CN108898622A - A kind of the representation of athletic method, apparatus and computer readable storage medium of heart - Google Patents

Abstract

The invention discloses the representation of athletic method, apparatus and computer readable storage medium of a kind of heart, and the MR image by extracting first phase, second phase is handled and constructs to obtain first phase heart graph structure A and second phase heart graph structure B_i, convex objective function is matched using two heart graph structure solution figures and obtains the vertex set with corresponding relationship, deformation function is constructed based on the vertex set with corresponding relationship, and using it to B_iIt carries out deformation and obtains new second phase heart graph structure B_i+1, calculate B_iWith B_i+1Changing value and judge whether changing value is greater than preset threshold, if then enabling i=i+1 and recycling step of the present invention, if otherwise using deformation function characterization heart movement.This method constructs to obtain after deformation function by the way of circulation, deformation also is carried out to original image structure using the deformation function that building obtains, the whether accurate of deformation function is further verified according to deformation changing value, therefore this method can more accurately characterize the movement of heart.

Description

A heart motion characterization method, device, and computer-readable storage medium

technical field

The present invention relates to the field of computer technology, and more specifically, to a heart motion characterization method, device and computer-readable storage medium.

Background technique

The use of cardiac image analysis to track the anatomical structure of the heart and the movement changes of heart lesion tissue, that is, to analyze the heart movement, is an important means for the diagnosis of heart disease and the formulation of treatment plans. The existing registration method based on B-spline free deformation and optical flow method Neither the method nor the shape-based structure can accurately characterize the motion of the heart, so how to use the structural characteristics of the myocardium to characterize the motion of the heart more accurately is an urgent problem to be solved.

Contents of the invention

The main purpose of the present invention is to provide a heart motion characterization method, aiming at solving the technical problem that the prior art cannot accurately characterize the heart motion.

In order to achieve the above object, the present invention provides a method for characterizing the motion of the heart, the method comprising:

Step 1. Extract the first phase magnetic resonance MR image and the second phase MR image of the heart, and use the fully convolutional neural network to segment the first phase MR image and the second phase MR image respectively to obtain the endocardial contour and cardiac Epicardial contour, and sampling on the endocardial contour and epicardial contour to obtain several cardiogram vertices;

Step 2. Connect the center of the epicardium contour in the center of the first phase MR image with the apex of the cardiogram on the first phase MR image to obtain the structure A of the first phase cardiogram; connect the center of the epicardium contour in the center of the second phase MR image with the The vertex of the cardiogram on the two-phase MR image, to obtain the second phase cardiogram structure B _i ; the initial value of i is 1;

Step 3, use the first phase cardiogram structure A and the second phase cardiogram structure B _i to solve the graph matching convex objective function, and obtain the corresponding relationship between the vertices of the first phase cardiogram structure A and the second phase cardiogram structure B _i set of vertices;

Step 4, constructing a deformation function based on the set of vertices with the corresponding relationship, and using the deformation function to deform the second phase cardiogram structure B _i to obtain a new second phase cardiogram structure B _i+1 ;

Step 5. Calculate the change value between the second phase cardiogram structure B _i and the new second phase cardiogram structure B _i+1 . When the change value is greater than the preset threshold, set i=i+1 and return to step 3; Otherwise, the deformation function is used to characterize the motion of the heart.

Optionally, the first phase cardiogram structure A and the second phase cardiogram structure Bi are represented by quadruples {P1, E1, G1, H1}, {P2, E2, G2, H2} respectively;

In the first phase cardiogram structure A, it is a set of point features of n1 _{dp -} dimensional cardiogram vertices;

In the first phase cardiogram structure _A , m1 d _e -dimensional edge feature set;

G ₁ and H ₁ are the collections of the starting point and end point of the cardiogram constituting a side in the first phase cardiogram structure A respectively, The vertex set consists of endocardial vertices and epicardial vertices. .

In the first phase cardiogram structure A, it is a set of point features of n _{2 dp} -dimensional cardiogram vertices;

In the first phase cardiogram structure A, m ₂ d _e -dimensional edge feature sets;

G ₂ and H ₂ are the collections of the starting point and end point of the cardiogram constituting a side in the second phase cardiogram structure B _i respectively,

Optionally, in step 1, sampling is performed on the endocardial contour and epicardial contour to obtain several endocardial vertices and epicardial vertices respectively, then the graph matching convex objective function in step 3 includes:

in, Ax-b=0;

x=vec(K _p ), y=vec(Y), p ₂ =vec(P ₂ ), e ₂ =vec(E ₂ ), vec is a vectorization operator;

for Hadama, is the Kronecker product, λ, γ are constants;

1 _m×n is an m×n matrix whose elements are all 1, I _n is an n×n identity matrix;

n ₁ and n ₂ are the numbers of the central endocardial vertices and epicardial vertices of the first phase cardiogram structure A and the second phase cardiogram structure Bi, respectively;

Optionally, the set of vertices with corresponding relationship includes the set of vertices U={u _i , i=1,...K} of the first phase cardiogram structure A and the set of vertices of the second phase cardiogram structure B _i Vertex set V={v _i , i=1,...,K}; the point feature set includes the vertex positions of the heart map vertices, and step 4 includes:

The iterative threshold shrinkage algorithm is used to solve the objective function to obtain the affine deformation coefficient and the elastic transformation coefficient. The objective function is:

φ is the radial basis function, ||*|| ₂ is the Euclidean distance, z=[α ₁ ,...α _K ,β ₀ ,β ₁ ,β ₂ ] ^T , the elastic transformation coefficient is α=[α ₁ , ...α _K ] ^T , the affine deformation coefficient is β=[β ₀ ,β ₁ ,β ₂ ] ^T , D _c =[0,1,1] ^T , λ ₁ and λ ₂ are constants, R is the field of real numbers;

Using the affine deformation coefficient, the elastic transformation coefficient and the radial basis function to construct the deformation function:

Use the deformation function to map the vertex positions in the point feature set P2 of the _second phase _cardiogram structure Bi to obtain a new vertex, and connect the points of the epicardial contour in the second phase MR image Centering with the new vertex, a new second phase cardiogram structure B _i+1 is obtained.

Optionally, in step 5, the change value is the change value between the point positions before and after the mapping.

Further, the present invention also provides a cardiac motion characterization device, which includes a processor, a memory, and a communication bus;

The communication bus is used to realize the connection communication between the processor and the memory;

The processor is used to execute one or more programs stored in the memory to achieve the following steps:

Step 1. Extract the first phase magnetic resonance MR image and the second phase MR image of the heart, and use the fully convolutional neural network to segment the first phase MR image and the second phase MR image respectively to obtain the endocardial contour and cardiac Epicardial contour, and sampling on the endocardial contour and epicardial contour to obtain several cardiogram vertices;

Step 2. Connect the center of the epicardium contour in the center of the first phase MR image with the apex of the cardiogram on the first phase MR image to obtain the structure A of the first phase cardiogram; connect the center of the epicardium contour in the center of the second phase MR image with the The vertex of the cardiogram on the two-phase MR image, to obtain the second phase cardiogram structure B _i ; the initial value of i is 1;

Step 3. Using the first phase cardiogram structure A and the second phase cardiogram structure B _i to solve the graph matching convex objective function, and obtain the vertices of the first phase cardiogram structure A and the second phase cardiogram structure B _i The set of vertices with corresponding relationship in ;

Step 4. Construct a deformation function based on the vertex set, and use the deformation function to deform the second phase cardiogram structure B _i to obtain a new second phase cardiogram structure B _i+1 ;

Step 5. Calculate the change value between the second phase cardiogram structure B _i and the new second phase cardiogram structure B _i+1 . When the change value is greater than the preset threshold, set i=i+1 and return to step 3; Otherwise, the deformation function is used to characterize the motion of the heart.

Optionally, the processor is used to execute one or more programs stored in the memory, so as to realize: using the quaternion {P ₁ , E ₁ , G ₁ , H ₁ }, {P ₂ , E ₂ , G ₂ , H ₂ } represents the first phase cardiogram structure A and the second phase cardiogram structure B _i ;

In the first phase cardiogram structure A, it is a set of point features of n1 _{dp -} dimensional cardiogram vertices;

In the first phase cardiogram structure _A , m1 d _e -dimensional edge feature set;

G ₁ and H ₁ are the collections of the starting point and end point of the cardiogram constituting a side in the first phase cardiogram structure A respectively, The vertex set consists of endocardial vertices and epicardial vertices. .

Be the point feature collection of the cardiogram vertices of n _{2 dp} dimensions in the second phase cardiogram structure B _i ;

Be the edge feature set of m ₂ d _e dimensions in the second phase cardiogram structure B _i ;

G ₂ and H ₂ are the collections of the starting point and end point of the cardiogram constituting a side in the second phase cardiogram structure B _i respectively,

Optionally, in step 1, sampling is performed on the endocardial contour and epicardial contour to obtain several endocardial vertices and epicardial vertices respectively, then the graph matching convex objective function in step 3 includes:

in, Ax-b=0;

x=vec(K _p ), y=vec(Y), p ₂ =vec(P ₂ ), e ₂ =vec(E ₂ ), vec is a vectorization operator;

for Hadama, is the Kronecker product, λ, γ are constants;

1 _m×n is an m×n matrix whose elements are all 1, I _n is an n×n identity matrix;

n ₁ and n ₂ are the numbers of the central endocardial vertices and epicardial vertices of the first phase cardiogram structure A and the second phase cardiogram structure Bi, respectively;

Optionally, the corresponding set of vertices includes the vertex set U={u _i , i=1,...K} of the first phase cardiogram structure A and the second phase cardiogram structure B _i 's vertex set V={v _i , i=1,...,K}; the point feature set includes the vertex position of the heart map vertex;

The processor is used to execute one or more programs stored in the memory to achieve step 4:

The iterative threshold shrinkage algorithm is used to solve the objective function to obtain the affine deformation coefficient and the elastic transformation coefficient. The objective function is:

φ is the radial basis function, ||*|| ₂ is the Euclidean distance, z=[α ₁ ,...α _K ,β ₀ ,β ₁ ,β ₂ ] ^T , the elastic transformation coefficient is α=[α ₁ , ...α _K ] ^T , the affine deformation coefficient is β=[β ₀ ,β ₁ ,β ₂ ] ^T , D _c =[0,1,1] ^T , λ ₁ and λ ₂ are constants, R is the field of real numbers;

Using the affine deformation coefficient, the elastic transformation coefficient and the radial basis function to construct the deformation function:

Use the deformation function to map the vertex positions in the point feature set P2 of the _second phase _cardiogram structure Bi to obtain a new vertex, and connect the center of the epicardial contour of the second phase MR image with the new vertex to obtain a new vertex Second Phase Cardiogram Structure B _i+1

Further, the present invention also provides a computer-readable storage medium. The computer-readable storage medium stores one or more programs, and one or more programs can be executed by one or more processors to realize the above heart Steps of the motion characterization method.

Beneficial effect

The present invention provides a heart motion characterization method, device and computer-readable storage medium, which can realize the characterization of heart motion through the following steps:

Step 1. Extract the first phase magnetic resonance MR image and the second phase MR image of the heart, and use the fully convolutional neural network to segment the first phase MR image and the second phase MR image respectively to obtain the endocardial contour and cardiac Epicardial contour, and sampling on the endocardial contour and epicardial contour to obtain several endocardial vertices and epicardial vertices;

Step 2. Connect the center of the epicardial contour in the center of the first phase MR image with the endocardial apex and the epicardial apex to obtain the first phase cardiogram structure A; connect the center of the epicardial contour in the second phase MR image with the endocardium Membrane vertex and epicardium vertex to obtain the second phase cardiogram structure B _i ; the initial value of i is 1;

Step 3, use the first phase cardiogram structure A and the second phase cardiogram structure B _i to solve the graph matching convex objective function, and obtain the corresponding relationship between the vertices of the first phase cardiogram structure A and the second phase cardiogram structure B _i set of vertices;

Step 4. Construct a deformation function based on the corresponding vertex set, and use the deformation function to deform the second phase cardiogram structure B _i to obtain a new second phase cardiogram structure B _i+1 ;

Step 5. Calculate the change value between the second phase cardiogram structure B _i and the new second phase cardiogram structure B _i+1 . When the change value is greater than the preset threshold, set i=i+1 and return to step 3; Otherwise, the deformation function is used to characterize the motion of the heart.

After the heart motion characterization method uses a cyclic method to construct the deformation function, it also uses the constructed deformation function to deform the heart map structure obtained last time, calculates the change value of the heart map structure before and after deformation, and then verifies the deformation function. Whether it is accurate, so as to obtain a deformation function that can characterize the motion of the heart, this method can more accurately characterize the motion of the heart compared with the prior art.

Description of drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings that need to be used in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only These are some embodiments of the present invention. For those skilled in the art, other drawings can also be obtained according to these drawings without creative work.

FIG. 1 is a basic flowchart of a heart motion characterization method provided by an embodiment of the present invention;

Fig. 2 is the heart graph structure that existing Delaunay triangulation algorithm divides into pieces;

Fig. 3 is the structure of the cardiograph of the embodiment of the present invention;

Fig. 4 is a schematic structural diagram of a heart motion characterization device provided by an embodiment of the present invention.

Detailed ways

In order to make the purpose, features and advantages of the present invention more obvious and understandable, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described The embodiments are only some of the embodiments of the present invention, but not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative efforts belong to the protection scope of the present invention.

Please refer to Fig. 1, Fig. 1 is the basic flowchart of the motion characterization method of the heart provided by this embodiment, the method includes:

S101. Extract the first phase MR image and the second phase MR image, and segment and sample them respectively to obtain a number of cardiogram vertices.

Specifically, the step S101 includes: extracting the first phase magnetic resonance MR image and the second phase MR image of the heart, and using a fully convolutional neural network to segment the first phase MR image and the second phase MR image respectively to obtain the first phase MR image. The endocardial contour and epicardial contour of the left ventricle of the image, and the left ventricle endocardial contour and epicardial contour of the second phase MR image, and in the first phase MR image, the second phase MR image of the cardiac Sampling is performed on the endocardial contour and the epicardial contour to obtain several cardiogram vertices on the corresponding phase MR image.

It should be understood that the fully convolutional neural network in step S101 is obtained by training the fully convolutional neural network with a large amount of cardiac image data with manual labeling results. Currently, the available cardiac image data include images from Toronto Children's Hospital The 33 cardiac short-axis MR images provided by the diagnostic department, this data set provides the segmentation results of the left ventricle in 20 phases, and can provide about 5,000 training data; another training set is the heart image data of the University of Toronto, Canada, The data has a total of 45 patient data, each of which provides the segmentation results of the left ventricle at the end of systole and end of diastole, and can provide about 500 training data. Segment the MR image by training the fully convolutional neural network to obtain the endocardial and epicardial contours of the left ventricle. In some other examples, in order to further improve the segmentation accuracy, solve the over-segmentation of the full convolutional network segmentation. In the case of ellipse detection, the segmentation accuracy can be improved.

S102. Connect the center of the epicardial contour on the same MR image with the apex of the cardiogram to obtain a first phase cardiogram structure A and a second phase cardiogram structure B _i .

The S102 step includes: connecting the center of the central epicardium contour of the first phase MR image with the apex of the cardiogram on the first phase MR image to obtain the first phase cardiogram structure A; connecting the center of the central epicardium contour of the second phase MR image and the cardiogram vertex on the second phase MR image to obtain the second phase cardiogram structure B _i ; the initial value of i is 1;

Referring to Fig. 2, it is the heart diagram structure formed by the method in the embodiment of the present invention, which is different from the heart diagram structure obtained by the Delaunay triangulation algorithm as shown in Fig. The graph structure describes the near-circular structure of the left ventricle. Its connecting edges have anisotropic directionality, while their lengths are similar. Each edge has a clear geometric interpretation, and the edge similarity measure has an important influence on the graph similarity. With a clear auxiliary meaning, it is very suitable for describing the anatomy of the left ventricle.

In some other examples, the first phase cardiogram structure A and the second phase cardiogram structure B _i can respectively adopt the quadruple {P ₁ , E ₁ , G ₁ , H ₁ }, {P ₂ , E ₂ , G ₂ , H ₂ } represent;

In the first phase cardiogram structure A, it is a set of point features of n1 _{dp -} dimensional cardiogram vertices;

In the first phase cardiogram structure _A , m1 d _e -dimensional edge feature set;

G ₁ and H ₁ are the collections of the starting point and end point of the cardiogram constituting a side in the first phase cardiogram structure A respectively, The vertex set consists of endocardial vertices and epicardial vertices. .

Be the point feature collection of the cardiogram vertices of n _{2 dp} dimensions in the second phase cardiogram structure B _i ;

Be the edge feature set of m ₂ d _e dimensions in the second phase cardiogram structure B _i ;

G ₂ and H ₂ are the collections of the starting point and end point of the cardiogram constituting a side in the second phase cardiogram structure B _i respectively,

It should be understood that the point feature set here is a set of feature vectors of all vertices of the heart map; the edge feature set is a set of all edges, where the edges are formed by connecting the center of the epicardial contour and the vertices of the heart map ; G ₁ and H ₁ are two matrices, respectively representing the vertices at both ends of each edge on the first phase graph structure, G ₂ and H ₂ are two matrices, respectively representing the two ends of each edge on the second phase graph structure end point.

S103 , using the first phase cardiogram structure A and the second phase cardiogram structure B _i to solve the graph matching convex objective function to obtain a set of vertices with corresponding relationships.

Specifically, S103 includes using the first phase cardiogram structure A and the second phase cardiogram structure B _i to solve the graph matching convex objective function to obtain the vertex of the first phase cardiogram structure A and the second phase cardiogram structure B _i The set of vertices corresponding to the relationship.

It should be understood that, in another example, the several cardiac graph vertices obtained in step 1 include endocardial vertices and epicardial vertices, and the graph matching convex objective function in step 3 includes:

in, Ax-b=0;

x=vec(K _p ), y=vec(Y), p ₂ =vec(P ₂ ), e ₂ =vec(E ₂ ), vec is a vectorization operator;

for Hadama, is the Kronecker product, λ, γ are constants;

1 _m×n is an m×n matrix whose elements are all 1, I _n is an n×n identity matrix;

n ₁ and n ₂ are respectively the number of intima vertices and epicardial vertices in the center of the first phase cardiogram structure A and the second phase cardiogram structure B _i ;

It should be understood that the sub-matrices K _p and K _q can be used to represent the vertex correspondence degree and edge correspondence degree of the first phase cardiogram structure A and the second phase cardiogram structure B _i respectively, and the first phase cardiogram structure A and the second phase cardiogram structure A The point feature sets of the biphasic cardiogram structure B _i are respectively P ₁ and P ₂ , and the point-to-point correspondence matrix K _p can use the function ||P ₂ -P ₁ K _p || ² under the condition of satisfying the permutation matrix To measure the degree of deviation between point sets, similarly, the edge sets E ₁ and E ₂ of two graphs and the edge correspondence matrix K _q can also be used to measure the degree of deviation between edges ||E ₂ -E ₁ K _q | | ² , where

S104. Construct a deformation function based on the corresponding vertex set, and use the deformation function to deform the second phase cardiogram structure B _i to obtain a new second phase cardiogram structure B _i+1 .

In another example, the set of vertices with corresponding relationship includes the set of vertices U={u _i , i=1,...K} of the first phase cardiogram structure A and the set of vertices of the second phase cardiogram structure B _i V={v _i ,i=1,...,K}, the vertex set here is a subset of the point feature set, and it should be understood that the point feature set also includes the vertex position of the heart map vertex, at this time , construct a deformation function based on the vertex set, and use the deformation function to deform the second phase cardiogram structure Bi, and the steps of obtaining a new second phase cardiogram structure Bi+1 include:

S1041, using the iterative threshold shrinkage algorithm to solve the objective function to obtain the affine deformation coefficient and the elastic transformation coefficient, the objective function is:

φ is the radial basis function, ||*|| ₂ is the Euclidean distance, z=[α ₁ ,...α _K ,β ₀ ,β ₁ ,β ₂ ] ^T , the elastic transformation coefficient is α=[α ₁ , ...α _K ] ^T , the affine deformation coefficient is β=[β ₀ ,β ₁ ,β ₂ ] ^T , D _c =[0,1,1] ^T , λ ₁ and λ ₂ are constants, R is the field of real numbers;

S1042. Using the affine deformation coefficient, the elastic transformation coefficient and the radial basis function to construct the deformation function:

S1043. Use the deformation function to map the vertex positions in the point feature set P2 of the _second phase cardiogram structure B _i to obtain a new vertex, and connect the center of the epicardial contour of the second phase MR image with the new vertex to obtain New second phase cardiogram structure B _i+1 .

What needs to be understood is that R is a field of real numbers, R ² means a two-dimensional plane; u means a point on a two-dimensional plane, u1 means a point on a first-dimensional plane, and u2 means a point on a second-dimensional plane.

S105. Calculate the change value between the second phase cardiogram structure B _i and the new second phase cardiogram structure B _i+1 , and determine whether the change value is greater than a preset threshold.

Specifically, when the change value is greater than the preset threshold, set i=i+1, and return to S103; otherwise, perform S106: use a deformation function to characterize the motion of the heart.

In some other examples, the change value in S106 is the change value between the positions of the second phase cardiogram structure B _i and the new second phase cardiogram structure B _i+1 of the apex of the cardiogram structure B i+1.

The heart motion characterization method provided in this embodiment adopts a cyclic method to construct the deformation function, and then uses the constructed deformation function to deform the original image structure, and further checks whether the deformation function is accurate according to the deformation change value. Therefore, the method Compared with the prior art, the motion of the heart can be accurately characterized. At the same time, this method can not only solve the problem of cardiac motion model estimation, but also obtain the dynamic segmentation results of cardiac images through the deformation model, so as to assist clinical applications such as cardiac disease diagnosis, cardiac function measurement, and analysis.

The present invention also provides a cardiac motion characterization device, as shown in FIG. 4 , which includes a processor 41, a memory 42 and a communication bus 43, wherein:

The communication bus 43 is used to realize connection and communication between the processor 41 and the memory 42;

The processor 41 is used to execute the programs stored in the memory 42, so as to realize each step of the above-mentioned cardiac motion characterization method.

This embodiment also provides a computer-readable storage medium, which is characterized in that the computer-readable storage medium stores one or more programs, and one or more programs can be executed by one or more processors to achieve the above The steps of the described cardiac motion characterization method.

It should be noted that, for the sake of simplicity of description, the aforementioned method embodiments are expressed as a series of action combinations, but those skilled in the art should know that the present invention is not limited by the described action sequence. Because of the present invention, certain steps may be performed in other orders or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification belong to preferred embodiments, and the actions and modules involved are not necessarily required by the present invention.

In the above-mentioned embodiments, the descriptions of each embodiment have their own emphases. For the parts that are not described in detail in a certain embodiment, you can refer to the relevant descriptions of other embodiments. At the same time, the serial numbers of the above-mentioned embodiments of the present invention are for description only On behalf of the advantages and disadvantages of the embodiment, those of ordinary skill in the art can also make many forms under the enlightenment of the present invention without departing from the purpose of the present invention and the scope protected by the claims, and these all belong to the protection of the present invention within.

Claims (10)

1. A motion characterization method of the heart, comprising: Step 1. Extract the first phase magnetic resonance MR image and the second phase MR image of the heart, and use a fully convolutional neural network to segment the first phase MR image and the second phase MR image respectively to obtain the endocardial contour of the left ventricle and epicardial contour, and sampling is performed on the endocardial contour and epicardial contour to obtain several cardiogram vertices; Step 2. Connect the center of the epicardial contour in the first phase MR image with the apex of the cardiogram on the first phase MR image to obtain a first phase cardiogram structure A; connect the second The center of the epicardial contour in the phase MR image and the apex of the cardiogram on the second phase MR image are obtained to obtain a second phase cardiogram structure B _i ; the initial value of i is 1; Step 3, use the first phase cardiogram structure A and the second phase cardiogram structure B _i to solve the graph matching convex objective function, and obtain the vertices of the first phase cardiogram structure A and the second phase cardiogram structure B i A set of vertices with corresponding relationships in the graph structure B _i ; Step 4. Constructing a deformation function based on the corresponding vertex set, and using the deformation function to deform the second phase cardiogram structure B _i to obtain a new second phase cardiogram structure B _i+1 ; Step 5. Calculate the change value between the second phase cardiogram structure B _i and the new second phase cardiogram structure B _i+1 , and when the change value is greater than a preset threshold, let i=i+1 , return to step 3; otherwise, use the deformation function to characterize the motion of the heart. 2. The motion characterization method of the heart according to claim 1, wherein the first phase cardiogram structure A and the second phase cardiogram structure B _i respectively adopt the quadruple {P ₁ , E ₁ , G ₁ , H ₁ }, {P ₂ , E ₂ , G ₂ , H ₂ } represent; said In the first phase cardiogram structure A, it is a set of point features of n1 _{dp -} dimensional cardiogram vertices; said In the first phase cardiogram structure _A , m1 d _e -dimensional edge feature sets; Said G ₁ and H ₁ are respectively the collection of starting points and end points of cardiograms constituting sides in said first phase cardiogram structure A, said said It is a point feature set of n ₂ d _p -dimensional cardiogram vertices in the second phase cardiogram structure B _i ; said It is a set of side features of m ₂ d _e dimensions in the second phase cardiogram structure B _i ; Said G ₂ and H ₂ are respectively the set of starting points and end points of the cardiograms constituting edges in said second phase cardiogram structure B _i , said 3. The motion characterization method of the heart as claimed in claim 2, wherein the heart map vertex in the step 1 includes an endocardium vertex and an epicardium vertex, and the graph matching convex target in the step 3 Functions include: in, Ax-b=0; The x=vec(K _p ), y=vec(Y), p ₂ =vec(P ₂ ), e ₂ =vec(E ₂ ), the vec is a vectorization operator; said For the Hadamard product, the is the Kronecker product, and the λ and γ are constants; The 1 _m×n is an m×n matrix whose elements are all 1, and the I _n is an n×n identity matrix; The n ₁ and n ₂ are respectively the number of intima vertices and epicardial vertices in the centers of the first phase cardiogram structure A and the second phase cardiogram structure B _i ; said said 4. The motion characterization method of the heart as claimed in any one of claims 1-3, wherein the set of vertices with corresponding relations comprises the set of vertices U={u _i of the first phase cardiogram structure A ,i=1,...K} and the vertex set V={v _i ,i=1,...,K} of the second phase cardiogram structure B _i ; the point feature set includes the heart Vertex positions of graph vertices; Said step 4 includes: The iterative threshold shrinkage algorithm is used to solve the objective function to obtain the affine deformation coefficient and the elastic transformation coefficient, and the objective function is: said The φ is the radial basis function, the ||*|| ₂ is the Euclidean distance, the z=[α ₁ ,...α _K ,β ₀ ,β ₁ ,β ₂ ] ^T , the elastic transformation The coefficient is α=[α ₁ ,...α _K ] ^T , the affine deformation coefficient is β=[β ₀ ,β ₁ ,β ₂ ] ^T , and the D _c =[0,1,1] ^T , the λ ₁ and λ ₂ are constants, the The R is a real number field; Using the affine deformation coefficient, the elastic transformation coefficient and the radial basis function to construct a deformation function: Use the deformation function to map the vertex positions in the point feature set P2 of the _second phase _cardiogram structure Bi to obtain a new vertex, and connect the points of the epicardial contour in the second phase MR image Centering with the new vertex, a new second phase cardiogram structure B _i+1 is obtained. 5. The motion characterization method of heart as claimed in claim 4, is characterized in that, in described step 5, described change value is described second phase cardiogram structure _Bi and described new second phase cardiogram The change value of the position of the vertices of the cardiogram of structure B _i+1 . 6. A heart motion characterization device, characterized in that the device comprises a processor, a memory and a communication bus; The communication bus is used to realize connection and communication between the processor and the memory; The processor is used to execute one or more programs stored in the memory, so as to realize the following steps: Step 1. Extract the first phase magnetic resonance MR image and the second phase MR image of the heart, and use a fully convolutional neural network to segment the first phase MR image and the second phase MR image respectively to obtain the endocardial contour of the left ventricle and epicardial contour, and sampling is performed on the endocardial contour and epicardial contour to obtain several cardiogram vertices; Step 2. Connect the center of the epicardial contour in the first phase MR image with the apex of the cardiogram on the first phase MR image to obtain a first phase cardiogram structure A; connect the second The center of the epicardial contour in the phase MR image and the apex of the cardiogram on the second phase MR image are obtained to obtain a second phase cardiogram structure B _i ; the initial value of i is 1; Step 3, use the first phase cardiogram structure A and the second phase cardiogram structure B _i to solve the graph matching convex objective function, and obtain the vertices of the first phase cardiogram structure A and the second phase cardiogram structure B i A set of vertices with corresponding relationships in the graph structure B _i ; Step 4. Constructing a deformation function based on the corresponding vertex set, and using the deformation function to deform the second phase cardiogram structure B _i to obtain a new second phase cardiogram structure B _i+1 ; Step 5. Calculate the change value between the second phase cardiogram structure B _i and the new second phase cardiogram structure B _i+1 , and when the change value is greater than a preset threshold, let i=i+1 , return to step 3; otherwise, use the deformation function to characterize the motion of the heart. 7. The cardiac motion characterization device according to claim 6, wherein the processor is used to execute one or more programs stored in the memory, so as to realize: using the quadruple {P ₁ , E ₁ , G ₁ , H ₁ }, {P ₂ , E ₂ , G ₂ , H ₂ } represent the first phase cardiogram structure A and the second phase cardiogram structure B _i ; said In the first phase cardiogram structure A, it is a set of point features of n1 _{dp -} dimensional cardiogram vertices; said In the first phase cardiogram structure _A , m1 d _e -dimensional edge feature sets; Said G ₁ and H ₁ are respectively the collection of starting points and end points of cardiograms constituting sides in said first phase cardiogram structure A, said said It is a point feature set of n ₂ d _p -dimensional cardiogram vertices in the second phase cardiogram structure B _i ; said It is a set of side features of m ₂ d _e dimensions in the second phase cardiogram structure B _i ; Said G ₂ and H ₂ are respectively the set of starting points and end points of the cardiograms constituting edges in said second phase cardiogram structure B _i , said 8. The motion characterization device of the heart as claimed in claim 7, wherein the heart graph vertex in the step 1 comprises an endocardium vertex and an epicardium vertex, and the graph matching convex object in the step 3 Functions include: in, Ax-b=0; The x=vec(K _p ), y=vec(Y), p ₂ =vec(P ₂ ), e ₂ =vec(E ₂ ), the vec is a vectorization operator; said For the Hadamard product, the is the Kronecker product, and the λ and γ are constants; The 1 _m×n is an m×n matrix whose elements are all 1, and the I _n is an n×n identity matrix; The n ₁ and n ₂ are respectively the number of intima and epicardium vertices in the centers of the first phase cardiogram structure A and the second phase cardiogram structure Bi; said said 9. The motion characterization device of the heart according to any one of claims 6-8, wherein the set of vertices with corresponding relationship comprises the set of vertices U={u _i of the first phase cardiogram structure A ,i=1,...K} and the vertex set V={v _i ,i=1,...,K} of the second phase cardiogram structure B _i ; the point feature set includes the heart Vertex positions of graph vertices; The processor is used to execute one or more programs stored in the memory, so as to realize the step 4: The iterative threshold shrinkage algorithm is used to solve the objective function to obtain the affine deformation coefficient and the elastic transformation coefficient, and the objective function is: said The φ is the radial basis function, the ||*|| ₂ is the Euclidean distance, the z=[α ₁ ,...α _K ,β ₀ ,β ₁ ,β ₂ ] ^T , the elastic transformation The coefficient is α=[α ₁ ,...α _K ] ^T , the affine deformation coefficient is β=[β ₀ ,β ₁ ,β ₂ ] ^T , and the D _c =[0,1,1] ^T , the λ ₁ and λ ₂ are constants, the The R is a real number field; Using the affine deformation coefficient, the elastic transformation coefficient and the radial basis function to construct a deformation function: Use the deformation function to map the vertex positions in the point feature set P2 of the _second phase _cardiogram structure Bi to obtain a new vertex, and connect the points of the epicardial contour in the second phase MR image Centering with the new vertex, a new second phase cardiogram structure B _i+1 is obtained. 10. A computer-readable storage medium, characterized in that the computer-readable storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to realize the The steps of the cardiac motion characterization method described in any one of 1 to 5 are required.