A dynamic magnetic resonance imaging method, device and storage medium based on second-order generalized total variation and nuclear norm
By using second-order generalized total variation and nuclear norm models to constrain sparse and low-rank matrices in dynamic magnetic resonance imaging, the image reconstruction accuracy and artifact problems in dynamic magnetic resonance imaging in the prior art are solved, and the more High-quality image reconstruction.
Patent Information
- Application Number
- CN202210513142.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-05-12
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2042-05-12
AI Technical Summary
The prior art has a step effect in dynamic magnetic resonance imaging, resulting in general image reconstruction accuracy, poor edge information processing, and insufficient effect of suppressing motion artifacts.
Using the method based on second-order generalized total variation and kernel norm, the original matrix of dynamic magnetic resonance imaging is constrained by sparse matrix and low-rank matrix. Through iterative solution of the original dual algorithm, the second-order generalized total variation and kernel norm are constructed. Model, optimize sparse and low rank matrices to generate high-quality reconstructed images.
Effectively suppress the step effect, improve image edge reconstruction effect, improve spatial artifact suppression, and generate higher quality dynamic magnetic resonance imaging.
Smart Images

Figure CN114972079B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of medical imaging technology, and in particular to a dynamic magnetic resonance imaging method, device and storage medium based on second-order generalized total variation and nuclear norm. Background Art
[0002] Dynamic MRI can observe the structure and function of the internal organs of the human body and is an important indirect diagnostic method in medicine. It contains a set of magnetic resonance images of a single target over a period of time, and is a magnetic resonance signal that covers both time and space information. Among them, the temporal and spatial resolution is a particularly important indicator for dynamic MRI images. High-standard temporal and spatial resolution can effectively improve the accuracy of clinical diagnosis.
[0003] Since a lot of scanning time is often required in the actual magnetic resonance imaging process, coupled with interference from external factors, the resolution of the final image is not ideal. To address this problem, researchers applied compressed sensing imaging technology to magnetic resonance images. Dynamic magnetic resonance images are regarded as the superposition of low-rank components and sparse components. Different constraints can be used to regularize the sparse components so that after subtracting the background information, the remaining part of the image will be sparser than the original image itself. With the application of compressed sensing imaging technology, many scholars have gradually noticed the low-rank property of the compressed sensing imaging technology matrix. They are no longer limited to studying sparsity alone, and try to combine low-rank with sparsity for research. For dynamic magnetic resonance images, images of different time frames in the same layer of space can be regarded as a column vector as a column of the space-time matrix, and a low-rank matrix is constructed from multiple time frames, thereby transforming the dynamic magnetic resonance image reconstruction problem into a low-rank matrix recovery problem, effectively improving the reconstruction speed and efficiency.
[0004] However, for the existing low-rank and sparse decomposition models in compressed sensing imaging technology, the low-rank components are represented by the nuclear norm, and the sparse components are represented by the sum of the absolute values of all elements of the matrix. Due to the staircase effect of this model when processing dynamic magnetic resonance images, the grayscale values in some areas are usually the same, resulting in the first-order derivative of the grayscale values in these areas being 0, making the brightness of each area different and the image contours overly sharp, resulting in general reconstruction accuracy of dynamic magnetic resonance imaging, and there is still room for improvement in the reconstruction of image edge information and the suppression of motion artifacts. Summary of the invention
[0005] The purpose of the present invention is to overcome the deficiencies in the prior art and to provide a dynamic magnetic resonance imaging method, device and storage medium based on second-order generalized total variation and nuclear norm, which can have a good effect of suppressing spatial artifacts while retaining the edge information of the image, thereby ensuring the quality of the reconstructed image.
[0006] To achieve the above object, the present invention is implemented by adopting the following technical solutions:
[0007] In a first aspect, the present invention provides a dynamic magnetic resonance imaging method based on second-order generalized total variation and nuclear norm, the method comprising:
[0008] Obtaining the original matrix of dynamic magnetic resonance imaging;
[0009] By using the second-order generalized total variation and nuclear norm to constrain the sparse matrix and low-rank matrix of the original matrix respectively, a second-order generalized total variation and nuclear norm model is constructed;
[0010] Iteratively solving the second-order generalized total variation and nuclear norm models using a primal-dual algorithm to obtain an optimal low-rank matrix and a sparse matrix;
[0011] The optimal low-rank matrix and sparse matrix are combined to obtain the reconstructed image of dynamic magnetic resonance.
[0012] In combination with the first aspect, preferably, the initial expression for constructing the second-order generalized total variation and nuclear norm model is:
[0013]
[0014] Where E is the data collection operator, L is the low-rank matrix, ‖L‖ * represents the nuclear norm of the low-rank matrix L, S is a sparse matrix, represents the second-order generalized total variation, represents the second-order generalized total variation of the sparse matrix S; d is the undersampled spatiotemporal data; β is the penalty parameter; α represents the weight coefficient.
[0015] In combination with the first aspect, preferably, the data acquisition operator is mapped to a finite-dimensional Shilbit space, and the initial expression (1) of the second-order generalized total variation and nuclear norm model is simplified to:
[0016]
[0017] Where: U and V represent two different finite-dimensional space vectors, V = U 3 ; represents the gradient, ν represents the symmetric tensor, α1 represents the weight coefficient of the gradient term, α0 represents the weight coefficient of the symmetric gradient term, and E represents the symmetric gradient; || ||1 represents the L1 norm, which is the sum of the absolute values of each element in the matrix; Represents the square of the F-norm of the matrix, which is the square root of the sum of the squares of all elements in the matrix.
[0018] In combination with the first aspect, preferably, the iteratively solving the second-order generalized total variation and nuclear norm model using the primal-dual algorithm comprises the following steps:
[0019] Initialize the model parameters of the constructed second-order generalized total variation and nuclear norm models;
[0020] constructing the primal-dual saddle point expression of the second-order generalized total variation and nuclear norm model using the primal-dual algorithm;
[0021] Based on the primal-dual saddle point expression, updating the sparse matrix and the low-rank matrix by iteratively solving the dual variables and the singular value thresholds respectively;
[0022] When the preset termination condition or the preset number of iterations is reached during the iterative solution process, the iteration is stopped, and the optimal low-rank matrix and sparse matrix are updated at this time.
[0023] In combination with the first aspect, preferably, the primal-dual saddle point expression of the second-order generalized total variation and nuclear norm model constructed by using the primal-dual algorithm is:
[0024]
[0025] Where Y is a finite-dimensional Shilbit space, p represents the dual variable of the gradient, q represents the dual variable of the symmetric gradient, λ represents the dual variable of the data acquisition operator, <,> represents the inner product of two matrices, and I is the indicator function of the convex set,|| || ∞ Represents the infinite norm.
[0026] In combination with the first aspect, preferably, updating the sparse matrix and the low-rank matrix by iteratively solving the dual variables and the singular value thresholds based on the primal dual saddle point expression respectively comprises:
[0027] The first-order symmetric divergence operator is used to calculate and update the sparse matrix through formula (4):
[0028] S b+1 =S b -τ(E * λ b+1 -div1p b+1 ) (4)
[0029] Where div1 represents the first-order symmetric divergence operator, τ is a non-negative constant; E * represents the adjoint matrix; b represents the number of iterations, b=1,2…; S b represents the sparse matrix updated in the bth iteration, S b+1 represents the sparse matrix updated at the b+1th iteration, λ b+1 represents the dual variable of the data acquisition operator updated in the b+1th iteration; pb+1 represents the dual variable of the gradient updated at the b+1th iteration; where, through the projection operator P x To solve and update the dual variable p of the gradient, the dual variable q of the symmetric gradient and the dual variable λ of the data acquisition operator, P x The expression is:
[0030]
[0031] Where t is a set of infinite norms less than the number of shear transformation subbands; x represents any dual variable of the gradient p, the symmetric gradient q, and the data acquisition operator λ;
[0032] The singular value threshold method is used to calculate and update the low-rank matrix through formula (6):
[0033] L b+1 =S j (L b -τE * r b+1 ) (6)
[0034] Where: L b+1 The low-rank matrix updated in the b+1th iteration; L b The low-rank matrix updated in the bth iteration, r b+1 The rank of the low-rank matrix updated in the b+1th iteration; j represents the threshold; S j Contraction operator, S j Obtained through the singular value threshold operator SVT: SVT = US j (Σ)V T , where UΣV T represents the singular value decomposition of the low-rank matrix, U, Σ, and V represent the submatrices generated after the singular value decomposition of the low-rank matrix, and V T represents the transpose of the matrix V;
[0035] The calculation is repeated iteratively until the preset termination condition is reached or the number of iterations b reaches the preset value, and finally the optimal low-rank matrix and sparse matrix are obtained.
[0036] In combination with the first aspect, preferably, the preset termination condition is:
[0037] τσ||K|| 2 <1 (7)
[0038] Where σ is a non-negative constant; K is a matrix consisting of gradients and symmetric gradients, the identity matrix, and linear continuous operators in the low-rank matrix.
[0039] In a second aspect, the present invention provides a dynamic magnetic resonance imaging device based on second-order generalized total variation and nuclear norm, the device comprising:
[0040] An acquisition module, used for acquiring an original matrix of dynamic magnetic resonance imaging;
[0041] A construction module, used for constraining the sparse matrix and the low-rank matrix of the original matrix by using the second-order generalized total variation and the nuclear norm, so as to construct a second-order generalized total variation and nuclear norm model;
[0042] An iterative solution module, used for iteratively solving the second-order generalized total variation and nuclear norm model using a primal-dual algorithm to obtain an optimal low-rank matrix and a sparse matrix;
[0043] The image reconstruction module is used to combine the obtained optimal low-rank matrix and sparse matrix to obtain a reconstructed image of dynamic magnetic resonance.
[0044] In a third aspect, the present invention further provides a dynamic magnetic resonance imaging device based on second-order generalized total variation and nuclear norm, comprising a processor and a storage medium;
[0045] The storage medium is used to store instructions;
[0046] The processor is used to operate according to the instructions to execute the steps of the dynamic magnetic resonance imaging method based on second-order generalized total variation and nuclear norm as described in any one of the first aspects.
[0047] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the dynamic magnetic resonance imaging method based on second-order generalized total variation and nuclear norm as described in any one of the first aspects.
[0048] Compared with the prior art, the present invention has the following beneficial effects:
[0049] The present invention provides a method for constraining the sparse matrix and the low-rank matrix in the original matrix of dynamic magnetic resonance imaging by using the second-order generalized total variation and the nuclear norm, respectively, to construct a second-order generalized total variation and nuclear norm model, and to update and optimize the sparse matrix and the low-rank matrix by using the primal dual algorithm, so that the step effect can be better suppressed in the reconstruction of dynamic magnetic resonance images, the edge reconstruction effect of the image is smoother, and the effect of suppressing spatial artifacts is better while retaining the edges, thereby generating higher quality images. BRIEF DESCRIPTION OF THE DRAWINGS
[0050] Figure 1 A schematic flow chart of a dynamic magnetic resonance imaging method based on second-order generalized total variation and nuclear norm provided in an embodiment of the present invention;
[0051] Figure 2 A structural principle block diagram of a dynamic magnetic resonance imaging device based on second-order generalized total variation and nuclear norm provided by an embodiment of the present invention;
[0052] Figure 3 This is a comparison diagram of the reconstructed image results of the method proposed in the present invention, the kt SLR method, and the kt RPCA method in the cardiac perfusion dataset PINCAT and the cardiac trunk dataset cardiac. DETAILED DESCRIPTION
[0053] The technical solution of the present invention is described in detail below through the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present application and the specific features in the embodiments are detailed descriptions of the technical solution of the present application, rather than limitations on the technical solution of the present application. In the absence of conflict, the embodiments of the present application and the technical features in the embodiments can be combined with each other.
[0054] The term "and / or" in this article is only a description of the association relationship of the associated objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship.
[0055] Embodiment 1:
[0056] like Figure 1 As shown, the embodiment of the present invention introduces a dynamic magnetic resonance imaging method flow based on second-order generalized total variation and nuclear norm, which specifically includes the following steps:
[0057] Step 1: Obtain the original matrix of dynamic magnetic resonance imaging;
[0058] Step 2: by using the second-order generalized total variation and nuclear norm to constrain the sparse matrix and low-rank matrix of the original matrix respectively, a second-order generalized total variation and nuclear norm model is constructed;
[0059] The initial expression for constructing the second-order generalized total variation and nuclear norm model is:
[0060]
[0061] Where E is the data collection operator, L is the low-rank matrix, ‖L‖ * represents the nuclear norm of the low-rank matrix L, S is a sparse matrix, represents the second-order generalized total variation, represents the second-order generalized total variation of the sparse matrix S; d is the undersampled spatiotemporal data; β is the penalty parameter; α represents the weight coefficient.
[0062] As an embodiment of the present invention, in this step, the data acquisition operator is mapped to a finite-dimensional Shilbit space. After the mapping operation, the initial expressions of the second-order generalized total variation and nuclear norm model can be derived as follows:
[0063]
[0064] Where: U and V represent two different finite-dimensional space vectors, V = U 3 ; represents the gradient, ν represents the symmetric tensor, α1 represents the weight coefficient of the gradient term, α0 represents the weight coefficient of the symmetric gradient term, and E represents the symmetric gradient; || ||1 represents the L1 norm, which is the sum of the absolute values of each element in the matrix; Represents the square of the F-norm of the matrix, which is the square root of the sum of the squares of all elements in the matrix.
[0065] Step 3: Iteratively solve the second-order generalized total variation and nuclear norm models using the primal-dual algorithm to obtain the optimal low-rank matrix and sparse matrix;
[0066] Step 4: Combining the obtained optimal low-rank matrix and sparse matrix is the reconstructed image of dynamic magnetic resonance.
[0067] As an embodiment of the present invention, in step 3, the primal-dual algorithm is used to iteratively solve the second-order generalized total variation and nuclear norm model, including the following steps:
[0068] Step a: Initialize the model parameters of the constructed second-order generalized total variation and nuclear norm models;
[0069] Step b: constructing the primal-dual saddle point expression of the second-order generalized total variation and nuclear norm model using the primal-dual algorithm;
[0070] The primal-dual saddle point expression is shown in formula (3):
[0071]
[0072] Where Y is a finite-dimensional Shilbit space, p represents the dual variable of the gradient, q represents the dual variable of the symmetric gradient, λ represents the dual variable of the data acquisition operator, <,> represents the inner product of two matrices, and I is the indicator function of the convex set,|| || ∞ Represents the infinite norm.
[0073] Step c: Based on the primal-dual saddle point expression, the sparse matrix and the low-rank matrix are updated by iteratively solving the dual variables and the singular value thresholds respectively;
[0074] Step d: When a preset termination condition or a preset number of iterations is reached during the iterative solution process, the iteration is stopped, and the optimal low-rank matrix and sparse matrix are updated at this time.
[0075] As an embodiment of the present invention, by cyclically solving the minimization problem of dual variables, the dual variables include dual variables of gradients, dual variables of symmetric gradients, and dual variables of data acquisition operators, thereby obtaining a low-rank matrix L and a sparse matrix S respectively; specifically, in step c, based on the original dual saddle point expression, the sparse matrix and the low-rank matrix are updated by iteratively solving the dual variables and the singular value threshold respectively; specifically including:
[0076] The first-order symmetric divergence operator is used to calculate and update the sparse matrix through formula (4):
[0077] S b+1 =S b -τ(E * λ b+1 -div1p b+1 ) (4)
[0078] Where div1 represents the first-order symmetric divergence operator, τ is a non-negative constant; E * represents the adjoint matrix; b represents the number of iterations, b=1,2…; S b represents the sparse matrix updated in the bth iteration, S b+1 represents the sparse matrix updated at the b+1th iteration, λ b+1 represents the dual variable of the data acquisition operator updated in the b+1th iteration; p b+1 represents the dual variable of the gradient updated at the b+1th iteration; where, through the projection operator P x To solve and update the dual variable p of the gradient, the dual variable q of the symmetric gradient and the dual variable λ of the data acquisition operator, P x The expression is:
[0079]
[0080] Where t is a set of infinite norms less than the number of shear transformation subbands; x represents any dual variable of the gradient p, the symmetric gradient q, and the data acquisition operator λ;
[0081] The singular value threshold method is used to calculate and update the low-rank matrix through formula (6):
[0082] L b+1 =S j (L b -τE * r b+1 ) (6)
[0083] Where: L b+1 The low-rank matrix updated in the b+1th iteration; L b The low-rank matrix updated in the bth iteration, r b+1 The rank of the low-rank matrix updated in the b+1th iteration; j represents the threshold; S j Contraction operator, S j Obtained through the singular value threshold operator SVT: SVT = US j (Σ)V T , where UΣV T represents the singular value decomposition of the low-rank matrix, U, Σ, and V represent the submatrices generated after the singular value decomposition of the low-rank matrix, and V T represents the transpose of the matrix V;
[0084] The above formulas (4), (5) and (6) are iterated repeatedly until the algorithm meets the preset termination condition or the number of iterations b reaches the preset value, and finally the optimal low-rank matrix and sparse matrix are obtained, thereby restoring the target image.
[0085] Specifically, the model parameters of the second-order generalized total variation and nuclear norm model constructed by initialization in the embodiment of the present invention include the weight coefficient of the gradient term, the weight coefficient of the symmetric gradient term, the number of iterations, and the constants τ and σ; the weight coefficient α1 of the gradient term is set to 0.004, the weight coefficient α0 of the symmetric gradient term is set to 0.006, the preset number of iterations b is taken as 500, and τ and σ are both set to 0.25; in addition, the preset termination condition is:
[0086] τσ||K|| 2 <1 (7)
[0087] Where σ is a non-negative constant; K is a matrix consisting of gradients and symmetric gradients, the identity matrix, and linear continuous operators in the low-rank matrix.
[0088] It should be noted that the image reconstruction of the dynamic magnetic resonance imaging method based on the second-order generalized total variation and nuclear norm provided in the embodiment of the present invention is completed based on the Matlab platform; in order to compare the reconstruction performance of each algorithm, this application uses the cardiac perfusion PINCAT data set and the cardiac trunk cardiac data set in dynamic magnetic resonance imaging for simulation experiments; the resolution size of the PINCAT data set is set to 128×128, and the time frame number is set to 50 frames; the resolution size of the cardiac data set is set to 190×90, and the time frame number is set to 70 frames; the sampling method is radial sampling The sampling mode is used, and the sampling acceleration factor is 6.6. For each frame, the radial trajectory interval is uniform, and each time frame is rotated by a small random angle to increase the incoherence. The model parameters of the second-order generalized total variation and nuclear norm models are set as follows: α0=0.006, α1=0.004, β=0.5, σ=0.25, τ=0.25. Under the premise of ensuring the convergence of the algorithm, the preset value of the number of iterations b is set to 500. The signal error rate SER and structural similarity SSIM are used as the performance indicators of the reconstructed image. The larger the values of SER and SSIM, the better the image performance.
[0089] As shown in Table 1, the results of simulation tests are given for the dynamic magnetic resonance imaging method based on second-order generalized total variation and nuclear norm provided by an embodiment of the present invention, the k-tRPCA method proposed by Trémoulhéac BR et al. in “Low-rank and sparse reconstruction indynamic magnetic resonance imaging via proximal splitting methods”, and the kt SLR method proposed by Lingala SG, HuY, DiBella E et al. in “celerated dynamic MRI exploitingsparsity and low-rank structure: kt SLR”. It can be seen from Table 1 that the SSIM values of the method proposed in the present invention on both PINCAT and cardiac data sets are higher than the SSIM values obtained by the existing kt SLR method and kt RPCA method. In the PINCAT data set, the SER value obtained by the method proposed in the present invention is higher than that of the traditional kt RPCA. In the cardiac data set, the method proposed in the present invention has the highest SER value.
[0090] Table 1 Experimental results on PINCAT and cardiac datasets
[0091]
[0092] As shown in Table 2, a comparison is given between the method proposed in the present invention and the kt SLR method and the kt RPCA method under different acceleration factors of the cardiac data set in the simulation experiment. It can be seen from the data in Table 2 that in the reconstruction of the cardiac data set, for different acceleration factors, the method proposed in the present application has higher SSIM values and SER values than the other two existing methods, so the image quality obtained by the method proposed in the present invention is higher.
[0093] Table 2 Comparison of different acceleration factors under cardiac dataset
[0094]
[0095]
[0096] like Figure 3 As shown in FIG. 1 , the reconstruction image results of the method proposed in the present invention, the kt SLR method and the kt RPCA method are compared in the PINCAT dataset and the cardiac dataset; Figure 3 It can be seen that the method proposed in the present invention is superior to the other two models in suppressing artifacts and retaining edge information for the PINCAT dataset; and it produces the least artifacts on the cardiac dataset and the reconstructed image is smoother.
[0097] In summary, the dynamic magnetic resonance imaging method based on second-order generalized total variation and nuclear norm provided in the embodiments of the present invention has shown in simulation experiments on the PINCAT dataset and the cardiac dataset that: compared with the existing kt SLR method and k-tRPCA method, the method proposed in the present invention is superior to the existing traditional methods in terms of ensuring edge information and suppressing artifacts, and has higher structural similarity and signal error rate under different acceleration factors, which can effectively improve the quality of the reconstructed image.
[0098] Embodiment 2:
[0099] like Figure 2 As shown, an embodiment of the present invention provides a dynamic magnetic resonance imaging device based on second-order generalized total variation and nuclear norm, which can be used to implement the method described in embodiment 1, and the device includes:
[0100] An acquisition module, used for acquiring an original matrix of dynamic magnetic resonance imaging;
[0101] A construction module, used for constraining the sparse matrix and the low-rank matrix of the original matrix by using the second-order generalized total variation and the nuclear norm, so as to construct a second-order generalized total variation and nuclear norm model;
[0102] An iterative solution module, used for iteratively solving the second-order generalized total variation and nuclear norm model using a primal-dual algorithm to obtain an optimal low-rank matrix and a sparse matrix;
[0103] The image reconstruction module is used to combine the obtained optimal low-rank matrix and sparse matrix to obtain a reconstructed image of dynamic magnetic resonance.
[0104] The smart contract-based blockchain transfer node incentive device provided in the embodiment of the present invention and the dynamic magnetic resonance imaging method based on second-order generalized total variation and nuclear norm provided in Example 1 are based on the same technical concept, and can produce the beneficial effects described in Example 1. For the contents not described in detail in this embodiment, please refer to Example 1.
[0105] Embodiment three:
[0106] An embodiment of the present invention provides a dynamic magnetic resonance imaging device based on second-order generalized total variation and nuclear norm, including a processor and a storage medium;
[0107] The storage medium is used to store instructions;
[0108] The processor is used to operate according to the instructions to execute the steps of any method in the first embodiment.
[0109] Embodiment 4:
[0110] An embodiment of the present invention provides a computer-readable storage medium on which a computer program is stored, characterized in that when the computer program is executed by a processor, the steps of any method in the first embodiment are implemented.
[0111] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.
[0112] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0113] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0114] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.
[0115] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.
Claims
1. A dynamic magnetic resonance imaging method based on second-order generalized total variation and nuclear norm, characterized in that: The method comprises: Obtaining the original matrix of dynamic magnetic resonance imaging; By using the second-order generalized total variation and nuclear norm to constrain the sparse matrix and low-rank matrix of the original matrix respectively, a second-order generalized total variation and nuclear norm model is constructed; Iteratively solving the second-order generalized total variation and nuclear norm models using a primal-dual algorithm to obtain an optimal low-rank matrix and a sparse matrix; The optimal low-rank matrix and sparse matrix are combined to form a reconstructed image of dynamic magnetic resonance; The iterative solution of the second-order generalized total variation and nuclear norm model using the primal-dual algorithm comprises the following steps: Initialize the model parameters of the constructed second-order generalized total variation and nuclear norm models; constructing the primal-dual saddle point expression of the second-order generalized total variation and nuclear norm model using the primal-dual algorithm; Based on the primal-dual saddle point expression, updating the sparse matrix and the low-rank matrix by iteratively solving the dual variables and the singular value thresholds respectively; When the preset termination condition or the preset number of iterations is reached during the iterative solution process, the iteration is stopped, and the optimal low-rank matrix and sparse matrix are updated at this time; The primal-dual saddle point expression of the second-order generalized total variation and nuclear norm model constructed using the primal-dual algorithm is: Where Y is a finite-dimensional Shilbit space, p represents the dual variable of the gradient, q represents the dual variable of the symmetric gradient, λ represents the dual variable of the data acquisition operator, and <,> represents the inner product of two matrices. is the indicator function of a convex set,|||| ∞ represents the infinite norm; Wherein, based on the original dual saddle point expression, the sparse matrix and the low-rank matrix are updated by iteratively solving the dual variables and the singular value thresholds, respectively, including: The sparse matrix is updated by using the first-order symmetric divergence operator through formula (2): S b+1 =S b -τ(E * l b+1 -div1p b+1 ) (2) Where div1 represents the first-order symmetric divergence operator, τ is a non-negative constant; E * represents the adjoint matrix; b represents the number of iterations, b=1,2…; S b represents the sparse matrix updated in the bth iteration, S b+1 represents the sparse matrix updated at the b+1th iteration, λ b+1 represents the dual variable of the data acquisition operator updated in the b+1th iteration; p b+1 Represents the dual variable of the gradient updated at the b+1th iteration; where, through the projection operator To solve and update the dual variable p of the gradient, the dual variable q of the symmetric gradient and the dual variable λ of the data acquisition operator, The expression is: Where t is a set of infinite norms less than the number of shear transformation subbands; x represents any dual variable of the gradient p, the symmetric gradient q, and the data acquisition operator λ; The singular value threshold method is used to calculate and update the low-rank matrix through formula (4): Where: L b+1 The low-rank matrix updated in the b+1th iteration; L b The low-rank matrix updated in the bth iteration, r b+1 The rank of the low-rank matrix updated in the b+1th iteration; j represents the threshold; Contraction operator, Obtained through the singular value threshold operator SVT: Among them, UΣV T represents the singular value decomposition of the low-rank matrix, U, Σ, and V represent the submatrices generated after the singular value decomposition of the low-rank matrix, and V T represents the transpose of the matrix V; The calculation is repeated iteratively until the preset termination condition is reached or the number of iterations b reaches the preset value, and finally the optimal low-rank matrix and sparse matrix are obtained.
2. The dynamic magnetic resonance imaging method based on second-order generalized total variation and nuclear norm according to claim 1, characterized in that: The initial expression for constructing the second-order generalized total variation and nuclear norm model is: Where E is the data collection operator, L is the low-rank matrix, ‖L‖ * represents the nuclear norm of the low-rank matrix L, S is a sparse matrix, represents the second-order generalized total variation, represents the second-order generalized total variation of the sparse matrix S; d is the undersampled spatiotemporal data; β is the penalty parameter; α represents the weight coefficient.
3. The dynamic magnetic resonance imaging method based on second-order generalized total variation and nuclear norm according to claim 2, characterized in that: The data acquisition operator is mapped to a finite-dimensional Shilbit space, and the initial expression (5) of the second-order generalized total variation and nuclear norm model is simplified to: Where: U and V represent two different finite-dimensional space vectors, V = U 3 ; represents the gradient, ν represents the symmetric tensor, α1 represents the weight coefficient of the gradient term, α0 represents the weight coefficient of the symmetric gradient term, and E represents the symmetric gradient; || ||1 represents the L1 norm, which is the sum of the absolute values of each element in the matrix; Represents the square of the F-norm of the matrix, which is the square root of the sum of the squares of all elements in the matrix.
4. The dynamic magnetic resonance imaging method based on second-order generalized total variation and nuclear norm according to claim 1, characterized in that: The preset termination condition is: ts||K|| 2 <1 (7) Where σ is a non-negative constant; K is a matrix consisting of gradients and symmetric gradients, the identity matrix, and linear continuous operators in the low-rank matrix.
5. A dynamic magnetic resonance imaging device based on second-order generalized total variation and nuclear norm, characterized in that: The device comprises: An acquisition module, used for acquiring an original matrix of dynamic magnetic resonance imaging; A construction module, used for constraining the sparse matrix and the low-rank matrix of the original matrix by using the second-order generalized total variation and the nuclear norm, so as to construct a second-order generalized total variation and nuclear norm model; An iterative solution module, used for iteratively solving the second-order generalized total variation and nuclear norm model using a primal-dual algorithm to obtain an optimal low-rank matrix and a sparse matrix; A reconstruction image module is used to combine the obtained optimal low-rank matrix and sparse matrix to obtain a reconstructed image of dynamic magnetic resonance; The iterative solution of the second-order generalized total variation and nuclear norm model using the primal-dual algorithm comprises the following steps: Initialize the model parameters of the constructed second-order generalized total variation and nuclear norm models; constructing the primal-dual saddle point expression of the second-order generalized total variation and nuclear norm model using the primal-dual algorithm; Based on the primal-dual saddle point expression, updating the sparse matrix and the low-rank matrix by iteratively solving the dual variables and the singular value thresholds respectively; When the preset termination condition or the preset number of iterations is reached during the iterative solution process, the iteration is stopped, and the optimal low-rank matrix and sparse matrix are updated at this time; The primal-dual saddle point expression of the second-order generalized total variation and nuclear norm model constructed using the primal-dual algorithm is: Where Y is a finite-dimensional Shilbit space, p represents the dual variable of the gradient, q represents the dual variable of the symmetric gradient, λ represents the dual variable of the data acquisition operator, and <,> represents the inner product of two matrices. is the indicator function of a convex set,|||| ∞ represents the infinite norm; Wherein, based on the original dual saddle point expression, the sparse matrix and the low-rank matrix are updated by iteratively solving the dual variables and the singular value thresholds, respectively, including: The sparse matrix is updated by using the first-order symmetric divergence operator through formula (2): S b+1 =S b -τ(E * l b+1 -div1p b+1 ) (2) Where div1 represents the first-order symmetric divergence operator, τ is a non-negative constant; E * represents the adjoint matrix; b represents the number of iterations, b=1,2…; S b represents the sparse matrix updated in the bth iteration, S b+1 represents the sparse matrix updated at the b+1th iteration, λ b+1 represents the dual variable of the data acquisition operator updated in the b+1th iteration; p b+1 Represents the dual variable of the gradient updated at the b+1th iteration; where, through the projection operator To solve and update the dual variable p of the gradient, the dual variable q of the symmetric gradient and the dual variable λ of the data acquisition operator, The expression is: Where t is a set of infinite norms less than the number of shear transformation subbands; x represents any dual variable of the gradient p, the symmetric gradient q, and the data acquisition operator λ; The singular value threshold method is used to calculate and update the low-rank matrix through formula (4): Where: L b+1 The low-rank matrix updated in the b+1th iteration; L b The low-rank matrix updated in the bth iteration, r b+1 The rank of the low-rank matrix updated in the b+1th iteration; j represents the threshold; Contraction operator, Obtained through the singular value threshold operator SVT: Among them, UΣV T represents the singular value decomposition of the low-rank matrix, U, Σ, and V represent the submatrices generated after the singular value decomposition of the low-rank matrix, and V T represents the transpose of the matrix V; The calculation is repeated iteratively until the preset termination condition is reached or the number of iterations b reaches the preset value, and finally the optimal low-rank matrix and sparse matrix are obtained.
6. A dynamic magnetic resonance imaging device based on second-order generalized total variation and nuclear norm, characterized in that: including processor and storage medium; The storage medium is used to store instructions; The processor is configured to operate according to the instructions to execute the steps of the dynamic magnetic resonance imaging method based on second-order generalized total variation and nuclear norm according to any one of claims 1 to 4.
7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the dynamic magnetic resonance imaging method based on second-order generalized total variation and nuclear norm as claimed in any one of claims 1 to 4 are implemented.
