CN103985122B - The dirtiest extracting method of based on heart CT image - Google Patents

Abstract

A kind of the dirtiest extracting method of based on cardiac CT image, it comprises the following steps: obtain cardiac CT image；Lung tissue, descending aorta tissue, chest wall tissue and the vertebral tissue removed in cardiac CT image obtain intermediate image；And remove the noise tissue in described intermediate image.The inventive method is from reverse angle, realize extracting the dirtiest purpose by progressively removing the non-cardiac tissue leaving a volume such as chest wall, pulmonary, vertebra and descending aorta, it has the advantages such as adaptivity is strong, operational efficiency is high, extraction effect is accurate, can quickly be embedded in existing medical network, it is achieved remote assistant diagnoses.

Description

The dirtiest extracting method of based on cardiac CT image

Technical field

The invention belongs to computer vision and image processing field, a kind of the dirtiest extraction side of based on cardiac CT image Method.

Background technology

The dirtiest being extracted in various fields has important using value, such as in cardiovascular disease diagnoses, it is provided that crown The three-dimensional visualization of tremulous pulse；When lung tumors radiotherapy planning is formulated, adjuvant radiotherapy doctor delimit target of prophylactic radiotherapy, avoid the health such as heart Tissue；And to chamber individually diagnosis etc. such as the dirtiest ventricle, atrium.

Currently for the method in the research employing extracting directly cardiac objects region mostly that the heart of CT image extracts, mainly utilize After injection contrast agent, the high-contrast of blood pool and its hetero-organization realizes in heart chamber.Method is roughly divided into two classes, and a class is to pass The Combinatorial Optimization of system two dimensional image dividing method, the commonly used K of having mean cluster, figure cut, ACM/ASM and fuzzy set Theoretical etc.；Another kind of is method based on three-dimensional statistical model, by a large amount of manual segmentation data are carried out statistical analysis, sets up Complete full heart model, then utilizes Model Matching to realize extracting the dirtiest.

Above-mentioned two class methods are all from problem, directly with cardiac objects region as object of study, it is desirable to utilize various side Method extracts whole heart from whole CT faultage image, but often causes heart edge tissues to lose or residual noise group The problem such as knit.

Summary of the invention

It is an object of the invention to for a kind of based on cardiac CT image the dirtiest new extracting method.

The technical solution used in the present invention is:

A kind of the dirtiest extracting method of based on cardiac CT image, comprises the following steps:

Obtain cardiac CT image；

Lung tissue, descending aorta tissue, chest wall tissue and the vertebral tissue removed in cardiac CT image obtain intermediate image； And

Remove the noise tissue in described intermediate image.

In based on cardiac CT image the dirtiest above-mentioned extracting method, it is preferable that the method removing lung tissue includes: Otsu threshold method is utilized to extract the main lung areas in cardiac CT image；By holes filling method to main lung areas Fracture and hole are filled with, it is thus achieved that complete lung areas.

In based on cardiac CT image the dirtiest above-mentioned extracting method, it is preferable that utilize connected component labeling and circle rate to remove Descending aorta tissue.

In based on cardiac CT image the dirtiest above-mentioned extracting method, it is preferable that the method removing chest wall tissue includes: The appointment source region that the setup action distance beyond chest wall maps is extracted in CT image；To described in CT field of view Appointment source region negates, it is thus achieved that target area；Utilization has target area described in symbol Euclidean distance mapping calculation to described appointment source The distance mapped image in region；The thickness of chest wall is obtained by the value in the SSED distance mapping graph that detection lung areas is covered Degree, removes distance in described distance mapped image and, less than the region of described thickness, obtains removing the masking-out of chest wall；And use institute State masking-out and remove the chest wall tissue in cardiac CT image.

In based on cardiac CT image the dirtiest above-mentioned extracting method, it is preferable that the method removing vertebral tissue includes: Threshold division, extracts the cortical bone tissue in cardiac CT image；With the top central point of descending aorta, bottom center point and Mean radius arranges the vertebra scope after thresholding in image；Each figure layer will be present in the cortical bone tissue in the range of described vertebra Superposition, it is thus achieved that go the masking-out of vertebra；And with described in go the masking-out of vertebra to remove the vertebral tissue in cardiac CT image.

In based on cardiac CT image the dirtiest above-mentioned extracting method, it is preferable that the method removing noise tissue includes: Ask for the SSED distance mapped image of pending image；It is characterized with zone radius and SSED image is carried out thresholding, remove Noise tissue；And along removing the outermost edges line of the image after noise tissue removes, respectively with marginal point as the center of circle, with noise The maximum radius of tissue is that radius draws circle, differentiates the source images in circle overlay area, the foreground point in Restorer varieties image.

The inventive method from reverse angle, with nontarget area (lung tissue, descending aorta tissue, chest wall tissue and Vertebral tissue) it is that object processes, by progressively removing the non-cardiac tissue leaving a volume such as chest wall, pulmonary, vertebra and descending aorta Realizing extracting the dirtiest purpose, it has the advantages such as adaptivity is strong, operational efficiency is high, extraction effect is accurate, it is possible to quickly It is embedded in existing medical network, it is achieved remote assistant diagnoses.

Accompanying drawing explanation

Fig. 1 is the flow chart of an embodiment the dirtiest extracting method of based on cardiac CT image；

Fig. 2 is for removing chest wall masking-out formation basic theory figure；

Fig. 3 is chest wall removal process schematic diagram；

Fig. 4 is the 3 d effect graph after removing chest wall；

Fig. 5 is for removing vertebra masking-out formation basic theory figure；

Fig. 6 is vertebra removal process schematic diagram；

Fig. 7 is the 3 d effect graph of the vertebral tissue removed；

Fig. 8 is the schematic diagram removing noise tissue；

Fig. 9 is noise tissue removal process schematic diagram.

Detailed description of the invention

The present invention is further described with embodiment below in conjunction with the accompanying drawings.These more detailed descriptions are intended to help and understand the present invention, And should not be taken to be limiting the present invention.According to present disclosure, it will be understood by those skilled in the art that and can need not Or all these specific detail can implement the present invention.And in other cases, in order to avoid innovation and creation are desalinated, the most in detail The well-known operating process of thin description.

This programme is from reverse angle, with nontarget area as object of study, by progressively removing chest wall, pulmonary, vertebra Realize extracting the dirtiest purpose with non-cardiac tissue leaving a volume such as descending aortas.

As it is shown in figure 1, this dirtiest extracting method of based on cardiac CT image comprises the following steps:

Step S1, obtains cardiac CT image.Utilize GE Light Speed VCT equipment, right bottom knuckle level to the heart The whole heart area of patient carries out tomoscan, and total figure number of plies is about 200 layers, and thickness is 0.625mm, and single image size is 512*512, pixel interval is 0.488281mm.

Step S2, removes the lung tissue in cardiac CT image.

A kind of method of preferable extraction lung tissue is: first, directly utilizes Otsu threshold method (OTSU method) and extracts the heart Main lung areas in dirty CT image.The interference organized due to pulmonary artery, pulmonary vein and bronchus etc. causes the master extracted Lung areas is wanted to there is the phenomenons such as fracture, hole, therefore, next with the holes filling method fracture to main lung areas And hole is filled with, it is thus achieved that complete lung areas.

Step S3, removes the descending aorta tissue in cardiac CT image.

One preferably method is, utilizes connected component labeling and circle rate to remove descending aorta tissue.More specifically, first, right Cardiac CT image carries out thresholding, preferably threshold value T=1200.Then utilize connected component labeling to the regional in image Carry out label, and extract the round rate geometric properties of each connected domain respectively.Owing to descending aorta is in the most circular, the most permissible Extract descending aorta tissue smoothly.

Step S4, removes the chest wall tissue in cardiac CT image.

In heart CT faultage image, chest wall presents banding substantially, with interior survey edge like two parallel curves outside chest wall. Utilization has symbol Euclidean distance mapping (SSED) to calculate the distance mapped image of chest wall each point, mapping graph of then adjusting the distance As carrying out thresholding operation, the template removing chest wall can be obtained, it is achieved remove chest wall tissue.

Based on this principle, a kind of method preferably removing chest wall tissue includes:

The appointment source region R that the setup action distance beyond chest wall maps is extracted in CT image_src, such as a in Fig. 2 Shown in dark parts in figure；

To described appointment source region R in CT field of view_srcNegate, it is thus achieved that target area R_obj, such as the b subgraph in Fig. 2 In dark parts shown in；

Utilization has symbol Euclidean distance mapping (SSED) to calculate described target area R_objTo described appointment source region R_srcDistance Mapping graph picture, such as the c subgraph in Fig. 2；

By the value in the SSED distance mapping graph that detection lung areas is covered, it is thus achieved that the thickness w of chest wall, remove described The distance region less than described thickness w in distance mapped image, such as the d subgraph in Fig. 2, obtains removing the masking-out of chest wall；

The chest wall tissue in cardiac CT image is removed by described masking-out.

Fig. 3 shows chest wall removal process, removes the state before chest wall during wherein subgraph a is heart CT faultage image, Subgraph b, c be masking-out be formation, subgraph d is to remove the state after chest wall.

Fig. 4 shows and removes the 3-D effect after chest wall, and wherein four subgraph a-d are respectively the state of different visual angles.

Step S5, removes the vertebral tissue in cardiac CT image.

Here use " masking-out superposition " to solve vertebra joint intermediate layer difficulty segmentation problem, thus realize removing vertebral tissue.Concrete side Method includes:

Threshold division, sets threshold value T=1600, extracts the cortical bone tissue in cardiac CT image, eliminates great majority simultaneously Blood vessel and ventricle interference tissue (blood vessel, the CT value about 1200 of ventricle)；

The vertebra scope arranged after thresholding in image with the top central point of descending aorta, bottom center point and mean radius.Logical Cross the statistical analysis to above-mentioned descending aorta, it is thus achieved that descending aorta top central point Start_center(i, j), bottom center point End_center(i, j), and mean radius R_avg, thus obtain the vertebra scope in image after thresholding $\mathrm{Scale}=\left\{\begin{array}{c}\mathrm{ImageWidth}>i>\min \{{\mathrm{Start}}_{\mathrm{Center}}.i,{\mathrm{End}}_{\mathrm{Center}}.m\}\\ 0<j<\max \{{\mathrm{Start}}_{\mathrm{Center}}.j,{\mathrm{End}}_{\mathrm{Center}}.n\}\end{array}\right.;$

Each figure layer will be present in the cortical bone tissue superposition in described vertebra scope Scale, it is thus achieved that go the masking-out of vertebra；

And with described in go the masking-out of vertebra to remove the vertebral tissue in cardiac CT image.

Fig. 5 shows the formation basic theory of the masking-out of vertebra.Wherein, subgraph a is vertebral structure, and three subgraphs b, c, d are The cross-sectional view of three vertebras joint, subgraph e is the masking-out removing vertebra obtained after superposition.

Fig. 6 shows vertebra removal process.Wherein, subgraph a is to remove prevertebral state in heart CT faultage image, son Figure b is the state after threshold division, and subgraph c, d are the state after eliminating most blood vessel and ventricle interference tissue.

Fig. 7 shows the 3-D effect of the vertebral tissue of removal.Wherein four subgraph a-d are respectively the state of different visual angles.

Step S6, removes noise tissue.

One preferably method is, is extended there being symbol Euclidean distance to map (SSED), makes full use of distance and maps anti- The position relationship of the noise group tissue region each point mirrored, reaches quick and precisely to remove the purpose of noise tissue.Concrete grammar includes:

Ask for the SSED distance mapped image of pending image, such as the subgraph b in Fig. 8.Along with mapping distance increases, Gao Ying (the closure edge line represented such as the subgraph c) in Fig. 8 is the most smoothened, it is assumed that need the discreet region removed to penetrate value Big width isSo maximum within the interference range in distance mapped image is little InBy extractingInterior zone just discreet region can be smoothed out；

WithSSED image is carried out thresholding, the region less than d is removed, the discreet region that is eliminated preliminary Result is (such as the subgraph c) in Fig. 8.That is, it is characterized with zone radius SSED image is carried out thresholding, remove noise tissue；

Owing to SSED image is overall to contract, the marginal point in non-interference district also can be to contract, so entering SSED After the direct thresholding of row, the edge in non-interference district can be caused also to be cut.For avoiding this problem, now need to carry out image restoration, Method is, the outermost edges line of the image after removing along removal noise tissue, respectively with marginal point as the center of circle, with noise tissue Maximum radius is that radius draws circle, differentiates the source images in circle overlay area, the foreground point in Restorer varieties image.Now by In having been removed by a little of small disturbance region, it will not be restored by this operation.In Fig. 8, subgraph d, e, f illustrate Image restoration process, subgraph a is the state before removing noise tissue.

Fig. 9 shows noise tissue removal process.Wherein, three groups, upper, middle and lower subgraph is three heart CT faultage images Removal process, removal process from left to right, the leftmost state for Noise tissue, rightmost for removing after noise tissue State.

Said method is from reverse angle, with nontarget area as object of study, and carries out multiple traditional images processing method Improving and optimize, it is achieved that the dirtiest extraction, it has the advantages such as adaptivity is strong, operational efficiency is high, extraction effect is accurate, Can quickly be embedded in existing medical network, it is achieved remote assistant diagnoses.

Claims (5)

1. the dirtiest extracting method of based on cardiac CT image, it is characterised in that the method comprises the following steps:

Obtain cardiac CT image；

Lung tissue, descending aorta tissue, chest wall tissue and the vertebral tissue removed in cardiac CT image obtain intermediate image； And

Removing the noise tissue in described intermediate image, the method removing noise tissue includes:

Ask for the SSED distance mapped image of pending image；

It is characterized with zone radius and SSED image is carried out thresholding, remove noise tissue；And

The outermost edges line of the image after removing along removal noise tissue, respectively with marginal point as the center of circle, with the maximum of noise tissue Radius is that radius draws circle, differentiates the source images in circle overlay area, the foreground point in Restorer varieties image.

The dirtiest extracting method of based on cardiac CT image the most according to claim 1, it is characterised in that remove lung The method of portion's tissue includes: utilize Otsu threshold method to extract the main lung areas in cardiac CT image；Use holes filling side Fracture and the hole of main lung areas are filled with by method, it is thus achieved that complete lung areas.

The dirtiest extracting method of based on cardiac CT image the most according to claim 1, it is characterised in that the company of utilization Logical field mark and circle rate remove descending aorta tissue.

The dirtiest extracting method of based on cardiac CT image the most according to claim 1, it is characterised in that remove breast The method of cavity wall tissue includes:

The appointment source region that the setup action distance beyond chest wall maps is extracted in CT image；

Described appointment source region is negated by CT field of view, it is thus achieved that target area；

Utilize and have target area described in symbol Euclidean distance mapping calculation to the distance mapped image of described appointment source region；

Obtained the thickness of chest wall by the value in the SSED distance mapping graph that detection lung areas is covered, remove described distance In mapping graph picture, distance is less than the region of described thickness, obtains removing the masking-out of chest wall；And

The chest wall tissue in cardiac CT image is removed by described masking-out.

The dirtiest extracting method of based on cardiac CT image the most according to claim 1, it is characterised in that remove vertebra The method of osseous tissue includes:

Threshold division, extracts the cortical bone tissue in cardiac CT image；

The vertebra scope arranged after thresholding in image with the top central point of descending aorta, bottom center point and mean radius；

Each figure layer will be present in the cortical bone tissue superposition in the range of described vertebra, it is thus achieved that go the masking-out of vertebra；And

The vertebral tissue in cardiac CT image is removed by the described masking-out removing vertebra.