Rib cage center line detecting device and methodTechnical field
The invention belongs to technical field of image processing, and in particular to a kind of rib cage center line detecting device and method.
Background technique
The doctor of image department would generally detect the rib cage transfer of fracture of rib and cancer by CT image.Each patient'sThe quantity of CT image is likely to be breached over one hundred, and in order to check lesion present in CT image and exception, doctor is generally required successivelyCT image is checked, dynamic change of the every rib cage on piece of different layers is tracked, visually to track rib cageThe change of sectional area.But manually CT image check it is relatively time consuming, and manually check be easy to cause fault and missRib cage is abnormal.
Extract rib cage center line can be used to enhance expansion rib cage visualization, can allow doctor more efficiently and effectivelyCheck that the correlation circumstance of rib cage, the center line of rib cage can be also used for positioning organ, registration lesion and guide interval change pointThe reference of correspondence between the serial thorax CT scan of analysis.
The detection method of existing rib cage center line may include following several methods:
1, rib cage is modeled as elongated tubular construction, and intrinsic using hessian matrix (Hessian) or structure tensorNetwork analysis carries out the detection of back voxel, detects tubulose knot by calculating the characteristic value of the hessian matrix of each point in 3d spaceStructure.But this method spent in terms of calculating it is very high, and possibly can not all obtain consistent result to all patients.
2, carry out rib cage center lines matching by template: the section of rib cage usually has certain rule, rib cage centerSpinal cord is partially dark, and the bone of rib edge is partially bright, and rib edge is oval, can match rib cage according to above-mentioned regular design templateAnd then center line is found, further according to needing to be modified center line.But in many cases, rib cage marrow may be than rib cage sideBoundary is darker, so that back voxel consistently can not be detected as rib cage central point.
3, center line is identified by the method for machine learning: extracts the Haar feature of CT image, passes through the number manually markedAccording to the prediction model of training centre's line, to each pixel prediction, it is the probability of center line.
4, pass through the method tracing center line of curve tracing: artificial setting seed automatically selects seed, using such as cardThe method for tracing of Kalman Filtering is tracked detected rib cage central point from a slice to next slice.ButThis method needs artificial initial seed point, On Local Fuzzy caused by the rib cage lesion for fracture etc or discontinuous heightIt is sensitive.
The robustness of prior art is poor, and for medical image, especially the data comprising lesion often change moreSample, the rule being manually arranged cannot handle abnormal conditions well；Calculating speed is slow, due to needing to handle more complicated exceptionSituation, the method for the prior art needs to be arranged more rule and step, slow so as to cause calculating speed.
Therefore, how to propose that a kind of technical solution to solve the above problems is that those skilled in the art need to solve at presentProblem.
Summary of the invention
In order to solve the above problem of the prior art, in order to solve, prior art poor robustness and calculating speed are slow to be askedTopic, the first aspect of the present invention provide a kind of rib cage center line detecting device, comprising:
Rib cage identification module is configured to rib cage region in preset rib cage identification model recognition target image,And the corresponding rib region of every rib cage is obtained according to the rib cage region；
Rib cage center line obtains module, is configured to obtain each institute in the corresponding three-dimension space image of the target imageThe corresponding multiple specific planes of rib region are stated, and calculate the center of gravity of corresponding rib region according to the specific plane, by instituteAll centers of gravity for stating each rib region, which are connected, obtains the center line of corresponding rib cage；
Wherein, the rib cage identification model is based on the Bone images demarcated in advance and using constructed by machine learning algorithmNeural network model；The specific plane is that Plane Angle is the plane preset polar angle and be parallel to preset reference axis,The angle value of the polar angle depends on the center of gravity according to the rib cage region to two-dimensional projection's point in corresponding rib regionCarry out the two-dimensional projection's point pole angular region obtained after polar coordinate transform.
In the optimal technical scheme of above-mentioned apparatus, the rib cage center line acquisition module is further configured to execute as followsOperation:
The center of gravity of the rib cage region and each rib region are projected into two-dimensional surface, respectively obtain center of gravity projectionThe rib cage view field of point and each rib region；
Polar coordinate transform is carried out to two-dimensional projection's point in each rib cage view field centered on the center of gravity subpoint；
It is corresponding that corresponding rib region is obtained according to the polar coordinates of each two-dimensional projection's point in each rib cage view fieldTwo-dimensional projection's point pole angular region.
In the optimal technical scheme of above-mentioned apparatus, the rib cage center line obtains module and is additionally configured to execute following behaviourMake:
According to each corresponding two-dimensional projection's point pole angular region of rib region and the extraction of preset angle stepThe corresponding multiple polar angles of each rib region.
In the optimal technical scheme of above-mentioned apparatus, the rib cage center line detecting device further includes center line amendment mouldBlock, the center line correction module are configured that
Center line acquired in module is obtained to the rib cage center line using preset curve smoothing method to carry out smoothlyProcessing.
In the optimal technical scheme of above-mentioned apparatus, the curve smoothing method is based on preset markov random fileThe curve processing method of model.
In the optimal technical scheme of above-mentioned apparatus, the center line correction module is further configured to shown according to the following formulaMethod center line is smoothed:
Wherein, the E (X) indicates that the energy function of the Markov random field model, the X indicate the center lineSome dimension (x, y or the dimension where z-axis) where upper all points, xiIt indicates to tie up in X for i-th point on the center lineThe coordinate of degree, NiIndicate the number of point adjacent with i-th point on the center line, xjIndicate on the center line with xiPoint phaseJ-th point adjacent of coordinate, x 'iIndicate the i-th point of initial coordinate on X-dimension, γ indicates x 'iShared weight, dIndicate preset distance.
In the optimal technical scheme of above-mentioned apparatus, the rib cage identification module is further configured to perform the following operations:
Obtain multiple connections that the rib cage region described in the corresponding three-dimension space image of the target image includesRegion, and the corresponding rib region of every rib cage is obtained according to each connected region.
In the optimal technical scheme of above-mentioned apparatus, the rib cage identification module is additionally configured to perform the following operations:
Obtain the three-dimensional space point that all Euclidean distances in the rib cage region are less than pre-determined distance；
The connected region of every rib cage is obtained according to acquired three-dimensional space point.
The second aspect of the present invention provides a kind of rib cage center line detecting method, comprising:
Based on rib cage region in preset rib cage identification model recognition target image, and according to where the rib cageRegion obtains the corresponding rib region of every rib cage；
It is corresponding multiple specific to obtain each rib region in the corresponding three-dimension space image of the target imagePlane, and the center of gravity of corresponding rib region is calculated according to the specific plane, by all centers of gravity of each rib regionIt is connected and obtains the center line of corresponding rib cage；
Wherein, the rib cage identification model is based on the Bone images demarcated in advance and using constructed by machine learning algorithmNeural network model；The specific plane is that Plane Angle is the plane preset polar angle and be parallel to preset reference axis,The angle value of the polar angle depends on the center of gravity according to the rib cage region to two-dimensional projection's point in corresponding rib regionCarry out the two-dimensional projection's point pole angular region obtained after polar coordinate transform.
In the optimal technical scheme of the above method, " obtaining in the corresponding three-dimension space image of the target imageThe corresponding multiple specific planes of each rib region, and calculate according to the specific plane weight of corresponding rib regionBefore the step of heart ", the method also includes:
The center of gravity of the rib cage region and each rib region are projected into two-dimensional surface, respectively obtain center of gravity projectionThe rib cage view field of point and each rib region；
Polar coordinate transform is carried out to two-dimensional projection's point in each rib cage view field centered on the center of gravity subpoint；
It is corresponding that corresponding rib region is obtained according to the polar coordinates of each two-dimensional projection's point in each rib cage view fieldTwo-dimensional projection's point pole angular region.
In the optimal technical scheme of the above method, " according to each two-dimensional projection in each rib cage view fieldAfter the step of corresponding two-dimensional projection's point pole angular region of the corresponding rib region of polar coordinates acquisition of point ", the method also includes:
According to each corresponding two-dimensional projection's point pole angular region of rib region and the extraction of preset angle stepThe corresponding multiple polar angles of each rib region.
In the optimal technical scheme of the above method, " all centers of gravity of each rib region is being connected and are obtaining phaseAnswer the center line of rib cage " the step of after, the method also includes:
Center line acquired in module is obtained to the rib cage center line using preset curve smoothing method to carry out smoothlyProcessing.
In the optimal technical scheme of the above method, the curve smoothing method is based on preset markov random fileThe curve processing method of model.
In the optimal technical scheme of the above method, " the rib cage center line is obtained using preset curve smoothing methodCenter line acquired in modulus block is smoothed " the step of include according to the following formula shown in method center line is carried out it is smoothProcessing:
Wherein, the E (X) indicates that the energy function of the Markov random field model, the X indicate the center lineSome dimension (x, y or the dimension where z-axis) where upper all points, xiIt indicates to tie up in X for i-th point on the center lineThe coordinate of degree, NiIndicate the number of point adjacent with i-th point on the center line, xjIndicate on the center line with xiPoint phaseJ-th point adjacent of coordinate, x 'iIndicate the i-th point of initial coordinate on X-dimension, γ indicates x 'iShared weight, dIndicate preset distance.
In the optimal technical scheme of the above method, " the corresponding rib of every rib cage is obtained according to the rib cage regionThe step of bone region " includes:
Obtain multiple connections that the rib cage region described in the corresponding three-dimension space image of the target image includesRegion, and the corresponding rib region of every rib cage is obtained according to each connected region.
In the optimal technical scheme of the above method, " institute in the corresponding three-dimension space image of the target image is obtainedState multiple connected regions that rib cage region includes " the step of include:
Obtain the three-dimensional space point that all Euclidean distances in the rib cage region are less than pre-determined distance；
The connected region of every rib cage is obtained according to acquired three-dimensional space point.
The third aspect of the present invention provides a kind of storage device, wherein be stored with a plurality of program, described program be suitable for byProcessor is loaded to execute rib cage center line detecting method described in any of the above embodiments.
The fourth aspect of the present invention provides a kind of control device, including processor and storage equipment, the storage equipmentSuitable for storing a plurality of program, described program is suitable for being loaded by the processor to execute rib cage center line described in any of the above embodimentsDetection method.
Compared with the immediate prior art, above-mentioned technical proposal is at least had the following beneficial effects:
1, rib cage region in target image, rib cage identification can accurately and efficiently be identified by rib cage identification modelModel is adapted to a variety of abnormal conditions in target image, and the rib cage region identified is more robust；
2, it obtains in target image passing through polar coordinate transform behind accurate rib cage region based on rib cage identification modelRib region center line can be converted to the computational problem of two-dimensional surface by method from the computational problem of three-dimensional planar, simplify meterTreatment process is calculated, calculating speed is improved.
Detailed description of the invention
Fig. 1 is the key step schematic diagram of an embodiment of the present invention rib cage center line detecting method；
Fig. 2 is the schematic diagram of the single rib cage connected region of an embodiment of the present invention；
Fig. 3 is the schematic diagram of the single rib cage center line of an embodiment of the present invention；
Fig. 4 is the schematic diagram of the revised center line of an embodiment of the present invention；
Fig. 5 is the primary structure schematic diagram of the rib cage center line detecting device of an embodiment of the present invention.
Specific embodiment
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with the embodiment of the present inventionIn attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment isA part of the embodiment of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the artEvery other embodiment obtained without making creative work, shall fall within the protection scope of the present invention
The preferred embodiment of the present invention described with reference to the accompanying drawings.It will be apparent to a skilled person that thisA little embodiments are used only for explaining technical principle of the invention, it is not intended that limit the scope of the invention.
Refering to attached drawing 1, Fig. 1 illustratively gives the key step of rib cage center line detecting method in the present embodiment.Such asShown in Fig. 1, rib cage center line detecting method includes the following steps: in the present embodiment
Step S101: based on rib cage region in preset rib cage identification model recognition target image.
In the embodiment of the present invention, rib cage identification model can be from the region where identifying rib cage in target image.SpecificallyGround, target image can be CT image, and rib cage identification model can be based on the Bone images demarcated in advance and utilize engineeringThe neural network model of algorithm building is practised, which can use UNet-like network structure, which includes on 4Sampling structure and 4 down-sampling structures.Target image is inputted into rib cage identification model, the convolutional layer of rib cage identification model can be adoptedConvolution operation is carried out with target image of the convolution kernel of 3*3 to input, extracts the feature of the target image of input convolutional layer；Rib cageReLU activation primitive can be used in identification model, removes the redundancy in image data, retains the feature of data most possibly；RibThe upsample operation of the maxpooling and size=2 of stride=2 can be used in the pond layer of bone identification model, will be defeatedThe characteristic pattern for entering pond layer reduces, and reduces the feature vector of convolutional layer output；The convolution kernel that 1*1 finally can be used carries out convolutionOperation is connected all features by full articulamentum, and the feature after connection is inputted softmax classifier, and target figure is exportedThe rib cage region as in.
In order to rib cage identification model can region preferably in recognition target image where rib cage, can be by marking in advanceFixed Bone images are trained rib cage identification model.Specifically, image translation, figure can be carried out to original target imageThe operation such as image rotation turn, image scaling and Image Reversal, under the premise of not changing target image content, to the picture of target imageElement carries out geometric transformation.To image pixel carry out geometric transformation after, to after geometric transformation image carry out gamma transformation, withThe operations such as machine cutting, Gaussian Blur and normalization, so that rib cage identification model to be trained can preferably identify rib cage instituteIn region.In the training process, rib cage identification model output to be trained can be judged using dice loss loss functionThe degree of closeness of rib cage region and true rib cage region.Rib cage identification model after training can be accurately from meshThe region where rib cage is identified in logo image, it is subsequent rapidly to be obtained in rib cage based on accurate rib cage regionHeart line improves the speed of operation.In addition, rib cage identification model is adapted to a variety of abnormal conditions, the rib cage institute identifiedIt is more robust in region.
The rib cage identification model of the embodiment of the present invention can also use VGG in addition to that can use UNet-like network structureNetwork structure or ResNet network structure etc., the present invention is it is not limited here.
Step S102: multiple connected regions that rib cage region includes in the corresponding three-dimension space image of target image are obtainedDomain, and the corresponding rib region of every rib cage is obtained according to each connected region.
With continued reference to attached drawing 2, Fig. 2 illustratively gives the schematic diagram of single rib cage connected region.It is identified by rib cageMultiple connected regions that region in model recognition target image where rib cage includes, the corresponding rib cage of each connected region.In order to obtain the center line of every rib cage, every rib cage in the region where rib cage can be separated, to every rib cage listSolely analyzed.Specifically, the three-dimensional space point in available rib cage region, and judge any two in three-dimensional space pointWhether the Euclidean distance of point is less than or equal to pre-determined distance (Euclidean distance of such as two points is 1), if so, can be determined that this two o'clockIt is connection, the point of several connections may be constructed the connected region of single rib cage.
Step S103: obtaining the corresponding multiple specific planes of each rib region, calculates corresponding rib cage according to specific planeAll centers of gravity of each rib region are connected and obtain the center line of corresponding rib cage by the center of gravity in region.
In embodiments of the present invention, after obtaining rib cage region, the first center of gravity of rib cage region is obtained, by rib cageThe first center of gravity and each rib region of region project to two-dimensional surface, respectively obtain the first center of gravity subpoint and eachFirst center of gravity subpoint is denoted as C by the rib cage view field of rib region, and two-dimensional surface can be x/y plane.By each rib cageAfter region projections to two-dimensional surface, the starting point of rib cage view field and the angle of end point are obtained, α and β can be denoted as respectively,Centered on the first center of gravity subpoint C, pole seat is carried out according to polar angle range (α, β) of the predetermined angle step-length to rib cage view fieldMark transformation, obtains the corresponding multiple polar angles of each rib region.Assuming that current angular is θ, predetermined angle step-length is ω, preset angleDegree step-length shares 10 corresponding values in current angular θ, solves corresponding second center of gravity of this 10 values, and so on,Available multiple second centers of gravity in polar angle range (α, β).
The corresponding multiple specific planes of each rib region in the corresponding three-dimension space image of target image are obtained, with currentFor angle is θ, the plane parallel with z-axis is constructed based on corresponding second center of gravity of the angle, which is specific plane, shouldThe intersection point of plane and rib region is the z-axis coordinate of the second center of gravity, and multiple second centers of gravity of each rib region are mutually got continuouslyTo the center line of the rib cage.With continued reference to attached drawing 3, Fig. 3 illustratively gives the schematic diagram of single rib cage center line.Pass through poleThe method of coordinate transform can ask rib region center line from the calculating that the computational problem of three-dimensional planar is converted to two-dimensional surfaceTopic, simplifies algorithm process process, improves calculating speed.
Step S104: center line is smoothed using preset curve smoothing method.
The above-mentioned center line being calculated is independently to calculate the separation of every rib cage, there is no in view of different rib cages itBetween correlation, certain points may not on centerline on the center line being finally calculated, therefore center line is not smooth.In addition, the rib region of rib cage identification model identification is likely to occur deviation, it may appear that some abnormal points, these abnormal points can deviateNormal position is farther out.In order to be modified to center line, it can use preset curve smoothing method and center line carried out smoothlyProcessing.Specifically, curve smoothing method, which can be, is modified center line based on preset Markov random field model, canTo be modified according to method shown in formula (1) to center line:
Wherein, E (X) indicates that the energy function of Markov random field model, X indicate where point all on center lineSome dimension (x, y or the dimension where z-axis), xiIndicate the i-th point of coordinate in X-dimension on center line, NiIndicate center lineThe number of the upper point adjacent with i-th point, xjIndicate center line on xiJ-th point adjacent of coordinate of point, x 'iIt indicates i-thInitial coordinate of the point on X-dimension, γ indicate x 'iShared weight, d indicate preset distance.
The first part (part on the plus sige left side) of energy function requires being closer for adjacent central point, it is adjacent inThe number of heart point is to determine a factor of line smoothing degree；Second part (part on the right of the plus sige) requirement of energy functionIt is too far that the distance of central point cannot deviate initial point, and d indicates preset distance, can be to avoid deviateing too far different of normal central lineChang Dian is also only the distance of d at a distance from center line if deviateing too far point.It, can be with by the constraint of energy functionKeep center line more smooth and insensitive to abnormal point.With continued reference to attached drawing 4, Fig. 4 illustratively gives revised centerThe schematic diagram of line.
Although each step is described in the way of above-mentioned precedence in above-described embodiment, this fieldTechnical staff is appreciated that the effect in order to realize the present embodiment, executes between different steps not necessarily in such order,It (parallel) execution simultaneously or can be executed with reverse order, these simple variations all protection scope of the present invention itIt is interior.
Based on above-mentioned rib cage center line detecting embodiment of the method, the present invention also provides a kind of rib cage center line detecting dressesIt sets.With reference to the accompanying drawing, which is illustrated.
Refering to attached drawing 5, Fig. 5 illustratively gives the primary structure of rib cage center line detecting device in the present embodiment.Such asShown in Fig. 5, the present embodiment rib cage center line detecting device includes that rib cage identification module 1 and rib cage center line obtain module 2.
Rib cage identification module 1 is configured to rib cage region in preset rib cage identification model recognition target image,And the corresponding rib region of every rib cage is obtained according to rib cage region；
Rib cage center line obtains module 2, is configured to obtain each rib cage in the corresponding three-dimension space image of target imageThe corresponding multiple specific planes in region, and the center of gravity of corresponding rib region is calculated according to specific plane, by each rib regionAll centers of gravity be connected obtain the center line of corresponding rib cage；
Wherein, rib cage identification model is based on the Bone images demarcated in advance and to utilize mind constructed by machine learning algorithmThrough network model；Specific plane is that Plane Angle is default polar angle and the plane for being parallel to preset reference axis, the angle of polar angleAngle value depends on after carrying out polar coordinate transform to two-dimensional projection's point in corresponding rib region according to the center of gravity of rib cage regionObtained two-dimensional projection's point pole angular region.
In a preferred embodiment of the embodiment of the present invention, rib cage center line obtains module 2 and is further configured to holdThe following operation of row:
The center of gravity of rib cage region and each rib region are projected into two-dimensional surface, respectively obtain center of gravity subpoint andThe rib cage view field of each rib region；
Polar coordinate transform is carried out to two-dimensional projection's point in each rib cage view field centered on center of gravity subpoint；
Corresponding rib region corresponding two is obtained according to the polar coordinates of two-dimensional projection's point each in each rib cage view fieldTie up subpoint polar angle range.
In a preferred embodiment of the embodiment of the present invention, rib cage center line obtains module 2 and is additionally configured to execute such asLower operation:
Each rib cage is extracted according to the corresponding two-dimensional projection's point pole angular region of each rib region and preset angle stepThe corresponding multiple polar angles in region.
In a preferred embodiment of the embodiment of the present invention, rib cage center line detecting device further includes center line amendmentModule, center line correction module are configured that
Center line acquired in module 2 is obtained to rib cage center line using preset curve smoothing method smoothly to be locatedReason.
In a preferred embodiment of the embodiment of the present invention, curve smoothing method be based on preset Markov withThe curve processing method of airport model.
In a preferred embodiment of the embodiment of the present invention, center line correction module is further configured to according to the following formulaShown in method center line is smoothed:
Wherein, E (X) indicates that the energy function of Markov random field model, X indicate where point all on center lineSome dimension (x, y or the dimension where z-axis), xiIndicate the i-th point of coordinate in X-dimension on center line, NiIndicate center lineThe number of the upper point adjacent with i-th point, xjIndicate center line on xiJ-th point adjacent of coordinate of point, x 'iIt indicates i-thInitial coordinate of the point on X-dimension, γ indicate x 'iShared weight, d indicate preset distance.
In a preferred embodiment of the embodiment of the present invention, rib cage identification module 1 is further configured to execute as followsOperation:
Multiple connected regions that rib cage region includes in the corresponding three-dimension space image of target image are obtained, andThe corresponding rib region of every rib cage is obtained according to each connected region.
In a preferred embodiment of the embodiment of the present invention, rib cage identification module 1 is additionally configured to perform the following operations:
Obtain the three-dimensional space point that all Euclidean distances in rib cage region are less than pre-determined distance；
The connected region of every rib cage is obtained according to acquired three-dimensional space point.
Further, it is based on above method embodiment, the present invention also provides a kind of storage devices, wherein the storage deviceIt is stored with a plurality of program and these programs may be adapted to be loaded by processor to execute the rib cage center of above method embodimentLine detecting method.
Still further, being based on above method embodiment, the present invention also provides a kind of control device, the control device packetsInclude processor and storage equipment；Storage equipment may be adapted to store a plurality of program and these programs may be adapted to be added by processorIt carries to execute the rib cage center line detecting method of above method embodiment.
Person of ordinary skill in the field can be understood that for convenience of description and succinctly, the present invention fillsThe specific work process and related description for setting embodiment, can refer to corresponding processes in the foregoing method embodiment, and with it is upperMethod beneficial effect having the same is stated, details are not described herein.
Those skilled in the art should be able to recognize that, side described in conjunction with the examples disclosed in the embodiments of the present disclosureMethod step, can be realized with electronic hardware, computer software, or a combination of the two, in order to clearly demonstrate electronic hardware andThe interchangeability of software generally describes each exemplary composition and step according to function in the above description.TheseFunction is executed actually with electronic hardware or software mode, specific application and design constraint depending on technical solution.Those skilled in the art can use different methods to achieve the described function each specific application, but this realityNow it should not be considered as beyond the scope of the present invention.
It should be noted that description and claims of this specification and term " first " in above-mentioned attached drawing, "Two " etc. be to be used to distinguish similar objects, rather than be used to describe or indicate specific sequence or precedence.It should be understood that thisThe data that sample uses can be interchanged in appropriate circumstances, so that the embodiment of the present invention described herein can be in addition at thisIn illustrate or description those of other than sequence implement.
So far, it has been combined preferred embodiment shown in the drawings and describes technical solution of the present invention, still, this fieldTechnical staff is it is easily understood that protection scope of the present invention is expressly not limited to these specific embodiments.Without departing from thisUnder the premise of the principle of invention, those skilled in the art can make equivalent change or replacement to the relevant technologies feature, theseTechnical solution after change or replacement will fall within the scope of protection of the present invention.