CN104569880B - A kind of magnetic resonance fast imaging method and system - Google Patents

Abstract

The present invention proposes a kind of magnetic resonance fast imaging method and system, wherein, the method includes：Obtain the MR data of lack sampling；According to the MR data of lack sampling, using multiple convolution kernel functions, the K space data corresponding to various magnetic resonance image (MRI) features to be asked is calculated；K space data according to corresponding to various magnetic resonance image (MRI) features to be asked, using compressed sensing reconstruction magnetic resonance image (MRI) feature, obtains the characteristics of image rebuild；The characteristics of image of reconstruction is merged, the magnetic resonance image (MRI) rebuild is generated.Magnetic resonance fast imaging method proposed by the present invention and system are first with the characteristics of image corresponding to the multiple convolution kernel functions of compressed sensing reconstruction, these characteristics of image are more sparse, or with higher signal to noise ratio, thus be more conducive to be rebuild with compressed sensing technology；Finally, the merging of characteristics of image can further suppress the noise and artifact in the characteristics of image rebuild, and then realize high-quality image reconstruction.

Description

A kind of magnetic resonance fast imaging method and system

Technical field

The present invention relates to mr imaging technique field, more particularly to a kind of magnetic resonance fast imaging method and system.

Background technology

In recent years, asked based on the focus that the MR imaging method of compressed sensing technology is FastMRI field Topic.Compressed sensing technology mainly uses the openness to realize that lack sampling is rebuild of signal.

The physical mechanism of nuclear magnetic resonance can be expressed as：

D=F ρ+n (formula 1)

Wherein, d represents the MR data gathered on magnetic resonance device, and F represents Fourier-encoded matrix, and ρ is to wait to ask Magnetic resonance image (MRI), n is generally assumed to be white complex gaussian noise.Compressive sensing theory is thought, if signal is in a certain transform domain It is sparse, then by a kind of and noncoherent sampling configuration of the transform domain, it is possible to accurate from a small amount of sampled data Reconstruct real signal.Also, signal is more sparse, the sampled data of needs is fewer.Magnetic based on compressed sensing technology Resonant imaging method can be expressed as：

Wherein,The nuclear magnetic resonance based on compressed sensing reconstruction is represented, ρ represents magnetic resonance image (MRI) to be asked, and y is k The lack sampling data in space, F_uFor the Fourier-encoded matrix of lack sampling, Ψ is sparse transformation matrix (such as wavelet transformation), λ For regularization coefficient, | | | |₁L1 norms are sought in expression, | | | |₂L2 norms are sought in expression.

However, medical magnetic resonance image is often simply highly compressible, and and non-critical is sparse.It is this it is openness not Noise in being enough to and sampling produces serious artifact in causing the image rebuild, and reduces the image quality of image, so as to Limit compression and know practical application of the sense technology in accelerating magnetic resonance imaging.

The content of the invention

Based on the mr imaging technique of compressed sensing, producing ratio is understood in image sparse deficiency and in the case of making an uproar More serious image artifacts.For foregoing problems, the present invention extracts the feature of image, these figures using a series of convolution kernel functions As feature need it is more sparse than magnetic resonance image (MRI) or with higher signal to noise ratio.Specifically, the method is first with compressed sensing Reconstruction these characteristics of image, then the characteristics of image that these reconstructions are obtained is merged, to generate the image of reconstruction.

To reach above-mentioned purpose, the present invention proposes a kind of magnetic resonance fast imaging method, including：Obtain the magnetic of lack sampling Resonance data；According to the MR data of the lack sampling, using multiple convolution kernel functions, various magnetic resonance figures to be asked are calculated K space data as corresponding to feature；K space data according to corresponding to various magnetic resonance image (MRI) features to be asked, profit With compressed sensing reconstruction magnetic resonance image (MRI) feature, the characteristics of image rebuild is obtained；The characteristics of image of the reconstruction is carried out Merge, generate the magnetic resonance image (MRI) rebuild.

To reach above-mentioned purpose, the invention allows for a kind of magnetic resonance fast imaging system, including：MR data is obtained Delivery block, for obtaining the MR data of lack sampling；K space data computing module, for common according to the magnetic of the lack sampling Shake data, using multiple convolution kernel functions, calculate various corresponding K space datas of magnetic resonance image (MRI) feature to be asked；Image is special Reconstruction module is levied, for the K space data according to corresponding to various magnetic resonance image (MRI) features to be asked, using compressed sensing Reconstruction magnetic resonance image (MRI) feature, obtains the characteristics of image rebuild；Reconstruction image generation module, for by the figure of the reconstruction As feature is merged, the magnetic resonance image (MRI) rebuild is generated.

Magnetic resonance fast imaging method proposed by the present invention and system are first with the multiple convolution of compressed sensing reconstruction Characteristics of image corresponding to kernel function, these characteristics of image or more sparse, or with higher signal to noise ratio, thus be more conducive to Rebuild with compressed sensing technology；Finally, during the merging of characteristics of image can further suppress the characteristics of image rebuild Noise and artifact, and then realize high-quality image reconstruction.

Description of the drawings

Accompanying drawing described herein is used for providing a further understanding of the present invention, constitutes the part of the application, not Constitute limitation of the invention.In the accompanying drawings：

Magnetic resonance fast imaging method flow charts of the Fig. 1 for one embodiment of the invention.

Structural representations of the Fig. 2 for the magnetic resonance fast imaging system of one embodiment of the invention.

Specific embodiment

Hereinafter coordinate schema and presently preferred embodiments of the present invention, the present invention is expanded on further to reach predetermined goal of the invention institute The technological means taken.

The present invention proposes a kind of magnetic resonance fast imaging method and system, compares the compression sense of more existing utilization standard The imaging effect of the mr imaging technique known, the present invention are right first with the multiple convolution kernel function institutes of compressed sensing reconstruction The characteristics of image answered；Again the characteristics of image of reconstruction is merged, the magnetic resonance image (MRI) rebuild is obtained, and then is realized high-quality Image reconstruction.

Magnetic resonance fast imaging method flow charts of the Fig. 1 for one embodiment of the invention.As shown in figure 1, the method includes：

Step 1, obtains the MR data of lack sampling；

Step 2, according to the MR data of lack sampling, using multiple convolution kernel functions, calculates various magnetic resonance to be asked K space data corresponding to characteristics of image；

Step 3, the K space data according to corresponding to various magnetic resonance image (MRI) features to be asked, using compressed sensing technology Magnetic resonance image (MRI) feature is rebuild, the characteristics of image rebuild is obtained；

Step 4, the characteristics of image of reconstruction is merged, and generates the magnetic resonance image (MRI) rebuild.

Further, in step 2, the K space data corresponding to various magnetic resonance image (MRI) features to be asked is：

y₁=F { K₁}⊙y

y₂=F { K₂}⊙y

y_p=F { K_p}⊙y

Wherein, y₁、y₂、…、y_pRepresent the K space data corresponding to various magnetic resonance image (MRI) features to be asked, K₁、K₂、…、 K_pMultiple convolution kernel functions are represented, ⊙ is accorded with for point multiplication operation, F { } represents Fourier transform, magnetic resonance numbers of the y for lack sampling According to numbers of the p for convolution kernel function.

Wherein, convolution kernel function can be edge extracting kernel function or gaussian kernel function.Using edge extracting kernel function weight The characteristics of image built can be more sparse, has higher signal to noise ratio using the characteristics of image that gaussian kernel function is rebuild.Therefore, compare For the method for directly reconstructing magnetic resonance image (MRI) of background technology formula 2, the present invention can be more smart using compressed sensing technology True reconstructs these characteristics of image.

In step 3, based on compressed sensing reconstruction magnetic resonance image (MRI) character representation it is：

Wherein,Represent the characteristics of image rebuild, ρ₁、ρ₂、…、ρ_pRepresent magnetic resonance image (MRI) to be asked Feature, F_uFor the Fourier-encoded matrix of lack sampling, λ₁、λ₂、…、λ_pFor regularization coefficient, Ψ represents sparse transformation matrix, | |·||₁L1 norms are sought in expression, | | | |₂L2 norms are sought in expression；

Wherein, magnetic resonance image (MRI) feature is expressed as with the relation of magnetic resonance image (MRI)：

ρ₁=K₁*ρ

ρ₂=K₂*ρ

ρ_p=K_p*ρ

Wherein, * is convolution operator, and ρ is magnetic resonance image (MRI) to be asked.

In step 4, mainly the reconstruction image feature in step 3 is merged, generates the magnetic resonance image (MRI) rebuild. Wherein, the merging method for utilizing in the present embodiment is the method using regularization constraint, i.e., complete reconstruction image is by with lower section Formula is obtained：

Wherein,For the magnetic resonance image (MRI) rebuild, function R () is for constraining the openness of magnetic resonance image (MRI), γ₁For control Make the openness regularization parameter of magnetic resonance image (MRI) to be asked, γ₂For the figure for controlling magnetic resonance image (MRI) feature to be asked with rebuild As the regularization parameter of similarity between feature, α₁、α₂、…、α_pRepresent the weight of each magnetic resonance image (MRI) feature.

Function R () for constraining the openness of magnetic resonance image (MRI), wherein, R (ρ)=| | Ψ ρ | |₁。

In other embodiments, the merging method for utilizing in step 4 is not limited to that, be given in above-described embodiment only It is one of optimized method.

Just reconstruction image can be obtained by above-mentioned steps 1-4, the quality of the reconstruction image will be much better than directly using pressure The image that contracting cognition technology (formula 2) is obtained.

Based on same inventive concept, a kind of magnetic resonance fast imaging system in the embodiment of the present invention, is additionally provided, it is such as following Embodiment.Due to the principle of the system solve problem it is similar to said method, therefore the enforcement of the system may refer to it is above-mentioned The enforcement of method, repeats part and repeats no more.Used below, term " unit " or " module " can realize predetermined function Software and/or hardware combination.Although the device described by following examples is preferably with software realizing, hardware, Or the realization of the combination of software and hardware is also what is may and be contemplated.

Structural representations of the Fig. 2 for the magnetic resonance fast imaging system of one embodiment of the invention.As shown in Fig. 2 the system Including：

MR data acquisition module 1, for obtaining the MR data of lack sampling；

K space data computing module 2, for the MR data according to lack sampling, using multiple convolution kernel functions, calculates K space data corresponding to various magnetic resonance image (MRI) features to be asked；

Characteristics of image rebuilds module 3, for the K space data according to corresponding to various magnetic resonance image (MRI) features to be asked, Using compressed sensing reconstruction magnetic resonance image (MRI) feature, the characteristics of image rebuild is obtained；

Reconstruction image generation module 4, for the characteristics of image of reconstruction is merged, generates the magnetic resonance image (MRI) rebuild.

In K space data computing module 2, the K space data corresponding to various magnetic resonance image (MRI) features to be asked is：

y₁=F { K₁}⊙y

y₂=F { K₂}⊙y

y_p=F { K_p}⊙y

Wherein, y₁、y₂、…、y_pRepresent the K space data corresponding to various magnetic resonance image (MRI) features to be asked, K₁、K₂、…、 K_pMultiple convolution kernel functions are represented, ⊙ is accorded with for point multiplication operation, F { } represents Fourier transform, magnetic resonance numbers of the y for lack sampling According to numbers of the p for convolution kernel function.

Characteristics of image is rebuild in module 3, based on compressed sensing reconstruction magnetic resonance image (MRI) character representation is：

Wherein,Represent the characteristics of image rebuild, ρ₁、ρ₂、…、ρ_pRepresent magnetic resonance image (MRI) to be asked Feature, F_uFor the Fourier-encoded matrix of lack sampling, λ₁、λ₂、…、λ_pFor regularization coefficient, Ψ represents sparse transformation matrix, | |·||₁L1 norms are sought in expression, | | | |₂L2 norms are sought in expression；

Wherein, magnetic resonance image (MRI) feature is expressed as with the relation of magnetic resonance image (MRI)：

ρ₁=K₁*ρ

ρ₂=K₂*ρ

ρ_p=K_p*ρ

Wherein, * is convolution operator, and ρ is magnetic resonance image (MRI) to be asked.

In reconstruction image generation module 4, complete reconstruction image is obtained by the following manner：

Wherein,For the magnetic resonance image (MRI) rebuild, function R () is for constraining the openness of magnetic resonance image (MRI), γ₁For control Make the openness regularization parameter of magnetic resonance image (MRI) to be asked, γ₂For the figure for controlling magnetic resonance image (MRI) feature to be asked with rebuild As the regularization parameter of similarity between feature, α₁、α₂、…、α_pThe weight of each magnetic resonance image (MRI) feature is represented respectively.

Function R () for constraining the openness of magnetic resonance image (MRI), wherein, R (ρ)=| | Ψ ρ | |₁。

Magnetic resonance fast imaging method proposed by the present invention and system are first with the multiple convolution of compressed sensing reconstruction Characteristics of image corresponding to kernel function, these characteristics of image or more sparse, or with higher signal to noise ratio, thus be more conducive to Rebuild with compressed sensing technology；Finally, during the merging of characteristics of image can further suppress the characteristics of image rebuild Noise and artifact, and then realize high-quality image reconstruction.

Particular embodiments described above, has been carried out to the purpose of the present invention, technical scheme and beneficial effect further in detail Describe bright, the be should be understood that specific embodiment that the foregoing is only the present invention, the guarantor being not intended to limit the present invention in detail Shield scope, all any modification, equivalent substitution and improvements within the spirit and principles in the present invention, done etc., should be included in this Within the protection domain of invention.

Claims (12)

1. a kind of magnetic resonance fast imaging method, it is characterised in that include：

Obtain the MR data of lack sampling；

According to the MR data of the lack sampling, using multiple convolution kernel functions, various magnetic resonance image (MRI) to be asked are calculated special Levy corresponding K space data；

K space data according to corresponding to various magnetic resonance image (MRI) features to be asked, using compressed sensing reconstruction magnetic Resonance image feature, obtains the characteristics of image rebuild；

The characteristics of image of the reconstruction is merged, the magnetic resonance image (MRI) rebuild is generated.

2. magnetic resonance fast imaging method as claimed in claim 1, it is characterised in that according to the magnetic resonance number of the lack sampling According to, using multiple convolution kernel functions, the K space data being calculated corresponding to various magnetic resonance image (MRI) features to be asked, including：

K space data corresponding to various magnetic resonance image (MRI) features to be asked is：

y₁=F { K₁}⊙y

y₂=F { K₂}⊙y

y_p=F { K_p}⊙y

Wherein, y₁、y₂、…、y_pRepresent the K space data corresponding to various magnetic resonance image (MRI) features to be asked, K₁、K₂、…、K_pTable Show multiple convolution kernel functions, ⊙ is accorded with for point multiplication operation, F { } represents Fourier transform, MR data of the y for lack sampling, p is The number of convolution kernel function.

3. magnetic resonance fast imaging method as claimed in claim 2, it is characterised in that the convolution kernel function is edge extracting Kernel function or gaussian kernel function.

4. magnetic resonance fast imaging method as claimed in claim 2, it is characterised in that according to various magnetic resonance to be asked K space data corresponding to characteristics of image, using compressed sensing reconstruction magnetic resonance image (MRI) feature, obtains the image rebuild special Levy, including：

Based on compressed sensing reconstruction magnetic resonance image (MRI) character representation it is：

$\begin{array}{c}{\widehat{\mathrm{\ρ}}}_1=\underset{{\mathrm{\ρ}}_1}{\arg \min }\left|\right|y_1-F_u{\mathrm{\ρ}}_1|{|}_2^2+{\mathrm{\λ}}_1|\left|{\mathrm{\Ψ\ρ}}_1\right|{|}_1\\ {\widehat{\mathrm{\ρ}}}_2=\underset{{\mathrm{\ρ}}_2}{\arg \min }\left|\right|y_2-F_u{\mathrm{\ρ}}_2|{|}_2^2+{\mathrm{\λ}}_2|\left|{\mathrm{\Ψ\ρ}}_2\right|{|}_1\\ .\\ .\\ .\\ {\widehat{\mathrm{\ρ}}}_p=\underset{{\mathrm{\ρ}}_p}{\arg \min }\left|\right|y_p-F_u{\mathrm{\ρ}}_p|{|}_2^2+{\mathrm{\λ}}_p|\left|{\mathrm{\Ψ\ρ}}_p\right|{|}_1\end{array}$

Wherein,Represent the characteristics of image rebuild, ρ₁、ρ₂、…、ρ_pMagnetic resonance image (MRI) feature to be asked is represented, F_uFor the Fourier-encoded matrix of lack sampling, λ₁、λ₂、…、λ_pFor regularization coefficient, Ψ represents sparse transformation matrix, | | | |₁Table Show and seek L1 norms, | | | |₂L2 norms are sought in expression；

Wherein, magnetic resonance image (MRI) feature is expressed as with the relation of magnetic resonance image (MRI)：

ρ₁=K₁*ρ

ρ₂=K₂*ρ

ρ_p=K_p*ρ

Wherein, * is convolution operator, and ρ is magnetic resonance image (MRI) to be asked.

5. magnetic resonance fast imaging method as claimed in claim 4, it is characterised in that the characteristics of image of the reconstruction is carried out Merge, generate the magnetic resonance image (MRI) rebuild, including：

The magnetic resonance image (MRI) of reconstruction is obtained by the following manner：

$\begin{array}{c}\widehat{\mathrm{\ρ}}=\underset{\mathrm{\ρ}}{\arg \min }\left|\right|y-F_u\mathrm{\ρ}|{|}_2^2+{\mathrm{\γ}}_{1}R\left(\mathrm{\ρ}\right)+\\ {\mathrm{\γ}}_1\left({\mathrm{\α}}_1\left|\right|K_1*\mathrm{\ρ}-{\widehat{\mathrm{\ρ}}}_1|{|}_2^2+{\mathrm{\α}}_2||K_2*\mathrm{\ρ}-{\widehat{\mathrm{\ρ}}}_2|{|}_2^2+\mathrm{...}+{\mathrm{\α}}_p\left|\right|K_p*\mathrm{\ρ}-{\widehat{\mathrm{\ρ}}}_p|{|}_2^2\right)\end{array}$

Wherein,For the magnetic resonance image (MRI) rebuild, function R () is for constraining the openness of magnetic resonance image (MRI), γ₁Treat for control The openness regularization parameter of the magnetic resonance image (MRI) asked, γ₂Image to control magnetic resonance image (MRI) feature to be asked with rebuild is special The regularization parameter of similarity, α between levying₁、α₂、…、α_pThe weight of each magnetic resonance image (MRI) feature is represented respectively.

6. magnetic resonance fast imaging method as claimed in claim 5, it is characterised in that the function R () is for constraining magnetic Openness, the R (ρ)=| | Ψ ρ | | of resonance image₁。

7. a kind of magnetic resonance fast imaging system, it is characterised in that include：

MR data acquisition module, for obtaining the MR data of lack sampling；

K space data computing module, for the MR data according to the lack sampling, using multiple convolution kernel functions, calculates Various corresponding K space datas of magnetic resonance image (MRI) feature to be asked；

Characteristics of image rebuilds module, for the K space data according to corresponding to various magnetic resonance image (MRI) features to be asked, profit With compressed sensing reconstruction magnetic resonance image (MRI) feature, the characteristics of image rebuild is obtained；

Reconstruction image generation module, for the characteristics of image of the reconstruction is merged, generates the magnetic resonance image (MRI) rebuild.

8. magnetic resonance fast imaging system as claimed in claim 7, it is characterised in that in the K space data computing module, K space data corresponding to various magnetic resonance image (MRI) features to be asked is：

y₁=F { K₁}⊙y

y₂=F { K₂}⊙y

y_p=F { K_p}⊙y

Wherein, y₁、y₂、…、y_pRepresent the K space data corresponding to various magnetic resonance image (MRI) features to be asked, K₁、K₂、…、K_pTable Show multiple convolution kernel functions, ⊙ is accorded with for point multiplication operation, F { } represents Fourier transform, MR data of the y for lack sampling, p is The number of convolution kernel function.

9. magnetic resonance fast imaging system as claimed in claim 8, it is characterised in that the convolution kernel function is edge extracting Kernel function or gaussian kernel function.

10. magnetic resonance fast imaging system as claimed in claim 8, it is characterised in that in described image feature reconstruction module, Using compressed sensing reconstruction magnetic resonance image (MRI) character representation it is：

$\begin{array}{c}{\widehat{\mathrm{\ρ}}}_1=\underset{{\mathrm{\ρ}}_1}{\arg \min }\left|\right|y_1-F_u{\mathrm{\ρ}}_1|{|}_2^2+{\mathrm{\λ}}_1|\left|{\mathrm{\Ψ\ρ}}_1\right|{|}_1\\ {\widehat{\mathrm{\ρ}}}_2=\underset{{\mathrm{\ρ}}_2}{\arg \min }\left|\right|y_2-F_u{\mathrm{\ρ}}_2|{|}_2^2+{\mathrm{\λ}}_2|\left|{\mathrm{\Ψ\ρ}}_2\right|{|}_1\\ .\\ .\\ .\\ {\widehat{\mathrm{\ρ}}}_p=\underset{{\mathrm{\ρ}}_p}{\arg \min }\left|\right|y_p-F_u{\mathrm{\ρ}}_p|{|}_2^2+{\mathrm{\λ}}_p|\left|{\mathrm{\Ψ\ρ}}_p\right|{|}_1\end{array}$

Wherein,Represent the characteristics of image rebuild, ρ₁、ρ₂、…、ρ_pMagnetic resonance image (MRI) feature to be asked is represented, F_uFor the Fourier-encoded matrix of lack sampling, λ₁、λ₂、…、λ_pFor regularization coefficient, Ψ represents sparse transformation matrix, | | | |₁Table Show and seek L1 norms, | | | |₂L2 norms are sought in expression；

Wherein, magnetic resonance image (MRI) feature is expressed as with the relation of magnetic resonance image (MRI)：

ρ₁=K₁*ρ

ρ₂=K₂*ρ

ρ_p=K_p*ρ

Wherein, * is convolution operator, and ρ is magnetic resonance image (MRI) to be asked.

11. magnetic resonance fast imaging systems as claimed in claim 10, it is characterised in that the reconstruction image generation module In, the magnetic resonance image (MRI) of reconstruction is obtained by the following manner：

$\begin{array}{c}\widehat{\mathrm{\ρ}}=\underset{\mathrm{\ρ}}{\arg \min }\left|\right|y-F_u\mathrm{\ρ}|{|}_2^2+{\mathrm{\γ}}_{1}R\left(\mathrm{\ρ}\right)+\\ {\mathrm{\γ}}_1\left({\mathrm{\α}}_1\left|\right|K_1*\mathrm{\ρ}-{\widehat{\mathrm{\ρ}}}_1|{|}_2^2+{\mathrm{\α}}_2||K_2*\mathrm{\ρ}-{\widehat{\mathrm{\ρ}}}_2|{|}_2^2+\mathrm{...}+{\mathrm{\α}}_p\left|\right|K_p*\mathrm{\ρ}-{\widehat{\mathrm{\ρ}}}_p|{|}_2^2\right)\end{array}$

Wherein,For the magnetic resonance image (MRI) rebuild, function R () is for constraining the openness of magnetic resonance image (MRI), γ₁Treat for control The openness regularization parameter of the magnetic resonance image (MRI) asked, γ₂Image to control magnetic resonance image (MRI) feature to be asked with rebuild is special The regularization parameter of similarity, α between levying₁、α₂、…、α_pThe weight of each magnetic resonance image (MRI) feature is represented respectively.

12. magnetic resonance fast imaging systems as claimed in claim 11, it is characterised in that the function R () is for constraining Magnetic resonance image (MRI) it is openness, wherein, R (ρ)=| | Ψ ρ | |₁。