Non-Cartesian trajectory stacking reconstruction method and device and magnetic resonance imaging system
The non-Cartesian trajectory stacking reconstruction method addresses aliasing artifacts and calculation inefficiencies by mapping undersampled data to a Cartesian grid and fitting a GRAPPA kernel, enhancing MR image quality and speed.
Patent Information
- Application Number
- CN202310799350.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-06-30
- Publication Date
- 2024-12-31
AI Technical Summary
Existing 3D non-Cartesian MRI reconstruction methods suffer from aliasing artifacts and require time-consuming calculations due to undersampling in the layer direction, especially in star-stacked datasets, which are not efficiently addressed by current algorithms like GRAPPA, SENSE, and SPIRiT.
A method and device that utilize a non-Cartesian trajectory stacking reconstruction technique, mapping undersampled imaging data to a Cartesian grid, fitting a GRAPPA kernel using Cartesian calibration data, and performing inverse FFT to reduce aliasing artifacts and speed up reconstruction, while maintaining image quality.
Significantly reduces aliasing artifacts and calculation time in MR images reconstructed from undersampled non-Cartesian trajectories, improving image quality and reconstruction speed.
Smart Images
Figure CN119224667A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of MRI (Magnetic Resonance Imaging) technology, in particular to a non-Cartesian trajectory stacking reconstruction method, device and MRI system. Background Art
[0002] A stack-of-stars, or radial stack, is a 3D (three-dimensional) trajectory used in MRI. It is a 3D non-Cartesian trajectory. It combines radial sampling in the plane and linear sampling in the slice direction (i.e., Cartesian sampling). Sequences using this trajectory are often called StarVIBE (Star Volume Interpolated Breath-hold Examination) sequences.
[0003] Figure 1 1 is an example diagram of a conventional Cartesian coordinate system and a star-shaped stacking trajectory, where 11 is a conventional Cartesian coordinate system. As shown in 11, in the Cartesian coordinate system, multiple parallel acquisition lines are distributed in the plane of each layer, and multiple acquisition points are evenly distributed on each acquisition line, and each acquisition point is located on a Cartesian grid; 12 is a star-shaped stacking trajectory, as shown in 12, multiple acquisition lines are radially distributed in the plane of each layer, and all acquisition lines start from the center of the layer and end at the edge of the layer. The acquisition lines at the same position in different layers are parallel to each other. Wherein, the x direction is the readout direction, the y direction is the phase encoding direction, and the z direction is the layer selection direction.
[0004] In practical applications of magnetic resonance imaging, there is always a demand for faster scans. A common approach is to skip some acquisition lines and reconstruct the image from the undersampled dataset using appropriate methods. As a 3D non-Cartesian trajectory, star stacks may be undersampled in the plane and / or in the slice direction, both of which usually require 3D non-Cartesian reconstruction methods.
[0005] The 3D non-Cartesian GRAPPA (Generalized Auto Calibrating Partially Parallel Acquisitions) algorithm can be used to reconstruct undersampled star-stacked datasets. Unlike the common algorithm for GRAPPA (i.e., the Cartesian algorithm), there is no single GRAPPA kernel that is globally applicable throughout the entire K-space. Instead, different "local" kernels must be applied to different locations in K-space, which is time consuming. In addition, since the acquired data points are not exactly on a Cartesian grid, some approximations must be made, which may increase signal distortion on the reconstructed image.
[0006] Algorithms based on 3D SENSE (SENSitivity Enconding) are another common alternative. It explicitly uses coil sensitivity maps to reconstruct the image, which are usually extracted through calibration scans. The main problem with the SENSE algorithm is that it is very sensitive to aliasing artifacts compared to the GRAPPA algorithm. Aliasing on the edges of the image will cause artifacts in the center. To avoid this problem, a larger FOV (Field Of View) or oversampling or saturation bands are usually used to eliminate aliasing, but this will increase the acquisition time.
[0007] The SPIRiT (iTerative Self-consistent Parallel Imaging Reconstruction from arbitrary k-space) algorithm and the derived ESPIRiT (anEigenvalue approach to SPIRiT) algorithm can also be used for 3D non-Cartesian reconstruction. They can reconstruct missing k-space lines in an iterative manner (GRAPPA type) or reconstruct a fully sampled image (SENSE type). Usually, some hyperparameters (e.g., regularization strength) must be determined empirically, and the reconstruction time will be more or less longer than the direct method, depending on the number of iterations. Summary of the invention
[0008] In view of this, on the one hand, an embodiment of the present invention proposes a non-Cartesian trajectory stacking reconstruction method and device to reduce the aliasing artifacts in the MR image reconstructed from the imaging data acquired by the non-Cartesian trajectory stacking acquisition method with undersampling in the layer direction, reduce the reconstruction calculation amount, and improve the reconstruction speed; on the other hand, an MRI system is proposed to reduce the aliasing artifacts in the MR image reconstructed from the imaging data acquired by the non-Cartesian trajectory stacking acquisition method with undersampling in the layer direction, reduce the reconstruction calculation amount, and improve the reconstruction speed.
[0009] A non-Cartesian trajectory stacking reconstruction method, the method comprising:
[0010] The three-dimensional imaging data of the target part is collected by using a non-Cartesian trajectory stacking acquisition method, wherein full sampling is performed in the plane of each layer, and undersampling is performed in the layer direction, and the collected three-dimensional imaging data is filled into the three-dimensional K space to obtain a three-dimensional K space imaging data set undersampled in the layer direction;
[0011] Acquire three-dimensional Cartesian calibration data of the target part, and fill the acquired three-dimensional Cartesian calibration data into the three-dimensional K space to obtain a three-dimensional K space Cartesian calibration data set;
[0012] Mapping the imaging data of each layer in the three-dimensional K-space imaging data set under-sampled in the layer direction to a Cartesian grid respectively, to obtain a three-dimensional K-space Cartesian imaging data set under-sampled in the layer direction;
[0013] The generalized self-calibration parallel acquisition GRAPPA kernel was obtained by fitting the three-dimensional K-space Cartesian calibration data set;
[0014] The GRAPPA kernel is used to reconstruct the undersampled 3D K-space Cartesian imaging dataset in the slice direction to obtain a fully sampled 3D K-space imaging dataset.
[0015] A reconstructed magnetic resonance MR image is obtained based on the fully sampled three-dimensional K-space imaging data set.
[0016] The step of mapping the imaging data of each layer in the three-dimensional K-space imaging data set undersampled in the layer direction to a Cartesian grid comprises:
[0017] Using a data-driven grid algorithm, the imaging data of each layer in the three-dimensional K-space imaging data set that is undersampled in the layer direction is interpolated to a Cartesian grid; or,
[0018] A two-dimensional non-uniform fast Fourier transform NUFF is performed on the imaging data of each layer in the three-dimensional K-space imaging data set with under-sampling in the layer direction to obtain an image, and then a fast Fourier transform is performed on the image to obtain two-dimensional K-space Cartesian data.
[0019] The method of using a data-driven grid algorithm to interpolate the imaging data of each layer in the three-dimensional K-space imaging data set undersampled in the layer direction to the Cartesian grid further includes:
[0020] The radial density compensation algorithm is used to reduce the signal amplitude of the imaging data located in the central area of the three-dimensional K space, and to increase the signal amplitude of the imaging data located in the edge area of the three-dimensional K space.
[0021] The method of acquiring three-dimensional imaging data of a target part by adopting a non-Cartesian trajectory stacking acquisition method includes:
[0022] Use the same acquisition trajectory type for different layers of the target area; or,
[0023] Use different acquisition trajectory types for different layers of the target area;
[0024] Furthermore, when different acquisition trajectory types are used for different layers of the target part, the acquisition trajectories of different layers are aligned or not aligned.
[0025] The non-Cartesian trajectory stacking is a star stacking or a spiral stacking.
[0026] The unit coordinate size of the three-dimensional K space corresponding to the three-dimensional K space imaging dataset undersampled in the slice direction and the K space corresponding to the three-dimensional K space Cartesian calibration dataset in each direction is the same;
[0027] Furthermore, filling the acquired three-dimensional Cartesian calibration data into the three-dimensional K space includes:
[0028] The acquired 3D Cartesian calibration data is filled into the central area of the 3D K space.
[0029] A non-Cartesian trajectory stacking reconstruction device, the device comprising:
[0030] An imaging data acquisition module is used to acquire three-dimensional imaging data of a target part by adopting a non-Cartesian trajectory stacking acquisition method, wherein full sampling is performed in the plane of each layer, and undersampling is performed in the layer direction, and the acquired three-dimensional imaging data is filled into the three-dimensional K space to obtain a three-dimensional K space imaging data set undersampled in the layer direction;
[0031] A calibration data acquisition module is used to acquire three-dimensional Cartesian calibration data of a target part, and fill the acquired three-dimensional Cartesian calibration data into a three-dimensional K space to obtain a three-dimensional K space Cartesian calibration data set;
[0032] A mapping module is used to map the imaging data of each layer in the three-dimensional K-space imaging data set under-sampled in the layer direction to a Cartesian grid, and after all the imaging data are mapped, a three-dimensional K-space Cartesian imaging data set under-sampled in the layer direction is obtained;
[0033] The generalized self-calibration parallel acquisition GRAPPA kernel fitting module is used to fit the GRAPPA kernel using the three-dimensional K-space Cartesian calibration data set;
[0034] The reconstruction module is used to reconstruct the three-dimensional K-space Cartesian imaging data set that is undersampled in the layer direction using the GRAPPA kernel to obtain a fully sampled three-dimensional K-space imaging data set; and perform inverse fast Fourier transform on the fully sampled three-dimensional K-space imaging data set to obtain a reconstructed magnetic resonance MR image.
[0035] The mapping module maps the imaging data of each layer in the three-dimensional K-space imaging data set undersampled in the layer direction to a Cartesian grid, including:
[0036] Using a data-driven grid algorithm, the imaging data of each layer in the three-dimensional K-space imaging data set that is undersampled in the layer direction is interpolated to a Cartesian grid; or,
[0037] A two-dimensional non-uniform fast Fourier transform NUFF is performed on the imaging data of each layer in the three-dimensional K-space imaging data set with under-sampling in the layer direction to obtain an image, and then a fast Fourier transform is performed on the image to obtain two-dimensional K-space Cartesian data.
[0038] The unit coordinate size of the three-dimensional K space corresponding to the three-dimensional K space imaging data set undersampled in the layer direction obtained by the imaging data acquisition module and the K space corresponding to the three-dimensional K space Cartesian calibration data set obtained by the calibration data acquisition module in each direction is the same, and,
[0039] The calibration data acquisition module fills the acquired three-dimensional Cartesian calibration data into the three-dimensional K space, including: filling the acquired three-dimensional Cartesian calibration data into the central area of the three-dimensional K space.
[0040] A magnetic resonance imaging (MRI) system comprises any one of the above-mentioned non-Cartesian trajectory stacking reconstruction devices.
[0041] In an embodiment of the present invention, the imaging data of each layer acquired by a non-Cartesian trajectory stacking acquisition method are respectively mapped to a Cartesian grid to convert a three-dimensional K-space imaging data set that is under-sampled in the layer direction into a three-dimensional K-space Cartesian imaging data set that is under-sampled in the layer direction, and then the three-dimensional K-space Cartesian calibration data set is used to fit the GRAPPA kernel, and the three-dimensional K-space Cartesian imaging data set that is under-sampled in the layer direction is reconstructed using the GRAPPA kernel to obtain a fully sampled three-dimensional K-space imaging data set, and IFFT is performed on the fully sampled three-dimensional K-space imaging data set to obtain a reconstructed MR image, thereby greatly reducing aliasing artifacts in the MR image reconstructed from the imaging data acquired by a non-Cartesian trajectory stacking acquisition method that is under-sampled in the layer direction, improving the MR image quality, and reducing the amount of reconstruction calculations, thereby improving the reconstruction speed. BRIEF DESCRIPTION OF THE DRAWINGS
[0042] The preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings so that the above and other features and advantages of the present invention will be more apparent to those skilled in the art. In the accompanying drawings:
[0043] Figure 1 Example diagrams for conventional Cartesian coordinate systems and star stacking trajectories;
[0044] Figure 2 A flowchart of a non-Cartesian trajectory stacking reconstruction method provided by an embodiment of the present invention;
[0045] Figure 3 It is a schematic diagram of a process of non-Cartesian trajectory stacking reconstruction in an application example of the present invention;
[0046] Figure 4 A schematic diagram of an MR image reconstructed using an existing method from full-sample imaging data of a volunteer's brain acquired by a star stack acquisition method;
[0047] Figure 5 A schematic diagram of an MR image obtained by using IFFT to reconstruct undersampled imaging data of a volunteer's brain acquired by a star stack acquisition method;
[0048] Figure 6 A schematic diagram of an MR image obtained by using the reconstruction method provided by an embodiment of the present invention to reconstruct undersampled imaging data of a volunteer's brain acquired by a star stack acquisition method;
[0049] Figure 7 A schematic structural diagram of a non-Cartesian trajectory stacking and reconstruction device provided by an embodiment of the present invention.
[0050] The reference numerals are as follows:
[0051] DETAILED DESCRIPTION
[0052] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention is further described in detail with reference to the following embodiments.
[0053] Figure 2 A flowchart of a non-Cartesian trajectory stacking reconstruction method provided by an embodiment of the present invention, wherein the specific steps are as follows:
[0054] Step 201: Acquire three-dimensional imaging data of the target part using a non-Cartesian trajectory stacking acquisition method, wherein full sampling is performed in the plane of each layer, and undersampling is performed in the layer direction, and the acquired three-dimensional imaging data is filled into the three-dimensional K space to obtain a three-dimensional K space imaging data set that is undersampled in the layer direction.
[0055] For example, if the number of layers is 72 and the acceleration factor is 3, then when acquiring the three-dimensional imaging data of the target part, only the imaging data of layers 1, 4, 7, ..., 70 are acquired.
[0056] Step 202: Acquire three-dimensional Cartesian calibration data of the target part, and fill the acquired three-dimensional Cartesian calibration data into the three-dimensional K space to obtain a three-dimensional K space Cartesian calibration data set.
[0057] The 3DK space corresponding to the 3D K space Cartesian calibration data set in step 202 has the same size of the unit coordinate (i.e., Δk (x, y, z)) in each direction as the 3DK space corresponding to the 3D K space imaging data set undersampled in the layer direction in step 201. The x direction is the readout direction, the y direction is the phase encoding direction, and the z direction is the layer selection direction.
[0058] In an optional embodiment, in step 202, filling the acquired three-dimensional Cartesian calibration data into the three-dimensional K space includes: filling the acquired three-dimensional Cartesian calibration data into the central area of the three-dimensional K space, that is, the three-dimensional Cartesian calibration data only covers the central area of the three-dimensional K space. The specific range of the central area can be set based on experience, etc. In practical applications, the typical size of the three-dimensional K space Cartesian calibration data set is 256 (x direction) × 24 (y direction) × 24 (z direction).
[0059] Step 203: Map the imaging data of each layer in the three-dimensional K-space imaging data set under-sampled in the layer direction to a Cartesian grid respectively, and complete the mapping of all imaging data to obtain a three-dimensional K-space Cartesian imaging data set under-sampled in the layer direction.
[0060] Step 204: The GRAPPA kernel is obtained by fitting the three-dimensional K-space Cartesian calibration data set.
[0061] The use of a three-dimensional K-space Cartesian calibration data set to fit the GRAPPA kernel is a mature technology and will not be described in detail here.
[0062] Step 205: Reconstruct the slice-direction under-sampled three-dimensional K-space Cartesian imaging dataset obtained in step 203 using the GRAPPA kernel to obtain a fully sampled three-dimensional K-space imaging dataset.
[0063] Step 206 : Obtain a reconstructed MR image according to the fully sampled three-dimensional K-space imaging data set obtained in step 205 .
[0064] In the above embodiment, the imaging data of each layer acquired by the non-Cartesian trajectory stacking acquisition method are respectively mapped to the Cartesian grid to convert the three-dimensional K-space imaging data set that is under-sampled in the layer direction into a three-dimensional K-space Cartesian imaging data set that is under-sampled in the layer direction, and then the three-dimensional K-space Cartesian calibration data set is used to fit the GRAPPA kernel, and the three-dimensional K-space Cartesian imaging data set that is under-sampled in the layer direction is reconstructed by the GRAPPA kernel to obtain a fully sampled three-dimensional K-space imaging data set, and the fully sampled three-dimensional K-space imaging data set is subjected to IFFT (Inverse Fast Fourier Transform) to obtain a reconstructed MR image, thereby greatly reducing the aliasing artifacts in the MR image reconstructed by the imaging data acquired by the non-Cartesian trajectory stacking acquisition method that is under-sampled in the layer direction, improving the MR image quality, and reducing the reconstruction calculation amount, thereby improving the reconstruction speed.
[0065] In an optional embodiment, in step 203, the imaging data of each layer in the three-dimensional K-space imaging data set that is undersampled in the layer direction is mapped to a Cartesian grid, including: using a data-driven grid algorithm to interpolate the imaging data of each layer in the three-dimensional K-space imaging data set that is undersampled in the layer direction to a Cartesian grid; or, performing a two-dimensional NUFF (Non-Uniform Fast Fourier Transform) on the imaging data of each layer in the three-dimensional K-space imaging data set that is undersampled in the layer direction to obtain an image, and then performing FFT on the image to obtain two-dimensional K-space Cartesian data.
[0066] In an optional embodiment, a data-driven grid algorithm is used to interpolate the imaging data of each layer in the three-dimensional K-space imaging data set that is undersampled in the layer direction to the Cartesian grid, and further includes: using a radial density compensation algorithm to reduce the signal amplitude of the imaging data located in the central area of the three-dimensional K-space, and to increase the signal amplitude of the imaging data located in the edge area of the three-dimensional K-space. In practical applications, the specific range of the central area and the edge area, as well as the specific adjustment size of the signal amplitude reduction and increase, can be determined based on experience, etc. The radial density compensation algorithm is such as: radial DCF (Density Compensation Function, density compensation function).
[0067] In an optional embodiment, in step 201, 3D imaging data of the target part is acquired by using a non-Cartesian trajectory stacking acquisition method, including: using the same acquisition trajectory type for different layers of the target part; or using different acquisition trajectory types for different layers of the target part, for example: the acquisition trajectory type of some layers is a radial trajectory, and the acquisition trajectory type of some layers is a spiral trajectory, etc. In addition, when different acquisition trajectory types are used for different layers of the target part, the acquisition trajectories of different layers are aligned or not aligned. Among them, when the acquisition trajectories of different layers are not aligned, the stripe artifacts caused by radial sampling can be reduced.
[0068] In an optional embodiment, the MR sequence used when acquiring 3D imaging data of the target part in step 201 and the MR sequence used when acquiring 3D Cartesian calibration data of the target part in step 202 are the same or similar types of sequences. For example, if the StarVIBE sequence is used in step 201, then the phase-scratched 3D GRE (Gradient Recalled Echo) sequence or VIBE sequence is used in step 202.
[0069] In an optional embodiment, the non-Cartesian trajectory stacking is a star stacking, or a spiral stacking, etc.
[0070] Figure 3 Schematic diagram of the process of non-Cartesian trajectory stacking reconstruction in an application example of the present invention. In this example, the non-Cartesian trajectory stacking is star stacking, and the acceleration factor = 3. Wherein: 31 is a schematic diagram of acquiring a three-dimensional K-space imaging data set by star stacking, wherein the solid line represents the acquired data and the dotted line represents the missing data; 32 is a schematic diagram of interpolating the three-dimensional K-space imaging data set obtained by the acquisition method of 31 to the Cartesian grid to obtain an undersampled three-dimensional K-space Cartesian imaging data set, wherein the solid line represents the acquired data and the dotted line represents the missing data; 33 is a schematic diagram of using the GRAPPA kernel to Figure 2 The diagram shown is a schematic diagram of reconstructing a fully sampled three-dimensional K-space Cartesian imaging dataset from an under-sampled three-dimensional K-space Cartesian imaging dataset.
[0071] Figure 4 A schematic diagram of an MR image reconstructed using an existing method from full-sample imaging data of a volunteer's brain acquired by a star stack acquisition method;
[0072] Figure 5 A schematic diagram of an MR image obtained by using IFFT to reconstruct undersampled imaging data of a volunteer's brain acquired by a star stack acquisition method;
[0073] Figure 6The figure is a schematic diagram of an MR image obtained by using the reconstruction method provided by an embodiment of the present invention to reconstruct undersampled imaging data of a volunteer's brain acquired by a star stacking acquisition method.
[0074] visible, Figure 5 There are more aliasing artifacts in Figure 6 Image quality and Figure 4 Get close and no aliasing artifacts are visible.
[0075] Figure 7 The structure diagram of a non-Cartesian trajectory stacking reconstruction device 70 provided in an embodiment of the present invention mainly includes: an imaging data acquisition module 71, a calibration data acquisition module 72, a mapping module 73, a GRAPPA kernel fitting module 74 and a reconstruction module 75, wherein:
[0076] The imaging data acquisition module 71 is used to acquire three-dimensional imaging data of the target part by adopting a non-Cartesian trajectory stacking acquisition method, wherein full sampling is performed in the plane of each layer, and under-sampling is performed in the layer direction, and the acquired three-dimensional imaging data is filled into the three-dimensional K space to obtain a three-dimensional K space imaging data set that is under-sampled in the layer direction, and the three-dimensional K space imaging data set that is under-sampled in the layer direction is sent to the mapping module 73.
[0077] The calibration data acquisition module 72 is used to acquire three-dimensional Cartesian calibration data of the target part, fill the acquired three-dimensional Cartesian calibration data into the three-dimensional K space to obtain a three-dimensional K space Cartesian calibration data set, and send the three-dimensional K space Cartesian calibration data set to the GRAPPA kernel fitting module 74.
[0078] The mapping module 73 is used to map the imaging data of each layer in the three-dimensional K-space imaging data set that is under-sampled in the layer direction to a Cartesian grid respectively, and after all the imaging data are mapped, a three-dimensional K-space Cartesian imaging data set that is under-sampled in the layer direction is obtained, and the three-dimensional K-space Cartesian imaging data set that is under-sampled in the layer direction is sent to the reconstruction module 75.
[0079] The GRAPPA kernel fitting module 74 is used to obtain the GRAPPA kernel by fitting the three-dimensional K-space Cartesian calibration data set, and send the GRAPPA kernel to the reconstruction module 75 .
[0080] The reconstruction module 75 is used to reconstruct the three-dimensional K-space Cartesian imaging data set that is undersampled in the layer direction using the GRAPPA kernel to obtain a fully sampled three-dimensional K-space imaging data set; and perform IFFT on the fully sampled three-dimensional K-space imaging data set to obtain a reconstructed MR image.
[0081] In an optional embodiment, the mapping module 73 maps the imaging data of each layer in the three-dimensional K-space imaging data set that is undersampled in the layer direction to a Cartesian grid, respectively, including: using a data-driven grid algorithm to interpolate the imaging data of each layer in the three-dimensional K-space imaging data set that is undersampled in the layer direction to a Cartesian grid; or, performing a two-dimensional NUFF on the imaging data of each layer in the three-dimensional K-space imaging data set that is undersampled in the layer direction to obtain an image, and then performing FFT on the image to obtain two-dimensional K-space Cartesian data.
[0082] In an optional embodiment, the mapping module 73 uses a data-driven grid algorithm to interpolate the imaging data of each layer in the three-dimensional K-space imaging data set that is undersampled in the layer direction to the Cartesian grid, and further includes: using a radial density compensation algorithm to lower the signal amplitude of the imaging data located in the central area of the three-dimensional K-space, and to increase the signal amplitude of the imaging data located in the edge area of the three-dimensional K-space.
[0083] In an optional embodiment, the imaging data acquisition module 71 acquires three-dimensional imaging data of the target part using a non-Cartesian trajectory stacking acquisition method, including: using the same acquisition trajectory type for different layers of the target part; or, using different acquisition trajectory types for different layers of the target part; and, when using different acquisition trajectory types for different layers of the target part, the acquisition trajectories of different layers are aligned or not aligned.
[0084] In an optional embodiment, the imaging data acquisition module acquires three-dimensional imaging data of the target part using a non-Cartesian trajectory stacking acquisition method, including: acquiring three-dimensional imaging data of the target part using a star stacking or spiral stacking acquisition method.
[0085] In an optional embodiment, the three-dimensional K space corresponding to the three-dimensional K space imaging data set undersampled in the layer direction obtained by the imaging data acquisition module 71 and the K space corresponding to the three-dimensional K space Cartesian calibration data set obtained by the calibration data acquisition module 72 have the same size of unit coordinates in each direction, and the calibration data acquisition module 72 fills the acquired three-dimensional Cartesian calibration data into the three-dimensional K space, including: filling the acquired three-dimensional Cartesian calibration data into the central area of the three-dimensional K space.
[0086] An embodiment of the present invention further provides an MRI system, which includes the non-Cartesian trajectory stacking reconstruction device 70 as described in any of the above embodiments.
[0087] Those skilled in the art will appreciate that the features described in the various embodiments and / or claims of the present disclosure may be combined and / or combined in a variety of ways, even if such combinations or combinations are not explicitly described in the present application. In particular, without departing from the spirit and teachings of the present application, the features described in the various embodiments and / or claims of the present application may be combined and / or combined in a variety of ways, and all of these combinations and / or combinations fall within the scope disclosed in the present application.
[0088] Specific embodiments are used herein to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method and core ideas of the present application, and is not used to limit the present application. For those skilled in the art, changes can be made in the specific implementation methods and application scopes according to the ideas, spirits and principles of the present application, and any modifications, equivalent substitutions, improvements, etc. made by them should be included in the scope of protection of the present application.
Claims
1. A non-Cartesian trajectory stacking reconstruction method, characterized in that: The method includes: The three-dimensional imaging data of the target part is collected by using a non-Cartesian trajectory stacking acquisition method, wherein full sampling is performed in the plane of each layer, and undersampling is performed in the layer direction, and the collected three-dimensional imaging data is filled into the three-dimensional K space to obtain a three-dimensional K space imaging data set undersampled in the layer direction; Acquire three-dimensional Cartesian calibration data of the target part, and fill the acquired three-dimensional Cartesian calibration data into the three-dimensional K space to obtain a three-dimensional K space Cartesian calibration data set; Mapping the imaging data of each layer in the three-dimensional K-space imaging data set under-sampled in the layer direction to a Cartesian grid respectively, to obtain a three-dimensional K-space Cartesian imaging data set under-sampled in the layer direction; The generalized self-calibration parallel acquisition GRAPPA kernel was obtained by fitting the three-dimensional K-space Cartesian calibration data set; The GRAPPA kernel is used to reconstruct the undersampled 3D K-space Cartesian imaging dataset in the slice direction to obtain a fully sampled 3D K-space imaging dataset. A reconstructed magnetic resonance MR image is obtained based on the fully sampled three-dimensional K-space imaging data set.
2. The method according to claim 1, characterized in that The step of mapping the imaging data of each layer in the three-dimensional K-space imaging data set undersampled in the layer direction to a Cartesian grid comprises: Using a data-driven grid algorithm, the imaging data of each layer in the three-dimensional K-space imaging data set that is undersampled in the layer direction is interpolated to a Cartesian grid; or, A two-dimensional non-uniform fast Fourier transform NUFF is performed on the imaging data of each layer in the three-dimensional K-space imaging data set with under-sampling in the layer direction to obtain an image, and then a fast Fourier transform is performed on the image to obtain two-dimensional K-space Cartesian data.
3. The method according to claim 2, characterized in that The method of using a data-driven grid algorithm to interpolate the imaging data of each layer in the three-dimensional K-space imaging data set undersampled in the layer direction to the Cartesian grid further includes: The radial density compensation algorithm is used to reduce the signal amplitude of the imaging data located in the central area of the three-dimensional K space, and to increase the signal amplitude of the imaging data located in the edge area of the three-dimensional K space.
4. The method according to claim 1, characterized in that: The method of acquiring three-dimensional imaging data of a target part by adopting a non-Cartesian trajectory stacking acquisition method includes: Use the same acquisition trajectory type for different layers of the target area; or, Use different acquisition trajectory types for different layers of the target area; Furthermore, when different acquisition trajectory types are used for different layers of the target part, the acquisition trajectories of different layers are aligned or not aligned.
5. The method according to claim 1, characterized in that The non-Cartesian trajectory stacking is a star stacking or a spiral stacking.
6. The method according to claim 1, characterized in that The unit coordinate size of the three-dimensional K space corresponding to the three-dimensional K space imaging dataset undersampled in the slice direction and the K space corresponding to the three-dimensional K space Cartesian calibration dataset in each direction is the same; Furthermore, filling the acquired three-dimensional Cartesian calibration data into the three-dimensional K space includes: The acquired 3D Cartesian calibration data is filled into the central area of the 3D K space.
7. A non-Cartesian trajectory stacking reconstruction device (70), characterized in that: The device (70) comprises: An imaging data acquisition module (71) is used to acquire three-dimensional imaging data of a target part by adopting a non-Cartesian trajectory stacking acquisition method, wherein full sampling is performed in the plane of each layer, and undersampling is performed in the layer direction, and the acquired three-dimensional imaging data is filled into the three-dimensional K space to obtain a three-dimensional K space imaging data set undersampled in the layer direction; A calibration data acquisition module (72) is used to acquire three-dimensional Cartesian calibration data of the target part, and fill the acquired three-dimensional Cartesian calibration data into the three-dimensional K space to obtain a three-dimensional K space Cartesian calibration data set; A mapping module (73) is used to map the imaging data of each layer in the three-dimensional K-space imaging data set under-sampled in the layer direction to a Cartesian grid, and after all the imaging data are mapped, obtain a three-dimensional K-space Cartesian imaging data set under-sampled in the layer direction; A generalized self-calibration parallel acquisition GRAPPA kernel fitting module (74) is used to fit the GRAPPA kernel using a three-dimensional K-space Cartesian calibration data set; The reconstruction module (75) is used to reconstruct the three-dimensional K-space Cartesian imaging data set that is undersampled in the layer direction using the GRAPPA kernel to obtain a fully sampled three-dimensional K-space imaging data set; and perform inverse fast Fourier transform on the fully sampled three-dimensional K-space imaging data set to obtain a reconstructed magnetic resonance MR image.
8. The device (70) according to claim 7, characterized in that The mapping module (73) maps the imaging data of each layer in the three-dimensional K-space imaging data set undersampled in the layer direction to a Cartesian grid, including: Using a data-driven grid algorithm, the imaging data of each layer in the three-dimensional K-space imaging data set that is undersampled in the layer direction is interpolated to a Cartesian grid; or, A two-dimensional non-uniform fast Fourier transform NUFF is performed on the imaging data of each layer in the three-dimensional K-space imaging data set with under-sampling in the layer direction to obtain an image, and then a fast Fourier transform is performed on the image to obtain two-dimensional K-space Cartesian data.
9. The device (70) according to claim 7, characterized in that The three-dimensional K space corresponding to the three-dimensional K space imaging data set undersampled in the layer direction obtained by the imaging data acquisition module (71) and the K space corresponding to the three-dimensional K space Cartesian calibration data set obtained by the calibration data acquisition module (72) have the same size of unit coordinates in each direction, and, The calibration data acquisition module (72) fills the acquired three-dimensional Cartesian calibration data into the three-dimensional K space, including: filling the acquired three-dimensional Cartesian calibration data into the central area of the three-dimensional K space.
10. A magnetic resonance imaging (MRI) system, characterized in that: The MRI system comprises the non-Cartesian trajectory stack reconstruction device (70) according to any one of claims 7 to 9.