CN107154036A - Blood vessel segmentation method and its segmenting system based on sub-pix - Google Patents
Blood vessel segmentation method and its segmenting system based on sub-pix Download PDFInfo
- Publication number
- CN107154036A CN107154036A CN201710184201.5A CN201710184201A CN107154036A CN 107154036 A CN107154036 A CN 107154036A CN 201710184201 A CN201710184201 A CN 201710184201A CN 107154036 A CN107154036 A CN 107154036A
- Authority
- CN
- China
- Prior art keywords
- blood vessel
- segmentation
- sub
- edge
- pix
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10081—Computed x-ray tomography [CT]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30101—Blood vessel; Artery; Vein; Vascular
Landscapes
- Engineering & Computer Science (AREA)
- Quality & Reliability (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Apparatus For Radiation Diagnosis (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a kind of blood vessel segmentation method based on sub-pix, comprise the following steps：S1, the Bmp file sequences that Dicom file sequences are converted to computer general-purpose；S2, generation thick edge testing result；S3, by the thick edge testing result generation chained list record；S4, acquisition edge coordinate；S5, the three-dimensional blood vessel structure of acquisition；S6, using 26 neighborhoods Region growing segmentation is carried out to the three-dimensional blood vessel structure, realize that Interactive Segmentation goes out specified blood vessel structure.Present invention also offers a kind of blood vessel segmentation system based on sub-pix.Compared with correlation technique, the present invention has the advantages that：The more zone phenomenon in the tune window processing of Dicom files is solved, display effect is improved；The problem of threshold value chooses difficult in thick edge detection is solved, and vessel boundary becomes apparent from, and details is more perfect；Blood vessel segmentation can be reached subpixel accuracy；Interactive Ground Split specifies blood vessel structure.
Description
Technical field
The present invention relates to blood vessel segmentation technical field, more particularly to a kind of blood vessel segmentation method based on sub-pix and its
Segmenting system.
Background technology
In recent years, with the fast development of biomedical imaging technology and medical imaging device, the resolution ratio of medical image
Day by day improve so that medical image turns into one of field with the fastest developing speed among medical technology.Therefore, at medical image
Reason technology is of crucial importance in effect, the influence of medical research and clinical practice.Medical image segmentation is image Segmentation Technology in doctor
The important application in field, is an important part in Medical Image Processing, and important clinical reference can be provided to doctor
Value.In medical image segmentation, blood vessel segmentation is always the focus of research.
With the raising of human living standard, cardiovascular and cerebrovascular disease turns into the No.1 cause of the death of the mankind, to the morning of vascular diseases
Phase quantitative Diagnosis and risk assessment improve the passive protective physical fitness of the mankind, play very crucial work to extension mankind's life expectancy
With.X-ray angiogram (X-Ray Angiogram, XRA) is a kind of method that blood vessel is shown by contrast agent.From passing through CT
It is to carry out visualization of 3 d reconstructing blood vessel and the diagnosis of vascular diseases to scan the accurate vascular tree that is partitioned into obtained contrastographic picture
It is also the important leverage for carrying out auxiliary diagnosis and surgical operation with the important prerequisite for the treatment of.It is necessary that CT blood-vessel images carry out segmentation
Seek optimal feasible method according to the characteristics of CT images, to improve the accuracy of diagnosis identification.In view of blood vessel complexity
Morphosis, and situations such as noise present in imaging circumstances, contrast agent attenuation and skewness, cause in image often
The various problems such as blood vessel objective fuzzy and local form be special are deposited, so that can not be to destination object (the particularly tiny end of blood vessel
The tip) accurately split.
Therefore, it is necessary to provide a kind of new blood vessel segmentation method and its segmenting system based on sub-pix come on solving
State problem.
The content of the invention
The technical problem to be solved in the invention is to provide a kind of blood vessel segmentation method based on sub-pix and its segmentation system
System, it can accurately be split to blood vessel.
The invention provides a kind of blood vessel segmentation method based on sub-pix, comprise the following steps：
S1, the Dicom files sequence obtained to CT scan are by high position interception, inverse transformation and adjust window three links of processing
Pre-processed, the Dicom files sequence is converted to the Bmp file sequences of computer general-purpose；
S2, to the Bmp files sequence using Canny algorithms carry out thick edge detection, and generate thick edge detection knot
Really；
S3, thick edge testing result generation chained list recorded, and retain original gradation；
S4, the edge that sub-pix is carried out to the thick edge testing result recorded using Zernike squares position with
Obtain edge coordinate；
S5, according to the edge coordinate angiosomes is filled using unrestrained water filling algorithm, and obtained by Threshold segmentation
Three-dimensional blood vessel structure；
S6, acquisition initial seed point is clicked in the three-dimensional blood vessel structure by mouse, using 26 neighborhoods to described three
Tie up blood vessel structure and carry out Region growing segmentation, realize that Interactive Segmentation goes out specified blood vessel structure.
It is preferred that, in step sl, window processing links are adjusted by transformation for mula
And Center=C/RS-RI,
Input value in Width=W/RS-RI Min=Center-Width, Max=Center+Width is divided into three classes, divides mark
Temp=1024/ (Max-Min) × gm
The intercept and slope of quasi- associated images, respectively Pixel_In≤Min, Pixel_In >=Max and Other, its
In, C is window position, and W is window width, and RS, RI are respectively slope and intercept, gmFor the maximum show value of display, Pixel_In is defeated
The view data original value entered, Pixel_Out is the display gray shade value of output.
It is preferred that, in step s 2, chosen for the threshold value in Canny algorithms, with reference to Otsu algorithms, adaptive obtains
High-low threshold value is taken, the pixel in the non-maxima suppression image N (x, y) that Canny algorithms are obtained is divided into N1、N2And N3Three
Classification, expects according to the gray value of three classifications, defines inter-class variance σ2(k, t), according to formula
Obtain two optimal threshold k*And t*, wherein, N1Represent the non-edge point in non-maxima suppression image N (x, y), N2Represent
Need to judge the point of marginal point, N in non-maxima suppression image N (x, y)3Represent in non-maxima suppression image N (x, y)
Marginal point.
It is preferred that, in step s 4, the 7X7 templates for choosing Zernike squares carry out sub-pixel edge positioning.
It is preferred that, in step s 6, choose performance characteristic in blood vessel is specified by mouse and significantly put and be used as entering
The initial seed point of line direction guiding, regard the gray scale difference between initial seed point and the body image vegetarian refreshments of its neighborhood as new seed
The basis for estimation one of point；By the unit vector between each adjacent initial seed point and initial seed point to its neighborhood point direction
The scalar product of unit vector is used as new seed point basis for estimation two；Interactive mode is realized according to basis for estimation one and basis for estimation two
Blood vessel structure is specified in segmentation.
Present invention also offers a kind of blood vessel segmentation system based on sub-pix, including：
Image sequence pretreatment module, the Dicom files sequence for being obtained to CT scan passes through high position interception, contravariant
Change and adjust window to handle three links to be pre-processed, and the Bmp that the Dicom files sequence is converted into computer general-purpose is literary
Part sequence；
Sub-pixel edge detects blood vessel segmentation module, for using Canny algorithms and Zemike squares to Bmp files sequence
Handled to obtain three-dimensional blood vessel structure；And
Interactive Segmentation specifies blood vessel structure module, for carrying out region growing to three-dimensional blood vessel structure using 26 fields
Segmentation, realizes that Interactive Segmentation goes out specified blood vessel structure.
It is preferred that, the sub-pixel edge detection blood vessel segmentation module includes：
Thick edge detection module, for using Canny algorithms to carrying out thick edge detection to the Bmp files sequence；
Logging modle, the result generation chained list for thick edge to be detected is recorded, and retains original gradation；
Edge coordinate acquisition module, for carrying out sub- picture to the thick edge testing result recorded using Zernike squares
The edge of element positions to obtain edge coordinate；And
Three-dimensional blood vessel structure formation module, for using unrestrained water filling algorithm by angiosomes according to the edge coordinate
Filling, and three-dimensional blood vessel structure is obtained by Threshold segmentation.
Compared with correlation technique, the blood vessel segmentation method based on sub-pix and its segmenting system that the present invention is provided have
Following beneficial effect：
1) the tune window Processing Algorithm that is converted to Dicom files in Bmp files is improved：To conventional linear transformation
Formula is improved, and input value is associated with RS, RI, solves the more zone phenomenon that the pixel value after window processing is adjusted may occur,
Effectively improve display effect；
2) propose that Canny_Zernike moments methods carry out the accurate blood vessel structure of sub-pixel edge detection segmentation, after processing
The improved Canny algorithms of imagery exploitation carry out thick edge detection, the result then detected using Zernike squares to thick edge
Carry out the edge positioning of sub-pixel, can accurate Ground Split blood vessel tiny tip；
3) adaptive Canny operators thick edge detection algorithm is proposed, with reference to the think of of Otsu algorithms (maximum between-cluster variance)
Think, adaptive acquisition high-low threshold value, solve the problem of threshold value chooses difficult, and vessel boundary becomes apparent from, and details is more
It is perfect；
4) angiosomes is specified in segmentation that can be interactive, blood vessel segmentation can be reached into subpixel accuracy.
Brief description of the drawings
Technical scheme in order to illustrate the embodiments of the present invention more clearly, required in being described below to embodiment
The accompanying drawing used is briefly described, and drawings in the following description are only some embodiments of the present invention, general for this area
For logical technical staff, on the premise of not paying creative work, other accompanying drawings can also be obtained according to these accompanying drawings,
Wherein：
Fig. 1 is the structured flowchart of the blood vessel segmentation system of the invention based on sub-pix；
Sub-pixel edge shown in Fig. 2 Fig. 1 detects the structured flowchart of blood vessel segmentation module；
Fig. 3 is the workflow diagram of the blood vessel segmentation system of the invention based on sub-pix；
Fig. 4 is the flow chart of the blood vessel segmentation method of the invention based on sub-pix；
Fig. 5 (a) is without the pretreated original image of image sequence；
Fig. 5 (b) is by the pretreated Bmp images of image sequence；
Fig. 6 (a) is the schematic perspective view of the preferable sub-pixel edge model of the present invention；
Fig. 6 (b) is the floor map of the preferable sub-pixel edge model of the present invention；
Fig. 6 (c) is the Plane Rotation schematic diagram of the preferable sub-pixel edge model of the present invention；
Fig. 7 (a) is whole blood vessel structure figure；
Fig. 7 (b) to whole blood vessel structure shown in Fig. 6 (a) by blood vessel segmentation method of the present invention based on sub-pix and its
The specified blood vessel structure figure that segmenting system is partitioned into.
Embodiment
The technical scheme in the embodiment of the present invention will be clearly and completely described below, it is clear that described reality
Apply a part of embodiment that example is only the present invention, rather than whole embodiments.Based on the embodiment in the present invention, this area
All other embodiment that those of ordinary skill is obtained under the premise of creative work is not made, belongs to guarantor of the present invention
The scope of shield.
Referring to Fig. 1, the invention provides a kind of blood vessel segmentation system 100 based on sub-pix, including image sequence are pre-
Processing module 1, sub-pixel edge detection blood vessel segmentation module 2 and Interactive Segmentation specify blood vessel structure module 3.
Described image sequence pretreatment module 1 be used for CT scan obtain Dicom files sequence by a high position interception,
Inverse transformation and tune window handle three links and pre-processed, and the Dicom files sequence is converted into computer general-purpose
Bmp file sequences.It is improved, conventional linear transformation formula is improved, by input value by exchanging window Processing Algorithm
Associated with RS, RI, solve the more zone phenomenon that the pixel value after window processing is adjusted may occur, effectively improve display effect
Really, it is highlighted the key component of image.
Sub-pixel edge detection blood vessel segmentation module 2 be used for Bmp files sequence using Canny algorithms and
Zemike squares are handled to obtain three-dimensional blood vessel structure, so as to the tiny tip of accurate Ground Split blood vessel.In addition, proposing
Adaptive Canny algorithm thick edge detection algorithms, combine the thought of Otsu algorithms, and adaptive acquisition high-low threshold value is solved
The problem of threshold value chooses difficult, and vessel boundary becomes apparent from, and details is more perfect.
With reference to shown in Fig. 2, the sub-pixel edge detection blood vessel segmentation module 2 is specifically included：
Thick edge detection module 20, for carrying out thick edge detection to the Bmp files sequence using Canny algorithms；
Logging modle 21, the result generation chained list for thick edge to be detected is recorded, and retains original gradation；
Edge coordinate acquisition module 22, it is sub- for being carried out using Zernike squares to the thick edge testing result recorded
The edge of pixel positions to obtain edge coordinate；And
Three-dimensional blood vessel structure formation module 23, for using unrestrained water filling algorithm by area vasculosa according to the edge coordinate
Domain is filled, and obtains three-dimensional blood vessel structure by Threshold segmentation.
The Interactive Segmentation specifies blood vessel structure module 3 to be used to carry out region to three-dimensional blood vessel structure using 26 fields
Growth segmentation, realizes that Interactive Segmentation goes out specified blood vessel structure.
With reference to shown in Fig. 3, the blood vessel segmentation system 100 based on sub-pix that the present invention is provided first passes through the figure
As sequence pretreatment module 1 is pre-processed to image sequence, blood vessel segmentation module is then detected by the sub-pixel edge
2 carry out vessel boundary detection, and specially Pixel-level Canny algorithms coarse positioning edge and sub-pix Zernike squares is accurately positioned side
Edge, so as to obtain three-dimensional blood vessel structure, blood vessel structure module 3 is specified based on interactive mode finally by the Interactive Segmentation
Region growing, which is obtained, specifies blood vessel structure.
With reference to shown in Fig. 4, present invention also offers a kind of according to the above-mentioned blood vessel segmentation system 100 based on sub-pix
Dividing method, the dividing method comprises the following steps：
S1, the Dicom files sequence obtained to CT scan are by high position interception, inverse transformation and adjust window three links of processing
Pre-processed, the Dicom files sequence is converted to the Bmp file sequences of computer general-purpose；
S2, to the Bmp files sequence using Canny algorithms carry out thick edge detection, and generate thick edge detection knot
Really；
S3, thick edge testing result generation chained list recorded, and retain original gradation；
S4, the edge that sub-pix is carried out to the thick edge testing result recorded using Zernike squares position with
Obtain edge coordinate；
S5, according to the edge coordinate angiosomes is filled using unrestrained water filling algorithm, and obtained by Threshold segmentation
Three-dimensional blood vessel structure；
S6, acquisition initial seed point is clicked in the three-dimensional blood vessel structure by mouse, using 26 neighborhoods to described three
Tie up blood vessel structure and carry out Region growing segmentation, realize that Interactive Segmentation goes out specified blood vessel structure.
In step sl, Dicom file sequences are converted into Bmp files sequence mainly to comprise the following steps：
(1) Dicom file sequences are read, and preserve image data information (number of image frames, figure necessary to image is shown
Image width degree, picture altitude, distribution digit (Bits Allocated), storage bit number (Bits Stored), highest digit (High
Bit) and image window position (WindowsCenter), window width (WindowWidth), intercept (RescaleIntercept), tiltedly
Rate (RescaleSlope) etc.) and parameter information；
(2) image real time transfer, including high-order interception, inverse transformation and tune window handle three links；
(3) Bmp fileinfos are filled, the view data and parameter after processing are sequentially filled simultaneously according to Bmp file formats
Display.
So-called tune window processing (C/W), i.e. window width and window level adjust, and are window width and window place value according to precognition in principle, obtain
The window size (window width W) and center (window position C) of display are needed, so that when the value in window is converted into showing most
Value in bright and most dark scope, the gray value that will be above window upper limit is set to display gray scale peak, less than lower window edge
Gray value be set to gray scale minimum, changing window width and window level value moving window by dynamic shows different Gray Level Segmentss.
After window processing is adjusted, more zone phenomenon may occur for pixel value, and (pixel value after handling is beyond actual pixel
It is worth scope), the present invention is directed to such a situation, and conventional linear transformation formula is improved, and input value and RS, RI are closed
Connection, effectively improves the display effect of original algorithm, formula is as follows：
With
Center=C/RS-RI, Width=W/RS-RI
Min=Center-Width, Max=Center+Width
Temp=1024/ (Max-Min) × gm
Adjust window processing links that the input value in above-mentioned transformation for mula is divided into three classes, section of criteria for classifying associated images
Away from and slope, respectively Pixel_In≤Min, Pixel_In >=Max and Other, wherein, C be window position, W is window width, RS, RI
Respectively slope and intercept, gmFor the maximum show value of display, Pixel_In is the view data original value of input, Pixel_
Out is the display gray shade value of output.As shown in Fig. 5 (a) and Fig. 5 (b), three rings are handled by interception, inverse transformation and tune window
Save the contrast of the image and the image without pretreatment that are pre-processed, it is seen then that after pretreatment, effectively improve aobvious
Show effect, and pixel value does not cross the border, and is highlighted the key component of image.
In step s 2, chosen for the threshold value in Canny algorithms, with reference to Otsu algorithms, adaptive acquisition height threshold
Pixel in value, the non-maxima suppression image N (x, y) that Canny algorithms are obtained is divided into N1, N2And N3Three classifications, its
Middle N1In comprising gray value be { m1,m2,...,mkPixel, represent the non-edge point in artwork；N2In include gray value
For { mk+1,mk+2,...,mt, represent needs the point for judging marginal point in artwork；N3In comprising gray value be { mt+1,
mt+2,...,ml, represent the marginal point in artwork.In addition, setting pixel count total in image as N, gray value is miIt is corresponding
Number of pixels is ni, then its probability isThen whole interval gray value is desired forIn N1, N2, N3Gray value in three classes is expected：
Order
Inter-class variance, which can then be defined, is：
σ2(k, t)=p1(k)·(e1(k)-E)2+p2(k,t)·(e2(k,t)-E)2+p3(t)·(e3(t)-E)2
And following relation is set up：
p1(k)·e1(k)+p2(k,t)·e2(k,t)+p3(t)·e3(t)=E and p1(k)+p2(k,t)+p3(t)=1
Two optimal threshold k*And t*It is so that σ2(k, t) maximum value, optimal threshold is found with following formula：
(value is 1 to the process, because searching threshold value is nonsensical at 0 gray value, is increased by selecting first k value
Value is integer).Then, the t ownership increases in the range of more than k and less than l.Then, that k is increased into its is next
Value, t all values increase in the range of all values more than k again.The process is repeated, untill k=l-2.The processing
Result is a two-dimensional array σ2(k, t), final step is maximizing in the array.Corresponding to maximum k values and
T values are exactly optimal threshold k*And t*.If there is several maximums, then the value corresponding to k and t is averaged to obtain final threshold
Value.
In step s 4, the 7X7 templates for choosing Zernike squares carry out sub-pixel edge positioning.Image f (x, y) n ranks m
Secondary Zernike squares[ii]It is defined as：
If image rotationAngle, then the Zernike squares Z ' before rotatingnmWith the Zernike squares Z after rotationnmBetween have
Following relation：
As can be seen that simply phase angle is changed before and after image rotation from formula, and the modular invariance of Zernike squares, this
Individual is exactly the rotational invariance of Zernike squares.It can be easy to calculate edge parameters using postrotational Zernike squares, from
And realize the sub-pixel positioning of contrast fringes.
Edge detection algorithm principle based on Zernike squares is by setting up Zernike squares and preferable sub-pix model
4 edge parameters relation, solve square respectively and obtain 4 edge parameters of iconic model, then by parameter and default threshold
Value is compared judgement, and then is accurately positioned marginal point.Shown in preferable sub-pixel edge model such as Fig. 6 (a)~(c), wherein, k
For Gray step height, h is background gray scale, and the distance at edge is arrived centered on l (l ∈ [- 1,1]),For the folder between l and x-axis
Angle.Dotted line is ideal step edge in unit circle, and the gray scale of straight line both sides is respectively target (h+k) and background (k).Fig. 6 (c)
It is that Fig. 6 (b) turns clockwiseResult.
Model in Fig. 6 (a)~(c), calculates the anglec of rotationThe Zernike squares of different orders are expressed as afterwards：
3 parameters at edge can be obtained by above-mentioned formula, are expressed as：
Im[Z′11] it is odd function on y, thereforeTherefore
Under discrete digital image conditions, calculating for Zernike squares can be using template and the convolution of gradation of image.This
Text considers template effect, is sampled according to unit circle, in n × n pixel regions, and when template, movement is entered on image
During row convolution, template covering is n around template center2Individual pixel, now unit radius become in order toTherefore need unit
The vertical range l amplifications calculated on circleTimes, therefore the subpixel coordinates of pixel are：
Although Zernike square template n values are bigger, sub-pixel edge positioning precision is higher, also increases amount of calculation,
Therefore 7 × 7 templates of currently preferred selection Zernike square operators carry out sub-pixel edge positioning.
Adaptive threshold rim detection is being carried out using improved Canny algorithms to whole image, and thick edge is being detected
Result generation chained list record, retain original gradation after, obtained new images, using Zernike squares to being stored in array
In possible edge dot image relocation clock as edge, step is as follows：
(1) 7 × 7 coefficients are calculated；
(2) convolution is asked to obtain the Z of Zernike squares each pixel and coefficients00,Z11And Z20；
(3) according to above-mentioned formulaCalculate the anglec of rotation
(4) according to above-mentioned formulaAnd
Calculate parameter l and k；
(5) basis for estimation of sub-pixel edge, i.e. k >=k are determinedt∩l≤lt(kt,ltIt is used as judgment threshold, ktFor k matrixes
Take obtained by average, ltAverage is taken for l matrixesObtained by times), if pixel meets the condition, the pixel is marginal point,
Then above-mentioned formula is utilizedCalculate subpixel coordinates.Finally, according to the edge coordinate of acquisition
Angiosomes is filled using unrestrained water filling algorithm, three-dimensional blood vessel structure is then obtained by Threshold segmentation.
In step s 6, the selection on initial seed point：User clicks on acquisition by mouse in three-dimensional blood vessel structure
Several " key points ", in general specify a small amount of key point, in vascular bending area on the relative angiosomes for tending to straight line
More key point is specified in domain, to improve the accuracy of segmentation.These " key points ", as initial seed point, seed point is in three-dimensional
In the case of be referred to as voxel.An empty queue is set up simultaneously, by seed point press-in wherein, then queue is initialized to A1,A2,
...An, wherein A1,A2,...AnRepresent initial seed point or the seed region chosen.It is M × N pictures for a tomography size
The 3-D view of element, and initial seed point AiTo Ai+1In z-direction it is common H layers section, its volume representation be V=f (i, j,
H) | i=1,2 ..., M；J=1,2 ..., N；H=1,2 ... H }.
Performance characteristic is chosen by mouse in blood vessel is specified significantly to put as the initial kind guided for travel direction
It is sub-, using the gray scale difference between initial seed point and the body image vegetarian refreshments of its neighborhood as new seed point basis for estimation one；Will be each
Unit vector and the scalar product conduct of initial seed point to the unit vector in its neighborhood point direction between adjacent initial seed point
New seed point basis for estimation two；Realize that Interactive Segmentation specifies blood vessel structure according to basis for estimation one and basis for estimation two.
Iterative process：From seed point, study not yet treated body image vegetarian refreshments T.T in its neighborhood and be represented byWherein N (x) represents tissue points x neighborhood.For 3-D view, its neighborhood type point
For 6 neighborhoods, 18 neighborhoods and 26 neighborhoods.
Currently preferred that data are handled using 26 neighborhoods, the similitude between seed point and its neighborhood judges
According to being | G (p)-G (q) |<τ, wherein, G (p) represents the gray value at seed point p, and G (q) represents a certain neighborhood point q of seed point
The gray value at place, τ is selected threshold value.
In addition, seed point AiSpace coordinate be (xi,yi,zi), seed point Ai+1Space coordinate be (xi+1,yi+1,
zi+1), by seed point AiTo Ai+1The unit vector in direction is defined as vectorSeed point AiTo Ai+1On direction in H layers of section
Any seed point is defined as vector to the unit vector in not yet treated body image vegetarian refreshments T directions in its neighborhoodBy two to
The scalar product of amountAs the criterion for obtaining designated area, wherein δ is selected threshold value.I.e. basis for estimation is as follows
When test point meets above-mentioned basis for estimation formula, it is included into the point as new seed point in queue.With iteration
Progress, when more new seed points are not produced, then region growing terminates, and all pixels are constituted in current queue
Region be exactly segmentation result.
The present invention obtains specifying the tomograph of blood vessel, such as using the blood vessel CT data of the Anterolateral femoral of scanning
Shown in Fig. 7 (a) and (b), Fig. 7 (a) is whole blood vessel structure, and Fig. 7 (b) is the blood vessel structure specified, it is seen then that the present invention can be with
Accurately it is partitioned into the blood vessel structure specified.
Compared with correlation technique, the blood vessel segmentation method based on sub-pix and its segmenting system that the present invention is provided have
Following beneficial effect：
1) the tune window Processing Algorithm that is converted to Dicom files in Bmp files is improved：To conventional linear transformation
Formula is improved, and input value is associated with RS, RI, solves the more zone phenomenon that the pixel value after window processing is adjusted may occur,
Effectively improve display effect；
2) propose that Canny_Zernike moments methods carry out the accurate blood vessel structure of sub-pixel edge detection segmentation, after processing
The improved Canny algorithms of imagery exploitation carry out thick edge detection, the result then detected using Zernike squares to thick edge
Carry out the edge positioning of sub-pixel, can accurate Ground Split blood vessel tiny tip；
3) adaptive Canny operators thick edge detection algorithm is proposed, with reference to the think of of Otsu algorithms (maximum between-cluster variance)
Think, adaptive acquisition high-low threshold value, solve the problem of threshold value chooses difficult, and vessel boundary becomes apparent from, and details is more
It is perfect；
4) angiosomes is specified in segmentation that can be interactive, blood vessel segmentation can be reached into subpixel accuracy.
Embodiments of the invention are the foregoing is only, are not intended to limit the scope of the invention, it is every to utilize this hair
Equivalent structure or equivalent flow conversion that bright specification and accompanying drawing content are made, or be directly or indirectly used in other related
Technical field, is included within the scope of the present invention.
Claims (7)
1. a kind of blood vessel segmentation method based on sub-pix, it is characterised in that comprise the following steps：
S1, the Dicom files sequence obtained to CT scan are carried out pre- by high position interception, inverse transformation and tune window three links of processing
Processing, the Dicom files sequence is converted to the Bmp file sequences of computer general-purpose；
S2, using Canny algorithms thick edge detection is carried out to the Bmp files sequence, and generate thick edge testing result；
S3, thick edge testing result generation chained list recorded, and retain original gradation；
S4, the edge for carrying out sub-pix to the thick edge testing result recorded using Zernike squares are positioned to obtain
Edge coordinate；
S5, according to the edge coordinate angiosomes is filled using unrestrained water filling algorithm, and pass through Threshold segmentation and obtain three-dimensional
Blood vessel structure；
S6, acquisition initial seed point is clicked in the three-dimensional blood vessel structure by mouse, using 26 neighborhoods to the three-dimensional blood
Tubular construction carries out Region growing segmentation, realizes that Interactive Segmentation goes out specified blood vessel structure.
2. the blood vessel segmentation method according to claim 1 based on sub-pix, it is characterised in that in step sl, adjusts window
Processing links are by transformation for mulaWith
Input value in Center=C/RS-RI, Width=W/RS-RIMin=Center-Width, Max=Center+Width is drawn
It is divided into three classes, criteria for classifying Temp=1024/ (Max-Min) × gmThe intercept and slope of associated images, respectively Pixel_In
≤ Min, Pixel_In >=Max and Other, wherein, C is window position, and W is window width, and RS, RI are respectively slope and intercept, gmIt is aobvious
Show the maximum show value of device, Pixel_In is the view data original value of input, and Pixel_Out is the display gray shade value of output.
3. the blood vessel segmentation method according to claim 1 or 2 based on sub-pix, it is characterised in that in step s 2, right
Threshold value in Canny algorithms is chosen, with reference to Otsu algorithms, adaptive acquisition high-low threshold value, by Canny algorithms obtain it is non-
The pixel that maximum suppresses in image N (x, y) is divided into N1、N2And N3Three classifications, expect according to the gray value of three classifications,
Define inter-class variance σ2(k, t), according to formulaObtain two optimal threshold k*And t*, wherein, N1
Represent the non-edge point in non-maxima suppression image N (x, y), N2Represent needs to judge in non-maxima suppression image N (x, y)
The point of marginal point, N3Represent the marginal point in non-maxima suppression image N (x, y).
4. the blood vessel segmentation method according to claim 1 based on sub-pix, it is characterised in that in step s 4, chooses
The 7X7 templates of Zernike squares carry out sub-pixel edge positioning.
5. the blood vessel segmentation method according to claim 1 based on sub-pix, it is characterised in that in step s 6, pass through
Mouse is chosen performance characteristic in blood vessel is specified and significantly put as the initial seed point guided for travel direction, will initially plant
Son selects the gray scale difference between the body image vegetarian refreshments of its neighborhood as the basis for estimation one of new seed point；By each adjacent initial seed point
Between unit vector and initial seed point to the unit vector in its neighborhood point direction scalar product as new seed point judge according to
According to two；Realize that Interactive Segmentation specifies blood vessel structure according to basis for estimation one and basis for estimation two.
6. a kind of blood vessel segmentation system based on sub-pix, it is characterised in that including：
Image sequence pretreatment module, the Dicom files sequence for being obtained to CT scan passes through high position interception, inverse transformation and tune
Window handles three links and pre-processed, and the Dicom files sequence is converted to the Bmp file sequences of computer general-purpose；
Sub-pixel edge detects blood vessel segmentation module, to Bmp files sequence using Canny algorithms and Zemike squares
Manage to obtain three-dimensional blood vessel structure；And
Interactive Segmentation specifies blood vessel structure module, for carrying out Region growing segmentation to three-dimensional blood vessel structure using 26 fields,
Realize that Interactive Segmentation goes out specified blood vessel structure.
7. the segmenting system of the blood vessel segmentation method according to claim 6 based on sub-pix, it is characterised in that the Asia
Pixel edge detection blood vessel segmentation module includes：
Thick edge detection module, for using Canny algorithms to carrying out thick edge detection to the Bmp files sequence；
Logging modle, the result generation chained list for thick edge to be detected is recorded, and retains original gradation；
Edge coordinate acquisition module, for carrying out sub-pix to the thick edge testing result recorded using Zernike squares
Edge positions to obtain edge coordinate；And
Three-dimensional blood vessel structure formation module, for angiosomes to be filled using unrestrained water filling algorithm according to the edge coordinate,
And three-dimensional blood vessel structure is obtained by Threshold segmentation.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710184201.5A CN107154036A (en) | 2017-03-24 | 2017-03-24 | Blood vessel segmentation method and its segmenting system based on sub-pix |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710184201.5A CN107154036A (en) | 2017-03-24 | 2017-03-24 | Blood vessel segmentation method and its segmenting system based on sub-pix |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107154036A true CN107154036A (en) | 2017-09-12 |
Family
ID=59791908
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710184201.5A Pending CN107154036A (en) | 2017-03-24 | 2017-03-24 | Blood vessel segmentation method and its segmenting system based on sub-pix |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107154036A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107705303A (en) * | 2017-10-16 | 2018-02-16 | 长沙乐成医疗科技有限公司 | The dividing method of blood vessel on a kind of medical image |
| CN108629773A (en) * | 2018-05-10 | 2018-10-09 | 北京红云智胜科技有限公司 | The method for establishing the convolutional neural networks data set of training identification cardiovascular type |
| CN108765433A (en) * | 2018-05-31 | 2018-11-06 | 西京学院 | One kind is for carrying high-precision leafy area measurement method |
| CN108898578A (en) * | 2018-05-29 | 2018-11-27 | 杭州晟视科技有限公司 | A kind of processing method of medical image, device and computer storage medium |
| CN109166124A (en) * | 2018-11-20 | 2019-01-08 | 中南大学 | A kind of retinal vascular morphologies quantization method based on connected region |
| CN109410191A (en) * | 2018-10-18 | 2019-03-01 | 中南大学 | Optical fundus blood vessel localization method and its anaemia screening method based on OCT image |
| CN112734784A (en) * | 2021-01-28 | 2021-04-30 | 依未科技(北京)有限公司 | High-precision fundus blood vessel boundary determining method, device, medium and equipment |
| CN113470102A (en) * | 2021-06-23 | 2021-10-01 | 依未科技(北京)有限公司 | Method, device, medium and equipment for measuring fundus blood vessel curvature with high precision |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101474083A (en) * | 2009-01-15 | 2009-07-08 | 西安交通大学 | System and method for super-resolution imaging and multi-parameter detection of vascular mechanical characteristic |
| CN103400365A (en) * | 2013-06-26 | 2013-11-20 | 成都金盘电子科大多媒体技术有限公司 | Automatic segmentation method for lung-area CT (Computed Tomography) sequence |
| US8612890B2 (en) * | 2007-12-14 | 2013-12-17 | Koninklijke Philips N.V. | Labeling a segmented object |
-
2017
- 2017-03-24 CN CN201710184201.5A patent/CN107154036A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8612890B2 (en) * | 2007-12-14 | 2013-12-17 | Koninklijke Philips N.V. | Labeling a segmented object |
| CN101474083A (en) * | 2009-01-15 | 2009-07-08 | 西安交通大学 | System and method for super-resolution imaging and multi-parameter detection of vascular mechanical characteristic |
| CN103400365A (en) * | 2013-06-26 | 2013-11-20 | 成都金盘电子科大多媒体技术有限公司 | Automatic segmentation method for lung-area CT (Computed Tomography) sequence |
Non-Patent Citations (3)
| Title |
|---|
| 吴平: "DICOM医学图像格式转换与肺结节检测方法研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
| 周琳: "眼底图像中血管分割技术研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
| 唐路路等: "一种自适应阈值的Canny边缘检测算法", 《光电工程》 * |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107705303A (en) * | 2017-10-16 | 2018-02-16 | 长沙乐成医疗科技有限公司 | The dividing method of blood vessel on a kind of medical image |
| CN108629773A (en) * | 2018-05-10 | 2018-10-09 | 北京红云智胜科技有限公司 | The method for establishing the convolutional neural networks data set of training identification cardiovascular type |
| CN108898578A (en) * | 2018-05-29 | 2018-11-27 | 杭州晟视科技有限公司 | A kind of processing method of medical image, device and computer storage medium |
| CN108898578B (en) * | 2018-05-29 | 2020-12-08 | 杭州晟视科技有限公司 | Medical image processing method and device and computer storage medium |
| CN108765433A (en) * | 2018-05-31 | 2018-11-06 | 西京学院 | One kind is for carrying high-precision leafy area measurement method |
| CN109410191A (en) * | 2018-10-18 | 2019-03-01 | 中南大学 | Optical fundus blood vessel localization method and its anaemia screening method based on OCT image |
| CN109410191B (en) * | 2018-10-18 | 2022-03-25 | 中南大学 | OCT (optical coherence tomography) image-based fundus blood vessel positioning method and anemia screening method thereof |
| CN109166124A (en) * | 2018-11-20 | 2019-01-08 | 中南大学 | A kind of retinal vascular morphologies quantization method based on connected region |
| CN109166124B (en) * | 2018-11-20 | 2021-12-14 | 中南大学 | Retinal blood vessel morphology quantification method based on connected region |
| CN112734784A (en) * | 2021-01-28 | 2021-04-30 | 依未科技(北京)有限公司 | High-precision fundus blood vessel boundary determining method, device, medium and equipment |
| CN113470102A (en) * | 2021-06-23 | 2021-10-01 | 依未科技(北京)有限公司 | Method, device, medium and equipment for measuring fundus blood vessel curvature with high precision |
| CN113470102B (en) * | 2021-06-23 | 2024-06-11 | 依未科技(北京)有限公司 | Method, device, medium and equipment for measuring fundus blood vessel curvature with high precision |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107154036A (en) | Blood vessel segmentation method and its segmenting system based on sub-pix | |
| CN108268870B (en) | Multi-scale feature fusion ultrasonic image semantic segmentation method based on counterstudy | |
| CN111047572B (en) | Automatic spine positioning method in medical image based on Mask RCNN | |
| CN106340021B (en) | Blood vessel extraction method | |
| CN109509197B (en) | Method, device and equipment for dividing region of interest and storage medium | |
| CN102068281B (en) | Processing method for space-occupying lesion ultrasonic images | |
| CN109215032A (en) | The method and system of image segmentation | |
| CN107492071A (en) | Medical image processing method and equipment | |
| CN112308822B (en) | Intervertebral disc CT image detection method based on deep convolutional neural network | |
| CN111784628A (en) | End-to-end colorectal polyp image segmentation method based on effective learning | |
| CN110110808B (en) | Method and device for performing target labeling on image and computer recording medium | |
| CN111899165A (en) | Multi-task image reconstruction convolution network model based on functional module | |
| CN102525529B (en) | Medical image-processing apparatus and medical image processing method | |
| CN112381811B (en) | Method, device and equipment for realizing medical image data annotation | |
| CN112750137B (en) | Liver tumor segmentation method and system based on deep learning | |
| Pandey et al. | Segmentation of liver lesions with reduced complexity deep models | |
| CN113628230A (en) | Ventricular myocardium segmentation model training method, segmentation method and device in cardiac nuclear magnetic resonance image | |
| CN112116989A (en) | Multi-organ sketching method and device | |
| CN111127487B (en) | Real-time multi-tissue medical image segmentation method | |
| CN107146211A (en) | Retinal vascular images noise-reduction method based on line spread function and bilateral filtering | |
| CN104915989A (en) | CT image-based blood vessel three-dimensional segmentation method | |
| AU2019204365B1 (en) | Method and System for Image Segmentation and Identification | |
| CN115937163A (en) | Target region extraction method and system for SPECT lung perfusion imaging | |
| Mouzai et al. | Xray-Net: Self-supervised pixel stretching approach to improve low-contrast medical imaging | |
| CN106934785B (en) | Medical image segmentation method for liver model in robot virtual training system |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170912 |