Magnetic resonance sodium imaging image super-resolution method, device and processing equipment

Abstract

The invention provides a magnetic resonance sodium imaging image super-resolution method, a magnetic resonance sodium imaging image super-resolution device and processing equipment, which are used for configuring a novel super-resolution processing scheme and configuring a corresponding magnetic resonance sodium imaging image super-resolution model, so that the spatial resolution of a magnetic resonance sodium imaging image can be effectively improved, and the image super-resolution effect of the magnetic resonance sodium imaging image can be improved. In addition, the number of image layers can be effectively increased, the high-quality magnetic resonance sodium imaging super-resolution effect is achieved, data support of more comprehensive and detailed case information can be provided for related clinical work, and then it can be guaranteed that the corresponding medical service quality is further improved.

Description

Technical Field

[0001] This application relates to the field of magnetic resonance imaging, and particularly to a magnetic resonance sodium imaging super-resolution method, apparatus, and processing device. Background Art

[0002] In medical imaging, magnetic resonance sodium imaging (which can also be referred to as sodium element magnetic resonance imaging, 23 Na magnetic resonance imaging, etc.) technology has made significant progress in recent years. By detecting the distribution of sodium ions in the body, magnetic resonance sodium imaging has been applied to the diagnosis of diseases such as osteoarthritis, multiple sclerosis, and brain tumors. In these diseases, magnetic resonance sodium imaging can provide disease-related pathological information, thus helping doctors better understand the progression and impact of the diseases.

[0003] However, the application of magnetic resonance sodium imaging faces a series of technical challenges. Sodium nuclei will generate four energy levels in a static magnetic field (main magnetic field), and there are three possible transition modes between these energy levels. Due to 23 the interaction between the electric quadrupole moment of the Na atomic nucleus and the surrounding electric field gradient as well as macromolecules, these transition processes will decay rapidly, resulting in a double-exponential transverse relaxation phenomenon. Specifically, the relaxation time is relatively short, which directly affects the signal-to-noise ratio, resolution, and measurement accuracy of the image. Therefore, in clinical applications, due to the limitations of imaging equipment, the image resolution of magnetic resonance sodium imaging often fails to achieve the ideal effect.

[0004] To solve this problem, image super-resolution technology has been introduced. Image super-resolution technology aims to generate high-resolution images by processing low-resolution images, thereby improving the clarity and detail level of the images. In medical imaging, it can significantly enhance the image resolution, further improving the accuracy of disease diagnosis and reflecting more pathophysiological information.

[0005] However, existing super-resolution methods mainly focus on improving the resolution of a single image, that is, generating high-resolution images from a single low-resolution image. This method is effective when dealing with multi-layer scanned images, especially when dealing with low-resolution CT images, but faces challenges in the super-resolution of magnetic resonance sodium imaging.

[0006] In magnetic resonance sodium imaging, since the number of low-resolution image layers obtained during the imaging process is relatively small, it is very difficult to generate high-resolution images with more layers. This is mainly because traditional super-resolution techniques assume that the number of layers of the input and output images is the same, and in practical applications, this assumption often does not hold.

[0007] That is to say, the general image super-resolution method in the prior art is not suitable for the application scenario of magnetic resonance sodium imaging, and it is difficult to meet the image super-resolution requirements of magnetic resonance sodium imaging that take into account both high resolution and the number of image layers. Summary of the Invention

[0008] The present application provides a method, apparatus, and processing device for magnetic resonance sodium imaging image super-resolution, which configure a novel super-resolution processing scheme and accordingly configure a corresponding magnetic resonance sodium imaging image super-resolution model. In this way, not only can the spatial resolution of magnetic resonance sodium imaging images be effectively improved, but also the number of image layers can be effectively increased, achieving a high-quality magnetic resonance sodium imaging super-resolution effect. It can provide more comprehensive and detailed data support for relevant clinical work, and further ensure the improvement of the corresponding medical service quality.

[0009] In a first aspect, the present application provides a method for magnetic resonance sodium imaging image super-resolution, the method comprising: Obtaining an initial magnetic resonance sodium imaging image whose resolution is to be improved; Inputting the initial magnetic resonance sodium imaging image into a magnetic resonance sodium imaging image super-resolution model, wherein the magnetic resonance sodium imaging image super-resolution model is used to increase the number of layers and resolution of the input magnetic resonance sodium imaging image. The magnetic resonance sodium imaging image super-resolution model is pre-trained by sample magnetic resonance sodium imaging images. During the operation of the magnetic resonance sodium imaging image super-resolution model, the input magnetic resonance sodium imaging image is converted into a feature matrix F1 with a specification of w1, h1, n×64, where w represents width, h represents height, and n represents the number of layers. The feature matrix F1 is sampled into a feature matrix F2 with the same specification. The feature matrix obtained after weighting the feature matrix F2 is added to the feature matrix F1 to obtain a feature matrix F3. The size of the feature matrix F3 is enlarged to obtain a feature matrix F4 with a specification of w1×8, h1×8, n×8. The feature matrix F4 is downsampled to obtain a feature matrix F5 with a specification of w2, h2, n×8×r 2 where r represents a scaling factor, w1×8 = w2 / r 2 and h1×8 = h2 / r 2 The feature matrix F5 is convolved to obtain a feature matrix F6 with a specification of w2, h2, n, and the feature matrix F6 is converted into an output image; Extracting the target magnetic resonance sodium imaging image output by the magnetic resonance sodium imaging image super-resolution model.

[0010] In a second aspect, the present application provides a device for magnetic resonance sodium imaging image super-resolution, the device comprising: An acquisition unit for acquiring an initial magnetic resonance sodium imaging image whose resolution is to be improved; An input unit for inputting an initial magnetic resonance sodium imaging image into a magnetic resonance sodium imaging image super-resolution model, where the magnetic resonance sodium imaging image super-resolution model is used to increase the number of layers and resolution of the input magnetic resonance sodium imaging image, and the magnetic resonance sodium imaging image super-resolution model is pre-trained by sample magnetic resonance sodium imaging images. During the operation of the magnetic resonance sodium imaging image super-resolution model, the input magnetic resonance sodium imaging image is converted into a feature matrix F1 with specifications of w1, h1, n×64, where w represents width, h represents height, and n represents the number of layers. The feature matrix F1 is sampled into a feature matrix F2 of the same specification. The feature matrix after weighting the feature matrix F2 is added to the feature matrix F1 to obtain a feature matrix F3. The size of the feature matrix F3 is enlarged to obtain a feature matrix F4 with specifications of w1×8, h1×8, n×8. The feature matrix F4 is downsampled to obtain a feature matrix F5 with specifications of w2, h2, n×8×r 2 where r represents the scaling factor, and w1×8 = w2 / r 2 and h1×8 = h2 / r 2 The feature matrix F5 is convolved to obtain a feature matrix F6 with specifications of w2, h2, n, and the feature matrix F6 is converted into an output image; An extraction unit for extracting the target magnetic resonance sodium imaging image output by the magnetic resonance sodium imaging image super-resolution model.

[0011] In a third aspect, the present application provides a processing device, including a processor and a memory. A computer program is stored in the memory. When the processor calls the computer program in the memory, it executes the method provided in the first aspect of the present application or any possible implementation manner of the first aspect of the present application.

[0012] In a fourth aspect, the present application provides a computer-readable storage medium, which stores multiple instructions suitable for being loaded by a processor to execute the method provided in the first aspect of the present application or any possible implementation manner of the first aspect of the present application.

[0013] From the above content, the following beneficial effects of the present application can be obtained: For the magnetic resonance sodium imaging super-resolution target, the present application configures a novel super-resolution processing scheme and configures a corresponding magnetic resonance sodium imaging image super-resolution model accordingly. In this way, not only can the spatial resolution of the magnetic resonance sodium imaging image be effectively improved, but also the number of image layers can be effectively increased, realizing a high-quality magnetic resonance sodium imaging super-resolution effect, which can provide more comprehensive and detailed data support for relevant clinical work, and further ensure the improvement of the corresponding medical service quality. Description of the Drawings

[0014] To more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application. For those skilled in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0015] Figure 1 It is a schematic flowchart of a method for super-resolution of magnetic resonance sodium imaging images of the present application; Figure 2 It is a schematic diagram of a scenario of the overall model processing architecture of the present application; Figure 3 It is a schematic diagram of a scenario of the processing logic of the deep upsampling block of the present application; Figure 4 It is a schematic diagram of a scenario of the processing logic of the dense feature extraction block of the present application; Figure 5 It is a schematic structural diagram of a magnetic resonance sodium imaging image super-resolution device of the present application; Figure 6 It is a schematic structural diagram of a processing device of the present application. Specific embodiments

[0016] The following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, rather than all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present application.

[0017] The terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and do not have to be used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments described here can be implemented in an order different from that shown or described here. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or modules does not necessarily have to be limited to those steps or modules clearly listed, but may include other steps or modules not clearly listed or inherent to these processes, methods, products, or devices. In the present application, the naming or numbering of steps does not mean that the steps in the method flow must be executed in the time / logical sequence indicated by the naming or numbering. The named or numbered process steps can be changed in the execution order according to the technical purpose to be achieved, as long as the same or similar technical effects can be achieved.

[0018] The division of modules in this application is a logical division. In actual implementation, there may be other division methods. For example, multiple modules can be combined or integrated into another system, or some features can be ignored or not executed. Additionally, the displayed or discussed coupling, direct coupling, or communication connection between each other can be through some interfaces. The indirect coupling or communication connection between modules can be in an electrical or other similar form, which is not limited in this application. Moreover, the modules or sub-modules described as separate components may or may not be physically separated, may or may not be physical modules, or may be distributed to multiple circuit modules. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this application.

[0019] Before introducing the magnetic resonance sodium imaging image super-resolution method provided by this application, the background content involved in this application is first introduced.

[0020] The magnetic resonance sodium imaging image super-resolution method, device, and computer-readable storage medium provided by this application can be applied to a processing device, which is configured with a novel super-resolution processing scheme and accordingly configures a corresponding magnetic resonance sodium imaging image super-resolution model. In this way, not only can the spatial resolution of the magnetic resonance sodium imaging image be effectively improved, but also the number of image layers can be effectively increased, achieving a high-quality magnetic resonance sodium imaging super-resolution effect, which can provide more comprehensive and detailed data support for relevant clinical work, and further ensure that the corresponding medical service quality is further improved.

[0021] The magnetic resonance sodium imaging image super-resolution method mentioned in this application can be executed by a magnetic resonance sodium imaging image super-resolution device, or different types of processing devices such as a server, a physical host, or a user equipment (UE) that integrates the magnetic resonance sodium imaging image super-resolution device. Among them, the magnetic resonance sodium imaging image super-resolution device can be implemented in a hardware or software manner. The UE can specifically be a terminal device such as a smart phone, a tablet computer, a notebook computer, a desktop computer, or a personal digital assistant (PDA). The processing device can be set in the form of a device cluster.

[0022] It can be understood that in specific applications, the key point of the solution of this application lies in performing corresponding super-resolution processing on magnetic resonance sodium imaging images based on a pre-configured magnetic resonance sodium imaging image super-resolution model. In this case, when implementing the magnetic resonance sodium imaging image super-resolution method of this application or a processing device equipped with the application service corresponding to the magnetic resonance sodium imaging image super-resolution method of this application, usually only the required data processing capabilities need to be satisfied. The specific device types and specific device deployment forms involved are relatively flexible and can be flexibly configured according to actual needs. This application does not make specific limitations.

[0023] If further training of the magnetic resonance sodium imaging image super-resolution model is involved, further adaptive adjustments can be made according to the actual situation.

[0024] As an example, the processing device can be divided into two parts, that is, the processing device can include a first processing device that executes the training task of the magnetic resonance sodium imaging image super-resolution model and a second processing device that executes the application task of the magnetic resonance sodium imaging image super-resolution model.

[0025] Furthermore, in some cases, the solution of this application may further involve further data support work for auxiliary disease diagnosis such as lesion identification / localization, etc., then further data analysis and processing can be correspondingly performed, and further adaptive adjustments can be made to the processing device. Corresponding to the above example, in some cases, the processing device may further include a third processing device that executes data analysis and processing based on the super-resolution results of magnetic resonance sodium imaging images.

[0026] Furthermore, in some cases, the solution of this application may further involve further image display, then the processing device itself can be configured with a display screen (including a touch screen), or an external display device, or an external device with a display screen to meet the requirements of the image display function.

[0027] Next, the magnetic resonance sodium imaging image super-resolution method provided by this application will be introduced.

[0028] First, refer to Figure 1 , Figure 1 which shows a schematic flowchart of a process of the magnetic resonance sodium imaging image super-resolution method provided by this application. The magnetic resonance sodium imaging image super-resolution method provided by this application may specifically include the following steps S101 to step S103: Step S101, obtain an initial magnetic resonance sodium imaging image with a resolution to be improved; It can be understood that corresponding to the actual super-resolution requirements for magnetic resonance sodium imaging images, the magnetic resonance sodium imaging image with the currently required resolution to be improved or the resolution to be improved can be obtained.

[0029] For the convenience of description, in this application, the magnetic resonance sodium imaging images in different cases are respectively denoted as the initial magnetic resonance sodium imaging image, the sample magnetic resonance sodium imaging image, and the target magnetic resonance sodium imaging image.

[0030] Among them, the magnetic resonance imaging technology itself is an existing concept. In actual situations, the magnetic resonance sodium imaging images obtained by its scanning usually exist in the form of an image sequence. Considering that this is not the focus of the solution of this application, this application will not elaborate too much.

[0031] As an example, the initial magnetic resonance sodium imaging image obtained here can specifically be a low-resolution magnetic resonance sodium imaging image acquired in the sagittal plane of the bFFE-UTE sequence. The corresponding parameter settings are TR / TE = 8.0 / 0.28 ms, FOV = 260×260×90 mm, the slice gap is -3 mm, and the image sequence dimension is (w, h, n) = (60, 60, 30), where w represents the width, h represents the height, and n represents the number of slices.

[0032] Among them, for the acquisition and processing of the initial magnetic resonance sodium imaging image, it can either be manually entered or automatically obtained. The device can extract it from the local storage space or the storage space of other devices, or the device can receive the images sent by other devices. This is all acceptable, corresponding to the flexible application requirements in actual situations.

[0033] In addition, in the actual application of the solution of this application, it is usually carried out in the form of a work task, that is, the magnetic resonance sodium imaging image super-resolution task. The initiation of the task can either be manually entered or automatically initiated. The device can initiate it independently according to the preset autonomous initiation strategy / rules, or the device can receive the tasks sent by other devices. This is all acceptable. Similar to the above, it meets the diverse application requirements of the solution.

[0034] Step S102: Input the initial magnetic resonance sodium imaging image into the magnetic resonance sodium imaging image super-resolution model. The magnetic resonance sodium imaging image super-resolution model is used to increase the number of layers and resolution of the input magnetic resonance sodium imaging image. The magnetic resonance sodium imaging image super-resolution model is pre-trained by sample magnetic resonance sodium imaging images. During the operation of the magnetic resonance sodium imaging image super-resolution model, the input magnetic resonance sodium imaging image is converted into a feature matrix F1 with the specification of w1, h1, n×64, where w represents the width, h represents the height, and n represents the number of layers. The feature matrix F1 is sampled into a feature matrix F2 with the same specification. The feature matrix after weighting the feature matrix F2 is added to the feature matrix F1 to obtain a feature matrix F3. The size of the feature matrix F3 is enlarged to obtain a feature matrix F4 with the specification of w1×8, h1×8, n×8. The feature matrix F4 is downsampled to obtain a feature matrix F5 with the specification of w2, h2, n×8×r, where r represents the scaling factor, and w1×8 = w2 / r 2 and h1×8 = h2 / r 2 ; 2 The feature matrix F5 is convolved to obtain a feature matrix F6 with the specification of w2, h2, n, and the feature matrix F6 is converted into an output image; It can be seen that this application designs a novel magnetic resonance sodium imaging image super-resolution logic / processing solution, which can be specifically implemented by a corresponding processing model. In this regard, this application can pre-configure a corresponding magnetic resonance sodium imaging image super-resolution model, which is trained by sample magnetic resonance sodium imaging images in the prior model training work / session.

[0035] Specifically, referring to Figure 2 a schematic diagram of a scenario of the overall model processing architecture of this application shown, the processing architecture of the magnetic resonance sodium imaging image super-resolution model (or the asymmetric medical image super-resolution model) of this application, or the asymmetric medical image super-resolution method designed by this application, may have the following main contents: 1) The convolutional block performs preliminary image feature extraction, converting the input image sequence (low-resolution image) of the model into a matrix in the form of (w, h, n), so as to convert the input magnetic resonance sodium imaging image of the model into a feature matrix F1 (w1, h1, n×64), which corresponds to the input layer of the model; 2) The feature sampling block performs feature enhancement, sampling the feature matrix F1 into a feature matrix F2 (w1, h1, n×64) with the same specification; 3) The self-attention mechanism introduced by the self-attention gating is advanced to weight the feature matrix F2; 4) Corresponding to the matrix addition step of the self-attention mechanism, add the feature matrix (w1, h1, n×64) after weighting the feature matrix F2 to the feature matrix F1 to obtain the feature matrix F3 (w1, h1, n×64); 5) Enlarge the size of the feature matrix by the depth upsampling block, and enlarge the size of the feature matrix F3 to obtain the feature matrix F4 (w1×8, h1×8, n×8); 6) Downsample the feature matrix F4 to obtain the feature matrix F5 (w2, h2, n×8×r 2 ), where w1×8 = w2 / r 2 and h1×8 = h2 / r 2 , and r can be understood as an intermediate coefficient introduced in the parameter calculation process of this application; 7) Convolve the feature matrix F5 by the convolution block to obtain the feature matrix F6 (w2, h2, n); 8) Convert the feature matrix F6 into the output image (high-resolution image) as the model output. Specifically, this may involve model structures such as the fully connected layer in the corresponding model output step. Considering that this is not the focus of the solution of this application and that the output layer is also a mature concept in the neural network of the machine model, no specific expansion is made here.

[0036] Under the above processing mechanism, it can be seen that this application combines specific image specification configurations, magnifies and then restores the image through upsampling and downsampling to extract features, strengthens the original features, reduces the error of the generated layer image, and effectively increases the number of image layers while improving the resolution.

[0037] In terms of details, the feature sampling block involved in this application, its processing logic or working process, can sequentially include upsampling (on the input side), activation function, and downsampling (on the output side).

[0038] Correspondingly, as an exemplary embodiment, sample the feature matrix F1 into the feature matrix F2 of the same specification by the feature sampling block, which can specifically include: 2.1) Through the following formula, sample the feature matrix F1 with the specification of w1, h1, n×64 into the feature matrix F11 with the specification of w×r, h×r, n×64 / r 2 : , It can be seen that this is the upsampling step involved in the feature sampling block. Starting from the quantization formula, a more specific and practical implementation scheme is given.

[0039] Among them, mod is the modulo operator.

[0040] As an example, r that acts as a scaling factor can be taken as 2, i.e., r = 2. In this case, the element value at each position in the enlarged matrix X' is calculated from the original matrix X based on the scaling factor r = 2 using the above formula.

[0041] 2.2) Process the feature matrix F11 through the activation function ReLU to obtain the feature matrix F12; Among them, ReLU, that is, the Linear rectification function, is a mature concept or a common network structure in the neural network architecture. This application introduces it into the specific working logic / structure of the feature sampling block responsible for processing the feature matrix F1 into the feature matrix F2.

[0042] By inserting ReLU between the upsampling link above and the downsampling link below, in terms of specific implementation, it helps to improve the generalization ability and representation ability of the model.

[0043] 2.3) Downsample the feature matrix F12 into a feature matrix F2 with specifications of w, h, n×64.

[0044] After downsampling, the enhancement of the feature matrix F1 is completed, and the feature matrix F2 that can be input into subsequent processing is obtained.

[0045] Furthermore, for the depth upsampling block involved in this application, its processing logic or working process refers to Figure 3 A schematic diagram of a scenario showing the processing logic of the depth upsampling block of this application, which can sequentially include the first dense feature extraction block (corresponding to the input side), the first convolutional block, the second dense feature extraction block, the second convolutional block, the third dense feature extraction block, the third convolutional block, the first upsampling, the second upsampling, and the third upsampling (corresponding to the output side).

[0046] Correspondingly, as an exemplary embodiment, the depth upsampling block enlarges the size of the feature matrix F3 to obtain a feature matrix F4 with specifications of w1×8, h1×8, n×8. Specifically, it can include: After the feature matrix F3 with specifications of w1, h1, n×64 is sequentially subjected to depth feature extraction processing through 3 depth feature extraction units, 3 upsamplings are performed to obtain a feature matrix F4 with specifications of w1×8, h1×8, n×8. Among them, the depth feature extraction unit sequentially includes 1 dense feature extraction block and 1 convolutional block; The specifications of the feature matrices before and after the processing of the dense feature extraction block remain unchanged (the changes in the specifications of the feature matrices before and after the deep feature extraction unit are mainly achieved by the convolutional block). The specifications of the feature matrix after the first deep feature extraction process are w1, h1, n×128, the specifications of the feature matrix after the second deep feature extraction process are w1, h1, n×256, and the specifications of the feature matrix after the third deep feature extraction process are w1, h1, n×512; The specifications of the feature matrix after the first upsampling are w1×2, h1×2, n×128, the specifications of the feature matrix after the second upsampling are w1×4, h1×4, n×32, and the specifications of the feature matrix after the third upsampling are w1×8, h1×8, n×8.

[0047] It can be easily seen that in the embodiments herein, the present application designs a more complex processing result / logic for the deep upsampling process involved in the deep upsampling block in combination with the specific image specification configuration, that is, 3 groups of "dense feature extraction block + convolutional block" + 3 groups of upsampling, so as to better achieve the processing effect of the solution of expanding the feature matrix F3 to obtain the feature matrix F4.

[0048] For the dense feature extraction block among them, corresponding Figure 4 A schematic diagram of a scenario showing the processing logic of the dense feature extraction block of the present application. As an exemplary embodiment, there is: In the dense feature extraction block, the input of the dense feature extraction block is respectively connected to the inputs of the first convolutional block, the second convolutional block, and the third convolutional block. The output of the first convolutional block is respectively connected to the inputs of the second convolutional block and the third convolutional block. The output of the second convolutional block is connected to the input of the third convolutional block. The input of the feature sampling block is respectively connected to the input of the dense feature extraction block, the output of the first convolutional block, the output of the second convolutional block, and the output of the third convolutional block. The input of the self-attention gating is connected to the feature sampling block. The input of the dense feature extraction block, the output of the first convolutional block, the output of the second convolutional block, the output of the third convolutional block, and the output of the self-attention gating are added together as the output.

[0049] In contrast Figure 4 , from the perspective of specific feature matrix processing, there can be: 5.1) The input feature matrix F3 is subjected to feature extraction through the first convolutional block to obtain the feature matrix F31; After adding the feature matrix F3 and the feature matrix F31, the result is input into the second convolutional block for feature extraction to obtain the feature matrix F32; After adding the feature matrix F3, the feature matrix F31, and the feature matrix F32, the result is input into the third convolutional block for feature extraction to obtain the feature matrix F33; 5.4) After adding the feature matrices F3, F31, F32, and F33 through matrix addition, the feature matrix F34 is obtained. The feature matrix F34 is further enhanced through the feature sampling block and the self-attention gate to obtain the feature matrix F35; Among them, the self-attention mechanism involved in the process of processing the feature matrix F34 into the feature matrix F35 can refer to the processing involved in the process of processing the feature matrix F2 into the feature matrix F3 mentioned above.

[0050] 5.5) Add the feature matrices F34 and F35 through matrix addition to obtain the feature matrix F36 representing the depth features.

[0051] At this time, the feature matrix F36 can continue to be upsampled 3 times to improve the image resolution.

[0052] The above content is an introduction to the specific model structure / working logic involved in the magnetic resonance sodium imaging image super-resolution model of this application. It is easy to understand that in specific applications, as mentioned above, it may involve prior model training processing. In this regard, the method of this application can also involve corresponding model training links.

[0053] Correspondingly, as an exemplary embodiment, before inputting the initial magnetic resonance sodium imaging image into the magnetic resonance sodium imaging image super-resolution model in step S102, the method of this application can further include: Obtain a sample magnetic resonance sodium imaging image and make annotations; Based on the sample magnetic resonance sodium imaging image, train the initial model to obtain the magnetic resonance sodium imaging image super-resolution model.

[0054] Among them, the annotation process can be understood as configuring the theoretical or standard model processing result for the sample magnetic resonance sodium imaging image, that is, the high-resolution magnetic resonance sodium imaging image, to assist in training the model's pertinence for the super-resolution processing of the magnetic resonance sodium imaging image during the model training process.

[0055] In specific operations, the annotation process can be completed either manually or by handing it over to the corresponding automated annotation tool. When the automated annotation tool is specifically applied, the corresponding automated annotation logic also needs to be pre-configured.

[0056] As an example, the annotation of the sample magnetic resonance sodium imaging image can specifically also use a ready-made high-resolution image, such as a high-resolution magnetic resonance sodium imaging image collected in the sagittal plane of the bFFE-UTE sequence.

[0057] In this way, the completed annotated sample magnetic resonance sodium imaging image can be used for specific model training.

[0058] During the model training process, it generally includes the following content: In each model training session, a training sample is input into the model, enabling the model to perform super-resolution processing on the corresponding magnetic resonance sodium imaging (MRSI) image to achieve forward propagation. Then, based on the super-resolution result of the MRSI image output by the model, the loss function is calculated in combination with the annotation, and the model parameters are optimized according to the calculation result of the loss function to achieve backward propagation. After a large number of trainings like this, if the preset model training requirements such as the number of training times, training duration, or prediction accuracy are met, the model training can be completed, and a super-resolution model of MRSI images that can be put into practical applications can be obtained.

[0059] Among them, it can be understood that the specific model training architecture and the specific loss function adopted during the training process can either adopt existing solutions, or be further optimized and improved based on existing solutions, or novel self-developed solutions can also be adopted. These are all acceptable and can be configured according to the actual situation.

[0060] As an example, as a model training architecture, for sample MRSI images, they can be specifically divided into a training set, a validation set, and a test set according to 7:2:1 for input into the model training work.

[0061] In addition, during the model training session, other aspects of configuration work can also be involved, such as hyperparameters, model performance evaluation, etc. These can also be flexibly configured according to actual needs.

[0062] For model performance evaluation, this application can specifically use the existing index of Structural Similarity (SSIM) to evaluate the structural similarity between two images, and calculate the similarity between the model output image and the provided high-resolution image from three aspects: brightness, contrast, and structure. The corresponding specific formula is as follows: , , Among them, and are respectively the means of all pixel values in the matrix of image , is the covariance of image , and are respectively the variances of all pixel values in the matrix of image , and b is the image depth, and the specific value can be 8.

[0063] The value range of SSIM is [-1, 1]. The larger the value, the more similar it indicates, and the model performance is evaluated accordingly.

[0064] Additionally, to further improve the model training effect, the present application can further process the sample magnetic resonance sodium imaging images to obtain a more abundant sample size.

[0065] Correspondingly, as an exemplary embodiment, the obtaining of the sample magnetic resonance sodium imaging images specifically may include: Obtaining initial sample magnetic resonance sodium imaging images, where the initial sample magnetic resonance sodium imaging images are an image sequence; According to the preset number of high-resolution image layers corresponding to the model output, interpolating each adjacent layer of the low-resolution images in the initial sample magnetic resonance sodium imaging images. If k - 1 layers of images are inserted between adjacent layers, after performing matrix addition on the original adjacent layer image matrices, multiply them by 1 / k, 2 / k,..., (k - 1) / k respectively to obtain k - 1 matrices, and convert the k - 1 matrices into images to obtain the target sample magnetic resonance sodium imaging images with the number of layers being the preset number of high-resolution image layers as the sample magnetic resonance sodium imaging images, where k - 1 is the number of interpolation layers.

[0066] It can be understood that in this embodiment, through the sample secondary processing designed by the present application, based on interpolation, the number of image layers is effectively further increased and extended.

[0067] For the above content, it can also be more vividly understood in combination with the following set of examples involved in practical applications.

[0068] S1: Use the bFFE-UTE sequence to sagittally acquire low-resolution and corresponding high-resolution 23 Na MRI sequences. The acquisition parameters are set as TR / TE = 8.0 / 0.28 ms, FOV = 260×260×90 mm, the slice spacings of the low-resolution and high-resolution images are -3 mm and -1.5 mm respectively. Divide the acquired data into a training set, a validation set, and a test set according to 7:2:1. The dimension of the low-resolution image sequence is (w, h, n) = (60, 60, 30), indicating that the image width × height is 60×60, the number of layers is 30, and each layer of the image is an 8-bit single-channel image. The dimension of the high-resolution image sequence is (80, 80, 59); S2: Insert one layer between adjacent layer images of each low-resolution image sequence. The matrix of the newly inserted image is obtained by multiplying the matrix addition of the adjacent layer images by 1 / 2. Finally, expand the low-resolution image matrix from 30 layers to 59 layers; S3: Construct a magnetic resonance sodium imaging image super-resolution model, use the training set to train the model, and during the model training process, use the validation set to optimize the model parameters to obtain the model with the optimal AUC on the validation set. The image processing process of the model is specifically as follows: S31: Transform the expanded image sequence matrix through a convolution operation to obtain F1, where the size of F1 is (60, 60, 59×64). Perform reinforcement processing on F1 through a feature sampling block to obtain F2, and the size of F2 remains (60, 60, 59×64). The specific processing of the feature sampling block is as follows: S311: Enlarge the information through an upsampling algorithm. After upsampling, the size of the F1 matrix is (60×2, 60×2, 59×16); S312: Process the transformed matrix using the ReLU activation function to improve the generalization ability and representation ability of the model; S313: Perform the inverse operation of upsampling on the enlarged information through downsampling to obtain F2, and the size of F2 is (60, 60, 59×64); S32: After weighting F2 through a self-attention module, add it to F1 matrix-wise to obtain F3; S33: Enlarge the size of the feature matrix of F3 through a depth upsampling block to obtain F4 with a size of (60×8, 60×8, 59×8). Perform downsampling on F4 to obtain F5 with a size of (80, 80, 59×8×36). Further perform a convolution operation on F5 to obtain F6 with a size of (80, 80, 59). Convert it into an image to obtain the final high-resolution image sequence. The processing of the depth upsampling block is as follows: S331: Perform depth feature extraction through 3 groups of "dense feature extraction blocks + convolution blocks". After the first group of extraction, the size of the feature matrix is (60, 60, 59×128). After the second group of extraction, the size of the feature matrix is (60, 60, 59×256). After the third group of extraction, the size of the feature matrix is (60, 60, 59×512). The matrix size remains unchanged before and after the processing of the dense feature extraction block. Taking the dense feature extraction block in the first group as an example, the specific processing flow is as follows: S3311: Extract features from the input F3 through the first convolution block to obtain F31; S3312: After adding F3 and F31 matrix-wise, input them into the second convolution block for feature extraction to obtain F32; S3313: After adding F3, F31, and F32 matrix-wise, input them into the third convolution block for feature extraction to obtain F33; S3314: After adding F3, F31, F32, and F33 matrix-wise, obtain F34. Perform reinforcement processing on F34 using a feature sampling block and a self-attention gating to obtain F35.

[0069] S3315: Add F34 and F35 matrix-wise to obtain F36.

[0070] S332: Upsample F36 three times to improve the image resolution and obtain matrix F4. After the first upsampling, the size of the feature matrix is (60×2, 60×2, 59×128). After the second upsampling, the size of the feature matrix is (60×4, 60×4, 59×32). After the third upsampling, the size of the feature matrix is (60×8, 60×8, 59×8).

[0071] Step S103: Extract the target magnetic resonance sodium imaging image output by the magnetic resonance sodium imaging image super-resolution model.

[0072] After processing the current initial magnetic resonance sodium imaging image through the magnetic resonance sodium imaging image super-resolution model to obtain the corresponding target magnetic resonance sodium imaging image with effectively improved resolution and number of layers, the model will output it.

[0073] Correspondingly, the target magnetic resonance sodium imaging image output by the magnetic resonance sodium imaging image super-resolution model can be extracted.

[0074] At this time, it can be understood that for the target magnetic resonance sodium imaging image, which is the super-resolution result of the magnetic resonance sodium imaging image, the following operations can be carried out: local storage, off-site storage, result forwarding, prompt for completion of output processing, result display, or further data processing and analysis (such as the data support work for lesion recognition / localization mentioned above for auxiliary disease diagnosis). Obviously, the specific data application content involved is relatively flexible and can be adjusted according to the pre-set and real-time configured data application strategies / rules. The present application does not make specific limitations.

[0075] Finally, generally speaking, for the above solution content, aiming at the magnetic resonance sodium imaging super-resolution target, the present application configures a novel super-resolution processing solution and configures a corresponding magnetic resonance sodium imaging image super-resolution model accordingly. In this way, not only can the spatial resolution of the magnetic resonance sodium imaging image be effectively improved, but also the number of image layers can be effectively increased, achieving a high-quality magnetic resonance sodium imaging super-resolution effect, which can provide more comprehensive and detailed case information data support for relevant clinical work, and further ensure the further improvement of the corresponding medical service quality.

[0076] The above is the introduction of the magnetic resonance sodium imaging image super-resolution method provided by the present application. To facilitate the better implementation of the magnetic resonance sodium imaging image super-resolution method provided by the present application, the present application also provides a magnetic resonance sodium imaging image super-resolution device from the perspective of functional modules.

[0077] Refer to Figure 5 , Figure 5This is a schematic structural diagram of a magnetic resonance sodium imaging image super-resolution device in this application. In this application, the magnetic resonance sodium imaging image super-resolution device 500 may specifically include the following structures: An acquisition unit 501, configured to acquire an initial magnetic resonance sodium imaging image whose resolution is to be improved; An input unit 502, configured to input the initial magnetic resonance sodium imaging image into a magnetic resonance sodium imaging image super-resolution model. The magnetic resonance sodium imaging image super-resolution model is used to increase the number of layers and resolution of the input magnetic resonance sodium imaging image. The magnetic resonance sodium imaging image super-resolution model is pre-trained by sample magnetic resonance sodium imaging images. During the operation of the magnetic resonance sodium imaging image super-resolution model, the input magnetic resonance sodium imaging image is converted into a feature matrix F1 with a specification of w1, h1, n×64, where w represents the width, h represents the height, and n represents the number of layers. The feature matrix F1 is sampled into a feature matrix F2 with the same specification. The feature matrix obtained after weighting the feature matrix F2 is added to the feature matrix F1 to obtain a feature matrix F3. The size of the feature matrix F3 is enlarged to obtain a feature matrix F4 with a specification of w1×8, h1×8, n×8. The feature matrix F4 is downsampled to obtain a feature matrix F5 with a specification of w2, h2, n×8×r 2 where r represents a scaling factor, and w1×8 = w2 / r 2 and h1×8 = h2 / r 2 The feature matrix F5 is convolved to obtain a feature matrix F6 with a specification of w2, h2, n, and the feature matrix F6 is converted into an output image; An extraction unit 503, configured to extract the target magnetic resonance sodium imaging image output by the magnetic resonance sodium imaging image super-resolution model.

[0078] In an exemplary embodiment, sampling the feature matrix F1 into a feature matrix F2 with the same specification specifically includes: Through the following formula, the feature matrix F1 with a specification of w1, h1, n×64 is sampled into a feature matrix F11 with a specification of w×r, h×r, n×64 / r 2 : , where represents the output, represents the input, and mod represents the modulo operator; The feature matrix F11 is processed through the activation function ReLU to obtain a feature matrix F12; The feature matrix F12 is downsampled into a feature matrix F2 with a specification of w, h, n×64.

[0079] In yet another exemplary embodiment, the size of the feature matrix F3 is enlarged to obtain a feature matrix F4 with specifications of w1×8, h1×8, n×8, including: After the feature matrix F3 with specifications of w1, h1, n×64 is successively subjected to deep feature extraction processing through 3 deep feature extraction units, 3 times of upsampling are performed to obtain a feature matrix F4 with specifications of w1×8, h1×8, n×8. Among them, the deep feature extraction unit successively includes 1 dense feature extraction block and 1 convolutional block; The specifications of the feature matrix before and after the dense feature extraction block processing remain unchanged. The specifications of the feature matrix after the first deep feature extraction processing are w1, h1, n×128, the specifications of the feature matrix after the second deep feature extraction processing are w1, h1, n×256, and the specifications of the feature matrix after the third deep feature extraction processing are w1, h1, n×512; The specifications of the feature matrix after the first upsampling are w1×2, h1×2, n×128, the specifications of the feature matrix after the second upsampling are w1×4, h1×4, n×32, and the specifications of the feature matrix after the third upsampling are w1×8, h1×8, n×8.

[0080] In yet another exemplary embodiment, in the dense feature extraction block, the input of the dense feature extraction block is respectively connected to the inputs of the first convolutional block, the second convolutional block, and the third convolutional block. The output of the first convolutional block is respectively connected to the inputs of the second convolutional block and the third convolutional block. The output of the second convolutional block is connected to the input of the third convolutional block. The input of the feature sampling block is respectively connected to the input of the dense feature extraction block, the output of the first convolutional block, the output of the second convolutional block, and the output of the third convolutional block. The input of the self-attention gating is connected to the feature sampling block. The input of the dense feature extraction block, the output of the first convolutional block, the output of the second convolutional block, the output of the third convolutional block, and the output of the self-attention gating are added together as the output.

[0081] In yet another exemplary embodiment, the device further includes a training unit 504 for: Obtain a sample magnetic resonance sodium imaging image and make annotations; Based on the sample magnetic resonance sodium imaging image, train the initial model to obtain a magnetic resonance sodium imaging image super-resolution model.

[0082] In yet another exemplary embodiment, the obtaining unit 501 is specifically used for: Obtain an initial sample magnetic resonance sodium imaging image, where the initial sample magnetic resonance sodium imaging image is an image sequence; According to the preset number of high-resolution image layers corresponding to the model output, interpolation is performed on each adjacent layer of the low-resolution images in the initial sample magnetic resonance sodium imaging image. If k-1 layers of images are inserted between adjacent layers, after performing matrix addition on the original adjacent layer image matrices, they are respectively multiplied by 1 / k, 2 / k, ……, (k-1) / k to obtain k-1 matrices, and the k-1 matrices are converted into images to obtain the target sample magnetic resonance sodium imaging image with the number of layers being the preset number of high-resolution image layers, which is used as the sample magnetic resonance sodium imaging image, where k-1 is the number of interpolation layers.

[0083] In another exemplary embodiment, the device further includes a display unit 505 for: Display the target magnetic resonance sodium imaging image.

[0084] This application also provides a processing device from the perspective of the hardware structure. Refer to Figure 6 , Figure 6 shows a schematic structural diagram of the processing device of this application. Specifically, the processing device of this application may include a processor 601, a memory 602, and an input / output device 603. When the processor 601 executes the computer program stored in the memory 602, it implements the steps of the magnetic resonance sodium imaging image super-resolution method in the corresponding Figure 1 embodiment; or, when the processor 601 executes the computer program stored in the memory 602, it implements the functions of each unit in the corresponding Figure 5 embodiment. The memory 602 is used to store the computer program required for the processor 601 to execute the magnetic resonance sodium imaging image super-resolution method in the above Figure 1 corresponding embodiment.

[0085] Exemplarily, the computer program can be divided into one or more modules / units. One or more modules / units are stored in the memory 602 and executed by the processor 601 to complete this application. One or more modules / units can be a series of computer program instruction segments capable of performing specific functions, and these instruction segments are used to describe the execution process of the computer program in the computer device.

[0086] The processing device may include, but is not limited to, the processor 601, the memory 602, and the input / output device 603. Those skilled in the art can understand that the schematic diagram is only an example of the processing device and does not constitute a limitation on the processing device. It may include more or fewer components than shown in the figure, or combine certain components, or different components. For example, the processing device may further include a network access device, a bus, etc. The processor 601, the memory 602, the input / output device 603, etc. are connected through the bus.

[0087] The processor 601 can be a Central Processing Unit (CPU), or can also be other general-purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field-Programmable Gate Arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. The processor is the control center of the processing device and connects various parts of the entire device using various interfaces and circuits.

[0088] The memory 602 can be used to store computer programs and / or modules. The processor 601 realizes various functions of the computer device by running or executing the computer programs and / or modules stored in the memory 602, and by invoking the data stored in the memory 602. The memory 602 can mainly include a program storage area and a data storage area. Among them, the program storage area can store an operating system, application programs required for at least one function, etc.; the data storage area can store data created according to the use of the processing device, etc. In addition, the memory can include high-speed random access memory, and can also include non-volatile memory, such as a hard disk, memory, plug-in hard disk, Smart Media Card (SMC), Secure Digital (SD) card, Flash Card, at least one magnetic disk storage device, flash device, or other volatile solid-state storage devices.

[0089] When the processor 601 is used to execute the computer program stored in the memory 602, the following functions can be specifically realized: Obtain an initial magnetic resonance sodium imaging image whose resolution is to be improved; Input the initial magnetic resonance sodium imaging image into the magnetic resonance sodium imaging image super-resolution model. The magnetic resonance sodium imaging image super-resolution model is used to increase the number of layers and resolution of the input magnetic resonance sodium imaging image. The magnetic resonance sodium imaging image super-resolution model is pre-trained by sample magnetic resonance sodium imaging images. During the operation of the magnetic resonance sodium imaging image super-resolution model, convert the input magnetic resonance sodium imaging image into a feature matrix F1 with specifications of w1, h1, n×64, where w represents width, h represents height, and n represents the number of layers. Sample the feature matrix F1 into a feature matrix F2 with the same specifications. Add the feature matrix obtained after weighting the feature matrix F2 to the feature matrix F1 to obtain a feature matrix F3. Enlarge the size of the feature matrix F3 to obtain a feature matrix F4 with specifications of w1×8, h1×8, n×8. Downsample the feature matrix F4 to obtain a feature matrix F5 with specifications of w2, h2, n×8×r 2 where r represents the scaling factor, and w1×8 = w2 / r 2 and h1×8 = h2 / r 2 Convolve the feature matrix F5 to obtain a feature matrix F6 with specifications of w2, h2, n. Convert the feature matrix F6 into an output image; Extract the target magnetic resonance sodium imaging image output by the magnetic resonance sodium imaging image super-resolution model.

[0090] Those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working processes of the above-described magnetic resonance sodium imaging image super-resolution device, processing equipment, and their corresponding units can refer to the description of the magnetic resonance sodium imaging image super-resolution method in the Figure 1 corresponding embodiments, and will not be elaborated herein specifically.

[0091] Those of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructions or by controlling relevant hardware through instructions. The instructions can be stored in a computer-readable storage medium and loaded and executed by a processor.

[0092] Therefore, the present application provides a computer-readable storage medium storing multiple instructions that can be loaded by a processor to execute the steps of the magnetic resonance sodium imaging image super-resolution method in the present application as Figure 1 described in the corresponding embodiments. The specific operations can refer to the description of the magnetic resonance sodium imaging image super-resolution method in the Figure 1 corresponding embodiments and will not be elaborated herein.

[0093] Among them, the computer-readable storage medium may include: Read Only Memory (ROM), Random Access Memory (RAM), magnetic disk, optical disc, etc.

[0094] Due to the instructions stored in the computer-readable storage medium, the steps of the magnetic resonance sodium imaging image super-resolution method in the corresponding embodiments of the present application can be executed. Therefore, Figure 1 the beneficial effects achievable by the magnetic resonance sodium imaging image super-resolution method in the corresponding embodiments of the present application can be realized. For details, refer to the previous description and will not be repeated here. Figure 1

[0095] The magnetic resonance sodium imaging image super-resolution method, device, processing equipment, and computer-readable storage medium provided in the present application have been introduced in detail above. Specific examples are used in this article to elaborate on the principle and implementation manner of the present application. The description of the above embodiments is only used to help understand the core idea of the present application; at the same time, for those skilled in the art, according to the idea of the present application, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to the present application.​

Claims

1. A method for super-resolution of magnetic resonance sodium imaging images, characterized in that, The method includes: Obtaining an initial magnetic resonance sodium imaging image to be enhanced in resolution; Input the initial magnetic resonance sodium imaging image into a magnetic resonance sodium imaging image super-resolution model, where the magnetic resonance sodium imaging image super-resolution model is used to increase the number of layers and resolution of the input magnetic resonance sodium imaging image. The magnetic resonance sodium imaging image super-resolution model is pre-trained by sample magnetic resonance sodium imaging images. During the operation of the magnetic resonance sodium imaging image super-resolution model, convert the input magnetic resonance sodium imaging image into a feature matrix F1 with specifications of w1, h1, n×64, where w represents width, h represents height, and n represents the number of layers. Sample the feature matrix F1 into a feature matrix F2 with the same specifications. Add the feature matrix after weighting the feature matrix F2 to the feature matrix F1 to obtain a feature matrix F3. Enlarge the feature matrix size of the feature matrix F3 to obtain a feature matrix F4 with specifications of w1×8, h1×8, n×8. Downsample the feature matrix F4 to obtain a feature matrix F5 with specifications of w2, h2, n×8×r 2 , where r represents the scaling factor, and w1×8 = w2 / r 2 , h1×8 = h2 / r 2 . Convolve the feature matrix F5 to obtain a feature matrix F6 with specifications of w2, h2, n. Convert the feature matrix F6 into an output image; Extracting a target magnetic resonance sodium imaging image output by the super-resolution model of the magnetic resonance sodium imaging image.

2. The method according to claim 1, characterized in that, Sampling the feature matrix F1 into the feature matrix F2 of the same specification, specifically including: The feature matrix F1 with the specification of w1, h1, n×64 is sampled into a feature matrix F11 with the specification of w×r, h×r, n×64 / r through the following formula: 2 : ， Among them, represents the output, represents the input, and mod represents the modulo operator; Processing the feature matrix F11 through the activation function ReLU to obtain the feature matrix F12; Downsampling the feature matrix F12 into the feature matrix F2 with the specification of w, h, n×64.

3. The method according to claim 1, wherein Enlarging the feature matrix size of the feature matrix F3 to obtain the feature matrix F4 with the specification of w1×8, h1×8, n×8, including: After performing deep feature extraction processing on the feature matrix F3 with the specification of w1, h1, n×64 through 3 deep feature extraction units in sequence, performing 3 times of upsampling to obtain the feature matrix F4 with the specification of w1×8, h1×8, n×8, where the deep feature extraction unit includes 1 dense feature extraction block and 1 convolutional block in sequence; The specification of the feature matrix before and after being processed by the dense feature extraction block remains unchanged. The specification of the feature matrix after the first deep feature extraction processing is w1, h1, n×128, the specification of the feature matrix after the second deep feature extraction processing is w1, h1, n×256, and the specification of the feature matrix after the third deep feature extraction processing is w1, h1, n×512; The specification of the feature matrix after the first upsampling is w1×2, h1×2, n×128, the specification of the feature matrix after the second upsampling is w1×4, h1×4, n×32, and the specification of the feature matrix after the third upsampling is w1×8, h1×8, n×8.

4. The method according to claim 3, wherein In the dense feature extraction block, the input of the dense feature extraction block is respectively connected to the inputs of the first convolutional block, the second convolutional block, and the third convolutional block. The output of the first convolutional block is respectively connected to the inputs of the second convolutional block and the third convolutional block. The output of the second convolutional block is connected to the input of the third convolutional block. The input of the feature sampling block is respectively connected to the input of the dense feature extraction block, the output of the first convolutional block, the output of the second convolutional block, and the output of the third convolutional block. The input of the self-attention gating is connected to the feature sampling block. The input of the dense feature extraction block, the output of the first convolutional block, the output of the second convolutional block, the output of the third convolutional block, and the output of the self-attention gating are added together as the output.

5. The method according to claim 1, characterized in that, Before inputting the initial magnetic resonance sodium imaging image into the super-resolution model of the magnetic resonance sodium imaging image, the method further includes: Obtaining the sample magnetic resonance sodium imaging image and making annotations; Training the initial model based on the sample magnetic resonance sodium imaging image to obtain the super-resolution model of the magnetic resonance sodium imaging image.

6. The method according to claim 5, characterized in that, Obtaining the sample magnetic resonance sodium imaging image includes: Obtaining an initial sample magnetic resonance sodium imaging image, where the initial sample magnetic resonance sodium imaging image is an image sequence; According to the preset number of high-resolution image layers corresponding to the model output, interpolation is performed on each adjacent layer of the low-resolution images in the initial sample magnetic resonance sodium imaging image. If k-1 layers of images are inserted between adjacent layers, after performing matrix addition on the original adjacent layer image matrices, they are respectively multiplied by 1 / k, 2 / k,..., (k-1) / k to obtain k-1 matrices, and the k-1 matrices are converted into images to obtain a target sample magnetic resonance sodium imaging image with the number of layers being the preset number of high-resolution image layers, which is used as the sample magnetic resonance sodium imaging image, where k-1 is the number of interpolation layers.

7. The method according to claim 1, wherein The method further includes: Displaying the target magnetic resonance sodium imaging image.

8. A magnetic resonance sodium imaging image super-resolution device, characterized in that, The device includes: An acquisition unit, configured to acquire an initial magnetic resonance sodium imaging image whose resolution needs to be improved; An input unit for inputting the initial magnetic resonance sodium imaging image into a magnetic resonance sodium imaging image super-resolution model, wherein the magnetic resonance sodium imaging image super-resolution model is used to increase the number of layers and resolution of the input magnetic resonance sodium imaging image, and the magnetic resonance sodium imaging image super-resolution model is pre-trained by sample magnetic resonance sodium imaging images. During the operation of the magnetic resonance sodium imaging image super-resolution model, the input magnetic resonance sodium imaging image is converted into a feature matrix F1 with a specification of w1, h1, n×64, where w represents the width, h represents the height, and n represents the number of layers. The feature matrix F1 is sampled into a feature matrix F2 with the same specification. The feature matrix after weighting the feature matrix F2 is added to the feature matrix F1 to obtain a feature matrix F3. The size of the feature matrix F3 is enlarged to obtain a feature matrix F4 with a specification of w1×8, h1×8, n×8. The feature matrix F4 is downsampled to obtain a feature matrix F5 with a specification of w2, h2, n×8×r, where r represents the scaling factor and w1×8 = w2 / r 2 , h1×8 = h2 / r 2 , h1×8 = h2 / r 2 . The feature matrix F5 is convolved to obtain a feature matrix F6 with a specification of w2, h2, n, and the feature matrix F6 is converted into an output image; An extraction unit, configured to extract a target magnetic resonance sodium imaging image output by the super-resolution model of the magnetic resonance sodium imaging image.

9. A processing device, characterized in that, It includes a processor and a memory, and a computer program is stored in the memory. When the processor calls the computer program in the memory, it executes the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores multiple instructions, and the instructions are suitable for being loaded by a processor to execute the method according to any one of claims 1 to 7.

Images