A method and device for estimating cardiac mechanical parameters based on neural radiation field

Abstract

The present invention discloses a method and device for estimating cardiac mechanical parameters based on neural radiation field, which belongs to the field of data processing technology. The method includes constructing a cardiac mechanical model based on acquired cardiac motion geometric data; constructing a neural radiation field model for differential rendering of the cardiac mechanical model, inputting camera parameters and cardiac motion geometric data, and generating a rendered image that matches the actual observed data; performing loss solution and gradient return on the rendered image and the benchmark real data at each moment to obtain a preliminarily trained neural radiation field model; performing loss solution and gradient return on the rendered image and the benchmark real data at all moments to obtain a final trained neural radiation field model to estimate cardiac mechanical parameters. The present invention has significant advantages in terms of computational efficiency, experimental cost, personalization and real-time performance, estimation stability and robustness, and clinical application feasibility, and significantly improves the application value of cardiac mechanical analysis in individualized diagnosis and treatment.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a method and device for estimating cardiac mechanical parameters based on a neural radiation field. Background Art

[0002] In the diagnosis and treatment of heart diseases, accurate estimation of cardiac mechanical parameters is crucial. These parameters include elastic modulus, stress, strain, etc. of cardiac tissue, which can reflect the functional state and pathological changes of the heart. Although existing technologies have made some progress in the estimation of cardiac mechanical parameters, there are still many shortcomings.

[0003] Existing methods are difficult to achieve real-time, personalized mechanical parameter estimation of an individual heart, which limits their application in personalized diagnosis and treatment, including the following aspects:

[0004] (1) Traditional finite element analysis methods require a lot of computing resources and have high requirements for geometric models and initial conditions, making them difficult to apply in real time to clinical practice;

[0005] (2) Although experimental measurement methods can provide accurate mechanical parameters, their high cost and complex operation limit their widespread clinical application;

[0006] (3) Existing methods are difficult to achieve personalized mechanical parameter estimation of an individual heart, which limits their application in personalized diagnosis and treatment;

[0007] (4) Current optimization and estimation methods often fail to ensure robustness and stability when dealing with complex cardiac motion and mechanical behavior, resulting in unstable parameter estimation results. Therefore, it is urgent to design a cardiac mechanical parameter estimation method to address the shortcomings of existing technologies in terms of high computational complexity, experimental method limitations, personalization, slow computational speed, and insufficient robustness of the optimization process. Summary of the invention

[0008] In order to solve the problems existing in the prior art, the present invention provides the following technical solutions.

[0009] A first aspect of the present invention provides a method for estimating cardiac mechanical parameters based on a nerve radiation field, comprising:

[0010] S101, obtaining cardiac motion geometry data from existing databases and clinical data;

[0011] S102, constructing a cardiac mechanics model based on Taichi language (a high-performance parallel programming language) based on the acquired cardiac motion geometry data;

[0012] S103, constructing a neural radiation field model for differential rendering of the cardiac mechanics model, inputting camera parameters and cardiac motion geometry data into the neural radiation field model, and generating a rendering image that matches the actual observation data;

[0013] S104, performing loss solution and gradient return for the rendered image at each moment and the benchmark real data at the corresponding moment, updating the parameters of the neural radiation field model, obtaining a preliminarily trained neural radiation field model, and reconstructing and simulating the geometric structure of the heart using the preliminarily trained neural radiation field model to ensure that the preliminarily trained neural radiation field model can accurately reflect the original geometric shape of the heart;

[0014] S105, performing loss solution and gradient feedback on the rendered images at all times and their corresponding benchmark real data at all times, updating the parameters of the initially trained neural radiation field model, obtaining the final trained neural radiation field model, and using the final trained neural radiation field model to estimate cardiac mechanical parameters.

[0015] Preferably, the method further includes: S106, optimizing the gradient based on a loss function including pixel difference and structural similarity, and updating and optimizing the cardiac mechanics parameters.

[0016] Preferably, the cardiac motion geometry data includes vertex position data, tetrahedron vertex index data, fiber orientation data, myocardial fiber layer orientation data, ventricular base data, left endocardium data, right endocardium data and epicardium data.

[0017] Preferably, constructing a cardiac mechanics model based on Taichi language based on the acquired cardiac motion geometry data includes: constructing a cardiac mechanics model based on Taichi language using finite element analysis technology based on the acquired cardiac motion geometry data.

[0018] Preferably, constructing a cardiac mechanics model based on the Taichi language based on the acquired cardiac motion geometry data further includes: dividing the cardiac mechanics model using tetrahedral meshes, and performing mechanical simulation in combination with the directions of myocardial fibers and fiber layers.

[0019] A second aspect of the present invention provides a cardiac mechanical parameter estimation device based on a nerve radiation field, comprising:

[0020] A data acquisition module, used for acquiring cardiac motion geometry data from existing databases and clinical data;

[0021] A cardiac mechanics model building module, used to build a cardiac mechanics model based on Taichi language based on the acquired cardiac motion geometry data;

[0022] A neural radiation field model building module is used to build a neural radiation field model for differential rendering of a cardiac mechanics model, input camera parameters and cardiac motion geometry data into the neural radiation field model, and generate a rendering image that matches the actual observation data;

[0023] The neural radiation field model preliminary training module is used to solve the loss and return the gradient of the rendered image at each moment and the benchmark real data at the corresponding moment, update the parameters of the neural radiation field model, obtain the preliminary trained neural radiation field model, and use the preliminary trained neural radiation field model to reconstruct and simulate the geometric structure of the heart to ensure that the preliminary trained neural radiation field model can accurately reflect the original geometric shape of the heart;

[0024] The cardiac mechanics parameter estimation module is used to solve the loss and return the gradient of the rendered images at all times and their corresponding benchmark real data at all times, update the parameters of the initially trained neural radiation field model, obtain the final trained neural radiation field model, and use the final trained neural radiation field model to estimate the cardiac mechanics parameters.

[0025] Preferably, it also includes: a cardiac mechanics parameter update optimization module, which is used to optimize the gradient based on the loss function of pixel difference and structural similarity, and update and optimize the cardiac mechanics parameters.

[0026] A third aspect of the present invention provides a memory storing a plurality of instructions, wherein the instructions are used to implement a method for estimating cardiac mechanical parameters based on neural radiation fields as described in the first aspect.

[0027] The fourth aspect of the present invention provides an electronic device, comprising a processor and a memory connected to the processor, the memory storing a plurality of instructions, the instructions being loadable and executable by the processor so that the processor can execute a method for estimating cardiac mechanical parameters based on neural radiation fields as described in the first aspect.

[0028] Compared with the prior art, the method and device for estimating cardiac mechanical parameters based on neural radiation field provided by the present invention have the following beneficial effects:

[0029] 1. Significantly improved computational efficiency: Efficient encoding using neural radiation field technology with differentiable simulation and differentiable rendering reduces the high requirements for geometric models and initial conditions, greatly reducing computational complexity.

[0030] 2. Reduced reliance on expensive and complex experimental measurements: By learning the mechanical parameters of the heart from existing data, the reliance on experimental measurements with high costs and complex operations is reduced. The neural radiation field model can automatically extract and estimate cardiac mechanical parameters by learning from a large amount of existing data, thereby reducing the need for experimental equipment and operations.

[0031] 3. Personalized and real-time mechanical parameter estimation: It can update and optimize cardiac mechanical parameters in real time according to the specific data of the individual heart to meet the needs of personalized diagnosis and treatment. Differentiable rendering and optimization pipeline design enable the neural radiation field model to respond to changes in input data in real time and quickly adjust and update mechanical parameters through gradient optimization methods.

[0032] 4. Improve the robustness and stability of parameter estimation: When dealing with complex cardiac motion and mechanical behavior, the stability and robustness of the estimation results can be maintained. This application designs an end-to-end differentiable optimization process, combines physical constraints and prior knowledge, and improves the stability of the parameter estimation process and the reliability of the results through a reasonable loss function and optimization algorithm.

[0033] 5. Enhance the feasibility of clinical application: Due to the real-time and high efficiency of this application, its feasibility and application value in clinical practice are enhanced. This application uses advanced neural network technology and optimization methods to make the estimation process of cardiac mechanical parameters more efficient and accurate, and can be quickly applied and responded in clinical environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 This is a schematic diagram of the process of a cardiac mechanical parameter estimation method based on neural radiation field according to the present invention;

[0035] Figure 2 This is a functional structure diagram of a cardiac mechanical parameter estimation device based on neural radiation field described in the present invention. DETAILED DESCRIPTION

[0036] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0037] The method provided by the present invention can be implemented in the following terminal environment, and the terminal may include one or more of the following components: a processor, a memory, and a display screen. The memory stores at least one instruction, and the instruction is loaded and executed by the processor to implement the method described in the following embodiment.

[0038] The processor may include one or more processing cores. The processor uses various interfaces and lines to connect various parts of the entire terminal, and executes various functions of the terminal and processes data by running or executing instructions, programs, code sets or instruction sets stored in the memory, and calling data stored in the memory.

[0039] The memory may include random access memory (RAM) or read-only memory (ROM). The memory may be used to store instructions, programs, codes, code sets or instructions.

[0040] The display screen is used to display the user interface of each application.

[0041] In addition, those skilled in the art can understand that the structure of the above terminal does not constitute a limitation on the terminal, and the terminal may include more or fewer components, or combine certain components, or arrange the components differently. For example, the terminal also includes components such as a radio frequency circuit, an input unit, a sensor, an audio circuit, and a power supply, which will not be described in detail here.

[0042] Embodiment 1

[0043] like Figure 1 As shown, an embodiment of the present invention provides a method for estimating cardiac mechanical parameters based on a neural radiation field, comprising:

[0044] S101, obtaining cardiac motion geometry data from existing databases and clinical data.

[0045] Specifically, the cardiac motion geometry data is three-dimensional geometry data extracted from two-dimensional medical images, including vertex position data (representing the three-dimensional coordinates of each vertex in the cardiac mechanics model), tetrahedron vertex index data (tetrahedron units used to construct the cardiac mechanics model using finite element analysis technology), fiber orientation data (the main direction of myocardial fibers, used to simulate the biomechanical behavior of the heart), myocardial fiber sheet orientation data (representing the arrangement direction of the myocardial fiber layer, used to describe the arrangement of fibers within the layer), ventricular base data (the upper surface of the ventricle, as an important boundary condition of the heart model), left endocardium data (the inner wall of the left ventricle, used to simulate the deformation and stress distribution of the left ventricle), right endocardium data (the inner wall of the right ventricle, used to simulate the deformation and stress distribution of the right ventricle) and epicardium data (the outer membrane of the heart, as the outer boundary of the entire cardiac mechanics model).

[0046] S102, constructing a cardiac mechanics model based on Taichi language based on the acquired cardiac motion geometry data.

[0047] Specifically, based on the acquired cardiac motion geometry data, the finite element analysis technology was used to construct a cardiac mechanics model based on the Taichi language to simulate the deformation of the heart; the cardiac mechanics model was divided into tetrahedral grids, and mechanical simulation was performed in combination with the directions of myocardial fibers and fiber layers. During the simulation process, the grid density and time step were adjusted to ensure numerical stability and computational efficiency in the simulation process, so as to obtain a stable cardiac mechanics model.

[0048] S103, constructing a neural radiation field model for differential rendering of the cardiac mechanics model, inputting camera parameters and cardiac motion geometry data into the neural radiation field model, and generating a rendered image that matches the actual observation data, wherein the actual observation data is two-dimensional medical imaging data, including ultrasound, CT, MRI and other data.

[0049] S104, solving the loss and returning the gradient for the rendered image at each moment and the benchmark real data at the corresponding moment, updating the parameters of the neural radiation field model, obtaining a preliminarily trained neural radiation field model, and using the preliminarily trained neural radiation field model to reconstruct and simulate the geometric structure of the heart, ensuring that the preliminarily trained neural radiation field model can accurately reflect the original geometric shape of the heart.

[0050] Specifically, the benchmark real data is obtained based on the cardiac mechanics model, and the loss is solved and the gradient is returned for the rendering data at each moment and the benchmark real data at the corresponding moment, the parameters of the neural radiation field model are updated, and the preliminary trained neural radiation field model is obtained. The geometric structure of the heart is reconstructed and simulated using the preliminary trained neural radiation field model, wherein the loss is solved to calculate the loss between the rendered image at each moment and the benchmark real image at the corresponding moment, including the pixel difference and structural similarity of the image; the gradient is returned to return the loss gradient to the neural radiation field (NeRF) model through the back propagation algorithm to update the neural radiation field model; the geometric reconstruction and simulation is to reconstruct and simulate the geometric structure of the heart using the preliminary trained neural network model to ensure that the neural radiation field model can accurately reflect the original geometric shape of the heart.

[0051] S105, performing loss solution and gradient feedback on the rendered images at all times and their corresponding benchmark real data at all times, updating the parameters of the initially trained neural radiation field model, obtaining the final trained neural radiation field model, and using the final trained neural radiation field model to estimate cardiac mechanical parameters.

[0052] Specifically, loss solution and gradient feedback are performed on the rendered images at all times and their corresponding benchmark real data at all times, the parameters of the preliminary trained neural radiation field model are updated, and the final trained neural radiation field model is obtained. The final trained neural radiation field model is used to estimate the mechanical parameters of the heart; wherein, loss solution is to compare the rendered data at all times and their corresponding benchmark real data at all times, and calculate the overall loss function; gradient feedback is to use the back propagation algorithm to feedback the overall loss gradient to the preliminary trained model, and update the parameters of the preliminary trained model to obtain the final trained neural radiation field model; then the final trained neural radiation field model is used to estimate the mechanical parameters of the heart, including elastic modulus, stress, strain, etc., to ensure that the final trained neural radiation field model can accurately predict the mechanical behavior of the heart under different states.

[0053] S106, optimizing the gradient based on the loss function of pixel difference and structural similarity, and updating and optimizing the cardiac mechanics parameters.

[0054] Specifically, the gradient is optimized by a loss function based on pixel difference and structural similarity, and the cardiac mechanics parameters are updated and optimized. Considering the brightness, contrast and structural information of the image, a loss function based on pixel difference and structural similarity is adopted to enhance the ability to capture the image structure and details, so that the rendering result is closer to the actual observation data: using the Adam or SGD optimization algorithm, gradient optimization is performed by minimizing the loss function, and the parameters of the finally trained neural radiation field model are updated. The optimization process is iterated repeatedly to gradually reduce the error between the rendering results and the actual observation data. At the same time, after each gradient optimization, the cardiac mechanics parameters are updated in real time to ensure that the model can dynamically adapt to new actual observation data and improve the accuracy and stability of parameter estimation.

[0055] Embodiment 2

[0056] like Figure 2 As shown, another aspect of the present invention also includes a functional module architecture that is completely consistent with the aforementioned method flow, that is, an embodiment of the present invention further provides a cardiac mechanical parameter estimation device based on a neural radiation field, including:

[0057] A data acquisition module 201 is used to acquire cardiac motion geometry data from an existing database and clinical data;

[0058] A cardiac mechanics model building module 202, for building a cardiac mechanics model based on the Taichi language based on the acquired cardiac motion geometry data;

[0059] A neural radiation field model construction module 203 is used to construct a neural radiation field model for differential rendering of a cardiac mechanics model, input camera parameters and cardiac motion geometry data into the neural radiation field model, and generate a rendering image that matches the actual observation data;

[0060] The neural radiation field model preliminary training module 204 is used to solve the loss and return the gradient of the rendered image at each moment and the benchmark real data at the corresponding moment, update the parameters of the neural radiation field model, obtain the preliminary trained neural radiation field model, and use the preliminary trained neural radiation field model to reconstruct and simulate the geometric structure of the heart to ensure that the preliminary trained neural radiation field model can accurately reflect the original geometric shape of the heart;

[0061] The cardiac mechanics parameter estimation module 205 is used to solve the loss and return the gradient of the rendered images at all times and their corresponding benchmark real data at all times, update the parameters of the initially trained neural radiation field model, obtain the final trained neural radiation field model, and use the final trained neural radiation field model to estimate the cardiac mechanics parameters.

[0062] Preferably, the cardiac mechanical parameter estimation device based on the neural radiation field also includes: a cardiac mechanical parameter update optimization module, which is used to optimize the gradient based on the loss function of pixel difference and structural similarity, and update and optimize the cardiac mechanical parameters. The device can be implemented by the cardiac mechanical parameter estimation method based on the neural radiation field provided in the above-mentioned embodiment 1. The specific implementation method can be referred to the description in embodiment 1, which will not be repeated here.

[0063] The present invention also provides a memory storing a plurality of instructions, wherein the instructions are used to implement a method for estimating cardiac mechanical parameters based on a neural radiation field as described in the first embodiment.

[0064] The present invention also provides an electronic device, comprising a processor and a memory connected to the processor, wherein the memory stores a plurality of instructions, and the instructions can be loaded and executed by the processor so that the processor can execute a method for estimating cardiac mechanical parameters based on neural radiation fields as described in Example 1.

[0065] Although preferred embodiments of the present invention have been described, additional changes and modifications may be made to these embodiments by those skilled in the art once the basic inventive concepts are known. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention. Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A method for estimating cardiac mechanical parameters based on neural radiation fields. It is characterized in that include: S101, obtaining cardiac motion geometry data from existing databases and clinical data; S102, constructing a cardiac mechanics model based on Taichi language based on the acquired cardiac motion geometry data; S103, constructing a neural radiation field model for differential rendering of the cardiac mechanics model, inputting camera parameters and the cardiac motion geometry data into the neural radiation field model, and generating a rendering image matching the actual observation data; S104, performing loss solving and gradient return on the rendered image at each moment and the reference real data at the corresponding moment, updating the parameters of the neural radiation field model, obtaining a preliminarily trained neural radiation field model, and reconstructing and simulating the geometric structure of the heart using the preliminarily trained neural radiation field model to ensure that the preliminarily trained neural radiation field model can accurately reflect the original geometric shape of the heart; S105, performing loss solution and gradient feedback on the rendered images at all times and the corresponding benchmark real data at all times, updating the parameters of the initially trained neural radiation field model, obtaining the final trained neural radiation field model, and estimating cardiac mechanical parameters using the final trained neural radiation field model.

2. A method for estimating cardiac mechanical parameters based on neural radiation field as claimed in claim 1, It is characterized in that Also includes: S106, optimizing the gradient based on the loss function including pixel difference and structural similarity, updating and optimizing the cardiac mechanics parameters.

3. A method for estimating cardiac mechanical parameters based on neural radiation field as claimed in claim 1, It is characterized in that The cardiac motion geometry data includes vertex position data, tetrahedron vertex index data, fiber orientation data, myocardial fiber layer orientation data, ventricular base data, left endocardium data, right endocardium data and epicardium data.

4. A method for estimating cardiac mechanical parameters based on neural radiation field as claimed in claim 1, It is characterized in that Constructing a cardiac mechanics model based on Taichi language based on the acquired cardiac motion geometry data includes: constructing a cardiac mechanics model based on Taichi language using finite element analysis technology based on the acquired cardiac motion geometry data.

5. A method for estimating cardiac mechanical parameters based on neural radiation field as claimed in claim 4, It is characterized in that Constructing a cardiac mechanics model based on the Taichi language based on the acquired cardiac motion geometry data also includes: using tetrahedral meshes to divide the cardiac mechanics model, and performing mechanical simulation in combination with the directions of myocardial fibers and fiber layers.

6. A device for estimating cardiac mechanical parameters based on neural radiation field, It is characterized in that include: A data acquisition module, used for acquiring cardiac motion geometry data from existing databases and clinical data; A cardiac mechanics model building module, used for building a cardiac mechanics model based on Taichi language based on the acquired cardiac motion geometry data; A neural radiation field model construction module is used to construct a neural radiation field model for differential rendering of the cardiac mechanics model, input camera parameters and the cardiac motion geometry data into the neural radiation field model, and generate a rendering image that matches the actual observation data; A neural radiation field model preliminary training module is used to solve the loss and return the gradient of the rendered image at each moment and the benchmark real data at the corresponding moment, update the parameters of the neural radiation field model, obtain a preliminary trained neural radiation field model, and use the preliminary trained neural radiation field model to reconstruct and simulate the geometric structure of the heart to ensure that the preliminary trained neural radiation field model can accurately reflect the original geometric shape of the heart; The cardiac mechanics parameter estimation module is used to solve the loss and return the gradient of the rendered images at all times and the corresponding benchmark real data at all times, update the parameters of the initially trained neural radiation field model, obtain the final trained neural radiation field model, and use the final trained neural radiation field model to estimate the cardiac mechanics parameters.

7. A device for estimating cardiac mechanical parameters based on neural radiation field as claimed in claim 6, It is characterized in that Also includes: The cardiac mechanics parameter update optimization module is used to optimize the gradient based on the loss function of pixel difference and structural similarity, and update and optimize the cardiac mechanics parameters.

8. A memory, It is characterized in that A plurality of instructions are stored, and the instructions are used to implement a method for estimating cardiac mechanical parameters based on a neural radiation field as described in any one of claims 1-5.

9. An electronic device, It is characterized in that It includes a processor and a memory connected to the processor, wherein the memory stores a plurality of instructions, and the instructions can be loaded and executed by the processor so that the processor can execute a method for estimating cardiac mechanical parameters based on neural radiation fields as described in any one of claims 1-5.

Images