Data calibration method and device for system matrix

Data calibration of system matrix is ​​performed through pre-trained neural networks, and features are extracted by convolution processing and residual intensive network blocks, the problems of time-consuming and low quality of system matrix calibration in the prior art are solved, achieving high efficiency and low loss high resolution image reconstruction.

CN119048354APending Publication Date: 2024-11-29INST OF AUTOMATION CHINESE ACAD OF SCI
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411050278.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-01
Publication Date
2024-11-29

AI Technical Summary

Technical Problem

In the prior art, the calibration process of the system matrix takes a long time, resulting in low SM quality and large loss of reconstruction images. In particular, the compression perception method cannot effectively process position information.

Method used

The pre-trained neural network is used to calibrate the data of the system matrix, including digital coding units, feature extraction and nonlinear mapping units and upsampling reconstruction units. The features are extracted through convolution processing and residual intensive network blocks are extracted to achieve high spatial resolution Reconstruction of images.

Benefits of technology

It improves the calibration efficiency and quality of the system matrix, reduces the loss rate of the reconstructed image, can effectively process low-resolution images, and output high-resolution images.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119048354A_ABST
    Figure CN119048354A_ABST
Patent Text Reader

Abstract

The invention discloses a system matrix data calibration method and device, and relates to the technical field of system matrix data calibration, and the method comprises the steps: obtaining a target image corresponding to a target system matrix SM, inputting the target image corresponding to the target SM into a target SM calibration network, and obtaining a calibration result of the target image corresponding to the target SM, the target SM calibration network is a pre-trained neural network, and the target SM calibration network comprises a digital coding unit, a feature extraction and nonlinear mapping unit and an up-sampling reconstruction unit. According to the method, the low-resolution target image corresponding to the target SM can be calibrated to obtain the high-resolution target image corresponding to the target SM, the calibration efficiency of the target SM is high, the calibration quality is good, and finally the loss rate of the image reconstructed by the target SM is reduced.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of data calibration of a system matrix, and in particular to a method and a device for data calibration of a system matrix. Background Art

[0002] The system matrix (SM), also known as the projection matrix, is a key parameter in the image reconstruction process. It describes the relationship between the signals received by each detector in the imaging system and each voxel in the image. SM is widely used in magnetic particle imaging (MPI).

[0003] In related technologies, in order to reconstruct high-resolution MPI images, the system response of all grid points within the imaging field of view is generally obtained by directly measuring the SM, thereby constructing a high-resolution system matrix. However, this method requires a lot of time. In order to overcome the time-consuming problem, the SM is calibrated using the compressed sensing method, but the compressed sensing method cannot process the position information of the SM, resulting in low quality of the calibrated SM and large loss of the reconstructed image. Summary of the invention

[0004] In view of this, the present invention provides a method and device for calibrating data of a system matrix to solve the problem that the SM obtained by calibration in the related art has low quality and large loss of the reconstructed image.

[0005] According to a first aspect, an embodiment of the present disclosure provides a data calibration method for a system matrix, the method comprising:

[0006] Obtain the target image corresponding to the target system matrix SM;

[0007] Input the target image corresponding to the target SM into the target SM calibration network to obtain the calibration result of the target image corresponding to the target SM. The target SM calibration network is a pre-trained neural network. The target SM calibration network includes: a digital encoding unit, a feature extraction and nonlinear mapping unit, and an upsampling reconstruction unit;

[0008] Among them, the digital encoding unit is used to encode the position information of the target image corresponding to the target SM, the feature extraction and non-linear mapping unit is used to perform convolution processing on the target image corresponding to the target SM through N standard convolution kernels to obtain N first feature maps, and perform feature extraction and non-linear mapping on the N first feature maps through stacked residual dense network blocks RDB to obtain a second feature map, wherein the RDB is composed of a first dense block, a second dense block, and a third dense block connected in a residual connection manner, and the upsampling reconstruction unit is used to upsample the feature extraction information corresponding to the second feature map to obtain a third feature map, and then perform convolution filtering processing on the third feature map through a convolution filter to obtain a high spatial resolution image.

[0009] In an optional embodiment, the first dense block, the second dense block and the third dense block are each composed of five convolutional layers connected in sequence, wherein the first four convolutional layers are followed by a ReLU activation function layer; the input of the first dense block is the first feature map, the input of the second dense block is the output of the first dense block, and the input of the third dense block is the output of the second dense block.

[0010] In an optional implementation, the upsampling unit is composed of a nearest neighbor interpolation upsampling layer, a convolution layer, and a Relu activation function layer connected in sequence; and the convolution filter is composed of a convolution layer and a Relu activation function layer connected in sequence.

[0011] In an optional implementation, before acquiring the target image corresponding to the target system matrix SM, the method further includes:

[0012] Obtain a target image training set corresponding to the target SM for training the target SM calibration network.

[0013] In an optional implementation, obtaining a target image training set corresponding to the training target SM further includes:

[0014] Acquire a first original SM data set and a second original SM data set;

[0015] Preliminarily performing baseband removal, noise data removal, and RGB conversion processing on the first original SM data set and the second original SM data set, respectively, to obtain a preprocessed first resolution image and a second resolution image;

[0016] The first resolution image and the second resolution image are used as a target image training set corresponding to the target SM, wherein the second resolution image is a label image of the first resolution image.

[0017] In an optional implementation, inputting a target image corresponding to the target SM into a target SM calibration network to obtain a calibration result of the target image of the target SM includes:

[0018] Get the output result of the target SM calibration network;

[0019] The loss value between the output result and the true value label of the target image corresponding to the target SM is calculated, and the network parameters of the target SM calibration network are updated, and the target SM calibration network is trained cyclically until the preset number of training times is reached to end the training, so as to obtain a trained target SM calibration network.

[0020] In an optional embodiment, the target SM is a SM in magnetic particle imaging MPI.

[0021] According to a second aspect, an embodiment of the present disclosure provides a data calibration device for a system matrix, the device comprising:

[0022] SM data acquisition module, used to acquire the target image corresponding to the target system matrix SM;

[0023] The SM data calibration module is used to input the target image corresponding to the target SM into the target SM calibration network to obtain the calibration result of the target image corresponding to the target SM. The target SM calibration network is a pre-trained neural network. The target SM calibration network includes: a digital encoding unit, a feature extraction and nonlinear mapping unit, and an upsampling reconstruction unit;

[0024] Among them, the digital encoding unit is used to encode the position information of the target image corresponding to the target SM, the feature extraction and non-linear mapping unit is used to perform convolution processing on the target image corresponding to the target SM through N standard convolution kernels to obtain N first feature maps, and perform feature extraction and non-linear mapping on the residual dense network block RDB stacked with the N first feature maps to obtain a second feature map, wherein the RDB is composed of a first dense block, a second dense block, a third dense block and connected in a residual connection manner, and the upsampling reconstruction unit is used to upsample the feature extraction information corresponding to the second feature map to obtain a third feature map, and then perform convolution filtering processing on the third feature map through a convolution filter to obtain a high spatial resolution image.

[0025] According to a third aspect, an embodiment of the present disclosure provides a computer device, including:

[0026] The memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the data calibration method of the system matrix in the first aspect or any embodiment of the first aspect by executing the computer instructions.

[0027] According to a fourth aspect, an embodiment of the present disclosure provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the data calibration method for the system matrix in the first aspect or any one of the embodiments of the first aspect.

[0028] According to a fifth aspect, an embodiment of the present disclosure provides a computer program product, including computer instructions, which are used to enable a computer to execute the data calibration method of the system matrix in the first aspect or any implementation of the first aspect.

[0029] The technical solution of the present invention has the following advantages:

[0030] The present invention discloses a method and device for data calibration of a system matrix, and relates to the technical field of data calibration of a system matrix, wherein the method, after obtaining a target image corresponding to a target system matrix SM, inputs the target image corresponding to the target SM into a target SM calibration network to obtain a calibration result of the target image corresponding to the target SM, wherein the target SM calibration network is a pre-trained neural network, and the target SM calibration network includes: a digital encoding unit, a feature extraction and nonlinear mapping unit, and an upsampling reconstruction unit. The present invention can calibrate a low-resolution target image corresponding to the target SM to obtain a high-resolution target image corresponding to the target SM, and the calibration efficiency of the target SM is high and the calibration quality is good, and ultimately it is also beneficial to reduce the loss rate of the image reconstructed by the target SM. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0032] Figure 1 is a flow chart of a method for calibrating data of a system matrix according to an embodiment of the present invention;

[0033] Figure 2 is a low-resolution image corresponding to the SM and a high-resolution image corresponding to the SM according to an embodiment of the present invention;

[0034] Figure 3 is a schematic diagram of a target SM calibration network according to an embodiment of the present invention;

[0035] Figure 4 is a flow chart of another method for calibrating data of a system matrix according to an embodiment of the present invention;

[0036] Figure 5 is a flow chart of a data calibration method for another system matrix according to an embodiment of the present invention;

[0037] Figure 6 is a schematic diagram of a process for updating a target SM calibration network according to an embodiment of the present invention;

[0038] Figure 7 is a structural block diagram of a data calibration device for a system matrix according to an embodiment of the present invention;

[0039] Figure 8 It is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0040] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0041] According to an embodiment of the present invention, an embodiment of a data calibration method for a system matrix is ​​provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0042] In this embodiment, a data calibration method for a system matrix is ​​provided, which can be used in mobile terminals, such as mobile phones, tablet computers, etc. Figure 1 is a flow chart of a method for calibrating data of a system matrix according to an embodiment of the present invention. Figure 1 As shown, the process includes the following steps:

[0043] Step S101, obtaining a target image corresponding to a target system matrix SM.

[0044] Specifically, the target system matrix SM can be the system matrix data calibration method provided in the embodiment of the present disclosure applied to any SM detected by the MPI imaging device, and further convert the SM into a corresponding target image. Therefore, the target image corresponding to the target SM is the image converted from any SM.

[0045] The target image corresponding to the target system matrix SM can be obtained by RGB conversion. For example, first, the HSV color model is used to represent the phase That is, adopt Completed from Second, use the standard HSV to RGB transformation to convert the HSV color vector Convert to Finally, for the amplitude Encoding, linearly scaling the RGB color vector by magnitude to get the final Final Represents the target image corresponding to the target SM. Therefore, the above process is equivalent to converting the target SM from the complex domain to the RGB domain.

[0046] As a preferred example, the target SM is a SM in a magnetic particle imaging apparatus MPI.

[0047] Specifically, the application of SM in MPI is a breakthrough medical imaging technology. It visualizes the spatial concentration distribution and dynamic changes of magnetic nanoparticle tracers in the human body based on the response signal received by the external coil. It has the advantages of no imaging depth limitation, high sensitivity, linear quantification, no background signal interference, and no ionizing radiation hazards. It has gradually become the research frontier of the new generation of in vivo tomography and has broad clinical application prospects. In order to ensure the imaging quality and resolution of MPI, it is very important to calibrate the system matrix before SM performs MPI imaging. Therefore, the MPI image is calibrated by the method in the embodiment of the present disclosure, and then a high-resolution MPI image is obtained, which is conducive to the in-depth study of MPI.

[0048] Step S102, inputting the target image corresponding to the target SM into the target SM calibration network to obtain the calibration result of the target image corresponding to the target SM, wherein the target SM calibration network is a pre-trained neural network, and the target SM calibration network includes: a digital encoding unit, a feature extraction and nonlinear mapping unit, and an upsampling reconstruction unit;

[0049] Among them, the digital encoding unit is used to encode the position information of the target image corresponding to the target SM, the feature extraction and non-linear mapping unit is used to perform convolution processing on the target image corresponding to the target SM through N standard convolution kernels to obtain N first feature maps, and perform feature extraction and non-linear mapping on the N first feature maps through a stacked residual dense network block RDB to obtain a second feature map, wherein the RDB is composed of a first dense block, a second dense block, a third dense block and a residual connection method, and the upsampling reconstruction unit is used to upsample the feature extraction information corresponding to the second feature map to obtain a third feature map, and then perform convolution filtering processing on the third feature map through a convolution filter to obtain a high spatial resolution image.

[0050] Specifically, the target SM calibration network is a trained neural network. The target SM can obtain a corresponding target image after RGB conversion. The target image corresponding to the target SM can be a low-resolution image. The low-resolution image is input into the trained target SM calibration network for calibration to obtain a high-resolution target image. Figure 2 As shown, Figure 2 2a in is the low-resolution SM before inputting the target SM calibration network. Figure 2 2b is the target image corresponding to the high-resolution target SM output after the target SM calibration network is trained. Therefore, the target SM calibration network in the disclosed embodiment can calibrate the low-resolution target image corresponding to the target SM to obtain the high-resolution target image corresponding to the target SM, and the calibration efficiency of the target SM is high and the calibration quality is good, which is ultimately beneficial to reducing the loss rate of the image reconstructed by the target SM.

[0051] The target SM calibration network in the embodiment of the present disclosure includes: a digital encoding unit, a feature extraction and nonlinear mapping unit, and an upsampling reconstruction unit.

[0052] The digital encoding unit is used to encode the position information of the target image corresponding to the target SM. For example, the relative position information of the target image corresponding to the target SM is encoded using a trained target neural network to guide the unique position attribute information of the target SM, which is conducive to the efficient calibration of the image reconstructed by the target SM. The position encoding unit encodes the relative position information of the target image corresponding to the target SM input into the target SM calibration network and inputs it into the multi-dimensional channels of the target SM calibration network respectively.

[0053] Among them, the feature extraction and non-linear mapping unit is configured to perform convolution processing on the target image corresponding to the input target SM through N (preferably 64 in the embodiment of the present disclosure) standard convolution kernels to obtain N feature maps as the first feature map; each first feature map further completes feature extraction and non-linear mapping through a stacked RDB to obtain a second feature map; the stacked RDB is composed of dense blocks and residual connections. In the embodiment of the present disclosure, the RDB is preferably based on three dense blocks connected in sequence, which serve as the first dense block, the second dense block, and the third dense block respectively.

[0054] In an optional embodiment, if Figure 3 As shown in the figure, the first dense block, the second dense block and the third dense block are composed of five convolutional layers connected in sequence, where the first four convolutional layers are followed by a ReLU activation function layer; the input of the first dense block is the first feature map, the input of the second dense block is the output of the first dense block, and the input of the third dense block is the output of the second dense block.

[0055] In an optional implementation, the upsampling unit is composed of a nearest neighbor interpolation upsampling layer, a convolution layer, and a Relu activation function layer connected in sequence; and the convolution filter is composed of a convolution layer and a Relu activation function layer connected in sequence.

[0056] In this embodiment, a data calibration method for a system matrix is ​​provided, which can be used in mobile terminals, such as mobile phones, tablet computers, etc. Figure 4is a flow chart of a method for calibrating data of a system matrix according to an embodiment of the present invention. Figure 4 As shown, before the above step S1001, it also includes:

[0057] Step S100, obtaining a target image training set corresponding to a target SM for training a target SM calibration network.

[0058] Specifically, in the embodiment of the present disclosure, the target SM used for training the target SM calibration network is preferably from an OpenMPI dataset, and the target SM contains 37×37×37 voxels.

[0059] The disclosed embodiment obtains a target image training set corresponding to the target SM for training, in order to train the target SM calibration network, thereby achieving efficient calibration of any target SM and reducing the loss of the target SM in reconstructing the target image, thereby ultimately ensuring the quality of the target SM.

[0060] In this embodiment, a data calibration method for a system matrix is ​​provided, which can be used in mobile terminals, such as mobile phones, tablet computers, etc. Figure 5 is a flow chart of a method for calibrating data of a system matrix according to an embodiment of the present invention. Figure 5 As shown, step S100, a target image training set corresponding to a target SM for training a target SM calibration network is obtained. The process includes the following steps:

[0061] Step S1001: Acquire a first original SM data set and a second original SM data set.

[0062] Specifically, the original SM dataset comes from the OpenMPI dataset, which contains 37×37×37 voxels. Among them, the first original SM dataset can be a low-resolution original SM dataset, and the second original SM dataset can be a high-resolution original SM dataset. Among them, each SM in the second original SM dataset can be sampled at equal intervals according to a single-dimensional sampling factor of 2, 3, and 4 to obtain each SM in the first original SM dataset, that is, a low-resolution SM. The embodiment of the present disclosure finally forms a pair of low-resolution SM and high-resolution SM in this way, wherein the high-resolution SM corresponding to the low-resolution SM is used as the true value label.

[0063] Step S1002 , preliminarily remove the fundamental frequency, remove the noise data and perform RGB conversion processing on the first original SM data set and the second original SM data set, respectively, to obtain a preprocessed first resolution image and a preprocessed second resolution image.

[0064] Exemplarily, data with a frequency less than 8 kHz in the SM obtained from the OpenMPI data set is removed, that is, the fundamental frequency data is removed; then data with an SNR less than 3 in the SM is removed, that is, the noise data is removed; finally, the SM is converted from the complex domain to the RGB domain.

[0065] Step S1003: Use the first resolution image and the second resolution image as a target image training set corresponding to the target SM, wherein the second resolution image is a label image of the first resolution image.

[0066] Specifically, the first resolution image may be a low-resolution image, and the second resolution image may be a high-resolution image. Each low-resolution target image corresponding to the target SM corresponds to a high-resolution target image corresponding to the target SM, and the high-resolution target image corresponding to the target SM is used as a true value label to match the low-resolution target image corresponding to the target SM. For example, Figure 2 As shown, Figure 2 a in is a low-resolution image. Figure 2 b in the figure is a high-resolution image.

[0067] In an optional implementation, inputting a target image corresponding to the target SM into a target SM calibration network to obtain a calibration result corresponding to the target SM includes:

[0068] Step a1, obtaining the output result of the target SM calibration network.

[0069] Step a2, calculate the loss value between the output result and the true value label of the target image corresponding to the target SM, update the network parameters of the target SM calibration network, and cyclically train the target SM calibration network until the preset training times are reached to obtain a trained target SM calibration network.

[0070] Specifically, in the disclosed embodiment, it is preferred to use the normalized root mean square error (NRMSE) as an indicator to calculate the loss value of the model, perform back propagation, adjust the parameters, and update the parameters of the target SM calibration network. The optimization algorithm used is the adaptive moment estimation method (ADAM).

[0071] like Figure 6 As shown in FIG. 1 , a flow chart of updating the target SM calibration network is shown in FIG. Figure 6 In the process, step S602 to step S610 are repeated until a trained target SM calibration network is obtained.

[0072] In the present embodiment, a data calibration device for a system matrix is ​​also provided, which is used to implement the above-mentioned embodiments and preferred implementation modes, and the descriptions thereof will not be repeated. As used below, the term "module" can implement a combination of software and / or hardware of a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and contemplated.

[0073] This embodiment provides a data calibration device for a system matrix, such as Figure 7 As shown, including:

[0074] SM data acquisition module 71, used to acquire the target image corresponding to the target system matrix SM;

[0075] The SM data calibration module 72 is used to input the target image corresponding to the target SM into the target SM calibration network to obtain the calibration result of the target image corresponding to the target SM. The target SM calibration network is a pre-trained neural network. The target SM calibration network includes: a digital encoding unit, a feature extraction and nonlinear mapping unit, and an upsampling reconstruction unit;

[0076] Among them, the digital encoding unit is used to encode the position information of the target image corresponding to the target SM, the feature extraction and non-linear mapping unit is used to perform convolution processing on the target image corresponding to the target SM through N standard convolution kernels to obtain N first feature maps, and perform feature extraction and non-linear mapping on the residual dense network block RDB stacked with the N first feature maps to obtain a second feature map, wherein the RDB is composed of a first dense block, a second dense block, a third dense block and connected in a residual connection manner, and the upsampling reconstruction unit is used to upsample the feature extraction information corresponding to the second feature map to obtain a third feature map, and then perform convolution filtering processing on the third feature map through a convolution filter to obtain a high spatial resolution image.

[0077] In an optional implementation, the upsampling unit is composed of a nearest neighbor interpolation upsampling layer, a convolution layer, and a Relu activation function layer connected in sequence; and the convolution filter is composed of a convolution layer and a Relu activation function layer connected in sequence.

[0078] In an optional implementation manner, before the SM data acquisition module, it also includes:

[0079] The original data acquisition module is used to obtain a target image training set corresponding to the target SM for training the target SM calibration network.

[0080] In an optional implementation, the original data acquisition module includes:

[0081] An original data acquisition submodule, used to acquire a first original SM data set and a second original SM data set;

[0082] A data preprocessing submodule is used to perform baseband removal, noise data removal and RGB conversion processing on the first original SM data set and the second original SM data set in advance, respectively, to obtain a preprocessed first resolution image and a second resolution image;

[0083] The target data determination submodule is used to use the first resolution image and the second resolution image as a target image training set corresponding to the target SM, wherein the second resolution image is a label image of the first resolution image.

[0084] In an optional implementation, the SM data calibration module 72 includes:

[0085] A data acquisition submodule is used to obtain the output result of the target SM calibration network;

[0086] The data update submodule is used to output the loss value between the result and the true value label of the target image corresponding to the calculated target SM, update the network parameters of the target SM calibration network, and cyclically train the target SM calibration network until the preset number of training times is reached to terminate the training, so as to obtain a trained target SM calibration network.

[0087] In an optional embodiment, the target SM is a SM in magnetic particle imaging MPI.

[0088] The further functional description of each of the above modules and units is the same as that of the above corresponding embodiments and will not be repeated here.

[0089] The data calibration device of the system matrix in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.

[0090] An embodiment of the present invention further provides a computer device having the above-mentioned data calibration device for the system matrix.

[0091] See also Figure 8 , Figure 8 is a schematic diagram of the structure of a computer device provided by an optional embodiment of the present invention, such as Figure 8As shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components are connected to each other using different buses for communication, and can be installed on a common mainboard or installed in other ways as needed. The processor can process the instructions executed in the computer device, including instructions stored in or on the memory to display the graphical information of the GUI on an external input / output device (such as, a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple computer devices can be connected, and each device provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 8 A processor 10 is taken as an example.

[0092] The processor 10 may be a central processing unit, a network processor or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be a dedicated integrated circuit, a programmable logic device or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic or any combination thereof.

[0093] The memory 20 stores instructions executable by at least one processor 10, so that the at least one processor 10 executes the method shown in the above embodiment.

[0094] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely arranged relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0095] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid state drive; the memory 20 may also include a combination of the above types of memory.

[0096] The computer device also includes an input device 30 and an output device 40. The processor 10, the memory 20, the input device 30 and the output device 40 may be connected via a bus or other means. Figure 8 The example of connecting through bus is taken in the following.

[0097] The input device 30 can receive input digital or character information, and generate key signal input related to the user settings and function control of the computer device, such as a touch screen, a keypad, a mouse, a track pad, a touch pad, an indicator bar, one or more mouse buttons, a trackball, a joystick, etc. The output device 40 may include a display device, an auxiliary lighting device (e.g., an LED) and a tactile feedback device (e.g., a vibration motor), etc. The above-mentioned display device includes but is not limited to a liquid crystal display, a light emitting diode, a display and a plasma display. In some optional embodiments, the display device can be a touch screen.

[0098] The embodiment of the present invention also provides a computer-readable storage medium. The method according to the embodiment of the present invention can be implemented in hardware, firmware, or can be implemented as a computer code that can be recorded in a storage medium, or can be implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and will be stored in a local storage medium through a network download, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state hard disk, etc.; further, the storage medium can also include a combination of the above types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor, or hardware, the method shown in the above embodiment is implemented.

[0099] A part of the present invention may be applied as a computer program product, such as a computer program instruction, which, when executed by a computer, can call or provide the method and / or technical solution according to the present invention through the operation of the computer. Those skilled in the art should understand that the existence of the computer program instruction in a computer-readable medium includes, but is not limited to, a source file, an executable file, an installation package file, etc., and accordingly, the way in which the computer program instruction is executed by the computer includes, but is not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. Here, the computer-readable medium may be any available computer-readable storage medium or communication medium accessible to the computer.

[0100] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A data calibration method for a system matrix, It is characterized in that The method comprises: Obtain the target image corresponding to the target system matrix SM; Input the target image corresponding to the target SM into the target SM calibration network to obtain the calibration result of the target image corresponding to the target SM. The target SM calibration network is a pre-trained neural network. The target SM calibration network includes: a digital encoding unit, a feature extraction and nonlinear mapping unit, and an upsampling reconstruction unit; Among them, the digital encoding unit is used to encode the position information of the target image corresponding to the target SM, the feature extraction and non-linear mapping unit is used to perform convolution processing on the target image corresponding to the target SM through N standard convolution kernels to obtain N first feature maps, and perform feature extraction and non-linear mapping on the N first feature maps through stacked residual dense network blocks RDB to obtain a second feature map, wherein the RDB is composed of a first dense block, a second dense block, and a third dense block connected in a residual connection manner, and the upsampling and reconstruction unit is used to upsample the feature extraction information corresponding to the second feature map to obtain a third feature map, and then perform convolution filtering processing on the third feature map through a convolution filter to obtain a high spatial resolution image.

2. The data calibration method of the system matrix according to claim 1, It is characterized in that The first dense block, the second dense block and the third dense block are each composed of five convolutional layers connected in sequence, wherein the first four convolutional layers are followed by a ReLU activation function layer; the input of the first dense block is the first feature map, the input of the second dense block is the output of the first dense block, and the input of the third dense block is the output of the second dense block.

3. The data calibration method of the system matrix according to claim 1, It is characterized in that The upsampling unit is composed of a nearest neighbor interpolation upsampling layer, a convolution layer and a Relu activation function layer connected in sequence; the convolution filter is composed of a convolution layer and a Relu activation function layer connected in sequence.

4. The data calibration method of the system matrix according to claim 1, It is characterized in that Before acquiring the target image corresponding to the target system matrix SM, the method further includes: A target image training set corresponding to the target SM for training the target SM calibration network is obtained.

5. The data calibration method of the system matrix according to claim 4, It is characterized in that The step of obtaining a target image training set corresponding to the target SM for training the target SM calibration network further includes: Acquire a first original SM data set and a second original SM data set; Preliminarily performing baseband removal, noise data removal, and RGB conversion processing on the first original SM data set and the second original SM data set, respectively, to obtain a preprocessed first resolution image and a second resolution image; The first resolution image and the second resolution image are used as target image training sets corresponding to the target SM, wherein the second resolution image is a label image of the first resolution image.

6. The data calibration method of the system matrix according to claim 1, It is characterized in that Inputting the target image corresponding to the target SM into the target SM calibration network to obtain the calibration result of the target image of the target SM, including: Get the output result of the target SM calibration network; Calculate the loss value between the output result and the true value label of the target image corresponding to the target SM, update the network parameters of the target SM calibration network, and train the target SM calibration network in a loop until the preset number of training times is reached to terminate the training, so as to obtain the trained target SM calibration network.

7. The data calibration method of the system matrix according to any one of claims 1 to 6, It is characterized in that The target SM is the SM in the magnetic particle imaging device MPI.

8. A data calibration device for a system matrix, It is characterized in that The device comprises: SM data acquisition module, used to acquire the target image corresponding to the target system matrix SM; The SM data calibration module is used to input the target image corresponding to the target SM into the target SM calibration network to obtain the calibration result of the target image corresponding to the target SM. The target SM calibration network is a pre-trained neural network. The target SM calibration network includes: a digital encoding unit, a feature extraction and nonlinear mapping unit, and an upsampling reconstruction unit; Among them, the digital encoding unit is used to encode the position information of the target image corresponding to the target SM, the feature extraction and non-linear mapping unit is used to perform convolution processing on the target image corresponding to the target SM through N standard convolution kernels to obtain N first feature maps, and perform feature extraction and non-linear mapping on the N first feature maps through a stacked residual dense network block RDB to obtain a second feature map, wherein the RDB is composed of a first dense block, a second dense block, a third dense block and connected in a residual connection manner, and the upsampling reconstruction unit is used to upsample the feature extraction information corresponding to the second feature map to obtain a third feature map, and then perform convolution filtering processing on the third feature map through a convolution filter to obtain a high spatial resolution image.

9. A computer device, It is characterized in that include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the data calibration method of the system matrix according to any one of claims 1 to 7 by executing the computer instructions.

10. A computer-readable storage medium, It is characterized in that The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the data calibration method for the system matrix according to any one of claims 1 to 7.

11. A computer program product, It is characterized in that The method comprises computer instructions for causing a computer to execute the data calibration method of the system matrix according to any one of claims 1 to 7.