Joint design method of k-space trajectory with variable acceleration factor
Through the JOINT method, the variable acceleration factor k-space trajectory is designed in multi-frame imaging, which solves the contrast difference caused by variable acceleration factor sampling, realizes the consistency of T2 filtering effect between all frames, and improves the quantitative effect of magnetic resonance parameters and images quality.
Patent Information
- Application Number
- PCT/CN2024/090880
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2023-05-16
- Filing Date
- 2024-04-30
- Publication Date
- 2024-11-21
AI Technical Summary
In multi-frame imaging technology, the additional contrast difference caused by variable acceleration factor sampling and the application of variable acceleration factor strategies are difficult, especially when using sequences read out by long echo chains, resulting in inconsistent T2 filtering effects between different frames. Interferes with the measurement of magnetic resonance parameters.
A joint design method (JOINT method) for variable acceleration factor k-space trajectory is proposed. By automatically designing the sampling trajectory of undersampled image frames of variable acceleration factor, the T2 filtering effect between different frames is consistent, and uniform or variable density Poisson drop is adopted. The sampling mask ensures that the echo chain trajectory of the high-acceleration factor frame comes from the echo chain trajectory queue of the low-acceleration factor frame, and achieves consistency of the T2 filtering effect of all frames.
The effective implementation of the variable acceleration factor strategy in multi-frame imaging is achieved, providing short scanning time and high image quality, eliminating the T2 contrast difference between different frames, and improving the quantitative effect of magnetic resonance parameters.
Smart Images

Figure CN2024090880_21112024_PF_FP_ABST
Abstract
Description
Joint design method for k-space trajectories with variable acceleration factors Technical Field
[0001] The present invention belongs to the field of magnetic resonance imaging and can design a k-space trajectory for a magnetic resonance parameter measurement sequence requiring acquisition with a variable acceleration factor. Background Art
[0002] Magnetic resonance imaging (MRI) technology can measure tissue parameters including nuclear spin density, longitudinal relaxation time (T1), transverse relaxation time (T2), apparent diffusion coefficient (ADC), cerebral blood flow rate, magnetization transfer rate, etc. These parameters can be derived from the nuclear spin density of endogenous compounds in the body (such as 1 H), or from exogenous substances or tracers. Direct quantification of MRI parameters can provide more valuable clinical information about various pathologies in neurological, musculoskeletal, hepatic, and myocardial imaging. This includes techniques such as relaxation time imaging, diffusion imaging, perfusion imaging, functional magnetic resonance imaging (fMRI), and chemical exchange saturation transfer (CEST) imaging, which generally require the acquisition of multi-frame image datasets with pulse sequence parameter modulation. Target parameter estimation can be achieved by fitting the signal evolution in the multi-frame image dataset to the corresponding mathematical model.
[0003] In order to accelerate parameter measurement in multi-frame imaging, scholars have proposed many methods. For example, the Keyhole method combines the acquired low-resolution image and the high-resolution reference image to reduce the data required for each acquisition, but it sacrifices the high-frequency information of the image. The SLAM method can directly generate regional CEST measurements from a region of interest of arbitrary shape, with a much higher signal-to-noise ratio efficiency than traditional single-voxel and multi-voxel methods, but it loses information about the heterogeneity of tissue within the region. Model-based methods, such as k-wROSA, can directly estimate the asymmetric Z spectrum of interest from complete or incomplete measurements, but their acceleration capabilities are limited. The CS-CEST method based on compressed sensing can use the sparsity of the image in the transform domain to speed up, but complete random undersampling has not yet been achieved in practical applications. These methods can improve the efficiency of data acquisition, reduce scanning time, and also help improve image quality.
[0004] Multi-frame imaging usually acquires multiple frames at the same location, ranging from a few to dozens. Full k-space sampling of all frames would be too time-consuming, while using the same high acceleration factor for each frame would result in strong artifacts and SNR loss. Therefore, it is a reasonable approach to adopt a variable acceleration factor according to the importance of different frames, which can maximize the balance between signal-to-noise ratio and scanning speed. For example, the variable acceleration sensitivity encoding method (vSENSE) has recently been proposed to quantify acceleration parameters. The vSENSE method adopts a variable acceleration factor strategy to achieve higher acceleration and higher image quality. As an improvement to the vSENSE method, the KIPI method implements automatic calibration, does not require the acquisition of additional sensitivity maps, does not require full sampling frames, and allows for more flexible variable acceleration strategies.
[0005] On the other hand, due to the advantages of long echo chain readout in terms of acquisition efficiency and signal-to-noise ratio, it is widely used in the field of multi-frame imaging, such as TSE, EPI, etc. However, due to the T2 or T2 * Attenuation: An exponential decay filter is applied to k-space along the echo encoding direction. For echo tracks read outward from the center, this effect is equivalent to a low-pass filter in k-space. When using an echo train readout sequence for multi-frame imaging, a variable acceleration factor strategy results in different downsampling masks for image frames with different acceleration factors. Traditional methods calculate the echo train track separately for the downsampling mask for each acceleration factor, which results in different T2 filtering effects in different frames, introduces additional T2 contrast, and interferes with magnetic resonance parameter measurement.
[0006] Therefore, using a variable acceleration factor strategy on multi-frame imaging sequences with long echo train readouts remains challenging, and high-quality acceleration has yet to be achieved. Therefore, in the field of multi-frame imaging, it is of great significance to propose a method for designing k-space trajectories for variable acceleration factor acquisition to achieve a consistent T2 filtering effect across all image frames.
[0007] Summary of the Invention
[0008] The present invention aims to address the problems of additional contrast differences caused by variable acceleration factor sampling in multi-frame imaging technology and the difficulty in implementing variable acceleration factor strategies. It also provides a joint design method of trajectory with variable acceleration factor k-space trajectory (hereinafter referred to as JOINT method), which can automatically design sampling trajectories for undersampled image frames with variable acceleration factors and eliminate contrast differences between different frames caused by sampling.
[0009] The specific technical solutions adopted in the present invention are as follows:
[0010] A joint design method for k-space trajectories with variable acceleration factors is provided, for designing k-space trajectories for a multi-frame imaging sequence when acquiring undersampled image frames, wherein the undersampled image frames include a first undersampled image frame and a second undersampled image frame having an acceleration factor not less than 2, and the acceleration factor of the first undersampled image frame is not greater than the acceleration factor of the second undersampled image frame, wherein the readout module of the multi-frame imaging sequence is an echo chain readout;
[0011] The design method steps are as follows:
[0012] S1: For the first undersampled image frame, a sampling mask is designed according to a pre-set acceleration factor and echo train length;
[0013] S2: Determine the number of first echo chains required to acquire the first undersampled image frame based on the number of acquisition points in the sampling mask and the echo chain length, where the number of first echo chains is the rounded-up value of the ratio of the number of acquisition points in the sampling mask to the echo chain length; then, sequentially design an acquisition trajectory for each echo chain, and form a first echo chain trajectory queue from the trajectories of all echo chains of the first undersampled image frame;
[0014] S3: determining the number of second echo chains to be acquired for the second undersampled image frame, wherein the number of the second echo chains is an upward-rounded value of the product of the ratio of the first undersampled image frame acceleration factor to the second undersampled image frame acceleration factor and the number of the first echo chains;
[0015] S4: selecting echo chain trajectories that meet the second echo chain number from the first echo chain trajectory queue in sequence, and forming a second echo chain trajectory queue of the second under-sampled image frame, thereby completing the echo chain trajectory design.
[0016] Preferably, the first undersampled image frame and the second undersampled image frame are both two-dimensional images; the acceleration factor of the first undersampled image frame is preferably 2, and the acceleration factor of the second undersampled image frame is preferably 2 to 4.
[0017] Preferably, the first undersampled image frame and the second undersampled image frame are both three-dimensional images; the acceleration factor of the first undersampled image frame is preferably 4, and the acceleration factor of the second undersampled image frame is preferably 4 to 12.
[0018] Preferably, in S1, the sampling mask is a uniform downsampling mask or a variable density Poisson downsampling mask.
[0019] Preferably, in S1, a full sampling area is reserved at the center of the sampling mask.
[0020] Preferably, the density of the variable-density Poisson downsampling mask decreases from the center to the edge.
[0021] Preferably, in S2, the acquisition trajectory of each echo chain is designed to be read outward from the center of the k-space.
[0022] Preferably, in S2, the acquisition trajectory of each designed echo chain should be dispersed in the entire k-space.
[0023] Preferably, in S2, the first under-sampled image frame and the second under-sampled image frame are both three-dimensional images, and the acquisition trajectory of each echo chain of the first under-sampled image frame is designed as follows:
[0024] S21. Convert the coordinates of all N acquisition points in the sampling mask of the first undersampled image frame into polar coordinates (r, θ) and then sort all acquisition points in ascending order according to the polar radius r. Then, group the sorted acquisition point sequence based on the echo train length L and the number of first echo trains ShotNum1. The grouping rule is divided into two forms based on the remainder M of N divided by L:
[0025] If M=0, the N acquisition points in the acquisition point sequence are directly divided into L groups in sequence with 1 acquisition point per ShotNum;
[0026] If M>0, first divide the first ShotNum1*M acquisition points in the acquisition point sequence into M groups with each ShotNum1 acquisition point as a group, and then divide the remaining acquisition points into LM groups with each ShotNum1-1 acquisition point as a group, for a total of L groups;
[0027] S22. Based on the grouping results and the golden angle φ, the trajectories of the ShotNum1 echo chains are designed in sequence, where:
[0028] For 1≤t≤(ShotNum1-1), when designing the trajectory of the t-th echo chain, the j-th echo is sequentially assigned to the acquisition point p in the j-th group, 1≤j≤L, where the acquisition point p in the j-th group is the one with the polar angle θ closest to all the unassigned acquisition points in the j-th group. The acquisition points of , thus the trajectory of the t-th echo chain is formed by the allocated L acquisition points;
[0029] When designing the trajectory of the last t=ShotNum1 echo chain, if M=0, the jth echo is assigned to the only remaining unassigned acquisition point in the jth group, 1≤j≤L; if M>0, the jth echo is assigned to the only remaining unassigned acquisition point in the jth group, 1≤j≤M.
[0030] Preferably, in S4, the second echo chain trajectory queue is composed of the front echo chain trajectory in the first echo chain trajectory queue that meets the second echo chain number.
[0031] Preferably, the second under-sampled image frame has multiple frames, different second under-sampled image frames are allowed to have different acceleration factors, and each second under-sampled image frame needs to design a second echo chain trajectory queue according to its own acceleration factor.
[0032] Compared with the prior art, the present invention has the following beneficial effects:
[0033] The existing technology designs separate acquisition trajectories for undersampled image frames with different acceleration factors. These differences in trajectories lead to inconsistent T2 filtering between frames, which introduces additional T2 contrast and interferes with MRI parameter measurement. The present invention aligns the echo train trajectories for high-acceleration factor frames entirely with the echo train trajectory queue for low-acceleration factor frames. This ensures consistent T2 filtering across all frames and eliminates additional T2 contrast, enabling optimal acquisition.
[0034] Compared to existing technologies, this invention allows for arbitrary acceleration factor variations while also integrating the design of image frame acquisition trajectories for varying acceleration factors, thereby improving the quantitative performance of magnetic resonance parameters. This invention also enables the practical application of variable acceleration factor undersampling strategies, resulting in shorter scan times and higher image quality.
[0035] This method is applicable to any MRI acquisition technique requiring echo train readout, including but not limited to TSE (FSE), EPI, and variable flip angle TSE sequences (also known as SPACE sequences). Multi-frame imaging is any MRI technique requiring multiple acquisitions of the same imaging subject, including but not limited to relaxation time (T1, T2) imaging, functional MRI (fMRI), magnetic resonance spectroscopy (MRS), chemical exchange saturation transfer (CEST) imaging, diffusion imaging, and perfusion imaging. BRIEF DESCRIPTION OF THE DRAWINGS
[0036] Figure 1 shows a portion of the echo chain acquisition trajectories for the first undersampled image frame in an embodiment. The first row shows the results of the conventional method, and the second row shows the results of the JOINT method. From left to right, each column shows the trajectories of the first, third, fifth, and seventh echo chains, respectively. The color shades represent the acquisition order within the echo chain, with the lightest color representing the k-space line acquired by the first echo in the echo chain, and the darkest color representing the k-space line acquired by the last echo in the echo chain.
[0037] Figure 2 shows the acquisition trajectories of all echo chains designed for the first and second undersampled image frames using the conventional and JOINT methods in this embodiment. For the first undersampled image frame, both the conventional and JOINT methods use an acceleration factor (AF) of 2×2 (in the phase encoding and slice selection encoding directions, respectively), with a 15×15 fully sampled area at the center (A, D). For the second undersampled image frame, the conventional method uses an AF of 4×2 and four echo chains (B), while the JOINT method uses an AF of 9 and four echo chains (E). (C) and (F) are the differences in the order of echoes at the same location in the echo chain in the trajectory of the second undersampled image frame and the trajectory of the first undersampled image frame, with (C) for the conventional method and (F) for the JOINT method.
[0038] Figure 3 shows a layer from the 3.5 ppm and -3.5 ppm 3D-CEST source images of a healthy volunteer in the example. The first column shows the 3.5 ppm source image, the second column shows the -3.5 ppm source image, and the third column shows the difference between the 3.5 ppm and -3.5 ppm source images. The first row shows the reference method, a conventional GRAPPA image with an AF = 2×2. The second row shows data acquired using the conventional trajectory, where the 3.5 ppm frame is the first undersampled image frame with an AF = 2×2 and the -3.5 ppm frame is the second undersampled image frame with an AF = 4×2, with separate trajectories designed. The third row shows data acquired using the JOINT trajectory, where the 3.5 ppm frame is the first undersampled image frame with an AF = 2×2 and the -3.5 ppm frame is the second undersampled image frame with an AF = 9, with a joint trajectory designed. Both the conventional and JOINT methods use the KIPI method for reconstruction.
[0039] Figure 4 shows APTw parameter images of healthy volunteers in the examples. The first row shows the reference method (AC), using conventional GRAPPA with an AF = 2 × 2. The second row shows data (DF) acquired using a trajectory designed by the conventional method. The 3.5 ppm frame is the first undersampled image frame with an AF = 2 × 2, and the remaining six frames are the second undersampled image frames with an AF = 4 × 2. Separate trajectories were designed. The third row shows data acquired using the JOINT method trajectory. The 3.5 ppm frame is the first undersampled image frame with an AF = 2 × 2, and the remaining six frames are the second undersampled image frames with an AF = 9. Both the conventional and JOINT methods use KIPI for reconstruction.
[0040] FIG5 is a flow chart of the JOINT trajectory design method of the present invention. DETAILED DESCRIPTION
[0041] The present invention will be further described and illustrated below with reference to the accompanying drawings and specific embodiments. The technical features of each embodiment of the present invention may be combined accordingly, provided that there is no conflict between them.
[0042] Multi-frame imaging requires multiple acquisitions of the same imaging object, resulting in a long acquisition time. A variable acceleration factor strategy uses a low acceleration factor for image frames with low signal-to-noise ratios and a high acceleration factor for image frames with high signal-to-noise ratios, achieving an optimal balance between signal-to-noise ratio and acquisition speed. On the other hand, acquisition sequences with long echo train readout modules, such as EPI, TSE (FSE), and variable flip angle TSE sequences, are widely used in the multi-frame imaging field. When a variable acceleration factor strategy is used in these sequences, the T2 attenuation effect varies significantly between frames due to differences in echo train trajectories between frames with different acceleration factors. This introduces additional T2-related contrast, interfering with the quantification of magnetic resonance parameters.
[0043] By jointly designing echo train trajectories for frames with different acceleration factors, the present invention ensures that the echo train trajectories for high-acceleration factor frames are completely derived from the echo train trajectories for low-acceleration factor frames, ensuring that the T2 filtering effect is the same across frames with different acceleration factors. The process flow of the joint design method for variable acceleration factor k-space trajectories provided by the present invention is shown in Figure 5, and its specific implementation process is described in detail below.
[0044] This design method jointly designs k-space acquisition trajectories for undersampled image frames with varying acceleration factors on an MRI machine. The undersampled image frames are divided into first and second undersampled image frames. The first undersampled image frame is undersampled at a lower acceleration factor, while the second undersampled image frame is undersampled at a higher acceleration factor (AF ≥ 2). The acceleration factor of the second undersampled image frame is no less than that of the first undersampled image frame. There is only one first undersampled image frame; the remaining frames are second undersampled image frames.
[0045] It should be noted that the multiple frames of image data acquired by the MRI machine can be either 2D or 3D data, and the variable acceleration factor varies depending on the actual situation. If the first undersampled image frame and the second undersampled image frame are 2D images, the acceleration factor of the first undersampled image frame is preferably 2, while the acceleration factor of the second undersampled image frame is preferably 2 to 4. If the first undersampled image frame and the second undersampled image frame are 3D images, the acceleration factor of the first undersampled image frame is preferably 4, while the acceleration factor of the second undersampled image frame is preferably 4 to 12.
[0046] It should be noted that the present invention does not involve the design of acquisition in the readout direction. If it is a 2D image, the present invention designs the acquisition trajectory in the phase encoding direction; if it is a 3D image, the present invention designs the acquisition trajectory in the phase encoding and layer selection encoding directions.
[0047] The joint design method of the variable acceleration factor k-space trajectory in the present invention comprises the following specific steps:
[0048] S1: For the first undersampled image frame, a sampling mask is designed according to a pre-set acceleration factor and echo train length.
[0049] The sampling mask design for the first undersampled frame in the present invention can employ either a uniform downsampling mask or a variable-density Poisson downsampling mask. The present invention recommends retaining a smaller fully sampled area at the center of the mask. Furthermore, if a variable-density Poisson downsampling mask is used, the present invention recommends that the density of the variable-density Poisson downsampling decrease from the center toward the edges.
[0050] However, it should be noted that in this step, although the uniform downsampling mask or the variable density Poisson downsampling mask is recommended, other downsampling masks may also be used depending on the situation, such as a downsampling mask with a partial Fourier or a random downsampling mask.
[0051] S2: According to the number of acquisition points in the sampling mask and the echo chain length, determine the number of echo chains required to acquire the first undersampled image frame (recorded as the first echo chain number ShotNum1). The first echo chain number is the rounded-up value of the ratio of the number of acquisition points in the sampling mask to the echo chain length. Where N represents the number of acquisition points in the sampling mask, L represents the echo chain length, Then design the acquisition trajectory of each echo chain in turn. i , the queue composed of the trajectories of all ShotNum1 echo chains of the first under-sampled image frame is recorded as the first echo chain trajectory queue Ω1 = {tra i},i=1...ShotNum1.
[0052] The acquisition trajectory of each echo chain designed in the present invention is preferably read outward from the center of k-space. The acquisition trajectory of each designed echo chain should be dispersed throughout k-space as much as possible. However, it should be noted that in this step, while it is recommended that the acquisition trajectory of each echo chain be read outward from the center of k-space, other methods, such as random arrangement, can also be adopted depending on the actual situation.
[0053] In addition, in an embodiment of the present invention, in order to ensure that the acquisition trajectory of each designed echo chain should be dispersed as much as possible in the entire k-space, a design method for the echo chain acquisition trajectory of the first undersampled image frame based on the golden angle is designed for the phase encoding ky and layer selection encoding kz direction trajectories of three-dimensional imaging.
[0054] For ease of description, define M as the remainder of N divided by L, the golden angle or Based on this definition, when both the first undersampled image frame and the second undersampled image frame are three-dimensional images, the acquisition trajectory design method of each echo chain of the first undersampled image frame is as follows:
[0055] First, for 3D imaging, the coordinate positions (ky, kz) of all k-space points to be acquired, i.e., all N acquisition points in the sampling mask of the first undersampled image frame, are converted to polar coordinates (r, θ) and sorted in ascending order according to the polar radius r. The sorted acquisition point sequence is then grouped based on the echo train length L and the number of the first echo train, ShotNum1. The grouping rules are based on the remainder M when N divides L, which can be divided into two forms:
[0056] In the case where N is divisible by L (M=0), the N acquisition points in the acquisition point sequence are directly divided into a group of one acquisition point per ShotNum, which is exactly divided into L groups.
[0057] If N is not divisible by L (M>0), first divide all N acquisition points into two queues, A and B. The first ShotNum1*M acquisition points are assigned to queue A, and the remaining acquisition points are assigned to queue B. In queue A, each ShotNum1 acquisition point is grouped together, resulting in M groups. In queue B, each ShotNum1-1 acquisition point is grouped together, resulting in L groups. Therefore, in this case, there are also L groups, with the first M groups each containing ShotNum1 acquisition point, and the last L groups each containing ShotNum1-1 acquisition points.
[0058] Then, based on the grouping results and the golden angle φ, the trajectories of the ShotNum1 echo chains are designed in sequence, where:
[0059] First, determine the trajectory of the first echo chain: assign the jth echo (1≤j≤L) to the acquisition point p in the jth group. The acquisition point p in the jth group is the acquisition point with the polar angle θ closest to (j-1)φ among all unassigned acquisition points in the jth group. Thus, the L assigned acquisition points form the trajectory of the first echo chain. After determining the trajectory of the first echo chain, mark the assigned acquisition points in each group, reducing the number of unassigned acquisition points in each group by one.
[0060] By analogy, determine the trajectory of the t-th echo chain (2≤t≤(ShotNum1-1)): assign the j-th echo (1≤j≤L) to the acquisition point p in the j-th group above, where the acquisition point p in the j-th group is the one with the polar angle θ closest to all the unassigned acquisition points in the j-th group. After determining the t-th echo chain trajectory, the allocated acquisition points in each group are marked, and the number of unallocated acquisition points in each group will be reduced by 1.
[0061] Then determine the trajectory of the last echo chain, that is, the t=ShotNum1th echo chain, which can be divided into two cases:
[0062] For the case where N is divisible by L (M=0), there is one acquisition point remaining in each of the L groups, so the j-th echo is assigned to the only remaining unassigned acquisition point in the j-th group, where 1≤j≤L;
[0063] For the case where N is not divisible by L (M>0), there is one unassigned acquisition point in each of the first M groups and zero acquisition points in each of the last LM groups. Therefore, the j-th echo is assigned to the only remaining acquisition point in the j-th group, where 1≤j≤M.
[0064] S3: For the second undersampled image frame, determine the number of echo chains to be collected (recorded as the second echo chain number ShotNum2). The second echo chain number is the product of the ratio of the first undersampled image frame acceleration factor to the second undersampled image frame acceleration factor and the first echo chain number, rounded up to the integer value. AF1 and AF2 represent a first undersampled image frame acceleration factor and a second undersampled image frame acceleration factor, respectively.
[0065] It should be noted that the echo train lengths of the first and second under-sampled image frames of the present invention are the same.
[0066] S4: ShotNum2 echo chain trajectories are sequentially selected from the first echo chain trajectory queue of the first under-sampled image frame to form a second echo chain trajectory queue Ω2 of the second under-sampled image frame. At this point, the echo chain trajectory design of all under-sampled image frames is completed.
[0067] In the present invention, when constructing the second echo chain trajectory queue Ω2, it is necessary to select ShotNum2 echo chain trajectories in sequence from the first echo chain trajectory queue Ω1 of the first under-sampled image frame echo chain, where the selected position in the queue can be adjusted according to actual needs. Preferably, the front ShotNum2 echo chain trajectories in the Ω1 queue are selected as the Ω2 queue, but other methods can also be selected according to actual conditions. For example, ShotNum2 echo chain trajectories can be randomly selected from the Ω1 queue as the Ω2 queue.
[0068] It should be noted that the second undersampled image frame in the present invention has multiple frames, and different second undersampled image frames can have different acceleration factors. For each second undersampled image frame, the second echo chain number ShotNum2 needs to be determined based on its own acceleration factor and the second echo chain trajectory queue needs to be designed. Specifically, if the acceleration factor of a second undersampled image frame is X and the number of echo chains to be collected is ShotNum2_X, then ShotNum2_X echo chain trajectories need to be selected from the trajectory queue of the echo chain of the first undersampled image frame as the Ω2 queue of this frame; when the acceleration factor of another second undersampled image frame is Y and the number of echo chains to be collected is ShotNum2_Y, then ShotNum2_Y echo chain trajectories need to be selected from the trajectory queue of the echo chain of the first undersampled image frame as the Ω2 queue of this frame.
[0069] Therefore, the above S1 to S4 constitute the joint design method of the variable acceleration factor k-space trajectory of the present invention.
[0070] The following will further illustrate the technical effects that can be achieved by the aforementioned joint design method of k-space trajectories with variable acceleration factors of the present invention through an embodiment, so that those skilled in the art can better understand the essence of the present invention.
[0071] Example
[0072] This embodiment applies the joint design method for k-space trajectories with variable acceleration factors (i.e., the JOINT method) shown in S1 to S4 to a specific example. The specific data source and specific implementation of each step of this method are described in detail below:
[0073] 1. Trajectory design
[0074] All processing and analysis were performed offline using MATLAB (MathWorks, Natick, MA) software written on a PC (3.2 GHz).
[0075] The acquisition matrix size of the phase encoding and slice selection encoding directions is preset to be 80×72, the echo train length is 142, the first under-sampling image frame has an AF of 4, and the second under-sampling image frame has only one acceleration factor, AF of 9.
[0076] The trajectory design of this embodiment is divided into the following steps:
[0077] 1. Design a sampling mask for the first undersampled image frame. Use a uniform downsampling mask of AF = 2 × 2 (in both the phase encoding and slice selection encoding directions). Preserve a fully sampled central 15 × 15 area. Use 80 and 72 as the major and minor axes of an ellipse. No data is collected outside the ellipse. The resulting downsampling mask has 1276 sampling points.
[0078] 2. Based on the downsampling mask, nine echo chains are required. Note that 1276 divided by 142 equals 8 with a remainder of 140, so a total of nine echo chains are required. However, the actual number of echoes collected by the ninth echo chain is 140. The trajectory of each echo chain is designed to read outward from the inside of k-space, and each echo chain is evenly distributed throughout k-space. The specific method is as follows:
[0079] Convert all k-space locations (ky, kz) to be acquired into polar coordinates (r, θ). Arrange them in ascending order by r, then group them into groups of 9, for a total of 140 groups. Finally, group the last 16 locations into groups of 8, for a total of 2 groups, for a total of 142 groups. The first group of 9 locations corresponds to the first echo of all 9 echo chains, and the Xth group of 9 locations corresponds to the Xth echo of all 9 echo chains (1 ≤ X ≤ 140).
[0080] The following describes how all echoes on each echo chain are assigned to corresponding positions to ensure that the trajectory of each echo chain is relatively uniform: Defining the golden angle
[0081] First determine the trajectory of the first echo chain: fill the jth (1≤j≤142) echo into the position in the jth group whose angle θ is closest to ((j-1)*φ). After determining the trajectory of the first echo chain, remove the determined positions, and reduce the remaining positions in each group by one.
[0082] For the trajectory of the second echo chain: fill the jth (1≤j≤142) echo into the position in the jth group whose angle θ is closest to ((j-1)*φ+φ). After determining the trajectory of the second echo chain, remove the determined positions, and continue to reduce one position in each group.
[0083] And so on, for the trajectory of the mth (3≤m≤8) echo chain: select from the unassigned positions, fill the jth (1≤j≤142) echo into the position in the jth group whose angle θ is closest to ((j-1)*φ+m*(m-1) / 2*φ), and then remove the already determined positions.
[0084] After determining the trajectories of the eight echo chains, there is one remaining position in each of groups 1 to 140, and no remaining positions in groups 141 and 142. Therefore, for the trajectory of the ninth echo chain, the jth echo is placed in the only remaining position in group j, where 1≤j≤140.
[0085] At this point, the design of all echo chain trajectories of the first under-sampled image frame is completed.
[0086] 3. The number of echo trains of the second under-sampled image frame is calculated to be 4 / 9*9=4. Therefore, the first to fourth echo train trajectories in the echo train trajectory queue of the first under-sampled image frame are selected as acquisition trajectories of the second under-sampled image frame.
[0087] Thus, the echo train trajectory design of the second under-sampled image frame is completed.
[0088] 2. Data Collection
[0089] Human experiments were conducted on a 3T Siemens MRI scanner (MAGNETOM Prisma) with a 64-channel head receive coil. The human studies were approved by the local institutional review board. The CEST imaging sequence was used, and the MRI parameter measured was the amide proton transfer effect amplitude.
[0090] For the 3D human brain experiment, the SPACE-CEST sequence was used to acquire data with the following operating parameters: TE = 17 ms, TR = 3 s, FOV = 232 × 232 × 209 mm. 3 , resolution = 2.9 × 2.9 × 2.9 mm 3 Acquisition matrix size = 80 × 80 × 72 (phase encoding, readout encoding, and slice-selective encoding, respectively), echo train length = 142, NSA = 1.3. Seven CEST saturation offset frames were acquired for amide proton transfer-weighted (APTw) imaging, including (S0), ±3, ±3.5, and ±4 ppm. The +3.5 ppm frame was selected as the first undersampled image frame (AF = 4), and the remaining six frames were selected as the second undersampled image frames (AF = 9). The SPACE sequence, a variable flip angle TSE sequence, was acquired using the trajectory designed using the JOINT method described above, with an acquisition time of 2.5 minutes.
[0091] For comparison, a conventional method for acquiring variable acceleration data was performed. The +3.5 ppm frame had an AF of 2 × 2, while the remaining frames had an AF of 4 × 2. A 15 × 15 ACS region was retained in the center of the +3.5 ppm frame. A conventional GRAPPA (2 × 2) was used as a reference. For all seven frames, AF was 2 × 2 with a 24 × 24 ACS region in the center. The acquisition time was 5 minutes.
[0092] To assess the magnitude of the amide proton transfer effect, B0-corrected data were acquired using a GRE sequence with the same field of view, orientation, and resolution as the CEST sequence, with a TR of 30 ms. The GRE sequence was run in dual-echo mode, with TEs of 4.92 ms and 9.84 ms, respectively.
[0093] 3. Image reconstruction
[0094] The acquired data of the trajectory designed using the JOINT method of the present invention and the acquired data using the traditional method are both reconstructed using the KIPI method: the first undersampled image frame is first reconstructed using the GRAPPA method, and then the coil sensitivity map and the correction factor map of the second undersampled image frame are derived from it; the second undersampled image frame is reconstructed using CG-SENSE, and the image generated by CG-SENSE is multiplied point by point with the correction factor map to obtain the final reconstructed image.
[0095] The APTw parametric image is calculated as follows. First, the source image is registered to the first undersampled frame at 3.5 ppm. Second, the phase difference between the two GRE images acquired at different TEs is calculated as a B0 map. Third, based on the calculated B0 map, corrected +3.5-ppm and -3.5-ppm signal values are generated for each voxel. Finally, the APTw parametric image is generated by subtracting the corrected 3.5-ppm and +3.5-ppm images.
[0096] 4. Results Analysis
[0097] FIG1 shows the trajectory design for the first undersampled image frame. The trajectory of each echo chain in the traditional method is relatively concentrated (AD), while the method adopted in this example ensures that the trajectory of each echo chain is relatively uniform and dispersed (EH).
[0098] Figure 2 shows the trajectory design results for the first and second undersampled image frames using the conventional method and the proposed JOINT method. For the first undersampled image frame, both the conventional and proposed methods use an AF = 2×2, with a 15×15 fully sampled area at the center (A, D). For the second undersampled image frame, the conventional method uses an AF = 4×2 and four echo chains (B), while the JOINT method uses an AF = 9 and four echo chains (E). The conventional method designs the trajectories for the two undersampled image frames separately, resulting in different echo numbers at the same acquisition location, with a difference of up to 20 (C). The proposed JOINT method uses a joint design, and the echo chain trajectory for the second undersampled image frame is completely derived from the first undersampled image frame. Therefore, the echo numbers at the same acquisition location in the first and second undersampled image frames are exactly the same (F).
[0099] Figure 3 shows a layer from 3.5 ppm and -3.5 ppm 3D-CEST source images of a healthy volunteer. The first row shows the reference method, conventional GRAPPA (AC) with an AF = 2×2. The second row shows data acquired using a trajectory designed by the conventional method (DF), where the 3.5 ppm frame is the first undersampled image frame with AF = 2×2 and the -3.5 ppm frame is the second undersampled image frame with AF = 4×2, with separate trajectories designed. The third row shows data acquired using a trajectory designed by the joint method of the present invention (GI), where the 3.5 ppm frame is the first undersampled image frame with AF = 2×2 and the -3.5 ppm frame is the second undersampled image frame with AF = 9, with a joint trajectory designed. Both the conventional and joint method data were reconstructed using KIPI. It can be seen that due to the different T2 filtering effects of the first and second undersampled frames in the traditional method, the difference between the two frames (F) shows obvious T2-related signals, which is extremely different from the reference (C); while the JOINT method of the present invention makes the T2 filtering effect consistent in each frame through joint trajectory design, and the difference between the two frames (I) is consistent with the reference method (C) without additional artifacts.
[0100] Figure 4 shows whole-brain APTw parametric images of healthy volunteers. The conventional method results in significant artifacts (DF) due to the failure to maintain consistent T2 filtering across all frames. The 3.5 ppm frame is the first undersampled image frame with AF = 2 × 2, and the other six frames are the second undersampled image frame with AF = 2 × 4, with separate trajectories designed. However, the JOINT method of the present invention ensures consistent T2 filtering across all frames, resulting in highly consistent APTw images (GI) and reference results (AC). The 3.5 ppm frame is the first undersampled image frame with AF = 2 × 2, and the other six frames are the second undersampled image frame with AF = 9, with joint trajectories designed.
[0101] The embodiment described above is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Persons skilled in the art may make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, any technical solution obtained by equivalent substitution or equivalent transformation falls within the scope of protection of the present invention.
Claims
1. A joint design method of k-space trajectory with variable acceleration factor, used for designing k-space trajectory of a multi-frame imaging sequence when acquiring undersampled image frames, characterized in that: The undersampled image frames include a first undersampled image frame and a second undersampled image frame with an acceleration factor not less than 2, and the acceleration factor of the first undersampled image frame is not higher than the acceleration factor of the second undersampled image frame, and the readout module of the multi-frame imaging sequence is an echo chain readout; The design method steps are as follows: S1: for the first under-sampled image frame, designing a sampling mask according to a pre-set acceleration factor and echo chain length; S2: determining the number of first echo chains to be collected for the first under-sampled image frame according to the number of acquisition points in the sampling mask and the length of the echo chain, wherein the number of first echo chains is the rounded-up value of the ratio of the number of acquisition points in the sampling mask to the length of the echo chain; then sequentially designing the acquisition trajectory of each echo chain, and forming a first echo chain trajectory queue by the trajectories of all echo chains of the first under-sampled image frame; S3: for the second under-sampled image frame, determining the number of second echo chains to be collected, wherein the number of second echo chains is an upward rounded value of the product of the ratio of the first under-sampled image frame acceleration factor to the second under-sampled image frame acceleration factor multiplied by the number of first echo chains; S4: echo chain trajectories satisfying the second echo chain number are selected in sequence from the first echo chain trajectory queue, and a second echo chain trajectory queue of a second under-sampled image frame is formed, thereby completing the echo chain trajectory design.
2. The joint design method of variable acceleration factor k-space trajectory according to claim 1, characterized in that: The first under-sampled image frame and the second under-sampled image frame are both two-dimensional images; the acceleration factor of the first under-sampled image frame is preferably 2, and the acceleration factor of the second under-sampled image frame is preferably 2 to 4.
3. The joint design method of variable acceleration factor k-space trajectory according to claim 1, characterized in that: The first under-sampled image frame and the second under-sampled image frame are both three-dimensional images; the acceleration factor of the first under-sampled image frame is preferably 4, and the acceleration factor of the second under-sampled image frame is preferably 4 to 12.
4. The joint design method of variable acceleration factor k-space trajectory according to claim 1, characterized in that: In S1, the sampling mask is a uniform downsampling mask or a variable density Poisson downsampling mask.
5. The joint design method of variable acceleration factor k-space trajectory according to claim 1, characterized in that: In S1, a full sampling area is reserved at the center of the sampling mask.
6. The joint design method of variable acceleration factor k-space trajectory according to claim 4, characterized in that: The density of the variable density Poisson downsampling mask decreases from the center to the edge.
7. The joint design method of variable acceleration factor k-space trajectory according to claim 1, characterized in that: In S2, the acquisition trajectory of each designed echo chain is read outward from the center of the k-space, and the acquisition trajectory of each designed echo chain should be dispersed in the entire k-space.
8. The joint design method of variable acceleration factor k-space trajectory according to claim 1, characterized in that: In S2, the first under-sampled image frame and the second under-sampled image frame are both three-dimensional images, and the acquisition trajectory design method of each echo chain of the first under-sampled image frame is as follows: S21, converting the coordinates of all N acquisition points in the sampling mask of the first under-sampled image frame into polar coordinate (r, θ) form, and then arranging all acquisition points in ascending order according to the polar radius r, and then grouping the sorted acquisition point sequence based on the echo chain length L and the first echo chain number ShotNum1, and the grouping rule is divided into two forms based on the remainder M of N divided by L: If M=0, the N acquisition points in the acquisition point sequence are directly divided into L groups in order with 1 acquisition point per ShotNum; If M>0, first divide the first ShotNum1*M acquisition points in the acquisition point sequence into M groups with each ShotNum1 acquisition point as a group, and then divide the remaining acquisition points into LM groups with each ShotNum1-1 acquisition point as a group, for a total of L groups; S22. Based on the grouping results and the golden angle φ, the trajectories of the ShotNum1 echo chains are designed in sequence, where: For 1≤t≤(ShotNum1-1), when designing the trajectory of the t-th echo chain, the j-th echo is sequentially assigned to the acquisition point p in the j-th group, 1≤j≤L, where the acquisition point p in the j-th group is the one with the polar angle θ closest to all the unassigned acquisition points in the j-th group. The L acquisition points thus obtained by the allocation constitute the trajectory of the t-th echo chain; When designing the trajectory of the last t=ShotNum1th echo chain, if M=0, the jth echo is assigned to the only remaining unassigned acquisition point in the jth group, 1≤j≤L; if M>0, the jth echo is assigned to the only remaining unassigned acquisition point in the jth group, 1≤j≤M.
9. The joint design method of variable acceleration factor k-space trajectory according to claim 1, characterized in that: In S4, the second echo chain trajectory queue is composed of the front echo chain trajectory in the first echo chain trajectory queue that meets the second echo chain number.
10. The joint design method of variable acceleration factor k-space trajectory according to claim 1, characterized in that: The second under-sampled image frame has multiple frames, and different second under-sampled image frames are allowed to have different acceleration factors. Each second under-sampled image frame needs to design a second echo chain trajectory queue according to its own acceleration factor.
Citation Information
Patent Citations
Magnetic-resonance image acquisition method and magnetic resonance system
CN106680748A
Specification of an echo train length in magnetic resonance
CN107209239A
CEST (Chemical Exchange Saturation Transfer) image reconstruction method and device based on variable acceleration sensitivity encoding
CN109839607A
Imaging method and device combining k space and image space reconstruction
CN113009398A
Reconstruction of mr image data
CN113066141A
