CN106324537A - Super-fast sectional type single-shot water-fat separation method - Google Patents

Abstract

The invention discloses a super-fast sectional type single-shot water-fat separation method. Super-fast chemical shift imaging sequence, signal analysis model, echo amplitude attenuation compensation methods as well as rapid phase unwrapping and precision correction technologies are applied to a magnetic resonance imaging system, thereby realizing water-fat separation enhanced scanning and thin-layer water-fat separation imaging. The scanning efficiency can be improved substantially and image artifacts can be eliminated based on phase correction. And with a two-dimensional version, clinical application of chemical shift imaging in enhanced scanning can be realized. The method can be used for scanning of the infant or hyperactivity patient.

Description

A kind of supper-fast stagewise single-shot water fat separation method

Technical field

The present invention relates to medical nuclear magnetic resonance imaging technical field, particularly relate to a kind of supper-fast stagewise single-shot water fat Separation method.

Background technology

In medical magnetic resonance imaging (MRI), chemical shift imaging (also known as spectrum imaging) is than conventional selected fat magnetic Change saturated, water selective to excite and recover (STIR) pressure fat imaging technique with short TI and be more beneficial for the clear of pathological changes and surrounding tissue thereof Display, to radio-frequency field B₁Uniformity and magnetic field intensity do not have particular/special requirement the most not lose signal noise ratio (snr) of image, it is also possible to mensuration group Knit the relative scale of middle fat and water.Dixon water fat isolation technics is one of chemical shift imaging technology, based on spin echo 2 and 3 Dixon separate imaging of water and fat technology are applied relatively broad clinically, but sweep time sweeps for conventional T1 weighting The twice or three times retouched, reduces the efficiency of clinical scanning and diagnosis.At present, use single-shot gather simultaneously two homophases and 3 Dixon technology of one anti-phase echo have replaced the imaging system of tradition Dixon technology several famous brand names at home and abroad On put it over and realize routine clinical application, but up to now, separate imaging of water and fat technology is controlled owing to being difficult to sweep time System was never applied to T1 within 10 seconds clinically and weights enhanced ct scans, also cannot be existed by supper-fast scan mode Infant and hyperkinetic syndrome patient imaging procedure are avoided the interference of motion artifacts, to such an extent as to this technology presses fat technology relative to other Advantage cannot give full play to clinically.It practice, exciting of long echo string is conducive to adding fast scan speed with acquisition mode, In fast spin echo (FSE) and echo-planar imaging (EPI), the most just obtain application, but this excite and the side of collection The adjoint phase error of formula can cause multiple image artifacts, and this was once FSE and EPI routinizes difficult point that application faces through relatively Development efforts over a long time just solves.The phase place of magnetization vector and the time thereof for separate imaging of water and fat, water and fat constituent Dependent request radio-frequency pulse and gradient pulse technology accurately control, and the exciting with acquisition mode due to hardware of long echo string Imperfection and vortex field and Maxwell's field effect can produce significantly larger phase error accumulation.Magnet system and gradient The process modification of system has a maturation process the veryest long also to be limited by current industrial basis, and this is accomplished by strong Data processing technique solve these technical barriers.

Summary of the invention

It is an object of the invention to provide and a kind of realize three-dimensional stagewise single-shot N point Dixon water based on gtadient echo Fat method for separate imaging, can increase substantially scan efficiency and eliminate image artifacts by phase correction techniques.

The invention provides a kind of supper-fast stagewise single-shot water fat separation method to comprise the following steps:

Step S100: according to prescan needs, load apparent horizontal relaxation on industrial computer sequencer by control station The Henan timeCycle tests.

Step S200: load two-dimensional chemical displacement imaging sequence module on sequencer by control station.

2-dimensional gradient echo sequence basis utilize 2N positive-negative polarity frequency encoding gradient alternately formed by N group water The gtadient echo string that fat same phase echo and antiphase echo are alternately constituted；N number of homophase echo is carrying out Gradient Phase volume respectively Code, k-space is divided into N number of region correspondingly, and first echo fills central area, and next echo is sequentially filled next-door neighbour Region, and use part Fourier mode to reconstruct same phasor；N number of anti-phase echo carries out Gradient Phase coding respectively, therewith Correspondingly k-space is divided into N number of region, and first echo fills central area, and next echo is sequentially filled immediate area, and Part Fourier mode is used to reconstruct anti-phase figure.

3-dimensional gradient echo sequence basis utilize 2N positive-negative polarity frequency encoding gradient alternately formed by N number of water The gtadient echo string that fat same phase echo and N number of antiphase echo are alternately constituted；N number of homophase echo carries out Gradient Phase volume respectively Code, k-space is divided into N number of region correspondingly, and first echo fills central area, and next echo is sequentially filled next-door neighbour Region, then carries out 3 d-dem inverse Fourier transform and produces a series of thin layer homophase figures；N number of anti-phase echo carries out gradient respectively Phase code, k-space is divided into N number of region correspondingly, and first echo fills central area, and next echo is filled out successively Fill immediate area, then carry out 3 d-dem inverse Fourier transform and produce a series of thin layer homophase figures.

Step S300: utilize general theoretical model gtadient echo string set up in spatial frequency domain and image area, correlation Frequently during the TE that pulse excitation to first echo is formed, the impact of phase of echo and amplitude is corrected by magnetic field bump:

$I_{2N-2}^t=\underset{-\mathrm{\∞}}{\overset{+\mathrm{\∞}}{\∫}}\{\left(M_w\right(r)+M_f\left(r\right)e^{-i2\mathrm{\π}\mathrm{\Δ}f\·\[TE+(2N-2)\mathrm{\Δ}\mathrm{\τ}\]})e^{-(2N-1)\mathrm{\Δ}\mathrm{\τ}/T_2^{*}}{e}^{-i\[{\mathrm{\φ}}_0\left(r\right)+(2N-1)\mathrm{\φ}\left(r\right)\]}\}{d}^{3}r---\left(1\right)$

$I_{2N-1}^t=\underset{-\mathrm{\∞}}{\overset{+\mathrm{\∞}}{\∫}}\{\left(M_w\right(r)+M_f\left(r\right)e^{-i2\mathrm{\π}\mathrm{\Δ}f\·\[TE+(2N-2)\mathrm{\Δ}\mathrm{\τ}\]})e^{-2N\·\mathrm{\Δ}\mathrm{\τ}/T_2^{*}}{e}^{-i\[{\mathrm{\φ}}_0\left(r\right)+2N\·\mathrm{\φ}\left(r\right)\]}\}{d}^{3}r---\left(2\right)$

$S_0=(S_w-S_f)\·e^{i\stackrel{\‾}{\mathrm{\φ}}}---\left(11\right)$

$S_1=(S_w+S_f)\·A\·e^{i(\stackrel{\‾}{\mathrm{\φ}}+\mathrm{\φ})}---\left(12\right)$

Wherein, t refers to current phase code circulation step, and N arrives 2L scope round numbers, φ 1₀R () is proton magnetization vector Initial phase, φ (r) is the phase error mainly produced by Magnetic field inhomogeneity effect,Item is for carrying out excellent to echo amplitude Change and compensate.

S_wAnd S_fRepresent water and fat components in image pixel, they and M respectively_wAnd M_fIt is directly proportional,Including anti-phase figure Initial phase and the phase error that field inhomogeneities, vortex field and Maxwell field produce during TE and (2N-2) Δ τ,Main by first anti-phase echo (i.e. k_y=0 line) phase error during TE determines.

M_w(r) and M_fR () is expressed as water and the proton magnetization vector of fat in region, at three dimensions encoding strip Can be described as under part:

$M_w\left(r\right)=M_w^0{e}^{-i\mathrm{\γ}{\∫}_0^t\[x\·G_x\left(\mathrm{\τ}\right)+y\·G_y\left(\mathrm{\τ}\right)+z\·G_z\left(\mathrm{\τ}\right)\]d\mathrm{\τ}}---\left(3\right)$

$M_f\left(r\right)=M_f^0{e}^{-i\mathrm{\γ}{\∫}_0^t\[x\·G_x\left(\mathrm{\τ}\right)+y\·G_y\left(\mathrm{\τ}\right)+z\·G_z\left(\mathrm{\τ}\right)\]d\mathrm{\τ}}---\left(4\right);$

Step S400: water is as S_wWith fat as S_fCalculate generation according to the following formula:

$S_w=\frac{\mathrm{\κ}\·\left|S_0^{`}\right|+\left|S_1^{`}\right|/A}{2}$

$S_w=\frac{\left|S_1^{`}\right|/A-\mathrm{\κ}\·\left|S_0^{`}\right|}{2}.$

In step S100, echo amplitude apparent T2 based on imaging regionTest value compensates school Just, used test sequence includes the unit such as slice selective gradient, frequency encoding gradient and echo acquirement, and water mould is placed in imaging area Territory scope or direct location at human body are chosen on picture, as thick-layer, arrange echo time TE and take different time between 1ms and 1s Numerical value also meets TE=n/f condition, and n is natural number, then based on single exponential function, the echo amplitude of different TE is carried out non-thread Property matching obtain

In step S200, select k-space centrage, after one-dimensional inverse Fourier transform, amplitude is more than maximum 10% The phase place of complex data point carries out primary phase correction by linear fit method.

In step S300, from anti-phase echo I_2N-2Or homophase echo I_2N-1In choose two of which in frequency encoding gradient pole Property be positive and negative in the case of, the k-space centrage k gathered respectively_y=0 carries out one-dimensional discrete along frequency coding direction becomes against Fourier After changing, obtained the phase error of centrage by complex operation, the most accordingly to all k-space lines at one-dimensional discrete against Fourier The laggard line phase of leaf transformation corrects, thus eliminates the non-equilibrium effect of bipolarity frequency encoding gradient.

In step 300, from anti-phase echo I_2N-2Or homophase echo I_2N-1In choose two of which in frequency encoding gradient polarity In the case of positive and negative, gather central area k-space line respectively, carry out one-dimensional discrete along frequency coding direction respectively and become against Fourier Obtain the phase error of each line in central area by complex operation after changing and be averaging, more accordingly to all k-space lines one-dimensional The laggard line phase of discrete inverse Fourier transform corrects, and eliminates the non-equilibrium effect of bipolarity frequency encoding gradient.

In step S300, the k-space data that the homophase echo of twice collection is corresponding with anti-phase echo is carried out respectively Image reconstruction separates with water fat, and then water picture and fat are as being overlapped respectively, eliminate irreducible phase errors.

Beneficial effect: the present invention is mended by supper-fast chemical shift imaging sequence, signal analysis model, echo amplitude decay Compensation method solution quick with phase place twines and accurately alignment technique separates enhanced ct scans for realizing water fat on magnetic resonance imaging system With thin layer separate imaging of water and fat；Image artifacts is eliminated by phasing while can increasing substantially scan efficiency；Two dimension version Make chemical shift imaging can realize clinical practice in enhanced ct scans first it can also be used to baby and hyperkinetic syndrome patient sweep Retouch.

Accompanying drawing explanation

The T of Fig. 1 embodiment of the present invention₂ ^*Cycle tests.

Wherein, TE=n/ Δ f, Δ f are water fat chemical shift difference, and n is natural number, and TE is value between 1ms and 1s.

The two-dimentional stagewise single-shot water fat separation sequence of Fig. 2 embodiment of the present invention.

Wherein, sequence start-up portion gathers one group of 2 Dixon echo, and follow-up repetitive gathers 2 Dixon of N-1 group Echo, the time interval of homophase echo and anti-phase echo is TE=Δ τ=1/ Δ f/2, applies phase place and compile before each group echo samples Code gradient, the phase encoding gradient of after-applied polarity inversion of sampling, last is organized after echo acquirement terminates at three gradient directions All apply luffing and damage phase gradient to eliminate remaining Mxy.

The three-dimensional stagewise single-shot water fat separation sequence of Fig. 3 embodiment of the present invention.

Wherein, sequence start-up portion gathers one group of 2 Dixon echo, and follow-up repetitive gathers N-1 group two point Dixon echo, the time interval of homophase echo and anti-phase echo is TE=Δ τ=1/ Δ f/2, applies phase before each group echo samples Position coding gradient and select layer direction phase encoding gradient, the phase encoding gradient of after-applied polarity inversion of sampling and select layer direction phase Position coding gradient, it is remaining horizontal to eliminate that last group echo acquirement all applies luffing damage phase gradient at three gradient directions after terminating To magnetization vector.

Fig. 4 clinical scanning of the present invention workflow.

Fig. 5 present invention quickly accurate phase unwrapping algorithm flow chart.

The k-space filling mode of each imaging aspect in the case of the enhanced ct scans of Fig. 6 embodiment of the present invention.

Wherein, the solid line of signals collecting part represents reverse phase gradient echo, and dotted line represents homophase gtadient echo.

The k-space filling mode of each imaging aspect in the case of the non-reinforcing scanning of Fig. 7 embodiment of the present invention.

Wherein, the solid line of signals collecting part represents reverse phase gradient echo, and dotted line represents homophase gtadient echo.

Fig. 8 embodiment of the present invention two dimension stagewise single-shot water fat separation sequence.

Wherein, sequence start-up portion gathers one group of 3 Dixon echo, and follow-up repetitive gathers 2 Dixon of N-1 group Echo, the time interval of homophase echo and anti-phase echo is TE=Δ τ=1/ Δ f/2, applies phase place and compile before each group echo samples Code gradient, the phase encoding gradient of after-applied polarity inversion of sampling, last is organized after echo acquirement terminates at three gradient directions All apply luffing and damage phase gradient to eliminate remaining Mxy.

The three-dimensional stagewise single-shot water fat separation sequence of Fig. 9 embodiment of the present invention.

Wherein, sequence start-up portion gathers one group of 3 Dixon echo, and follow-up repetitive gathers 2 Dixon of N-1 group Echo, the time interval of homophase echo and anti-phase echo is TE=Δ τ=1/ Δ f/2, applies phase place and compile before each group echo samples Code gradient and select layer direction phase encoding gradient, the phase encoding gradient of after-applied polarity inversion of sampling and select layer direction phase place to compile Code gradient, last group echo acquirement all applies luffing at three gradient directions after terminating and damages phase gradient to eliminate remaining transverse magnetic Change vector.

Detailed description of the invention

Technical scheme and the technique effect reached for making to present invention solves the technical problem that, using are clearer, below The present invention is described in further detail in conjunction with the accompanying drawings and embodiments.It is embodied as it is understood that described herein Example is used only for explaining the present invention, rather than limitation of the invention.It also should be noted that, for the ease of describing, accompanying drawing In illustrate only part related to the present invention rather than full content.

One, the principle of the present invention

Medical magnetic resonance imaging instrument generally by magnet, industrial computer, control station, gradient coil, radio-frequency coil, radio-frequency (RF) power amplification and The hardware cells such as gradient power amplifier are constituted.It is each to control that pulse train module shown in Fig. 1 to Fig. 3 is loaded into industrial computer by control station Hardware cell realizes the exciting of human body magnetic resonance signal, space encoding and collection, and the data that control station main frame is installed process mould Block carries out primary phase correction, image reconstruction, phase unwrapping and field inhomogeneities (Δ B to magnetic resonance signal₀) correction, amplitude school Just separate with water fat.Whole data acquisition and processing (DAP) process performs automatically according to the scanning process shown in Fig. 4.

First, according to prescan needs, on industrial computer sequencer, the apparent horizontal stroke shown in Fig. 1 is loaded by control station To the relaxation timeCycle tests, this sequence includes the unit such as slice selective gradient, frequency encoding gradient and echo acquirement.First Human body location picture is chosen to, as thick-layer, arrange echo time TE and take different time numerical value between 1ms and 1s and meet TE =n/ Δ f condition (n is natural number), then the echo amplitude that different TE are gathered based onCarry out nonlinear fitting ObtainHere I is the magnetization vector of water fat, I₀For I in the initial value of equilibrium state.Then, according to the vortex field of gradient system Characteristic carries out preemphasis to gradient waveform.

Then, by control station load on sequencer the three dimensional chemical displacement imaging sequence module shown in Fig. 3 or Two-dimensional chemical displacement imaging sequence module shown in Fig. 2.This module includes 90 ° of excitation pulses, slice selective gradient G_s, phase code ladder Degree G_p2, positive-negative polarity frequency encoding gradient G_rAnd the unit such as echo collection, wherein three-dimensional version also includes selecting layer direction phase place Coding gradient G_p1, echo time TE=1/ Δ f/2 is set in sequence, and the time interval Δ between positive-negative polarity gradient is set τ=1/ Δ f/2.The a length of L of homophase (or anti-phase) echo of single-shot is set in sequential parameter table, frequency coding is set The data point in direction, phase-encoding direction and choosing layer direction is respectively Dim1, Dim2 and Dim3, and other imaging parameters can refer to often Rule GRE T1WI imaging requirements is arranged.This chemical shift imaging sequence carries out fractional scanning with repeatedly mode of excitation, each anti-phase (or homophase) echo carries out phase code by the way of shown in Fig. 6 or Fig. 7, and the total degree of phase code circulation is Dim2/L.k Space is divided into L region by L-value, and first homophase (or anti-phase) echo fills k-space central area, second homophase (or Anti-phase) echo filling immediate area, follow-up each homophase (or anti-phase) echo is sequentially filled each region of periphery.It addition, reading gradient K-space centrage (k under the conditions of polarity inversion_y=0) or the once and separately storage of central area or whole k-space repeated acquisition is One R-matrix.The k-space data corresponding with anti-phase image with phase images is stored respectively for two-dimensional complex number square on control station Battle array (Dim1 × Dim2) or three-dimensional complex matrix (Dim1 × Dim2 × Dim3).For three-dimensional version, control station data processing module First along selecting layer direction to carry out one-dimensional discrete Fourier transform, it is thus achieved that the k-space complex matrix of each imaging aspect.For each Imaging aspect, anti-phase echoWith homophase echoTheoretical modeling as follows:

$I_{2N-2}^t=\underset{-\mathrm{\∞}}{\overset{+\mathrm{\∞}}{\∫}}\{\left(M_w\right(r)+M_f\left(r\right)e^{-i2\mathrm{\π}\mathrm{\Δ}f\·\[TE+(2N-2)\mathrm{\Δ}\mathrm{\τ}\]})e^{-(2N-1)\mathrm{\Δ}\mathrm{\τ}/T_2^{*}}{e}^{-i\[{\mathrm{\φ}}_0\left(r\right)+(2N-1)\mathrm{\φ}\left(r\right)\]}\}{d}^{3}r---\left(1\right)$

$I_{2N-1}^t=\underset{-\mathrm{\∞}}{\overset{+\mathrm{\∞}}{\∫}}\{\left(M_w\right(r)+M_f\left(r\right)e^{-i2\mathrm{\π}\mathrm{\Δ}f\·\[TE+(2N-2)\mathrm{\Δ}\mathrm{\τ}\]})e^{-2N\·\mathrm{\Δ}\mathrm{\τ}/T_2^{*}}{e}^{-i\[{\mathrm{\φ}}_0\left(r\right)+2N\·\mathrm{\φ}\left(r\right)\]}\}{d}^{3}r---\left(2\right)$

In above formula, t refers to current phase code circulation step, and N arrives 2L scope round numbers, φ 1₀R () is proton magnetization vector Initial phase, φ (r) is the phase error mainly produced by Magnetic field inhomogeneity effect,Item is for carrying out excellent to echo amplitude Change and compensate, M_w(r) and M_fR () is expressed as water and the proton magnetization vector of fat in region, at three dimensions encoding condition Under can be described as:

$M_w\left(r\right)=M_w^0{e}^{-i\mathrm{\γ}{\∫}_0^t\[x\·G_x\left(\mathrm{\τ}\right)+y\·G_y\left(\mathrm{\τ}\right)+z\·G_z\left(\mathrm{\τ}\right)\]d\mathrm{\τ}}---\left(3\right)$

$M_f\left(r\right)=M_f^0{e}^{-i\mathrm{\γ}{\∫}_0^t\[x\·G_x\left(\mathrm{\τ}\right)+y\·G_y\left(\mathrm{\τ}\right)+z\·G_z\left(\mathrm{\τ}\right)\]d\mathrm{\τ}}---\left(4\right)$

In formula (1) and (2), the field inhomogeneities effect during TE is not ignored, this is because TE is in the many imagings of needs Actually cannot usually reach several milliseconds even more than ten milliseconds close to zero, and gtadient echo can not reunion in system Displacement study effect can not be met again an inhomogeneous broadening effect.Under the conditions of the Δ f/2 of TE=Δ τ=1/, formula (1) and (2) are reduced to

$I_{2N-2}^t=\[M_w\left(r\right)+M_f\left(r\right)\]e^{-(2N-1)\mathrm{\Δ}\mathrm{\τ}/T_2^{*}}{e}^{-i\[{\mathrm{\φ}}_0\left(r\right)+(2N-1)\mathrm{\φ}\left(r\right)\]}---\left(5\right)$

$I_{2N-1}^t=\[M_w\left(r\right)+M_f\left(r\right)\]e^{-2N\·\mathrm{\Δ}\mathrm{\τ}/T_2^{*}}{e}^{-i\[{\mathrm{\φ}}_0\left(r\right)+2N\·\mathrm{\φ}\left(r\right)\]}---\left(6\right)$

In above formula,

$I_0^t=\[M_w\left(r\right)-M_f\left(r\right)\]\·A\·e^{-i\[{\mathrm{\φ}}_0\left(r\right)+\mathrm{\φ}\left(r\right)\]}---\left(7\right)$

$I_1^t=\[M_w\left(r\right)+M_f\left(r\right)\]\·A^2\·e^{-i\[{\mathrm{\φ}}_0\left(r\right)+2\mathrm{\φ}\left(r\right)\]}---\left(8\right)$

Here we defineBased on above measureValue can be calculated A.

It addition, can determine that the first two echo is as follows with the relation of other echo from formula (5) to (8):

$I_{2N-2}^t=I_0^t{A}^{2N-2}{\left(e^{-i\mathrm{\φ}\left(r\right)}\right)}^{(2N-2)}---\left(9\right)$

$I_{2N-1}^t=I_1^t{A}^{2N-2}{\left(e^{-i\mathrm{\φ}\left(r\right)}\right)}^{(2N-2)}---\left(10\right)$

It is visible,WithBetween amplitude and phase relation be quite analogous toWithBetween amplitude and phase place close System, therefore can use similar data processing method to each group homophase in the echo of each radio-frequency drive with anti-phase echo.

Data processing module starts primary in the case of gradient system lacks vortex field and Maxwell field suppression function Phase correction module, first to anti-phase echoWith homophase echoTentatively carry out vortex field and Maxwell's field phase school Just, its implementation is as described below:

A. selectK_y=0 line is along after the one-dimensional discrete inverse Fourier transform of frequency coding direction and the amplitude that calculates is more than The phase place of the complex data point of the 10% of maximum.

B. gained phase value in step A is carried out linear fit, obtains slope Δ φ respectively₀With intercept θ₀。

C. selectK_y=0 line along the amplitude that calculates after the one-dimensional discrete inverse Fourier transform of frequency coding direction more than The phase place of the complex data point of the 10% of big value.

D. gained phase value in step C is carried out automatically linear fit, obtains slope Δ φ respectively₁With intercept θ₁。

E. to allIt is multiplied by after carrying out one-dimensional discrete inverse Fourier transform along frequency coding directionAgain Carry out one-dimensional discrete Fourier transform, update allWherein n is to number at frequency coding direction number strong point.

F. to allIt is multiplied by after carrying out one-dimensional discrete inverse Fourier transform along frequency coding direction Carry out one-dimensional discrete Fourier transform again and carry out time reversal and calculate conjugate complex number, updating all

Then, from anti-phase (or homophase) echo(orChoosing two of which in) in frequency encoding gradient polarity is K-space centrage (the k gathered respectively in the case of positive and negative_y=0) one-dimensional discrete inverse Fourier transform is carried out along frequency coding direction Rear acquisition array P and P`, and calculateHereRepresent P_i,jConjugate complex number, i and j represents number respectively Strong point is along frequency coding and the numbering of phase-encoding direction.By anti-phase (or homophase) echo(or) k-space filled Array carries out one-dimensional discrete inverse Fourier transform along frequency coding direction and is multiplied byOne is carried out again along phase-encoding direction Dimension discrete inverse Fourier transform, obtains the anti-phase image S after phasing₀(or with phase images S₁), finally to homophase (or anti-phase) Echo(or) the k-space array filled carries out same phase images S that two-dimensional discrete inverse Fourier transform obtains updating₁ (or anti-phase image S₀)。

The purpose of above-mentioned phasing is that the probability of happening reducing phase place winding is the most effective to ensure phase unwrapping algorithm And reduce operation time, because the phase unwrapping algorithm that MRI field is commonly used is (as polynomial fitting method, region growth method and branch are cut Method etc.) efficiency on the low side and when noise and sampled point deficiency be insufficient effectively.Then, data processing module starts Phase unwrapping and correction module, its theoretical basis and implementation are as described below:

Compare during previous anti-phase echo exists an extra Δ τ due to each homophase echo and amplitude fading occurs A and phase error phi, anti-phase figure S₀With same phasor S₁Can be described as:

$S_0=(S_w-S_f)\·e^{i\stackrel{\‾}{\mathrm{\φ}}}---\left(11\right)$

$S_1=(S_w+S_f)\·A\·e^{i(\stackrel{\‾}{\mathrm{\φ}}+\mathrm{\φ})}---\left(12\right)$

Here S_wAnd S_fRepresent water and fat components in image pixel, they and M respectively_wAnd M_fIt is directly proportional,Including anti-phase The initial phase of figure and the phase place that field inhomogeneities, vortex field and Maxwell field produce during TE and (2N-2) Δ τ Error,Main by first anti-phase echo (i.e. k_y=0 line) phase error during TE determines.

By the phase place in formula (11)Algorithm flow as shown in Figure 5 Carry out the most accurate phase unwrapping, then to anti-phase image S₀EliminateAs follows:

$S_0^{`}=S_0\·e^{-i\·\stackrel{\‾}{\mathrm{\φ}}}---\left(13\right)$

When with κ=Re (S₀`)/|S<sub>0</sub>` | determine (S_w-S_f) symbol time, formula (13) also can be written as

S₀`=κ | S₀|=S_w-S_f (14)

And to same phase images S₁Elimination phase place is as follows:

S₁`=| S₁|=(S_w+S_f)·A (15)

Then, water fat separation module calculates based on formula (14) and formula (15) and produces water as S_wWith fat as S_fAs follows:

$S_w=\frac{\mathrm{\&kappa;}\&CenterDot;\left|S_0^{`}\right|+\left|S_1^{`}\right|/A}{2}---\left(16\right)$

$S_w=\frac{\left|S_1^{`}\right|/A-\mathrm{\&kappa;}\&CenterDot;\left|S_0^{`}\right|}{2}---\left(17\right)$

So, we can be eliminated four sub-pictures of Magnetic field inhomogeneity effect, i.e. with phasor S₀`, anti-phase figure S<sub>1</sub>`, water As S_wWith fat as S_f。

In formula (16) and (17), the spatial distribution of A is the most negligible, but local magnetic susceptibility effect is relatively on high field intensity equipment In the case of great, the spatial distribution of A can not be ignored, and water fat otherwise may be caused to separate not exclusively, to this end, pulse train module Thering is provided the sequence shown in Fig. 8 and Fig. 9 standby, the Dixon echo of start-up portion three phase places of design of this sequence is for measuring A's Distribution, wherein the 3rd echo is only filled with Dim2/N bar, k-space central area k-space line, and data processing module is still by aforementioned side Formula carries out phasing and discrete inverse Fourier transform, and result is as follows:

$S_2^c=(S_w-S_f)\&CenterDot;A^2\&CenterDot;e^{i(\stackrel{\&OverBar;}{\mathrm{\&phi;}}+2\mathrm{\&phi;})}---\left(18\right).$

Meanwhile, after aforementioned phasingWithMiddle extraction k-space central area Dim2/L bar k-space line, Obtain after discrete inverse Fourier transform:

$S_0^c=(S_w-S_f)\&CenterDot;e^{-i\stackrel{\&OverBar;}{\mathrm{\&phi;}}}---\left(19\right)$

Low resolution A scattergram is obtained as follows from formula (18) and formula (19):

$A=\sqrt{\left|S_2^c\right|/\left|S_0^c\right|}---\left(20\right)$

Then it is Dim1 × Dim2 matrix by A linear interpolation, is so achieved with water fat based on formula (16) and formula (17) and fills The image separated.It is filled with twice owing to k-space center that anti-phase figure is corresponding is actual, therefore signal noise ratio (snr) of image obtains further Improve.

Above-mentioned chemical shift imaging technology is greatly improved the scan efficiency of chemical shift imaging and overcomes MRI machine The adverse effect of hardware deficiency.

Two, embodiment

The control station main frame of 1.5T medical magnetic resonance imager is installed at water fat discrete pulse block and data Reason module, the submodule such as separates including primary phase correction, image reconstruction, phase unwrapping and correction, amplitude correction and water fat, presses Automatically perform according to the scanning process shown in Fig. 4.

First, on industrial computer sequencer, the apparent T2 shown in Fig. 1 is loaded by control station Cycle tests, first is chosen to, as thick-layer, arrange echo time TE and take different time number between 1ms and 1s on the picture of human body location Value also meets TE=n/ Δ f condition (n is natural number), then the echo amplitude that different TE are gathered based onCarry out Nonlinear fitting obtainsThen, according to the vortex field characteristic of gradient system, gradient waveform is carried out preemphasis.Then, water fat Discrete pulse sequence and data processing module produce water fat separate picture after performing successively.

Embodiment 1: enhanced ct scans, two-dimensional imaging mode.

In the case of enhanced ct scans, with reference to the representative value of conventional SPGR sweep parameter clinically, the imaging shown in Fig. 2 is set Sequential parameter, such as Dim1=192, Dim2=161, TR=160ms, L=4, X%=60%, TE=Δ τ=4.4ms.Running should Two-dimensional imaging sequence as shown in Figure 6 mode gather data, and the k-space filling mode of the most anti-phase echo with the situation of L=4 is Example is described as follows:

A1.k space is divided into four storage areas (C, S, P and V), stores central area data and the week of k-space respectively The each area data in limit, as shown in Figure 6.

A2. first anti-phase echo fills the central area that k-space C represents.

A3. second anti-phase echo fills the neighboring area that k-space S represents.

A4. the 3rd anti-phase echo fills the neighboring area that k-space P represents.

A5. the 4th anti-phase echo fills the neighboring area that k-space V represents.

Correspondingly, the PE system of anti-phase echo as shown in Figure 6, is described as follows:

B1. set phase code sum as Dim2=2m+1, set k-space filling proportion as X%, then for upper half k simultaneously Space, coded number shared by each region of its periphery is m/4, and coded number shared by central area is m/4+1.Each region phase code ladder Degree excursion is respectively labeled as by order from top to bottom: P_M—P_m-m/4+1；P_m-m/4—P_m-m/2+1；P_m-m/2—P_m-3m/4+1； P_m-3m/4—P₁.Phase encoding gradient amplitude is from P_mTo P₁Successively decreasing successively, stepping is-G_p/m/2.Then, for first anti-phase time Ripple, during the circulation of each phase code, phase code is gradually set to P_m-3m/4To P₁；For second anti-phase echo, each phase place During coding cycle, phase code is gradually set to P_m-m/2To P_m-3m/4+1；For the 3rd anti-phase echo, each phase code circulates Time phase code be gradually set to P_m-m/4To P_m-m/2+1；For the 4th anti-phase echo, during the circulation of each phase code, phase place is compiled Code is gradually set to P_mTo P_m-m/4+1。

B2. for lower half k-space, being only filled with central area C, remaining region does not refills；Phase code is followed every time During ring, phase code is gradually set to P₂To P_mm, mm=Round [(2m+1) X%]；Under needs provide state of signal-to-noise, in The repeatable collection of C of heart region once and is again filled with, and carries out cumulative and average with the front data once gathered.

B3. phase code circulation execution order: in order to alleviate the impact of systematic jitters, the phase code in each region from The most gradually perform near k-space center, and be alternately performed in upper half k-space and lower half k-space, every of collection K-space line sequentially stores, it is thus achieved that two sizes are the matrix of Dim1 × mm, respectively constitutes real part and the void of plural array K Portion.

B4. when phase code go to gradient amplitude be zero time, i.e. k_yThe situation of=0, arranges reading gradient in imaging sequence Carry out polarity inversion, and again gather echo data, carry out complex conjugate after one-dimensional discrete inverse Fourier transform and store separately be One one-dimensional plural number array P.

Equally, for homophase echo, we still can use aforesaid way to carry out phase code and k-space and fill, and only the Four steps save, and vice versa.

Then, the k-space line (including real part and imaginary part) sequentially stored is arranged again by above-mentioned PE system Arrange and build plural array K` that size is Dim1 × Dim2, the most tentatively carry out vortex field and Maxwell field Phasing, and carry out phasing the most in the following manner:

C1. array K` is carried out one-dimensional discrete inverse Fourier transform along frequency coding direction, extract corresponding k<sub>y</sub>The row arrow of=0 Amount, it is thus achieved that one-dimensional plural number array P`.

C2. the phase error of array P and each element of P` is set as Δ φ_ij, can obtain

C3. will along frequency coding directionIt is multiplied by each row vector of array K`, and carries out one-dimensional discrete Fourier transform Rear renewal array K.

Finally, image reconstruction can zeroize in usual manner and carry out inverse Fourier transform again, uses following manner to enter here OK:

D1. from K` array, choose the k-space line corresponding to k-space central area C, and carry out inverse Fourier transform and obtain Array A.

D2. K` array is directly carried out inverse Fourier transform and obtains array B.

D3. to each element of array A and B according to C_i,j=exp [-i angle (A_i,j)]B_i,jObtain new array C.

D4. array C is carried out Fourier transform, and chooses corresponding upper half k-space according to k-space conjugation symmetry principle K-space line, seek its complex conjugate and fill array C remainder, it is thus achieved that new array D.

D5. array D is carried out inverse Fourier transform and complete image reconstruction.

D6. water fat separation is carried out according to formula (11)-(17).

Sweep time is calculated as follows:

Timing node according to enhanced ct scans selects m, L and X%, realizes image resolution ratio under conditions of ensureing signal to noise ratio Optimization.

The separate imaging of water and fat technology of the present invention can be implemented in the application in enhanced ct scans completely, either in low field In imaging system or High-Field imaging system.Further, in 1.5T imaging system, due to echo time TE=2.2ms, Wo Menke To arrange N=16, other parameter ibid, then T_scan=0.96 (second), this can realize the separate imaging of water and fat technology of the present invention Application in abdominal contrast enhancement scanning.

Embodiment 2: non-reinforcing scans, three-dimensional imaging mode.

In the case of non-reinforcing scans, with reference to the representative value of conventional 3D SPGR sweep parameter clinically, arrange shown in Fig. 3 Three-dimensional imaging sequential parameter, as Dim1=256, Dim2=160, Dim3=32, L=4, TR=60ms, TE=Δ τ= 4.4ms, within the time span of phase encoding gradient is limited in Δ τ/2.Run this three-dimensional imaging sequence and gather k-space number According to, the k-space filling mode of the most anti-phase echo is described as follows in case of M=4:

A1. by selecting layer direction phase code order extraction two dimension k-space, it is divided into four storage areas (C, S, P and V) point Yong Yu not store central area data and each region of periphery of k-space, as shown in Figure 7.

A2. first anti-phase echo fills the central area that k-space C represents.

A3. second anti-phase echo fills the neighboring area that k-space S represents.

A4. the 3rd anti-phase echo fills the neighboring area that k-space P represents.

A5. the 4th anti-phase echo fills the neighboring area that k-space V represents.

Correspondingly, the PE system of anti-phase echo is as it is shown in fig. 7, be described as follows:

B1. set phase code sum as Dim2=2m+1, then for upper half k-space, each region of its periphery is shared compiles Yardage is m/4, and coded number shared by central area is m/4+1.Each region phase encoding gradient excursion is by order from top to bottom It is respectively labeled as: P_m—P_m-m/4+1；P_m-m/4—P_m-m/2+1；P_m-m/2—P_m-3m/4+1；P_m-3m/4—P₁.Phase encoding gradient amplitude is from P_m To P₁Successively decreasing successively, stepping is-G_p/m/2.Then, for first anti-phase echo, the phase code during circulation of each phase code Gradually it is set to P_m-3m/4—P₁；For second anti-phase echo, during the circulation of each phase code, phase code is gradually set to P_m-m/2—P_m-3m/4+1；For the 3rd anti-phase echo, during the circulation of each phase code, phase code is gradually set to P_m-m/4— P_m-m/2+1；For the 4th anti-phase echo, during the circulation of each phase code, phase code is gradually set to P_m—P_m-m/4+1。

B2. for lower half k-space, each region phase encoding gradient excursion is by the labelling respectively of order from top to bottom For: P₂—P_3m/4；P_m-3m/4+1—P_m-m/2；P_m-m/2+1—P_m-m/4；P_m-m/4+1—P_m.Phase encoding gradient amplitude is from P₂To P_mSuccessively Being incremented by, stepping is+G_p/m/2.Then, for first anti-phase echo, during the circulation of each phase code, phase code is gradually arranged For P₂—P_m-3m/4；For second anti-phase echo, during the circulation of each phase code, phase code is gradually set to P_m-3m/4+1— P_m-m/2；For the 3rd anti-phase echo, during the circulation of each phase code, phase code is gradually set to P_m-m/2+1—P_m-m/4；For 4th anti-phase echo, during the circulation of each phase code, phase code is gradually set to P_m-m/4+1—P_m。

B3. phase code circulation execution order: in order to alleviate the impact of systematic jitters, the phase code in each region from The most gradually perform near k-space center, and be alternately performed in upper half k-space and lower half k-space, every of collection K-space line sequentially stores, it is thus achieved that size be the complex matrix K of Dim1 × Dim2, its real part and imaginary part size be Dim1 × Dim2。

B4. central area C repeats the circulation of phase code, and arranges reading gradient and carry out polarity inversion, again adopts Collection echo data, after one-dimensional discrete inverse Fourier transform, storage is array P separately.

B5. under three-dimensional imaging mode, for each phase loop (outer circulation), perform to select the phase code in layer direction to follow Ring (interior circulation), is incremented by the most successively and selects the amplitude of layer direction phase encoding gradient until cycle-index reaches Dim3, it is thus achieved that be big The little complex matrix K for Dim1 × Dim2 × Dim3, its real part and imaginary part size are Dim1 × Dim2 × Dim3.

Equally, for homophase echo, aforesaid way still can be used to carry out phase code for we and k-space is filled, but the Four steps save, and vice versa.

Finally, the k-space line (including real part and imaginary part) sequentially stored is arranged again by above-mentioned PE system Arrange and build new plural array K`, the most tentatively carry out vortex field and Maxwell field phasing, go forward side by side One step carries out phasing in the following manner:

C1. array K` is carried out one-dimensional discrete inverse Fourier transform along frequency coding direction, extract along frequency coding direction The first row vector of central area C carries out complex conjugate, it is thus achieved that one-dimensional plural number array P`.

C2. the phase error of array P and each element of P` is set as Δ φ_ij, can obtain And constitute one-dimension array X1.

C3. remaining each row vector of central area is repeated above-mentioned (1)-(2) data handling procedure, it is thus achieved that a series of arrays X1, X2, X3 ..., then these arrays are stored the most afterwards for array X.

C4. along frequency coding direction, array X is multiplied by each row vector of array K` and carries out one-dimensional discrete Fourier transform Rear renewal array K`.

Finally, array K` is carried out image reconstruction by two dimension or 3 d-dem inverse Fourier transform, and according to formula (11)- (17) water fat separation is carried out.

When L is set to other value, we can use similar fashion quickly to obtain water fat separate picture, imaging equally Time shortens to the 1/N of usual manner.Generally, in the case of L less (such as L=4), vortex field and field inhomogeneities will not be led Cause image and obvious geometric distortion occurs.In the imaging system of gradient system being equipped with anti-eddy current function, L may be configured as 8 very To 16, obvious geometric distortion is not had to occur.Sweep time is calculated as follows:

$T_{scan}=\&lsqb;2m+1+(\frac{2m}{L}+1)\&rsqb;\&CenterDot;TR\&CenterDot;\frac{1}{L}\&CenterDot;Dim3.$

Three, the innovative point of the present invention

In the present invention, use supper-fast chemical shift imaging sequence, signal analysis model, echo amplitude attenuation compensation method Solution quick with phase place twines and accurately alignment technique separates enhanced ct scans and thin layer for realizing water fat on magnetic resonance imaging system Separate imaging of water and fat.

2-dimensional gradient echo sequence basis utilize 2N positive-negative polarity frequency encoding gradient alternately formed by N group water The gtadient echo string that fat same phase echo and antiphase echo are alternately constituted.N number of homophase echo is carrying out Gradient Phase volume respectively Code, k-space is divided into N number of region correspondingly, and first echo fills central area, and next echo is sequentially filled next-door neighbour Region, and use part Fourier mode to reconstruct same phasor, thus improve scan efficiency N times.N number of anti-phase echo enters respectively Row Gradient Phase encodes, and k-space is divided into N number of region correspondingly, and first echo fills central area, next echo It is sequentially filled immediate area, and uses part Fourier mode to reconstruct anti-phase figure, thus improve scan efficiency N times.

3-dimensional gradient echo sequence basis utilize 2N positive-negative polarity frequency encoding gradient alternately formed by N number of The gtadient echo string that water fat same phase echo and N number of antiphase echo are alternately constituted.N number of homophase echo carries out Gradient Phase respectively Coding, k-space is divided into N number of region correspondingly, and first echo fills central area, and next echo is sequentially filled tightly Neighbouring region, then carries out 3 d-dem inverse Fourier transform and produces a series of thin layer homophase figures.N number of anti-phase echo carries out ladder respectively Degree phase code, k-space is divided into N number of region correspondingly, and first echo fills central area, and next echo is successively Fill immediate area, then carry out 3 d-dem inverse Fourier transform and produce a series of thin layer homophase figures；

At spatial frequency domain and image area gtadient echo string set up general theoretical model respectively:

$I_{2N-2}^t=\underset{-\mathrm{\&infin;}}{\overset{+\mathrm{\&infin;}}{\&Integral;}}\{\left(M_w\right(r)+M_f\left(r\right)e^{-i2\mathrm{\&pi;}\mathrm{\&Delta;}f\&CenterDot;\&lsqb;TE+(2N-2)\mathrm{\&Delta;}\mathrm{\&tau;}\&rsqb;})e^{-(2N-1)\mathrm{\&Delta;}\mathrm{\&tau;}/T_2^{*}}{e}^{-i\&lsqb;{\mathrm{\&phi;}}_0\left(r\right)+(2N-1)\mathrm{\&phi;}\left(r\right)\&rsqb;}\}{d}^{3}r---\left(1\right)$

$I_{2N-1}^t=\underset{-\mathrm{\&infin;}}{\overset{+\mathrm{\&infin;}}{\&Integral;}}\{\left(M_w\right(r)+M_f\left(r\right)e^{-i2\mathrm{\&pi;}\mathrm{\&Delta;}f\&CenterDot;\&lsqb;TE+(2N-2)\mathrm{\&Delta;}\mathrm{\&tau;}\&rsqb;})e^{-2N\&CenterDot;\mathrm{\&Delta;}\mathrm{\&tau;}/T_2^{*}}{e}^{-i\&lsqb;{\mathrm{\&phi;}}_0\left(r\right)+2N\&CenterDot;\mathrm{\&phi;}\left(r\right)\&rsqb;}\}{d}^{3}r---\left(2\right)$

$S_0=(S_w-S_f)\&CenterDot;e^{i\stackrel{\&OverBar;}{\mathrm{\&phi;}}}---\left(11\right)$

$S_1=(S_w+S_f)\&CenterDot;A\&CenterDot;e^{i(\stackrel{\&OverBar;}{\mathrm{\&phi;}}+\mathrm{\&phi;})}---\left(12\right)$

Respectively in order to describe magnetic field bump to spatial frequency domain and image area signal amplitude and the impact of phase place, and During the TE that radio frequency pulse excitation to first echo is formed by above-mentioned model, magnetic field bump is to phase of echo and amplitude Impact is corrected.

Echo amplitude apparent T2 based on imaging regionTest value compensates correction, survey used Examination sequence includes the unit such as slice selective gradient, frequency encoding gradient and echo acquirement, and water mould is placed in imaging region scope or Directly human body location as on be chosen to as thick-layer, echo time TE is set between 1ms and 1s, takes different time numerical value completely Foot TE=n/ Δ f condition (n is natural number), then carries out nonlinear fitting to the echo amplitude of different TE based on single exponential function Obtain

Selection k-space centrage is more than the complex data point of maximum 10% after your Fourier transformation one-dimensional to amplitude Phase place carries out primary phase correction by linear fit method；

From anti-phase (or homophase) echo I_2N-2(or I_2N-1Two of which is chosen in frequency encoding gradient polarity for just to forsake one's love in) K-space centrage (the k gathered respectively under condition_y=0) pass through after carrying out one-dimensional discrete inverse Fourier transform along frequency coding direction Complex operation obtains the phase error of centrage, laggard at one-dimensional discrete inverse Fourier transform to all k-space lines the most accordingly Line phase corrects, thus eliminates the effect of anisotropy of bipolarity frequency encoding gradient；

From anti-phase (or homophase) echo I_2N-2(or I_2N-1Two of which is chosen in frequency encoding gradient polarity for just to forsake one's love in) Gather central area k-space line under condition respectively, pass through after carrying out one-dimensional discrete inverse Fourier transform respectively along frequency coding direction Complex operation obtains the phase error of each line in central area and is averaging, more accordingly to all k-space lines at one-dimensional discrete against Fu Vertical leaf transformation laggard line phase correction, thus eliminate the effect of anisotropy of bipolarity frequency encoding gradient；

The k-space data that the homophase echo of twice collection is corresponding with anti-phase echo is carried out image reconstruction and water respectively Fat separates, and then water picture and fat are as being overlapped respectively, thus thoroughly eliminate irreducible phase errors.

The relevant phase error of magnetic field bump carries out phase by a kind of exact algorithm based on discontinuous path integral Position solution is eliminated by the plural number operation with phasor and anti-phase figure after twining again, and then water is as S_wWith fat as S_fCount according to the following formula Calculate and produce:

$S_w=\frac{\mathrm{\&kappa;}\&CenterDot;\left|S_0^{`}\right|+\left|S_1^{`}\right|/A}{2}$

$S_w=\frac{\left|S_1^{`}\right|/A-\mathrm{\&kappa;}\&CenterDot;\left|S_0^{`}\right|}{2}.$

Last it is noted that various embodiments above is only in order to illustrate technical scheme, it is not intended to limit；To the greatest extent The present invention has been described in detail by pipe with reference to foregoing embodiments, it will be understood by those within the art that: it is right Technical scheme described in foregoing embodiments is modified, or the most some or all of technical characteristic is carried out equivalent replaces Change, do not make the essence of appropriate technical solution depart from the scope of various embodiments of the present invention technical scheme.

Claims (6)

1. a supper-fast stagewise single-shot water fat separation method, it is characterised in that described supper-fast stagewise single swashs The fat separation method that floods comprises the following steps:

Step S100: according to prescan needs, when loading apparent transverse relaxation by control station on industrial computer sequencer BetweenCycle tests；

Step S200: load two-dimensional chemical displacement imaging sequence module on sequencer by control station；

2-dimensional gradient echo sequence basis utilize 2N positive-negative polarity frequency encoding gradient alternately formed by N group water fat same The gtadient echo string that phase place echo and antiphase echo are alternately constituted；N number of homophase echo carries out Gradient Phase coding the most respectively, with Correspondingly k-space be divided into N number of region, first echo fills central area, and next echo is sequentially filled immediate area, And use part Fourier mode to reconstruct same phasor；N number of anti-phase echo carries out Gradient Phase coding respectively, correspondingly K-space is divided into N number of region, and first echo fills central area, and next echo is sequentially filled immediate area, and uses portion Fourier mode is divided to reconstruct anti-phase figure；Or

3-dimensional gradient echo sequence basis utilize 2N positive-negative polarity frequency encoding gradient alternately formed by N number of water fat same The gtadient echo string that phase place echo and N number of antiphase echo are alternately constituted；N number of homophase echo carries out Gradient Phase coding respectively, K-space is divided into N number of region correspondingly, and first echo fills central area, and next echo is sequentially filled next-door neighbour district Territory, then carries out 3 d-dem inverse Fourier transform and produces a series of thin layer homophase figures；N number of anti-phase echo carries out gradient phase respectively Position coding, k-space is divided into N number of region correspondingly, and first echo fills central area, and next echo is sequentially filled Immediate area, then carries out 3 d-dem inverse Fourier transform and produces a series of thin layer homophase figures；

Step S300: utilize the general theoretical model set up gtadient echo string in spatial frequency domain and image area, corrects radio frequency During the TE that pulse excitation to first echo is formed, magnetic field bump is on phase of echo and the impact of amplitude:

$I_{2N-2}^t=\underset{-\mathrm{\&infin;}}{\overset{+\mathrm{\&infin;}}{\&Integral;}}\left\{\left(M_w\right(r)+M_f(r\left)e^{-i2\mathrm{\&pi;}\mathrm{\&Delta;}f\&CenterDot;\&lsqb;TE+(2N-2)\mathrm{\&Delta;}\mathrm{\&tau;}\&rsqb;}\right)e^{-(2N-1)\mathrm{\&Delta;}\mathrm{\&tau;}/T_2^{*}}{e}^{-i\&lsqb;{\mathrm{\&phi;}}_0\left(r\right)+\left(2N-1\right)\mathrm{\&phi;}\left(r\right)\&rsqb;}\right\}{d}^{3}r---\left(1\right)$

$I_{2N-1}^t=\underset{-\mathrm{\&infin;}}{\overset{+\mathrm{\&infin;}}{\&Integral;}}\left\{\left(M_w\right(r)+M_f(r\left)e^{-i2\mathrm{\&pi;}\mathrm{\&Delta;}f\&CenterDot;\&lsqb;TE+(2N-1)\mathrm{\&Delta;}\mathrm{\&tau;}\&rsqb;}\right)e^{-2N\&CenterDot;\mathrm{\&Delta;}\mathrm{\&tau;}/T_2^{*}}{e}^{-i\&lsqb;{\mathrm{\&phi;}}_0\left(r\right)+2N\&CenterDot;\mathrm{\&phi;}\left(r\right)\&rsqb;}\right\}{d}^{3}r---\left(2\right)$

$S_0=(S_w-S_f)\&CenterDot;e^{i\stackrel{\&OverBar;}{\mathrm{\&phi;}}}---\left(11\right)$

$S_1=(S_w+S_f)\&CenterDot;A\&CenterDot;e^{i(\stackrel{\&OverBar;}{\mathrm{\&phi;}}+\mathrm{\&phi;})}---\left(12\right)$

Wherein, t refers to current phase code circulation step, and N arrives 2L scope round numbers, φ 1₀R () is the initial of proton magnetization vector Phase place, φ (r) is the phase error mainly produced by Magnetic field inhomogeneity effect,Item is for being optimized benefit to echo amplitude Repay；

S_wAnd S_fRepresent water and fat components in image pixel, they and M respectively_wAnd M_fIt is directly proportional,Initial including anti-phase figure Phase place and the phase error that field inhomogeneities, vortex field and Maxwell field produce during TE and (2N-2) Δ τ,Main Will be by first anti-phase echo (i.e. k_y=0 line) phase error during TE determines；

M_w(r) and M_fR () is expressed as water and the proton magnetization vector of fat in region, under three dimensions encoding condition Can be described as:

$M_w\left(r\right)=M_w^0{e}^{-i\mathrm{\&gamma;}{\&Integral;}_0^t\&lsqb;x\&CenterDot;G_x\left(\mathrm{\&tau;}\right)+y\&CenterDot;G_y\left(\mathrm{\&tau;}\right)+z\&CenterDot;G_z\left(\mathrm{\&tau;}\right)\&rsqb;d\mathrm{\&tau;}}---\left(3\right)$

$M_f\left(r\right)=M_f^0{e}^{-i\mathrm{\&gamma;}{\&Integral;}_0^t\&lsqb;x\&CenterDot;G_x\left(\mathrm{\&tau;}\right)+y\&CenterDot;G_y\left(\mathrm{\&tau;}\right)+z\&CenterDot;G_z\left(\mathrm{\&tau;}\right)\&rsqb;d\mathrm{\&tau;}}---\left(4\right);$

Step S400: water is as S_wWith fat as S_fCalculate generation according to the following formula:

$S_w=\frac{\mathrm{\&kappa;}\&CenterDot;\left|S_0^{`}\right|+\left|S_1^{`}\right|/A}{2}$

$S_f=\frac{\left|S_1^{`}\right|/A-\mathrm{\&kappa;}\&CenterDot;\left|S_0^{`}\right|}{2}.$

One the most according to claim 1 supper-fast stagewise single-shot water fat separation method, it is characterised in that step In S100, echo amplitude apparent T2 based on imaging regionTest value compensates correction, used test sequence Row include the unit such as slice selective gradient, frequency encoding gradient and echo acquirement, and water mould is placed in imaging region scope or direct Human body location picture is chosen to, as thick-layer, arrange echo time TE and take different time numerical value between 1ms and 1s and meet TE =n/f condition, n is natural number, then based on single exponential function, the echo amplitude of different TE is carried out nonlinear fitting acquisition

One the most according to claim 1 supper-fast stagewise single-shot water fat separation method, it is characterised in that step In S200, selection k-space centrage is more than the complex data point of maximum 10% after one-dimensional inverse Fourier transform to amplitude Phase place carries out primary phase correction by linear fit method.

One the most according to claim 1 supper-fast stagewise single-shot water fat separation method, it is characterised in that step In S300, from anti-phase echo I_2N-2Or homophase echo I_2N-1In to choose two of which in frequency encoding gradient polarity be positive and negative situation The lower k-space centrage k gathered respectively_yAfter=0 carries out one-dimensional discrete inverse Fourier transform along frequency coding direction, by plural number Computing obtains the phase error of centrage, the most all k-space lines is carried out after one-dimensional discrete inverse Fourier transform phase Bit correction, thus eliminate the non-equilibrium effect of bipolarity frequency encoding gradient.

One the most according to claim 1 supper-fast stagewise single-shot water fat separation method, it is characterised in that step In 300, from anti-phase echo I_2N-2Or homophase echo I_2N-1In to choose two of which in frequency encoding gradient polarity be positive and negative situation Under, gather central area k-space line respectively, by multiple after carrying out one-dimensional discrete inverse Fourier transform respectively along frequency coding direction Number computings obtain the phase error of each line in central area and be averaging, more accordingly to all k-space lines at one-dimensional discrete against Fourier The laggard line phase of leaf transformation corrects, and eliminates the non-equilibrium effect of bipolarity frequency encoding gradient.

One the most according to claim 1 supper-fast stagewise single-shot water fat separation method, it is characterised in that step In S300, the k-space data that the homophase echo of twice collection is corresponding with anti-phase echo is carried out image reconstruction and water respectively Fat separates, and then water picture and fat are as being overlapped respectively, eliminate irreducible phase errors.