Echo planar imaging method, magnetic resonance system, computer equipment and storage medium

Abstract

The invention relates to a plane echo imaging method and system, computer equipment and a storage medium, and the method comprises the steps: obtaining a K-space data set obtained after the radio frequency excitation of a to-be-imaged region, the K-space data set comprising an echo time dimension, the K space data set is under-sampled along at least one phase coding direction and echo time dimension; recovering the K space data set along at least one phase coding direction and echo time dimension to obtain an updated K space data set; and reconstructing according to the updated K space data set to obtain a plurality of plane echo images. According to the method and the device, the problem that plane echo imaging cannot give consideration to improvement of the signal-to-noise ratio, the robustness and the calculation speed of the image in related technologies is solved, and the signal-to-noise ratio, the robustness and the calculation speed of the plane echo imaging are comprehensively improved.

Description

Technical Field

[0001] The present application relates to the field of magnetic resonance imaging, and in particular, to an echo planar imaging method, a magnetic resonance system, a computer device, and a storage medium. Background Art

[0002] Magnetic resonance imaging (MRI) is a diagnostic technique that utilizes the nuclear magnetic resonance phenomenon of a certain atomic nucleus in human tissues, processes the obtained radio frequency signals through an electronic computer, and reconstructs an image of a certain plane of the human body.

[0003] Echo Planar Imaging (EPI) belongs to a magnetic resonance imaging technique. Echo Planar Imaging can be obtained by reconstructing an echo planar sequence (abbreviated as EPI sequence). Traditional EPI sequence reconstruction algorithms: ① Linear K-space interpolation reconstruction algorithm; ② Low-rank iterative reconstruction algorithm. Among them, algorithm ① has a relatively fast calculation speed, but low signal-to-noise ratio and robustness. Algorithm ② has a relatively high signal-to-noise ratio and robustness, but a slow calculation speed.

[0004] In view of the problem in the related art that echo planar imaging cannot balance the improvement of image signal-to-noise ratio, robustness, and calculation speed, no effective solution has been proposed yet. Summary of the Invention

[0005] In the present embodiment, an echo planar imaging method, a magnetic resonance system, a computer device, and a storage medium are provided to solve the problem in the related art that echo planar imaging cannot balance the improvement of image signal-to-noise ratio, robustness, and calculation speed.

[0006] In a first aspect, in the present embodiment, an echo planar imaging method is provided, including:

[0007] Obtaining a K-space data set obtained by performing radio frequency excitation on an area to be imaged, where the K-space data set includes an echo time dimension, and the K-space data set is undersampled along at least one phase encoding direction and the echo time dimension;

[0008] Restoring the K-space data set along at least one phase encoding direction and the echo time dimension to obtain an updated K-space data set;

[0009] Reconstructing a plurality of echo planar images according to the updated K-space data set.

[0010] In one of the embodiments, the K-space data set includes a central region and an edge region, the central region is fully sampled, and the edge region is undersampled.

[0011] In one embodiment, recovering the K-space data set along at least one phase encoding direction and the echo time dimension includes:

[0012] Constructing a training data set according to the central region and training an initial model based on the training data set to obtain a trained model, where the trained model can represent the reconstruction convolution kernel corresponding to the reconstruction of the K-space data set;

[0013] Inputting the K-space data set into the trained model to obtain the updated K-space data set.

[0014] In one embodiment, constructing a training data set according to the central region and training an initial model based on the training data set to obtain a trained model includes:

[0015] Extracting the repeating units obtained by undersampling based on a preset undersampling pattern in the edge region;

[0016] Using the repeating units as reconstruction kernels, traversing the data in the central region, and taking the traversal results as the training data set.

[0017] In one embodiment, in the central region, the intervals along each phase encoding direction are the same, and the intervals along the echo time dimension are the same; and, in the edge region, the intervals along each phase encoding direction are the same, and the intervals along the echo time dimension are the same.

[0018] In one embodiment, the intervals between the central region and the edge region along the same phase encoding direction are consistent; and, the intervals between the central region and the edge region along the echo time dimension are consistent.

[0019] In one embodiment, the data in the central region is obtained by the following method:

[0020] Based on acquisitions with opposite polarities, obtaining two initial calibration data with echo intervals consistent with the target undersampled data of the multi-pulse echo;

[0021] Merging the two initial calibration data to generate the data of the central region with the eddy current effect removed.

[0022] In one embodiment, after obtaining a plurality of echo planar images, the method further includes:

[0023] Optimizing the plurality of echo planar images based on the estimated information of the signal change of the pixel points in the plurality of echo planar images with respect to the echo time to obtain corrected multi-echo images.

[0024] Second aspect, in this embodiment, a planar echo imaging method is provided, including:

[0025] Obtaining a K-space data set obtained by performing radio frequency excitation on the region to be imaged, where the K-space data set includes an echo time dimension, and the K-space data set is undersampled along at least one phase encoding direction and the echo time dimension;

[0026] Obtaining initial calibration data based on acquisitions with opposite polarities, and generating data for the central region with the influence of eddy currents removed according to the initial calibration data;

[0027] Restoring the K-space data set along at least one phase encoding direction and the echo time dimension based on the data of the central region to obtain an updated K-space data set;

[0028] Reconstructing a plurality of planar echo images according to the updated K-space data set.

[0029] In one of the embodiments, after obtaining a plurality of planar echo images, the method further includes:

[0030] Optimizing the plurality of planar echo images based on the estimated information of the signal change of the pixel points in the plurality of planar echo images with respect to the echo time to obtain corrected multi-echo images.

[0031] Third aspect, in this embodiment, a magnetic resonance system is provided, including:

[0032] A main magnet for generating a main magnetic field in the imaging space;

[0033] Gradient coils for generating gradient magnetic fields in the imaging space;

[0034] A radio frequency excitation unit for generating radio frequency excitation pulses, and performing radio frequency excitation on the region to be imaged in the imaging space according to the radio frequency excitation pulses to obtain magnetic resonance signals;

[0035] A radio frequency receiving unit for receiving the magnetic resonance signals;

[0036] An image reconstruction unit for performing the planar echo imaging method described in the first aspect or the second aspect based on the magnetic resonance signals.

[0037] Fourth aspect, in this embodiment, a computer device is provided, including a memory and a processor, where the memory stores a computer program, and when the processor executes the computer program, the steps of the planar echo imaging method described in the first aspect or the second aspect are implemented.

[0038] Fifth aspect, in this embodiment, a storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of the planar echo imaging method described in the first aspect or the second aspect above are implemented.

[0039] For the above planar echo imaging method, magnetic resonance system, computer device and storage medium, by expanding the K-space data set from K-space to K-Echo space, it can achieve the effect of obtaining multiple planar echo images with different echo times by only single reconstruction. And by acquiring the K-space data set with undersampling and introducing non-linear factors into the interpolation reconstruction algorithm in the K-Echo space based on deep learning, it can play a role in suppressing or avoiding artifacts. Finally, the reconstruction result is similar to that using the low-rank iterative reconstruction algorithm, but the calculation speed is faster. With such settings, the reconstruction quality is improved without increasing the reconstruction workload, and the image signal-to-noise ratio, robustness and calculation speed are enhanced.

[0040] Details of one or more embodiments of this application are set forth in the following drawings and description, so that other features, objects, and advantages of this application will become more concise and understandable. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The drawings described herein are used to provide a further understanding of this application, and constitute a part of this application. The illustrative embodiments and descriptions of this application are used to explain this application, and do not constitute an improper limitation of this application. In the drawings:

[0042] Figure 1 is a hardware structure block diagram of a terminal of the planar echo imaging method in an embodiment;

[0043] Figure 2 is a schematic structural diagram of a magnetic resonance system in an embodiment;

[0044] Figure 3 is a flowchart of the planar echo imaging method in an embodiment;

[0045] Figure 4 is a schematic diagram of two-dimensional K-space encoding in an embodiment;

[0046] Figure 5 is a flowchart of the model training method in an embodiment;

[0047] Figure 6 is a schematic diagram of using a reconstruction kernel to traverse fully sampled data in an embodiment;

[0048] Figure 7 is a schematic structural diagram of a convolutional neural network in an embodiment;

[0049] Figure 8 is a schematic diagram of a uniform undersampled space in an embodiment;

[0050] Figure 9 is a schematic diagram of initial calibration data acquisition in an embodiment;

[0051] Figure 10 is a schematic diagram of the generation of central region data in an embodiment;

[0052] Figure 11 is a flowchart of a planar echo imaging method in another embodiment;

[0053] Figure 12 is a flowchart of a planar echo imaging method in another embodiment;

[0054] Figure 13 is an internal structure diagram of a computer device in an embodiment. Detailed implementation manners

[0055] For a clearer understanding of the purpose, technical solution, and advantages of the present application, the present application will be described and explained below in conjunction with the accompanying drawings and embodiments.

[0056] Unless otherwise defined, the technical terms or scientific terms involved in the present application shall have the general meaning understood by those with ordinary skills in the technical field to which the present application belongs. In the present application, words such as "a", "one", "kind", "the", "these", etc. do not indicate a limitation in quantity, and they can be singular or plural. The terms "include", "comprise", "have" and any variants thereof involved in the present application are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or device including a series of steps or modules (units) is not limited to the listed steps or modules (units), but may include unlisted steps or modules (units), or may include other steps or modules (units) inherent in these processes, methods, products, or devices. The terms "connect", "be connected", "be coupled" and the like involved in the present application are not limited to physical or mechanical connections, but may include electrical connections, whether directly or indirectly. The term "plurality" involved in the present application refers to two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" may represent: A exists alone, A and B exist simultaneously, and B exists alone. Usually, the character " / " indicates that the objects associated before and after are in an "or" relationship. The terms "first", "second", "third", etc. involved in the present application are only used to distinguish similar objects and do not represent a specific order for the objects.

[0057] The method embodiments provided in this embodiment can be executed on a terminal, a computer, or a similar computing device. For example, running on a terminal Figure 1It is a hardware block diagram of a terminal for a planar echo imaging method according to an embodiment of the present application. As Figure 1 shown, the terminal may include one or more ( Figure 1 only one is shown in the figure) processors 1 and a memory 2 for storing data. Among them, the processor 1 may include, but is not limited to, a processing device such as a microprocessor MCU or a programmable logic device FPGA. The above terminal may further include a transmission device 3 for communication functions and an input / output device 4. Those of ordinary skill in the art can understand that Figure 1 the structure shown is only schematic and does not limit the structure of the above terminal. For example, the terminal may further include more or fewer components than Figure 1 shown in the figure, or have a different configuration from Figure 1 shown.

[0058] The memory 2 can be used to store computer programs. For example, software programs and modules of application software, such as the computer program corresponding to the planar echo imaging method in this embodiment. The processor 1 executes various functional applications and data processing by running the computer program stored in the memory 2, that is, implements the above method. The memory 2 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memories. In some instances, the memory 2 may further include a memory remotely set relative to the processor 1, and these remote memories can be connected to the terminal through a network. Examples of the above network include, but are not limited to, the Internet, an enterprise intranet, a local area network, a mobile communication network, and combinations thereof.

[0059] The transmission device 3 is used to receive or send data via a network. The above network includes a wireless network provided by the communication provider of the terminal. In one instance, the transmission device 3 includes a network adapter (abbreviated as NIC), which can be connected to other network devices through a base station and thus can communicate with the Internet. In one instance, the transmission device 3 may be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0060] In one embodiment, a magnetic resonance system is provided. Figure 2 It is a schematic structural diagram of the magnetic resonance system of this embodiment. As Figure 2 shown, the magnetic resonance system includes: a main magnet 101, gradient coils 102, a radio frequency excitation unit (a body volume coil 103, a local coil 104), a radio frequency receiving unit 105, and an image reconstruction unit 106.

[0061] The main magnet 101 is used to generate a main magnetic field in the imaging space 107. The main magnet 101 can be composed of superconducting coils, and in some cases, the main magnet 101 can also be composed of permanent magnets. During magnetic resonance imaging, the subject 108 lies on the bed 109, and by moving the bed 109, the subject 108 is moved into the imaging space 107.

[0062] The gradient coil 102 is used to generate a gradient magnetic field in the imaging space 107. The gradient coil 102 can be used to perform spatial encoding on magnetic resonance signals during magnetic resonance imaging.

[0063] The RF excitation unit includes a body volume coil 103 and a local coil 104, which are used to generate RF excitation pulses. According to the RF excitation pulses, the region of interest of the subject 108 in the imaging space 107 is RF-excited to obtain sampled data, which is called the K-space data set. According to the RF excitation, the subject 108 will generate corresponding magnetic resonance signals (i.e., RF signals) according to the magnetic resonance effect. During magnetic resonance imaging, one or more regions of interest of the subject 108 (such as the head region, thoracic region, arm region, shoulder region, leg region, or ankle region, etc.) can be RF-excited, and the magnetic resonance signals generated by the region of interest are received. When receiving the magnetic resonance signals generated by the subject 108 according to the excitation, it can be received by the body volume coil 103 or the local coil 104. There can be many RF receiving links. After the magnetic resonance signals are sent to the RF receiving unit 105, they are further sent to the image reconstruction unit 106 for image reconstruction to form a magnetic resonance image of the corresponding region of interest of the subject 108.

[0064] The image reconstruction unit 106 is used to perform an echo planar imaging method based on the magnetic resonance signals.

[0065] Continue to refer to Figure 2, in some embodiments, the magnetic resonance system may further include a communication bus 110, and the image reconstruction unit 106 is mounted on the communication bus 110. In addition, the communication bus 110 may also mount a processing unit 111, a display unit 112, an input / output device 113, a storage unit 114, and a communication port 115. The communication bus 110 is used to implement data transmission among these components, so as to control the magnetic resonance imaging process. Among them, the processing unit 111 may be composed of one or more processors. The display unit 112 may be a display provided for the user to display images. The input / output device 113 may be related devices such as a keyboard, a mouse, and a control box, which support input / output of corresponding data streams. The storage unit 114 may be a read-only memory (ROM), a random access memory (RAM), a hard disk, etc. The storage unit 114 may be used to store various data files required for processing and / or communication, as well as possible program instructions executed by the processing unit 111. The communication port 115 may implement data communication with external devices, image acquisition devices, databases, external storage, and image processing workstations, etc.

[0066] When the magnetic resonance system performs magnetic resonance imaging, it involves a series of radio frequency pulses and gradient magnetic fields that change over time. Among them, the radio frequency pulse refers to an electrical signal formed by controlling a radio frequency signal of a certain frequency using a switch to emit at intervals. The gradient magnetic field refers to the magnetic field intensity changing according to a certain law over time in a certain direction. The changes of the radio frequency pulse and the gradient magnetic field can be achieved through spatial encoding. Spatial encoding is also the process of writing magnetic resonance signals into an array in a specific space, such as the k-space. The k-space can be regarded as a digital array, representing the spatial frequency information in the MRI image. The k-space is also an object for storing the original magnetic resonance data. By performing a series of processes on the original magnetic resonance data in the k-space, a magnetic resonance image can be obtained.

[0067] Compared with the related art, the magnetic resonance system provided in this embodiment has improved image signal-to-noise ratio, robustness, and calculation speed in the echo planar imaging method running thereon. The echo planar imaging method of the present application will be introduced below.

[0068] In one embodiment, as Figure 3 shown, a flowchart of an echo planar imaging method is provided. Taking the method applied to the Figure 2 magnetic resonance system as an example for illustration, it includes the following steps:

[0069] Step S101, obtain a k-space data set obtained by radio frequency excitation of the to-be-imaged region. The k-space data set includes an echo time dimension, and the k-space data set is undersampled along at least one phase encoding direction and the echo time dimension.

[0070] The region to be imaged can be a region on the subject. According to radio frequency excitation, the subject will generate corresponding magnetic resonance signals according to the magnetic resonance effect. During magnetic resonance imaging, radio frequency excitation can be performed on one or more regions to be imaged of the subject, and the magnetic resonance signals generated by the region to be imaged can be received. Among them, the regions to be imaged include but are not limited to the head region, thoracic region, arm region, shoulder region, leg region, ankle region, etc. Among them, the radio frequency excitation can be multiple excitations, that is, the original data is divided into two or more acquisitions, and the data lines acquired by each excitation are interlaced with each other to obtain a K-space data set.

[0071] As introduced above, K-space is the filling space of the original data of MR signals with spatial positioning coding information. Each MR image has its corresponding K-space data. By performing Fourier transform on the data in K-space, the spatial positioning coding information in the original data can be decoded, that is, MR information with different signal intensities is assigned to corresponding spatial positions to reconstruct an MR image.

[0072] The K-space of this embodiment can be one-dimensional, two-dimensional or three-dimensional. As an example of this application, Figure 4 a two-dimensional K-space encoding schematic diagram is given. The two coordinates Kx and Ky of the two-dimensional K-space represent the frequency encoding and phase encoding directions of the MR signal respectively, and the two-dimensional MR image matrix is 256×256. As Figure 4 shown in a, during the acquisition process of the MR signal of the two-dimensional image, the phase encoding of each MR signal changes once, and the acquired MR signal fills a line in the Ky direction of the K-space. Therefore, the MR signal with spatial information is also called a phase encoding line. Figure 4 All 256 K-space lines are not drawn in a, and only a few are drawn for illustration. As Figure 4 shown in b, during the acquisition process of the MR signal of the two-dimensional image, the magnitude of the frequency encoding gradient field of each MR signal remains unchanged, and the direction will change, that is, the frequency encoding directions of two adjacent times are opposite, while the direction and field strength of the phase encoding gradient field change with a certain step. It should be noted that the lattice points in the K-space and the lattice points of the MR image are not in one-to-one correspondence, and each point in the K-space contains the full-layer information of the scanned slice. The K-space exhibits the characteristics of mirror symmetry in both the Kx and Ky directions.

[0073] The phase encoding direction, which corresponds to a specific axis in the K-space, determines the spatial resolution of the image. The echo time is the time interval from applying a radio frequency pulse excitation to detecting the first or subsequent free induction decay signal. In a multi-echo sequence, each echo corresponds to different data rows or data sets in the K-space, that is, different parts of the same K-space can be data obtained through different echo times.

[0074] In this embodiment, a data axis can be set in the acquired magnetic resonance data to represent the echo time dimension, similar to the Kx axis or Ky axis in the K-space. For convenience of description, the K-space including the echo time dimension is hereinafter defined as the K-Echo space (or K-echo space), where the one-dimensional K-space is defined as the Kx-Echo space, the two-dimensional K-space is defined as the Kx-Ky-Echo space, and the three-dimensional K-space is defined as the Kx-Ky-Kz-Echo space.

[0075] Step S102: Recover the K-space data set along at least one phase encoding direction and the echo time dimension to obtain an updated K-space data set.

[0076] Since the K-space data set is undersampled along at least one phase encoding direction and the echo time dimension, there are unknown points in the K-space data set. In order to reconstruct a relatively complete echo planar image, it is necessary to recover the K-space data set along at least one phase encoding direction and the echo time dimension, that is, to perform data recovery on the unknown points. During the data recovery process, a non-linear factor can be introduced into the K-Echo space interpolation reconstruction algorithm. For example, a trained model is used. The K-space data set is input into the trained model to obtain an updated K-space data set. The trained model can predict the unknown points in the undersampled data to recover the undersampled data.

[0077] Step S103: Reconstruct a plurality of echo planar images according to the updated K-space data set.

[0078] When reconstructing the image, by recovering the undersampled data, relatively complete sampling data is obtained. These relatively complete sampling data include magnetic resonance signals with phase and frequency characteristics. According to the corresponding relationship of the voxel encoding of each slice, the voxel signal magnitude is represented by the gray level, thereby forming an echo planar image.

[0079] In the above steps S101 to S103, by expanding the K-space data set from the K-space to the K-Echo space, it can play the role of obtaining multiple echo planar images with different echo times by only single reconstruction. By acquiring the K-space data set with undersampling and introducing non-linear factors into the K-Echo space interpolation reconstruction algorithm based on deep learning, it can play the role of suppressing or avoiding artifacts. Finally, the reconstruction result is similar to that using the low-rank iterative reconstruction algorithm, but the calculation speed is faster. With such settings, the reconstruction quality is improved without increasing the reconstruction workload, and the signal-to-noise ratio, robustness, and calculation speed of the image are enhanced.

[0080] In one embodiment, the K-space data set includes a central region and a peripheral region. The central region is fully sampled, and the peripheral region is undersampled. The central region corresponds to the low-frequency components of the MR image. The peripheral region corresponds to the high-frequency components of the MR image. Considering that the key information is usually located in the central region, therefore, differentiating the fully sampled central region and the undersampled peripheral region in the K-space data set helps to retain the key information. At the same time, adopting an undersampling strategy in the peripheral region can shorten the overall sampling time.

[0081] The above embodiment mentions that a trained model can be used to recover the undersampled data of the K-space data set. In this embodiment, a method for training a model is provided, and this training method can be executed before step S102 of the above embodiment. Figure 5 A flowchart of the model training method is provided, as Figure 5 shown, and this process includes the following steps:

[0082] Step S201: Extract the repeating units obtained by sampling based on a preset undersampling pattern in the peripheral region;

[0083] Step S202: Use the repeating units as reconstruction kernels, traverse the data in the central region, and use the traversal results as the training data set.

[0084] There are unknown points in the reconstruction kernel. Using the reconstruction kernel to traverse the data in the central region can be regarded as using a set of grids to enclose the fully sampled data, and all the fully sampled data are known points. Then, it can be known the data of the known points around the unknown points in this set of grids, that is, the data of the known points in the reconstruction kernel are obtained. When the reconstruction kernel traverses the fully sampled data, the data values of the corresponding points are added to the first data set and the second data set respectively every time it passes through a place. Exemplarily, Figure 6 A schematic diagram of using the reconstruction kernel to traverse the fully sampled data is given. On the left is a 3×3×3 reconstruction kernel, and on the right is the traversal result. The fully sampled data can be just a small part of the fully sampled scan data of the same subject, and this small part of the fully sampled scan data can be collected together with the undersampled data, that is, the calibration data collected before the EPI sequence downsampling scan.

[0085] Step S203: Train the initial model based on the training data set to obtain a trained model. The trained model can represent the reconstruction convolution kernel corresponding to the reconstruction of the K-space data set.

[0086] The input for model training is the data of known points in the reconstruction kernel, and the output of model training is the data of unknown points in the reconstruction kernel. By establishing a mapping relationship between the known points and the unknown points in the reconstruction kernel, based on the trained model, the data of the unknown points is predicted. In the training process of this embodiment, it is not necessary to collect complicated images for training, but only to train based on a small amount of fully sampled data, and the fully sampled data has a short acquisition time and a simple training process. Optionally, when training the model, in order to obtain a well-trained model, the process of data input and output in step S203 can be repeated in a loop until the model meets the preset convergence condition, and the training ends.

[0087] In one embodiment, the model can be a convolutional neural network, and the convolutional neural network includes at least one non-linear activation layer. The non-linear activation layer is used to perform non-linear transformation and add non-linear features to the convolutional neural network to train a model that conforms to the corresponding features. As an example, Figure 7 a schematic structural diagram of a convolutional neural network is given, as Figure 7 shown. The convolutional neural network includes a convolutional layer, a pooling layer, and a fully connected layer. Among them, the fully connected layer is at the end of the network and can be set in multiple layers; the convolutional layer and the pooling layer are paired one by one, and multiple convolution-pooling operations can be set. Each convolutional layer can include an activation function to assist in expressing complex features, and this activation function is the non-linear activation layer. This activation function can be a rectified linear unit (ReLU). Optionally, this activation function can also be a ReLU with a slope (Leaky ReLU), a randomized ReLU, etc. The convolutional neural network used in this embodiment includes at least one convolution-pooling operation and contains an activation function. In addition, the mean square error can be used as the residual function (loss function) of the convolutional neural network.

[0088] In one embodiment, the recovery of the K-space data set along at least one phase encoding direction and the echo time dimension can be achieved in the following way: construct a training data set according to the central region, and train an initial model based on the training data set to obtain a trained model. The trained model can represent the reconstruction convolution kernel corresponding to the reconstruction of the K-space data set; input the K-space data set into the trained model to obtain an updated K-space data set.

[0089] In this embodiment, the central region can be divided into a first data set and a second data set, where the first data set corresponds to the unknown points in the edge region, and the second data set corresponds to the known points in the edge region. Based on the first data set and the second data set, a training data set is constructed, and based on the mapping relationship between the first data set and the second data set, an initial model is constructed. In this embodiment, the training data set is constructed through the central region of the K-space data set and the initial model is trained. The amount of data required for the training process is small, avoiding the long data collection and training processes in traditional deep learning methods, achieving the effects of convenience, speed, and accuracy.

[0090] In one embodiment, during the process of constructing the training data set, the repeating units obtained by undersampling based on a preset undersampling pattern in the edge region can be extracted; using the repeating units as reconstruction kernels, the data in the central region is traversed, and the traversal results are used as the training data set.

[0091] In this embodiment, the preset undersampling pattern has regularity and repeatability. For example, for the Kx-Ky-Echo space, the original data resolution is m×m. It is possible to use the same preset undersampling pattern to sample the Kx-Ky-echo space for every a×b×c space, and finally piece them together in the same pattern to form a global undersampling pattern of m×m×a×b×c for convenient image reconstruction. Each a×b×c space in the undersampled data is a repeating unit. The lengths of the repeating units along the three phase encoding directions can be the same or different. The smallest repeating unit can be used as the reconstruction kernel, or a slightly larger repeating unit can be used as the reconstruction kernel. The larger the size of the reconstruction kernel, the greater the computational amount. Depending on the number of undersampling trajectories, the reconstruction kernel can be one or more.

[0092] In this embodiment, by extracting the repeating units and constructing the training data set based on them, the model can better understand and adapt to the undersampling pattern in the K-Echo space, thereby improving the accuracy during reconstruction.

[0093] In some embodiments, in the central region, the intervals along each phase encoding direction are the same, and the intervals along the echo time dimension are the same; and in the edge region, the intervals along each phase encoding direction are the same, and the intervals along the echo time dimension are the same. Among them, the interval along each phase encoding direction, which can also be called the K-space interval, refers to the distance between every two adjacent acquisition points in each phase encoding direction in the frequency domain. The interval along the echo time dimension, which can also be called the echo interval, refers to the time interval between the midpoints of two adjacent echoes in the echo train. The echo midpoint is the time point with the largest echo signal. For example, in the Kx-Ky-Echo space, it is ensured that ΔKx, ΔKy, and the echo interval always remain unchanged. Such a setting makes the preset undersampling pattern regular and repeatable, facilitating reconstruction using deep learning.

[0094] Furthermore, the intervals between the central region and the peripheral region along the same phase encoding direction are consistent; and the intervals between the central region and the peripheral region along the echo time dimension are consistent. For example, for the same subject under examination, in the Kx-Ky-Echo space, the ΔKx, ΔKy, and echo interval of the central region are consistent with the corresponding parameter intervals of the peripheral region. With such settings, the effectiveness of the training results is ensured. In practical applications, the consistency of the relevant parameter intervals between the central region and the peripheral region can be achieved by designing consistent EPI sequence parameters.

[0095] In one embodiment, the data of the central region is obtained in the following manner: based on acquisitions with opposite polarities, two initial calibration data with echo intervals consistent with the target under-sampled data of the multi-pulse echo are acquired; the two initial calibration data are combined to generate the data of the central region with the influence of eddy currents removed.

[0096] In this embodiment, the target under-sampled data is the data obtained by uniformly under-sampling in the k-echo space based on the multi-pulse EPI technology. This target under-sampled data can have the same polarity as the initial calibration data used as the training input in the training stage of the reconstruction model. Among them, Figure 8 is a schematic diagram of a uniformly under-sampled space, as Figure 8 shown. The horizontal axis of the box in the figure is the x-axis of the k-echo space, and the vertical axis is the y-axis of the k-echo space. Each arrow line represents a data acquisition, and the arrow direction represents the direction of data acquisition. The numbers 1 to 6 on the right side of some arrow lines represent the acquisition order. For example, as Figure 8 shown, the acquisition order in numerical order is successively right, left, right, left, right, left. The dotted lines divide different radio frequency pulse excitations, that is, shots. Taking the first shot as an example, after the pulse excitation, the data is acquired in sequence according to the numbers on the right side of the figure. After that, for each pulse excitation, the same applies. Finally, uniformly under-sampled parallel line-style data is obtained in the k-echo space. Among them, after each pulse excitation, the above data acquisition order is often repeated several times. For example, the acquisition order after the first shot excitation can be (1, 2, 3, 4, 5, 6, 1, 2, 3, 4, 5, 6,..., 1, 2, 3, 4, 5, 6). After repeating several times, the data acquisition process corresponding to this excitation is ended, thereby improving the data utilization rate. It should also be noted that the numbers on the right side of the above arrows are only examples and do not constitute specific limitations on the data acquisition order and number of times in actual acquisitions.

[0097] Initial calibration data can be acquired based on sequences such as echoplanar spectroscopic imaging (abbreviated as EPSI) or multi-echo gradient echo imaging (abbreviated as GRE) before collecting multi-pulse EPI data. The above initial calibration data can be full-acquisition data completely acquired after being excited by multiple pulses. Specifically, two types of initial calibration data can be acquired using an acquisition method with opposite readout gradients. Among them, using two types of initial calibration data with opposite readout gradients can make each acquired data have forward and reverse reading results, and the specific impact caused by eddy currents can be obtained by calculating the difference between these two results during the reconstruction process, so as to perform eddy current correction during the subsequent reconstruction process.

[0098] Specifically, two types of initial calibration data can be acquired using a method with opposite readout gradient directions and an echo spacing consistent with the data to be reconstructed. Among them, the purpose of using two types of initial calibration data with opposite readout gradients is to ensure that each data has forward and reverse reading results, and the specific impact caused by eddy currents can be obtained by calculating the difference between these two results during the reconstruction process, so as to perform eddy current correction during the subsequent reconstruction process. In addition, the purpose of the echo spacing being consistent with the multi-pulse EPI data to be reconstructed is also to make the calibration data and the subsequent data to be reconstructed completely spatially corresponding.

[0099] Figure 9 This is a schematic diagram of the acquisition of initial calibration data for this embodiment. As Figure 9 shown, among them, the serial number in the upper left corner of each inner box represents an echo, Figure 9 3 echoes are shown in Figure 9 and the ellipsis "..." in Figure 3 means that there can be more echoes. Among them, the horizontal axis in each inner box represents the x-axis of the k-echo space, and the vertical axis represents the y-axis of the k-echo space. For each radio frequency pulse excitation shot, a certain row in the k-echo space is acquired after the pulse excitation, and this row is continuously acquired several times, for example, 20 echoes of this row can be acquired. The waveform curve RF in the figure represents the radio frequency pulse. Figure 9 Each arrow in

[0100] Figure 10 can represent a pulse excitation, and the arrow direction represents the data acquisition direction. After multiple pulse excitations, a part of the full-acquisition data at the center of the k-echo space can be obtained, that is, the above initial calibration data. Among them, Figure 3The first type of initial calibration data and the second type of initial calibration data with opposite polarities. When collecting the initial calibration data, a certain row can be collected multiple times. Figure 10 The innermost boxes in it all represent the data acquisition in the k-echo space. The horizontal axis represents the x-axis of the k-echo space, and the vertical axis represents the y-axis of the k-echo space. The waveform curve RF in the figure represents the radio frequency pulse. The numerical labels in the upper left corner of each innermost box represent echo 1, echo 2, echo 3, and so on. Each arrow in the innermost box represents a pulse excitation, and the arrow direction represents the data acquisition direction. Figure 10 Multi-pulse excitation is shown in it.

[0101] In Figure 10 After obtaining the two types of initial calibration data, the data in the central region without eddy current influence can be generated based on these two types of initial calibration data. Then, one of the initial calibration data can be used as the training input of the initial model, and the data in the central region can be used as the training output of the initial model to train the initial model, so as to obtain a well-trained model. It can be understood that Figure 10 The number of echoes and the number of pulse excitations marked in it are only examples and do not constitute actual limitations. In addition, in addition to Figure 10 the method shown in it, other methods can also be used to combine the two different initial calibration data to generate the data in the central region without eddy current influence and without abnormal deviation between odd and even echoes. This embodiment does not make specific limitations on this.

[0102] In this embodiment, by using two types of initial calibration data with opposite readout gradient directions and the same echo interval as the echo interval of the multi-pulse under-sampled data to construct the data in the central region without eddy current influence, reliable training data can be provided for the subsequent training of the model. Therefore, this embodiment can improve the eddy current removal ability of the subsequent model, eliminate the abnormal difference between the odd-echo image and the even-echo image in the reconstruction result due to the eddy current influence, so that the reconstructed echo image can be easily quantitatively calculated, and the accuracy and stability of the echo image can be improved.

[0103] Further, in one embodiment, the initial calibration data is the full acquisition data selected from within a preset range of the echo space center obtained after multi-pulse excitation. Wherein, the echo space is the k-echo space, and within the preset range of the echo space center are several rows pre-set based on the actual application scenario and as close as possible to the k-echo space center. For several rows close to the k-echo space center, such as 24 rows or 32 rows close to the k-echo space center, the initial calibration data is repeatedly acquired to obtain the full acquisition data close to the k-echo space center. In addition, according to the requirements of the actual application scenario, other rows close to the k-echo space center can also be used as the preset range of the echo space center. By acquiring the full acquisition data within the preset range of the k-echo space center obtained after multi-pulse excitation as the initial calibration data, the accuracy of data reconstruction can be improved, enabling a trained complete model to obtain complete reconstruction data, and further obtaining multiple echo images through a single scan.

[0104] In one embodiment, after obtaining multiple planar echo images, the method further includes: optimizing the multiple planar echo images based on the estimated information of the signal change of pixel points in the multiple planar echo images with respect to the echo time to obtain corrected multi-echo images.

[0105] In this embodiment, given the known sequence timing diagram, the change of the signal of any pixel point with respect to the echo time is regular, for example, determined by factors such as its own relaxation time and proton density. Therefore, before performing the sequence scan, all possible evolution patterns of the signal change with respect to the echo time, that is, the above-mentioned estimated information, can be traversed through simulation, and singular value decomposition is used for dimensionality reduction in the dimension of the echo time to obtain the basis (usually reduced to 5 to 10 bases), and then the echo images obtained by the above reconstruction are iterated based on the basis. The specific iteration process can be expressed as:

[0106]

[0107] Wherein, ψ is the dimensionality reduction matrix, X is the unknown quantity (in the k-echo space data after dimensionality reduction, ψX is the original k-echo space data), S is the coil sensitivity matrix, F is the Fourier transform, and y is the echo image obtained in this embodiment.

[0108] Specifically, given known parameters such as relaxation time and magnetization vector per unit volume, the above-mentioned dimensionality reduction matrix ψ can be calculated based on the Bloch equation. Specifically:

[0109]

[0110] Wherein, M is the magnetization vector per unit volume, t is the time, B is the magnetic field, γ is the gyromagnetic ratio, M xy and Mz They are the transverse component and the longitudinal component of M respectively, which is M in the resting state z , and T2 and T1 are the transverse relaxation time and the longitudinal relaxation time respectively. Parameters such as the magnetization vector and the relaxation time in the human body have significant numerical ranges. After traversing all the parameter values within the corresponding numerical ranges and substituting them into the above Bloch equation, all possible magnetic resonance signal evolution curves can be obtained, that is, the estimation of all the above possible evolution modes of the signal with the echo time is completed. Then, after performing singular value decomposition on all the evolution curves and reducing the dimension to find the intrinsic relationship, the above reduction matrix ψ can be obtained.

[0111] Thus, through the above iterative process, this embodiment realizes the optimization of the echo image. This embodiment utilizes the low rank property of the multi-echo image in the time dimension, so that a multi-echo image with a higher signal-to-noise ratio and smoother in the time dimension can be obtained on the premise of slightly increasing the reconstruction time.

[0112] In one embodiment, another echo planar imaging method is provided. Figure 11 is the flowchart of the echo planar imaging method of this embodiment, as Figure 11 shown, and this process includes the following steps:

[0113] Step S301, obtain a K-space data set obtained by performing radio frequency excitation on the region to be imaged. The K-space data set includes an echo time dimension, and the K-space data set is undersampled along at least one phase encoding direction and the echo time dimension.

[0114] In this embodiment, a data axis can be set in the obtained magnetic resonance data center to represent the echo time dimension, similar to the Kx axis or the Ky axis in K-space. For the convenience of description, the K-space including the echo time dimension is hereinafter defined as the K-Echo space (or K-echo space), where the one-dimensional K-space is defined as the Kx-Echo space, the two-dimensional K-space is defined as the Kx-Ky-Echo space, and the three-dimensional K-space is defined as the Kx-Ky-Kz-Echo space. Specific examples related to this step can refer to the examples described in the foregoing embodiments and optional implementation manners, and will not be elaborated in this embodiment.

[0115] Step S302, obtain initial calibration data based on acquisitions with opposite polarities, and generate data for the central region after removing the influence of eddy currents according to the initial calibration data.

[0116] Specific examples related to this step can refer to the examples described in the foregoing embodiments and optional implementation manners, and will not be elaborated in this embodiment.

[0117] Step S303: Based on the data in the central region, the K-space data set is restored along at least one phase-encoding direction and the echo time dimension to obtain an updated K-space data set.

[0118] Since the K-space data set is undersampled along at least one phase-encoding direction and the echo time dimension, there are unknown points in the K-space data set. To reconstruct a relatively complete echo planar image, it is necessary to restore the K-space data set along at least one phase-encoding direction and the echo time dimension, that is, to perform data restoration on the unknown points. During the data restoration process, non-linear factors can be introduced into the K-Echo space interpolation reconstruction algorithm. For example, a training data set can be constructed based on the central region, and an initial model can be trained based on the training data set to obtain a trained model. The K-space data set is input into the trained model to obtain an updated K-space data set. The trained model can predict the unknown points in the undersampled data to restore the undersampled data.

[0119] Among them, since the data in the central region is the data obtained by removing the eddy current influence from the initial calibration data, the initial calibration data is used as the input and the data in the central region is used as the output to train the initial model, so that the trained model can simulate the mathematical relationship between the input data and the output data, that is, the mathematical relationship between the data containing the eddy current influence and the data after removing the eddy current influence. Furthermore, the trained model can be used to perform eddy current removal processing on the data. The above model can specifically be any one of the following: a generalized auto-calibrating partially parallel acquisitions (GRAPPA) model, a bipolar gradient pulse model, and a bipolar gradient echo model. In addition, other models can also be used based on the actual application scenario, and this embodiment does not make specific limitations. Exemplarily, the initial model can be trained based on the following formula to obtain a trained model:

[0120]

[0121] where S kx,ky,t represents the unknown point in the k-echo space, represents the known points within the preset neighborhood of the unknown point, Indicates the weight for predicting unknown points based on known points within a preset neighborhood. Based on the above training process, a trained complete model can accurately simulate the above weights. Among them, the k-echo space is the dual space of the ordinary space under Fourier transform and is mainly applied to the imaging analysis of magnetic resonance imaging. In this embodiment, when training the initial model, by considering odd echoes and even echoes separately, using the initial calibration data as the input data for training, and using the data in the central region with the eddy current effect removed as the reconstruction target, that is, the output data of training, to train the initial model, so that a model that can eliminate the offset between odd echoes and even echoes can be trained.

[0122] In this step, other specific examples related to the central region can refer to the examples described in the foregoing embodiments and alternative embodiments, and will not be elaborated in this embodiment.

[0123] Step S304, reconstruct a plurality of echo planar images based on the updated K-space data set.

[0124] When reconstructing an image, by restoring the under-sampled data, relatively complete sampling data is obtained. These relatively complete sampling data contain magnetic resonance signals with phase and frequency characteristics. According to the corresponding relationship of voxel encoding for each slice, the voxel signal magnitude is represented by gray levels, thereby forming an echo planar image. Based on step S303, for each unknown point in the k-echo space, its corresponding weight can be output. Therefore, based on a trained complete model, a weight model can be output for the data to be reconstructed. By performing a convolution operation on the data to be reconstructed and this weight model, complete reconstruction data can be obtained. In the complete reconstruction data, the same echo time has complete data. Performing an inverse Fourier transform on the data with the same echo time can obtain an echo image. Therefore, in this embodiment, multiple echo images can be obtained through a single scan. It can be understood that in the echo planar imaging sequence scan, due to the existence of eddy current effects, there is an offset between odd echoes and even echoes. The final reconstruction result of this embodiment can remove the eddy current effect, thereby eliminating the abnormal difference caused by eddy currents between odd echo images and even echo images, making the reconstructed multi-echo images smoother and without jitter in the echo dimension, which is beneficial for post-processing such as quantitative calculation, and the reconstruction accuracy, signal-to-noise ratio, and stability have all been significantly improved.

[0125] In the above steps S301 to S304, by expanding the K-space dataset from the K-space to the K-Echo space, it is possible to obtain multiple echo planar images with different echo times through only a single reconstruction. By obtaining the K-space dataset with undersampling and introducing non-linear factors into the interpolation reconstruction algorithm in the K-Echo space based on deep learning, it is possible to suppress or avoid artifacts. The final reconstruction result is similar to that using the low-rank iterative reconstruction algorithm, but the calculation speed is faster. By obtaining the initial calibration data, generating the data of the central region with the influence of eddy currents removed based on the initial calibration data, and restoring the undersampled data based on the data of the central region, it can remove the influence of eddy currents in the reconstructed echo images, thereby eliminating the offset between the odd-echo images and the even-echo images caused by the influence of eddy currents, obtaining smoother and non-jittery data in the echo dimension, which is convenient for quantitative calculation and improves the accuracy and stability of the echo images.

[0126] In one embodiment, in the process of reconstructing the updated K-space dataset based on the trained model, it may include: reconstructing the target undersampled data of the multi-pulse echo using the trained model to obtain the updated K-space dataset.

[0127] In this embodiment, the target undersampled data of the multi-pulse echo is input into the trained model to predict the unknown points in the target undersampled data to obtain the fully sampled data, and the echo image is reconstructed based on the fully sampled data. Among them, after predicting the target undersampled data, the above weight model can be obtained. Then, the obtained weight model is convolved on the target undersampled data, so that the fully sampled data can be obtained, and the complete echo image can be reconstructed. By predicting the unknown points in the target undersampled data using the trained model, the accuracy of subsequent reconstruction can be improved.

[0128] In addition, in one embodiment, after obtaining multiple echo planar images, the method further includes: optimizing the multiple echo planar images based on the estimated information of the signal change of the pixel points in the multiple echo planar images with the echo time to obtain the corrected multi-echo images. With this setting, the optimization of the echo images is realized. Other specific examples related to this step can refer to the examples described in the foregoing embodiments and alternative embodiments, and will not be elaborated in this embodiment.

[0129] In one embodiment, the above model is a model based on the parallel imaging reconstruction algorithm. Among them, the parallel imaging reconstruction algorithm is the above global auto-calibration partial parallel acquisition GRAPPA model. Figure 12 It is a flowchart of another echo planar imaging method of this embodiment. As Figure 12 shown, this echo planar imaging method includes the following steps:

[0130] Step S401, before acquiring the target under-sampled data of multi-pulse EPI, obtain two initial calibration data acquired based on opposite polarities and having the same echo spacing as the target under-sampled data of the multi-pulse echoes;

[0131] Step S402, merge the two initial calibration data to generate central region data with the eddy current effect removed;

[0132] Step S403, use one of the initial calibration data as the input data for training, and use the central region data as the output data for training to train the GRAPPA model to obtain a fully trained model;

[0133] Step S404, obtain the target under-sampled data of uniformly under-sampled multi-pulse EPI;

[0134] Step S405, input the target under-sampled data into the above-mentioned fully trained model to obtain a weight model;

[0135] Step S406, perform convolution of the weight model on the target under-sampled data to obtain fully sampled data, and obtain an echo image based on the fully sampled data;

[0136] Step S407, optimize multiple planar echo images based on the estimated information of the signal change of pixel points in the multiple planar echo images with the echo time, and obtain corrected multi-echo images.

[0137] The above Steps S401 to S407 can remove the eddy current effect in the reconstructed echo image, thereby eliminating the offset between the odd echo image and the even echo image caused by the eddy current effect, obtaining smoother and jitter-free data in the echo dimension, facilitating quantitative calculation, and improving the accuracy and stability of the echo image.

[0138] In one embodiment, a planar echo imaging device is further provided for implementing the above embodiment and preferred implementation manners, and those that have been described will not be repeated here. The following terms "module", "unit", "sub-unit", etc. can be a combination of software and / or hardware that can achieve a predetermined function. Although the device described in the following embodiments is preferably implemented in software, implementation in hardware, or a combination of software and hardware is also possible and contemplated.

[0139] The planar echo imaging device includes: a first acquisition module configured to acquire a K-space data set obtained by radio frequency excitation of an area to be imaged, the K-space data set including an echo time dimension and being undersampled along at least one phase encoding direction and the echo time dimension; a restoration module configured to restore the K-space data set along at least one phase encoding direction and the echo time dimension to obtain an updated K-space data set; and a reconstruction module configured to reconstruct a plurality of planar echo images based on the updated K-space data set.

[0140] In one embodiment, the restoration module includes: a training unit configured to construct a training data set based on a central region and train an initial model based on the training data set to obtain a trained model, the trained model being capable of characterizing a reconstruction convolution kernel corresponding to the reconstruction of the K-space data set; and an input unit configured to input the K-space data set into the trained model to obtain an updated K-space data set.

[0141] In one embodiment, the training unit includes: an extraction subunit configured to extract repeating units obtained by sampling based on a preset undersampling pattern in an edge region; and a traversal subunit configured to use the repeating units as reconstruction kernels to traverse data in the central region and use the traversal results as a training data set.

[0142] In one embodiment, the device further includes a second acquisition module configured to acquire two initial calibration data with echo intervals consistent with target undersampled data of multi-pulse echoes based on acquisitions with opposite polarities; and combine the two initial calibration data to generate data for the central region with eddy current effects removed.

[0143] In one embodiment, the device further includes: an optimization module configured to optimize the plurality of planar echo images based on estimated information about the variation of signals of pixel points in the plurality of planar echo images with respect to echo time to obtain corrected multi-echo images.

[0144] Based on the same inventive concept as the above planar echo imaging device, in one embodiment, another planar echo imaging device is provided, which includes:

[0145] an acquisition module configured to acquire a K-space data set obtained by radio frequency excitation of an area to be imaged, the K-space data set including an echo time dimension and being undersampled along at least one phase encoding direction and the echo time dimension;

[0146] a generation module configured to acquire initial calibration data based on acquisitions with opposite polarities and generate data for the central region with eddy current effects removed according to the initial calibration data;

[0147] A recovery module, configured to recover the K-space data set along at least one phase encoding direction and the echo time dimension based on the data in the central region, so as to obtain an updated K-space data set;

[0148] A reconstruction module, configured to reconstruct a plurality of echo planar images according to the updated K-space data set.

[0149] In one embodiment, the apparatus further includes: an optimization module, configured to optimize the plurality of echo planar images based on the estimated information of the signal of the pixel points in the plurality of echo planar images changing with the echo time, so as to obtain corrected multi-echo images.

[0150] It should be noted that the above-mentioned respective modules can be functional modules or program modules, and can be implemented either by software or by hardware. For the modules implemented by hardware, the above-mentioned respective modules can be located in the same processor; or the above-mentioned respective modules can also be located in different processors in any combined form.

[0151] In addition, in combination with the echo planar imaging method provided in the above-mentioned embodiments, a storage medium can also be provided to implement in this embodiment. A computer program is stored on the storage medium; when the computer program is executed by a processor, the following steps are implemented:

[0152] Obtain a K-space data set obtained by radio frequency excitation of the region to be imaged. The K-space data set includes an echo time dimension, and the K-space data set is undersampled along at least one phase encoding direction and the echo time dimension;

[0153] Recover the K-space data set along at least one phase encoding direction and the echo time dimension, so as to obtain an updated K-space data set;

[0154] Reconstruct a plurality of echo planar images according to the updated K-space data set.

[0155] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0156] Construct a training data set according to the central region, and train an initial model based on the training data set to obtain a trained model, and the trained model can represent the reconstruction convolution kernel corresponding to the reconstruction of the K-space data set;

[0157] Input the K-space data set into the trained model to obtain an updated K-space data set.

[0158] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0159] Extract the repeating units obtained by undersampled sampling based on a preset in the edge region;

[0160] Using the repeating unit as the reconstruction kernel, traverse the data in the central region, and use the traversal result as the training data set.

[0161] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0162] Based on acquisitions with opposite polarities, obtain two initial calibration data sets whose echo intervals are consistent with the target under-sampled data of the multi-pulse echo;

[0163] Merge the two initial calibration data sets to generate data for the central region with the eddy current effect removed.

[0164] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0165] Based on the estimated information of the signals of the pixel points in multiple planar echo images changing with the echo time, optimize the multiple planar echo images to obtain corrected multi-echo images.

[0166] In one embodiment, when the computer program is executed by a processor, the following steps are implemented:

[0167] Obtain a k-space data set obtained by performing radio frequency excitation on the region to be imaged. The k-space data set includes an echo time dimension, and the k-space data set is under-sampled along at least one phase encoding direction and the echo time dimension;

[0168] Obtain initial calibration data based on acquisitions with opposite polarities, and generate data for the central region with the eddy current effect removed according to the initial calibration data;

[0169] Based on the data in the central region, recover the k-space data set along at least one phase encoding direction and the echo time dimension to obtain an updated k-space data set;

[0170] Reconstruct multiple planar echo images according to the updated k-space data set.

[0171] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0172] Based on the estimated information of the signals of the pixel points in multiple planar echo images changing with the echo time, optimize the multiple planar echo images to obtain corrected multi-echo images.

[0173] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as Figure 13As shown in the figure. The computer device includes a processor, a memory, a communication interface, a display screen, and an input device connected via a system bus. Among them, 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 computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The communication interface of the computer device is used to communicate with external terminals in a wired or wireless manner, and the wireless manner can be achieved through WIFI, mobile cellular networks, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements a planar echo imaging method. 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 buttons, trackballs, or touchpads provided on the computer device housing, or an external keyboard, touchpad, or mouse, etc.

[0174] Those skilled in the art can understand that Figure 13 the structure shown in the figure is only a block diagram of some parts of the structure related to the solution of this application, and does not constitute a limitation on the computer device to which the solution of this application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements.

[0175] It should be understood that the specific embodiments described here are only used to explain this application, rather than to limit it. According to the embodiments provided in this application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of this application.

[0176] Obviously, the drawings are only some examples or embodiments of this application. For those of ordinary skill in the art, this application can also be applied to other similar situations based on these drawings without creative work. In addition, it can be understood that although the work done during the development process here may be complex and time-consuming, for those of ordinary skill in the art, certain design, manufacturing, or production changes based on the technical content disclosed in this application are only conventional technical means and should not be regarded as insufficient disclosure of this application.

[0177] The term "embodiment" in this application means that the specific features, structures, or characteristics described in connection with the embodiment may be included in at least one embodiment of this application. The phrase appears in various positions in the specification does not necessarily mean the same embodiment, nor does it mean being independent or alternative to other embodiments and mutually exclusive. Those of ordinary skill in the art can clearly or implicitly understand that the embodiments described in this application can be combined with other embodiments without conflict.

[0178] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. The acquisition, storage, use, processing, etc. of data in the embodiments of this application all comply with the relevant provisions of national laws and regulations.

[0179] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. 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 methods. Among them, any reference to a memory, database, or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memories. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can 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), etc. The databases involved in the embodiments provided in this application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., without limitation. The processors involved in the embodiments provided in this application can be general-purpose processors, central processors, graphics processors, digital signal processors, programmable logic devices, data processing logics based on quantum computing, etc., without limitation.

[0180] The above-described embodiments merely represent several implementation manners of the present application. The description thereof is relatively specific and detailed, but it should not be construed as a limitation on the scope of patent protection. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all fall within the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the appended claims.

Claims

1. A planar echo imaging method, characterized in that Including: Obtaining a K-space data set obtained by performing radio frequency excitation on an imaging region to be imaged, the K-space data set including an echo time dimension, and the K-space data set being undersampled along at least one phase encoding direction and the echo time dimension; Restoring the K-space data set along at least one phase encoding direction and the echo time dimension to obtain an updated K-space data set; Reconstructing a plurality of echo planar images according to the updated K-space data set.

2. The planar echo imaging method according to claim 1, wherein, The K-space data set includes a central region and a peripheral region, the central region is fully sampled, and the peripheral region is undersampled.

3. The planar echo imaging method according to claim 2, wherein Restoring the K-space data set along at least one phase encoding direction and the echo time dimension includes: Constructing a training data set according to the central region, and training an initial model based on the training data set to obtain a trained model, the trained model being capable of characterizing a reconstruction convolution kernel corresponding to the reconstruction of the K-space data set; Inputting the K-space data set into the trained model to obtain the updated K-space data set.

4. The planar echo imaging method according to claim 3, wherein Constructing a training data set according to the central region, and training an initial model based on the training data set to obtain a trained model includes: Extracting a repeating unit obtained by sampling based on a preset undersampling pattern in the peripheral region; Using the repeating unit as a reconstruction kernel, traversing the data in the central region, and using the traversal result as the training data set.

5. The planar echo imaging method according to claim 4, characterized in that In the central region, the intervals along each phase encoding direction are the same, and the intervals along the echo time dimension are the same; and in the peripheral region, the intervals along each phase encoding direction are the same, and the intervals along the echo time dimension are the same.

6. The planar echo imaging method according to claim 5, wherein, The intervals between the central region and the peripheral region along the same phase encoding direction are consistent; and the intervals between the central region and the peripheral region along the echo time dimension are consistent.

7. The planar echo imaging method according to claim 2, characterized in that The data of the central region is obtained by the following method: Based on acquisitions with opposite polarities, obtaining two initial calibration data with echo intervals consistent with the target undersampled data of multi-pulse echoes; Combining the two initial calibration data to generate the data of the central region with the influence of eddy currents removed.

8. The planar echo imaging method according to any one of claims 1 to 7, characterized in that, After obtaining a plurality of echo planar images, the method further includes: Optimizing the plurality of echo planar images based on the estimated information of the signal change of the pixel points in the plurality of echo planar images with respect to the echo time to obtain corrected multi-echo images.

9. A planar echo imaging method, characterized in that, Including: Obtaining a K-space data set obtained by performing radio frequency excitation on an imaging region to be imaged, the K-space data set including an echo time dimension, and the K-space data set being undersampled along at least one phase encoding direction and the echo time dimension; Obtaining initial calibration data based on acquisitions with opposite polarities, and generating data of a central region with the influence of eddy currents removed according to the initial calibration data; Restoring the K-space data set along at least one phase encoding direction and the echo time dimension based on the data of the central region to obtain an updated K-space data set; Reconstructing a plurality of echo planar images according to the updated K-space data set.

10. The planar echo imaging method according to claim 9, wherein After obtaining a plurality of planar echo images, the method further includes: Optimizing the plurality of planar echo images based on the estimated information of the signal of the pixel points in the plurality of planar echo images changing with the echo time, to obtain corrected multi-echo images.

11. A magnetic resonance system, characterized in that, Including: A main magnet for generating a main magnetic field in the imaging space; Gradient coils for generating gradient magnetic fields in the imaging space; An RF excitation unit for generating RF excitation pulses, and performing RF excitation on the region to be imaged in the imaging space according to the RF excitation pulses to obtain magnetic resonance signals; An RF reception unit for receiving the magnetic resonance signals; An image reconstruction unit for performing the planar echo imaging method according to any one of claims 1 to 10 based on the magnetic resonance signals.

12. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, the steps of the planar echo imaging method according to any one of claims 1 to 10 are implemented.

13. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, the steps of the planar echo imaging method according to any one of claims 1 to 10 are implemented.

Images