A high spatio-temporal resolution magnetic resonance metabolic imaging method and apparatus

Abstract

The present invention discloses a high spatio-temporal resolution magnetic resonance metabolic imaging method, including obtaining dynamic k-space metabolic imaging data ksp with complementary high-frequency information t ; and reconstructing the magnetic resonance metabolic imaging data to be finally solved by using the dynamic k-space metabolic imaging data with complementary high-frequency information and the chemical shift imaging data reconstruction model of multi-level and multi-dimensional similar problems. The present invention also discloses a high spatio-temporal resolution magnetic resonance metabolic imaging device, a storage medium, and a program product. The present invention can improve the undersampling ratio of magnetic resonance metabolic imaging, make full use of the prior information related to multiple levels of the spatial dimension and spectral dimension of dynamic metabolic imaging data, and greatly improve the spatio-temporal resolution of magnetic resonance metabolic imaging.

Description

Technical Field

[0001] The present invention relates to the field of magnetic resonance metabolic imaging, and in particular to a high spatio-temporal resolution magnetic resonance metabolic imaging method and device. Background Art

[0002] Magnetic resonance metabolic imaging is an important medical imaging technique that can simultaneously provide structural information and metabolic information of metabolites (such as glucose, lactate) in the tissue of interest, and has important value in clinical research. Since the metabolic duration is limited, a single metabolic imaging experiment often needs to collect multiple metabolic imaging data in the time dimension to obtain dynamic metabolic information. Therefore, it is of great significance to improve the spatio-temporal resolution of metabolic imaging data.

[0003] Currently, the gold standard for magnetic resonance metabolic imaging is chemical shift imaging (CSI). Since CSI performs phase encoding in the spatial direction and the spectral dimension is obtained through the evolution of chemical shift over time, CSI acquires three-dimensional data containing two-dimensional spatial distribution and one-dimensional spectral information. This three-dimensional characteristic enables CSI to simultaneously provide the structural information of the tissue and the distribution characteristics of metabolites. However, since CSI requires phase encoding in two directions, the acquisition process is time-consuming, which limits its application in dynamic metabolic research.

[0004] Compressed sensing is a technique that uses signal sparsity and irregular sampling to reconstruct high-quality data. Using the principle of compressed sensing, undersampling can be performed in the phase encoding dimension of CSI, and the information lost during the undersampling process can be recovered through algorithms such as sparse reconstruction, so as to improve the acquisition efficiency while ensuring the quality of metabolic imaging and obtain high spatio-temporal resolution metabolic information.

[0005] However, the existing technologies that introduce compressed sensing methods into CSI have the following problems: 1. A single undersampling phase encoding matrix is used for each collection of metabolic imaging data during the metabolic process. This approach limits the acquisition of high-frequency information, resulting in the loss of details in dynamic metabolic images and also limiting the undersampling ratio of the undersampling phase encoding matrix. 2. The metabolic imaging data are regarded as independent units for reconstruction, and the prior information of multi-level similarity in the image dimension (spatial dimension) and spectral dimension of adjacent dynamic metabolic imaging data is not fully utilized. Therefore, the existing technologies have limitations in the undersampling ratio and insufficient utilization of the prior information of dynamic metabolic imaging data, ultimately resulting in difficulty in significantly improving the spatio-temporal resolution.

[0006] Other high spatio-temporal resolution magnetic resonance metabolic imaging methods mainly include the following two categories, but they all have limitations in obtaining full-spectrum metabolic information, and the technical challenges they face restrict their practical applications: 1. Model-based fast metabolic imaging. This method uses prior knowledge such as chemical shift to reduce the number of echoes to be collected and speed up the acquisition speed. However, due to the constraints of multiple factors such as T2 relaxation and noise, the selection of the optimal echo spacing is extremely challenging.

[0007] 2. Metabolite-specific imaging. This method uses simultaneous spatial spectral excitation pulses to selectively excite only the required compounds, but this method faces two technical challenges: for metabolites with similar chemical shifts, its resolution ability is limited; at the same time, the technical implementation depends on a high-performance radiofrequency excitation system, which poses high requirements for hardware equipment. Summary of the Invention

[0008] In view of the problems existing in the prior art, such as low spatio-temporal resolution, limited undersampling ratio, and insufficient utilization of prior information of dynamic metabolic imaging data, the present invention proposes a high spatio-temporal resolution magnetic resonance metabolic imaging method and device for obtaining high spatio-temporal resolution magnetic resonance metabolic imaging data.

[0009] The above object of the present invention is achieved by the following means:

[0010] A high spatio-temporal resolution magnetic resonance metabolic imaging method includes:

[0011] Step 1: Obtain dynamic k-space metabolic imaging data ksp t , t = 1, 2,..., T, where T is the number of dynamic k-space metabolic imaging data,

[0012] Step 2: Use the dynamic k-space metabolic imaging data and the chemical shift imaging data reconstruction model to reconstruct and obtain the magnetic resonance metabolic imaging data to be solved for the final target.

[0013] As described above, the k-space metabolic imaging data ksp t is obtained by undersampling in the phase encoding dimension based on an undersampled phase encoding matrix. Data is collected in the central part of the undersampled phase encoding matrix, and data is randomly collected in the peripheral part of the undersampled phase encoding matrix. The randomly undersampled positions in the periphery of different dynamic undersampled phase encoding matrices are not exactly the same.

[0014] In Step 2 as described above, the k-space metabolic imaging data ksp t is input into the chemical shift imaging data reconstruction model. When the preset iteration stop condition is reached after cyclic iteration, the magnetic resonance metabolic imaging data to be solved for the final target is obtained.

[0015] As described above, the chemical shift imaging data reconstruction model is based on the following formula:

[0016] ,

[0017] where x t is the t-th dynamic magnetic resonance metabolic imaging data to be solved, t is the dynamic sequence number,

[0018] E is the undersampled Fourier transform operator,

[0019] ksp t is the k-space metabolic imaging data of the t-th dynamic,

[0020] ‖‖2 represents the L2 norm,

[0021] is the regularization parameter when traversing the window and merging the dynamic dimensions of the magnetic resonance metabolic imaging data x t in the i-th hierarchical manner, S is the total number of hierarchies,

[0022] is the nuclear norm of the feature matrix B in ,

[0023] i is the hierarchy sequence number, n is the image block sequence number, N i is the total number of image blocks obtained by dividing the magnetic resonance metabolic imaging data x t in the i-th hierarchy.

[0024] As described above, the traversing window and dynamic dimension merging of the magnetic resonance metabolic imaging data x t include:

[0025] Starting from the vertex of the magnetic resonance metabolic imaging data x t , traverse the magnetic resonance metabolic imaging data x t with the traversing window divided in the i-th hierarchy to obtain N i image blocks b n , and the dimension size of each image block b n is [j i , k i , l i , J and K are the image dimensions of the magnetic resonance metabolic imaging data x t , j i , k i are the image dimensions of the image block corresponding to the i-th hierarchy, L is the spectral dimension of the magnetic resonance metabolic imaging data x t , l i is the spectral dimension of the image block corresponding to the i-th hierarchy, and the image block b nThe rows of the image dimension planes corresponding to the two image dimensions are concatenated in sequence to obtain a one-dimensional vector of the image dimension plane, and then the one-dimensional vectors of the image dimension planes corresponding to the respective spectral dimensions of the image block are concatenated in sequence to obtain a one-dimensional vector of the image block. According to the dynamic order, the respective dynamic one-dimensional vectors of the image blocks at the same position are concatenated to obtain the feature matrix B in 。

[0026] A high spatio-temporal resolution magnetic resonance metabolic imaging method further includes: for each magnetic resonance metabolic imaging data x t corresponding feature matrix B in perform SVD decomposition. After SVD decomposition of each B in matrix, a series of singular values are obtained correspondingly , where the largest singular value after SVD decomposition of each B in matrix is denoted as the corresponding largest singular value , retain the singular values greater than in after SVD decomposition of each feature matrix B × and perform SVD reconstruction to obtain the matrix P in , the matrices P t corresponding to the respective magnetic resonance metabolic imaging data x in are obtained by the inverse operation of traversing window and dynamic dimension merging to obtain the magnetic resonance metabolic imaging matrix L with low-rank constraint t , calculate the updated magnetic resonance metabolic imaging data x t 1 ,

[0027] ,

[0028] ,

[0029] ,

[0030] wherein, is the k-space domain residual return to the magnetic resonance metabolic imaging domain to obtain the image domain residual, is the undersampled Fourier inverse transform operator, is the k-space domain residual, and E is the undersampled Fourier transform operator.

[0031] As described above, the preset iteration stop condition is to reach the iteration number, or the value of is less than the set threshold .

[0032] A high spatio-temporal resolution magnetic resonance metabolic imaging device includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the above-mentioned steps 1 and 2 are implemented.

[0033] A computer-readable storage medium stores a computer program thereon. When the computer program is executed by a processor, the above-mentioned steps 1 and 2 are implemented.

[0034] A computer program product includes a computer program. When the computer program is executed by a processor, the above-mentioned steps 1 and 2 are implemented.

[0035] Compared with the prior art, the present invention has the following advantages:

[0036] 1. In the present invention, the central part of the undersampled phase encoding matrix is used for data acquisition, and the peripheral part of the undersampled phase encoding matrix is randomly sampled. The randomly undersampled positions of the peripheries of the undersampled phase encoding matrices with different dynamics are not exactly the same, which can further increase the undersampling ratio and thus improve the spatio-temporal resolution of metabolic imaging. When the above sampling method is not adopted, all the undersampled phase encoding matrices are the same, and the high-frequency information collected at adjacent time points cannot complement each other. In the present invention, different undersampled phase encoding matrices are used for each dynamic, and the high-frequency information collected by each undersampled phase encoding matrix is complementary and has dynamic correlation. Therefore, the undersampling ratio can be further increased, and a high spatio-temporal resolution metabolic image can still be obtained by reconstructing using the chemical shift imaging data reconstruction model for multi-level and multi-dimensional similar problems.

[0037] 2. Reconstructing data using the chemical shift imaging data reconstruction model for multi-level and multi-dimensional similar problems can make full use of the prior information of multi-level similarity in the image dimension (spatial dimension) and spectral dimension of adjacent dynamic metabolic imaging data, and further improve the spatio-temporal resolution.

[0038] 3. This method is based on the traditional chemical shift imaging sequence CSI, has low requirements for hardware, is not easily affected by factors such as magnetic field inhomogeneity, has strong stability, and is more easily developed for practical applications; it can obtain full-spectrum metabolic information and is suitable for exploratory metabolic research. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 is a schematic flowchart of the present invention;

[0040] Figure 2 is a collection diagram of k-space metabolic imaging data ksp of dynamics with complementary high-frequency information t ;

[0041] Figure 3Attention tissue map of the mouse attention tissue metabolism model in Example 1;

[0042] Figure 4 Comparison diagram between the target magnetic resonance metabolism imaging data map of the present invention and the CSI magnetic resonance metabolism image in Example 1. Detailed implementation manners

[0043] To facilitate the understanding and implementation of the present invention by those of ordinary skill in the art, the present invention will be further described in detail below in conjunction with embodiments. It should be understood that the embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0044] Example 1

[0045] In this example, high spatio-temporal resolution magnetic resonance metabolism imaging is performed on a mouse attention tissue metabolism model. A high spatio-temporal resolution magnetic resonance metabolism imaging method includes:

[0046] Step 1: Obtain dynamic k-space metabolism imaging data ksp with complementary high-frequency information at a high under-sampling ratio, where t = 1, 2, …, T, and T is the number of dynamic k-space metabolism imaging data. t , t = 1, 2, …, T, and T is the number of dynamic k-space metabolism imaging data.

[0047] The high-frequency information of the dynamic k-space metabolism imaging data kspt, t = 1, 2, …, T, can be complementary, which is mainly based on the k-space characteristics of chemical shift imaging CSI. The data in the outer part of the k-space of chemical shift imaging CSI contains high-frequency information, and the high-frequency information determines the details and resolution of the magnetic resonance metabolism imaging data. By using an under-sampling phase encoding matrix with the same under-sampling ratio in different dynamics and different randomly under-sampled positions in the periphery, the k-space metabolism imaging data ksp collected in each dynamic t can contain different high-frequency information. According to the characteristics of dynamic magnetic resonance metabolism imaging, the k-space metabolism imaging data ksp collected in adjacent dynamics t has a certain similarity in high-frequency information. This similarity enables the high-frequency information of the k-space metabolism imaging data between different dynamics to be mutually complementary after reconstruction using a chemical shift imaging data reconstruction model for multi-level and multi-dimensional similar problems, not only effectively improving the reconstruction quality at a high under-sampling ratio, but also maximizing the retention of the spatial details and resolution of the magnetic resonance metabolism imaging data.

[0048] In the mouse attention tissue metabolism model, T = 30 dynamic k-space metabolism imaging data are required, where T is the number of dynamic k-space metabolism imaging data. Then, 30 different under-sampling phase encoding matrices (with the same under-sampling ratio and different randomly under-sampled positions in the periphery) need to be generated as mask t(t = 1, 2, ……, T), mask t represents the undersampled phase encoding matrix under the t-th dynamic state.

[0049] The k-space of chemical shift imaging CSI has three dimensions. Phase encoding is performed in two dimensions, and chemical shift frequency encoding is performed in the other dimension. The undersampled phase encoding matrix of the present invention performs undersampling in the dimension of phase encoding.

[0050] The undersampled phase encoding matrix consists of 0s and 1s. 0 represents not collecting the k-space data of chemical shift imaging CSI at the phase encoding position corresponding to this image dimension (spatial dimension), and 1 represents collecting the k-space data of chemical shift imaging CSI at the phase encoding position corresponding to this image dimension (spatial dimension). The values of the central part of the undersampled phase encoding matrix (accounting for about one-tenth of the entire undersampled phase encoding matrix) are all 1, that is, data collection is performed on the central part of the undersampled phase encoding matrix. The values of the peripheral part (the part except the central part) of the undersampled phase encoding matrix are randomly 0, that is, data collection is randomly performed on the peripheral part of the undersampled phase encoding matrix.

[0051] According to the undersampled phase encoding matrix, the data of the central part of the k-space of chemical shift imaging CSI are all collected, while the data of the peripheral part of the k-space of chemical shift imaging CSI are randomly collected. And the randomly undersampled positions of the undersampled phase encoding matrices used in different dynamic experiments are not exactly the same.

[0052] Calculation of undersampling ratio: The number of 1s in the undersampled phase encoding matrix is m1, the size of the undersampled phase encoding matrix is M×N, and M and N are the sizes of the undersampled phase encoding matrix in the two phase encoding dimensions respectively. Then the undersampling ratio = M×N / m1. In this embodiment, the undersampling ratio is 9.

[0053] According to the order of the undersampled phase encoding matrix, collect the k-space data of chemical shift imaging CSI at the phase encoding positions where the value in the undersampled phase encoding matrix is 1, and obtain the k-space metabolic imaging data ksp with an undersampling ratio of 9 t (t = 1, 2, ……, T), ksp t represents the k-space metabolic imaging data under the t-th dynamic state.

[0054] The order of the undersampled phase encoding matrix is carried out sequentially from mask t (t = 1, 2, ……, T). The phase encoding acquisition order in each mask t is first from left to right, and then from top to bottom.

[0055] The undersampled phase encoding matrix mask under the t-th dynamic state tThe data collected in the interior is placed into the k-space according to the corresponding phase encoding positions to obtain the k-space metabolic imaging data ksp at the t-th dynamic state. t 。

[0056] Step 2: Reconstruct the final target magnetic resonance metabolic imaging data by using the dynamic k-space metabolic imaging data with complementary high-frequency information and the chemical shift imaging data reconstruction model for multi-level and multi-dimensional similarity problems.

[0057] The chemical shift imaging data reconstruction model for multi-level and multi-dimensional similarity problems performs operations of S-level partition traversal windows and dynamic dimension merging on the image dimension (spatial dimension) and spectral dimension of the magnetic resonance metabolic imaging data to obtain the feature matrix B in (i = 1, 2, ……, S; n = 1, 2, ……, N i ), and constrains its low rank to make the most of the prior information of multi-level similarity of adjacent dynamic magnetic resonance metabolic imaging data in the image dimension (spatial dimension) and spectral dimension, and further improve the spatio-temporal resolution.

[0058] (1) The chemical shift imaging data reconstruction model for multi-level and multi-dimensional similarity problems (Equation 1) can find the magnetic resonance metabolic imaging data x t (t = 1, 2, ……, T):

[0059] (1)

[0060] Where

[0061] x t is the magnetic resonance metabolic imaging data of the t-th dynamic state to be solved, and t is the dynamic serial number.

[0062] E is the undersampled Fourier transform operator.

[0063] ksp t is the k-space metabolic imaging data of the t-th dynamic state, t = 1, 2, ……, T.

[0064] ‖‖2 represents the L2 norm.

[0065] is the regularization parameter for traversing the window and merging the dynamic dimensions of the magnetic resonance metabolic imaging data x i in the i-th level S t way, generally using empirical values, and S is the total number of levels.

[0066] S is the total number of levels. The traversing window and dynamic dimension merging (S-level partition) of the magnetic resonance metabolic imaging data x t of the t-th dynamic state includes the following steps:

[0067] The t-th dynamic magnetic resonance metabolic imaging data x t is a three-dimensional data, and the sizes of the three dimensions are denoted as J, K, and L respectively. For a magnetic resonance metabolic imaging data x of size [J, K, L] t perform the i-th level S i = [j i , k i , l i , where J and K are the image dimensions of the magnetic resonance metabolic imaging data x t , j i , k i are the image dimensions of the image block corresponding to the i-th level, and L is the spectral dimension of the magnetic resonance metabolic imaging data x t , l i is the spectral dimension of the image block corresponding to the i-th level: that is, starting from the vertex of the magnetic resonance metabolic imaging data x t , with the size j i , k i , l i divided at the i-th level as the three-dimensional size of the traversal window, use the traversal window to traverse the magnetic resonance metabolic imaging data x t to obtain N i image blocks, n is the image block serial number, and N i is the total number of image blocks obtained by dividing the magnetic resonance metabolic imaging data x t at the i-th level. The dimension (two image dimensions and one spectral dimension) of each image block b n is [j i , k i , l i . Concatenate the rows of the image dimension plane where the two image dimensions of the image block b n are located in sequence to obtain a one-dimensional vector of the image dimension plane, and then concatenate the one-dimensional vectors of the image dimension planes corresponding to the spectral dimensions of each image block in sequence to obtain a one-dimensional vector of the image block. The size of the one-dimensional vector of the image block is j i ×k i ×l i . Concatenate the one-dimensional vectors of the dynamic image blocks corresponding to the image blocks in the same position in dynamic order to obtain the feature matrix B in . At this time, the matrix size of the obtained feature matrix B in is [j i ×k i ×l i , T]. is the nuclear norm of the feature matrix B in , which is equal to the sum of all singular values of the feature matrix B in . Through the nuclear norm That is, the multi-level and multi-dimensional similar prior information of the dynamic k-space metabolic imaging data, and the complementary of the dynamic high-frequency information is completed through the similarity constraint.

[0068] Magnetic resonance metabolic imaging data x t is a three-dimensional data, where the first two dimensions represent the image dimensions and the third dimension is the spectral dimension. According to the characteristics of chemical shift imaging: within a certain range near the spectral peak of a compound, the chemical shift images are similar. Therefore, according to the full width at half maximum (FWHM) of the spectral peak and the spectral resolution, the size of the hierarchical division in the spectral dimension is adjusted during hierarchical division.

[0069] For example: if the FWHM of the spectral peak is 90 hz and the spectral resolution is 15 hz, then the block size in the spectral dimension is: FWHM of the spectral peak / spectral resolution = 6.

[0070] (2) Substitute the known data into Equation (1) and solve it iteratively.

[0071] The target magnetic resonance metabolic imaging data to be solved is to input the k-space metabolic imaging data ksp t (t = 1, 2, ……, T) into the chemical shift imaging data reconstruction model formula (1) of the multi-level and multi-dimensional similar problem, and after cyclic iteration, when the preset iteration stop condition is reached (the number of iterations is reached, or less than the set threshold ), the target magnetic resonance metabolic imaging data x is obtained.

[0072] The reconstruction iteration process is as follows:

[0073] The initial magnetic resonance metabolic imaging data x t (t = 1, 2, ……, T) can be directly obtained by directly performing Fourier transform on the k-space metabolic imaging data ksp t (t = 1, 2, ……, T) after filling it with zeros.

[0074] Perform the above operations of traversing the window and dynamic dimension merging on the magnetic resonance metabolic imaging data x t (t = 1, 2, ……, T) to obtain the feature matrix B in , and perform singular value soft threshold processing on the feature matrix B t corresponding to each magnetic resonance metabolic imaging data x in , that is, first perform SVD decomposition on the feature matrix B t corresponding to each magnetic resonance metabolic imaging data x in . After SVD decomposition of each B in matrix, a series of singular values are obtained , where the largest singular value after SVD decomposition of each B in matrix is denoted as the corresponding largest singular value , retain each feature matrix B in After performing SVD decomposition, the values greater than × of the singular values are used for SVD reconstruction to obtain matrix P in , for each magnetic resonance metabolic imaging data x t corresponding feature matrix B in After performing the above operations on all of them, for each magnetic resonance metabolic imaging data x t corresponding P in matrices, through the inverse operation of traversing window and dynamic dimension merging, the low-rank constrained magnetic resonance metabolic imaging matrix L t (t = 1, 2, ……, T) is obtained.

[0075] Calculate the k-space domain residual for each dynamic (t = 1, 2, ……, T):

[0076] , where E is the undersampled Fourier transform operator.

[0077] Return the k-space domain residual (t = 1, 2, ……, T) to the magnetic resonance metabolic imaging domain to obtain the image domain residual (t = 1, 2, ……, T):

[0078] ,

[0079] is the undersampled inverse Fourier transform operator.

[0080] Then the updated magnetic resonance metabolic imaging data x t 1 (t = 1, 2, ……, T) is:

[0081] ,

[0082] Calculate and judge whether the value of is less than the set threshold , if the value of is less than, stop the iteration, and the updated magnetic resonance metabolic imaging data (t = 1, 2, ……, T) is the magnetic resonance metabolic imaging data for the desired final target solution. If the value of is greater than or equal to the threshold then use the updated magnetic resonance metabolic imaging data as the new magnetic resonance metabolic imaging data x t , and continue the iterative update according to the above process until is less than , or the number of iterations is greater than the set number of iterations q, and the finally updated magnetic resonance metabolic imaging data The magnetic resonance metabolic imaging data solved as the final target, and the set number of iterations q can be set according to actual needs. The value can be set according to actual needs.

[0083] Figure 3 is the magnetic resonance structural image of the mouse attention tissue metabolism model.

[0084] Figure 4 This is the magnetic resonance metabolic imaging comparison chart of the same mouse attention tissue model between the present invention and chemical shift imaging CSI. It can be clearly seen that the present invention improves the spatio-temporal resolution of magnetic resonance metabolic imaging by about 5 times compared with CSI (the spatio-temporal resolution of CSI is 3.5mm×3.5mm×6mm, one dynamic every 10 minutes; the spatio-temporal resolution of the present invention is 2.1mm×2.1mm×6mm, one dynamic every 6 minutes). The magnetic resonance metabolic image obtained by this method fits better with the structural image of the attention tissue.

[0085] Those of ordinary skill in the art can understand that to implement steps 1 and 2 in the above embodiments, it can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processing flows of the above respective steps.

[0086] Embodiment 2

[0087] In one embodiment, a high spatio-temporal resolution magnetic resonance metabolic imaging device is further provided, including a memory and a processor. A computer program is stored in the memory, and when the processor executes the computer program, it implements the processing flows of steps 1 and 2 in the above Embodiment 1.

[0088] Embodiment 3

[0089] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, it implements the processing flows of steps 1 and 2 in the above Embodiment 1.

[0090] Embodiment 4

[0091] In one embodiment, a computer program product is provided, including a computer program. When the computer program is executed by a processor, it implements the processing flows of steps 1 and 2 in the above Embodiment 1.

[0092] It should be noted that the embodiments described in the present invention are only illustrative of the spirit of the present invention. Those skilled in the art to which the present invention pertains may make various modifications or supplements to the described embodiments or use similar means for substitution, but will not deviate from the spirit of the present invention or exceed the scope defined by the appended claims.

Claims

1. A high spatio-temporal resolution magnetic resonance metabolic imaging method, characterized in that, including: Step 1: Obtain dynamic k-space metabolic imaging data ksp t , where t = 1, 2, …, T, and T is the number of dynamic k-space metabolic imaging data Step 2: Reconstruct the magnetic resonance metabolic imaging data to be finally solved by using the dynamic k-space metabolic imaging data and the chemical shift imaging data reconstruction model, In the step 2, the k-space metabolic imaging data ksp t is input into the chemical shift imaging data reconstruction model. When the preset iteration stop condition is reached after cyclic iteration, the magnetic resonance metabolic imaging data to be finally solved is obtained. The chemical shift imaging data reconstruction model is based on the following formula: ， where x t is the t-th dynamic magnetic resonance metabolic imaging data to be solved for the target, and t is the dynamic sequence number E is an undersampled Fourier transform operator, ksp t is the k-space metabolic imaging data for the t-th dynamic ‖‖2 represents the L2 norm, is the regularization parameter when traversing windows and merging dynamic dimensions for magnetic resonance metabolic imaging data x in the i-th hierarchical manner, and S is the total number of hierarchies, t ​ is the nuclear norm of the feature matrix B in and i is the level serial number, n is the image block serial number, and N i is the total number of image blocks obtained by partitioning the magnetic resonance metabolic imaging data x t at the i-th level.

2. The high spatio-temporal resolution magnetic resonance metabolic imaging method according to claim 1, wherein, The k-space metabolic imaging data ksp t It is obtained by undersampling in the phase encoding dimension based on an undersampled phase encoding matrix. Data is collected from the central part of the undersampled phase encoding matrix, and data is randomly collected from the peripheral part of the undersampled phase encoding matrix. The randomly undersampled positions in the periphery of the undersampled phase encoding matrix with different dynamics are not exactly the same.

3. The high spatio-temporal resolution magnetic resonance metabolic imaging method according to claim 1, characterized in that The traversal window and dynamic dimension merging of the magnetic resonance metabolic imaging data x t are performed as follows: Starting from the vertex of the magnetic resonance metabolic imaging data x t and traversing the magnetic resonance metabolic imaging data x t with a traversal window divided at the i-th level, N i image patches b n are obtained. The dimension size of each image patch b n is [j i , k i , l i . J and K are the image dimensions of the magnetic resonance metabolic imaging data x t . j i and k i are the image dimensions of the image patch corresponding to the i-th level. L is the spectral dimension of the magnetic resonance metabolic imaging data x t . l i is the spectral dimension of the image patch corresponding to the i-th level. The rows in the image dimension plane where the two image dimensions of the image patch b n are located are concatenated in sequence to obtain a one-dimensional vector of the image dimension plane. Then, the one-dimensional vectors of the image dimension planes corresponding to the respective spectral dimensions of the image patch are concatenated in sequence to obtain a one-dimensional vector of the image patch. According to the dynamic order, the one-dimensional vectors of the respective dynamic image patches corresponding to the image patches at the same position are concatenated to obtain the feature matrix B in .

4. A high spatio-temporal resolution magnetic resonance metabolic imaging method according to claim 3, characterized in that, further including: For each magnetic resonance metabolic imaging data x t The corresponding feature matrix B in Perform SVD decomposition on each B in After the SVD decomposition of each B matrix, a series of singular values are obtained , where for each B in The largest singular value after the SVD decomposition of the matrix is denoted as the corresponding largest singular value Retain each feature matrix B in After performing SVD decomposition, the values greater than × of the singular values and perform SVD reconstruction to obtain the matrix P in For each magnetic resonance metabolic imaging data x t The corresponding matrix P in According to the reverse operation of traversing window and dynamic dimension merging, obtain the low-rank constrained magnetic resonance metabolic imaging matrix L t Calculate the updated magnetic resonance metabolic imaging data x t 1 , ， ， ， Among them, is the residual in the k-space domain returns to the magnetic resonance metabolic imaging domain to obtain the image domain residual, is the undersampled Fourier inverse transform operator, is the residual in the k-space domain, and E is the undersampled Fourier transform operator.

5. A high spatio-temporal resolution magnetic resonance metabolic imaging method according to claim 4, characterized in that The preset iteration stop condition is to reach the number of iterations, or the value of is less than the set threshold .

6. A high spatio-temporal resolution magnetic resonance metabolic imaging device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, it implements step 1 and step 2 described in any one of claims 1 to 5.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements step 1 and step 2 described in any one of claims 1 to 5.

8. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements step 1 and step 2 described in any one of claims 1 to 5.

Images