CN109509199A - A kind of medical image Organizational Intelligence dividing method based on three-dimensional reconstruction - Google Patents
A kind of medical image Organizational Intelligence dividing method based on three-dimensional reconstruction Download PDFInfo
- Publication number
- CN109509199A CN109509199A CN201811177386.8A CN201811177386A CN109509199A CN 109509199 A CN109509199 A CN 109509199A CN 201811177386 A CN201811177386 A CN 201811177386A CN 109509199 A CN109509199 A CN 109509199A
- Authority
- CN
- China
- Prior art keywords
- point
- image
- pixel
- value
- threshold
- 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
- 238000000034 method Methods 0.000 title claims abstract description 33
- 230000011218 segmentation Effects 0.000 claims abstract description 20
- 230000009466 transformation Effects 0.000 claims abstract description 12
- 238000001514 detection method Methods 0.000 claims abstract description 4
- 238000003709 image segmentation Methods 0.000 claims description 10
- 238000001914 filtration Methods 0.000 claims description 6
- 238000009499 grossing Methods 0.000 claims description 6
- 238000003708 edge detection Methods 0.000 claims description 5
- 238000000354 decomposition reaction Methods 0.000 claims description 2
- 230000003247 decreasing effect Effects 0.000 claims 1
- 230000000694 effects Effects 0.000 description 5
- 230000008901 benefit Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 239000003814 drug Substances 0.000 description 2
- 230000008707 rearrangement Effects 0.000 description 2
- 238000000638 solvent extraction Methods 0.000 description 2
- 230000007704 transition Effects 0.000 description 2
- 101100338885 Arabidopsis thaliana HHL1 gene Proteins 0.000 description 1
- 101100135632 Arabidopsis thaliana PAR2 gene Proteins 0.000 description 1
- 241001269238 Data Species 0.000 description 1
- 230000002776 aggregation Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- 210000004556 brain Anatomy 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 241000894007 species Species 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/40—Image enhancement or restoration using histogram techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- 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
- G06T2207/10016—Video; Image sequence
- G06T2207/10021—Stereoscopic video; Stereoscopic image sequence
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
- Apparatus For Radiation Diagnosis (AREA)
Abstract
The invention discloses a kind of medical image Organizational Intelligence dividing method based on three-dimensional reconstruction is pre-processed to image and is adjusted window comprising steps of 1) inputting DICOM medical image sequences；2) Wavelet transformation is carried out to image, image data is reduced to the 1/4 of original image；3) grey level histogram is counted, initial seed point and growing threshold are extracted；4) three-dimensional edges detection is carried out to image, obtains edge contour figure；5) gray scale and marginal information are combined, three-dimensional spatial area growth is carried out；6) area attribute after segmentation is calculated, seed point and growing threshold are optimized.The present invention passes through statistic histogram, threshold value needed for initial seed point needed for region growing can be automatically extracted and growth criterion, 3D region growth is carried out then in conjunction with the edge contour information of image, and initial seed point and growing threshold are continued to optimize by iteration growth, so as to improve the result of segmentation.Meanwhile the present invention is reduced image data to the 1/4 of original image, by Wavelet transformation so as to shorten operation time.
Description
Technical field
The present invention relates to the technical fields of medical image segmentation, refer in particular to a kind of medical image group based on three-dimensional reconstruction
Knit intelligent scissor method.
Background technique
Medical image segmentation refers to according to a certain specific character, will have the area of different meanings in medical image
Domain separates, and identical region is made to meet region consistency.The dividing method of common medical image includes: based on threshold value
Dividing method, the dividing method based on region, the dividing method based on edge and the dividing method based on cluster.
Dividing method based on threshold value is using one or a set of gray threshold, by the image data mark in threshold range
It is denoted as area-of-interest, the image data outside range is labeled as background data, achievees the purpose that image segmentation with this.Based on threshold value
Dividing method, have it is simple, quickly, the low advantage of computation complexity；But the selection of threshold value is affected by noise big, it is difficult to choose
Degree is high；Simultaneously as merely with the grayscale information of image, it is difficult to divide the high area-of-interest of gray scale similarity.And medicine
Image generally has the characteristic that intensity profile is uneven, each tissue gray scale similarity is high, and therefore, which is being used for medical image
Segmentation of Multi-target when, segmentation effect is poor.
Dividing method based on region, the methods of main inclusion region growth, split degree method, level set；Wherein, due to
It calculates simply, region-growing method is widely used in medical image segmentation.But traditional algorithm of region growing high degree
Dependent on initial seed point selection and grow criterion setting, it is difficult to realize the automatic segmentation of multizone.
Dividing method based on edge, be using between different zones, often exist more violent grey scale change this
Characteristic comes the edge between detection zone and carries out image segmentation.This method is for edge gray value transition than more significant and noise
The segmentation effect of lesser simple image is ideal；For structure is complicated and presence is compared with the medical image of very noisy, then can produce
The case where raw pseudo-edge or missing edges.
Dividing method based on cluster is to indicate the pixel in image space with corresponding feature space point, according to it
Feature space is split in the aggregation of feature space, they are then mapped back into original image image space, obtains segmentation result.
Dividing method based on cluster mainly includes following difficult point: the determination of cluster class number；Determine the validity criterion of cluster；It is poly-
The setting of class initial value；Computational complexity is larger.Traditional clustering algorithm does not consider the spatial information of image data, makes an uproar to image
Sound is very sensitive.
Currently, traditional Medical image segmentation algorithm still remains following problem:
1. being split using priori knowledge, for complicated medical image, any single partitioning algorithm is all difficult
To obtain satisfactory segmentation result.
2. most partitioning algorithm is all based on the segmentation that two-dimensional image data carries out, image is not accounted in three-dimensional
Location information in space.
3. current Algorithm on Automatic Segmentation is difficult to meet the needs of segmentation precision, therefore cutting procedure usually introduces manually
Interactive information divides low efficiency.
Summary of the invention
It is an object of the invention to overcome the shortcomings of the prior art, a kind of medicine shadow based on three-dimensional reconstruction is proposed
As Organizational Intelligence dividing method, this method can automatically extract region growing institute by statistic histogram using three-dimensional data
Threshold value needed for the initial seed point and growth criterion that need carries out region growing then in conjunction with the edge contour information of image, and
Initial seed point and growing threshold are continued to optimize by iteration growth, so as to improve the result of segmentation.Meanwhile this method is by small
Wave variation, image data was reduced to the 1/4 of original image, so as to shorten operation time.
To achieve the above object, a kind of technical solution provided by the present invention are as follows: medical image group based on three-dimensional reconstruction
Knit intelligent scissor method, comprising the following steps:
1) DICOM medical image sequences are inputted, window is pre-processed and adjusted to image；
2) Wavelet transformation is carried out to image, image data is reduced to the 1/4 of original image；
3) grey level histogram is counted, initial seed point and growing threshold are extracted；
4) three-dimensional edges detection is carried out to image, obtains edge contour figure；
5) gray scale and marginal information are combined, three-dimensional spatial area growth is carried out；
6) area attribute after segmentation is calculated, seed point and growing threshold are optimized.
In step 1), one group of DICOM format medical image sequences is inputted, image sequence is adjusted in pretreatment stage
Window and Gaussian smoothing.So-called tune window turns the gray value being located in window in image data that is, according to the window width of precognition and window position
Be changed to value when display in most bright and most dark range, be set to higher than the part of window tonal range it is most bright, be lower than window gray scale model
Part in enclosing is set to most dark.Adjust window formula as follows:
Wherein, F'xyzFor gray value of the pixel after adjusting window, F at three dimensional space coordinate (x, y, z)xyzFor the pixel
Original gray value, C is window position, and W is window width, and Max is the maximum gradation value that display is shown.
For the image sequence after tune window, Gaussian smoothing is carried out, i.e., image is weighted and averaged, each pixel
Value, the especially value of itself and other interior pixels of neighborhood obtain after being weighted averagely.
Three-dimensional Gaussian function is defined in above formula, wherein G (x, y, z) indicates the Gaussian template being calculated, (x, y, z)
The relative coordinate of other pixels and central pixel point in expression field, σ indicate the standard deviation of Gaussian Profile.
In step 2), two layers of three-dimensional discrete wavelet variation are carried out to image using " db1 " wavelet basis, that is, to image
X, Y, the filtering that Z-direction carries out low-and high-frequency respectively decomposes, and every layer of Wavelet transformation will all generate 8 components, wherein low comprising 1
Frequency component and 7 high fdrequency components continue the low frequency component that first layer generates to do filtering decomposition, finally obtain 15 components.It takes
The low frequency component generated after second layer Wavelet transformation carries out subsequent image segmentation, image data can be reduced to original image
1/4.
In step 3), the pixel grey scale of statistical picture generates grey level histogram, scans histogram, obtains each sense
The gray value and its growing threshold of interest region initial seed point, detail are as follows:
3.1) grey level histogram is generated according to the following formula, defines peak set P, valley set V；
In formula, H (i) indicates that gray scale is the percentage of the total pixel number of pixel number Zhan of i, wherein i ∈ 0,1,2,
3 ..., Max-1 }, X, Y, Z indicate image in three-dimensional space X, and Y, the pixel sum in Z-direction, x, y, z indicates that pixel exists
The coordinate of three-dimensional space, FxyzIndicate gray value of the image pixel at (x, y, z), Equalxyz(i) for judging at (x, y, z)
Pixel grey scale whether be equal to i, if being equal to, return to 1, otherwise, return 0；
3.2) the corresponding point of maximum value of histogram is found as first peak point, and recording the corresponding gray value of point is
p0, and be added among set P；
3.3) find next peak point, corresponding grey scale value is p', Ying Man: centered on p', t is the range of step-length
It is interior, that is, tonal range is in the range of [p'- t, p'] and [p', p'+t], monotonic increase and list is presented in grayscale image respectively
Adjust decrement states；
3.4) step 3.3) is repeated until finding all peak points and their corresponding gray scale value set P, from small arrival
Order rearrangement column set P in element；
3.5) set P is traversed, gray value corresponding to the smallest point of histogram value between every two peak point is sequentially found
As valley, set V is formed；
3.6) set V is traversed from small to large, in pairs as growing threshold, is found between two valleies according to set P
Peak value p, randomly choosing three gray values is the pixel of p as initial seed point.
In step 4), three-dimensional Sobel edge detection operator is designed, calculates the pixel gradient of image, threshold value is set, it is high
In the edge pixel that the pixel of threshold value is considered as image, it is set to 1, is otherwise set to 0, the edge contour of medical image is obtained with this
Figure.
In step 5), using the initial seed point and its corresponding growing threshold for counting acquisition in histogram, in conjunction with figure
The edge contour information of picture carries out the region growing of three-dimensional space, that is, is determining whether point to be grown meets appraisal of growth criterion
When, gray scale bound needed for not only meeting growth, also needing to meet rim condition, specific step is as follows:
5.1) initial seed point extracted in step 3) is pressed into seed stack, the corresponding growing threshold of record seed point；
5.2) judge whether seed stack is empty, is that sky then terminates to grow；It is not sky, then the seed point of stack top is taken out, with this
Point is initial point, searches for its six fields pixel (x, y, z), centered on (x, y, z), generates the super voxel of 3*3*3, using super body
The average gray value of element is as the gray value at (x, y, z) pixel；
5.3) judge whether the gray value of (x, y, z) meets the corresponding growing threshold of seed point, if satisfied, judging that the point is
No is marginal point, and marginal point is divided into three kinds of situations: 1. the point is not marginal point, then with seed point labeled as the same area, and at
For new seed point, it is pressed into seed stack；2. the point is marginal point, but next point in the direction of growth is still marginal point, then
It is labeled as the same area with seed point, and becomes new seed point and is pressed into storehouse；3. be marginal point, and in the direction of growth under
Some non-edge points, the direction stop growing；
5.4) step 5.2) is repeated, until growth terminates.
In step 6), after each seed point growth, the flat of the area-of-interest of its generation is calculated according to the following formula
Equal gray scale and its inter-class variance size with background area；
In formula, μROIIndicate the average gray of area-of-interest, NROIIndicate the number of pixels that area-of-interest includes, μBTable
Show the average gray of background area, NBIndicate the number of pixels that background area includes, FxyzIndicate the ash at pixel (x, y, z)
Angle value, θ indicate that the inter-class variance in right interest region and background area, μ indicate image ensemble average gray scale；
Merge the initial seed point and growing threshold of area-of-interest similar in average gray；Set minimum inter-class variance threshold
Value judges whether there is the area-of-interest that inter-class variance is less than the threshold value, and if it exists, then assert that the cut zone is meaningless,
Merge the region initial seed point and growing threshold to adjacent average gray area-of-interest；According to new seed point and life
Long threshold value regrows, until there is no meaningless cut zone or reaching highest the number of iterations, finally returns that comprising more
The segmented image of class label.
Compared with prior art, the present invention have the following advantages that with the utility model has the advantages that
1. the present invention is based on 3 d image datas to be split, the spatial positional information of image is effectively utilized, it can be with
Obtain better segmentation effect.
2. the present invention solves the disadvantage that algorithm of region growing relies on artificial selected seed point and design growth criterion, can
The gray threshold that automatic selected seed point and its growth need to meet, realizes the automation of segmentation.
3. the present invention combines the edge contour information of image on the basis of region growing, it ensure that and chosen in automation
The inaccurate situation of growing threshold under the accuracy rate that grows, improve the accuracy of segmentation.
4. the present invention carries out region growing by iteration, calculates average gray and inter-class variance improves segmentation result, into one
Step reduces automation and chooses the invalid influence of seed point and growing threshold to segmentation result in obtained seed point, improves segmentation
Accuracy.
5. the present invention reduces image data using Wavelet transformation, reduce calculation amount and runing time, so that algorithm has
There is better real-time.
Detailed description of the invention
Fig. 1 is logical flow diagram of the present invention.
Fig. 2 is the Gaussian template figure that the present invention uses.
Fig. 3 is that two layers of 3 D wavelet that the present invention uses change schematic diagram.
Fig. 4 is the Sobel edge detection operator structure chart that the present invention uses.
Fig. 5 is the region growing flow chart that the present invention designs.
Specific embodiment
The present invention is further explained in the light of specific embodiments.
As shown in Figure 1, the medical image Organizational Intelligence dividing method provided by the present invention based on three-dimensional reconstruction, including such as
Lower step:
The first step, input one group of DICOM format medical image sequences, pretreatment stage to image sequence carry out adjust window and
Gaussian smoothing.So-called tune window is by the grayvalue transition in window is located in image data that is, according to the window width of precognition and window position
Value when display in most bright and most dark range, be set to higher than the part of window tonal range it is most bright, lower than in window tonal range
Part be set to it is most dark.Adjust window formula as follows:
Wherein, F'xyzFor gray value of the pixel after adjusting window, F at three dimensional space coordinate (x, y, z)xyzFor the pixel
Original gray value, C is window position, and W is window width, and Max is the maximum gradation value that display is shown.
In this example, the brain medical image sequences of DICOM format are inputted, C=50 is arranged in size 512*512*200,
W=90, Max=256.For the image sequence after tune window, Gaussian smoothing is carried out, i.e., image is weighted and averaged, each
The value of pixel, the especially value of itself and other interior pixels of neighborhood obtain after being weighted averagely.
Three-dimensional Gaussian function is defined in above formula, wherein G (x, y, z) indicates the coefficient for the Gaussian template being calculated,
(x, y, z) indicates the relative coordinate of other pixels and central pixel point in field, and σ indicates the standard deviation of Gaussian Profile.According to
The size of the σ value of setting determines smooth effect, and value is bigger, then smooth effect is more obvious.As shown in Figure 2, it is shown that a 3*3*
σ=0.8 is arranged in the coordinate of each point in 3 Gaussian template, this example, and coordinate value each in Fig. 2 is substituted into Gaussian function, meter
Obtained G (x, y, z) value is the weight coefficient of Gaussian template.
Second step carries out the variation of two layers of three-dimensional discrete wavelet to image using " db1 " wavelet basis, that is, to the X of image,
Y, the filtering that Z-direction carries out low-and high-frequency respectively are decomposed, and every layer of Wavelet transformation will all generate 8 components, wherein including 1 low frequency point
Amount and 7 high fdrequency components continue to do filtering to the low frequency component that first layer generates and decompose, finally obtain it is shown in Fig. 3 [LLL2,
(LLH2, LHL2, HLL2, LHH2, HLH2, HHL2, HHH2), (LLH1, LHL1, HLL1, LHH1, HLH1, HHL1, HHH1)] altogether
15 components.Take the low frequency component generated after second layer Wavelet transformation, i.e., " LLL2 " component carries out subsequent image segmentation, can be with
Image data is reduced to the 1/4 of original image.In this example, the image of 512*512*200 size by two layers of three-dimensional from
After dissipating Wavelet transformation, output image size is 128*128*40.
Third step, the pixel grey scale of statistical picture generate grey level histogram, scan histogram, it is interested to obtain each
The gray value and its growing threshold of region initial seed point, detail are as follows:
3.1, grey level histogram is generated according to the following formula, defines peak set P, valley set V；
In formula, H (i) indicates that gray scale is the percentage of the total pixel number of pixel number Zhan of i, wherein i ∈ 0,1,2,
3 ..., 255 }, X=128, Y=128, Z=40, x, y, z indicate coordinate of the pixel in three-dimensional space, FxyzIndicate image pixel
Gray value at (x, y, z), if gray value is i, Equalxyz(i) it is equal to 1, the pixel that gray scale is i in corresponding H (i)
Number plus 1.
3.2, the corresponding point of maximum value of histogram is found as first peak point, and recording the corresponding gray value of point is
p0, and be added among set P；
3.3, next peak point is found, corresponding grey scale value is p', Ying Man: centered on p', step-length t=10 is set,
In the range of namely tonal range is [p'- 10, p'] and [p', p'+10], monotonic increase and dullness is presented in grayscale image respectively
Decrement states；
3.4, step 3.3 is repeated until finding all peak points and their corresponding gray scale value set P, from small arrival
Order rearrangement arranges the element in set P；
3.5, set P is traversed, gray value corresponding to the smallest point of histogram value between every two peak point is sequentially found
As valley, set V is formed.
3.6, set V is traversed from small to large, in pairs as growing threshold, is found between two valleies according to set P
Peak value p, randomly choosing three gray values is the pixel of p as initial seed point.
4th step designs each of the Sobel edge detection operator of three-dimensional shown in Fig. 4, traversal image three-dimensional space
Pixel calculates the gradient of the point using Sobel edge detection operator.Threshold value is set, and the pixel higher than threshold value is considered as figure
The edge pixel of picture is set to 1, is otherwise set to 0, obtains the edge contour figure of medical image with this.
5th step, using the initial seed point and its corresponding growing threshold for counting acquisition in histogram, in conjunction with image
Edge contour information carries out the region growing of three-dimensional space shown in fig. 5, that is, is determining whether point to be grown meets appraisal of growth
When criterion, gray scale bound needed for not only meeting growth also needs to meet rim condition.Specific step is as follows:
5.1, the initial seed point extracted in third step is pressed into seed stack, the corresponding growing threshold of record seed point；
5.2, judge whether seed stack is empty, is that sky then terminates to grow；It is not sky, then takes out the seed point (x of stack top0,y0,
z0), using the point as initial point, search for its six fields pixel (x, y, z), (x, y, z) ∈ { (x0-1,y0,z0), (x0+1,y0,z0),
(x0,y0-1,z0),(x0,y0+1,z0),(x0,y0,z0-1),(x0,y0-1,z0+ 1) }, centered on (x, y, z), 3*3*3 is generated
Super voxel, using the average gray value of super voxel, i.e., centered on (x, y, z) add 26 pixels of surrounding average gray
Value is as the gray value at (x, y, z) pixel；
5.3, judge whether the gray value of (x, y, z) meets the corresponding growing threshold of seed point, if satisfied, judging that the point is
No is marginal point, and marginal point is divided into three kinds of situations: 1. the point is not marginal point, then with seed point labeled as the same area, and at
For new seed point, it is pressed into seed stack；2. the point is marginal point, but next point in the direction of growth is still marginal point, then
It is labeled as the same area with seed point, and becomes new seed point and is pressed into storehouse；3. be marginal point, and in the direction of growth under
Some non-edge points, the direction stop growing；
5.4, step 5.2 is repeated, until growth terminates.
6th step calculates the average ash of the area-of-interest of its generation after each seed point growth according to the following formula
Degree and its inter-class variance size with background area.
In formula, μROIIndicate the average gray of area-of-interest, NROIIndicate the number of pixels that area-of-interest includes, μBTable
Show the average gray of background area, NBIndicate the number of pixels that background area includes, FxyzIndicate the ash at pixel (x, y, z)
Angle value, θ indicate that the inter-class variance in right interest region and background area, μ indicate image ensemble average gray scale.
Merge the initial seed point and growing threshold of area-of-interest similar in average gray；This example is set between infima species
Variance threshold values are 5, judge whether there is area-of-interest and background area inter-class variance less than 5, and if it exists, then assert that the sense is emerging
The segmentation in interesting region is meaningless, merge the region initial seed point and growing threshold to adjacent average gray region of interest
Domain；It is regrowed according to new seed point and growing threshold, is changed until meaningless cut zone is not present or reaches highest
Generation number finally returns that the segmented image comprising multiclass label.
Embodiment described above is only the preferred embodiments of the invention, is not limited thereto and makes practical range of the invention,
Therefore all shapes according to the present invention, change made by principle, it should all be included within the scope of protection of the present invention.
Claims (7)
1. a kind of medical image Organizational Intelligence dividing method based on three-dimensional reconstruction, which comprises the following steps:
1) DICOM medical image sequences are inputted, window is pre-processed and adjusted to image；
2) Wavelet transformation is carried out to image, image data is reduced to the 1/4 of original image；
3) grey level histogram is counted, initial seed point and growing threshold are extracted；
4) three-dimensional edges detection is carried out to image, obtains edge contour figure；
5) gray scale and marginal information are combined, three-dimensional spatial area growth is carried out；
6) area attribute after segmentation is calculated, seed point and growing threshold are optimized.
2. a kind of medical image Organizational Intelligence dividing method based on three-dimensional reconstruction according to claim 1, feature exist
In: in step 1), input one group of DICOM format medical image sequences, pretreatment stage to image sequence carry out adjust window and
Gaussian smoothing, wherein so-called tune window, i.e., according to the window width of precognition and window position, the gray value being located in window in image data
Be converted to value when display in most bright and most dark range, be set to higher than the part of window tonal range it is most bright, be lower than window gray scale
Part in range is set to most dark；Adjust window formula as follows:
Wherein, F'xyzFor gray value of the pixel after adjusting window, F at three dimensional space coordinate (x, y, z)xyzFor the original of the pixel
Beginning gray value, C are window position, and W is window width, and Max is the maximum gradation value that display is shown；
For the image sequence after tune window, Gaussian smoothing is carried out, i.e., image is weighted and averaged, the value of each pixel,
Especially the value of itself and other interior pixels of neighborhood obtains after being weighted averagely；
Three-dimensional Gaussian function is defined in above formula, wherein G (x, y, z) indicates that the Gaussian template being calculated, (x, y, z) indicate
The relative coordinate of other pixels and central pixel point in field, σ indicate the standard deviation of Gaussian Profile.
3. a kind of medical image Organizational Intelligence dividing method based on three-dimensional reconstruction according to claim 1, feature exist
In: in step 2), the variation of two layers of three-dimensional discrete wavelet is carried out to image using " db1 " wavelet basis, that is, to the X of image,
Y, the filtering that Z-direction carries out low-and high-frequency respectively are decomposed, and every layer of Wavelet transformation will all generate 8 components, wherein including 1 low frequency point
Amount and 7 high fdrequency components continue the low frequency component that first layer generates to do filtering decomposition, finally obtain 15 components；Take second
The low frequency component generated after layer Wavelet transformation carries out subsequent image segmentation, can reduce image data to the 1/ of original image
4。
4. a kind of medical image Organizational Intelligence dividing method based on three-dimensional reconstruction according to claim 1, feature exist
In: in step 3), the pixel grey scale of statistical picture generates grey level histogram, scans histogram, obtains each region of interest
The gray value and its growing threshold of domain initial seed point, detail are as follows:
3.1) grey level histogram is generated according to the following formula, defines peak set P, valley set V；
In formula, H (i) indicates that gray scale is the percentage of the total pixel number of pixel number Zhan of i, wherein i ∈ 0,1,2,3 ...,
Max-1 }, X, Y, Z indicate image in three-dimensional space X, and Y, the pixel sum in Z-direction, x, y, z indicates pixel in three-dimensional space
Between coordinate, FxyzIndicate gray value of the image pixel at (x, y, z), Equalxyz(i) for judging the pixel at (x, y, z)
Whether gray scale is equal to i, if being equal to, returns to 1, otherwise, returns to 0；
3.2) the corresponding point of maximum value of histogram is found as first peak point, and recording the corresponding gray value of point is p0, and
It is added among set P；
3.3) next peak point is found, corresponding grey scale value is p', Ying Man: centered on p', in the range of t is step-length,
It is exactly tonal range in the range of [p'- t, p'] and [p', p'+t], monotonic increase and monotone decreasing is presented in grayscale image respectively
State；
3.4) step 3.3) is repeated, until finding all peak points and their corresponding gray scale value set P, from the suitable of small arrival
Element in sequence permutatation set P；
3.5) set P is traversed, gray value conduct corresponding to the smallest point of histogram value between every two peak point is sequentially found
Valley forms set V；
3.6) set V is traversed from small to large, is used as growing threshold in pairs, the peak value between two valleies is found according to set P
P, randomly choosing three gray values is the pixel of p as initial seed point.
5. a kind of medical image Organizational Intelligence dividing method based on three-dimensional reconstruction according to claim 1, feature exist
In: in step 4), three-dimensional Sobel edge detection operator is designed, calculates the pixel gradient of image, threshold value is set, is higher than threshold
The pixel of value is considered as the edge pixel of image, is set to 1, is otherwise set to 0, obtains the edge contour figure of medical image with this.
6. a kind of medical image Organizational Intelligence dividing method based on three-dimensional reconstruction according to claim 1, feature exist
In: in step 5), using the initial seed point and its corresponding growing threshold for counting acquisition in histogram, in conjunction with the side of image
The region growing of edge profile information progress three-dimensional space, that is, when determining whether point to be grown meets appraisal of growth criterion, not only
Gray scale bound needed for meeting growth, also needs to meet rim condition, the specific steps are as follows:
5.1) initial seed point extracted in step 3) is pressed into seed stack, the corresponding growing threshold of record seed point；
5.2) judge whether seed stack is empty, is that sky then terminates to grow；It is not sky, then takes out the seed point of stack top, be with the point
Initial point searches for its six fields pixel (x, y, z), centered on (x, y, z), the super voxel of 3*3*3 is generated, using super voxel
Average gray value is as the gray value at (x, y, z) pixel；
5.3) judge (x, y, z) gray value whether meet the corresponding growing threshold of seed point, if satisfied, judge the point whether be
Marginal point, marginal point are divided into three kinds of situations: 1. the point is not marginal point, then the same area is labeled as with seed point, and becomes new
Seed point, be pressed into seed stack；2. the point is marginal point, but next point in the direction of growth is still marginal point, then and plants
Son point is labeled as the same area, and becomes new seed point and be pressed into storehouse；3. being marginal point, and the next point in the direction of growth
Non-edge point, the direction stop growing；
5.4) step 5.2) is repeated, until growth terminates.
7. a kind of medical image Organizational Intelligence dividing method based on three-dimensional reconstruction according to claim 1, feature exist
In: in step 6), after each seed point growth, the average gray of the area-of-interest of its generation is calculated according to the following formula
And its inter-class variance size with background area；
In formula, μROIIndicate the average gray of area-of-interest, NROIIndicate the number of pixels that area-of-interest includes, μBIndicate back
The average gray of scene area, NBIndicate the number of pixels that background area includes, FxyzIndicate the gray value at pixel (x, y, z),
θ indicates that the inter-class variance in right interest region and background area, μ indicate image ensemble average gray scale；
Merge the initial seed point and growing threshold of area-of-interest similar in average gray；Variance threshold values between setting infima species,
Judge whether there is the area-of-interest that inter-class variance is less than the threshold value, and if it exists, then assert that the cut zone is meaningless, merge
The initial seed point in the region and growing threshold to adjacent average gray area-of-interest；According to new seed point and growth threshold
Value regrows, until there is no meaningless cut zone or reaching highest the number of iterations, finally returns that comprising multiclass mark
The segmented image of label.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811177386.8A CN109509199B (en) | 2018-10-10 | 2018-10-10 | Medical image organization intelligent segmentation method based on three-dimensional reconstruction |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811177386.8A CN109509199B (en) | 2018-10-10 | 2018-10-10 | Medical image organization intelligent segmentation method based on three-dimensional reconstruction |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109509199A true CN109509199A (en) | 2019-03-22 |
| CN109509199B CN109509199B (en) | 2021-07-20 |
Family
ID=65746371
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811177386.8A Expired - Fee Related CN109509199B (en) | 2018-10-10 | 2018-10-10 | Medical image organization intelligent segmentation method based on three-dimensional reconstruction |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109509199B (en) |
Cited By (36)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110060268A (en) * | 2019-04-19 | 2019-07-26 | 哈尔滨理工大学 | A kind of 3 d medical images edge extracting method |
| CN110415246A (en) * | 2019-08-06 | 2019-11-05 | 东北大学 | A kind of analysis method of stomach fat ingredient |
| CN110458847A (en) * | 2019-07-05 | 2019-11-15 | 心医国际数字医疗系统(大连)有限公司 | Automatic coronary artery segmentation and center line extraction method based on CTA image |
| CN110544235A (en) * | 2019-07-31 | 2019-12-06 | 华南理工大学 | Flexible circuit board image area identification method based on differential geometry |
| CN110610481A (en) * | 2019-08-06 | 2019-12-24 | 深圳市旭东数字医学影像技术有限公司 | Automatic brain stem segmentation method and system based on medical image |
| CN110706245A (en) * | 2019-10-18 | 2020-01-17 | 合肥工业大学 | Image edge detection method of GM (1,1) prediction model based on accumulation method |
| CN111028258A (en) * | 2019-11-14 | 2020-04-17 | 中国科学院力学研究所 | Self-adaptive threshold value extraction method for large-scale gray level image |
| CN111127636A (en) * | 2019-12-24 | 2020-05-08 | 诸暨市人民医院 | Intelligent desktop-level three-dimensional diagnosis system for complex intra-articular fracture |
| CN111145353A (en) * | 2019-12-27 | 2020-05-12 | 华南理工大学 | Method for generating 3D point cloud through image segmentation and grid characteristic point extraction algorithm |
| CN111369580A (en) * | 2020-03-03 | 2020-07-03 | 深圳市安健科技股份有限公司 | Intelligent segmentation method and terminal for X-ray image |
| CN111402421A (en) * | 2020-03-17 | 2020-07-10 | 上海志唐健康科技有限公司 | Liver three-dimensional reconstruction method and device, computer equipment and storage medium |
| CN111598883A (en) * | 2020-05-20 | 2020-08-28 | 重庆工程职业技术学院 | Calibration label equipment for acquiring cloud data medical image and working method |
| CN111832563A (en) * | 2020-07-17 | 2020-10-27 | 江苏大学附属医院 | Intelligent breast tumor identification method based on ultrasonic image |
| WO2020238817A1 (en) * | 2019-05-24 | 2020-12-03 | Shanghai United Imaging Healthcare Co., Ltd. | Systems and methods for processing x-ray images |
| CN112288678A (en) * | 2019-09-11 | 2021-01-29 | 中北大学 | Pseudo-color display method and system for weld image |
| CN112665619A (en) * | 2020-12-17 | 2021-04-16 | 安徽中元新材料技术有限公司 | Automatic extraction device and method for high-sensitivity sensing mode |
| CN113096116A (en) * | 2021-04-28 | 2021-07-09 | 东南大学 | Automatic detection method for performance of medical ultrasonic equipment based on gray scale image |
| CN113177948A (en) * | 2021-04-15 | 2021-07-27 | 中国海洋大学 | Medical image automatic window adjusting method based on semantic segmentation and application |
| CN113192100A (en) * | 2021-05-24 | 2021-07-30 | 大连大学 | Method for acquiring edge path of key characteristic area of time-sharing overlapped plant image |
| CN113538424A (en) * | 2021-09-15 | 2021-10-22 | 江苏景瑞农业科技发展有限公司 | Wood board joint classification data identification method based on artificial intelligence |
| CN113744849A (en) * | 2021-08-31 | 2021-12-03 | 华中科技大学 | Semi-automatic labeling method and system for CT image |
| CN114066923A (en) * | 2021-11-29 | 2022-02-18 | 长春工业大学 | 3D Otsu threshold segmentation method based on iteration and dimension decomposition |
| CN114299081A (en) * | 2021-12-16 | 2022-04-08 | 北京朗视仪器股份有限公司 | Maxillary sinus CBCT image segmentation method and device, storage medium and electronic equipment |
| CN114549675A (en) * | 2020-11-25 | 2022-05-27 | 天津工业大学 | Bone model reconstruction method based on tomography medical image |
| CN114842038A (en) * | 2022-04-14 | 2022-08-02 | 深圳市医未医疗科技有限公司 | Image data feature automatic generation method and device based on image omics |
| CN114972178A (en) * | 2022-04-08 | 2022-08-30 | 大连理工大学 | Shot blasting surface coverage rate obtaining method |
| CN115100198A (en) * | 2022-08-24 | 2022-09-23 | 江苏思锐装备科技有限公司 | Secondary oil removal intensity control method and system based on oil removal effect evaluation |
| CN115187592A (en) * | 2022-09-08 | 2022-10-14 | 山东奥洛瑞医疗科技有限公司 | Image segmentation method suitable for nuclear magnetic resonance image |
| CN115200797A (en) * | 2022-09-19 | 2022-10-18 | 山东超华环保智能装备有限公司 | Leakage detection system for zero leakage valve |
| CN115345893A (en) * | 2022-09-15 | 2022-11-15 | 吉林大学中日联谊医院 | Ovarian tissue canceration region segmentation method based on image processing |
| CN113764072B (en) * | 2021-05-13 | 2023-04-18 | 腾讯科技（深圳）有限公司 | Medical image reconstruction method, device, equipment and storage medium |
| CN116402816A (en) * | 2023-06-08 | 2023-07-07 | 中国人民解放军海军青岛特勤疗养中心 | Physical examination CT image data management method and system |
| CN116912282A (en) * | 2023-06-05 | 2023-10-20 | 北京纳通医用机器人科技有限公司 | Three-dimensional segmentation method, device, equipment and storage medium based on medical image |
| CN116996673A (en) * | 2023-09-28 | 2023-11-03 | 深圳市铁军智能科技有限公司 | Intelligent cloud management system based on passing in and out management and equipment running state |
| CN118196128A (en) * | 2024-03-20 | 2024-06-14 | 南京筑卫医学科技有限公司 | Finite area growth segmentation method based on adaptive gradient |
| CN118485667A (en) * | 2024-07-15 | 2024-08-13 | 南充市中心医院 | Auxiliary diagnosis method and system based on ultrasonic image intellectualization |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101576997A (en) * | 2009-06-19 | 2009-11-11 | 西安电子科技大学 | Abdominal organ segmentation method based on secondary three-dimensional region growth |
| CN104376551A (en) * | 2014-08-25 | 2015-02-25 | 浙江工业大学 | Color image segmentation method integrating region growth and edge detection |
| CN107346545A (en) * | 2017-05-22 | 2017-11-14 | 沈阳工业大学 | Improved confinement growing method for the segmentation of optic cup image |
-
2018
- 2018-10-10 CN CN201811177386.8A patent/CN109509199B/en not_active Expired - Fee Related
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101576997A (en) * | 2009-06-19 | 2009-11-11 | 西安电子科技大学 | Abdominal organ segmentation method based on secondary three-dimensional region growth |
| CN104376551A (en) * | 2014-08-25 | 2015-02-25 | 浙江工业大学 | Color image segmentation method integrating region growth and edge detection |
| CN107346545A (en) * | 2017-05-22 | 2017-11-14 | 沈阳工业大学 | Improved confinement growing method for the segmentation of optic cup image |
Non-Patent Citations (4)
| Title |
|---|
| GUANGHUA FU ET AL.: "Segmentation of Spinal Canal Region in CT Images using 3D Region Growing Technique", 《2018 INTERNATIONAL CONFERENCE ON INFORMATION AND COMMUNICATION TECHNOLOGY ROBOTICS》 * |
| 张馨: "被动式太赫兹图像增强及目标检测算法研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
| 王光飞 等: "DICOM医学图像的调窗技术及DIB显示", 《医疗装备》 * |
| 韩宝如，林尔敏: "离散小波变换在医学图像中的应用", 《电子测试》 * |
Cited By (58)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110060268A (en) * | 2019-04-19 | 2019-07-26 | 哈尔滨理工大学 | A kind of 3 d medical images edge extracting method |
| WO2020238817A1 (en) * | 2019-05-24 | 2020-12-03 | Shanghai United Imaging Healthcare Co., Ltd. | Systems and methods for processing x-ray images |
| CN110458847A (en) * | 2019-07-05 | 2019-11-15 | 心医国际数字医疗系统(大连)有限公司 | Automatic coronary artery segmentation and center line extraction method based on CTA image |
| CN110458847B (en) * | 2019-07-05 | 2022-05-13 | 心医国际数字医疗系统(大连)有限公司 | Automatic coronary artery segmentation and center line extraction method based on CTA image |
| CN110544235A (en) * | 2019-07-31 | 2019-12-06 | 华南理工大学 | Flexible circuit board image area identification method based on differential geometry |
| CN110544235B (en) * | 2019-07-31 | 2022-12-16 | 华南理工大学 | Flexible circuit board image area identification method based on differential geometry |
| CN110415246A (en) * | 2019-08-06 | 2019-11-05 | 东北大学 | A kind of analysis method of stomach fat ingredient |
| CN110610481A (en) * | 2019-08-06 | 2019-12-24 | 深圳市旭东数字医学影像技术有限公司 | Automatic brain stem segmentation method and system based on medical image |
| CN112288678B (en) * | 2019-09-11 | 2022-05-27 | 中北大学 | Pseudo-color display method and system for welding seam image |
| CN112288678A (en) * | 2019-09-11 | 2021-01-29 | 中北大学 | Pseudo-color display method and system for weld image |
| CN110706245A (en) * | 2019-10-18 | 2020-01-17 | 合肥工业大学 | Image edge detection method of GM (1,1) prediction model based on accumulation method |
| CN111028258A (en) * | 2019-11-14 | 2020-04-17 | 中国科学院力学研究所 | Self-adaptive threshold value extraction method for large-scale gray level image |
| CN111028258B (en) * | 2019-11-14 | 2023-05-16 | 中国科学院力学研究所 | Large-scale gray level image self-adaptive threshold value extraction method |
| CN111127636A (en) * | 2019-12-24 | 2020-05-08 | 诸暨市人民医院 | Intelligent desktop-level three-dimensional diagnosis system for complex intra-articular fracture |
| CN111127636B (en) * | 2019-12-24 | 2023-07-25 | 诸暨市人民医院 | Intelligent complex intra-articular fracture desktop-level three-dimensional diagnosis system |
| CN111145353B (en) * | 2019-12-27 | 2023-05-23 | 华南理工大学 | Method for generating 3D point cloud through image segmentation and grid feature point extraction algorithm |
| CN111145353A (en) * | 2019-12-27 | 2020-05-12 | 华南理工大学 | Method for generating 3D point cloud through image segmentation and grid characteristic point extraction algorithm |
| CN111369580A (en) * | 2020-03-03 | 2020-07-03 | 深圳市安健科技股份有限公司 | Intelligent segmentation method and terminal for X-ray image |
| CN111369580B (en) * | 2020-03-03 | 2023-06-20 | 深圳市安健科技股份有限公司 | Intelligent segmentation method and terminal for X-ray image |
| CN111402421B (en) * | 2020-03-17 | 2021-12-28 | 上海志唐健康科技有限公司 | Liver three-dimensional reconstruction method and device, computer equipment and storage medium |
| CN111402421A (en) * | 2020-03-17 | 2020-07-10 | 上海志唐健康科技有限公司 | Liver three-dimensional reconstruction method and device, computer equipment and storage medium |
| CN111598883A (en) * | 2020-05-20 | 2020-08-28 | 重庆工程职业技术学院 | Calibration label equipment for acquiring cloud data medical image and working method |
| CN111598883B (en) * | 2020-05-20 | 2023-05-26 | 重庆工程职业技术学院 | Calibration label equipment for acquiring cloud data medical images and working method |
| CN111832563A (en) * | 2020-07-17 | 2020-10-27 | 江苏大学附属医院 | Intelligent breast tumor identification method based on ultrasonic image |
| CN114549675A (en) * | 2020-11-25 | 2022-05-27 | 天津工业大学 | Bone model reconstruction method based on tomography medical image |
| CN112665619B (en) * | 2020-12-17 | 2022-05-20 | 安徽中元新材料技术有限公司 | Automatic extraction device and method for high-sensitivity sensing mode |
| CN112665619A (en) * | 2020-12-17 | 2021-04-16 | 安徽中元新材料技术有限公司 | Automatic extraction device and method for high-sensitivity sensing mode |
| CN113177948A (en) * | 2021-04-15 | 2021-07-27 | 中国海洋大学 | Medical image automatic window adjusting method based on semantic segmentation and application |
| CN113177948B (en) * | 2021-04-15 | 2023-02-03 | 中国海洋大学 | Medical image automatic window adjusting method based on semantic segmentation and application |
| CN113096116A (en) * | 2021-04-28 | 2021-07-09 | 东南大学 | Automatic detection method for performance of medical ultrasonic equipment based on gray scale image |
| CN113764072B (en) * | 2021-05-13 | 2023-04-18 | 腾讯科技（深圳）有限公司 | Medical image reconstruction method, device, equipment and storage medium |
| CN113192100B (en) * | 2021-05-24 | 2023-07-28 | 大连大学 | Time-sharing overlapped plant image key feature area edge path acquisition method |
| CN113192100A (en) * | 2021-05-24 | 2021-07-30 | 大连大学 | Method for acquiring edge path of key characteristic area of time-sharing overlapped plant image |
| CN113744849B (en) * | 2021-08-31 | 2024-03-19 | 华中科技大学 | Semi-automatic labeling method and system for CT image |
| CN113744849A (en) * | 2021-08-31 | 2021-12-03 | 华中科技大学 | Semi-automatic labeling method and system for CT image |
| CN113538424A (en) * | 2021-09-15 | 2021-10-22 | 江苏景瑞农业科技发展有限公司 | Wood board joint classification data identification method based on artificial intelligence |
| CN114066923A (en) * | 2021-11-29 | 2022-02-18 | 长春工业大学 | 3D Otsu threshold segmentation method based on iteration and dimension decomposition |
| CN114299081A (en) * | 2021-12-16 | 2022-04-08 | 北京朗视仪器股份有限公司 | Maxillary sinus CBCT image segmentation method and device, storage medium and electronic equipment |
| CN114299081B (en) * | 2021-12-16 | 2023-02-17 | 北京朗视仪器股份有限公司 | Maxillary sinus CBCT image segmentation method, maxillary sinus CBCT image segmentation device, maxillary sinus CBCT storage medium and electronic equipment |
| CN114972178B (en) * | 2022-04-08 | 2024-08-13 | 大连理工大学 | Shot blasting surface coverage rate obtaining method |
| CN114972178A (en) * | 2022-04-08 | 2022-08-30 | 大连理工大学 | Shot blasting surface coverage rate obtaining method |
| CN114842038B (en) * | 2022-04-14 | 2023-05-16 | 深圳市医未医疗科技有限公司 | Image data characteristic automatic generation method and device based on image histology |
| CN114842038A (en) * | 2022-04-14 | 2022-08-02 | 深圳市医未医疗科技有限公司 | Image data feature automatic generation method and device based on image omics |
| CN115100198B (en) * | 2022-08-24 | 2022-12-13 | 江苏思锐装备科技有限公司 | Secondary oil removal intensity control method and system based on oil removal effect evaluation |
| CN115100198A (en) * | 2022-08-24 | 2022-09-23 | 江苏思锐装备科技有限公司 | Secondary oil removal intensity control method and system based on oil removal effect evaluation |
| CN115187592A (en) * | 2022-09-08 | 2022-10-14 | 山东奥洛瑞医疗科技有限公司 | Image segmentation method suitable for nuclear magnetic resonance image |
| CN115187592B (en) * | 2022-09-08 | 2022-12-09 | 山东奥洛瑞医疗科技有限公司 | Image segmentation method suitable for nuclear magnetic resonance image |
| CN115345893A (en) * | 2022-09-15 | 2022-11-15 | 吉林大学中日联谊医院 | Ovarian tissue canceration region segmentation method based on image processing |
| CN115345893B (en) * | 2022-09-15 | 2023-02-21 | 吉林大学中日联谊医院 | Ovarian tissue canceration region segmentation method based on image processing |
| CN115200797B (en) * | 2022-09-19 | 2022-12-16 | 山东超华环保智能装备有限公司 | Leakage detection system for zero leakage valve |
| CN115200797A (en) * | 2022-09-19 | 2022-10-18 | 山东超华环保智能装备有限公司 | Leakage detection system for zero leakage valve |
| CN116912282A (en) * | 2023-06-05 | 2023-10-20 | 北京纳通医用机器人科技有限公司 | Three-dimensional segmentation method, device, equipment and storage medium based on medical image |
| CN116402816A (en) * | 2023-06-08 | 2023-07-07 | 中国人民解放军海军青岛特勤疗养中心 | Physical examination CT image data management method and system |
| CN116402816B (en) * | 2023-06-08 | 2023-08-15 | 中国人民解放军海军青岛特勤疗养中心 | Physical examination CT image data management method and system |
| CN116996673A (en) * | 2023-09-28 | 2023-11-03 | 深圳市铁军智能科技有限公司 | Intelligent cloud management system based on passing in and out management and equipment running state |
| CN116996673B (en) * | 2023-09-28 | 2023-11-24 | 深圳市铁军智能科技有限公司 | Intelligent cloud management system based on passing in and out management and equipment running state |
| CN118196128A (en) * | 2024-03-20 | 2024-06-14 | 南京筑卫医学科技有限公司 | Finite area growth segmentation method based on adaptive gradient |
| CN118485667A (en) * | 2024-07-15 | 2024-08-13 | 南充市中心医院 | Auxiliary diagnosis method and system based on ultrasonic image intellectualization |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109509199B (en) | 2021-07-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109509199A (en) | A kind of medical image Organizational Intelligence dividing method based on three-dimensional reconstruction | |
| CN110503649B (en) | Liver segmentation method based on spatial multi-scale U-net and superpixel correction | |
| CN105844285B (en) | Cucumber disease recognition methods and device based on image information | |
| CN104809723B (en) | The three-dimensional CT image for liver automatic division method of algorithm is cut based on super voxel and figure | |
| US8224089B2 (en) | Information processing device and information processing method, image processing device and image processing method, and computer program | |
| CN101576997B (en) | Abdominal organ segmentation method based on secondary three-dimensional region growth | |
| CN110060235A (en) | A kind of thyroid nodule ultrasonic image division method based on deep learning | |
| CN104933709B (en) | Random walk CT lung tissue image automatic segmentation methods based on prior information | |
| CN104217213B (en) | A kind of medical image multistage sorting technique based on symmetric theory | |
| CN110264479B (en) | Three-dimensional image segmentation method based on random walk and level set | |
| CN109840913A (en) | The method and system of lump segmentation in a kind of mammography X | |
| CN107680110B (en) | Inner ear three-dimensional level set segmentation method based on statistical shape model | |
| CN108537751A (en) | A kind of Thyroid ultrasound image automatic segmentation method based on radial base neural net | |
| CN102930538B (en) | The self-adaptive projection method method that Gauss potential and spatial histogram merge | |
| Panda et al. | Automated brain tumor detection using discriminative clustering based MRI segmentation | |
| Tyagi et al. | Performance comparison and analysis of medical image segmentation techniques | |
| CN103761727A (en) | Robust image segmentation method based on self-adaption Gaussian mixture model | |
| CN109285176B (en) | Brain tissue segmentation method based on regularization graph segmentation | |
| Borzooie et al. | Modified density-based data clustering for interactive liver segmentation | |
| Wisaeng et al. | Brain tumor segmentation using fuzzy otsu threshold morphological algorithm | |
| CN111986216B (en) | RSG liver CT image interactive segmentation algorithm based on neural network improvement | |
| Kaur et al. | Intensity based image segmentation using wavelet analysis and clustering techniques | |
| Wang | Research of image segmentation algorithm based on wavelet transform | |
| Mehena et al. | Medical image segmentation and detection of MR images based on spatial multiple-kernel fuzzy C-means algorithm | |
| Wang et al. | Current situation and review of image segmentation |
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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20210720 |