WO2019104702A1 - 一种基于自适应联合稀疏编码的并行磁共振成像方法、装置及计算机可读介质 - Google Patents

一种基于自适应联合稀疏编码的并行磁共振成像方法、装置及计算机可读介质 Download PDF

Info

Publication number
WO2019104702A1
WO2019104702A1 PCT/CN2017/114172 CN2017114172W WO2019104702A1 WO 2019104702 A1 WO2019104702 A1 WO 2019104702A1 CN 2017114172 W CN2017114172 W CN 2017114172W WO 2019104702 A1 WO2019104702 A1 WO 2019104702A1
Authority
WO
WIPO (PCT)
Prior art keywords
image
channel
module
formula
matrix
Prior art date
Application number
PCT/CN2017/114172
Other languages
English (en)
French (fr)
Inventor
王珊珊
梁栋
谭莎
刘新
郑海荣
Original Assignee
深圳先进技术研究院
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 深圳先进技术研究院 filed Critical 深圳先进技术研究院
Priority to PCT/CN2017/114172 priority Critical patent/WO2019104702A1/zh
Priority to US16/760,956 priority patent/US11668777B2/en
Publication of WO2019104702A1 publication Critical patent/WO2019104702A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G01MEASURING; TESTING
    • G01RMEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R33/00Arrangements or instruments for measuring magnetic variables
    • G01R33/20Arrangements or instruments for measuring magnetic variables involving magnetic resonance
    • G01R33/44Arrangements or instruments for measuring magnetic variables involving magnetic resonance using nuclear magnetic resonance [NMR]
    • G01R33/48NMR imaging systems
    • G01R33/54Signal processing systems, e.g. using pulse sequences ; Generation or control of pulse sequences; Operator console
    • G01R33/56Image enhancement or correction, e.g. subtraction or averaging techniques, e.g. improvement of signal-to-noise ratio and resolution
    • G01R33/561Image enhancement or correction, e.g. subtraction or averaging techniques, e.g. improvement of signal-to-noise ratio and resolution by reduction of the scanning time, i.e. fast acquiring systems, e.g. using echo-planar pulse sequences
    • G01R33/5611Parallel magnetic resonance imaging, e.g. sensitivity encoding [SENSE], simultaneous acquisition of spatial harmonics [SMASH], unaliasing by Fourier encoding of the overlaps using the temporal dimension [UNFOLD], k-t-broad-use linear acquisition speed-up technique [k-t-BLAST], k-t-SENSE
    • GPHYSICS
    • G01MEASURING; TESTING
    • G01RMEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R33/00Arrangements or instruments for measuring magnetic variables
    • G01R33/20Arrangements or instruments for measuring magnetic variables involving magnetic resonance
    • G01R33/44Arrangements or instruments for measuring magnetic variables involving magnetic resonance using nuclear magnetic resonance [NMR]
    • G01R33/48NMR imaging systems
    • G01R33/54Signal processing systems, e.g. using pulse sequences ; Generation or control of pulse sequences; Operator console
    • G01R33/56Image enhancement or correction, e.g. subtraction or averaging techniques, e.g. improvement of signal-to-noise ratio and resolution
    • G01R33/5608Data processing and visualization specially adapted for MR, e.g. for feature analysis and pattern recognition on the basis of measured MR data, segmentation of measured MR data, edge contour detection on the basis of measured MR data, for enhancing measured MR data in terms of signal-to-noise ratio by means of noise filtering or apodization, for enhancing measured MR data in terms of resolution by means for deblurring, windowing, zero filling, or generation of gray-scaled images, colour-coded images or images displaying vectors instead of pixels
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T11/002D [Two Dimensional] image generation
    • G06T11/003Reconstruction from projections, e.g. tomography
    • G06T11/008Specific post-processing after tomographic reconstruction, e.g. voxelisation, metal artifact correction
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/10Image acquisition modality
    • G06T2207/10072Tomographic images
    • G06T2207/10088Magnetic resonance imaging [MRI]
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/20Special algorithmic details
    • G06T2207/20048Transform domain processing
    • G06T2207/20056Discrete and fast Fourier transform, [DFT, FFT]
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/20Special algorithmic details
    • G06T2207/20081Training; Learning

Definitions

  • the present invention relates to the field of magnetic resonance imaging, and in particular to a parallel magnetic resonance imaging method, apparatus and computer readable medium based on adaptive joint sparse coding.
  • the calibration-free reconstruction method does not require special calibration information to estimate sensitivity. They reconstruct images in the K-space or image domain through the relationship between the channels.
  • the proposed calibration-free reconstruction method now only utilizes joint sparsity directly from the conversion domain.
  • the present invention proposes a parallel magnetic resonance imaging method and apparatus based on adaptive joint sparse coding, which can improve sparsity based on block sparsity and can further improve the acceleration multiple. To ensure the quality of reconstruction.
  • the invention provides a parallel magnetic resonance imaging method based on adaptive joint sparse coding, comprising the following steps:
  • Step a Construct a parallel magnetic resonance imaging model based on calibration-free joint sparse coding, ie reconstruction model, which is defined as:
  • V represents the reconstructed image
  • represents the objective function
  • F M represents the Fourier transform
  • v j represents the image of each channel
  • the matrix of all the channels v j merged together is V
  • f j represents K-space data
  • M represents the size of the atom
  • N represents the vectorized size of the image
  • P represents the number of atoms
  • subscript F represents the Fourier transform
  • represents the data fitting weight
  • represents the joint sparse regularization weight
  • 2,1 means joint sparse term
  • Step b Temporarily fix X, and solve the dictionary D by using the gradient descent method.
  • represents the learning rate and k represents the number of iterations.
  • Step c updating the joint sparse coefficient X
  • Step d update v j by the following formula
  • Step e Update the K-space data by the following formula:
  • Step f performing inverse Fourier transform on the K-space data to update v j again
  • Step g Obtain an updated image based on the image v j of all channels.
  • constructing a reconstruction model is to solve one Minimum objective function, where
  • the norm is a data fitting term.
  • Norm represents sparse representation error
  • the mixed norm represents a joint sparse constraint between each channel.
  • step a the initialization X 0 , V 0 is included , A step of.
  • the atoms are normalized to avoid ambiguity in the dictionary atomic scale.
  • updating v j further comprises the following steps:
  • Step d1 Obtain the following formula according to the least squares method:
  • I a diagonal matrix
  • the value of the element on the diagonal is equal to the number of blocks in which the image pixels of the corresponding position overlap; when it is assumed that these blocks are mapped at the edge of the image, the values of the elements on the diagonal are all equal,
  • M
  • Step d2 The image domain is converted to the Fourier domain by normalizing the full-Fourier Fourier coding matrix to obtain:
  • I a diagonal matrix of 0 and 1, where 1 represents the position of the point taken in K space, ⁇ represents the number of pixel repetitions, and I N represents the unit matrix of N*N, a Fourier measurement indicating zero padding of the j-th channel;
  • Step d3 Transform the formula in step d2 into:
  • Fv j (k x , k y ) represents the updated value of the K-space position (k x , k y ), Represents the zero-padded K-space measurement of the j-th channel; ⁇ represents a small scalar, and ⁇ represents the set of points that the K-space has acquired.
  • Step d4 v j is obtained by inverse Fourier transform by the data in the frequency domain of the formula (9) of the step c3.
  • step g the squares of the images v j of all the channels are summed, and then squared to obtain an updated image.
  • Another aspect of the present invention provides a parallel magnetic resonance imaging apparatus based on adaptive joint sparse coding, comprising: a model construction module, and constructing a parallel magnetic resonance imaging model based on calibration-free joint sparse coding, that is, a reconstruction model,
  • the model is defined as:
  • V represents the reconstructed image
  • represents the objective function
  • F M represents the Fourier transform
  • v j represents the image of each channel
  • the matrix of all the channels v j merged together is V
  • f j represents K-space data
  • M represents the size of the atom
  • N represents the vectorized size of the image
  • P represents the number of atoms
  • subscript F represents the Fourier transform
  • represents the data fitting weight
  • represents the joint sparse regularization weight
  • 2,1 means joint sparse term
  • the update module includes: a module for updating the joint sparse code X, a module for updating the dictionary D, a module for updating the image v j of each channel, and a module for updating the K-space data f j ;
  • represents the learning rate and k represents the number of iterations.
  • the module that updates the image v j of each channel updates v j by the following formula,
  • the module that updates the K-space data f j updates the K-space data by the following formula:
  • the imaging module obtains an updated image based on the image v j of all channels.
  • the imaging device further comprises: an initialization module, configured to initialize X 0 , V 0 ,
  • constructing a reconstruction model is solving one Minimum objective function, where
  • the norm is a data fitting term.
  • Norm represents sparse representation error,
  • the mixed norm represents a joint sparse constraint between each channel.
  • the update module normalizes the atom when updating the dictionary D to avoid the dictionary atomic scale blur.
  • the module for updating the image v j of each channel further performs the following steps:
  • Step d1 Obtain the following formula according to the least squares method:
  • I a diagonal matrix
  • the value of the element on the diagonal is equal to the number of blocks in which the pixel of the corresponding position overlaps; when it is assumed that these blocks are mapped at the edge of the image, the values of the elements on the diagonal are all equal.
  • M
  • Step d2 The image domain is converted to the Fourier domain by normalizing the full-Fourier Fourier coding matrix to obtain:
  • I a diagonal matrix of 0 and 1, where 1 represents the position of the point taken in K space, ⁇ represents the number of pixel repetitions, and I N represents the unit matrix of N*N, a Fourier measurement indicating zero padding of the j-th channel;
  • Step d3 Transform the formula in step d2 into:
  • Fv j (k x , k y ) represents the updated value of the K-space position (k x , k y ), Represents the zero-padded K-space measurement of the j-th channel; ⁇ represents a small scalar, and ⁇ represents the set of points that the K-space has acquired.
  • Step d4 v j is obtained by inverse Fourier transform by the data in the frequency domain of the formula (9) of the step d3.
  • the imaging module sums the squares of the images v j of all channels and then squares to obtain an updated image.
  • the present invention also provides a computer readable medium having a program stored therein, the program being computer executable to cause a computer to perform the parallel magnetic resonance imaging method based on adaptive joint sparse coding described above Each step.
  • the method of the invention is utilized
  • the joint sparseness of the norm development channel, while developing information sparsity can be free of calibration, and the method has strong robustness.
  • FIG. 1 is a flow chart of a parallel magnetic resonance imaging method based on adaptive joint sparse coding of the present invention.
  • FIG. 2 is a block diagram of a parallel magnetic resonance imaging apparatus based on adaptive joint sparse coding of the present invention.
  • step a construct a parallel magnetic co-synthesis based on joint-free sparse coding.
  • the vibration imaging model is the reconstruction model, which is defined as:
  • V represents the reconstructed image
  • represents the objective function
  • F M represents the Fourier transform
  • v j represents the image of each channel
  • the matrix of all the channels v j merged together is V
  • f j represents K-space data
  • M represents the size of the atom
  • N represents the vectorized size of the image
  • P represents the number of atoms
  • subscript F represents the Fourier transform
  • represents the data fitting weight
  • represents the joint sparse regularization weight
  • 2,1 means joint sparse term
  • step a constructing a reconstruction model is to solve one Minimum objective function, where
  • the norm is a data fitting term.
  • Norm represents sparse representation error
  • the mixed norm represents a joint sparse constraint between each channel.
  • the method further includes initializing X 0 , V 0 , A step of.
  • step b temporarily fix X, and solve the dictionary D by using the gradient descent method.
  • represents the learning rate and k represents the number of iterations.
  • step b the atoms are normalized to avoid ambiguity in the dictionary atomic scale.
  • step c update the joint sparse coefficient X until the inner layer iteration reaches the specified number of times
  • q is the number of loops in which the inner loop is nested
  • B is the auxiliary variable
  • is the auxiliary parameter
  • step d update the image v j of each channel by the following formula
  • updating v j further includes the following steps:
  • Step d1 Obtain the following formula according to the least squares method:
  • I a diagonal matrix
  • the value of the element on the diagonal is equal to the number of blocks in which the image pixels of the corresponding position overlap; when it is assumed that these blocks are mapped at the edge of the image, the values of the elements on the diagonal are all equal,
  • M
  • Step d2 The image domain is converted to the Fourier domain by normalizing the full-Fourier Fourier coding matrix to obtain:
  • I a diagonal matrix of 0 and 1, where 1 represents the position of the point taken in K space, ⁇ represents the number of pixel repetitions, and I N represents the unit matrix of N*N, a Fourier measurement indicating zero padding of the j-th channel;
  • Step d3 Transform the formula in step d2 into:
  • Fv j (k x , k y ) represents the updated value of the K-space position (k x , k y ), Represents the zero-padded K-space measurement of the j-th channel; ⁇ represents a small scalar, and ⁇ represents the set of points that the K-space has acquired.
  • Step d4 v j is obtained by inverse Fourier transform by the data in the frequency domain of the formula (9) of the step d3.
  • step e update the K-space data by the following formula:
  • step f performing inverse Fourier transform on the K-space data to update v j again
  • step g is performed: summing the squares of the images v j of all the channels, and then prescribing to obtain the updated image.
  • the method proposed by the present invention can achieve high-quality magnetic resonance image reconstruction with 4 times acceleration through the overall model and algorithm.
  • the joint sparse norm can better suppress the noise
  • the dictionary update can adaptively acquire the structural information of the image
  • the accurate iterative loop solution of the overall algorithm can achieve the reconstruction quality of the minimum error.
  • the present invention also provides a computer readable medium having a program stored therein that is computer executable to cause a computer to perform processing including the steps described above.
  • the present invention also provides a parallel joint magnetic resonance imaging apparatus based on adaptive joint sparse coding for the above method.
  • the device includes different modules for implementing the various steps mentioned in the above method.
  • the apparatus can include a model building module, an update module, and an imaging module.
  • the model building module constructs a parallel magnetic resonance imaging model based on calibration-free joint sparse coding, ie reconstruction model.
  • the update module includes a module for updating the joint sparse code X, a module for updating the dictionary D, a module for updating the image v j of each channel, and a module for updating the K-space data f j .
  • the imaging module obtains an updated image based on the image v j of all channels.
  • the imaging device further comprises: an initialization module, configured to initialize X 0 , V 0 ,
  • constructing a reconstruction model is to solve one Minimum objective function, where
  • the norm is a data fitting term.
  • Norm represents sparse representation error
  • the mixed norm represents a joint sparse constraint between each channel.
  • the update module updates the dictionary D
  • the atom needs to be normalized to avoid the dictionary atomic scale blur.
  • the imaging module sums the squares of the images v j of all channels and then squares to obtain an updated image.

Landscapes

  • Physics & Mathematics (AREA)
  • Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • High Energy & Nuclear Physics (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Radiology & Medical Imaging (AREA)
  • Signal Processing (AREA)
  • Condensed Matter Physics & Semiconductors (AREA)
  • Theoretical Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Magnetic Resonance Imaging Apparatus (AREA)

Abstract

一种基于自适应联合稀疏编码的并行磁共振成像方法、装置以及计算机可读介质。本发明方法通过求解一个l 2-l F-l 2 , 1最小目标函数,l 2范数是数据拟合项,l F范数代表稀疏表达误差,l 2,1混合范数代表每个通道之间的联合稀疏约束;再采用分而治之的方法和对应的算法更新稀疏矩阵、字典和K空间数据,最后,通过所有通道的均方根和求得重建图像。利用了l 2,1范数开发通道的联合稀疏,开发信息稀疏性的同时,可以免校准,且该方法有较强的鲁棒性。

Description

一种基于自适应联合稀疏编码的并行磁共振成像方法、装置及计算机可读介质 技术领域
本发明涉及磁共振成像领域,尤其涉及一种基于自适应联合稀疏编码的并行磁共振成像方法、装置以及计算机可读介质。
背景技术
目前,为了提高磁共振成像速度,基于压缩感知的并行磁共振成像比较流行。基于如何使用灵敏度信息,可把重建方法大概分为三类:精确校准,自动校准和免校准。免校准方法结合先验信息,在灵敏度信息上具有较强的鲁棒性。自动校准不依赖于准确的预估计灵敏度信息。K空间自校准灵敏度方法有GRAPPA,L1-SPIRiT等,这些方法的图像重建质量较好。但随着加速倍数提高,这些算法的性能随之降低。针对该缺陷,提出了JSENSE,Sparse BLIP等算法,在估计灵敏度信息的同时还能估计欠采图像。然而同时估计是一个非凸问题,对灵敏度的初始值非常敏感,且计算量非常复杂。免校准的重建方法不需要特殊的校准信息来估计灵敏度。它们在K空间或图像域通过通道间的关系来重建图像。然而,现在提出的免校准重建方法只是直接从转换域利用联合稀疏性。
发明内容
针对现有技术的上述问题,本发明提出了一种基于自适应联合稀疏编码的并行磁共振成像方法及装置,其基于块的稀疏性,能够提高稀疏性,且能在进一步提高加速倍数的同时,保证重建质量。
本发明提供了一种基于自适应联合稀疏编码的并行磁共振成像方法,包括以下步骤:
步骤a:构建一个基于免校准的联合稀疏编码的并行磁共振成像模型即重建模型,所述模型定义为:
Figure PCTCN2017114172-appb-000001
式中,V代表重建的图像,
Figure PCTCN2017114172-appb-000002
代表过完备(P>>M)的字典,θ表示目标函数,
Figure PCTCN2017114172-appb-000003
代表稀疏矩阵,
Figure PCTCN2017114172-appb-000004
表示从第j-th通道的图像提取第l-th图像块,FM代表傅里叶变换,vj表示每个通道的图像,所有通道的图像vj合并在一起组成的矩阵即为V,fj代表K空间数据,
Figure PCTCN2017114172-appb-000005
代表块提取矩阵,M表示原子的尺寸,N表示图像的向量化尺寸,P表示原子的个数,下标F表示傅里叶变换,λ表示数据拟合项权重,β表示联合稀疏正则化权重,||·||2,1表示联合稀疏项,
Figure PCTCN2017114172-appb-000006
式中,
Figure PCTCN2017114172-appb-000007
表示第j-th通道的第l-th提取块中第p-th像素;
步骤b:临时固定X,采用梯度下降的方法求解字典D,
Figure PCTCN2017114172-appb-000008
式中γ表示学习率,k表示迭代次数,
由于
Figure PCTCN2017114172-appb-000009
该公式是重建目标函数对字典D的倒数,H表示共轭转置,
因此获得下式:
Dk+1=Dk-γ(DX-RlV)XH   (3)
步骤c:更新联合稀疏系数X;
步骤d:通过下式更新vj
Figure PCTCN2017114172-appb-000010
步骤e:通过下述公式更新K空间数据:
Figure PCTCN2017114172-appb-000011
其中,
Figure PCTCN2017114172-appb-000012
表示第j个通道的第k次迭代更新值;
步骤f:对K空间数据进行傅里叶反变换再次更新vj,获得
Figure PCTCN2017114172-appb-000013
以及
步骤g:根据所有通道的图像vj,获得更新后的图像。
优选地,步骤a中,构建重建模型即求解一个
Figure PCTCN2017114172-appb-000014
最小目标函数,其中,
Figure PCTCN2017114172-appb-000015
范数是数据拟合项,
Figure PCTCN2017114172-appb-000016
范数代表稀疏表达误差,
Figure PCTCN2017114172-appb-000017
混合范数代表每个通道之间的联合稀疏约束。
优选地,步骤a中,包括初始化X0,V0,
Figure PCTCN2017114172-appb-000018
的步骤。
优选地,步骤b中,对原子进行归一化处理,以避免字典原子尺度模糊。
优选地,步骤d中,更新vj进一步包括如下步骤:
步骤d1:根据最小二乘法获得下式:
Figure PCTCN2017114172-appb-000019
式中,
Figure PCTCN2017114172-appb-000020
是一个对角矩阵,对角线上元素值等于对应位置图像像素重叠的块的个数;当假设这些块映射在图像的边缘,则对角线上元素值全部 相等,
Figure PCTCN2017114172-appb-000021
特殊情况下,当填充的距离为单位距离1时,ω=M;
步骤d2:图像域通过归一化全采傅里叶编码矩阵转换到傅里叶域,获得:
Figure PCTCN2017114172-appb-000022
式中,
Figure PCTCN2017114172-appb-000023
是0和1的对角矩阵,1代表K空间采到的点的位置,ω表示像素重复的个数,IN表示N*N的单位矩阵,
Figure PCTCN2017114172-appb-000024
表示第j-th通道零填充的傅里叶测量值;
步骤d3:将步骤d2中的公式变换成:
Figure PCTCN2017114172-appb-000025
式中,Fvj(kx,ky)表示K空间位置(kx,ky)的更新值,
Figure PCTCN2017114172-appb-000026
表示第j-th通道的零填充K空间测量值;η表示一个小标量,Ω表示K空间已经采到的点的集合,
Figure PCTCN2017114172-appb-000027
Figure PCTCN2017114172-appb-000028
步骤d4:通过步骤c3的公式(9)中频域中的数据通过逆傅里叶变换求得vj
优选地,步骤g中,对所有通道的图像vj的平方进行求和,然后开方,获得更新后的图像。
本发明的另一个方面提供了一种基于自适应联合稀疏编码的并行磁共振成像装置,包括:模型构建模块,构建一个基于免校准的联合稀疏编码的并行磁共振成像模型即重建模型,所述模型定义为:
Figure PCTCN2017114172-appb-000029
式中,V代表重建的图像,
Figure PCTCN2017114172-appb-000030
代表过完备(P>>M)的字典,θ表示目标函数,
Figure PCTCN2017114172-appb-000031
代表稀疏矩阵,
Figure PCTCN2017114172-appb-000032
表示从第j-th通道的图像提取第l-th图像块,FM代表傅里叶变换,vj表示每个通道的图像,所有通道的图像vj合并在一起组成的矩阵即为V,fj代表K空间数据,
Figure PCTCN2017114172-appb-000033
代表块提取矩阵,M表示原子的尺寸,N表示图像的向量化尺寸,P表示原子的个数,下标F表示傅里叶变换,λ表示数据拟合项权重,β表示联合稀疏正则化权重,||·||2,1表示联合稀疏项,
Figure PCTCN2017114172-appb-000034
式中,
Figure PCTCN2017114172-appb-000035
表示第j-th通道的第l-th提取块中第p-th像素;
更新模块,包括:更新联合稀疏编码X的模块、更新字典D的模块、更新每一个通道的图像vj的模块以及更新K空间数据fj的模块;其中,
更新字典D的模块临时固定X,采用梯度下降的方法求解字典D,
Figure PCTCN2017114172-appb-000036
式中γ表示学习率,k表示迭代次数,
由于
Figure PCTCN2017114172-appb-000037
该公式是重建目标函数对字典D的倒数,H表示共轭转置,
因此获得下式:
Dk+1=Dk-γ(DX-RlV)XH   (3)
更新每一个通道的图像vj的模块通过下式更新vj
Figure PCTCN2017114172-appb-000038
更新K空间数据fj的模块通过下述公式更新K空间数据:
Figure PCTCN2017114172-appb-000039
其中,
Figure PCTCN2017114172-appb-000040
表示第j个通道的第k次迭代更新值;
并且对K空间数据进行傅里叶反变换再次更新vj,获得
Figure PCTCN2017114172-appb-000041
成像模块,根据所有通道的图像vj,获得更新后的图像。
优选地,该成像装置还包括:初始化模块,用于初始化X0,V0,
Figure PCTCN2017114172-appb-000042
优选地,构建重建模型即求解一个
Figure PCTCN2017114172-appb-000043
最小目标函数,其中,
Figure PCTCN2017114172-appb-000044
范数是数据拟合项,
Figure PCTCN2017114172-appb-000045
范数代表稀疏表达误差,
Figure PCTCN2017114172-appb-000046
混合范数代表每个通道之间的联合稀疏约束。
优选地,所述更新模块在更新字典D时,对原子进行归一化处理,以避免字典原子尺度模糊。
优选地,更新每一个通道的图像vj的模块进一步执行如下步骤:
步骤d1:根据最小二乘法获得下式:
Figure PCTCN2017114172-appb-000047
式中,
Figure PCTCN2017114172-appb-000048
是一个对角矩阵,对角线上元素值等于对应位置图像像素重叠的块的个数;当假设这些块映射在图像的边缘,则对角线上元素值全部相等,
Figure PCTCN2017114172-appb-000049
特殊情况下,当填充的距离为单位距离1时,ω=M;
步骤d2:图像域通过归一化全采傅里叶编码矩阵转换到傅里叶域,获得:
Figure PCTCN2017114172-appb-000050
式中,
Figure PCTCN2017114172-appb-000051
是0和1的对角矩阵,1代表K空间采到的点的位置,ω表示像素重复的个数,IN表示N*N的单位矩阵,
Figure PCTCN2017114172-appb-000052
表示第j-th通道零填充的傅里叶测量值;
步骤d3:将步骤d2中的公式变换成:
Figure PCTCN2017114172-appb-000053
式中,Fvj(kx,ky)表示K空间位置(kx,ky)的更新值,
Figure PCTCN2017114172-appb-000054
表示第j-th通道的零填充K空间测量值;η表示一个小标量,Ω表示K空间已经采到的点的集合,
Figure PCTCN2017114172-appb-000055
Figure PCTCN2017114172-appb-000056
步骤d4:通过步骤d3的公式(9)中频域中的数据通过逆傅里叶变换求得vj
优选地,所述成像模块对所有通道的图像vj的平方进行求和,然后开方,获得更新后的图像。
本发明还提供了一种计算机可读介质,该计算机可读介质具有存储在其中的程序,该程序是计算机可执行的以使计算机执行上述的基于自适应联合稀疏编码的并行磁共振成像方法的各步骤。
有益效果
本方发明方法利用了
Figure PCTCN2017114172-appb-000057
范数开发通道的联合稀疏,开发信息稀疏性的同时,可以免校准,且该方法有较强的鲁棒性。
附图说明
图1是本发明的基于自适应联合稀疏编码的并行磁共振成像方法的流程图。
图2是本发明的基于自适应联合稀疏编码的并行磁共振成像装置的框图。
具体实施方式
在下列说明中,为了提供对本发明的彻底了解而提出许多具体细节。本发明可在不具有部分或所有这些具体细节的情况下实施。在其他情况下,为了不对本发明造成不必要的混淆,不详述众所周知的过程操作。虽然本发明将结合具体实施例来进行说明,但应当理解的是,这并非旨在将本发明限制于这些实施例。
下面根据图1对本发明的一个实施例的基于自适应联合稀疏编码的并行磁共振成像方法进行详细说明。
首先,步骤a,构建一个基于免校准的联合稀疏编码的并行磁共 振成像模型即重建模型,所述模型定义为:
Figure PCTCN2017114172-appb-000058
式中,V代表重建的图像,
Figure PCTCN2017114172-appb-000059
代表过完备(P>>M)的字典,θ表示目标函数,
Figure PCTCN2017114172-appb-000060
代表稀疏矩阵,
Figure PCTCN2017114172-appb-000061
表示从第j-th通道的图像提取第l-th图像块,FM代表傅里叶变换,vj表示每个通道的图像,所有通道的图像vj合并在一起组成的矩阵即为V,fj代表K空间数据,
Figure PCTCN2017114172-appb-000062
代表块提取矩阵,M表示原子的尺寸,N表示图像的向量化尺寸,P表示原子的个数,下标F表示傅里叶变换,λ表示数据拟合项权重,β表示联合稀疏正则化权重,||·||2,1表示联合稀疏项,
Figure PCTCN2017114172-appb-000063
式中,
Figure PCTCN2017114172-appb-000064
表示第j-th通道的第l-th提取块中第p-th像素。
这里,步骤a中,构建重建模型即求解一个
Figure PCTCN2017114172-appb-000065
最小目标函数,其中,
Figure PCTCN2017114172-appb-000066
范数是数据拟合项,
Figure PCTCN2017114172-appb-000067
范数代表稀疏表达误差,
Figure PCTCN2017114172-appb-000068
混合范数代表每个通道之间的联合稀疏约束。步骤a中,还包括初始化X0,V0,
Figure PCTCN2017114172-appb-000069
的步骤。
接着进行步骤b:临时固定X,采用梯度下降的方法求解字典D,
Figure PCTCN2017114172-appb-000070
式中γ表示学习率,k表示迭代次数,
由于
Figure PCTCN2017114172-appb-000071
该公式是重建目标函数对字典D的倒数,H表示共轭转置,因此获得下式:
Dk+1=Dk-γ(DX-RlV)XH   (3)
这里,步骤b中,对原子进行归一化处理,以避免字典原子尺度模糊。
接着进行步骤c:更新联合稀疏系数X,直到内层迭代达到规定次数,
Figure PCTCN2017114172-appb-000072
其中,q是内循环又嵌套的一层循环的次数,B表示辅助变量,α表示辅助参数。
接下来进行步骤d:通过下式更新每一个通道的图像vj
Figure PCTCN2017114172-appb-000073
这里,步骤d中,更新vj进一步包括如下步骤:
步骤d1:根据最小二乘法获得下式:
Figure PCTCN2017114172-appb-000074
式中,
Figure PCTCN2017114172-appb-000075
是一个对角矩阵,对角线上元素值等于对应位置图像像素重叠的块的个数;当假设这些块映射在图像的边缘,则对角线上元素值全部 相等,
Figure PCTCN2017114172-appb-000076
特殊情况下,当填充的距离为单位距离1时,ω=M;
步骤d2:图像域通过归一化全采傅里叶编码矩阵转换到傅里叶域,获得:
Figure PCTCN2017114172-appb-000077
式中,
Figure PCTCN2017114172-appb-000078
是0和1的对角矩阵,1代表K空间采到的点的位置,ω表示像素重复的个数,IN表示N*N的单位矩阵,
Figure PCTCN2017114172-appb-000079
表示第j-th通道零填充的傅里叶测量值;
步骤d3:将步骤d2中的公式变换成:
Figure PCTCN2017114172-appb-000080
式中,Fvj(kx,ky)表示K空间位置(kx,ky)的更新值,
Figure PCTCN2017114172-appb-000081
表示第j-th通道的零填充K空间测量值;η表示一个小标量,Ω表示K空间已经采到的点的集合,
Figure PCTCN2017114172-appb-000082
Figure PCTCN2017114172-appb-000083
步骤d4:通过步骤d3的公式(9)中频域中的数据通过逆傅里叶变换求得vj
接下来进行步骤e:通过下述公式更新K空间数据:
Figure PCTCN2017114172-appb-000084
其中,
Figure PCTCN2017114172-appb-000085
表示第j个通道的第k次迭代更新值。
接着进行步骤f:对K空间数据进行傅里叶反变换再次更新vj,获得
Figure PCTCN2017114172-appb-000086
最后进行步骤g:对所有通道的图像vj的平方进行求和,然后开方,获得更新后的图像。
现有的并行磁共振成像方法中,一般利用二维随机欠采样,该采样模式会使得直接反傅里叶变换的图像有类噪声的伪影且细节丢失与模糊。而本发明提出的方法可以通过整体模型与算法实现4倍加速的高质量磁共振图像重建。具体地,联合稀疏范数可以较好地抑制噪声,字典更新可以自适应地获取图像的结构信息,而整体算法的精确迭代循环求解可以实现最小误差的重建质量。
此外,本发明还提供了一种计算机可读介质,该计算机可读介质具有存储在其中的程序,该程序是计算机可执行的,以使计算机执行包括上述各步骤的处理。
本发明还提供了一种用于上述方法的基于自适应联合稀疏编码的并行磁共振成像装置。该装置包括了不同模块用于实现上述方法中提及的各个步骤。
下面根据图2说明本发明的一个实施例的成像装置。例如,该装置可以包括模型构建模块、更新模块和成像模块。其中,模型构建模块,构建一个基于免校准的联合稀疏编码的并行磁共振成像模型即重建模型。更新模块,包括更新联合稀疏编码X的模块、更新字典D的模块、更新每一个通道的图像vj的模块以及更新K空间数据fj的模块。成像模块,根据所有通道的图像vj,获得更新后的图像。优选地,该成像装置还进一步包括:初始化模块,用于初始化X0,V0,
Figure PCTCN2017114172-appb-000087
上述模型构建模块中,构建重建模型即求解一个
Figure PCTCN2017114172-appb-000088
最小目标函数,其中,
Figure PCTCN2017114172-appb-000089
范数是数据拟合项,
Figure PCTCN2017114172-appb-000090
范数代表稀疏表达误差,
Figure PCTCN2017114172-appb-000091
混合范数代表每个通道之间的联合稀疏约束。
优选地,上述更新模块在更新字典D时,需要对原子进行归一化处理,以避免字典原子尺度模糊。
优选地,上述成像模块对所有通道的图像vj的平方进行求和,然后开方,从而获得更新后的图像。
尽管已经根据优选的实施方案对本发明进行了说明,但是存在落入本发明范围之内的改动、置换以及各种替代等同方案。还应当注意的是,存在多种实现本发明的方法和系统的可选方式。因此,意在将随附的权利要求书解释为包含落在本发明的主旨和范围之内的所有这些改动、置换以及各种替代等同方案。

Claims (13)

  1. 一种基于自适应联合稀疏编码的并行磁共振成像方法,其特征在于,包括以下步骤:
    步骤a:构建一个基于免校准的联合稀疏编码的并行磁共振成像模型即重建模型,所述模型定义为:
    Figure PCTCN2017114172-appb-100001
    式中,V代表重建的图像,
    Figure PCTCN2017114172-appb-100002
    代表过完备(P>>M)的字典,θ表示目标函数,
    Figure PCTCN2017114172-appb-100003
    代表稀疏矩阵,
    Figure PCTCN2017114172-appb-100004
    表示从第j-th通道的图像提取第l-th图像块,FM代表傅里叶变换,vj表示每个通道的图像,所有通道的图像vj合并在一起组成的矩阵即为V,fj代表K空间数据,
    Figure PCTCN2017114172-appb-100005
    代表块提取矩阵,M表示原子的尺寸,N表示图像的向量化尺寸,P表示原子的个数,下标F表示傅里叶变换,λ表示数据拟合项权重,β表示联合稀疏正则化权重,||·||2,1表示联合稀疏项,
    Figure PCTCN2017114172-appb-100006
    式中,
    Figure PCTCN2017114172-appb-100007
    表示第j-th通道的第l-th提取块中第p-th像素;
    步骤b:临时固定X,采用梯度下降的方法求解字典D,
    Figure PCTCN2017114172-appb-100008
    式中γ表示学习率,k表示迭代次数,
    由于
    Figure PCTCN2017114172-appb-100009
    该公式是重建目标函数对字典D的倒数,H表示共轭转置,
    因此获得下式:
    Dk+1=Dk-γ(DX-RlV)XH    (3)
    步骤c:更新联合稀疏系数X;
    步骤d:通过下式更新vj
    Figure PCTCN2017114172-appb-100010
    步骤e:通过下述公式更新K空间数据:
    Figure PCTCN2017114172-appb-100011
    其中,
    Figure PCTCN2017114172-appb-100012
    表示第j个通道的第k次迭代更新值;
    步骤f:对K空间数据进行傅里叶反变换再次更新vj,获得
    Figure PCTCN2017114172-appb-100013
    以及
    步骤g:根据所有通道的图像vj,获得更新后的图像。
  2. 根据权利要求1所述的方法,其特征在于,步骤a中,构建重建模型即求解一个
    Figure PCTCN2017114172-appb-100014
    最小目标函数,其中,
    Figure PCTCN2017114172-appb-100015
    范数是数据拟合项,
    Figure PCTCN2017114172-appb-100016
    范数代表稀疏表达误差,
    Figure PCTCN2017114172-appb-100017
    混合范数代表每个通道之间的联合稀疏约束。
  3. 根据权利要求1所述的方法,其特征在于,步骤a中,包括初始化X0,V0,
    Figure PCTCN2017114172-appb-100018
    的步骤。
  4. 根据权利要求1所述的方法,其特征在于,步骤b中,对原子进行归一化处理,以避免字典原子尺度模糊。
  5. 根据权利要求1所述的方法,其特征在于,步骤d中,更新vj进一步包括如下步骤:
    步骤d1:根据最小二乘法获得下式:
    Figure PCTCN2017114172-appb-100019
    式中,
    Figure PCTCN2017114172-appb-100020
    是一个对角矩阵,对角线上元素值等于对应位置图像像素重叠的块的个数;当假设这些块映射在图像的边缘,则对角线上元素值全部相等,
    Figure PCTCN2017114172-appb-100021
    特殊情况下,当填充的距离为单位距离1时,ω=M;
    步骤d2:图像域通过归一化全采傅里叶编码矩阵转换到傅里叶域,获得:
    Figure PCTCN2017114172-appb-100022
    式中,
    Figure PCTCN2017114172-appb-100023
    是0和1的对角矩阵,1代表K空间采到的点的位置,ω表示像素重复的个数,IN表示N*N的单位矩阵,
    Figure PCTCN2017114172-appb-100024
    表示第j-th通道零填充的傅里叶测量值;
    步骤d3:将步骤d2中的公式变换成:
    Figure PCTCN2017114172-appb-100025
    式中,Fvj(kx,ky)表示K空间位置(kx,ky)的更新值,
    Figure PCTCN2017114172-appb-100026
    表示第j-th通道的零填充K空间测量值;η表示一个小标量,Ω表示K空间已经采到的点的集合,
    Figure PCTCN2017114172-appb-100027
    步骤d4:通过步骤d3的公式(9)中频域中的数据通过逆傅里叶变换求得vj
  6. 根据权利要求1所述的方法,其特征在于,步骤g中,对所有通道的图像vj的平方进行求和,然后开方,获得更新后的图像。
  7. 一种基于自适应联合稀疏编码的并行磁共振成像装置,其特征在于,包括:
    模型构建模块,构建一个基于免校准的联合稀疏编码的并行磁共振成像模型即重建模型,所述模型定义为:
    Figure PCTCN2017114172-appb-100028
    式中,V代表重建的图像,
    Figure PCTCN2017114172-appb-100029
    代表过完备(P>>M)的字典,θ表示目标函数,
    Figure PCTCN2017114172-appb-100030
    代表稀疏矩阵,
    Figure PCTCN2017114172-appb-100031
    表示从第j-th通道的图像提取第l-th图像块,FM代表傅里叶变换,vj表示每个通道的图像,所有通道的图像vj合并在一起组成的矩阵即为V,fj代表K空间数据,
    Figure PCTCN2017114172-appb-100032
    代表块提取矩阵,M表示原子的尺寸,N表示图像的向量化尺寸,P表示原子的个数,下标F表示傅里叶变换,λ表示数据拟合项权重,β表示联合稀疏正则化权重,||·||2,1表示联合稀疏项,
    Figure PCTCN2017114172-appb-100033
    式中,
    Figure PCTCN2017114172-appb-100034
    表示第j-th通道的第l-th提取块中第p-th像素;
    更新模块,包括:更新联合稀疏编码X的模块、更新字典D的模块、更新每一个通道的图像vj的模块以及更新K空间数据fj的模块;其中, 更新字典D的模块临时固定X,采用梯度下降的方法求解字典D,
    Figure PCTCN2017114172-appb-100035
    式中γ表示学习率,k表示迭代次数,
    由于
    Figure PCTCN2017114172-appb-100036
    该公式是重建目标函数对字典D的倒数,H表示共轭转置,
    因此获得下式:
    Dk+1=Dk-γ(DX-RlV)XH    (3)
    更新每一个通道的图像vj的模块通过下式更新vj
    Figure PCTCN2017114172-appb-100037
    更新K空间数据fj的模块通过下述公式更新K空间数据:
    Figure PCTCN2017114172-appb-100038
    其中,
    Figure PCTCN2017114172-appb-100039
    表示第j个通道的第k次迭代更新值;
    并且对K空间数据进行傅里叶反变换再次更新vj,获得
    Figure PCTCN2017114172-appb-100040
    成像模块,根据所有通道的图像vj,获得更新后的图像。
  8. 根据权利要求7所述的成像装置,其特征在于,还包括:初始化模块,用于初始化X0,V0,
    Figure PCTCN2017114172-appb-100041
  9. 根据权利要求7所述的成像装置,其特征在于,构建重建模型即求解一个
    Figure PCTCN2017114172-appb-100042
    最小目标函数,其中,
    Figure PCTCN2017114172-appb-100043
    范数是数据拟合项,
    Figure PCTCN2017114172-appb-100044
    范数代表稀疏表达误差,
    Figure PCTCN2017114172-appb-100045
    混合范数代表每个通道之间的联合稀疏约束。
  10. 根据权利要求7所述的成像装置,其特征在于,所述更新模块在更新字典D时,对原子进行归一化处理,以避免字典原子尺度模糊。
  11. 根据权利要求7所述的成像装置,其特征在于,更新每一个通道的图像vj的模块进一步执行如下步骤:
    步骤d1:根据最小二乘法获得下式:
    Figure PCTCN2017114172-appb-100046
    式中,
    Figure PCTCN2017114172-appb-100047
    是一个对角矩阵,对角线上元素值等于对应位置图像像素重叠的块的个数;当假设这些块映射在图像的边缘,则对角线上元素值全部相等,
    Figure PCTCN2017114172-appb-100048
    特殊情况下,当填充的距离为单位距离1时,ω=M;
    步骤d2:图像域通过归一化全采傅里叶编码矩阵转换到傅里叶域,获得:
    Figure PCTCN2017114172-appb-100049
    式中,
    Figure PCTCN2017114172-appb-100050
    是0和1的对角矩阵,1代表K空间采到的点的位置,ω表示像素重复的个数,IN表示N*N的单位矩阵,
    Figure PCTCN2017114172-appb-100051
    表示第j-th通道零填充的傅里叶测量值;
    步骤d3:将步骤d2中的公式变换成:
    Figure PCTCN2017114172-appb-100052
    式中,Fvj(kx,ky)表示K空间位置(kx,ky)的更新值,
    Figure PCTCN2017114172-appb-100053
    表示第j-th通道的零填充K空间测量值;η表示一个小标量,Ω表示K空间已经采到的点的集合,
    Figure PCTCN2017114172-appb-100054
    步骤d4:通过步骤d3的公式(9)中频域中的数据通过逆傅里叶变换求得vj
  12. 根据权利要求7所述的成像装置,其特征在于,所述成像模块对所有通道的图像vj的平方进行求和,然后开方,获得更新后的图像。
  13. 一种计算机可读介质,该计算机可读介质具有存储在其中的程序,该程序是计算机可执行的以使计算机执行权利要求1-6中任一项所述的基于自适应联合稀疏编码的并行磁共振成像方法的各步骤。
PCT/CN2017/114172 2017-12-01 2017-12-01 一种基于自适应联合稀疏编码的并行磁共振成像方法、装置及计算机可读介质 WO2019104702A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
PCT/CN2017/114172 WO2019104702A1 (zh) 2017-12-01 2017-12-01 一种基于自适应联合稀疏编码的并行磁共振成像方法、装置及计算机可读介质
US16/760,956 US11668777B2 (en) 2017-12-01 2017-12-01 Adaptive joint sparse coding-based parallel magnetic resonance imaging method and apparatus and computer readable medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2017/114172 WO2019104702A1 (zh) 2017-12-01 2017-12-01 一种基于自适应联合稀疏编码的并行磁共振成像方法、装置及计算机可读介质

Publications (1)

Publication Number Publication Date
WO2019104702A1 true WO2019104702A1 (zh) 2019-06-06

Family

ID=66665359

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2017/114172 WO2019104702A1 (zh) 2017-12-01 2017-12-01 一种基于自适应联合稀疏编码的并行磁共振成像方法、装置及计算机可读介质

Country Status (2)

Country Link
US (1) US11668777B2 (zh)
WO (1) WO2019104702A1 (zh)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP7203287B1 (ja) * 2020-06-08 2023-01-12 グァンチョウ コンピューテーショナル スーパー-レゾリューション バイオテック カンパニー リミテッド 画像処理のためのシステム及び方法
CN116522269B (zh) * 2023-06-28 2023-09-19 厦门炬研电子科技有限公司 一种基于Lp范数非平稳信号稀疏重建的故障诊断方法

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103049923A (zh) * 2012-12-10 2013-04-17 深圳先进技术研究院 磁共振快速成像的方法
CN104376198A (zh) * 2014-11-04 2015-02-25 中国科学院深圳先进技术研究院 自适应磁共振并行成像方法和装置
CN104484886A (zh) * 2014-12-31 2015-04-01 深圳先进技术研究院 一种mr图像的分割方法及装置
CN105931242A (zh) * 2016-04-22 2016-09-07 河海大学 基于字典学习和时间梯度的动态核磁共振图像重建方法
EP3132742A1 (en) * 2014-04-30 2017-02-22 Samsung Electronics Co., Ltd. Magnetic resonance imaging device and method for generating magnetic resonance image
CN108154484A (zh) * 2017-12-01 2018-06-12 深圳先进技术研究院 一种基于自适应联合稀疏编码的并行磁共振成像方法、装置及计算机可读介质

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103049923A (zh) * 2012-12-10 2013-04-17 深圳先进技术研究院 磁共振快速成像的方法
EP3132742A1 (en) * 2014-04-30 2017-02-22 Samsung Electronics Co., Ltd. Magnetic resonance imaging device and method for generating magnetic resonance image
CN104376198A (zh) * 2014-11-04 2015-02-25 中国科学院深圳先进技术研究院 自适应磁共振并行成像方法和装置
CN104484886A (zh) * 2014-12-31 2015-04-01 深圳先进技术研究院 一种mr图像的分割方法及装置
CN105931242A (zh) * 2016-04-22 2016-09-07 河海大学 基于字典学习和时间梯度的动态核磁共振图像重建方法
CN108154484A (zh) * 2017-12-01 2018-06-12 深圳先进技术研究院 一种基于自适应联合稀疏编码的并行磁共振成像方法、装置及计算机可读介质

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
TAN, SHA ET AL.: "The Parallel Magnetic Resonance Reconstruction Method Based on Adaptive Sparse Representation", JOURNAL OF INTEGRATION TECHNOLOGY, vol. 5, no. 3, 15 May 2016 (2016-05-15) *

Also Published As

Publication number Publication date
US20200256942A1 (en) 2020-08-13
US11668777B2 (en) 2023-06-06

Similar Documents

Publication Publication Date Title
Tezcan et al. MR image reconstruction using deep density priors
US10902651B2 (en) Systems and methods for magnetic resonance image reconstruction
Pierre et al. Multiscale reconstruction for MR fingerprinting
US9734601B2 (en) Highly accelerated imaging and image reconstruction using adaptive sparsifying transforms
RU2568929C1 (ru) Способ и система для быстрой реконструкции изображения мрт из недосемплированных данных
US10638951B2 (en) Systems and methods for magnetic resonance imaging
US8335955B2 (en) System and method for signal reconstruction from incomplete data
US10690740B2 (en) Sparse reconstruction strategy for multi-level sampled MRI
US9430854B2 (en) System and method for model consistency constrained medical image reconstruction
US8942445B2 (en) Method and system for correction of lung density variation in positron emission tomography using magnetic resonance imaging
US9939509B2 (en) Variable density incoherent spatiotemporal acquisition (VISTA) for highly accelerated magnetic resonance imaging
WO2020114329A1 (zh) 磁共振快速参数成像方法及装置
WO2009098371A2 (fr) Procede de reconstruction d'un signal a partir de mesures experimentales perturbees et dispositif de mise en oeuvre
CN108154484B (zh) 一种基于自适应联合稀疏编码的并行磁共振成像方法、装置及计算机可读介质
US11867786B2 (en) Parameter map determination for time domain magnetic resonance
US11035919B2 (en) Image reconstruction using a colored noise model with magnetic resonance compressed sensing
CN108765514B (zh) 一种ct图像重建的加速方法及装置
CN112991483B (zh) 一种非局部低秩约束的自校准并行磁共振成像重构方法
US11776171B2 (en) Systems and methods for magnetic resonance image reconstruction
US9858689B1 (en) Fast and memory efficient redundant wavelet regularization with sequential cycle spinning
WO2019104702A1 (zh) 一种基于自适应联合稀疏编码的并行磁共振成像方法、装置及计算机可读介质
EP2791907B1 (fr) Procédé de reconstruction d'un signal en imagerie médicale à partir de mesures expérimentales perturbées, et dispositif d'imagerie médicale mettant en uvre ce procédé
CN117011673A (zh) 基于噪声扩散学习的电阻抗层析成像图像重建方法和装置
Guan et al. MRI reconstruction using deep energy-based model
Xu et al. Nonlocal low-rank and prior image-based reconstruction in a wavelet tight frame using limited-angle projection data

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 17933343

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 17933343

Country of ref document: EP

Kind code of ref document: A1

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 22.09.2020)

122 Ep: pct application non-entry in european phase

Ref document number: 17933343

Country of ref document: EP

Kind code of ref document: A1