesthetic simulation training management method of any one of claims 1-6.

9. A computer readable storage medium having stored thereon a computer program, characterized in that: The computer program is executed by the processor to realize the steps of the VR technology-based real scene interactive anesthetic simulation training management method of any one of claims 1-6.

Images