CN103310449A - Lung segmentation method based on improved shape model - Google Patents
Lung segmentation method based on improved shape model Download PDFInfo
- Publication number
- CN103310449A CN103310449A CN2013102376861A CN201310237686A CN103310449A CN 103310449 A CN103310449 A CN 103310449A CN 2013102376861 A CN2013102376861 A CN 2013102376861A CN 201310237686 A CN201310237686 A CN 201310237686A CN 103310449 A CN103310449 A CN 103310449A
- Authority
- CN
- China
- Prior art keywords
- point
- shape
- lung
- gray scale
- image
- 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
Images
Landscapes
- Apparatus For Radiation Diagnosis (AREA)
Abstract
Disclosed is a lung segmentation method based on an improved shape model. The method includes building a prior model of the lung contour, and then segmenting a lung region by utilizing grayscale and shape similarity information and combining image characteristics. In some images, initial positions are probably too far away from an actual boundary, so that when the grayscale and shape similarity information is utilized for segmentation, a search region does not cover the lung boundary. Therefore, an ASM (active shape model) algorithm is used for modifying the lung boundary on the basis of first segmentation, the situation that partial point search falls into local extremum is improved, and better search results are obtained.
Description
Technical field: the present invention relates to a kind of lung dividing method, especially a kind of lung dividing method based on improving shape.
Background technology: lung cancer is now to one of malignant tumour of human health risk maximum.Particularly since nearly half a century, the environment that causes along with air pollution constantly worsens and the rolling up of the smoking size of population, and the incidence of disease of various countries' lung cancer and case fatality rate are all in rapid rising.The death number of the whole America lung cancer in 2012 is 160,259, occupies the first place of all cancer mortalities, is higher than the mortality ratio of other cancer far away.China's lung cancer death case load was 493,348 in 2008, compared with 30.83 ten thousand of 2004-2005, and mortality ratio improves nearly 60%.The Cancer in China prevention determines that with control planning outline (2004~2010) lung cancer is the emphasis of China's prevention and control of cancer.
Because lung belongs to the inside of human body internal organs, most lung cancer are at first just silently growth in health, and the patient is without any sensation.When the patient went to a doctor because of clinical symptoms such as cough, spitting of blood and pectoralgias, great majority had been in middle and advanced stage, had missed the best opportunity for the treatment of.Studies show that, the early diagnosis of lung cancer can reduce the mortality ratio of patients with lung cancer, and the survival rate in 10 years is higher than 90% behind the early stage of lung cancer operation in patients.So the early diagnosis of lung cancer is the key that improves the patients with lung cancer survival rate with treatment.
In recent years, digitized x-ray photography (Digital Radiography) makes traditional X-ray photographic technology enter digital area.It is compared with traditional X-ray photographic, has the higher quality of image, comprises the characteristics such as more image information.Simultaneously, because that the DR image has an imaging device is simple, cost is low, compares the characteristics such as dose radiation is low with CT, so be the main imaging mode of screening lung cancer.Because accurately cutting apart of lung zone is that the lung tubercle automatically detects the basis with auxiliary diagnosis, it therefore is one of focus of studying of scholars.The people such as Xu propose the lung Region Segmentation Algorithm based on image characteristic analysis.Determine lung top and border, thoracic cavity by the second derivative of profile first, determine the mediastinum right margin according to boundary gradient again, determine starting point search mediastinum left margin by rule, this algorithm is realized the lung zone is cut apart roughly, and precision is lower.The people such as Ginneken use active shape model, and active appearance models reaches multiresolution pixel sorting algorithm the lung zone is cut apart, and is used alternatingly shape and half-tone information, easily is absorbed in local extremum.The people such as Shi propose the lung Region Segmentation Algorithm based on the deformation model of general population and the peculiar lung of patient shape Statistics information.The people such as Soleymanpour use first the nonlinear filter based on the equilibrium of self-adaptation contrast that original image is strengthened, and the recycling region growing algorithm obtains initial lung zone to morphological operation and repairs the more accurate segmentation result of acquisition.The people such as Liu Yan have proposed the lung partitioning algorithm based on flexible morphology and clustering algorithm, and the method utilizes flexible erosion operator to the dilating effect of dark picture areas, cut apart but segmentation performance too relies on the first time.The people such as Shi Zhenghao improve Fuzzy C-Means Cluster Algorithm, adopt based on what examine to cut apart the lung zone apart from the alternative European cluster of normal form.
Summary of the invention: for above-mentioned the deficiencies in the prior art, the invention provides a kind of lung dividing method based on improving shape.
For achieving the above object, the technical solution used in the present invention is: based on the lung dividing method that improves shape, concrete steps are:
One, determining of model initial profile position: comprise the mark training set, the alignment training set, prior model set up three major parts.
1. mark training set: use along the point on border and come the mark training set, point comprises following three classes: a, target-marking have the point of application-specific part.For example, in the faceform, the point of expression eye center.Sharper keen turning on the expression border is such as the point at canthus; The point of b, the irrelevant applying portion of mark, such as the peak of target on specific direction, or the extreme value place of curvature; C, fill out the point between a class point or 2 class points.
2. training set aligns: in order to compare the identical point on the difformity, they must be about one group of coordinate axis alignment.Operation by convergent-divergent, rotation and translation makes the training shapes alignment, and they are alignd closely as far as possible.If x
iThe vector of n point in i shape in the training set, x
i=(x
I0, y
I0, x
I1, y
I1..., x
Ik, y
Ik..., x
In-1, y
In-1)
TWherein, (x
Ij, y
Ij) be j point in i the shape.
Given two similar shape x
iAnd x
j, select anglec of rotation θ, convergent-divergent s, translation (t
x, t
y), then to represent the anglec of rotation be that θ and scaling are the conversion of s to M (s, θ) [x], x
iBe mapped as M (s, θ) [x
j]+t minimizes following weighted sum:
E
j=(x
i-M(s,θ)[x
j]-t)
TW(x
i-M(s,θ)[x
j]-t) (1)
Wherein,
3. the foundation of prior model: after the shape vector registration process in the training set, just can utilize the method for principal component analysis (PCA) to find out statistical information and the rule of change of shape, do like this efficient that can improve algorithm.
If average shape is
Each sample with respect to the shape vector covariance of the deviation formation of average shape is after the alignment
Calculate the eigen vector of this covariance, and with the eigenwert Sp that sorts
k=λ
kp
kWherein, λ
kRepresent the eigenwert that k is large.
λ
kLarger, its corresponding p
kData of description point changing pattern is just more important.T important changing pattern forms new main shaft system p before selecting
s, then allow any one shape in the shape territory to be similar to by the weighted sum that average shape adds main shaft system and one group control parameter,
Wherein, p
s=(p
1p
2... p
t) be the matrix that front t proper vector forms, b
s=(b
1b
2... b
t)
TIt is weight vector.
Through principal component analysis, get a front t eigenwert and characteristic of correspondence vector according to descending, the target object deformation that a front t eigenwert is determined accounts for all 2n eigenwert and determines that the ratio of target object deformation total amount is not less than V (general V gets 0.98).
Two, the pulmonary parenchyma of intensity-based information and shape information is cut apart: the present invention utilizes gray scale and the shape information of image simultaneously, so that the border gray scale that searches, shape information are similar to training image.
(1) characteristic image: owing to can more give prominence to the variation of gray scale based on the characteristic image of derivative, this paper utilizes characteristic image to obtain the gray scale cost of candidate point and each candidate point of frontier point.
The present invention adopts 6 kinds of characteristic images: (1-2) x, y direction single order partial derivative image, expression x, y direction grey scale change; (3-4) x, y direction second-order partial differential coefficient image, expression x, y direction grey scale change speed; (5) x, y direction mixed partial derivative image; (6) x, y direction second-order partial differential coefficient and image, this value is larger, shows that grey scale change speed is faster herein, for the possibility on lung border larger.
(2) candidate point of frontier point: for each point on initial lung border, calculate in all characteristic images in this point search zone the similarity degree of respective point gray scale in the gray scale of all pixels and training characteristics image.Select the point of 30 similarity degree maximums, as the candidate point of this frontier point.Similarity degree is the mahalanobis distance of this surrounding pixel point gray scale respective point surrounding pixel point gray scale set in the training sample characteristic image in all characteristic images, is defined as:
Wherein,
For on characteristic image, with a p
iBe the center of circle, r
cBe the n on the circle of radius
cThe gray scale of individual point,
Be respectively average and the covariance of the surrounding pixel point gray scale of i frontier point in j the characteristic image of training image.N is the characteristic image sum, and value is 6 here.
(3) cut apart based on the lung of dynamic programming
1. gray scale similarity cost: in the frontier point region of search, the gray scale similarity cost of pixel is the similarity degree of the surrounding pixel point gray scale of corresponding frontier point in this surrounding pixel point gray scale and the training image, as shown in Equation (2).The h of certain boundary candidates point in the test pattern
iBe worth littlely, show that the similarity of the intensity profile of point around this point and corresponding frontier point training sample is higher.
2. shape similarity cost: the shape similarity cost of i frontier point is defined as in the image:
Wherein, v
i=p
I+1-p
i, represent the shape facility of i frontier point,
The average and the covariance that represent respectively i frontier point shape facility in all training images.
3. the Optimal Boundary of intensity-based and shape similarity information search
For i frontier point p in the test pattern
i, in the region of search of appointment, exist m to have less gray scale similarity cost candidate point, then n frontier point will produce the gray scale cost matrix of a n * m:
The search optimal profile is looked for an optimal path (every row selects an element in the Matrix C) exactly, and in the time of along routing, the summation of gray scale and shape similarity cost is minimum, that is:
Wherein, γ is gray scale and shape similarity cost coefficient.Adjust the γ value, so that these two kinds of costs are brought into play roughly the same effect in the boundary search process.
Three, revise based on the lung border of ASM algorithm: because in some image, initial position may be excessively far away apart with actual boundary, when utilizing gray scale and shape similarity information to cut apart, the region of search does not cover the lung border.Therefore, the present invention improves the situation that the part point search is absorbed in local extremum by using the ASM algorithm cutting apart for the first time correction lung border, basis, obtains more excellent Search Results.
Revise the stage on the border, the intensity profile by utilizing frontier point gradient direction in all characteristic images of test sample book and weighted sum (the being mahalanobis distance) minimum of all characteristic image frontier point gradient direction intensity profile of training sample are revised the lung border.
For i point in the test pattern, can open at j and find one in the characteristic image centered by this point, length is 2m+l pixel, and direction is the derivation outline line of this place boundary normal direction.If the vector of the local gray level on the outline line after any point standardization is g
s, then this point whether be the border optimum point can by mahalanobis distance in all characteristic images and measure, that is:
F (g
s) value is less, shows the distribution of gray scale on the more approximate lung of the distribution real border point normal outline line of gray scale on this some place normal outline line, this put for the possibility of Optimal Boundary point larger.Thus, by searching f (g at the frontier point outline line
s) point of minimum value, can obtain optimum lung border.
After revising the boundary, the attitude of adjustment model and form parameter are determined final segmentation result.Calculate form parameter adjustment amount dx by formula (7):
M(s(1+ds),(θ+dθ))[x+dx]+(X
c+dX
c)=(X+dX) (7)
Can get
dx=M((s(1+ds))
-1,-(θ+dθ))[M(s,θ)[x]+dX-dX
c]-x (8)
Wherein, zoom factor 1+ds, twiddle factor 1+d θ can obtain with the reposition X+dX that search obtains by mating current some x.The form parameter that is obtained by formula (8), with this parameter adjustment initial profile result usually and shape inconsistent.We wish to find db, so that
By adjusting form parameter b+db, make b
k+ db
k In the scope, come the constraint shapes model.The variation dX of the amount of being adjusted
c, dY
c, ds, d θ and form parameter variable quantity db use these variable quantities and come undated parameter in iterative process, obtain suitable matching result.
The present invention sets up first the prior model of lung profile, and recycling gray scale and shape similarity information combining image feature are to the lung Region Segmentation.Because in some image, initial position may be excessively far away apart with actual boundary, when utilizing gray scale and shape similarity information to cut apart, the region of search does not cover the lung border.Therefore, the present invention improves the situation that the part point search is absorbed in local extremum by using active shape model (Active Shape Model, ASM) algorithm cutting apart for the first time correction lung border, basis, obtains more excellent Search Results.
Description of drawings:
Fig. 1 is the DR image that uses three kinds of algorithm segmentation results.
Fig. 2 is FB(flow block) of the present invention.
Embodiment:
Below in conjunction with the accompanying drawing in the embodiment of the invention, the technical scheme in the embodiment of the invention is carried out clear, intactly description, obviously, described embodiment only is the present invention's part embodiment, rather than whole embodiment.Give the embodiment among the present invention, those of ordinary skills belong to the scope of protection of the invention not making the every other embodiment that obtains under the creative work prerequisite.
The common data base that the present invention uses Japanese reflection technology association (Japanese Society of Radiological Technology) to set up.This database is comprised of 247 PA (Posterior-Anterior) image.Wherein, 93 normal pictures, 154 tuberculous images.Image is to obtain (Konica, Tokyo, Japan) by LD-4500 or LD-5500 laser film digitizer.Size is the 2048*2048 pixel, and each pixel is 12bits, and the yardstick of each pixel is 0.175*0.175mm
2
Interpretation
When the initial candidate point number of (1) lung frontier point is 20, (2) shape cost weight is 1*10
6(3) right lung 4-16, the region of search of 27-39 gauge point is centered by this point, the square area of 70*70 pixel, all the other gauge points are the 90*90 pixel region, left lung 4-16, the region of search of 36-39 gauge point is centered by this point, the square area of 70*70 pixel, all the other gauge points are the 120*120 pixel region, when (4) using 6 kinds of characteristic image search lung borders, obtain best segmentation performance, the overlap of cutting apart reaches 88.90.
1, the segmentation result analysis of different candidate point numbers
In the region of search of each gauge point, select the point of 10,20,30,40 gray scale Least-costs as lung boundary candidates point.If the candidate point number very little, in then the relative slightly large true lung frontier point of gray scale cost may be not included in; If candidate point is too many, then can introduce the noise spots such as noise, rib edge, reduce segmentation performance.Table 1 is in other parameter one regularly, the lung segmentation performance of different candidate point numbers.
The result that the different candidate point numbers of table 1 are cut apart
The candidate point number | 10 | 20 | 30 | 40 |
Overlap(%) | 87.63 | 88.90 | 88.53 | 88.64 |
2, the segmentation result analysis of different weights
Regulate gray scale and shape cost weight and can make the two be in the close order of magnitude, reach the effect of regulating grey scale change power and edge smoothing intensity.If the gray scale cost is excessive, segmentation result is affected by noise larger; If the shape effect is excessive, can search for less than the lung border with grey scale change.Therefore, select suitable weight coefficient very important.Table 2 is certain in other parameter, and gray scale cost weight is 1, and shape cost weight is respectively 1 * 10
3, 1 * 10
4, 1 * 10
5, 1 * 10
6, 1 * 10
7The time, the performance that lung is cut apart.
The segmentation result of table 2 difformity cost weight
Shape cost weight | 1×10 3 | 1×10 4 | 1×10 5 | 1×10 6 | 1×10 7 |
Overlap(%) | 87.05 | 86.79 | 86.29 | 88.90 | 86.19 |
3, the segmentation result analysis of different scale region of search
Put in the corresponding region of search at each initial boundary, seek the candidate boundary point with true lung frontier point half-tone information similarity maximum.If the region of search is excessive, comprise the borderline regions such as rib, produce a large amount of noise spots; If the region of search is too small, may not cover true lung border, make search be absorbed in local extremum.
Table 3 is that other parameter is certain, the performance that the lung that change region of search area obtains is cut apart.Wherein, L
i, R
i(i=1,2 ..., 42) i gauge point region of search of the left and right lung of the expression length of side.
The segmentation result analysis of different characteristic image
Table 4 obtains segmentation performance for utilizing characteristic image and being combined into the row bound search.Therefrom can find out, when using 6 kinds of characteristic images simultaneously, take full advantage of the change information of image all directions gray scale, obtain best segmentation result.
Different partitioning algorithm Performance Ratios
For (1) intensity-based and shape similarity partitioning algorithm, (2) utilize the intensity-based of characteristic image and shape similarity partitioning algorithm to reach these three kinds of partitioning algorithms that (3) utilize moving shape model correction segmentation result on this basis, segmentation result is as shown in table 5.Fig. 1 is for using the DR image of three kinds of algorithm segmentation results.(a) group is undistinguishable segmentation result, and (b) group is characteristic segmentation result, (c) segmentation result of group for feature being arranged and revising the boundary.
The segmentation result of the different regions of search of table 3 length of side
Table 4 is asked the method for characteristic image
Three kinds of partitioning algorithm Performance Ratios of table 5
Three kinds of partitioning algorithms | Without characteristic image | Characteristic image is arranged | Characteristic image and Edge retouch are arranged |
Overlap(%) | 82.18 | 87.99 | 88.90 |
Claims (6)
1. based on the lung dividing method that improves shape, concrete steps are:
One, determining of model initial profile position: comprise the mark training set, the alignment training set, prior model set up three major parts;
1. mark training set: use along the point on border and come the mark training set, point comprises following three classes: a, target-marking have the point of application-specific part, and the point of b, the irrelevant applying portion of mark, is filled out the point between a class point or 2 class points at c;
2. training set aligns: the operation by convergent-divergent, rotation and translation makes the training shapes alignment, and they are alignd closely as far as possible, establishes x
iThe vector of n point in i shape in the training set, x
i=(x
I0, y
I0, x
I1, y
I1..., x
Ik, y
Ik..., x
In-1, y
In-1)
T, wherein, (x
Ij, y
Ij) be j point in i the shape,
Given two similar shape x
iAnd x
j, select anglec of rotation θ, convergent-divergent s, translation (t
x, t
y), then to represent the anglec of rotation be that θ and scaling are the conversion of s to M (s, θ) [x], x
iBe mapped as M (s, θ) [x
j]+t minimizes following weighted sum:
E
j=(x
i-M(s,θ)[x
j]-t)
TW(x
i-M(s,θ)[x
j]-t) (1)
Wherein,
3. the foundation of prior model: after the shape vector registration process in the training set, utilize the method for principal component analysis (PCA) to find out statistical information and the rule of change of shape;
If average shape is
Each sample with respect to the shape vector covariance of the deviation formation of average shape is after the alignment
Calculate the eigen vector of this covariance, and with the eigenwert Sp that sorts
k=λ
kp
k, wherein, λ
kRepresent the eigenwert that k is large, λ
kLarger, its corresponding p
kData of description point changing pattern is just more important, and t important changing pattern forms new main shaft system p before selecting
s, then allow any one shape in the shape territory to be similar to by the weighted sum that average shape adds main shaft system and one group control parameter,
Wherein, p
s=(p
1p
2... p
t) be the matrix that front t proper vector forms, b
s=(b
1b
2... b
t)
TBe weight vector, through principal component analysis, get a front t eigenwert and characteristic of correspondence vector according to descending, the target object deformation that a front t eigenwert is determined accounts for all 2n eigenwert and determines that the ratio of target object deformation total amount is not less than V;
Two, the pulmonary parenchyma of intensity-based information and shape information is cut apart: utilize simultaneously gray scale and the shape information of image, so that the border gray scale that searches, shape information are similar to training image;
(1) characteristic image: utilize characteristic image to obtain the gray scale cost of candidate point and each candidate point of frontier point, adopt 6 kinds of characteristic images: (1-2) x, y direction single order partial derivative image, expression x, y direction grey scale change; (3-4) x, y direction second-order partial differential coefficient image, expression x, y direction grey scale change speed; (5) x, y direction mixed partial derivative image; (6) x, y direction second-order partial differential coefficient and image, this value is larger, shows that grey scale change speed is faster herein, for the possibility on lung border larger;
(2) candidate point of frontier point: for each point on initial lung border, calculate in all characteristic images in this point search zone the similarity degree of respective point gray scale in the gray scale of all pixels and training characteristics image, similarity degree is the mahalanobis distance of this surrounding pixel point gray scale respective point surrounding pixel point gray scale set in the training sample characteristic image in all characteristic images, is defined as:
Wherein,
For on characteristic image, with a p
iBe the center of circle, r
cBe the n on the circle of radius
cThe gray scale of individual point,
Be respectively average and the covariance of the surrounding pixel point gray scale of i frontier point in j the characteristic image of training image, N is the characteristic image sum, and value is 6 here;
(3) cut apart based on the lung of dynamic programming
1. gray scale similarity cost: in the frontier point region of search, the gray scale similarity cost of pixel is the similarity degree of the surrounding pixel point gray scale of corresponding frontier point in this surrounding pixel point gray scale and the training image, as shown in Equation (2), the h of certain boundary candidates point in the test pattern
iBe worth littlely, show that the similarity of the intensity profile of point around this point and corresponding frontier point training sample is higher;
2. shape similarity cost: the shape similarity cost of i frontier point is defined as in the image:
Wherein, v
i=p
I+1-p
i, represent the shape facility of i frontier point,
The average and the covariance that represent respectively i frontier point shape facility in all training images;
3. the Optimal Boundary of intensity-based and shape similarity information search
For i frontier point p in the test pattern
i, in the region of search of appointment, exist m to have less gray scale similarity cost candidate point, then n frontier point will produce the gray scale cost matrix of a n * m:
The search optimal profile is looked for an optimal path exactly, and in the time of along routing, the summation of gray scale and shape similarity cost is minimum, that is:
Wherein, γ is gray scale and shape similarity cost coefficient, adjusts the γ value, so that these two kinds of costs are brought into play roughly the same effect in the boundary search process;
Three, revise based on the lung border of ASM algorithm: by using the ASM algorithm cutting apart for the first time correction lung border, basis, improve the situation that the part point search is absorbed in local extremum, obtain more excellent Search Results;
Revise the stage on the border, the intensity profile by utilizing frontier point gradient direction in all characteristic images of test sample book and the weighted sum of all characteristic image frontier point gradient direction intensity profile of training sample, i.e. mahalanobis distance, minimum is revised the lung border;
For i point in the test pattern, can open at j and find one in the characteristic image centered by this point, length is 2m+l pixel, and direction is the derivation outline line of this place boundary normal direction, if the vector of the local gray level after any point standardization is g on the outline line
s, then this point whether be the border optimum point can by mahalanobis distance in all characteristic images and measure, that is:
F (g
s) value is less, shows the distribution of gray scale on the more approximate lung of the distribution real border point normal outline line of gray scale on this some place normal outline line, this is put for the possibility of Optimal Boundary point is larger, by searching f (g at the frontier point outline line
s) point of minimum value, can obtain optimum lung border;
After revising the boundary, the attitude of adjustment model and form parameter are determined final segmentation result, calculate form parameter adjustment amount dx by formula (7):
M(s(1+ds),(θ+dθ))[x+dx]+(X
c+dX
c)=(X+dX) (7)
Can get
dx=M((s(1+ds))
-1,-(θ+dθ))[M(s,θ)[x]+dX-dX
c]-x (8)
Wherein, zoom factor 1+ds, twiddle factor 1+d θ can obtain with the reposition X+dX that search obtains by mating current some x, the form parameter that is obtained by formula (8), with this parameter adjustment initial profile result usually and shape inconsistent, find db so that
By adjusting form parameter b+db, make b
k+ db
k In the scope, come the constraint shapes model, the variation dX of the amount of being adjusted
c, dY
c, ds, d θ and form parameter variable quantity db use these variable quantities and come undated parameter in iterative process, obtain suitable matching result.
2. the lung dividing method based on improving shape as claimed in claim 1, it is characterized in that: 1. the target-marking described in the mark training set has the point of application-specific part in the faceform, the point of expression eye center, or sharper keen turning on the expression border.
3. the lung dividing method based on improving shape as claimed in claim 2, it is characterized in that: sharper keen turning is the point at canthus on the described expression border.
4. the lung dividing method based on improving shape as claimed in claim 1, it is characterized in that: 1. the point of the irrelevant applying portion of the mark described in the mark training set is the peak of target on specific direction, or the extreme value place of curvature.
5. the lung dividing method based on improving shape as claimed in claim 1, it is characterized in that: described V gets 0.98.
6. the lung dividing method based on improving shape as claimed in claim 1 is characterized in that: in the candidate point of step (2) frontier point, select the point of 20 similarity degree maximums as the candidate point of this frontier point.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310237686.1A CN103310449B (en) | 2013-06-13 | 2013-06-13 | Based on the lung dividing method improving shape |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310237686.1A CN103310449B (en) | 2013-06-13 | 2013-06-13 | Based on the lung dividing method improving shape |
Publications (2)
Publication Number | Publication Date |
---|---|
CN103310449A true CN103310449A (en) | 2013-09-18 |
CN103310449B CN103310449B (en) | 2016-03-02 |
Family
ID=49135626
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201310237686.1A Expired - Fee Related CN103310449B (en) | 2013-06-13 | 2013-06-13 | Based on the lung dividing method improving shape |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103310449B (en) |
Cited By (16)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103294281A (en) * | 2013-04-28 | 2013-09-11 | 广州视睿电子科技有限公司 | Touch identification method and device of optical imaging touch screen |
CN105719278A (en) * | 2016-01-13 | 2016-06-29 | 西北大学 | Organ auxiliary positioning segmentation method based on statistical deformation model |
CN106611416A (en) * | 2015-10-19 | 2017-05-03 | 上海联影医疗科技有限公司 | Method and apparatus for lung segmentation in medical image |
CN107220965A (en) * | 2017-05-05 | 2017-09-29 | 上海联影医疗科技有限公司 | A kind of image partition method and system |
CN107230204A (en) * | 2017-05-24 | 2017-10-03 | 东北大学 | A kind of method and device that the lobe of the lung is extracted from chest CT image |
CN107424162A (en) * | 2017-05-05 | 2017-12-01 | 上海联影医疗科技有限公司 | A kind of image partition method and system |
WO2018201437A1 (en) * | 2017-05-05 | 2018-11-08 | 上海联影医疗科技有限公司 | Image segmentation method and system |
CN109117847A (en) * | 2018-08-31 | 2019-01-01 | 阿里巴巴集团控股有限公司 | Component dividing method and device for vehicle damage identification |
US10482604B2 (en) | 2017-05-05 | 2019-11-19 | Shanghai United Imaging Healthcare Co., Ltd. | Systems and methods for image processing |
CN110555853A (en) * | 2019-08-07 | 2019-12-10 | 杭州深睿博联科技有限公司 | Method and device for segmentation algorithm evaluation based on anatomical priors |
CN111524081A (en) * | 2020-04-24 | 2020-08-11 | 科大讯飞股份有限公司 | Lung image angle correction method and device, electronic equipment and storage medium |
CN111738980A (en) * | 2020-05-14 | 2020-10-02 | 上海依智医疗技术有限公司 | Medical image display method, computer equipment and storage medium |
CN112233131A (en) * | 2020-10-22 | 2021-01-15 | 广州极飞科技有限公司 | Method, device and equipment for dividing block and storage medium |
CN112470166A (en) * | 2019-06-18 | 2021-03-09 | 香港理工大学 | Method and system for segmenting overlapping cytoplasms in medical images |
CN112489064A (en) * | 2020-12-14 | 2021-03-12 | 桂林电子科技大学 | Panorama segmentation method based on edge scaling correction |
US10949977B2 (en) | 2017-06-30 | 2021-03-16 | Shanghai United Imaging Healthcare Co., Ltd. | Methods and systems for image segmentation |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2006054269A2 (en) * | 2004-11-19 | 2006-05-26 | Koninklijke Philips Electronics, N.V. | System and method for false positive reduction in computer-aided detection (cad) using a support vector machine (svm) |
CN102324109A (en) * | 2011-09-26 | 2012-01-18 | 上海理工大学 | Method for three-dimensionally segmenting insubstantial pulmonary nodule based on fuzzy membership model |
CN103035009A (en) * | 2012-12-19 | 2013-04-10 | 长春工业大学 | Pulmonary nodule edge rebuilding and partitioning method based on computed tomography (CT) image |
-
2013
- 2013-06-13 CN CN201310237686.1A patent/CN103310449B/en not_active Expired - Fee Related
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2006054269A2 (en) * | 2004-11-19 | 2006-05-26 | Koninklijke Philips Electronics, N.V. | System and method for false positive reduction in computer-aided detection (cad) using a support vector machine (svm) |
CN102324109A (en) * | 2011-09-26 | 2012-01-18 | 上海理工大学 | Method for three-dimensionally segmenting insubstantial pulmonary nodule based on fuzzy membership model |
CN103035009A (en) * | 2012-12-19 | 2013-04-10 | 长春工业大学 | Pulmonary nodule edge rebuilding and partitioning method based on computed tomography (CT) image |
Cited By (26)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103294281A (en) * | 2013-04-28 | 2013-09-11 | 广州视睿电子科技有限公司 | Touch identification method and device of optical imaging touch screen |
CN103294281B (en) * | 2013-04-28 | 2017-08-11 | 广州视睿电子科技有限公司 | touch identification method and device of optical imaging touch screen |
CN106611416A (en) * | 2015-10-19 | 2017-05-03 | 上海联影医疗科技有限公司 | Method and apparatus for lung segmentation in medical image |
CN105719278A (en) * | 2016-01-13 | 2016-06-29 | 西北大学 | Organ auxiliary positioning segmentation method based on statistical deformation model |
CN105719278B (en) * | 2016-01-13 | 2018-11-16 | 西北大学 | A kind of medical image cutting method based on statistics deformation model |
CN107424162A (en) * | 2017-05-05 | 2017-12-01 | 上海联影医疗科技有限公司 | A kind of image partition method and system |
WO2018201437A1 (en) * | 2017-05-05 | 2018-11-08 | 上海联影医疗科技有限公司 | Image segmentation method and system |
CN107220965A (en) * | 2017-05-05 | 2017-09-29 | 上海联影医疗科技有限公司 | A kind of image partition method and system |
US11935246B2 (en) | 2017-05-05 | 2024-03-19 | Shanghai United Imaging Healthcare Co., Ltd. | Systems and methods for image segmentation |
US10482604B2 (en) | 2017-05-05 | 2019-11-19 | Shanghai United Imaging Healthcare Co., Ltd. | Systems and methods for image processing |
US11170509B2 (en) | 2017-05-05 | 2021-11-09 | Shanghai United Imaging Healthcare Co., Ltd. | Systems and methods for image segmentation |
CN107230204A (en) * | 2017-05-24 | 2017-10-03 | 东北大学 | A kind of method and device that the lobe of the lung is extracted from chest CT image |
CN107230204B (en) * | 2017-05-24 | 2019-11-22 | 东北大学 | A kind of method and device for extracting the lobe of the lung from chest CT image |
US10949977B2 (en) | 2017-06-30 | 2021-03-16 | Shanghai United Imaging Healthcare Co., Ltd. | Methods and systems for image segmentation |
US11710242B2 (en) | 2017-06-30 | 2023-07-25 | Shanghai United Imaging Healthcare Co., Ltd. | Methods and systems for image segmentation |
CN109117847A (en) * | 2018-08-31 | 2019-01-01 | 阿里巴巴集团控股有限公司 | Component dividing method and device for vehicle damage identification |
CN112470166A (en) * | 2019-06-18 | 2021-03-09 | 香港理工大学 | Method and system for segmenting overlapping cytoplasms in medical images |
CN110555853A (en) * | 2019-08-07 | 2019-12-10 | 杭州深睿博联科技有限公司 | Method and device for segmentation algorithm evaluation based on anatomical priors |
CN110555853B (en) * | 2019-08-07 | 2022-07-19 | 杭州深睿博联科技有限公司 | Method and device for segmentation algorithm evaluation based on anatomical priors |
CN111524081A (en) * | 2020-04-24 | 2020-08-11 | 科大讯飞股份有限公司 | Lung image angle correction method and device, electronic equipment and storage medium |
CN111524081B (en) * | 2020-04-24 | 2023-10-10 | 讯飞医疗科技股份有限公司 | Lung image angle correction method, device, electronic equipment and storage medium |
CN111738980A (en) * | 2020-05-14 | 2020-10-02 | 上海依智医疗技术有限公司 | Medical image display method, computer equipment and storage medium |
CN111738980B (en) * | 2020-05-14 | 2023-08-04 | 北京深睿博联科技有限责任公司 | Medical image display method, computer equipment and storage medium |
CN112233131A (en) * | 2020-10-22 | 2021-01-15 | 广州极飞科技有限公司 | Method, device and equipment for dividing block and storage medium |
CN112489064A (en) * | 2020-12-14 | 2021-03-12 | 桂林电子科技大学 | Panorama segmentation method based on edge scaling correction |
CN112489064B (en) * | 2020-12-14 | 2022-03-25 | 桂林电子科技大学 | Panorama segmentation method based on edge scaling correction |
Also Published As
Publication number | Publication date |
---|---|
CN103310449B (en) | 2016-03-02 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103310449B (en) | Based on the lung dividing method improving shape | |
CN106485695B (en) | Medical image Graph Cut dividing method based on statistical shape model | |
CN111401480B (en) | Novel mammary gland MRI automatic auxiliary diagnosis method based on fusion attention mechanism | |
US9760935B2 (en) | Method, system and computer program product for generating recommendations for products and treatments | |
CN102354397B (en) | Method for reconstructing human facial image super-resolution based on similarity of facial characteristic organs | |
CN103985133B (en) | Search method and system for optimal splicing lines among images based on graph-cut energy optimization | |
US7187811B2 (en) | Method for image resolution enhancement | |
CN109754404A (en) | A kind of lesion segmentation approach end to end based on more attention mechanism | |
CN107169974A (en) | It is a kind of based on the image partition method for supervising full convolutional neural networks more | |
CN106611416B (en) | Method and device for segmenting lung in medical image | |
CN109584251A (en) | A kind of tongue body image partition method based on single goal region segmentation | |
CN102682449B (en) | Automatic segmentation method based on self-adaptive external force level set for magnetic resonance images (MRIs) of soft tissue and realization method thereof | |
CN102509286B (en) | Target region sketching method for medical image | |
CN109859184A (en) | A kind of real-time detection of continuous scanning breast ultrasound image and Decision fusion method | |
CN106157249A (en) | Based on the embedded single image super-resolution rebuilding algorithm of optical flow method and sparse neighborhood | |
CN106096654A (en) | A kind of cell atypia automatic grading method tactful based on degree of depth study and combination | |
CN112862824A (en) | Novel coronavirus pneumonia focus detection method, system, device and storage medium | |
CN104732546B (en) | The non-rigid SAR image registration method of region similitude and local space constraint | |
JP7312510B1 (en) | Whole-slide pathological image classification system and construction method considering tumor microenvironment | |
CN110021019B (en) | AI-assisted hair thickness distribution analysis method for AGA clinical image | |
CN104794693A (en) | Human image optimization method capable of automatically detecting mask in human face key areas | |
CN104408731B (en) | Region graph and statistic similarity coding-based SAR (synthetic aperture radar) image segmentation method | |
CN106462974A (en) | Optimization of parameters for segmenting an image | |
Li et al. | Novel iterative attention focusing strategy for joint pathology localization and prediction of MCI progression | |
Ma et al. | Coarse-to-fine segmentation of organs at risk in nasopharyngeal carcinoma radiotherapy |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20160302 |