CN103033782B - The method of parallel MR imaging device and imaging thereof - Google Patents

Abstract

A kind of parallel MR imaging device, comprises multiple imaging band, acquisition module, initialization module, computing module and rebuilds module.Wherein: acquisition module is used for owing to the utilization of each imaging band the mode of adopting and owes to adopt matrix d according to owing to adopt factor collection _i(i > 0, the numbering for imaging band); Initialization module is for obtaining the initializing sense degree matrix s of initialisation image matrix ρ and described imaging band _i(i > 0, the numbering for imaging band); Computing module is used for adopting matrix d according to described owing _i, initialisation image matrix ρ and initializing sense degree matrix s _iutilize conjugate gradient algorithm to be optimized iterative to constraint function, obtain rebuilding image array ρ and sensitivity matrix s _i; Rebuild module to be used for rebuilding image according to described reconstruction image array ρ and described sensitivity matrix si.By parallel MR imaging device of the present invention to image reconstruction, effectively raise image taking speed, and snr loss is little.

Description

The method of parallel MR imaging device and imaging thereof

Technical field

The present invention relates to medical skill, particularly relate to a kind of method of parallel MR imaging device and imaging thereof.

Background technology

MRI has become one of important means of clinical medicine inspection, for clinical medicine provides very valuable diagnostic message.MRI technology is compared with other Medical Imaging Technology, there is the advantages such as radiationless harm, multi-faceted and multiparameter imaging, it is very responsive to the inspection of soft tissue, can not only show the shape information of human anatomic structure, and can also reflect some Physiology and biochemistry information of tissue.But the image taking speed of MRI is slower, physiological motion in imaging process in examinee's health all can make image fog, contrast distortion, the requirement of the fast imagings such as heart dynamic imaging, cerebral function imaging, human motion imaging and cardiovascular and cerebrovascular cannot be met, therefore how high-resolutionly realize the key that FastMRI has become MRI technical development and application.

In the MRI investigation of nearest three more than ten years, researcher proposes a variety of fast imaging method.The fast imaging sequences in sampling time is such as shortened according to MRI principle, as Echo-plane imaging (EPI:Echo planar Imaging), screw propeller imaging, the gtadient echo read soon and fast spin-echo sequence etc.But in most cases, we adopt these fast imaging sequences to shorten the overall sampling time by gathering a small amount of data and combining, and then obtain complete K space data by subsequent reconstruction.Typically there are parallel MR imaging (Parallel MRI:pMRI), compressed sensing imaging (Compressed Sensing:CS) etc.Parallel imaging can be divided into again based on self-correcting parallel acquisition (the Generalized autocalibratingpartially parallel acquisitio ns:GRAPPA) algorithm of K spatial manipulation and encode (Sensitivity encoding Sense) imaging according to handkerchief Preece Generalized sampling theory based on the susceptibility of image area process, when we to the sensitivity matrix in Sense algorithm estimate comparatively accurately time, in parallel imaging methods, it comparatively GRAPPA can obtain better picture quality.But sensitivity is obtained by the pre-surface sweeping of some low resolution images or some separation, is difficult to like this obtain sensitivity matrix accurately, is therefore difficult to reconstruct high-quality image.The people such as Uecker propose the nonlinear iteration method (Joint estimation of coilsensitivities and image by nonlinear iterative methods:JSENSE) of coil sensitivity and image being carried out to Combined estimator, the method is a kind of effective ways obtaining high-quality Sense reconstruction, but along with owing the increase of adopting the factor, it is very fast that the signal to noise ratio (S/N ratio) of rebuilding image follows traditional Sense method equally to decline.Address this problem, have researcher some prior imformations to be incorporated in pMRI as bound term.

CS utilizes the openness MRI that improves of image to gather, and the method has been combined with pMRI at present and has accelerated MRI image taking speed faster.We are divided into three major types using this associated methods: a kind of is utilize coil sensitivity map to rebuild as the L1 norm restriction of prior imformation, such as sparseSENSE, in sparseSENSE, coil sensitivity angle value is rebuild the same with traditional Sense, first the prescanned data owing data or the separation of adopting is utilized to obtain sensitivity matrix, and then rebuild image in conjunction with CS sparse constraint, but, this method is the same with traditional Sense, sensitivity estimate inaccurate time, still there is larger artifact.The second is in GRAPPA, add CS sparse constraint, and this method utilizes prior acquisition data to obtain some weighted values to estimate not adopting data, but, in the middle of utilization, entirely adopt data go the non-image data of estimated edge, can not estimate very accurately.The third is as sparseSENSE, only simultaneously image and sensitivity matrix rebuild out, according to prior-constrained difference, such can be divided into a variety of method again, in reconstruction formula, have plenty of and use final image openness as prior-constrained, have plenty of and adopt coil sensitivity slickness or openness as prior-constrained, but the loss of image to-noise ratio is larger.

Summary of the invention

In view of this, be necessary to provide a kind of image acquisition fast and signal noise ratio (snr) of image loses little imaging device and formation method thereof.

A kind of parallel MR imaging device provided by the invention, comprises multiple imaging band, acquisition module, initialization module, computing module and rebuilds module.Wherein: acquisition module is used for owing to the utilization of each imaging band the mode of adopting and owes to adopt matrix d according to owing to adopt factor collection _i(i > 0, the numbering for imaging band); Initialization module is for obtaining the initializing sense degree matrix s of initialisation image matrix ρ and described imaging band _i(i > 0, the numbering for imaging band); Computing module is used for adopting matrix d according to described owing _i, initialisation image matrix ρ and initializing sense degree matrix s _iutilize conjugate gradient algorithm to be optimized iterative to constraint function, obtain rebuilding image array ρ and sensitivity matrix s _i; Rebuild module to be used for according to described reconstruction image array ρ and described sensitivity matrix s _irebuild image.

The present invention also provides a kind of formation method of parallel MR imaging device, and described parallel MR imaging device comprises multiple imaging band, comprises the following steps: owe to the utilization of each imaging band the mode of adopting and owe to adopt matrix d according to owing to adopt factor collection _i; Obtain the initializing sense degree matrix s of initialisation image matrix ρ and described imaging band _i; Matrix d is adopted according to described owing _i, initialisation image matrix ρ and initializing sense degree matrix s _iutilize conjugate gradient algorithm to be optimized iterative to constraint function, obtain rebuilding image array ρ and sensitivity matrix s _i; Image s is rebuild according to described reconstruction image array ρ and sensitivity matrix _i.

Parallel MR imaging device in the present invention and formation method thereof are owed the mode of adopting collection by utilization and are owed to adopt matrix, and obtain initialisation image matrix ρ and initializing sense degree matrix s _iafter utilize conjugate gradient algorithm to be optimized iterative to constraint function, to realize, to the reconstruction of image, effectively raise image taking speed, and snr loss being little.

Accompanying drawing explanation

Fig. 1 is the module map of parallel MR imaging device in an embodiment of the present invention;

Fig. 2 is for utilizing the process flow diagram of the method for the parallel MR imaging device imaging shown in Fig. 1 in an embodiment of the present invention;

Fig. 3 is original image;

Fig. 4 A is the image after utilizing sparseSENSE to rebuild Fig. 3;

Fig. 4 B is the image after utilizing the method in the present invention to rebuild Fig. 3.

Embodiment

Be described below in detail embodiments of the invention, the example of described embodiment is shown in the drawings, and wherein same or similar label represents same or similar element or has element that is identical or similar functions from start to finish.Being exemplary below by the embodiment be described with reference to the drawings, only for explaining the present invention, and can not limitation of the present invention being interpreted as.

In describing the invention, term " interior ", " outward ", " longitudinal direction ", " transverse direction ", " on ", D score, " top ", the orientation of the instruction such as " end " or position relationship be based on orientation shown in the drawings or position relationship, be only the present invention for convenience of description instead of require that the present invention with specific azimuth configuration and operation, therefore must can not be interpreted as limitation of the present invention.

Refer to Fig. 1, Figure 1 shows that the module map of parallel MR imaging device 10 in an embodiment of the present invention.

In the present embodiment, parallel MR imaging device 10 comprises: acquisition module 102, initialization module 104, computing module 106, reconstruction module 108, processor 110 and storer 112, arranging module 102, acquisition module 104, judge module 106 and splitting module 108 is stored in storer 112, and processor 110 is for the modules in execute store 112.

In the present embodiment, parallel MR imaging device 10 also comprises multiple imaging band (not shown), and wherein, each imaging band is made up of coil.

In the present embodiment, acquisition module 102 is owed to adopt matrix di (i > 0, the numbering for imaging band) according to owing to adopt factor collection for owing to the utilization of each imaging band the mode of adopting.In the present embodiment, each channel acquisition deficient size of adopting matrix di is out the same, but deficient just full benefit of adopting no collection in matrix di is out zero.

In the present embodiment, the described mode of adopting of owing refers to and all gathers center section in described imaging band, the random lack sampling of peripheral part interlacing collection or both sides or radial sparse sampling or spiral sparse sampling.

Initialization module 104 is for obtaining the initializing sense degree matrix s of initialisation image matrix ρ and described imaging band _i(i > 0, the numbering for imaging band).

In the present embodiment, described initializing sense degree matrix s _i, described initialisation image matrix ρ number identical with described port number.

In the present embodiment, described initialisation image matrix ρ to be value be entirely 1 matrix.

In the present embodiment, initializing sense degree matrix s _ifor be worth be entirely 0 matrix.

Computing module 106 is for adopting matrix d according to described owing _i, initialisation image matrix ρ and initializing sense degree matrix s _iutilize conjugate gradient algorithm to be optimized iterative to constraint function, obtain rebuilding image array ρ and sensitivity matrix s _i.

In the present embodiment, described computing module 106 utilizes the solving model of conjugate gradient algorithm $\underset{\mathrm{\ρ},{\hat{s}}_1,\·\·\·,{\hat{s}}_n}{\arg \min }{\left|\right|\left(\begin{array}{c}{d}_1\\ d_2\\ .\\ .\\ .\\ d_n\end{array}\right)-\left(\begin{array}{c}\mathrm{PF\ρ}.M{\hat{s}}_1\\ \mathrm{PF\ρ}.M{\hat{s}}_2\\ .\\ .\\ .\\ \mathrm{PF\ρ}.M{\hat{s}}_n\end{array}\right)\left|\right|}_2+\mathrm{\λ}{\mathrm{\Σ}}_{i=1}^n{|\mathrm{W\ρ}.M{\hat{s}}_i|}_1$ Iterative, wherein, F is that Fourier changes, and P owes to adopt the factor, and W is sparse change operator, and described M Suo Bailiefu figures son as smoothing operator, for: $M=(1/{(1+|\left|\stackrel{\→}{k}\right|\left|\right)}^{1/2}).$

In other embodiments of the present invention, described computing module 106 also can utilize the solving model of conjugate gradient algorithm $\underset{\mathrm{\ρ},{\hat{s}}_1,\·\·\·,{\hat{s}}_n}{\arg \min }{\left|\right|\left(\begin{array}{c}{d}_1\\ d_2\\ .\\ .\\ .\\ d_n\end{array}\right)-\left(\begin{array}{c}\mathrm{PF\ρ}.M{\hat{s}}_1\\ \mathrm{PF\ρ}.M{\hat{s}}_2\\ .\\ .\\ .\\ \mathrm{PF\ρ}.M{\hat{s}}_n\end{array}\right)\left|\right|}_2+{\mathrm{\λ}}_1{\mathrm{\Σ}}_{i=1}^n{|\mathrm{W\ρ}.M{\hat{s}}_i|}_1+$ ${\mathrm{\λ}}_2{\mathrm{\Σ}}_{i=1}^n\mathrm{TV}(\mathrm{\ρ}.M{\hat{s}}_i)$ Carry out iterative.

Rebuild module 108 for rebuilding image according to described reconstruction image array ρ and described sensitivity matrix si.

Refer to Fig. 2, Figure 2 shows that in an embodiment of the present invention the process flow diagram of the method utilizing parallel MR imaging device 10 imaging shown in Fig. 1.

In step S200, acquisition module 102 is owed to the utilization of each imaging band the mode of adopting and is owed to adopt matrix according to owing to adopt factor collection.In the present embodiment, each channel acquisition deficient size of adopting matrix di is out the same, but deficient just full benefit of adopting no collection in matrix di is out zero.

In the present embodiment, the described mode of adopting of owing refers to and all gathers center section in described imaging band, the random lack sampling of peripheral part interlacing collection or both sides or radial sparse sampling or spiral sparse sampling.

In step S210, initialization module 104 obtains the initializing sense degree matrix s of initialisation image matrix ρ and described imaging band _i.

In the present embodiment, described initializing sense degree matrix s _i, described initialisation image matrix ρ number identical with described port number.

In the present embodiment, described initialisation image matrix ρ to be value be entirely 1 matrix.

In the present embodiment, initializing sense degree matrix s _ifor be worth be entirely 0 matrix.

In step S220, computing module 106 adopts matrix d according to described owing _i, initialisation image matrix ρ and initializing sense degree matrix s _iutilize conjugate gradient algorithm to be optimized iterative to constraint function, obtain rebuilding image array ρ and sensitivity matrix s _i.

In the present embodiment, described computing module 106 utilizes the solving model of conjugate gradient algorithm $\underset{\mathrm{\ρ},{\hat{s}}_1,\·\·\·,{\hat{s}}_n}{\arg \min }{\left|\right|\left(\begin{array}{c}{d}_1\\ d_2\\ .\\ .\\ .\\ d_n\end{array}\right)-\left(\begin{array}{c}\mathrm{PF\ρ}.M{\hat{s}}_1\\ \mathrm{PF\ρ}.M{\hat{s}}_2\\ .\\ .\\ .\\ \mathrm{PF\ρ}.M{\hat{s}}_n\end{array}\right)\left|\right|}_2+\mathrm{\λ}{\mathrm{\Σ}}_{i=1}^n{|\mathrm{W\ρ}.M{\hat{s}}_i|}_1$ Iterative, wherein, F is that Fourier changes, and P owes to adopt the factor, and W is sparse change operator, and described M Suo Bailiefu figures son as smoothing operator, for: $M=(1/{(1+|\left|\stackrel{\→}{k}\right|\left|\right)}^{1/2}).$

In other embodiments of the present invention, described computing module 106 also can utilize the solving model of conjugate gradient algorithm $\underset{\mathrm{\ρ},{\hat{s}}_1,\·\·\·,{\hat{s}}_n}{\arg \min }{\left|\right|\left(\begin{array}{c}{d}_1\\ d_2\\ .\\ .\\ .\\ d_n\end{array}\right)-\left(\begin{array}{c}\mathrm{PF\ρ}.M{\hat{s}}_1\\ \mathrm{PF\ρ}.M{\hat{s}}_2\\ .\\ .\\ .\\ \mathrm{PF\ρ}.M{\hat{s}}_n\end{array}\right)\left|\right|}_2+{\mathrm{\λ}}_1{\mathrm{\Σ}}_{i=1}^n{|\mathrm{W\ρ}.M{\hat{s}}_i|}_1+$ ${\mathrm{\λ}}_2{\mathrm{\Σ}}_{i=1}^n\mathrm{TV}(\mathrm{\ρ}.M{\hat{s}}_i)$ Carry out iterative.

In step S230, rebuild module 108 and rebuild image s according to described reconstruction image array ρ and sensitivity matrix _i.

Refer to Fig. 3, Fig. 4 A and Fig. 4 B, wherein, Figure 3 shows that original image, Fig. 4 A is depicted as the image after utilizing sparseSENSE to rebuild Fig. 3, the method in invention shown in Fig. 4 B Fig. 3 is rebuild after image.

In the present embodiment, the original image shown in Fig. 3 is the Shepp-Logan image of 256 × 256 sizes of Noise.

In Figure 4 A, employing owes to adopt factor R is successively 3,5,7,9, utilizes sparseSENSE original image to be rebuild to the reconstruction image obtained.

In figure 4b, employing owes to adopt factor R is successively 3,5,7,9, utilizes the method in the present invention original image to be rebuild to the reconstruction image obtained.

Following table is depicted as the AP value deck watch rebuilding image:

Owe to adopt factor R	3	5	7	9
					The inventive method	0.045796	0.050110	0.057300	0.061012
sparseSENSE	0.100678	0.114854	0.121454	0.131453

Therefore can find out that the image utilizing sparseSENSE method to rebuild out is relative to the method utilized in the present invention, create larger artifact, and the method for parallel MR imaging device 10 provided by the present invention imaging has carried out smoothing denoising to image, effect clearly, by the comparison of the image AP value in table, also can find out, the method for parallel MR imaging device 10 imaging that provides is provided can reconstruct high signal-to-noise ratio image in situation fast.

Parallel MR imaging device 10 in embodiment of the present invention and formation method thereof are owed the mode of adopting collection by utilization and are owed to adopt matrix, and obtain initialisation image matrix ρ and initializing sense degree matrix s _iafter utilize conjugate gradient algorithm to be optimized iterative to constraint function, to realize, to the reconstruction of image, effectively raise image taking speed, and snr loss being little.

Although the present invention is described with reference to current better embodiment; but those skilled in the art will be understood that; above-mentioned better embodiment is only used for the present invention is described; not be used for limiting protection scope of the present invention; any within the spirit and principles in the present invention scope; any modification of doing, equivalence replacement, improvement etc., all should be included within the scope of the present invention.

Claims (10)

1. a parallel MR imaging device, comprises multiple imaging band, it is characterized in that, also comprise:

Acquisition module, owes to adopt matrix d according to owing to adopt factor collection for owing to the utilization of each imaging band the mode of adopting _i, wherein i>0 is the numbering of imaging band;

Initialization module, for obtaining the initializing sense degree matrix s of initialisation image matrix ρ and described imaging band _i, wherein i>0 is the numbering of imaging band;

Computing module, for adopting matrix d according to described owing _i, initialisation image matrix ρ and initializing sense degree matrix s _iutilize conjugate gradient algorithm to be optimized iterative to constraint function, obtain rebuilding image array ρ and sensitivity matrix s _i, wherein said computing module utilizes the solving model of conjugate gradient algorithm $\underset{\mathrm{\ρ},{\hat{s}}_1,...,{\hat{s}}_n}{\arg \min }{\left|\right|\left(\begin{array}{c}{d}_1\\ d_2\\ .\\ .\\ .\\ d_n\end{array}\right)-\left(\begin{array}{c}\mathrm{PF\ρ}.M{\hat{s}}_1\\ \mathrm{PF\ρ}.M{\hat{s}}_2\\ .\\ .\\ .\\ \mathrm{PF\ρ}.M{\hat{s}}_n\end{array}\right)\left|\right|}_2+\mathrm{\λ}{\mathrm{\Σ}}_{i=1}^n{|\mathrm{W\ρ}.M{\hat{s}}_i|}_1$ Iterative, described M Suo Bailiefu figures son as smoothing operator, for:

Rebuild module, for according to described reconstruction image array ρ and described sensitivity matrix s _irebuild image.

2. parallel MR imaging device as claimed in claim 1, is characterized in that, the described mode of adopting of owing refers to and all gathers center section in described imaging band, and peripheral part interlacing gathers.

3. parallel MR imaging device as claimed in claim 1, is characterized in that, described initializing sense degree matrix s _i, described initialisation image matrix ρ number identical with described port number.

4. parallel MR imaging device as claimed in claim 1, is characterized in that, described initialisation image matrix ρ to be value be entirely 1 matrix.

5. parallel MR imaging device as claimed in claim 1, is characterized in that, initializing sense degree matrix s _ifor be worth be entirely 0 matrix.

6. a formation method for parallel MR imaging device, described parallel MR imaging device comprises multiple imaging band, comprising:

Owe to the utilization of each imaging band the mode of adopting to owe to adopt matrix d according to owing to adopt factor collection _i;

Obtain the initializing sense degree matrix s of initialisation image matrix ρ and described imaging band _i;

Matrix d is adopted according to described owing _i, described initialisation image matrix ρ and described initializing sense degree matrix s _iutilize conjugate gradient algorithm to be optimized iterative to constraint function, obtain rebuilding image array ρ and sensitivity matrix s _i, wherein utilize the solving model of conjugate gradient algorithm $\underset{\mathrm{\ρ},{\hat{s}}_1,...,{\hat{s}}_n}{\arg \min }{\left|\right|\left(\begin{array}{c}{d}_1\\ d_2\\ .\\ .\\ .\\ d_n\end{array}\right)-\left(\begin{array}{c}\mathrm{PF\ρ}.M{\hat{s}}_1\\ \mathrm{PF\ρ}.M{\hat{s}}_2\\ .\\ .\\ .\\ \mathrm{PF\ρ}.M{\hat{s}}_n\end{array}\right)\left|\right|}_2+\mathrm{\λ}{\mathrm{\Σ}}_{i=1}^n{|\mathrm{W\ρ}.M{\hat{s}}_i|}_1$ Iterative, described M Suo Bailiefu figures son as smoothing operator, for:

Image s is rebuild according to described reconstruction image array ρ and sensitivity matrix _i.

7. method as claimed in claim 6, is characterized in that, the described mode of adopting of owing refers to and all gathers center section in described imaging band, and peripheral part interlacing gathers.

8. method as claimed in claim 6, is characterized in that, described initializing sense degree matrix s _i, described initialisation image matrix ρ number identical with described port number.

9. method as claimed in claim 6, is characterized in that, described initialisation image matrix ρ to be value be entirely 1 matrix.

10. method as claimed in claim 6, is characterized in that, initializing sense degree matrix s _ifor be worth be entirely 0 matrix.