Magnetic resonance imaging method, apparatus, computer device, medium and program product

Abstract

This application relates to a magnetic resonance imaging method, apparatus, computer device, medium, and program product. The method obtains an initial K-space dataset of a subject; inputs the initial K-space dataset into a machine learning model, which supplements missing data from undersampled radiological data lines and outputs a target K-space dataset; and reconstructs the target K-space dataset to generate a magnetic resonance image of the subject. The magnetic resonance imaging method provided by this application can reduce sampling time while improving the signal-to-noise ratio and contrast of the magnetic resonance image.

Description

Technical Field

[0001] The present application relates to the field of magnetic resonance imaging technology, and in particular to a magnetic resonance imaging method, apparatus, computer equipment, medium and program product. Background Art

[0002] Magnetic resonance imaging (MRI) uses signals generated by the resonance of atomic nuclei within a strong magnetic field to reconstruct images. MRI is widely used because it does not require the injection of contrast agents and is ionizing radiation-free. During MRI, a silent sequence can be used to collect signals to reduce noise during data acquisition. However, the signals collected using this silent sequence are padded into K-space data, resulting in a low signal-to-noise ratio for MRI images generated using this K-space data.

[0003] Existing techniques can improve the signal-to-noise ratio (SNR) of MRI images by increasing the acquisition time of the silent sequence. However, this increase in acquisition time introduces motion artifacts and increases discomfort for the subject. Reducing the number of sampling points during signal acquisition can also reduce motion artifacts, but this results in a lower SNR in the MRI image. Existing techniques cannot guarantee that reducing the number of sampling points will simultaneously improve the SNR of MRI images. Summary of the Invention

[0004] Based on this, it is necessary to provide a magnetic resonance imaging method, apparatus, computer equipment, medium and program product to address the above technical problems.

[0005] In a first aspect, an embodiment of the present application provides a magnetic resonance imaging method, the method comprising:

[0006] Acquiring an initial K-space data set of the detection object, the initial K-space data set including undersampled radiation data lines, and missing data on the undersampled radiation data lines;

[0007] The initial K-space dataset is input into the machine learning model, which supplements the missing data on the undersampled radiological data line and outputs the target K-space dataset;

[0008] The target K-space data set is reconstructed to generate a magnetic resonance image of the detection object.

[0009] In one embodiment, the undersampled radiation data line includes a first data segment and a second data segment, the first data segment is located in the central area of ​​the initial K-space data set, the second data segment is located in other areas outside the central area of ​​the initial K-space data set, and there are unfilled data points in the second data segment; wherein the central area refers to the area covering the center of the initial K-space data set with the radiation data line corresponding to the first data segment as the radius.

[0010] In one embodiment, the density of data points filled in the first data segment is greater than the density of data points filled in the second data segment.

[0011] In one embodiment, the undersampled radiation data line includes a first undersampled radiation data line and a second undersampled radiation data line; the radius corresponding to the first undersampled radiation data line is different from the radius corresponding to the second undersampled radiation data line.

[0012] In a second aspect, an embodiment of the present application provides a magnetic resonance imaging method, the method comprising:

[0013] Acquiring magnetic resonance signals of the subject and filling the magnetic resonance signals in K-space to obtain an initial K-space dataset; the radius of the K-space is a first radius, and a filling trajectory corresponding to at least a portion of the data of the initial K-space dataset has a second radius, which is smaller than the first radius;

[0014] Inputting the initial K-space dataset into the machine learning model, the machine learning model extending the filling trajectory corresponding to at least a portion of the data of the initial K-space dataset to a first radius, and outputting a target K-space dataset;

[0015] The target K-space data set is reconstructed to generate a magnetic resonance image of the detection object.

[0016] In one embodiment, the K space includes a central area and other areas outside the central area, the central area is an area with a second radius as a radius, covering the center of the K space, and the sampling rate corresponding to the central area of ​​the K space satisfies the Nyquist sampling theorem.

[0017] In a third aspect, an embodiment of the present application provides a magnetic resonance imaging apparatus, the apparatus comprising:

[0018] A first acquisition module is used to acquire an initial K-space data set of the detection object, where the initial K-space data set includes undersampled radiation data lines, where there is missing data on the undersampled radiation data lines;

[0019] A first determination module is used to input the initial K-space dataset into the machine learning model, and the machine learning model supplements the missing data on the undersampled radiological data line and outputs the target K-space dataset;

[0020] The first reconstruction module is used to reconstruct the target K-space data set to generate a magnetic resonance image of the detection object.

[0021] In a fourth aspect, an embodiment of the present application provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the method provided in the above embodiment are implemented.

[0022] In a fifth aspect, an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon, which implements the steps of the method provided in the above embodiment when the computer program is executed by a processor.

[0023] In a sixth aspect, an embodiment of the present application further provides a computer program product, comprising a computer program, which implements the steps of the method provided in the above embodiment when executed by a processor.

[0024] The present invention provides a magnetic resonance imaging method, apparatus, computer device, medium, and program product. The method obtains an initial K-space dataset of a subject; inputs the initial K-space dataset into a machine learning model, which supplements the missing data on the undersampled radiation data lines in the initial K-space dataset and outputs a target K-space dataset; and reconstructs the target K-space dataset to generate a magnetic resonance image of the subject. The initial K-space dataset obtained by the magnetic resonance imaging method provided by the present invention includes undersampled radiation data lines, which can reduce the number of sampling points and thus the sampling time. At the same time, the target K-space dataset can be obtained by supplementing the missing data on the undersampled radiation data lines using the machine learning model. The target K-space dataset contains recovered K-space high-frequency data, and the magnetic resonance image of the subject generated by reconstructing the target K-space dataset has a higher contrast, that is, the quality of the magnetic resonance image is higher. In other words, the magnetic resonance imaging method provided by the present invention can ensure that the signal-to-noise ratio (quality) and contrast of the magnetic resonance image are improved while reducing the sampling time. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the conventional technology, the following briefly introduces the drawings required for use in the embodiments or the conventional technology descriptions. Obviously, the drawings described below are only some embodiments of the present application. For different technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0026] Figure 1 A schematic diagram of an application scenario of a magnetic resonance imaging method provided by an embodiment;

[0027] Figure 2 A schematic flow chart of the steps of a magnetic resonance imaging method provided by one embodiment;

[0028] Figure 3 A flowchart illustrating the steps of a machine learning model training process provided in one embodiment;

[0029] Figure 4 A schematic flow chart of the steps for determining a training sample according to an embodiment;

[0030] Figure 5 A schematic diagram of a gold standard K-space dataset provided in one embodiment;

[0031] Figure 6 A schematic diagram of a training sample provided for one embodiment;

[0032] Figure 7 A schematic diagram of a partial area of ​​a training sample provided by an embodiment;

[0033] Figure 8 A schematic flow chart of the steps of a magnetic resonance imaging method provided in another embodiment;

[0034] Figure 9 A schematic flow chart of the steps of a magnetic resonance imaging method provided in another embodiment;

[0035] Figure 10 A schematic structural diagram of a magnetic resonance imaging apparatus provided by one embodiment;

[0036] Figure 11 A schematic structural diagram of a magnetic resonance imaging apparatus provided in another embodiment;

[0037] Figure 12 A schematic diagram of the structure of a computer device provided in one embodiment. DETAILED DESCRIPTION

[0038] To make the above-mentioned objects, features, and advantages of the present application more clearly understood, the specific embodiments of the present application are described in detail below with reference to the accompanying drawings. The following description sets forth many specific details to facilitate a full understanding of the present application. However, the present application can be implemented in many other ways than those described herein, and those skilled in the art can make similar improvements without violating the scope of the present application. Therefore, the present application is not limited to the specific embodiments disclosed below.

[0039] The serial numbers assigned to the components in this document, such as "first", "second", etc., are only used to distinguish the objects described and do not have any order or technical meaning.

[0040] The magnetic resonance imaging method provided in the embodiments of the present application can be applied to the application scenario shown in Figure 1. This application environment includes a terminal 102 and a magnetic resonance imaging device 104. Terminal 102 can communicate with magnetic resonance imaging device 104 via a network. Terminal 102 can be, but is not limited to, various personal computers, laptop computers, and tablet computers. This embodiment does not limit the specific structure of magnetic resonance imaging device 104.

[0041] The following specific embodiments describe in detail the technical solution of this application and how the technical solution of this application solves the technical problem. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of this application will be described below in conjunction with the accompanying drawings.

[0042] See Figure 2 One embodiment of the present application provides a magnetic resonance imaging method, which is applied to Figure 1 The terminal in FIG. 1 is used as an example to illustrate the method. The magnetic resonance imaging method includes the following steps:

[0043] Step 200: Acquire an initial K-space data set of the detection object, where the initial K-space data set includes undersampled radiation data lines, and there is missing data on the undersampled radiation data lines.

[0044] The initial K-space data set refers to the K-space data set obtained by the terminal by filling the K-space according to the magnetic resonance signal corresponding to the detected object. Figure 1 The magnetic resonance device in the K space obtains and stores it in the memory of the terminal. When the terminal fills the magnetic resonance signal into the K space, the filling trajectory adopted by the terminal is a radial trajectory. That is to say, the K space includes many radial data lines. For example, the initial K space data set can be obtained by radial golden angle sparse sampling. Each radial data line extends radially from the center of the K space, and each radial data line includes multiple data points (encoding position points of the echo signal). The terminal fills the magnetic resonance signal into the data points on each radial data line. The data points on each radial data line correspond to the data points at the time of full sampling, and the sampling rate of full sampling satisfies the Nyquist sampling theorem.

[0045] The initial K-space dataset includes fully sampled and undersampled radiographic data lines. A fully sampled radiographic data line is one in which all data points on the radiographic line are filled with the magnetic resonance signal; an undersampled radiographic data line is one in which only a portion of the data points on the radiographic line are filled with the magnetic resonance signal. This means that some data points on the undersampled radiographic data line are missing (unfilled) compared to the fully sampled radiographic data line.

[0046] In an optional embodiment, the under-sampled radiation data line may be obtained by filling the data points on the radiation line with the magnetic resonance signal at fixed intervals, or may be obtained by randomly filling the data points on the radiation data line with the magnetic resonance signal.

[0047] Step 210: Input the initial K-space dataset into the machine learning model, and the machine learning model supplements the missing data on the undersampled radiological data line and outputs the target K-space dataset.

[0048] The initial k-space dataset contains missing data on the undersampled radiographic lines. After obtaining the initial k-space dataset, the terminal inputs it into a pre-trained machine learning model. The machine learning model then supplements the missing data on the undersampled radiographic lines in the initial k-space dataset to output a target k-space dataset. The radiographic lines in the target k-space dataset are all filled with data.

[0049] Step 220: Reconstruct the target K-space data set to generate a magnetic resonance image of the detection object.

[0050] After obtaining the target K-space dataset, the terminal performs 3D reconstruction on it to obtain a magnetic resonance image corresponding to the detected object. This embodiment does not limit the specific method of reconstructing the target K-space dataset, as long as its function can be achieved.

[0051] The magnetic resonance imaging method provided in the embodiments of the present application comprises obtaining an initial k-space dataset of a subject; inputting the initial k-space dataset into a machine learning model, which supplements the missing data on undersampled radioactive data lines in the initial k-space dataset to output a target k-space dataset; and reconstructing the target k-space dataset to generate a magnetic resonance image of the subject. The initial k-space dataset obtained by the magnetic resonance imaging method provided in the embodiments of the present application includes undersampled radioactive data lines, thereby reducing the number of sampling points and, therefore, the sampling time. This in turn reduces the impact of physiological and non-physiological motion on the sampled data, thereby improving the signal-to-noise ratio (SNR) of the reconstructed magnetic resonance image. Furthermore, by supplementing the missing data on the undersampled radioactive data lines using the machine learning model, a target k-space dataset is obtained. The target k-space dataset includes recovered high-frequency K-space data, and the magnetic resonance image of the subject generated by reconstructing the target k-space dataset has a higher contrast, i.e., a higher-quality magnetic resonance image. In other words, the magnetic resonance imaging method provided in the embodiments of the present application can ensure that the signal-to-noise ratio and contrast of the magnetic resonance image are improved while reducing the sampling time.

[0052] In one embodiment, the undersampled radiation data line includes a first data segment and a second data segment, the first data segment is located in the central area of ​​the initial K-space dataset, the second data segment is located in other areas outside the central area of ​​the initial K-space dataset, and there are unfilled data points in the second data segment; wherein the central area refers to the area covering the center of the initial K-space dataset with the radiation data line corresponding to the first data segment as the radius.

[0053] The initial K-space dataset may include a central area and an area (peripheral area) other than the central area, and the central area refers to an area with a preset length as a radius that covers the center of the initial K-space dataset. The undersampled radiation data line in the initial K-space dataset includes a first data segment and a second data segment. The first data segment refers to a line segment on the radiation data line that is filled with the first data, and the second data segment refers to a line segment on the radiation data line that is filled with the second data. There are unfilled data points in the second data segment, that is, the second data does not completely fill the data points on the radiation data line corresponding to the second data segment. This embodiment does not limit the filling amount of the first data on the radiation data line corresponding to the first data segment. The first data segment is located in the central area of ​​the initial K-space dataset, that is, the radius of the radiation data line corresponding to the first data segment is the same as the preset length; the second data segment is located in the peripheral area of ​​the initial K-space dataset. In other words, there are unfilled data points in the peripheral area of ​​the initial K-space dataset.

[0054] In one embodiment, the density of data points populated in the first data segment is greater than the density of data points populated in the second data segment. That is, the first amount of data points populated on the radiation data line corresponding to the first data segment is greater than the second amount of data points populated on the radiation data line corresponding to the second data segment. The density of data points populated in the central region of the undersampled radiation data line in the initial k-space dataset is greater than the density of data points populated in the peripheral region. This embodiment does not impose any restrictions on the specific density of data points populated in the first data segment or the specific density of data points populated in the second data segment, as long as the functionality is achieved.

[0055] In a specific embodiment, all data points on the radiation data line corresponding to the first data segment are filled, and data points on the radiation data line corresponding to the second data point are partially filled according to a preset interval (one is filled for every other encoding position point of the echo signal).

[0056] In another specific embodiment, every other data point on the radiation data line corresponding to the second data segment is filled, and every other data point on the radiation data line corresponding to the second data segment is filled.

[0057] In one embodiment, the undersampled radiation data line includes a first undersampled radiation data line and a second undersampled radiation data line; the radius corresponding to the first undersampled radiation data line is different from the radius corresponding to the second undersampled radiation data line. The first undersampled radiation data line refers to the radiation data line corresponding to the first data segment, and the second undersampled radiation data line refers to the radiation data line corresponding to the second data segment. The radius corresponding to the first undersampled radiation data line is the length of the first undersampled radiation data line, and the radius of the second undersampled radiation data line is the length of the second undersampled radiation data line. The length of the first undersampled radiation data line may be smaller than the length of the second undersampled radiation data line, or the length of the first undersampled radiation data line may be greater than the length of the second undersampled radiation data line. The radius corresponding to the first undersampled radiation data line and the radius corresponding to the second undersampled radiation data line can be determined based on the radius of the initial K-space data set, and this embodiment does not impose any restrictions on this.

[0058] In this embodiment, various situations of under-sampled radiation data lines in the initial K-space data set are proposed. The missing data on the under-sampled radiation data lines in various situations can be supplemented by a machine learning model to obtain a target K-space data set, which can improve the applicability and reliability of the magnetic resonance imaging method provided in this application.

[0059] In one embodiment, Figure 3 As shown, the training process of a machine learning model can include:

[0060] Step 300: Acquire a gold standard K-space dataset, and determine training samples based on the gold standard K-space dataset.

[0061] The gold standard K-space dataset refers to a K-space dataset obtained by fully sampling and filling the data points on all the radiation data lines in the K-space dataset according to the magnetic resonance signal, and the sampling process corresponding to the gold standard K-space dataset is not affected by motion or the influence of motion is suppressed. For example, a 3D golden angle radial stack-of-stars (SOS) scanning sequence can be used to obtain the gold standard K-space dataset, which has a good suppression effect on motion artifacts. After obtaining the gold standard K-space dataset, the terminal can obtain training samples by extracting or assigning data in the gold standard K-space dataset.

[0062] In an optional embodiment, if Figure 4 As shown, methods for determining training samples based on the gold standard K-space dataset may include the following:

[0063] Step 400: retain the data of the central region of the gold standard K-space dataset and set the data of other regions of the gold standard K-space dataset except the central region to zero to obtain a training sample. The central region of the gold standard K-space dataset refers to an area covering the center of the gold standard K-space dataset within a preset range.

[0064] That is to say, the K-space data sets corresponding to the training samples are all undersampled radiation data lines, and the first data segment of the undersampled radiation data line is a data segment obtained by full sampling, that is, the data points on the radiation data line corresponding to the first data segment are all filled; the data points on the radiation data line corresponding to the second data segment are not filled.

[0065] Step 410: retain the data of the central region of the gold standard K-space dataset, extract the data of other regions of the gold standard K-space dataset except the central region, and set the remaining data of other regions to zero to obtain training samples.

[0066] That is, the K-space dataset corresponding to the training sample may all be undersampled radiation data lines, the first data segment of the undersampled radiation data line is a data segment obtained through full sampling, and the radiation data line corresponding to the second data segment of the undersampled radiation data line has unfilled data points. The K-space dataset corresponding to the training sample may include undersampled radiation data lines and fully sampled radiation data lines, and the fully sampled radiation data line may also include two data segments, one of which is located in the center area of ​​the K-space dataset, and the data points on the radiation data line corresponding to these two data segments are all filled; the first data segment of the undersampled radiation data line is a data point obtained through full sampling, and the radiation data line corresponding to the second data segment of the undersampled radiation data line has unfilled data points.

[0067] Step 420: retain the data on the radiation data lines in the gold standard K-space dataset according to a preset interval, and set the data on the remaining radiation data lines in the gold standard K-space dataset to zero to obtain training samples.

[0068] That is to say, the K-space data sets corresponding to the training samples are all undersampled radiation data lines, and the data points on the radiation data lines corresponding to the first data segment and the second data segment of the undersampled radiation data lines are all filled according to the preset intervals.

[0069] The gold standard K-space dataset is Figure 5 The training samples of the machine learning model are as follows. Figure 6 As shown in FIG, the training sample is obtained by retaining the central area of ​​the gold standard K-space dataset and extracting and assigning values ​​to the data of other areas of the gold standard K-space dataset except the central area. Figure 6 The enlarged image of some areas of the training samples is shown in Figure 7 shown.

[0070] Step 310: Use the training samples and the gold standard K-space dataset to train the initial machine learning model to obtain a machine learning model.

[0071] The training samples used by the terminal can be samples obtained by at least one of the methods for determining training samples based on the gold standard K-space dataset provided in the above embodiments. The terminal can train the initial machine learning model using the obtained training samples and the gold standard K-space dataset to obtain a machine learning model.

[0072] In an optional embodiment, a Gaussian noise signal can be added to the obtained training samples, so that the obtained training samples are more in line with practical applications, and the machine learning model obtained through training is more accurate. Using the machine learning model, a more accurate target K-space data set can be obtained.

[0073] In an optional embodiment, during the training of the initial machine learning model, the idea of ​​perceptual loss can be used to use the initial machine learning model to obtain information about the gold standard K-space data set in advance and construct a loss function, which can improve training efficiency and accuracy.

[0074] In one embodiment, the machine learning model may be a model set consisting of multiple sub-models, each sub-model corresponding to a different scan time. Training samples for the multiple sub-models corresponding to different scan times may be obtained by cropping or extracting the gold standard K-space dataset to varying degrees, or by acquiring gold standard K-space datasets corresponding to different time lengths and cropping or extracting each gold standard K-space dataset separately.

[0075] In an alternative embodiment, see Figure 8 The steps of inputting the initial K-space dataset into the machine learning model, supplementing the missing data of the undersampled radiographic data lines by the machine learning model, and outputting the target K-space dataset may include:

[0076] Step 800: Obtain the scan time corresponding to the initial K-space data set to obtain the target scan time.

[0077] The scan time corresponding to the initial K-space dataset, i.e., the scan time required to fill the magnetic resonance signals of the initial K-space dataset, may be sent by the magnetic resonance device to the terminal, or may be obtained by the terminal from the magnetic resonance device.

[0078] Step 810: Determine a sub-model corresponding to the target scan time from the machine learning model according to the target scan time.

[0079] Each sub-model in the machine learning model has a corresponding scan time. The terminal can search the machine learning module for a sub-model with the same scan time as the target scan time.

[0080] Step 820: Input the initial K-space dataset into the sub-model corresponding to the target scanning time to obtain the target K-space dataset.

[0081] The terminal inputs the initial K-space data set into a sub-model corresponding to the target scan time. The sub-model can obtain the target K-space data set by supplementing the missing data on the undersampled data radiation data line in the initial K-space data set.

[0082] In another optional embodiment, an evaluation system can be constructed within the machine learning model, which can consider aspects such as image fidelity and image contrast. Thus, after the terminal inputs the initial K-space dataset into the machine learning model, each sub-model within the machine learning model supplements the missing data of the undersampled radiological data lines in the initial K-space dataset, thereby obtaining multiple supplemented K-space datasets. These multiple supplemented K-space datasets are then evaluated by the evaluation system to obtain a target K-space dataset.

[0083] In this embodiment, the machine learning model can adapt to different scan times. That is, medical personnel can flexibly select different scan times based on actual conditions, thereby improving the applicability of the magnetic resonance imaging method provided by this application. Furthermore, the machine learning model has a certain degree of fault tolerance, meaning that changes in scan time have little impact on the quality of the magnetic resonance images. This can improve the quality of the magnetic resonance images obtained using the magnetic resonance imaging method provided by this application.

[0084] See Figure 9 In one embodiment, the present application provides a magnetic resonance imaging method, which is applied to Figure 1 The terminal in FIG. 1 is used as an example to illustrate the method. The magnetic resonance imaging method includes the following steps:

[0085] Step 900: Acquire magnetic resonance signals of the subject and fill the magnetic resonance signals in K-space to obtain an initial K-space dataset; the radius of the K-space is a first radius, and the filling trajectory corresponding to at least part of the data of the initial K-space dataset has a second radius, which is smaller than the first radius.

[0086] application Figure 1The MRI device in the apparatus obtains MRI signals from the subject and transmits them to the terminal. The MRI signals can be directly transmitted from the MRI device to the terminal, or the terminal can obtain them from the MRI device when needed. After obtaining the MRI signals from the subject, the terminal fills the K-space with them to obtain an initial K-space dataset.

[0087] When the terminal fills the magnetic resonance signal into K-space, the filling trajectory used is a radial trajectory. That is, the K-space includes many radial data lines, each of which extends radially from the center of the K-space. Each radial data line includes multiple data points, and the terminal fills the magnetic resonance signal into the data points on each radial data line. The radius of the K-space is a first radius, and the filling trajectory corresponding to at least part of the data in the initial K-space dataset has a second radius. In other words, there are some radial data lines in the initial K-space dataset whose radius is smaller than the first radius, or all radial data lines whose radius is smaller than the first radius. When the radius of the radial data line is smaller than the first radius, the data points filled on the radial data line are smaller than the data points filled during full sampling, and there will be missing data in the initial K-space dataset.

[0088] Step 910: Input the initial K-space dataset into the machine learning model. The machine learning model extends the filling trajectory corresponding to at least part of the data of the initial K-space dataset to a first radius and outputs the target K-space dataset.

[0089] After obtaining the initial K-space dataset, the terminal inputs it into the machine learning model. The machine learning model then extends the filling trajectory corresponding to at least a portion of the data in the initial K-space dataset to the first radius. Specifically, the machine learning model supplements the missing data from the second radius to the first radius in the filling trajectory corresponding to at least a portion of the data in the initial K-space dataset, thereby obtaining the target K-space dataset. A description of the machine learning model can be found in the detailed description of the above embodiment and is not repeated here.

[0090] Step 920: Reconstruct the target K-space data set to generate a magnetic resonance image of the detection object.

[0091] After obtaining the target K-space dataset, the terminal performs 3D reconstruction on it to obtain a magnetic resonance image corresponding to the detected object. This embodiment does not limit the specific method of reconstructing the target K-space dataset, as long as its function can be achieved.

[0092] The magnetic resonance imaging method provided in an embodiment of the present application obtains magnetic resonance signals from a subject and fills the magnetic resonance signals into K-space to obtain an initial K-space dataset; inputs the initial K-space dataset into a machine learning model, and the machine learning model extends the filling trajectory corresponding to at least a portion of the data in the initial K-space dataset to a first radius, outputting a target K-space dataset; and reconstructs the target K-space dataset to generate a magnetic resonance image of the subject. The filling trajectory corresponding to at least a portion of the data in the initial K-space dataset obtained by the magnetic resonance imaging method provided in an embodiment of the present application has a second radius, i.e., there is missing data in the initial K-space dataset. This reduces the number of sampling points and thus the sampling time. Simultaneously, the filling trajectory corresponding to at least a portion of the data in the initial K-space dataset is extended to the first radius by the machine learning model to obtain a target K-space dataset. The target K-space dataset includes recovered high-frequency K-space data. Thus, the magnetic resonance image of the subject generated by reconstructing the target K-space dataset has a higher contrast, i.e., a higher quality magnetic resonance image. In other words, the magnetic resonance imaging method provided in an embodiment of the present application can ensure that the signal-to-noise ratio (quality) and contrast of the magnetic resonance image are improved while reducing the sampling time.

[0093] In one embodiment, the K space includes a central region and other regions outside the central region. The central region is a region with a second radius covering the center of the K space, and the sampling rate corresponding to the central region of the K space complies with the Nyquist sampling law. In other words, the data points in the central region of the initial K space dataset are all filled; some data points in the peripheral region of the initial K space dataset are not filled, and the peripheral region of the initial K space dataset is not filled. That is, the initial K space dataset includes undersampled radiation data lines, and the undersampled radiation data lines include a first undersampled radiation data line and a second undersampled reflection data line. The radius of the first undersampled radiation data line is the second radius. The data points of the first undersampled radiation data line are all filled, and some data points on the second undersampled radiation data line are not filled.

[0094] It should be understood that, although the various steps in the flowcharts involved in the various embodiments described above are displayed in sequence according to the instructions of the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the flowcharts involved in the various embodiments described above can include multiple steps or multiple stages, and these steps or stages are not necessarily executed and completed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of steps or stages in other steps.

[0095] Based on the same inventive concept, embodiments of the present application further provide a magnetic resonance imaging apparatus for implementing the aforementioned magnetic resonance imaging method. The solution provided by this apparatus is similar to the solution described in the aforementioned method. Therefore, the specific limitations of one or more magnetic resonance imaging apparatus embodiments provided below can be found in the above-described limitations of the magnetic resonance imaging method and will not be further elaborated here.

[0096] In one embodiment, Figure 10 As shown, a magnetic resonance imaging device 10 is provided, which includes: a first acquisition module 11, a first determination module 12 and a first reconstruction module 13, wherein:

[0097] The first acquisition module 11 is used to acquire an initial K-space data set of the detection object, where the initial K-space data set includes undersampled radiation data lines, where there are missing data on the undersampled radiation data lines;

[0098] The first determination module 12 is used to input the initial K-space data set into the machine learning model, and the machine learning model supplements the missing data on the undersampled radiological data line and outputs the target K-space data set;

[0099] The first reconstruction module 13 is used to reconstruct the target K-space data set to generate a magnetic resonance image of the detection object.

[0100] In one embodiment, the undersampled radiation data line includes a first data segment and a second data segment, the first data segment is located in the central area of ​​the initial K-space dataset, the second data segment is located in other areas outside the central area of ​​the initial K-space dataset, and there are unfilled data points in the second data segment; wherein the central area refers to the area covering the center of the initial K-space dataset with the radiation data line corresponding to the first data segment as the radius.

[0101] In one embodiment, the density of data points filled in the first data segment is greater than the density of data points filled in the second data segment.

[0102] In one embodiment, the undersampled radiation data line includes a first undersampled radiation data line and a second undersampled radiation data line; the radius corresponding to the first undersampled radiation data line is different from the radius corresponding to the second undersampled radiation data line.

[0103] In one embodiment, Figure 11 As shown, a magnetic resonance imaging device 20 is provided, which includes: a second acquisition module 21, a second determination module 22 and a second reconstruction module 23, wherein:

[0104] The second acquisition module 21 is configured to acquire magnetic resonance signals of the subject and fill the magnetic resonance signals in a K-space to obtain an initial K-space dataset; the radius of the K-space is a first radius, and a filling trajectory corresponding to at least a portion of the data in the initial K-space dataset has a second radius, which is smaller than the first radius;

[0105] The second determination module 22 is configured to input the initial K-space dataset into the machine learning model, and the machine learning model extends the filling trajectory corresponding to at least a portion of the data in the initial K-space dataset to a first radius, and outputs a target K-space dataset;

[0106] The second reconstruction module 23 is used to reconstruct the target K-space data set to generate a magnetic resonance image of the detection object.

[0107] In one embodiment, the K space includes a central area and other areas outside the central area. The central area is an area with the second radius as the radius, covering the center of the K space, and the sampling rate corresponding to the central area of ​​the K space satisfies the Nyquist sampling theorem.

[0108] Each module in the magnetic resonance imaging apparatus 10 and the magnetic resonance imaging apparatus 20 may be implemented in whole or in part by software, hardware, or a combination thereof. Each module may be embedded in or independent of a processor in a computer device in the form of hardware, or may be stored in a memory in the form of software in the computer device so that the processor can call and execute the corresponding operations of each module.

[0109] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Figure 12 As shown. The computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a magnetic resonance imaging method is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad provided on the computer device housing, or an external keyboard, touchpad or mouse.

[0110] Those skilled in the art will understand that Figure 12The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0111] In one embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the following steps are performed:

[0112] Acquiring an initial K-space data set of the detection object, the initial K-space data set including undersampled radiation data lines, and missing data on the undersampled radiation data lines;

[0113] The initial K-space dataset is input into the machine learning model, which supplements the missing data on the undersampled radiological data line and outputs the target K-space dataset;

[0114] The target K-space data set is reconstructed to generate a magnetic resonance image of the detection object.

[0115] In one embodiment, the undersampled radiation data line includes a first data segment and a second data segment, the first data segment is located in the central area of ​​the initial K-space dataset, the second data segment is located in other areas outside the central area of ​​the initial K-space dataset, and there are unfilled data points in the second data segment; wherein the central area refers to the area covering the center of the initial K-space dataset with the radiation data line corresponding to the first data segment as the radius.

[0116] In one embodiment, the density of data points filled in the first data segment is greater than the density of data points filled in the second data segment.

[0117] In one embodiment, the undersampled radiation data line includes a first undersampled radiation data line and a second undersampled radiation data line; the radius corresponding to the first undersampled radiation data line is different from the radius corresponding to the second undersampled radiation data line.

[0118] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0119] Acquiring magnetic resonance signals of the subject and filling the magnetic resonance signals in K-space to obtain an initial K-space dataset; the radius of the K-space is a first radius, and a filling trajectory corresponding to at least a portion of the data of the initial K-space dataset has a second radius, which is smaller than the first radius;

[0120] Inputting the initial K-space dataset into the machine learning model, the machine learning model extending the filling trajectory corresponding to at least a portion of the data in the initial K-space dataset to a first radius, and outputting a target K-space dataset;

[0121] The target K-space data set is reconstructed to generate a magnetic resonance image of the detection object.

[0122] In one embodiment, the K space includes a central area and other areas outside the central area. The central area is an area with the second radius as the radius, covering the center of the K space, and the sampling rate corresponding to the central area of ​​the K space satisfies the Nyquist sampling theorem.

[0123] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0124] An initial K-space data set of the object to be examined is obtained, where the initial K-space data set includes undersampled radiation data lines, and there are missing data on the undersampled radiation data lines;

[0125] The initial K-space dataset is input into the machine learning model, which supplements the missing data on the undersampled radiological data line and outputs the target K-space dataset;

[0126] The target K-space data set is reconstructed to generate a magnetic resonance image of the detection object.

[0127] In one embodiment, the undersampled radiation data line includes a first data segment and a second data segment, the first data segment is located in the central area of ​​the initial K-space dataset, the second data segment is located in other areas outside the central area of ​​the initial K-space dataset, and there are unfilled data points in the second data segment; wherein the central area refers to the area covering the center of the initial K-space dataset with the radiation data line corresponding to the first data segment as the radius.

[0128] In one embodiment, the density of data points filled in the first data segment is greater than the density of data points filled in the second data segment.

[0129] In one embodiment, the undersampled radiation data line includes a first undersampled radiation data line and a second undersampled radiation data line; the radius corresponding to the first undersampled radiation data line is different from the radius corresponding to the second undersampled radiation data line.

[0130] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0131] Acquiring magnetic resonance signals of the subject and filling the magnetic resonance signals in K-space to obtain an initial K-space dataset; the radius of the K-space is a first radius, and a filling trajectory corresponding to at least a portion of the data of the initial K-space dataset has a second radius, which is smaller than the first radius;

[0132] Inputting the initial K-space dataset into the machine learning model, the machine learning model extending the filling trajectory corresponding to at least a portion of the data in the initial K-space dataset to a first radius, and outputting a target K-space dataset;

[0133] The target K-space data set is reconstructed to generate a magnetic resonance image of the detection object.

[0134] In one embodiment, the K space includes a central area and other areas outside the central area. The central area is an area with the second radius as the radius, covering the center of the K space, and the sampling rate corresponding to the central area of ​​the K space satisfies the Nyquist sampling theorem.

[0135] In one embodiment, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the following steps:

[0136] An initial K-space data set of the object to be examined is obtained, where the initial K-space data set includes undersampled radiation data lines, and there are missing data on the undersampled radiation data lines;

[0137] The initial K-space dataset is input into the machine learning model, which supplements the missing data on the undersampled radiological data line and outputs the target K-space dataset;

[0138] The target K-space data set is reconstructed to generate a magnetic resonance image of the detection object.

[0139] In one embodiment, the undersampled radiation data line includes a first data segment and a second data segment, the first data segment is located in the central area of ​​the initial K-space dataset, the second data segment is located in other areas outside the central area of ​​the initial K-space dataset, and there are unfilled data points in the second data segment; wherein the central area refers to the area covering the center of the initial K-space dataset with the radiation data line corresponding to the first data segment as the radius.

[0140] In one embodiment, the density of data points filled in the first data segment is greater than the density of data points filled in the second data segment.

[0141] In one embodiment, the undersampled radiation data line includes a first undersampled radiation data line and a second undersampled radiation data line; the radius corresponding to the first undersampled radiation data line is different from the radius corresponding to the second undersampled radiation data line.

[0142] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0143] Acquiring magnetic resonance signals of the subject and filling the magnetic resonance signals in K-space to obtain an initial K-space dataset; the radius of the K-space is a first radius, and a filling trajectory corresponding to at least a portion of the data of the initial K-space dataset has a second radius, which is smaller than the first radius;

[0144] Inputting the initial K-space dataset into the machine learning model, the machine learning model extending the filling trajectory corresponding to at least a portion of the data in the initial K-space dataset to a first radius, and outputting a target K-space dataset;

[0145] The target K-space data set is reconstructed to generate a magnetic resonance image of the detection object.

[0146] In one embodiment, the K space includes a central area and other areas outside the central area. The central area is an area with the second radius as the radius, covering the center of the K space, and the sampling rate corresponding to the central area of ​​the K space meets the Nyquist sampling rate.

[0147] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, database or other media used in the embodiments provided in this application may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory may include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processor involved in the various embodiments provided herein may be, but are not limited to, a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic unit, a data processing logic unit based on quantum computing, and the like.

[0148] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0149] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present application. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, and these modifications and improvements fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.

Claims

1. A magnetic resonance imaging method, characterized in that: The method comprises: Acquiring an initial K-space data set of the subject, the initial K-space data set comprising undersampled radiation data lines, wherein the undersampled radiation data lines have missing data; Inputting the initial K-space dataset into a machine learning model, wherein the machine learning model supplements the missing data on the undersampled radiation data line and outputs a target K-space dataset; the machine learning model is a model set consisting of multiple sub-models, each sub-model corresponding to a different scan time; Reconstructing the target K-space data set to generate a magnetic resonance image of the detection object; The initial K-space dataset is input into a machine learning model, and the machine learning model supplements the missing data on the undersampled radiation data line to output a target K-space dataset, including: Obtain the scanning time corresponding to the initial K-space dataset to obtain the target scanning time; determine the sub-model corresponding to the target scanning time from the machine learning model according to the target scanning time; input the initial K-space dataset into the sub-model corresponding to the target scanning time to obtain the target K-space dataset.

2. The method according to claim 1, characterized in that The undersampled radiation data line includes a first data segment and a second data segment, wherein the first data segment is located in a central area of ​​the initial K-space dataset, and the second data segment is located in other areas outside the central area of ​​the initial K-space dataset, and there are unfilled data points in the second data segment.

3. The method according to claim 2, characterized in that The density of data points filled in the first data segment is greater than the density of data points filled in the second data segment.

4. The method according to claim 1, wherein The under-sampled radiation data lines include a first under-sampled radiation data line and a second under-sampled radiation data line; a radius corresponding to the first under-sampled radiation data line is different from a radius corresponding to the second under-sampled radiation data line.

5. A magnetic resonance imaging method, characterized in that: The method comprises: Acquiring magnetic resonance signals of a detection object and filling the magnetic resonance signals in a K-space to obtain an initial K-space dataset; wherein the radius of the K-space is a first radius, and a filling trajectory corresponding to at least a portion of the data of the initial K-space dataset has a second radius, and the second radius is smaller than the first radius; Inputting the initial K-space dataset into a machine learning model, wherein the machine learning model extends a filling trajectory corresponding to at least a portion of the data in the initial K-space dataset to the first radius and outputs a target K-space dataset; the machine learning model is a model set consisting of a plurality of sub-models, each sub-model corresponding to a different scan time; Reconstructing the target K-space data set to generate a magnetic resonance image of the detection object; Inputting the initial K-space dataset into a machine learning model, wherein the machine learning model extends a filling trajectory corresponding to at least a portion of the data in the initial K-space dataset to the first radius and outputs a target K-space dataset, includes: Obtain the scanning time corresponding to the initial K-space dataset to obtain the target scanning time; determine the sub-model corresponding to the target scanning time from the machine learning model according to the target scanning time; input the initial K-space dataset into the sub-model corresponding to the target scanning time to obtain the target K-space dataset.

6. The method according to claim 5, characterized in that The K space includes a central area and other areas outside the central area, the central area is an area with the second radius as a radius, covering the center of the K space, and the sampling rate corresponding to the central area of ​​the K space satisfies the Nyquist sampling theorem.

7. A magnetic resonance imaging apparatus, characterized in that: The device comprises: A first acquisition module is configured to acquire an initial K-space dataset of the detection object, wherein the initial K-space dataset includes undersampled radiation data lines, and the undersampled radiation data lines have missing data; a first determination module, configured to input the initial K-space dataset into a machine learning model, wherein the machine learning model supplements the missing data on the undersampled radiation data line and outputs a target K-space dataset; the machine learning model is a model set consisting of a plurality of sub-models, each sub-model corresponding to a different scan time; A first reconstruction module is used to reconstruct the target K-space data set to generate a magnetic resonance image of the detection object; The first determination module is specifically used to obtain the scanning time corresponding to the initial K-space data set to obtain the target scanning time; determine the sub-model corresponding to the target scanning time from the machine learning model according to the target scanning time; input the initial K-space data set into the sub-model corresponding to the target scanning time to obtain the target K-space data set.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images