Multi-channel electron paramagnetic resonance imaging method, system, equipment and medium

Abstract

The invention relates to the technical field of medical image reconstruction, in particular to a multi-channel electron paramagnetic resonance imaging method, system and device and a medium, and the method comprises the steps: constructing a data separation constraint-based dual-weighted nuclear TV optimization model for a multi-channel electron paramagnetic resonance image; wherein the dual-weighted nuclear TV norm of a set voxel in the multichannel electron paramagnetic resonance image is equal to the sum of self-adaptive weighted singular values corresponding to a Jacobian matrix weighted by the dual-weighted nuclear TV norm, and in the reconstruction process of the multichannel electron paramagnetic resonance image based on the nuclear TV, solving a dual-weighted nuclear TV optimization model by using a Chambbel-Pock algorithm. An obtained reconstruction result is used as a reconstructed electron paramagnetic resonance image. According to the method, the structural similarity and low-rank prior information of electron paramagnetic resonance images of different channels can be combined, local information of the images is smoothed adaptively, the edge protection capability of an algorithm is enhanced, and the reconstruction efficiency and quality of the multi-channel electron paramagnetic resonance images are improved.

Description

Technical Field

[0001] The invention relates to the technical field of medical image reconstruction, and in particular to a multi-channel electron paramagnetic resonance imaging method, system, equipment and medium. Background Art

[0002] The oxygen concentration in human tissues is of great significance to human life, health, and disease prevention. Studies have found that tumor cells induce hypoxia through various mechanisms, such as high metabolic rate and high oxygen consumption, thereby creating a chronic hypoxic environment, activating the hypoxia-inducible factor signaling pathway, and causing tumor growth to accelerate and become more invasive. Therefore, low-oxygen or hypoxic areas are common inside tumors. These areas are less sensitive to radiation, which reduces the efficiency of radiotherapy. Efficient radiotherapy requires high-dose radiation irradiation in low-oxygen areas and low-dose radiation irradiation in high-oxygen areas. Electron Paramagnetic Resonance Imaging (EPRI), as an advanced oxygen imaging technology, can effectively measure the distribution of oxygen concentration inside tumors. EPRI images can guide precise radiotherapy, which is of great significance to the treatment efficiency of cancer patients.

[0003] Electron Paramagnetic Resonance Imaging (EPRI), as a mainstream tumor oxygen imaging method, can reconstruct the oxygen concentration image inside the tumor with high spatial resolution and high density resolution. However, the main disadvantage and technical bottleneck of EPRI is its slow scanning speed. In order to speed up the scanning speed, an effective way is to reconstruct EPRI images in sparse projection data. However, the EPRI images reconstructed by traditional analytical methods using sparse projection data have obvious artifacts, resulting in reduced image availability. Therefore, this technology aims to study an optimization method for high-precision sparse reconstruction of EPRI.

[0004] The total variation (TV) minimization algorithm is a classic EPRI sparse reconstruction method based on compressed sensing (CS), which can reconstruct images with high accuracy from sparsely sampled projection data. However, in the reconstruction task of multi-channel EPRI, traditional EPRI imaging methods including filtered back-projection (FBP) algorithm and TV-related optimization algorithms only consider the characteristics of a single channel image, ignoring the similarities between the images of each channel.

[0005] The kernel TV algorithm considers the structural similarity and low-rank prior information between multiple T1-weighted images to achieve multi-channel EPRI sparse reconstruction. However, in the kernel TV model based on data separation constraints, there is an imbalance between the data fidelity term and the kernel TV regularization term, resulting in a low convergence rate or iterative divergence. In addition, the kernel TV algorithm fails to adaptively consider information in different gradient directions. At the same time, different singular values ​​in the kernel TV norm contain different amounts of image information in the image, and no adaptive adjustment is made based on the contribution of different singular values. Summary of the invention

[0006] The technical problem to be solved by the present invention is to provide a multi-channel electron paramagnetic resonance imaging method, system, device and medium to address the technical problem of inaccurate EPRI image reconstruction by the current TV algorithm in view of the deficiencies in the above-mentioned prior art.

[0007] The objective of the present invention is achieved by the following technical solutions: In a first aspect, the present invention provides a multi-channel electron paramagnetic resonance imaging method, comprising: For multi-channel electron paramagnetic resonance images, a doubly weighted kernel TV optimization model based on data separation constraints is constructed; the doubly weighted kernel TV optimization model is obtained according to the doubly weighted kernel TV norm of the data separation constraints of the multi-channel electron paramagnetic resonance images; the doubly weighted kernel TV norm of the voxels set in the multi-channel electron paramagnetic resonance images is equal to the sum of the adaptive weighted singular values ​​corresponding to the Jacobian matrix weighted by the doubly weighted kernel TV norm; In the process of reconstructing a multi-channel electron paramagnetic resonance image based on nuclear TV, the Chambolle-Pock algorithm is used to solve the double-weighted nuclear TV optimization model, and the solution obtained is used as the reconstructed electron paramagnetic resonance image. As a further improvement of the present invention, a dual weighted kernel TV optimization model based on data separation constraints specifically includes:

[0008] In the formula, Representing multi-channel images The doubly weighted kernel TV norm of ; Represents the adaptive weight coefficient along the gradient direction; Represents the adaptive weight coefficient of each singular value; is the solution of the optimization model, i.e., the reconstructed electron paramagnetic resonance image; is a non-zero balance factor; Represents the difference between the actual measured projection data and the optimized model norm, Indicates the data tolerance.

[0009] As a further improvement of the present invention, the doubly weighted kernel TV norm of a voxel in a multi-channel electron paramagnetic resonance image is set equal to the sum of the adaptively weighted singular values ​​corresponding to its weighted Jacobian matrix, specifically including: The Jacobian matrix of each voxel position of the electron paramagnetic resonance image is defined as:

[0010] In the formula, the adaptive weight coefficient They are defined as:

[0011]

[0012]

[0013] In the formula, is the adaptive weight coefficient in the x direction, is the adaptive weight coefficient in the y direction, is the adaptive weight coefficient in the z direction, x, y, z represent the three directions in the spatial coordinate system, D is the gradient transformation matrix, i, j, k represent the position information of the voxel relative to the spatial coordinate system, is a weight coefficient greater than 0, used to control the strength of each iteration.

[0014] As a further improvement of the present invention, the dual weighted kernel TV optimization model is solved based on the Chambolle-Pock algorithm, specifically including: Initializing parameters, including model parameters and algorithm parameters; Introducing a dual variable to transform the double-weighted kernel TV optimization model into a convex conjugate function; The proximal mapping calculation is performed on the convex conjugate function, and the singular values ​​of the image to be reconstructed and the voxel Jacobian matrix are continuously and alternately updated until convergence to obtain the optimal electron paramagnetic resonance image reconstruction result.

[0015] As a further improvement of the present invention, the dual weighted kernel TV optimization model is converted into a convex conjugate function according to the Chambolle-Pock algorithm framework, and the conversion process specifically includes:

[0016]

[0017]

[0018]

[0019] Where x is the original variable, u is the electron paramagnetic resonance image to be reconstructed, A is the imaging system matrix, and the value of the system matrix is ​​obtained according to the three-dimensional ray-driven method. At a specific angle, the ray passes through the three-dimensional discrete image, and the intersection area with the three-dimensional voxel array is the ray weight; p and q are dual variables; To transform the doubly weighted kernel TV optimization model into a function of the dual variable p, To transform the doubly weighted kernel TV optimization model into a function of the dual variable q, is the balancing factor, and the original variable is the electron paramagnetic resonance image to be reconstructed.

[0020] As a further improvement of the present invention, the convex conjugate function is:

[0021]

[0022] In the formula, is the singular value vector of the singular value decomposition of the weighted Jacobian matrix, is a convex conjugate function with respect to the dual variable p, is a convex conjugate function with respect to the dual variable q, Indicates data tolerance, is the L2 norm of the dual variable p.

[0023] As a further improvement of the present invention, the proximal mapping calculation result is:

[0024]

[0025]

[0026] In the formula, For function The proximal mapping of for The proximal mapping of For operator The proximal mapping of yes The singular value decomposition of ; represents the soft threshold scaling of singular values, is a non-zero constant, Indicates the data tolerance.

[0027] In a second aspect, the present invention provides a multi-channel electron paramagnetic resonance imaging system for implementing the above-mentioned multi-channel electron paramagnetic resonance imaging method, comprising: An optimization model construction module; for multi-channel electron paramagnetic resonance images, a double-weighted kernel TV optimization model based on data separation constraints is constructed; the double-weighted kernel TV optimization model is obtained according to the double-weighted kernel TV norm of the multi-channel electron paramagnetic resonance image; the double-weighted kernel TV norm of the voxel set in the multi-channel electron paramagnetic resonance image is equal to the sum of the adaptive weighted singular values ​​corresponding to the Jacobian matrix weighted by the double-weighted kernel TV norm; The optimization model solving module solves the double-weighted nuclear TV optimization model based on the Chambolle-Pock algorithm during the reconstruction of the multi-channel electron paramagnetic resonance image based on the nuclear TV, and the obtained solution is used as the reconstructed electron paramagnetic resonance image.

[0028] In a third aspect, the present invention provides a computer-readable storage medium storing one or more programs, wherein the one or more programs include instructions, which, when executed by a computing device, enable the computing device to perform the above-mentioned multi-channel electron paramagnetic resonance imaging method.

[0029] In a fourth aspect, the present invention provides a computing device, comprising: One or more processors, a memory and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs include steps for executing the above-mentioned multi-channel electron paramagnetic resonance imaging method.

[0030] The beneficial effects of the present invention are as follows: the multi-channel electron paramagnetic resonance imaging method provided by the present invention. By introducing the dual-weighted kernel TV norm to construct an optimization model, the different features of the multi-channel electron paramagnetic resonance image can be considered in a targeted manner. Different weights can be assigned to different structures or regions in the image, so that important image features (such as lesion areas, specific tissue structures, etc.) can be better preserved during the reconstruction process. Compared with traditional reconstruction methods, key features will not be lost due to reasons such as excessive smoothing, thereby improving the accuracy and availability of the reconstructed image for subsequent diagnosis or analysis. The Jacobian matrix is ​​used to describe the gradient information of the electron paramagnetic resonance image, so that the optimization model can more accurately capture the changes in the image in all directions. In the reconstruction process, effective use is made of the edges and details that help to restore the image. The Chambolle-Pock algorithm is used to solve the dual-weighted kernel TV optimization model, and the algorithm has the advantages of fast convergence speed and good stability. Compared with some traditional solution algorithms, a more accurate solution result can be obtained in a shorter time, thereby improving the reconstruction efficiency and quality of multi-channel electron paramagnetic resonance images.

[0031] Furthermore, multi-channel data often contain richer information. Through means such as data separation constraints, we can fully explore and utilize these multi-channel information to avoid mutual interference between different channel data, so that the reconstructed image can comprehensively reflect the advantages of multi-channel data, providing strong support for more comprehensive and in-depth analysis of electron paramagnetic resonance images.

[0032] Furthermore, the addition of data separation constraints helps to further standardize and optimize the reconstruction process. It can reasonably separate and process data of different properties or sources in the reconstruction model, so that each part of the data can play its maximum role under appropriate constraints, thereby improving the quality and accuracy of the entire reconstructed image and better meeting the requirements of practical applications (such as medical diagnosis, material research, etc.) for electron paramagnetic resonance images. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0034] Figure 1 The embodiment of the present invention is a 3D six-ball digital phantom example diagram.

[0035] Figure 2 The embodiment of the present invention is a slice image of 8 channels of a 3D six-ball digital phantom.

[0036] FIG3 is a representative slice image of a 3D six-ball digital phantom reconstructed through reverse crime research in an embodiment of the present invention, (a) is a 3D digital phantom image, (b) is a reconstructed 3D digital phantom image, (c) is an N / 2 transverse slice image, and (d) is an N / 4 transverse slice image.

[0037] Figure 4 is an iterative graph of the 3D six-ball digital model reconstructed for reverse crime research according to an embodiment of the present invention, (a) is the convergence index iterative graph of NOE, representing the standardized object error; (b) is the convergence index iterative graph of NDE, representing the standardized data difference; (c) is the convergence index iterative graph corresponding to NTVE, representing the standardized TV difference.

[0038] Figure 5 The embodiments of the present invention reconstruct representative slice images of electron paramagnetic resonance images by different methods.

[0039] Figure 6The embodiments of the present invention are representative slice images of electron paramagnetic resonance images reconstructed by different methods.

[0040] Figure 7 The embodiments of the present invention reconstruct representative slice images of a 3D digital phantom through different methods.

[0041] Figure 8 It is a schematic diagram of a flow chart in an embodiment of the present invention.

[0042] Fig. 9 It is a schematic diagram of the structure of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0043] In order to make the purpose and technical solution of the present invention clearer and easier to understand, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. The specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0044] The technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings and specific embodiments, wherein the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments.

[0045] Example 1 like Figure 1-Figure 8 As shown, this embodiment provides a multi-channel electron paramagnetic resonance imaging method based on minimization of nuclear total variation. The following is a specific implementation method.

[0046] For multi-channel electron paramagnetic resonance images, a doubly weighted kernel TV optimization model based on data separation constraints is constructed; the doubly weighted kernel TV optimization model is obtained according to the doubly weighted kernel TV norm of the data separation constraints of the multi-channel electron paramagnetic resonance images; the doubly weighted kernel TV norm of the set voxel in the multi-channel electron paramagnetic resonance image is equal to the sum of the adaptive weighted singular values ​​corresponding to the Jacobian matrix weighted by the doubly weighted kernel TV norm.

[0047] The double-weighted kernel TV optimization model based on data separation constraints includes:

[0048] In the formula, Representing multi-channel images The dual-weighted total nuclear variation (TV norm) ); Represents the adaptive weight coefficient along the gradient direction; Represents the adaptive weight coefficient of each singular value; is the solution of the optimization model, i.e., the reconstructed electron paramagnetic resonance image; is a non-zero balancing factor; in inequality constraints, Represents the difference between the actual measured projection data and the optimized model norm, Represents the data tolerance, which is used to describe the noise level and imaging system inconsistencies.

[0049] Among them, the doubly weighted kernel TV norm is defined as:

[0050] Among them, for the dual-weighted nuclear TV imaging model of multi-channel electron paramagnetic resonance imaging EPRI, it is defined that at the specified position The Jacobian matrix of the multi-channel electron paramagnetic resonance image is set to have a double-weighted kernel TV norm of a voxel equal to the sum of the adaptively weighted singular values ​​corresponding to its weighted Jacobian matrix, including:

[0051] Among them, the adaptive weight coefficient Defined as:

[0052]

[0053]

[0054] In the formula, is the adaptive weight coefficient in the x direction, is the adaptive weight coefficient in the y direction, is the adaptive weight coefficient in the z direction, x, y, z represent the three directions in the spatial coordinate system, D is the gradient transformation matrix, i, j, k represent the position information of the voxel relative to the spatial coordinate system, is a weight coefficient greater than 0, which is used to control the intensity of each iteration and is consistent with the adaptive weight coefficient.

[0055] The weighted nuclear norm of a specific voxel of a multi-channel electron paramagnetic resonance image is equal to the sum of the adaptively weighted singular values ​​corresponding to its weighted Jacobian matrix, that is:

[0056] represents the singular value vector of the singular value decomposition of the weighted Jacobian matrix, is the adaptive weight coefficient of each singular value, and the adaptive weight coefficient of each singular value is defined as:

[0057] Through the above analysis, it can be seen that the double weighted kernel TV performs a second adaptive weighting on the singular values ​​of the weighted Jacobian matrix on the basis of adaptively considering the information of different gradient directions, making full use of the contribution of different singular values ​​and the information of different gradient directions to improve the algorithm's ability to protect edges and image details.

[0058] In the process of reconstructing multi-channel electron paramagnetic resonance images based on nuclear TV, the double-weighted nuclear TV optimization model is solved based on the Chambolle-Pock algorithm, and the solution is used as the reconstructed electron paramagnetic resonance image.

[0059] In the multi-channel EPRI image reconstruction algorithm based on kernel TV, the Chambolle-Pock algorithm is used to solve the dual-weighted kernel TV optimization model. Specifically, it includes: Initialize the parameters, including model parameters and algorithm parameters; the parameters are mainly divided into model parameters and algorithm parameters, and the model parameters determine the solution of the optimization algorithm. Mainly includes: system matrix , projection method, gradient transformation matrix and data tolerance .

[0060] In this embodiment, the system matrix The voxel-driven projection method is used to construct the data. 0. The parameters related to the kernel TV algorithm affect the convergence speed and path of the algorithm. The Chambolle-Pock solution process includes: , and It is a parameter inherent to the algorithm. It is determined by calculation.

[0061] The dual variable is introduced to transform the doubly weighted kernel TV optimization model into a convex conjugate function.

[0062] Among them, according to the Chambolle-Pock algorithm form, the double weighted kernel TV optimization model is converted:

[0063]

[0064]

[0065]

[0066] In the formula, x is the original variable, u is the electron paramagnetic resonance image to be reconstructed, A is the imaging system matrix, and p and q are dual variables; To transform the doubly weighted kernel TV optimization model into a function of the dual variable p, To transform the doubly weighted kernel TV optimization model into a function of the dual variable q, is the balance factor. The value of the imaging system matrix is ​​obtained according to the three-dimensional ray-driven method. At a specific angle, the ray passes through the three-dimensional discrete image, and the intersection area with the three-dimensional voxel array is the ray weight. The specific angle is determined according to the current imaging physical environment, setting requirements, etc. The original variable is the electron paramagnetic resonance image to be reconstructed.

[0067] According to the derivation process of the CP algorithm example, and , and G(x), solving for the convex conjugate function:

[0068]

[0069] In the formula, is the singular value vector of the singular value decomposition of the weighted Jacobian matrix, is a convex conjugate function with respect to the dual variable p, is a convex conjugate function with respect to the dual variable q, Indicates data tolerance, is the L2 norm with respect to the dual variable p.

[0070] The proximal mapping calculation is performed on the convex conjugate function, and the singular values ​​of the image to be reconstructed and the voxel Jacobian matrix are updated until convergence to obtain the optimal electron paramagnetic resonance image reconstruction result.

[0071] in , and The proximal mapping calculation results are as follows:

[0072]

[0073]

[0074] In the formula, For function The proximal mapping of for The proximal mapping of For operator The proximal mapping of yes The singular value decomposition of . represents the soft threshold scaling of singular values, which is in the form of , is a non-zero constant, Indicates the data tolerance.

[0075] In this embodiment, an algorithm pseudo code is designed according to the above-mentioned reconstruction method example, as shown in Table 1, Table 2 and Table 3.

[0076] Table 1 Algorithm pseudocode

[0077] Table 2 gives the pseudo code of the wSV function. The weighted singular value matrix obtained by the singular value decomposition of the Jacobian matrix is ​​weighted, and then the soft threshold scaling is performed according to the weighted singular values. Indicates taking the absolute value of each value in the singular value matrix. Table 3 gives the pseudo code of the wTV function, which mainly implements the adaptive weighting operation of gradients in different directions.

[0078] Table 2 wSV function pseudo code

[0079] Table 3 wTV function pseudo code

[0080] In this embodiment, the data collected from the physical model by the EPR imager are used for the experiment. The EPR imager is a JIVA-25EPR oxygen imager. The main magnetic field strength is 250G, the resonant frequency is 720MHZ, and the resonator volume is The performance of the proposed method in sparse reconstruction is evaluated through real data experiments. In this experiment, the physical phantom consists of a 10 ml bottle filled with 1 mol / L trityl solution and sealed.

[0081] First, the projection data of the physical phantom are collected from 208 projection angles using equal solid angle sampling. The length of the projection is 3.4 cm and there are 80 measurements. The sparse data set is evenly distributed in the maximum margin projection method. The sparse data at 52, 104, 156 and 208 angles are obtained from the complete data set within the solid angle range. The sparse reconstruction of the T1-weighted images of the 8 channels of EPRI is performed based on the acquired real sparse data set, where the image size of each channel is , the voxel size is .

[0082] For sparse reconstruction of physical phantoms, the balance factor of the algorithm is set to .in, and The adaptive weight coefficient of the algorithm is set to The number of iterations was set to 4000. In actual EPRI research, it is usually impossible to obtain real images. Therefore, the images reconstructed at 208 projection feet using the Filtered Back-Projection (FBP) algorithm were used as reference images. Calculate data tolerance . Analyze the T1 mapping curve of 3D-EPRI voxels and set the first to fifth channels , Channel 6 to Channel 8 .

[0083] The data tolerance is:

[0084] In order to evaluate the image quality of the physical phantom reconstructed by the proposed algorithm, in addition to visually observing the reconstructed image, two evaluation indicators, rNOE and CNR, are designed for quantitative evaluation of the reconstructed image.

[0085] For the reconstruction of 3D digital phantoms, NOE and contrast noise ratio (CNR) are used to evaluate the reconstruction accuracy of the reconstructed image, that is, the image quality. For real data reconstruction, rNOE and CNR are used as evaluation indicators. They are defined as follows:

[0086]

[0087] in, represents the image reconstructed by the algorithm, and Indicates the region of interest (ROI) containing the signal in the reconstructed image and the ROI containing the background in the reconstructed image. and Represents the standard deviation of the image value within the ROI. The value is as follows: In actual data reconstruction, the true value image is unknown to the image. Therefore, when calculating the rNOE value according to the above formula, it is defined under 208 projection angles. The real image reconstructed by the algorithm is the reference image . Reference image reconstructed from actual projection data There is noise, so it can only be used as an indicator to evaluate the image reconstruction quality in real EPRI sparse reconstruction.

[0088] The size is 3D six-ball digital phantom for anti-crime experiment 3D digital phantom Figure 1 As shown, Figure 2The eight channel images of the 3D digital model are given. The model parameters and algorithm parameters in this experiment are set. In the anti-crime research, the number of projections is set to 828, and the projection data is . Use equal solid angle sampling mode. Determine the optimal algorithm parameters and balance factors through experimental verification and analysis. , and the adaptive weight coefficient is , the data tolerance in the simulation experiment is In the simulation experiment, the true value image is known. The TV value of the true value image is taken. The number of iterations is set to 15000. For simplicity, the experimental results only show the slice image of the eighth channel of the digital phantom as the object of qualitative observation. In the reverse crime research, the convergence condition is designed within the range of computer accuracy as follows:

[0089]

[0090]

[0091] Figure 3(b) shows The algorithm reconstructed the image of the 3D six-ball phantom at 828 angles, and the reconstructed image is exactly the same as the real image shown in Figure 3 (a). Figures 3 (c) and (d) show the comparison of the voxel contour curves of the slice image of the six-ball phantom at N / 2 and N / 4. The contour curve of the reconstructed image completely overlaps with the contour curve of the real image. Through qualitative observation, it is found that The algorithm can completely reconstruct the image, and the reverse crime is successful. Figure 4 shows the trend chart of the three convergence indicators of the algorithm. From Figure 4 (a), Figure 4 (b) and Figure 4 (c), it can be seen that the values ​​of the three indicators are all less than . The defined reverse crime convergence condition is satisfied, indicating that the proposed The algorithm's imaging model, optimization algorithm and its code implementation are correct.

[0092] The quality of the algorithm-reconstructed images was evaluated by visual observation and quantitative analysis. Figure 5 and Figure 6 Demonstrating the eighth channel of the algorithm reconstructing the physical model and The slice image, the window size is Tables 4 and 5 respectively give the rNOE and CNR of the eighth channel image of the physical model reconstructed by different algorithms. In Tables 4-5, at 208 angles, The rNOE values ​​of the algorithms were not compared.

[0093] Table 4. The results of FBP, , and The algorithm reconstructs the NOE value of the physical model.

[0094]

[0095] Table 5. FBP, , and The algorithm reconstructs the CNR value of the physical model.

[0096]

[0097] exist Figure 5 Compared with other algorithms, the crescent-shaped structure of the FBP algorithm is the worst due to the influence of noise in the projection. In particular, at 52 projection angles, the bottle image reconstructed by the FBP algorithm cannot even be distinguished from the entire slice image. By observing the images reconstructed by the other three algorithms, it can be found that no matter what projection angle is used, The crescent-shaped image structure reconstructed by the algorithm is clearer and more uniform. Figure 6 Four algorithms are given to reconstruct the eighth channel of the physical model. It can be clearly observed that under the same window, the image reconstructed by the FBP algorithm contains a lot of noise and has the worst imaging effect. The image background reconstructed by the other three algorithms is cleaner, but a closer look reveals that and The image reconstructed by the algorithm has overly smoothed edges and lost a lot of edge details. The algorithm retains more image details and has better edge retention. In the experiment, the trityl solution in the small bottle is evenly distributed, so the image color of the high-density small bottle should be uniform. By comparing the images of the small bottle reconstructed by different algorithms, it is found that using the same projection data, The images reconstructed by this algorithm are more uniform than those reconstructed by other algorithms.

[0098] In addition to the above visual observation analysis, this embodiment also provides FBP, , and The algorithm makes quantitative analysis based on the rNOE value and CNR value calculated by the formula. It can be seen from Table 4 that The rNOE values ​​of the proposed algorithm are lower than those of other algorithms, which indicates that the proposed algorithm has the ability to accurately reconstruct images from sparse projection data. Table 5 shows the CNR values ​​of each algorithm. From the calculation results, it can be seen that no matter what sparse angle the projection data is, The CNR values ​​of the algorithms are higher than those of other algorithms. The algorithm can accurately reconstruct low-contrast images. This further illustrates the application of multi-channel EPRI The algorithm can achieve accurate sparse reconstruction, compared to The algorithm has better edge preservation and noise suppression capabilities, and can better achieve fast EPRI sparse reconstruction.

[0099] In addition, from Figure 7 It can be seen that at different angles of 52, 104, 156 and 208, when the number of iterations reaches 5000. The algorithm cannot visually reconstruct the image at any angle, and the reconstructed image has obvious artifacts. The reconstructed images of the algorithm at 156 and 208 projection angles are basically consistent with the real images, and the naked eye can no longer distinguish them. Although the reconstructed images at 52 and 104 projection angles have artifacts, their adaptive weights can adaptively smooth the image edges, making the image edges better preserved. The overall imaging effect is better than However, for the algorithm proposed in this paper The algorithm can reconstruct the image at any sparse angle, which proves that the algorithm further improves The edge protection capability of the algorithm will not reduce The ability of the algorithm to suppress artifacts.

[0100] Example 2 like Figure 2 As shown, this embodiment provides a multi-channel electron paramagnetic resonance imaging system, which is used to implement the multi-channel electron paramagnetic resonance imaging method in the above-mentioned embodiment 1. The system mainly includes: an optimization model building module and an optimization model solving module.

[0101] Optimization model construction module; for multi-channel electron paramagnetic resonance images, a double-weighted kernel TV optimization model based on data separation constraints is constructed; the double-weighted kernel TV optimization model is obtained according to the double-weighted kernel TV norm of the multi-channel electron paramagnetic resonance image; the double-weighted kernel TV norm of the voxel set in the multi-channel electron paramagnetic resonance image is equal to the sum of the adaptive weighted singular values ​​corresponding to the Jacobian matrix weighted by the double-weighted kernel TV norm; The optimization model solving module solves the double-weighted nuclear TV optimization model based on the Chambolle-Pock algorithm during the reconstruction of multi-channel electron paramagnetic resonance images based on nuclear TV, and the solution obtained is used as the reconstructed electron paramagnetic resonance image.

[0102] The specific implementation methods of each module have been described in Example 1 and will not be repeated here.

[0103] Example 3 In another embodiment of the present invention, the present invention also provides a storage medium, specifically a computer-readable storage medium (Memory), which is a memory device in a terminal device for storing programs and data. It is understandable that the computer-readable storage medium here can include both the built-in storage medium in the terminal device and the extended storage medium supported by the terminal device, and can be any tangible medium containing or storing a program, which can be used by an instruction execution system, device or device or used in combination with it. The computer-readable storage medium provides a storage space, which stores the operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by the processor are also stored in the storage space, and these instructions can be one or more computer programs (including program codes). It should be noted that more specific examples (non-exhaustive list) of the computer-readable storage medium here include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0104] Computer readable storage media also include data signals propagated in baseband or as part of a carrier wave, which carry readable program codes. Such propagated data signals can take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The readable storage medium can also be any readable medium other than a readable storage medium, which can send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, device, or device. The program code contained on the readable storage medium can be transmitted using any appropriate medium, including but not limited to wireless, wired, optical cable, RF, etc., or any suitable combination of the above.

[0105] Program code for performing the operations of the present invention may be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java, C++, etc., and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as a separate software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server. In the case of a remote computing device, the remote computing device may be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., through the Internet using an Internet service provider).

[0106] The processor may load and execute one or more instructions stored in the computer-readable storage medium to implement the corresponding steps of the multi-channel electron paramagnetic resonance imaging method in the above embodiment; the processor may load and execute the following steps: For multi-channel electron paramagnetic resonance images, a doubly weighted kernel TV optimization model based on data separation constraints is constructed; In the process of reconstructing a multi-channel electron paramagnetic resonance image based on nuclear TV, the double-weighted nuclear TV optimization model is solved based on the Chambolle-Pock algorithm, and the solution obtained is used as the reconstructed electron paramagnetic resonance image.

[0107] Example 4 See also Fig. 9 , the terminal device is a computer device, and the computer device 60 of this embodiment includes: a processor 61, a memory 62, and a computer program 63 stored in the memory 62 and executable on the processor 61. When the computer program 63 is executed by the processor 61, the multi-channel electron paramagnetic resonance imaging method in the embodiment is implemented. To avoid repetition, it is not described one by one here. Alternatively, when the computer program 63 is executed by the processor 61, the functions of each model / unit in the computing system in the multi-channel electron paramagnetic resonance imaging method in Embodiment 1 are implemented. To avoid repetition, it is not described one by one here.

[0108] The computer device 60 may be a computing device such as a desktop computer, a notebook, a PDA, or a cloud server. The computer device 60 may include, but is not limited to, a processor 61 and a memory 62. Those skilled in the art will appreciate that Fig. 9This is only an example of the computer device 60 and does not constitute a limitation of the computer device 60. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the computer device may also include input and output devices, network access devices, buses, etc.

[0109] The processor 61 may be a central processing unit (CPU), other general-purpose processors, central processing units, graphics processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, data processing logic based on quantum computing, discrete hardware components, etc. A general-purpose processor may be a microprocessor or the processor may be any conventional processor, etc.

[0110] The memory 62 may be an internal storage unit of the computer device 60, such as a hard disk or memory of the computer device 60. The memory 62 may also be an external storage device of the computer device 60, such as a plug-in hard disk, a smart memory card (SmartMediaCard, SMC), a secure digital (SecureDigital, SD) card, a flash card (FlashCard), etc. equipped on the computer device 60.

[0111] Furthermore, the memory 62 may include both an internal storage unit of the computer device 60 and an external storage device. The memory 62 is used to store computer programs and other programs and data required by the computer device. The memory 62 may also be used to temporarily store data that has been output or is to be output.

[0112] Any reference to a memory, database or other medium used in the embodiments provided in this application may include at least one of a non-volatile and a volatile memory. Non-volatile memory may include a read-only memory (ROM), a magnetic tape, a floppy disk, a flash memory, an optical memory, a high-density embedded non-volatile memory, a resistive random access memory (ReRAM), a magnetic random access memory (MRAM), a ferroelectric random access memory (FRAM), a phase change memory (PCM), a graphene memory, etc. Volatile memory may include a random access memory (RAM) or an external cache memory, etc. As an illustration and not limitation, RAM may be in various forms, such as a static random access memory (SRAM) or a dynamic random access memory (DRAM), etc.

[0113] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

Claims

1. A multi-channel electron paramagnetic resonance imaging method, characterized in that: include: For multi-channel electron paramagnetic resonance images, a doubly weighted kernel TV optimization model based on data separation constraints is constructed; the doubly weighted kernel TV optimization model is obtained according to the doubly weighted kernel TV norm of the data separation constraints of the multi-channel electron paramagnetic resonance images; the doubly weighted kernel TV norm of the voxels set in the multi-channel electron paramagnetic resonance images is equal to the sum of the adaptive weighted singular values ​​corresponding to the Jacobian matrix weighted by the doubly weighted kernel TV norm; In the process of reconstructing a multi-channel electron paramagnetic resonance image based on nuclear TV, the Chambolle-Pock algorithm is used to solve the double-weighted nuclear TV optimization model, and the solution obtained is used as the reconstructed electron paramagnetic resonance image.

2. The multi-channel electron paramagnetic resonance imaging method according to claim 1, characterized in that: The double-weighted kernel TV optimization model based on data separation constraints includes: In the formula, Representing multi-channel images The doubly weighted kernel TV norm of ; Represents the adaptive weight coefficient along the gradient direction; Represents the adaptive weight coefficient of each singular value; is the solution of the optimization model, i.e., the reconstructed electron paramagnetic resonance image; is a non-zero balance factor; Represents the difference between the actual measured projection data and the optimized model norm, Indicates the data tolerance.

3. The multi-channel electron paramagnetic resonance imaging method according to claim 1, characterized in that: The doubly weighted kernel TV norm of a voxel in a multi-channel electron paramagnetic resonance image is set equal to the sum of the adaptively weighted singular values ​​corresponding to its weighted Jacobian matrix, specifically including: The Jacobian matrix of each voxel position of the electron paramagnetic resonance image is defined as: In the formula, the adaptive weight coefficient They are defined as: In the formula, is the adaptive weight coefficient in the x direction, is the adaptive weight coefficient in the y direction, is the adaptive weight coefficient in the z direction, x, y, z represent the three directions in the spatial coordinate system, D is the gradient transformation matrix, i, j, k represent the position information of the voxel relative to the spatial coordinate system, is a weight coefficient greater than 0, used to control the strength of each iteration.

4. The multi-channel electron paramagnetic resonance imaging method according to claim 1, characterized in that: The dual-weighted kernel TV optimization model is solved based on the Chambolle-Pock algorithm, specifically including: Initializing parameters, including model parameters and algorithm parameters; Introducing a dual variable to transform the double-weighted kernel TV optimization model into a convex conjugate function; The proximal mapping calculation is performed on the convex conjugate function, and the singular values ​​of the image to be reconstructed and the voxel Jacobian matrix are continuously and alternately updated until convergence to obtain the optimal electron paramagnetic resonance image reconstruction result.

5. The multi-channel electron paramagnetic resonance imaging method according to claim 4, characterized in that: The dual-weighted kernel TV optimization model is converted into a convex conjugate function according to the Chambolle-Pock algorithm framework. The conversion process specifically includes: Where x is the original variable, u is the electron paramagnetic resonance image to be reconstructed, A is the imaging system matrix, and the value of the system matrix is ​​obtained according to the three-dimensional ray-driven method. At a specific angle, the ray passes through the three-dimensional discrete image, and the intersection area with the three-dimensional voxel array is the ray weight; p and q are dual variables; To transform the doubly weighted kernel TV optimization model into a function of the dual variable p, To transform the doubly weighted kernel TV optimization model into a function of the dual variable q, is the balancing factor, and the original variable is the electron paramagnetic resonance image to be reconstructed.

6. The multi-channel electron paramagnetic resonance imaging method according to claim 5, characterized in that: The convex conjugate function is: In the formula, is the singular value vector of the singular value decomposition of the weighted Jacobian matrix, is a convex conjugate function with respect to the dual variable p, is a convex conjugate function with respect to the dual variable q, Indicates data tolerance, is the L2 norm with respect to the dual variable p.

7. The multi-channel electron paramagnetic resonance imaging method according to claim 6, characterized in that: The calculation result of the proximal mapping is: In the formula, For function The proximal mapping of for The proximal mapping of For operator The proximal mapping of yes The singular value decomposition of ; represents the soft threshold scaling of singular values, is a non-zero constant, Indicates the data tolerance.

8. A multi-channel electron paramagnetic resonance imaging system, used to implement the multi-channel electron paramagnetic resonance imaging method according to any one of claims 1 to 7, characterized in that: include: An optimization model construction module; for multi-channel electron paramagnetic resonance images, a double-weighted kernel TV optimization model based on data separation constraints is constructed; the double-weighted kernel TV optimization model is obtained according to the double-weighted kernel TV norm of the multi-channel electron paramagnetic resonance image; the double-weighted kernel TV norm of the voxel set in the multi-channel electron paramagnetic resonance image is equal to the sum of the adaptive weighted singular values ​​corresponding to the Jacobian matrix weighted by the double-weighted kernel TV norm; The optimization model solving module solves the double-weighted nuclear TV optimization model based on the Chambolle-Pock algorithm during the reconstruction of the multi-channel electron paramagnetic resonance image based on the nuclear TV, and the obtained solution is used as the reconstructed electron paramagnetic resonance image.

9. A computer-readable storage medium storing one or more programs, characterized in that: The one or more programs include instructions, which, when executed by a computing device, cause the computing device to perform the multi-channel electron paramagnetic resonance imaging method according to any one of claims 1 to 7.

10. A computing device, characterized in that include: One or more processors, a memory and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs include steps for executing the multi-channel electron paramagnetic resonance imaging method according to any one of claims 1 to 7.

Images