CN108335304A - A kind of aortic aneurysm dividing method of abdominal CT scan sequence image - Google Patents
A kind of aortic aneurysm dividing method of abdominal CT scan sequence image Download PDFInfo
- Publication number
- CN108335304A CN108335304A CN201810122546.2A CN201810122546A CN108335304A CN 108335304 A CN108335304 A CN 108335304A CN 201810122546 A CN201810122546 A CN 201810122546A CN 108335304 A CN108335304 A CN 108335304A
- Authority
- CN
- China
- Prior art keywords
- aorta
- image
- aortic aneurysm
- abdominal
- scan sequence
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/194—Segmentation; Edge detection involving foreground-background segmentation
-
- 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
- G06T2207/30096—Tumor; Lesion
-
- 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)
- 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 present invention relates to a kind of aortic aneurysm dividing methods of abdominal CT scan sequence image, including five parts:Image preprocessing;Aortic blood tube cavity extracts;Aorta segmentation;Aortic aneurysm is extracted;Three-dimensional reconstruction aortic aneurysm.The sample data that the present invention does not need any known label carries out learning training, applied widely;Intravascular space and aorta profile are extracted using different partitioning algorithms respectively, then obtained aorta removal intravascular space region is obtained into hollow aorta inner and outer wall aneurysm profile, and aorta segmentation is proposed a kind of new to cut algorithm based on figure prior-constrained sequence;Take full advantage of distinctive incidence relation and similitude between CT image sequences;Can be good at the less divided brought situations such as solving aortic aneurysm lesion and over-segmentation influences.The present invention can provide the aortic aneurysm model of 3D printing for in-vitro simulated perform the operation, and facilitate the formulation and planning of operation prediction scheme, reduce the high risk of operation on aorta.
Description
Technical field
The present invention relates to automatic measure on line and biomedical image analysis field, more specifically to a kind of abdominal CT
The aortic aneurysm dividing method of scanning sequence image, more particularly to by based on transmitting prior-constrained semi-automatic active between sequence
Arteries and veins dividing method is particularly applicable to but is not limited to medical image segmentation task dispatching.
Background technology
With the improvement of people ' s living standards with the change of eating habit, cardiovascular morbidity disease class statistics in present
The trend of liter, and aortic aneurysm is a kind of angiocardiopathy more dangerous at present, aortoclasia can be caused to cause extremely when serious
It dies.Blood in aortic aneurysm aorta lumen enters the middle level of aorta wall by the breach of inner membrance by is formed, previous more titles
It is now to rename as aortic aneurysm hemotoncus or aortic aneurysm separation, abbreviation aortic aneurysm for aortic aneurysm aneurysm more.Aortic aneurysm illness
Morbidity is relatively anxious, serious symptom, poor prognosis and peri-operative mortality rate are high, be it is clinical it is common be easy to be failed to pinpoint a disease in diagnosis, the disease of mistaken diagnosis
One of.
Such disease effective solution scheme is based on Endovascular placement holder isolation operative treatment at present, attending physician
In the preoperative it will be clear that the structural information for arteries and veins Endovascular of having the initiative in hands, as aorta inner wall break location, range size and
The clinical information such as quantity.In view of aortic aneurysm in relation to the high risk and complexity performed the operation, based on medical image, (such as electronics calculates
Machine tomoscan:The schemes such as operation prediction scheme formulation CT), surgery planning, in-vitro simulated operation can effectively improve operation
Success rate reduces the work difficulty of surgical risk and medical staff.Realizing that the important prerequisite of these purposes is can be from medicine
Aortic aneurysm is accurately partitioned into image, the solid shape of three-dimensional reconstruction aorta inside cavity aneurysm film is intuitive to present
Morphosis relationship between aorta inner and outer wall, each angle, comprehensive presentation aortic aneurysm lesion anatomic characteristic.
But existing medical image reconstruction software cannot show the structure of aorta inner cavity aneurysm wall at present, and
Due to aorta form is changeable, branch vessel is numerous, the obscurity boundary of outer wall between angiolithic degeneration, aneurysm hemotoncus and artery,
The factors such as complexity existing for disease itself carry out greatly difficulty to the segmentation band of aortic aneurysm, thus, aorta segmentation is main
The technological difficulties of aneurysm segmentation, be easy to cause aorta over-segmentation and less divided as a result, being brought greatly to aorta segmentation
Challenge.
Invention content
It is an object of the invention to overcome the deficiencies of the prior art and provide one kind capable of accurately extracting aorta vessel
Inner cavity, aorta profile, the aortic aneurysm dividing method of the abdominal CT scan sequence image for row aorta tumor three-dimensional reconstruction of going forward side by side.
Technical scheme is as follows:
A kind of aortic aneurysm dividing method of abdominal CT scan sequence image, steps are as follows:
1) abdominal CT scan sequence image is pre-processed;Wide window and wide position, interception including adjusting image are interested
Image-region, and the image-region to being truncated to carries out resampling to pre-set dimension;
2) pretreated CT scan sequence image is carried using Threshold Segmentation Algorithm and three-dimensional largest connected region method
Take out aortic blood tube cavity;
3) use cuts the automanual aorta segmentation of algorithm progress based on prior-constrained figure is transmitted between sequence;
4) aorta segmentation results are rejected into intravascular space, obtains aortic aneurysm and vascular wall;
5) according to the three-D profile of aortic aneurysm and vascular wall after rejecting intravascular space, three-dimensional aortic aneurysm mould is reconstructed
Type.
Preferably, if the abdominal CT scan sequence image of input is unenhanced phase CT scan sequence image, by step
After rapid 2) processing, step 3) is directly carried out;If the abdominal CT scan sequence image of input is enhancing phase CT scan sequence chart
Picture carries out pixel filling then after step 2) processing to the aortic blood tube cavity in abdominal CT scan sequence image.
Preferably, the method for pixel filling is filled using adjacent pixel value, by target area to be split
Pixel value meet and be uniformly distributed.
Preferably, in step 3), based on being transmitted between sequence, the step of prior-constrained figure cuts algorithm is as follows:
3.1) pre-segmentation of aorta initial slice layer;
3.2) shape prior template is transmitted between sequence;
3.3) it builds and solves the figure based on shape constraining and cut model.
Preferably, in step 3.1), the slice tomographic image of preset condition is chosen as initial slice layer;Use interactive mode
Figure cuts algorithm, and handmarking's aorta target and background region obtains the histogram model of foreground and background, in conjunction with histogram artwork
Type establishes the S-T figures based on pixel, and the global optimum that S-T figures are acquired using max-flow min-cut algorithm is divided, and obtains initial
The aorta segmentation results of layer
It is described as follows preferably, figure to be solved in initial slice layer cuts model:
Wherein, vectorialIndicate the label value of all pixels point in initial slice,Indicate the label value of p-th of pixel in initial slice layer,Indicate that the pixel belongs to foreground, instead
Then belong to background;Area itemEach pixel label value is counted by the histogram model of scene area before and after handmarking
Respectively 0 and 1 probability value, border itemIndicate that neighbor pixel is assigned the penalty term of different labels, control targe
Segmenting edge smoothness;λ0For no negative coefficient weight, border item and area item value ratio in energy function are controlled;
When the figure, which cuts pattern function, reaches minimum, the value vector of all pixels pointFor aorta global optimum point
Cut result.
Preferably, in step 3.2), the mode being registrated based on light stream between sequence transmits aorta shape prior template:It is logical
The Optic flow information crossed between adjacent image obtains the position offset ω of all pixels point, by known aorta profile in adjacent layer
It is registrated to obtain the aorta shape prior template of current slice image in conjunction with coordinate shift amount, is the aorta segmentation of present image
Shape priors are provided, less divided or over-segmentation phenomenon are reduced.
Preferably, shape prior registration formula is as follows:
Shapen(x, y)=An-1(x+μ,y+v);
Wherein, vectorial ShapenIndicate the bianry image model of the aorta shape prior of n-th layer image, x and y indicate the
The pixel point coordinates of n-layer image, μ and ν indicate that the coordinate both horizontally and vertically of corresponding pixel points in n-th layer image is inclined respectively
Shifting amount, An-1For the aorta segmentation results of the (n-1)th adjacent tomographic image.
Preferably, in step 3.3), it is as follows that the figure based on shape constraining cuts model:
Wherein, area item R (An) by the corresponding front and back scape histogram model system of aorta segmentation results of initial slice layer
Meter;Shape prior bound term S (A are addedn), count n-th layer image on all pixels point belong to aorta foreground prior probability it
With;λ1And λ2For no negative coefficient weight;The figure, which is solved, using max-flow min-cut algorithm cuts model, when the model reaches minimum,
VectorFor the aorta segmentation results of n-th layer image.
Preferably, step 3) further, further includes the post-processing of step 3.4) aorta segmentation results:Expand actively
The arteries and veins region number of plies and level set three-dimensional smoothing processing, specially:From initial n0Layer starts, and iterative solution is based on shape between sequence
The figure of constraint cuts model, obtains aorta segmentation profile corresponding with list entries, by open-close operation, level set and most Dalian
Logical region is corrected, and final three-dimensional aorta segmentation results are obtained.
Beneficial effects of the present invention are as follows:
The aortic aneurysm dividing method of abdominal CT scan sequence image of the present invention, does not need any known label
Sample data carries out learning training, applied widely;Intravascular space and aorta profile are calculated using different segmentations respectively
Method is extracted, and obtained aorta removal intravascular space region is then obtained hollow aorta inner and outer wall aneurysm wheel
Exterior feature, and aorta segmentation is proposed a kind of new to cut algorithm based on figure prior-constrained sequence;Take full advantage of CT image sequences
Distinctive incidence relation and similitude between row;The less divided that brings of situations such as can be good at solving aortic aneurysm lesion and excessively
Cut influence.The present invention can provide the aortic aneurysm model of 3D printing for in-vitro simulated perform the operation, facilitate operation prediction scheme formulation and
Planning, reduces the high risk of operation on aorta.
It is of the present invention that algorithm is cut based on figure prior-constrained between sequence, it is only necessary to before artificial simple marking initiation layer
Background obtains initial aorta segmentation results, transmits aorta shape prior template by the incidence relation between sequence, finally asks
Solution cuts model based on the figure that shape prior constrains and obtains global optimum's aorta segmentation results.Relative to traditional figure segmentation method,
The shape priors of addition can be very good to solve to influence caused by the factors such as edge blurry, calcification, lesion.Since CT schemes
As there is each histoorgan of sequence consecutive variations and similitude, each layer of aorta shape prior can pass through contiguous slices
The aorta segmentation contour registration of layer obtains, and need not train the data structure aorta shape template library with label in advance,
It is improved based on priori between sequence about without complicated data collection and training relative to the method learnt based on training sample
Beam figure cuts the applicability of algorithm.
Description of the drawings
Fig. 1 is the overall flow schematic diagram of the aortic aneurysm dividing method of the present invention;
Fig. 2 is the flow diagram that algorithm is cut based on figure prior-constrained between sequence;
Fig. 3 is the result schematic diagram post-processed to aorta profile;
Fig. 4 is the aortic aneurysm model schematic of three-dimensional reconstruction;
Fig. 5 is the result schematic diagram for the aorta segmentation that aorta segmentation method obtains.
Specific implementation mode
The present invention is further described in detail with reference to the accompanying drawings and embodiments.
The present invention can not provide the deficiency of effective aortic aneurysm segmentation to solve the prior art, provide a kind of semi-automatic
The aortic aneurysm dividing method of the abdominal CT scan sequence image of segmentation, does not need the sample data of any known label
Training is practised, intravascular space and aorta profile are extracted using different partitioning algorithms respectively, then remove obtained aorta
Except intravascular space region obtains hollow aorta inner and outer wall aneurysm profile, it is finally based on marching cube
The iso-surface patch algorithm of (Marching cubes) obtains aortic aneurysm surfaces externally and internally structural model, and the 3D printing model carries out external
The analysis of the schemes such as simulation operation.
A kind of aortic aneurysm dividing method of abdominal CT scan sequence image includes mainly image preprocessing;Aortic blood
Tube cavity extracts;Aorta segmentation;Aortic aneurysm is extracted;Three-dimensional reconstruction aortic aneurysm and etc., it is specific as follows:
1) abdominal CT scan sequence image is pre-processed;Including the wide window for adjusting image and wide position, its gray scale is adjusted
To a certain range, interested image-region and the number of plies are intercepted, and the image-region to being truncated to carries out resampling to presetting
Size.
2) pretreated CT scan sequence image is carried using Threshold Segmentation Algorithm and three-dimensional largest connected region method
Take out aortic blood tube cavity.
Due in enhancing phase CT scan sequence image, contrast agent can be with flowing in blood in the blood vessels chamber, intravascular space
With other tissue regions there is apparent difference in brightness, aortic blood tube cavity contrast is enhanced by contrast agent, is had
Go out aortic blood tube cavity profile conducive to by Threshold Segmentation Algorithm and three-dimensional largest connected extracted region.Therefore, it is possible to use
Partitioning algorithm based on threshold value extracts aortic blood tube cavity profile, and higher threshold value brush choosing is arranged and removes its hetero-organization and the back of the body
The information such as scape impurity.Since the angiosomes of its hetero-organization can also influence the extraction of aortic blood tube cavity, using three-dimensional maximum
Connected region can remove the disconnected angiosomes of its hetero-organization, and aortic blood tube cavity wheel is obtained by smoothly equal post-processings
It is wide.
3) use cuts the automanual aorta segmentation of algorithm progress based on prior-constrained figure is transmitted between sequence.
Algorithm, which is cut, based on figure prior-constrained between sequence solves edge blurry, inner wall that aorta segmentation faced and outer
Less divided and over-segmentation problem, key step caused by the factors such as calcification, the aneurysm lesion of wall is as follows:
3.1) pre-segmentation of aorta initial slice layer;
3.2) shape prior template is transmitted between sequence;
3.3) it builds and solves the figure based on shape constraining and cut model.
And following steps can be increased according to specific implementation demand:
In step 3.1) advance row aorta blood vessel pixel filling, the step 3.4) aorta point after step 3.3)
Cut the post-processing of result.
Wherein, if the abdominal CT scan sequence image of input is unenhanced phase CT scan sequence image, pass through step 2)
After processing, step 3) is directly carried out.
If the abdominal CT scan sequence image of input is enhancing phase CT scan sequence image, handled by step 2)
Afterwards, pixel filling is carried out to the aortic blood tube cavity in abdominal CT scan sequence image.Due to the thinner thickness of aorta wall,
Enhancing phase aorta images region area with brighter blood vessel is accounted for smaller, causes target area to be unevenly distributed, shadow
Ring the accuracy of aorta segmentation results;Therefore, brighter intravascular space can be passed through the pixel value of adjacent aorta wall
It is filled, the pixel value of target area to be split is met and is uniformly distributed, improve the segmentation efficiency of algorithm.
In step 3.1), the slice tomographic image of preset condition is chosen as initial slice layer, i.e., by image preprocessing
In CT scan sequence image afterwards, the slicing layer that aorta contour area is larger, edge clear and calcification degree are less is chosen
Image is as initial slice layer.Algorithm is cut using interactive map, artificial a small amount of label aorta target and background region obtains
The histogram model of foreground and background is established the S-T figures based on pixel in conjunction with histogram model, is calculated using max-flow min-cut
Method acquires global optimum's segmentation of S-T figures, obtains the aorta segmentation results of initiation layer
Figure to be solved cuts model and is described as follows in initial slice layer:
Wherein, vectorialIndicate the label value of all pixels point in initial slice,Indicate the label value of p-th of pixel in initial slice layer,Indicate that the pixel belongs to foreground, instead
Then belong to background;Area itemEach pixel label value is counted by the histogram model of scene area before and after handmarking
Respectively 0 and 1 probability value, border itemIndicate that neighbor pixel is assigned the penalty term of different labels, control targe
Segmenting edge smoothness;λ0For no negative coefficient weight, border item and area item value ratio in energy function are controlled;
When the figure, which cuts pattern function, reaches minimum, the value vector of all pixels pointFor aorta global optimum point
Cut result.
In step 3.2), the mode being registrated based on light stream between sequence transmits aorta shape prior template:Due to CT scan
Each histoorgan between sequence image has continuous minor change and similitude, can utilize the adjacent association between image sequence
Relationship;The position offset ω of all pixels point is obtained by the Optic flow information between adjacent image, by known master in adjacent layer
Artery profile combination coordinate shift amount is registrated to obtain the aorta shape prior template of current slice image, is the master of present image
Artery segmentation provides shape priors, reduces since aorta boundaries are fuzzy or missing, aneurysm lesion are caused
Less divided or over-segmentation phenomenon.
Aorta shape as adjacent layer after each layer of aorta segmentation profile combination coordinate shift amount registration is first
It tests, it is as follows that shape prior is registrated formula:
Shapen(x, y)=An-1(x+μ,y+ν);
Wherein, vectorial ShapenIndicate the bianry image model of the aorta shape prior of n-th layer image, x and y indicate the
The pixel point coordinates of n-layer image, μ and ν indicate that the coordinate both horizontally and vertically of corresponding pixel points in n-th layer image is inclined respectively
Shifting amount, An-1For the aorta segmentation results of the (n-1)th adjacent tomographic image.
In step 3.3), it is as follows that the figure based on shape constraining cuts model:
Wherein, area item R (An) by the corresponding front and back scape histogram model system of aorta segmentation results of initial slice layer
Meter;Shape prior bound term S (A are addedn), count n-th layer image on all pixels point belong to aorta foreground prior probability it
With;λ1And λ2For no negative coefficient weight;The figure, which is solved, using max-flow min-cut algorithm cuts model, when the model reaches minimum,
VectorFor the aorta segmentation results of n-th layer image.
Step 3) further includes the post-processing of step 3.4) aorta segmentation results further:Expand aorta regions layer
Number and level set three-dimensional smoothing processing.
Since the CT sequence images of input are tomoscan sectioning images, in arch of aorta section portion due to cross section
The slice relationship of angle can have the case where zones vanishes, lead to three-dimensional reconstruction aortic aneurysm arch of aorta section up and down
It is partially flat, it needs to obtain aorta segmentation knot to the end by expanding the aorta regions number of plies and level set three-dimensional smoothing processing
Fruit.
Specially:From initial n0Layer start, between sequence iteratively solve the figure based on shape constraining cut model, obtain with it is defeated
Enter the corresponding aorta segmentation profile of sequence, by post-processings such as open-close operation, level set and largest connected region amendments, obtains
Final three-dimensional aorta segmentation results.
4) aorta segmentation results are rejected into intravascular space, obtains aortic aneurysm and vascular wall.Aortic aneurysm lesion exists
Between aorta outer wall and inner wall, do not include aortic blood tube cavity, needs to reject the solid aorta profile being partitioned into
Intravascular space part obtains hollow aortic aneurysm, facilitates the stenter to implant of in-vitro simulated operation.
Above-mentioned steps have respectively obtained the extraction of aortic blood tube cavity, aorta profile, and aorta regions include actively
Arteries and veins intravascular space, aortic aneurysm are made of intravascular space film, middle film and outer membrane, then the extraction of aortic aneurysm only needs to extract
Aorta regions removal intravascular space region.
5) according to the three-D profile of aortic aneurysm and vascular wall after rejecting intravascular space, three-dimensional aortic aneurysm mould is reconstructed
Type.The three-dimensional reconstruction of aortic aneurysm is realized using the iso-surface patch algorithm based on marching cube, it is intuitive that aortic aneurysm outer wall is presented
Morphosis relationship between profile and inner wall, each angle, comprehensive presentation aortic aneurysm lesion anatomic characteristic, it is pre- to perform the operation
Case is formulated, surgery planning and in-vitro simulated operation provide corresponding aortic aneurysm model.
Embodiment
The present embodiment is for inputting enhancing phase CT scan sequence image, as shown in Figure 1, steps are as follows:
Step 1, data collection and pretreatment
Step 1.1, data collection
Data are derived from the CT scan sequential image data (enhancing phase) of four aortic aneurysm patients of certain hospital's heart surgical department, own
The interfloor distance of CT scanning sequence images is 0.625~1.25mm, and the sequence number of plies ranging from 257~1025, image size is
512*512, each pixel space size ranging from 0.70~0.75mm in image.
Step 1.2, tonal range interception, the acquisition of interest region and image magnification
It is observed that wide window and wide position are respectively set to 45 and 405, respectively organize discrimination larger in image and possess it greatly
Part details, therefore gradation of image is intercepted in the range, in then pixel value linear transformation to [0,255] space.It carries out
Aorta interest region intercepts, i.e., according to the CT 3-D scanning sequence images of input, analyzes and intercept the approximate region of aorta,
The effective aorta sequence number of plies ranging from 250~380, can remove other inactive area layers and tissue, improve computational efficiency.
Resampling is amplified after the interception image of interest region, is 200*200 by the aorta regions resampling of interception to image size, is expanded
Aorta regions area accounting improves Target Segmentation precision.
Step 2, aortic blood tube cavity is divided
Choosing CT scan sequence enhances phase image, and intravascular space shows the larger area of brightness value due to the influence of contrast agent
Area image, setting threshold value are 0.7 times of image maximum gradation value, and three-dimensional extracts largest connected region and obtains three-dimensional aorta vessel
Inner cavity removes other brighter blood vessels or tissue regions.
Step 3, aorta segmentation
Step 3.1, aortic blood inside pipe wall pixel filling
Choosing CT scan sequence enhances phase image, and aortic blood tube cavity is expanded with the border circular areas that radius is 1
Afterwards, it is filled the region using the pixel value of neighbouring aorta wall, as shown in Figure 2.
Step 3.2, the aorta pre-segmentation of initiation layer
Initial slice layer is selected from the sequence image after filling, cuts algorithm using interactive map, first artificial a small amount of mark
Remember aorta target and background region, obtains the histogram model of foreground and background.The figure established in initiation layer cuts model description
As follows:
The value vector of all pixels pointIt is then aorta global optimum segmentation result, as shown in Figure 2.
Step 3.3, light stream registration transmits shape prior model
The aorta profile of known initiation layerIt is then registrated adjacent to the aorta shape prior of the slicing layer of initiation layer public
Formula is as follows:
Shapen(x, y)=An-1(x+μ,y+v);
CT scan image sequence is in addition to initiation layer, and the aorta shape prior profile of every other image sequence is all by upper
Light stream registration Algorithm iteration is stated to acquire.
Step 3.4, the figure based on shape constraining cuts algorithm segmentation aorta
To removing initial slice layer in the sequence image after filling, each tomographic image cuts algorithm using the figure based on shape constraining
Solve aorta optimum segmentation result.Figure based on shape constraining cuts model and is described as follows:
Wherein, shape prior bound term S (An) be described as follows:
AnIndicate that the vector of n-th layer image all pixels point label in sequence image, P are all pixels in n-th layer image
Point set.Model is cut to the figure based on shape constraining still to solve using max-flow min-cut method, when the model reaches minimum
When, vectorFor the aorta segmentation results of n-th layer image, as shown in Figure 2.
Step 3.5, aorta segmentation results post-process
Since the CT sequence images of input are tomoscan sectioning images, in arch of aorta section portion due to cross section
The slice relationship of angle can have the case where zones vanishes, lead to three-dimensional reconstruction aortic aneurysm arch of aorta section up and down
It is partially flat (as shown in Figure 3), it needs to expand 5 layers or so up and down using level set to the aortic arch area of segmentation.Due to human body
The influence of the factors such as arterial wall lesion and calcification, need by the aorta profile of entire segmentation of sequence image using open-close operation,
Hole is filled and the post-processing operations such as largest connected region, obtains smoother aorta outer wall profile.
Step 4, the extraction of aortic aneurysm and three-dimensional reconstruction
Above-mentioned steps have respectively obtained the extraction of aortic blood tube cavity, aorta profile, and aorta regions include actively
Arteries and veins intravascular space, aortic aneurysm are made of intravascular space film, middle film and outer membrane, then the extraction of aortic aneurysm only needs to extract
Aorta regions removal intravascular space region.Three-dimensional reconstruction aortic aneurysm uses the iso-surface patch algorithm based on marching cube
Realize that the three-dimensional reconstruction of aortic aneurysm, the intuitive morphosis presented between aortic aneurysm outer wall profile and inner wall close (such as Fig. 4 institutes
Show), by the anatomic characteristic of each angle, comprehensive presentation aortic aneurysm lesion, for the formulation of operation prediction scheme, surgery planning and body
Outer simulation operation provides corresponding aortic aneurysm model.
In order to prove the reliability of the invention that cut algorithm based on prior-constrained figure sequence proposed between aorta segmentation, figure
5 show the aorta cross section segmentation result of different patients, the results show that for there are the aortas of multiple regions (in Fig. 3
(a)), arterial wall calcification ((d) and (e) in Fig. 3), smeared out boundary ((b) and (c) in Fig. 3) and aortic aneurysm hemotoncus
((d) in Fig. 3, (e) and (f)) all obtains good segmentation result.
Above-described embodiment is intended merely to illustrate the present invention, and is not used as limitation of the invention.As long as according to this hair
Bright technical spirit is changed above-described embodiment, modification etc. will all be fallen in the scope of the claims of the present invention.
Claims (10)
1. a kind of aortic aneurysm dividing method of abdominal CT scan sequence image, which is characterized in that steps are as follows:
1) abdominal CT scan sequence image is pre-processed;Wide window and wide position, the interested figure of interception including adjusting image
As region, and the image-region to being truncated to carry out resampling to pre-set dimension;
2) pretreated CT scan sequence image is extracted using Threshold Segmentation Algorithm and three-dimensional largest connected region method
Aortic blood tube cavity;
3) use cuts the automanual aorta segmentation of algorithm progress based on prior-constrained figure is transmitted between sequence;
4) aorta segmentation results are rejected into intravascular space, obtains aortic aneurysm and vascular wall;
5) according to the three-D profile of aortic aneurysm and vascular wall after rejecting intravascular space, three-dimensional aortic aneurysm model is reconstructed.
2. the aortic aneurysm dividing method of abdominal CT scan sequence image according to claim 1, which is characterized in that if
The abdominal CT scan sequence image of input is unenhanced phase CT scan sequence image, then after step 2) processing, is directly walked
It is rapid 3);It is right after step 2) processing if the abdominal CT scan sequence image of input is enhancing phase CT scan sequence image
Aortic blood tube cavity in abdominal CT scan sequence image carries out pixel filling.
3. the aortic aneurysm dividing method of abdominal CT scan sequence image according to claim 2, which is characterized in that pixel
The method of filling is filled using neighborhood pixels value, and the pixel value of target area to be split is met and is uniformly distributed.
4. the aortic aneurysm dividing method of abdominal CT scan sequence image according to claim 1, which is characterized in that step
3) in, based on being transmitted between sequence, the step of prior-constrained figure cuts algorithm is as follows:
3.1) pre-segmentation of aorta initial slice layer;
3.2) shape prior template is transmitted between sequence;
3.3) it builds and solves the figure based on shape constraining and cut model.
5. the aortic aneurysm dividing method of abdominal CT scan sequence image according to claim 4, which is characterized in that step
3.1) in, the slice tomographic image of preset condition is chosen as initial slice layer;Algorithm is cut using interactive map, handmarking is actively
Arteries and veins target and background region, obtains the histogram model of foreground and background, and the S-T based on pixel is established in conjunction with histogram model
Figure, the global optimum that S-T figures are acquired using max-flow min-cut algorithm are divided, and the aorta segmentation results of initiation layer are obtained
6. the aortic aneurysm dividing method of abdominal CT scan sequence image according to claim 5, which is characterized in that initial
Figure to be solved cuts model and is described as follows in slicing layer:
Wherein, vectorialIndicate the label value of all pixels point in initial slice,
Indicate the label value of p-th of pixel in initial slice layer,Indicate that the pixel belongs to foreground, it is on the contrary then belong to the back of the body
Scape;Area itemIt is respectively 0 and 1 to count each pixel label value by the histogram model of scene area before and after handmarking
Probability value, border itemIndicate that neighbor pixel is assigned the penalty term of different labels, control targe segmenting edge is flat
Slippery;λ0For no negative coefficient weight, border item and area item value ratio in energy function are controlled;
When the figure, which cuts pattern function, reaches minimum, the value vector of all pixels pointDivide for aorta global optimum and ties
Fruit.
7. the aortic aneurysm dividing method of abdominal CT scan sequence image according to claim 4, which is characterized in that step
3.2) in, the mode being registrated based on light stream between sequence transmits aorta shape prior template:Believed by the light stream between adjacent image
Breath obtains the position offset ω of all pixels point, and known aorta profile combination coordinate shift amount in adjacent layer is matched will definitely
To the aorta shape prior template of current slice image, shape priors are provided for the aorta segmentation of present image, are dropped
Low less divided or over-segmentation phenomenon.
8. the aortic aneurysm dividing method of abdominal CT scan sequence image according to claim 7, which is characterized in that shape
It is as follows that priori is registrated formula:
Shapen(x, y)=An-1(x+μ,y+v);
Wherein, vectorial ShapenIndicate that the bianry image model of the aorta shape prior of n-th layer image, x and y indicate n-th layer figure
The pixel point coordinates of picture, μ and ν indicate the coordinate shift amount both horizontally and vertically of corresponding pixel points in n-th layer image respectively,
An-1For the aorta segmentation results of the (n-1)th adjacent tomographic image.
9. the aortic aneurysm dividing method of abdominal CT scan sequence image according to claim 4, which is characterized in that step
3.3) in, it is as follows that the figure based on shape constraining cuts model:
Wherein, area item R (An) by the corresponding front and back scape histogram model statistics of aorta segmentation results of initial slice layer;Add
Enter shape prior bound term S (An), it counts all pixels point on n-th layer image and belongs to the sum of aorta foreground prior probability;λ1With
λ2For no negative coefficient weight;The figure, which is solved, using max-flow min-cut algorithm cuts model, when the model reaches minimum, vectorFor the aorta segmentation results of n-th layer image.
10. the aortic aneurysm dividing method of abdominal CT scan sequence image according to claim 9, which is characterized in that step
It is rapid 3) further, further include the post-processing of step 3.4) aorta segmentation results:Expand the aorta regions number of plies and level set
Three-dimensional smoothing processing, specially:From initial n0Layer starts, and the figure based on shape constraining is iteratively solved between sequence and cuts model, is obtained
To aorta segmentation profile corresponding with list entries, corrects, obtained most by open-close operation, level set and largest connected region
Whole three-dimensional aorta segmentation results.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810122546.2A CN108335304B (en) | 2018-02-07 | 2018-02-07 | Aortic aneurysm segmentation method of abdominal CT scanning sequence image |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810122546.2A CN108335304B (en) | 2018-02-07 | 2018-02-07 | Aortic aneurysm segmentation method of abdominal CT scanning sequence image |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108335304A true CN108335304A (en) | 2018-07-27 |
CN108335304B CN108335304B (en) | 2021-08-03 |
Family
ID=62928624
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810122546.2A Active CN108335304B (en) | 2018-02-07 | 2018-02-07 | Aortic aneurysm segmentation method of abdominal CT scanning sequence image |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108335304B (en) |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109118501A (en) * | 2018-08-03 | 2019-01-01 | 上海电气集团股份有限公司 | Image processing method and system |
CN109447967A (en) * | 2018-10-26 | 2019-03-08 | 强联智创(北京)科技有限公司 | A kind of dividing method and system of intracranial aneurysm image |
CN110136217A (en) * | 2019-03-28 | 2019-08-16 | 青岛大学附属医院 | CT image for liver enhances processing method and system |
CN111445473A (en) * | 2020-03-31 | 2020-07-24 | 复旦大学 | Vascular membrane accurate segmentation method and system based on intravascular ultrasound image sequence multi-angle reconstruction |
CN111489434A (en) * | 2020-03-18 | 2020-08-04 | 创业慧康科技股份有限公司 | Medical image three-dimensional reconstruction method based on three-dimensional graph cut |
CN111862304A (en) * | 2020-06-30 | 2020-10-30 | 西安增材制造国家研究院有限公司 | Method and device for segmenting inferior vena cava and abdominal aorta based on skeleton guidance |
CN111951277A (en) * | 2020-07-28 | 2020-11-17 | 杭州电子科技大学 | Coronary artery segmentation method based on CTA image |
CN112348860A (en) * | 2020-10-27 | 2021-02-09 | 中国科学院自动化研究所 | Vessel registration method, system and device for endovascular aneurysm surgery |
CN113160265A (en) * | 2021-05-13 | 2021-07-23 | 四川大学华西医院 | Construction method of prediction image for brain corpus callosum segmentation for corpus callosum state evaluation |
CN113223704A (en) * | 2021-05-20 | 2021-08-06 | 吉林大学 | Auxiliary diagnosis method for computed tomography aortic aneurysm based on deep learning |
CN113516624A (en) * | 2021-04-28 | 2021-10-19 | 武汉联影智融医疗科技有限公司 | Determination of puncture forbidden zone, path planning method, surgical system and computer equipment |
CN115919464A (en) * | 2023-03-02 | 2023-04-07 | 四川爱麓智能科技有限公司 | Tumor positioning method, system and device and tumor development prediction method |
CN117438092A (en) * | 2023-12-20 | 2024-01-23 | 杭州脉流科技有限公司 | Intracranial aneurysm rupture risk prediction device, computer device, and storage medium |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102573638A (en) * | 2009-10-13 | 2012-07-11 | 新加坡科技研究局 | A method and system for segmenting a liver object in an image |
CN103544686A (en) * | 2013-10-25 | 2014-01-29 | 天津工业大学 | Method for detecting eye fundus image microaneurysm based on phase equalization |
CN103985123A (en) * | 2014-05-17 | 2014-08-13 | 清华大学深圳研究生院 | Abdominal aortic aneurysm outer boundary segmentation method based on CTA images |
US9358099B2 (en) * | 2009-11-30 | 2016-06-07 | Biflow Medical Ltd. | Method of implanting a stent graft and creating a fenestration therein |
-
2018
- 2018-02-07 CN CN201810122546.2A patent/CN108335304B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102573638A (en) * | 2009-10-13 | 2012-07-11 | 新加坡科技研究局 | A method and system for segmenting a liver object in an image |
US9358099B2 (en) * | 2009-11-30 | 2016-06-07 | Biflow Medical Ltd. | Method of implanting a stent graft and creating a fenestration therein |
CN103544686A (en) * | 2013-10-25 | 2014-01-29 | 天津工业大学 | Method for detecting eye fundus image microaneurysm based on phase equalization |
CN103985123A (en) * | 2014-05-17 | 2014-08-13 | 清华大学深圳研究生院 | Abdominal aortic aneurysm outer boundary segmentation method based on CTA images |
Non-Patent Citations (2)
Title |
---|
KWON S T 等: "Interaction of expanding abdominal aortic aneurysm with surrounding tissue: Retrospective CT image studies", 《JOURNAL OF NATURE & SCIENCE》 * |
刘耀辉 等: "基于MITK的CT序列图像模糊连接度分割算法研究", 《湘南学院学报》 * |
Cited By (18)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109118501A (en) * | 2018-08-03 | 2019-01-01 | 上海电气集团股份有限公司 | Image processing method and system |
CN109447967A (en) * | 2018-10-26 | 2019-03-08 | 强联智创(北京)科技有限公司 | A kind of dividing method and system of intracranial aneurysm image |
CN110136217A (en) * | 2019-03-28 | 2019-08-16 | 青岛大学附属医院 | CT image for liver enhances processing method and system |
CN111489434A (en) * | 2020-03-18 | 2020-08-04 | 创业慧康科技股份有限公司 | Medical image three-dimensional reconstruction method based on three-dimensional graph cut |
CN111445473B (en) * | 2020-03-31 | 2023-04-28 | 复旦大学 | Vascular membrane accurate segmentation method and system based on intravascular ultrasound image sequence multi-angle reconstruction |
CN111445473A (en) * | 2020-03-31 | 2020-07-24 | 复旦大学 | Vascular membrane accurate segmentation method and system based on intravascular ultrasound image sequence multi-angle reconstruction |
CN111862304A (en) * | 2020-06-30 | 2020-10-30 | 西安增材制造国家研究院有限公司 | Method and device for segmenting inferior vena cava and abdominal aorta based on skeleton guidance |
CN111862304B (en) * | 2020-06-30 | 2023-09-05 | 西安增材制造国家研究院有限公司 | Method and device for segmenting inferior vena cava and abdominal aorta based on skeleton guidance |
CN111951277A (en) * | 2020-07-28 | 2020-11-17 | 杭州电子科技大学 | Coronary artery segmentation method based on CTA image |
CN111951277B (en) * | 2020-07-28 | 2024-03-12 | 杭州电子科技大学 | Coronary artery segmentation method based on CTA image |
CN112348860A (en) * | 2020-10-27 | 2021-02-09 | 中国科学院自动化研究所 | Vessel registration method, system and device for endovascular aneurysm surgery |
CN113516624A (en) * | 2021-04-28 | 2021-10-19 | 武汉联影智融医疗科技有限公司 | Determination of puncture forbidden zone, path planning method, surgical system and computer equipment |
CN113160265A (en) * | 2021-05-13 | 2021-07-23 | 四川大学华西医院 | Construction method of prediction image for brain corpus callosum segmentation for corpus callosum state evaluation |
CN113223704A (en) * | 2021-05-20 | 2021-08-06 | 吉林大学 | Auxiliary diagnosis method for computed tomography aortic aneurysm based on deep learning |
CN115919464B (en) * | 2023-03-02 | 2023-06-23 | 四川爱麓智能科技有限公司 | Tumor positioning method, system, device and tumor development prediction method |
CN115919464A (en) * | 2023-03-02 | 2023-04-07 | 四川爱麓智能科技有限公司 | Tumor positioning method, system and device and tumor development prediction method |
CN117438092A (en) * | 2023-12-20 | 2024-01-23 | 杭州脉流科技有限公司 | Intracranial aneurysm rupture risk prediction device, computer device, and storage medium |
CN117438092B (en) * | 2023-12-20 | 2024-03-22 | 杭州脉流科技有限公司 | Intracranial aneurysm rupture risk prediction device, computer device, and storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN108335304B (en) | 2021-08-03 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108335304A (en) | A kind of aortic aneurysm dividing method of abdominal CT scan sequence image | |
CN109035255B (en) | Method for segmenting aorta with interlayer in CT image based on convolutional neural network | |
US11514572B2 (en) | Automatic image segmentation methods and analysis | |
US11783498B2 (en) | Feature-based registration method | |
US6754376B1 (en) | Method for automatic segmentation of medical images | |
CN107230206B (en) | Multi-mode data-based 3D pulmonary nodule segmentation method for hyper-voxel sequence lung image | |
Aykac et al. | Segmentation and analysis of the human airway tree from three-dimensional X-ray CT images | |
CN105118056B (en) | Coronary artery blood vessel extraction method based on three dimensional morphology | |
CN103606152B (en) | DSA vascular image segmentation method based on SIFT feature point clustering and Boolean different operation | |
CN105279759B (en) | The abdominal cavity aortic aneurysm outline dividing method constrained with reference to context information arrowband | |
CN108324300B (en) | Method and apparatus for vessel segmentation | |
CN102968783B (en) | Automatically the method and system of skeleton are split from abdomen image data | |
CN104851108B (en) | Arteria hepatica dividing method based on CT images | |
CN103985122B (en) | The dirtiest extracting method of based on heart CT image | |
CN101373479A (en) | Method and system for searching computer picture of mammary gland x-ray radiography | |
JP2011517986A (en) | Automatic detection and accurate segmentation of abdominal aortic aneurysms | |
CN110448319A (en) | Based on radiography image and blood flow velocity calculation method coronarius | |
CN105139377A (en) | Rapid robustness auto-partitioning method for abdomen computed tomography (CT) sequence image of liver | |
JP2007202957A (en) | Wall motion measuring device and medical image diagnostic apparatus | |
CN107527341A (en) | The processing method and system of angiographic image | |
Fan et al. | Lung nodule detection based on 3D convolutional neural networks | |
CN107595390A (en) | A kind of real-time matching fusion method of ultrasonic image and CT images | |
CN111179298A (en) | CT image-based three-dimensional lung automatic segmentation and left-right lung separation method and system | |
Lyu et al. | Dissected aorta segmentation using convolutional neural networks | |
CN104915989B (en) | Blood vessel three-dimensional dividing method based on CT images |
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 |