CN109712163A - Coronary artery extracting method, device, image processing workstations and readable storage medium storing program for executing - Google Patents

Coronary artery extracting method, device, image processing workstations and readable storage medium storing program for executing Download PDF

Info

Publication number
CN109712163A
CN109712163A CN201811478488.3A CN201811478488A CN109712163A CN 109712163 A CN109712163 A CN 109712163A CN 201811478488 A CN201811478488 A CN 201811478488A CN 109712163 A CN109712163 A CN 109712163A
Authority
CN
China
Prior art keywords
coronary artery
voxel
volume data
sampling
trunk
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
Application number
CN201811478488.3A
Other languages
Chinese (zh)
Other versions
CN109712163B (en
Inventor
沈建华
王晓东
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shanghai United Imaging Healthcare Co Ltd
Original Assignee
Shanghai United Imaging Healthcare Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Shanghai United Imaging Healthcare Co Ltd filed Critical Shanghai United Imaging Healthcare Co Ltd
Priority to CN201811478488.3A priority Critical patent/CN109712163B/en
Publication of CN109712163A publication Critical patent/CN109712163A/en
Application granted granted Critical
Publication of CN109712163B publication Critical patent/CN109712163B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Landscapes

  • Apparatus For Radiation Diagnosis (AREA)
  • Magnetic Resonance Imaging Apparatus (AREA)

Abstract

This application involves a kind of coronary artery extracting method, device, image processing workstations and readable storage medium storing program for executing.The described method includes: comparing preset coronary artery trunk model and the heart volume data comprising target coronary artery, registration parameter is obtained；According to the coronary artery trunk location information and the registration parameter of the coronary artery trunk model, target coronary artery trunk volume data is extracted from the heart volume data；According to region growing model, region growing processing is carried out using the target coronary artery trunk volume data as seed voxels, obtains the target coronary artery in the heart volume data.Complete target coronary artery trunk volume data can be extracted from heart volume data using this method, and extract complete target coronary artery.

Description

Coronary artery extracting method, device, image processing workstations and readable storage medium storing program for executing
Technical field
This application involves technical field of medical image processing, at a kind of coronary artery extracting method, device, image Manage work station and readable storage medium storing program for executing.
Background technique
Cardiovascular disease is the higher disease of morbidity and mortality, and has the characteristics such as morbidity is anxious, concealment is strong, because This realizes that the diagnosis of heart disease has highly important clinical meaning.Currently, along with the image taking speed and scanning accuracy of CT Raising, CT medical image has been widely used in cardiac work up and medical diagnosis on disease.At the same time, in order to painstaking effort are better achieved The inspection of pipe disease, heart coronary artery is extracted to be widely used as a kind of supplementary means；It can accurately extract coronary artery Trunk profile and left and right coronary artery tree profile, the observation that doctor can be allowed more convenient by the coronary artery of extraction is narrow, calcification and spot Situations such as block, to provide foundation to the early prevention of cardiovascular disease and diagnosis for doctor.
Divide field in heart coronary artery, may be generally based upon CT shadowgraph technique and radiography is carried out to coronary artery, obtain comprising coronary artery Heart volume data, then the tubulose feature in heart volume data is enhanced using the technology of the similar gloomy enhancing in sea, finally Result based on enhancing carries out region growth, to complete the segmentation of coronary artery.
However, the cutting techniques increased based on region are coronary angiography is not good enough or hat when radiography is with the presence of interruption Arteries and veins trunk extracts incomplete problem, eventually leads to coronary artery and extracts incomplete problem.
Summary of the invention
Based on this, it is necessary in view of the above technical problems, provide a kind of coronary artery extracting method for capableing of complete extraction coronary artery, Device, image processing workstations and readable storage medium storing program for executing.
In a first aspect, a kind of coronary artery extracting method, which comprises
Preset coronary artery trunk model and the heart volume data comprising target coronary artery are compared, registration parameter is obtained；
According to the coronary artery trunk location information and the registration parameter of the coronary artery trunk model, from the heart volume data In extract target coronary artery trunk volume data；
According to region growing model, carried out the target coronary artery trunk volume data as seed voxels at region growing Reason, obtains the target coronary artery in the heart volume data.
The preset coronary artery trunk model of the comparison and the heart body number comprising target coronary artery in one of the embodiments, According to obtaining registration parameter, comprising:
The cardiac position information of the heart volume data is obtained from the heart volume data；
The cardiac position information of the cardiac position information of the coronary artery trunk model and the heart volume data is compared It is right, obtain the registration parameter.
The coronary artery trunk location information according to the coronary artery trunk model and described match in one of the embodiments, Quasi- parameter extracts target coronary artery trunk volume data from the heart volume data, comprising:
The coronary artery trunk location information is converted according to the registration parameter, obtains transformed coronary artery trunk position Confidence breath；
It is extracted from the heart volume data and the transformed matched volume data of coronary artery trunk location information, work For target coronary artery trunk volume data.
The target coronary artery trunk volume data is made according to region growing model described in one of the embodiments, Region growing processing is carried out for seed voxels, before obtaining the target coronary artery in the heart volume data, the method also includes:
Predeterminable area around each benchmark voxel of the target coronary artery trunk volume data is sampled, and obtains including institute State multiple sampling voxels including benchmark voxel；
Calculate the probability value that each sampling voxel belongs to coronary artery center voxel；The coronary artery center voxel is coronary artery center Voxel on line；
According to the probability value of each sampling voxel, one is chosen respectively from the corresponding multiple sampling voxels of each benchmark voxel A sampling voxel forms new target coronary artery trunk volume data.
The probability value for calculating each sampling voxel and belonging to coronary artery center voxel in one of the embodiments, packet It includes:
Sea gloomy enhancing processing is carried out to the heart volume data, obtains enhancing treated heart volume data；
According to the heart volume data before enhancing treated the heart volume data and enhancing processing, each sampling voxel is calculated The changing value of voxel value after enhancing processing and before enhancing processing；
According to the changing value of the voxel value of each sampling voxel, calculates each sampling voxel and belong to coronary artery center voxel Probability value.
The probability value for calculating each sampling voxel and belonging to coronary artery center voxel in one of the embodiments, packet It includes:
Each sampling voxel is inputted into preset machine learning model, each sampling voxel is exported and belongs to coronary artery center The probability value of voxel；The training sample of the machine learning model includes: positive sample and negative sample, and the input of the positive sample is Voxel on coronary artery center line exports as probability value 1, voxel of the input of the negative sample for coronary artery center line periphery, output For probability value 0.
It is described default around each benchmark voxel of the target coronary artery trunk volume data in one of the embodiments, Region is sampled, and multiple sampling voxels including the benchmark voxel are obtained, comprising:
For each benchmark voxel, each datum level where each benchmark voxel is determined；Each datum level is vertical The corresponding coronary artery trunk center line of the target coronary artery trunk volume data；
It is sampled on each datum level according to pre-set radius and preset step-length, obtains including that the benchmark voxel exists Interior multiple sampling voxels.
The probability value according to each sampling voxel in one of the embodiments, it is corresponding from each benchmark voxel A sampling voxel is chosen in multiple sampling voxels respectively, forms new target coronary artery trunk volume data, comprising:
According to the probability value and preset probability function of each sampling voxel, from the corresponding multiple samplings of each benchmark voxel A sampling voxel is chosen in voxel respectively；The value of the corresponding probability function of each sampling voxel of the selection is maximum, described general Rate function includes: the positive coefficient linear combination of the first probability function and the second probability function, and first probability function is to choose Each sampling voxel the sum of probability value, second probability function is the sum of the probability value that each segmentation belongs to coronary artery segmentation；Institute State the heart volume data in each sampling voxel for be respectively segmented into selection between each group neighbouring sample voxel；
Obtain the new target coronary artery trunk volume data being made of each sampling voxel of the selection.
Each segmentation belongs to the calculation of the probability value of coronary artery segmentation in one of the embodiments, comprising:
Heart volume data between each group neighbouring sample voxel is sampled, multiple slight bar voxels are obtained；
It calculates the multiple slight bar voxel and belongs to the probability value of coronary artery center voxel, and calculate each probability value Characteristic value belongs to the probability value of coronary artery segmentation as the segmentation.
In one of the embodiments, the method also includes:
Calculate the radius of deflection in the sampling voxel respectively chosen between each group neighbouring sample voxel；The radius of deflection For distance of the one group of neighbouring sample voxel between two subpoints on datum level, the vertical target of the datum level The corresponding coronary artery trunk center line of coronary artery trunk volume data, and by any sampling body in one group of neighbouring sample voxel Element；
If the radius of deflection between one group of neighbouring sample voxel is greater than default bias radius threshold, to described one group Two sampling voxels in neighbouring sample voxel are chosen again.
It is described according to region growing model in one of the embodiments, using the target coronary artery trunk volume data as Seed voxels carry out region growing processing, obtain the target coronary artery in the heart volume data, comprising:
Using each sampling voxel of the target coronary artery trunk volume data as seed voxels, based on default growth conditions, carry out Region growing processing, obtains the initial target coronary artery in the heart volume data；The default growth conditions includes: and seed body The probability value that the voxel of element connection belongs to coronary artery is greater than or equal to the first probability threshold value；
Micronization processes are carried out to the initial target coronary artery, obtain target coronary artery skeleton；
Based on preset cutting condition, cutting processing is carried out to the target coronary artery skeleton, the target coronary artery after being cut Skeleton；The preset cutting condition includes: that the voxel of the target coronary artery skeleton belongs to the probability value of coronary artery less than the second probability Threshold value, wherein first probability threshold value is less than the second probability threshold value；
Region growth is carried out to the target coronary artery skeleton after the cutting according to preset distance field, obtains the heart body Target coronary artery in data.
Second aspect, a kind of coronary artery extraction element, described device include:
Coronary artery Model registration module, for comparing preset coronary artery trunk model and the heart body number comprising target coronary artery According to obtaining registration parameter；
Coronary artery trunk extraction module, for the coronary artery trunk location information and the registration according to the coronary artery trunk model Parameter extracts target coronary artery trunk volume data from the heart volume data；
Coronary artery trunk pop-in upgrades is used for according to region growing model, using the target coronary artery trunk volume data as kind Daughter element carries out region growing processing, obtains the target coronary artery in the heart volume data.
The third aspect, a kind of image processing workstations, including memory and processor, the memory are stored with computer Program, the processor perform the steps of when executing the computer program
Preset coronary artery trunk model and the heart volume data comprising target coronary artery are compared, registration parameter is obtained；
According to the coronary artery trunk location information and the registration parameter of the coronary artery trunk model, from the heart volume data In extract target coronary artery trunk volume data；
According to region growing model, carried out the target coronary artery trunk volume data as seed voxels at region growing Reason, obtains the target coronary artery in the heart volume data.
Fourth aspect, a kind of readable storage medium storing program for executing are stored thereon with computer program, and the computer program is by processor It is performed the steps of when execution
Preset coronary artery trunk model and the heart volume data comprising target coronary artery are compared, registration parameter is obtained；
According to the coronary artery trunk location information and the registration parameter of the coronary artery trunk model, from the heart volume data In extract target coronary artery trunk volume data；
According to region growing model, carried out the target coronary artery trunk volume data as seed voxels at region growing Reason, obtains the target coronary artery in the heart volume data.
Above-mentioned coronary artery extracting method, device, image processing workstations and readable storage medium storing program for executing, image processing workstations can be with By comparing preset coronary artery trunk model with comprising the heart volume data of target coronary artery, registration parameter is obtained；Then basis is matched The coronary artery trunk location information of quasi- parameter and coronary artery trunk model, can extract target coronary artery trunk from heart volume data Volume data, because the coronary artery trunk in coronary artery trunk model is complete, the target coronary artery trunk volume data extracted It is complete, therefore is based ultimately upon complete target coronary artery trunk volume data and carries out the target coronary artery that region growing is handled It is complete；In short, the coronary artery extracting method of the present embodiment can extract complete target coronary artery master from heart volume data Stem body data, and extract complete target coronary artery.
Detailed description of the invention
Fig. 1 is the applied environment figure of coronary artery extracting method in one embodiment；
Fig. 2 a is the flow diagram of coronary artery extracting method in one embodiment；
Fig. 2 b is the schematic diagram of coronary artery trunk model in one embodiment；
Fig. 3 is the flow diagram of the acquisition process of registration parameter in one embodiment；
Fig. 4 a is the flow diagram of the extraction process of target coronary artery trunk volume data in one embodiment；
Fig. 4 b is the figure one of the schematic cross-section of one embodiment cardiac volume data；
Fig. 5 a is the flow diagram of the optimization process of target coronary artery trunk volume data in one embodiment；
Fig. 5 b is the schematic diagram of the optimization process of target coronary artery trunk volume data in one embodiment；
Fig. 5 c is the figure two of the schematic cross-section of one embodiment cardiac volume data；
Fig. 6 a is the flow diagram of the optimization extraction process of target coronary artery in one embodiment；
Fig. 6 b is the figure three of the schematic cross-section of one embodiment cardiac volume data；
Fig. 7 is the structural block diagram of coronary artery extraction element in one embodiment；
Fig. 8 is the structural block diagram of coronary artery extraction element in another embodiment；
Fig. 9 is the internal structure chart of image processing workstations in one embodiment.
Specific embodiment
It is with reference to the accompanying drawings and embodiments, right in order to which the objects, technical solutions and advantages of the application are more clearly understood The application is further elaborated.It should be appreciated that specific embodiment described herein is only used to explain the application, not For limiting the application.
Coronary artery extracting method provided by the present application, can be applied in application environment as shown in Figure 1.Wherein, PACS (Picture Archiving and Communication Systems, image archiving and communication system) server 102 difference With at least one medical scanning apparatus 103, image processing workstations 101 connect, connection type can for network connection or DICOM (Digital Imaging and Communications in Medicine, digital imaging and communications in medicine) connection Deng.Wherein, medical scanning apparatus 103 can be CT equipment, PET-CT equipment (Positron Emission Computed Tomography), MRI Equipment (magnetic resonance imaging), ultrasonic device etc., can be by various ways such as emphasis scannings after unenhanced, injection contrast agent to tested Object is scanned, and obtains the heart scanning image sequence comprising coronary artery；The available medical scanning apparatus hair of PACS server The heart scanning image sequence that send simultaneously is saved in digitized form, heart scanning image sequence can also be sent to image Processing workstation；Image processing workstations can carry out the processing such as three-dimensional reconstruction to heart scanning image sequence, obtain comprising hat The heart volume data of arteries and veins, and it is partitioned into coronary artery.Certainly, image processing method provided by the present application also can be applied to medicine and sweep Retouch equipment 103 and application environment that image processing workstations 101 are directly connected in, be not necessarily to PACS server；In short, the present embodiment This is not intended to limit.
In one embodiment, as shown in Figure 2 a, a kind of coronary artery extracting method is provided, is applied in Fig. 1 in this way Image processing workstations for be illustrated, comprising the following steps:
S201 compares preset coronary artery trunk model and the heart volume data comprising target coronary artery, obtains registration parameter.
Coronary artery trunk model can be the cardiac structure model for including coronary artery trunk, the coronary artery trunk in coronary artery trunk model Location information is known, it can be understood as the three-dimensional coordinate of each point of coronary artery trunk (or the center line for being interpreted as coronary artery trunk) It is known；Wherein, cardiac structure model may include complete heart structure, can also to include part of heart structure, such as Aorta structure.The schematic diagram of the coronary artery trunk model referring to shown in Fig. 2 b shows the three-dimensional knot of aorta and coronary artery trunk Structure, wherein coronary artery trunk may include right hat, left hand and Zuo Jiangzhi.It is understood that coronary artery in practice is in addition to coronary artery It can also include the coronary artery ramuscule that extension, bifurcated go out on coronary artery trunk outside trunk；For different subjects (same to species) For, the three-dimensional structure of coronary artery ramuscule is different, and the three-dimensional structure of coronary artery trunk is similar, i.e. coronary artery trunk Location information is similar.
It should be noted that the coronary artery trunk model in the present embodiment can be and the infraspecific coronary artery trunk of subject Model can be generally the standard coronary artery trunk model of authoritative institution's publication；Meanwhile it can be according to the different coronary arterys of subject Type selects the coronary artery trunk model with the coronary artery type matching of subject, for example, coronary artery type can be divided into right advantage Type, balanced type, left advantage type etc..In addition, coronary artery trunk model is also possible to the coronary artery master for subject stored in advance Dry model is also possible to use for example, it may be passing through the coronary artery trunk model that other detection modes obtain before subject The coronary artery trunk model that the coronary artery extracting method of the present embodiment obtains.In short, the present embodiment is not intended to limit this.
It is understood that can also include that each heart is special for the ease of comparing, in coronary artery trunk model generally Sign, such as the feature more easy to identify such as cardiac silhouette feature, heart chamber feature, heart aorta feature, convenient for each heart Coronary artery is positioned on the basis of feature.In the present embodiment, image processing workstations can by preset coronary artery trunk model with comprising The heart volume data of target coronary artery is compared based on the same heart features of the two.Illustratively, for the coronary artery master of Fig. 2 b For dry model, image processing workstations are using coronary artery trunk model as template, using image registration algorithm by heart volume data It is registrated with the coronary artery trunk model, allows the aorta feature in heart volume data and the master in coronary artery trunk model Artery characteristics are overlapped, and obtain corresponding registration parameter at this time.Wherein, above-mentioned registration can be similar variation, correspondingly, registration ginseng Number may include similar running parameter, which can become for rotationally-varying parameter, translation running parameter and scaling Change one of parameter or a variety of combinations.
S202, according to the coronary artery trunk location information and the registration parameter of the coronary artery trunk model, from the heart Target coronary artery trunk volume data is extracted in volume data.
In the present embodiment, which can be registration parameter of the heart volume data relative to coronary artery trunk model, Then it is understood that image processing workstations can convert above-mentioned heart volume data according to above-mentioned registration parameter, become Heart volume data after changing should be overlapped with the same heart features between above-mentioned coronary artery trunk model, i.e. the coronary artery trunk of the two Also it should be overlapped, therefore image processing workstations can be according to above-mentioned coronary artery trunk location information from above-mentioned transformed heart body It obtains with the matched volume data of coronary artery trunk location information in data as target coronary artery trunk volume data.
It should be noted that above-mentioned coronary artery trunk location information can be the three-dimensional coordinate of each point of composition coronary artery trunk, Then image processing workstations can obtain matched with the three-dimensional coordinate of the premises from above-mentioned transformed heart volume data Voxel, then target coronary artery trunk volume data includes above-mentioned each matched voxel；Target coronary artery trunk volume data can also include upper The voxel around each matched voxel in predeterminable area is stated, can also include in each matched voxel between each group adjacent voxels The voxel that is sampled of heart volume data.It is understood that each point in heart volume data with composition coronary artery trunk The corresponding point of three-dimensional coordinate be point on target coronary artery trunk, therefore above-mentioned matched voxel can be above-mentioned target coronary artery Voxel belonging to point on trunk is also possible to the distance between point on above-mentioned target coronary artery trunk in pre-determined distance threshold value Within the obtained voxel of voxel or other matching rules.
It is raw to carry out region using the target coronary artery trunk volume data as seed voxels according to region growing model by S203 Long processing, obtains the target coronary artery in the heart volume data.
Illustratively, each voxel of the target coronary artery trunk volume data can be respectively seed by image processing workstations Voxel, judges whether the voxel being connected to each seed voxels meets default growth conditions, if satisfied, then the voxel belongs to target hat Arteries and veins, and the voxel for belonging to target coronary artery is re-used as seed voxels and carries out region growing processing, in this way, constantly by region growing The voxel of obtained new target coronary artery carries out region growing processing as new seed voxels, until all and new seed bodies Until the voxel of element connection is unsatisfactory for default growth conditions, the target coronary artery in heart volume data is finally obtained.Certainly, above-mentioned area For domain growth process generally there are limited area, which can be the region for including above-mentioned target coronary artery trunk volume data, It can also entire heart volume data region.
Above-mentioned default growth conditions may include: the voxel being connected to seed voxels voxel value and above-mentioned target coronary artery master The difference of dry Feature element value is less than preset threshold, and the Feature element value of above-mentioned target coronary artery trunk can be target coronary artery master The characteristic values such as median, the average value of voxel value of each voxel of stem body data；Above-mentioned default growth conditions is also possible to and plants The probability value that the voxel of daughter element connection belongs to coronary artery is greater than or equal to the first probability threshold value, the above-mentioned body being connected to seed voxels Described in the calculation that element belongs to the probability value of coronary artery sees below, which is not described herein again.In short, the present embodiment is not intended to limit State growth conditions.
The coronary artery extracting method of the present embodiment, image processing workstations can by compare preset coronary artery trunk model with Heart volume data comprising target coronary artery, obtains registration parameter；Then according to the coronary artery master of registration parameter and coronary artery trunk model Dry location information can extract target coronary artery trunk volume data from heart volume data, because in coronary artery trunk model Coronary artery trunk is complete, therefore the target coronary artery trunk volume data extracted is also complete, therefore is based ultimately upon complete It is also complete that target coronary artery trunk volume data, which carries out the target coronary artery that region growing is handled,；In short, the hat of the present embodiment Arteries and veins extracting method can extract complete target coronary artery trunk volume data from heart volume data, and extract complete target Coronary artery.
Referring to shown in Fig. 3, what is involved is a kind of acquisition process of registration parameter for the present embodiment, specifically, above-mentioned S201 can To include:
S301 obtains the cardiac position information of the heart volume data from the heart volume data.
Above-mentioned cardiac position information can be aortic position information, cardiac silhouette location information etc., heart chamber position Information etc..Image processing workstations can use generalised Hough transform, and preset heart mesh model is projected to heart body number Heart chamber location information is obtained in, and according to edge iterative algorithm；Alternatively, it is rigid to carry out heart to heart volume data Body registration process obtains the cardiac position information of heart volume data；Certainly, image processing workstations can also be special according to other hearts Heart in sign identification heart volume data, to obtain cardiac position information, the present embodiment is not intended to limit this.
S302, by the cardiac position information of the cardiac position information of the coronary artery trunk model and the heart volume data into Row compares, and obtains the registration parameter.
Above-mentioned cardiac position information can be coordinate of each characteristic point of a certain heart features in three-dimensional system of coordinate.Example Property, based on the same characteristic point in the same heart features both in coronary artery trunk model and heart volume data in three-dimensional coordinate The comparison of coordinate in system can know translation parameters between the two；Same heart features based on the two are in three-dimensional coordinate The comparison of same characteristic dimension (such as aorta caliber) in system, can know zooming parameter between the two；Based on the two Same heart features in relative bearing of same group of two characteristic points in three-dimensional system of coordinate comparison, can know the two Between rotation parameter.In short, the image processing workstations of the present embodiment can be obtained by comparison methods such as method for registering images Registration parameter between coronary artery trunk model and heart volume data, the registration parameter may include zooming parameter, rotation parameter, put down Shifting parameter etc., rotation parameter generally can be 3 × 3 spin matrix, and zooming parameter is generally one-dimensional zoom factor.
In short, the coronary artery extracting method of the present embodiment can cardiac position information and heart body based on coronary artery trunk model The cardiac position information of data, obtains the registration parameter between coronary artery trunk model and heart volume data.
Referring to shown in Fig. 4 a, what is involved is a kind of extraction process of target coronary artery trunk volume data for the present embodiment, specifically, Above-mentioned S202 may include:
S401 converts the coronary artery trunk location information according to the registration parameter, obtains transformed coronary artery Trunk location information.
Illustratively, transformed coronary artery trunk location information rrIt can be with are as follows:
rr=sR (rl)+r0
Wherein, rlFor the coronary artery trunk location information before transformation；In registration parameter, s is zoom factor, and R is spin matrix, r0For translation parameters.
S402 is extracted and the transformed matched body number of coronary artery trunk location information from the heart volume data According to as target coronary artery trunk volume data.
Wherein, transformed coronary artery trunk location information rrCoronary artery trunk location information in corresponding heart volume data, therefore Image processing workstations can be according to rrIt is extracted from heart volume data and rrIn each matched volume data of three-dimensional coordinate (or body Element), as target coronary artery trunk volume data.Similarly, referring to the description of S202, above-mentioned target coronary artery trunk volume data be can wrap The volume data around above-mentioned each matched volume data in predeterminable area is included, can also include to each group phase in each matched volume data The volume data that heart volume data between adjacent volume data is sampled.
Referring to a kind of schematic cross-section of heart volume data shown in Fig. 4 b, extracted wherein being shown on the section Multiple voxels (voxel of right hat is not present on the section) of left descending branch and left hand in target coronary artery trunk volume data, and Actual left descending branch and left hand.As it can be seen that the target coronary artery trunk volume data extracted, is overlapped or connects with actual target coronary artery Closely, i.e., the present embodiment be based on coronary artery trunk model, just position obtained target coronary artery trunk volume data be it is believable, can be used as Actual target coronary artery trunk volume data, or subsequent optimization process provides a believable initial target coronary artery trunk Volume data.
In short, the coronary artery extracting method of the present embodiment coronary artery trunk location information based on coronary artery trunk model and can match Quasi- parameter extracts complete target coronary artery trunk volume data from heart volume data.
Referring to shown in Fig. 5 a, the present embodiment what is involved is the optimization process of the target coronary artery trunk volume data to extraction, Before S203, the coronary artery extracting method can also include:
S501, the predeterminable area around each benchmark voxel of the target coronary artery trunk volume data are sampled, are obtained Multiple sampling voxels including the benchmark voxel.
Each benchmark voxel of above-mentioned target coronary artery trunk volume data can be the voxel in each target coronary artery trunk volume data, It is also possible to the voxel sampled to target coronary artery trunk volume data.Illustratively, above-mentioned predeterminable area can be with Centered on said reference voxel, using the line between above-mentioned reference body element and its adjacent reference voxel as the cylindrical region of axis, on The radius for stating cylindrical region can be the maximum radius of coronary artery trunk, the length of above-mentioned cylindrical region can for each benchmark voxel it Between average distance.
It is understood that above-mentioned each equal very little of benchmark voxel, can regard a datum mark, the base as in calculating process It can be on schedule any point in said reference voxel, preferably central point；Similarly, sampling voxel etc. hereinafter can also To regard corresponding sampled point as.
Optionally, the predeterminable area around each benchmark voxel of the target coronary artery trunk volume data is adopted Sample obtains multiple sampling voxels including the benchmark voxel, may include: to determine institute for each benchmark voxel State each datum level where each benchmark voxel；The vertical corresponding coronary artery master of the target coronary artery trunk volume data of each datum level Dry center line；It is sampled on each datum level according to pre-set radius and preset step-length, obtains including the benchmark voxel Multiple sampling voxels inside.
Image processing workstations can be directed to each benchmark voxel, according to pre- on the datum level where above-mentioned each benchmark voxel If radius and preset step-length are sampled, using the point on the datum level where benchmark voxel as the center of circle, using pre-set radius as radius, It can determine a round sampling face (being equivalent to imaginary coronary artery section), be adopted on the circle sampling face with preset step-length Sample a, for example, point can be sampled by horizontal spacing and longitudinal pitch of preset step-length, using same using the center of circle as starting point Mode samples next point in turn, and the horizontal spacing and longitudinal pitch between finally obtained sampled point are above-mentioned preset step-length. Certainly, above example is only a kind of illustrative sample mode.
It is understood that according to pre- on the datum level where each benchmark voxel of above-mentioned target coronary artery trunk volume data If radius and preset step-length are sampled, it is equivalent to and is sampled on imaginary coronary artery section, compared to other sample modes, It is bigger to sample the probability value that voxel is coronary artery center, that is, samples more efficient.
Referring to shown in Fig. 5 b, target coronary artery trunk volume data can be presented as the form of the center line of target coronary artery trunk, Voxel in target coronary artery trunk volume data is presented as each benchmark voxel on the center line of target coronary artery trunk.Such as Fig. 5 b institute Show, for the i-th benchmark voxel, determines the i-th datum level where the i-th benchmark voxel；I-th datum level is perpendicular to the i-th benchmark voxel The segmentation of the coronary artery trunk center line at place；It can be sampled on the i-th datum level, available includes the i-th benchmark voxel 4 sampling voxels inside.Similarly, it for the (i-1)-th benchmark voxel, can up-sample to obtain including in the (i-1)-th datum level 3 sampling voxels including i-1 benchmark voxel.
S502 calculates the probability value that each sampling voxel belongs to coronary artery center voxel；The coronary artery center voxel is hat Voxel on arteries and veins center line.
Because the corresponding voxel value of different tissues is different, simplest one kind is achieved in that image processing workstations can To calculate the voxel value of each sampling voxel and the difference of standard unit value according to the corresponding standard unit value in coronary artery center；Then According to the transformational relation between preset difference and probability value, the difference of the voxel value of each sampling voxel and standard unit value is turned It is changed to the probability value that each sampling voxel belongs to coronary artery center voxel.It is understood that above-mentioned difference is smaller, above-mentioned probability value is got over Greatly.
In one embodiment, the S502 may include: to carry out the gloomy enhancing processing in sea to the heart volume data, obtain To enhancing treated heart volume data；According to the heart body number before enhancing treated the heart volume data and enhancing processing According to the changing value of voxel value of each sampling voxel of calculating after enhancing processing and before enhancing processing；According to each sampling voxel Voxel value changing value, calculate each sampling voxel and belong to the probability value of coronary artery center voxel.Sea gloomy enhancing processing can be with Enhance the tubulose feature in heart volume data, i.e., the voxel value for belonging to the voxel of coronary artery in heart volume data can become larger.At image Managing work station can be according to the transformational relation between preset changing value and probability value, by the variation of the voxel value of each sampling voxel Value is converted to the probability value that each sampling voxel belongs to coronary artery center voxel；Above-mentioned transformational relation can be a kind of functional form, It can be the form of corresponding table.It is understood that above-mentioned changing value is bigger, above-mentioned probability value is bigger.Illustratively, above-mentioned body The voxel value of element can generally be characterized by the gray value of each voxel, and above-mentioned transformational relation embodies are as follows: sampling voxel belongs to coronary artery Probability value=min (changing value of the gray value of sampling voxel, 100)/100 of center voxel, that is, think the gray scale for sampling voxel When being worth changing value greater than 100, the probability value that sampling voxel belongs to coronary artery center voxel is 1；If the gray value of a certain sampling voxel Changing value be 10, then the sampling voxel is that belong to the probability value of coronary artery center voxel be 0.1；If the gray scale of a certain sampling voxel The changing value of value is 105, then the sampling voxel is that belong to the probability value of coronary artery center voxel be 1.
In one embodiment, the S502 may include: that each sampling voxel is inputted preset machine learning Model exports the probability value that each sampling voxel belongs to coronary artery center voxel；The training sample packet of the machine learning model Include: positive sample and negative sample, the input of the positive sample are the voxel on coronary artery center line, are exported as probability value 1, the negative sample This input is the voxel on coronary artery center line periphery, is exported as probability value 0.Above-mentioned machine learning model can be neural network mould Type, or regression analysis model etc., the present embodiment is not intended to limit this.
The present embodiment can calculate sampling voxel using modes such as the gloomy enhancing in sea or machine learning models and belong in coronary artery The accuracy of judgement can be improved in the probability value of heart voxel, so that the target coronary artery trunk volume data that optimization obtains is more acurrate.
S503, according to the probability value of each sampling voxel, from the corresponding multiple sampling voxels of each benchmark voxel respectively A sampling voxel is chosen, new target coronary artery trunk volume data is formed.
It is understood that the purpose of the present embodiment is the probability value for belonging to coronary artery center voxel according to each sampling voxel, It chooses each sampling voxel and forms new target coronary artery trunk volume data, so that new target coronary artery trunk volume data belongs to coronary artery Probability is as big as possible.Therefore, in one embodiment, image processing workstations multiple can be adopted from each benchmark voxel is corresponding The maximum sampling voxel of probability value is chosen in sample voxel respectively, forms new target coronary artery trunk volume data.
As shown in Figure 5 b, image processing module can be directed to the (i-1)-th benchmark voxel, from the (i-1)-th benchmark voxel corresponding the The maximum sampling voxel X of probability value is selected in 3 sampling voxels on i-1 datum level；For the i-th benchmark voxel, from the i-th base The maximum sampling voxel Y of probability value is selected in 4 sampling voxels on corresponding i-th datum level of quasi- voxel；In this way, from each base Quasi- voxel is corresponding to sample the sampling voxel chosen respectively in voxel, forms new target coronary artery trunk volume data.
A kind of schematic cross-section of heart volume data referring to shown in Fig. 5 c, wherein after showing optimization on the section Multiple voxels of left descending branch and left hand in target coronary artery trunk volume data.As it can be seen that compared to the mesh before optimizing shown in Fig. 4 b Coronary artery trunk volume data is marked, optimized target coronary artery trunk volume data has been fully located at actual target coronary artery region In, i.e., it is more acurrate.
In short, the present embodiment can belong to the probability value of coronary artery according to each sampling voxel, the probability for belonging to coronary artery is chosen Biggish sampling voxel forms new target coronary artery trunk volume data, realizes the optimization to target coronary artery trunk volume data, improves The accuracy of the target coronary artery trunk volume data of extraction.
Optionally, the new target coronary artery trunk volume data obtained in the present embodiment can be used as the new of subject Coronary artery trunk model can obtain more targeted more preferably target coronary artery trunk volume data, reduce subsequent target hat The calculation amount of arteries and veins trunk optimization process and processing time.
Optionally, in order to further optimize to target coronary artery trunk volume data, the present embodiment is further related to selection The factor that volume data between sampling voxel belongs to the probability value of coronary artery considers that specifically, the S503 may include: according to institute The probability value and preset probability function for stating each sampling voxel, choose respectively from the corresponding multiple sampling voxels of each benchmark voxel One sampling voxel；The value of the corresponding probability function of each sampling voxel of the selection is maximum, and the probability function includes: first The positive coefficient linear combination of probability function and the second probability function, first probability function are the general of each sampling voxel chosen The sum of rate value, second probability function are the sum of the probability value that each segmentation belongs to coronary artery segmentation；It is described to be respectively segmented into selection Heart volume data in each sampling voxel between each group neighbouring sample voxel；What acquisition was made of each sampling voxel of the selection New target coronary artery trunk volume data.
Illustratively, above-mentioned probability function f can be indicated are as follows:
f(P0,P1..,PN)=k1 × Σ Pi+k2×ΣSeg(Pi,Pi-1)
It should be noted that P0,P1..,PNIt is the probability value for each sampling voxel chosen, Σ PiFor the first probability function, ΣSeg(Pi,Pi-1) it is the second probability function, k1 is the coefficient of the first probability function, and k2 is the coefficient of the second probability function, and K1 and k2 are all larger than 0.
Wherein, Seg (Pi,Pi-1) be in each sampling voxel chosen adjacent sampling voxel i and sampling voxel (i-1) it Between volume data belong to coronary artery segmentation probability value.In one embodiment, image processing workstations can be first to sampling body Volume data between plain i and sampling voxel (i-1) is sampled, and each slight bar voxel is obtained；Then each slight bar is calculated The Feature element value of voxel, and according to the voxel value and probability value of sampling voxel i, and the voxel value with sampling voxel (i-1) And probability value, the corresponding probability value of Feature element value is obtained as Seg (P by difference arithmetici,Pi-1) value.Another real It applies in mode, for adjacent sampling voxel i and sampling voxel (i-1), image processing workstations can calculate above-mentioned multiple thin Divide sampling voxel to belong to the probability value of coronary artery center voxel, and calculate the characteristic value of each probability value, belongs to hat as the segmentation The probability value of arteries and veins segmentation, i.e. Seg (Pi,Pi-1) value.It is understood that above-mentioned slight bar voxel belongs to coronary artery centerbody The calculation of the probability value of element is referred to the description in above-mentioned 502, and which is not described herein again.
Referring to shown in Fig. 5 b, for sampling voxel X adjacent in each sampling voxel of selection and sampling voxel Y, Seg (PY,PX) for sampling voxel X and the heart volume data between voxel Y can be sampled belong to the probability value of coronary artery center voxel, it can be with The average value or median of the probability value of three slight bar voxels between above-mentioned sampling voxel X and sampling voxel Y etc. are special Value indicative.
The present embodiment is it is further contemplated that the volume data between the sampling voxel chosen belongs to the probability value of coronary artery with excellent Change target coronary artery trunk volume data, realization advanced optimizes target coronary artery trunk volume data, improves the target coronary artery of extraction The accuracy of trunk volume data.
In addition, because coronary artery trunk should be it is continuous, in order to avoid the jump of coronary artery trunk, need to control adjacent two Radius of deflection between a sampling voxel controls the continuity between two neighboring sampling voxel, therefore the present embodiment further relates to The successional process that coronary artery trunk volume data is controlled by control radius of deflection, can specifically include: calculate each choosing Radius of deflection in the sampling voxel taken between each group neighbouring sample voxel；The radius of deflection is one group of neighbouring sample body Distance of the element between two subpoints on datum level, vertically the target coronary artery trunk volume data is corresponding for the datum level Coronary artery trunk center line, and by any sampling voxel in one group of neighbouring sample voxel；If one group of neighbouring sample Radius of deflection between voxel is greater than default bias radius threshold, then to two sampling bodies in one group of neighbouring sample voxel Element is chosen again.
Referring to shown in Fig. 5 b, for sampling voxel X adjacent in each sampling voxel of selection and sampling voxel Y, adopt Subpoint of the sample voxel Y on the (i-1)-th datum level through over-sampling voxel X is Y ', and samples voxel X on the (i-1)-th datum level Subpoint be itself, therefore the distance between Y ' and X shown in Fig. 5 b be sample voxel X and sampling voxel Y between Radius of deflection.If above-mentioned radius of deflection is greater than default bias radius threshold, it is meant that adjacent in each sampling voxel of selection to adopt Sample voxel X and sampling voxel Y between coronary artery trunk there may be jumps, do not meet continuous feature, may choose it is wrong, therefore It needs to choose again.Image processing workstations can select probability value from the corresponding 3 samplings voxel of the (i-1)-th benchmark voxel Secondary big sampling voxel replacement sampling voxel X, alternatively, selecting probability value from the corresponding 4 samplings voxel of the i-th benchmark voxel Secondary big sampling voxel replacement sampling voxel Y, re-starts the judgement of the condition of continuity, if satisfied, can then determine from (i-1)-th The sampling voxel chosen in datum level and the sampling voxel chosen from the i-th datum level；If not satisfied, being then referred to above Description is chosen again, can also re-start sampling；When the sampling voxel of all selections meets the above-mentioned condition of continuity, then most The sampling voxel determined eventually forms new target coronary artery trunk volume data.
In addition, in one embodiment, above-mentioned probability function may include: the first probability function, the second probability function and The linear combination of third offset function, wherein third offset function can be expressed as Σ Dev (Pi,Pi-1), wherein the first probability The coefficient of function and the second probability function is greater than 0, and the coefficient of third offset function is less than 0, Dev (Pi,Pi-1) it is each of selection Sample the size of the radius of deflection between sampling voxel i and sampling voxel (i-1) adjacent in voxel.
In short, above-described embodiment can control the continuity of coronary artery trunk volume data by control radius of deflection, avoid Coronary artery trunk volume data is in the presence of that jump is this and does not meet biocompatibility characteristics, advanced optimizes mentioning for coronary artery trunk volume data It takes, improves the accuracy that coronary artery extracts.
Referring to shown in Fig. 6 a, the present embodiment is what is involved is according to target coronary artery trunk volume data, at Low threshold growth Reason obtains initial target coronary artery, then micronization processes obtain target coronary artery skeleton, then are carried out using high threshold to target coronary artery skeleton It cuts, it is final that region growth is carried out to the target coronary artery skeleton after cutting using distance field, the process of target coronary artery is obtained, is realized Pair the optimization of target coronary artery extract；Specifically, above-mentioned S203 may include:
S601 is based on default growth conditions using each sampling voxel of the target coronary artery trunk volume data as seed voxels, Region growing processing is carried out, the initial target coronary artery in the heart volume data is obtained；The default growth conditions includes: and plants The probability value that the voxel of daughter element connection belongs to coronary artery is greater than or equal to the first probability threshold value.
It is understood that above-mentioned voxel, which belongs to calculating for the probability value of coronary artery, can enhance the side handled using sea is gloomy Formula, can also be by the way of machine learning model or other way, specific implementation are referred to retouching for above-mentioned S502 It states.It is understood that using machine learning model compared to other way, for the gloomy enhancing processing mode in sea, Ke Yigeng The false edge and impurity of good inhibition coronary artery.Because although tubulose feature can be enhanced in the gloomy enhancing processing in sea, for coronary artery The feature of the tubular edge on periphery can not identify false tubular edge or true tubular edge, therefore meeting very well Enhance false tubular edge, and using false tubular edge as coronary artery edge, thus the target coronary artery extracted is easy to deposit In false edge and impurity.And machine learning model can be then trained by training sample, which may include: Positive sample and negative sample, the input of the positive sample are the voxel on coronary artery, are exported as probability value 1, the input of the negative sample For the voxel on coronary artery periphery, export as probability value 0；Therefore machine learning model be can be to the false tubulose side on coronary artery periphery What the feature of edge was effectively identified, the false edge and impurity of coronary artery can be inhibited.
S602 carries out micronization processes to the initial target coronary artery, obtains target coronary artery skeleton.
Generally, it carries out algorithm used by micronization processes and is properly termed as image framework algorithm, such as k3m algorithm, Zhang-Suen algorithm etc..In one embodiment, image processing module can be using lesser scale (such as the ruler of voxel Degree) it is split along the centerline direction of initial target coronary artery, it is equivalent to and initial target coronary artery is divided into coronary artery one by one Section, the coronary artery section are made of each coronary artery voxel for belonging to the coronary artery section；Image processing module can choose each coronary artery and cut Voxel of the voxel in bosom as target coronary artery skeleton in face forms target coronary artery skeleton.Specifically, image processing module can Using the voxel by the geometric center in each coronary artery section as target coronary artery skeleton, each hat can also constantly be corroded based on k3m algorithm Arteries and veins section extracts the last one voxel voxel as target coronary artery skeleton of each coronary artery section during corrosion treatment. Above-mentioned micronization processes can relocate to obtain the center line of more accurate target coronary artery, i.e. target coronary artery skeleton, convenient for subsequent Centered on target coronary artery skeleton, region growth is carried out according to distance field.
S603 is based on preset cutting condition, carries out cutting processing to the target coronary artery skeleton, the target after being cut Coronary artery skeleton；The preset cutting condition includes: that the voxel of the target coronary artery skeleton belongs to the probability value of coronary artery less than second Probability threshold value, wherein first probability threshold value is less than the second probability threshold value.
It is understood that when the voxel of target coronary artery skeleton belongs to the probability value of coronary artery less than the second probability threshold value, Then the voxel is cut, the lower voxel of probability for belonging to coronary artery can be reduced.Can effectively it inhibit in coronary artery segmentation Leakage, that is, inhibit due to region growing leakage generate impurity.
S604 carries out region growth to the target coronary artery skeleton after the cutting according to preset distance field, obtains described Target coronary artery in heart volume data.
Above-mentioned preset distance field can be understood as preset coronary artery caliber parameter, the coronary artery caliber closer apart from aorta General bigger, the coronary artery caliber remoter apart from aorta is generally smaller.Growth process is carried out compared to based on region growing model, The possibility revealed again is avoided using preset distance field, realizes the optimization at coronary artery edge.
Referring to a kind of schematic cross-section of heart volume data shown in Fig. 6 b, extracted wherein being shown on the section The left descending branch and left hand of target coronary artery.As it can be seen that the target coronary artery extracted is essentially coincided with actual target coronary artery, that is, realize The accurate segmentation of coronary artery.
Therefore, in the present embodiment, image processing workstations can carry out growth process based on Low threshold, to avoid omission Coronary artery feature guarantees the integrality that coronary artery extracts；Then after the initial target coronary artery refinement obtained to growth process, for The target coronary artery skeleton arrived, is cut based on high threshold, inhibits leakage；Region increasing is finally carried out based on preset distance field again It is long, and be not based on region growing model and carry out growth process, the possibility revealed again is avoided, in short, above-mentioned treatment process is It is complementary, utmostly guarantee the integrality that coronary artery extracts, while further suppressing leakage, improves the accurate of coronary artery extraction Property.
Although should be understood that Fig. 2 a, each step in the flow chart of 3,4a, 5a, 6a according to arrow instruction according to Secondary display, but these steps are not that the inevitable sequence according to arrow instruction successively executes.Unless having herein explicitly Bright, there is no stringent sequences to limit for the execution of these steps, these steps can execute in other order.Moreover, Fig. 2 a, At least part step in 3,4a, 5a, 6a may include multiple sub-steps perhaps these sub-steps of multiple stages or stage It is not necessarily to execute completion in synchronization, but can execute at different times, these sub-steps or stage hold Row sequence is also not necessarily and successively carries out, but can be with the sub-step or stage of other steps or other steps at least A part executes in turn or alternately.
In one embodiment, as shown in fig. 7, providing a kind of coronary artery extraction element, comprising: coronary artery Model registration module 71, coronary artery trunk extraction module 72 and coronary artery trunk pop-in upgrades 73, in which:
Coronary artery Model registration module 71, for comparing preset coronary artery trunk model and the heart body number comprising target coronary artery According to obtaining registration parameter；
Coronary artery trunk extraction module 72, for according to the coronary artery trunk location information of the coronary artery trunk model and described matching Quasi- parameter extracts target coronary artery trunk volume data from the heart volume data；
Coronary artery trunk pop-in upgrades 73, for according to region growing model, using the target coronary artery trunk volume data as Seed voxels carry out region growing processing, obtain the target coronary artery in the heart volume data.
Optionally, referring to shown in Fig. 8, on the basis of coronary artery extraction element shown in Fig. 7, coronary artery Model registration module 71 May include:
Cardiac position acquiring unit 711, for obtaining the cardiac of the heart volume data from the heart volume data Confidence breath；
Registration unit 712 is compared, for by the cardiac position information of the coronary artery trunk model and the heart volume data Cardiac position information be compared, obtain the registration parameter.
Optionally, referring to shown in Fig. 8, coronary artery trunk extraction module 72 may include:
Coronary artery trunk position acquisition unit 721, for according to the registration parameter to the coronary artery trunk location information into Row transformation, obtains transformed coronary artery trunk location information；
Coronary artery trunk extraction unit 722, for being extracted from the heart volume data and the transformed coronary artery master The dry matched volume data of location information, as target coronary artery trunk volume data.
Optionally, referring to shown in Fig. 8, coronary artery extraction element can also include:
Coronary artery trunk sampling module 74, for default around each benchmark voxel of the target coronary artery trunk volume data Region is sampled, and multiple sampling voxels including the benchmark voxel are obtained；
Coronary artery probability evaluation entity 75 belongs to the probability value of coronary artery center voxel for calculating each sampling voxel；Institute Stating coronary artery center voxel is the voxel on coronary artery center line；
Coronary artery trunk optimization module 76, it is corresponding from each benchmark voxel for the probability value according to each sampling voxel A sampling voxel is chosen in multiple sampling voxels respectively, forms new target coronary artery trunk volume data.
Optionally, referring to shown in Fig. 8, coronary artery probability evaluation entity 75 may include:
Enhance processing unit 751, is handled for carrying out the gloomy enhancing in sea to the heart volume data, obtaining enhancing, treated Heart volume data；
Changing value computing unit 752, for according to the heart before enhancing treated the heart volume data and enhancing processing Dirty volume data calculates the changing value of voxel value of each sampling voxel after enhancing processing and before enhancing processing；
Coronary artery probability calculation unit 753 calculates each described for the changing value according to each voxel value for sampling voxel Sampling voxel belongs to the probability value of coronary artery center voxel.
Optionally, coronary artery probability evaluation entity 75 may include:
Machine learning computing unit exports each institute for each sampling voxel to be inputted preset machine learning model State the probability value that sampling voxel belongs to coronary artery center voxel；The training sample of the machine learning model includes: positive sample and bears Sample, the input of the positive sample are the voxel on coronary artery center line, are exported as probability value 1, the input of the negative sample is hat The voxel on arteries and veins center line periphery exports as probability value 0.
Optionally, referring to shown in Fig. 8, coronary artery trunk sampling module 74 may include:
Datum level determination unit 741 determines each base where each benchmark voxel for being directed to each benchmark voxel Quasi- face；The vertical corresponding coronary artery trunk center line of the target coronary artery trunk volume data of each datum level；
Coronary artery trunk sampling unit 742, for being adopted on each datum level according to pre-set radius and preset step-length Sample obtains multiple sampling voxels including the benchmark voxel.
Optionally, referring to shown in Fig. 8, coronary artery trunk optimization module 76 may include:
Sample voxel selection unit 761, for according to it is described it is each sampling voxel probability value and preset probability function, from A sampling voxel is chosen respectively in the corresponding multiple sampling voxels of each benchmark voxel；Each sampling voxel of the selection is corresponding The value of probability function is maximum, and the probability function includes: the positive coefficient linear combination of the first probability function and the second probability function, First probability function is the sum of the probability value for each sampling voxel chosen, and second probability function is that each segmentation belongs to hat The sum of the probability value of arteries and veins segmentation；The heart body being respectively segmented into each sampling voxel of selection between each group neighbouring sample voxel Data；
Coronary artery trunk optimizes unit 762, the new target coronary artery that each sampling voxel for obtaining by the selection forms Trunk volume data.
Optionally, sampling voxel selection unit 761 specifically can be used for the heart between each group neighbouring sample voxel Dirty volume data is sampled, and multiple slight bar voxels are obtained；It calculates the multiple slight bar voxel and belongs to coronary artery centerbody The probability value of element, and the characteristic value of each probability value is calculated, the probability value of coronary artery segmentation is belonged to as the segmentation.
Optionally, referring to shown in Fig. 8, coronary artery extraction element can also include:
Radius of deflection computing module 77, for calculating in the sampling voxel respectively chosen between each group neighbouring sample voxel Radius of deflection；The radius of deflection be one group of neighbouring sample voxel between two subpoints on datum level away from From the datum level vertically corresponding coronary artery trunk center line of the target coronary artery trunk volume data, and process one group of phase Any sampling voxel in neighbour's sampling voxel；
Voxel optimization module 78 is sampled, if being greater than for the radius of deflection between one group of neighbouring sample voxel default inclined Radius threshold is moved, then two sampling voxels in one group of neighbouring sample voxel are chosen again.
Optionally, referring to shown in Fig. 8, coronary artery trunk pop-in upgrades 73 may include:
Region growing unit 731, for using each sampling voxel of the target coronary artery trunk volume data as seed voxels, base In default growth conditions, region growing processing is carried out, the initial target coronary artery in the heart volume data is obtained；The default life Elongate member includes: that the voxel being connected to seed voxels belongs to the probability value of coronary artery more than or equal to the first probability threshold value；
Refinement processing unit 732 obtains target coronary artery skeleton for carrying out micronization processes to the initial target coronary artery；
Processing unit 733 is cut, for carrying out cutting processing to the target coronary artery skeleton based on preset cutting condition, Target coronary artery skeleton after being cut；The preset cutting condition includes: that the voxel of the target coronary artery skeleton belongs to coronary artery Probability value less than the second probability threshold value, wherein first probability threshold value is less than the second probability threshold value；
Distance field Growth Units 734, for being carried out according to preset distance field to the target coronary artery skeleton after the cutting Region increases, and obtains the target coronary artery in the heart volume data.
The coronary artery extraction element of the present embodiment, image processing workstations can by compare preset coronary artery trunk model with Heart volume data comprising target coronary artery, obtains registration parameter；Then according to the coronary artery master of registration parameter and coronary artery trunk model Dry location information can extract target coronary artery trunk volume data from heart volume data, because in coronary artery trunk model Coronary artery trunk is complete, therefore the target coronary artery trunk volume data extracted is also complete, therefore is based ultimately upon complete It is also complete that target coronary artery trunk volume data, which carries out the target coronary artery that region growing is handled,；In short, the hat of the present embodiment Arteries and veins extracting method can extract complete target coronary artery trunk volume data from heart volume data, and extract complete target Coronary artery.
Specific about coronary artery extraction element limits the restriction that may refer to above for coronary artery extracting method, herein not It repeats again.Modules in above-mentioned coronary artery extraction element can be realized fully or partially through software, hardware and combinations thereof.On Stating each module can be embedded in the form of hardware or independently of in the processor in computer equipment, can also store in a software form In memory in computer equipment, the corresponding operation of the above modules is executed in order to which processor calls.
In one embodiment, a kind of image processing workstations are provided, internal structure chart can be as shown in Figure 9.It should Image processing workstations include processor, memory, network interface, display screen and the input unit connected by system bus. Wherein, processor is for providing calculating and control ability；Memory includes non-volatile memory medium, built-in storage, this is non-easily The property lost storage medium is stored with operating system and computer program, which is the operation system in non-volatile memory medium The operation of system and computer program provides environment, and a kind of coronary artery extraction side is realized when which is executed by processor Method；Network interface is used to communicate with external terminal by network connection；Display screen can be liquid crystal display or electronic ink Water display screen；Input unit can be the touch layer covered on display screen, be also possible to be arranged on image processing workstations shell Key, trace ball or Trackpad, can also be external keyboard, Trackpad or mouse etc..
It will be understood by those skilled in the art that structure shown in Fig. 9, only part relevant to application scheme is tied The block diagram of structure does not constitute the restriction for the computer equipment being applied thereon to application scheme, specific computer equipment It may include perhaps combining certain components or with different component layouts than more or fewer components as shown in the figure.
In one embodiment, a kind of computer equipment, including memory and processor are provided, is stored in memory Computer program, the processor perform the steps of when executing computer program
Preset coronary artery trunk model and the heart volume data comprising target coronary artery are compared, registration parameter is obtained；
According to the coronary artery trunk location information and the registration parameter of the coronary artery trunk model, from the heart volume data In extract target coronary artery trunk volume data；
According to region growing model, carried out the target coronary artery trunk volume data as seed voxels at region growing Reason, obtains the target coronary artery in the heart volume data.
In one embodiment, a kind of readable storage medium storing program for executing is provided, computer program, computer program are stored thereon with It is performed the steps of when being executed by processor
Preset coronary artery trunk model and the heart volume data comprising target coronary artery are compared, registration parameter is obtained；
According to the coronary artery trunk location information and the registration parameter of the coronary artery trunk model, from the heart volume data In extract target coronary artery trunk volume data；
According to region growing model, carried out the target coronary artery trunk volume data as seed voxels at region growing Reason, obtains the target coronary artery in the heart volume data.
Those of ordinary skill in the art will appreciate that realizing all or part of the process in above-described embodiment method, being can be with Relevant hardware is instructed to complete by computer program, the computer program can be stored in a non-volatile computer In read/write memory medium, the computer program is when being executed, it may include such as the process of the embodiment of above-mentioned each method.Wherein, To any reference of memory, storage, database or other media used in each embodiment provided herein, Including non-volatile and/or volatile memory.Nonvolatile memory may include read-only memory (ROM), programming ROM (PROM), electrically programmable ROM (EPROM), electrically erasable ROM (EEPROM) or flash memory.Volatile memory may include Random access memory (RAM) or external cache.By way of illustration and not limitation, RAM is available in many forms, Such as static state RAM (SRAM), dynamic ram (DRAM), synchronous dram (SDRAM), double data rate sdram (DDRSDRAM), enhancing Type SDRAM (ESDRAM), synchronization link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic ram (DRDRAM) and memory bus dynamic ram (RDRAM) etc..
Each technical characteristic of above embodiments can be combined arbitrarily, for simplicity of description, not to above-described embodiment In each technical characteristic it is all possible combination be all described, as long as however, the combination of these technical characteristics be not present lance Shield all should be considered as described in this specification.
The several embodiments of the application above described embodiment only expresses, the description thereof is more specific and detailed, but simultaneously It cannot therefore be construed as limiting the scope of the patent.It should be pointed out that coming for those of ordinary skill in the art It says, without departing from the concept of this application, various modifications and improvements can be made, these belong to the protection of the application Range.Therefore, the scope of protection shall be subject to the appended claims for the application patent.

Claims (14)

1. a kind of coronary artery extracting method, which is characterized in that the described method includes:
Preset coronary artery trunk model and the heart volume data comprising target coronary artery are compared, registration parameter is obtained；
According to the coronary artery trunk location information and the registration parameter of the coronary artery trunk model, mentioned from the heart volume data Take out target coronary artery trunk volume data；
According to region growing model, region growing processing is carried out using the target coronary artery trunk volume data as seed voxels, is obtained To the target coronary artery in the heart volume data.
2. the method according to claim 1, wherein described compare preset coronary artery trunk model and include target The heart volume data of coronary artery, obtains registration parameter, comprising:
The cardiac position information of the heart volume data is obtained from the heart volume data；
The cardiac position information of the cardiac position information of the coronary artery trunk model and the heart volume data is compared, is obtained To the registration parameter.
3. the method according to claim 1, wherein the coronary artery trunk position according to the coronary artery trunk model Confidence breath and the registration parameter, extract target coronary artery trunk volume data from the heart volume data, comprising:
The coronary artery trunk location information is converted according to the registration parameter, obtains transformed coronary artery trunk position letter Breath；
Extracted from the heart volume data with the transformed matched volume data of coronary artery trunk location information, as mesh Mark coronary artery trunk volume data.
4. the method according to claim 1, wherein the target is preced with according to region growing model described Arteries and veins trunk volume data carries out region growing processing as seed voxels, before obtaining the target coronary artery in the heart volume data, The method also includes:
Predeterminable area around each benchmark voxel of the target coronary artery trunk volume data is sampled, and obtains including the base Multiple sampling voxels including quasi- voxel；
Calculate the probability value that each sampling voxel belongs to coronary artery center voxel；The coronary artery center voxel is on coronary artery center line Voxel；
According to the probability value of each sampling voxel, one is chosen respectively from the corresponding multiple sampling voxels of each benchmark voxel and is adopted Sample voxel forms new target coronary artery trunk volume data.
5. according to the method described in claim 4, it is characterized in that, each sampling voxel of calculating belongs to coronary artery centerbody The probability value of element, comprising:
Sea gloomy enhancing processing is carried out to the heart volume data, obtains enhancing treated heart volume data；
According to the heart volume data before enhancing treated the heart volume data and enhancing processing, calculates each sampling voxel and increasing The changing value of voxel value after the reason of strength and before enhancing processing；
According to the changing value of the voxel value of each sampling voxel, calculates each sampling voxel and belong to the general of coronary artery center voxel Rate value.
6. according to the method described in claim 4, it is characterized in that, each sampling voxel of calculating belongs to coronary artery centerbody The probability value of element, comprising:
Each sampling voxel is inputted into preset machine learning model, each sampling voxel is exported and belongs to coronary artery center voxel Probability value；The training sample of the machine learning model includes: positive sample and negative sample, and the input of the positive sample is coronary artery Voxel on center line exports as probability value 1, and the input of the negative sample is the voxel on coronary artery center line periphery, and it is general for exporting Rate value 0.
7. according to the method described in claim 4, it is characterized in that, each benchmark in the target coronary artery trunk volume data Predeterminable area around voxel is sampled, and multiple sampling voxels including the benchmark voxel are obtained, comprising:
For each benchmark voxel, each datum level where each benchmark voxel is determined；Each datum level is vertically described The corresponding coronary artery trunk center line of target coronary artery trunk volume data；
It is sampled, is obtained including the benchmark voxel according to pre-set radius and preset step-length on each datum level Multiple sampling voxels.
8. according to the method described in claim 4, it is characterized in that, the probability value according to each sampling voxel, from each A sampling voxel is chosen respectively in the corresponding multiple sampling voxels of benchmark voxel, forms new target coronary artery trunk volume data, Include:
According to the probability value and preset probability function of each sampling voxel, from the corresponding multiple sampling voxels of each benchmark voxel It is middle to choose a sampling voxel respectively；The value of the corresponding probability function of each sampling voxel of the selection is maximum, the probability letter Number includes: the positive coefficient linear combination of the first probability function and the second probability function, and first probability function is each of selection The sum of the probability value of voxel is sampled, second probability function is the sum of the probability value that each segmentation belongs to coronary artery segmentation；It is described each It is segmented into the heart volume data in each sampling voxel of selection between each group neighbouring sample voxel；
Obtain the new target coronary artery trunk volume data being made of each sampling voxel of the selection.
9. according to the method described in claim 8, it is characterized in that, each segmentation belongs to the calculating of the probability value of coronary artery segmentation Mode, comprising:
Heart volume data between each group neighbouring sample voxel is sampled, multiple slight bar voxels are obtained；
It calculates the multiple slight bar voxel and belongs to the probability value of coronary artery center voxel, and calculate the feature of each probability value Value belongs to the probability value of coronary artery segmentation as the segmentation.
10. according to the method described in claim 4, it is characterized in that, the method also includes:
Calculate the radius of deflection in the sampling voxel respectively chosen between each group neighbouring sample voxel；The radius of deflection is institute State distance of one group of neighbouring sample voxel between two subpoints on datum level, the vertical target coronary artery of the datum level The corresponding coronary artery trunk center line of trunk volume data, and by any sampling voxel in one group of neighbouring sample voxel；
If the radius of deflection between one group of neighbouring sample voxel is greater than default bias radius threshold, adjacent to described one group Two sampling voxels in sampling voxel are chosen again.
11. the method according to claim 1, wherein described according to region growing model, by the target coronary artery Trunk volume data carries out region growing processing as seed voxels, obtains the target coronary artery in the heart volume data, comprising:
Using each sampling voxel of the target coronary artery trunk volume data as seed voxels, based on default growth conditions, region is carried out Growth process obtains the initial target coronary artery in the heart volume data；The default growth conditions includes: to connect with seed voxels The probability value that logical voxel belongs to coronary artery is greater than or equal to the first probability threshold value；
Micronization processes are carried out to the initial target coronary artery, obtain target coronary artery skeleton；
Based on preset cutting condition, cutting processing is carried out to the target coronary artery skeleton, the target coronary artery skeleton after being cut； The preset cutting condition includes: that the voxel of the target coronary artery skeleton belongs to the probability value of coronary artery less than the second probability threshold value, Wherein, first probability threshold value is less than the second probability threshold value；
Region growth is carried out to the target coronary artery skeleton after the cutting according to preset distance field, obtains the heart volume data In target coronary artery.
12. a kind of coronary artery extraction element, which is characterized in that described device includes:
Coronary artery Model registration module is obtained for comparing preset coronary artery trunk model with comprising the heart volume data of target coronary artery To registration parameter；
Coronary artery trunk extraction module, for according to the coronary artery trunk location information of the coronary artery trunk model and registration ginseng Number, extracts target coronary artery trunk volume data from the heart volume data；
Coronary artery trunk pop-in upgrades is used for according to region growing model, using the target coronary artery trunk volume data as seed body Element carries out region growing processing, obtains the target coronary artery in the heart volume data.
13. a kind of image processing workstations, including memory and processor, the memory are stored with computer program, special The step of sign is, the processor realizes any one of claims 1 to 11 the method when executing the computer program.
14. a kind of readable storage medium storing program for executing, is stored thereon with computer program, which is characterized in that the computer program is processed The step of device realizes method described in any one of claims 1 to 11 when executing.
CN201811478488.3A 2018-12-05 2018-12-05 Coronary artery extraction method, device, image processing workstation and readable storage medium Active CN109712163B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201811478488.3A CN109712163B (en) 2018-12-05 2018-12-05 Coronary artery extraction method, device, image processing workstation and readable storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201811478488.3A CN109712163B (en) 2018-12-05 2018-12-05 Coronary artery extraction method, device, image processing workstation and readable storage medium

Publications (2)

Publication Number Publication Date
CN109712163A true CN109712163A (en) 2019-05-03
CN109712163B CN109712163B (en) 2021-05-18

Family

ID=66254617

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201811478488.3A Active CN109712163B (en) 2018-12-05 2018-12-05 Coronary artery extraction method, device, image processing workstation and readable storage medium

Country Status (1)

Country Link
CN (1) CN109712163B (en)

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111292314A (en) * 2020-03-03 2020-06-16 上海联影智能医疗科技有限公司 Coronary artery segmentation method, coronary artery segmentation device, image processing system and storage medium
CN111312374A (en) * 2020-01-21 2020-06-19 上海联影智能医疗科技有限公司 Medical image processing method, device, storage medium and computer equipment
CN112700421A (en) * 2021-01-04 2021-04-23 推想医疗科技股份有限公司 Coronary image classification method and device
CN113096141A (en) * 2021-04-19 2021-07-09 推想医疗科技股份有限公司 Coronary artery segmentation method and coronary artery segmentation device
CN113192099A (en) * 2021-04-30 2021-07-30 上海联影医疗科技股份有限公司 Tissue extraction method, device, equipment and medium
WO2022185644A1 (en) * 2021-03-03 2022-09-09 富士フイルム株式会社 Image processing device, image processing device operation method, program for operating image processing device, and learning method

Citations (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1489114A (en) * 2003-09-09 2004-04-14 北京交通大学 Diveided image coding and decoding method with arbitrary shape region segmentation
US20100061611A1 (en) * 2008-09-11 2010-03-11 Siemens Corporate Research, Inc. Co-registration of coronary artery computed tomography and fluoroscopic sequence
CN102737376A (en) * 2012-03-31 2012-10-17 常熟市支塘镇新盛技术咨询服务有限公司 Improved region growing method applied to coronary artery angiography image segmentation
CN102945328A (en) * 2012-11-22 2013-02-27 北京理工大学 Simulation method for X ray angiography image based on graphic processing unit (GPU) parallel computation
CN103810363A (en) * 2012-11-09 2014-05-21 上海联影医疗科技有限公司 Blood vessel seed point selecting method and blood vessel extracting method in angiography
CN104166979A (en) * 2013-12-31 2014-11-26 上海联影医疗科技有限公司 Blood vessel extracting method
CN104282009A (en) * 2013-07-02 2015-01-14 上海联影医疗科技有限公司 Coronary artery extraction method
CN104346799A (en) * 2013-08-01 2015-02-11 上海联影医疗科技有限公司 Method for extracting spinal marrow in CT (Computed Tomography) image
CN104851108A (en) * 2015-06-09 2015-08-19 武汉联影医疗科技有限公司 Hepatic artery segmentation method based on CT image
CN104978725A (en) * 2014-04-03 2015-10-14 上海联影医疗科技有限公司 Method and device for dividing coronary artery
CN104978440A (en) * 2014-04-03 2015-10-14 上海联影医疗科技有限公司 Heart model building method, heart model registration and heart multi-plane reconstruction method
CN104992437A (en) * 2015-06-29 2015-10-21 北京欣方悦医疗科技有限公司 Coronary artery three-dimensional image segmentation method
CN106296660A (en) * 2016-07-28 2017-01-04 北京师范大学 A kind of full-automatic coronary artery dividing method
CN107527341A (en) * 2017-08-30 2017-12-29 上海联影医疗科技有限公司 The processing method and system of angiographic image

Patent Citations (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1489114A (en) * 2003-09-09 2004-04-14 北京交通大学 Diveided image coding and decoding method with arbitrary shape region segmentation
US20100061611A1 (en) * 2008-09-11 2010-03-11 Siemens Corporate Research, Inc. Co-registration of coronary artery computed tomography and fluoroscopic sequence
CN102737376A (en) * 2012-03-31 2012-10-17 常熟市支塘镇新盛技术咨询服务有限公司 Improved region growing method applied to coronary artery angiography image segmentation
CN103810363A (en) * 2012-11-09 2014-05-21 上海联影医疗科技有限公司 Blood vessel seed point selecting method and blood vessel extracting method in angiography
CN102945328A (en) * 2012-11-22 2013-02-27 北京理工大学 Simulation method for X ray angiography image based on graphic processing unit (GPU) parallel computation
CN104282009A (en) * 2013-07-02 2015-01-14 上海联影医疗科技有限公司 Coronary artery extraction method
CN104346799A (en) * 2013-08-01 2015-02-11 上海联影医疗科技有限公司 Method for extracting spinal marrow in CT (Computed Tomography) image
CN104166979A (en) * 2013-12-31 2014-11-26 上海联影医疗科技有限公司 Blood vessel extracting method
CN104978725A (en) * 2014-04-03 2015-10-14 上海联影医疗科技有限公司 Method and device for dividing coronary artery
CN104978440A (en) * 2014-04-03 2015-10-14 上海联影医疗科技有限公司 Heart model building method, heart model registration and heart multi-plane reconstruction method
CN104851108A (en) * 2015-06-09 2015-08-19 武汉联影医疗科技有限公司 Hepatic artery segmentation method based on CT image
CN104992437A (en) * 2015-06-29 2015-10-21 北京欣方悦医疗科技有限公司 Coronary artery three-dimensional image segmentation method
CN106296660A (en) * 2016-07-28 2017-01-04 北京师范大学 A kind of full-automatic coronary artery dividing method
CN107527341A (en) * 2017-08-30 2017-12-29 上海联影医疗科技有限公司 The processing method and system of angiographic image

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
HANNAH SOFIAN: "Segmentation and detection of media adventitia coronary artery boundary in medical imaging intravascular ultrasound using otsu thresholding", 《2015 INTERNATIONAL CONFERENCE ON BIOSIGNAL ANALYSIS, PROCESSING AND SYSTEMS (ICBAPS)》 *
WEIFENG LI ET AL.: "Automatic recognition and segmentation of coronary artery lumen based on snake model in CTA data", 《2017 IEEE 2ND INFORMATION TECHNOLOGY, NETWORKING, ELECTRONIC AND AUTOMATION CONTROL CONFERENCE (ITNEC)》 *
周辉 等: "一种分割冠状动脉X射线造影图像的有效方法", 《生物医学工程与临床》 *

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111312374A (en) * 2020-01-21 2020-06-19 上海联影智能医疗科技有限公司 Medical image processing method, device, storage medium and computer equipment
CN111312374B (en) * 2020-01-21 2024-03-22 上海联影智能医疗科技有限公司 Medical image processing method, medical image processing device, storage medium and computer equipment
CN111292314A (en) * 2020-03-03 2020-06-16 上海联影智能医疗科技有限公司 Coronary artery segmentation method, coronary artery segmentation device, image processing system and storage medium
CN111292314B (en) * 2020-03-03 2024-05-24 上海联影智能医疗科技有限公司 Coronary artery segmentation method, device, image processing system and storage medium
CN112700421A (en) * 2021-01-04 2021-04-23 推想医疗科技股份有限公司 Coronary image classification method and device
WO2022185644A1 (en) * 2021-03-03 2022-09-09 富士フイルム株式会社 Image processing device, image processing device operation method, program for operating image processing device, and learning method
CN113096141A (en) * 2021-04-19 2021-07-09 推想医疗科技股份有限公司 Coronary artery segmentation method and coronary artery segmentation device
CN113096141B (en) * 2021-04-19 2022-01-11 推想医疗科技股份有限公司 Coronary artery segmentation method and coronary artery segmentation device
CN113192099A (en) * 2021-04-30 2021-07-30 上海联影医疗科技股份有限公司 Tissue extraction method, device, equipment and medium
CN113192099B (en) * 2021-04-30 2023-02-17 上海联影医疗科技股份有限公司 Tissue extraction method, device, equipment and medium

Also Published As

Publication number Publication date
CN109712163B (en) 2021-05-18

Similar Documents

Publication Publication Date Title
CN109712163A (en) Coronary artery extracting method, device, image processing workstations and readable storage medium storing program for executing
CN109544566B (en) Coronary image segmentation method, device, computer equipment and storage medium
US10229517B2 (en) Method and apparatus for automated determination of contours in iterative reconstruction of image data
CN102422307B (en) For method, system, device and computer program that interactive liver vessel and biliary system are assessed
US7822246B2 (en) Method, a system and a computer program for integration of medical diagnostic information and a geometric model of a movable body
CN111325759B (en) Vessel segmentation method, apparatus, computer device, and readable storage medium
CN107784647A (en) Liver and its lesion segmentation approach and system based on multitask depth convolutional network
CN107563983A (en) Image processing method and medical imaging devices
CN111080584B (en) Quality control method for medical image, computer device and readable storage medium
CN109478327B (en) Method for automatic detection of systemic arteries in Computed Tomography Angiography (CTA) of arbitrary field of view
CN107622492A (en) Lung splits dividing method and system
EP2120208A1 (en) Method and system for lesion segmentation
JP2004105737A (en) Integrated image recording method for heart magnetism resonance perfusion data
US11847547B2 (en) Method and system for generating a centerline for an object, and computer readable medium
CN109727242B (en) Blood vessel center line extraction method, device, computer equipment and storage medium
CN110619635B (en) Hepatocellular carcinoma magnetic resonance image segmentation system and method based on deep learning
CN113538471B (en) Plaque segmentation method, plaque segmentation device, computer equipment and storage medium
CN111145200B (en) Blood vessel center line tracking method combining convolutional neural network and cyclic neural network
CN114693671A (en) Lung nodule semi-automatic segmentation method, device, equipment and medium based on deep learning
CN111383236B (en) Method, apparatus and computer-readable storage medium for labeling regions of interest
CN111738975B (en) Image identification method and image identification device
CN114419032B (en) Method and device for segmenting the endocardium and/or the epicardium of the left ventricle of the heart
US20220198667A1 (en) Method and device for extracting blood vessel wall
CN114010227A (en) Right ventricle characteristic information identification method and device
Vogelsang et al. Model-based analysis of chest radiographs

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
CB02 Change of applicant information
CB02 Change of applicant information

Address after: 201807 Shanghai City, north of the city of Jiading District Road No. 2258

Applicant after: Shanghai Lianying Medical Technology Co., Ltd

Address before: 201807 Shanghai City, north of the city of Jiading District Road No. 2258

Applicant before: SHANGHAI UNITED IMAGING HEALTHCARE Co.,Ltd.

GR01 Patent grant
GR01 Patent grant