CN105796102B - Realize the method and system of water fat separation - Google Patents
Realize the method and system of water fat separation Download PDFInfo
- Publication number
- CN105796102B CN105796102B CN201410849730.9A CN201410849730A CN105796102B CN 105796102 B CN105796102 B CN 105796102B CN 201410849730 A CN201410849730 A CN 201410849730A CN 105796102 B CN105796102 B CN 105796102B
- Authority
- CN
- China
- Prior art keywords
- block
- pixel
- sub
- seed
- point
- 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.)
- Active
Links
Landscapes
- Magnetic Resonance Imaging Apparatus (AREA)
Abstract
A kind of method for realizing the separation of water fat, the method includes：Acquire several magnetic resonance image that echo time interval is not waited；Several described magnetic resonance image are divided into the sub-block of multiple and different resolution ratio；Multiple sub-blocks are chosen with the minimum in error of fitting respectively and chooses seed block corresponding with resolution ratio respectively as sub-block field figure estimated value corresponding with resolution ratio, and according to the field figure estimated value；The pixel seed point under different resolution under corresponding original image resolution is obtained, and the pixel seed point is merged to obtain final pixel seed point using multiple seed blocks corresponding with resolution ratio；The true field map values of residual pixel point are chosen using the final pixel seed point；Using the true field map values of all pixels point, several described magnetic resonance image are carried out with calculating and respectively obtains water figure and fat figure.Using this method, the efficiency for choosing true field map values can be effectively improved and effective water fat separation is carried out by real field map values.Further, it would be desirable to provide a kind of systems for realizing the separation of water fat.
Description
Technical field
The present invention relates to nmr imaging technique fields, more particularly to a kind of method for realizing the separation of water fat and are
System.
Background technology
Water fat isolation technics has in clinical magnetic resonance imaging (Magnetic Resonance Imaging, abbreviation MRI)
It and is widely applied value, such as in high level main field B0 biased fields (map values f on the spotB, unit Hz) under fat suppression.It passes
The water fat separation method of system is the separation algorithm of the iterative fitting pixel-by-pixel (Iteration based on more echo MR signals
Decomposition of water and fat with Echo Asymmetry and Least-squares, referred to as
IDEAL), but the algorithm is easy to converge to a local minimum and causes water fat point instead when handling high level B0 figure, and
And for single composition pixel, as contained only water or fat in pixel, which may also can obtain the separating resulting of mistake, because
F at this timeBThere are two solution values for tool, one of them is right value (fB,true) it is true field map values, and another is that water fat point is anti-
It is worth (fB,swap), the two solutions can not correctly be distinguished on the basis of single pixel.
In order to overcome this shortcoming, different researchers proposes the estimation that a variety of methods carry out field figure, main to wrap
Include following three kinds：
First, the complexor (phasor) of each pixel, i.e. b=exp (i2 π f are acquired firstBΔ t), wherein Δ
T is echo time interval, as previously mentioned, due to fBThere are two solutions, therefore b is also two value btrueAnd bswap, Ran Houxuan
It takes signal-to-noise ratio high and shows as the close pixel of water fat ratio as rational pixel seed point, then increased using part and calculated
Method is not to estimating that the correct complexor value of pixel is chosen.
2nd, estimate its complexor using image as an entirety first, then divide the image into multiple with overlapping
By the use of Such phase complex vector as initial value, field figure complexor is carried out using IDEAL algorithms to sub-block for the sub-block in region
Estimation, and so on, sub-block is finely divided again, estimates complexor, until obtaining satisfied water fat separating resulting.
3rd, based on 3 points of equidistant MR images, each pixel is obtained using golden cut algorithm on low-resolution image
The candidate solution of field map values, the foundation spatial position of pixel and the sorting position selected pixels seed point of field map values candidate solution, with
Part increases strategy and completes field map values estimation under the resolution ratio, using the estimated value carry out next resolution ratio field figure estimation and
Part increases, until the field figure estimation under highest resolution (i.e. original image resolution) is detached with water fat.
There are shortcoming in above-mentioned three kinds of methods, method one is on the field figure image of single resolution ratio according to pixel signal intensities
Multiple pixels are had chosen as pixel seed point with water fat content ratio, but the imaging section of tissue is spatially separating for containing
, the field map values that more pixel seed points in this single resolution ratio are not enough to complete entire position increase, that is to say, that certain discrete
Field map values in tissue cannot access correct estimation.Method two and method three have used multiresolution strategy, difference point
There is very high correlation, the initial value of high-rise resolution ratio field figure estimation is too dependent on the estimation of low layer resolution ratio between resolution
Value, if the actual field figure of high-rise resolution ratio has acute variation, above-mentioned initial value will cause the high level field figure to converge to one
The minimum of a mistake.Method three acquires 3 points of equidistant MR images, then obtains pixel using golden cut algorithm
Field figure candidate solution, since the field map values of equidistant MR Image estimations have periodically, each the candidate solution of pixel has more
It is a, this efficiency that will reduce true field map values selection.In addition method one and method three have only obtained the complexor of field figure,
Without obtaining true field map values.
Therefore, how to effectively improve the efficiency for choosing true field map values and carried out by real field map values effective
Water fat detaches, a current or technical barrier.
Invention content
Based on this, it is necessary to for above-mentioned technical problem, provide a kind of effectively improve and choose the efficiency of true field map values simultaneously
And the method and system for realizing the separation of water fat of effective water fat separation is carried out by real field map values.
A kind of method for realizing the separation of water fat, the method includes：
Acquire several magnetic resonance image that echo time interval is not waited；
Several described magnetic resonance image are divided into the sub-block of multiple and different resolution ratio；
The minimum chosen respectively to multiple sub-blocks in error of fitting is estimated as sub-block field figure corresponding with resolution ratio
Evaluation, and seed block corresponding with resolution ratio is chosen according to the field figure estimated value respectively；
Corresponding original image resolution under different resolution is obtained using multiple seed blocks corresponding with resolution ratio
Under pixel seed point, and the pixel seed point is merged to obtain final pixel seed point；
The true field map values of residual pixel point are chosen using the final pixel seed point；
Using the true field map values of all pixels point, several described magnetic resonance image calculate respectively obtain water figure and
Fat figure.
In one of the embodiments, it is described acquisition the echo time interval not wait several magnetic resonance image the step of it
Afterwards, it further includes：
The field map values distributed area of pixel in several described magnetic resonance image is carried out discrete；
Error of fitting corresponding with the pixel is calculated to the field map values obtained after discrete respectively；
Solve field figure estimated value of the minimum in the error of fitting as the pixel.
The minimum chosen respectively to multiple sub-blocks in error of fitting is as the son in one of the embodiments,
Block field figure estimated value corresponding with resolution ratio, and seed block corresponding with resolution ratio is chosen according to the field figure estimated value respectively
Step includes：
The error of fitting of all pixels point in the sub-block is corresponded to resolution ratio to be overlapped respectively, after obtaining multiple superpositions
Error of fitting；
The field that the minimum in the error of fitting after the multiple superposition corresponds to resolution ratio as the sub-block is solved respectively
Figure estimated value；
According to the threshold value of the pixel amplitude, the field figure estimated value that resolution ratio is corresponded to using the sub-block is chosen not respectively
With the seed block under resolution ratio.
The threshold value according to the pixel amplitude in one of the embodiments, utilizes the field figure estimated value point
Xuan Qu be under different resolution seed block the step of include：
According to the threshold value of the pixel amplitude, the sub-block is corresponded into the field figure estimated value of resolution ratio as the sub-block
The candidate solution of field map values；
With reference to the signal-to-noise ratio of the pixel, choose in the candidate solution that only there are one the sub-blocks of field figure estimated value to make respectively
For seed block, multiple seed blocks under different resolution are obtained.
It is described in one of the embodiments, to obtain different resolution using multiple seed blocks corresponding with resolution ratio
Under pixel seed point under corresponding original image resolution, and the pixel seed point is merged to obtain final pixel seed point
The step of include：
Using the seed block as starting point, increase the tactful true field map values for choosing sub-block to be estimated using part, obtain difference
Sub-block after increasing under resolution ratio；
Using the true field map values of the sub-block after described increase as initial value, to all pixels point of the sub-block after the growth
Field map values selected, obtain the pixel seed point under corresponding original image resolution under different resolution；
The pixel seed point of original image resolution corresponding under the different resolution is merged, obtains final picture
Plain seed point.
A kind of system for realizing the separation of water fat, the system comprises：
Acquisition module, for acquiring several magnetic resonance image that echo time interval is not waited；
Division module, for several described magnetic resonance image to be divided into the sub-block of multiple and different resolution ratio；
First chooses module, for multiple sub-blocks to be chosen with the minimum in error of fitting respectively as the sub-block with dividing
The corresponding field figure estimated value of resolution, and seed block corresponding with resolution ratio is chosen according to the field figure estimated value respectively；
Acquisition module, for obtaining corresponding original under different resolution using multiple seed blocks corresponding with resolution ratio
Pixel seed point under beginning image resolution ratio, and the pixel seed point is merged to obtain final pixel seed point；
Second chooses module, for choosing the true field map values of residual pixel point using the final pixel seed point；
Separation module, for using the true field map values of the pixel seed point, being carried out to several described magnetic resonance image
Calculating respectively obtains water figure and fat is schemed.
In one of the embodiments, the system also includes：
Discrete block carries out discrete for the field map values distributed area to pixel in several described magnetic resonance image；
Fitting module, for calculating error of fitting corresponding with the pixel to the field map values obtained after discrete respectively；
First solves module, for solving field figure estimation of the minimum in the error of fitting as the pixel
Value.
The first selection module includes in one of the embodiments,：
Laminating module is overlapped respectively for the error of fitting of all pixels point in the sub-block to be corresponded to resolution ratio,
Obtain the error of fitting after multiple superpositions；
Second solves module, for solving the minimum in the error of fitting after the multiple superposition respectively as the son
Block corresponds to the field figure estimated value of resolution ratio；
Seed block chooses module, and for the threshold value according to the pixel amplitude, resolution ratio is corresponded to using the sub-block
Field figure estimated value chooses the seed block under different resolution respectively.
The seed block is chosen module and is included in one of the embodiments,：
The sub-block for the threshold value according to the pixel amplitude, is corresponded to the field of resolution ratio by candidate solution selecting unit
Candidate solution of the figure estimated value as sub-block field map values；
Seed block selection unit for the signal-to-noise ratio with reference to the pixel, is chosen there was only one in the candidate solution respectively
The sub-block of a figure estimated value obtains multiple seed blocks under different resolution as seed block.
The acquisition module includes in one of the embodiments,：
Increase module, for using the seed block as starting point, increasing the tactful true field for choosing sub-block to be estimated using part
Map values obtain the sub-block after increasing under different resolution；
Pixel seed point acquisition module, for using the true field map values of the sub-block after described increase as initial value, to described
The field map values of all pixels point of sub-block after growth are selected, and obtain corresponding original image resolution under different resolution
Under pixel seed point；
Final pixel seed point acquisition module, for by the picture of original image resolution corresponding under the different resolution
Plain seed point merges, and obtains final pixel seed point.
The method and system of above-mentioned realization water fat separation, several magnetic resonance figures not waited by acquiring echo time interval
Picture；Several magnetic resonance image are divided into the sub-block of multiple and different resolution ratio；Multiple sub-blocks are chosen in error of fitting respectively
Minimum is chosen according to field figure estimated value corresponding with resolution ratio respectively as sub-block field figure estimated value corresponding with resolution ratio
Seed block；The picture under different resolution under corresponding original image resolution is obtained using multiple seed blocks corresponding with resolution ratio
Plain seed point, and pixel seed point merged to obtain final pixel seed point；Residual pixel is chosen using final pixel seed point
The true field map values of point；Using the true field map values of all pixels point, several magnetic resonance image are carried out with calculating and respectively obtains water
Figure and fat figure.Due to several magnetic resonance image not waited using echo time interval, which thereby enhance true field map values and choose
Efficiency, due to obtaining final pixel seed point using the seed block under different resolution, and clicked using final pixel seed
The true field map values of each residual pixel point are taken, by the true field map values of each pixel, thus by several magnetic resonance image
Middle water figure and fatty figure separate, it is thus achieved that carrying out effective water fat separation by obtaining real field map values.
Description of the drawings
Fig. 1 is the flow chart that water fat separation method is realized in one embodiment；
Fig. 2 is the procedure chart that the separation of water fat is realized in one embodiment；
Fig. 3 is neck and the result figure of ankle water fat separation in one embodiment；
Fig. 4 is pixel fit error curve figure in one embodiment；
Fig. 5 is the fit error curve figure of seed block in one embodiment；
Fig. 6 is the structure diagram that water fat piece-rate system is realized in one embodiment；
Fig. 7 is the structure diagram that water fat piece-rate system is realized in another embodiment；
Fig. 8 is the first structure diagram for choosing module in one embodiment；
Fig. 9 is the structure diagram that seed block chooses module in one embodiment；
Figure 10 is the structure diagram of acquisition module in one embodiment.
Specific embodiment
In order to make the purpose , technical scheme and advantage of the present invention be clearer, with reference to the accompanying drawings and embodiments, it is right
The present invention is further elaborated.It should be appreciated that specific embodiment described herein is only used to explain the present invention, not
For limiting the present invention.
In one embodiment, as shown in Figure 1, providing a kind of method for realizing the separation of water fat, this method includes：
Step 102, several magnetic resonance image that acquisition echo time interval is not waited.
Each echo forms a width magnetic resonance (Magnetic Resonance, abbreviation MR) image, this image can also
Referred to as original image.Several original images are acquired respectively by multiple echoes, between the echo time (echo time, abbreviation TE)
Every unequal (alternatively referred to as non-equidistant).Preferably, echo be 3, the corresponding echo time be respectively TE1, TE2 and
TE3, wherein, TE2-TE1 ≠ TE3-TE2.Specifically, can (fast low angle shot, low-angle excitation be fast by FLASH
Fast gtadient echo) more echo sequences scanning measurand, MR images are acquired, sweep parameter mainly includes：Repetition time TR
(repeat time, abbreviation TR), echo time TE, matrix size, thickness and flip angle.Using measurand as neck and ankle
For, the corresponding numerical value such as following table of corresponding each sweep parameter：
Step 104, several magnetic resonance image are divided into the sub-block of multiple and different resolution ratio.
Every width original image all has certain resolution ratio, is divided into original image according to the resolution ratio of original image more
A different sub-block.If the resolution ratio of original image is higher, the sub-block of a variety of different resolutions can be divided.It is for example, former
The resolution ratio of beginning image is 512 × 512, the number of pixels that the corresponding resolution ratio, that is, sub-block of sub-block includes can be 8 × 8,16 ×
16 and 32 × 32, thus original image is divided into the sub-block of 3 kinds of different resolutions, then the sub-block under 3 kinds of different resolutions
Number is respectively 64 × 64,32 × 32 and 16 × 16.If the resolution ratio of original image is relatively low, such as 256 × 256, then sub-block pair
The resolution ratio answered can be 8 × 8,16 × 16, thus original image is divided into the sub-block of 2 kinds of different resolutions.
Step 106, multiple sub-blocks are chosen with the minimum in error of fitting respectively as sub-block field corresponding with resolution ratio
Figure estimated value, and seed block corresponding with resolution ratio is chosen according to field figure estimated value respectively.
Each sub-block includes multiple pixels, and each pixel has corresponding field graph model error of fitting, referred to as
For error of fitting.The error of fitting of all pixels point in each sub-block is overlapped, the error of fitting after superposition is sub-block
Error of fitting.Minimum is chosen in the fit error curve of sub-block as sub-block field figure estimation corresponding with resolution ratio
Value.Select noise higher and only there are one minimum i.e. only there are one the sub-block of field figure estimated value as seed block.For example,
Original image includes the sub-block of 8 × 8,16 × 16 and 32 × 32 this 3 kinds of different resolutions, then needing will be under this 3 kinds of resolution ratio
Sub-block is overlapped operation one by one, obtains this corresponding field figure estimated value of 3 kinds of resolution ratio.It is required for selecting under each resolution ratio
Take corresponding seed block.
Step 108, corresponding original image point under different resolution is obtained using multiple seed blocks corresponding with resolution ratio
Pixel seed point under resolution, and pixel seed point merged to obtain final pixel seed point.
Under each resolution ratio, each seed block can obtain corresponding pixel seed point.Specifically, using seed block as rise
Point increases strategy using part and the true field map values of other sub-blocks is chosen, the sub-block after being increased.After all growths
Sub-block in pixel be the corresponding pixel seed point of original image resolution.Sub-block after increasing under different resolution is all
It can correspond to obtain pixel seed point, these different pixel seed points are merged, are obtained under original image resolution most
Whole pixel seed point.For example, resolution ratio, which is respectively sub-block after 8 × 8,16 × 16 and 32 × 32 growth, respectively obtains 3 kinds not
Same pixel seed point, this 3 kinds different pixel seed points are merged, then it is (original can to obtain 512 × 512 resolution ratio
Image resolution ratio) final pixel seed point.
Step 110, the true field map values of residual pixel point are chosen using final pixel seed point.
Using final pixel seed point as starting point, increase strategy to residual pixel point using the part under original image resolution
Field map values chosen.Residual pixel point refers to remove final pixel seed point under original image resolution in all pixels point
Pixel.Specifically, using the field map values of the final pixel seed point around residual pixel point as initial value, from the residual pixel
True field map values of the numerical value nearest from initial value as the residual pixel point are selected in the field map values candidate solution of point.It is if remaining
When having multiple final pixel seed points around pixel, using the average value of the field map values of multiple final pixel seed points as the residue
The initial value of pixel, and the final pixel seed point around the residual pixel point is more, then and the pixel seed point just has more
High priority carries out the selection of field map values.Residual pixel point and final pixel seed point merging composition are estimated newly most by above-mentioned
Whole pixel seed point repeats the above steps until whole residual pixel points are completed with local growth processing, acquisition all pixels point
True field map values.
Step 112, using the true field map values of all pixels point, several magnetic resonance image is carried out with calculating and respectively obtains water
Figure and fat figure.
The true field map values chosen using each pixel, several magnetic resonance image not waited echo time interval are
Original image respectively obtains water figure by following formula (1) and fat is schemed, is achieved in the separation of water fat.ρ=A+diag(ψ
(fB))S (1)
Wherein A+,Definition such as formula (4), ρ=[ρw, ρf] for plural number, the water figure that is as calculated and
Fat figure.Whole process can be with as shown in Fig. 2, wherein K indicates the different resolution ratio of K kinds.For example, measurand for neck and
Ankle, what is obtained is the separation of water fat as a result, as shown in figure 3, wherein (a) neck sagittal plane separating resulting；(b) neck coronal-plane point
From result；(c) ankle sagittal plane separating resulting.
In the present embodiment, by acquire echo time interval not wait several magnetic resonance image；By several magnetic resonance image
It is divided into the sub-block of multiple and different resolution ratio；Multiple sub-blocks are chosen with the minimum in error of fitting respectively as sub-block with differentiating
The corresponding field figure estimated value of rate, and seed block corresponding with resolution ratio is chosen according to field figure estimated value respectively；Using multiple with dividing
The corresponding seed block of resolution obtains the pixel seed point under corresponding original image resolution under different resolution, and by pixel kind
Son point, which merges, obtains final pixel seed point；The true field map values of residual pixel point are chosen using final pixel seed point；It utilizes
The true field map values of all pixels point carry out several magnetic resonance image calculating and respectively obtain water figure and fat figure.Due to using
Several magnetic resonance image that echo time interval is not waited which thereby enhance the efficiency that true field map values are chosen, due to utilizing difference
Seed block under resolution ratio obtains final pixel seed point, and the true field of each pixel is chosen using final pixel seed point
Map values, by the true field map values of each pixel, so as to water figure in several magnetic resonance image and fatty figure be separated, thus
It realizes and effective water fat separation is carried out by obtaining real field map values.
In one embodiment, it after the step of several magnetic resonance image that acquisition echo time interval is not waited, further includes：
The field map values distributed area of pixel in several magnetic resonance image is carried out discrete；The field map values obtained after discrete are calculated respectively
Error of fitting corresponding with pixel；Solve field figure estimated value of the minimum in error of fitting as pixel.
In the present embodiment, pixel in corresponding magnetic resonance image, that is, original image is chosen according to imaging parameters and imaging position
Field map values distributed area.Imaging parameters include repetition time TR, echo time TE, matrix size, thickness, flip angle etc..
Imaging position refers to the measurand of magnetic resonance imaging, such as neck, ankle.
Several magnetic resonance image that echo time interval is not waited are introduced into magnetic resonance signal model simultaneously, calculate each pixel
The corresponding field graph model error of fitting (abbreviation error of fitting) of point.Wherein, equation below may be used in magnetic resonance signal model：
Wherein N is echo number；ρwAnd ρfIt is the intensity value of water and fat, for plural number；fFIt is chemistry of the fat relative to water
Displacement is -3.5ppm, if in 3T (tesla, tesla's (International System of Units derived unit of magnetic induction intensity)) magnetic resonance system
In system, then fF=-431Hz；fBIt is the field map values of magnetostatic field B0.
If the content difference of water and fat is bigger in tissue, by taking pure water as an example, ρf=0, then formula (2) be reduced to：
From formula (3) as can be seen that there are two solve for field map values in the case of pure water：True solution fB(i.e. fB,true) i.e. true
Real field map values and water number ρw；Reversion solutionIt is worth with fatLikewise, for pure fat tissue
There are two solve for meeting：True solution fB(i.e. fB,tru) and fatty value ρf；Reversion solution(i.e. fB,swap) and water numberAnd
For the close tissue of water fat ratio, only one group of solution：Field map values fB, water number ρwWith fatty value ρf。
Using formula (4), solution makes error of fitting E (fB) reach minimum value, that is, field figure estimated value：
Wherein, subscript "+" represents M-P generalized inverses, | | | |2Expression takes L2 norms, and diag (X) represents the member using vector X
Element construction diagonal matrix, S=[S (t1),S(t2),…,S(tN)]H, A is the matrix of N × 2：
Each pixel has corresponding field map values, selects the distributed area of one section of field map values, and carries out discretization, example
If the distributed area of field map values is -1000~1000, by this distributed area progress discretization obtain it is discrete after field map values.According to
Illuminated (4) obtains a field graph model fit error curve, chooses the field map values that fit error curve is made to reach local minimum, i.e.,
For the corresponding field figure estimated value of pixel, according to the threshold value of pixel amplitude, using field figure estimated value as pixel field of points map values
Candidate solution.For the pixel that water fat content difference is larger, such as pure water or pure fat pixel, obtained best estimate is 2
It is a, F1 and F2 are expressed as, for the close pixel of water fat content, best estimate is 2, it is believed that F1=F2.
For example, the equal MR images of the MR images and 3 echo time intervals not waited at 3 echo time intervals are respectively adopted, obtain
To pixel field graph model fit error curve (abbreviation fit error curve) as shown in figure 4, wherein abscissa represent it is discrete
The field map values of change, ordinate represent error of fitting.Curve 1 is pixel in the MR images not waited using 3 echo time intervals
Fit error curve, curve 2 are the error of fitting of pixel in the MR images using 3 echo time intervals equal (i.e. equidistantly)
Curve.Straight line is pixel amplitude thresholds.Wherein, it is respectively f comprising two minimums in curve 11(fB,tru) and f2
(fB,swap), 3 minimums, respectively f are included in curve 21(fB,tru)、f2(fB,swap) and f3(fB,tru), wherein each minimum
Value all has periodically, and each period includes 2 minimum, that is, f1And f2, the period is f3With f1Between interval.It can from Fig. 4
With, it is evident that the MR images not waited as a result of echo time interval carry out the field figure estimated value of selected pixels point, so as to subtract
The number of minimum in pixel fit error curve is lacked, the number of field figure estimated value is limited to 2 or 1, is effectively carried
Height chooses the efficiency of true field map values.
In one embodiment, multiple sub-blocks are chosen with the minimum in error of fitting respectively as sub-block and resolution ratio pair
The field figure estimated value answered, and the step of seed block corresponding with resolution ratio is chosen according to field figure estimated value respectively include：By sub-block
The error of fitting of middle all pixels point corresponds to resolution ratio and is overlapped respectively, obtains the error of fitting after multiple superpositions；It asks respectively
Solve the field figure estimated value that the minimum in the error of fitting after multiple superpositions corresponds to resolution ratio as sub-block；According to pixel amplitude
Threshold value, the field figure estimated value that resolution ratio is corresponded to using sub-block chooses seed block under different resolution respectively.
In the present embodiment, by the error of fitting E (f of all pixels point in sub-block each under different resolutionB) be overlapped,
Error of fitting after being superimposedWherein r represents the pixel in sub-block R.Pixel amplitude also may be used
To be known as magnetic resonance image signal amplitude.Threshold value can be according to the 90th percentile of the amplitude of all pixels point in original image
Value, which is multiplied by 0.1 to obtain.Threshold value can be 40% or 50% of signal amplitude etc..Root in one of the embodiments,
According to the threshold value of pixel amplitude, the step of choosing the seed block under different resolution respectively using field figure estimated value, includes：According to
Sub-block is corresponded to candidate solution of the field figure estimated value as sub-block field map values of resolution ratio by the threshold value of pixel amplitude；With reference to pixel
The signal-to-noise ratio of point is chosen in candidate solution only there are one the sub-block of field figure estimated value as seed block, obtains different resolution respectively
Under multiple seed blocks.With reference to the signal-to-noise ratio of pixel, refer to choose by the sub-block of higher signal-to-noise ratio.The height of signal-to-noise ratio is
Opposite, in the present embodiment, a certain proportion of amplitude of all pixels point in sub-block can be more than threshold value, then it is assumed that the sub-block has
There is high s/n ratio.Pixel amplitude such as in all pixels point in the sub-block of one 16 × 16 70% is more than the threshold value, then the son
Block has high s/n ratio.If the sub-block only there are one field figure estimated value, i.e., error of fitting after sub-block superposition only there are one
Minimum, then the sub-block can be used as seed block.As shown in figure 5, the fit error curve for seed block.Its cathetus is threshold
It is worth, only there are one minimums in curve.For the low sub-block of signal-to-noise ratio, i.e. a certain proportion of amplitude of all pixels point is less than threshold value
Sub-block, the selection without field map values.
In one embodiment, it is obtained using multiple seed blocks corresponding with resolution ratio corresponding original under different resolution
Pixel seed point under image resolution ratio, and the step of pixel seed point is merged and obtains final pixel seed point include：With kind
Sub-block is starting point, increases the tactful true field map values for choosing sub-block to be estimated using part, obtains after increasing under different resolution
Sub-block；Using the true field map values of the sub-block after increasing as initial value, to the field map values of all pixels point of the sub-block after growth into
Row selection, obtains the pixel seed point under corresponding original image resolution under different resolution；It will be corresponding under different resolution
The pixel seed point of original image resolution merge, obtain final pixel seed point.
In the present embodiment, sub-block to be estimated refers to remove except seed block in all sub-blocks under each resolution ratio, and signal-to-noise ratio is high
Sub-block.Part increases strategy and refers to using the seed block field map values around sub-block to be estimated as initial value, from the candidate of sub-block to be estimated
True field map values of the numerical value nearest from initial value as the sub-block to be estimated are selected in solution.If there are multiple kinds around sub-block to be estimated
During sub-block, using the average value of the field map values of multiple seed blocks as the initial value of the sub-block to be estimated, and around sub-block to be estimated
Seed block is more, then the sub-block to be estimated just has the selection that higher priority carries out field map values.Seed block and local increasing is utilized
The sub-block after estimating sub-block and may be collectively referred to as increasing of long strategy.Using the sub-block after growth as new seed block, repeat
Above-mentioned steps until completing local growth processing to all sub-blocks to be estimated, obtain the true field map values of corresponding sub block.Utilize part
Increase strategy, obtain the sub-block after increasing under different resolution.
Using the true field map values of the sub-block after increasing as initial value, to the field map values of all pixels point of the sub-block after growth
It is selected, specifically, selecting the numerical value nearest from initial value in the best field figure estimated value of pixel from sub-block as this
The true field map values of pixel.Corresponding pixel is exactly the pixel seed under original image resolution in sub-block after all growths
Point.What there is the position of pixel seed point under different resolution is identical, some differences, and the identical multiple pixel seed points in position can be with
It can be regarded as a pixel seed point, if the pixel obtains different field map values under different resolution, abandon the picture of the position
Element is used as pixel seed point.It, thus will be all as the pixel seed point of corresponding position for the different pixel seed point in position
Pixel seed point under different resolution merges, and then obtains final pixel seed point.Due to son each under different resolution
The selection that the selection of block field map values, sub-block locally increase all pixels field map values in strategy and sub-block is all independent from each other,
It is exactly that processing between each resolution ratio is no any overlapping, so as to eliminate field figure that may be present between different resolution
It is worth the influence of saltus step, ensures the correctness that pixel seed point selects under original image resolution.
In one embodiment, as shown in fig. 6, providing a kind of system for realizing the separation of water fat, which includes：Acquisition
Module 602, division module 604, first choose module 606, acquisition module 608, second chooses module 610 and separation module 612,
Wherein：
Acquisition module 602, for acquiring several magnetic resonance image that echo time interval is not waited.
Division module 604, for several magnetic resonance image to be divided into the sub-block of multiple and different resolution ratio.
First chooses module 606, for multiple sub-blocks to be chosen with the minimum in error of fitting respectively as sub-block with dividing
The corresponding field figure estimated value of resolution, and seed block corresponding with resolution ratio is chosen according to field figure estimated value respectively.
Acquisition module 608, for obtaining corresponding original under different resolution using multiple seed blocks corresponding with resolution ratio
Pixel seed point under beginning image resolution ratio, and pixel seed point merged to obtain final pixel seed point.
Second chooses module 610, for choosing the true field map values of residual pixel point using final pixel seed point.
Separation module 612, for using the true field map values of all pixels point, to several magnetic resonance image calculate and divide
Water figure is not obtained and fat is schemed.
In one embodiment, as shown in fig. 7, the system further includes：Discrete block 614, fitting module 616 and first are asked
Module 618 is solved, wherein：
Discrete block 614 carries out discrete for the field map values distributed area to pixel in several magnetic resonance image.
Fitting module 616, for calculating error of fitting corresponding with pixel to the field map values obtained after discrete respectively.
First solves module 618, for solving field figure estimated value of the minimum in error of fitting as pixel.
In one embodiment, as shown in figure 8, the first selection module 606 includes：Laminating module 616, second solves module
626 and seed block choose module 636, wherein：
Laminating module 616 for the error of fitting of all pixels point in sub-block to be corresponded to resolution ratio is overlapped, obtains respectively
Error of fitting to after multiple superpositions.
Second solves module 626, for solving the minimum in the error of fitting after multiple superpositions respectively as sub-block pair
Answer the field figure estimated value of resolution ratio.
Seed block chooses module 636, and for the threshold value according to pixel amplitude, the field figure that resolution ratio is corresponded to using sub-block is estimated
Evaluation chooses the seed block under different resolution respectively.
In one embodiment, as shown in figure 9, seed block selection module 636 includes：Candidate solution selecting unit 636a and kind
Sub-block selection unit 636b, wherein：
Candidate solution selecting unit 636a, for the threshold value according to pixel amplitude, the field figure that sub-block is corresponded to resolution ratio is estimated
Candidate solution of the evaluation as sub-block field map values.
Seed block selection unit 636b for the signal-to-noise ratio with reference to the pixel, chooses only one in candidate solution respectively
The sub-block of a figure estimated value obtains multiple seed blocks under different resolution as seed block.
In one embodiment, as shown in Figure 10, acquisition module 608 includes：Increase module 618, pixel seed point obtains
Module 628 and final pixel seed point acquisition module 638, wherein：
Increase module 618, for using seed block as starting point, increasing the tactful true field figure for choosing sub-block to be estimated using part
Value obtains the sub-block after increasing under different resolution.
Pixel seed point acquisition module 628, for using the true field map values of the sub-block after increasing as initial value, after growth
The field map values of all pixels point of sub-block selected, obtain the picture under corresponding original image resolution under different resolution
Plain seed point.
Final pixel seed point acquisition module 638, for by the picture of original image resolution corresponding under different resolution
Plain seed point merges, and obtains final pixel seed point.
Above example only expresses the several embodiments of the present invention, and description is more specific and detailed, but can not
Therefore it is interpreted as the limitation to the scope of the claims of the present invention.It should be pointed out that for those of ordinary skill in the art,
Without departing from the inventive concept of the premise, various modifications and improvements can be made, these belong to the protection model of the present invention
It encloses.Therefore, the protection domain of patent of the present invention should be determined by the appended claims.
Claims (10)
1. a kind of method for realizing the separation of water fat, the method includes：
Acquire several magnetic resonance image that echo time interval is not waited；The magnetic resonance image has corresponding resolution ratio；
Each width magnetic resonance image is divided into the sub-block with multiple and different resolution ratio；
Multiple sub-blocks are chosen with minimum in error of fitting respectively as sub-block field figure estimated value corresponding with resolution ratio,
And seed block corresponding with resolution ratio is chosen according to the field figure estimated value respectively；
It is obtained under different resolution under corresponding original image resolution using multiple seed blocks corresponding with resolution ratio
Pixel seed point, and the pixel seed point is merged to obtain final pixel seed point；
The true field map values of residual pixel point are chosen using the final pixel seed point；
Using the true field map values of all pixels point, several described magnetic resonance image are carried out with calculating and respectively obtains water figure and fat
Figure.
2. according to the method described in claim 1, it is characterized in that, several magnetic resonance that the acquisition echo time interval is not waited
After the step of image, further include：
The field map values distributed area of pixel in several magnetic resonance image is carried out discrete；
Error of fitting corresponding with pixel in the magnetic resonance image is calculated to the field map values obtained after discrete respectively；
Solve field figure estimated value of the minimum as pixel in the magnetic resonance image in the error of fitting.
3. the according to the method described in claim 2, it is characterized in that, pole multiple sub-blocks chosen respectively in error of fitting
Small value is chosen according to the field figure estimated value and resolution ratio respectively as corresponding with the resolution ratio field figure estimated value of the sub-block
The step of corresponding seed block, includes：
The error of fitting of all pixels point in the sub-block is corresponded to resolution ratio to be overlapped respectively, obtains the plan after multiple superpositions
Close error；
The minimum in the error of fitting after the multiple superposition is solved respectively the field figure of resolution ratio is corresponded to as the sub-block estimate
Evaluation；
According to the threshold value of pixel amplitude, the field figure estimated value that resolution ratio is corresponded to using the sub-block chooses different resolution respectively
Under seed block.
4. according to the method described in claim 3, it is characterized in that, the threshold value according to the pixel amplitude, utilizes institute
A step of figure estimated value chooses the seed block under different resolution respectively is stated to include：
According to the threshold value of pixel amplitude, the sub-block is corresponded into the field figure estimated value of resolution ratio as sub-block field map values
Candidate solution；
With reference to the signal-to-noise ratio of pixel in sub-block, choose in the candidate solution that only there are one the sub-block conducts of field figure estimated value respectively
Seed block obtains multiple seed blocks under different resolution.
5. according to the method described in claim 4, it is characterized in that, described utilize multiple seed blocks corresponding with resolution ratio
The pixel seed point under corresponding original image resolution under different resolution is obtained, and the pixel seed point is merged to obtain
The step of final pixel seed point, includes：
Using the seed block as starting point, increase the tactful true field map values for choosing sub-block to be estimated using part, obtain different resolutions
Sub-block after increasing under rate；
Using the true field map values of the sub-block after described increase as initial value, to the field of all pixels point of the sub-block after the growth
Map values are selected, and obtain the pixel seed point under corresponding original image resolution under different resolution；
The pixel seed point of original image resolution corresponding under the different resolution is merged, obtains final pixel kind
Sub- point.
6. a kind of system for realizing the separation of water fat, which is characterized in that the system comprises：
Acquisition module, for acquiring several magnetic resonance image that echo time interval is not waited；The magnetic resonance image, which has, to be corresponded to
Resolution ratio；
Division module, for each width magnetic resonance image to be divided into the sub-block with multiple and different resolution ratio；
First chooses module, for multiple sub-blocks to be chosen with the minimum in error of fitting respectively as the sub-block and resolution ratio
Corresponding field figure estimated value, and seed block corresponding with resolution ratio is chosen according to the field figure estimated value respectively；
Acquisition module, for obtaining corresponding original graph under different resolution using multiple seed blocks corresponding with resolution ratio
Merge to obtain final pixel seed point as the pixel seed point under resolution ratio, and by the pixel seed point；
Second chooses module, for choosing the true field map values of residual pixel point using the final pixel seed point；
Separation module, for using the true field map values of all pixels point, calculating difference to be carried out to several described magnetic resonance image
It obtains water figure and fat is schemed.
7. system according to claim 6, which is characterized in that the system also includes：
Discrete block, for in several magnetic resonance image the field map values distributed area of pixel carry out it is discrete；
Fitting module, for calculating plan corresponding with pixel in the magnetic resonance image to the field map values obtained after discrete respectively
Close error；
First solves module, for solving field of the minimum in the error of fitting as pixel in the magnetic resonance image
Figure estimated value.
8. system according to claim 7, which is characterized in that the first selection module includes：
Laminating module for the error of fitting of all pixels point in the sub-block to be corresponded to resolution ratio is overlapped, obtains respectively
Error of fitting after multiple superpositions；
Second solves module, for solving the minimum in the error of fitting after the multiple superposition respectively as the sub-block pair
Answer the field figure estimated value of resolution ratio；
Seed block chooses module, and for the threshold value according to pixel amplitude, the field figure that resolution ratio is corresponded to using the sub-block is estimated
Value chooses the seed block under different resolution respectively.
9. system according to claim 8, which is characterized in that the seed block is chosen module and included：
The sub-block for the threshold value according to pixel amplitude, is corresponded to the field figure estimated value of resolution ratio by candidate solution selecting unit
Candidate solution as sub-block field map values；
Seed block selection unit, for combining the signal-to-noise ratio of pixel in sub-block, choose respectively in the candidate solution only there are one
The sub-block of field figure estimated value obtains multiple seed blocks under different resolution as seed block.
10. system according to claim 9, which is characterized in that the acquisition module includes：
Increase module, for using the seed block as starting point, increasing the tactful true field map values for choosing sub-block to be estimated using part,
Obtain the sub-block after increasing under different resolution；
Pixel seed point acquisition module, for using the true field map values of the sub-block after described increase as initial value, to the growth
The field map values of all pixels point of sub-block afterwards are selected, and are obtained under different resolution under corresponding original image resolution
Pixel seed point；
Final pixel seed point acquisition module, for by the pixel kind of original image resolution corresponding under the different resolution
Son point merges, and obtains final pixel seed point.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410849730.9A CN105796102B (en) | 2014-12-30 | 2014-12-30 | Realize the method and system of water fat separation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410849730.9A CN105796102B (en) | 2014-12-30 | 2014-12-30 | Realize the method and system of water fat separation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105796102A CN105796102A (en) | 2016-07-27 |
| CN105796102B true CN105796102B (en) | 2018-07-03 |
Family
ID=56421272
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410849730.9A Active CN105796102B (en) | 2014-12-30 | 2014-12-30 | Realize the method and system of water fat separation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105796102B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR101844514B1 (en) * | 2016-09-02 | 2018-04-02 | 삼성전자주식회사 | Magnetic resonance imaging apparatus and method of obtaining magnetic resonance image |
| CN111047597B (en) * | 2019-12-30 | 2023-04-25 | 中国科学院武汉物理与数学研究所 | Multi-echo water-fat separation method based on deep learning |
| CN113470032B (en) * | 2021-05-25 | 2022-10-18 | 上海东软医疗科技有限公司 | Water-fat separation method and device based on magnetic resonance imaging and computer equipment |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102736047A (en) * | 2011-04-13 | 2012-10-17 | 深圳迈瑞生物医疗电子股份有限公司 | Magnetic resonance system, water fat separation imaging method thereof, and device thereof |
| CN103513202A (en) * | 2012-06-16 | 2014-01-15 | 上海联影医疗科技有限公司 | DIXON water-fat separation method in magnetic resonance imaging |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP4251763B2 (en) * | 2000-08-11 | 2009-04-08 | 株式会社日立メディコ | Magnetic resonance imaging system |
| US6603990B2 (en) * | 2001-08-10 | 2003-08-05 | Toshiba America Mri, Inc. | Separation and identification of water and fat MR images at mid-field strength with reduced T2/T2* weighting |
-
2014
- 2014-12-30 CN CN201410849730.9A patent/CN105796102B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102736047A (en) * | 2011-04-13 | 2012-10-17 | 深圳迈瑞生物医疗电子股份有限公司 | Magnetic resonance system, water fat separation imaging method thereof, and device thereof |
| CN103513202A (en) * | 2012-06-16 | 2014-01-15 | 上海联影医疗科技有限公司 | DIXON water-fat separation method in magnetic resonance imaging |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105796102A (en) | 2016-07-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US8023705B2 (en) | Method for reconstructing image from echo planar imaging sequence | |
| EP2798364B1 (en) | Mr imaging with suppression of flow artefacts | |
| CN104068859A (en) | Method and magnetic resonance system to generate multiple magnetic resonance images | |
| CN105796102B (en) | Realize the method and system of water fat separation | |
| CN103513202A (en) | DIXON water-fat separation method in magnetic resonance imaging | |
| CN107076819B (en) | Dixon MR imaging with suppression of flow artifacts | |
| CN103487773A (en) | System for reconstruction of virtual frequency selective inversion MR images | |
| CN104545914B (en) | Method for separate imaging of water and fat and system | |
| EP2495579B1 (en) | Method for generating 2D or 3D maps of MRI T1 and T2 relaxation times | |
| Gao et al. | Improving the temporal resolution of functional MR imaging using keyhole techniques | |
| Zhu et al. | Arterial spin labeling perfusion MRI signal denoising using robust principal component analysis | |
| JP4826332B2 (en) | Magnetic resonance measuring device | |
| JP2014200687A (en) | Method and magnetic resonance system to implement multiecho measurement sequence | |
| CN106137198B (en) | Magnetic resonance imaging method and device | |
| CN108377641B (en) | Magnetic resonance vessel wall imaging method and apparatus | |
| CN105809662B (en) | The image water fat separation method and system of magnetic resonance imaging | |
| JP6637053B2 (en) | Method and system for susceptibility weighted magnetic resonance imaging | |
| CN107589387B (en) | Magnetic resonance imaging method and apparatus | |
| Xue et al. | Motion compensated magnetic resonance reconstruction using inverse-consistent deformable registration: application to real-time cine imaging | |
| KR20130046517A (en) | Method and mri device for correcting distortion in an epi image | |
| CN104918546A (en) | Magnetic resonance imaging device and processing method thereof | |
| CN105467342A (en) | Method and device for reconstructing magnetic resonance multichannel acquired image | |
| US20120235682A1 (en) | Method and apparatus for acquiring magnetic resonance imaging signals | |
| US7187170B1 (en) | Multiple acquisition phase-sensitive SSFP for species separating in MRI | |
| TWI529405B (en) | Method and apparatus for acquiring magnetic resonance imaging signals |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |