CN110766768A - Magnetic resonance image reconstruction method, device, equipment and medium - Google Patents
Magnetic resonance image reconstruction method, device, equipment and medium Download PDFInfo
- Publication number
- CN110766768A CN110766768A CN201911011736.8A CN201911011736A CN110766768A CN 110766768 A CN110766768 A CN 110766768A CN 201911011736 A CN201911011736 A CN 201911011736A CN 110766768 A CN110766768 A CN 110766768A
- Authority
- CN
- China
- Prior art keywords
- data
- sub
- magnetic resonance
- image reconstruction
- solution
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 72
- 238000012549 training Methods 0.000 claims abstract description 39
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 32
- 238000013178 mathematical model Methods 0.000 claims abstract description 29
- 230000006870 function Effects 0.000 claims description 44
- 238000013527 convolutional neural network Methods 0.000 claims description 23
- 238000005457 optimization Methods 0.000 claims description 10
- 238000013528 artificial neural network Methods 0.000 claims description 9
- 238000004364 calculation method Methods 0.000 claims description 8
- 238000011478 gradient descent method Methods 0.000 claims description 7
- 238000013499 data model Methods 0.000 claims description 6
- 238000004590 computer program Methods 0.000 claims description 4
- 238000000354 decomposition reaction Methods 0.000 claims description 3
- 239000000284 extract Substances 0.000 claims description 3
- 238000005070 sampling Methods 0.000 abstract description 23
- 238000010586 diagram Methods 0.000 description 6
- 230000003287 optical effect Effects 0.000 description 6
- 238000012545 processing Methods 0.000 description 5
- 238000003384 imaging method Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 238000002595 magnetic resonance imaging Methods 0.000 description 3
- 230000001133 acceleration Effects 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 230000002093 peripheral effect Effects 0.000 description 2
- 230000000644 propagated effect Effects 0.000 description 2
- 102100029114 Fatty-acid amide hydrolase 2 Human genes 0.000 description 1
- 101000918490 Homo sapiens Fatty-acid amide hydrolase 2 Proteins 0.000 description 1
- 208000036266 acromesomelic dysplasia 3 Diseases 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000003759 clinical diagnosis Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000002059 diagnostic imaging Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000000835 fiber Substances 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 231100000957 no side effect Toxicity 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 230000008707 rearrangement Effects 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T11/00—2D [Two Dimensional] image generation
- G06T11/003—Reconstruction from projections, e.g. tomography
- G06T11/005—Specific pre-processing for tomographic reconstruction, e.g. calibration, source positioning, rebinning, scatter correction, retrospective gating
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/30—Assessment of water resources
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Magnetic Resonance Imaging Apparatus (AREA)
Abstract
The embodiment of the invention discloses a magnetic resonance image reconstruction method, a device, equipment and a medium. The method comprises the following steps: acquiring undersampled magnetic resonance data; and inputting the magnetic resonance data into an image reconstruction model based on an alternating direction multiplier algorithm to obtain a reconstructed target magnetic resonance image, wherein the image reconstruction model is a result of pre-training a mathematical model of which both a data fidelity term and a regular term are indefinite terms. The embodiment of the invention solves the problems that in the process of reconstructing a magnetic resonance image, a data fidelity item in a model applicable to an ADMM-net method is represented by a 2 norm between a reconstructed k space and a sampling point, is not the most effective data consistency guarantee method and is not applicable to all image reconstruction conditions; the image quality after the image reconstruction based on the ADMM algorithm can be improved.
Description
Technical Field
The embodiments of the present invention relate to medical imaging technologies, and in particular, to a magnetic resonance image reconstruction method, apparatus, device, and medium.
Background
Magnetic resonance images human tissue using static and radio frequency magnetic fields, which not only provides rich tissue contrast, but also has no side effects on the human body, thus becoming a powerful tool for medical clinical diagnosis.
In order to improve the magnetic resonance imaging speed and the imaging quality, a deep learning method is mostly adopted to reconstruct images, for example, a neural network is utilized to learn the optimal parameters required by reconstruction from a large amount of training data or directly learn the mapping relation from under-acquired data to fully-acquired images, so that the imaging quality and the acceleration multiple which are better than those of the traditional parallel imaging or compressed sensing method are obtained.
The ADMM algorithm, namely an alternative direction multiplier method, is a calculation framework for solving an optimization problem, and is suitable for solving a distributed convex optimization problem. The ADMM algorithm decomposes a large global problem into a plurality of smaller, more easily solved local sub-problems by a Decomposition-Coordination (decomplexing-Coordination) process, and obtains a solution to the large global problem by coordinating the solutions of the sub-problems. The ADMM-net method combining deep learning and the ADMM algorithm learns the parameters in the algorithm by adopting a deep neural network, and solves the problems of difficult parameter adjustment and long iteration time in the optimization problem.
However, the data fidelity term in the model to which the ADMM-net method is applied is characterized by using the 2 norm between the reconstructed k space and the sampling point, and this least square constraint is established on the premise of linear unbiased estimation, and is not the most effective data consistency guarantee method, and is not applicable to all image reconstruction situations, so that the imaging quality of the reconstructed image still needs to be improved.
Disclosure of Invention
The embodiment of the invention provides a magnetic resonance image reconstruction method, a magnetic resonance image reconstruction device, magnetic resonance image reconstruction equipment and a magnetic resonance image reconstruction medium, so that the network freedom degree of a neural network is improved, more prior information is learned, and the image quality is improved.
In a first aspect, an embodiment of the present invention provides a magnetic resonance image reconstruction method, including:
acquiring undersampled magnetic resonance data;
and inputting the magnetic resonance data into an image reconstruction model based on an alternating direction multiplier algorithm to obtain a reconstructed target magnetic resonance image, wherein the image reconstruction model is a result of pre-training a mathematical model of which both a data fidelity term and a regular term are indefinite terms.
Optionally, the process of training the image reconstruction model includes:
acquiring fully sampled magnetic resonance data, and extracting at least one group of undersampled data from the fully sampled magnetic resonance data to obtain at least one group of data pairs of the undersampled data and the fully sampled magnetic resonance data;
inputting the undersampled data into a mathematical model of which the data fidelity term and the regular term are both indefinite terms;
decomposing a mathematical model in which the data fidelity term and the regular term are both indefinite terms into a first subproblem, a second subproblem and a third subproblem based on an alternating direction multiplier algorithm, wherein the third subproblem is a constraint condition of a solution of the first subproblem and the second subproblem;
solving the first sub-problem and the second sub-problem by adopting a gradient descent method;
and aiming at the solution of the first sub-problem and the solution of the second sub-problem, determining each parameter value in the solution of the first sub-problem and the solution of the second sub-problem through a convolutional neural network iterative computation method, and finishing the training of the image reconstruction model.
Optionally, the determining, by a convolutional neural network iterative computation method, parameter values in the solution of the first sub-problem and the solution of the second sub-problem includes:
fitting a first-order partial derivative function of a data fidelity term function in the solution of the first sub-problem and a first-order partial derivative function of a data regular term function in the solution of the second sub-problem by adopting a convolutional neural network, wherein the initial value of each parameter in the solution of the first sub-problem and the solution of the second sub-problem is an empirical value;
and determining the numerical value of each parameter in the solution of the first subproblem and the solution of the second subproblem through preset iteration times until the difference value between the reconstructed image obtained through the data model and the reconstructed image corresponding to the fully sampled magnetic resonance data meets the loss function.
Optionally, in each iterative computation, the neural network structure includes three modules, namely a reconstruction layer, an optimization layer, and a parameter update layer.
In a second aspect, an embodiment of the present invention further provides a magnetic resonance image reconstruction apparatus, including:
a data acquisition module for acquiring undersampled magnetic resonance data;
and the image reconstruction module is used for inputting the magnetic resonance data into an image reconstruction model based on an alternating direction multiplier algorithm so as to obtain a reconstructed target magnetic resonance image, wherein the image reconstruction model is a result of pre-training a mathematical model of which both a data fidelity term and a regular term are indefinite terms.
Optionally, the apparatus further includes a model training module, configured to train the image reconstruction model; the model training module specifically comprises:
the sample data acquisition submodule acquires fully sampled magnetic resonance data, extracts at least one group of undersampled data from the fully sampled magnetic resonance data, and obtains at least one group of data pairs of the undersampled data and the fully sampled magnetic resonance data;
the sample input submodule is used for inputting the undersampled data into a mathematical model of which the data fidelity term and the regular term are both indefinite terms;
the decomposition calculation sub-module is used for decomposing a mathematical model of which the data fidelity term and the regular term are both indefinite terms into a first sub-problem, a second sub-problem and a third sub-problem based on an alternating direction multiplier algorithm, wherein the third sub-problem is a constraint condition of the solutions of the first sub-problem and the second sub-problem;
a subproblem solving submodule for solving the first subproblem and the second subproblem by using a gradient descent method;
and the parameter solving submodule is used for determining parameter values in the solutions of the first sub-problem and the second sub-problem through a convolutional neural network iterative computation method aiming at the solutions of the first sub-problem and the second sub-problem, and finishing the training of the image reconstruction model.
Optionally, the parameter solving submodule is specifically configured to:
fitting a first-order partial derivative function of a data fidelity term function in the solution of the first sub-problem and a first-order partial derivative function of a data regular term function in the solution of the second sub-problem by adopting a convolutional neural network, wherein the initial value of each parameter in the solution of the first sub-problem and the solution of the second sub-problem is an empirical value;
and determining the numerical value of each parameter in the solution of the first subproblem and the solution of the second subproblem through preset iteration times until the difference value between the reconstructed image obtained through the data model and the reconstructed image corresponding to the fully sampled magnetic resonance data meets the loss function.
Optionally, in each iterative computation, the neural network structure includes three modules, namely a reconstruction layer, an optimization layer, and a parameter update layer.
In a third aspect, an embodiment of the present invention further provides a computer device, where the computer device includes:
one or more processors;
a memory for storing one or more programs;
when executed by the one or more processors, cause the one or more processors to implement a magnetic resonance image reconstruction method as provided by any of the embodiments of the invention.
In a fourth aspect, embodiments of the present invention further provide a computer-readable storage medium, on which a computer program is stored, which when executed by a processor, implements a magnetic resonance image reconstruction method as provided in any of the embodiments of the present invention.
The embodiment of the invention inputs undersampled magnetic resonance data into an image reconstruction model based on an alternating direction multiplier algorithm, wherein the image reconstruction model is a result of pre-training a mathematical model with uncertain data fidelity terms and regular terms, so as to obtain a reconstructed target magnetic resonance image, and the problem that the data fidelity terms in the model suitable for the ADMM-net method are represented by 2 norms between a reconstructed k space and sampling points, are not the most effective data consistency guarantee method and are not suitable for all image reconstruction conditions is solved; the image quality after the image reconstruction based on the ADMM algorithm can be improved.
Drawings
Fig. 1 is a flowchart of a magnetic resonance image reconstruction method according to a first embodiment of the present invention;
FIG. 2a is a flowchart of an image reconstruction model training method according to a second embodiment of the present invention;
FIG. 2b is a schematic diagram of a convolutional neural network structure according to a second embodiment of the present invention;
FIG. 3 is a comparison graph of reconstruction effects of image reconstruction using the trained image reconstruction model and image reconstruction using other algorithms in the second embodiment of the present invention;
fig. 4 is a schematic structural diagram of a magnetic resonance image reconstruction apparatus according to a third embodiment of the present invention;
fig. 5 is a schematic structural diagram of a computer device in the fourth embodiment of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings and examples. It is to be understood that the specific embodiments described herein are merely illustrative of the invention and are not limiting of the invention. It should be further noted that, for the convenience of description, only some of the structures related to the present invention are shown in the drawings, not all of the structures.
Example one
Fig. 1 is a flowchart of a magnetic resonance image reconstruction method according to an embodiment of the present invention, which is applicable to medical image reconstruction.
As shown in fig. 1, the magnetic resonance image reconstruction method specifically includes the following steps:
and S110, acquiring undersampled magnetic resonance data.
Specifically, the undersampled magnetic resonance data is undersampled magnetic resonance K-space data obtained by scanning the magnetic resonance imaging apparatus in a preset scanning manner.
The K space is also called fourier space, and is a filling space of magnetic resonance signal original data with space positioning coding information, and each magnetic resonance image has its corresponding K space data lattice. The undersampled K-space data is not the data of all the sampling points, so that the time for sampling the data can be reduced.
And S120, inputting the magnetic resonance data into an image reconstruction model based on an alternating direction multiplier algorithm to obtain a reconstructed target magnetic resonance image, wherein the image reconstruction model is a result of pre-training a mathematical model of which both a data fidelity term and a regular term are indefinite terms.
And inputting the acquired undersampled K space data into a pre-trained image reconstruction model, wherein the output of the model is a reconstructed target image.
In the training process of the model, undersampled magnetic resonance data and corresponding full-sampling magnetic resonance data are used as a data sample pair, the undersampled magnetic resonance data are input into a mathematical model of which both a fidelity term and a regular term of the data to be trained are indefinite terms, and image reconstruction model parameters which can meet the image quality requirement are determined through iterative calculation of a convolutional neural network, so that the trained image reconstruction model is obtained. It can be understood that, in the process of model training, the training sample includes a plurality of data sample pairs, each sample pair may be a plurality of undersampled magnetic resonance data rearranged from a set of fully sampled magnetic resonance data according to a preset rule, and each undersampled magnetic resonance data and the set of fully sampled magnetic resonance data form a data sample pair. Alternatively, the plurality of sets of fully sampled magnetic resonance data may correspond to one set of undersampled magnetic resonance data, respectively, so as to obtain a plurality of sample data pairs.
Specifically, meeting the image quality requirement means that the difference between two reconstructed images reaches a minimum value compared with a reconstructed image obtained by inputting undersampled magnetic resonance data into an image reconstruction model and performing image reconstruction by using fully-sampled magnetic resonance data corresponding to the undersampled magnetic resonance data. After learning through the neural network, an image reconstruction model meeting the conditions can be obtained, and the model training process is completed.
According to the technical scheme, undersampled magnetic resonance data are input into an image reconstruction model based on an alternating direction multiplier algorithm, the image reconstruction model is a result of pre-training of a mathematical model with uncertain data fidelity terms and regular terms, and a reconstructed target magnetic resonance image is obtained, so that the problems that the data fidelity terms in the model suitable for the ADMM-net method are represented by 2 norms between a reconstructed k space and sampling points, the data fidelity terms are not the most effective data consistency guarantee method, and the method is not suitable for all image reconstruction conditions are solved; the image quality after the image reconstruction based on the ADMM algorithm can be improved.
Example two
Fig. 2a is a flowchart of an image reconstruction model training method according to a second embodiment of the present invention, and this embodiment further illustrates a process of training an image reconstruction model based on the above embodiment.
As shown in fig. 2a, the process of training the image reconstruction model specifically includes the following steps:
s210, acquiring fully sampled magnetic resonance data, and extracting at least one group of undersampled data from the fully sampled magnetic resonance data to obtain at least one group of data pairs of the undersampled data and the fully sampled magnetic resonance data.
The method comprises the steps that a model training sample collection process is carried out, and for the fully sampled K-space data, sampling data of corresponding sampling lines are selected from the fully sampled data according to a preset rule to obtain under-sampled data. . Illustratively, in the full sampling process, there are 256 sampling lines, and if 4 times of acceleration sampling is required, that is, 64 sampling lines are required, then undersampling refers to selecting 64 sampling lines from the 256 sampling lines for sampling. Specifically, at least one group of sampling data of 64 sampling lines can be extracted from the group of full-sampling data of 256 sampling lines according to a preset rule to be used as undersampled magnetic resonance data, so that a sample data pair consisting of at least one group of undersampled magnetic resonance data and corresponding full-sampling magnetic resonance data can be obtained.
S220, inputting the undersampled data into a mathematical model with the data fidelity term and the regular term both being indefinite terms.
Specifically, since the image reconstruction model is based on the ADMM algorithm, the mathematical model in which the data fidelity term and the regularization term are both indefinite terms can be expressed as: minmF (Am, F) + lambda R (m), wherein m is an image to be reconstructed, F is undersampled k-space data, A represents an encoding matrix, an undersampled Fourier transform operator is represented in single-channel magnetic resonance imaging, lambda is a regular parameter, R (m) is a regular function, and F (Am, F) is a data fidelity term function. The F (Am, F) function is taken as a data fidelity term function, so that the general situation is considered, the method is a more effective data consistency guarantee method, and different from an ADMM-net method applicable model, on the premise that least square constraint is established in linear unbiased estimation, the data fidelity term is represented by adopting 2 norms between a reconstructed k space and sampling points.
S230, decomposing a mathematical model of which the data fidelity term and the regular term are both indefinite terms into a first subproblem, a second subproblem and a third subproblem based on an alternating direction multiplier algorithm, wherein the third subproblem is a constraint condition of a solution of the first subproblem and the second subproblem.
The process of decomposing the mathematical model is to introduce a z variable, which can be understood as a de-noised image of m. Under the premise that m is equal to z, the original mathematical model is decomposed into three unconstrained subproblems. Wherein,
in order to be the first sub-problem,
to the second sub-problem, argmaxβ<β,m-z>Is the third sub-problem.
S240, solving the first sub-problem and the second sub-problem by adopting a gradient descent method.
After solving the first and second sub-problems by the gradient descent method, the formula of each solution can be expressed as:
wherein i and k are the internal loop times of the first sub-problem and the second sub-problem respectively, and n is the iteration time of the ADMM algorithm. Gamma ray1、γ2、μ1And mu2The parameters for each item in the subproblem are given initial values during the calculation of the algorithm. The initial value may be an empirical value. F 'and R' are the first order partial derivatives of the functions F and R, i.e., the first order partial derivatives of the data fidelity term function and the regularization function.
And S250, determining parameter values in the solutions of the first sub-problem and the second sub-problem by a convolutional neural network iterative computation method aiming at the solutions of the first sub-problem and the second sub-problem, and finishing the training of the image reconstruction model.
Specifically, the method comprises the following steps: fitting a convolutional neural network to the first order partial derivative function of the data fidelity term function in the solution of the first sub-problem and the first order partial derivative function of the data regular term function in the solution of the second sub-problem, i.e. replacing the functions F 'and R' in the formula in step S240 with a convolutional neural network CNN, wherein each parameter (γ) in the solutions of the first and second sub-problems is1、γ2、μ1And mu2) Is an empirical value and can be expressed as the following equation:
and determining the numerical value of each parameter in the solution of the first subproblem and the solution of the second subproblem through preset iteration times until the difference value between the reconstructed image obtained through the data model and the reconstructed image corresponding to the fully sampled magnetic resonance data meets the loss function.
Specifically, a network structure (i.e., an ADMM-net-general network structure) of solutions of sub-problems subjected to convolutional neural network fitting may be output as m satisfying the requirement through n times of iterative computation with f as an input as shown in fig. 2 b. The procedure of the second iteration (iter-2) calculation is taken as an example:
in each iteration, the ADMM-net-generator consists of three modules: a reconstruction layer M, an optimization layer Z and a parameter update layer P. Wherein the structure of the reconstruction layer M and the optimization layer Z are consistent, S represents the addition operation, and the output is S(n,k)＝μ1 (n,k-1)s(n,k-1)+μ2 (n,k-1)s(n,0)-c(n,k-1)；c(n,k-1)Representing the output of the (k-1) th CNN module in the nth iteration. The related parameters of the external iteration times n, the internal circulation times k, i and the CNN are all set by a user. In a specific embodiment, the number of outer iterations is 15, the number of inner loops is 1, the number of layers of each CNN convolution is 3, the size of the convolution kernel is 3 × 3, and the activation function is the Relu function. Since the magnetic resonance signal is a complex signal, all data is processed in two channels, a real part and an imaginary part. During training, the loss function is defined as the mean square error:whereinFor reconstructed images output via the network, xrefAn image reconstructed for the fully sampled magnetic resonance data corresponding to f.
The image reconstruction is performed by using the image reconstruction model obtained through the training process, and the comparison between the image reconstruction effect and the effect of the reconstructed image obtained by using other algorithms can refer to a comparison graph shown in fig. 3.
According to the technical scheme, the mathematical model with the uncertain data fidelity terms and the uncertain regular terms is trained through the learning process of the convolutional neural network, so that the image reconstruction model meeting the loss function requirement is obtained, the application range of the image reconstruction model based on the ADMM algorithm is wider, and the image reconstruction quality is improved.
EXAMPLE III
Fig. 4 is a schematic structural diagram of a magnetic resonance image reconstruction apparatus according to a third embodiment of the present invention, and the magnetic resonance image reconstruction apparatus according to this embodiment is suitable for use in medical image reconstruction.
As shown in fig. 4, the magnetic resonance image reconstruction apparatus specifically includes: a data acquisition module 410 and an image reconstruction module 420.
Wherein, the data acquisition module 410 is configured to acquire undersampled magnetic resonance data; and an image reconstruction module 420, configured to input the magnetic resonance data into an image reconstruction model based on an alternating direction multiplier algorithm to obtain a reconstructed target magnetic resonance image, where the image reconstruction model is a result of pre-training a mathematical model in which a data fidelity term and a regular term are both indefinite terms.
According to the technical scheme, undersampled magnetic resonance data are input into an image reconstruction model based on an alternating direction multiplier algorithm, the image reconstruction model is a result of pre-training of a mathematical model with uncertain data fidelity terms and regular terms, and a reconstructed target magnetic resonance image is obtained, so that the problems that the data fidelity terms in the model applicable to the ADMM-net method are represented by 2 norms between a reconstructed k space and sampling points, the data fidelity terms are not the most effective data consistency guarantee method, and the method is not applicable to all image reconstruction conditions are solved; the image quality of the reconstructed image based on the AMDD algorithm can be improved.
Optionally, the magnetic resonance image reconstruction apparatus further includes a model training module, configured to train the image reconstruction model; the model training module specifically comprises:
the sample data acquisition submodule acquires fully sampled magnetic resonance data, extracts at least one group of undersampled data from the fully sampled magnetic resonance data, and obtains at least one group of data pairs of the undersampled data and the fully sampled magnetic resonance data;
the sample input submodule is used for inputting the undersampled data into a mathematical model of which the data fidelity term and the regular term are both indefinite terms;
the decomposition calculation sub-module is used for decomposing a mathematical model of which the data fidelity term and the regular term are both indefinite terms into a first sub-problem, a second sub-problem and a third sub-problem based on an alternating direction multiplier algorithm, wherein the third sub-problem is a constraint condition of the solutions of the first sub-problem and the second sub-problem;
a subproblem solving submodule for solving the first subproblem and the second subproblem by using a gradient descent method;
and the parameter solving submodule is used for determining parameter values in the solutions of the first sub-problem and the second sub-problem through a convolutional neural network iterative computation method aiming at the solutions of the first sub-problem and the second sub-problem, and finishing the training of the image reconstruction model.
Optionally, the parameter solving submodule is specifically configured to:
fitting a first-order partial derivative function of a data fidelity term function in the solution of the first sub-problem and a first-order partial derivative function of a data regular term function in the solution of the second sub-problem by adopting a convolutional neural network, wherein the initial value of each parameter in the solution of the first sub-problem and the solution of the second sub-problem is an empirical value;
and determining the numerical value of each parameter in the solution of the first subproblem and the solution of the second subproblem through preset iteration times until the difference value between the reconstructed image obtained through the data model and the reconstructed image corresponding to the fully sampled magnetic resonance data meets the loss function.
Optionally, in each iterative computation, the neural network structure includes three modules, namely a reconstruction layer, an optimization layer, and a parameter update layer.
The magnetic resonance image reconstruction device provided by the embodiment of the invention can execute the magnetic resonance image reconstruction method provided by any embodiment of the invention, and has corresponding functional modules and beneficial effects for executing the magnetic resonance image reconstruction method.
Example four
Fig. 5 is a schematic structural diagram of a computer device according to a fourth embodiment of the present invention. FIG. 5 illustrates a block diagram of an exemplary computer device 12 suitable for use in implementing embodiments of the present invention. The computer device 12 shown in FIG. 5 is only an example and should not bring any limitations to the functionality or scope of use of embodiments of the present invention.
As shown in FIG. 5, computer device 12 is in the form of a general purpose computing device. The components of computer device 12 may include, but are not limited to: one or more processors or processing units 16, a system memory 28, and a bus 18 that couples various system components including the system memory 28 and the processing unit 16.
The system memory 28 may include computer system readable media in the form of volatile memory, such as Random Access Memory (RAM)30 and/or cache memory 32. Computer device 12 may further include other removable/non-removable, volatile/nonvolatile computer system storage media. By way of example only, storage system 34 may be used to read from and write to non-removable, nonvolatile magnetic media (not shown in FIG. 5, and commonly referred to as a "hard drive"). Although not shown in FIG. 5, a magnetic disk drive for reading from and writing to a removable, nonvolatile magnetic disk (e.g., a "floppy disk") and an optical disk drive for reading from or writing to a removable, nonvolatile optical disk (e.g., a CD-ROM, DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to bus 18 by one or more data media interfaces. System memory 28 may include at least one program product having a set (e.g., at least one) of program modules that are configured to carry out the functions of embodiments of the invention.
A program/utility 40 having a set (at least one) of program modules 42 may be stored, for example, in system memory 28, such program modules 42 including, but not limited to, an operating system, one or more application programs, other program modules, and program data, each of which examples or some combination thereof may comprise an implementation of a network environment. Program modules 42 generally carry out the functions and/or methodologies of the described embodiments of the invention.
The processing unit 16 executes programs stored in the system memory 28 to perform various functional applications and data processing, for example, to implement the steps of a magnetic resonance image reconstruction method provided by the embodiment of the present invention, the method including:
acquiring undersampled magnetic resonance data;
and inputting the magnetic resonance data into an image reconstruction model based on an alternating direction multiplier algorithm to obtain a reconstructed target magnetic resonance image, wherein the image reconstruction model is a result of pre-training a mathematical model of which both a data fidelity term and a regular term are indefinite terms.
Of course, those skilled in the art can understand that the processor may also implement the technical solution of the interface authority configuration method provided by any embodiment of the present invention.
EXAMPLE five
This fifth embodiment provides a computer readable storage medium, on which a computer program is stored, which when executed by a processor, implements the steps of a magnetic resonance image reconstruction method as provided by any of the embodiments of the present invention, the method comprising:
acquiring undersampled magnetic resonance data;
and inputting the magnetic resonance data into an image reconstruction model based on an alternating direction multiplier algorithm to obtain a reconstructed target magnetic resonance image, wherein the image reconstruction model is a result of pre-training a mathematical model of which both a data fidelity term and a regular term are indefinite terms.
Computer storage media for embodiments of the invention may employ any combination of one or more computer-readable media. The computer readable medium may be a computer readable signal medium or a computer readable storage medium. The computer-readable storage medium may be, for example but not limited to: an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (a non-exhaustive list) of the computer readable storage medium would include the following: an electrical connection having one or more wires, a portable computer diskette, 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 disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the context of this document, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.
A computer readable signal medium may include a propagated data signal with computer readable program code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated data signal may take many forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A computer readable signal medium may also be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device.
Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, fiber optic cable, RF, etc., or any suitable combination of the foregoing.
Computer program code for carrying out operations for aspects of the present invention may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, Smalltalk, C + + or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet service provider).
It will be understood by those skilled in the art that the modules or steps of the invention described above may be implemented by a general purpose computing device, they may be centralized on a single computing device or distributed across a network of computing devices, and optionally they may be implemented by program code executable by a computing device, such that it may be stored in a memory device and executed by a computing device, or it may be separately fabricated into various integrated circuit modules, or it may be fabricated by fabricating a plurality of modules or steps thereof into a single integrated circuit module. Thus, the present invention is not limited to any specific combination of hardware and software.
It is to be noted that the foregoing is only illustrative of the preferred embodiments of the present invention and the technical principles employed. It will be understood by those skilled in the art that the present invention is not limited to the particular embodiments described herein, but is capable of various obvious changes, rearrangements and substitutions as will now become apparent to those skilled in the art without departing from the scope of the invention. Therefore, although the present invention has been described in greater detail by the above embodiments, the present invention is not limited to the above embodiments, and may include other equivalent embodiments without departing from the spirit of the present invention, and the scope of the present invention is determined by the scope of the appended claims.
Claims (10)
1. A magnetic resonance image reconstruction method, comprising:
acquiring undersampled magnetic resonance data;
and inputting the magnetic resonance data into an image reconstruction model based on an alternating direction multiplier algorithm to obtain a reconstructed target magnetic resonance image, wherein the image reconstruction model is a result of pre-training a mathematical model of which both a data fidelity term and a regular term are indefinite terms.
2. The method of claim 1, wherein the image reconstruction model training process comprises:
acquiring fully sampled magnetic resonance data, and extracting at least one group of undersampled data from the fully sampled magnetic resonance data to obtain at least one group of data pairs of the undersampled data and the fully sampled magnetic resonance data;
inputting the undersampled data into a mathematical model of which the data fidelity term and the regular term are both indefinite terms;
decomposing a mathematical model in which the data fidelity term and the regular term are both indefinite terms into a first subproblem, a second subproblem and a third subproblem based on an alternating direction multiplier algorithm, wherein the third subproblem is a constraint condition of a solution of the first subproblem and the second subproblem;
solving the first sub-problem and the second sub-problem by adopting a gradient descent method;
and aiming at the solution of the first sub-problem and the solution of the second sub-problem, determining each parameter value in the solution of the first sub-problem and the solution of the second sub-problem through a convolutional neural network iterative computation method, and finishing the training of the image reconstruction model.
3. The method of claim 2, wherein determining the respective parameter values in the solution of the first sub-problem and the solution of the second sub-problem by a convolutional neural network iterative computation method comprises:
fitting a first-order partial derivative function of a data fidelity term function in the solution of the first sub-problem and a first-order partial derivative function of a data regular term function in the solution of the second sub-problem by adopting a convolutional neural network, wherein the initial value of each parameter in the solution of the first sub-problem and the solution of the second sub-problem is an empirical value;
and determining the numerical value of each parameter in the solution of the first subproblem and the solution of the second subproblem through preset iteration times until the difference value between the reconstructed image obtained through the data model and the reconstructed image corresponding to the fully sampled magnetic resonance data meets the loss function.
4. The method of claim 3, wherein the neural network structure comprises three modules, namely a reconstruction layer, an optimization layer and a parameter updating layer, in each iterative calculation.
5. A magnetic resonance image reconstruction apparatus, characterized by comprising:
a data acquisition module for acquiring undersampled magnetic resonance data;
and the image reconstruction module is used for inputting the magnetic resonance data into an image reconstruction model based on an alternating direction multiplier algorithm so as to obtain a reconstructed target magnetic resonance image, wherein the image reconstruction model is a result of pre-training a mathematical model of which both a data fidelity term and a regular term are indefinite terms.
6. The apparatus of claim 5, further comprising a model training module for training the image reconstruction model; the model training module specifically comprises:
the sample data acquisition submodule acquires fully sampled magnetic resonance data, extracts at least one group of undersampled data from the fully sampled magnetic resonance data, and obtains at least one group of data pairs of the undersampled data and the fully sampled magnetic resonance data;
the sample input submodule is used for inputting the undersampled data into a mathematical model of which the data fidelity term and the regular term are both indefinite terms;
the decomposition calculation sub-module is used for decomposing a mathematical model of which the data fidelity term and the regular term are both indefinite terms into a first sub-problem, a second sub-problem and a third sub-problem based on an alternating direction multiplier algorithm, wherein the third sub-problem is a constraint condition of the solutions of the first sub-problem and the second sub-problem;
a subproblem solving submodule for solving the first subproblem and the second subproblem by using a gradient descent method;
and the parameter solving submodule is used for determining parameter values in the solutions of the first sub-problem and the second sub-problem through a convolutional neural network iterative computation method aiming at the solutions of the first sub-problem and the second sub-problem, and finishing the training of the image reconstruction model.
7. The apparatus of claim 6, wherein the parameter solving submodule is specifically configured to:
fitting a first-order partial derivative function of a data fidelity term function in the solution of the first sub-problem and a first-order partial derivative function of a data regular term function in the solution of the second sub-problem by adopting a convolutional neural network, wherein the initial value of each parameter in the solution of the first sub-problem and the solution of the second sub-problem is an empirical value;
and determining the numerical value of each parameter in the solution of the first subproblem and the solution of the second subproblem through preset iteration times until the difference value between the reconstructed image obtained through the data model and the reconstructed image corresponding to the fully sampled magnetic resonance data meets the loss function.
8. The apparatus of claim 7, wherein the neural network structure comprises three modules, namely a reconstruction layer, an optimization layer and a parameter update layer, in each iterative computation.
9. A computer device, characterized in that the computer device comprises:
one or more processors;
a memory for storing one or more programs;
when executed by the one or more processors, cause the one or more processors to implement a magnetic resonance image reconstruction method as claimed in any one of claims 1-4.
10. A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the magnetic resonance image reconstruction method as set forth in any one of claims 1-4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911011736.8A CN110766768B (en) | 2019-10-23 | 2019-10-23 | Magnetic resonance image reconstruction method, device, equipment and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911011736.8A CN110766768B (en) | 2019-10-23 | 2019-10-23 | Magnetic resonance image reconstruction method, device, equipment and medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110766768A true CN110766768A (en) | 2020-02-07 |
| CN110766768B CN110766768B (en) | 2023-08-04 |
Family
ID=69333181
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911011736.8A Active CN110766768B (en) | 2019-10-23 | 2019-10-23 | Magnetic resonance image reconstruction method, device, equipment and medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110766768B (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111537930A (en) * | 2020-04-09 | 2020-08-14 | 深圳先进技术研究院 | Gradient field control method, gradient field control device, magnetic resonance imaging equipment and medium |
| CN113129404A (en) * | 2021-04-20 | 2021-07-16 | 上海科技大学 | Iterative photoacoustic image reconstruction method accelerated by deep learning |
| CN113269849A (en) * | 2020-07-23 | 2021-08-17 | 上海联影智能医疗科技有限公司 | Method and apparatus for reconstructing magnetic resonance |
| CN113298902A (en) * | 2021-05-13 | 2021-08-24 | 中国科学院深圳先进技术研究院 | Method for reconstructing magnetic resonance image in convoluted field of view, computer device and storage medium |
| WO2021253671A1 (en) * | 2020-06-18 | 2021-12-23 | 中国科学院深圳先进技术研究院 | Magnetic resonance cine imaging method and apparatus, and imaging device and storage medium |
| WO2022193378A1 (en) * | 2021-03-17 | 2022-09-22 | 中国科学院深圳先进技术研究院 | Image reconstruction model generation method and apparatus, image reconstruction method and apparatus, device, and medium |
| CN116228913A (en) * | 2023-05-06 | 2023-06-06 | 杭州师范大学 | Processing method and device for magnetic resonance image data and storage medium |
| WO2023102782A1 (en) * | 2021-12-08 | 2023-06-15 | 深圳先进技术研究院 | Reconstruction method and system for quantitative imaging of magnetic resonance parameters |
| WO2023108787A1 (en) * | 2021-12-16 | 2023-06-22 | 中国科学院深圳先进技术研究院 | Gradient-domain matrix elastic network infimal convolution regularization accelerated magnetic resonance parameter imaging method, and terminal device and computer storage medium |
| WO2023116633A1 (en) * | 2021-12-21 | 2023-06-29 | 中国科学院深圳先进技术研究院 | Detail-fidelity multi-scale deep learning magnetic resonance dynamic image reconstruction method |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108335339A (en) * | 2018-04-08 | 2018-07-27 | 朱高杰 | A kind of magnetic resonance reconstruction method based on deep learning and convex set projection |
| CN109325985A (en) * | 2018-09-18 | 2019-02-12 | 上海联影智能医疗科技有限公司 | MR image reconstruction method, apparatus and computer readable storage medium |
| CN109350061A (en) * | 2018-11-21 | 2019-02-19 | 成都信息工程大学 | MR imaging method based on depth convolutional neural networks |
| CN109671129A (en) * | 2018-12-14 | 2019-04-23 | 深圳先进技术研究院 | A kind of the dynamic magnetic resonance image method for reconstructing and device of auto-adaptive parameter study |
| CN109872376A (en) * | 2019-02-20 | 2019-06-11 | 四川大学华西医院 | A kind of method, apparatus and readable storage medium storing program for executing for rebuilding dynamic magnetic resonance image |
| CN110288672A (en) * | 2019-06-28 | 2019-09-27 | 闽江学院 | A kind of compressed sensing MR image reconstruction method based on the dense network of ultra-deep |
-
2019
- 2019-10-23 CN CN201911011736.8A patent/CN110766768B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108335339A (en) * | 2018-04-08 | 2018-07-27 | 朱高杰 | A kind of magnetic resonance reconstruction method based on deep learning and convex set projection |
| CN109325985A (en) * | 2018-09-18 | 2019-02-12 | 上海联影智能医疗科技有限公司 | MR image reconstruction method, apparatus and computer readable storage medium |
| CN109350061A (en) * | 2018-11-21 | 2019-02-19 | 成都信息工程大学 | MR imaging method based on depth convolutional neural networks |
| CN109671129A (en) * | 2018-12-14 | 2019-04-23 | 深圳先进技术研究院 | A kind of the dynamic magnetic resonance image method for reconstructing and device of auto-adaptive parameter study |
| CN109872376A (en) * | 2019-02-20 | 2019-06-11 | 四川大学华西医院 | A kind of method, apparatus and readable storage medium storing program for executing for rebuilding dynamic magnetic resonance image |
| CN110288672A (en) * | 2019-06-28 | 2019-09-27 | 闽江学院 | A kind of compressed sensing MR image reconstruction method based on the dense network of ultra-deep |
Non-Patent Citations (2)
| Title |
|---|
| 徐宗本等: "求解反问题的一个新方法:模型求解与范例学习结合", 《中国科学:数学》 * |
| 袁太文等: "非均匀磁共振压缩成像的交替方向乘子法", 《仪器仪表学报》 * |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111537930A (en) * | 2020-04-09 | 2020-08-14 | 深圳先进技术研究院 | Gradient field control method, gradient field control device, magnetic resonance imaging equipment and medium |
| CN111537930B (en) * | 2020-04-09 | 2021-09-21 | 深圳先进技术研究院 | Magnetic resonance parameter imaging method and equipment based on gradient waveform adjustment |
| WO2021253671A1 (en) * | 2020-06-18 | 2021-12-23 | 中国科学院深圳先进技术研究院 | Magnetic resonance cine imaging method and apparatus, and imaging device and storage medium |
| CN113269849A (en) * | 2020-07-23 | 2021-08-17 | 上海联影智能医疗科技有限公司 | Method and apparatus for reconstructing magnetic resonance |
| WO2022193378A1 (en) * | 2021-03-17 | 2022-09-22 | 中国科学院深圳先进技术研究院 | Image reconstruction model generation method and apparatus, image reconstruction method and apparatus, device, and medium |
| CN113129404A (en) * | 2021-04-20 | 2021-07-16 | 上海科技大学 | Iterative photoacoustic image reconstruction method accelerated by deep learning |
| CN113298902A (en) * | 2021-05-13 | 2021-08-24 | 中国科学院深圳先进技术研究院 | Method for reconstructing magnetic resonance image in convoluted field of view, computer device and storage medium |
| WO2023102782A1 (en) * | 2021-12-08 | 2023-06-15 | 深圳先进技术研究院 | Reconstruction method and system for quantitative imaging of magnetic resonance parameters |
| WO2023108787A1 (en) * | 2021-12-16 | 2023-06-22 | 中国科学院深圳先进技术研究院 | Gradient-domain matrix elastic network infimal convolution regularization accelerated magnetic resonance parameter imaging method, and terminal device and computer storage medium |
| WO2023116633A1 (en) * | 2021-12-21 | 2023-06-29 | 中国科学院深圳先进技术研究院 | Detail-fidelity multi-scale deep learning magnetic resonance dynamic image reconstruction method |
| CN116228913A (en) * | 2023-05-06 | 2023-06-06 | 杭州师范大学 | Processing method and device for magnetic resonance image data and storage medium |
| CN116228913B (en) * | 2023-05-06 | 2023-08-22 | 杭州师范大学 | Processing method and device for magnetic resonance image data and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110766768B (en) | 2023-08-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110766769B (en) | Magnetic resonance image reconstruction method, device, equipment and medium | |
| CN110766768B (en) | Magnetic resonance image reconstruction method, device, equipment and medium | |
| US11604984B2 (en) | Systems and methods for machine learning based modeling | |
| CN109671129B (en) | Dynamic magnetic resonance image reconstruction method and device for adaptive parameter learning | |
| US10452979B2 (en) | Convolution neural network training apparatus and method thereof | |
| DE112020004167T5 (en) | VIDEO PREDICTION USING ONE OR MORE NEURAL NETWORKS | |
| WO2020114329A1 (en) | Fast magnetic resonance parametric imaging and device | |
| CN110717958A (en) | Image reconstruction method, device, equipment and medium | |
| CN112907439A (en) | Supine position and prone position mammary gland image registration method based on deep learning | |
| KR102428725B1 (en) | Method and program for imaging quality improving | |
| CN113569891A (en) | Training data processing device, electronic equipment and storage medium of neural network model | |
| CN111856362A (en) | Magnetic resonance imaging method, device, system and storage medium | |
| WO2022193379A1 (en) | Image reconstruction model generation method and apparatus, image reconstruction method and apparatus, device, and medium | |
| US11460528B2 (en) | MRI reconstruction with image domain optimization | |
| CN110728732A (en) | Image reconstruction method, device, equipment and medium | |
| CN114972118B (en) | Noise reduction method and device for inspection image, readable medium and electronic equipment | |
| US11216692B2 (en) | Systems and methods for coupled representation using transform learning for solving inverse problems | |
| KR102462513B1 (en) | Method, program, and apparatus for processing three dimentional medical data | |
| CN115115736A (en) | Image artifact removing method, device and equipment and storage medium | |
| CN115115722A (en) | Image reconstruction model generation method, image reconstruction device, image reconstruction equipment and medium | |
| CN113592973A (en) | Magnetic resonance image reconstruction method and device based on multi-frequency complex convolution | |
| EP4053752A1 (en) | Cnn-based image processing | |
| CN117745880B (en) | Medical image filling method, device, equipment and medium for multidimensional nonlinear transformation | |
| EP4375918A1 (en) | Image processing | |
| CN116579414B (en) | Model training method, MRI thin layer data reconstruction method, device and equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |