CN105701815A - Magnetic resonance perfusion imaging postprocessing method and system thereof - Google Patents

Abstract

The invention provides a magnetic resonance perfusion imaging postprocessing method and a system thereof. The method comprises the following steps of perfusing image filtering to an original brain, removing a brain edge and acquiring a filtered brain perfusion image; receiving an area selection instruction of a user, acquiring a selected area of a middle artery area in the filtered brain perfusion image and outputting a weighted artery input curve assessing an artery input function; according to an improved Gamma function, fitting the weighted artery input curve and a concentration time curve respectively so as to obtain the optimized artery input curve and the concentration time curve; according to an improved singular matrix decomposition method based on a non-parametric model, solving a relative quantification parameter of perfusion. Through carrying out weighted optimization on AIF and a solving matrix, sensitivity to a noise is reduced; a simplified and effective Gamma function is used to carry out fitting, a non-linear problem is converted into linear solution, and a post-treatment speed is accelerated.

Description

A kind of MR perfusion imaging post-processing approach and system

Technical field

The present invention relates to MR perfusion imaging technical field, particularly relate to a kind of MR perfusion imaging post-processing approach and system。

Background technology

Along with the development of science and technology, magnetic resonance imaging system application at home and abroad is more and more universal。Nuclear magnetic resonance (MagneticResonanceImaging, it is called for short MRI) it is utilize the hydrogen atom spreading all over whole body in human body to be subject to exciting of radio-frequency pulse in additional high-intensity magnetic field, produce nmr phenomena, through spatial encoding techniques, detect with detector and accept the NMR signal released with electromagnetic form, by the NMR signal of detection input computer, processing conversion through data, the form finally respectively organized by human body forms image。The diseases such as partes corporis humani position, the neoplastic lesion of each system, vascular lesions, infection, wound, congenital development deformity and retrograde affection can be carried out checking and diagnose by nuclear magnetic resonance from different angles, by regulate magnetic field can unrestricted choice desired profile, other imaging technique can be obtained and do not caned close or be difficult to the image close to position。Further, nuclear magnetic resonance has without advantages such as ionizing radiation injury, soft tissue contrast height, image resolution ratio height, imaging parameters and scan position selection are flexible, so being widely used in clinical diagnose。Modern medicine iconography has been not content with the research to pathological change form change in recent years, and even develops in cerebral function imaging direction to reflection tissue and organ physiology with pathological change。Wherein brain magnetic resonance Perfusion Imaging (Perfusion-weightedMagneticResonanceImaging, PWI) it is one of the method for cerebral function imaging, in order to reflect that in tissue, blood capillary distribution and blood perfusion state are thus disclosing the biological behaviour of the cerebral tumor；Become the important medical means observing tumor change。

Magnetosensitive contrast medium Perfusion Imaging is formation method conventional at present。When paramagnetic contrast's agent is by during by inspection tissue, it is possible to shorten the T2* relaxation time of tissue, be reflected in the signal on corresponding T2* Perfusion Imaging and reduce；After paramagnetic contrast's agent is passed through, signal intensity recovers gradually。In T2* relaxation rate of change and blood there is linear relationship in contrast medium concentrations within the specific limits, can calculate evaluation relative quantification parameter relative blood volume (rCBV) of Perfusion Imaging, relative blood flow speed (rCBF), mean transit time (MTT), initial time of advent (T0) in conjunction with haemodynamic model through post processing。Existing post-processing approach carries out solving semi-quantitative parameters mainly by the convolution relation of arterial input function (AIF) Yu concentration-time signal。

Wherein the defining method of AIF has two classes, the non-automatic selection of the first kind；Equations of The Second Kind is based on automatically selecting of statistical method and fuzzy C-means clustering method。Automatically select and decrease user operation, but its amount of calculation and accuracy still treat rigorous proof。In order to improve computational accuracy, most researcheres propose based on Gamma Function Fitting Cot curve, and the method is high to image quality requirements, and the region error of fitting relatively low for signal to noise ratio is bigger；And nonlinear fitting operation time is longer。The calculating of rCBV, rCBF is mainly based upon the nonparametric model of singular value decomposition (SVD) and analyzes method and the analysis method based on parameter model。Wherein need to select suitable mathematical model in advance based on the method for parameter model, if selecting improper to often result in bigger calculating error。Therefore there is bigger difficulty in actual applications。

Therefore, prior art need to improve and development。

Summary of the invention

In view of above-mentioned the deficiencies in the prior art part, it is an object of the invention to provide a kind of MR perfusion imaging post-processing approach and system, aim to solve the problem that the post-processing approach of contrast medium Perfusion Imaging in prior art carries out solving semi-quantitative parameters mainly by the convolution relation of arterial input function (AIF) Yu concentration-time signal, solving result is chosen by mathematical model to be affected, and image quality requirements is high, it is impossible to the problem fast and accurately obtaining semi-quantitative parameters。

In order to achieve the above object, this invention takes techniques below scheme:

A kind of MR perfusion imaging post-processing approach, wherein, said method comprising the steps of:

A, to original brain perfusion image filter, remove brain edge, obtain filtering hindbrain perfusion image；

B, receive the regional choice instruction of user, obtain the selection area in the medium-sized artery district of brain perfusion image after the filtering, the weighting tremulous pulse input curve of output assessment arterial input function；

C, according to improve Gamma function respectively weighting tremulous pulse input curve and Cot curve are fitted, obtain optimize tremulous pulse input curve and Cot curve；

D, solve the relative quantification parameter of perfusion according to the decomposition of singular matrix method based on nonparametric model improved。

Described MR perfusion imaging post-processing approach, wherein, described step B specifically includes:

B1, receive user regional choice instruction, obtain the selection area in the medium-sized artery district of brain perfusion image after the filtering；

B2, concentration-time signal to the selection area in medium-sized artery district are weighted smoothing, and obtain the arterial input function calculated for parameter, and export weighting tremulous pulse input curve according to arterial input function。

Described MR perfusion imaging post-processing approach, wherein, described step C specifically includes:

C1, according to the maximum of Cot curve, the rising stage, decline phase Improvement of estimation the initial value of Gamma Function Fitting；

C2, at [T₀,T_max+ n] parameters optimization of Gamma function that is improved in conjunction with the initial value of Gamma Function Fitting in curve ranges；Wherein T₀It is the initial time of advent, T_maxBeing time corresponding to peak-peak, n participates in counting of matching after peak value。

C3, according to the Gamma function of this improvement to weighting tremulous pulse input curve and Cot curve matching, obtain the tremulous pulse input curve and the Cot curve that optimize。

Described MR perfusion imaging post-processing approach, wherein, described step D specifically includes:

D1, according to improve based on the matrix A in the decomposition of singular matrix method of nonparametric model be weighted combination, and matrix A is carried out SVD decompose obtain diagonal matrix W, orthogonal matrix V, triangular matrix U；

D2, processed according to threshold strategies after diagonal matrix S, and determine the relative quantification parameter of perfusion according to diagonal matrix S；Wherein S=1/W。

A kind of MR perfusion imaging after-treatment system, wherein, including:

Filtration module, for original brain perfusion image is filtered, removes brain edge, obtains filtering hindbrain perfusion image；

Region is selected and output module, for receiving the regional choice instruction of user, obtains the selection area in the medium-sized artery district of brain perfusion image after the filtering, the weighting tremulous pulse input curve of output assessment arterial input function；

Curve fitting module, for respectively weighting tremulous pulse input curve and Cot curve being fitted according to the Gamma function improved, obtains the tremulous pulse input curve and the Cot curve that optimize；

Parameter acquisition module, for the relative quantification parameter irrigated according to the decomposition of singular matrix system solution based on nonparametric model improved。

Described MR perfusion imaging after-treatment system, wherein, described region is selected and output module specifically includes:

Unit is selected in region, for receiving the regional choice instruction of user, obtains the selection area in the medium-sized artery district of brain perfusion image after the filtering；

Weighting output unit, for being weighted smoothing to the concentration-time signal of the selection area in medium-sized artery district, obtains the arterial input function calculated for parameter；Wherein, described arterial input function is weighting tremulous pulse input curve。

Described MR perfusion imaging after-treatment system, wherein, described curve fitting module specifically includes:

Evaluation unit, for according to the maximum of Cot curve, the rising stage, decline phase Improvement of estimation the initial value of Gamma Function Fitting；

Gamma function parameter acquiring unit, at [T₀,T_max+ n] in scope, the initial value in conjunction with evaluation unit output obtains the Gamma function parameter for curve matching；Wherein T₀It is the initial time of advent, T_maxBeing time corresponding to peak-peak, n participates in counting of matching after peak value。

Curve fitting unit, for the Gamma function according to this improvement to weighting tremulous pulse input curve and Cot curve matching, obtains the tremulous pulse input curve and the Cot curve that optimize。

Described MR perfusion imaging after-treatment system, wherein, described parameter acquisition module specifically includes:

Resolving cell, for according to improve based on the matrix A in the decomposition of singular matrix method of nonparametric model be weighted combination, and matrix A is carried out SVD decompose obtain diagonal matrix W, orthogonal matrix V, triangular matrix U；

Relative quantification parameter acquiring unit, for the diagonal matrix S after being processed according to threshold strategies, and determines the relative quantification parameter of perfusion according to diagonal matrix S；Wherein S=1/W。

MR perfusion imaging post-processing approach of the present invention and system, method includes: original brain perfusion image is filtered, and removes brain edge, obtains filtering hindbrain perfusion image；Receive the regional choice instruction of user, obtain the selection area in the medium-sized artery district of brain perfusion image after the filtering, the weighting tremulous pulse input curve of output assessment arterial input function；Weighting tremulous pulse input curve and Cot curve are fitted by Gamma function according to improving respectively, and the tremulous pulse input curve and the Cot curve that obtain optimization solve the relative quantification parameter of perfusion according to the decomposition of singular matrix method based on nonparametric model improved。The present invention, by AIF and solution matrix are weighted optimization, reduces the sensitivity to noise, adopts simplification and effective Gamma function to be fitted, nonlinear problem is converted into linear solution, accelerates post processing speed。

Accompanying drawing explanation

Fig. 1 is the flow chart of MR perfusion imaging post-processing approach preferred embodiment of the present invention。

Fig. 2 is the result schematic diagram assessing AIF in MR perfusion imaging post-processing approach of the present invention。

Fig. 3 is the parameter mapping graph that MR perfusion imaging post-processing approach of the present invention is embodied as。

Fig. 4 is the functional block diagram of MR perfusion imaging after-treatment system preferred embodiment of the present invention。

Detailed description of the invention

The present invention provides a kind of MR perfusion imaging post-processing approach and system, and for making the purpose of the present invention, technical scheme and effect clearly, clearly, developing simultaneously referring to accompanying drawing, the present invention is described in more detail for embodiment。Should be appreciated that specific embodiment described herein is only in order to explain the present invention, is not intended to limit the present invention。

Referring to Fig. 1, it is the flow chart of MR perfusion imaging post-processing approach preferred embodiment of the present invention。As it is shown in figure 1, described MR perfusion imaging post-processing approach includes:

Step S100, to original brain perfusion image filter, remove brain edge, obtain filtering hindbrain perfusion image。

In embodiments of the invention, by the Filtering Processing in step S100, remove background noise, and remove skull in conjunction with threshold method, reduce the calculating of regions of non-interest, it is possible to accelerate the computing of post processing。

Step S200, receive the regional choice instruction of user, obtain the selection area in the medium-sized artery district of brain perfusion image after the filtering, the weighting tremulous pulse input curve of output assessment arterial input function；

Step S300, according to improve Gamma function respectively weighting tremulous pulse input curve and Cot curve are fitted, obtain optimize tremulous pulse input curve and Cot curve；

Step S400, according to improve the decomposition of singular matrix method based on nonparametric model solve obtain perfusion relative quantification parameter。

Concrete, the relative quantification parameter of perfusion includes relative blood volume (rCBV), relative blood flow speed (rCBF), mean transit time (MTT) and arrives time to peak (TTP)

Further, described step S200 specifically includes:

Step S201, receive user regional choice instruction, obtain the selection area in the medium-sized artery district of brain perfusion image after the filtering。

In step s 201, choosing medium-sized artery region by man-machine interactively mode on perfusion image, choosing of this region can be rectangle, ellipse, irregular, but chosen area is not easily excessive。This embodiment is chosen the rectangular area of 3*3。

Step S202, concentration-time signal to the selection area in medium-sized artery district are weighted smoothing, and obtain the arterial input function calculated for parameter, i.e. output weighting tremulous pulse input curve。

Namely the curve that the every bit of the selection area in medium-sized artery district is corresponding is carried out smothing filtering, reduces influence of noise, be then weighted the signal of each point again obtaining weighting tremulous pulse input curve。

For reducing the backflow response of contrast medium, the present invention adopts the Gamma function of improvement the arterial input function in step S202 is fitted tremulous pulse input curve Ca (t) obtaining calculating eventually for parameter, refer to Fig. 2。

Gamma Function Fitting has great importance in Perfusion Imaging, and its expression formula is as follows:

G (t)=A* (t-D)^B*e^(-(t-D)/C)

\*MERGEFORMAT(1)

Wherein A, D, B, C are the parameters optimization needing estimation；T is time variable。But existing approximating method many employings non-linear method, calculates the time longer。The present invention proposes the Gamma approximating method of a kind of improvement, adopts linear solution, accelerate post processing speed and calculating effect meets clinical needs。

Further, described step S300 specifically includes:

Step S301, according to the maximum of Cot curve, the rising stage, decline phase Improvement of estimation the initial value of Gamma Function Fitting；

Initial value [the A of Gamma function in the invention process example₀,B₀,C₀,D₀] it is estimated as follows:

A₀Take peak-peak and the max (S) of Cot curve Ct；

D₀Take first corner position of Cot curve Ct and contrast medium initially enters time of tissue；

C₀Take signal curve Ct in the time corresponding with D0 in moment that declines；

B₀=(Tmax-T0)/C0；Tmax is the time that A0 is corresponding。

Step S302, at [T₀,T_max+ n] weighting tremulous pulse input curve is carried out linear fit by initial value according to the Gamma Function Fitting improved in scope, obtains the tremulous pulse input curve optimized；Wherein T₀It is the initial time of advent, T_maxBeing time corresponding to peak-peak, n participates in counting of matching after peak value。

In step s 302, the initial value [A of Gamma function is inputted₀,B₀,C₀,D₀] it is optimized unknown parameter。Crossing the principle of component to accelerate contrast medium head in optimal speed and combination perfusion, the present invention chooses [T₀,T_max+ n] data in scope carry out linear least square and are fitted。Wherein n participates in counting of matching after most peak value。In specific embodiment of the invention, n chooses the 3rd time point after peak-peak。

Cot curve C (t) and tremulous pulse input curve C in the Perfusion Imaging of magnetic susceptibility contrast medium_aThere is convolution relation in (t):

$C\left(t\right)=C_a\left(t\right)\⊗h\left(t\right)={\∫}_0^t{C}_a\left(\mathrm{\τ}\right)h\left(\mathrm{\τ}\right)d\mathrm{\τ}---\left(2\right)$

Introduce contrast medium residual concentration function R (t) in the tissue and blood flow rate F_t；According to formula (2) then:

$C\left(t_i\right)=F_t{\∫}_0^{t_i}{C}_a\left(\mathrm{\τ}\right)R(t-\mathrm{\τ})d\mathrm{\τ}\≈\mathrm{\Δ}t\underset{i=0}{\overset{j}{\Σ}}{C}_a\left(t_j\right)R(t_j-t_i)---\left(3\right)$

Δ t is the dynamic interval of Perfusion Imaging；Formula (3) is expressed as follows with matrix:

$\mathrm{\Δ}t\left(\begin{array}{cccc}{C}_a\left(t_1\right)& 0& ...& 0\\ C_a\left(t_2\right)& C_a\left(t_1\right)& ...& 0\\ ...& ...& ...& ...\\ C_a\left(t_n\right)& C_a\left(t_{n-1}\right)& ...& C_a\left(t_1\right)\end{array}\right)\·\left(\begin{array}{c}R\left(t_1\right)\\ R\left(t_2\right)\\ \mathrm{...}\\ R\left(t_n\right)\end{array}\right)=\left(\begin{array}{c}C\left(t_1\right)\\ C\left(t_2\right)\\ \mathrm{...}\\ C\left(t_n\right)\end{array}\right)---\left(4\right)$

Order $A=\left(\begin{array}{cccc}{C}_a\left(t_1\right)& 0& ...& 0\\ C_a\left(t_2\right)& C_a\left(t_1\right)& ...& 0\\ ...& ...& ...& ...\\ C_a\left(t_n\right)& C_a\left(t_{n-1}\right)& ...& C_a\left(t_1\right)\end{array}\right),b=\left(\begin{array}{c}R\left(t_1\right)\\ R\left(t_2\right)\\ \mathrm{...}\\ R\left(t_n\right)\end{array}\right),C=\left(\begin{array}{c}C\left(t_1\right)\\ C\left(t_2\right)\\ \mathrm{...}\\ C\left(t_n\right)\end{array}\right)$ Then its vector representation is:

A b=C (5)

Matrix A is carried out decomposition of singular matrix (SVD), is expressed as follows:

A^-1=V*W*U^T(6)

Wherein W is diagonal matrix, and V is orthogonal matrix, and U is triangular matrix, then

B=V*W* (U^T*C)(7)

Show according to many literature research, can by the F in the tissue of the maximum agent as a comparison in vector b_t。But in above-mentioned result, noise is comparatively sensitive, this process is improved by the factor present invention in step S400。

Further, described step S400 specifically includes:

Step S401, according to improve based on the matrix A in the decomposition of singular matrix method of nonparametric model be weighted combination, and matrix A is carried out SVD decompose obtain diagonal matrix W, orthogonal matrix V, triangular matrix U。

Matrix A is weighted combination, and first the linear optimization thought of integrating step S300 utilizes the rising stage of AIF curve, decline phase to choose the concentration signal new matrix of composition of special time scope $M=\left(\begin{array}{cccc}{C}_a\left(t_l\right)& 0& ...& 0\\ C_a\left(t_2\right)& C_a\left(t_l\right)& ...& 0\\ ...& ...& ...& ...\\ C_a\left(t_m\right)& C_a\left(t_{m-1}\right)& ...& C_a\left(t_l\right)\end{array}\right);$ Wherein D0≤l≤Total (t)；D0≤m≤n；Total (t) represents the overall time that contrast medium passes through。

Note a_ijIt is the element of matrix M after improving, as satisfied 0≤j≤i:

a_ij=Δ t (C_a(t_i-j)+4*Ca(t_i-j+1)+Ca(t_i-j+2))/6(8)

A in other situations_ij=0；Then formula (5) is changed to

M b=C (9)

Formula (9) is carried out SVD decomposition and obtains diagonal matrix W, orthogonal matrix V, triangular matrix U。

Step S402, processed according to threshold strategies after diagonal matrix S, and determine the relative quantification parameter of perfusion according to diagonal matrix S；Wherein S=1/W。

Diagonal matrix S after utilizing threshold strategies to be processed；And S=1/W；In the invention process example, threshold value chooses 0.2*max (W)；This threshold value is according to magnetic resonance acquisition system experience gained in the present invention。Then formula (7) can be written as:

B=V*S* (U^T*C)(10)

The pertinent literature so calculated according to perfusion can obtain rCBF=max (b)。Thus can obtain parameters variable successively according to the relation of perfusion relative quantification parameter, Fig. 3 is to one of result that clinical patient Perfusion Imaging data process in the invention process example。

Visible, the present invention, by AIF and solution matrix are weighted optimization, reduces the sensitivity to noise, adopts simplification and effective Gamma function to be fitted, nonlinear problem is converted into linear solution, accelerates post processing speed。

Based on said method embodiment, present invention also offers a kind of MR perfusion imaging after-treatment system。As shown in Figure 4, described MR perfusion imaging after-treatment system includes:

Filtration module 100, for original brain perfusion image is filtered, removes brain edge, obtains filtering hindbrain perfusion image；

Region is selected and output module 200, for receiving the regional choice instruction of user, obtains the selection area in the medium-sized artery district of brain perfusion image after the filtering, the weighting tremulous pulse input curve of output assessment arterial input function；

Curve fitting module 300, for respectively weighting tremulous pulse input curve and Cot curve being fitted according to the Gamma function improved, obtains the tremulous pulse input curve and the Cot curve that optimize；

Parameter acquisition module 400, for the relative quantification parameter irrigated according to the decomposition of singular matrix system solution based on nonparametric model improved。

Further, in described MR perfusion imaging after-treatment system, described region is selected and output module 200 specifically includes:

Unit is selected in region, for receiving the regional choice instruction of user, obtains the selection area in the medium-sized artery district of brain perfusion image after the filtering；

Weighting output unit, for being weighted smoothing to the concentration-time signal of the selection area in medium-sized artery district, obtains the arterial input function calculated for parameter, i.e. output weighting tremulous pulse input curve。

Further, in described MR perfusion imaging after-treatment system, described curve fitting module 300 specifically includes:

Evaluation unit, for according to the maximum of Cot curve, the rising stage, decline phase Improvement of estimation the initial value of Gamma Function Fitting；

Gamma function parameter acquiring unit, at [T₀,T_max+ n] in scope, the initial value in conjunction with evaluation unit output obtains the Gamma function parameter for curve matching。；Wherein T₀It is the initial time of advent, T_maxBeing time corresponding to peak-peak, n participates in counting of matching after peak value。

Curve fitting unit, for weighting tremulous pulse input curve and Cot curve matching, obtaining the tremulous pulse input curve and the Cot curve that optimize。

Further, in described MR perfusion imaging after-treatment system, described parameter acquisition module 400 specifically includes:

Resolving cell, for according to improve based on the matrix A in the decomposition of singular matrix method of nonparametric model be weighted combination, and matrix A is carried out SVD decompose obtain diagonal matrix W, orthogonal matrix V, triangular matrix U；

Relative quantification parameter acquiring unit, for the diagonal matrix S after being processed according to threshold strategies, and determines the relative quantification parameter of perfusion according to diagonal matrix S；Wherein S=1/W。

In sum, the invention provides a kind of MR perfusion imaging post-processing approach and system, method includes: original brain perfusion image is filtered, and removes brain edge, obtains filtering hindbrain perfusion image；Receive the regional choice instruction of user, obtain the selection area in the medium-sized artery district of brain perfusion image after the filtering, the weighting tremulous pulse input curve of output assessment arterial input function；Weighting tremulous pulse input curve and Cot curve are fitted by Gamma function according to improving respectively, and the tremulous pulse input curve and the Cot curve that obtain optimization solve the relative quantification parameter of perfusion according to the decomposition of singular matrix method based on nonparametric model improved。The present invention, by AIF and solution matrix are weighted optimization, reduces the sensitivity to noise, adopts simplification and effective Gamma function to be fitted, nonlinear problem is converted into linear solution, accelerates post processing speed。

It is understood that for those of ordinary skills, it is possible to it is equal to replacement according to technical scheme and present inventive concept or is changed, and all these are changed or replace the scope of the claims that all should belong to appended by the present invention。

Claims (8)

1. a MR perfusion imaging post-processing approach, it is characterised in that said method comprising the steps of:

A, to original brain perfusion image filter, remove brain edge, obtain filtering hindbrain perfusion image；

B, receive the regional choice instruction of user, obtain the selection area in the medium-sized artery district of brain perfusion image after the filtering, the weighting tremulous pulse input curve of output assessment arterial input function；

C, according to improve Gamma function respectively weighting tremulous pulse input curve and Cot curve are fitted, obtain optimize tremulous pulse input curve and Cot curve；

D, solve the relative quantification parameter of perfusion according to the decomposition of singular matrix method based on nonparametric model improved。

2. MR perfusion imaging post-processing approach according to claim 1, it is characterised in that described step B specifically includes:

B1, receive user regional choice instruction, obtain the selection area in the medium-sized artery district of brain perfusion image after the filtering；

B2, concentration-time signal to the selection area in medium-sized artery district are weighted smoothing, and obtain the arterial input function calculated for parameter, and export weighting tremulous pulse input curve according to arterial input function。

3. MR perfusion imaging post-processing approach according to claim 1, it is characterised in that described step C specifically includes:

C1, according to the maximum of Cot curve, the rising stage, decline phase Improvement of estimation the initial value of Gamma Function Fitting；

C2, at [T₀,T_max+ n] parameters optimization of Gamma function that is improved in conjunction with the initial value of Gamma Function Fitting in curve ranges；Wherein T₀It is the initial time of advent, T_maxBeing time corresponding to peak-peak, n participates in counting of matching after peak value；

C3, according to the Gamma function of this improvement to weighting tremulous pulse input curve and Cot curve matching, obtain the tremulous pulse input curve and the Cot curve that optimize。

4. MR perfusion imaging post-processing approach according to claim 1, it is characterised in that described step D specifically includes:

D1, according to improve based on the matrix A in the decomposition of singular matrix method of nonparametric model be weighted combination, and matrix A is carried out SVD decompose obtain diagonal matrix W, orthogonal matrix V, triangular matrix U；

D2, processed according to threshold strategies after diagonal matrix S, and determine the relative quantification parameter of perfusion according to diagonal matrix S；Wherein S=1/W。

5. a MR perfusion imaging after-treatment system, it is characterised in that including:

Filtration module, for original brain perfusion image is filtered, removes brain edge, obtains filtering hindbrain perfusion image；

Region is selected and output module, for receiving the regional choice instruction of user, obtains the selection area in the medium-sized artery district of brain perfusion image after the filtering, the weighting tremulous pulse input curve of output assessment arterial input function；

Curve fitting module, for respectively weighting tremulous pulse input curve and Cot curve being fitted according to the Gamma function improved, obtains the tremulous pulse input curve and the Cot curve that optimize；

Parameter acquisition module, for the relative quantification parameter irrigated according to the decomposition of singular matrix system solution based on nonparametric model improved。

6. MR perfusion imaging after-treatment system according to claim 5, it is characterised in that described region is selected and output module specifically includes:

Unit is selected in region, for receiving the regional choice instruction of user, obtains the selection area in the medium-sized artery district of brain perfusion image after the filtering；

Weighting output unit, for being weighted smoothing to the concentration-time signal of the selection area in medium-sized artery district, obtains the arterial input function calculated for parameter；Wherein, described arterial input function is weighting tremulous pulse input curve。

7. MR perfusion imaging after-treatment system according to claim 5, it is characterised in that described curve fitting module specifically includes:

Evaluation unit, for according to the maximum of Cot curve, the rising stage, decline phase Improvement of estimation the initial value of Gamma Function Fitting；

Gamma function parameter acquiring unit, at [T₀,T_max+ n] in scope, the initial value in conjunction with evaluation unit output obtains the Gamma function parameter for curve matching；Wherein T₀It is the initial time of advent, T_maxBeing time corresponding to peak-peak, n participates in counting of matching after peak value。

Curve fitting unit, for the Gamma function according to this improvement to weighting tremulous pulse input curve and Cot curve matching, obtains the tremulous pulse input curve and the Cot curve that optimize。

8. MR perfusion imaging after-treatment system according to claim 5, it is characterised in that described parameter acquisition module specifically includes:

Resolving cell, for according to improve based on the matrix A in the decomposition of singular matrix method of nonparametric model be weighted combination, and matrix A is carried out SVD decompose obtain diagonal matrix W, orthogonal matrix V, triangular matrix U；

Relative quantification parameter acquiring unit, for the diagonal matrix S after being processed according to threshold strategies, and determines the relative quantification parameter of perfusion according to diagonal matrix S；Wherein S=1/W。