CN104599244B - The denoising method and system of diffusion tensor imaging - Google Patents
The denoising method and system of diffusion tensor imaging Download PDFInfo
- Publication number
- CN104599244B CN104599244B CN201410816610.9A CN201410816610A CN104599244B CN 104599244 B CN104599244 B CN 104599244B CN 201410816610 A CN201410816610 A CN 201410816610A CN 104599244 B CN104599244 B CN 104599244B
- Authority
- CN
- China
- Prior art keywords
- diffusion
- denoising
- dispersion
- weighted
- tensor
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Landscapes
- Magnetic Resonance Imaging Apparatus (AREA)
Abstract
The present invention provides the denoising methods and system of a kind of diffusion tensor imaging, Gaussian distribution nature of its method based on Diffusion-weighted imaging model and sampling noise, using the openness of average diffusivity, using the method for maximum a-posteriori estimation directly as the dispersion tensor matrix after the denoising corresponding to K space data obtains each spatial position.The influence that the present invention can estimate dispersion tensor to avoid the error of image denoising, can more effectively inhibit the noise in dispersion tensor, improve the estimated accuracy of dispersion tensor.
Description
Technical field
The present invention relates to mr imaging technique, more particularly to a kind of denoising method of diffusion tensor imaging
And system.
Background technology
Diffusion tensor (Diffusion Tensor Imaging DTI) is in diffusion-weighted imaging (Diffusion
Weighted Imaging DWI) on the basis of the new imaging method that grows up, utilize the Diffusion anisotropy of hydrone
Be imaged, can it is lossless from microscopic fields evaluation institutional framework integrality, for the prevention of disease, diagnosis and treat provide more
More information.But compared to other mr imaging techniques, DTI needs longer sweep time, and signal-to-noise ratio is relatively low.
In order to improve the signal-to-noise ratio of diffusion tensor, current relatively straightforward method be averaged by multiple repairing weld and
Reduce the method in K spatial samplings region.These methods have certain application in practice, but can increase sweep time and influence empty
Between resolution ratio.Another common method is after K spacescan data are obtained, and reconstructs diffusion weighted images first, then
Denoising is carried out to image with the method for signal processing again, finally, by image after denoising calculate dispersion tensor and it is various more
Parameter is dissipated, to generate diffusion tensor figure.This method application is wider, but the systematic error during image denoising is possible to
It can be further transferred in subsequent tensor computation, and then influence the quality of various dispersion parameter images.
Invention content
Based on this, it is necessary to for the problems of the prior art, provide a kind of denoising side of diffusion tensor imaging
Method and system, the influence that can be estimated to avoid the error of image denoising dispersion tensor can more effectively inhibit disperse
Noise in amount improves the estimated accuracy of dispersion tensor.
The present invention provides a kind of denoising method of diffusion tensor imaging, including：
Image data acquisition step：Obtain the K space data corresponding to diffusion-weighted MR imaging image；
Denoising step：Gaussian distribution nature based on Diffusion-weighted imaging model and sampling noise, using averagely
Diffusivity it is openness, each spatial position institute is obtained by the K space data using the method for maximum a-posteriori estimation
Dispersion tensor matrix after corresponding denoising；
Dispersion parameter calculates step：Based on the dispersion tensor matrix after the denoising, dispersion parameter figure is obtained.
The denoising step includes in one of the embodiments,：
Gaussian distribution nature based on Diffusion-weighted imaging model and sampling noise, utilizes the dilute of average diffusivity
Property is dredged, denoising function model is built using the method for maximum a-posteriori estimation, the denoising function model is referring to such as following public affairs
Formula (1)：
Formula (1)
Wherein,Represent the estimated value of dispersion tensor matrix, MD is average diffusivity；R () is to act on average disperse
The sparse constraint function of rate, λ are corresponding regularization parameter；dmFor the K space data corresponding to m-th of diffusion weighted images；F
Represent Fourier-encoded matrix；Represent m-th of diffusion weighted images, wherein, I0Indicate the reference of no diffusion-weighted
Image,For the phase of m-th of diffusion weighted images, b is the diffusion-weighted factor, gmIt is corresponding to m-th of diffusion weighted images
Diffusion gradient vector gm=(gxm,gym,gzm)T；
Using the K space data, the formula (1) is solved, after obtaining the denoising corresponding to each spatial position
Dispersion tensor matrix.
The sparse constraint function constraint is averaged the openness of diffusivity in one of the embodiments,.
The sparse constraint function utilizes L in one of the embodiments,1Model letter constrains average diffusivity in sparse change
It changes openness in domain.
The sparse constraint function is obtained by calling the following formula to calculate in one of the embodiments,：
Wherein, Ψ represents sparse transformation operator, ‖ ‖1Expression takes L1Norm, D1、D2、D3Respectively dispersion tensor matrix
The element of leading diagonal.
Based on the above method, the present invention also provides a kind of denoising system of diffusion tensor imaging, including：
Image data acquisition module, for obtaining the K space data corresponding to diffusion-weighted MR imaging image；
Denoising module for the Gaussian distribution nature based on Diffusion-weighted imaging model and sampling noise, utilizes
Averagely diffusivity is openness, each space bit is obtained by the K space data using the method for maximum a-posteriori estimation
Put the dispersion tensor matrix after corresponding denoising；And
Dispersion parameter computing module, for based on the dispersion tensor matrix after the denoising, obtaining dispersion parameter figure.
The denoising module includes in one of the embodiments,：
Model construction unit, for based on Diffusion-weighted imaging model and sampling noise Gaussian distribution nature,
Using the openness of average diffusivity, denoising function model, the denoising letter are built using the method for maximum a-posteriori estimation
Exponential model is referring to such as following formula (1)：
Formula (1)
Wherein,Represent the estimated value of dispersion tensor matrix, MD is average diffusivity；R () is to act on average disperse
The sparse constraint function of rate, λ are corresponding regularization parameter；dmFor the K space data corresponding to m-th of diffusion weighted images；F
Represent Fourier-encoded matrix；Represent m-th of diffusion weighted images, wherein, I0Indicate the reference of no diffusion-weighted
Image,For the phase of m-th of diffusion weighted images, b is the diffusion-weighted factor, gmIt is corresponding to m-th of diffusion weighted images
Diffusion gradient vector gm=(gxm,gym,gzm)T；With
Matrix Solving unit for utilizing the K space data, solves the formula (1), obtains each spatial position
Dispersion tensor matrix after corresponding denoising.
The present invention proposed using the openness of average diffusivity, directly by the K space data of acquisition to dispersion tensor into
The step of method of row denoising imaging, this method has skipped over traditional image denoising, the error of image denoising is avoided to disperse
The influence of tensor estimation, can more effectively inhibit the noise in dispersion tensor, improve the estimated accuracy of dispersion tensor.
Description of the drawings
Fig. 1 is one embodiment flow diagram of the method for the present invention；
Fig. 2 is another embodiment flow diagram of the method for the present invention；
Fig. 3 is one embodiment structure diagram of present system；
Fig. 4 is another example structure schematic diagram of present system.
Specific embodiment
The present invention is based on Diffusion-weighted imaging technologies, utilize dispersion tensor model and the intrinsic of diffusivity MD that be averaged
Characteristic (such as openness), the theoretical frame based on maximum a-posteriori estimation, after directly obtaining denoising by the K space data acquired
Dispersion tensor, so as to fulfill the denoising of diffusion tensor imaging.Diffusion tensor imaging mentioned herein is that have
Not in a kind of new method of diffusion-weighted imaging, can be used for describing brain structure.For example, if Magnetic resonance imaging
The hydrogen atom tracked in hydrone, then diffusion tensor be according to the drawing of hydrone moving direction, dispersion tensor into
As figure (presentation mode is different from pervious image) can reveal that brain tumor how to influence nerve cell connect, guide healthcare givers into
Row operation on brain.It can also disclose the related subtle abnormality of same apoplexy, multiple sclerosis, schizophrenia, Dyslexia
Variation.
The physical mechanism of magnetic resonance imaging can be expressed as following formula (2-1)：
D=F ρ+n formula (2-1)
Wherein, the signal that d expressions are acquired on magnetic resonance device, F represent Bo Liye encoder matrixs, and ρ is the figure of magnetic resonance reconstruction
Picture, n are generally assumed to be white complex gaussian noise.
In Diffusion-weighted imaging, m-th of diffusion weighted images ρmImaging model can write following formula
(2-2)：
Formula (2-2) wherein, I0Indicate the reference picture of no diffusion-weighted,For m-th more
The phase of weighted image is dissipated, b is the diffusion-weighted factor (constant), gmIt is the diffusion gradient corresponding to m-th of diffusion weighted images
Vectorial gm=(gxm,gym,gzm)T。
For diffusion-weighted imaging, only there are one one scalar values of dispersion coefficient to describe disperse attribute, and in disperse
In amount imaging, the movement of hydrone on each direction can be fully described for dispersion tensor and hydrone moves in those directions
Correlation.And tensor is substantially exactly the direction vector figure of a secondary three dimensions, can have direction in brain image to show
Property white matter fiber tract in hydrone movement selectivity.Dispersion tensor is usually represented by following formula (2-3) matrix, simple below
Referred to as dispersion tensor matrix.
Above-mentioned dispersion tensor matrix D is symmetrical matrix, can further will more in order to visually state dispersion tensor matrix
It dissipates tensor matrix and is considered as an oval ball (ellipsoid), characteristic value is represented along the minimum and maximum axis of disperse ellipsoid
Dispersion coefficient.(do not change so three characteristic values of dispersion tensor matrix are most basic rotational invariants with disperse direction
Become).They are the main dispersion coefficients measured along three change in coordinate axis direction.These three reference axis are that tissue is intrinsic.Each originally
For value indicative in connection with an eigenvector, this eigenvector is also that tissue is intrinsic.Three eigenvectors of dispersion tensor matrix
It is mutually perpendicular to, and constructs the part of each pixel with reference to fiber frame.In each pixel, characteristic value arranges from big to small：
λ1=maximum dispersion coefficient, λ2=middle rank dispersion coefficient, λ3=minimum dispersion coefficient.λ1Represent the disperse for being parallel to machine direction
Coefficient, λ2And λ3Represent lateral dispersion coefficient.Characteristic value λ corresponding to each pixel1、λ2、λ3Hereafter referred to collectively as dispersion tensor is special
Value indicative.
Denoising method below in conjunction with the attached drawing diffusion tensor imaging that the present invention will be described in detail is provided and it is
The specific embodiment of system.
As shown in Figure 1, a kind of denoising method of diffusion tensor imaging provided in this embodiment, including following step
Suddenly：
Image data acquisition step 110：Obtain the K space data corresponding to diffusion-weighted MR imaging image；
Denoising step 120：Gaussian distribution nature based on Diffusion-weighted imaging model and sampling noise, using flat
Diffusivity is openness, each spatial position is obtained by the K space data using the method for maximum a-posteriori estimation
Dispersion tensor matrix after corresponding denoising；
Dispersion parameter calculates step 130：Based on the dispersion tensor matrix after denoising, dispersion parameter figure is obtained.
Based on above-described embodiment, in above-mentioned steps 110, the K space data obtained can pass through echo-planar imaging
(echo planar imaging, EPI) technology and parallel imaging technique (parallel imaging) obtain.Certainly the present invention
Step 110 is primarily to obtain K space data, without acquisition technique used in limitation.
Based on above-described embodiment, as shown in Fig. 2, above-mentioned denoising step 120 can include in specific implementation it is following two
Step.
Step 121, based on Diffusion-weighted imaging model (i.e. above-mentioned formula (2-1) and (2-2)) and sampling noise
Gaussian distribution nature, using the openness of average diffusivity MD, denoising letter is built using the method for maximum a-posteriori estimation
Exponential model, this denoising function model is referring to such as following formula (1)：
Formula (1)
Wherein,Represent the estimated value of dispersion tensor matrix, MD is average diffusivity；R () is to act on average disperse
The sparse constraint function of rate, λ are corresponding regularization parameter；dmFor the K space data corresponding to m-th of diffusion weighted images；F
Represent Fourier-encoded matrix；Represent m-th of diffusion weighted images, wherein, I0Indicate the ginseng of no diffusion-weighted
Examine image,For the phase of m-th of diffusion weighted images, b is the diffusion-weighted factor, gmIt is that m-th of diffusion weighted images institute is right
The diffusion gradient vector g answeredm=(gxm,gym,gzm)T；
Step 122, using K space data, the formula (1) is solved, obtains the denoising corresponding to each spatial position
Dispersion tensor matrix afterwards.
The MD mentioned in above-mentioned steps 121, i.e., averagely diffusivity (mean diffusivity, MD), reflection molecule are whole
Disperse horizontal (size of average ellipsoid) and disperse resistance overall condition.MD only represents the size of disperse, and with disperse
Direction is unrelated.MD is bigger, and contained free water molecule is then more in tissue.Shown in for example following formula (2-4) of calculation formula：
MD=(λ1+λ2+λ3The formula of)/3 (2-4)
Wherein, λ1, λ2, λ3Three dispersion tensor matrix exgenvalues for dispersion tensor matrix D.
The present invention is based on sampling noise Gaussian distributeds, with reference to above-mentioned formula (2-1) and (2-2), utilize average disperse
The inherent characteristic (such as openness) of rate, gives the denoising function model described in above-mentioned formula (1), to characterize dispersion tensor
Maximum a-posteriori estimation.
And above-mentioned formula (2-4) is based on,Wherein D1、D2、D3Respectively disperse
The element of tensor matrix leading diagonal, then above-mentioned formula (1) following formula (1-2) can be changed into.
Formula (1-2)
For formula (1-2), it is above-mentioned using maximum a-posteriori estimation method when, introduce average diffusivity it is sparse about
Beam.Such as：Above-mentioned sparse constraint function constraint is averaged the openness of diffusivity, preferably by L1Model letter constrains average diffusivity
It is openness in sparse transform-domain.
In one embodiment of the invention, the sparse constraint function is obtained by calling the following formula (3) to calculate.
The sparse constraint function R (MD) for acting on average diffusivity i.e. in above-mentioned formula (1) is expressed as above-mentioned formula (3)
Content, D1、D2、D3The respectively element of dispersion tensor matrix leading diagonal, operator Ψ represent certain sparse transformation, such as small echo
Transformation etc..‖·‖1Expression takes L1 norms.L1 norm ‖ ‖ are utilized in the present embodiment1For constraining average diffusivity in sparse transformation
It is openness in domain.So far, the Denoising Problems of dispersion tensor are changed into reference to formula (3), using numerical optimization to above-mentioned public affairs
The Solve problems of formula (1), and for the common method for solving of above-mentioned formula (1) have Nonlinear Conjugate Gradient Methods, simulated annealing,
Bregman algorithms, FPC (Fixed-Point Continuation) algorithm, L1-magic algorithms, L1-LS algorithms, newton decline
Method, genetic algorithm etc., do not elaborate herein.The present invention is not limited to only with L1 norms, can also using other model letters come
It realizes, for details, reference can be made to above-mentioned formula (3) and carry out similar setting.
Based on above-described embodiment, above-mentioned dispersion parameter calculates step 130 and includes the following steps：
According to the dispersion tensor matrix after denoising, the corresponding dispersion parameter of dispersion parameter figure is calculated；Disperse mentioned herein
Parameter is primarily referred to as 5 basic parameters mentioned below.
For diffusion tensor technology, it is also necessary to various dispersion parameters are calculated according to above-mentioned dispersion tensor matrix D, it is as follows
It states bright.
(1) MD, i.e., averagely diffusivity (mean diffusivity MD), the disperse of MD reflection molecule entirety are horizontal (average
The size of ellipsoid) and disperse resistance overall condition.MD only represents the size of disperse, and unrelated with the direction of disperse.MD is bigger,
Contained free water molecule is then more in tissue.Shown in for example following formula (2-4) of calculation formula：
MD=(λ1+λ2+λ3The formula of)/3 (2-4)
(2) FA, i.e. Fractional anisotropy index are the ratios that hydrone anisotropy ingredient accounts for entire dispersion tensor, it
Variation range from 0~1.0 represents that disperse is unrestricted, for example, cerebrospinal fluid FA values close to 0；There is side for fairly regular
The tissue of tropism, FA values are more than 0, such as cerebral white matter fiber FA values are close to 1.The calculation formula of FA values following formula as follows
(2-5)：
(3) RA, relative anisotropies index are the anisotropic segment of dispersion tensor and dispersion tensor isotropism part
Ratio, its variation range is from 0 (isotropism disperse) to √ 2 (infinite anisotropy).The calculation formula of RA is following public affairs
Formula (2-6)：
(4) VR, i.e. volume ratio index are the ratio of ellipsoid and sphere volume.Since its variation range is (i.e. each from 1
To same sex disperse) to 0, so, it is clinically more likely to using 1/VR.For example following formula (2-6) of the calculation formula of VR：
In above-mentioned formula, λ1,λ2,λ3For dispersion tensor matrix D three dispersion tensor matrix exgenvalues (it is aforementioned to have explained,
Refer to preceding description).
(5) dispersion tensor track (the trace of the diffusion tensor), is an invariant parameter,
Tr (D)=D1+D2+D3.Wherein, D1、D2、D3The elements in a main diagonal for matrix shown in above-mentioned formula (2-3).
Fig. 1 or Fig. 2 is the flow diagram of one embodiment of the invention.Although it should be understood that the stream of Fig. 1 or Fig. 2
Each step in journey figure shows successively according to the instruction of arrow, but these steps be not it is inevitable indicated according to arrow it is suitable
Sequence performs successively.Unless expressly state otherwise herein, there is no stringent sequences to limit for the execution of these steps, can be with
Other sequences perform.Moreover, at least part step in Fig. 1 or Fig. 2 can include multiple sub-steps or multiple stages,
These sub-steps or stage are not necessarily to perform completion in synchronization, but can be performed at different times, are held
Row sequence is also not necessarily and carries out successively, but can either the sub-step of other steps or stage are combined reality with other steps
It applies.Above realization method of each embodiment only just for corresponding steps in illustrating is expounded, and is then being patrolled
In the case of volume not contradicting, above-mentioned each embodiment be can be combined with each other and form new technical solution, and this is new
Technical solution still in the open scope of present embodiment.
Based on the above method, the present invention also provides a kind of denoising system of diffusion tensor imaging, such as Fig. 3 institutes
Show, which includes：
Image data acquisition module 210, for obtaining the K space data corresponding to diffusion-weighted MR imaging image；
Denoising module 220, for the Gaussian distribution nature based on Diffusion-weighted imaging model and sampling noise, profit
It is openness with average diffusivity, each space is obtained by the K space data using the method for maximum a-posteriori estimation
Dispersion tensor matrix after denoising corresponding to position；And
Dispersion parameter computing module 230, for based on the dispersion tensor matrix after denoising, obtaining dispersion parameter figure.
Above-mentioned image data acquisition module 210 is mainly used for the step 110 in method shown in Fig. 1, above-mentioned denoising module 220
The step 120 being mainly used in method shown in Fig. 1, above-mentioned dispersion parameter computing module 230 are mainly used in method shown in Fig. 1
Step 130.Therefore the present embodiment system in each function module specific implementation referring to above-mentioned related step 110 to step
Rapid 130 explanation.
Based on above system, in one embodiment of the invention, above-mentioned denoising module 220 includes：
Model construction unit 221, for the Gaussian Profile based on Diffusion-weighted imaging model and sampling noise
Using the openness of average diffusivity, denoising function model is built using the method for maximum a-posteriori estimation for matter；With
Matrix Solving unit 222 for utilizing the K space data, solves above-mentioned denoising function model, obtains each
Dispersion tensor matrix after denoising corresponding to spatial position.Here denoising function model is referring to above-mentioned formula (1).
Above-mentioned model construction unit 221 is mainly used for the step 121 in method shown in Fig. 2, above-mentioned Matrix Solving unit 222
The step 122 being mainly used in method shown in Fig. 2.Therefore the specific implementation of each function module in the present embodiment referring to
The above-mentioned explanation in relation to step 121 to step 122.
Through the above description of the embodiments, those skilled in the art can be understood that above-described embodiment side
Method can add the mode of required general hardware platform to realize by software, naturally it is also possible to by hardware, but in many cases
The former is more preferably embodiment.Based on such understanding, technical scheme of the present invention substantially in other words does the prior art
Going out the part of contribution can be embodied in the form of software product, which is stored in a non-volatile meter
In calculation machine readable storage medium storing program for executing (such as ROM, magnetic disc, CD), used including some instructions so that a station terminal equipment (can be hand
Machine, computer, server or network equipment etc.) perform system structure and method described in each embodiment of the present invention.
In conclusion the present invention directly passes through the K spaces of acquisition with the inherent characteristic (such as openness) of average diffusivity
Dispersion tensor matrix after data acquisition denoising compared with prior art, is not needed to by increasing sampling number or reducing K skies
Between sampling area method, it is possible to improve the signal-to-noise ratio of diffusion tensor, and be not necessary to add by the disperse to reconstruction
Weight graph picture carries out denoising to obtain dispersion tensor matrix and related dispersion parameter, so as to eliminate the error in image denoising
Influence to dispersion tensor estimation, effectively inhibits the noise in dispersion tensor, improves the precision of dispersion tensor estimation.
Embodiment described above only expresses the several embodiments of the present invention, and description is more specific and detailed, but simultaneously
Cannot the limitation to the scope of the claims of the present invention therefore be interpreted as.It should be pointed out that for those of ordinary skill in the art
For, without departing from the inventive concept of the premise, various modifications and improvements can be made, these belong to the guarantor of the present invention
Protect range.Therefore, the protection domain of patent of the present invention should be determined by the appended claims.
Claims (8)
1. a kind of denoising method of diffusion tensor imaging, including：
Image data acquisition step：Obtain the K space data corresponding to diffusion-weighted MR imaging image；
Denoising step：Gaussian distribution nature based on Diffusion-weighted imaging model and sampling noise, utilizes average disperse
Rate it is openness, using the method for maximum a-posteriori estimation as corresponding to the K space data obtains each spatial position
Denoising after dispersion tensor matrix；
Dispersion parameter calculates step：Based on the dispersion tensor matrix after the denoising, dispersion parameter figure is obtained；
The denoising step includes：
Gaussian distribution nature based on Diffusion-weighted imaging model and sampling noise, utilizes the sparse of average diffusivity
Property, denoising function model is built using the method for maximum a-posteriori estimation, the denoising function model is referring to such as following formula
(1)：
Wherein,Represent the estimated value of dispersion tensor matrix, MD is average diffusivity；R () is act on average diffusivity dilute
Constraint function is dredged, λ is corresponding regularization parameter；dmFor the K space data corresponding to m-th of diffusion weighted images；F represents Fu
Vertical leaf encoder matrix；Represent m-th of diffusion weighted images, wherein, I0Indicate the reference picture of no diffusion-weighted,For the phase of m-th of diffusion weighted images, b is the diffusion-weighted factor, gmIt is the disperse corresponding to m-th of diffusion weighted images
Gradient vector gm=(gxm,gym,gzm)T, using the K space data, the formula (1) is solved, obtains each spatial position
Dispersion tensor matrix after corresponding denoising.
2. the denoising method of diffusion tensor imaging according to claim 1, which is characterized in that the sparse constraint
Function constraint is averaged the openness of diffusivity.
3. the denoising method of diffusion tensor imaging according to claim 1, which is characterized in that the sparse constraint
Function utilizes L1Model letter is openness in sparse transform-domain to constrain average diffusivity.
4. the denoising method of diffusion tensor imaging according to claim 1, which is characterized in that the sparse constraint
Function is obtained by calling the following formula to calculate：
Wherein, Ψ represents sparse transformation operator, | | | |1Expression takes L1Norm, D1、D2、D3Respectively dispersion tensor matrix master
Cornerwise element.
5. a kind of denoising system of diffusion tensor imaging, which is characterized in that the system comprises：
Image data acquisition module, for obtaining the K space data corresponding to diffusion-weighted MR imaging image；
Denoising module, for the Gaussian distribution nature based on Diffusion-weighted imaging model and sampling noise, using averagely
Diffusivity it is openness, each spatial position institute is obtained by the K space data using the method for maximum a-posteriori estimation
Dispersion tensor matrix after corresponding denoising；And
Dispersion parameter computing module, for based on the dispersion tensor matrix after the denoising, obtaining dispersion parameter figure；
The denoising module includes：
Model construction unit for the Gaussian distribution nature based on Diffusion-weighted imaging model and sampling noise, utilizes
Averagely diffusivity is openness, and denoising function model, the denoising Function Modules are built using the method for maximum a-posteriori estimation
Type is referring to such as following formula (1)：
Wherein,Represent the estimated value of dispersion tensor matrix, MD is average diffusivity；R () is act on average diffusivity dilute
Constraint function is dredged, λ is corresponding regularization parameter；dmFor the K space data corresponding to m-th of diffusion weighted images；F represents Fu
Vertical leaf encoder matrix；Represent m-th of diffusion weighted images, wherein, I0Indicate the reference picture of no diffusion-weighted,For the phase of m-th of diffusion weighted images, b is the diffusion-weighted factor, gmIt is the disperse corresponding to m-th of diffusion weighted images
Gradient vector gm=(gxm,gym,gzm)T；
Matrix Solving unit for utilizing the K space data, solves the formula (1), and it is right to obtain each spatial position institute
Dispersion tensor matrix after the denoising answered.
6. the denoising system of diffusion tensor imaging according to claim 5, which is characterized in that the sparse constraint
Function constraint is averaged the openness of diffusivity.
7. the denoising system of diffusion tensor imaging according to claim 5, which is characterized in that the sparse constraint
Function utilizes L1Model letter is openness in sparse transform-domain to constrain average diffusivity.
8. the denoising system of diffusion tensor imaging according to claim 5, which is characterized in that the sparse constraint
Function is obtained by calling the following formula to calculate：
Wherein, Ψ represents sparse transformation operator, | | | |1Expression takes L1Norm, D1、D2、D3Respectively dispersion tensor matrix master
Cornerwise element.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410816610.9A CN104599244B (en) | 2014-12-23 | 2014-12-23 | The denoising method and system of diffusion tensor imaging |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410816610.9A CN104599244B (en) | 2014-12-23 | 2014-12-23 | The denoising method and system of diffusion tensor imaging |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104599244A CN104599244A (en) | 2015-05-06 |
| CN104599244B true CN104599244B (en) | 2018-06-15 |
Family
ID=53124994
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410816610.9A Active CN104599244B (en) | 2014-12-23 | 2014-12-23 | The denoising method and system of diffusion tensor imaging |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104599244B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10324154B2 (en) * | 2015-05-13 | 2019-06-18 | General Electric Company | Generalized spherical deconvolution in diffusion magnetic resonance imaging |
| CN106646300B (en) * | 2016-09-23 | 2019-07-16 | 中国科学院武汉物理与数学研究所 | A kind of hyperpolarized gas nuclear magnetic resonance image processing method |
| CN108090937B (en) * | 2018-02-09 | 2021-03-19 | 奥泰医疗系统有限责任公司 | Self-adaptive correction method for magnetic resonance diffusion weighted imaging |
| CN108765332B (en) * | 2018-05-23 | 2022-05-06 | 成都信息工程大学 | Ellipse search window and parameter self-adaptive non-local mean denoising method |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102663701A (en) * | 2011-12-12 | 2012-09-12 | 中国科学院深圳先进技术研究院 | Method and system for reconstructing magnetic resonance parameters |
| CN103356193A (en) * | 2013-07-19 | 2013-10-23 | 哈尔滨工业大学深圳研究生院 | Method and system for performing rapid diffusion tensor imaging under compression sensing framework |
| CN103595414A (en) * | 2012-08-15 | 2014-02-19 | 王景芳 | Sparse sampling and signal compressive sensing reconstruction method |
| CN103985099A (en) * | 2014-05-30 | 2014-08-13 | 成都信息工程学院 | Dispersion tensor magnetic resonance image tensor domain non-local mean denoising method |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160061923A1 (en) * | 2013-04-03 | 2016-03-03 | Choukri Mekkaoui | Sheet tractography using diffusion tensor mri |
| CN103705239B (en) * | 2013-12-05 | 2016-01-20 | 深圳先进技术研究院 | Magnetic resonance parameters formation method and system |
-
2014
- 2014-12-23 CN CN201410816610.9A patent/CN104599244B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102663701A (en) * | 2011-12-12 | 2012-09-12 | 中国科学院深圳先进技术研究院 | Method and system for reconstructing magnetic resonance parameters |
| CN103595414A (en) * | 2012-08-15 | 2014-02-19 | 王景芳 | Sparse sampling and signal compressive sensing reconstruction method |
| CN103356193A (en) * | 2013-07-19 | 2013-10-23 | 哈尔滨工业大学深圳研究生院 | Method and system for performing rapid diffusion tensor imaging under compression sensing framework |
| CN103985099A (en) * | 2014-05-30 | 2014-08-13 | 成都信息工程学院 | Dispersion tensor magnetic resonance image tensor domain non-local mean denoising method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104599244A (en) | 2015-05-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Lin et al. | Artificial intelligence for MR image reconstruction: an overview for clinicians | |
| Tourbier et al. | An efficient total variation algorithm for super-resolution in fetal brain MRI with adaptive regularization | |
| CN104379058B (en) | Merge the multiple shooting scan protocols of multiplexing sensitivity encoding (MUSE) for high-resolution MRI | |
| Raffelt et al. | Reorientation of fiber orientation distributions using apodized point spread functions | |
| CN104574298B (en) | A kind of noise-reduction method of more b values diffusion weightings images based on mutual information | |
| Benning et al. | Phase reconstruction from velocity-encoded MRI measurements–a survey of sparsity-promoting variational approaches | |
| Poot et al. | Super‐resolution for multislice diffusion tensor imaging | |
| Reisert et al. | Fiber continuity: an anisotropic prior for ODF estimation | |
| Gramfort et al. | Denoising and fast diffusion imaging with physically constrained sparse dictionary learning | |
| Holme et al. | ENLIVE: an efficient nonlinear method for calibrationless and robust parallel imaging | |
| CN104599244B (en) | The denoising method and system of diffusion tensor imaging | |
| CN104635188B (en) | K space reconstructions method and MR imaging method | |
| Kames et al. | Rapid two-step dipole inversion for susceptibility mapping with sparsity priors | |
| Mani et al. | qModeL: A plug‐and‐play model‐based reconstruction for highly accelerated multi‐shot diffusion MRI using learned priors | |
| KR20220070502A (en) | Maxwell Parallel Imaging | |
| Hossbach et al. | Deep learning‐based motion quantification from k‐space for fast model‐based magnetic resonance imaging motion correction | |
| CN105931242B (en) | Dynamic nuclear magnetic resonance (DNMR) image rebuilding method based on dictionary learning and time gradient | |
| Alhamud et al. | Motion artifact reduction in pediatric diffusion tensor imaging using fast prospective correction | |
| Christiansen et al. | Total variation regularization of matrix‐valued images | |
| CN104545916B (en) | Diffusion tensor imaging denoising method and system | |
| EP3953725A1 (en) | Parameter map determination for time domain magnetic resonance | |
| Landman et al. | Diffusion tensor imaging at low SNR: nonmonotonic behaviors of tensor contrasts | |
| Goodwin-Allcock et al. | How can spherical CNNs benefit ML-based diffusion MRI parameter estimation? | |
| Dai et al. | Multi‐band multi‐shot diffusion MRI reconstruction with joint usage of structured low‐rank constraints and explicit phase mapping | |
| CN104504657B (en) | Method and device for de-noising magnetic resonance diffusion tensor |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant |