CN107146228A - A kind of super voxel generation method of brain magnetic resonance image based on priori - Google Patents
A kind of super voxel generation method of brain magnetic resonance image based on priori Download PDFInfo
- Publication number
- CN107146228A CN107146228A CN201710175040.3A CN201710175040A CN107146228A CN 107146228 A CN107146228 A CN 107146228A CN 201710175040 A CN201710175040 A CN 201710175040A CN 107146228 A CN107146228 A CN 107146228A
- Authority
- CN
- China
- Prior art keywords
- mrow
- msub
- msup
- image
- mroot
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20112—Image segmentation details
- G06T2207/20156—Automatic seed setting
-
- 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
Abstract
The invention discloses a kind of super voxel generation method of brain magnetic resonance image based on priori, based on K means clustering algorithms, utilization space distance, the weighting of image pixel intensities and priori are used as final distance metric, image pixel is clustered, brain MRI image is divided into a series of uniform and image border of preferably fitting super voxel.Priori of the invention by incorporating brain different tissues, design one kind is new to estimate operator, builds a kind of super voxel generation method of robust, realizes the super voxel segmentation to brain magnetic resonance image, can reduce influence of the picture noise to segmentation result.Compared with existing super voxel generation method, the inventive method is more efficient, and border compactness is higher, can preferably suppress the influence of noise.
Description
Technical field
The present invention relates to digital image processing techniques field, it is related to the processing method of brain magnetic resonance image, specifically
Say, be to be related to a kind of super voxel generation method of brain magnetic resonance image based on priori.
Background technology
Super-pixel (superpixel) or super voxel (supervoxel) are that a kind of image fast-developing in recent years is located in advance
Reason technology, it refers to subregion that is local, consistent, can keeping certain Local Structure of Image feature in image,
Compared to the elementary cell in traditional treatment method --- pixel, super-pixel is more beneficial for extraction and the structural information of local feature
Expression, and the computation complexity of subsequent treatment can be greatly lowered, in computer vision field especially image segmentation
It is widely applied.In addition, super-pixel is also introduced into target following, human body attitude estimation, target identification, significance analysis etc.
Other computer vision research fields.
2003, Ren et al. proposed the concept of super-pixel first, and was applied in image segmentation.By more than ten
The development in year, many classical super-pixel generating algorithms are successively suggested.Current existing super-pixel generating algorithm is roughly divided into
Two classes：A kind of is the algorithm based on graph theory, and its basic thought is to regard the pixel in image as node of graph, and is assigned between node
Side with weights, then figure is divided using various segmentation criterions, so as to form super-pixel (as Shi et al. is proposed
Normalized cuts algorithms)；Another is the algorithm risen based on gradient, and its basic thought is from initial pixel cluster
Start, using gradient method iterated revision cluster result until meeting the condition of convergence, so as to form super-pixel (such as Achanta et al.
The SLIC algorithms of proposition).
But, for different types of image, current super-pixel generating algorithm it is universal super-pixel quantity, tight ness rating with
Split between quality, algorithm practicality the problem of there is mutually restriction.Meanwhile, special objective is also difficult to obtain preferable segmentation
As a result.Especially for MRI noises are big, weak boundary the characteristics of, above-mentioned algorithm is not very applicable, especially when image is by noise
Interference than it is more serious when, cannot get preferable super-pixel generation effect.
The content of the invention
To solve the above problems, the invention discloses a kind of super voxel generation of brain magnetic resonance image based on priori
Method, based on K-means clustering algorithms, the weighting of utilization space distance, image pixel intensities and priori is right as distance metric
Image pixel is clustered, and brain magnetic resonance image is divided into a series of uniform and image border of preferably fitting super body
Element.
In order to achieve the above object, the present invention provides following technical scheme：
A kind of super voxel generation method of brain magnetic resonance image based on priori, comprises the following steps：
Step 1, probability collection of illustrative plates is registrated to individual space
The spatial alternation that template image is deformed into target image is obtained using registration, will be with template using the spatial alternation
The corresponding known results of image, i.e. profile information, are mapped to target image, and probability collection of illustrative plates is registrated into individual space, so that
Obtain on individual space, the probability distribution of different voxels；
Step 2, the initialization of seed point, specifically includes following steps：
Step 2-1, it is assumed that image one has N number of voxel, the super voxel number for expecting generation is M, then each super voxel
Size is：
V=N/M；
Step 2-2, wherein M also illustrate that the number of initial seed point, and the length of side of each super voxel is：
In above formula, L is iteratively step-length each time；
Step 2-3, calculates being averaged for each seed point and the gray value of the area pixel of surrounding 3 × 3 × 3, as with this
The gray value I of seed point, calculation formula is as follows：
Wherein (xi, yi, zi) seed point i coordinate, I are represented respectivelyxyzIt is the pixel value under (x, y, z) coordinate；
Step 3, Weighted distance is calculated as between voxel and seed point estimate operator；
Step 3-1, if Weighted distance is D, D includes three parts：Image pixel intensities dI, space length dSWith priori dA,
Use D0The distance of these three parts is represented, specific formula is as follows：
In above formula, dI、dSCalculation formula it is as follows：
In above formula, IjAnd IiThe image pixel intensities at seed point j and at tissue points i are represented respectively,
In above formula, (xj, yj, zj) and (xi, yi, zi) respectively represent seed point j and tissue points i coordinate；
dAValue obtained by image registration；
Step 3-2, by dI, dSAnd dADifference divided by a coefficient WI, WAAnd WAThe Weighted distance D ' corrected：
Wherein,N is the voxel sum of image, and M is desired super voxel number, coefficient WI
And WARepresented respectively with constant m and λ, D ' is expressed as follows：
Step 3-3, last Weighted distance D is obtained after D ' is simplified：
Step 4, voxel cluster is generated by super voxel to each seed point using improved k-means clustering methods, specifically
Comprise the following steps：
Step 4-1, chooses the M tissue points for being located at region 2L × 2L × 2L centers, then each point exists in the picture
The minimum place of Grad is moved in 3 × 3 × 3 region centered on using it and is used as initial seed point；
In step 4-2, the formula zoning obtained using step 3-3 each tissue points to seed point Weighted distance D,
And tissue points are grouped into closest seed point that class；
Step 4-3, once all tissue points are all classified into that seed point nearest from it, then calculates each again
The cluster centre of class is used as such new seed point；
Step 4-4, the 4-2 that repeats the above steps, step 4-3 processes are until the cluster centre that is newly generated and between the last time
Error no longer change or less than some threshold value, algorithm stops.
Further, in the step 1 template image include three probability distribution standard form CFS cerebrospinal fluid image,
GM grey matters image, WM white matter images.
Further, it is that each seed point distributes a single label in the step 2-2.
Further, in the step 3-3, simplify process D ' in step 3-2 is multiplied by into coefficient lambda m.
Compared with prior art, the invention has the advantages that and beneficial effect：
Priori of the invention by incorporating brain different tissues, design one kind is new to estimate operator, builds a kind of Shandong
The super voxel generation method of rod, realizes and the super voxel of brain magnetic resonance image is split, and can reduce picture noise and segmentation is tied
The influence of fruit.Compared with existing super voxel generation method, the inventive method is more efficient, and border compactness is higher, can be preferably
Ground suppresses the influence of noise.
Brief description of the drawings
The super voxel generation method steps flow chart of the brain magnetic resonance image based on priori that Fig. 1 provides for the present invention
Figure.
Fig. 2 is the closest interpolation schematic diagram of two dimensional image gray scale.
Fig. 3 (a) is the point schematic diagram in routine k-means algorithms calculating view picture figure, and Fig. 3 (b) is that this method calculates part
The point schematic diagram in region.
Fig. 4 (a) is the big Typical AVM axial view of Noise, and Fig. 4 (b) is that this method is based on the super voxel that Fig. 4 (a) is obtained
Image.
Fig. 5 (a) is the big Typical AVM coronal-plane figure of Noise, and Fig. 5 (b) is that this method is based on the super voxel that Fig. 5 (a) is obtained
Image.
Fig. 6 (a) is the big Typical AVM sagittal view of Noise, and Fig. 6 (b) is that this method is based on the super voxel that Fig. 6 (a) is obtained
Image.
Fig. 7 is that this method is compared figure with the other method time.
Fig. 8 (a) is big Typical AVM axial view, and Fig. 8 (b) is big Typical AVM coronal-plane figure, and Fig. 8 (c) is that big Typical AVM is axially regarded
Scheme, Fig. 8 (d) is that this method is based on the super voxel image that Fig. 8 (a) is obtained, and Fig. 8 (e) is this method based on surpassing that Fig. 8 (b) is obtained
Voxel image, Fig. 8 (f) is that this method is based on the super voxel image that Fig. 8 (c) is obtained,
Embodiment
The technical scheme provided below with reference to specific embodiment the present invention is described in detail, it should be understood that following specific
Embodiment is only illustrative of the invention and is not intended to limit the scope of the invention.
The super voxel generation method of a kind of brain magnetic resonance image based on priori that the present invention is provided, by three tissues
Probability distribution collection of illustrative plates be registrated to individual template, after selected seed point, fusion priori design is a kind of new to estimate calculation
Son, super voxel is generated by voxel cluster to seed point.Idiographic flow of the present invention is as shown in figure 1, comprise the following steps：
Step 1, probability collection of illustrative plates is registrated to individual space.
The spatial alternation that template image is deformed into target image is obtained using registration, then will be with using this spatial alternation
The corresponding known results of template image, i.e. profile information, are mapped to target image.The present invention is using by three probability distribution
The method for registering that standard form CFS (cerebrospinal fluid image), GM (grey matter image), WM (white matter image) are registrated to individual template is used
Method (Klein S, Staring M, Murphy K, Viergever MA, Pluim JP.elastix:a toolbox for
intensity-based medical image registration.IEEE transactions on medical
imaging.2010；29(1):Method for registering disclosed in 196-205.) completes probability collection of illustrative plates and is registrated to individual space, so that
Onto individual space, the probability distribution of different voxels.
Step 2, the initialization of seed point.Specifically include following steps：
Step 2-1, it is assumed that image one has N number of voxel, the super voxel number for expecting generation is M.Then each super voxel
Size is
V=N/M
Step 2-2, wherein M also illustrate that the number of initial seed point (i.e. central point).Each the length of side of super voxel is
In above formula, L is iteratively step-length each time, is that each seed point distributes a single label.
Step 2-3, meanwhile, in order to reduce the influence of noise, calculate the area pixel of each seed point and surrounding 3 × 3 × 3
Gray value be averaged, as the gray value I with the seed point, calculation formula is as follows：
Wherein (xi, yi, zi) seed point i coordinate, I are represented respectivelyxyzIt is the pixel value under (x, y, z) coordinate.
Step 3, Weighted distance is calculated as between voxel and seed point estimate operator.
Step 3-1, to each tissue points i, is calculated between cluster centre j (i.e. seed point) closest therewith respectively
Similarity degree, the tissue points are assigned to by the label of most like seed point.By the continuous iteration process, until convergence.It is therein
" similarity degree " is measured by Weighted distance.If Weighted distance is D, three parts are included：Image pixel intensities, space length and
Priori.The distance of these three parts can be represented simply as, D is used0To represent, specific formula is as follows：
Wherein：dI：Image pixel intensities, dS：Space length, dA：Priori.This three part will produce difference to cluster result
Influence.Work as dSWeight it is larger when, the super voxel of generation is more compact, but the compactness on border will reduce.Similarly, when
dIWeight it is larger when, super voxel can fit the border district of image well, while also to become comparison irregular for super voxel.
dAThen control picture noise generates the influence of result to super voxel.
dI、dSCalculation formula it is as follows：
In above formula, IjAnd IiThe image pixel intensities at seed point j and at tissue points i are represented respectively.
In above formula, (xj, yj, zj) and (xi, yi, zi) respectively represent seed point j and tissue points i coordinate.
dAImage registration of the value then in step 1 obtain.
Step 3-2, by dI, dSAnd dADifference divided by a coefficient WI, WSAnd WAThe Weighted distance D corrected, with D ' come
Represent：
Wherein,N is the voxel sum of image, and M is desired super voxel number.For not
Same image, dIAnd dACalculating be very different.In this example, this two-part coefficient WIAnd WARespectively with constant m and λ come table
Show.Then D ' can be expressed as follows：
Step 3-3, above formula is simplified and (is multiplied by and last Weighted distance D is obtained after coefficient lambda m)：
Step 4, voxel cluster is generated into super voxel to each seed point.Specifically include following steps：
Step 4-1, chooses the M tissue points positioned at region 2L × 2L × 2L centers and is used as seed point in the picture；
In step 4-2, the formula zoning obtained using step 3-3 each tissue points to seed point Weighted distance D,
And tissue points are grouped into " distance " nearest seed point that class.
Traditional K-means algorithms are to calculate the distance for arriving each seed point in entire image a little, and this algorithm is
Calculate all tissue points in one piece of regional area and, to the distance of seed point, shown in such as Fig. 3 (b), considerably reduce number of computations,
So as to substantially increase the efficiency of algorithm..Because the size of each super voxel is L × L × L, therefore regional area is set to
Size is 2L × 2L × 2L centered on seed point.
Step 4-3, recalculates the barycenter of each obtained class.Once all tissue points are all classified into from it
That nearest seed point, the then cluster centre for calculating each class again is used as such new seed point.
Step 4-4, the 4-2 that repeats the above steps, step 4-3 processes are until the cluster centre that is newly generated and between the last time
Error no longer change or less than some threshold value, algorithm stops.
Using super voxel image such as Fig. 4 (b) institutes of big Typical AVM axial view generation of the inventive method based on Noise
Show, shown in super voxel image such as Fig. 5 (b) of the big Typical AVM coronal-plane figure generation based on Noise, the brain based on Noise
Shown in super voxel image such as Fig. 6 (b) of MRI sagittal views generation.It can be seen that, for containing noisy image, what the present invention was obtained
Super voxel image boundary compactness is high, can preferably suppress the influence of noise.The super voxel life for the robust that the present invention is built
Into method, influence of the picture noise to segmentation result can be reduced.
And for muting image, we compare the present invention with two kinds of existing methods, Fig. 7 for the present invention and
Existing two methods generate the time that the super voxel of different numbers is spent, it can be seen that the efficiency highest of this algorithm.Fig. 8 is this
Invent obtained super voxel image, it is seen that border compactness is very high.
Technological means disclosed in the present invention program is not limited only to the technological means disclosed in above-mentioned embodiment, in addition to
Constituted technical scheme is combined by above technical characteristic.It should be pointed out that for those skilled in the art
For, under the premise without departing from the principles of the invention, some improvements and modifications can also be made, these improvements and modifications are also considered as
Protection scope of the present invention.
Claims (4)
1. the super voxel generation method of a kind of brain magnetic resonance image based on priori, it is characterised in that comprise the following steps：
Step 1, probability collection of illustrative plates is registrated to individual space
The spatial alternation that template image is deformed into target image is obtained using registration, will be with template image using the spatial alternation
Corresponding known results, i.e. profile information, are mapped to target image, and probability collection of illustrative plates is registrated into individual space, so as to obtain
On individual space, the probability distribution of different voxels；
Step 2, the initialization of seed point, specifically includes following steps：
Step 2-1, it is assumed that image one has N number of voxel, the super voxel number for expecting generation is M, then the size of each super voxel
For：
V=N/M；
Step 2-2, wherein M also illustrate that the number of initial seed point, and the length of side of each super voxel is：
<mrow>
<mi>L</mi>
<mo>=</mo>
<mroot>
<mi>V</mi>
<mn>3</mn>
</mroot>
<mo>=</mo>
<mroot>
<mrow>
<mi>N</mi>
<mo>/</mo>
<mi>M</mi>
</mrow>
<mn>3</mn>
</mroot>
</mrow>
In above formula, L is iteratively step-length each time；
Step 2-3, calculates being averaged for each seed point and the gray value of the area pixel of surrounding 3 × 3 × 3, as with the seed
The gray value I of point, calculation formula is as follows：
<mrow>
<mi>I</mi>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<mn>27</mn>
</mfrac>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>x</mi>
<mo>=</mo>
<msub>
<mi>x</mi>
<mi>i</mi>
</msub>
<mo>-</mo>
<mn>1</mn>
</mrow>
<mrow>
<msub>
<mi>x</mi>
<mi>i</mi>
</msub>
<mo>+</mo>
<mn>1</mn>
</mrow>
</munderover>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>y</mi>
<mo>=</mo>
<msub>
<mi>y</mi>
<mi>i</mi>
</msub>
<mo>-</mo>
<mn>1</mn>
</mrow>
<mrow>
<msub>
<mi>y</mi>
<mi>i</mi>
</msub>
<mo>+</mo>
<mn>1</mn>
</mrow>
</munderover>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>z</mi>
<mo>=</mo>
<msub>
<mi>z</mi>
<mi>i</mi>
</msub>
<mo>-</mo>
<mn>1</mn>
</mrow>
<mrow>
<msub>
<mi>z</mi>
<mi>i</mi>
</msub>
<mo>+</mo>
<mn>1</mn>
</mrow>
</munderover>
<msub>
<mi>I</mi>
<mrow>
<mi>x</mi>
<mi>y</mi>
<mi>z</mi>
</mrow>
</msub>
</mrow>
Wherein (xi, yi, zi) seed point i coordinate, I are represented respectivelyxyzIt is the pixel value under (x, y, z) coordinate；
Step 3, Weighted distance is calculated as between voxel and seed point estimate operator；
Step 3-1, if Weighted distance is D, D includes three parts：Image pixel intensities dI, space length dSWith priori dA, use D0
The distance of these three parts is represented, specific formula is as follows：
<mrow>
<msub>
<mi>D</mi>
<mn>0</mn>
</msub>
<mo>=</mo>
<mroot>
<mrow>
<msup>
<msub>
<mi>d</mi>
<mi>I</mi>
</msub>
<mn>2</mn>
</msup>
<mo>+</mo>
<msup>
<msub>
<mi>d</mi>
<mi>S</mi>
</msub>
<mn>2</mn>
</msup>
<mo>+</mo>
<msup>
<msub>
<mi>d</mi>
<mi>A</mi>
</msub>
<mn>2</mn>
</msup>
</mrow>
<mn>2</mn>
</mroot>
</mrow>
In above formula, dI、dSCalculation formula it is as follows：
<mrow>
<msub>
<mi>d</mi>
<mi>I</mi>
</msub>
<mo>=</mo>
<mroot>
<mrow>
<msup>
<msub>
<mi>I</mi>
<mi>j</mi>
</msub>
<mn>2</mn>
</msup>
<mo>-</mo>
<msup>
<msub>
<mi>I</mi>
<mi>i</mi>
</msub>
<mn>2</mn>
</msup>
</mrow>
<mn>2</mn>
</mroot>
</mrow>
In above formula, IjAnd IiThe image pixel intensities at seed point j and at tissue points i are represented respectively,
<mrow>
<msub>
<mi>d</mi>
<mi>S</mi>
</msub>
<mo>=</mo>
<mroot>
<mrow>
<msup>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<mi>j</mi>
</msub>
<mo>-</mo>
<msub>
<mi>x</mi>
<mi>i</mi>
</msub>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<mo>+</mo>
<msup>
<mrow>
<mo>(</mo>
<msub>
<mi>y</mi>
<mi>j</mi>
</msub>
<mo>-</mo>
<msub>
<mi>y</mi>
<mi>i</mi>
</msub>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<mo>+</mo>
<msup>
<mrow>
<mo>(</mo>
<msub>
<mi>z</mi>
<mi>j</mi>
</msub>
<mo>-</mo>
<msub>
<mi>z</mi>
<mi>i</mi>
</msub>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
</mrow>
<mn>2</mn>
</mroot>
</mrow>
In above formula, (xj, yj, zj) and (xi, yi, zi) respectively represent seed point j and tissue points i coordinate；
dAValue obtained by image registration；
Step 3-2, by dI, dSAnd dADifference divided by a coefficient WI, WSAnd WAThe Weighted distance D ' corrected：
<mrow>
<msup>
<mi>D</mi>
<mo>&prime;</mo>
</msup>
<mo>=</mo>
<mroot>
<mrow>
<msup>
<mrow>
<mo>(</mo>
<mfrac>
<msub>
<mi>d</mi>
<mi>I</mi>
</msub>
<msub>
<mi>W</mi>
<mi>I</mi>
</msub>
</mfrac>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<mo>+</mo>
<msup>
<mrow>
<mo>(</mo>
<mfrac>
<msub>
<mi>d</mi>
<mi>S</mi>
</msub>
<msub>
<mi>W</mi>
<mi>S</mi>
</msub>
</mfrac>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<mo>+</mo>
<msup>
<mrow>
<mo>(</mo>
<mfrac>
<msub>
<mi>d</mi>
<mi>A</mi>
</msub>
<msub>
<mi>W</mi>
<mi>A</mi>
</msub>
</mfrac>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
</mrow>
<mn>2</mn>
</mroot>
</mrow>
Wherein,N is the voxel sum of image, and M is desired super voxel number, coefficient WIAnd WA
Represented respectively with constant m and λ, D ' is expressed as follows：
<mrow>
<msup>
<mi>D</mi>
<mo>&prime;</mo>
</msup>
<mo>=</mo>
<mroot>
<mrow>
<msup>
<mrow>
<mo>(</mo>
<mfrac>
<msub>
<mi>d</mi>
<mi>I</mi>
</msub>
<mi>m</mi>
</mfrac>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<mo>+</mo>
<msup>
<mrow>
<mo>(</mo>
<mfrac>
<msub>
<mi>d</mi>
<mi>S</mi>
</msub>
<mi>L</mi>
</mfrac>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<mo>+</mo>
<msup>
<mrow>
<mo>(</mo>
<mfrac>
<msub>
<mi>d</mi>
<mi>A</mi>
</msub>
<mi>&lambda;</mi>
</mfrac>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
</mrow>
<mn>2</mn>
</mroot>
</mrow>
Step 3-3, last Weighted distance D is obtained after D ' is simplified：
<mrow>
<mi>D</mi>
<mo>=</mo>
<mroot>
<mrow>
<msup>
<mi>&lambda;</mi>
<mn>2</mn>
</msup>
<msup>
<mrow>
<mo>(</mo>
<msub>
<mi>d</mi>
<mi>I</mi>
</msub>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<mo>+</mo>
<msup>
<mi>&lambda;</mi>
<mn>2</mn>
</msup>
<msup>
<mi>m</mi>
<mn>2</mn>
</msup>
<msup>
<mrow>
<mo>(</mo>
<mfrac>
<msub>
<mi>d</mi>
<mi>S</mi>
</msub>
<mi>L</mi>
</mfrac>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<mo>+</mo>
<msup>
<mi>m</mi>
<mn>2</mn>
</msup>
<msup>
<mrow>
<mo>(</mo>
<msub>
<mi>d</mi>
<mi>A</mi>
</msub>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
</mrow>
<mn>2</mn>
</mroot>
</mrow>
Step 4, voxel cluster is generated to each seed point by super voxel using improved k-means clustering methods, specifically included
Following steps：
Step 4-1, chooses the M tissue points for being located at region 2L × 2L × 2L centers, then by each point with it in the picture
Centered on 3 × 3 × 3 region in be moved to the minimum place of Grad as initial seed point；
Each tissue points is to the Weighted distance D of seed point, and handle in step 4-2, the formula zoning obtained using step 3-3
Tissue points are grouped into closest seed point that class；
Step 4-3, once all tissue points are all classified into that seed point nearest from it, then calculates each class again
Cluster centre is used as such new seed point；
Step 4-4, the 4-2 that repeats the above steps, step 4-3 processes are until the cluster centre being newly generated and the mistake between the last time
Difference no longer changes or less than some threshold value, and algorithm stops.
2. the super voxel generation method of the brain magnetic resonance image according to claim 1 based on priori, its feature exists
In：Standard form CFS cerebrospinal fluid image, GM grey matter image, WM of the template image including three probability distribution are white in the step 1
Matter image.
3. the super voxel generation method of the brain magnetic resonance image according to claim 1 based on priori, its feature exists
In：It is each seed point distribution label in the step 2-2.
4. the super voxel generation method of the brain magnetic resonance image according to claim 1 based on priori, its feature exists
In：In the step 3-3, simplify process D ' in step 3-2 is multiplied by into coefficient lambda m.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710175040.3A CN107146228B (en) | 2017-03-22 | 2017-03-22 | A kind of super voxel generation method of brain magnetic resonance image based on priori knowledge |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710175040.3A CN107146228B (en) | 2017-03-22 | 2017-03-22 | A kind of super voxel generation method of brain magnetic resonance image based on priori knowledge |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107146228A true CN107146228A (en) | 2017-09-08 |
| CN107146228B CN107146228B (en) | 2019-10-08 |
Family
ID=59783605
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710175040.3A Active CN107146228B (en) | 2017-03-22 | 2017-03-22 | A kind of super voxel generation method of brain magnetic resonance image based on priori knowledge |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107146228B (en) |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108171009A (en) * | 2017-12-21 | 2018-06-15 | 东南大学 | Causality detection method between EEG signals based on the estimation of two-dimentional Parameters of Autoregressive Models |
| CN108596236A (en) * | 2018-04-18 | 2018-09-28 | 东南大学 | It is a kind of to roll into a ball partition method based on the thalamic nuclei of global connection features and geodesic distance |
| CN108921816A (en) * | 2018-05-16 | 2018-11-30 | 东南大学 | A kind of cerebral function partition method based on super voxel and spectral clustering |
| CN109035265A (en) * | 2018-09-26 | 2018-12-18 | 重庆邮电大学 | A kind of novel structure brain connection map construction method |
| CN109285176A (en) * | 2018-10-08 | 2019-01-29 | 东南大学 | A kind of cerebral tissue dividing method cut based on regularization figure |
| CN109472263A (en) * | 2018-10-12 | 2019-03-15 | 东南大学 | A kind of brain magnetic resonance image dividing method of the global and local information of combination |
| CN109886944A (en) * | 2019-02-02 | 2019-06-14 | 浙江大学 | A kind of white matter high signal intensity detection and localization method based on multichannel chromatogram |
| CN110638477A (en) * | 2018-06-26 | 2020-01-03 | 佳能医疗系统株式会社 | Medical image diagnosis apparatus and alignment method |
| CN110728685A (en) * | 2019-09-20 | 2020-01-24 | 东南大学 | Brain tissue segmentation method based on diagonal voxel local binary pattern texture operator |
| CN110751664A (en) * | 2019-09-29 | 2020-02-04 | 东南大学 | Brain tissue segmentation method based on hyper-voxel matching |
| CN110811622A (en) * | 2019-11-12 | 2020-02-21 | 北京大学 | Individual structure connection brain atlas drawing method based on diffusion magnetic resonance imaging fiber bundle tracking technology |
| CN112435261A (en) * | 2020-11-16 | 2021-03-02 | 辽宁师范大学 | Multi-modal MRI brain tumor segmentation method based on hyper-voxels |
| CN113288181A (en) * | 2021-06-21 | 2021-08-24 | 杭州电子科技大学 | Individual template reconstruction method based on steady-state visual evoked potential electroencephalogram signal identification |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103353987A (en) * | 2013-06-14 | 2013-10-16 | 山东大学 | Superpixel segmentation method based on fuzzy theory |
| CN103544697A (en) * | 2013-09-30 | 2014-01-29 | 南京信息工程大学 | Hypergraph spectrum analysis based image segmentation method |
| US20150138191A1 (en) * | 2013-11-19 | 2015-05-21 | Thomson Licensing | Method and apparatus for generating superpixels |
| CN104809723A (en) * | 2015-04-13 | 2015-07-29 | 北京工业大学 | Three-dimensional liver CT (computed tomography) image automatically segmenting method based on hyper voxels and graph cut algorithm |
-
2017
- 2017-03-22 CN CN201710175040.3A patent/CN107146228B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103353987A (en) * | 2013-06-14 | 2013-10-16 | 山东大学 | Superpixel segmentation method based on fuzzy theory |
| CN103544697A (en) * | 2013-09-30 | 2014-01-29 | 南京信息工程大学 | Hypergraph spectrum analysis based image segmentation method |
| US20150138191A1 (en) * | 2013-11-19 | 2015-05-21 | Thomson Licensing | Method and apparatus for generating superpixels |
| CN104809723A (en) * | 2015-04-13 | 2015-07-29 | 北京工业大学 | Three-dimensional liver CT (computed tomography) image automatically segmenting method based on hyper voxels and graph cut algorithm |
Cited By (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108171009A (en) * | 2017-12-21 | 2018-06-15 | 东南大学 | Causality detection method between EEG signals based on the estimation of two-dimentional Parameters of Autoregressive Models |
| CN108596236A (en) * | 2018-04-18 | 2018-09-28 | 东南大学 | It is a kind of to roll into a ball partition method based on the thalamic nuclei of global connection features and geodesic distance |
| CN108921816A (en) * | 2018-05-16 | 2018-11-30 | 东南大学 | A kind of cerebral function partition method based on super voxel and spectral clustering |
| CN110638477A (en) * | 2018-06-26 | 2020-01-03 | 佳能医疗系统株式会社 | Medical image diagnosis apparatus and alignment method |
| CN110638477B (en) * | 2018-06-26 | 2023-08-11 | 佳能医疗系统株式会社 | Medical image diagnosis device and alignment method |
| CN109035265A (en) * | 2018-09-26 | 2018-12-18 | 重庆邮电大学 | A kind of novel structure brain connection map construction method |
| CN109285176A (en) * | 2018-10-08 | 2019-01-29 | 东南大学 | A kind of cerebral tissue dividing method cut based on regularization figure |
| CN109285176B (en) * | 2018-10-08 | 2021-11-02 | 东南大学 | Brain tissue segmentation method based on regularization graph segmentation |
| CN109472263A (en) * | 2018-10-12 | 2019-03-15 | 东南大学 | A kind of brain magnetic resonance image dividing method of the global and local information of combination |
| CN109472263B (en) * | 2018-10-12 | 2021-06-15 | 东南大学 | Global and local information combined brain magnetic resonance image segmentation method |
| CN109886944B (en) * | 2019-02-02 | 2019-11-22 | 浙江大学 | A kind of white matter high signal intensity detection and localization method based on multichannel chromatogram |
| CN109886944A (en) * | 2019-02-02 | 2019-06-14 | 浙江大学 | A kind of white matter high signal intensity detection and localization method based on multichannel chromatogram |
| CN110728685A (en) * | 2019-09-20 | 2020-01-24 | 东南大学 | Brain tissue segmentation method based on diagonal voxel local binary pattern texture operator |
| CN110728685B (en) * | 2019-09-20 | 2023-04-07 | 东南大学 | Brain tissue segmentation method based on diagonal voxel local binary pattern texture operator |
| CN110751664A (en) * | 2019-09-29 | 2020-02-04 | 东南大学 | Brain tissue segmentation method based on hyper-voxel matching |
| CN110751664B (en) * | 2019-09-29 | 2022-11-18 | 东南大学 | Brain tissue segmentation method based on hyper-voxel matching |
| CN110811622A (en) * | 2019-11-12 | 2020-02-21 | 北京大学 | Individual structure connection brain atlas drawing method based on diffusion magnetic resonance imaging fiber bundle tracking technology |
| CN112435261A (en) * | 2020-11-16 | 2021-03-02 | 辽宁师范大学 | Multi-modal MRI brain tumor segmentation method based on hyper-voxels |
| CN112435261B (en) * | 2020-11-16 | 2023-06-16 | 辽宁师范大学 | Multi-mode MRI brain tumor segmentation method based on super-voxels |
| CN113288181A (en) * | 2021-06-21 | 2021-08-24 | 杭州电子科技大学 | Individual template reconstruction method based on steady-state visual evoked potential electroencephalogram signal identification |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107146228B (en) | 2019-10-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107146228A (en) | A kind of super voxel generation method of brain magnetic resonance image based on priori | |
| CN108776969B (en) | Breast ultrasound image tumor segmentation method based on full convolution network | |
| CN108734659B (en) | Sub-pixel convolution image super-resolution reconstruction method based on multi-scale label | |
| CN111798462B (en) | Automatic delineation method of nasopharyngeal carcinoma radiotherapy target area based on CT image | |
| Chakraborty et al. | Deformable boundary finding influenced by region homogeneity | |
| CN110992262A (en) | Remote sensing image super-resolution reconstruction method based on generation countermeasure network | |
| Zhang et al. | LU-NET: An improved U-Net for ventricular segmentation | |
| CN109903319B (en) | Multi-resolution-based fast iteration closest point registration algorithm | |
| Aranguren et al. | Improving the segmentation of magnetic resonance brain images using the LSHADE optimization algorithm | |
| CN109767459B (en) | Novel fundus image registration method | |
| CN108876769B (en) | Left auricle CT image segmentation method | |
| CN113570627B (en) | Training method of deep learning segmentation network and medical image segmentation method | |
| CN107845065A (en) | Super-resolution image reconstruction method and device | |
| CN110570394A (en) | medical image segmentation method, device, equipment and storage medium | |
| CN109285176B (en) | Brain tissue segmentation method based on regularization graph segmentation | |
| CN108765287A (en) | A kind of image super-resolution method based on non-local mean | |
| Rashid et al. | Single MR image super-resolution using generative adversarial network | |
| Lohit et al. | Modified total Bregman divergence driven picture fuzzy clustering with local information for brain MRI image segmentation | |
| Zhang et al. | A novel medical image registration method based on mutual information and genetic algorithm | |
| CN113838104B (en) | Registration method based on multispectral and multimodal image consistency enhancement network | |
| Ahmed et al. | 3D U-Net: fully convolutional neural network for automatic brain tumor segmentation | |
| CN115359140A (en) | Automatic multi-feature region-of-interest delineation system and method based on neural network | |
| CN104240291B (en) | A kind of image segmentation method for reconstructing and system based on nuclear magnetic resonance image sequence | |
| Zheng et al. | SurfNN: Joint reconstruction of multiple cortical surfaces from magnetic resonance images | |
| Zyuzin et al. | Generation of echocardiographic 2D images of the heart using cGAN |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |