WO2020135014A1

WO2020135014A1 - Method for building medical imaging model, device, apparatus, and storage medium

Info

Publication number: WO2020135014A1
Application number: PCT/CN2019/124238
Authority: WO
Inventors: 王珊珊; 梁栋; 柯子文; 刘新
Original assignee: 深圳先进技术研究院
Priority date: 2018-12-27
Filing date: 2019-12-10
Publication date: 2020-07-02
Also published as: CN111383741B; CN111383741A

Abstract

A method for building a medical imaging model, a device, an apparatus, and a storage medium. The method comprises: obtaining undersampled k-space data of a medical image to serve as a training sample (101); inputting the training sample into a pre-built original neural network to perform training (102); and using the trained original neural network as a target medical imaging model (103).

Description

Method, device, equipment and storage medium for establishing medical imaging model

This application requires the priority of the Chinese patent application filed on December 27, 2018 with the Chinese Patent Office, application number 201811611862.2. The entire contents of this application are incorporated by reference in this application.

Technical field

The present application relates to the technical field of magnetic resonance imaging, for example, to a method, device, equipment and storage medium for establishing a medical imaging model.

Background technique

Magnetic resonance cardiac imaging is a non-invasive imaging technique that can provide rich spatial and temporal information for clinical diagnosis. Due to the limitations of MRI physics and hardware, MRI cardiac film imaging is often accompanied by shortcomings such as long scanning time and slow imaging speed. Therefore, on the premise of ensuring the imaging quality, accelerated magnetic resonance cardiac film imaging is particularly important.

Accelerated magnetic resonance cardiac film imaging methods commonly used in related technologies include parallel imaging, compressed sensing technology, and deep learning methods. For example, dynamic generalized automatic calibration partial parallel acquisition (TGRAPPA), adaptive sensitivity coding (TSENSE) using time filters, focal underdetermination system (ktFOCUSS) using time-frequency sparsity, and Kalki method using dynamic redundancy (kt SLR), low rank sparse matrix (L+S), etc. This type of method uses the spatial information of the data to fill the undersampled K-space data. In the field of magnetic resonance cardiac imaging, magnetic resonance dynamic imaging (D5C5) and convolutional recurrent neural network (CRNN) based on cascaded convolutional network can also be used in the field of magnetic resonance cardiac imaging. These two methods use neural networks, You can directly learn the mapping relationship from undersampled images to fully sampled images. Traditional parallel imaging or compressed sensing technology does not make use of big data priors, and this iterative optimization method is often time-consuming and difficult to select parameters. The neural network methods (D5C5, CRNN) based on deep learning also have obvious shortcomings. They all construct the entire network in the image domain, and do not make full use of the frequency domain information. The methods in the related art cannot reconstruct magnetic resonance cardiac film images more accurately.

Summary of the invention

Embodiments of the present invention provide a method, device, equipment, and storage medium for establishing a medical imaging model, so as to achieve more accurate and faster reconstruction of a K-space under-sampled medical image.

An embodiment of the present invention provides a method for establishing a medical imaging model. The method includes:

Obtain K-space under-sampled data of medical images as training samples;

Input the training samples to the pre-built original neural network for training, wherein the original neural network includes a frequency domain network FDN and an image domain network SDN, the frequency domain network FDN and the image domain network SDN Connected by inverse Fourier transform IFFT;

The trained original neural network is used as the target medical imaging model.

An embodiment of the present invention also provides a device for establishing a medical imaging model. The device includes:

The training sample acquisition module is set to acquire K-space undersampled data of medical images as training samples;

A training module configured to input the training samples into the pre-built original neural network for training, wherein the original neural network includes a frequency domain network FDN and an image domain network SDN, and the frequency domain network FDN and the The image domain network SDN is connected by inverse Fourier transform IFFT;

The target medical imaging model determination module is set to use the trained original neural network as the target medical imaging model.

An embodiment of the present invention also provides a device, which includes:

One or more processors;

Memory for storing one or more programs,

When the one or more programs are executed by the one or more processors, the one or more processors implement the medical imaging model establishment method described in any of the embodiments of the present invention.

An embodiment of the present invention also provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the method for establishing a medical imaging model according to any of the embodiments of the present invention is implemented.

The technical solution of the embodiment of the present invention acquires K-space under-sampled data of medical images as training samples, and can directly learn the mapping relationship from under-sampled images to fully sampled images. Furthermore, the training samples are input to the pre-built original neural network for training, wherein the original neural network includes a frequency domain network FDN and an image domain network SDN, the frequency domain network FDN and the image domain network The SDN is connected by inverse Fourier transform IFFT, which can make full use of the frequency domain information and image domain information of the image, and establish the medical imaging model more accurately. Furthermore, the trained original neural network is used as the target medical imaging model to realize the rapid and accurate reconstruction of the under-sampled medical images. The above technical solution solves the traditional parallel imaging or compressed sensing technology, which does not use big data priors, is time-consuming and cumbersome to adjust parameters, and the neural network method based on deep learning cannot fully utilize the frequency domain information and cannot accurately undersample. The problem of image reconstruction is realized, which can learn the characteristics of the frequency domain and the image domain at the same time, fully combine the information of the frequency domain and the image domain, and quickly and accurately reconstruct the undersampled medical image, which can avoid time-consuming iterative solution steps and cumbersome The process of adjusting the parameters.

BRIEF DESCRIPTION

1a is a flowchart of a method for establishing a medical imaging model provided in Embodiment 1 of the present invention;

1b is a schematic structural diagram of an original neural network provided in Embodiment 1 of the present invention;

2a is a flowchart of a method for establishing a medical imaging model provided in Embodiment 2 of the present invention;

2b is a comparison of results of different magnetic resonance cardiac film reconstruction methods provided in Example 2 of the present invention;

3 is a flowchart of an apparatus for establishing a medical imaging model provided in Embodiment 3 of the present invention;

4 is a schematic structural diagram of a device provided in Embodiment 4 of the present invention.

detailed description

The present application will be further described in detail below with reference to the drawings and embodiments. It can be understood that the specific embodiments described herein are only used to explain the present application, rather than limit the present application. In addition, it should be noted that, in order to facilitate description, the drawings only show parts, but not all structures related to the present application.

Example one

FIG. 1a is a flowchart of a method for establishing a medical imaging model according to Embodiment 1 of the present invention. This embodiment is applicable to the case of establishing a medical imaging model, and is particularly suitable for establishing an imaging model of K-space under-sampled data. The method may be executed by a device for establishing a medical imaging model, and the device may be implemented by hardware and/or software. The device may be integrated into a device (such as a computer) for execution, and specifically includes the following steps:

Step 101: Acquire K-space under-sampled data of a medical image as a training sample.

Illustratively, the medical image may be a magnetic resonance cardiac movie image. K-space is the dual space of ordinary space under Fourier transform. It is mainly used in the imaging analysis of magnetic resonance imaging. Others such as the design of RF waveforms in magnetic resonance imaging and the preparation of the initial state in quantum computing also use the concept of k-space.

K-space under-sampled data refers to under-sampled K-space data.

This step obtains K-space undersampled data for model training.

In addition, the full-sampled medical image corresponding to the K-space under-sampled data of the medical image should also be obtained, which is used to calculate the loss of the model later.

Step 102: Input the training samples to the pre-built original neural network for training. Wherein, the original neural network includes a frequency domain network FDN and an image domain network SDN, and the frequency domain network FDN and the image domain network SDN are connected by an inverse Fourier transform IFFT.

Optionally, the frequency domain network includes a first preset number of frequency domain modules Fnet, wherein each frequency domain module Fnet includes a second preset number of three-dimensional convolution layers 3D Conv and a frequency domain data consistency layer KDC .

The frequency domain module Fnet may also only include the second preset number of three-dimensional convolution layers 3D Conv, excluding the frequency domain data consistency layer KDC.

Optionally, the training sample is used as the input of the first three-dimensional convolutional layer of the first frequency domain module;

Using the output result of the previous three-dimensional convolutional layer of the frequency domain module as the input of the next three-dimensional convolutional layer of the frequency domain module;

The output result of the last three-dimensional convolutional layer of the frequency domain module is used as the input of the frequency domain data consistency layer KDC of the frequency domain module, and the output result of the frequency domain data consistency layer KDC is used as the frequency domain module's output Output result

The output result of the previous frequency domain module is used as the input of the next frequency domain module, and the output result of the last frequency domain module is used as the output result of the frequency domain network.

The output result of the frequency domain network FDN undergoes the inverse Fourier transform IFFT as the input of the image domain network SDN.

Suppose that the frequency domain network is composed of M frequency domain modules Fnet _m (Fnet _m , m=1,...,M), each frequency domain module contains L 3D convolutional layers (3DConv) and one frequency domain data is consistent Layer (KDC). The forward process of the frequency domain network can be expressed by the following formula:

The first frequency domain module (m=1):

Subsequent frequency domain modules (m=2,...,M)

Among them, σ represents a nonlinear activation function, which can be a nonlinear activation function commonly used in neural network models. k _u represents the K-space undersampled data of the input medical image,

They are the convolution kernel and offset term of the lth convolutional layer in the mth frequency domain module, l=1,...,L,m=1,...,M.

Represents the output of the lth convolution layer in the mth frequency domain module. Except for the last convolutional layer of each frequency domain module, all other convolutional layers of each frequency domain module are activated by a nonlinear activation function σ. After extracting features through the convolution layer, you get

Then, the frequency domain data consistency layer KDC is used to correct the k-space predicted by the network, as shown in formula (9).

Among them, KDC is used to perform the frequency domain data consistency operation, the formula is as follows:

Indicate right

The result of the correction. Let the set of K-space undersampled data coordinates of all collected medical images be Ω. If the k-space coordinates (k _x , k _y ) are within the set Ω, then

It will be corrected by the real collected k-space points. λ is used to control the degree of data consistency. If λ→∞, the actual sampling point can be directly replaced

The corresponding point.

The final output of the frequency domain network FDN is

Correct

Then the inverse Fourier transform can be used to obtain the image domain data S ₀ , which is also the input of the image domain network, as shown in the following formula (10).

If the frequency domain module does not include KDC, the final output of the frequency domain network FDN is

Optionally, the image domain network SDN includes a third preset number of image domain modules Snet, wherein each image domain module includes a fourth preset number of three-dimensional convolution layers 3D Conv, an image domain data consistency layer IDC And a residual connection.

Each image domain module may contain a fourth preset number of three-dimensional convolutional layers 3D Conv and a residual connection, may not contain image domain data consistency layer IDC.

Optionally, the output result of the frequency domain network after inverse Fourier transform IFFT is used as the input of the first three-dimensional convolution layer of the first image domain module;

Using the output result of the previous three-dimensional convolutional layer of the image domain module as the input of the next three-dimensional convolutional layer of the image domain module;

After summing the output of the last three-dimensional convolutional layer of the image domain module and the input of the first three-dimensional convolutional layer of the image domain module, input the image domain data consistent layer IDC of the image domain module , The output result of the image domain data consistency layer IDC is used as the output result of the image domain module;

The output result of the previous image domain module is used as the input of the next image domain module, and the output result of the last image domain module is used as the output result of the image domain network.

The forward process of the image domain network can be expressed by the following formula, assuming that the image domain network contains N image domain modules Snet _n (Snet _n , n=1,...,N):

The first image domain module (n=1):

Subsequent image domain modules (n=2,...,N)

Among them, IDC is used to perform the image domain data consistency operation, the formula is as follows:

These are the convolution kernel and offset term of the lth convolutional layer in the nth image domain module, l=1,...,L,n=1,...,N.

Is the output of the lth convolution layer in the nth image domain. Except for the last (Lth) convolutional layer, all other convolutional layers are activated by a nonlinear activation function σ. After extraction features convolutional layer is introduced residual learning (implemented via residual connections), the equation (19) is S _n is the result of learning the residual (i.e., residual connection). Then the image domain data S _n coherency operation (IDC). IDC has more conversion between the frequency domain and the image domain than KDC, that is, formula (21), and then the image domain data consistency operation is performed by formulas (22)-(23). λ is used to control the degree of data consistency. S _n is the result of performing IDC on S _n . If the image domain module does not include IDC, then the output result of the sum of the output of the last three-dimensional convolutional layer of the image domain module and the input of the first three-dimensional convolutional layer of the image domain module is taken as The output result of the image domain module.

After the frequency domain network (FDN) and the image domain network (SDN), the result S _{N of the} reconstruction of the K-space undersampled data of the medical image can be obtained.

The frequency domain network is used to predict the fully sampled k-space, the image domain network is used to extract image features, and the two networks are connected by inverse Fourier transform. At the same time, the frequency domain network and the image domain network can use the data consistency layer to correct the k-space data.

Step 103: Use the trained original neural network as a target medical imaging model.

The original neural network obtained after the training samples are trained as the target medical imaging model can be used to reconstruct the K-space under-sampled data of medical images.

Exemplarily, the original neural network of the embodiment of the present invention is shown in FIG. 1b. The method of this embodiment is used for magnetic resonance cardiac film imaging. The input of the network is K-space undersampled data, and the output is a reconstructed magnetic resonance cardiac film image. The original neural network consists of a frequency domain network (FDN) and an image domain network (SDN), which are connected by an inverse Fourier transform (IFFT). The frequency domain network is composed of M frequency domain modules (Fnet), and each frequency domain module includes a second preset number of 3D convolution layers (3D Conv) and a frequency domain data consistency layer (KDC). The image domain network is composed of N image domain modules (Snet), and each image domain module includes a fourth preset number of 3D convolutional layers (3D Conv), an image domain data consistency layer (IDC), and a residual connection .

Example 2

FIG. 2a is a flowchart of a method for establishing a medical imaging model provided by Embodiment 2 of the present invention. In this embodiment, based on the foregoing embodiment, optionally, the training sample is input to the pre-built The original neural network for training includes: selecting a set number of training samples; sequentially obtaining a training sample and inputting it into a pre-established original frequency domain network to obtain a preliminary output result, and inputting the preliminary output result after inverse Fourier transform Go to the original image domain network to get the model output; return to perform the operation of obtaining a training sample input to the original frequency domain network until the preset training end condition is reached.

On this basis, in some embodiments, the method of establishing a medical imaging model further includes: acquiring K-space under-sampled data to be imaged; inputting the K-space under-sampled data to be imaged into the target medical imaging model after training In the process, the reconstructed medical image is obtained.

As shown in FIG. 2a, the method of this embodiment specifically includes the following steps:

Step 201: Acquire K-space under-sampled data of a medical image as a training sample.

Step 202: Select a set number of training samples.

Step 203: Obtain a training sample in sequence and input it into a pre-established original frequency domain network to obtain a preliminary output result, and then input the inverse Fourier transform to the original image domain network to obtain a model output result.

Step 204: Determine whether the preset training end condition is reached. If yes, go to step 205. If no, go back to step 203.

The preset training end condition is that the loss of the neural network of a preset number of training samples of a preset number of training samples (for example, 98%) reaches a preset threshold.

The loss function can be a loss function commonly used in neural network models, for example, the loss function can be expressed as

Ploss means loss, S _N means the final reconstruction result of under-sampled K-space data after frequency domain network, IFFT, image domain network, that is, the output result of the image domain network, that is, the output result of the Nth image domain module (you can not Consider the role of IDC in the image domain data consistency layer), S represents the fully sampled image corresponding to the K-space undersampled data; if KDC and IDC are considered, then

Ploss here refers to a single training sample. Assuming that there are multiple samples, the Ploss of each sample is calculated and summed using the above formula for calculating Ploss to obtain the total loss.

It should be noted that multiple training samples can be trained in parallel during training. The corresponding loss is to calculate and sum the Ploss of each sample using the above formula for calculating Ploss. The corresponding training end condition is that the sum of the loss of multiple samples reaches the pre- Stop training when setting a threshold.

S represents the fully sampled image corresponding to the K-space undersampled data. By optimizing the loss function to optimize the neural network model, the target medical imaging model is obtained.

Step 205: Use the trained original neural network as the target medical imaging model.

Step 206: Acquire K-space under-sampled data to be imaged; input the K-space under-sampled data to be imaged into the target medical imaging model after training to obtain a reconstructed medical image.

Taking magnetic resonance cardiac imaging as an example, in order to demonstrate the effectiveness of the embodiments of the present invention on magnetic resonance cardiac imaging, the method of the embodiments of the present invention is compared with the current mainstream compressed sensing and deep learning methods. At a factor of 4 times, the reconstruction results are shown in Figure 2b. Figure 2b shows a comparison of the results of different magnetic resonance cardiac film reconstruction methods. The four methods are: the time-frequency sparsity of the focal underdetermination system k-t FOCUSS, the dynamic redundancy Kalki method k-t SLR, the magnetic resonance dynamic imaging D5C5 based on the cascaded convolution network, and the method of this embodiment. (a) indicates a fully-captured image, (b) indicates a sampling template, (c) indicates a zero-fill image, (d) indicates a kt FOCUSS reconstruction result, (e) indicates a kt SLR reconstruction result, (f) indicates a D5C5 reconstruction result, ( g) represents the reconstruction result of the method proposed in the embodiment of the present invention; (h), (i), (j), (k) are the reconstruction corresponding to (d), (e), (f), (g) respectively The residual map between the result and the fully sampled image (a), where the smaller the residual, the better the reconstruction effect. It can be seen from the experimental results that the method of the embodiment of the present invention has the best reconstruction result for magnetic resonance cardiac film imaging. This fully illustrates the effectiveness of the method of the embodiments of the present invention.

The technical solution of this embodiment selects a set number of training samples; sequentially obtains a training sample and inputs it into a pre-established original frequency domain network to obtain a preliminary output result, and performs inverse Fourier transform on the preliminary output result and inputs it to Obtain the output of the model in the original image domain network; return to perform the operation of obtaining a training sample input to the original frequency domain network until the preset training end condition is reached, the network model can be optimized, and the network model can be used more accurately For image reconstruction. Furthermore, the K-space under-sampling data to be imaged is obtained; the K-space under-sampling data to be imaged is input into the target medical imaging model after training to obtain a reconstructed medical image, and the K-space under-sampling data is directly used for reconstruction.

Example Three

3 is a schematic structural diagram of a medical imaging model creation device provided in Embodiment 3 of the present invention. The apparatus for establishing a medical imaging model provided by an embodiment of the present invention can execute the method for establishing a medical imaging model provided by any embodiment of the present application. The specific structure of the apparatus is as follows: a training sample acquisition module 31, a training module 32, and target medicine The imaging model determination module 33.

Based on the above technical solution, the training module 32 may be specifically set as:

The frequency domain network includes a first preset number of frequency domain modules Fnet, wherein each frequency domain module Fnet includes a second preset number of three-dimensional convolution layers 3D Conv and a frequency domain data consistency layer KDC.

Based on the above technical solution, the training module 32 may specifically be set as follows: the image domain network SDN includes a third preset number of image domain modules Snet, wherein each image domain module includes a fourth preset number of three-dimensional volumes Multilayer 3D Conv, an image domain data consistency layer IDC, and a residual connection.

Based on the above technical solution, the device for establishing a medical imaging model may further include: a frequency domain network module and an image domain network module.

A frequency domain network module configured to use the training sample as an input of the first three-dimensional convolutional layer of the first frequency domain module;

The image domain network module is configured to use the output of the frequency domain network after inverse Fourier transform IFFT as the input of the first three-dimensional convolution layer of the first image domain module;

Based on the above technical solution, the training module 32 may be specifically configured to: select a set number of training samples;

Obtain a training sample in turn and input it into a pre-established original frequency domain network to obtain a preliminary output result, and perform an inverse Fourier transform on the preliminary output result and input it into the original image domain network to obtain a model output result;

Return to perform the operation of obtaining a training sample input to the original frequency domain network until the preset training end condition is reached.

Based on the above technical solution, the device for establishing the medical imaging model may further include a reconstruction module.

The reconstruction module is set to obtain K-space under-sampled data to be imaged;

The K-space under-sampled data to be imaged is input into the target medical imaging model after training to obtain a reconstructed medical image.

The apparatus for establishing a medical imaging model provided by an embodiment of the present invention can execute the method for establishing a medical imaging model provided by any embodiment of the present application, and has functional modules and beneficial effects corresponding to the execution method.

Example 4

4 is a schematic structural diagram of a device according to Embodiment 4 of the present invention. As shown in FIG. 4, the device includes a processor 40, a memory 41, an input device 42, and an output device 43; the number of processors 40 in the device may be One or more, one processor 40 is taken as an example in FIG. 4; the processor 40, the memory 41, the input device 42 and the output device 43 in the device may be connected through a bus or other means, and FIG. 4 is taken as an example through a bus connection .

The memory 41 is a computer-readable storage medium that can be used to store software programs, computer executable programs, and modules, such as program instructions/modules corresponding to the method of establishing a medical imaging model in the embodiment of the present invention (for example, the The training sample acquisition module 31, the training module 32 and the target medical imaging model determination module 33 in the establishment device). The processor 40 executes various functional applications and data processing of the device by running software programs, instructions, and modules stored in the memory 41, that is, implementing the above-described method of establishing a medical imaging model.

The memory 41 may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system and application programs required by at least one function; the storage data area may store data created according to the use of the terminal, and the like. In addition, the memory 41 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one magnetic disk storage device, a flash memory device, or other non-volatile solid-state storage devices. In some examples, the memory 41 may further include memories remotely provided with respect to the processor 40, and these remote memories may be connected to the device through a network. Examples of the aforementioned network include, but are not limited to, the Internet, intranet, local area network, mobile communication network, and combinations thereof.

The input device 42 can be used to receive the K-space under-sampled data of the input medical image and generate signal input related to the user settings and function control of the device. The output device 43 may include a display device such as a display screen.

Example 5

Embodiment 5 of the present invention also provides a storage medium containing computer-executable instructions, which when executed by a computer processor is used to perform a method for establishing a medical imaging model, the method includes:

Obtain K-space under-sampled data of medical images as training samples;

Of course, a storage medium containing computer-executable instructions provided by an embodiment of the present invention is not limited to the method operation described above, and can also execute the medical imaging model provided by any embodiment of the present application. Related operations in the establishment method.

Through the above description of the embodiments, those skilled in the art can clearly understand that the present application can be implemented by software and necessary general hardware, and of course can also be implemented by hardware, but in many cases the former is a better embodiment . Based on this understanding, the technical solutions of the present application can essentially be embodied in the form of software products, and the computer software products can be stored in computer-readable storage media, such as computer floppy disks, Read-only memory (Read-Only Memory, ROM), random access memory (Random Access Memory, RAM), flash memory (FLASH), hard disk or optical disc, etc., including several instructions to make a computer device (which can be a personal computer, Server, or network equipment, etc.) to execute the method described in each embodiment of the present application.

It is worth noting that in the embodiment of the above medical imaging model building device, the various units and modules included are only divided according to functional logic, but it is not limited to the above division, as long as the corresponding functions can be achieved; In addition, the specific names of the functional units are only for the purpose of distinguishing each other, and are not used to limit the protection scope of the present application.

Note that the above are only optional embodiments of the present application and applied technical principles. Those skilled in the art will understand that the present application is not limited to the optional embodiments described herein, and that those skilled in the art can make various obvious changes, readjustments, and substitutions without departing from the scope of protection of the present application. Therefore, although the present application has been described in more detail through the above embodiments, the present application is not limited to the above embodiments, and may include other equivalent embodiments without departing from the concept of the present application. The scope is determined by the scope of the appended claims.

Claims

A method for establishing a medical imaging model includes:

Obtain K-space under-sampled data of medical images as training samples;

Input the training samples to the pre-built original neural network for training, wherein the original neural network includes a frequency domain network FDN and an image domain network SDN, the frequency domain network FDN and the image domain network SDN Connected by inverse Fourier transform IFFT;

The trained original neural network is used as the target medical imaging model.
The method according to claim 1, wherein the frequency domain network includes a first preset number of frequency domain modules Fnet, wherein each frequency domain module Fnet includes a second preset number of three-dimensional convolution layers 3D Conv and A frequency domain data consistency layer KDC.
The method according to claim 1, wherein the image domain network SDN includes a third preset number of image domain modules Snet, wherein each image domain module includes a fourth preset number of three-dimensional convolution layers 3D Conv, An image domain data consistency layer IDC and a residual connection.
The method of claim 2, further comprising:

Using the training sample as the input of the first three-dimensional convolutional layer of the first frequency domain module;

Using the output result of the previous three-dimensional convolutional layer of the frequency domain module as the input of the next three-dimensional convolutional layer of the frequency domain module;

The output result of the last three-dimensional convolutional layer of the frequency domain module is used as the input of the frequency domain data consistency layer KDC of the frequency domain module, and the output result of the frequency domain data consistency layer KDC is used as the frequency domain module's output Output result

The output result of the previous frequency domain module is used as the input of the next frequency domain module, and the output result of the last frequency domain module is used as the output result of the frequency domain network.
The method of claim 2, further comprising:

Taking the output of the frequency domain network after inverse Fourier transform IFFT as the input of the first three-dimensional convolutional layer of the first image domain module;

Using the output result of the previous three-dimensional convolutional layer of the image domain module as the input of the next three-dimensional convolutional layer of the image domain module;

After summing the output of the last three-dimensional convolutional layer of the image domain module and the input of the first three-dimensional convolutional layer of the image domain module, input the image domain data consistent layer IDC of the image domain module , The output result of the image domain data consistency layer IDC is used as the output result of the image domain module;

The output result of the previous image domain module is used as the input of the next image domain module, and the output result of the last image domain module is used as the output result of the image domain network.
The method according to claim 1, wherein the inputting the training samples to the pre-built original neural network for training includes:

Select a set number of training samples;

Obtaining a training sample in turn and inputting it into a pre-established original frequency domain network to obtain a preliminary output result, inverse Fourier transforming the preliminary output result and inputting it into the original image domain network to obtain a model output result;

Return to perform the operation of obtaining a training sample input to the original frequency domain network until the preset training end condition is reached.
The method of claim 1, further comprising:

Obtain the K-space undersampled data to be imaged;

The K-space under-sampled data to be imaged is input into the target medical imaging model after training to obtain a reconstructed medical image.
A device for establishing a medical imaging model includes:

The training sample acquisition module is set to acquire K-space under-sampled data of medical images as training samples;

A training module configured to input the training samples into the pre-built original neural network for training, wherein the original neural network includes a frequency domain network FDN and an image domain network SDN, and the frequency domain network FDN and the The image domain network SDN is connected by inverse Fourier transform IFFT;

The target medical imaging model determination module is set to use the trained original neural network as the target medical imaging model.
A device, including:

One or more processors;

Memory for storing one or more programs,

When the one or more programs are executed by the one or more processors, the one or more processors implement the medical imaging model establishment method according to any one of claims 1-7.
A computer-readable storage medium on which a computer program is stored, characterized in that when the program is executed by a processor, the method for establishing a medical imaging model according to any one of claims 1-7 is realized.