Fast multi-contrast magnetic resonance diffusion imaging and reconstruction method and system

Abstract

The present invention provides a fast multi-contrast magnetic resonance diffusion imaging and reconstruction method and system, including: using a keyhole echo plane sequence to undersample k-space data of multiple diffusion-weighted images through multiple receiving coils; constructing an objective function including a multi-coil image fidelity term and an image regularization term based on the coil-space-contrast domain tensor low-rank (Low-RankTensorConstraint); minimizing the objective function through an optimization algorithm, reconstructing the diffusion images of all receiving coils and obtaining relevant quantitative parameter images according to the specific way of contrast change. The present invention implements undersampling of k-space through a keyhole-EPI sequence, improves the resolution of multi-contrast diffusion imaging, shortens the echo time, thereby improving the deformation artifacts of the image and reducing the imaging time; avoids the step of solving the coil sensitivity map, reduces the image artifacts caused by variable-density EPI imaging, and improves the robustness and accuracy of the reconstruction result.

Description

Technical Field

[0001] The invention relates to the field of magnetic resonance diffusion imaging, and in particular to a fast multi-contrast magnetic resonance diffusion imaging and reconstruction method and system. Background Art

[0002] Changes in the diffusion capacity of water molecules in human tissues can reflect the composition of physiological tissues and pathological conditions. Diffusion MRI (dMRI) is an imaging technique that can non-invasively examine the diffusion capacity of living tissues. It is widely used in quantitative studies of microvascular perfusion and diffusion effects in vivo, and plays an important role in the early diagnosis of diseases such as acute stroke, liver disease, cancer, and neuroscience research. The calculation of quantitative images of diffusion effects depends on multiple diffusion-weighted images (DWI) with different b values, directions, diffusion times, and other parameters. Since multiple images with different contrasts need to be obtained, diffusion MRI requires a long scanning time. At the same time, due to the addition of diffusion gradients, the signal attenuation rate is fast, the acquisition resolution is limited, and the image distortion is large.

[0003] There are two main types of existing MRI diffusion imaging methods.

[0004] The first type is the single-shot echo planar sequence. This method completes the data acquisition of the entire K space after one excitation, has a short scanning time, and is robust to motion. However, due to the long echo chain, it leads to low resolution, low signal-to-noise ratio, and large image distortion.

[0005] The second type is the multiple-excitation echo planar sequence. This type of method uses a multiple-excitation strategy to collect part of the K-space data each time, and the data obtained by multiple excitations are merged into a complete K-space, which can achieve higher spatial resolution and image quality. However, the diffusion encoding gradient causes phase changes between different excitations, and there may be motion between multiple excitations; directly merging these data will cause image artifacts. In addition, for diffusion quantitative imaging applications such as ADC, DTI, DKI, and IVIM, multiple images with multiple excitations need to be obtained to achieve high resolution, which increases the scanning time.

[0006] The patent document (application number 201210388038.1) discloses a fast diffusion magnetic resonance imaging and reconstruction method, comprising the following steps: (S1) acquiring signals of a target in N diffusion weighted directions by means of a multi-channel coil; (S2) merging the obtained complementary K-space data to obtain full-sampled K-space data K C (S3) based on the images of different diffusion weighted directions corresponding to the K-space data and the fully sampled K-space data KC The corresponding image is preliminarily reconstructed to obtain a preliminary image and (S4) regularized reconstruction is performed based on the preliminary image, and iterated until convergence to obtain the desired final image (I 1 ,…,I N ). The invention uses ordinary rather than keyhole echo planar imaging sequences for data acquisition, and needs to merge the acquired diffusion images in K space, and also needs to solve the coil sensitivity parameters. It is difficult to avoid the reconstruction error caused by the merging, and it is also impossible to avoid the inaccuracy of the reconstruction caused by the inaccurate estimation of the coil sensitivity parameters.

[0007] Patent document CN103675737A (application number: CN201310659202.2) discloses a diffusion magnetic resonance imaging and reconstruction method. The method includes: S1, using multiple channel coils, adopting multiple excitation diffusion imaging to collect signals from the target to obtain k-space data; S2, calculating the coil sensitivity map and performing iterative initialization; S3, based on the collected k-space data, the calculated coil sensitivity map and the initialization parameters, based on the POCS algorithm, iteratively reconstructing the required diffusion image. The invention adopts a multiple excitation strategy, and it is necessary to first merge the collected multi-coil diffusion images in K space, and it is not possible to reduce the imaging time in multi-contrast diffusion imaging by changing the contrast. Summary of the invention

[0008] In view of the defects in the prior art, the present invention provides a fast multi-contrast magnetic resonance diffusion imaging and reconstruction method and system, which can effectively improve the resolution, reduce deformation artifacts, and improve the scanning efficiency.

[0009] A fast multi-contrast magnetic resonance diffusion imaging and reconstruction method provided by the present invention comprises:

[0010] Step S1: Undersampling k-space data of multiple variable contrast diffusion weighted images using a keyhole echo planar sequence through multiple receiving coils;

[0011] Step S2: constructing an objective function including a multi-coil image fidelity term and an image regularization term based on the low rank of coil, space and contrast domain tensors according to the acquisition results;

[0012] Step S3: Minimize the objective function through an optimization algorithm, reconstruct the diffusion images of all receiving coils, and obtain relevant quantitative parameter images according to the specific way of contrast change.

[0013] Preferably, in step S1:

[0014] The number of receiving coils is the number of coil channels greater than or equal to 1;

[0015] The variable contrast includes: the contrast parameters between different TRs of the echo planar sequence can be arbitrarily changed, including but not limited to arbitrary changes in b value, diffusion encoding direction, and diffusion time;

[0016] For a specific contrast parameter, the arbitrary change means that the contrast parameter between different TRs remains unchanged or is different;

[0017] The keyhole-EPI sequence uses the following variable density k-space acquisition:

[0018] In the central k-space region, the undersampling rate R C Intensive collection;

[0019] In the surrounding k-space region, the sampling rate R P Uniformly shift undersampling so that adjacent TRs acquire cyclically shifted phase encoding lines;

[0020] The variable density k-space acquisition of the keyhole-EPI sequence is implemented in the following manner:

[0021] Determine the gradient momentum across a single phase encoding line; and based on the calculated results, calculate other required blip gradient momentums across different phase encoding lines; arrange the required blip gradients of corresponding momentum in time sequence according to the preset sampling trajectory; based on the above calculation results, use bipolar EPI gradients to collect k-space data.

[0022] Preferably, in step S2:

[0023] The image to be solved is a variable contrast diffusion-weighted image under multiple coils, rather than a diffusion-weighted image after merging multiple coils;

[0024] The objective function does not include coil sensitivity, and solving the objective function does not require solving coil sensitivity first or performing self-calibration of parallel imaging;

[0025] The image fidelity term based on the low rank of coil, space and contrast domain tensors uses the low rank of multi-coil variable contrast diffusion weighted images to constrain the range of variation of the image to be solved, including constraints based on the low rank of tensors in the global space and constraints based on the low rank of tensors in the local space.

[0026] The objective function is:

[0027]

[0028] Where y represents the multi-dimensional K-space data obtained after preprocessing, D represents the undersampling operator, F represents the Fourier transform operator, x represents the multi-coil multi-contrast image to be reconstructed, α is the regularization coefficient, and B ix is the coil, space and contrast three-dimensional tensor centered on the i-th pixel, R is the rank operator of the three-dimensional tensor based on Tucker decomposition or PARAFAC decomposition, N v is a preset value. If a global low-rank constraint is used, then N v =1, if the local low-rank constraint is used, then N v Can be equal to the number of pixels.

[0029] Preferably, in step S3:

[0030] Its objective function can be solved by any optimization algorithm, including gradient descent, conjugate gradient method, quasi-Newton method, alternating multiplier method, proximal gradient method, projected gradient method, and coordinate descent method, to reconstruct the diffusion images of all receiving coils and obtain relevant quantitative parameter images according to the specific application scenario.

[0031] Preferably, the imaging method can be used in diffusion imaging application scenarios including: diffusion weighted imaging under multiple excitations, apparent diffusion coefficient quantitative imaging, diffusion tensor imaging, diffusion kurtosis imaging, and intravoxel incoherent motion imaging.

[0032] A fast multi-contrast magnetic resonance diffusion imaging and reconstruction system provided by the present invention comprises:

[0033] Module M1: Undersample k-space data of multiple variable contrast diffusion-weighted images using keyhole echo planar sequence through multiple receiving coils;

[0034] Module M2: Based on the acquisition results, construct an objective function including a multi-coil image fidelity term and an image regularization term based on the low rank of coil, space and contrast domain tensors;

[0035] Module M3: Minimize the objective function through the optimization algorithm, reconstruct the diffusion images of all receiving coils and obtain relevant quantitative parameter images according to the specific way of contrast change.

[0036] Preferably, in the module M1:

[0037] The number of receiving coils is the number of coil channels greater than or equal to 1;

[0038] The variable contrast includes: the contrast parameters between different TRs of the echo planar sequence can be arbitrarily changed, including but not limited to arbitrary changes in b value, diffusion encoding direction, and diffusion time;

[0039] For a specific contrast parameter, the arbitrary change means that the contrast parameter between different TRs remains unchanged or is different;

[0040] The keyhole-EPI sequence uses the following variable density k-space acquisition:

[0041] In the central k-space region, the undersampling rate R C Intensive collection;

[0042] In the surrounding k-space region, the sampling rate R P Uniformly shift undersampling so that adjacent TRs acquire cyclically shifted phase encoding lines;

[0043] The variable density k-space acquisition of the keyhole-EPI sequence is implemented in the following manner:

[0044] Determine the gradient momentum across a single phase encoding line; and based on the calculated results, calculate other required blip gradient momentums across different phase encoding lines; arrange the required blip gradients of corresponding momentum in time sequence according to the preset sampling trajectory; based on the above calculation results, use bipolar EPI gradients to collect k-space data.

[0045] Preferably, in the module M2:

[0046] The image to be solved is a variable contrast diffusion-weighted image under multiple coils, rather than a diffusion-weighted image after merging multiple coils;

[0047] The objective function does not include coil sensitivity, and solving the objective function does not require solving coil sensitivity first or performing self-calibration of parallel imaging;

[0048] The image fidelity term based on the low rank of coil, space and contrast domain tensors utilizes the low rank of multi-coil variable contrast diffusion weighted image tensors to constrain the range of change of the image to be solved, including constraints based on the low rank of tensors in the global space and constraints based on the low rank of tensors in the local space.

[0049] The objective function is:

[0050]

[0051] Where y represents the multi-dimensional K-space data obtained after preprocessing, D represents the undersampling operator, F represents the Fourier transform operator, x represents the multi-coil multi-contrast image to be reconstructed, α is the regularization coefficient, and B i x is the coil, space and contrast three-dimensional tensor centered on the i-th pixel, R is the rank operator of the three-dimensional tensor based on Tucker decomposition or PARAFAC decomposition, N v is a preset value. If a global low-rank constraint is used, then N v =1, if the local low-rank constraint is used, then N v Can be equal to the number of pixels.

[0052] Preferably, in the module M3:

[0053] Its objective function can be solved by any optimization algorithm, including gradient descent, conjugate gradient method, quasi-Newton method, alternating multiplier method, proximal gradient method, projected gradient method, and coordinate descent method, to reconstruct the diffusion images of all receiving coils and obtain relevant quantitative parameter images according to the specific application scenario.

[0054] Preferably, the imaging method can be used in diffusion imaging application scenarios including: diffusion weighted imaging under multiple excitations, apparent diffusion coefficient quantitative imaging, diffusion tensor imaging, diffusion kurtosis imaging, and intravoxel incoherent motion imaging.

[0055] Compared with the prior art, the present invention has the following beneficial effects:

[0056] 1. Compared with single-shot echo planar imaging technology, the present invention reduces the acquisition time of the EPI echo chain through the keyhole echo planar imaging method, and can achieve higher resolution, higher signal-to-noise ratio, and smaller image distortion;

[0057] 2. Compared with multiple-excitation planar echo imaging technologies, the present invention introduces contrast changes during the multiple excitation process to obtain multiple diffusion-weighted images with different contrasts and corresponding quantitative images in a shorter imaging time. By using low-rank tensor constraints in the coil-space-contrast domain, the step of solving the coil sensitivity map is avoided, the image artifacts caused by variable-density EPI imaging are reduced, and the robustness and accuracy of the reconstruction results are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:

[0059] Figure 1 A flowchart of a diffusion magnetic resonance imaging and reconstruction method according to an embodiment of the present application;

[0060] Figure 2 A schematic diagram of a keyhole echo planar sequence according to an embodiment of the present application;

[0061] Figure 3 A schematic diagram of a keyhole echo planar sequence trajectory of a specific embodiment of the present application;

[0062] Figure 4 A flowchart of iterative reconstruction in a diffusion magnetic resonance imaging and reconstruction method according to a specific embodiment of the present application;

[0063] Figure 5 A schematic diagram comparing a specific embodiment of the present application with the retrospective experimental reconstruction results of the prior art;

[0064] Figure 6 It is a schematic diagram comparing the reconstruction results of a prospective experiment of a specific embodiment of the present application with those of the prior art. DETAILED DESCRIPTION

[0065] The present invention is described in detail below in conjunction with specific embodiments. The following embodiments will help those skilled in the art to further understand the present invention, but are not intended to limit the present invention in any form. It should be noted that, for those of ordinary skill in the art, several changes and improvements can also be made without departing from the concept of the present invention. These all belong to the protection scope of the present invention.

[0066] Embodiment 1:

[0067] According to a fast multi-contrast magnetic resonance diffusion imaging and reconstruction method provided by the present invention, Figure 1-Figure 6 ,include:

[0068] Step S1: Undersampling k-space data of multiple variable contrast diffusion weighted images using a keyhole echo planar sequence through multiple receiving coils;

[0069] Specifically, in step S1:

[0070] The number of receiving coils is the number of coil channels greater than or equal to 1;

[0071] The variable contrast includes: the contrast parameters between different TRs of the echo planar sequence can be arbitrarily changed, including but not limited to arbitrary changes in b value, diffusion encoding direction, and diffusion time;

[0072] For a specific contrast parameter, the arbitrary change means that the contrast parameter between different TRs remains unchanged or is different;

[0073] The keyhole-EPI sequence uses the following variable density k-space acquisition:

[0074] In the central k-space region, the undersampling rate R C Intensive collection;

[0075] In the surrounding k-space region, the sampling rate R P Uniformly shift undersampling so that adjacent TRs acquire cyclically shifted phase encoding lines;

[0076] The variable density k-space acquisition of the keyhole-EPI sequence is implemented in the following manner:

[0077] Determine the gradient momentum across a single phase encoding line; and based on the calculated results, calculate other required blip gradient momentums across different phase encoding lines; arrange the required blip gradients of corresponding momentum in time sequence according to the preset sampling trajectory; based on the above calculation results, use bipolar EPI gradients to collect k-space data.

[0078] Step S2: constructing an objective function including a multi-coil image fidelity term and an image regularization term based on the low rank of coil, space and contrast domain tensors according to the acquisition results;

[0079] Specifically, in step S2:

[0080] The image to be solved is a variable contrast diffusion-weighted image under multiple coils, rather than a diffusion-weighted image after merging multiple coils;

[0081] The objective function does not include coil sensitivity, and solving the objective function does not require solving coil sensitivity first or performing self-calibration of parallel imaging;

[0082] The image fidelity term based on the low rank of coil, space and contrast domain tensors uses the low rank of multi-coil variable contrast diffusion weighted images to constrain the range of variation of the image to be solved, including constraints based on the low rank of tensors in the global space and constraints based on the low rank of tensors in the local space.

[0083] The objective function is:

[0084]

[0085] Where y represents the multi-dimensional K-space data obtained after preprocessing, D represents the undersampling operator, F represents the Fourier transform operator, x represents the multi-coil multi-contrast image to be reconstructed, α is the regularization coefficient, and B i x is the coil, space and contrast three-dimensional tensor centered on the i-th pixel, R is the rank operator of the three-dimensional tensor based on Tucker decomposition or PARAFAC decomposition, N v is a preset value. If a global low-rank constraint is used, then N v =1, if the local low-rank constraint is used, then N v Can be equal to the number of pixels.

[0086] Step S3: Minimize the objective function through an optimization algorithm, reconstruct the diffusion images of all receiving coils, and obtain relevant quantitative parameter images according to the specific way of contrast change.

[0087] Specifically, in step S3:

[0088] Its objective function can be solved by any optimization algorithm, including gradient descent, conjugate gradient method, quasi-Newton method, alternating multiplier method, proximal gradient method, projected gradient method, and coordinate descent method, to reconstruct the diffusion images of all receiving coils and obtain relevant quantitative parameter images according to the specific application scenario.

[0089] Specifically, the imaging method can be used in diffusion imaging application scenarios including: diffusion weighted imaging under multiple excitations, apparent diffusion coefficient quantitative imaging, diffusion tensor imaging, diffusion kurtosis imaging, and intravoxel incoherent motion imaging.

[0090] A fast multi-contrast magnetic resonance diffusion imaging and reconstruction system provided by the present invention comprises:

[0091] Module M1: Undersample k-space data of multiple variable contrast diffusion-weighted images using keyhole echo planar sequence through multiple receiving coils;

[0092] Specifically, in the module M1:

[0093] The number of receiving coils is the number of coil channels greater than or equal to 1;

[0094] The variable contrast includes: the contrast parameters between different TRs of the echo planar sequence can be arbitrarily changed, including but not limited to arbitrary changes in b value, diffusion encoding direction, and diffusion time;

[0095] For a specific contrast parameter, the arbitrary change means that the contrast parameter between different TRs remains unchanged or is different;

[0096] The keyhole-EPI sequence uses the following variable density k-space acquisition:

[0097] In the central k-space region, the undersampling rate R C Intensive collection;

[0098] In the surrounding k-space region, the sampling rate R P Uniformly shift undersampling so that adjacent TRs acquire cyclically shifted phase encoding lines;

[0099] The variable density k-space acquisition of the keyhole-EPI sequence is implemented in the following manner:

[0100] Determine the gradient momentum across a single phase encoding line; and based on the calculated results, calculate other required blip gradient momentums across different phase encoding lines; arrange the required blip gradients of corresponding momentum in time sequence according to the preset sampling trajectory; based on the above calculation results, use bipolar EPI gradients to collect k-space data.

[0101] Module M2: Based on the acquisition results, construct an objective function including a multi-coil image fidelity term and an image regularization term based on the low rank of coil, space and contrast domain tensors;

[0102] Specifically, in the module M2:

[0103] The image to be solved is a variable contrast diffusion-weighted image under multiple coils, rather than a diffusion-weighted image after merging multiple coils;

[0104] The objective function does not include coil sensitivity, and solving the objective function does not require solving coil sensitivity first or performing self-calibration of parallel imaging;

[0105] The image fidelity term based on the low rank of coil, space and contrast domain tensors uses the low rank of multi-coil variable contrast diffusion weighted images to constrain the range of variation of the image to be solved, including constraints based on the low rank of tensors in the global space and constraints based on the low rank of tensors in the local space.

[0106] The objective function is:

[0107]

[0108] Where y represents the multi-dimensional K-space data obtained after preprocessing, D represents the undersampling operator, F represents the Fourier transform operator, x represents the multi-coil multi-contrast image to be reconstructed, α is the regularization coefficient, and B i x is the coil, space and contrast three-dimensional tensor centered on the i-th pixel, R is the rank operator of the three-dimensional tensor based on Tucker decomposition or PARAFAC decomposition, N v is a preset value. If a global low-rank constraint is used, then N v =1, if the local low-rank constraint is used, then N v Can be equal to the number of pixels.

[0109] Module M3: Minimize the objective function through the optimization algorithm, reconstruct the diffusion images of all receiving coils and obtain relevant quantitative parameter images according to the specific way of contrast change.

[0110] Specifically, in the module M3:

[0111] Its objective function can be solved by any optimization algorithm, including gradient descent, conjugate gradient method, quasi-Newton method, alternating multiplier method, proximal gradient method, projected gradient method, and coordinate descent method, to reconstruct the diffusion images of all receiving coils and obtain relevant quantitative parameter images according to the specific application scenario.

[0112] Specifically, the imaging method can be used in diffusion imaging application scenarios including: diffusion weighted imaging under multiple excitations, apparent diffusion coefficient quantitative imaging, diffusion tensor imaging, diffusion kurtosis imaging, and intravoxel incoherent motion imaging.

[0113] Embodiment 2:

[0114] Embodiment 2 is a preferred example of Embodiment 1, and is used to illustrate the present invention in more detail.

[0115] A fast multi-contrast magnetic resonance diffusion imaging and reconstruction method provided by the present invention comprises:

[0116] Step S1: Undersampling k-space data of multiple variable contrast diffusion weighted images using a keyhole echo planar (keyhole-EPI) sequence through multiple receiving coils;

[0117] Step S2: constructing an objective function including a multi-coil image fidelity term and an image regularization term based on the low rank of the coil-space-contrast domain tensor;

[0118] Step S3: Minimize the objective function through an optimization algorithm, reconstruct the diffusion images of all receiving coils, and obtain relevant quantitative parameter images according to the specific way of contrast change.

[0119] Preferably, in step S1:

[0120] The number of receiving coils is greater than or equal to 1;

[0121] The variable contrast includes: the contrast parameters between different TRs of the echo planar sequence can be arbitrarily changed, including but not limited to arbitrary changes in b value, diffusion encoding direction, and diffusion time;

[0122] For a specific contrast parameter, the arbitrary change means that the contrast parameter between different TRs remains unchanged or is different.

[0123] Preferably, the keyhole-EPI sequence uses the following variable density k-space acquisition:

[0124] In the central k-space region, the undersampling rate R C Intensive collection;

[0125] In the surrounding k-space region, the sampling rate R P Uniformly shift undersampling so that adjacent TRs acquire cyclically shifted phase encoding lines;

[0126] The variable density k-space acquisition of the keyhole-EPI sequence is implemented in the following manner:

[0127] Determine the gradient momentum across a single phase encoding line; and based on the calculated results, calculate other required blip gradient momentums across different phase encoding lines; arrange the required blip gradients of corresponding momentum in time sequence according to the preset sampling trajectory; based on the above calculation results, use bipolar EPI gradients to collect k-space data.

[0128] Preferably, in step S2:

[0129] The image to be solved is a variable contrast diffusion-weighted image under multiple coils, rather than a diffusion-weighted image after merging multiple coils.

[0130] The objective function does not include coil sensitivity, and solving the objective function does not require solving the coil sensitivity first or performing self-calibration of parallel imaging.

[0131] The image fidelity term based on the low rank of the coil-space-contrast domain tensor utilizes the low rank of the multi-coil variable contrast diffusion weighted image tensor to constrain the range of change of the image to be solved, including constraints based on the low rank of the tensor in the global space and constraints based on the low rank of the tensor in the local space.

[0132] Specifically, according to one embodiment of the present invention, when using the local tensor low rank as the low rank constraint, the constructed objective function can be described as:

[0133]

[0134] Where y represents the multi-dimensional K-space data obtained after preprocessing, D represents the undersampling operator, F represents the Fourier transform operator, x represents the multi-coil multi-contrast image to be reconstructed, α is the regularization coefficient, and B i x is the coil-space-contrast three-dimensional tensor centered at the i-th pixel, R is the rank operator of the three-dimensional tensor based on Tucker decomposition or PARAFAC decomposition, N v is a pre-set value. If a global low-rank constraint is used, N v =1, if the local low-rank constraint is used, then N v Can be equal to the number of pixels.

[0135] According to the above-mentioned fast multi-contrast magnetic resonance diffusion imaging and reconstruction method of the present invention, it is possible to avoid solving coil sensitivity map parameters while avoiding directly merging K spaces with different phases, thereby improving the quality of reconstructed images and allowing higher resolution and higher acquisition efficiency.

[0136] According to one embodiment of the present invention, the objective function can be rewritten as the following formula by constructing auxiliary variables:

[0137]

[0138] B i x=M i

[0139] M i =g i × l A i × 2 B i × 3 C i , j = 1, 2, ..., N v

[0140] Where y represents the multi-dimensional K-space data obtained after preprocessing, D represents the undersampling operator, F represents the Fourier transform operator, and x represents the multi-coil multi-contrast image to be reconstructed. α is the regularization coefficient, B i x is the coil-space-contrast three-dimensional tensor centered at the i-th pixel, R is the rank operator of the three-dimensional tensor based on Tucker decomposition or PARAFAC decomposition, N v is a pre-set value. If a global low-rank constraint is used, N v =1, if the local low-rank constraint is used, then N v M can be equal to the number of pixels. i is an auxiliary variable of type Tensor, obtained through PARAFAC or Tucker decomposition, M i =g i × 1 A i × 2 B i × 3 C i , i=1,2,…,N v

[0141] According to one embodiment of the present invention, the objective function can be rewritten as an augmented Lagrangian form, such as the following formula:

[0142]

[0143] Where y represents the multi-dimensional K-space data obtained after preprocessing, D represents the undersampling operator, F represents the Fourier transform operator, and x represents the multi-coil multi-contrast image to be reconstructed. α is the regularization coefficient, B i x is the coil-space-contrast three-dimensional tensor centered at the i-th pixel, R is the rank operator of the three-dimensional tensor based on Tucker decomposition or PARAFAC decomposition, N v is a pre-set value. If a global low-rank constraint is used, N v =1, if the local low-rank constraint is used, then N v M can be equal to the number of pixels. Re is a real number operator.i is an auxiliary variable of type Tensor, obtained through PARAFAC or Tucker decomposition, M i =g i × 1 A i × 2 B i × 3 C i , i=1,2,…,N v , μ is the augmented Lagrangian coefficient, λ i is the Lagrange multiplier.

[0144] Preferably, in step S3:

[0145] Its objective function can be solved by any optimization algorithm, including gradient descent, conjugate gradients, quasi-Newton, alternating direction method of multipliers, proximal gradient, projected gradient, coordinate descent and other algorithms, to reconstruct the diffusion images of all receiving coils and obtain relevant quantitative parameter images according to specific application scenarios.

[0146] This imaging method can be used for various diffusion imaging related application scenarios, such as diffusion weighted imaging under multiple excitations, Apparent Diffusion Coefficient (ADC) quantitative imaging, Diffusion Tensor Imaging, Diffusion Kurtosis Imaging, and Intravoxel Incoherent Motion Imaging (IVIM).

[0147] The obtaining of relevant quantitative parameter images according to the application scenario means that when it is necessary to obtain a quantitative parameter image, the parameter image and the diffusion image are the final results; when it is not necessary to obtain a quantitative parameter image, the reconstructed diffusion image is the final result.

[0148] Specifically, in one embodiment of the present invention, the objective function is iteratively minimized by the alternating multiplier method, and the calculation steps of the alternating multiplier method are as follows:

[0149] S31 Initialization: Before formal iterative reconstruction, it is necessary to initialize the diffusion image to be solved, that is, determine the initial value of the image. In the embodiment of the present invention, the initial value can be set to 0. For the convenience of description, it is described in the formula as: x selects the initial point x0 .

[0150] S32 solves sub-problems separately: In ADMM, it is necessary to solve the sub-problems that only have to be solved diffusion image variables, auxiliary variables and Lagrange multiplier variables. Steps S321-323 respectively describe the solution details of the corresponding sub-problems mentioned above.

[0151] Step S321: Solve for x k , there is an analytical solution

[0152]

[0153] where H stands for conjugate transpose.

[0154] Step S322: Solving The sub-goals are as follows

[0155]

[0156] Preferably, first perform a PARAFAC or Tucker decomposition of the tensor

[0157]

[0158] Then use the singular value soft thresholding method:

[0159]

[0160]

[0161]

[0162] Step S323: Solving

[0163]

[0164] After completing the above steps, S33 automatically updates the coil image of each channel and performs a loop iteration. The iteration stops when one of the following conditions is met: (1) the number of iterations is greater than 50; (2) the infinite norm of the gradient is less than 1*10 -10 ; (3) Step size is less than 1*10 -12 . Determine whether convergence or the maximum number of iterations is reached. If the iteration stop condition is not met, re-enter step S32 to continue iteration. If the iteration stop condition is met, the iteration is terminated and a magnetic resonance diffusion weighted image is obtained.

[0165] S34 solves diffusion quantitative imaging such as ADC, DTI, DKI, and IVIM according to the type of diffusion weighted parameters collected and actual needs. If quantitative images do not need to be solved, the magnetic resonance diffusion weighted image is the final result.

[0166] Embodiment 3:

[0167] Embodiment 3 is a preferred example of Embodiment 1, and is used to illustrate the present invention in more detail.

[0168] The purpose of this application is to propose a fast multi-contrast magnetic resonance diffusion imaging and reconstruction method.

[0169] The overall steps of the fast multi-contrast magnetic resonance diffusion imaging and reconstruction method provided by the present application are as follows: Figure 1 As shown, it includes: step S1: using multiple receiving coils, using a keyhole echo plane (keyhole-EPI) sequence to undersample the k-space data of multiple variable contrast diffusion weighted images; step S2: constructing an objective function including a multi-coil image fidelity term and an image regularization term based on the low rank of the coil-space-contrast domain tensor; step S3: minimizing the objective function through an optimization algorithm, reconstructing the diffusion images of all receiving coils and obtaining relevant quantitative parameter images according to the specific way of contrast change.

[0170] In the following description of the fast high-resolution diffusion magnetic resonance imaging and reconstruction method of the embodiment of the present application, the coil-space-contrast domain local tensor low rank constraint (Locally Low Rank Tensor, LLRT) is combined with the data fidelity term to form the objective function, and the alternating direction method of multipliers (Alternating Direction Method of Multipliers, ADMM) is preferably used as the iterative reconstruction algorithm to reconstruct the multi-contrast diffusion weighted image and obtain the ADC image.

[0171] Combine the following Figure 2 , 3, 4 describe the above steps in detail.

[0172] (S101) acquiring k-space by a keyhole echo planar sequence under any number of receiving coils; Figure 3 Schematic diagram of the variable contrast keyhole echo planar sequence. Figure 3 In the case of unit phase encoding, the momentum of K space is M u , the momentum of the Blip gradient in the center K space is M u , the momentum of the surrounding K-space Blip gradient is R p M u ,R pis the undersampling rate of the surrounding K space. The momentum difference of the Prephase gradient of adjacent TRs in the phase encoding direction is R c M u / 2. To achieve dense acquisition in the central k-space area, and undersampling rate R in the surrounding k-space area P Uniform shift acquisition. The contrast parameters between different TRs, such as b value, diffusion encoding direction, diffusion time, etc., can be changed arbitrarily. Figure 2 TR1 and TR2 in have different diffusion contrast parameters. Figure 4 A more detailed embodiment is given in FIG. 2. The b value varies continuously between different TRs.

[0173] (S102) Preprocess the acquired K-space data and perform Nyquist ghost correction and trapezoidal sampling correction. Nyquist ghost correction and trapezoidal sampling correction can use any known method. In this embodiment, the method integrated with the acquisition machine uMR790 is used for processing.

[0174] (S103) The objective equation is constructed by combining the coil-space-contrast domain tensor low rank regularization term and the data fidelity term. In this example, the low rank of the tensor (Locally Low Rank Tensor, LLRT) of the image to be sought in the coil-space-contrast domain is preferably combined with the data fidelity term to form the objective function, and the alternating direction multiplier method is selected as the iterative reconstruction algorithm to obtain the desired diffusion image.

[0175] It should be noted that the present invention does not need to obtain coil sensitivity parameters during reconstruction;

[0176] Specifically, the objective function can be described as:

[0177]

[0178] Where y represents the multi-dimensional K-space data obtained after preprocessing, D represents the undersampling operator, F represents the Fourier transform operator, x represents the multi-coil multi-contrast image to be reconstructed, α is the regularization coefficient, and B i x is the coil-space-contrast three-dimensional tensor centered at the i-th pixel, R is the rank operator of the three-dimensional tensor based on Tucker decomposition or PARAFAC decomposition, N v is a pre-set value. If a global low-rank constraint is used, N v =1, if the local low-rank constraint is used, then N vAccording to the above-mentioned fast multi-contrast magnetic resonance diffusion imaging and reconstruction method of the present invention, it is possible to avoid solving coil sensitivity map parameters while avoiding directly merging K spaces with different phases, thereby improving the quality of reconstructed images and allowing higher resolution and higher acquisition efficiency.

[0179] (S104) In the embodiment of the present application, the alternating direction multiplier method is used as an iterative reconstruction algorithm to obtain the required diffusion image. Before using the above method, auxiliary variables need to be constructed and the objective function is rewritten as the following formula.

[0180]

[0181] B i x=M i

[0182] M i =g i × 1 A i × 2 B i × 3 C i , i=1,2,…,N v

[0183] Where y represents the multi-dimensional K-space data obtained after preprocessing, D represents the undersampling operator, F represents the Fourier transform operator, and x represents the multi-coil multi-contrast image to be reconstructed. α is the regularization coefficient, B i x is the coil-space-contrast three-dimensional tensor centered at the i-th pixel, R is the rank operator of the three-dimensional tensor based on Tucker decomposition or PARAFAC decomposition, N v is a pre-set value. If a global low-rank constraint is used, N v =1, if the local low-rank constraint is used, then N v M can be equal to the number of pixels. i is an auxiliary variable of type Tensor, obtained through PARAFAC or Tucker decomposition, M i =g i × 1 A i × 2 B i × 3 C i , i=1,2,…,N v

[0184] According to one embodiment of the present invention, the objective function can be rewritten as an augmented Lagrangian form as follows:

[0185]

[0186] Where y represents the multi-dimensional K-space data obtained after preprocessing, D represents the undersampling operator, F represents the Fourier transform operator, and x represents the multi-coil multi-contrast image to be reconstructed. α is the regularization coefficient, B i x is the coil-space-contrast three-dimensional tensor centered at the i-th pixel, R is the rank operator of the three-dimensional tensor based on Tucker decomposition or PARAFAC decomposition, N v is a pre-set value. If a global low-rank constraint is used, N v =1, if the local low-rank constraint is used, then N v M can be equal to the number of pixels. Re is a real number operator. i is an auxiliary variable of type Tensor, obtained through PARAFAC or Tucker decomposition, M i =g i × 1 A i × 2 B i × 3 C i , i=1,2,…,N v , μ is the augmented Lagrangian coefficient, λ i is the Lagrange multiplier.

[0187] (S105) Before formal iterative reconstruction, it is necessary to initialize the diffusion image to be solved, that is, determine the initial value of the image. In the embodiment of the present invention, the initial value can be set to 0. For the convenience of description, it is described in the formula as: x selects the initial point x 0 .

[0188] (S106) Solve the sub-problems separately. In ADMM, it is necessary to solve the sub-problems that only have to be solved diffusion image variables, auxiliary variables and Lagrange multiplier variables. The following sub-steps describe the solution details of the corresponding sub-problems mentioned above.

[0189] Solving for x k , there is an analytical solution

[0190]

[0191] where H stands for conjugate transpose.

[0192] Solution The sub-goals are as follows

[0193]

[0194] Preferably, first perform a PARAFAC or Tucker decomposition of the tensor

[0195]

[0196] Then use the singular value soft thresholding method:

[0197]

[0198]

[0199]

[0200] Solution

[0201]

[0202] (S107) After completing the above steps, the coil image of each channel is automatically updated and iterated. The iteration stops when one of the following conditions is met: (1) the number of iterations is greater than 50; (2) the infinite norm of the gradient is less than 1*10 -10 ; (3) Step size is less than 1*10 -12 . Determine whether convergence has occurred or the maximum number of iterations has been reached.

[0203] (S108) If the iteration stop condition is not met, re-enter step S106 to continue iteration.

[0204] (S109) According to the type of diffusion weighted parameters collected and actual needs, diffusion quantitative imaging such as ADC, DTI, DKI, IVIM, etc. is solved. If it is not necessary to solve the quantitative image, the magnetic resonance diffusion weighted image is the final result.

[0205] The above discloses an embodiment to implement different structures of the present application; the described embodiment is only a part of the embodiments of the present application, not all of the embodiments. The different structures of the present invention can be implemented by other methods and combinations.

[0206] The following examples are carried out using the present invention, and the advantages of the present invention can be more fully reflected in these examples.

[0207] A control experiment of head multi-b value MRI diffusion imaging was conducted on a healthy subject using a variable contrast multiple excitation echo planar sequence and a variable contrast keyhole echo planar sequence imaging method according to an embodiment of the present application. Oral and written informed consent was obtained before the experiment. The scanning device was a United Imaging Medical uMR790 3T MRI. A 32-channel head coil was used to scan the head from b value 0-1000s / mm 2 At 50s / mm 2 The DWI full sampling K-space data of 21 different b values ​​were collected at intervals, and 3 diffusion directions were collected at each b value. Other sequence scanning parameters were:

[0208] Field of view / image size / TR / TE / bandwidth / repetition times / flip angle = 240×240 mm 2 / 128×128 / 3000ms / 132.3ms / 1790Hz / 1 / 90°. Then, the central k-space area is densely sampled, and the surrounding k-space area is undersampled by R P = 4; the central k-space region is densely sampled, and the surrounding k-space region is undersampled by R P =8, scan prospective undersampling data. Other sequence scanning parameters are: field of view / image size / TR / TE / bandwidth / repetition times / flip angle = 240×240mm 2 / 240×240 / 3000ms / 100.4ms / 1790Hz / 1 / 90°.

[0209] Figure 5 The surrounding k-space undersampling ratio R is shown at low resolution. P =4 and 8, the reconstruction results of the retrospective experiment are compared. The comparison methods are named EPI+SPA-LLR and kEPI+LLR, and the reconstruction result according to the embodiment is kEPI+LLRT (variable contrast keyhole-EPI). As can be seen from the figure, kEPI+LLRT shows a better image (closer to the gold standard) than kEPI+LLR and EPI+SPA-LLR, especially EPI+SPA-LLR has large errors in some areas.

[0210] Figure 6 The surrounding k-space undersampling ratio R is shown at high resolution. P =4 and 8, the comparison of the prospective experimental reconstruction results and the corresponding diffusion parameter images. The comparison methods are named EPI+SPA-LLR and kEPI+LLR; the result reconstructed according to the embodiment is kEPI+LLRT. EPI+SPA-LLR has the highest intensity of residual artifacts in the global image. The kEPI+LLRT constructed according to the implementation accurately reflects the details of the image, has no residual artifacts, and has a higher signal-to-noise ratio.

[0211] Thus, the present invention realizes fast high-resolution magnetic resonance diffusion imaging and reconstruction, and performs K-space acquisition by using an arbitrarily variable contrast keyhole echo plane sequence, and reconstructs using the low rank of diffusion-weighted images in the image dimension, contrast dimension, and coil dimension, thereby avoiding phase and motion artifacts caused by direct merging of k-space, and avoiding solving coil sensitivity parameters. The robustness is improved, higher resolution and higher signal-to-noise ratio are achieved, and imaging time can be significantly reduced in applications such as diffusion quantitative imaging. A balance is achieved between resolution and acquisition efficiency, which is of great significance in clinical applications.

[0212] Those skilled in the art know that, in addition to implementing the system, device and its various modules provided by the present invention in a purely computer-readable program code, it is entirely possible to implement the same program in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers and embedded microcontrollers by logically programming the method steps. Therefore, the system, device and its various modules provided by the present invention can be considered as a hardware component, and the modules included therein for implementing various programs can also be considered as structures within the hardware component; the modules for implementing various functions can also be considered as both software programs for implementing the method and structures within the hardware component.

[0213] The above describes the specific embodiments of the present invention. It should be understood that the present invention is not limited to the above specific embodiments, and those skilled in the art can make various changes or modifications within the scope of the claims, which does not affect the essence of the present invention. In the absence of conflict, the embodiments of the present application and the features in the embodiments can be combined with each other arbitrarily.

Claims

1. A fast multi-contrast magnetic resonance diffusion imaging and reconstruction method, It is characterized in that include: Step S1: Under-sampling k-space data of multiple variable contrast diffusion weighted images using a keyhole-EPI sequence through multiple receiving coils; Step S2: construct an objective function including a multi-coil image fidelity term and an image regularization term based on coil, space and contrast domain tensor low-rank (Low-Rank Tensor Constraint) according to the acquisition results; Step S3: minimizing the objective function through an optimization algorithm, reconstructing the diffusion images of all receiving coils and obtaining relevant quantitative parameter images according to the specific way of contrast change; In step S1: The keyhole-EPI sequence uses the following variable density k-space acquisition: Dense acquisition at the undersampling rate R_C in the central k-space region; Undersampling is uniformly shifted in the surrounding k-space region at an undersampling rate R_P so that adjacent TRs acquire cyclically shifted phase encoding lines; In step S2: The image to be solved is a variable contrast diffusion-weighted image under multiple coils, rather than a diffusion-weighted image after merging multiple coils; The image fidelity term based on the low rank of coil, space and contrast domain tensors uses the low rank of multi-coil variable contrast diffusion weighted images to constrain the range of variation of the image to be solved, including constraints based on the low rank of tensors in the global space (Globally Low-Rank Tensor Constraint) and constraints based on the low rank of tensors in the local space (Locally Low-Rank Tensor Constraint); The objective function is: Where y represents the multi-dimensional K-space data obtained after preprocessing, D represents the undersampling operator, F represents the Fourier transform operator, x represents the multi-coil multi-contrast image to be reconstructed, α is the regularization coefficient, and B i x is the coil, space and contrast three-dimensional tensor centered on the i-th pixel, R is the rank operator of the three-dimensional tensor based on Tucker decomposition or PARAFAC decomposition, N v is a preset value. If a global low-rank constraint is used, then N v =1, if the local low-rank constraint is used, then N v Can be equal to the number of pixels.

2. The fast multi-contrast magnetic resonance diffusion imaging and reconstruction method according to claim 1, It is characterized in that In step S1: The number of receiving coils is the number of coil channels greater than or equal to 1; The variable contrast includes: the contrast parameters between different TRs of the echo planar sequence can be arbitrarily changed, including arbitrary changes in b value, diffusion encoding direction, and diffusion time; For a specific contrast parameter, the arbitrary change means that the contrast parameter between different TRs remains unchanged or is different; The variable density k-space acquisition of the keyhole-EPI sequence is implemented as follows: Determine the gradient momentum across a single phase encoding line; and based on the calculated results, calculate other required blip gradient momentums across different phase encoding lines; arrange the required blip gradients of corresponding momentum in time sequence according to the preset sampling trajectory; based on the above calculation results, use bipolar EPI gradients to collect k-space data.

3. The fast multi-contrast magnetic resonance diffusion imaging and reconstruction method according to claim 1, It is characterized in that In step S2: The objective function does not include coil sensitivity, and solving the objective function does not require solving the coil sensitivity first or performing self-calibration of parallel imaging.

4. The fast multi-contrast magnetic resonance diffusion imaging and reconstruction method according to claim 1, It is characterized in that In step S3: Its objective function can be solved by any optimization algorithm, including gradient descent, conjugate gradient method, quasi-Newton method, alternating multiplier method, proximal gradient method, projected gradient method, and coordinate descent method, to reconstruct the diffusion images of all receiving coils and obtain relevant quantitative parameter images according to the specific application scenario.

5. The fast multi-contrast magnetic resonance diffusion imaging and reconstruction method according to claim 1, Features: This method can be used in diffusion imaging application scenarios including: diffusion-weighted imaging under multiple excitations, quantitative imaging of apparent diffusion coefficient, diffusion tensor imaging, diffusion kurtosis imaging, and intravoxel incoherent motion imaging.

6. A fast multi-contrast magnetic resonance diffusion imaging and reconstruction system, It is characterized in that include: Module M1: Undersample k-space data of multiple variable contrast diffusion-weighted images using keyhole-EPI sequence through multiple receiving coils; Module M2: Based on the acquisition results, construct an objective function including a multi-coil image fidelity term and an image regularization term based on the low rank of coil, space and contrast domain tensors; Module M3: Minimize the objective function through the optimization algorithm, reconstruct the diffusion images of all receiving coils and obtain relevant quantitative parameter images according to the specific way of contrast change; In the module M1: The keyhole-EPI sequence uses the following variable density k-space acquisition: In the central k-space region, the undersampling rate R C Intensive collection; In the surrounding k-space region, the sampling rate R P Uniformly shift undersampling so that adjacent TRs acquire cyclically shifted phase encoding lines; In the module M2: The image to be solved is a variable contrast diffusion-weighted image under multiple coils, rather than a diffusion-weighted image after merging multiple coils; The image fidelity term based on the low rank of coil, space and contrast domain tensors uses the low rank of multi-coil variable contrast diffusion weighted images to constrain the range of variation of the image to be solved, including constraints based on the low rank of tensors in the global space and constraints based on the low rank of tensors in the local space. The objective function is: Where y represents the multi-dimensional K-space data obtained after preprocessing, D represents the undersampling operator, F represents the Fourier transform operator, x represents the multi-coil multi-contrast image to be reconstructed, α is the regularization coefficient, and B i x is the coil, space and contrast three-dimensional tensor centered on the i-th pixel, R is the rank operator of the three-dimensional tensor based on Tucker decomposition or PARAFAC decomposition, N v is a preset value. If a global low-rank constraint is used, then N v =1, if the local low-rank constraint is used, then N v Can be equal to the number of pixels.

7. The fast multi-contrast magnetic resonance diffusion imaging and reconstruction system according to claim 6, It is characterized in that In the module M1: The number of receiving coils is the number of coil channels greater than or equal to 1; The variable contrast includes: the contrast parameters between different TRs of the echo planar sequence can be arbitrarily changed, including arbitrary changes in b value, diffusion encoding direction, and diffusion time; For a specific contrast parameter, the arbitrary change means that the contrast parameter between different TRs remains unchanged or is different; The variable density k-space acquisition of the keyhole-EPI sequence is implemented as follows: Determine the gradient momentum across a single phase encoding line; and based on the calculated results, calculate other required blip gradient momentums across different phase encoding lines; arrange the required blip gradients of corresponding momentum in time sequence according to the preset sampling trajectory; based on the above calculation results, use bipolar EPI gradients to collect k-space data.

8. The fast multi-contrast magnetic resonance diffusion imaging and reconstruction system according to claim 6, It is characterized in that In the module M2: The objective function does not include coil sensitivity, and solving the objective function does not require solving the coil sensitivity first or performing self-calibration of parallel imaging.

9. The fast multi-contrast magnetic resonance diffusion imaging and reconstruction system according to claim 6, It is characterized in that In the module M3: Its objective function can be solved by any optimization algorithm, including gradient descent, conjugate gradient method, quasi-Newton method, alternating multiplier method, proximal gradient method, projected gradient method, and coordinate descent method, to reconstruct the diffusion images of all receiving coils and obtain relevant quantitative parameter images according to the specific application scenario.

10. The fast multi-contrast magnetic resonance diffusion imaging and reconstruction system according to claim 6, Features: The system can be used for diffusion imaging applications including: diffusion-weighted imaging under multiple excitations, quantitative imaging of apparent diffusion coefficient, diffusion tensor imaging, diffusion kurtosis imaging, and intravoxel incoherent motion imaging.

Images