Example Shape Matching and Feature-Driven Heartbeat Simulation Method and Terminal

Abstract

The present invention discloses a method and a terminal for simulating heartbeats based on example shape matching and feature driving, which reconstruct a three-dimensional heart model; input an electrocardiogram signal, extract features from the electrocardiogram signal by using a slope threshold method to obtain electrocardiogram feature data, and use the electrocardiogram feature data to drive the heartbeat of the three-dimensional heart model; use a method based on example shape matching to simulate the deformation that occurs during the heartbeat of the heart; use a feature-driven method to make the three-dimensional heart model continuously repeat the process of example shape matching; and construct a heartbeat simulation system. The present invention uses the methods of example-based shape matching and feature driving to simulate the deformation during the heart movement process, while taking into account the realistic simulation of the heartbeat, increases the control of the movement state within the cardiac cycle, realizes personalized simulation of the heartbeat, and meets the real-time performance and realism of the heartbeat simulation.

Description

Technical Field

[0001] The present invention relates to the technical field of human organ simulation in virtual surgery, and in particular to a heart beat simulation method and terminal based on example shape matching and feature drive. Background Art

[0002] The beating of the human heart is a very complex physiological process, which includes the contraction changes of myocardial cells, the generation and propagation of bioelectric signals and other physiological system reactions. However, in terms of the surface movement of the heartbeat, it can be understood as a deformation movement of periodic contraction and relaxation. The simulation of the human heartbeat must be able to truly and reasonably reflect the law of the heartbeat and provide users with realistic visual effects. Specifically, in terms of real-time performance, it can truly reflect the periodic law of the heartbeat within a visual feedback cycle time period; in terms of realism, reasonable geometric deformation must occur during the heartbeat process, in line with the basic deformation law of the heartbeat.

[0003] Dawson et al. used key frame technology to simulate heart beating. This technology first stores the deformation at each key moment of the heart beating process in advance, and then plays the stored deformation in a loop during simulation. Therefore, this method can only simulate normal heart beating, and the deformation effect is limited. Seemann et al. used a method similar to cellular automata to simulate the deformation behavior of the heart. This method takes into account the physiological characteristics of heart tissue, the transmission of cardiac bioelectric signals, the direction of myocardial fibers and other factors. Starting from the physiological factors of heart beating, it can simulate the physiological movement laws of the heart during movement, and is generally used for computer-aided diagnosis and auxiliary treatment. However, such simulation methods have a very large amount of calculation and cannot meet real-time requirements. It is difficult to use them for real-time simulation in virtual surgery. Wang Yanzhen et al. proposed a real-time heartbeat simulation method based on a composite spring oscillator model based on the basic laws of human heartbeat. Although this method simplifies the calculation, it does not take into account the behavior of the microscopic physiological layer of the heart. For the simulation of heartbeat, it is far from enough to obtain the visual effect of heartbeat. What is more important is to be able to observe the physiological health of the heart during heartbeat.

[0004] Therefore, the simulation of heart beats still needs further optimization and innovation in terms of real-time and realism. Simply simulating heart beats lacks richness in system functions and practicality. Therefore, it is of great research significance to take into account the simulation of realistic heart beats and the physiological health information of heart movement and use the characteristic signal of electrocardiogram to drive the heart beat. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide a heart beating simulation method and terminal based on example shape matching and feature drive to meet the real-time and realism of heart beating simulation.

[0006] To solve the above technical problems, the technical solution adopted by the present invention is as follows:

[0007] A method for simulating heart beats based on example shape matching and feature driving, comprising the steps of:

[0008] S1. Reconstruct a three-dimensional heart model;

[0009] S2. Input an electrocardiogram signal, and use a slope threshold method to extract features of the electrocardiogram signal to obtain electrocardiogram feature data, and use the electrocardiogram feature data to drive the heart beats of the three-dimensional heart model;

[0010] S3. Use a method based on example shape matching to simulate the deformation that occurs during heart beats;

[0011] S4. Use a feature-driven method to make the three-dimensional heart model continuously repeat the process of example shape matching;

[0012] S5. Construct a heart beat simulation system.

[0013] Further, the step S1 specifically includes:

[0014] S11. Extract a heart region data set from a heart picture set in the VHP (Visible Human Project) data set and preprocess it;

[0015] S12. Segment the heart contour in the preprocessed heart region data set by using a region growing method with a defined threshold;

[0016] S13. Use a method of drawing by sub-region to realize the reconstruction of a three-dimensional patch model;

[0017] S14. Use a graphics interface to realize the rendering of the three-dimensional heart model.

[0018] Further, the step S3 specifically includes:

[0019] S31. Use the x coordinate to represent the position of a particle. After the particle responds to environmental collisions and external forces, a set of example deformations are obtained, which are respectively represented by x 0 ,…,x k ;

[0020] S32. Assume that the discrete representation of an elastic solid in a tetrahedral mesh is n nodes and c elements. Then let X∈R 3n represent the position vector of the undeformed object, and let x∈R 3n represent the configuration of the deformed object;

[0021] In the system's stable shape matching algorithm, the transformation matrix A can be obtained r is the minimum value of the squared energy:

[0022] ∑ i m i ||A r q i -p i || 2 ;

[0023] where q i = x 0 i - c 0 r represents the initial position of the particle relative to the center of gravity, and p i = x i - c r represents the current position of the particle relative to the center of gravity;

[0024] Regarding the transformation matrix A r as an approximation of the deformation gradient of the local region r, its tensor component S r is the right stretch tensor, that is: And use the tensor component S r ∈ R 6 to quantify the given deformation configuration x;

[0025] Convert the k example deformations obtained in step S31 into deformation space description variables S i = S(x i ), where S represents the graphical mapping, and x i represents the current example deformation, thus forming an example manifold;

[0026] S33. Continuously approximate by minimizing the following quadratic energy expression:

[0027]

[0028] where the vector norm |·|ζ is used to match the sum of the basic tensor two-norms, and the objective function W(x w , w) is used as the interpolation energy, and its minimization defines the projection ∏: R 6m → ζ to the mapping of, where R represents the pure rotation matrix, m represents the space dimension, and the mapping of its interpolation segment S(w), w ∈ [0, 1] is an example segment S(x w ) of the achievable manifold ζ under the projection ∏;

[0029] S34. Take x w as the stationary state. If S 0 = S(x 0) represents the initialization deformation descriptor of the elastic object, S k = S(x k ) represents the k-th example deformation descriptor, then the convex combination closest to the current pose descriptor S = S(x) is obtained That is, it is necessary to calculate the weight parameters w1,…,w n , and the minimization of the quadratic energy expression can be rewritten as:

[0030]

[0031] The solution w T = (w1,…,w n ) 2 can be obtained as:

[0032] w = (L T L) -1 L T (S - S0);

[0033] where L = (S1 - S0…S n - S0) ∈ R 6mn is a constant during the simulation, and thus the value of w0 is

[0034] S35. Obtain the next preferred deformation state;

[0035] S36. Repeat steps S33 to S35 to obtain the simulation of the entire example-based deformation process. When all the example shapes are projected onto the example manifold, the object will undergo a series of continuous deformations, constituting a motion animation.

[0036] Furthermore, the specific steps of step S4 include:

[0037] S41. Describe the dynamic deformation behavior of the heart model;

[0038] The deformation behavior of the heart is represented by W(X, x) with elastic characteristics for control, where X represents the position of the vertex that has not undergone deformation, and x represents the position of the vertex after deformation, combined with the internal deformation driving force and the external force f ext , then the dynamic deformation behavior of the three-dimensional heart model can be described by the following motion equation:

[0039]

[0040] where M is the mass matrix, is the nodal acceleration;

[0041] Let g(x, t) represent the vector-valued objective function, and the state optimization formula for the dynamic deformation behavior of the three-dimensional heart model is:

[0042]

[0043] S42. Describe the static deformation behavior of the heart model;

[0044] Use a simplified linear subspace to configure the static deformation behavior of the three-dimensional heart model. The configuration formula of the linear subspace is as follows:

[0045] X(p) = X0 + Lp;

[0046] Where X0 is the initial state, L is the linear mapping between the static form coordinates X and the control variable p. The state optimization formula for the static deformation behavior of the three-dimensional heart model is:

[0047]

[0048] S42. Describe a controller for heart movement;

[0049] Convert the target of the deformation movement into a target sequence and use g(x,t) to convert it into an algebraic function of the state variables. Then the vector-valued objective function is the result of connecting linear constraints of any number of vertex positions, and its expression is as follows:

[0050] g(x,t) = [w1g(x,t), w2g(x,t), …, w n g(x,t)] T ;

[0051] g i (x,t) = ∑ j w xj x j -s i (t);

[0052] Where s(t) represents a target function that changes with time, and the weight w i is used to weigh the importance of the target g i relative to other targets.

[0053] To solve the above technical problems, another technical solution adopted by the present invention is:

[0054] A heart-beating simulation terminal based on example shape matching and feature driving, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the heart-beating simulation method based on example shape matching and feature driving as described above.

[0055] The beneficial effects of the present invention are as follows: The method and terminal for simulating heart beats based on example shape matching and feature driving utilize the method of example-based shape matching and feature driving to simulate the deformation during the heart movement process. While taking into account the realistic simulation of heart beats, it increases the control over the movement state within the cardiac cycle, realizes personalized simulation of heart beats, and meets the real-time and realistic requirements of heart beat simulation. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 It is a schematic flowchart of the method for simulating heart beats based on example shape matching and feature driving according to an embodiment of the present invention;

[0057] Figure 2 It is a schematic diagram of an achievable manifold;

[0058] Figure 3 It is an example interpolation diagram with different weights.

[0059] Figure 4 It is a schematic structural diagram of the terminal for simulating heart beats based on example shape matching and feature driving according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0060] To describe in detail the technical content, achieved objectives and effects of the present invention, the following is described in conjunction with the embodiments and accompanied by the drawings.

[0061] Embodiment 1

[0062] Please refer to Figures 1 to 3 As shown, the method for simulating heart beats based on example shape matching and feature driving includes the steps:

[0063] S1. Reconstruct a three-dimensional heart model;

[0064] Among them, the step S1 specifically includes:

[0065] S11. Extract the heart region dataset from the heart picture set in the VHP dataset and preprocess it;

[0066] S12. Segment the heart contour in the preprocessed heart region dataset by using the region growing method with a defined threshold;

[0067] S13. Realize the reconstruction of the three-dimensional patch model by using the method of sub-region drawing;

[0068] S14. Render the three-dimensional heart model by using the graphics interface.

[0069] Thus, a heart model with a relatively high degree of authenticity is provided for the heart beat simulation process.

[0070] S2. Input an electrocardiogram (ECG) signal, extract features from the ECG signal using a slope threshold method to obtain ECG feature data, and use the ECG feature data to drive the heartbeat of the three-dimensional heart model;

[0071] In this embodiment, since it is difficult to record the change in the main frequency of ventricular fibrillation within a short time period using the fast Fourier transform, and wavelet transform is prone to introducing errors in detecting wave groups caused by arrhythmia. To obtain the law of heartbeat, a slope threshold method is used to extract features from the ECG signal, which can directly classify according to the result data of the extracted features and can extract the features of the ECG signal according to requirements. At the same time, during the simulation of heartbeat, the detailed law of movement within the cardiac cycle needs to be considered. Therefore, the time interval and slope change of each waveform feature point are what need to be studied in the simulation of heartbeat. Therefore, based on the consideration of the details of the heartbeat simulation, a method of the slope threshold type is selected to extract the ECG feature signal.

[0072] S3. Use a method based on example shape matching to simulate the deformation that occurs during the heartbeat of the heart;

[0073] Among them, step S3 specifically includes:

[0074] S31. Represent the position of a particle with the x coordinate. When the particle responds to environmental collisions and external forces, a set of example deformations are obtained, which are represented by x 0 ,…,x k respectively;

[0075] S32. Assume that the discrete representation of an elastic solid in a tetrahedral mesh is known as n nodes and c elements. Then let X∈R 3n represent the position vector of the undeformed object, and let x∈R 3n represent the configuration of the deformed object; the given deformed configuration x can be quantified using the right Cauchy - Green component C(X,x) = F F of the rotation invariant or the Green tensor E(X,x) = (C(X,x) - I) / 2. T F or the Green tensor E(X,x) = (C(X,x) - I) / 2 to quantify.

[0076] In the system - stable shape - matching algorithm, it can be known that the transformation matrix A r is the minimum value of the squared energy:

[0077] ∑ i m i ||A r q i -p i || 2 ;

[0078] Among them, where q i = x0 i -c 0 r represents the initial position of the particle relative to the center of gravity, p i = x i -c r represents the current position of the particle relative to the center of gravity;

[0079] In continuum mechanics, the deformation gradient is defined as follows: Taking a certain material point inside the object as the center of gravity, its initial position and current position are X cm and x cm , respectively. Another material point is infinitely close to this center, and its initial position and current position are X1 and x1, respectively. Then the correlation between this material point and the center point can be described as:

[0080] x cm - x1 = F(X cm - X1);

[0081] where F becomes the deformation gradient tensor. If the center point is set as the centroid, and adding that F approximates the deformation of the particles in this region, then F is the minimum value of the following quadratic energy, that is:

[0082] ∑ i w i ||Fq i - p i || 2 ;

[0083] where w i is the weight coefficient. At this time, setting the value of the weight to m i , then the two formulas are exactly the same. In the shape matching algorithm for system stability, in order to obtain the optimal rotation matrix, the common approach is to directly set the weight coefficient equal to the mass. Thus, the transformation matrix A r as an approximation of the deformation gradient of the local region r, its tensor component S r is the right stretch tensor, that is: In the simulation based on example shape matching, in order to measure the similarity between deformations, the tensor component S r ∈ R 6 is used to quantify the given deformation configuration x;

[0084] If x(X) is piecewise linear in elements, the Green tensor is invariant in each tetrahedron. Except for structures with inverted elements, the 6m vector S(x) T =(S T 1,…,S T m ) ∈ R 6mFully capable of encoding any specified deformation. The deformation encoded in this case is a unique deformation descriptor, especially in the case where the deformation remains invariant under basic rotation. Each deformation can be mapped to a deformation descriptor, and an effective deformation can also be calculated from any deformation description because the inconsistent target positions estimated through the overlapping regions are mean-mixed. The reconfigurable descriptor space, i.e., the mirror mapping x → S(x), is a realizable manifold As Figure 2 shown

[0085] Since the shape descriptor is linear, it means that the linear combination of two deformation descriptors is generally reconfigurable, i.e., it corresponds to the pose of any deformation. If two specified example poses x1 and x2 are known, their descriptors can be interpolated between these two examples. In the R 6m space, S1 = S(x1) and S2 = S(x2), then we have:

[0086] S(w) = (1 - w)S1 + wS2;

[0087] where w is the interpolation weight. This method linearly inserts the stretching and rotation components for each element and can perform smooth interpolation for all elements, as Figure 3 shown

[0088] In fact, the length of any line segment inside the interpolated element is constrained by its corresponding length.

[0089] Convert the k example deformations obtained in step S31 into deformation space description variables S i = S(x i ), where S represents the graphic mapping and x i represents the current example deformation, thereby forming an example manifold;

[0090] S33. Continuously approximate by minimizing the following quadratic energy expression:

[0091]

[0092] where the vector norm |·|ζ is used to match the sum of the basic tensor two-norms, and the objective function W(x w , w) serves as the interpolation energy. Its minimization defines the projection ∏: R 6m → ζ to the mapping, where R represents the pure rotation matrix, m represents the space dimension, and the mapping of its interpolation line segment S(w), w ∈ [0, 1] is an example line segment S(xw )

[0093] S34. Set x w as the static state. If S 0 = S(x 0 ) represents the initialization deformation descriptor of the elastic object, and S k = S(x k ) represents the k-th example deformation descriptor, then the convex combination closest to the current pose descriptor S = S(x) is obtained , that is, the weight parameters w1,..., w n need to be calculated. The minimization of the quadratic energy expression can be rewritten as:

[0094]

[0095] The solution w T = (w1,..., w n ) 2 can be obtained as:

[0096] w = (L T L) -1 L T (S - S0);

[0097] where L = (S1 - S0... S n - S0) ∈ R 6mn is a constant during the simulation, and thus the value of w0 is

[0098] S35. Obtain the next preferred deformation state;

[0099] S36. Repeat steps S33 to S35 to obtain the simulation of the entire example-based deformation process. When all example shapes are projected onto the example manifold, the object will undergo a series of continuous deformations to form a motion animation.

[0100] S4. Use the feature-driven method to make the three-dimensional heart model continuously repeat the process of example shape matching;

[0101] Among them, the specific steps of step S4 include:

[0102] S41. Describe the dynamic deformation behavior of the heart model;

[0103] If the deformation behavior of the heart is represented by W(X, x) with elastic characteristics for control, where X represents the position of the undeformed vertex and x represents the position of the deformed vertex, combined with the internal deformation driving force and the external force f ext , then the dynamic deformation behavior of the three-dimensional heart model can be described by the following motion equation:

[0104]

[0105] Among them, M is the mass matrix, is the nodal acceleration; since during the simulation of the heart beating process, the heart model does not interact with the outside world such as collisions and frictions, the external force f ext can be ignored. In this case, the contraction and relaxation deformation movement of the heart model can be considered as the result of internal deformation. Therefore, it can be considered that X is the only control parameter, that is, the heart model will generate the deformation required for movement according to the configuration of its preferred deformation space.

[0106] To standardize this method, let g(x, t) represent the vector-valued objective function, and the components of this function can measure the deviation of the given time movement target. The current goal is to find the undeformed and deformed shape configurations that can minimize the distance of the movement target while satisfying the above-mentioned movement equations. Therefore, the state optimization formula for the dynamic deformation behavior of the three-dimensional heart model is:

[0107]

[0108] S42. Describe the static deformation behavior of the heart model;

[0109] In a general setting, each vertex in the static form is an independent control variable of the optimization problem described by the above movement equation. To solve this problem, a simplified linear subspace is used to configure the static deformation behavior of the three-dimensional heart model, and the configuration formula of the linear subspace is as follows:

[0110] X(p) = X0 + Lp;

[0111] where X0 is the initial state, L is the linear mapping between the static form coordinates X and the control variable p, and the state optimization formula for the static deformation behavior of the three-dimensional heart model is:

[0112]

[0113] S42. Describe a controller for heart movement;

[0114] Convert the target of the deformation movement into a target sequence, and use g(x, t) to convert it into an algebraic function of the state variables. Then the vector-valued objective function is the result of connecting linear constraints of any number of vertex positions, and its expression form is as follows:

[0115] g(x, t) = [w1g(x, t), w2g(x, t), …, w n g(x, t)] T ;

[0116] gi (x, t) = ∑ j w xj x j -s i (t);

[0117] where s(t) represents an objective function that varies with time, and the weight w i is used to weigh the importance of the objective g i relative to other objectives. Due to the existence of the weight w xj , each objective constraint g i will affect a subset of the object vertices. For the centroid constraint used in this chapter, the value of the vertex weight w xj is determined by m j / ∑i. For vertices not affected by a specific objective, the weight is set to zero.

[0118] S5. Construct a heart beating simulation system.

[0119] Example 2

[0120] Please refer to Figure 4 as shown. Based on the example shape matching and feature-driven heart beating simulation terminal, it includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the example shape matching and feature-driven heart beating simulation method in Example 1.

[0121] The above are only the embodiments of the present invention, and do not limit the patent scope of the present invention. Any equivalent transformation made using the description and drawings of the present invention, or directly or indirectly applied in the relevant technical fields, shall be equally included in the patent protection scope of the present invention.

Claims

1. A method for simulating heartbeats based on example shape matching and feature driving, characterized in that Including the steps: S1. Reconstruct a three-dimensional heart model; S2. Input an electrocardiogram signal, and use the slope threshold method to extract features from the electrocardiogram signal to obtain electrocardiogram feature data, and use the electrocardiogram feature data to drive the heartbeat of the three-dimensional heart model; S3. Use the method based on example shape matching to simulate the deformation that occurs during the heartbeat; S4. Use the feature-driven method to make the three-dimensional heart model continuously repeat the process of example shape matching; S5. Build a heartbeat simulation system; The specific steps of step S4 include: S41. Describe the dynamic deformation behavior of the heart model; If the deformation behavior of the heart is represented by W(X, x) with elastic properties for control, then represents the position of the vertex where no deformation has occurred, represents the position of the vertex after deformation, combined with the internal deformation driving force and the external force , then the dynamic deformation behavior of the three-dimensional heart model can be described by the following motion equation: ； Among them, is the mass matrix, is the nodal acceleration; Let represent the objective function of vector values, and the state optimization formula for the dynamic deformation behavior of the three-dimensional heart model is as follows: ； S42. Describe the static deformation behavior of the heart model; Use a simplified linear subspace to configure the static deformation behavior of the three-dimensional heart model. The configuration formula of the linear subspace is as follows: ； Where X0 is the initial state, L is the linear mapping between the static form coordinates X and the control variable p. The state optimization formula of the static deformation behavior of the three-dimensional heart model is: ； S42. Describe a controller for heart movement; Convert the target of the deformation motion into a target sequence and use Converted into an algebraic function of state variables, the vector-valued objective function is the result of connecting linear constraints of any number of vertex positions, and its expression is as follows: ； ； Among them, represents a time-varying objective function, and the weight is used to weigh the objective with respect to the importance relative to other objectives.

2. The method for simulating heart beats based on example shape matching and feature driving according to claim 1, wherein The specific steps of step S1 include: S11. Extract the heart region dataset from the heart picture set in the VHP dataset and preprocess it; S12. Use the region growing method with a defined threshold to segment the heart contour in the preprocessed heart region dataset; S13. Use the method of drawing by region to realize the reconstruction of the three-dimensional patch model; S14. Use the graphics interface to realize the rendering of the three-dimensional heart model.

3. The method for simulating heart beats based on example shape matching and feature driving according to claim 2, wherein, The specific steps of step S3 include: S31. Represent the position of the particle with the x - coordinate. After the particle responds to environmental collisions and external forces, a set of exemplary deformations are obtained, which are represented by respectively; S32. Suppose that the discrete representation of an elastic solid in a tetrahedral mesh is known to be \(n\) nodes and \(c\) elements. Then let denote the position vector of the undeformed body, and let denote the configuration of the deformed body; In the stable shape matching algorithm of the system, it can be known that the transformation matrix is the minimum value of the squared energy: ； Among them, among them represents the initial position of the particle relative to the center of gravity, represents the current position of the particle relative to the center of gravity; Take the transformation matrix as an approximation of the deformation gradient of the local region , whose tensor components are the right stretch tensors, i.e.: , and use the tensor components to quantify the given deformed configuration x; Convert the k example deformations obtained in step S31 into deformation space description variables S in turn i = S(x i ), where S represents a graphical mapping, and x i represents the current example deformation, thereby forming an example manifold; S33. Continuously approximate by the following quadratic energy minimization expression: ； Among them, the vector norm is used to match the sum of the basic tensor two-norms, and the objective function serves as the interpolation energy, and its minimization defines the projection : onto , where represents a pure rotation matrix, represents the spatial dimension, and its interpolation line segment is mapped under the projection to be an example line segment of the achievable manifold ; ; S34. Set as the static state. If represents the initialization deformation descriptor of the elastic object, represents the k-th example deformation descriptor, then the convex combination closest to the current pose descriptor is obtained. That is, it is necessary to calculate the weight parameter . The minimization of the quadratic energy expression can be rewritten as: ； The solution can be obtained as follows: ； Among them, is a constant during the simulation, so as to obtain the value of ; S35. Obtain the next preferred deformation state; S36. Repeat steps S33 to S35 to obtain the simulation of the entire example-based deformation process. When all example forms are projected onto the example manifold, the object will undergo a series of continuous deformations to form a motion animation.

4. A cardiac beating simulation terminal based on example shape matching and feature driving, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that: When the processor executes the computer program, it implements the steps of the example shape matching and feature-driven heartbeat simulation method as described in any one of claims 1-3.

Images