Magnetic particle imaging reconstruction method based on calibration magnetic field generation system matrix

Abstract

The invention belongs to the technical field of magnetic particle imaging, relates to a magnetic particle imaging reconstruction method based on a calibration magnetic field generation system matrix, and aims to solve the problems that a traditional magnetic particle imaging reconstruction method is poor in applicability in asymmetric bilateral structure magnetic particle imaging equipment (A-MPI) and cannot perform imaging in any region. The method comprises the following steps: performing grid division on an imaging area, calibrating magnetic field intensity, and obtaining magnetic field gradient distribution of each point; obtaining a theoretical system matrix according to a mathematical calculation mode; randomly moving the positions of the non-magnetic field points for scanning to obtain actually measured magnetic field gradient distribution and a system matrix, and constructing a data set; expanding the data set based on the simulation system; constructing a deep learning model, and inputting the extended data set and the measured data into the training deep learning model; and inputting a theoretical system matrix at any position, and obtaining an accurate system matrix by using the trained model. According to the invention, the system matrix at any position can be obtained under the A-MPI equipment, so that imaging at any position is realized.

Description

Background Art

[0002] Magnetic particle imaging technology was first proposed in 2005. It uses superparamagnetic iron oxide particles (SPIOs) as contrast agents for imaging, which can generate tomographic images with high sensitivity and high contrast, and has the characteristics of quantifiability and no depth attenuation. As a new type of molecular imaging technology, it is playing an increasingly important role in biomedical research.

[0003] The magnetic particle imaging device with an asymmetric bilateral structure is designed to solve the problems of limited imaging range of existing magnetic nanoparticle imaging devices and insufficient imaging depth for the human torso position. Its main structure includes a fixed end and a handheld end. The fixed end is used to generate a uniformly variable homogeneous magnetic field and drive the free point of the magnetic field to move along the depth direction. The field-free point is generated by the permanent magnet and the magnetic field generating coil at the handheld end. In the imaging method for a large field of view, the position of the field-free point can be moved by moving the handheld end to achieve scanning of different regions. During the imaging scan, targeted local imaging or covering overall imaging of the object to be measured can be performed according to the position of the handheld end, and imaging at different depths can be achieved according to the magnitude of the applied current.

[0004] With the development of magnetic particle imaging technology, its commonly used reconstruction methods are mainly system matrix (SM) and X-Space. The X-Space reconstruction method is based on the time-domain signal of the non-linear response signal for reconstruction. However, for a device with an asymmetric bilateral structure magnetic particle imaging device (A-MPI) that has characteristics such as a small gradient, orthogonal excitation, and orthogonal reception, using X-Space reconstruction is likely to generate large image artifacts. The imaging performance of this method depends to a large extent on the quality of the point spread function. Usually, when the gradient is large, the point spread function performs better. However, although the A-MPI device has a large imaging field of view, the rapid decrease in the selected field gradient will reduce the imaging quality, making the X-Space method unable to be effectively applied to the image reconstruction of A-MPI devices.

[0005] In the reconstruction method based on the system matrix, the ways to obtain the system matrix include model-based system matrix calibration and measurement-based system matrix calibration. In the model-based calibration method, due to the complex magnetization behavior of SPIOs in a complex magnetic field environment, it is extremely challenging to accurately establish a physical model. Currently, although the measurement-based acquisition method is commonly used and relatively accurate, it requires moving the sampling points throughout the imaging field of view to obtain the particle response characteristics at different positions, and the measurement process is extremely time-consuming. For A-MPI devices, which have a large imaging field of view and can image at any position and depth, theoretically, an infinite number of system matrices will be generated. If the traditional measurement method is used to calibrate the system matrix every time a position is moved during measurement, it is almost impossible to implement in actual operation, resulting in the difficulty of implementing the traditional reconstruction method based on system matrix calibration on A-MPI devices, and there is an urgent need for innovation and improvement. Summary of the Invention

[0006] To solve the above problems in the prior art, that is, to solve the problems of poor applicability of the magnetic particle imaging reconstruction method in the prior art in A-MPI devices and the inability to image any region, the first aspect of the present invention proposes a magnetic particle imaging reconstruction method for generating a system matrix based on a calibration magnetic field, which is applied to an asymmetric bilateral structure magnetic particle imaging device, and changes the traditional measurement-based system matrix calibration to magnetic field calibration. The method includes the following steps: S100. Divide the imaging space into grids, calibrate the magnetic field intensity at each point in the grid, and determine the overall magnetic field gradient distribution and the excitation magnetic field intensity; S200. Use the second-kind Chebyshev polynomial to determine the mathematical expression of the ideal system matrix at each point and obtain the theoretical system matrix; S300. Randomly select a position, calibrate the actual system matrix within the imaging field of view at this position, and combine the overall magnetic field gradient distribution to form a data pair of magnetic field gradient distribution - system matrix, and construct a data set; S400. Add actual noise through a simulation system to expand the data set; S500. Construct a U-net network model and pre-train the U-net network model using the expanded data set; S600. Fine-tune the pre-trained U-net network model using the data pair of magnetic field gradient distribution - system matrix to obtain a fine-tuned U-net network model; S700. Input the magnetic field gradient value and the theoretical system matrix at any position to be measured into the fine-tuned U-net network model to obtain the system matrix at the position to be measured, and generate a reconstructed image based on the system matrix.

[0007] In some preferred embodiments, the method for calibrating the magnetic field intensity is as follows: S110. Calibrate the divergent magnetic field of the handheld permanent magnet, grid the uniform magnetic field of the fixed-end permanent magnet, and calibrate the uniform magnetic field intensity at each grid point; S120. Based on the uniform magnetic field intensity, calculate the magnetic field gradient value at each grid through fitting; S130. Interpolate the magnetic field gradient distribution, and define the gradient values at each position according to the position of each grid point; S140. Calibrate the excitation magnetic field intensity of the handheld excitation coil at the grid; S150. Determine the overall magnetic field gradient distribution by combining the gradient value and the excitation magnetic field intensity.

[0008] In some preferred embodiments, a theoretical system matrix is obtained by the following method: Express the two-dimensional system matrix using the tensor product of the second-kind Chebyshev polynomials: ; where k , l represent the orders of the Chebyshev polynomials, x , y represent FOV the two-dimensional coordinates; different orders correspond to the frequency components in the actual system matrix; the coordinates correspond to the positions in the actual system matrix, that is, FOV the number of calibration points of the system matrix; Let the excitation frequencies in the x and y directions satisfy , then: ; ; where represents the value of the system matrix with frequency at the coordinate , represents the entire system matrix diagram corresponding to the frequency i ; f x 、f y are respectively x , y the excitation frequencies in the directions, and

[0009] In some preferred embodiments, the U-net network model includes an encoder and a decoder. The encoder uses ResNet18, and the number of layers of the decoder is the same as that of the encoder. When the U-net network model outputs, feature splicing is performed, and the features output by different layers in the encoder are fused with the corresponding layers in the decoder.

[0010] In some preferred embodiments, a dataset is constructed as follows: S310. Select the magnetic field parameters of the grid corresponding to the required imaging area according to the position of a single imaging, and obtain the system matrix at this position based on this magnetic field distribution through actual measurement; S320. Randomly select the positions of multiple field-free points, and measure the system matrices under different magnetic field parameters respectively to obtain the actual system matrix ; where the system matrix diagram corresponding to the i frequency in is ;

[0011] S330. Combine the actual system matrix with the magnetic field gradient value at the corresponding position in the overall magnetic field gradient distribution to form a data pair of magnetic field gradient distribution - system matrix, and then construct a dataset. In some preferred embodiments, the dataset is augmented as follows: ; S410. Simulate the response signal of magnetic nanoparticles based on the Langevin model to generate a simulated system matrix corresponding to the actual system matrix ; S420. Subtract the simulated system matrix from the actual system matrix to obtain the actual noise distribution ; S430. Set the gradient distribution of the simulated magnetic field to the magnetic field gradient distribution randomly sampled from the overall magnetic field gradient distribution, simulate the system matrix again and add the i actual noise distribution to obtain the simulated overall magnetic field system matrix ; the system matrix diagram corresponding to the

[0012] frequency in

[0013] In some preferred embodiments, the U-net network model includes an encoder and a decoder. When the U-net network model outputs, feature splicing is performed, and the features output by different layers in the encoder are fused with the corresponding layers in the decoder.

[0014] Advantages of the present invention: Based on magnetic fields for calibration, the present invention generates a system matrix at any position from the magnetic field distribution through a established deep learning model, making it possible to obtain the system matrix at any position under an A-MPI device, thereby enabling imaging at any position, greatly enhancing the flexibility of magnetic particle imaging, and meeting diverse imaging requirements, such as comprehensive imaging of complex human structures or precise imaging of specific small regions; Through the optimization of the preliminary theoretical system matrix by the deep learning model, an accurate system matrix is obtained, which helps to improve the accuracy of image reconstruction, and further enhances the quality of magnetic particle imaging, providing clearer and more accurate image information for medical diagnosis and biological research; It avoids the cumbersome operation of moving sampling points in the entire imaging field of view when obtaining the system matrix based on traditional measurements, reduces the measurement cost, saves a large amount of time, improves the imaging efficiency, makes the magnetic particle imaging technology more practical and clinically applicable, and is conducive to the wide promotion of this technology. Description of the Drawings

[0015] By reading the detailed description of the non-limiting embodiments with reference to the following drawings, other features, objectives, and advantages of the present application will become more apparent: Figure 1 is a flowchart of the magnetic particle imaging reconstruction method for generating a system matrix based on a calibrated magnetic field according to the present invention; Figure 2 is a flowchart of expanding the data set in an embodiment of the present invention; Figure 3 is a flowchart of obtaining the system matrix at any position based on the U-net network model in an embodiment of the present invention. Detailed Embodiments

[0016] The following further details the present application with reference to the drawings and embodiments. It can be understood that the specific embodiments described herein are only for explaining the related invention and not for limiting the invention. Additionally, it should be noted that for the sake of description, only parts related to the relevant invention are shown in the drawings.

[0017] It should be noted that, without conflict, the embodiments in the present application and the features in the embodiments can be combined with each other. The present application will be described in detail below with reference to the drawings and embodiments.

[0018] The A-MPI has the device characteristics of small gradient, orthogonal excitation, and orthogonal reception. Aiming at the problem that the traditional reconstruction method based on system matrix calibration cannot be directly applied to A-MPI, the present invention provides a magnetic particle imaging reconstruction method based on generating a system matrix from a calibrated magnetic field. The present invention changes the traditional measurement-based system matrix calibration to magnetic field calibration, generates the system matrix at any position from the magnetic field distribution through the established deep learning model, generates an accurate system matrix based on the overall magnetic field gradient distribution and the theoretical system matrix, realizes imaging at any position, and makes it possible to obtain the system matrix at any position under the A-MPI device.

[0019] A magnetic particle imaging reconstruction method based on generating a system matrix from a calibrated magnetic field according to the present invention includes the following steps: S100. Divide the imaging space into grids, calibrate the magnetic field intensity at each point in the grids, and determine the overall magnetic field gradient distribution and the excitation magnetic field intensity; S200. Use the second-kind Chebyshev polynomial to determine the mathematical expression of the ideal system matrix at each point and obtain the theoretical system matrix; S300. Randomly select a position, calibrate the actual system matrix within the imaging field of view at this position, and combine the overall magnetic field gradient distribution to form a data pair of magnetic field gradient distribution - system matrix, and construct a data set; S400. Add actual noise through a simulation system to expand the data set; S500. Construct a U-net network model and pre-train the U-net network model using the expanded data set; S600. Fine-tune the pre-trained U-net network model using the data pair of magnetic field gradient distribution - system matrix to obtain a fine-tuned U-net network model; S700. Input the magnetic field gradient value and the theoretical system matrix at any position to be measured into the fine-tuned U-net network model to obtain the system matrix at the position to be measured, and generate a reconstructed image based on the system matrix.

[0020] To more clearly illustrate the magnetic particle imaging reconstruction method based on generating a system matrix from a calibrated magnetic field according to the present invention, the following combines Figure 1 to detail each step in the embodiments of the present invention.

[0021] The magnetic particle imaging reconstruction method based on generating a system matrix from a calibrated magnetic field in the first embodiment of the present invention includes the following steps S100 to S700, and is applied to an asymmetric bilateral structure magnetic particle imaging device. Each step is described in detail as follows: S100. Divide the imaging space into grids, calibrate the magnetic field intensity at each point in the grids, and determine the overall magnetic field gradient distribution and the excitation magnetic field intensity. The method is as follows: S110. Calibrate the divergent magnetic field of the handheld permanent magnet , and grid the uniform magnetic field of the fixed - end permanent magnet to calibrate the uniform magnetic field intensity at each grid point; S120. Based on the uniform magnetic field intensity, calculate the magnetic field gradient value at each grid through fitting ; S130. Interpolate the magnetic field gradient distribution and define the gradient value at each position according to the position of each grid point; S140. Calibrate the excitation magnetic field intensity of the handheld excitation coil at the grid, defined as ; Then the total magnetic field at each position is ; S150. Determine the overall magnetic field gradient distribution by combining the gradient value and the excitation magnetic field intensity.

[0022] In a two - dimensional plane, the grid size is divided according to the actual imaging field of view (FOV). In this embodiment, assume that the FOV of A - MPI is 18 * 18 cm, and the grid is divided with a step size of 5 mm. Then the number of grid points is 37 * 37 = 1369. Calibrate the uniform magnetic field intensity at each grid point; then calculate the magnetic field gradient value at each grid through fitting. Next, interpolate the magnetic field gradient distribution to increase the resolution to 1 mm, and the number of grid points is 181 * 181 = 32761; define the gradient value at each position according to the position of each grid point as , where i = 1, 2, …… 32761; then the total magnetic field is , since has two components in the x and y directions, so the final magnetic field gradient distribution is a matrix of 181 * 181 * 2.

[0023] S200. Use the second - kind Chebyshev polynomial to determine the mathematical expression of the system matrix and obtain the theoretical system matrix. The method is as follows: Select the order of the Chebyshev polynomial according to the number of system matrix frequency points selected in the actual system matrix. Secondly, according to the size of the actual calibration FOV, generate the number of grid points. According to the formula, obtain the theoretical system matrix expressed by the Chebyshev polynomial at different orders at each calibration point.

[0024] Preferably, the method for obtaining the theoretical system matrix is as follows: The second - kind Chebyshev polynomial represents the system matrix expression under uniform excitation, uniform reception, and ideal particle model: ; Assuming that the excitation magnetic field is uniform, the difference in the system matrix between different positions comes from the different gradient values. The above equation shows the one-dimensional Chebyshev expression. Under the Lissajou scanning trajectory, the two-dimensional system matrix is expressed by the tensor product of the second-kind Chebyshev polynomials: ; where, k and l represent the orders of the Chebyshev polynomials, x and y represent the two-dimensional coordinates under FOV ; Different orders correspond to the row vectors in the actual system matrix, i.e., the frequency components; the coordinates correspond to the columns of the actual system matrix, i.e., the number of calibration points of the system matrix under FOV ; Let x and y direction excitation frequencies satisfy , then: ; ; where, represents the value of the system matrix at the coordinate of frequency , represents the entire system matrix diagram corresponding to the frequency i ; f x 、f y are x and y direction excitation frequencies respectively, is the density parameter of the Lissajous trajectory during scanning.

[0025] S300. Randomly select positions, calibrate the actual system matrix within the imaging field of view at these positions, and combine with the overall magnetic field gradient distribution obtained in step S100 to form data pairs of magnetic field gradient distribution - system matrix, and construct a data set.

[0026] Preferably, the method for constructing the data set is as follows: S310. Select the magnetic field parameters of the corresponding grid in the required imaging area according to the position of a single imaging, and obtain the system matrix at this position based on this magnetic field distribution through actual measurement; S320. Randomly select positions of multiple field-free points, and measure the system matrices under different magnetic field parameters respectively to obtain the actual system matrix ; where, represents the system matrix diagram corresponding to frequency i in the actual system matrix.

[0027] S330. Combine the actual system matrix with the magnetic field gradient values at the corresponding positions in the overall magnetic field gradient distribution to form data pairs of magnetic field gradient distribution - system matrix, and further construct a data set.

[0028] Preferably, in this embodiment, it is assumed that the size of the single - imaging field of view is 2 * 2 cm, then the corresponding magnetic field gradient distribution map is a 21 * 21 * 2 matrix I; it is assumed that the resolution during system matrix calibration is 1 mm, and k modulation frequencies are taken for one system matrix, then the system matrix is a 21 * 21 * k matrix, where represents the frequency i corresponding system matrix diagram.

[0029] S400. Augment the data set by adding actual noise through a simulation system. The method is as follows: S410. Based on Langevin the model, simulate the response signal of magnetic nanoparticles to generate a simulation system matrix corresponding to the actual system matrix ; wherein, the magnetic field distribution setting of the simulation system is the same as that of the actual measurement; S420. Subtract the simulation system matrix from the actual system matrix to obtain the actual noise distribution ; S430. Set the gradient distribution of the simulation magnetic field to the magnetic field gradient distribution randomly sampled from the overall magnetic field gradient distribution, simulate again to obtain a system matrix and add to obtain a simulated overall magnetic field system matrix ; where represents the system matrix diagram corresponding to the frequency i in the simulated overall magnetic field system matrix; S440. Combine the overall magnetic field gradient distribution with the simulated overall magnetic field system matrix to obtain data pairs of magnetic field gradient distribution - simulated system matrix, and complete the data set augmentation.

[0030] Based on Langevin the model to simulate the response signal of magnetic nanoparticles, those skilled in the art can implement it based on general knowledge and experience, so it is not specifically described in this specification.

[0031] Preferably, in this embodiment, the Langevin function is used to fit the particle response, and the simulated signal is used to generate a corresponding simulated measured system matrix , and the parameter settings of the simulation are kept consistent with those during actual measurement.

[0032] Build a U-net network model and pre-train the U-net network model using the augmented dataset. The U-net network model includes an encoder and a decoder. The encoder uses ResNet18, and the number of layers of the decoder is the same as that of the encoder. When the U-net network model outputs, feature splicing is performed to fuse the features output by different layers in the encoder with the corresponding layers in the decoder.

[0033] Preferably, in this embodiment, the ResNet18 includes four residual modules. The first residual module contains two convolutional layers, and the number of output channels is 64. The number of output channels of the second residual module is 128. The number of output channels of the third residual module is 256. The number of output channels of the fourth residual module is 512. Each decoding module gradually restores the resolution through upsampling operations and further extracts features through specific convolutional operations, which are successively implemented by 4 decoding layers in total to ensure the effective fusion of features in the spatial dimension and semantic information. Finally, the output of the decoder passes through a convolutional layer to generate a result with the same size as the input image.

[0034] Through feature splicing, the features output by different layers in the encoder are fused with the corresponding layers in the decoder, so as to retain more detailed information in the upsampling stage.

[0035] Preferably, when pre-training the U-net network model using the augmented dataset, first preprocess the input data. The method is as follows: For each layer of normalize, convert it into a single-channel image, and scale the image to the specified size; add a Chebyshev polynomial theoretical representation of the system matrix with a dimension corresponding to the frequency i to the overall magnetic field gradient distribution normalize, convert it into a three-channel RGB image, and scale the image to the specified size.

[0036] Further preferably, the output of the U-net network model is a single-channel system matrix map of a certain layer or a multi-channel multi-layer system matrix map.

[0037] By inputting the augmented data, measured data, etc. into the trained deep learning model, the relationship between the magnetic field gradient distribution under each magnetic field grid and the measured system matrix is fitted.

[0038] In this embodiment, the specified size is 64*64. Normalize each layer of the system matrix to the range of [0, 255], convert it into a single-channel image, and finally scale the image to 64*64 size; add a Chebyshev polynomial theoretical representation of the system matrix with a dimension corresponding to the frequency i to the magnetic field gradient distribution ​, and then it is similarly normalized to the range of [0, 255] to become a three-channel RGB image. Finally, the image is scaled to a size of 64*64. There are two ways for the input and output of the model: Method 1: The input of the model is an image formed by concatenating the magnetic field gradient of 64*64*3 and the theoretical system matrix in the channel dimension. The output of the model is a single-channel system matrix image of a certain layer with a size of 64*64*1. This method requires training k models.

[0039] Method 2: The input of the model is an image formed by concatenating the magnetic field gradient of 64*64*3 and the theoretical system matrix. The output of the model is a multi-channel and multi-layer system matrix image with a size of 64*64*k.

[0040] Use the measured data set for data augmentation, extract the noise distribution of the measured data, and combine with the simulation results to construct augmented data close to the measured data.

[0041] S600. Use the real data (the data pair of magnetic field gradient distribution - system matrix) obtained in step S300 to fine-tune the pre-trained U-net network model to obtain the fine-tuned U-net network model and determine the model weights.

[0042] Pre-training on the augmented data and fine-tuning on the measured data are beneficial to the convergence of the model.

[0043] S700. Input the magnetic field gradient value and the theoretical system matrix at any position to be measured into the fine-tuned U-net network model. This network can generate a system matrix approximating the measured current position, obtain the system matrix at the position to be measured, and based on the system matrix and the detected signal, obtain the reconstructed image based on the system matrix reconstruction method.

[0044] Although the above steps are described in the above sequential order in the above embodiments, those skilled in the art can understand that in order to achieve the effects of this embodiment, different steps do not have to be executed in such an order. They can be executed simultaneously (in parallel) or in a reversed order, and these simple changes are all within the protection scope of the present invention.

[0045] The magnetic particle imaging reconstruction system for generating a system matrix based on a calibrated magnetic field according to the second embodiment of the present invention, the system includes a signal acquisition device and a central processing device; The signal acquisition device includes an asymmetric bilateral structure magnetic particle imaging device. The asymmetric bilateral structure magnetic particle imaging device includes a fixed-end magnetic field generating coil, a hand-held permanent magnet, a hand-held excitation coil, and a hand-held receiving coil. The signal acquisition device is configured to divide the imaging space into grids, calibrate the magnetic field intensity at each point in the grids, and obtain the measurement signal; The central processing device includes a CPU and a GPU; the central processing device includes a matrix generation module, a model construction module, a model fine-tuning module, a dataset construction module, an image reconstruction module, and a simulation module; The matrix generation module is configured to determine the overall magnetic field gradient distribution and the excitation magnetic field intensity according to the calibration data, use the second-kind Chebyshev polynomial to determine the mathematical expression of the ideal system matrix at each point, and obtain the theoretical system matrix; determine the actual system matrix according to the measurement signal; and is further configured to input the magnetic field gradient value at any position to be measured and the theoretical system matrix into the fine-tuned U-net network model to obtain the system matrix at the position to be measured; The model construction module is configured to construct a U-net network model and pre-train the U-net network model using the expanded dataset; The model fine-tuning module is configured to fine-tune the pre-trained U-net network model using the magnetic field gradient distribution-system matrix data pair to obtain the fine-tuned U-net network model; The dataset construction module is configured to combine the overall magnetic field gradient distribution to form a magnetic field gradient distribution-system matrix data pair and construct a dataset; The simulation module is configured to add actual noise to the dataset through a simulation system; The image reconstruction module is configured to generate a reconstructed image based on the system matrix at the position to be measured.

[0046] It should be noted that the magnetic particle imaging reconstruction system for generating a system matrix based on a calibrated magnetic field provided in the above embodiment is only illustrated by the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules according to needs, that is, the modules or steps in the embodiments of the present invention can be further decomposed or combined. For example, the modules in the above embodiment can be combined into one module, or further split into multiple sub-modules to complete all or part of the functions described above. For the names of the modules and steps involved in the embodiments of the present invention, they are only used to distinguish each module or step and are not regarded as an improper limitation of the present invention.

[0047] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working process and related descriptions of the system described above can refer to the corresponding process in the foregoing method embodiment and will not be repeated here.

[0048] An electronic device according to the third embodiment of the present invention includes: At least one processor; and A memory communicatively connected to at least one of the processors; wherein, The memory stores instructions executable by the processor, and the instructions are used to be executed by the processor to implement the above-mentioned magnetic particle imaging reconstruction method for generating a system matrix based on a calibration magnetic field.

[0049] A computer-readable storage medium according to a fourth embodiment of the present invention, the computer-readable storage medium stores computer instructions, and the computer instructions are used to be executed by a computer to implement the above-mentioned magnetic particle imaging reconstruction method for generating a system matrix based on a calibration magnetic field.

[0050] Those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working processes and related descriptions of the above-described electronic device and computer-readable storage medium can refer to the corresponding processes in the foregoing method embodiments, and will not be elaborated herein.

[0051] Those skilled in the art should be able to realize that the modules and method steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, computer software, or a combination of the two. The programs corresponding to the software modules and method steps can be placed in a random access memory (RAM), internal memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium well known in the technical field. To clearly illustrate the interchangeability of electronic hardware and software, the composition and steps of each example have been generally described according to functions in the above description. Whether these functions are executed in the form of electronic hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0052] Computer program code for performing the operations of the present application can be written in one or more programming languages or combinations thereof. The above programming languages include object-oriented programming languages - such as Java, Smalltalk, C++, and also include conventional procedural programming languages - such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, executed as an independent software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any type of network - including a local area network (LAN) or a wide area network (WAN) - or can be connected to an external computer (for example, by using an Internet service provider to connect through the Internet).

[0053] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architectures, functions, and operations of systems, methods, and computer program products according to various embodiments of the present application. In this regard, each block in the flowchart or block diagram may represent a module, a segment of a program, or a part of code that contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions noted in the blocks may occur in a different order than that noted in the accompanying drawings. For example, two consecutive blocks shown may actually be executed substantially in parallel, and they may sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, and combinations of blocks in the block diagram and / or flowchart, can be implemented by a dedicated hardware-based system that performs the specified functions or operations, or can be implemented by a combination of dedicated hardware and computer instructions.

[0054] The terms "first", "second", etc. are used to distinguish similar objects and are not used to describe or represent a particular order or sequence.

[0055] The term "comprising" or any other similar term is intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus / device that comprises a series of elements includes not only those elements but also other elements not expressly listed, or elements that are inherent to those processes, methods, articles, or apparatus / device.

[0056] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the protection scope of the present invention.

Claims

1. A magnetic particle imaging reconstruction method based on generating a system matrix with a calibrated magnetic field, applied to an asymmetric bilateral structure magnetic particle imaging device, characterized in that, The method includes the following steps: S100. Divide the imaging space into grids, calibrate the magnetic field intensity at each point in the grids, and determine the overall magnetic field gradient distribution and the excitation magnetic field intensity; S200. Use the second kind of Chebyshev polynomial to determine the mathematical expression of the ideal system matrix at each point and obtain the theoretical system matrix; S300. Randomly select positions, calibrate the actual system matrix within the imaging field of view at these positions, and combine the overall magnetic field gradient distribution to form data pairs of magnetic field gradient distribution - system matrix, and construct a data set; S400. Add actual noise through a simulation system to expand the data set; S500. Construct a U-net network model and pre-train the U-net network model using the expanded data set; S600. Fine-tune the pre-trained U-net network model using the data pairs of magnetic field gradient distribution - system matrix to obtain a fine-tuned U-net network model; S700. Input the magnetic field gradient value and the theoretical system matrix at any position to be measured into the fine-tuned U-net network model to obtain the system matrix at the position to be measured, and generate a reconstructed image based on the system matrix.

2. The magnetic particle imaging reconstruction method for generating a system matrix based on a calibration magnetic field according to claim 1, wherein The method for the magnetic field intensity calibration is as follows: S110. Calibrate the divergent magnetic field of the handheld permanent magnet, grid the uniform magnetic field of the fixed-end permanent magnet, and calibrate the uniform magnetic field intensity at each grid point; S120. Based on the uniform magnetic field intensity, obtain the magnetic field gradient values at each grid through fitting calculation; S130. Interpolate the magnetic field gradient distribution and define the gradient values at each position according to the position of each grid point; S140. Calibrate the excitation magnetic field intensity of the handheld excitation coil at the grid; S150. Combine the gradient values and the excitation magnetic field intensity to determine the overall magnetic field gradient distribution.

3. The magnetic particle imaging reconstruction method for generating a system matrix based on a calibration magnetic field according to claim 1, wherein The method for obtaining the theoretical system matrix is as follows: Express the two-dimensional system matrix using the tensor product of the second-kind Chebyshev polynomials as follows: ； Among them, k , l represent the order of the Chebyshev polynomial, x , y represent FOV the two-dimensional coordinates under; Different orders correspond to the frequency components in the actual system matrix; The coordinates correspond to the positions in the actual system matrix, that is FOV the number of calibration points of the system matrix under. Let x and y the direction excitation frequency satisfy , then: ； ； Among them, represents the system matrix at the frequency of with coordinates value, represents the frequency i corresponding to the entire system matrix diagram; f x 、f y are respectively x , y the excitation frequencies in the directions of is the density parameter of the Lissajous trajectory during scanning.

4. The magnetic particle imaging reconstruction method for generating a system matrix based on a calibration magnetic field according to claim 2, wherein The method for constructing the data set is as follows: S310. Select the magnetic field parameters of the corresponding grids in the required imaging area according to the position of a single imaging, and obtain the system matrix at this position based on this magnetic field distribution through actual measurement; S320. Randomly select the positions of multiple field-free points, measure the system matrices under different magnetic field distributions respectively, and obtain the actual system matrix ; Among them, the mid-frequency i corresponding system matrix diagram is ; S330. Combine the actual system matrix with the magnetic field gradient values at the corresponding positions in the overall magnetic field gradient distribution to form data pairs of magnetic field gradient distribution - system matrix, and then construct a data set.

5. The magnetic particle imaging reconstruction method for generating a system matrix based on a calibration magnetic field according to claim 4, wherein The method for expanding the data set is as follows: S410. Based on Langevin the model to simulate the response signal of magnetic nanoparticles and generate a simulation system matrix corresponding to the actual system matrix ; S420. Subtract the simulation system matrix from the actual system matrix to obtain the actual noise distribution ; S430. Set the gradient distribution of the simulated magnetic field to the magnetic field gradient distribution randomly sampled from the overall magnetic field gradient distribution, and simulate again to obtain the system matrix and add , to obtain the simulated overall magnetic field system matrix ; Among them, the system matrix diagram corresponding to the intermediate frequency i is ; S440. Combine the overall magnetic field gradient distribution with the simulated overall magnetic field system matrix to obtain data pairs of magnetic field gradient distribution - simulated system matrix, and complete the expansion of the data set.

6. The magnetic particle imaging reconstruction method for generating a system matrix based on a calibration magnetic field according to claim 5, wherein The magnetic field distribution setting of the simulation system is the same as that of the actual measurement.

7. The magnetic particle imaging reconstruction method for generating a system matrix based on a calibration magnetic field according to claim 1, characterized in that, The U-net network model includes an encoder and a decoder. When the U-net network model outputs, feature splicing is performed, and the features output from different layers in the encoder are fused with the corresponding layers in the decoder.

8. The magnetic particle imaging reconstruction method for generating a system matrix based on a calibration magnetic field according to any one of claims 5-6, characterized in that When pre-training the U-net network model using the expanded data set, first preprocess the input data. The method is as follows: For each layer of the normalize it, convert it into a single-channel image, and scale the image to the specified size; ​ Add a theoretical representation of the Chebyshev polynomial of the system matrix with dimension corresponding to frequency i to the overall magnetic field gradient distribution , then perform normalization, convert it into an RGB image with three channels, and scale the image to the specified size and specifications.

9. The magnetic particle imaging reconstruction method for generating a system matrix based on a calibration magnetic field according to claim 8, characterized in that The said and are both normalized to the range of [0, 255].

10. The magnetic particle imaging reconstruction method for generating a system matrix based on a calibration magnetic field according to claim 9, wherein The output of the U-net network model is a single-channel system matrix map of a certain layer or a multi-channel multi-layer system matrix map.

Images