CN105303577B - A kind of method of brain white matter integrity imaging - Google Patents
A kind of method of brain white matter integrity imaging Download PDFInfo
- Publication number
- CN105303577B CN105303577B CN201510751670.1A CN201510751670A CN105303577B CN 105303577 B CN105303577 B CN 105303577B CN 201510751670 A CN201510751670 A CN 201510751670A CN 105303577 B CN105303577 B CN 105303577B
- Authority
- CN
- China
- Prior art keywords
- mrow
- msup
- msub
- distribution function
- richardson
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10088—Magnetic resonance imaging [MRI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30016—Brain
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Magnetic Resonance Imaging Apparatus (AREA)
Abstract
The method for solving the partial volume effect of brain white matter integrity imaging using hybrid response kernel function, comprises the following steps:Brain MR data is read, obtains the magnetic resonance signal S (g) for applying gradient direction g, does not apply the magnetic resonance signal S of gradient direction0And gradient direction data, required area-of-interest is chosen, and calculate diffusive attenuation signal S (the g)/S in the region0;Using Richardson-Lucy iterative algorithms by diffusive attenuation signal S (g)/S of each voxel in area-of-interest0The model that there is elliposoidal to be distributed is modeled as one by one, and increase total variation regularization and remove influence of noise, dispersal direction distribution function f (v) is obtained by Richardson-Lucy iterative algorithms, isotropic diffusion value f is isolated from f (v)(m+1)With the fiber distribution function fod of anisotropy parameter.The present invention relates to probability theory and the theory of total variation regularization, compared to conventional method, calculating speed is fast, imaging angle high resolution, can distinguish ectocinerea and white matter of brain, and experiment effect is good.
Description
Technical field
The present invention relates to a kind of method of brain white matter integrity imaging, is based especially on the hybrid response kernel function solutions of RL algorithms
The certainly partial volume effect in fiber imaging, utilizes diffusion-weighted magnetic resonance imaging (Diffusion Weighted Magnetic
Resonance Imaging, DW-MRI) data and multiple response Kernel-Based Methods combination RL algorithms carry out machine direction distribution function
Fitting, so as to obtain the direction of fiber, and obtain more accurately in grey matter and the obvious intersection of white matter partial volume effect
Machine direction.The machine direction for making to obtain is more conducive to the tracking of fiber.Belong to medical imaging, Nervous System Anatomy field.
Background technology
Nuclear magnetic resonance (MRI) be it is a kind of be widely used in medical imaging without diffusivity method, as unique live body
Non-invasive method, it is helping people to obtain clinical neuromechanism information and is understanding the function and connection between cerebral cortex region
System etc. has played huge effect.The trend of brain white matter integrity exists close with mental disorder and cranial surgery medical conditions
Contact, development, schizophrenia, congenital research with acquired leukoencephalopathy and dementia etc. of these information for brain carry
New application prospect is supplied.Fiber imaging algorithm based on diffusion-weighted magnetic resonance imaging (DW-MRI) can be from DW-MRI data
Machine direction information is obtained, foundation is provided for clinical medicine diagnosis, new method is provided for brain scientific research.
In all kinds of MRI methods, diffusion tensor imaging (Diffusion Tensor Imaging, DTI) is more important
One kind, for a variety of brain diseases clinical diagnosises being currently known, DTI technologies have all played irreplaceable effect.But pass
The DTI methods of system are assumed only to contain a fiber in voxel, therefore can not differentiate the complicated fiber such as intersection, bottleneck, scattered
Structure, and often there are intersection, branch or the complex situations of fusion in the nerve fibre of human brain, become the machine direction that DTI is reconstructed
It must not know.
In order to overcome DTI inherent limitation, high angular resolution diffusion magnetic resonance imaging (HARDI) technology is arisen at the historic moment.
On basis based on HARDI technologies, it is proposed that the method for multiple fiber reconstruct, such as:Q-ball, diffusion spectrum imaging
(Diffusion Spectrum Imaging, DSI), sphere deconvolution (Spherical deconvolution, SD) etc..From
At present, although every kind of method all solves the imaging problem of complicated white matter few fibers, most of HARDI well
Method does not explain the influence that the partial volume effect of non-white matter (grey matter and cerebrospinal fluid) is imaged to fiber.The field is still at present
There is not a kind of mathematical modeling for truly solving non-white confrontation brain fiber Imaging.
The content of the invention
In order to overcome weak point above-mentioned in the prior art, present invention proposition is a kind of to utilize multiple response based on RL algorithms
Kernel function handles the imaging method of the partial volume effect of non-white matter, so that the weight of the fiber of white matter and grey matter intersection
Structure goes out the machine direction of high resolution and low error, and further distinguishes white matter of brain and ectocinerea region.
(1) brain MR data is read, the magnetic resonance diffusion signal S (g) for applying that gradient direction is g is obtained, does not apply
The magnetic resonance diffusion signal S of gradient direction0And gradient direction data, the data of collection are pre-processed, including:High frequency is filtered
Ripple, spatial noise reduction, remove linear drift etc.;Area-of-interest needed for choosing, and calculate the diffusive attenuation signal S in the region
(g)/S0;
(2) Richardson-Lucy iterative algorithms are utilized by the diffusive attenuation signal S of each voxel in area-of-interest
(g)/S0The model that there is elliposoidal to be distributed is modeled as one by one, and is increased total variation regularization and removed influence of noise, modeling process
It is as follows:
2.1) voxel model micro-structure scheme:By diffusive attenuation signal S (g)/S0It is assumed to be along the signal for rebuilding vector v and rings
Kernel function R (v, g) and dispersal direction distribution function f (v) is answered in Spherical Surface S2On convolution:
R (v, g) represents hybrid response kernel function, and it is responded each in kernel function and ectocinerea using the single fiber of white matter of brain
To same sex response kernel function composition, can effectively solve the partial volume effect in brain white matter integrity imaging, g={ gi∈R1×3
| i=1 ..., n } it is Diffusion direction, v={ vj∈R1×3| j=1 ..., m } it is vectorial to rebuild;Its mathematical modeling is:
R (v, g)=[Rani Risot]
Wherein,Represent that anisotropy response kernel function and isotropism are rung in voxel respectively
Kernel function is answered, b is diffusion-sensitive coefficient;Anisotropy response kernel function RaniIt is by the response core group along m reconstruction direction v
Into each core that responds is identical round pie, and simply their distribution arrangement is different;And isotropism response kernel function only by
One individually responds core composition, and its shape is spherical shape;Represent that diffusion is carried out along a principal direction,In all directions, its diffusion is consistent, and wherein α, β represent fiber diffusion, and its numerical value can be in related text
Middle access is offered to obtain;
2.2) a kind of mathematical modeling based on Richardson-Lucy iterative algorithms is as follows:
Because the data acquisition of magnetic resonance signal is unsatisfactory, often with certain noise, in order to overcome as much as possible
The influence that noise is rebuild to machine direction, using the Richardson-Lucy iterative algorithms in the image recovery method of classics, have
The nonnegativity of the machine direction distribution function of the guarantee reconstruct of effect, reduces the quantity for reconstructing pseudo- peak;The mathematical modeling is by most
The method of maximum-likelihood estimation, which calculates, to be obtained;If diffusion-weighted magnetic resonance signals have n Diffusion direction, and along m
Rebuild vector to be rebuild, then its mathematical modeling is:
Wherein k is iterations, and i represents the i-th row of vector, f(k)It is the column vector that length that kth time iteration obtains is m,
Represent that S is the length for including HARDI signals along the m dispersal direction distribution function rebuild direction and be evenly distributed on sphere
For n column vector;
2.3) the TV regularization models based on Richardson-Lucy iterative algorithms are as follows:
Richardson-Lucy algorithms have limitation in itself, although the algorithm can suppress noise to a certain extent
The influence to imaging;But with the increase of iterations, the optimal solution being calculated by Richardson-Lucy algorithms can quilt
Noise is controlled;In order to overcome the harmful effect that iterations increase is brought, total variation is added on the basis of this algorithm
Regularization term suppresses noise, and its mathematical modeling is:
Wherein,
It is the total variation regularization term of kth time iteration;▽f(k)It is the image dispersal direction distribution letter of kth time iteration
Number f(k)Gradient, div represent divergence, λ is regularization parameter;
(3) dispersal direction distribution function f (v), dispersal direction distribution function f (v) calculating side are obtained by iterative calculation
Method comprises the following steps:
3.1) the uniform sampling m discrete points in unit sphere, this m reconstruction vector v, meter are obtained by origin of the centre of sphere
Fiber response kernel function R (v, g) value is calculated, obtains n × (m+1) circulant matrix;
3.2) analogue data analog simulation is utilized, iterative initial value is set;Make f(0)Letter is distributed for the dispersal direction of isotropic
Number, its amplitude are arranged to 1;Because f(0)Initial value is positive, is naturally met so the nonnegativity of algorithm limits;It can pass through
The selected λ value of experiment, carrys out the influence of equilibrium criterion item and regular terms to iterative algorithm;
3.3) voxel of region of interest is pre-processed using the RL algorithms without regular terms;Obtain the diffusion of each voxel
Direction distribution function f, the initial propagations direction distribution function value as regularization RL algorithms;
3.4) stopping criterion for iteration is set:First, iterations;First, iteration error, the iteration error is made to be:
So iterations is more than optimal iterations or iteration error d < ε, (general ε takes 10-3) it is used as iteration ends
Condition;(for analogue data, it is known that optimal dispersal direction distribution function solution f, f in institute's above formula iteration error(k)=f;)
3.5) by iteration, optimal dispersal direction distribution function f has been obtained, it is the column vector of m+1 row, wherein finally
The value f of one row(m+1)Relative scale size in as each voxel shared by isotropic diffusion;And preceding m row are formed in voxel
Anisotropic fiber direction distribution function fod, i.e. machine direction distribution function fod are dispersal direction distribution function f preceding m-1
;It is known that by experiment:Work as f(m+1)During > θ, the diffusion in voxel is in isotropism;The weight of machine direction is not carried out to it
Structure, and for f(m+1)< θ voxel, fitting machine direction distribution function fod direction is emulated using MATLAB;
3.6) three-dimensional imaging is carried out to machine direction distribution function fod in perceptive construction on mathematics, and it is fine by searching for
The extreme point in direction distribution function value is tieed up to obtain the principal direction of fiber.
The present invention is exactly to solve the influence that non-white matter part is imaged to brain fiber using the RL algorithms of multiple response function.This
Invention is related to the theory of maximal possibility estimation and Medical Image Processing;
It is an advantage of the invention that:Calculating speed is fast, imaging angle high resolution, and it is high to calculate robustness.According to the author's experience
Apparently, income effect of the present invention is the current field best effects.
Brief description of the drawings
Fig. 1 is analogue data result figure of the present invention.Wherein, analogue data is produced by following formula:
Wherein f represents i-th with the ratio shared by fiber,f1=0.5, f2=0.5, S0=1, b=3000s/
mm2, diffusion tensor D characteristic value is:λ1=1.8 × 10-3mm2/ s, λ2=0.3 × 10-3mm2/s,λ3=0.3 × 10-3mm2/s。
81 equally distributed diffusion-weighted magnetic resonance imaging directions in hemisphere face, hemisphere face sampling number is 321, first in figure
Row represents angle, and the second row represents machine direction, and the third line represents imaging model, and black line illustrates the direction of two fibers (to pass through
Calculate diffusion peak value to obtain).
Fig. 2 is actual clinical effect data figure of the present invention.Real data comes from Harvard University's hospital attached to a medical college
(Brigham and Women ' s Hospital, Brockton VA Hospital, McLean Hospital), utilizes 3-TGE
The brain data that system and double echo plane imaging sequence extract from true human brain, data acquisition parameters are:TR=
17000ms, TE=78ms.Voxel amount is 144 × 144 × 85, is that 24cm. is cut parallel to 85 axial directions of AC-PC lines into image field
Piece, per thickness degree 1.7mm. from 51 different gradient direction scan datas, b=900s/mm2,8 b=0's of diffusion-sensitive coefficient
Scan data.
Embodiment
The present invention is further illustrated below in conjunction with the accompanying drawings.
A kind of method of brain white matter integrity imaging of the present invention, comprises the following steps:
(1) brain MR data is read, the magnetic resonance diffusion signal S (g) for applying that gradient direction is g is obtained, does not apply
The magnetic resonance diffusion signal S of gradient direction0And gradient direction data, the data of collection are pre-processed, including:High frequency is filtered
Ripple, spatial noise reduction, remove linear drift etc.;Area-of-interest needed for choosing, and calculate the diffusive attenuation signal S in the region
(g)/S0;
(2) Richardson-Lucy iterative algorithms are utilized by the diffusive attenuation signal S of each voxel in area-of-interest
(g)/S0The model that there is elliposoidal to be distributed is modeled as one by one, and is increased total variation regularization and removed influence of noise, modeling process
It is as follows:
2.1) voxel model micro-structure scheme:By diffusive attenuation signal S (g)/S0It is assumed to be along the signal for rebuilding vector v and rings
Kernel function R (v, g) and dispersal direction distribution function f (v) is answered in Spherical Surface S2On convolution:
R (v, g) represents hybrid response kernel function, and it is responded each in kernel function and ectocinerea using the single fiber of white matter of brain
To same sex response kernel function composition, can effectively solve the partial volume effect in brain white matter integrity imaging, g={ gi∈R1×3
| i=1 ..., n } it is Diffusion direction, v={ vj∈R1×3| j=1 ..., m } it is vectorial to rebuild;Its mathematical modeling is:
R (v, g)=[Rani Risot]
Wherein,Represent that anisotropy response kernel function and isotropism are rung in voxel respectively
Kernel function is answered, b is diffusion-sensitive coefficient;Anisotropy response kernel function RaniIt is by the response core group along m reconstruction direction v
Into each core that responds is identical round pie, and simply their distribution arrangement is different;And isotropism response kernel function only by
One individually responds core composition, and its shape is spherical shape;Represent that diffusion is carried out along a principal direction,In all directions, its diffusion is consistent, and wherein α, β represent fiber diffusion, and its numerical value can be in related text
Middle access is offered to obtain;
2.2) a kind of mathematical modeling based on Richardson-Lucy iterative algorithms is as follows:
Because the data acquisition of magnetic resonance signal is unsatisfactory, often with certain noise, in order to overcome as much as possible
The influence that noise is rebuild to machine direction, using the Richardson-Lucy iterative algorithms in the image recovery method of classics, have
The nonnegativity of the machine direction distribution function of the guarantee reconstruct of effect, reduces the quantity for reconstructing pseudo- peak;The mathematical modeling is by most
The method of maximum-likelihood estimation, which calculates, to be obtained;If diffusion-weighted magnetic resonance signals have n Diffusion direction, and along m
Rebuild vector to be rebuild, then its mathematical modeling is:
Wherein k is iterations, and i represents the i-th row of vector, f(k)It is the column vector that length that kth time iteration obtains is m,
Represent that S is the length for including HARDI signals along the m dispersal direction distribution function rebuild direction and be evenly distributed on sphere
For n column vector;
2.3) the TV regularization models based on Richardson-Lucy iterative algorithms are as follows:
Richardson-Lucy algorithms have limitation in itself, although the algorithm can suppress noise to a certain extent
The influence to imaging;But with the increase of iterations, the optimal solution being calculated by Richardson-Lucy algorithms can quilt
Noise is controlled;In order to overcome the harmful effect that iterations increase is brought, total variation is added on the basis of this algorithm
Regularization term suppresses noise, and its mathematical modeling is:
Wherein,
It is the total variation regularization term of kth time iteration;It is the image dispersal direction distribution letter of kth time iteration
Number f(k)Gradient, div represent divergence, λ is regularization parameter;
(3) dispersal direction distribution function f (v), dispersal direction distribution function f (v) calculating side are obtained by iterative calculation
Method comprises the following steps:
3.1) the uniform sampling m discrete points in unit sphere, this m reconstruction vector v, meter are obtained by origin of the centre of sphere
Fiber response kernel function R (v, g) value is calculated, obtains n × (m+1) circulant matrix;
3.2) analogue data analog simulation is utilized, iterative initial value is set;Make f(0)Letter is distributed for the dispersal direction of isotropic
Number, its amplitude are arranged to 1;Because f(0)Initial value is positive, is naturally met so the nonnegativity of algorithm limits;It can pass through
The selected λ value of experiment, carrys out the influence of equilibrium criterion item and regular terms to iterative algorithm;
3.3) voxel of region of interest is pre-processed using the RL algorithms without regular terms;Obtain the diffusion of each voxel
Direction distribution function f, the initial propagations direction distribution function value as regularization RL algorithms;
3.4) stopping criterion for iteration is set:First, iterations;First, iteration error, the iteration error is made to be:
So iterations is more than optimal iterations or iteration error d < ε, (general ε takes 10-3) it is used as iteration ends
Condition;(for analogue data, it is known that optimal dispersal direction distribution function solution f, f in institute's above formula iteration error(k)=f;)
3.5) by iteration, optimal dispersal direction distribution function f has been obtained, it is the column vector of m+1 row, wherein finally
The value f of one row(m+1)Relative scale size in as each voxel shared by isotropic diffusion;And preceding m row are formed in voxel
Anisotropic fiber direction distribution function fod, i.e. machine direction distribution function fod are dispersal direction distribution function f preceding m-1
;It is known that by experiment:Work as f(m+1)During > θ, the diffusion in voxel is in isotropism;The weight of machine direction is not carried out to it
Structure, and for f(m+1)< θ voxel, fitting machine direction distribution function fod direction is emulated using MATLAB;
3.6) three-dimensional imaging is carried out to machine direction distribution function fod in perceptive construction on mathematics, and it is fine by searching for
The extreme point in direction distribution function value is tieed up to obtain the principal direction of fiber.
Claims (3)
1. a kind of method of brain white matter integrity imaging, its imaging method comprise the following steps:
(1) brain MR data is read, the magnetic resonance diffusion signal S (g) for applying that gradient direction is g is obtained, does not apply gradient
The magnetic resonance diffusion signal S in direction0And gradient direction data, the data of collection are pre-processed, including:High frequency filter, it is empty
Between noise reduction, remove linear drift;Area-of-interest needed for choosing, and calculate diffusive attenuation signal S (the g)/S in the region0;
(2) using Richardson-Lucy iterative algorithms by the diffusive attenuation signal S (g) of each voxel in area-of-interest/
S0The model that there is elliposoidal to be distributed is modeled as one by one, and increases total variation regularization and removes influence of noise, and modeling process is as follows:
2.1) voxel model micro-structure scheme:By diffusive attenuation signal S (g)/S0It is assumed to be along the signal response core for rebuilding vector v
Function R (v, g) and dispersal direction distribution function f (v) is in Spherical Surface S2On convolution:
<mrow>
<mi>S</mi>
<mrow>
<mo>(</mo>
<mi>g</mi>
<mo>)</mo>
</mrow>
<mo>/</mo>
<msub>
<mi>S</mi>
<mn>0</mn>
</msub>
<mo>=</mo>
<mi>R</mi>
<mrow>
<mo>(</mo>
<mi>v</mi>
<mo>,</mo>
<mi>g</mi>
<mo>)</mo>
</mrow>
<mo>&CircleTimes;</mo>
<mi>f</mi>
<mrow>
<mo>(</mo>
<mi>v</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<msub>
<mo>&Integral;</mo>
<msup>
<mi>S</mi>
<mn>2</mn>
</msup>
</msub>
<mi>R</mi>
<mrow>
<mo>(</mo>
<mi>v</mi>
<mo>,</mo>
<mi>g</mi>
<mo>)</mo>
</mrow>
<mi>f</mi>
<mrow>
<mo>(</mo>
<mi>v</mi>
<mo>)</mo>
</mrow>
<mi>d</mi>
<mi>v</mi>
</mrow>
R (v, g) represents fiber response kernel function, and it is responded each to same in kernel function and ectocinerea using the single fiber of white matter of brain
Property response kernel function composition, can effectively solve brain white matter integrity imaging in partial volume effect, g={ gi∈R1×3| i=
1 ..., n } it is Diffusion direction, v={ vj∈R1×3| j=1 ..., m } it is vectorial to rebuild;Its mathematical modeling is:
R (v, g)=[Rani Risot]
Wherein,Anisotropy response kernel function and isotropism response core in voxel are represented respectively
Function, b are diffusion-sensitive coefficients;Anisotropy response kernel function RaniIt is made up of the response core along m reconstruction direction v, often
Individual response core is all identical round pie, and simply their distribution arrangement is different;And isotropism response kernel function is only by one
Individually response core composition, its shape is spherical shape;Represent that diffusion is carried out along a principal direction,In all directions, its diffusion is consistent, and wherein α, β represent fiber diffusion;
2.2) a kind of mathematical modeling based on Richardson-Lucy iterative algorithms is as follows:
Because the data acquisition of magnetic resonance signal is unsatisfactory, often with certain noise, in order to overcome noise as much as possible
To machine direction rebuild influence, using classics image recovery method in Richardson-Lucy iterative algorithms, effectively
Ensure the nonnegativity of the machine direction distribution function of reconstruct, reduce the quantity for reconstructing pseudo- peak;It should be changed based on Richardson-Lucy
Mathematical modeling for algorithm is to calculate to obtain by the method for maximal possibility estimation;If diffusion-weighted magnetic resonance signals have n
Diffusion direction, and rebuild vector along m and rebuild, then its mathematical modeling is:
<mrow>
<mi>f</mi>
<msup>
<msub>
<mrow>
<mo>(</mo>
<mi>v</mi>
<mo>)</mo>
</mrow>
<mi>i</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>k</mi>
<mo>+</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</msup>
<mo>=</mo>
<mi>f</mi>
<msup>
<msub>
<mrow>
<mo>(</mo>
<mi>v</mi>
<mo>)</mo>
</mrow>
<mi>i</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>k</mi>
<mo>)</mo>
</mrow>
</msup>
<mfrac>
<msub>
<mrow>
<mo>&lsqb;</mo>
<mi>R</mi>
<msup>
<mrow>
<mo>(</mo>
<mi>v</mi>
<mo>,</mo>
<mi>g</mi>
<mo>)</mo>
</mrow>
<mi>T</mi>
</msup>
<mi>S</mi>
<mo>&rsqb;</mo>
</mrow>
<mi>i</mi>
</msub>
<msub>
<mrow>
<mo>&lsqb;</mo>
<mi>R</mi>
<msup>
<mrow>
<mo>(</mo>
<mi>v</mi>
<mo>,</mo>
<mi>g</mi>
<mo>)</mo>
</mrow>
<mi>T</mi>
</msup>
<mi>R</mi>
<mrow>
<mo>(</mo>
<mi>v</mi>
<mo>,</mo>
<mi>g</mi>
<mo>)</mo>
</mrow>
<mi>f</mi>
<msup>
<mrow>
<mo>(</mo>
<mi>v</mi>
<mo>)</mo>
</mrow>
<mrow>
<mo>(</mo>
<mi>k</mi>
<mo>)</mo>
</mrow>
</msup>
<mo>&rsqb;</mo>
</mrow>
<mi>i</mi>
</msub>
</mfrac>
</mrow>
Wherein k is iterations, and i represents the i-th row of vector, f(k)It is that the length that kth time iteration obtains is m column vector, represents
The dispersal direction distribution function rebuild direction along m and be evenly distributed on sphere, S are that the length comprising HARDI signals is n
Column vector;
2.3) the TV regularization models based on Richardson-Lucy iterative algorithms are as follows:
Richardson-Lucy algorithms have limitation in itself, although the Richardson-Lucy algorithms can be to a certain degree
The upper influence to imaging for suppressing noise;But with the increase of iterations, it is calculated by Richardson-Lucy algorithms
Optimal solution can be controlled by noise;In order to overcome the harmful effect that iterations increase is brought, calculated in this Richardson-Lucy
Total variation regularization term is added on the basis of method to suppress noise, its mathematical modeling is:
<mrow>
<mi>f</mi>
<msup>
<msub>
<mrow>
<mo>(</mo>
<mi>v</mi>
<mo>)</mo>
</mrow>
<mi>i</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>k</mi>
<mo>+</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</msup>
<mo>=</mo>
<mi>f</mi>
<msup>
<msub>
<mrow>
<mo>(</mo>
<mi>v</mi>
<mo>)</mo>
</mrow>
<mi>i</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>k</mi>
<mo>)</mo>
</mrow>
</msup>
<mfrac>
<msub>
<mrow>
<mo>&lsqb;</mo>
<mi>R</mi>
<msup>
<mrow>
<mo>(</mo>
<mi>v</mi>
<mo>,</mo>
<mi>g</mi>
<mo>)</mo>
</mrow>
<mi>T</mi>
</msup>
<mi>S</mi>
<mo>&rsqb;</mo>
</mrow>
<mi>i</mi>
</msub>
<msub>
<mrow>
<mo>&lsqb;</mo>
<mi>R</mi>
<msup>
<mrow>
<mo>(</mo>
<mi>v</mi>
<mo>,</mo>
<mi>g</mi>
<mo>)</mo>
</mrow>
<mi>T</mi>
</msup>
<mi>R</mi>
<mrow>
<mo>(</mo>
<mi>v</mi>
<mo>,</mo>
<mi>g</mi>
<mo>)</mo>
</mrow>
<mi>f</mi>
<msup>
<mrow>
<mo>(</mo>
<mi>v</mi>
<mo>)</mo>
</mrow>
<mrow>
<mo>(</mo>
<mi>k</mi>
<mo>)</mo>
</mrow>
</msup>
<mo>&rsqb;</mo>
</mrow>
<mi>i</mi>
</msub>
</mfrac>
<mo>&times;</mo>
<msup>
<msub>
<mi>TV</mi>
<mi>i</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>k</mi>
<mo>)</mo>
</mrow>
</msup>
</mrow>
Wherein,
<mrow>
<msup>
<msub>
<mi>TV</mi>
<mi>i</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>k</mi>
<mo>)</mo>
</mrow>
</msup>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<mrow>
<mn>1</mn>
<mo>-</mo>
<mi>&lambda;</mi>
<mi>d</mi>
<mi>i</mi>
<mi>v</mi>
<mrow>
<mo>(</mo>
<mfrac>
<mrow>
<mo>&dtri;</mo>
<msub>
<mrow>
<mo>&lsqb;</mo>
<msup>
<mi>f</mi>
<mrow>
<mo>(</mo>
<mi>k</mi>
<mo>)</mo>
</mrow>
</msup>
<mo>&rsqb;</mo>
</mrow>
<mi>i</mi>
</msub>
</mrow>
<mrow>
<mo>&lsqb;</mo>
<msub>
<mrow>
<mo>|</mo>
<mrow>
<mo>&dtri;</mo>
<msup>
<mi>f</mi>
<mrow>
<mo>(</mo>
<mi>k</mi>
<mo>)</mo>
</mrow>
</msup>
</mrow>
<mo>|</mo>
</mrow>
<mi>i</mi>
</msub>
<mo>&rsqb;</mo>
</mrow>
</mfrac>
<mo>)</mo>
</mrow>
</mrow>
</mfrac>
</mrow>
TVi (k)It is the total variation regularization term of kth time iteration;It is the image dispersal direction distribution function f of kth time iteration(k)
Gradient, div represent divergence, λ is regularization parameter;
(3) dispersal direction distribution function f (v), dispersal direction distribution function f (v) computational methods bag are obtained by iterative calculation
Include following steps:
3.1) the uniform sampling m discrete points in unit sphere, this m reconstruction vector v is obtained by origin of the centre of sphere, is calculated fine
Dimension response kernel function R (v, g) value, obtains n × (m+1) circulant matrix;
3.2) analogue data analog simulation is utilized, iterative initial value is set;Make f(0)For the dispersal direction distribution function of isotropic,
Its amplitude is arranged to 1;Because f(0)Initial value is positive, is naturally met so the nonnegativity of algorithm limits;It is selected by testing
λ value, carry out the influence of equilibrium criterion item and regular terms to iterative algorithm;
3.3) voxel of region of interest is pre-processed using the Richardson-Lucy algorithms without regular terms;Obtain each
The dispersal direction distribution function f of voxel, the initial propagations direction distribution function as regularization Richardson-Lucy algorithms
Value;
3.4) stopping criterion for iteration is set:First, iterations;First, iteration error, the iteration error is made to be:
<mrow>
<mi>d</mi>
<mo>=</mo>
<mfrac>
<mrow>
<mo>|</mo>
<mo>|</mo>
<msup>
<mi>f</mi>
<mrow>
<mo>(</mo>
<mi>k</mi>
<mo>+</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</msup>
<mo>-</mo>
<msup>
<mi>f</mi>
<mrow>
<mo>(</mo>
<mi>k</mi>
<mo>)</mo>
</mrow>
</msup>
<mo>|</mo>
<mo>|</mo>
</mrow>
<mrow>
<mo>|</mo>
<mo>|</mo>
<msup>
<mi>f</mi>
<mrow>
<mo>(</mo>
<mi>k</mi>
<mo>)</mo>
</mrow>
</msup>
<mo>|</mo>
<mo>|</mo>
</mrow>
</mfrac>
</mrow>
So iterations is more than optimal iterations or iteration error d < ε take 10 as stopping criterion for iteration, ε-3;For
Analogue data, it is known that optimal dispersal direction distribution function f, f in institute's above formula iteration error(k)=f;
3.5) by iteration, optimal dispersal direction distribution function f has been obtained, it is the column vector of m+1 row, and wherein last is arranged
Value f(m+1)Relative scale size in as each voxel shared by isotropic diffusion;And preceding m row form in voxel respectively to
Foreign fiber direction distribution function fod, i.e. machine direction distribution function fod are dispersal direction distribution function f preceding m-1 items;When
f(m+1)During > θ, the diffusion in voxel is in isotropism;The reconstruct of machine direction is not carried out to it, and for f(m+1)< θ body
Element, fitting machine direction distribution function fod direction is emulated using MATLAB;
3.6) three-dimensional imaging is carried out to machine direction distribution function fod in perceptive construction on mathematics, and by searching for fiber side
Extreme point into distribution function value obtains the principal direction of fiber.
2. the method as described in claim 1, it is characterised in that:Exist simultaneously in each region of described 2.1) midbrain each
To the same sex and anisotropy parameter, hybrid response kernel function fully utilizes isotropism and anisotropy response kernel function, can
To remove isotropic diffusion part in white matter of brain so that the fiber of imaging has more preferable anisotropy, there is more preferable imaging
Effect;White matter of brain and ectocinerea region can more be further discriminated between out.
3. the method as described in claim 1, it is characterised in that:Regularization parameter initial value can neither in described step (2)
It is too small can not be too big:If λ is too small, Richardson-Lucy algorithmic procedures are mainly controlled by data item model, regularization
Optimization Progress is slower;If λ is too big, Richardson-Lucy algorithmic procedures are then mainly controlled by regular terms, gained knot
Fruit can deviate true solution;Preferably regularization parameter is selected by the method for many experiments setting different parameters value.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510751670.1A CN105303577B (en) | 2015-11-09 | 2015-11-09 | A kind of method of brain white matter integrity imaging |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510751670.1A CN105303577B (en) | 2015-11-09 | 2015-11-09 | A kind of method of brain white matter integrity imaging |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105303577A CN105303577A (en) | 2016-02-03 |
CN105303577B true CN105303577B (en) | 2017-12-05 |
Family
ID=55200799
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510751670.1A Active CN105303577B (en) | 2015-11-09 | 2015-11-09 | A kind of method of brain white matter integrity imaging |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105303577B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110197498B (en) * | 2019-04-21 | 2023-09-01 | 渤海大学 | Non-convex regularized deconvolution-based intra-voxel multi-fiber trend estimation method |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6697538B1 (en) * | 1999-07-30 | 2004-02-24 | Wisconsin Alumni Research Foundation | Apparatus for producing a flattening map of a digitized image for conformally mapping onto a surface and associated method |
CN103279633A (en) * | 2013-03-26 | 2013-09-04 | 浙江工业大学 | Brain fiber three-dimensional display method based on diffusion-weighted magnetic resonance data |
-
2015
- 2015-11-09 CN CN201510751670.1A patent/CN105303577B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6697538B1 (en) * | 1999-07-30 | 2004-02-24 | Wisconsin Alumni Research Foundation | Apparatus for producing a flattening map of a digitized image for conformally mapping onto a surface and associated method |
CN103279633A (en) * | 2013-03-26 | 2013-09-04 | 浙江工业大学 | Brain fiber three-dimensional display method based on diffusion-weighted magnetic resonance data |
Non-Patent Citations (2)
Title |
---|
A swarm tracking approach for stochastic white matter tractography;Yuanjing Feng et al;《2011 IEEE International Symposium on Biomedical Imaging: From Nano to Macro》;20110609;803-807 * |
一种重建脑白质内神经纤维的新算法;赵欣 等;《生物医学工程与临床》;20061231(第S1期);33 * |
Also Published As
Publication number | Publication date |
---|---|
CN105303577A (en) | 2016-02-03 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Wilkins et al. | Fiber estimation and tractography in diffusion MRI: development of simulated brain images and comparison of multi-fiber analysis methods at clinical b-values | |
Fogtmann et al. | A unified approach to diffusion direction sensitive slice registration and 3-D DTI reconstruction from moving fetal brain anatomy | |
Zhang et al. | Characterization of U-shape streamline fibers: Methods and applications | |
Smith et al. | Anatomically-constrained tractography: improved diffusion MRI streamlines tractography through effective use of anatomical information | |
Bastiani et al. | Human cortical connectome reconstruction from diffusion weighted MRI: the effect of tractography algorithm | |
Westin et al. | Processing and visualization for diffusion tensor MRI | |
Wig et al. | An approach for parcellating human cortical areas using resting-state correlations | |
US9940712B2 (en) | Quantitating disease progression from the MRI images of multiple sclerosis patients | |
Raffelt et al. | Apparent fibre density: a novel measure for the analysis of diffusion-weighted magnetic resonance images | |
WO2024083057A1 (en) | Disease prediction system using graph convolutional neural network and based on multi-modal magnetic resonance imaging | |
Suh et al. | Influence of anisotropic conductivity in the skull and white matter on transcranial direct current stimulation via an anatomically realistic finite element head model | |
Campbell et al. | Potential and limitations of diffusion MRI tractography for the study of language | |
CN104323777B (en) | A kind of removing method of diffusion magnetic resonance imaging moving artifact | |
CN105842642B (en) | Based on kurtosis tensor score anisotropic microstructure feature extracting method and device | |
Liu et al. | Improved low-rank filtering of magnetic resonance spectroscopic imaging data corrupted by noise and $ B_0 $ field inhomogeneity | |
CN104267361A (en) | Composite reconstruction method for self-adaptation quantitative magnetisability distribution diagram based on structural feature | |
Nowinski et al. | Three-dimensional interactive and stereotactic human brain atlas of white matter tracts | |
Johnson et al. | Huntington's disease: brain imaging in Huntington's disease | |
CN108898135A (en) | A kind of cerebral limbic system's map construction method | |
Kamiya | Deep learning technique for musculoskeletal analysis | |
CN111612866A (en) | Quantitative magnetic susceptibility imaging reconstruction method and system, storage medium and terminal | |
CN107146235B (en) | A kind of ambient field minimizing technology of the magnetic resonance phase figure based on adaptive convolution kernel | |
CN105488757B (en) | A kind of method of the sparse reconstruction of brain fiber | |
Karimi et al. | A machine learning-based method for estimating the number and orientations of major fascicles in diffusion-weighted magnetic resonance imaging | |
CN105303577B (en) | A kind of method of brain white matter integrity imaging |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
TR01 | Transfer of patent right |
Effective date of registration: 20211227 Address after: 310053 2301a, building B, 4760 Jiangnan Avenue, Puyan street, Binjiang District, Hangzhou, Zhejiang Patentee after: Yuenaoyunfu medical information technology (Zhejiang) Co.,Ltd. Address before: 310014 No. 18 Chao Wang Road, Xiacheng District, Zhejiang, Hangzhou Patentee before: ZHEJIANG University OF TECHNOLOGY |
|
TR01 | Transfer of patent right |