CN110286344A - A kind of rapid magnetic-resonance variable-resolution imaging method, system and readable medium - Google Patents
A kind of rapid magnetic-resonance variable-resolution imaging method, system and readable medium Download PDFInfo
- Publication number
- CN110286344A CN110286344A CN201810529328.0A CN201810529328A CN110286344A CN 110286344 A CN110286344 A CN 110286344A CN 201810529328 A CN201810529328 A CN 201810529328A CN 110286344 A CN110286344 A CN 110286344A
- Authority
- CN
- China
- Prior art keywords
- data
- frequency
- space
- frequency section
- section
- Prior art date
- Legal status (The legal status 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 status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R33/00—Arrangements or instruments for measuring magnetic variables
- G01R33/20—Arrangements or instruments for measuring magnetic variables involving magnetic resonance
- G01R33/44—Arrangements or instruments for measuring magnetic variables involving magnetic resonance using nuclear magnetic resonance [NMR]
- G01R33/48—NMR imaging systems
- G01R33/4818—MR characterised by data acquisition along a specific k-space trajectory or by the temporal order of k-space coverage, e.g. centric or segmented coverage of k-space
- G01R33/482—MR characterised by data acquisition along a specific k-space trajectory or by the temporal order of k-space coverage, e.g. centric or segmented coverage of k-space using a Cartesian trajectory
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R33/00—Arrangements or instruments for measuring magnetic variables
- G01R33/20—Arrangements or instruments for measuring magnetic variables involving magnetic resonance
- G01R33/44—Arrangements or instruments for measuring magnetic variables involving magnetic resonance using nuclear magnetic resonance [NMR]
- G01R33/48—NMR imaging systems
- G01R33/54—Signal processing systems, e.g. using pulse sequences ; Generation or control of pulse sequences; Operator console
- G01R33/56—Image enhancement or correction, e.g. subtraction or averaging techniques, e.g. improvement of signal-to-noise ratio and resolution
- G01R33/561—Image 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
- G01R33/5611—Parallel 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
Abstract
The present invention relates to mr imaging technique field, a kind of rapid magnetic-resonance variable-resolution imaging method, system and readable medium are disclosed.The main technical scheme is that: integrating parallel imaging and compressed sensing technology, data are acquired using parallel imaging acquisition method and variable density Descartes's sampling method, obtain three low frequency part, intermediate-frequency section and high frequency section frequency range parts, and three kinds of convolution kernels of different sizes is respectively adopted, convolution, reconstruction image are carried out to the data of different frequency range.The beneficial effects of the practice of the present invention mainly has: can preferably inhibit in parallel imaging as accelerate multiple become larger and caused by noise amplification phenomenon, can be imaged under higher accelerations multiple, raising image taking speed；It can make the image obtained that there is the resolution ratio of different resolution and area-of-interest compared with more high image quality is also more preferable elsewhere.
Description
Technical field
The present invention relates to mr imaging technique fields, in particular to a kind of rapid magnetic-resonance variable-resolution imaging side
Method, system and readable medium.
Background technique
Magnetic resonance is imaged tissue using magnetostatic field and RF magnetic field, it provides not only tissue contrast abundant
Degree, and it is harmless, therefore become a kind of strong tool of medical clinic applications.But image taking speed is always slowly
It restricts its fast-developing big bottleneck and improves scanning speed how under the premise of image quality is clinically-acceptable, thus
It is particularly important to reduce sweep time.Commercial fast imaging techniques are mainly parallel imaging at present, such as the automatic calibrated section of broad sense
Parallel acquisition (GRAPPA), susceptibility coding (SENSE) etc., the spatial information of receiving coil is all utilized in such method, to fill out
Fill the K space data for owing to adopt.
Parallel imaging is a kind of fast imaging method using multi-channel coil acquisition signal, utilizes list in phased-array coil
The spatial sensitivities difference of a receiving coil carrys out space encoder information, phase code step number necessary to encoding is reduced, to obtain
Obtain faster scanning speed.The data acquisition of parallel imaging is that uniform density acquires, and is influenced by factors such as magnetic field bumps,
The acceleration multiple of parallel imaging will not be very high, and usually at 2-3 times, and with the increase for accelerating multiple, parallel imaging, which will appear, makes an uproar
The phenomenon that sound amplifies.
Research and development by many years to technology is suggested there are many parallel imaging technique, and the difference of these technologies exists
In the mode that Coil sensitivity information is used.SMASH(Simultaneous Acquisition of Spatial
Harmonics), the methods of SENSE needs accurate sensitivity information, but is difficult accurately to obtain sensitivity information, and in image
In reconstruction, even if the sensitivity information error of very little may all make reconstruction image artifact occur, reconstructed image quality is seriously affected.
And implicit being rebuild using Coil sensitivity information such as AUTO-SMASH (Auto-calibrating SMASH), GRAPPA is schemed
Picture can be avoided and be difficult to the problem of obtaining accurate Coil sensitivity information, but when accelerating multiple higher, have noise amplification
The phenomenon that, therefore accelerate multiple cannot be too high.
After compressed sensing technology occurs, there is researcher to be combined it with Paraller imaging algorithm, compressive sensing theory is logical
Cross the sparse prior by image, reduce sample number, can obtain than Traditional parallel be imaged higher acceleration multiple with more preferably
Picture quality, but accelerate multiple it is higher when still suffer from noise amplification phenomenon generate.
" Zhang Jiuming, Guo Shuxu, Wang Miaoshi, Zhong Fei: compressed sensing calibrates parallel imaging algorithm for reconstructing [J] meter to document automatically
Calculation machine application, 2014,34 (5): 1491-1493,1502 " in there is proposition to combine compressed sensing technology and parallel imaging method, and
And it is also proposed that it can preferably inhibit the generation of noise phenomenon, but in the document using LI-SPIRiT algorithm progress data reconstruction
The imaging method being previously mentioned is to be rebuild using the same convolution kernel to k-space, all LI-SPIRiT is used to calculate all data
Method is rebuild, and is only capable of obtaining the image with single resolution ratio.
Magnetic resonance data acquisition mode can be divided into Descartes's sampling and non-Cartesian sampling.For Cartesian sampling pattern,
Data have been fallen on cartesian grid, and so there is no need to pre-process to data；And for non cartesian data, then need logarithm
According to being pre-processed, fall in it on cartesian grid, current processing method is mainly gridding method (gridding), i.e.,
The method for carrying out convolution using a convolution kernel to fill k-space.
In conclusion although the methods of Paraller imaging algorithm, Traditional parallel imaging-compressed sensing combined techniques can be improved and sweep
Speed is retouched, but when accelerating multiple too high, still will affect picture quality, and current imaging method is only capable of obtaining with single
The image of resolution ratio.
Summary of the invention
The technical problem to be solved by the present invention is to accelerate image taking speed, and obtain how under the premise of guaranteeing picture quality
Obtain the image of multiresolution.
In order to solve the above-mentioned technical problem, the present invention discloses a kind of rapid magnetic-resonance variable-resolution imaging side first
Method, technical solution are implemented:
A kind of rapid magnetic-resonance variable-resolution imaging method, is adopted using parallel imaging acquisition method and variable density Descartes
Sample method acquires data, obtains three low frequency part, intermediate-frequency section and high frequency section frequency range parts, and three kinds of sizes are respectively adopted not
Same convolution kernel carries out convolution, reconstruction image to the data of different frequency range；Specifically includes the following steps:
S1: being divided into three adjacent frequency range parts along phase-encoding direction for k-space, i.e., low frequency part, intermediate-frequency section and
High frequency section adopts the low frequency part entirely, acquires data using parallel imaging acquisition mode to the intermediate-frequency section, right
The high frequency section carries out variable density acquisition；
S2: the data of three frequency range collected to step S1 carry out k-space reconstruction, obtain reconstruction image.
Preferably, the data acquisition modes of the high frequency section are that frequency coding direction is adopted entirely, and phase-encoding direction is adopted
The stochastical sampling that collection follows compressed sensing is theoretical.
Preferably, the imaging method combines parallel imaging method and compressed sensing technology carries out k-space reconstruction.
Preferably, to the data of high frequency section, k-space reconstruction is carried out using L1-SPIRiT method.
Preferably, in step S1, the cutoff frequency of the low frequency part, intermediate-frequency section and high frequency section is respectively k1、k2、
k3, low frequency part, intermediate-frequency section, high frequency section frequency range be respectively 0≤kIt is low≤k1,k1<kIn≤k2,k2<kIt is high≤k3, setting
Sample rate is respectively R1、R2、R3, wherein R1=1, R3>R2>R1；The data of the low frequency part acquisition are as automatic in k-space
Calibration signal.
Preferably, the data acquisition modes of the intermediate-frequency section be frequency coding direction adopt entirely, phase-encoding direction every
(R2- 1) line acquires a line, until reaching cutoff frequency k2。
Preferably, step S2 is further comprising the steps of:
S2.1: frequency part and the intermediate-frequency section data are rebuild, and obtaining tool, there are two the figures in different resolution region
Picture, reconstruction process are as follows:
Si=fi*ci, i=1,2 (1)
Wherein, fiIndicate the k-space data for owing to adopt, ciIndicate convolution kernel, SiIndicate filled k-space, i=1,2 difference
Indicate the low frequency part and intermediate-frequency section of k-space, symbol * indicates convolution operation.
Preferably, the convolution kernel size c of the low frequency part1=4R1, the convolution kernel size c of the intermediate-frequency section2=
4R2。
Preferably, further comprising the steps of:
S2.2: in the automatic calibration signal data obtained in step S1, weight system is obtained by solving following formula
Number:
Wherein m indicates that m-th of coil, r indicate position, RrIt is all k-spaces in some neighborhood around a selection position r
The operation of point, gjmIt is the weight vectors that all the points obtain out of position r surrounding neighbors,It is gjmConjugate transposition；
S2.3: entire k-space is reconstructed by solving following optimization problem:
Wherein G is to be obtained from the automatic calibration data comprising gjmSPIRiT operation, I be unit matrix, x indicate
The k-space of reconstruction, D are to choose all operations for having acquired data, and y is sampled data, and ∈ is for controlling data fidelity.
Preferably, incoherent artifact is removed by applying canonical constraint, specifically includes the following steps:
S2.4: apply canonical constraint on the basis of (3) formula:
Wherein Ψ indicates wavelet transformation, λ1、λ2Respectively regular parameter,
(4) formula of solution, reconstructs high frequency k-space.
Preferably, further comprising the steps of:
S2.5: convolution kernel c is used2Convolution is carried out to the high frequency k-space data reconstructed in step S2.4, is obtained new
K-space high-frequency data；
S2.6: to k-space data S obtained in step S2.11、S2With high-frequency data obtained in step S2.5 directly into
Row inversefouriertransform obtains the reconstruction image with Resolutions.
Secondly the present invention further discloses a kind of rapid magnetic-resonance variable-resolution imaging system, including data acquisition module；
The data acquisition module is used to k-space being divided into three adjacent parts, i.e. low frequency part, intermediate frequency portion along phase-encoding direction
Point and high frequency section, the low frequency part is adopted entirely, to the intermediate-frequency section using parallel imaging acquisition mode acquisition number
According to, variable density acquisition is carried out to the high frequency section, the data acquisition modes of the high frequency section are that frequency coding direction is adopted entirely,
The stochastical sampling that the acquisition of phase-encoding direction follows compressed sensing is theoretical；The low frequency part, intermediate-frequency section and high frequency section
Cutoff frequency be respectively k1、k2、k3, low frequency part, intermediate-frequency section, high frequency section frequency range be respectively 0≤kIt is low≤k1,k1
<kIn≤k2,k2<kIt is high≤k3, setting sample rate is respectively R1、R2、R3, wherein R1=1, R3>R2>R1；The low frequency part acquisition
Data are as the automatic calibration signal in k-space；The data acquisition modes of the intermediate-frequency section are that frequency coding direction is adopted entirely, phase
Position coding direction is every (R2- 1) line acquires a line, until reaching cutoff frequency k2。
It preferably, further include data reconstruction module；
The data reconstruction module is had for rebuilding to the low frequency part and the intermediate-frequency section data
The image in two different resolution regions, reconstruction process are as follows:
Si=fi*ci, i=1,2 (1)
Wherein, fiIndicate the k-space data for owing to adopt, ciIndicate convolution kernel, SiIndicate filled k-space, i=1,2 difference
Indicate the low frequency part and intermediate-frequency section of k-space, symbol * indicates convolution operation；
The convolution kernel size c of the low frequency part1=4R1, the convolution kernel size c of the intermediate-frequency section2=4R2；
In the automatic calibration signal data, weight coefficient is obtained by solving following formula:
Wherein m indicates that m-th of coil, r indicate position, RrIt is all k-spaces in some neighborhood around a selection position r
The operation of point, gjmIt is the weight vectors that all the points obtain out of position r surrounding neighbors,It is gjmConjugate transposition；
Entire k-space is reconstructed by solving following optimization problem:
Wherein G is to be obtained from the automatic calibration data comprising gjmSPIRiT operation, I be unit matrix, x indicate
The k-space of reconstruction, D are to choose all operations for having acquired data, and y is sampled data, and ∈ is for controlling data fidelity；
When considering the prior information of image, apply canonical constraint on the basis of (3) formula:
Wherein Ψ indicates wavelet transformation, λ1、λ2Respectively regular parameter,
(4) formula of solution, reconstructs high frequency k-space；
Use convolution kernel c2Convolution is carried out to the high frequency k-space data that the above method reconstructs, obtains new k-space
High-frequency data；
To the k-space data S of low frequency part and high frequency section that reconstruction obtains1、S2It is directly carried out with k-space high-frequency data
Inversefouriertransform obtains the reconstruction image with Resolutions.
The present invention also provides a kind of computer-readable medium, which has the program being stored therein, should
Program is that computer is executable so that computer executes each step of above-mentioned rapid magnetic-resonance variable-resolution imaging method.
The beneficial effects of the practice of the present invention mainly has:
1, can preferably inhibit in parallel imaging as accelerate multiple become larger and caused by noise amplification phenomenon, can be more
It is imaged under high acceleration multiple, improves image taking speed；
2, it can make the image obtained that there is the resolution ratio of different resolution and area-of-interest compared with higher figure elsewhere
Image quality amount is also more preferable；
It 3, can be by further decreasing hits, to improve scanning speed for uninterested regional resolution.
Detailed description of the invention
Technical solution for a better understanding of the invention, can refer to it is following, for being carried out to the prior art or embodiment
The attached drawing of explanation.These attached drawings will carry out brief displaying to section Example or prior art related products or method.This
The essential information of a little attached drawings is as follows:
Fig. 1 is the flow chart of rapid magnetic-resonance variable-resolution imaging method in one embodiment；
Fig. 2 is the schematic illustration of SPIRiT algorithm；
Fig. 3 is the data arrangement schematic diagram of k-space.
Specific embodiment
Present technical solution in the embodiment of the present invention or beneficial effect make further expansion description, it is clear that are retouched
The embodiment stated is only some embodiments of the invention, and and not all.
It should be pointed out that the proposition of the invention, primarily to solving in mr imaging technique field, accordingly
Problem of the existing technology, so the invention is especially suitable for the subdivision field, but not meaning the invention
The applicable range of technical solution institute it is therefore limited, those skilled in the art can be as needed, in various concrete application occasions
Reasonably implemented.
In some embodiments, understood referring to attached drawing, a kind of rapid magnetic-resonance variable-resolution imaging method is adopted
Data are acquired with parallel imaging acquisition method and variable density Descartes's sampling method, obtain low frequency part, intermediate-frequency section and radio-frequency head
Divide three frequency range parts, and three kinds of convolution kernels of different sizes are respectively adopted, convolution is carried out to the data of different frequency range, rebuilds figure
Picture；Specifically includes the following steps:
S1: being divided into three adjacent frequency range parts along phase-encoding direction for k-space, i.e., low frequency part, intermediate-frequency section and
High frequency section adopts the low frequency part entirely, acquires data using parallel imaging acquisition mode to the intermediate-frequency section, right
The high frequency section carries out variable density acquisition, and the data acquisition modes of the high frequency section are that frequency coding direction is adopted entirely, phase
The stochastical sampling that the acquisition of coding direction follows compressed sensing is theoretical；
S2: the data of three frequency range collected to step S1 carry out k-space reconstruction, obtain reconstruction image.
In a preferred embodiment, the imaging method combines parallel imaging method and compressed sensing technology carries out k
Space reconstruction.
Compressed sensing technology and parallel imaging method combination can be obtained into higher acceleration multiple, when being further reduced sampling
Between, and low frequency part, the data of three frequency range parts of intermediate-frequency section and high frequency section will be acquired in data acquisition phase, it uses
Three kinds of convolution kernels of different sizes carry out convolution to the data of these three frequency range parts respectively, can obtain multiple resolutions
Image.
In some preferred embodiments, to the data of high frequency section, k-space reconstruction is carried out using L1-SPIRiT method.
Such as Fig. 2, SPIRiT algorithm is first to be rebuild not in k-space domain with the linear weighted function factor that self calibration data are calculated
Then the data point of sampling obtains the image of each coil by inverse Fourier transform, merges finally by by a coil image
At reconstruction image, rebuilds data and the consistency of sampled data and the consistency of automatic calibration process is the basis of SPIRiT.
GAPPA algorithm is similar to SPIRiT algorithm, does not use k-space data, and weighted and generated by the k-space data in its field,
But the two the difference lies in that the data that SPIRiT algorithm weights use not only have the data of sampling, generated also by iteration
Non- sampled data.SPIRiT algorithm can preferably determine the reconstruction relationship between non-sampled point and sampled point.
On this basis, the structure of reconstruction image obtained is more clear LI-SPIRiT algorithm, and noise can obtain more
Effective to inhibit, when accelerating multiple higher, the quality of reconstruction image is more increased.
In document, " Zhang Jiuming, Guo Shuxu, Wang Miaoshi, Zhong Fei: compressed sensing calibrates parallel imaging algorithm for reconstructing [J] automatically
Computer application, 2014,34 (5): 1491-1493,1502 " in there is proposition to combine compressed sensing technology and parallel imaging method,
And it is also proposed that carrying out data reconstruction using LI-SPIRiT algorithm, but the imaging method being previously mentioned in the document is using same
A convolution kernel rebuilds k-space, all uses LI-SPIRiT algorithm to rebuild all data, is only capable of obtaining one point
The image of resolution merely provides the method for inhibiting to cause noise amplification phenomenon due to accelerating multiple to become larger in parallel imaging,
The advantages of present invention utilizes LI-SPIRiT algorithms also proposed carry out k sky using different size of convolution kernel on this basis
Between rebuild, obtain the image with different resolution, and only rebuild to high-frequency data using LI-SPIRiT algorithm, make
The resolution ratio for obtaining area-of-interest is higher compared with other regions.
In some preferred embodiments, in step S1, the cutoff frequency of the low frequency part, intermediate-frequency section and high frequency section
Rate is respectively k1、k2、k3, setting sample rate is respectively R1、R2、R3, wherein R1=1, R3>R2>R1；The low frequency part acquisition
Data are as the automatic calibration signal in k-space.
Value those skilled in the art of cutoff frequency can set according to actual needs, in a specific embodiment
In, low frequency part, intermediate-frequency section, high frequency section frequency range be respectively 0≤kIt is low≤k1,k1<kIn≤k2,k2<kIt is high≤k3.Low frequency
Partially, intermediate-frequency section and the cutoff frequency k of high frequency section1、k2、k3It can be set as k1=(0~0.1) kmax, k2=(0.2~
0.5)kMax,k3=kmax, kmaxValue be determined according to scan protocols.k1、k2、k3With kmaxRelationship those skilled in the art
It can also specifically be needed voluntarily to select according to image scene.The scan protocols of different scanning portions are different, with specific reference to scanning scene
Carry out set.K after scan protocols determinemaxAlso it can determine therewith, kMax is theoretical=N/FOV, wherein N is phase code step
Number, corresponding to the size in encoder matrix along phase directional, FOV is that scan vision, N and FOV are determined by scan protocols, specifically
In scanning, FOV is determined, determines phase code step number N according to required resolution ratio, it is thus determined that N is to have determined
kMax is theoretical, in actual scanning, General N value is 192-512.But since low frequency part is located at the special data arrangement at k-space center
Mode, actually kMax is theoreticalHalf be negative value, the part of negative value needs to take absolute value, therefore k in actual conditionsmaxIt is in fact
kMax is theoreticalHalf, as shown in figure 3, k1、k2、k3Also according to the actual situation in kmaxIt determines.
In a specific embodiment, the cutoff frequency of low frequency part, intermediate-frequency section and high frequency section can be set as
k1=0.03kmax, k2=0.28kmax, k3=kmax。
The value setting of sample rate is related to image quality, and those skilled in the art need to set according to actual needs,
R1=1 is unique value, since when data acquire, energy is concentrated mainly on k-space center low frequency part, therefore, in order to ensure that figure
The contrast of picture, general k space center part can select to adopt entirely, R2=2~3, R3=4~8, in a specific embodiment,
The sample rate of intermediate-frequency section and high frequency section can be set as R2=2, R3=4；In the disclosure, intermediate-frequency section is to take parallel
It owes the mode adopted to carry out, so can generally select 2 times or 3 times to owe to adopt to guarantee image quality；For high frequency section, Ke Yixuan
It selects 4 times, 5 times, 6 times, 7 times or 8 times to owe to adopt, if the resolution ratio of image is higher, can choose 6 times, 7 times or 8 times and owe to adopt.
In some preferred embodiments, the data acquisition modes of the intermediate-frequency section are that frequency coding direction is adopted entirely, phase
Position coding direction is every (R2- 1) line acquires a line, until reaching cutoff frequency k2。
In some preferred embodiments, step S2 is further comprising the steps of:
S2.1: rebuilding the low frequency part and the intermediate-frequency section data, and obtaining tool, there are two different resolutions
The image in region, reconstruction process are as follows:
Si=fi*ci, i=1,2 (1)
Wherein, fiIndicate the k-space data for owing to adopt, ciIndicate convolution kernel, SiIndicate filled k-space, i=1,2 difference
Indicate the low frequency part and intermediate-frequency section of k-space, symbol * indicates convolution operation.
In some preferred embodiments, the convolution kernel size c of the low frequency part1=4R1, the volume of the intermediate-frequency section
Product core size c2=4R2。
The size of convolution kernel is very big on the influence of the filling effect of k-space, low to adopt if the width of convolution kernel is arranged too small
The part of sample rate cannot be then filled up completely；If the width of convolution kernel is arranged too big, serious diffraction effect may cause.
Since low frequency part is different with the sample rate of intermediate-frequency section, to obtain optimal filling effect, different size is used
Convolution kernel c1、c2Low frequency part and intermediate-frequency section data are handled respectively.
It is further comprising the steps of in some preferred embodiments:
S2.2: in the automatic calibration signal data obtained in step sl, weight is obtained by solving following formula
Coefficient:
Wherein m indicates that m-th of coil, r indicate position, RrIt is all k-spaces in some neighborhood around a selection position r
The operation of point, gjmIt is the weight vectors that all the points obtain out of position r surrounding neighbors,It is gjmConjugate transposition；
Some point that formula (2) encodes k-space can be raw by all the points in its neighborhood (including sampling and unsampled point)
At.
S2.3: entire k-space is reconstructed by solving following optimization problem:
Wherein G is to be obtained from the automatic calibration data comprising gjmSPIRiT operation, I be unit matrix, x indicate
The k-space of reconstruction, D are to choose all operations for having acquired data, and y is sampled data, and ∈ is for controlling data fidelity.
For the data of high frequency section, due to using simultaneously using compressed sensing based stochastical sampling mode
Method of the row imaging in conjunction with compressed sensing rebuilds high-frequency data, and used method is L1-SPIRiT method, specifically
Reconstruction process is the prior art, can refer to document " Lustig M, Pauly JM.SPIRiT:iterative self-
consistent parallel imaging reconstruction from arbitrary k-space.Magn Reson
Med 2010；64:457–471".
It, can be by applying canonical constraint removal when considering the prior information of image in some preferred embodiments
Incoherent artifact, specifically includes the following steps:
S2.4: apply canonical constraint on the basis of (3) formula:
Wherein Ψ indicates wavelet transformation, λ1、λ2Respectively regular parameter,
(4) formula is solved using projections onto convex sets (POCS), reconstructs high frequency k-space,
It is as follows using the solution process of POCS method:
Input: y: the k-space data adopted is owed
D: the operation that k-space has acquired data is chosen
Dc: the operation that k-space does not acquire data is chosen
G: the SPIRiT operation matrix obtained from calibration data
WavWeight:L1 canonical constrained parameters
ErrToll: stop condition
Output: Xk: all channel k-space datas of reconstruction
Algorithm: X0=DTy；K=0
As e > errToll,
{
K=k+1；
Xk=Dc (GXk-1+Xk-1)+Dy；
X=Ψ { IFFT (Xk)}；
X=softThresh (X, wavWeight)；
Xk=FFT (ΨTX)；
E=| | Xk-Xk-1||；
Xk-1=Xk
}
Those skilled in the art can also be used other methods and solve (4) formula, and method for solving is not innovative point institute of the invention
, therefore convex set projection method is only schematically illustrated in the present embodiment.
It is further comprising the steps of in some preferred embodiments:
S2.5: convolution kernel c is used2Convolution is carried out to the high frequency k-space data reconstructed in step S2.4, is obtained new
K-space high-frequency data；
S2.6: to k-space data S obtained in step S2.11、S2With high-frequency data obtained in step S2.5 directly into
Row inversefouriertransform obtains the reconstruction image with Resolutions.
For the high resolution of the stripe region of obtained picture centre in other regions, stripe region is area-of-interest,
Resolution ratio more high image quality is also more preferable；Uninterested regional resolution is lower, can be by further for these regions
Hits is reduced, scanning imagery speed is improved.
In some embodiments, invention further discloses a kind of rapid magnetic-resonance variable-resolution imaging systems, including number
According to acquisition module；The data acquisition module is used to k-space being divided into three adjacent parts, i.e. low frequency along phase-encoding direction
Partially, intermediate-frequency section and high frequency section acquire the low frequency part and the intermediate-frequency section using parallel imaging acquisition mode
Data carry out variable density acquisition to the high frequency section, and the data acquisition modes of the high frequency section are that frequency coding direction is complete
It adopts, the stochastical sampling that the acquisition of phase-encoding direction follows compressed sensing is theoretical；The low frequency part, intermediate-frequency section and radio-frequency head
The cutoff frequency divided is respectively k1、k2、k3, setting sample rate is respectively R1、R2、R3, wherein R1=1, R3>R2>R1；The low frequency
The data of part acquisition are as the automatic calibration signal in k-space；The data acquisition modes of the intermediate-frequency section are frequency coding
Direction is adopted entirely, and phase-encoding direction is every (R2- 1) line acquires a line, until reaching cutoff frequency k2。
It further include data reconstruction module in some preferred embodiments；
The data reconstruction module is had for rebuilding to the low frequency part and the intermediate-frequency section data
The image in two different resolution regions, reconstruction process are as follows:
S=fi*ci, i=1,2 (1)
Wherein, fiIndicate the k-space data for owing to adopt, ciIndicate convolution kernel, SiIndicate filled k-space, i=1,2 difference
Indicate the low frequency part and intermediate-frequency section of k-space, symbol * indicates convolution operation；
The convolution kernel size c of the low frequency part1=4R1, the convolution kernel size c of the intermediate-frequency section2=4R2；
In the automatic calibration signal data, weight coefficient is obtained by solving following formula:
Wherein m indicates that m-th of coil, r indicate position, RrIt is all k-spaces in some neighborhood around a selection position r
The operation of point, gjmIt is the weight vectors that all the points obtain out of position r surrounding neighbors,It is gjmConjugate transposition；
Entire k-space is reconstructed by solving following optimization problem:
Wherein G is to be obtained from the automatic calibration data comprising gjmSPIRiT operation, I be unit matrix, x indicate
The k-space of reconstruction, D are to choose all operations for having acquired data, and y is sampled data, and ∈ is for controlling data fidelity；
When considering the prior information of image, apply canonical constraint on the basis of (3) formula:
Wherein Ψ indicates wavelet transformation, λ1、λ2Respectively regular parameter,
(4) formula is solved using projections onto convex sets, reconstructs high frequency k-space；
Use convolution kernel c2Convolution is carried out to the high frequency k-space data reconstructed in above-mentioned steps, it is empty to obtain new k
Between high-frequency data；
To the k-space data S of low frequency part and high frequency section that reconstruction obtains1、S2It is directly carried out with k-space high-frequency data
Inversefouriertransform obtains the reconstruction image with Resolutions.
In some embodiments, invention further discloses a kind of computer-readable medium, which has storage
In program wherein, which is that computer is executable so that computer executes above-mentioned rapid magnetic-resonance variable-resolution imaging
Each step of method.
Finally it should be pointed out that embodiment cited hereinabove, is more typical, preferred embodiment of the invention, only
For being described in detail, explaining technical solution of the present invention, in order to reader's understanding, the protection scope being not intended to limit the invention
Or application.Therefore, within the spirit and principles in the present invention any modification, equivalent replacement, improvement and so on and obtain
Technical solution should be all included within protection scope of the present invention.
Claims (14)
1. a kind of rapid magnetic-resonance variable-resolution imaging method, it is characterised in that: using parallel imaging acquisition method and become close
It spends Descartes's sampling method and acquires data, obtain three low frequency part, intermediate-frequency section and high frequency section frequency range parts, and be respectively adopted
Three kinds of convolution kernels of different sizes carry out convolution, reconstruction image to the data of different frequency range；
Specifically includes the following steps:
S1: k-space is divided into three adjacent frequency range parts, i.e. low frequency part, intermediate-frequency section and high frequency along phase-encoding direction
The low frequency part is adopted in part entirely, data is acquired using parallel imaging acquisition mode to the intermediate-frequency section, to described
High frequency section carries out variable density acquisition；
S2: the data of three frequency range collected to step S1 carry out k-space reconstruction, obtain reconstruction image.
2. imaging method according to claim 1, it is characterised in that: the data acquisition modes of the high frequency section are frequency
Coding direction is adopted entirely, and the stochastical sampling that the acquisition of phase-encoding direction follows compressed sensing is theoretical.
3. imaging method according to claim 1, it is characterised in that: the imaging method combines parallel imaging method and pressure
Contracting cognition technology carries out k-space reconstruction.
4. imaging method according to claim 1, it is characterised in that: to the data of high frequency section, using the side L1-SPIRiT
Method carries out k-space reconstruction.
5. imaging method according to claim 1, it is characterised in that:
In step S1, the cutoff frequency of the low frequency part, intermediate-frequency section and high frequency section is respectively set as k1、k2、k3, low frequency
Partially, intermediate-frequency section, high frequency section frequency range be respectively 0≤kIt is low≤k1, k1<kIn≤k2, k2<kIt is high≤k3, set sample rate
Respectively R1、R2、R3, wherein R1=1, R3>R2>R1；
The data of the low frequency part acquisition are as the automatic calibration signal in k-space.
6. imaging method according to claim 5, it is characterised in that:
The data acquisition modes of the intermediate-frequency section are that frequency coding direction is adopted entirely, and phase-encoding direction is every (R2- 1) line is adopted
Collect a line, until reaching cutoff frequency k2。
7. imaging method according to claim 5, it is characterised in that:
Step S2 is further comprising the steps of:
S2.1: rebuilding the low frequency part and the intermediate-frequency section data, and obtaining tool, there are two different resolution regions
Image, reconstruction process is as follows:
Si=fi*ci, i=1,2 (1)
Wherein, fiIndicate the k-space data for owing to adopt, ciIndicate convolution kernel, SiIndicate filled k-space, i=1,2 respectively indicate
The low frequency part and intermediate-frequency section of k-space, symbol * indicate convolution operation.
8. imaging method according to claim 7, it is characterised in that: the convolution kernel size c of the low frequency part1=4R1,
The convolution kernel size c of the intermediate-frequency section2=4R2。
9. imaging method according to claim 8, it is characterised in that:
Step S2 is further comprising the steps of:
S2.2: in the automatic calibration signal data obtained in step sl, weight coefficient is obtained by solving following formula:
Wherein m indicates that m-th of coil, r indicate position, RrIt is all k-space points in some neighborhood around a selection position r
Operation, gjmIt is the weight vectors that all the points obtain out of position r surrounding neighbors,It is gjmConjugate transposition；
S2.3: entire k-space is reconstructed by solving following optimization problem:
Wherein G is to be obtained from the automatic calibration data comprising gjmSPIRiT operation, I be unit matrix, x indicate rebuild
K-space, D is to choose all operations for having acquired data, and y is sampled data, and ∈ is for controlling data fidelity.
10. imaging method according to claim 9, it is characterised in that: incoherent artifact is removed by applying canonical constraint,
Specifically includes the following steps:
S2.4: apply canonical constraint on the basis of (3) formula:
Wherein ψ indicates wavelet transformation, λ1、λ2Respectively regular parameter,
(4) formula of solution, reconstructs high frequency k-space.
11. imaging method according to claim 10, it is characterised in that:
It is further comprising the steps of:
S2.5: convolution kernel c is used2Convolution is carried out to the high frequency k-space data reconstructed in step S2.4, it is empty to obtain new k
Between high-frequency data；
S2.6: the k-space data S that step S2.1 is arrived1、S2It is directly carried out in anti-Fu with high-frequency data obtained in step S2.5
Leaf transformation obtains the reconstruction image with Resolutions.
12. a kind of rapid magnetic-resonance variable-resolution imaging system, it is characterised in that:
Including data acquisition module；
The data acquisition module is used to k-space being divided into three adjacent parts along phase-encoding direction, i.e., low frequency part, in
Frequency part and high frequency section, adopt the low frequency part entirely, are adopted to the intermediate-frequency section using parallel imaging acquisition mode
Collect data, variable density acquisition is carried out to the high frequency section, the data acquisition modes of the high frequency section are frequency coding direction
Quan Cai, the stochastical sampling that the acquisition of phase-encoding direction follows compressed sensing are theoretical；
The cutoff frequency of the low frequency part, intermediate-frequency section and high frequency section is respectively k1、k2、k3, low frequency part, intermediate-frequency section,
The frequency range of high frequency section is respectively 0≤kIt is low≤k1, k1< kIn≤k2, k2< kIt is high≤k3, setting sample rate is respectively R1、R2、
R3, wherein R1=1, R3> R2> R1；
The data of the low frequency part acquisition are as the automatic calibration signal in k-space；
The data acquisition modes of the intermediate-frequency section are that frequency coding direction is adopted entirely, and phase-encoding direction is every (R2- 1) line is adopted
Collect a line, until reaching cutoff frequency k2。
13. imaging system according to claim 12, it is characterised in that:
It further include data reconstruction module；
The data reconstruction module obtains there are two tools for rebuilding to the low frequency part and the intermediate-frequency section data
The image in different resolution region, reconstruction process are as follows:
Si=fi*ci, i=1,2 (1)
Wherein, fiIndicate the k-space data for owing to adopt, ciIndicate convolution kernel, SiIndicate filled k-space, i=1,2 respectively indicate
The low frequency part and intermediate-frequency section of k-space, symbol * indicate convolution operation；
The convolution kernel size c of the low frequency part1=4R1, the convolution kernel size c of the intermediate-frequency section2=4R2；
In the automatic calibration signal data, weight coefficient is obtained by solving following formula:
Wherein m indicates that m-th of coil, r indicate position, RrIt is all k-space points in some neighborhood around a selection position r
Operation, gjmIt is the weight vectors that all the points obtain out of position r surrounding neighbors,It is gjmConjugate transposition；
Entire k-space is reconstructed by solving following optimization problem:
Wherein G is to be obtained from the automatic calibration data comprising gjmSPIRiT operation, I be unit matrix, x indicate rebuild
K-space, D is to choose all operations for having acquired data, and y is sampled data, and ∈ is for controlling data fidelity；
When considering the prior information of image, apply canonical constraint on the basis of (3) formula:
Wherein ψ indicates wavelet transformation, λ1、λ2Respectively regular parameter,
(4) formula of solution, reconstructs high frequency k-space；
Use convolution kernel c2Convolution is carried out to the high frequency k-space data reconstructed in above-mentioned steps, it is high to obtain new k-space
Frequency evidence；
To the k-space data S of low frequency part and high frequency section that reconstruction obtains1、S2Anti- Fu is directly carried out with k-space high-frequency data
In leaf transformation, obtain the reconstruction image with Resolutions.
14. a kind of computer-readable medium, it is characterised in that: the computer media has the program being stored therein, the program
It is that computer is executable so that computer perform claim requires 1~11 any rapid magnetic-resonance variable-resolution imaging
Each step of method.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810203130 | 2018-03-13 | ||
| CN2018102031303 | 2018-03-13 |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110286344A true CN110286344A (en) | 2019-09-27 |
| CN110286344B CN110286344B (en) | 2021-11-30 |
Family
ID=68001045
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810529328.0A Active CN110286344B (en) | 2018-03-13 | 2018-05-29 | Rapid magnetic resonance variable resolution imaging method, system and readable medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110286344B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113933773A (en) * | 2021-09-30 | 2022-01-14 | 深圳先进技术研究院 | Magnetic resonance imaging method, system, terminal and storage medium based on deep learning |
| CN115774830A (en) * | 2022-10-28 | 2023-03-10 | 贵州大学 | Rapid mapping method for sparse interference array |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1827039A (en) * | 2005-02-28 | 2006-09-06 | 西门子（中国）有限公司 | MRI parallel imaging method utilizing sensitivity encoding in frequency domain |
| US20100171496A1 (en) * | 2009-01-05 | 2010-07-08 | Hyunwook Park | Higher-order generalized series parallel imaging method for acquiring high spatio-temporal resolution functional magnetic resonance images |
| CN102928796A (en) * | 2012-09-28 | 2013-02-13 | 清华大学 | Fast-diffused magnetic resonance imaging and restoring method |
| CN103675737A (en) * | 2013-12-06 | 2014-03-26 | 清华大学 | Diffusion magnetic resonance imaging and reconstruction method |
| CN106526511A (en) * | 2016-10-21 | 2017-03-22 | 杭州电子科技大学 | SPEED magnetic resonance imaging method based on k space center ghost positioning |
-
2018
- 2018-05-29 CN CN201810529328.0A patent/CN110286344B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1827039A (en) * | 2005-02-28 | 2006-09-06 | 西门子（中国）有限公司 | MRI parallel imaging method utilizing sensitivity encoding in frequency domain |
| US20100171496A1 (en) * | 2009-01-05 | 2010-07-08 | Hyunwook Park | Higher-order generalized series parallel imaging method for acquiring high spatio-temporal resolution functional magnetic resonance images |
| CN102928796A (en) * | 2012-09-28 | 2013-02-13 | 清华大学 | Fast-diffused magnetic resonance imaging and restoring method |
| CN103675737A (en) * | 2013-12-06 | 2014-03-26 | 清华大学 | Diffusion magnetic resonance imaging and reconstruction method |
| CN106526511A (en) * | 2016-10-21 | 2017-03-22 | 杭州电子科技大学 | SPEED magnetic resonance imaging method based on k space center ghost positioning |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113933773A (en) * | 2021-09-30 | 2022-01-14 | 深圳先进技术研究院 | Magnetic resonance imaging method, system, terminal and storage medium based on deep learning |
| CN115774830A (en) * | 2022-10-28 | 2023-03-10 | 贵州大学 | Rapid mapping method for sparse interference array |
| CN115774830B (en) * | 2022-10-28 | 2023-09-29 | 贵州大学 | Rapid imaging method for sparse interference array |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110286344B (en) | 2021-11-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10634753B2 (en) | MR imaging with motion detection | |
| US20190257905A1 (en) | Highly-scalable image reconstruction using deep convolutional neural networks with bandpass filtering | |
| US10996306B2 (en) | MRI system and method using neural network for detection of patient motion | |
| US8638096B2 (en) | Method of autocalibrating parallel imaging interpolation from arbitrary K-space sampling with noise correlations weighted to reduce noise of reconstructed images | |
| Chen et al. | Fast algorithms for image reconstruction with application to partially parallel MR imaging | |
| CN106997034B (en) | Based on the magnetic resonance diffusion imaging method rebuild using Gauss model as example integration | |
| WO2005047919A1 (en) | Method for generating fast magnetic resonance images | |
| CN110133556B (en) | Magnetic resonance image processing method, device, equipment and storage medium | |
| CN106471389A (en) | The reduction of the artifact causing of moving between exciting in MRI due to exciting more | |
| US20090196478A1 (en) | Auto calibration parallel imaging reconstruction method from arbitrary k-space sampling | |
| US11181598B2 (en) | Multi-contrast MRI image reconstruction using machine learning | |
| CN112526423B (en) | Parallel magnetic resonance imaging algorithm based on conjugation and interlayer information | |
| WO2014115043A1 (en) | Metal resistant mr imaging | |
| CN112014782B (en) | Magnetic resonance imaging method, device and computer storage medium | |
| CN108717171A (en) | A kind of compressed sensing Low-field magnetic resonance imaging algorithm | |
| CN110286344A (en) | A kind of rapid magnetic-resonance variable-resolution imaging method, system and readable medium | |
| US6745064B2 (en) | Magnetic resonance method for forming a fast dynamic imaging | |
| KR101575798B1 (en) | Apparatus and method for magnetic resonance image processing | |
| CN105184755A (en) | Parallel magnetic resonance imaging high quality reconstruction method based on self-consistency and containing combined total variation | |
| KR101836235B1 (en) | Method And Apparatus for Magnetic Resonance Imaging | |
| WO2003040744A1 (en) | Magnetic resonance method for forming a fast dynamic image | |
| US20170322279A1 (en) | Multi-shot echo planar imaging using reordered segments and rf excitation pulse phase and slice profiles matched across interleaves | |
| CN108333545A (en) | MR image reconstruction method based on high-pass filtering | |
| CN112557981A (en) | Improved algorithm for parallel magnetic resonance imaging | |
| WO2004099810A1 (en) | Undersampled magnetic resonance imaging |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |