CN107563983A - Image processing method and medical imaging devices - Google Patents
Image processing method and medical imaging devices Download PDFInfo
- Publication number
- CN107563983A CN107563983A CN201710899166.5A CN201710899166A CN107563983A CN 107563983 A CN107563983 A CN 107563983A CN 201710899166 A CN201710899166 A CN 201710899166A CN 107563983 A CN107563983 A CN 107563983A
- Authority
- CN
- China
- Prior art keywords
- image
- vessel
- coronary artery
- blood vessel
- center line
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Landscapes
- Image Analysis (AREA)
Abstract
The embodiment of the present invention provides a kind of image processing method and medical imaging devices, applied to technical field of image processing, improves the order of accuarcy of the identification to left and right coronary artery image to a certain extent.Image processing method provided in an embodiment of the present invention, including:Obtain primitive vessel 3-D scanning image;The primitive vessel 3-D scanning image is handled, obtains specifying blood vessel candidate region;The specified blood vessel candidate region is handled, obtains the center line of the specified blood vessel candidate region;Along the trend of the center line, each sample point is obtained on center line perpendicular to the two dimensional slice data of the center line;The two dimensional slice data is inputted into housebroken neutral net to be learnt, obtains learning outcome;Determine to specify blood-vessel image according to several learning outcomes.
Description
Technical field
The present invention relates to technical field of image processing, more particularly to a kind of image processing method and medical imaging devices.
Background technology
In recent years, the morbidity and mortality of angiocardiopathy just increase year by year, it has also become dead master in world wide
Reason is wanted, in various angiocardiopathies, coronary artery disease occupies very high ratio in the death rate.With the hair of technology
Exhibition, can be well to former by coronary artery using CTA (computed tomography angiography, angiogram)
Heart disease makes diagnosis caused by.
Coronary artery includes arteria coroaria sinistra and arteria coronaria dextra, and they send from the sustainer of base of heart respectively, to
The direction extension of the apex of the heart, for envelope on the surface of heart, it is thinner more arrive the apex of the heart direction coronary artery away from root of coronary artery, is moved based on coronal
The labyrinth of arteries and veins, and coronary artery is wrapped in pericardium so that accurate extraction coronary artery turns into non-in the image of CTA generations
A Chang Guanjian step.
Because before CTA images are obtained, doctor is needed to give patient injection contrast agent, and contrast agent can be flowed into close to coronal dynamic
In the chamber of arteries and veins, the problem of using instrumentation along with doctor, accurately extraction coronary artery images can be influenceed from CTA images
Accuracy.
The content of the invention
The embodiment of the present invention provides a kind of image processing method and medical imaging devices, improves and is carried from CTA images
Take accuracy coronarius.
In a first aspect, the embodiment of the present invention provides a kind of image processing method, including:
Obtain primitive vessel 3-D scanning image;
The primitive vessel 3-D scanning image is handled, obtains specifying blood vessel candidate region;
The specified blood vessel candidate region is handled, obtains the center line of the specified blood vessel candidate region;
Along the trend of the center line, each sample point is obtained on center line perpendicular to the two dimension slicing number of the center line
According to;
The two dimensional slice data is inputted into housebroken neutral net to be learnt, obtains learning outcome;
Determine to specify blood-vessel image according to several learning outcomes.
Aspect as described above and any possible implementation, it is further provided a kind of implementation, specified blood vessel are
Coronary artery, methods described also include:
Coronary artery in the specified blood-vessel image is handled, moved with removing non-coronary in each vessel branch
The region of arteries and veins.
Aspect as described above and any possible implementation, it is further provided a kind of implementation, specified to described
Coronary artery in blood-vessel image, which carries out processing, to be included:
Determine the center line coronarius;
Bifurcation and end points are determined on the center line coronarius;
The center line is divided into some sections according to the bifurcation and end points;
The specified blood-vessel image is divided into the region and coronary artery region of non-coronary artery according to described some sections.
Aspect as described above and any possible implementation, it is further provided a kind of implementation, specified blood vessel are
Coronary artery, methods described also include:
Coronary artery in the specified blood-vessel image is handled, to remove the non-hat in the specified blood-vessel image
Shape arteries point.
Aspect as described above and any possible implementation, it is further provided a kind of implementation, in the acquisition
Before primitive vessel 3-D scanning image, methods described also includes:
Neutral net is trained using positive and negative samples, obtains the housebroken neutral net.
Aspect as described above and any possible implementation, it is further provided a kind of implementation, the nerve net
Network includes convolutional layer, pond layer, Nonlinear Mapping layer, full articulamentum and classification layer, can by the housebroken neutral net
Determine the class probability value of the two dimensional slice data.
Aspect as described above and any possible implementation, it is further provided a kind of implementation, the classification are general
Rate value includes the probability that the two dimensional slice data belongs to specified blood vessel.
Aspect as described above and any possible implementation, it is further provided a kind of implementation, the nerve net
Network uses perception-memory-decision model.
Second aspect, the embodiment of the present invention also provide a kind of image processing method, including:
Obtain primitive vessel 3-D scanning image;
Determine to specify blood vessel candidate region from the primitive vessel 3-D scanning image;
The specified blood vessel candidate region is divided into multiple two dimensional slice datas;
Multiple two dimensional slice datas are inputted into housebroken neutral net to be learnt, obtain learning outcome;
Determine to specify blood-vessel image in the specified blood vessel candidate region according to several learning outcomes.
The third aspect, the embodiment of the present invention also provide a kind of medical imaging devices, and the equipment includes:
Processor;
For storing the memory of the processor-executable instruction;
The processor is configured as:
Obtain primitive vessel 3-D scanning image;
Determine to specify blood vessel candidate region from the primitive vessel 3-D scanning image;
The specified blood vessel candidate region is divided into multiple two dimensional slice datas;
Multiple two dimensional slice datas are inputted into housebroken neutral net to be learnt, obtain learning outcome;
Determine to specify blood-vessel image in the specified blood vessel candidate region according to several learning outcomes.
Image processing method and medical imaging devices medical imaging devices provided in an embodiment of the present invention, by original
The coronary artery candidate region that blood vessel 3-D scanning image obtains by processing is handled, and obtains the center line of candidate region,
Then along the trend of center line, obtaining each sample point on center line perpendicular to center line there is the two dimension for specifying size to cut
Sheet data, housebroken neutral net finally is input to using two dimensional slice data as input data, by neural network learning
Afterwards, non-coronary arteriosomes in primitive vessel 3-D scanning image can be effectively removed, realizes that left and right is coronarius accurate
Identification, solve the problems, such as accurately to extract accuracy coronarius from CTA images in the prior art relatively low.
Brief description of the drawings
In order to illustrate more clearly about the embodiment of the present invention or technical scheme of the prior art, below will be to embodiment or existing
There is the required accompanying drawing used in technology description to be briefly described, it should be apparent that, drawings in the following description are the present invention
Some embodiments, for those of ordinary skill in the art, without having to pay creative labor, can be with root
Other accompanying drawings are obtained according to these accompanying drawings.
Fig. 1 is the flow chart of medical image processing method embodiment provided in an embodiment of the present invention;
Fig. 2 specifies blood vessel candidate region center line result schematic diagram for provided in an embodiment of the present invention obtain;
Fig. 3 is another flow chart of medical image processing method embodiment provided in an embodiment of the present invention;
Fig. 4 is another flow chart of medical image processing method embodiment provided in an embodiment of the present invention;
Fig. 5 is candidate's connected domain schematic diagram provided in an embodiment of the present invention;
Fig. 6 is candidate's connected domain result schematic diagram after deep learning provided in an embodiment of the present invention;
Fig. 7 is that coronary artery provided in an embodiment of the present invention finally extracts result schematic diagram;
Fig. 8 is another flow chart of medical image processing method embodiment provided in an embodiment of the present invention;
Fig. 9 is the schematic diagram of a scenario of medical image processing method embodiment provided in an embodiment of the present invention.
Embodiment
To make the purpose, technical scheme and advantage of the embodiment of the present invention clearer, below in conjunction with the embodiment of the present invention
In accompanying drawing, the technical scheme in the embodiment of the present invention is clearly and completely described, it is clear that described embodiment is
Part of the embodiment of the present invention, rather than whole embodiments.Based on the embodiment in the present invention, those of ordinary skill in the art
The every other embodiment obtained under the premise of creative work is not made, belongs to the scope of protection of the invention.
The term used in embodiments of the present invention is only merely for the purpose of description specific embodiment, and is not intended to be limiting
The present invention." one kind ", " described " and "the" of singulative used in the embodiment of the present invention and appended claims
It is also intended to including most forms, unless context clearly shows that other implications.
It should be appreciated that term "and/or" used herein is only a kind of incidence relation for describing affiliated partner, represent
There may be three kinds of relations, for example, A and/or B, can be represented:Individualism A, while A and B be present, individualism B these three
Situation.In addition, character "/" herein, it is a kind of relation of "or" to typically represent forward-backward correlation object.
Existing target or tissue segmentation are generally included based on gray value model or based on appearance model etc., but for
The target or tissue of complex shape, cutting procedure are easily affected by various factors.In view of the above-mentioned problems, the application proposes one
Kind medical image processing method, it may include training (study) stage and detection (prediction) stage, wherein:Training stage can be from
Line process, same person under inspection or the same body part of different persons under inspection, the number with modality images composition can be utilized during this period
According to storehouse, training obtains including the model or parameter of multiple features;Detection-phase is in line process, is included according to what training obtained
The model of multiple features detects to medical image to be detected, obtains the target organ mark or fixed for current medical image
Position.
In certain embodiments, database may include the statistical measurement knots such as the position of anatomical structure, function, person under inspection's information
Fruit, by training process can by the optimal fitting between the minimum Optimization Prediction on database and target or tissue learning or
Mapping in person's training objective or tissue and medical image between each pixel, or target or tissue are obtained most by training process
The model based on form and/or the model based on pixel of matching.Certainly, the 3 d medical images in database can also be through
Expert marks the medical image of (mark), by these marks come the automatic key feature learnt on focus or target organ
And its characteristic.It should be noted that the feature that manually selects of the aspect ratio learnt automatically is more representative and universality, it is based on
Automatically the target organ precision that the Feature Selection learnt obtains is higher.
In certain embodiments, the training stage includes:Obtain multiple 3 d medical images;From multiple 3 d medical images point
Multiple candidate regions are cut, the plurality of candidate region corresponds to one or more target sites or destination organization;In multiple candidate regions
The center line of target site is extracted in domain, and the 3 d medical images according to corresponding to center line by multiple candidate regions are divided,
Obtain two-dimensional slice image data set;Using two-dimensional slice image data set as training sample, and using training sample to depth
Learning neural network carries out learning training, and the deep learning neutral net after the completion of training can be used for segmentation test 3 D medical figure
Target site as in.Further, target site can be the tubular tissues such as artery, vein, tracheae.
In embodiments of the present invention, above-mentioned medical image processing method, it is three-dimensional available for the primitive vessel to heart area
Scan image is handled, and obtains coronary artery region image, so that doctor is judged coronary artery disease, it is determined that sick
Become position and estimating to lesion degree.
Fig. 1 is the flow chart of medical image processing method embodiment provided in an embodiment of the present invention, and 3 d medical images can
Select heart CT angiographic images (CT angiography, CTA) or heart MR angiographic images (Magnetic
Resonance angiography, MRA), illustrate below by taking CTA as an example.As shown in figure 1, image provided in an embodiment of the present invention
Processing method may include steps of:
101st, primitive vessel 3-D scanning image is obtained.
Doctor utilizes CT scan after contrast agent is injected in the blood vessels for detected personnel
(Computed Tomography, CT) equipment is scanned to being detected personnel, tomoscan image is obtained, then to tomography
Scan image is rebuild, and obtains the primitive vessel 3-D scanning image of target area or tissue.
Can be the primitive vessel 3-D scanning image of heart area during a concrete implementation.
102nd, primitive vessel 3-D scanning image is handled, obtains specifying blood vessel candidate region.
During a concrete implementation, when primitive vessel 3-D scanning image is three-dimensional for the primitive vessel of heart area
During scan image, obtain and specify blood vessel candidate region to be carried out as follows explanation with exemplary:
Processing procedure can be:First, the position of sustainer in primitive vessel 3-D scanning image is determined, and to original graph
As carrying out enhancing filtering, obtain strengthening filtering image, in this step, become according to the feature of sustainer in original image and Hough
Change, determine the position of sustainer in original image, wherein, sustainer is characterized as the gray value and shape facility of sustainer.
In CTA images, the feature of sustainer is:The gray value of sustainer is typically between 250-550;Aortic root is in CTA images
On be approximately standard border circular areas and radius between 10-25cm.According to the characteristic information of sustainer in above CTA images,
Using Hough transformation a center of circle is determined in aortic root;Using the center of circle as starting point, a circle is obtained by region growing
Region;Each tomographic image all does certain deformation, the wheel of each tomographic image corresponding region based on a tomographic image thereon afterwards
Exterior feature is still closing, stops finding the envelope of the tomographic image when area of closed outline on a certain secondary tomographic image becomes big suddenly
Close contour line;Contour area on every tomographic image is overlapped, that is, has obtained the segmentation result of sustainer in CTA images.
Secondly, according to the position of sustainer, primitive vessel 3-D scanning image and the gray scale for strengthening filtering image, it is determined that
Left and right initiation region coronarius.Specifically, the profile of sustainer is outwards expanded into m millimeters, ring is formed between sustainer
Shape structure.Due to the branch that left and right coronary artery is sustainer, positioned at the both sides of sustainer, therefore it may only be necessary to around sustainer
The pixel of a part, it is not necessary to handle entire image.Then, calculating each pixel in cyclic structure turns into blood
The probability of pipe point.Aortic position can be combined, original image gray scale and enhancing filtering image gray scale determine that certain point turns into hat
The probability of shape arteries point.Then, the pixel that probability is more than predetermined threshold is puncta vasculosa, and counts the connected region of puncta vasculosa
Domain.Then, according to the gray value of the position relationship and connected region of connected region and sustainer, each connected region is calculated
Condition value, and according to condition value and connected region and the position relationship of sustainer, carry out descending sort.In sustainer or so
Both sides, the preceding M connected region after sorting is chosen respectively, and region growing is carried out to preceding M connected region successively, if S
The growth district of individual connected region reaches predetermined threshold value, then as the initiation region of respective side, wherein, S≤M.
Then, based on the default segmentation threshold segmentation enhancing filtering image of at least one, obtain corresponding to default segmentation threshold
Candidate region.Specifically, N number of default segmentation threshold is chosen, row threshold division is entered to enhancing filtering image respectively, is obtained corresponding
N number of Threshold segmentation result.Then, in N number of Threshold segmentation result, the extension positioned at left and right coronary artery initiation region is removed
Pixel outside region, then obtain candidate region corresponding to N number of threshold value.
Finally, candidate region and left and right initiation region coronarius according to corresponding to default segmentation threshold, it is determined that left,
The candidate region of arteria coronaria dextra.Specifically, calculate respectively and be located at the sustainer left side corresponding to each N number of default segmentation threshold
Candidate region and the volume of arteria coroaria sinistra initiation region and each N number of default segmentation threshold corresponding to be located at sustainer
The candidate region on the right and the volume of arteria coronaria dextra initiation region.Then, according to the descending order of N number of default segmentation threshold,
The default segmentation threshold that the rate of change of volume corresponding to adjacent thresholds is maximum is as left and right optimal threshold coronarius
Value, then the candidate region in segmentation figure picture corresponding to left and right optimal threshold coronarius is left and right candidate regions coronarius
Domain.
In embodiments of the present invention, other methods for obtaining specified blood vessel candidate region in the prior art can also be used.
103rd, to specifying blood vessel candidate region to handle, obtain specifying the center line of blood vessel candidate region.
Specify the extracting method of blood vessel candidate region vessel centerline to use from medical image and be partitioned into blood vessel, based on refinement
Method obtains vessel bone stringing using the erosion operation of mathematical morphology, specifically refers to Pal á gyi K, Balogh E, Kuba
A,et al.A sequential 3D thinning algorithm and its medical applications[C]//
Biennial International Conference on Information Processing in Medical
Imaging.Springer,Berlin,Heidelberg,2001:409-415。
In embodiments of the present invention, can also be by finding the most short of surface into telorism's value identical voxel
These tissue points are connected and composed vessel centerline by the maximum tissue points of distance value.
During a concrete implementation, coronary artery starting connected domain (region) is determined from candidate's connected domain.Work as the left side
When at most 5 specified blood vessel candidates starting connected domains are respectively remained with the right, based on specified blood vessel candidate region, refinement side is utilized
Exemplary as described below of method of method extraction center line:
1) vascular tree is classified, the attribute of any point on skeleton line is determined using thinning method.Exemplarily, utilize
Thinning method determines the connectedness of single pixel in the related neighborhood of skeleton line, when the point only has 1 neighborhood point, then is defined as holding
Point, blood vessel starting point or blood vessel distal point are shown as on blood vessel;When the point has two neighborhood points, then it is defined as commonly connecting
Contact, the intermediate point of blood vessel is shown as on blood vessel;When the point has 3 neighborhood points, vascular bifurcation point is shown as on blood vessel.
Further, when the node on skeleton is connected or is indirectly connected by other nodes on skeleton, then judge that these nodes belong to same
It is a kind of;When the node on skeleton is not attached to and is not indirectly connected with other nodes on skeleton, then judge that these nodes are not belonging to
Same class.
2) make any one end points of a vascular tree as initial point, judge its whether computing, if it is not, performing
Step 3);If it is, continue to select the end points of other vascular trees, until the processing of all vascular trees is completed.
3) node searching is carried out along vascular tree, and judges the category of node on multiple neighborhood middle skeleton lines of present node
Property, when being determined as node or bifurcation, continue executing with;When it is end points to judge, step 4) is performed;
4) also uncalculated branch point is judged whether, if so, then deleting the point and return to step 3) continue executing with;Such as
Fruit is no, then illustrates that the node searching process of vascular tree terminates, and return to step 2).
Alternatively, based on the above method, the also extractable root node based on blood vessel integral structure characteristic, according to root
Node carries out the correction of vessel centerline.As shown in Fig. 2 Fig. 2 specifies blood vessel candidate regions for provided in an embodiment of the present invention obtain
Domain center line result schematic diagram, the center line are the collection for specifying multiple two-dimensional slice image central points corresponding to blood vessel candidate region
Close.
In actual applications, other acquisitions in the prior art can also be used to specify the side of blood vessel candidate region center line
Method.
104th, along the trend of center line, each sample point is obtained on center line perpendicular to the two dimensional slice data of center line.
In embodiments of the present invention, the center line for how obtaining blood vessel is described in foregoing teachings, obtains vessel centerline
Purpose be determine two dimension slicing cutting position, specifically, can be obtained along the cabling of center line perpendicular to center line
Two dimensional slice data, during a concrete implementation, it is 64*64 pixels or 32*32 pixels to specify size, sampled point
It can be set, be utilized according to actual conditions, multiple sampled points be set according to appointed interval distance, because the cabling of blood vessel is
It is unfixed, therefore, it is necessary first to determine normal direction of each sample point perpendicular to center line, it is determined that mode be, currently
Sampled point and its direction n-th point (n=5) determination forward moved towards on centerline along center line, are determining normal direction
Afterwards, it is possible to be determined perpendicular to the direction of center line, finally obtain corresponding two dimensional slice data.
Hereby it is achieved that primitive vessel 3-D scanning image is converted into two dimensional slice data, for inputting into model
Carry out neural network learning.It is pointed out that two dimensional slice data compares 3-D scanning when neural network learning is with processing
Image can reduce the difficulty of sample process;Two dimensional slice data is obtained by 3-D scanning image, sample size can be increased.
105th, two dimensional slice data is inputted into housebroken neutral net to be learnt, obtains learning outcome.
In embodiments of the present invention, two dimensional slice data is inputted into final neutral net carry out the destination of study and be to determine
Whether each two dimensional slice data is target area or tissue, therefore, each two dimensional slice data is inputted into final nerve
After network is learnt, corresponding result can be obtained.In one embodiment, it is as a result two kinds, Yi Zhongshi, the two dimension slicing
Data belong to target area or tissue probability, and another kind is, the two dimensional slice data is not belonging to target area or tissue
Probability.
During a concrete implementation, when primitive vessel 3-D scanning image is heart area, two dimensional slice data
For the two dimensional slice data related to coronary artery, then, each two dimensional slice data is inputted into final neutral net
After habit, obtain the two dimensional slice data belong to probability coronarius or the two dimensional slice data be not belonging to it is coronarius general
Rate.Further, the attribute of two dimensional slice data is can determine that according to the probable value of learning outcome.Such as, for any two dimension slicing
Data, after housebroken Processing with Neural Network, the exportable probability for belonging to coronary artery (positive sample) is 0.7, is not belonging to
The probability of coronary artery (negative sample) is 0.3, that is, the probability for belonging to positive sample is more than the probability for belonging to negative sample, thus can determine that
Two dimensional slice data belongs to coronary artery.And for example, for a two dimensional slice data, after housebroken Processing with Neural Network,
The probability that the exportable probability for belonging to coronary artery (positive sample) is 0.2, is not belonging to coronary artery (negative sample) is 0.8, that is, is belonged to
It is more than the probability for belonging to positive sample in the probability of negative sample, thus can determine that two dimensional slice data is not belonging to coronary artery or non-hat
Shape artery.Similarly, for the two dimensional slice data of any housebroken neutral net of input, it may be determined that two dimensional slice data category
In coronary artery or non-coronary artery.It is pointed out that in the application using two dimensional slice data in neural network learning and
Can be from the real property of multi-angle reflected sample compared to 3-D scanning image during processing, the structure for learning to obtain is more reliable.
106th, determined to specify blood-vessel image according to several learning outcomes.
Because in foregoing teachings, the two dimensional slice data of acquisition has certain continuity and systematicness, therefore, when by number
Individual two dimensional slice data inputs after final neutral net is learnt and can obtain multiple class probability values, and the class probability value characterizes
Belong to target area or nontarget area, the two dimensional slice data for wherein belonging to target area or tissue is classified as one kind, will
The two dimensional slice data for being not belonging to target area or tissue is classified as one kind.
Then, all two dimensional slice datas for belonging to target area or tissue are combined, obtain specifying vessel graph
Picture.
During a concrete implementation, when primitive vessel 3-D scanning image is heart area, two dimensional slice data
For the two dimensional slice data related to coronary artery, then, two dimension slicing coronarius will be belonged in all two dimensional slice datas
Data are combined, and obtain coronary artery images.
On the basis of foregoing teachings, following method flow is also provided in the embodiment of the present invention, for being specified to what is obtained
Blood-vessel image is handled, and obtains more accurate, fine specified blood-vessel image.Especially, when specified blood vessel is coronary artery
When, as shown in figure 3, Fig. 3 is another flow chart of medical image processing method embodiment provided in an embodiment of the present invention, in step
After 106, it can also comprise the following steps in the embodiment of the present invention:
107th, to specifying the coronary artery in blood-vessel image to handle, moved with removing non-coronary in each vessel branch
The region of arteries and veins.
Alternatively, include to specifying the coronary artery in blood-vessel image to carry out processing:Determine center line coronarius;
Bifurcation and end points are determined on center line coronarius;Center line is divided into some sections according to bifurcation and end points;According to
Some sections are drawn the region and coronary artery region that specified blood-vessel image is divided into non-coronary artery.In embodiments of the present invention, for
The coronary artery in blood-vessel image is specified to be handled, the point of center line coronarius, if having in its field in more than two
Point on heart line, the point are defined as bifurcation;If having the point on 1 center line in its neighborhood, the point is defined as end points,
Center line is divided into some sections by bifurcation and end points jointly, is the ratio that coronary artery puncta vasculosa accounts for puncta vasculosa in section in every section of blood vessel of statistics
Example.Such as:If it is that coronary artery puncta vasculosa accounts for the ratio of puncta vasculosa in section more than 65% in this section of blood vessel, it is determined that this section of blood vessel is hat
Shape arterial blood pipeline section.
In another embodiment, judge that a certain section of blood vessel can be in the following way for Coronary Vessel Segment:
Since the bifurcation or end points of the center line corresponding to the starting point of this section of blood vessel, if according to neural network learning
As a result continuously there is the non-coronary artery that length is more than 5mm, then it is assumed that the part may not be coronary artery, then removes the part simultaneously
The end for making the part is new starting point, is judged again since starting point;Or if if according to neural network learning result
There is length and then think that the part is not necessarily coronary artery more than 20mm non-coronary artery in this section of blood vessel, labeled as non-hat
Shape artery.Further, also the coronary artery determined after aforesaid operations can further be handled, remove coronary arterial tree compared with
Short subbranch (being usually 10mm).
In another embodiment, it is contemplated that whole coronary artery vascular tree is obtained on the basis of candidate's connected domain, may be wrapped
It is coronal to remove also to specifying the coronary artery in blood-vessel image to handle in the present embodiment containing some non-coronary artery puncta vasculosas
Non-coronary arteries point in arterial tree:From whole vessel tree extraction center line and it is divided into multiple branch/segmentations;Calculate every
The length of the one non-coronary artery part of segmentation tail end, gone if the length of the non-coronary artery part of segmentation tail end is more than given threshold (such as 5mm)
Fall the tail end.
On the basis of foregoing teachings, as shown in figure 4, Fig. 4 is medical image processing method provided in an embodiment of the present invention
Another flow chart of embodiment, before step 101, the embodiment of the present invention also comprises the following steps:
100th, neutral net is trained using positive negative sample, obtains housebroken neutral net.
In one embodiment, training examples number comes from the patient (subject) of at least 26, from the hat of each patient
Arteries and veins blood vessel three-dimensional communication extracts positive sample image in domain, and negative sample image is extracted from non-coronary artery blood vessel connected region, positive and negative
Sample image totally 10 ten thousand or so.1,000,000 can be expanded to by data, data amplification mode is to be revolved Vascular Slice image
Turn and translate.The size of positive and negative sample image is the two dimensional image of the individual pixels of 32*32 (32-64 can), all sectioning images
Resolution ratio be unified for 0.25mm (between 0.2-0.6 can).Using the original CT values of image as input, to initial nerve
Network is trained.
Wherein, initial neutral net is trained using positive negative sample, including:
64 convolution kernels are set in first layer, each convolution kernel size is 5*5, and two dimensional slice data and convolution kernel are carried out
After convolution algorithm, 64 first layer characteristic patterns are obtained, each first layer characteristic pattern size is 32*32;
Nonlinear Mapping is carried out to first layer characteristic pattern using linear unit function is corrected in the second layer, obtains second layer spy
Sign figure;
Chi Huahe is set in third layer, the size of each pond core is 3*3, carries out pond to second layer characteristic pattern, obtains
64 third layer characteristic patterns, each third layer characteristic pattern size are 16*16;
64 convolution kernels are set at the 4th layer, each convolution kernel size is 5*5, and third layer characteristic pattern and convolution kernel are carried out
After convolution algorithm, 64 the 4th layer of characteristic patterns are obtained, each 4th layer of characteristic pattern size is 16*16;
Nonlinear Mapping is carried out to the 4th layer of characteristic pattern using linear unit function is corrected in layer 5, obtains layer 5 spy
Sign figure;
Chi Huahe is set in layer 6, the size of each pond core is 3*3, carries out pond to layer 5 characteristic pattern, obtains
64 layer 6 characteristic patterns, each layer 6 characteristic pattern size are 8*8;
128 convolution kernels are set in layer 7, each convolution kernel size is 5*5, and layer 6 characteristic pattern is entered with convolution kernel
After row convolution algorithm, 64 layer 7 characteristic patterns are obtained, each layer 7 characteristic pattern size is 8*8;
Nonlinear Mapping is carried out to layer 7 characteristic pattern using linear unit function is corrected at the 8th layer, obtains the 8th layer of spy
Sign figure;
Chi Huahe is set at the 9th layer, the size of each pond core is 3*3, carries out pond to the 8th layer of characteristic pattern, obtains
128 the 9th layer of characteristic patterns, each 9th layer of characteristic pattern size is 4*4;
128 convolution kernels are set at the tenth layer, each convolution kernel size is 4*4, and the 9th layer of characteristic pattern is connected entirely
Processing, obtains the tenth layer of characteristic pattern, each ten layer of characteristic pattern size is 1*1;
2 convolution kernels are set in eleventh floor, each convolution kernel size is 1*1, and the tenth layer of characteristic pattern is connected entirely
Processing, obtains eleventh floor characteristic pattern, each ten layer of characteristic pattern size is 1*1;
The difference between predicted value and actual value is calculated in Floor 12, gradient is returned by back-propagation algorithm
Pass, the weight of each layer of renewal and biasing.
In training process, the Loss values of training set and checking collection persistently reduce, when the Loss values of checking collection no longer reduce
Deconditioning, over-fitting is prevented, take out the grader of the neural network model as Vascular Slice at the moment.
In one example, Processing with Neural Network medical image can use perception-memory-decision model (perception-
Memory-judgment, PMJ).In perception stage, preliminary feature extraction can be carried out to medical image;, can in memory stage
Learn to obtain the excessively complete dictionary of target organ using depth convolutional network;In the decision phase, excessively complete dictionary is regard as three-dimensional
The foundation of medical image target organ extraction, the pipes such as blood vessel are extracted to single 3 D medical image.
Further, perception stage is also referred to as feature extraction, that is, determines whether the point in computer extraction image belongs to
One characteristics of image, it is based primarily upon following feature:A variety of images including medical image have its intrinsic characteristic, figure
The part statistical property of picture is identical, namely can also be used in the feature of part study in other parts, so
In all positions of the image, identical learning characteristic can be used.Alternatively, system can the preferred training sample to input enter
Row pretreatment, then obtains weights using line decoder pre-training.
In one embodiment, perception stage can use unsupervised learning algorithm or backpropagation (back
Propagation, BP) algorithm own coding neutral net.Exemplarily, for an own coding neutral net, it may include defeated
Enter layer, hidden layer and output layer, network can be used full connecting mode, learns by own coding neutral net between layers
Identity function hW,b(x)≈x.Alternatively, if the unit number in hidden layer is less than the unit number of input data, equivalent to acquisition
The sparse matrix of input data;, can also be openness by introducing if the data number of hidden layer is more than or equal to input layer
Limit to obtain the sparse matrix of input signal.In this embodiment, the penalty of sparse autoencoder network is expressed as:
Wherein,For error term, constructed with L2 moulds;It is just
Then item, for preventing over-fitting;For penalty factor, β controls the weight of openness penalty factor.It is represented by:
Wherein, ρ is openness parameter, and be one close to zero value;s2For the number of hidden neuron in hidden layer;
J is each neuron in hidden layer;To imply neural j average active degree, it is formulated as:
Wherein,Represent activity of the input for x own coding neutral net hidden neuron j.
In another embodiment, the sparse own coding neutral net that perception stage can be formed based on uniform enconding network,
The sparse own coding neutral net includes input layer, hidden layer and output layer, and neuron all uses identical excitation function.Three
In the sparse own coding neutral net of layer, the calculation formula of output neuron is respectively:
z(3)=W(2)a(2)+b(2)
a(3)=f (z(3))
The output of network is a(3), equal to excitation function f output.Alternatively, the excitation function in sparse autoencoder network
Usually Sigmoid functions, output valve scope are [0,1], accordingly, a(3)Scope be also [0,1].
Further, uniform enconding network is to use identity function in output layer as excitation function, and in hidden layer still
Autoencoder network of the Sigmoid functions for excitation function is used, now output layer meets:
a(3)=f (z(3))=z(3)=W(2)a(2)+b(2)
In this particular embodiment, with r × c large-size images x1.A × b is selected first from large-size images
Small image pattern x2 trains sparse own coding, and k feature can be calculated according to equation below:
F=σ (W(1)x2+b(1))
Wherein, W (1) represents the weight of visual layer unit;B (1) represents the deviation between implicit unit.For each
The small image pattern x2 of a × b, can as corresponding to calculating above-mentioned formula characteristic value, further, to the feature of each small image pattern
Value makees convolution, you can obtains the eigenmatrix after k × (r-a+1) × (c-b+1) individual convolution.
It should be noted that image has a kind of attribute with " nature static ", represent in the useful spy of an image-region
Sign is likely to be applied to another region.Therefore, in order to describe large-size images, some of image on a region can be calculated
The average value or maximum of special characteristic, class statistic is carried out to the feature of diverse location.The operation of above-mentioned cluster is pond
Change, when the average value for calculating some special characteristic of the image on a region, the corresponding pond that is averaged;When calculating image is at one
The maximum of some special characteristic on region, corresponding maximum pond.After convolution feature is obtained, it is also necessary to determine pond region
Size obtain the convolution feature behind pond.For example, when the size in pond region is m × n, it is possible to which convolution feature is divided
On the disjoint range for being m × n to several sizes, then Chi Huahou is obtained with the average characteristics or maximum feature in these regions
Convolution feature.
Depth convolutional neural networks model has been used to include 5 layers of convolutional neural networks model in this particular embodiment, this 5
Layer convolutional neural networks model includes:Convolutional layer, pond layer, convolutional layer, pond layer and full context layer.In this embodiment, it is right
Any one in two-dimensional slice image data set, the process of depth convolutional neural networks processing are:
1) two-dimensional slice image input convolutional layer, the two-dimensional slice image size are 64 × 64, are instructed in advance using perception stage
The convolution kernel got to 36 5 × 5 sizes carries out convolution to input picture, obtains the Feature Mapping figure of 36 64 × 64 sizes;
2) pond layer, using the window of 3 × 3 sizes to 36 characteristic pattern ponds in convolutional layer, 36 32 × 32 are obtained
Feature Mapping figure;
3) convolutional layer, the image block set of one or more 5 × 5 sizes is obtained to 36 image samplings of pond layer, so
The weights of 64 5 × 5 are obtained to this set training using sparse autoencoder network afterwards, using the weights as convolution kernel, with
36 image convolutions of pond layer obtain the Feature Mapping figure of 64 24 × 24 sizes.The measure taken in the application is by 36
An image convolution of every three works, twice, adjacent 3 of first time phase selection, second of phase selection is every 3 of 2 units, finally for circulation
Obtain Feature Mapping figure (36-3+1)+(36-3 × 2)=64.
4) pond layer, the Feature Mapping figure of 64 8 × 8 is obtained using the window pond of 3 × 3 sizes.
5) full context layer.Training dataset used herein has amounted to 1300 images, after S4, whole net
The Feature Mapping figure of network is 1300 × 64 × 8 × 8, is represented for the input picture of the size of each 64 × 64, can be with
Obtain the mapping graph of 64 8 × 8 sizes.By 1300 × 64 × 8 × 8 Data Dimensionality Reduction obtain (1300 × 64) × (8 × 8)=
83200 × 64, then train final dictionary by exporting for 64 sparse own coding grid.
Image processing method provided in an embodiment of the present invention, by being obtained to primitive vessel 3-D scanning image by processing
Coronary artery candidate region handled, obtain the center line of candidate region, then along the trend of center line, obtain center
On line each sample point perpendicular to center line have specify size two dimensional slice data, finally using two dimensional slice data as
Input data is input to housebroken neutral net, after neural network learning, can effectively remove primitive vessel three-dimensional
Non-coronary arteriosomes in scan image, realize that left and right is coronarius and accurately identify solve in the prior art from CTA images
In accurately extract the problem of accuracy coronarius is relatively low.
Training sample selection and source:
Training examples number comes from the patient (subject) of 26, is carried from the coronary artery blood vessel three-dimensional communication domain of each patient
Positive sample image is taken out, negative sample image, positive and negative sample image totally 10 ten thousand or so are extracted from non-coronary artery blood vessel connected region.
1,000,000 can be expanded to by data, data amplification mode is that Vascular Slice image is carried out into rotation and translation.Positive and negative sample image
Size be the two dimensional image of the individual pixels of 32*32 (32-64 can), the resolution ratio of all sectioning images is unified for 0.25mm
(between 0.2-0.6 can).Using the original CT values of image as input, it is trained.
Neutral net is set:
Neutral net uses convolutional neural networks (CNN), and optimized algorithm is using stochastic gradient descent method (SGD) renewal power
Weight.Totally 12 layers of the convolutional neural networks, wherein there is three convolutional layers, three Nonlinear Mapping layers, three pond layers, two connect entirely
Connect layer, a Loss layer.
First layer is convolutional layer, and effect is the extraction feature from input picture, 64 convolution kernels of setting, and each convolution kernel is big
Small is 5*5, after input picture and convolution kernel are carried out into convolution algorithm, obtains 64 characteristic patterns of first layer, size 32*32;
The second layer is Nonlinear Mapping layer, and effect is non-linear to being added in neutral net, and accelerates convergence rate.Make
Nonlinear Mapping is carried out to first layer characteristic pattern with linear unit function (Relu) is corrected, obtains second layer characteristic pattern;
Third layer is pond layer, and effect is to reduce image size and reduction noise.The size of pond core is 3*3, to
Two layers of characteristic pattern carry out pond, and the method in pond is to take the maximum in 3*3 pixel frames, obtains third layer characteristic pattern, size is
16*16 pixel, number 64;
64 convolution kernels are set at the 4th layer, each convolution kernel size is 5*5, obtains the 4th layer of 64 characteristic pattern, size
For 16*16;
Nonlinear Mapping is carried out to the 4th layer of characteristic pattern using linear unit function is corrected in layer 5, obtains layer 5 spy
Sign figure;
Layer 6 is pond layer, and the size of each pond core is 3*3, carries out pond to layer 5 characteristic pattern, obtains the 6th
Layer characteristic pattern, size are 8*8 pixels, and number is 64;
128 convolution kernels are set in layer 7, each convolution kernel size is 5*5, obtains layer 7 characteristic pattern;
Nonlinear Mapping is carried out to layer 7 characteristic pattern using linear unit function is corrected at the 8th layer, obtains the 8th layer of spy
Sign figure;
In the 9th layer of setting, the size of each pond core is 3*3, carries out pond to the 8th layer of characteristic pattern, obtains the 9th layer
Characteristic pattern, size 4*4, number 128;
128 convolution kernels are set at the tenth layer, the size of each convolution kernel is 4*4, and the 9th layer of characteristic pattern is connected entirely
Processing is connect, obtains the tenth layer of characteristic pattern, size 1*1;
2 convolution kernels are set in eleventh floor, the size of each convolution kernel is 1*1, and the tenth layer of characteristic pattern is connected entirely
Processing is connect, obtains eleventh floor characteristic pattern;
Floor 12 is softmax loss layers, calculates the difference between predicted value and actual value, gradient is passed through reverse
Propagation algorithm (BP algorithm) is returned, the weight (weight) of each layer of renewal and biasing (bias).
In training process, the Loss values of training set and checking collection persistently reduce, when the Loss values of checking collection no longer reduce
Deconditioning, over-fitting is prevented, take out the grader of the neural network model as Vascular Slice at the moment.Will in test process
Floor 12 is changed to softmax layers, and eleventh floor characteristic pattern is inputted to this layer and carries out classification prediction, can obtain input picture
It is blood vessel and the probability of non-vascular, so as to draw classification results.
Using 1:Fig. 5 is candidate's connected domain schematic diagram provided in an embodiment of the present invention, as shown in figure 5, from candidate's connected domain
Middle determination coronary artery starting connected domain, when the left side and the right respectively remain with most 5 candidates originate connected domain when, extract these connections
The center line in domain, two dimension slicing is done, input network, obtain learning outcome.Fig. 6 is candidate's connected domain provided in an embodiment of the present invention
The result schematic diagram after deep learning, as shown in fig. 6, according to learning outcome, remove each branch non-coronary artery part end to end, if
The continuous non-coronary artery for thering is length to be more than 5mm of head, then it is assumed that the part may not be the coronary artery temporary removal part, if length is more than
The non-coronary arterys of 20mm then think that the part is not necessarily coronary artery blood vessel, labeled as non-coronary artery.Meanwhile remove shorter small of coronary branches
Branch (10mm).Fig. 7 is that coronary artery provided in an embodiment of the present invention finally extracts result schematic diagram, as shown in fig. 7, finally,
The ratio that coronary artery puncta vasculosa accounts in remaining coronary artery puncta vasculosa in all connected domains, and whole connected domain is counted, retains connection
The most connected domain of coronary artery puncta vasculosa in domain, if first most long connected domain is not the most connected domain of coronary artery points and company
Logical domain points are more or less the same, and within length about 10mm, then similarly retain first most long connected domain.Using 2:Remove coronary artery
Non- coronary artery puncta vasculosa in tree.Whole coronary artery vascular tree is obtained on the basis of connected domain is originated, wherein some non-coronary arterys can be included
Puncta vasculosa.Whole vessel tree extraction center line and branch, removing the non-coronary artery part of each tail end, (tail end length is more than 5mm
When), or remove when branch length is less than 10mm, while detect endpiece coronary artery blood vessel ratio also to remove less than 0.5.Retain
What is come is exactly coronary artery blood vessel.
The embodiment of the present invention also provides a kind of image processing method, and Fig. 8 is at medical image provided in an embodiment of the present invention
Another flow chart of embodiment of the method is managed, Fig. 9 is the scene of medical image processing method embodiment provided in an embodiment of the present invention
Schematic diagram, as shown in Figure 8 and Figure 9, image processing method provided in an embodiment of the present invention may include steps of:
801st, primitive vessel 3-D scanning image is obtained.
The specific implementation process of step 801 refers to step 101, and its realization principle is approximate with process, is not being gone to live in the household of one's in-laws on getting married herein
State.
802nd, determine to specify blood vessel candidate region from primitive vessel 3-D scanning image.
The specific implementation process of step 802 refers to step 102, and its realization principle is approximate with process, is not being gone to live in the household of one's in-laws on getting married herein
State.
803rd, specified blood vessel candidate region is divided into multiple two dimensional slice datas.
Due to being that 3-D view is handled in abovementioned steps, because 3-D view is carried out in neural network model
Processing, its processing speed is partially slow, therefore, for speed up processing, the intractability of sample is reduced, by specified blood vessel candidate regions
Domain carries out division processing, is divided into multiple two dimensional slice datas, then again inputs two dimensional slice data in neural network model
Learnt.Also, by three-dimensional image segmentation into multiple two dimensional slice datas, the quantity of sample can also be increased.
804th, multiple two dimensional slice datas are inputted into housebroken neutral net to be learnt, obtains learning outcome.
The specific implementation process of step 804 refers to step 104, and its realization principle is approximate with process, is not being gone to live in the household of one's in-laws on getting married herein
State.
805th, determined to specify blood-vessel image in specified blood vessel candidate region according to several learning outcomes.
The specific implementation process of step 805 refers to step 105, and its realization principle is approximate with process, is not being gone to live in the household of one's in-laws on getting married herein
State.
Image processing method provided in an embodiment of the present invention, by being obtained to primitive vessel 3-D scanning image by processing
Coronary artery candidate region handled, multiple two dimensional slice datas are then defeated using two dimensional slice data as input data
Enter to housebroken neutral net, can effectively remove non-coronary arteriosomes in primitive vessel 3-D scanning image, realize
Left and right is coronarius to be accurately identified, solve accurately extracted from CTA images in the prior art accuracy coronarius compared with
The problem of low.In order to realize above method flow, the embodiment of the present invention also provides a kind of medical imaging devices, and the equipment includes
Processor and the memory for storing processor-executable instruction.
Wherein, processor is configured as:
Obtain primitive vessel 3-D scanning image;
Determine to specify blood vessel candidate region from primitive vessel 3-D scanning image;
Specified blood vessel candidate region is divided into multiple two dimensional slice datas;
Multiple two dimensional slice datas are inputted into housebroken neutral net to be learnt, obtain learning outcome;
Specified blood-vessel image is determined in specified blood vessel candidate region according to several learning outcomes.
The medical imaging devices of the present embodiment, it can be used for the technical scheme for performing embodiment of the method shown in Fig. 8, it is realized
Principle is similar with technique effect, and here is omitted.
It is apparent to those skilled in the art that for convenience and simplicity of description, the system of foregoing description,
The specific work process of device and unit, the corresponding process in preceding method embodiment is may be referred to, will not be repeated here.
In several embodiments provided by the present invention, it should be understood that disclosed system, apparatus and method can be with
Realize by another way.For example, device embodiment described above is only schematical, for example, the unit
Division, only a kind of division of logic function, can there is other dividing mode, for example, multiple units or group when actually realizing
Part can combine or be desirably integrated into another system, or some features can be ignored, or not perform.It is another, it is shown
Or the mutual coupling discussed or direct-coupling or communication connection can be by some interfaces, device or unit it is indirect
Coupling or communication connection, can be electrical, mechanical or other forms.
The unit illustrated as separating component can be or may not be physically separate, show as unit
The part shown can be or may not be physical location, you can with positioned at a place, or can also be distributed to multiple
On NE.Some or all of unit therein can be selected to realize the mesh of this embodiment scheme according to the actual needs
's.
In addition, each functional unit in each embodiment of the present invention can be integrated in a processing unit, can also
That unit is individually physically present, can also two or more units it is integrated in a unit.Above-mentioned integrated list
Member can both be realized in the form of hardware, can also be realized in the form of hardware adds SFU software functional unit.
The above-mentioned integrated unit realized in the form of SFU software functional unit, can be stored in one and computer-readable deposit
In storage media.Above-mentioned SFU software functional unit is stored in a storage medium, including some instructions are causing a computer
It is each that device (can be personal computer, server, or network equipment etc.) or processor (Processor) perform the present invention
The part steps of embodiment methods described.And foregoing storage medium includes:USB flash disk, mobile hard disk, read-only storage (Read-
Only Memory, ROM), random access memory (Random Access Memory, RAM), magnetic disc or CD etc. it is various
Can be with the medium of store program codes.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all essences in the present invention
God any modification, equivalent substitution and improvements done etc., should be included within the scope of protection of the invention with principle.
Claims (10)
1. a kind of image processing method, it is characterised in that methods described includes:
Obtain primitive vessel 3-D scanning image;
The primitive vessel 3-D scanning image is handled, obtains specifying blood vessel candidate region;
The specified blood vessel candidate region is handled, obtains the center line of the specified blood vessel candidate region;
Along the trend of the center line, each sample point is obtained on center line perpendicular to the two dimensional slice data of the center line;
The two dimensional slice data is inputted into housebroken neutral net to be learnt, obtains learning outcome;
Determine to specify blood-vessel image according to several learning outcomes.
2. according to the method for claim 1, it is characterised in that it is coronary artery to specify blood vessel, and methods described also includes:
Coronary artery in the specified blood-vessel image is handled, to remove non-coronary artery in each vessel branch
Region.
3. according to the method for claim 2, it is characterised in that at the coronary artery in the specified blood-vessel image
Reason includes:
Determine the center line coronarius;
Bifurcation and end points are determined on the center line coronarius;
The center line is divided into some sections according to the bifurcation and end points;
The specified blood-vessel image is divided into the region and coronary artery region of non-coronary artery according to described some sections.
4. according to the method for claim 1, it is characterised in that it is coronary artery to specify blood vessel, and methods described also includes:
Coronary artery in the specified blood-vessel image is handled, moved with removing the non-coronary in the specified blood-vessel image
Arteries and veins puncta vasculosa.
5. according to the method for claim 1, it is characterised in that before the acquisition primitive vessel 3-D scanning image,
Methods described also includes:
Neutral net is trained using positive and negative samples, obtains the housebroken neutral net.
6. according to the method for claim 5, it is characterised in that the neutral net includes convolutional layer, pond layer, non-linear
Mapping layer, full articulamentum and classification layer, the classification of the two dimensional slice data is can determine that by the housebroken neutral net
Probable value.
7. according to the method for claim 6, it is characterised in that the class probability value includes the two dimensional slice data category
In the probability for specifying blood vessel.
8. according to the method for claim 5, it is characterised in that the neutral net uses perception-memory-decision model.
A kind of 9. image processing method, it is characterised in that including:
Obtain primitive vessel 3-D scanning image;
Determine to specify blood vessel candidate region from the primitive vessel 3-D scanning image;
The specified blood vessel candidate region is divided into multiple two dimensional slice datas;
Multiple two dimensional slice datas are inputted into housebroken neutral net to be learnt, obtain learning outcome;
Determine to specify blood-vessel image in the specified blood vessel candidate region according to several learning outcomes.
10. a kind of medical imaging devices, it is characterised in that the equipment includes:
Processor;
For storing the memory of the processor-executable instruction;
The processor is configured as:
Obtain primitive vessel 3-D scanning image;
Determine to specify blood vessel candidate region from the primitive vessel 3-D scanning image;
The specified blood vessel candidate region is divided into multiple two dimensional slice datas;
Multiple two dimensional slice datas are inputted into housebroken neutral net to be learnt, obtain learning outcome;
Determine to specify blood-vessel image in the specified blood vessel candidate region according to several learning outcomes.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710899166.5A CN107563983B (en) | 2017-09-28 | 2017-09-28 | Image processing method and medical imaging device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710899166.5A CN107563983B (en) | 2017-09-28 | 2017-09-28 | Image processing method and medical imaging device |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107563983A true CN107563983A (en) | 2018-01-09 |
CN107563983B CN107563983B (en) | 2020-09-01 |
Family
ID=60982095
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710899166.5A Active CN107563983B (en) | 2017-09-28 | 2017-09-28 | Image processing method and medical imaging device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107563983B (en) |
Cited By (58)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108304845A (en) * | 2018-01-16 | 2018-07-20 | 腾讯科技(深圳)有限公司 | Image processing method, device and storage medium |
CN108447046A (en) * | 2018-02-05 | 2018-08-24 | 龙马智芯(珠海横琴)科技有限公司 | The detection method and device of lesion, equipment, computer readable storage medium |
CN108932715A (en) * | 2018-07-13 | 2018-12-04 | 北京红云智胜科技有限公司 | A kind of optimization method of the coronarogram segmentation based on deep learning |
CN109272514A (en) * | 2018-10-05 | 2019-01-25 | 数坤(北京)网络科技有限公司 | The sample evaluation method and model training method of coronary artery parted pattern |
CN109325948A (en) * | 2018-10-09 | 2019-02-12 | 数坤(北京)网络科技有限公司 | A kind of coronary artery dividing method and device based on special area optimization |
CN109389606A (en) * | 2018-09-30 | 2019-02-26 | 数坤(北京)网络科技有限公司 | A kind of coronary artery dividing method and device |
CN109523560A (en) * | 2018-11-09 | 2019-03-26 | 成都大学 | A kind of three-dimensional image segmentation method based on deep learning |
CN109523547A (en) * | 2018-12-21 | 2019-03-26 | 四川大学华西医院 | A kind of method and device of image tubercle detection |
CN109840483A (en) * | 2019-01-11 | 2019-06-04 | 深圳大学 | A kind of method and device of landslide fissure detection and identification |
CN109872314A (en) * | 2019-02-20 | 2019-06-11 | 数坤(北京)网络科技有限公司 | A kind of Optimized Segmentation method and apparatus based on center line |
CN109978888A (en) * | 2019-02-18 | 2019-07-05 | 平安科技(深圳)有限公司 | A kind of image partition method, device and computer readable storage medium |
CN110060263A (en) * | 2018-04-24 | 2019-07-26 | 深圳科亚医疗科技有限公司 | Dividing method, segmenting device, segmenting system and the computer-readable medium of medical image |
CN110136137A (en) * | 2019-04-02 | 2019-08-16 | 成都真实维度科技有限公司 | A method of angiosomes segmentation is carried out based on faulted scanning pattern data set |
CN110148113A (en) * | 2019-04-02 | 2019-08-20 | 成都真实维度科技有限公司 | A kind of lesion target area information labeling method based on tomoscan diagram data |
CN110148114A (en) * | 2019-04-02 | 2019-08-20 | 成都真实维度科技有限公司 | A kind of deep learning model training method based on 2D faulted scanning pattern data set |
CN110148124A (en) * | 2019-05-21 | 2019-08-20 | 中山大学 | Throat recognition methods, device, system, storage medium and equipment |
CN110310315A (en) * | 2018-03-21 | 2019-10-08 | 北京猎户星空科技有限公司 | Network model training method, device and object pose determine method, apparatus |
CN110428431A (en) * | 2019-07-12 | 2019-11-08 | 广东省人民医院(广东省医学科学院) | A kind of dividing method, device, equipment and the storage medium of cardiologic medical image |
CN110517757A (en) * | 2018-05-21 | 2019-11-29 | 美国西门子医疗系统股份有限公司 | The medical ultrasound image of tuning |
WO2020001217A1 (en) * | 2018-06-27 | 2020-01-02 | 东南大学 | Segmentation method for dissected aorta in ct image based on convolutional neural network |
CN110675444A (en) * | 2019-09-26 | 2020-01-10 | 东软医疗系统股份有限公司 | Method and device for determining head CT scanning area and image processing equipment |
CN110717487A (en) * | 2018-07-13 | 2020-01-21 | 神经分析私人有限公司 | Method and system for identifying cerebrovascular abnormalities |
CN110796653A (en) * | 2019-10-31 | 2020-02-14 | 北京市商汤科技开发有限公司 | Image processing and neural network training method, device, equipment and medium |
CN110991339A (en) * | 2019-12-02 | 2020-04-10 | 太原科技大学 | Three-dimensional puckery palate identification method adopting circular spectrum |
WO2020087732A1 (en) * | 2018-11-02 | 2020-05-07 | 无锡祥生医疗科技股份有限公司 | Neural network-based method and system for vein and artery identification |
WO2020088469A1 (en) * | 2018-11-01 | 2020-05-07 | 腾讯科技(深圳)有限公司 | Medical image identifying method, model training method, and computer device |
CN111126242A (en) * | 2018-10-16 | 2020-05-08 | 腾讯科技(深圳)有限公司 | Semantic segmentation method, device and equipment for lung image and storage medium |
CN111134727A (en) * | 2018-11-02 | 2020-05-12 | 无锡祥生医疗科技股份有限公司 | Puncture guiding system for vein and artery identification based on neural network |
CN111145137A (en) * | 2018-11-02 | 2020-05-12 | 无锡祥生医疗科技股份有限公司 | Vein and artery identification method based on neural network |
CN111144163A (en) * | 2018-11-02 | 2020-05-12 | 无锡祥生医疗科技股份有限公司 | Vein and artery identification system based on neural network |
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 |
CN111627023A (en) * | 2020-04-27 | 2020-09-04 | 数坤(北京)网络科技有限公司 | Method and device for generating coronary artery projection image and computer readable medium |
CN111681211A (en) * | 2020-05-18 | 2020-09-18 | 沈阳先进医疗设备技术孵化中心有限公司 | Blood vessel image processing method and device |
CN111681226A (en) * | 2020-06-09 | 2020-09-18 | 上海联影医疗科技有限公司 | Target tissue positioning method and device based on blood vessel identification |
CN111754476A (en) * | 2019-06-19 | 2020-10-09 | 北京昆仑医云科技有限公司 | Method and system for disease quantitative modeling of anatomical tree structures |
CN112446911A (en) * | 2019-08-29 | 2021-03-05 | 阿里巴巴集团控股有限公司 | Centerline extraction, interface interaction and model training method, system and equipment |
CN112614141A (en) * | 2020-12-18 | 2021-04-06 | 深圳市德力凯医疗设备股份有限公司 | Method and device for planning blood vessel scanning path, storage medium and terminal equipment |
CN112734907A (en) * | 2020-12-30 | 2021-04-30 | 华东师范大学 | Ultrasonic or CT medical image three-dimensional reconstruction method |
WO2021114636A1 (en) * | 2020-05-29 | 2021-06-17 | 平安科技(深圳)有限公司 | Multimodal data-based lesion classification method, apparatus, device, and storage medium |
CN112997215A (en) * | 2018-09-05 | 2021-06-18 | Ai医疗股份公司 | Method and system for automatically regionalizing blood vessels of medical images by using machine learning and image processing algorithms |
CN113177928A (en) * | 2021-05-18 | 2021-07-27 | 数坤(北京)网络科技股份有限公司 | Image identification method and device, electronic equipment and storage medium |
CN113239992A (en) * | 2021-04-28 | 2021-08-10 | 深圳睿心智能医疗科技有限公司 | Blood vessel classification method and device |
CN113256564A (en) * | 2021-04-28 | 2021-08-13 | 深圳睿心智能医疗科技有限公司 | Method and device for extracting catheter parameters in medical image |
CN113516753A (en) * | 2020-06-01 | 2021-10-19 | 阿里巴巴集团控股有限公司 | Image processing method, device and equipment |
CN113744215A (en) * | 2021-08-24 | 2021-12-03 | 清华大学 | Method and device for extracting center line of tree-shaped lumen structure in three-dimensional tomography image |
US11200976B2 (en) | 2019-08-23 | 2021-12-14 | Canon Medical Systems Corporation | Tracking method and apparatus |
WO2022000977A1 (en) * | 2020-06-29 | 2022-01-06 | 苏州润迈德医疗科技有限公司 | Deep learning-based aortic image acquisition system |
CN113954360A (en) * | 2021-10-25 | 2022-01-21 | 华南理工大学 | 3D printing product anti-counterfeiting method based on embedded identification code multi-process application |
CN113974667A (en) * | 2021-11-02 | 2022-01-28 | 东北大学 | Automatic positioning device and method for TAVI preoperative key target |
CN114041761A (en) * | 2021-10-27 | 2022-02-15 | 北京医准智能科技有限公司 | Method and device for judging origin of coronary artery and computer readable medium |
CN114119602A (en) * | 2021-12-20 | 2022-03-01 | 深圳科亚医疗科技有限公司 | Method, apparatus and storage medium for object analysis of medical images |
CN114359205A (en) * | 2021-12-29 | 2022-04-15 | 推想医疗科技股份有限公司 | Head and neck blood vessel analysis method and device, storage medium and electronic equipment |
CN114862850A (en) * | 2022-07-06 | 2022-08-05 | 深圳科亚医疗科技有限公司 | Target detection method, device and medium for blood vessel medical image |
CN115019014A (en) * | 2022-06-15 | 2022-09-06 | 北京理工大学 | Four-dimensional blood vessel reconstruction method and mechanical calculation method |
CN115049590A (en) * | 2022-05-17 | 2022-09-13 | 北京医准智能科技有限公司 | Image processing method and device, electronic equipment and storage medium |
CN115588012A (en) * | 2022-12-13 | 2023-01-10 | 四川大学 | Pelvic artery blood vessel segmentation method, system, storage medium and terminal |
CN116862877A (en) * | 2023-07-12 | 2023-10-10 | 新疆生产建设兵团医院 | Scanning image analysis system and method based on convolutional neural network |
CN113744215B (en) * | 2021-08-24 | 2024-05-31 | 清华大学 | Extraction method and device for central line of tree-shaped lumen structure in three-dimensional tomographic image |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20100092053A1 (en) * | 2008-10-10 | 2010-04-15 | Kabushiki Kaisha Toshiba | Medical image processor and medical image processing method |
CN102521873A (en) * | 2011-11-22 | 2012-06-27 | 中国科学院深圳先进技术研究院 | Blood vessel modeling method |
CN103514597A (en) * | 2012-06-28 | 2014-01-15 | 株式会社东芝 | Image processing device |
CN103961135A (en) * | 2013-02-04 | 2014-08-06 | 通用电气公司 | System and method for detecting guide pipe position in three-dimensional ultrasonic image |
-
2017
- 2017-09-28 CN CN201710899166.5A patent/CN107563983B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20100092053A1 (en) * | 2008-10-10 | 2010-04-15 | Kabushiki Kaisha Toshiba | Medical image processor and medical image processing method |
CN102521873A (en) * | 2011-11-22 | 2012-06-27 | 中国科学院深圳先进技术研究院 | Blood vessel modeling method |
CN103514597A (en) * | 2012-06-28 | 2014-01-15 | 株式会社东芝 | Image processing device |
CN103961135A (en) * | 2013-02-04 | 2014-08-06 | 通用电气公司 | System and method for detecting guide pipe position in three-dimensional ultrasonic image |
Non-Patent Citations (1)
Title |
---|
王钏: "基于卷积神经网络的血管图像分割", 《中国优秀硕士学位论文全文数据库》 * |
Cited By (86)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108304845A (en) * | 2018-01-16 | 2018-07-20 | 腾讯科技(深圳)有限公司 | Image processing method, device and storage medium |
CN108304845B (en) * | 2018-01-16 | 2021-11-09 | 腾讯科技(深圳)有限公司 | Image processing method, device and storage medium |
CN108447046A (en) * | 2018-02-05 | 2018-08-24 | 龙马智芯(珠海横琴)科技有限公司 | The detection method and device of lesion, equipment, computer readable storage medium |
CN110310315A (en) * | 2018-03-21 | 2019-10-08 | 北京猎户星空科技有限公司 | Network model training method, device and object pose determine method, apparatus |
CN112070776A (en) * | 2018-04-24 | 2020-12-11 | 深圳科亚医疗科技有限公司 | Medical image segmentation method, segmentation device, segmentation system and computer readable medium |
CN110060263A (en) * | 2018-04-24 | 2019-07-26 | 深圳科亚医疗科技有限公司 | Dividing method, segmenting device, segmenting system and the computer-readable medium of medical image |
CN110060263B (en) * | 2018-04-24 | 2020-11-13 | 深圳科亚医疗科技有限公司 | Medical image segmentation method, segmentation device, segmentation system and computer readable medium |
CN110517757A (en) * | 2018-05-21 | 2019-11-29 | 美国西门子医疗系统股份有限公司 | The medical ultrasound image of tuning |
CN110517757B (en) * | 2018-05-21 | 2023-08-04 | 美国西门子医疗系统股份有限公司 | Tuned medical ultrasound imaging |
WO2020001217A1 (en) * | 2018-06-27 | 2020-01-02 | 东南大学 | Segmentation method for dissected aorta in ct image based on convolutional neural network |
CN110717487A (en) * | 2018-07-13 | 2020-01-21 | 神经分析私人有限公司 | Method and system for identifying cerebrovascular abnormalities |
CN108932715B (en) * | 2018-07-13 | 2022-06-07 | 北京红云智胜科技有限公司 | Deep learning-based coronary angiography image segmentation optimization method |
CN108932715A (en) * | 2018-07-13 | 2018-12-04 | 北京红云智胜科技有限公司 | A kind of optimization method of the coronarogram segmentation based on deep learning |
CN112997215A (en) * | 2018-09-05 | 2021-06-18 | Ai医疗股份公司 | Method and system for automatically regionalizing blood vessels of medical images by using machine learning and image processing algorithms |
CN109389606A (en) * | 2018-09-30 | 2019-02-26 | 数坤(北京)网络科技有限公司 | A kind of coronary artery dividing method and device |
CN109389606B (en) * | 2018-09-30 | 2019-12-27 | 语坤(北京)网络科技有限公司 | Coronary artery segmentation method and device |
CN109272514B (en) * | 2018-10-05 | 2021-07-13 | 数坤(北京)网络科技股份有限公司 | Sample evaluation method and model training method of coronary artery segmentation model |
CN109272514A (en) * | 2018-10-05 | 2019-01-25 | 数坤(北京)网络科技有限公司 | The sample evaluation method and model training method of coronary artery parted pattern |
CN109325948A (en) * | 2018-10-09 | 2019-02-12 | 数坤(北京)网络科技有限公司 | A kind of coronary artery dividing method and device based on special area optimization |
CN111126242B (en) * | 2018-10-16 | 2023-03-21 | 腾讯科技(深圳)有限公司 | Semantic segmentation method, device and equipment for lung image and storage medium |
CN111126242A (en) * | 2018-10-16 | 2020-05-08 | 腾讯科技(深圳)有限公司 | Semantic segmentation method, device and equipment for lung image and storage medium |
US11961226B2 (en) | 2018-11-01 | 2024-04-16 | Tencent Technology (Shenzhen) Company Limited | Medical image recognition method, model training method, and computer device |
WO2020088469A1 (en) * | 2018-11-01 | 2020-05-07 | 腾讯科技(深圳)有限公司 | Medical image identifying method, model training method, and computer device |
CN111144163A (en) * | 2018-11-02 | 2020-05-12 | 无锡祥生医疗科技股份有限公司 | Vein and artery identification system based on neural network |
CN111145137B (en) * | 2018-11-02 | 2023-08-15 | 无锡祥生医疗科技股份有限公司 | Vein and artery identification method based on neural network |
WO2020087732A1 (en) * | 2018-11-02 | 2020-05-07 | 无锡祥生医疗科技股份有限公司 | Neural network-based method and system for vein and artery identification |
CN111134727B (en) * | 2018-11-02 | 2022-12-20 | 无锡祥生医疗科技股份有限公司 | Puncture guiding system for vein and artery identification based on neural network |
CN111134727A (en) * | 2018-11-02 | 2020-05-12 | 无锡祥生医疗科技股份有限公司 | Puncture guiding system for vein and artery identification based on neural network |
CN111145137A (en) * | 2018-11-02 | 2020-05-12 | 无锡祥生医疗科技股份有限公司 | Vein and artery identification method based on neural network |
CN111144163B (en) * | 2018-11-02 | 2023-11-21 | 无锡祥生医疗科技股份有限公司 | Vein and artery identification system based on neural network |
CN109523560A (en) * | 2018-11-09 | 2019-03-26 | 成都大学 | A kind of three-dimensional image segmentation method based on deep learning |
CN109523547A (en) * | 2018-12-21 | 2019-03-26 | 四川大学华西医院 | A kind of method and device of image tubercle detection |
CN109840483A (en) * | 2019-01-11 | 2019-06-04 | 深圳大学 | A kind of method and device of landslide fissure detection and identification |
WO2020168648A1 (en) * | 2019-02-18 | 2020-08-27 | 平安科技(深圳)有限公司 | Image segmentation method and device, and computer-readable storage medium |
CN109978888B (en) * | 2019-02-18 | 2023-07-28 | 平安科技(深圳)有限公司 | Image segmentation method, device and computer readable storage medium |
CN109978888A (en) * | 2019-02-18 | 2019-07-05 | 平安科技(深圳)有限公司 | A kind of image partition method, device and computer readable storage medium |
CN109872314A (en) * | 2019-02-20 | 2019-06-11 | 数坤(北京)网络科技有限公司 | A kind of Optimized Segmentation method and apparatus based on center line |
CN110148113A (en) * | 2019-04-02 | 2019-08-20 | 成都真实维度科技有限公司 | A kind of lesion target area information labeling method based on tomoscan diagram data |
CN110148114A (en) * | 2019-04-02 | 2019-08-20 | 成都真实维度科技有限公司 | A kind of deep learning model training method based on 2D faulted scanning pattern data set |
CN110136137A (en) * | 2019-04-02 | 2019-08-16 | 成都真实维度科技有限公司 | A method of angiosomes segmentation is carried out based on faulted scanning pattern data set |
CN110148124A (en) * | 2019-05-21 | 2019-08-20 | 中山大学 | Throat recognition methods, device, system, storage medium and equipment |
CN111754476A (en) * | 2019-06-19 | 2020-10-09 | 北京昆仑医云科技有限公司 | Method and system for disease quantitative modeling of anatomical tree structures |
CN110428431A (en) * | 2019-07-12 | 2019-11-08 | 广东省人民医院(广东省医学科学院) | A kind of dividing method, device, equipment and the storage medium of cardiologic medical image |
CN110428431B (en) * | 2019-07-12 | 2022-12-16 | 广东省人民医院(广东省医学科学院) | Method, device and equipment for segmenting cardiac medical image and storage medium |
US11200976B2 (en) | 2019-08-23 | 2021-12-14 | Canon Medical Systems Corporation | Tracking method and apparatus |
CN112446911A (en) * | 2019-08-29 | 2021-03-05 | 阿里巴巴集团控股有限公司 | Centerline extraction, interface interaction and model training method, system and equipment |
CN110675444B (en) * | 2019-09-26 | 2023-03-31 | 东软医疗系统股份有限公司 | Method and device for determining head CT scanning area and image processing equipment |
CN110675444A (en) * | 2019-09-26 | 2020-01-10 | 东软医疗系统股份有限公司 | Method and device for determining head CT scanning area and image processing equipment |
CN110796653A (en) * | 2019-10-31 | 2020-02-14 | 北京市商汤科技开发有限公司 | Image processing and neural network training method, device, equipment and medium |
CN110796653B (en) * | 2019-10-31 | 2022-08-30 | 北京市商汤科技开发有限公司 | Image processing and neural network training method, device, equipment and medium |
CN110991339A (en) * | 2019-12-02 | 2020-04-10 | 太原科技大学 | Three-dimensional puckery palate identification method adopting circular spectrum |
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 |
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 |
CN111627023A (en) * | 2020-04-27 | 2020-09-04 | 数坤(北京)网络科技有限公司 | Method and device for generating coronary artery projection image and computer readable medium |
CN111627023B (en) * | 2020-04-27 | 2021-02-09 | 数坤(北京)网络科技有限公司 | Method and device for generating coronary artery projection image and computer readable medium |
CN111681211A (en) * | 2020-05-18 | 2020-09-18 | 沈阳先进医疗设备技术孵化中心有限公司 | Blood vessel image processing method and device |
CN111681211B (en) * | 2020-05-18 | 2024-03-08 | 东软医疗系统股份有限公司 | Vascular image processing method and device |
WO2021114636A1 (en) * | 2020-05-29 | 2021-06-17 | 平安科技(深圳)有限公司 | Multimodal data-based lesion classification method, apparatus, device, and storage medium |
CN113516753A (en) * | 2020-06-01 | 2021-10-19 | 阿里巴巴集团控股有限公司 | Image processing method, device and equipment |
CN111681226A (en) * | 2020-06-09 | 2020-09-18 | 上海联影医疗科技有限公司 | Target tissue positioning method and device based on blood vessel identification |
WO2022000977A1 (en) * | 2020-06-29 | 2022-01-06 | 苏州润迈德医疗科技有限公司 | Deep learning-based aortic image acquisition system |
CN112614141A (en) * | 2020-12-18 | 2021-04-06 | 深圳市德力凯医疗设备股份有限公司 | Method and device for planning blood vessel scanning path, storage medium and terminal equipment |
CN112614141B (en) * | 2020-12-18 | 2023-09-19 | 深圳市德力凯医疗设备股份有限公司 | Vascular scanning path planning method and device, storage medium and terminal equipment |
CN112734907A (en) * | 2020-12-30 | 2021-04-30 | 华东师范大学 | Ultrasonic or CT medical image three-dimensional reconstruction method |
CN112734907B (en) * | 2020-12-30 | 2022-07-08 | 华东师范大学 | Ultrasonic or CT medical image three-dimensional reconstruction method |
CN113239992B (en) * | 2021-04-28 | 2024-05-07 | 深圳睿心智能医疗科技有限公司 | Blood vessel classification method and device |
CN113239992A (en) * | 2021-04-28 | 2021-08-10 | 深圳睿心智能医疗科技有限公司 | Blood vessel classification method and device |
CN113256564B (en) * | 2021-04-28 | 2024-03-01 | 深圳睿心智能医疗科技有限公司 | Catheter parameter extraction method and device in medical image |
CN113256564A (en) * | 2021-04-28 | 2021-08-13 | 深圳睿心智能医疗科技有限公司 | Method and device for extracting catheter parameters in medical image |
CN113177928A (en) * | 2021-05-18 | 2021-07-27 | 数坤(北京)网络科技股份有限公司 | Image identification method and device, electronic equipment and storage medium |
CN113744215A (en) * | 2021-08-24 | 2021-12-03 | 清华大学 | Method and device for extracting center line of tree-shaped lumen structure in three-dimensional tomography image |
CN113744215B (en) * | 2021-08-24 | 2024-05-31 | 清华大学 | Extraction method and device for central line of tree-shaped lumen structure in three-dimensional tomographic image |
CN113954360A (en) * | 2021-10-25 | 2022-01-21 | 华南理工大学 | 3D printing product anti-counterfeiting method based on embedded identification code multi-process application |
CN114041761B (en) * | 2021-10-27 | 2022-12-09 | 北京医准智能科技有限公司 | Method, device and computer readable medium for judging origin of coronary artery |
CN114041761A (en) * | 2021-10-27 | 2022-02-15 | 北京医准智能科技有限公司 | Method and device for judging origin of coronary artery and computer readable medium |
CN113974667A (en) * | 2021-11-02 | 2022-01-28 | 东北大学 | Automatic positioning device and method for TAVI preoperative key target |
CN114119602A (en) * | 2021-12-20 | 2022-03-01 | 深圳科亚医疗科技有限公司 | Method, apparatus and storage medium for object analysis of medical images |
CN114359205B (en) * | 2021-12-29 | 2022-11-01 | 推想医疗科技股份有限公司 | Head and neck blood vessel analysis method and device, storage medium and electronic equipment |
CN114359205A (en) * | 2021-12-29 | 2022-04-15 | 推想医疗科技股份有限公司 | Head and neck blood vessel analysis method and device, storage medium and electronic equipment |
CN115049590B (en) * | 2022-05-17 | 2023-03-10 | 北京医准智能科技有限公司 | Image processing method and device, electronic equipment and storage medium |
CN115049590A (en) * | 2022-05-17 | 2022-09-13 | 北京医准智能科技有限公司 | Image processing method and device, electronic equipment and storage medium |
CN115019014A (en) * | 2022-06-15 | 2022-09-06 | 北京理工大学 | Four-dimensional blood vessel reconstruction method and mechanical calculation method |
CN115019014B (en) * | 2022-06-15 | 2024-05-31 | 北京理工大学 | Four-dimensional vascular reconstruction method and mechanical calculation method |
CN114862850A (en) * | 2022-07-06 | 2022-08-05 | 深圳科亚医疗科技有限公司 | Target detection method, device and medium for blood vessel medical image |
CN115588012A (en) * | 2022-12-13 | 2023-01-10 | 四川大学 | Pelvic artery blood vessel segmentation method, system, storage medium and terminal |
CN116862877A (en) * | 2023-07-12 | 2023-10-10 | 新疆生产建设兵团医院 | Scanning image analysis system and method based on convolutional neural network |
Also Published As
Publication number | Publication date |
---|---|
CN107563983B (en) | 2020-09-01 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107563983A (en) | Image processing method and medical imaging devices | |
CN109035255B (en) | Method for segmenting aorta with interlayer in CT image based on convolutional neural network | |
CN105741251B (en) | A kind of blood vessel segmentation method of Hepatic CT A sequence images | |
CN111369528B (en) | Coronary artery angiography image stenosis region marking method based on deep convolutional network | |
CN109478327B (en) | Method for automatic detection of systemic arteries in Computed Tomography Angiography (CTA) of arbitrary field of view | |
CN108765385B (en) | Double-source CT coronary artery automatic extraction method | |
CN110706225B (en) | Tumor identification system based on artificial intelligence | |
CN112308846B (en) | Blood vessel segmentation method and device and electronic equipment | |
CN112991346B (en) | Training method and training system for learning network for medical image analysis | |
CN106056596A (en) | Fully-automatic three-dimensional liver segmentation method based on local apriori information and convex optimization | |
Snaauw et al. | End-to-end diagnosis and segmentation learning from cardiac magnetic resonance imaging | |
González et al. | Computer aided detection for pulmonary embolism challenge (CAD-PE) | |
Kumar et al. | Lung nodule segmentation using unet | |
KR101625955B1 (en) | Method of classifying artery and vein of organ | |
He et al. | Automatic left ventricle segmentation from cardiac magnetic resonance images using a capsule network | |
Ma et al. | A coronary artery segmentation method based on region growing with variable sector search area | |
Zarvani et al. | Residual learning: a new paradigm to improve deep learning-based segmentation of the left ventricle in magnetic resonance imaging cardiac images | |
Li et al. | Left ventricle segmentation by combining convolution neural network with active contour model and tensor voting in short-axis MRI | |
Zheng et al. | Abdominal aortic aneurysm segmentation with a small number of training subjects | |
Khaled et al. | Automatic fuzzy-based hybrid approach for segmentation and centerline extraction of main coronary arteries | |
Zhao et al. | Automatic aortic dissection centerline extraction via morphology-guided CRN tracker | |
Hoori et al. | Automatic Deep Learning Segmentation and Quantification of Epicardial Adipose Tissue in Non-Contrast Cardiac CT scans | |
Cheng et al. | Automatic centerline detection of small three-dimensional vessel structures | |
CN114519722A (en) | Carotid artery extraction method based on convolutional neural network | |
Kim et al. | Automatic CAC voxel classification with multi-scale CNN architecture |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
CP01 | Change in the name or title of a patent holder | ||
CP01 | Change in the name or title of a patent holder |
Address after: 201807 Shanghai city Jiading District Industrial Zone Jiading Road No. 2258 Patentee after: Shanghai Lianying Medical Technology Co., Ltd Address before: 201807 Shanghai city Jiading District Industrial Zone Jiading Road No. 2258 Patentee before: SHANGHAI UNITED IMAGING HEALTHCARE Co.,Ltd. |