WO2023108484A1

WO2023108484A1 - Magnetic resonance imaging method based on wave gradient coding field and deep learning model

Info

Publication number: WO2023108484A1
Application number: PCT/CN2021/138359
Authority: WO
Inventors: 郑海荣; 王海峰; 梁栋; 刘新; 刘聪聪; 崔卓须
Original assignee: 中国科学院深圳先进技术研究院
Priority date: 2021-12-15
Filing date: 2021-12-15
Publication date: 2023-06-22

Abstract

The present application discloses a magnetic resonance imaging method and apparatus based on a wave (Wave-CAIPI) gradient coding field and a deep learning model, a device, and a storage medium. The method comprises: introducing a deep generative model (DGM) into a magnetic resonance reconstruction model of the Wave-CAIPI gradient coding field to accelerate magnetic resonance imaging; performing conjugate symmetry on physical coil channel data by means of a virtual coil concept (VCC) to generate VCC channel data; and merging the physical coil channel data and the VCC channel data to reconstruct a geometric factor calculation model. According to the solution provided by the present application, a VCC (Wave-CAIPI) technology in a Wave-CAIPI gradient coding occasion is combined with the DGM, the advantage of reducing the number of system conditions by the Wave-CAIPI and the VCC is utilized, and g-factor suitable for reconstruction is smaller and more uniform, such that a reconstructed graph has a higher signal-to-noise ratio, and the step of needing a large amount of training data to train network parameters in a conventional convolutional neural network is shortened.

Description

Magnetic Resonance Imaging Method Based on Wave Gradient Encoded Field and Deep Learning Model

technical field

The invention relates to the technical field of magnetic resonance imaging, in particular to a magnetic resonance imaging method, device, equipment and storage medium based on a wave gradient encoding field and a deep learning model.

Background technique

Long scan times are inherent in conventional magnetic resonance imaging (MRI). Among the existing magnetic resonance imaging acceleration methods, parallel imaging is an effective method of accelerating imaging. Parallel imaging is a method that utilizes the spatial encoding capability of multi-channel phased array coils to reduce the number of encoding steps of the gradient magnetic field to achieve acceleration. In practice, parallel imaging achieves the purpose of under-sampling through redundant information (prior information) during multi-channel k-space. In parallel imaging, reconstruction methods can be divided into two categories, one is based on image domain anti-aliasing or solving methods, such as PILS, SENSE and ESPIRiT, etc.; the other is based on k-space data filling methods, such as SMASH, GRAPPA and SPIRiT et al. In the reconstruction method of the image domain, the coil sensitivity information (CSM) is estimated by using the acquired k-space reference line (ACS) at first, and the aliasing of the image is generated after the accelerated (under-sampled) k-space becomes the image domain, and then the image is aliased by using CSM performs anti-aliasing; the method based on k-space data filling is similar to the method in the image domain, and the convolution kernel (kernel) used to fill the original data of k-space is estimated through the collected ACS lines. Whether it is an acceleration method based on the image domain or k-space, it is necessary to collect ACS lines for estimating CSM or kernel. Then collecting ACS takes a lot of time in practice, and after the ACS line is collected, it takes a lot of time to estimate CSM or kernel using commonly used algorithms (such as ESPIRiT, etc.), which is not friendly to practical applications.

A parallel imaging method using Wave-CAIPI (where CAIPI stands for Controlled Phase Shift or Misalignment Operation in Phase and Slice Direction) with controlled aliasing has a remarkable effect in 3D imaging by causing Aliasing, making full use of coil sensitivity changes in three directions, significantly high-magnification accelerated 3D imaging with reduced geometry factors and residual aliasing artifacts. Wave-CAIPI combines Beam Phase Encoding (BPE) and Controlled Aliasing Parallel Imaging (CAIPIRINHA), which can significantly utilize the CSM information aliased in three directions in space, and use the extended SENSE model to solve, the system has a smaller condition number , so that the reconstructed geometric factor is smaller and the distribution is more uniform, so that the reconstructed image has a higher signal-to-noise ratio. Although Wave-CAIPI can achieve a higher acceleration factor while taking into account a higher signal-to-noise ratio, the reconstruction process still needs to collect ACS to estimate CSM. This process still does not avoid the time-consuming process based on the image domain or k-space. It is unfriendly in clinical application.

Contents of the invention

In view of the above-mentioned defects or deficiencies in the prior art, it is desired to provide a magnetic resonance imaging method, device, equipment and storage medium based on the Wave-CAIPI gradient encoding field and deep learning model.

In the first aspect, the embodiment of the present application provides a magnetic resonance imaging method based on the Wave-CAIPI deep learning model, the method includes: introducing a deep generative network model into the magnetic resonance reconstruction model of the Wave-CAIPI gradient encoding field, and accelerating the magnetic resonance imaging method. Resonance imaging; make conjugate symmetry to the physical coil channel data through VCC to generate VCC channel data; merge the physical coil channel data and VCC channel data to reconstruct the geometric factor calculation model.

In one of the embodiments, before introducing the deep generation network model into the magnetic resonance reconstruction model of the Wave-CAIPI gradient encoding field, the method further includes: using parallel imaging to acquire multiple images corresponding to different channels in accelerated magnetic resonance imaging Under-sampled k-space, based on different coil sensitivity information, the artifact-free image of the reconstructed under-sampled k-space is obtained.

In one of the embodiments, the introduction of the deep generation network model in the magnetic resonance reconstruction model of the Wave-CAIPI gradient encoding field accelerates the magnetic resonance imaging, including: setting up the encoding model of the Wave gradient field (Wave) in the magnetic resonance imaging: wave(x,y,z)=Em(x,y,z), where E is the encoding matrix,

M represents the offset aliasing caused by the controlled aliasing parallel imaging sampling mode in the phase encoding direction, F _x represents the Fourier transform in the x direction, S represents the coil sensitivity information, Psf(k _x ,y,z) is the effect of the Wave-CAIPI gradient field; through wave(x,y,z)=Em(x,y,z) and

get

Wherein, Psf[x,y,z] is the inverse Fourier transform of Psf(k _x ,y,z) in the readout direction.

In one of the embodiments, the VCC is used to perform conjugate symmetry on the physical coil channel data to generate the VCC channel data, including: through wave(x, y, z)=Em(x, y, z) and

get

Among them, wave*(x,y,z) represents the data expanded by VCC,

Indicates the Psf that expands the number of channels on the basis of the existing Psf[x,y,z].

In one of the embodiments, after the VCC channel data is generated, the method further includes: introducing a deep generation model that does not require training.

In one of the embodiments, the merging of the physical coil channel data and the VCC channel data to reconstruct the geometric factor calculation model includes: using the formula

Get the geometric factor g-factor after VCC expansion, where,

Indicates the coding matrix after VCC expansion.

In the second aspect, the embodiment of the present application also provides a magnetic resonance imaging device based on the Wave-CAIPI gradient encoding field and the deep learning model, the device includes: an introduction unit for magnetic resonance reconstruction in the Wave-CAIPI gradient encoding field The model introduces a deep generation network model to accelerate magnetic resonance imaging; the generation unit is used to perform conjugate symmetry on the physical coil channel data through VCC to generate VCC channel data; the reconstruction unit is used to merge the physical coil channel data and VCC channel data Rebuild the geometry factor calculation model.

In the third aspect, the embodiment of the present application also provides a computer device, including a memory, a processor, and a computer program stored in the memory and operable on the processor. When the processor executes the program, it implements the The method described in any one of the descriptions of the examples.

In a fourth aspect, the embodiment of the present application also provides a computer device, a computer-readable storage medium, on which a computer program is stored, and the computer program is used for: when the computer program is executed by a processor, the computer program according to the present application is implemented. The method described in any one of the descriptions of the examples.

Beneficial effects of the present invention:

The magnetic resonance imaging method based on the Wave-CAIPI gradient coding field and the deep learning model provided by the present invention combines Wave-CAIPI, VCC and DGM, which not only utilizes the advantages of Wave-CAIPI and VCC to reduce the system condition number, but is applicable The reconstructed g-factor is smaller and more uniform, so that the reconstructed graphics have a higher signal-to-noise ratio, and there is no need to collect ACS lines to estimate CSM or kernel during the solution process, which greatly shortens the traditional convolutional neural network. The data collection time of training data training network parameters avoids the errors introduced by inaccurate data collection, and does not require time-consuming estimation of CSM or kernel in traditional reconstruction, which can be more suitable for clinical needs.

Description of drawings

Other characteristics, objects and advantages of the present application will become more apparent by reading the detailed description of non-limiting embodiments made with reference to the following drawings:

Fig. 1 shows the schematic flow chart of the magnetic resonance imaging method based on Wave-CAIPI gradient encoding field and deep learning model provided by the embodiment of the present application;

Fig. 2 shows an exemplary structural block diagram of a magnetic resonance imaging device 200 based on a Wave-CAIPI gradient encoding field and a deep learning model according to an embodiment of the present application;

FIG. 3 shows a schematic structural diagram of a computer system suitable for implementing a terminal device according to an embodiment of the present application;

Fig. 4 shows the structural diagram of the depth generation model provided by the embodiment of the present application;

FIG. 5 shows a schematic diagram of the simulation results combined with Wave-CAIPI and VCC provided by the embodiment of the present application;

Fig. 6 shows a schematic diagram of the reconstruction and comparison results of WV-DGM and WV-SENSE (Wave-VCC SENSE) provided in the embodiment of the present application on brain data.

Detailed ways

In order to make the above objects, features and advantages of the present invention more comprehensible, specific implementations of the present invention will be described in detail below in conjunction with the accompanying drawings. In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention. However, the present invention can be implemented in many other ways different from those described here, and those skilled in the art can make similar improvements without departing from the connotation of the present invention, so the present invention is not limited by the specific embodiments disclosed below.

In describing the present invention, it should be understood that the terms "center", "longitudinal", "transverse", "length", "width", "thickness", "upper", "lower", "front", " Back", "Left", "Right", "Vertical", "Horizontal", "Top", "Bottom", "Inner", "Outer", "Clockwise", "Counterclockwise", "Axial" , "radial", "circumferential" and other indicated orientations or positional relationships are based on the orientations or positional relationships shown in the drawings, which are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying the referred device or Elements must have certain orientations, be constructed and operate in certain orientations, and therefore should not be construed as limitations on the invention.

In addition, the terms "first" and "second" are used for descriptive purposes only, and cannot be interpreted as indicating or implying relative importance or implicitly specifying the quantity of indicated technical features. Thus, the features defined as "first" and "second" may explicitly or implicitly include at least one of these features. In the description of the present invention, "plurality" means at least two, such as two, three, etc., unless otherwise specifically defined.

In the present invention, unless otherwise clearly specified and limited, terms such as "installation", "connection", "connection" and "fixation" should be understood in a broad sense, for example, it can be a fixed connection or a detachable connection , or integrated; it may be mechanically connected or electrically connected; it may be directly connected or indirectly connected through an intermediary, and it may be the internal communication of two components or the interaction relationship between two components, unless otherwise specified limit. Those of ordinary skill in the art can understand the specific meanings of the above terms in the present invention according to specific situations.

In the present invention, unless otherwise clearly specified and limited, the first feature may be in direct contact with the first feature or the first and second feature may be in direct contact with the second feature through an intermediary. touch. Moreover, "above", "above" and "above" the first feature on the second feature may mean that the first feature is directly above or obliquely above the second feature, or simply means that the first feature is higher in level than the second feature. "Below", "beneath" and "beneath" the first feature may mean that the first feature is directly below or obliquely below the second feature, or simply means that the first feature is less horizontally than the second feature.

It should be noted that when an element is referred to as being “fixed on” or “disposed on” another element, it may be directly on the other element or there may be an intervening element. When an element is referred to as being "connected to" another element, it can be directly connected to the other element or intervening elements may also be present. As used herein, the terms "vertical", "horizontal", "upper", "lower", "left", "right" and similar expressions are for the purpose of illustration only and are not intended to represent the only embodiments.

Please refer to FIG. 1 , which shows a schematic flowchart of a magnetic resonance imaging method based on a Wave-CAIPI gradient encoding field and a deep learning model provided by an embodiment of the present application.

As shown in Figure 1, the method includes:

Step 110, introducing a deep generative network model into the magnetic resonance reconstruction model of the Wave-CAIPI gradient encoding field to accelerate magnetic resonance imaging;

Step 120, performing conjugate symmetry on the physical coil channel data through VCC to generate VCC channel data;

In step 130, the physical coil channel data and the VCC channel data are merged to reconstruct a geometric factor calculation model.

Using the above technical solution, combining Wave-CAIPI, VCC and DGM, it not only takes advantage of the advantages of Wave-CAIPI and VCC to reduce the system condition number, but the g-factor applicable to reconstruction is smaller and more uniform, so that the reconstructed graphics have more It has a high signal-to-noise ratio, and does not need to collect ACS lines to estimate CSM or kernel during the solution process, thereby greatly shortening the data collection time that requires a large amount of training data to train network parameters in traditional convolutional neural networks, and avoiding the need for data collection. Accurate and introduced errors do not require time-consuming estimation of CSM or kernel in traditional reconstruction, which can be more suitable for clinical needs.

In some embodiments, before introducing the deep generative network model in the magnetic resonance reconstruction model of the Wave-CAIPI gradient encoding field in the present application, the method further includes: using parallel imaging to acquire multiple images corresponding to different channels in accelerated magnetic resonance imaging The undersampled k-space of , by using different coil sensitivity information, obtains the artifact-free image of the reconstructed undersampled k-space.

Specifically, in accelerated magnetic resonance imaging, parallel imaging is an effective way to reduce the geometric factor of accelerated imaging, which simultaneously acquires multiple under-sampled k-spaces corresponding to different channels, and the coil sensitivity distribution of each channel is Different, so there is a difference in coil sensitivity encoding between each k-space, that is, there is complementary information, by using different coil sensitivity information, an image without artifacts can be reconstructed from multiple under-sampled k-spaces , in essence, parallel imaging mainly uses redundant information (prior information) between different coil k-spaces to achieve the purpose of undersampling and accelerated acquisition. The signal model is as follows,

in,

(2) where g _y (t) and g _z (t) are a pair of Wave-CAIPI gradient fields with a phase difference of π/2 during readout.

In some embodiments, the deep generation network model is introduced into the magnetic resonance reconstruction model of the Wave-CAIPI gradient encoding field in this application to accelerate magnetic resonance imaging, including: establishing a coding model of the Wave-CAIPI gradient field in magnetic resonance imaging: wave(x,y,z)=Em(x,y,z), where E is the encoding matrix,

M represents the offset aliasing caused by the controllable aliasing parallel imaging sampling mode in the phase encoding direction, f _x represents the Fourier transform in the x direction, S represents the coil sensitivity information, Psf(k _x ,y,z) is the effect of the Wave-CAIPI gradient field; through wave(x,y,z)=Em(x,y,z) and

get

Specifically, in MRI, the effect of the Wave-CAIPI gradient field can be represented by Psf, which is a three-dimensional phase diagram with periodic sinusoidal changes, where t is linearly corresponding to the code k _x of the k-space readout direction, Psf(t,y,z) can also be written in discrete form Psf(k _x ,y,z), and Psf(k _x ,y,z) is Fourier transformed in the y and z directions to become Psf(k _x , k _y , k _z ), it can describe the offset of the Wave-CAIPI track relative to the Cartesian track, and perform an inverse Fourier transform of Psf(k _x ,y, z) in the readout direction, which becomes Psf (x, y, z), which can describe the diffusion effect of the Wave-CAIPI gradient field in the readout direction. The encoding model of Wave-CAIPI can be simplified as, wave(x,y,z)=Em(x,y,z)(3), where E is the encoding matrix,

(4), where S denotes the CSM, F _x denotes the Fourier transform in the x direction, and M denotes the offset aliasing caused by the CAIPIRINHA sampling pattern in the phase encoding direction. In fact, formulas (3) and (4) can be equivalently written as,

(5), where Psf[x,y,z] is where Psf(k _x ,y,z) is the inverse Fourier transform in the readout direction. By introducing the Wave-CAIPI spatial encoding gradient into the encoding matrix, combined with CAIPIRINHA, pixels can be moved (aliased) in 3D space, thereby reducing the condition number of the solution system and shrinking the geometric factor (g-factor).

In some embodiments, the conjugate symmetry of the physical coil channel data through VCC in this application is used to generate VCC channel data, including: through wave(x, y, z)=Em(x, y, z) and

get

Among them, wave*(x,y,z) represents the data expanded by VCC,

Specifically, the use of Wave-CIAPI to accelerate MRI imaging can effectively reduce the g-factor and improve SNR. VCC (VCC) technology is another parallel imaging method to improve the system condition number of the encoding operator. VCC mainly uses the target The background phase and the coil phase provide additional encoding capabilities. The specific method is to generate the data of the virtual conjugate coil channel by performing conjugate symmetry on the data of the physical coil channel, and merge the two together for reconstruction, which can further Reduce the number of conditions of the system, reduce the geometric factor, and improve the image SNR. After introducing the VCC technology in Wave-CAIPI, and combining (3) and (4), it can be obtained,

(6) Among them, wave*(x,y,z) represents the data expanded by VCC,

Indicates the Psf that expands the number of channels on the basis of the original Psf[x,y,z].

In some embodiments, the combination of the physical coil channel data and the VCC channel data in the present application to reconstruct the geometric factor calculation model includes: through the formula

Get the geometric factor g-factor after VCC expansion, where,

Indicates the coding matrix after VCC expansion.

Specifically, the g-factor after VCC extension is obtained through (5) as follows,

(7), where,

Represents the coding matrix after VCC extension, (7) represents the g-factor calculation model after VCC extension, the calculation method of the original g-factor is similar to (7), except that E is the model without VCC extension. By comparing traditional g-factor and

It can be obtained that the model after VCC expansion will effectively reduce the g-factor and improve the signal-to-noise ratio of image reconstruction.

In some embodiments, after the VCC channel data is generated in the present application, the method further includes: introducing a deep generation model that does not require training.

Specifically, in the solution model in (6), the solution method based on the image domain is dominant. While using the advantages of Wave-CAIPI, it still needs to estimate the CSM or kernel (k-space data completion method) , it takes a certain amount of time to collect ACS in clinical practice, and it is estimated that CSM or kernel will also take a lot of time. In the present invention, while retaining the advantages of Wave-CAIPI, a deep generative model (DGM) that does not require training is introduced, combined with (1), (3) and (4), the model proposed by the present invention is named as Wave-VCC-DGM,

Among them, G(ξ) represents DGM, ξ represents fixed random noise, and the output result after G(ξ) operation is the image to be solved. The core idea is to use the network parameters of DGM to fit the image to be solved. DGM It is a simple convolution generator (of course, in this invention, it is not limited to the DGM network architecture. DGM is only a model suitable for the image domain. If it is solved in k-space, the model needs to be changed to a network that conforms to k-space solution Structure), consisting of upsampling (Upsampling), class convolution (Convolution-like), BatchNormalization (BN), and activation function layers, the structure is shown in Figure 4.

The deep generative network model used in the present invention can make an effective approximation to the traditional CSM or k-space kernel based on the image domain, and contains all the previous assumptions. The model itself is low-parameterized and does not involve Convolution, and has a simple structure. Although DMG does not use convolution operations, its structure is closely related to convolutional neural networks. Specifically, the network combines pixels between different channels. Since the coupling between linearly arranged pixels is not provided, Therefore, DGM is not a convolution operation in the traditional sense, but similar to convolution operations in convolutional neural networks, weights are shared between spatial locations. The coupling between pixels in DMG comes from upsampling. In (8), it can be seen that the present invention combines DGM, Wave-CAIPI and VCC to propose a new network model Wave-VCC-DGM that does not need to be trained, which can not only utilize the reduction system of Wave-CAIPI The advantage of the condition number, and the advantage of VCC combined with the target background phase, further reduces the g-factor of the reconstructed image and improves the SNR of the image.

In addition, DGM can take advantage of the advantage of not requiring training, and directly use ADAM (not limited to ADAM) to solve directly to meet the requirements of final image fitting, which can greatly reduce the cost of collecting training data in clinical practice. Time, and does not need to consume a lot of time for the estimation of CSM, or the calculation of the kernel, which is very friendly for rapid implementation in the clinic. The Wave-VCC-DGM model can make parallel imaging more flexible. In the two traditional methods of solving parallel imaging (based on k-space and based on image domain), it will increase the complexity of the solution process, such as calculation based on k-space The kernel and the estimated CSM based on the image domain are relatively complicated to calculate the kernel in the non-Cartesian coordinate system (Non-Cartesian), but the model proposed in the present invention can avoid this problem, because the present invention uses the Calibration-free method Solving, greatly reducing the complexity of the solving process.

The present invention proves through experiment, Wave-VCC has better effect, first uses the Wave coding of emulation (Psf is emulation), the result is shown in Figure 5, can see that compared with the traditional image-based domain (taking SENSE as an example) The reconstruction results of , combined with Wave-CAIPI and VCC have better results. The data is uniformly sampled from the template, achieving a 4x speedup.

In Figure 5, the simulated Wave coding model is used for preliminary experiments, and the traditional reconstruction based on the image domain (taking SENSE as an example) is carried out respectively. Using a 4-fold accelerated uniform sampling mask, you can see that Wave-CAIPI and VCC (Wave- VCC) combined results have a better reconstruction signal-to-noise ratio. The present invention has also carried out the reconstruction test of data encoded by real Wave space. The Wave encoding amplitude is 1.2mT/m, and the number of cycles is 7. This is just an example, and Wave is not limited to the above parameters. The result is shown in Figure 6 , the results show that WV-DGM has a better effect than traditional WV-SENSE.

Further, referring to FIG. 2 , FIG. 2 shows an exemplary structural block diagram of a magnetic resonance imaging apparatus 200 based on a Wave gradient encoding field and a deep learning model according to an embodiment of the present application.

As shown in Figure 2, the device includes:

The introduction unit 210 is used to introduce a deep generation network model into the magnetic resonance reconstruction model of the Wave-CAIPI gradient encoding field to accelerate magnetic resonance imaging;

The generating unit 220 is configured to perform conjugate symmetry on the physical coil channel data through the VCC to generate VCC channel data;

The reconstruction unit 230 is configured to combine the physical coil channel data and the VCC channel data to reconstruct the geometric factor calculation model.

It should be understood that the units or modules recorded in the device 200 correspond to the steps in the method described with reference to FIG. 1 . Therefore, the operations and features described above for the method are also applicable to the device 200 and the units contained therein, and will not be repeated here. The apparatus 200 may be pre-implemented in the browser of the electronic device or other security applications, and may also be loaded into the browser of the electronic device or its security applications by downloading or other means. The corresponding units in the apparatus 200 may cooperate with the units in the electronic device to implement the solutions of the embodiments of the present application.

Referring now to FIG. 3 , it shows a schematic structural diagram of a computer system 300 suitable for implementing a terminal device or a server according to an embodiment of the present application.

As shown in FIG. 3 , a computer system 300 includes a central processing unit (CPU) 301 that can operate according to a program stored in a read-only memory (ROM) 302 or loaded from a storage section 308 into a random-access memory (RAM) 303 Instead, various appropriate actions and processes are performed. In the RAM 303, various programs and data necessary for the operation of the system 300 are also stored. The CPU 301 , ROM 302 , and RAM 303 are connected to each other via a bus 304 . An input/output (I/O) interface 305 is also connected to the bus 304 .

The following components are connected to the I/O interface 305: an input section 306 including a keyboard, a mouse, etc.; an output section 307 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker; a storage section 308 including a hard disk, etc. and a communication section 309 including a network interface card such as a LAN card, a modem, or the like. The communication section 309 performs communication processing via a network such as the Internet. A drive 310 is also connected to the I/O interface 305 as needed. A removable medium 311, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is mounted on the drive 310 as necessary so that a computer program read therefrom is installed into the storage section 308 as necessary.

In particular, according to an embodiment of the present disclosure, the process described above with reference to FIG. 1 may be implemented as a computer software program. For example, an embodiment of the present disclosure includes a magnetic resonance imaging method based on a Wave-CAIPI gradient encoding field and a deep learning model, which includes a computer program tangibly embodied on a machine-readable medium, the computer program including a method for executing Program code for the method in Figure 1. In such an embodiment, the computer program may be downloaded and installed from a network via communication portion 309 and/or installed from removable media 311 .

The flowchart and block diagrams in the Figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, program segment, or part of code that includes one or more logical functions for implementing specified executable instructions. It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or they may sometimes be executed in the reverse order, depending upon the functionality involved. It should also be noted that each block of the block diagrams and/or flowchart illustrations, and combinations of blocks in the block diagrams and/or flowchart illustrations, can be implemented by a dedicated hardware-based system that performs the specified functions or operations , or may be implemented by a combination of dedicated hardware and computer instructions.

The units or modules involved in the embodiments described in the present application may be implemented by means of software or by means of hardware. The described units or modules can also be set in a processor, for example, it can be described as: a processor includes a first sub-region generating unit, a second sub-region generating unit, and a display area generating unit. Wherein, the names of these units or modules do not constitute limitations on the units or modules themselves in some cases, for example, the display area generation unit can also be described as "used to generate The cell of the display area of the text".

As another aspect, the present application also provides a computer-readable storage medium, which may be the computer-readable storage medium contained in the aforementioned devices in the above-mentioned embodiments; computer-readable storage media stored in the device. The computer-readable storage medium stores one or more programs, and the aforementioned programs are used by one or more processors to execute the text generation method applied to transparent window envelopes described in this application.

The above description is only a preferred embodiment of the present application and an illustration of the applied technical principle. Those skilled in the art should understand that the scope of the invention involved in this application is not limited to the technical solution formed by the specific combination of the above-mentioned technical features, and should also cover the technical solutions formed by the above-mentioned technical features or without departing from the aforementioned inventive concept. Other technical solutions formed by any combination of equivalent features. For example, a technical solution formed by replacing the above-mentioned features with technical features with similar functions disclosed in (but not limited to) this application.

Claims

A magnetic resonance imaging method based on a wave gradient encoding field and a deep learning model, characterized in that the method comprises:

Introduce a deep generative network model into the magnetic resonance reconstruction model of the wave gradient encoding field to accelerate magnetic resonance imaging;

Make conjugate symmetry to the physical coil channel data through the virtual conjugate coil, and generate the virtual conjugate coil channel data;

The physical coil channel data and the virtual conjugate coil channel data are combined to reconstruct the geometric factor calculation model.
According to the magnetic resonance imaging method based on the wave gradient encoding field and the deep learning model according to claim 1, it is characterized in that, before the introduction of the deep generation network model in the magnetic resonance reconstruction model of the wave gradient encoding field, the method also includes:

Using parallel imaging to acquire multiple undersampled k-spaces corresponding to different channels in Accelerated Magnetic Resonance Imaging,

Based on different coil sensitivity information, artifact-free images of reconstructed under-sampled k-space are obtained.
According to the magnetic resonance imaging method based on the wave gradient encoding field and the deep learning model according to claim 2, it is characterized in that, in the magnetic resonance reconstruction model of the wave gradient encoding field, the depth generation network model is introduced to accelerate the magnetic resonance imaging, including :

Establish the encoding model of the wave gradient field in magnetic resonance imaging: wave(x,y,z)=Em(x,y,z), where E is the encoding matrix,
M represents the offset aliasing caused by the controlled aliasing parallel imaging sampling mode in the phase encoding direction, F x represents the Fourier transform in the x direction, S represents the coil sensitivity information, Psf(k x ,y,z) is the effect of wave gradient field;

By wave(x,y,z)=Em(x,y,z) and
get
Wherein, Psf[x,y,z] is the inverse Fourier transform of Psf(k x ,y,z) in the readout direction.
According to the magnetic resonance imaging method based on the wave gradient encoding field and the deep learning model according to claim 3, it is characterized in that, the virtual conjugate coil is used to perform conjugate symmetry on the physical coil channel data to generate virtual conjugate coil channel data, include:

By wave(x,y,z)=Em(x,y,z) and
get
Among them, wave*(x,y,z) represents the data expanded by the virtual conjugate coil,
Indicates the Psf that expands the number of channels on the basis of the existing Psf[x,y,z].
According to the magnetic resonance imaging method based on the wave gradient encoding field and the deep learning model according to claim 4, it is characterized in that, after the virtual conjugate coil channel data is generated, the method also includes:

Introduce training-free deep generative models.
According to the magnetic resonance imaging method based on the wave gradient encoding field and the deep learning model according to claim 4, it is characterized in that, the physical coil channel data and the virtual conjugate coil channel data are merged to reconstruct the geometric factor calculation model, including:

by formula
Get the geometric factor g-factor after VCC expansion, where,
Indicates the coding matrix after VCC expansion.
A magnetic resonance imaging device based on a wave gradient encoding field and a deep learning model, characterized in that the device includes:

The introduction unit is used to introduce a deep generation network model into the magnetic resonance reconstruction model of the wave gradient encoding field to accelerate magnetic resonance imaging;

A generation unit is used to perform conjugate symmetry on the physical coil channel data through the virtual conjugate coil to generate virtual conjugate coil channel data;

The reconstruction unit is used to combine the physical coil channel data and the virtual conjugate coil channel data to reconstruct the geometric factor calculation model.
According to claim 7, based on the magnetic resonance imaging device based on wave gradient coding field and deep learning model, it is characterized in that, before the introduction of deep generation network model in the magnetic resonance reconstruction model of wave gradient coding field, the device also includes :

Using parallel imaging to acquire multiple undersampled k-spaces corresponding to different channels in Accelerated Magnetic Resonance Imaging,

Based on different coil sensitivity information, artifact-free images of reconstructed under-sampled k-space are obtained.
A computer device, comprising a memory, a processor, and a computer program stored on the memory and operable on the processor, characterized in that, when the processor executes the program, it implements any of claims 1-6 described method.
A computer-readable storage medium having stored thereon a computer program for:

When the computer program is executed by the processor, the method according to any one of claims 1-6 is implemented.