Magnetic resonance imaging permanent magnet design method based on graph neural network and related device

Abstract

The invention discloses a magnetic resonance imaging permanent magnet design method based on a graph neural network and a related device, and relates to the technical field of magnetic resonance imaging, and the method comprises the steps: obtaining the structure parameters of an initial magnetic resonance imaging permanent magnet, and the structure parameters comprise an adjacent matrix and a parameter matrix; constructing a graph neural network model according to the structure parameters of the initial magnetic resonance imaging permanent magnet; according to the graph neural network model and the adjacent matrix, optimizing the parameter matrix to obtain an optimized parameter matrix; and finally, according to the optimized parameter matrix, adjusting the initial magnetic resonance imaging permanent magnet to obtain a designed magnetic resonance imaging permanent magnet. According to the method, the magnetic resonance imaging permanent magnet is designed by adopting the graph neural network model and aiming at optimizing the magnetic field uniformity based on the graph data of a non-European structure of the adjacent matrix and the parameter matrix, so that the magnetic field uniformity of the magnetic resonance imaging permanent magnet can be effectively improved.

Description

Technical Field

[0001] The present application relates to the field of magnetic resonance imaging technology, and in particular to a magnetic resonance imaging permanent magnet design method and related devices based on graph neural network. Background Art

[0002] Compared with medical imaging methods such as X-ray and computed tomography (CT), magnetic resonance imaging (MRI) has the advantages of no radiation and good soft tissue imaging contrast, and plays a huge role in medical diagnosis and life science development. The main magnet is the most basic and important component of the magnetic resonance imager, and its performance directly affects the quality of the magnetic resonance image.

[0003] The development of MRI is dominated by superconducting type. Although it has achieved clearer imaging, its solenoid main magnet system is bulky and difficult to carry out on-site diagnosis and monitoring near patients. There are risks such as liquid helium leakage, strong magnetic field burns, and stimulation and damage to peripheral nerves. In addition, it is expensive and rare. The above problems seriously limit the accessibility of MRI and make it difficult to meet the urgent demand for MRI detection in the context of a deeply aging society.

[0004] In recent years, ultra-low field magnetic resonance imaging technology has attracted more and more attention. Compared with the solenoid-type main magnet, the main magnet used in ultra-low field magnetic resonance imaging is generally made of rare earth permanent magnet material. Compared with the solenoid-type magnet, the uniformity of the main magnetic field of the permanent magnet is poor, the geometric accuracy is insufficient, and the image distortion is serious. The existing permanent magnet is generally composed of multiple magnetic blocks arranged according to a certain theory. In order to obtain a more uniform main magnetic field, researchers have optimized the arrangement of magnetic blocks. For example, Cooley et al. designed a Halbach-type magnet geometry model for head magnetic resonance imaging, which supports three types of magnetic blocks: air, N52 NdFeB magnets, and N54 NdFeB magnets, and optimized the selection of each magnetic block position in the geometric model using genetic algorithms. Compared with the main magnetic field uniformity of 137,870ppm obtained when a single model of magnetic blocks is fully filled, Cooley et al. optimized the main magnetic field uniformity to 13,669ppm. Reilly et al. designed a head magnet geometry model that supports the optimization of the magnetic ring radius, and iteratively adjusted the magnetic ring radius through genetic algorithms. Finally, in a 20cm spherical space, using about 38kg of N48 NdFeB magnets, they achieved a main magnetic field uniformity of 2400ppm. Based on the main magnet designed by Reilly et al., Tewari et al. further used genetic algorithms to optimize the magnetization angle of the magnetic blocks in each magnetic ring, which improved the uniformity of the main magnet by about 18%.

[0005] However, in existing research on permanent magnet design, the magnet geometry model supports a single degree of optimization freedom, and the uniformity optimization algorithm is generally a genetic algorithm. This leads to the following limitations: (1) The single degree of freedom of the magnet geometry model limits the upper limit of the performance of the uniformity of the main magnetic field. Each magnetic block can only be adjusted in a single dimension, and the room for improvement is limited; (2) Although multi-degree-of-freedom adjustment can be achieved through cascade optimization, the existing magnet geometry model still lacks the ability to jointly optimize multiple degrees of freedom, and still has certain limitations on the uniformity of the main magnetic field; (3) The genetic algorithm lacks the use of the geometric relationship information of the magnetic blocks in the magnet, resulting in limited performance of the optimization algorithm. Summary of the invention

[0006] The purpose of this application is to provide a method and related device for designing a permanent magnet for magnetic resonance imaging based on a graph neural network. By performing multi-degree-of-freedom joint optimization of the permanent magnet for magnetic resonance imaging based on the geometric relationship of the magnetic blocks, the magnetic field uniformity of the permanent magnet for magnetic resonance imaging can be effectively improved.

[0007] To achieve the above objectives, this application provides the following solutions:

[0008] In a first aspect, the present application provides a method for designing a permanent magnet for magnetic resonance imaging based on a graph neural network, comprising:

[0009] Acquire the structural parameters of the initial magnetic resonance imaging permanent magnet; the initial magnetic resonance imaging permanent magnet is composed of a plurality of magnetic blocks; the structural parameters include a parameter matrix and an adjacency matrix; the parameter matrix is ​​a matrix including the three-dimensional center coordinates and three-dimensional rotation angles of each magnetic block; the adjacency matrix is ​​a matrix indicating whether there is an adjacent relationship between any two magnetic blocks.

[0010] A graph neural network model is constructed based on the structural parameters of the initial magnetic resonance imaging permanent magnet.

[0011] According to the graph neural network model and the adjacency matrix, the parameter matrix is ​​optimized with the goal of optimizing the uniformity of the magnetic field to obtain an optimized parameter matrix.

[0012] The initial magnetic resonance imaging permanent magnet is adjusted according to the optimized parameter matrix to obtain a designed magnetic resonance imaging permanent magnet.

[0013] In a second aspect, the present application provides a magnetic resonance imaging permanent magnet design device based on a graph neural network, comprising:

[0014] The structural parameter acquisition module is used to acquire the structural parameters of the initial magnetic resonance imaging permanent magnet; the initial magnetic resonance imaging permanent magnet is composed of a plurality of magnetic blocks; the structural parameters include an adjacency matrix and a parameter matrix; the parameter matrix is ​​a matrix including the three-dimensional center coordinates and three-dimensional rotation angles of each magnetic block; the adjacency matrix is ​​a matrix indicating whether there is an adjacent relationship between any two magnetic blocks.

[0015] A graph neural network model building module is used to build a graph neural network model based on the structural parameters of the initial magnetic resonance imaging permanent magnet.

[0016] The parameter optimization module is used to optimize the parameter matrix according to the graph neural network model and the adjacency matrix with the goal of optimizing the magnetic field uniformity to obtain an optimized parameter matrix.

[0017] The adjustment module is used to adjust the initial magnetic resonance imaging permanent magnet according to the optimized parameter matrix to obtain a designed magnetic resonance imaging permanent magnet.

[0018] In a third aspect, the present application provides a computer device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the method for designing a permanent magnet for magnetic resonance imaging based on a graph neural network as described in the first aspect.

[0019] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method for designing a permanent magnet for magnetic resonance imaging based on a graph neural network as described in the first aspect.

[0020] In a fifth aspect, the present application provides a computer program product, comprising a computer program, which, when executed by a processor, implements the steps of the method for designing a permanent magnet for magnetic resonance imaging based on a graph neural network as described in the first aspect.

[0021] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0022] The present application provides a method and related device for designing a permanent magnet for magnetic resonance imaging based on a graph neural network, by obtaining the structural parameters of the initial permanent magnet for magnetic resonance imaging; the structural parameters include an adjacency matrix and a parameter matrix, the parameter matrix is ​​a matrix including the three-dimensional center coordinates and three-dimensional rotation angles of each magnetic block, and the adjacency matrix is ​​a matrix indicating whether there is an adjacent relationship between any two magnetic blocks; and according to the structural parameters of the initial permanent magnet for magnetic resonance imaging, a graph neural network model is constructed; and then according to the graph neural network model and the adjacency matrix, the parameter matrix is ​​optimized to obtain an optimized parameter matrix; finally, according to the optimized parameter matrix, the initial permanent magnet for magnetic resonance imaging is adjusted to obtain a designed permanent magnet for magnetic resonance imaging. In other words, the present application uses a graph neural network model based on the graph data of the non-Euclidean structure of the adjacency matrix and the parameter matrix of the initial permanent magnet for magnetic resonance imaging, with the goal of optimizing the uniformity of the magnetic field, to design the permanent magnet for magnetic resonance imaging, and realizes the joint optimization of the multi-degree-of-freedom of the permanent magnet for magnetic resonance imaging based on the geometric relationship of the magnetic blocks, which can effectively improve the uniformity of the magnetic field of the permanent magnet for magnetic resonance imaging constructed by discrete magnetic blocks. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0024] Figure 1 A schematic diagram of a flow chart of a method for designing a permanent magnet for magnetic resonance imaging based on a graph neural network provided in one embodiment of the present application;

[0025] Figure 2 Another schematic diagram of a flow chart of a method for designing a permanent magnet for magnetic resonance imaging based on a graph neural network provided in one embodiment of the present application;

[0026] FIG3( a ) is a front view of an optimized Halbach type permanent magnet provided in one embodiment of the present application;

[0027] FIG3( b ) is a side view of an optimized Halbach type permanent magnet provided in an embodiment of the present application;

[0028] FIG3( c ) is a three-dimensional view of an optimized Halbach type permanent magnet provided in an embodiment of the present application;

[0029] FIG4( a ) is a three-dimensional field diagram of the magnetic field generated by the optimized Halbach type permanent magnet provided in one embodiment of the present application;

[0030] FIG4( b ) is a two-dimensional cross-sectional view of the magnetic field generated by the optimized Halbach type permanent magnet provided in one embodiment of the present application;

[0031] FIG5( a ) is a schematic diagram of the deflection angle distribution of the magnetic blocks in the magnetic ring with a frequency of 2 of the optimized Halbach type permanent magnet provided in one embodiment of the present application relative to the Halbach angle;

[0032] FIG5( b ) is a schematic diagram of the distribution of deflection angles of the magnetic blocks in the magnetic ring relative to the Halbach angle of the optimized Halbach type permanent magnet having frequencies of 2 and 4 provided in one embodiment of the present application;

[0033] FIG6( a ) is a statistical histogram of the displacement of a magnetic block of an optimized Halbach permanent magnet in the x-axis direction provided in an embodiment of the present application;

[0034] FIG6( b ) is a statistical histogram of the displacement of the magnetic block of the optimized Halbach permanent magnet in the y-axis direction provided in one embodiment of the present application;

[0035] FIG6( c ) is a statistical histogram of the displacement of the magnetic block of the optimized Halbach type permanent magnet in the z-axis direction provided in one embodiment of the present application;

[0036] Figure 7 A schematic diagram of functional modules of a permanent magnet design device for magnetic resonance imaging based on a graph neural network provided in one embodiment of the present application;

[0037] Figure 8 A schematic diagram of functional modules of a permanent magnet design device for magnetic resonance imaging based on a graph neural network provided in another embodiment of the present application;

[0038] Fig. 9 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0039] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0040] The magnetic resonance imaging permanent magnet in this application refers to the permanent magnet in the magnetic resonance imager. As the most basic and core component of the magnetic resonance imager, the permanent magnet needs to provide a sufficiently uniform main magnetic field to reduce image distortion and ensure the spatial resolution and imaging quality of the image. However, in the prior art, Euclidean data is usually used for magnet modeling, lacking an accurate description of the geometric relationship between magnetic blocks; the degrees of freedom supported by the magnet geometric model are relatively single, and the optimization space is limited; the optimization algorithm is relatively traditional and has poor performance. In response to the above problems, this application proposes a method for designing a permanent magnet for magnetic resonance imaging based on a graph neural network. Permanent magnet modeling is performed using graph data with a non-Euclidean structure, a magnet geometric model that supports multi-degree-of-freedom joint optimization is constructed, and a graph neural network is used in the high-dimensional solution space provided by the magnet geometric model, with optimization uniformity as the main goal, to achieve intelligent design of permanent magnets. This application can significantly optimize the magnetic field uniformity of a permanent magnet constructed by discrete magnetic blocks.

[0041] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0042] In an exemplary embodiment, Figure 1 As shown, a method for designing a permanent magnet for magnetic resonance imaging based on a graph neural network is provided, and the method includes the following steps 101 to 104. Among them:

[0043] Step 101, obtaining the structural parameters of an initial magnetic resonance imaging permanent magnet; the initial magnetic resonance imaging permanent magnet is composed of a plurality of magnetic blocks; the structural parameters include a parameter matrix and an adjacency matrix; the parameter matrix is ​​a matrix including the three-dimensional center coordinates and three-dimensional rotation angles (i.e., three-dimensional posture) of each of the magnetic blocks; the adjacency matrix is ​​a matrix indicating whether there is an adjacent relationship between any two magnetic blocks.

[0044] This application proposes a magnet geometry model that supports multi-degree-of-freedom optimization (i.e., the parameter matrix and adjacency matrix of a magnetic resonance imaging permanent magnet) to achieve joint and continuous optimization of multiple degrees of freedom in the subsequent training process of a graph neural network.

[0045] In this embodiment, the multiple degrees of freedom are the three-dimensional center coordinates and three-dimensional rotation angles of the magnetic block (i.e., six degrees of freedom). Each small magnetic block can rotate around its center in three orthogonal directions, and a space is left between the initial positions of each magnetic block to support the displacement of the small magnetic block in the three orthogonal directions.

[0046] In some examples, for a 0 attributes (corresponding to d 0 degrees of freedom), we can get the parameter matrix Using the adjacency matrix To describe the spatial relationship of the magnets. This process is called permanent magnet modeling of non-Euclidean data. Considering that magnets that are closer have stronger spatial correlation, if magnets i and j are close to each other, the corresponding value A(i,j) in the adjacency matrix is ​​assigned a larger value, otherwise it is assigned a smaller value.

[0047] Step 102: construct a graph neural network model (GNN) according to the structural parameters of the initial magnetic resonance imaging permanent magnet.

[0048] Step 103, based on the graph neural network model and the adjacency matrix, the parameter matrix is ​​optimized with the goal of optimizing the magnetic field uniformity to obtain an optimized parameter matrix.

[0049] Step 104: adjusting the initial magnetic resonance imaging permanent magnet according to the optimized parameter matrix to obtain a designed magnetic resonance imaging permanent magnet.

[0050] In some embodiments, step 103 specifically includes:

[0051] Step 103.1, setting a constraint condition; the constraint condition is a moving threshold of the magnetic block; the moving threshold is a value not greater than the gap between the magnetic blocks at the initial moment.

[0052] In order to ensure that the magnets do not interfere with each other during the optimization process, the thresholds for limiting the movement of the magnets in three orthogonal directions are not greater than the gaps between the initial magnet positions in the magnet geometry model.

[0053] Step 103.2, optimizing the parameter matrix according to the graph neural network model, the constraint conditions and the adjacency matrix to obtain an optimized parameter matrix.

[0054] In some embodiments, step 103.2 is specifically as follows: inputting the adjacency matrix and the parameter matrix into the graph neural network model for forward propagation, and truncating the parameter matrix output by the graph neural network model according to the constraint conditions to obtain an optimized parameter matrix.

[0055] The above process is further described below:

[0056] Step ①, for the parameter matrix X and the adjacency matrix A, input the weight parameter as W (1) The first convolutional layer of the graph gets d 1 Dimensional vector d 1 is the weight parameter W (1) The length of the size.

[0057] Step ②, for the parameter matrix H (1)and the adjacency matrix A, which is input with a weight parameter W (2) The second graph convolution layer of 2 Vector d 2 is the weight parameter W (2) The length of the size.

[0058] Step ③, for the parameter matrix H (2) and the adjacency matrix A, which is input with a weight parameter W (3) The third graph convolution layer of the graph has a dimension of d 3 Vector d 3 is the weight parameter W (3) The length of the size.

[0059] Step ④, for the parameter matrix H (3) and the adjacency matrix A, which is input with the weight parameter W (4) The fourth convolutional layer of the graph has a dimension of d 4 Vector d 4 is the weight parameter W (4) The length of the size.

[0060] Among them, X to H (4) The dimension keeps the trend of increasing first and then decreasing: d 0 -d 1 ↑-d 2 ↑-d 3 ↑-d 4 ↓.

[0061] Step ⑤, H (4) Input a fully connected layer FCL, the output dimension is the same as d 0 Consistent parameter matrix

[0062]

[0063] Step ⑥, according to the constraints * Perform truncation processing to obtain the optimized parameter matrix.

[0064] Step 103.3, determining the magnetic field uniformity of the preset area according to the magnetic resonance imaging permanent magnet corresponding to the optimized parameter matrix.

[0065] Step 103.4, based on the magnetic field uniformity of the preset area, a gradient back propagation algorithm is used to update the parameters of the graph neural network model to obtain an updated graph neural network model.

[0066] Step 103.5, update the graph neural network model to the updated graph neural network model, and update the parameter matrix to the optimized parameter matrix, and return to step 103.2; until the magnetic field uniformity in the preset area converges, and the optimized parameter matrix is ​​obtained.

[0067] That is to say, the general process of optimizing the uniformity of the main magnetic field based on GNN is as follows: ①, input the construction parameters (parameter matrix) of the permanent magnet with non-Euclidean structure and the corresponding adjacency matrix into GNN, construct the spatial relationship between the magnetic blocks through GNN, and mine its deep features and dependencies, and output the optimized permanent magnet construction parameters (parameter matrix); ②, calculate the three-dimensional magnetic field and its corresponding uniformity according to the output permanent magnet construction parameters (parameter matrix); ③, use the magnetic field uniformity as the loss function, perform gradient back propagation, and update the parameters of GNN; ④, replace the original input with the permanent magnet construction parameters output by GNN in step ①; ⑤ repeat steps ① to ④ until the uniformity converges, and finally obtain the optimal permanent magnet construction parameters to complete the optimization of magnetic field uniformity. Finally, the permanent magnet design is completed according to the permanent magnet construction parameters output by GNN.

[0068] Among them, the loss value is the difference between the magnetic field uniformity and 0, so the value of the magnetic field uniformity itself is the loss value.

[0069] In some embodiments, step 103.3 specifically includes:

[0070] Step 103.3.1, according to the Biot-Savart law and the magnetic dipole model, calculate the magnetic field generated by the magnetic blocks in the magnetic resonance imaging permanent magnet corresponding to the optimized parameter matrix in the preset area.

[0071] Specifically, according to the Biot-Savart Law, the magnetic field generated by a current-carrying wire at a point in space is:

[0072]

[0073] Where B is the three-dimensional vector magnetic field, μ 0 is the vacuum permeability (4π×10 -7 T·m / A), I is the current in the wire, dl is the vector representing the small segment of the wire, r is the vector distance from the wire to the observation point, and its unit vector is By integrating along the wire path C, we can obtain the magnetic field generated by the wire at the observation point.

[0074] In order to calculate the three-dimensional magnetic field generated by the magnetic block, the magnetic block can be regarded as an equivalent magnetic dipole of a small circular current, and its corresponding magnetic field is:

[0075]

[0076] Where B(r) is the magnetic field generated by the magnetic dipole at position r, m = M·V, M and V represent the magnetization intensity and the volume of the magnetic block respectively. The field diagram of a single magnetic block is calculated by this formula.

[0077] The calculation process can be assisted by simulation. First, a permanent magnet model is established in the simulation software. In the permanent magnet model, the posture and polarization direction of the magnetic block are determined by rotating a single magnetic block around the axis of the field of view (FOV), and then rotating it around the axis of the center of gravity of the magnetic block itself. Based on the posture of the single magnetic block, the corresponding posture of the remaining magnetic blocks on the same magnetic ring is determined; then, the polarization direction of the magnetic block is matched with its posture through an affine transformation that is the same as the posture transformation of the magnetic block.

[0078] The magnetic ring is composed of several discrete magnetic blocks. Here, we first determine a single magnetic block, and then obtain the remaining magnetic blocks through operations similar to "copy" and "paste". The ring arrangement of several magnetic blocks constitutes a magnetic ring. See Figure 3(a)-Figure 3(c) .

[0079] The field diagrams of the magnetic blocks of all magnetic rings are calculated using formula (2), and the main magnetic field in the field of view is obtained by linear superposition.

[0080] In some embodiments, the side length of the selected magnetic block is 2.54 cm, and the grade is NdFeB N57. At the same time, in order to reduce the quantization error of the affine transformation, the dimension of the three-dimensional field map within the FOV is expanded to 100 3 In this embodiment, the FOV is a spherical area with a diameter of 45 cm.

[0081] Step 103.3.2, linearly superimposing the magnetic fields generated by the magnetic blocks in the preset area to obtain the magnetic field generated by the magnetic resonance imaging permanent magnet in the preset area corresponding to the optimized parameter matrix.

[0082] Step 103.3.3, determining the magnetic field uniformity of the preset area according to the magnetic field generated by the magnetic resonance imaging permanent magnet in the preset area corresponding to the optimized parameter matrix.

[0083] In some embodiments, types of the magnetic resonance imaging permanent magnet include Halbach type, C type and H type.

[0084] In some embodiments, when the MRI permanent magnet is of Halbach type, the middle magnetic ring of the MRI permanent magnet has a large aperture, and the two side magnetic rings have small apertures; the magnetic ring is a ring formed by arranging a number of magnetic blocks.

[0085] In some embodiments, the size of the adjacency matrix A is N×N, where N is the number of magnetic blocks in the magnetic resonance imaging permanent magnet; when magnetic blocks i and j in the magnetic resonance imaging permanent magnet are adjacent, the value of the corresponding element A(i, j) in the adjacency matrix is ​​1; when magnetic blocks i and j in the magnetic resonance imaging permanent magnet are not adjacent, the corresponding element A(i, j) in the adjacency matrix is ​​0.

[0086] For example, the magnet geometry model has 2594 magnet positions. When modeling non-Euclidean data for permanent magnets, for 2594 magnets with 6 attributes (degrees of freedom), the parameter matrix can be obtained: Using the adjacency matrix To describe the spatial relationship of the magnets. If magnet i is adjacent to any magnet j, the corresponding value A(i,j) in the adjacency matrix is ​​assigned a value of 1, and the rest of the elements in row i are assigned a value of 0.

[0087] In this embodiment, the appearance of the permanent magnet is as follows: Figure 3(a)-Figure 3(c) Permanent magnets can generate magnetic fields up to 10th order, and their three-dimensional field diagrams and corresponding two-dimensional cross-sectional diagrams are shown in Figure 4(a)-Figure 4(b) As shown. The deflection angle of the magnetic block in the optimized permanent magnet ring relative to the Halbach angle has multiple frequency components. Figure 5(a) shows the deflection angle distribution of the magnetic ring with a frequency of 2, and Figure 5(b) shows the deflection angle distribution of the magnetic ring containing both frequencies 2 and 4. The frequency here refers to the number of times the rotation angle of the magnetic block in the corresponding magnetic ring is periodically repeated within the azimuth range of 0 to 360 degrees. The overall displacement and deflection angle distribution of the optimized permanent magnet block are basically symmetrical. Figure 6(a)-Figure 6(c) A statistical histogram of the magnet displacement is shown.

[0088] The present application also provides an application scenario, which applies the above-mentioned magnetic resonance imaging permanent magnet design method based on graph neural network. Specifically: the magnetic resonance imaging permanent magnet design method based on graph neural network provided in this embodiment can be applied in the magnetic resonance imaging permanent magnet design scenario. The magnetic resonance imaging permanent magnet design scenario includes an initial magnetic resonance imaging permanent magnet determination link and an initial magnetic resonance imaging permanent magnet optimization link; after the initial magnetic resonance imaging permanent magnet is determined in the initial magnetic resonance imaging permanent magnet determination link, the initial magnetic resonance imaging permanent magnet optimization link is entered, and the structural parameters of the initial magnetic resonance imaging permanent magnet are optimized in the initial magnetic resonance imaging permanent magnet optimization link, and the initial magnetic resonance imaging permanent magnet is adjusted according to the optimized structural parameters. The magnetic resonance imaging permanent magnet design method based on graph neural network provided in this embodiment belongs to the initial magnetic resonance imaging permanent magnet optimization link. Specifically according to the structural parameters of the initial magnetic resonance imaging permanent magnet, a graph neural network model is constructed, and the parameter matrix is ​​optimized according to the graph neural network model and the adjacency matrix to obtain an optimized parameter matrix. According to the optimized parameter matrix, the initial magnetic resonance imaging permanent magnet is adjusted to obtain a designed magnetic resonance imaging permanent magnet.

[0089] Based on the same inventive concept, the embodiment of the present application also provides a magnetic resonance imaging permanent magnet design device based on a graph neural network for implementing the magnetic resonance imaging permanent magnet design method based on a graph neural network. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more embodiments of the magnetic resonance imaging permanent magnet design device based on a graph neural network provided below can refer to the limitations of the magnetic resonance imaging permanent magnet design method based on a graph neural network above, and will not be repeated here.

[0090] In an exemplary embodiment, Figure 7 As shown, a magnetic resonance imaging permanent magnet design device based on a graph neural network is provided, comprising:

[0091] The structural parameter acquisition module M01 is used to obtain the structural parameters of the initial magnetic resonance imaging permanent magnet; the initial magnetic resonance imaging permanent magnet is composed of a plurality of magnetic blocks; the structural parameters include an adjacency matrix and a parameter matrix; the parameter matrix is ​​a matrix including the three-dimensional center coordinates and three-dimensional rotation angles of each magnetic block; the adjacency matrix is ​​a matrix indicating whether there is an adjacent relationship between any two magnetic blocks.

[0092] The graph neural network model construction module M02 is used to construct a graph neural network model based on the structural parameters of the initial magnetic resonance imaging permanent magnet.

[0093] The parameter optimization module M03 is used to optimize the parameter matrix according to the graph neural network model and the adjacency matrix, with the goal of optimizing the magnetic field uniformity, to obtain an optimized parameter matrix.

[0094] The adjustment module M04 is used to adjust the initial magnetic resonance imaging permanent magnet according to the optimized parameter matrix to obtain a designed magnetic resonance imaging permanent magnet.

[0095] In an exemplary embodiment, Figure 8 As shown, the present application also provides another magnetic resonance imaging permanent magnet design device based on graph neural network, including:

[0096] Get module M11 for constructing data of non-Euclidean structure to describe permanent magnets.

[0097] The output module M12 is used to input the non-Euclidean structure data into the graph neural network model, and output the permanent magnet parameters after optimization; the permanent magnet parameters include: the three-dimensional posture of each magnetic block in the magnet geometric model.

[0098] The determination module M13 is used to determine the three-dimensional spatial structure of the permanent magnet according to the permanent magnet parameters.

[0099] Compared with the prior art, this application has the following advantages:

[0100] 1. This application proposes a permanent magnet modeling method based on non-Euclidean structure data, which can effectively model and deeply explore the spatial relationship of magnetic blocks through the adjacency matrix, thereby improving the performance of the final designed magnet.

[0101] 2. This application proposes a magnet geometry model that supports multi-degree-of-freedom optimization. This model can jointly optimize the continuous three-dimensional posture parameters of the magnetic block. The optimization space that can be achieved by this model covers all the optimization spaces that can be achieved by existing single-degree-of-freedom optimization methods. By greatly expanding the optimization space, the performance upper limit of magnet optimization is significantly improved.

[0102] 3. This application proposes the application of graph neural network (GNN) to permanent magnet design, and pioneered the introduction of GNN technology into the design and optimization process of ultra-low field magnetic resonance imaging (ULF MRI) permanent magnets, filling the research gap in this field. The adjacency matrix is ​​used to model the geometric relationship of magnetic blocks with non-Euclidean structures, and multi-dimensional joint optimization is achieved in the multi-degree-of-freedom optimization space, breaking through the problem that traditional genetic algorithms are difficult to achieve effective exploration in the multi-degree-of-freedom optimization space and the performance of the designed magnets is limited.

[0103] 4. The permanent magnet optimized in this application can generate a uniform magnetic field up to order 10, which is significantly better than the permanent magnet designed by the existing algorithm.

[0104] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Fig. 9 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the structural parameters of the permanent magnet. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a permanent magnet design method for magnetic resonance imaging based on a graph neural network is implemented.

[0105] Those skilled in the art will understand that Fig. 9 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components. In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0106] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0107] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0108] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0109] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0110] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0111] The technical features of the above embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0112] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for designing permanent magnets for magnetic resonance imaging based on graph neural networks, characterized in that: The magnetic resonance imaging permanent magnet design method based on graph neural network includes: Acquire structural parameters of an initial magnetic resonance imaging permanent magnet; the initial magnetic resonance imaging permanent magnet is composed of a plurality of magnetic blocks; the structural parameters include a parameter matrix and an adjacency matrix; the parameter matrix is ​​a matrix including the three-dimensional center coordinates and three-dimensional rotation angles of each magnetic block; the adjacency matrix is ​​a matrix indicating whether there is an adjacent relationship between any two magnetic blocks; Constructing a graph neural network model according to the structural parameters of the initial magnetic resonance imaging permanent magnet; According to the graph neural network model and the adjacency matrix, the parameter matrix is ​​optimized with the goal of optimizing the magnetic field uniformity to obtain an optimized parameter matrix; The initial magnetic resonance imaging permanent magnet is adjusted according to the optimized parameter matrix to obtain a designed magnetic resonance imaging permanent magnet.

2. The method for designing a permanent magnet for magnetic resonance imaging based on a graph neural network according to claim 1, characterized in that: According to the graph neural network model and the adjacency matrix, the parameter matrix is ​​optimized with the goal of optimizing the magnetic field uniformity to obtain an optimized parameter matrix, which specifically includes: Setting a constraint condition; the constraint condition is a moving threshold of the magnetic block; the moving threshold is a value not greater than the gap between the magnetic blocks at the initial moment; According to the graph neural network model, the constraint conditions and the adjacency matrix, the parameter matrix is ​​optimized to obtain an optimized parameter matrix; Determining the magnetic field uniformity of a preset area according to the magnetic resonance imaging permanent magnet corresponding to the optimized parameter matrix; According to the magnetic field uniformity of the preset area, a gradient back propagation algorithm is used to update the parameters of the graph neural network model to obtain an updated graph neural network model; Update the graph neural network model to the updated graph neural network model, and update the parameter matrix to the optimized parameter matrix, return to step 10 to optimize the parameter matrix according to the graph neural network model, the constraints and the adjacency matrix to obtain the optimized parameter matrix; until the magnetic field uniformity in the preset area converges to obtain the optimized parameter matrix.

3. The method for designing a permanent magnet for magnetic resonance imaging based on a graph neural network according to claim 2, characterized in that: Determining the magnetic field uniformity of a preset area according to the magnetic resonance imaging permanent magnet corresponding to the optimized parameter matrix specifically includes: According to the Biot-Savart law and the magnetic dipole model, calculating the magnetic field generated by the magnetic blocks in the magnetic resonance imaging permanent magnet corresponding to the optimized parameter matrix in the preset area; Linearly superimposing the magnetic fields generated by the magnetic blocks in the preset area to obtain the magnetic field generated by the magnetic resonance imaging permanent magnet in the preset area corresponding to the optimized parameter matrix; The magnetic field uniformity of the preset area is determined according to the magnetic field generated by the magnetic resonance imaging permanent magnet in the preset area corresponding to the optimized parameter matrix.

4. The method for designing a permanent magnet for magnetic resonance imaging based on a graph neural network according to claim 1, characterized in that: The types of the magnetic resonance imaging permanent magnets include Halbach type, C type and H type.

5. The method for designing a permanent magnet for magnetic resonance imaging based on a graph neural network according to claim 1, characterized in that: When the MRI permanent magnet is of Halbach type, the aperture of the middle magnetic ring of the MRI permanent magnet is large, and the apertures of the magnetic rings on both sides are small; the magnetic ring is a ring formed by arranging a number of magnetic blocks.

6. The method for designing a permanent magnet for magnetic resonance imaging based on a graph neural network according to claim 1, characterized in that: The size of the adjacency matrix A is N×N, where N is the number of magnetic blocks in the magnetic resonance imaging permanent magnet; when magnetic blocks i and j in the magnetic resonance imaging permanent magnet are adjacent, the value of the corresponding element A(i, j) in the adjacency matrix is ​​1; when magnetic blocks i and j in the magnetic resonance imaging permanent magnet are not adjacent, the corresponding element A(i, j) in the adjacency matrix is ​​0.

7. A permanent magnet design device for magnetic resonance imaging based on graph neural network, characterized in that: The magnetic resonance imaging permanent magnet design device based on graph neural network includes: A structural parameter acquisition module, used to acquire structural parameters of an initial magnetic resonance imaging permanent magnet; the initial magnetic resonance imaging permanent magnet is composed of a plurality of magnetic blocks; the structural parameters include an adjacency matrix and a parameter matrix; the parameter matrix is ​​a matrix including the three-dimensional center coordinates and three-dimensional rotation angles of each magnetic block; the adjacency matrix is ​​a matrix indicating whether there is an adjacent relationship between any two magnetic blocks; A graph neural network model building module, used to build a graph neural network model according to the structural parameters of the initial magnetic resonance imaging permanent magnet; A parameter optimization module, used to optimize the parameter matrix according to the graph neural network model and the adjacency matrix with the goal of optimizing the magnetic field uniformity, so as to obtain an optimized parameter matrix; The adjustment module is used to adjust the initial magnetic resonance imaging permanent magnet according to the optimized parameter matrix to obtain a designed magnetic resonance imaging permanent magnet.

8. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method for designing a permanent magnet for magnetic resonance imaging based on a graph neural network as described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for designing a permanent magnet for magnetic resonance imaging based on a graph neural network as described in any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method for designing a permanent magnet for magnetic resonance imaging based on a graph neural network as described in any one of claims 1 to 6 is implemented.

Images