CN106611413A - Image segmentation method and system - Google Patents
Image segmentation method and system Download PDFInfo
- Publication number
- CN106611413A CN106611413A CN201611092346.4A CN201611092346A CN106611413A CN 106611413 A CN106611413 A CN 106611413A CN 201611092346 A CN201611092346 A CN 201611092346A CN 106611413 A CN106611413 A CN 106611413A
- Authority
- CN
- China
- Prior art keywords
- region
- voxel
- pixel
- target area
- image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20092—Interactive image processing based on input by user
- G06T2207/20101—Interactive definition of point of interest, landmark or seed
-
- 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
- G06T2207/30061—Lung
- G06T2207/30064—Lung nodule
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Apparatus For Radiation Diagnosis (AREA)
Abstract
The invention relates to an image segmentation method and system. The method comprises the following steps: initially positioning medical images to obtain a positioning area; preprocessing the positioning area to obtain a target area, wherein the target area contains a nodular area and a background area, and the nodular area consists of a solid area and a surrounding area which surrounds the solid area; processing the target area in combination with a Gaussian mixture model to obtain a probability graph of the target area; processing the probability graph according to a morphological model to determine the solid area and the surrounding area of a nodule; and fusing the solid area and the surrounding area to obtain an image segmentation result. According to the method and system provided by the invention, different types of nodules can be accurately segmented, and the follow-up diagnostic analysis can be effectively improved.
Description
【Technical field】
The present invention relates to field of medical image processing, more particularly to a kind of image partition method and its system.
【Background technology】
Pulmonary carcinoma is global mortality rate highest cancer.Although medical diagnosiss and treatment level are improved constantly, 5 years of pulmonary carcinoma
Survival rate also only has 15% or so, and early discovery, early treatment are the main paths for improving pulmonary carcinoma cure rate.Lung neoplasm is pulmonary carcinoma
Early manifestation form.The accurate segmentation result of Lung neoplasm can effectively reflect pathology and the morphological characteristic of tuberosity, and then
User is helped to carry out diagnostic analysiss to focus.Realize that Lung neoplasm is automatically analyzed is with the key technology of identification:Lung neoplasm segmentation, inspection
The a series of images such as survey, diagnosis process, analyze and understand the research and application of algorithm.
Nodule segmentation is carried out using level set algorithm or multi-scale thresholds method in prior art, but Lung neoplasm form
Different, size is different, distributing position is indefinite, easily with other organizing links, some tissues of density and pulmonary are similar, such as lung
Tuberosity has all kinds of shapes such as solid nodules, mixed type ground glass tuberosity, ground glass tuberosity (ground-glass nodule, GGN)
State, it is impossible to accurately identify Lung neoplasm simply by morphological method, especially ground glass tuberosity as a kind of pernicious probability most
A big class tuberosity, form top edge is irregular, presents in CT images and obscures thin shadow, the HU (Hounsfield in CT
Unit, Heng Shi unit) Distribution value is extensive and texture form is changeable, calculated based on edge by level set algorithm (Level Set) etc.
Method is difficult to accurately recognize Lung neoplasm.In addition gray value of the Lung neoplasm in CT images disobeys Gauss distribution, and multi thresholds method is easy
Leakage is caused to cause segmentation result inaccurate.
Therefore, the present invention provides a kind of image partition method, to improve the accuracy of nodule segmentation.
【The content of the invention】
The technical problem to be solved is to provide a kind of dividing method and its system of medical image, for splitting
Different types of Lung neoplasm, can effectively improve the accuracy of Lung neoplasm segmentation, and then beneficial to user is to the diagnosis of focus and divides
Analysis.
To solve above-mentioned technical problem, the present invention provides a kind of dividing method of medical image, comprises the steps：
Initial alignment is carried out to medical image, positioning region is obtained；
Positioning region described in pretreatment, obtains target area, and the target area includes knuckle areas and background area, institute
State knuckle areas to constitute by reality region and around the peripheral region in the real property region；
Based on the gauss hybrid models process target area, the probability graph of the target area is obtained；
According to the Morphological Model process probability graph, to determine the real property region and peripheral region of tuberosity；
Merge the real property region and peripheral region, obtain image segmentation result.
Further, initial alignment is carried out to medical image including following any one mode：
Positioning region described in manually determined；Or
It is determined that through the long axis of the positioning region, the selected seed point on long axis is true using region growing methods
The fixed positioning region；Or
Determine the positioning region based on pre-set programs.
Further, positioning region described in the pretreatment includes：
The positioning region is carried out to be based on the enhanced Threshold segmentation of extra large gloomy point, the target area is obtained.
Further, it is described that the probability graph is processed according to Morphological Model, including：
Probability graph described in models treated is strengthened using the gloomy point in sea, obtaining the gloomy point in sea strengthens figure；
Based on each voxel or pixel in the probability graph, the corresponding voxel or picture in the gloomy point enhancing figure in the sea is determined
Element；
If the value of described arbitrary voxel or pixel in the probability graph is more than first threshold, and Hai Sen points increase
The value of corresponding voxel or pixel is more than Second Threshold in strong figure, then the voxel or pixel belong to the real property region；
Otherwise, the voxel or pixel belong to background area.
Further, the span of the first threshold for (0,1) between constant, the value of the Second Threshold
Scope is the constant between (0,100).
Further, it is described based on the gauss hybrid models process target area, also include：
Probability graph described in models treated is strengthened using extra large gloomy line, obtaining extra large gloomy line strengthens figure；
Based on each voxel or pixel in the probability graph, the corresponding voxel or picture in the extra large gloomy line enhancing figure is determined
Element；
The value of the voxel in the probability graph or pixel is compared with first threshold；
The value of the voxel in the extra large gloomy line enhancing figure or pixel is compared with the 3rd threshold value；
If the value of described arbitrary voxel or pixel in the probability graph is more than first threshold, and the Hai Sen lines increase
The value of corresponding voxel or pixel is less than the 3rd threshold value in strong figure, then the voxel or pixel belong to the peripheral region；
Otherwise, the voxel or pixel belong to background area.
Further, the span of the 3rd threshold value be span for (0,50) between constant.
Further, the target area, the probability graph, the gloomy point enhancing in the sea and the extra large gloomy line strengthen in figure
The locus of arbitrary voxel or pixel are to correspond.
To solve above-mentioned technical characteristic, the present invention also provides a kind of image segmentation system, including:
Input block, for obtaining medical image；
Processing unit, for processing the medical image, obtains the segmentation result of target area, and the target area includes
Knuckle areas and background area, the knuckle areas are constituted by reality region and around the peripheral region in the real property region；
And output unit and memory element；
The processing unit includes tuberosity determination subelement, for processing the target area based on gauss hybrid models,
Obtain the probability graph of the target area；According to the Morphological Model process probability graph, with determine tuberosity real property region and
Peripheral region.
Further, the cutting unit also includes：
Locator unit, for carrying out initial alignment to medical image, obtains positioning region；
Pretreatment subelement, for positioning region described in pretreatment, obtains target area.
Compared with prior art, the present invention obtains target area using initial alignment and pretreatment, to enhance tuberosity
The contrast in reality region, reduces the amount of calculation of follow-up nodule segmentation, improves splitting speed；Institute is processed based on gauss hybrid models
State target area, acquisition probability figure, according to the Morphological Model process probability graph so that effectively split and protect the reality of tuberosity
Property region, and effectively extract the peripheral region of tuberosity and remove the main noise of the nodule segmentations such as blood vessel, trachea and lung wall.This
Invention provides image partition method highly versatile, high precision, can accurately split different types of tuberosity, beneficial to subsequent user pair
The diagnosis and analysis of focus.
【Description of the drawings】
Fig. 1 is the schematic flow sheet of image partition method in one embodiment of the invention；
Fig. 2 is the schematic flow sheet of image partition method in further embodiment of this invention；
Fig. 3 a～3c is the result schematic diagram of image partition method in one embodiment of the invention；
Fig. 4 is the structural representation of image segmentation in one embodiment of the invention；
Fig. 5 is the structural representation of processing unit in one embodiment of the invention.
【Specific embodiment】
Elaborate many details in order to fully understand the present invention in the following description.But the present invention can be with
Much it is different from alternate manner described here to implement, those skilled in the art can be in the situation without prejudice to intension of the present invention
Under do similar popularization, therefore the present invention is not embodied as being limited by following public.
The present invention is described in detail using schematic diagram, when the embodiment of the present invention is described in detail, for purposes of illustration only, the signal
Figure is embodiment, and its here should not limit the scope of protection of the invention.
Embodiment 1
Effectively, accurately split the technical problem of different types of Lung neoplasm in order to solve prior art, improve user couple
Focus diagnosis, the accuracy of analysis, provide a kind of image partition method, the such as flow process of Fig. 1 image partition methods in the present embodiment
Shown in schematic diagram, methods described comprises the steps：
Execution step S101：Initial alignment is carried out to medical image, positioning region is obtained.The medical image medical science includes
But it is not limited by the imaging system scanning collection of all kinds of mode and obtains three-dimensional or two dimensional image, it is also possible to is by such as storing
Image archiving and communication system (Picture Archiving and Communication S indulge stems, PACS) etc. it is internal or
External storage system transmission is obtained.The mode include but is not limited to nuclear magnetic resonance (MRI), magnetic resonance angiography (MRA),
Computed tomography (CT), positron emission computerized tomography (Positron Emission Tomograph are indulged, PET) etc. one
Plant or various combinations.The form of the medical image can include but is not limited to jpeg image format, tiff image form,
GIF picture formats, FPX picture formats, DICOM picture formats etc..For example, involved in the present embodiment pulmonary's medical image can
With the CT images by computed tomography (CT) equipment Jing after implementing scanning to human body.The medical image can also be logical
Cross and processed in input Computer Image Processing equipment, based on Threshold segmentation, clustering algorithm, histogram divion model, maximum
Medical image after the process of the methods such as inter-class variance parted pattern, the lung CT image for for example being obtained by dividing processing can be with
Follow-up Lung neoplasm segmentation is carried out based on the lung areas in image, so that user observes and uses.Computer Image Processing sets
It is standby to use hardware based on general computer equipment.In one embodiment, can be calculated by performing to be arranged on
The image processing program on processor in machine equipment is come the image procossing needed for realizing.At this moment, can be by advance by image
Processing routine is installed in computer equipment or by the way that image processing program is recorded in disk, magneto-optic disk, CD, quasiconductor
Image processing program is issued in memorizer etc. or on the network and be installed to image processing program in computer equipment to realize
Computer Image Processing equipment.In another embodiment, computer graphic can be realized by the hardware of such as logic circuit
As diagnostic process equipment.In another embodiment, it is also possible to Computer Image Processing is realized by combination hardware and software and is set
It is standby.
The initial alignment is used to obtain one or more positioning regions, to reduce the amount of calculation of subsequent singulation process, carries
The treatment effeciency of high target area acquisition process.The initial alignment target area for example in pulmonary's medical image, due in pulmonary
Medical image includes Pulmonary Vascular, bronchus and the organizational structure such as lung wall and Lung neoplasm, pulmonary parenchyma, and wherein Lung neoplasm is in space shape
Reality region and peripheral region can be divided in state, reality region is in mostly globoid, in highlighted in CT image Lung neoplasms
State, can be by naked eyes identification approximate region；Peripheral region is generally edge blurry, is similar to the irregular form such as burr shape
Region, or even edge is irregular curve.Therefore, user or background computer processing equipment can be according to the real property of tuberosity
Region, using initial alignment positioning region is obtained.The positioning region includes suspected nodular lesion area, there is blood vessel, pulmonary parenchyma etc.
Noise, the positioning region can be by circumscribed rectangular body frame (Computer reprocessing equipment plane is shown as rectangle frame) described
Positioning region is shown on medical image, such as in 256*256*256 lung CT images, by initial alignment, obtain 70mm ×
The positioning region that the boundary rectangle frame of 70mm sizes shows, the length of side of the boundary rectangle frame can be according to required segmentation
Object is preset, such as splitting Lung neoplasm in lung CT medical image, because the general diameter of tuberosity is in 3-30mm, if
Determine circumscribed rectangular body frame 70mm × 70mm, the positioning region for completely including tuberosity can be met.It is described initial fixed in the present embodiment
Position can be it is manual, it is full automatic, or manually and automatically method complements each other.For example can pass through：User is according to Jing
Test to be worth and outline positioning region manually in the lung CT image；Or user is rule of thumb given through the Shi Xing areas of doubtful tuberosity
The long axis in domain, according to the long axis selected seed point, the region based on threshold value increases the acquisition positioning region；Or base
Obtain described in the pre-set programs of the software and hardware combining such as the hardware devices such as computer processor or computer graphical processing equipment
Positioning region, the pre-set programs can be the journey that aforementioned hardware equipment or software and hardware combining equipment are performed according to special algorithm
Sequence.
Execution step S102：Positioning region described in pretreatment, obtains target area.The target area includes knuckle areas
And background area, the knuckle areas are the knuckle areas for determining, by reality region and around the real property region
Region is constituted.The real property region can represent the body shape region of tuberosity, and the peripheral region can represent the side of tuberosity
Edge region.The target area is obtained by positioning region described in pretreatment, the region of subsequent singulation can be further reduced, is carried
The accuracy and speed of high subsequent singulation.
In the present embodiment, the pretreatment includes but is not limited to pretreatment enhancing, interpolation processing, Morphological scale-space, image
The combination of one or more in segmentation, noise remove, smoothing processing etc..Enhancement process can the including but not limited to gloomy point increasing in sea
By force, the combination of one or more in extra large gloomy line enhancing etc..Interpolation processing can make voxel size in area-of-interest uniform.Shape
State process can adopt the element with certain morphosiss to go to process the shape in area-of-interest to reach analysis and know
The purpose of other target.Morphologic processing method can include but is not limited in expansion, burn into opening operation, closed operation etc.
Plant or several combinations.Noise remove can remove the interference brought due to machine noise, target motion etc. in area-of-interest.
The method of noise remove can include but is not limited to the one kind in mean filter, Wiener filtering, morphologic filtering, medium filtering etc.
Or several combinations.For example, can by projecting the real property region in knuckle areas based on the enhanced Threshold segmentation of extra large gloomy point, with
The noises such as the trachea and lung wall, blood vessel of pulmonary are different from, so that it is determined that the target area.In the present embodiment, the target
Region can be shown by boundary rectangle frame.
Execution step S103:Based on the gauss hybrid models process target area, the probability of the target area is obtained
Figure；Gauss hybrid models are the non-morphologic parted patterns of a class, and variously-shaped candidate target can pass through Gaussian Mixture
Model is preferably split, it is adaptable to different types of nodule segmentation.Due to gauss hybrid models need to prior probability,
The parameters such as average, standard deviation are calculated and obtained, and calculative variable is less so that the speed of service is very fast.Through Gaussian Mixture mould
The type process target area, can be divided into the pixel or voxel in image more than two classes or two classes.
Because the voxel or pixel of tuberosity and the voxel or pixel of background submit to Gauss distribution, the present embodiment adopts Gauss
The mixed model process target area, to the probability graph for obtaining the target area, for judging target area in each
Whether voxel is tuberosity.
As shown in formula (1), wherein, π is the prior probability of Gaussian Mixture distribution；μ, Λ are respectively average and standard deviation.
The joint probability density function of gauss hybrid models is shown in formula 1, μ can be obtained with iterative by EM algorithmsk,πk,Λk,
Then p (x) is tried to achieve, so as to obtain the probit of the voxel or pixel, each voxel or picture in shown target area is finally given
The probit of element.The value of each voxel or pixel represents the probability whether voxel or pixel be tuberosity in the probability graph.
In the probability graph locus of each voxel or pixel correspond in the target area voxel or
The position of pixel, the value of each voxel or the value of pixel represent whether the voxel or pixel are tuberosity in the probability graph
Probability.
Execution step S104, according to the Morphological Model process probability graph, to determine real property region and the surrounding of tuberosity
Region.The morphological segment model can include that the gloomy point in sea strengthens model, extra large gloomy line and strengthens model, multiple dimensioned Gaussian template
With model, multi-scale morphology filtering model etc..Hai Sendian strengthens model and can be used for strengthening the such as round dot figure in image
Or the region such as class round dot figure.The gloomy line in sea strengthens model and can be used for strengthening the linear regions in image.Multiple dimensioned Gaussian template
The process of Matching Model segmentation figure picture can be considered based on the form of process object.By taking the process of nodule segmentation as an example, many chis
Degree Gaussian template Matching Model can be based on the similar round form of Lung neoplasm and consider.Multi-scale morphology filtering model can be adopted
Various Mathematical Morphology Methods are filtered operation to image, to reach the purpose for strengthening candidate target.Based on edge algorithms
Parted pattern can be level set algorithm parted pattern.
For example, it is the real property region of acquisition tuberosity, the present embodiment is based on probability graph described in extra large gloomy enhancement process.Based on institute
Each voxel or the pixel in probability graph is stated, corresponding voxel or pixel in the gloomy point enhancing figure in the sea is determined.Institute is chosen one by one
Arbitrary voxel or pixel in probability graph are stated, if the probit of the voxel or pixel is more than first threshold, also, the target area
The value of the value of the voxel of correspondence position or pixel is more than Second Threshold in domain, then voxel described in labelling or pixel belong to Shi Xing areas
Domain, is otherwise labeled as background area.The background area includes but is not limited to the noises such as blood vessel, trachea and lung wall.Described first
The span of threshold value for (0, constant 1), for example, it is 0.5 that first threshold is chosen in the present embodiment, sets the Second Threshold
It is in order to highlight spherical real property region to suppress the noises such as blood vessel, so as to reach the mesh in protection tuberosity reality region
, the span of the Second Threshold is for (0, constant 100), the present embodiment can choose Second Threshold for 15.This enforcement
In example, the value of the voxel or the value of pixel can represent the color value of certain voxel or pixel in image.For black white image,
The value of voxel or the value of pixel can represent the gray value of certain pixel or voxel in image.It is to obtain reality in the present embodiment
Parameter value required for the extra large gloomy point enhancing in region can be differently configured from the gloomy point in sea in foregoing pre-treatment step strengthens the target area
The parameter value in domain.
To obtain the peripheral region of tuberosity, the present embodiment can be based on probability graph described in extra large gloomy line enhancement process.Based on institute
Each voxel or the pixel in probability graph is stated, corresponding voxel or pixel in the extra large gloomy line enhancing figure is determined.Before choosing one by one
Arbitrary voxel or pixel in step probability graph are stated, if the probit of the voxel is more than first threshold, also, it is described in extra large gloomy line
The value for strengthening correspondence position voxel or pixel in image is less than the 3rd threshold value, then voxel described in labelling or pixel belong to peripheral region
Domain, is otherwise labeled as background area.The linear shape structure that can strengthen in image is strengthened by extra large gloomy line in the present embodiment.Example
Such as, for the process of segmentation Lung neoplasm, being strengthened by extra large gloomy line can strengthen the pipe such as the probability graph medium vessels, trachea and lung wall
Shape structure so that the main noise such as blood vessel is in highlighted state.The selection of first threshold described in the present embodiment (0,1), for example, this reality
It is 0.5 to apply and choose in example first threshold, and setting the 3rd threshold value can remove making an uproar for the tubular structures such as blood vessel, trachea and lung wall
Sound, extracts the tuberosity of non-spherical in irregular shape, i.e. peripheral region.The span of the 3rd threshold value for (0,50), this
Embodiment can choose Second Threshold for 5.In the present embodiment, the value of the voxel or the value of pixel can represent certain in image
The color value of voxel.For black white image, the value of voxel or the value of pixel can represent the ash of certain pixel or voxel in image
Angle value.
In the present embodiment, the target area, the probability graph, the gloomy point in the sea strengthen and the extra large gloomy line enhancing figure
In arbitrary voxel or pixel locus be correspond.
Execution step S105, merges the real property region and peripheral region, obtains image segmentation result.For example to lung CT
Split tuberosity in image, the described real property region obtained based on abovementioned steps is spherical nodule segmentation result, peripheral region
It is mainly in irregular shape, ill-defined nodule segmentation result, according to the voxel for being labeled as reality region or peripheral region or
The position of pixel positions the voxel or pixel in lung CT image, obtains final nodule segmentation result, the nodule segmentation
As a result can be binary image (mask image).
In the present embodiment, obtaining target area based on initial alignment and pretreatment strengthens the contrast in tuberosity reality region,
The amount of calculation of follow-up nodule segmentation is reduced, splitting speed is improved；Based on the gauss hybrid models process target area, bag is obtained
Probability graph containing different shape tuberosity；According to the Morphological Model process probability graph, not only effectively split and protect tuberosity
Reality region, while effectively extract the peripheral region of tuberosity and remove the main noise of the nodule segmentations such as blood vessel, trachea and lung wall,
Improve the accuracy of nodule segmentation result.
Embodiment 2
Understandable to enable above-mentioned purpose, feature and advantage to become apparent from, the present embodiment provides a kind of for lung CT three
The nodule segmentation method of dimension image, to obtain different types of Lung neoplasm, as shown in Fig. 2 method flow diagrams, comprises the steps：
Execution step S201：Initial alignment is carried out to medical image, positioning region is obtained.In the present embodiment, the medical science
Image is pulmonary's medical image, and the medical image can be implemented to scan by computed tomography (CT) equipment Jing to human body
The original CT image for obtaining afterwards, as shown in Figure 3 a.Or the original CT image is input in Computer Image Processing equipment
Processed, the lung CT image needed for being obtained based on methods such as Threshold segmentation, clustering algorithms, as shown in Figure 3 b.
The initial alignment, to reduce the amount of calculation of subsequent singulation process, improves segmentation efficiency to obtain positioning region,
Concrete initial alignment method is as it was previously stated, will not be described here.
Execution step S202：The positioning region is carried out to be based on the enhanced Threshold segmentation of extra large gloomy point, the target is obtained
Region.In the lung CT image, Pulmonary Vascular, bronchus and lung wall and Lung neoplasm are closely similar in grey level distribution, therefore
So that clinically for the judgement of Lung neoplasm easily produces mistaken diagnosis or fails to pinpoint a disease in diagnosis.However, blood vessel, bronchus and lung wall and Lung neoplasm
There are certain difference, Pulmonary Vascular, bronchus and lung wall etc. often to present tubular structure in spatial shape, by human body
Anatomy understands that intrapulmonary blood vessel and trachea and lung wall can construct complete vascular tree, trachea and lung wall tree according to connectedness,
And the form of Lung neoplasm can be divided into as previously mentioned reality region and peripheral region.Lung neoplasm can refer to generation in pulmonary parenchyma
Circular or similar round (in 3-D view can be spherical or spherical) intrapulmonary of the single or multiple diameters less than 30mm is fine and close
Shade, is mainly shown as on the medical image the impermeable shadow of intrapulmonary, has more clear and definite edge, is such as in smooth, lobulated, spine shape
Projection or burr.It (can be ball in 3-D view that the real property region of Lung neoplasm can be the circular or similar round of Lung neoplasm
Shape is spherical) region, the peripheral region of Lung neoplasm can be edge blurry, be similar to the region of the irregular form such as burr shape,
Or the irregular curve in edge.Therefore pretreatment can be passed through effectively strengthens the real property region of tuberosity, is in highlighted shape on image
State, the contrast of the noise such as enhancing and blood vessel, trachea and lung wall, while further reducing the processing region of subsequent singulation, is improved
The accuracy and speed of subsequent singulation.
In the present embodiment, row is based on the enhanced Threshold segmentation of extra large gloomy point to be included：Process described fixed by interpolation method first
Position region, each voxel resolution (voxel spacing) in the x, y and z directions in positioning region described in normalization, for example
It is 1mm, 1mm and 1mm that resolution on arbitrary voxel x, y and z direction is passed through into interpolation processing.Then, to each voxel in x, y
Gaussian smoothing is carried out with the value on z directions, to be highlighted globoid region, the highlight regions include doubtful tuberal region
The spherical real property region in domain, such as tuberosity.The sea of each voxel in the target area of aforementioned processing is calculated according to formula (2)
Gloomy matrix, wherein, fxxLead for any voxel second order in the x direction, fxyIt is on the basis of x directions one are situated between inverse, to ask y directions
The second order derivative；The second order of one different directions is led and is built into Hessian matrix, try to achieve eigenvalue λ1、λ2And λ3.By feature
Value λ1、λ2And λ3Substitute into formula (3) and try to achieve a little enhanced value Zdot, i.e., the value of correspondence voxel in target area.Finally, by extra large gloomy point
Enhancement value returns former resolution in resolution inverse interpolation in the x, y and z directions, and acquisition is highlighted the extra large gloomy point in spherical region
Strengthen image.
Otherwise it is 0
Image is strengthened based on the extra large gloomy point for obtaining and enters row threshold division.Due to the target area by extra large gloomy enhancement process
In, spherical region is in highlighted state, i.e., extra large gloomy point value is larger, so as to strengthen spherical region and pulmonary trachea and lung wall,
The noises such as blood vessel contrast in the picture, improves the accuracy of nodule segmentation.The nodule segmentation can be based on Threshold segmentation
Obtain, the threshold value can take (0,100) in the range of constant, for example selected threshold be 5.The target area can be by outer
Connect rectangle frame to show, the area of the boundary rectangle frame of the target area can be less than the boundary rectangle frame of the positioning region,
The knuckle areas so that the boundary rectangle frame of the target area is further fitted, effectively improve the operation speed of follow-up nodule segmentation
Degree.
Execution step S203：Based on the target area in the gauss hybrid models process boundary rectangle frame, obtain described
The probability graph of target area.Abovementioned steps S103 have described the acquisition modes of the probability graph in detail, will not be described here.Through Gauss
The mixed model process target area, can be divided into the voxel in image more than two classes or two classes.In the present embodiment, lead to
Cross the gauss hybrid models and judge that whether each voxel is as tuberosity in target area.The value of each voxel in the probability graph
Represent the probability whether voxel is tuberosity.
Execution step S204～S210：During nodule segmentation, because knuckle areas come in every shape, including the reality of Lung neoplasm
Property region can be Lung neoplasm circular or similar round (can be spherical or spherical in 3-D view) region, Lung neoplasm
Peripheral region can be edge blurry, be similar to the region of the irregular form such as burr shape, or the irregular curve in edge.Meanwhile,
Because the regions such as lung wall, blood vessel are presented highlight regions in lung CT image, by possible in the probability graph that abovementioned steps are obtained
Comprising the tissue such as part blood vessel, lung wall, the accuracy of nodule segmentation is affected, it is therefore desirable to described general according to Morphological Model process
Rate figure, obtains respectively the real property region and peripheral region of tuberosity.
In the present embodiment, obtaining the real property region of tuberosity includes：Using probability graph described in the gloomy point enhancing models treated in sea, obtain
Take the gloomy point enhancing figure in sea.Based on each voxel in the probability graph, the corresponding voxel in the gloomy point enhancing figure in the sea is determined.By
One chooses arbitrary voxel in the probability graph, if the probit of the voxel is more than in first threshold, also, the target area
The value of the voxel of correspondence position is more than Second Threshold, then voxel described in labelling belongs to reality region, is otherwise labeled as background area.
The background area includes but is not limited to the noises such as blood vessel, trachea and lung wall.The span of the first threshold for (0,1),
For example, it is 0.5 that first threshold is chosen in the present embodiment, and it is to highlight spherical Shi Xing areas to set the Second Threshold
Domain to suppress the noises such as blood vessel, so as to reach the purpose in protection tuberosity reality region, the span of the Second Threshold for (0,
100), the present embodiment can choose Second Threshold for 15.It is that the extra large gloomy point for obtaining reality region strengthens required in the present embodiment
Parameter value can be differently configured from foregoing pre-treatment step the gloomy point in sea and strengthen the parameter value of the target area.
In the present embodiment, obtaining the peripheral region of tuberosity includes；Using probability graph described in extra large gloomy line enhancing models treated, obtain
Take extra large gloomy line enhancing figure.The gloomy line in sea strengthens figure can be obtained based on formula (4).Then, based on the probability graph in per individual
Element, determines the corresponding voxel in the extra large gloomy line enhancing figure.Arbitrary voxel in abovementioned steps probability graph is chosen one by one, if the body
The probit of element is more than first threshold, also, the voxel for strengthening correspondence position in image in extra large gloomy line is less than the 3rd threshold value,
Then voxel described in labelling belongs to peripheral region, and otherwise voxel described in labelling belongs to background area.
ifλ1<0,λ2<0；Otherwise it is 0 formula (4)
In lung CT image, Pulmonary Vascular, trachea lung wall are the false-positive main sources of Lung neoplasm, cause follow-up Lung neoplasm to examine
The mistaken diagnosis of survey and fail to pinpoint a disease in diagnosis；Pulmonary Vascular, bronchus and lung wall and Lung neoplasm are closely similar in grey level distribution, in spatial shape,
Pulmonary Vascular and bronchus and lung wall etc. often present tubular structure.Being strengthened by extra large gloomy line in the present embodiment can strengthen described
The tubular structures such as probability graph medium vessels, trachea and lung wall so that the main noise such as blood vessel is in highlighted state.Described in the present embodiment
(0,1), for example, it is 0.5 that first threshold is chosen in the present embodiment, and setting the 3rd threshold value can remove blood for first threshold selection
The noise of the tubular structures such as pipe, trachea and lung wall, while retaining the knuckle areas of non-spherical in irregular shape, i.e. peripheral region
Domain.The span of the 3rd threshold value is for (0,50), the present embodiment can choose Second Threshold for 5.
In the present embodiment, the target area, the probability graph, the gloomy point in the sea strengthen and the extra large gloomy line enhancing figure
In arbitrary voxel or pixel locus be correspond.
Execution step S311：The described real property region obtained based on abovementioned steps is spherical nodule segmentation result, week
Enclose region mainly in irregular shape, ill-defined nodule segmentation result, according to being labeled as reality region or peripheral region
The position of voxel or pixel positions the voxel or pixel in lung CT image, obtains final nodule segmentation result, such as Fig. 3 b
It is shown.It is appreciated that the intermediate result shown in Fig. 3 is merely illustrative, the intermediate result of various embodiments of the present invention is not meant to
The specific modality being in Fig. 3.
In the present embodiment, to split lung CT 3-D view in tuberosity as a example by, obtained based on initial alignment and pretreatment
Target area, and the contrast in tuberosity reality region is enhanced, the amount of calculation of follow-up nodule segmentation is reduced, improve segmentation speed
Degree；Based on the gauss hybrid models process target area, the probability graph comprising different shape tuberosity is obtained；According to morphology mould
The type process probability graph, by probability graph described in extra large gloomy enhancement process, the real property region of tuberosity is split and protected to effectively effect,
By probability graph described in extra large gloomy line enhancement process, effectively extract the peripheral region of tuberosity and remove the knot such as blood vessel, trachea and lung wall
The main noise of section segmentation, improves the accuracy of nodule segmentation result.
Embodiment 3
Image segmentation system is provided in solve above-mentioned technical problem the present embodiment.Described image segmenting system can include
One or more processing units, one or more memory element, one or more input blocks, one or more output units,
Can be between unit it is distributed can also be it is centralized, can be it is local can also be long-range.
In certain embodiments, the input block can respectively from imaging device, data base, memory element or external
Receive the data for each sending at equipment.Data herein, can be medical datas.The medical data can be medical science figure
Picture.The medical image can include but is not limited to x-ray image, CT images, PET image, MRI image, ultrasonoscopy, electrocardio
The combination of one or more in figure, electroencephalogram etc..The medical image can be two-dimentional (2D, two-dimensional) figure
Picture, or three-dimensional (3D, three-dimensional) image.The form of the medical image can be including but not limited to
Joint Photographic Experts Group (JPEG) picture formats, Tagged Image File Format (TIFF)
Picture format, Graphics Interchange Format (GIF) picture formats, Kodak Flash PiX (FPX) image panes
Formula, Digital Imaging and Communications in Medicine (DICOM) picture formats etc..The input of data
Mode can be handwriting input, gesture input, image input, phonetic entry, video input, electromagnetic wave input etc. in one kind or
Several combinations.The information for being received, can be stored in data base, it is also possible to which storage is in the memory unit, it is also possible to by
Reason unit is analyzed or processes.The input block can include but is not limited to character input device (for example, keyboard), optics
Arrangement for reading (for example, optical mark reader, optical character reader), graphic input device (for example, Genius mouse, action bars,
Light pen), image input device (for example, video camera, scanner, facsimile machine), simulation imput device (for example, language analog digital conversion
Identifying system) etc. in the combination of one or more.
The output unit can export treated data.Data herein, can be the most terminations of image segmentation
Intermediate data in fruit, or image segmentation process.For example, during image segmentation, processing unit can be to defeated
The medical image for entering is processed, analyzed, and during this, intermediate data can include obtaining the positioning area that initial alignment is obtained
The data such as domain, the segmentation result of target area that pretreatment is obtained, final result is the segmentation result of tuberosity.The form of data can
The combination of one or more in include but is not limited to text, audio frequency, video, picture etc..The data of output can be sent to
External equipment, it is also possible to do not send.The output data not sent can be stored in the memory unit.The output unit can be wrapped
In including but being not limited to display device, printing device, drawing apparatuss, image output system, voice output system, magnetic recording equipment etc.
The combination of one or more.In certain embodiments, some external equipments can simultaneously play a part of to be input into and export, example
Such as, desktop computer, notebook, smart mobile phone, panel computer, personal digital assistant (personal digital
Assistance, PDA) etc..
The memory element can store from processing unit, input block, output unit data.The memory element
It can be the external equipment of internal system, or system.The memory element can be actually existed in system, also may be used
To complete corresponding function by cloud computing platform.
The processing unit can process related data.Processing unit can obtain number from input block or memory element
According to.The processing unit can preserve the data after process into data base or memory element, it is also possible to send single to output
Unit is used for data output.In certain embodiments, the mode of data processing can be included but is not limited to data in processing unit
The combination of one or more in being stored, classified, being screened, being changed, being calculated, being retrieved, being predicted, being trained etc..In some enforcements
In example, processing unit can include but is not limited to central processing unit (Central Processing Unit (CPU)), specially should
With integrated circuit (Application Specific Integrated Circuit (ASIC)), dedicated instruction processor
(Application Specific Instruction Set Processor (ASIP)), concurrent physical processor (Physics
Processing Unit (PPU)), digital signal processor (Digital Processing Processor (DSP)), scene
Programmable gate array (Field-Programmable Gate Array (FPGA)), PLD
One kind in (Programmable Logic Device (PLD)), processor, microprocessor, controller, microcontroller etc. or
Several combinations.
It should be noted that above-mentioned processing unit can be actually existed in image segmentation system, it is also possible to by cloud meter
Calculate platform and complete corresponding function.Wherein, cloud computing platform includes but is not limited to storage-type cloud platform, the Yi Chu based on data storage
Manage the calculation type cloud platform based on data and take into account data storage and the comprehensive cloud computing platform for processing.The cloud that system is used
Platform can be public cloud, private clound, community cloud or mixed cloud etc..For example, according to actual needs, some medical science that system is received
Image, can be calculated by cloud platform and/or be stored.Other medical images, can by local diagnosis unit and/or
System database is calculated and/or stored.
It is as shown in Figure 4 and Figure 5 segmentation Lung neoplasm system in one embodiment of the invention according to some embodiments of the present application
Structural representation, the system includes：Input block U100, memory element U200, processing unit U300, and output unit
U400。
The input block U100, for obtaining medical image.Medical image medical science described in the medical image include but
The imaging system scanning collection for being not limited by all kinds of mode obtains three-dimensional or two dimensional image, it is also possible to by such as storing be shadow
As the inside such as archiving and communication system (Picture Archiving and Communication S indulge stems, PACS) or outer
Portion's storage system transmission is obtained.The mode includes but is not limited to nuclear magnetic resonance (MRI), magnetic resonance angiography (MRA), meter
The one kind such as calculation machine tomoscan (CT), positron emission computerized tomography (Positron Emission Tomograph are indulged, PET)
Or various combinations.Medical image can be sent to memory element U200 and make storage process by the input block U100, also may be used
Image procossing is carried out to transmit to processing unit U300.
The processing unit U300, for processing the medical image, obtains the segmentation result of target area, the present embodiment
In, the processing unit U300 includes pre-determined bit subelement U301, pretreatment subelement U302, and tuberosity determination subelement
U303。
Locator unit U301, for carrying out initial alignment to medical image, obtains positioning region.It is described initial fixed
, to reduce the amount of calculation of subsequent singulation process, target area acquisition process is improved for obtaining one or more positioning regions in position
Treatment effeciency.
The pretreatment subelement 302 is used for positioning region described in pretreatment, obtains target area.The target area bag
Containing knuckle areas and background area, the knuckle areas are the tuberosity focal area for determining, by reality region and around described
The peripheral region in reality region is constituted.Pretreatment can be carried out to initial data and/or area-of-interest.In some embodiments
In, the pretreatment can include the combination of one or more in enhancing, interpolation processing, Morphological scale-space, noise remove etc..
The purpose of Primary Location can be that the substantially area that doubtful tuberosity is located is determined in the medical image or in the positioning region
Domain is acted on simplifying the process of nodule segmentation and playing place mat as follow-up determination candidate target.Primary Location can be automatic
, it is automanual, manual etc..In certain embodiments, the data through pretreatment can be sent to other units or son
It is further processed in unit.For example, the data through pretreatment can be sent to tuberosity determination subelement, to determine knot
The segmentation result of section.The target area is obtained by positioning region described in pretreatment, subsequent singulation can be further reduced
Region, improves the accuracy and speed of subsequent singulation.
Tuberosity determination subelement U303, for processing the target area based on gauss hybrid models, obtains described
The probability graph of target area；According to the Morphological Model process probability graph, to determine the real property region and peripheral region of tuberosity.
Based on the gauss hybrid models process target area, the probability graph comprising different shape tuberosity is obtained；According to Morphological Model
The probability graph is processed, not only effectively splits and protect the real property region of tuberosity, while effectively extracting the peripheral region of tuberosity simultaneously
The main noise of the nodule segmentations such as blood vessel, trachea and lung wall is removed, the accuracy of nodule segmentation result is improved.
Memory element U200 can be the equipment with store function.The data that storage input block U100 is collected
The various data produced in (for example the medical image that, imaging device shoots) and the U300 work of meter processing unit.The storage is single
First U200 can be local, or long-range.Memory element U200 can by after information digitalization again with utilize
The storage device of electricity, the mode such as magnetically or optically is stored.Memory element U200 may also be used for depositing various information examples
Such as program and data.Data base 120 can be the equipment using electric energy mode storage information, such as various memorizeies, random
Access memorizer (Random Access Memory (RAM)), read only memory (Read Only Memory (ROM)) etc..With
On the storage device that refers to simply list some examples, the storage device that can be used in image segmentation system working environment is simultaneously
It is not limited to this.
In certain embodiments, output unit U400 can be to processing unit U300 input datas, it is also possible to reception processing
The data of unit U300 outputs, such as nodule segmentation result, and by the data of output in forms such as numeral, character, image, sound
Show.
It should be noted that between input block U100, memory element U200, processing unit U300, output unit U400
Connection or communication can be wired, or wireless.
Above for the description of segmenting system system, only for convenience of description, the application can not be limited in and lift enforcement
Within the scope of example.It is appreciated that for a person skilled in the art, after the principle for understanding the system, may not carry on the back
In the case of this principle, combination in any is carried out to unit, or composition subsystem is connected with other units, to implementing
Various amendments and change on said method and systematic difference field form and details.For example, memory element U200 can be
Cloud computing platform with data storage function, including but not limited to public cloud, private clound, community cloud and mixed cloud etc..Such as
Such deformation, within the protection domain of the application.
Although the present invention is disclosed as above with preferred embodiment, so it is not limited to the present invention, any this area skill
Art personnel, without departing from the spirit and scope of the present invention, when a little modification and perfect, therefore the protection model of the present invention can be made
Enclose when by being defined that claims are defined.
Claims (10)
1. a kind of image partition method, it is characterised in that comprise the steps：
Initial alignment is carried out to medical image, positioning region is obtained；
Positioning region described in pretreatment, obtains target area, and the target area includes knuckle areas and background area, the knot
Section region is constituted by reality region and around the peripheral region in the real property region；
Based on the gauss hybrid models process target area, the probability graph of the target area is obtained；
According to the Morphological Model process probability graph, to determine the real property region and peripheral region of tuberosity；
Merge the real property region and peripheral region, obtain image segmentation result.
2. image partition method as claimed in claim 1, it is characterised in that described initial alignment is carried out to medical image to include
Any one mode below：
Positioning region described in manually determined；Or
It is determined that through the long axis of the positioning region, the selected seed point on long axis determines institute using region growing methods
State positioning region；Or
Determine the positioning region based on pre-set programs.
3. image partition method as claimed in claim 1, it is characterised in that positioning region described in the pretreatment includes：
The positioning region is carried out to be based on the enhanced Threshold segmentation of extra large gloomy point, the target area is obtained.
4. the dividing method of medical image as claimed in claim 1, it is characterised in that described that institute is processed according to Morphological Model
Probability graph is stated, including：
Probability graph described in models treated is strengthened using the gloomy point in sea, obtaining the gloomy point in sea strengthens figure；
Based on each voxel or pixel in the probability graph, the corresponding voxel or pixel in the gloomy point enhancing figure in the sea is determined；
If the value of described arbitrary voxel or pixel in the probability graph is more than first threshold, and the gloomy point in the sea strengthens in figure
The value of corresponding voxel or pixel is more than Second Threshold, then the voxel or pixel belong to the real property region；
Otherwise, the voxel or pixel belong to background area.
5. the dividing method of medical image as claimed in claim 4, it is characterised in that the span of the first threshold is
(0,1) between constant, the span of the Second Threshold for (0,100) between constant.
6. the dividing method of medical image as claimed in claim 4, it is characterised in that described based on gauss hybrid models process
The target area, also includes：
Probability graph described in models treated is strengthened using extra large gloomy line, obtaining extra large gloomy line strengthens figure；
Based on each voxel or pixel in the probability graph, the corresponding voxel or pixel in the extra large gloomy line enhancing figure is determined；
The value of the voxel in the probability graph or pixel is compared with first threshold；
The value of the voxel in the extra large gloomy line enhancing figure or pixel is compared with the 3rd threshold value；
If the value of described arbitrary voxel or pixel in the probability graph is more than first threshold, and the Hai Sen lines strengthen figure
In corresponding voxel or pixel value be less than the 3rd threshold value, then the voxel or pixel belong to the peripheral region；
Otherwise, the voxel or pixel belong to background area.
7. the dividing method of medical image as claimed in claim 6, it is characterised in that the span of the 3rd threshold value is
(0, constant 50).
8. the image partition method as described in claim 1,4 or 6 any one, it is characterised in that the target area, described general
The locus of arbitrary voxel or pixel are to correspond in rate figure, the gloomy point enhancing in the sea and the extra large gloomy line enhancing figure.
9. a kind of image segmentation system, it is characterised in that include:
Input block, for obtaining medical image；
Processing unit, for processing the medical image, obtains the segmentation result of target area, and the target area includes tuberosity
Region and background area, the knuckle areas are constituted by reality region and around the peripheral region in the real property region；
And output unit and memory element；
The processing unit includes tuberosity determination subelement, for processing the target area based on gauss hybrid models, obtains
The probability graph of the target area；According to the Morphological Model process probability graph, to determine real property region and the surrounding of tuberosity
Region.
10. computer-aided detection system as claimed in claim 9, it is characterised in that the cutting unit also includes：
Locator unit, for carrying out initial alignment to medical image, obtains positioning region；
Pretreatment subelement, for positioning region described in pretreatment, obtains target area.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611092346.4A CN106611413A (en) | 2016-11-30 | 2016-11-30 | Image segmentation method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611092346.4A CN106611413A (en) | 2016-11-30 | 2016-11-30 | Image segmentation method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106611413A true CN106611413A (en) | 2017-05-03 |
Family
ID=58636281
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611092346.4A Pending CN106611413A (en) | 2016-11-30 | 2016-11-30 | Image segmentation method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106611413A (en) |
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107292889A (en) * | 2017-06-14 | 2017-10-24 | 上海联影医疗科技有限公司 | A kind of method of lesion segmentation, system and computer-readable recording medium |
| CN107464253A (en) * | 2017-07-10 | 2017-12-12 | 北京小米移动软件有限公司 | eyebrow location method and device |
| CN107610145A (en) * | 2017-07-26 | 2018-01-19 | 同济大学 | A kind of automatic pancreas dividing method based on adaptive threshold and template matches |
| CN107818577A (en) * | 2017-10-26 | 2018-03-20 | 滁州学院 | A kind of Parts Recognition and localization method based on mixed model |
| CN108154503A (en) * | 2017-12-13 | 2018-06-12 | 西安交通大学医学院第附属医院 | A kind of leucoderma state of an illness diagnostic system based on image procossing |
| CN108805883A (en) * | 2018-06-08 | 2018-11-13 | Oppo广东移动通信有限公司 | A kind of image partition method, image segmentation device and electronic equipment |
| CN108961281A (en) * | 2018-03-28 | 2018-12-07 | 研靖信息科技（上海）有限公司 | A kind of image partition method and equipment based on 3D voxel data image |
| CN109741360A (en) * | 2019-01-07 | 2019-05-10 | 上海联影医疗科技有限公司 | A kind of Bones and joints dividing method, device, terminal and readable medium |
| CN109801298A (en) * | 2019-01-21 | 2019-05-24 | 刘鹏程 | Specific aim disease treatment platform |
| CN109829920A (en) * | 2019-02-25 | 2019-05-31 | 上海商汤智能科技有限公司 | Image processing method and device, electronic equipment and storage medium |
| CN110246144A (en) * | 2019-06-17 | 2019-09-17 | 吉林大学第一医院 | A kind of the blood vessel Enhancement Method and system of lung CT image |
| CN110610481A (en) * | 2019-08-06 | 2019-12-24 | 深圳市旭东数字医学影像技术有限公司 | Automatic brain stem segmentation method and system based on medical image |
| CN111369580A (en) * | 2020-03-03 | 2020-07-03 | 深圳市安健科技股份有限公司 | Intelligent segmentation method and terminal for X-ray image |
| CN111383191A (en) * | 2019-12-11 | 2020-07-07 | 北京深睿博联科技有限责任公司 | Image processing method and device for repairing blood vessel fracture |
| CN113243923A (en) * | 2021-05-19 | 2021-08-13 | 中科芯未来微电子科技成都有限公司 | Method and device for improving accuracy of electrocardiogram monitoring |
| CN113940701A (en) * | 2021-10-11 | 2022-01-18 | 什维新智医疗科技(上海)有限公司 | Trachea and blood vessel detection device based on thyroid gland |
| CN114187252A (en) * | 2021-12-03 | 2022-03-15 | 推想医疗科技股份有限公司 | Image processing method and device, and method and device for adjusting detection frame |
| CN114972286A (en) * | 2022-06-09 | 2022-08-30 | 广州医科大学附属第一医院（广州呼吸中心） | Pulmonary nodule auxiliary positioning system based on image processing |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030035507A1 (en) * | 2001-08-17 | 2003-02-20 | Li-Yueh Hsu | Computer-aided diagnosis system for thoracic computer tomography images |
| CN105488800A (en) * | 2015-11-30 | 2016-04-13 | 上海联影医疗科技有限公司 | Feature extracting method and computer-aided diagnosis method and system |
| CN105513036A (en) * | 2014-09-26 | 2016-04-20 | 上海联影医疗科技有限公司 | Three-dimensional CT image segmentation method and three-dimensional CT image segmentation device |
| CN105976367A (en) * | 2016-04-29 | 2016-09-28 | 上海联影医疗科技有限公司 | Image segmentation method, pulmonary nodule detection method and computer-aided detection system |
-
2016
- 2016-11-30 CN CN201611092346.4A patent/CN106611413A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030035507A1 (en) * | 2001-08-17 | 2003-02-20 | Li-Yueh Hsu | Computer-aided diagnosis system for thoracic computer tomography images |
| CN105513036A (en) * | 2014-09-26 | 2016-04-20 | 上海联影医疗科技有限公司 | Three-dimensional CT image segmentation method and three-dimensional CT image segmentation device |
| CN105488800A (en) * | 2015-11-30 | 2016-04-13 | 上海联影医疗科技有限公司 | Feature extracting method and computer-aided diagnosis method and system |
| CN105976367A (en) * | 2016-04-29 | 2016-09-28 | 上海联影医疗科技有限公司 | Image segmentation method, pulmonary nodule detection method and computer-aided detection system |
Cited By (25)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107292889A (en) * | 2017-06-14 | 2017-10-24 | 上海联影医疗科技有限公司 | A kind of method of lesion segmentation, system and computer-readable recording medium |
| CN107292889B (en) * | 2017-06-14 | 2020-09-25 | 上海联影医疗科技有限公司 | Tumor segmentation method, system and readable medium |
| CN107464253A (en) * | 2017-07-10 | 2017-12-12 | 北京小米移动软件有限公司 | eyebrow location method and device |
| CN107464253B (en) * | 2017-07-10 | 2020-11-20 | 北京小米移动软件有限公司 | Eyebrow positioning method and device |
| CN107610145A (en) * | 2017-07-26 | 2018-01-19 | 同济大学 | A kind of automatic pancreas dividing method based on adaptive threshold and template matches |
| CN107610145B (en) * | 2017-07-26 | 2020-08-14 | 同济大学 | Automatic pancreas segmentation method based on adaptive threshold and template matching |
| CN107818577A (en) * | 2017-10-26 | 2018-03-20 | 滁州学院 | A kind of Parts Recognition and localization method based on mixed model |
| CN108154503A (en) * | 2017-12-13 | 2018-06-12 | 西安交通大学医学院第附属医院 | A kind of leucoderma state of an illness diagnostic system based on image procossing |
| CN108961281A (en) * | 2018-03-28 | 2018-12-07 | 研靖信息科技（上海）有限公司 | A kind of image partition method and equipment based on 3D voxel data image |
| CN108805883A (en) * | 2018-06-08 | 2018-11-13 | Oppo广东移动通信有限公司 | A kind of image partition method, image segmentation device and electronic equipment |
| CN109741360A (en) * | 2019-01-07 | 2019-05-10 | 上海联影医疗科技有限公司 | A kind of Bones and joints dividing method, device, terminal and readable medium |
| CN109741360B (en) * | 2019-01-07 | 2022-02-22 | 上海联影医疗科技股份有限公司 | Bone joint segmentation method, device, terminal and readable medium |
| CN109801298A (en) * | 2019-01-21 | 2019-05-24 | 刘鹏程 | Specific aim disease treatment platform |
| CN109829920A (en) * | 2019-02-25 | 2019-05-31 | 上海商汤智能科技有限公司 | Image processing method and device, electronic equipment and storage medium |
| CN109829920B (en) * | 2019-02-25 | 2021-06-15 | 上海商汤智能科技有限公司 | Image processing method and device, electronic equipment and storage medium |
| CN110246144A (en) * | 2019-06-17 | 2019-09-17 | 吉林大学第一医院 | A kind of the blood vessel Enhancement Method and system of lung CT image |
| CN110610481A (en) * | 2019-08-06 | 2019-12-24 | 深圳市旭东数字医学影像技术有限公司 | Automatic brain stem segmentation method and system based on medical image |
| CN111383191A (en) * | 2019-12-11 | 2020-07-07 | 北京深睿博联科技有限责任公司 | Image processing method and device for repairing blood vessel fracture |
| CN111383191B (en) * | 2019-12-11 | 2024-03-08 | 北京深睿博联科技有限责任公司 | Image processing method and device for vascular fracture repair |
| CN111369580A (en) * | 2020-03-03 | 2020-07-03 | 深圳市安健科技股份有限公司 | Intelligent segmentation method and terminal for X-ray image |
| CN111369580B (en) * | 2020-03-03 | 2023-06-20 | 深圳市安健科技股份有限公司 | Intelligent segmentation method and terminal for X-ray image |
| CN113243923A (en) * | 2021-05-19 | 2021-08-13 | 中科芯未来微电子科技成都有限公司 | Method and device for improving accuracy of electrocardiogram monitoring |
| CN113940701A (en) * | 2021-10-11 | 2022-01-18 | 什维新智医疗科技(上海)有限公司 | Trachea and blood vessel detection device based on thyroid gland |
| CN114187252A (en) * | 2021-12-03 | 2022-03-15 | 推想医疗科技股份有限公司 | Image processing method and device, and method and device for adjusting detection frame |
| CN114972286A (en) * | 2022-06-09 | 2022-08-30 | 广州医科大学附属第一医院（广州呼吸中心） | Pulmonary nodule auxiliary positioning system based on image processing |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106611413A (en) | Image segmentation method and system | |
| Liu et al. | Multi-view multi-scale CNNs for lung nodule type classification from CT images | |
| Roth et al. | Deep convolutional networks for pancreas segmentation in CT imaging | |
| Yu et al. | Liver vessels segmentation based on 3d residual U-NET | |
| CN105976367B (en) | Image partition method, pulmonary nodule detection method and its computer-aided detection system | |
| CN110310287B (en) | Automatic organ-at-risk delineation method, equipment and storage medium based on neural network | |
| Liu et al. | Spatial feature fusion convolutional network for liver and liver tumor segmentation from CT images | |
| CN109389129A (en) | A kind of image processing method, electronic equipment and storage medium | |
| CN109215033A (en) | The method and system of image segmentation | |
| Gao et al. | On combining morphological component analysis and concentric morphology model for mammographic mass detection | |
| Shaukat et al. | Computer-aided detection of lung nodules: a review | |
| Liu et al. | A CADe system for nodule detection in thoracic CT images based on artificial neural network | |
| CN109978037A (en) | Image processing method, model training method, device and storage medium | |
| Pan et al. | Mitosis detection techniques in H&E stained breast cancer pathological images: A comprehensive review | |
| CN109363698A (en) | A kind of method and device of breast image sign identification | |
| Ahmad et al. | Efficient liver segmentation from computed tomography images using deep learning | |
| Bergtholdt et al. | Pulmonary nodule detection using a cascaded SVM classifier | |
| Liu et al. | IOUC-3DSFCNN: Segmentation of brain tumors via IOU constraint 3D symmetric full convolution network with multimodal auto-context | |
| Song et al. | Kidney segmentation in CT sequences using SKFCM and improved GrowCut algorithm | |
| Liao et al. | A segmentation method for lung parenchyma image sequences based on superpixels and a self-generating neural forest | |
| CN110689525A (en) | Method and device for recognizing lymph nodes based on neural network | |
| Guo et al. | High performance lung nodule detection schemes in CT using local and global information | |
| Li et al. | 3D tumor detection in automated breast ultrasound using deep convolutional neural network | |
| Shan et al. | SCA-Net: A spatial and channel attention network for medical image segmentation | |
| Zheng et al. | Automatic liver tumour segmentation in CT combining FCN and NMF-based deformable model |
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: 20170503 |