Dynamic magnetic resonance imaging method for implicit nerve representation low-rank tensor

Abstract

The invention discloses a dynamic magnetic resonance imaging method of implicit nerve representation low-rank tensor. The method comprises the following steps: learning continuous feature representation of three-dimensional discrete coordinate points by using an implicit neural network for a magnetic resonance image; dividing the continuous features into a plurality of overlapped and non-overlapped basic cubes through a non-local method of continuous representation, and performing similarity clustering on the basic cubes to construct continuous group representation in a unified multi-dimensional space; decomposing the continuous group representation through a coupling low-rank function, and training to obtain a reconstruction model; and for actually acquired k-space data, obtaining a reconstructed image by using the trained reconstruction model. According to the invention, the quality and efficiency of heart movie image reconstruction are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of magnetic resonance imaging, and more specifically, to a dynamic magnetic resonance imaging method of implicit neural representation of low-rank tensors. Background Art

[0002] Magnetic resonance cardiac dynamic movie imaging has important clinical significance for the diagnosis and treatment of cardiovascular diseases. By non-invasively observing cardiac motion, the volume of cardiac chambers, ejection fraction and myocardial mass can be accurately quantified, thus providing important diagnostic basis for doctors, but this method takes a long time to scan.

[0003] In recent years, deep learning-based methods have been successfully applied to dynamic cardiac MRI reconstruction, but most deep learning-based reconstruction methods rely on supervised learning training that requires a large amount of fully sampled data. In dynamic cardiac MRI imaging, it is a huge challenge to obtain fully sampled k-space data. Due to the long scanning time, it is easy to cause fluctuations in breathing or heart rate, resulting in a decrease in image quality. Unsupervised / self-supervised learning methods do not require a large amount of labeled data and can directly learn the prior information of the image from the undersampled data to achieve high-quality image reconstruction. For example, the self-supervised learning method uses a physics-guided self-supervised learning strategy to divide the available undersampled data into training and validation subsets, and only relies on the existing undersampled data for optimization during the training process. These methods can achieve high acceleration rates while maintaining high reconstruction quality. However, the iterative nature of this type of method easily leads to time-consuming calculations of the algorithm. In the unsupervised learning method, some researchers use untrained deep neural networks to guide the network to reconstruct dynamic MR images from accelerated acquired data by combining the regularization term of the physical model. This method can effectively reconstruct dynamic MR images under different acceleration factors without relying on a large amount of training data. However, the initial conditions and network parameter settings of this type of method have a significant impact on the final results. If they are improperly set, they may converge to a suboptimal solution or produce artifacts. Some researchers have introduced implicit neural representation (INR) networks into MRI image reconstruction tasks, reformulating the image reconstruction problem as an optimization problem of a continuous function, reducing the dependence on a large amount of external training data and combining different prior regularizations to restore images from sparsely sampled images. Although the INR-based reconstruction method can have good reconstruction performance in low-dimensional reconstruction tasks, it often requires higher computing resources in high-dimensional reconstruction tasks (such as magnetic resonance cardiac dynamic movie image reconstruction, where the model must learn representations in both the spatial and temporal domains), and the training speed is slow, and the model is prone to overfitting.

[0004] In summary, the existing unsupervised or self-supervised fast MRI cardiac dynamic movie reconstruction methods still have limitations in performance under high acceleration rates, especially under high undersampling rates, the reconstruction quality will be significantly reduced. Summary of the invention

[0005] The purpose of the present invention is to overcome the defects of the prior art and provide a dynamic magnetic resonance imaging method for implicit neural representation of low-rank tensors. The method comprises the following steps:

[0006] For magnetic resonance images, an implicit neural network is used to learn the continuous feature representation of three-dimensional discrete coordinate points;

[0007] Dividing the continuous feature into a plurality of overlapping and non-overlapping basic cubes by a non-local method of continuous representation, and performing similarity clustering on the basic cubes to construct a continuous group representation in a unified multidimensional space;

[0008] Decomposing the continuous group representation by coupling low-rank functions, and then training to obtain a reconstruction model;

[0009] For the actually acquired k-space data, the reconstructed image is obtained using the trained reconstruction model.

[0010] Compared with the prior art, the advantages of the present invention are that it proposes a magnetic resonance cardiac dynamic image reconstruction model based on implicit neural representation and low-rank tensor decomposition, combined with non-local feature decomposition technology based on continuous representation, which can directly restore high-quality images from high-acceleration k-space data. In addition, the present invention uses an unsupervised training model, without the need for fully sampled training data, to restore high-quality cardiac movie images from high-multiple undersampled data, significantly shortening the image reconstruction time.

[0011] Further features and advantages of the present invention will become apparent from the following detailed description of exemplary embodiments of the present invention with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate embodiments of the invention and, together with the description, serve to explain the principles of the invention.

[0013] Figure 1 is a flow chart of a dynamic magnetic resonance imaging method for implicit neural representation of low-rank tensors according to one embodiment of the present invention;

[0014] Figure 2 is a schematic diagram of a non-local decomposition operation of a continuous representation according to an embodiment of the present invention;

[0015] Figure 3It is a general framework diagram of a magnetic resonance cardiac dynamic movie imaging method using implicit neural representation of low-rank tensors according to an embodiment of the present invention;

[0016] Figure 4 is a schematic diagram comparing the effects of image reconstruction according to an embodiment of the present invention and the prior art;

[0017] In the accompanying figure, loss-loss; zerofilling-zero filling; Simlar cubes-similar cubes; Projectedresults-projected results; Training-training; Inference-inference. DETAILED DESCRIPTION

[0018] Various exemplary embodiments of the present invention will now be described in detail with reference to the accompanying drawings. It should be noted that the relative arrangement of components and steps, numerical expressions and numerical values ​​set forth in these embodiments do not limit the scope of the present invention unless otherwise specifically stated.

[0019] The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the invention, its application, or uses.

[0020] Technologies, methods, and equipment known to ordinary technicians in the relevant art may not be discussed in detail, but where appropriate, the technologies, methods, and equipment should be considered as part of the specification.

[0021] In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not limiting. Therefore, other examples of the exemplary embodiments may have different values.

[0022] It should be noted that like reference numerals and letters refer to similar items in the following figures, and therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.

[0023] In general, the dynamic magnetic resonance imaging method of implicit neural representation of low-rank tensors provided by the present invention first uses an implicit neural network to learn a continuous feature representation of a three-dimensional discrete coordinate point; then, the feature is divided into multiple overlapping and non-overlapping basic cubes through a non-local method of continuous representation, and similarity clustering is performed on them to construct a continuous group representation in a unified multidimensional space; finally, the continuous group is efficiently represented by coupling low-rank function decomposition. This method takes into account both the similarity of features between groups and the uniqueness of features within groups, and can achieve fast and high-quality reconstruction of cardiac movie images.

[0024] Specifically, see Figure 1 As shown, the provided method for dynamic magnetic resonance imaging of a low-rank tensor with implicit neural representation includes the following steps:

[0025] Step S110, constructing an implicit neural network for implicitly representing the MR image to learn the continuous feature representation of three-dimensional discrete coordinate points.

[0026] Assuming that Y and X are the k-space data of the multi-coil under-sampled with C coils and the corresponding image to be reconstructed, the reconstruction model for reconstructing X from the under-sampled data is:

[0027]

[0028] Among them, Y j represents the actual undersampled k-space data of the jth coil, A j It represents the encoding operator on the jth coil, A is the encoding operator, which can be expressed by A=MFC, M is the data undersampling operator, F is the Fourier operator, C is the multi-channel coil sensitivity matrix, is the regularization operator. represents the regularization prior of the image to be reconstructed, and λ is the regularization coefficient.

[0029] In the unsupervised learning based INR method, the input features are usually (For cardiac movie data, N = 3) spatial coordinate vector. In order to model the internal continuity of cardiac movie images, in one embodiment, its intensity is represented as a continuous function of three-dimensional spatial coordinates and parameterized using a multilayer perceptron (MLP). For example, X(θ) is used to learn the continuous features through the parameterized continuous function, and the total variation (TV) term is used as the regularization term. The INR-based reconstruction model can be rewritten as:

[0030]

[0031] Among them, R TV represents the total variation (TV) regularization term, and the parameter θ characterizes the internal structural characteristics of the cardiac cine MR image. At the same time, the TV term is introduced to suppress noise and retain the edge details of the image, thereby achieving reconstruction.

[0032] Step S120 , dividing the features into a plurality of overlapping and non-overlapping basic cubes by a non-local method of continuous representation, and performing similarity clustering on the basic cubes to construct a continuous group representation in a unified multidimensional space.

[0033] Step S120 is used to implement non-local decomposition based on continuous representation. Specifically, firstly, given an observation coordinate set For the dth dimension (d=1,…,N), the minimum and maximum values ​​of the observed coordinates are ad 、b d , therefore, the entire N-dimensional continuous domain space is expressed as:

[0034]

[0035] In one embodiment, a continuous representation-based nonlocal (CRNL) self-similarity metric function h(·) can be used to compactly and efficiently represent the image to be reconstructed in the N-dimensional continuous domain space, and an implicit neural network can be used. To approximate it. For example, by optimizing the following equation:

[0036]

[0037] Among them, h -1 (·) represents the inverse operation of the CRNL method, ψ represents the regularization function, and v represents the initial coordinates of the observation.

[0038] The continuous domain Divide into multiple non-overlapping basic units by dimension. If δ d To represent the number of partition intervals of the corresponding dimension, p represents the hyperparameter that determines the size of the basic unit, then it can be combined into T overlapping and L non-overlapping cubes, whose volume V and number T / L are:

[0039] V=pδ1×pδ2×pδ3×…×(b d -a d ) (5)

[0040] T=(n1-p+1)(n2-p+1)(n3-p+1) (6)

[0041]

[0042] Among them, δ1, δ2 and δ3 represent the interval size of dividing the continuous space in each dimension, and n1, n2 and n3 represent the number of sub-regions after the continuous space is divided in each dimension.

[0043] Figure 2 is a schematic diagram of the nonlocal decomposition operation of continuous representation, where n1=n2=5, p=2, D t represents a basic continuous cube, represents the key continuous cube, Indicates that from the lth key cube The sets of overlapping and non-overlapping cubes constructed in this way are and Next, the global autocorrelation between consecutive cubes is constructed. Specifically, for each non-overlapping consecutive cube Hope in overlapping cube collection Find the most similar S consecutive cubes. and D t The similarity is measured by the function value f on the grid coordinates of the two θ The Euclidean distance is calculated and expressed as:

[0044]

[0045] Among them, v key and v is and D t According to the above distance index, select S sets. In and Collection Then, each and the S most similar cubes D t Stack them and write them into a continuous group. In this way, we can get L continuous groups. To make each successive set of observation coordinates lie in the same cubic space, each original observation coordinate can be adjusted, for example (i.e. similar cube The observation coordinate set in ) is concatenated with the cube index s to construct a new coordinate vector w v,s , expressed as:

[0046]

[0047] Among them, the definition and They are Minimum coordinate value in the 1st, 2nd, and 3rd dimensions.

[0048] Finally, the lth All new coordinates w in v,s Collect them to get a new set of observation coordinates:

[0049]

[0050] And define a new multivariable function:

[0051]

[0052] Step S130, characterizing the continuous group by coupling low-rank function decomposition, and then obtaining a reconstruction model.

[0053] For example, low-rank tensor function decomposition is used to efficiently fit multivariate functions of each continuous group. Specifically, the shared factor function is introduced N+1 factor functions (MLP networks) are shared among all groups, while each group retains an independent core tensor C l Finally, a reconstruction model of magnetic resonance cardiac dynamic movie imaging based on implicit neural representation of low-rank tensor can be obtained:

[0054]

[0055] Among them, × d represents the tensor product operation of the dth mode, s l (v) represents the low-rank tensor function of the lth continuous group, N represents the dimension of the continuous space, C l represents the core matrix of the lth consecutive group.

[0056] Step S140, using the set loss function to train the reconstruction model.

[0057] During the training process, the encoding operator The continuous representation result is projected back into k-space and is converted to k-space by using the data consistency loss. and total variation (TV) regularization The model parameters are optimized jointly, and the final loss function is composed of data consistency and image regularization terms.

[0058] Figure 3 This is the overall framework of the magnetic resonance cardiac dynamic movie imaging method using implicit neural representation of low-rank tensors, where Figure 3 (a) is a schematic diagram of the basic cube division. Figure 3 (b) is a schematic diagram of grouping. Figure 3 (c) is a schematic diagram of coupled low-rank tensor decomposition. Represents the encoding operator, in the training phase In the inference stage

[0059] Step S150 , obtaining a reconstruction result using the trained reconstruction model for the actually acquired k-space data.

[0060] In the model reasoning or application stage, in order to further improve the reconstruction quality, an additional data consistency constraint step is introduced in the k-space domain. Figure 3As shown in the reasoning section of , the predicted k-space undersampling data is replaced by the measured k-space undersampling data to ensure data consistency. Subsequently, the images of each coil are obtained by applying a 2D inverse Fourier transform to the updated combined k-space data. These coil images are then combined using the adaptive coil combination method to finally generate a high-quality reconstructed image.

[0061] To further verify the effect of the present invention, the reconstruction results of the present invention (labeled as Ours) are compared with those of the prior art at high acceleration rates of 12, 16, and 21. The prior art includes:

[0062] Unsupervised algorithm ConvDecode combined with physical model (KPSlavkova, et al., "An untrained deep learning method for reconstructing dynamic MR images from accelerated model-based data," Magnetic Resonance in Medicine, vol. 89, no. 4, pp. 1617–1633, 2023.);

[0063] Supervised algorithm Modl (HK Aggarwal, MP Mani, and M. Jacob, "MoDL: Model-based deep learning architecture for inverse problems," IEEE Transactions on Medical Imaging, vol. 38, no. 2, pp. 394–405, 2018.);

[0064] Supervised algorithm DCNet (Huang, W., Ke, Z., Cui, Z.X., Cheng, J., Qiu, Z., Jia, S., & Liang, D., "Deep low-rank plus sparse network for dynamic MR imaging," Med. Image Anal., vol. 73, p. 102190, 2021.);

[0065] The supervised algorithm LSNet (Cheng, J., Chen, H., Ke, Z., Huang, W., Qiu, Z., & Liang, D., "Learning data consistency and its application to dynamic MR imaging," IEEE Trans. Med. Imag., vol. 40, no. 11, pp. 3140–3153, 2021.) was used to compare the reconstruction results at high acceleration rates of A = 12, 16, and 21. The experimental results are shown in Figure 2. Figure 4 shown.

[0066] Figure 4 : is a schematic diagram comparing the effects of the present invention and the prior art, and the first frame reconstructed when the undersampling factor is R=12, 16, and 21 and the YT reverse dynamic heart movie image are compared. Figure 4 It can be seen that compared with the prior art, the image reconstructed by the present invention has the smallest error and the reconstructed heart structure is clearest.

[0067] In summary, compared with the prior art, the present invention has the following advantages:

[0068] 1) In the image restoration task, the present invention compactly models the complex image pattern by decomposing the low-rank tensor function based on INR, and fully exploits the low-rank characteristics of each dimension of the high-dimensional image, so as to efficiently restore a clear image;

[0069] 2) The present invention develops a magnetic resonance cardiac dynamic movie imaging method based on unsupervised learning of implicit neural representation of low-rank tensors. It adopts a non-local method based on continuous representation to compactly represent the cardiac movie data to be reconstructed, and combines a coupled low-rank function decomposition method to efficiently fit the represented observation data, further improving the quality of image reconstruction.

[0070] 3) It has been verified that the use of the present invention can significantly improve the reconstruction performance in cardiac movie dynamic MRI reconstruction tasks. For example, high-quality images can be reconstructed from more than 20 times undersampled data, further shortening the cardiac MRI scanning time.

[0071] The present invention may be a system, a method and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for causing a processor to implement various aspects of the present invention.

[0072] Computer readable storage medium can be a tangible device that can hold and store instructions used by an instruction execution device. Computer readable storage medium can be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination thereof. More specific examples (non-exhaustive list) of computer readable storage medium include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disk read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanical encoding device, for example, a punch card or a convex structure in a groove on which instructions are stored, and any suitable combination thereof. The computer readable storage medium used here is not interpreted as a transient signal itself, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagated by a waveguide or other transmission medium (for example, a light pulse by an optical fiber cable), or an electrical signal transmitted by a wire.

[0073] The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to each computing / processing device, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network can include copper transmission cables, optical fiber transmissions, wireless transmissions, routers, firewalls, switches, gateway computers, and / or edge servers. The network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in the computer-readable storage medium in each computing / processing device.

[0074] The computer program instructions for performing the operation of the present invention may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages, such as Smalltalk, C++, Python, etc., and conventional procedural programming languages, such as "C" language or similar programming languages. Computer-readable program instructions may be executed entirely on a user's computer, partially on a user's computer, as an independent software package, partially on a user's computer, partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., using an Internet service provider to connect via the Internet). In some embodiments, an electronic circuit, such as a programmable logic circuit, a field programmable gate array (FPGA), or a programmable logic array (PLA), may be personalized by utilizing the state information of the computer-readable program instructions, and the electronic circuit may execute the computer-readable program instructions, thereby realizing various aspects of the present invention.

[0075] Various aspects of the present invention are described herein with reference to the flow charts and / or block diagrams of the methods, devices (systems) and computer program products according to embodiments of the present invention. It should be understood that each box of the flow chart and / or block diagram and the combination of each box in the flow chart and / or block diagram can be implemented by computer-readable program instructions.

[0076] These computer-readable program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, thereby producing a machine, so that when these instructions are executed by the processor of the computer or other programmable data processing device, a device that implements the functions / actions specified in one or more boxes in the flowchart and / or block diagram is generated. These computer-readable program instructions can also be stored in a computer-readable storage medium, and these instructions cause the computer, programmable data processing device, and / or other equipment to work in a specific manner, so that the computer-readable medium storing the instructions includes a manufactured product, which includes instructions for implementing various aspects of the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0077] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device so that a series of operating steps are performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to implement the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0078] The flow charts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the systems, methods and computer program products according to multiple embodiments of the present invention. In this regard, each box in the flow chart or block diagram can represent a part of a module, a program segment or an instruction, and a part of the module, a program segment or an instruction contains one or more executable instructions for realizing the specified logical function. In some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flow chart, and the combination of the boxes in the block diagram and / or flow chart can be implemented by a dedicated hardware-based system that performs the specified function or action, or can be implemented by a combination of dedicated hardware and computer instructions. It is well known to those skilled in the art that it is equivalent to implement it by hardware, implement it by software, and implement it by combining software and hardware.

[0079] Embodiments of the present invention have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or technical improvements in the marketplace, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein. The scope of the present invention is defined by the appended claims.

Claims

1. A dynamic magnetic resonance imaging method for implicit neural representation of low-rank tensors, comprising the following steps: For magnetic resonance images, an implicit neural network is used to learn the continuous feature representation of three-dimensional discrete coordinate points; Dividing the continuous feature into a plurality of overlapping and non-overlapping basic cubes by a non-local method of continuous representation, and performing similarity clustering on the basic cubes to construct a continuous group representation in a unified multidimensional space; Decomposing the continuous group representation by coupling low-rank functions, and then training to obtain a reconstruction model; For the actually acquired k-space data, the reconstructed image is obtained using the trained reconstruction model.

2. The method according to claim 1, characterized in that The continuous group representation in the unified multidimensional space is constructed according to the following steps: Given a set of observation coordinates The minimum and maximum values ​​of the observed coordinates for the dth dimension are a d 、b d , and then the entire N-dimensional continuous domain space is expressed as: Where d = 1,…,N, N is the number of dimensions; Based on the continuous representation of non-local self-similarity metric function h(·) to represent the image to be reconstructed in N-dimensional continuous domain space, and using implicit neural network To approximate, the optimization equation is set to: The continuous domain D fθ Divided into multiple non-overlapping basic cubes according to the dimension, and then combined into T overlapping cubes and L non-overlapping cubes, where the volume V and number of cubes are expressed as: V=pδ1×pδ2×pδ2×…×(b d -a d ) T=(n1-p+1)(n2-p+1)(n3-p+1) Among them, δ d Represents the number of division intervals of the corresponding dimension, p represents the hyperparameter that determines the size of the basic cube, and it can be combined into T overlapping and L non-overlapping cubes. δ1, δ2 and δ3 represent the interval size of dividing the continuous space in each dimension, and n1, n2 and n3 represent the number of sub-regions after the continuous space is divided in each dimension. For each non-overlapping contiguous cube In a collection of overlapping cubes Find the most similar S consecutive cubes, where two similar cubes and D t The similarity is measured by the function value f on the grid coordinates of the two θ distance measurement; Select S sets according to the calculated distance In and Collection The most similar cube in each and the S most similar cubes D t Stacking is performed to form a continuous group to obtain L continuous groups Adjust Similarity Cube The values ​​of the first three dimensions of the observation coordinate set in are concatenated with the index s of the cube to construct a new coordinate vector w v,s , expressed as: The first All new coordinates w in v,s Collect them to get a new set of observation coordinates: And define a new multivariable function: Among them, the definition and They are Minimum coordinate value in the 1st, 2nd, and 3rd dimensions.

3. The method according to claim 2, characterized in that The distance is the Euclidean distance, expressed as: Among them, v key and v is and D t The coordinates in a set that correspond one to one.

4. The method according to claim 1, characterized in that: The implicit neural network is constructed based on a multi-layer perceptron.

5. The method according to claim 2, characterized in that: The reconstruction model is expressed as: Among them, × d represents the tensor product operation of the dth mode, s l (v) represents the low-rank tensor function of the lth continuous group, N represents the dimension of the continuous space, C l represents the core matrix of the lth consecutive group, R TV represents the total variation regularization term.

6. The method according to claim 1, characterized in that In the reconstruction model training stage, the continuous representation result is projected back to the k-space through the encoding operator A=MFC. Where M is the data undersampling operator, F is the Fourier operator, and C is the multi-channel coil sensitivity matrix.

7. The method according to claim 1, characterized in that The loss function for training the reconstruction model includes data consistency loss and image regularization term loss.

8. A computer-readable storage medium having a computer program stored thereon, wherein: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

9. A computer device comprising a memory and a processor, wherein a computer program capable of running on the processor is stored in the memory, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

Images