CN103456004B - Articular cartilage partitioning method and articular cartilage partitioning system based on image sheet structure enhancement - Google Patents
Articular cartilage partitioning method and articular cartilage partitioning system based on image sheet structure enhancement Download PDFInfo
- Publication number
- CN103456004B CN103456004B CN201310325364.2A CN201310325364A CN103456004B CN 103456004 B CN103456004 B CN 103456004B CN 201310325364 A CN201310325364 A CN 201310325364A CN 103456004 B CN103456004 B CN 103456004B
- Authority
- CN
- China
- Prior art keywords
- cartilage
- pixel
- image
- laminated structure
- region
- 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.)
- Active
Links
Abstract
The invention aims at providing an articular cartilage partitioning method and articular cartilage partitioning system based on image sheet structure enhancement, and belongs to the technical field of medical image processing. The articular cartilage partitioning method comprises the first step of extracting the cartilage detection area of a magnetic resonance image, the second step of calculating the characteristic function of the sheet structure inside the cartilage detection area, the third step of carrying out gray level enhancement on the sheet structure inside the cartilage detection area, and the fourth step of communicating threshold values to extract the cartilage. Therefore, the efficiency of algorithm research and development of the automatic cartilage segmentation technology of the magnetic resonance image is improved, and the application of the partitioning algorithm to different quality data is improved.
Description
Technical field
The present invention relates to technical field of medical image processing, more particularly to it is a kind of based on the enhanced joint of image laminated structure
The method and its system of cartilage segmentation.
Background technology
In MRI (Magnetic Resonance Imaging, NMR (Nuclear Magnetic Resonance)-imaging) image, articular cartilage and cartilage are all
The differentiation of Selvage Stitch is relatively obscured, and its contrast is very low, and cartilage thickness is relatively thin, and from from image cartilage at some positions
It is that, in gray scale connected state, these are all that the automatic segmentation of articular cartilage causes great difficulty with the tissue such as ligament.
In recent study, the algorithm split with regard to cartilage automatically is probed into and is being carried out always, and some are automatic with regard to cartilage
Or the method for semi-automatic segmentation is also suggested, the cartilage partitioning algorithm of wherein most is all based on statistical shape model or pattern
In terms of identification.The country there is presently no the Patents that articular cartilage is automatically or semi-automatically split.
Current cartilage automatic segmentation algorithm mainly includes that volumetric pixel is classified and counts appearance model this two big class.In body image
In this kind of algorithm of element classification, mainly cartilage boundary and image background are separated by machine learning algorithm, such as KNN (K-
Nearest Neighbours, K closest node algorithm) it is commonly used for cartilage segmentation.For dividing for this dependence machine learning
Class algorithm, needs substantial amounts of training sample, and needs expert's manual segmentation to go out many joint samples, the quality of sample image and
Effect will affect the effect of machine learning, can so consume the substantial amounts of time and efforts of algorithm development person.
Statistics appearance model has preferable effect in bone segmentation and cartilage segmentation, and this method needs also exist for professional person
Substantial amounts of skeleton and cartilage are partitioned into, and need training dataset to obtain an initialization effect.But this method is at some
In the case that there is cartilage disorder in joint, segmentation precision is not high.
One in Harbin Institute of Technology is entitled《Using the semi-automatic knee cartilage segmentation of variational method》(author：
Zhao Yun rocs School of Computer Science and Technology in September, 2010) paper in, author provides one for the semi-automatic segmentation of cartilage
The segmentation framework of interactive operation.The method have selected to be split to cartilage based on the border finding method of variable model.This
Kind of method belongs to interactively semi-automatic partition method, needs Manual description some profiles, needs to consume more time.
It is entitled in Nanyang Technological University one《Automatic knee cartilage segmentation
from multi-contrast MR images using support vector machine classification
with spatial dependencies》(Authors：Kunlei Zhang,Wenmiao Lu,Pina Marziliano)
(《Based on spatial coherence and the knee cartilage automatic division method for many contrast magnetic resonance image (MRI) of support vector machine》Make
Person：Kunlei Zhang, Wenmiao Lu, Pina Marziliano) paper in, author using support vector machine method
Cartilage is extracted.Method for being classified with extraction feature, needs to collect substantial amounts of sample and carries out learning training,
There is great difficulty to different quality well adapting to property of data to reach.
In summary, the existing technology based on the enhanced articular cartilage segmentation of image laminated structure shows in actual use
So there is inconvenience and defect, it is therefore necessary to being improved.
The content of the invention
For above-mentioned defect, it is an object of the invention to provide a kind of be based on the enhanced articular cartilage of image laminated structure
The method and its system of segmentation, to realize improve the efficiency of algorithm research and development on the automatic cutting techniques of magnetic resonance image (MRI) cartilage, with
And adaptability of the partitioning algorithm to different quality data.
To achieve these goals, the present invention provides a kind of side based on the enhanced articular cartilage segmentation of image laminated structure
Method, methods described comprise the steps：
The cartilage detection zone of A, extraction of magnetic resonance image；
Laminated structure characteristic function in B, the calculating cartilage detection zone；
C, grey level enhancement is carried out to the laminated structure in the cartilage detection zone；
D, carry out Threshold-connected to extract cartilage.
According to described method, step A includes：
The gray threshold scope of joint bony areas in A1, setting magnetic resonance image (MRI), count whole magnetic resonance image (MRI) each
The gray value of each point of the pixel in radius neighborhood R；
If A2, each gray value put in above-mentioned pixel radius neighborhood, will be upper all in skeleton threshold range
State pixel and be set to 1；Remaining situation is then set to 0, acquires image X；
Described image X is done n times dilation operation according to cartilage thickness, obtains image Y by A3, estimation cartilage maximum gauge；
A4, a pixel on selection joint skeleton in described image Y, extracting has connected relation with this pixel
Region, be labeled as region Z, the region Z is the detection zone of the articular cartilage.
According to described method, step B includes：
B1, construction Gaussian filter Gσ, σ is Gaussian filter variance, to Gaussian filter GσSecond dervative is sought, G is usedσ's
Each pixel in second dervative and the magnetic resonance image (MRI) region Z carries out convolution and obtains Ixx、Ixy、Ixz、Iyy、Iyz、Izz；
B2, composition Hessian matrixesCalculate three eigenvalue λs of H1、λ2、λ3, and by described three
Individual eigenvalue sorts according to order from small to large：λ1＜ λ2＜ λ3；
B3, construction lamellar architectural feature discriminant function Vs(λ1, λ2, λ3), the Hessian of each pixel value of function pair
Matrix exgenvalue carries out structural determination, and for laminated structure region, the Hessian matrix eigenvalue distributions rule should meet
[λ1＜ 0, λ2≈ 0, λ3≈0]；
B4, the laminated structure characteristic function for calculating all pixels point in the region Z, wherein H (λ1) it is to weigh λ1Effect effect
The function of fruit.
According to described method, step C includes：
C1, laminated structure characteristic function L (λ) to pixel in the region Z carry out consolidation process；
C2, pixel in the region Z is carried out into grey level enhancement according to the laminated structure characteristic function of each pixel, made
The region for belonging to laminated structure is strengthened.
According to described method, step D includes：
The substantially gray threshold scope of the articular cartilage after D1, setting grey level enhancement；
D2, the seed point chosen in the articular cartilage, extract the portion for having Threshold-connected relation with the seed point
Point, the part is the partition data of articular cartilage described in image.
In order to realize another goal of the invention of the present invention, present invention also offers a kind of enhanced based on image laminated structure
The system of articular cartilage segmentation, the system include：
Extraction module, for the cartilage detection zone of extraction of magnetic resonance image；
Computing module, for calculating laminated structure characteristic function in the cartilage detection zone；
Strengthen module, for grey level enhancement being carried out to the laminated structure in the cartilage detection zone；
Connectivity module, for carrying out Threshold-connected to extract cartilage.
According to described system, the extraction module includes：
Statistic submodule, for setting the gray threshold scope of joint bony areas in magnetic resonance image (MRI), counts whole magnetic
The gray value of each point of each pixel of resonance image in radius neighborhood R；
First acquisition submodule, if for each point in the above-mentioned pixel radius neighborhood gray value all in skeleton threshold
In the range of value, then above-mentioned pixel is set to into 1；Remaining situation is then set to 0, acquires image X；
Described image X, for estimating cartilage maximum gauge, is done n times expansion according to cartilage thickness by the second acquisition submodule
Computing, obtains image Y；
Extracting sub-module, for a pixel on the skeleton of joint is chosen in described image Y, extracts and this pixel
Point has the region of connected relation, is labeled as region Z, and the region Z is the detection zone of the articular cartilage.
According to described system, the computing module includes：
Construction submodule, for constructing Gaussian filter Gσ, σ is Gaussian filter variance, to Gaussian filter GσSeek second order
Derivative, uses GσSecond dervative and the magnetic resonance image (MRI) region Z in each pixel carry out convolution and obtain Ixx、Ixy、Ixz、
Iyy、Iyz、Izz；
Sorting sub-module, for constituting Hessian matrixesCalculate three eigenvalue λs of H1、λ2、
λ3, and three eigenvalues are sorted according to order from small to large：λ1＜ λ2＜ λ3；
Judging submodule, for constructing lamellar architectural feature discriminant function Vs(λ1, λ2, λ3), the function pair each pixel
The Hessian matrix exgenvalues of value carry out structural determination, for laminated structure region, the Hessian matrix eigenvalue distributions
Rule should meet [λ1＜ 0, λ2≈ 0, λ3≈0]；
Calculating sub module, for calculating the laminated structure characteristic function of all pixels point in the region Z, wherein H (λ1)
It is to weigh λ1The function of action effect.
According to described system, the enhancing module includes：
Regular submodule, for carrying out at consolidation to laminated structure characteristic function L (λ) of pixel in the region Z
Reason；
Strengthen submodule, for pixel in the region Z is entered according to the laminated structure characteristic function of each pixel
Row grey level enhancement, is strengthened the region for belonging to laminated structure.
According to described system, the connectivity module includes：
Setting submodule, for setting the substantially gray threshold scope of the articular cartilage after grey level enhancement；
Submodule is chosen, for choosing a seed point in the articular cartilage, extraction has threshold value with the seed point
The part of connected relation, the part are the partition data of articular cartilage described in image.
Cartilage detection zone of the present invention by extraction of magnetic resonance image；Then calculate lamellar in the cartilage detection zone
Architectural feature function；Then grey level enhancement is carried out to the laminated structure in the cartilage detection zone；Threshold-connected is carried out finally
To extract cartilage.Thus, on the automatic cutting techniques of MRI cartilages, algorithm research and development elapsed time is longer at present for solution, partitioning algorithm
The problem not strong to different quality data adaptability, proposes a kind of based on cutting object gradation of image shape information, without the need for training
The method of study carries out extracting directly to cartilage, and proposition can be according to the adjustable multiple parameters of different pieces of information quality, to carry
The adaptability that high algorithm is split to various qualitative data cartilages.
Description of the drawings
Fig. 1 is the system knot based on the enhanced articular cartilage segmentation of image laminated structure that first embodiment of the invention is provided
Structure schematic diagram；
Fig. 2 be the present invention second and third, four, five embodiments provide based on the enhanced articular cartilage of image laminated structure point
The system structure diagram cut；
Fig. 3 is the method stream based on the enhanced articular cartilage segmentation of image laminated structure that sixth embodiment of the invention is provided
Cheng Tu；
Fig. 4 is the method stream based on the enhanced articular cartilage segmentation of image laminated structure that one embodiment of the invention is provided
Cheng Tu；
Fig. 5 A are knee joint femoral cartilage artworks in one embodiment of the invention；
Fig. 5 B are extraction femur design sketchs in one embodiment of the invention；
Fig. 5 C are femoral cartilage detection zone design sketchs in one embodiment of the invention；
Fig. 5 D are laminated structure output effect figures in one embodiment of the invention；
Fig. 5 E are knee joint femoral cartilage extraction effect figures in one embodiment of the invention；
Fig. 6 A are Patella Cartilage and perienchyma's artworks in one embodiment of the invention
Fig. 6 B are patella extraction effect figures in one embodiment of the invention；
Fig. 6 C are Patella Cartilage detection zone design sketchs in one embodiment of the invention；
Fig. 6 D are laminated structure output effect figures in one embodiment of the invention；
Fig. 6 E are Patella Cartilage extraction effect figures in one embodiment of the invention.
Specific embodiment
In order that the objects, technical solutions and advantages of the present invention become more apparent, it is below in conjunction with drawings and Examples, right
The present invention is further elaborated.It should be appreciated that specific embodiment described herein is only to explain the present invention, and
It is not used in the restriction present invention.
Referring to Fig. 1, in the first embodiment of the present invention there is provided based on the enhanced articular cartilage point of image laminated structure
The system 100 cut, the system 100 based on the enhanced articular cartilage segmentation of image laminated structure include：
Extraction module 10, for the cartilage detection zone of extraction of magnetic resonance image；
Computing module 20, for calculating laminated structure characteristic function in the cartilage detection zone；
Strengthen module 30, for grey level enhancement being carried out to the laminated structure in the cartilage detection zone；
Connectivity module 40, for carrying out Threshold-connected to extract cartilage.
What the embodiment was provided belongs to medical image based on the system 100 of the enhanced articular cartilage segmentation of image laminated structure
Process, Artificial is intelligent, computer medical science assisting in diagnosis and treatment systems technology field, more particularly to soft to joint using computer
Bone carries out the technology of automatic segmentation.In this embodiment, it is contemplated that articular cartilage particularity in shape, employ a kind of contracting
Little area-of-interest simultaneously carries out laminated structure enhanced method cartilage is split automatically to image.It is in this technology, first
First extraction module 10 is split to joint skeleton, then cartilage region is determined by bony areas, to reach
Downscaled images detection range and the effect for eliminating other interference tissues.Then computing module 20 passes through in cartilage detection range
Hessian matrix exgenvalues carry out characteristic statisticses to the shape of each pixel composition of each pixel and certain neighborhood and divide
Class, strengthening module 30 carries out grey level enhancement by the pixel for meeting strip-like features again, and last connectivity module 40 is connected with gray threshold
It is logical to realize that cartilage is split, the cutting techniques strong adaptability.
Referring to Fig. 2, in the second embodiment of the present invention, extraction module 10 includes：
Statistic submodule 11, for setting the gray threshold scope of joint bony areas in magnetic resonance image (MRI), statistics is whole
The gray value of each point of each pixel of magnetic resonance image (MRI) in radius neighborhood R；
First acquisition submodule 12, if for each point in the above-mentioned pixel radius neighborhood gray value all in skeleton
In threshold range, then above-mentioned pixel is set to into 1；Remaining situation is then set to 0, acquires image X；
Described image X, for estimating cartilage maximum gauge, is done n times according to cartilage thickness swollen by the second acquisition submodule 13
Swollen computing, obtains image Y；
Extracting sub-module 14, for a pixel on the skeleton of joint is chosen in described image Y, extracts and this picture
Vegetarian refreshments has the region of connected relation, is labeled as region Z, and the region Z is the detection zone of the articular cartilage.
In this embodiment, therefore, using a kind of improved Threshold-connected algorithm can rapid extraction joint bony areas,
And degrees of expansion regulation can be carried out to cartilage detection zone according to the substantially thickness of cartilage.The algorithm can determine rapidly tightly
The ambital cartilaginous areas of close bag, and the interference of the ligament for being connected with cartilage and being difficult to differentiate between can be effectively removed, it is cartilage
Fast Segmentation carries out detection basis.
Referring to Fig. 2, in the third embodiment of the present invention, computing module 20 includes：
Construction submodule 21, for constructing Gaussian filter Gσ, σ is Gaussian filter variance, to Gaussian filter GσAsk two
Order derivative, uses GσSecond dervative and the magnetic resonance image (MRI) region Z in each pixel carry out convolution and obtain Ixx、Ixy、
Ixz、Iyy、Iyz、Izz；
Sorting sub-module 22, for constituting Hessian matrixesCalculate three eigenvalue λs of H1、
λ2、λ3, and three eigenvalues are sorted according to order from small to large：λ1＜ λ2＜ λ3；
Judging submodule 23, for constructing lamellar architectural feature discriminant function Vs(λ1, λ2, λ3), the function pair each picture
The Hessian matrix exgenvalues of plain value carry out structural determination, for laminated structure region, the Hessian matrix exgenvalues point
Cloth rule should meet [λ1＜ 0, λ2≈ 0, λ3≈0]；
Calculating sub module 24, for calculating the laminated structure characteristic function of all pixels point in the region Z, wherein H
(λ1) it is to weigh λ1The function of action effect.
In this embodiment, cartilage can be adapted to just using the laminated structure feature decision function of Hessian matrixes special
Different laminated structure, effectively extracts to cartilage, and in the presence of adjustable parameter, laminated structure feature decision function can
The data of various quality are correspondingly acted on, so as to strengthen the adaptability of cartilage identification.
Referring to Fig. 2, in the fourth embodiment of the present invention, strengthening module 30 includes：
Regular submodule 31, for carrying out consolidation to laminated structure characteristic function L (λ) of pixel in the region Z
Process；
Strengthen submodule 32, for by pixel in the region Z according to each pixel laminated structure characteristic function
Grey level enhancement is carried out, is strengthened the region for belonging to laminated structure.
Referring to Fig. 2, in the fifth embodiment of the present invention, connectivity module 40 includes：
Setting submodule 41, for setting the substantially gray threshold scope of the articular cartilage after grey level enhancement；
Submodule 42 is chosen, for choosing a seed point in the articular cartilage, extraction has threshold with the seed point
The part of value connected relation, the part are the partition data of articular cartilage described in image.
In this embodiment, directly the gray feature of image is differentiated, is started with to cartilage from the second dervative of gray scale
Make a distinction with image background, so effective detection can be carried out to cartilage boundary directly, when image second order derivative is extracted, use
Adjustable Gauss yardstick can be correspondingly differentiated to the cartilage boundary of different pieces of information quality.
In above-mentioned multiple embodiments, multiple modules of the system can be software or hardware, and software and hardware mould
Block.Solved at present in MRI (magnetic based on the system 100 of the enhanced articular cartilage segmentation of image laminated structure by practical above-mentioned
Resonance image) algorithm research and development elapsed time is longer on the automatic cutting techniques of cartilage, and partitioning algorithm is to different quality data adaptability
Not strong problem, proposes a kind of based on cutting object gradation of image shape information, without the need for training the system for learning to carry out cartilage
Extracting directly, and proposition can be according to the adjustable multiple parameters of different pieces of information quality, to improve algorithm to various qualitative datas
The adaptability of cartilage segmentation.
Referring to Fig. 3, in the sixth embodiment of the present invention, there is provided a kind of soft based on the enhanced joint of image laminated structure
The method of bone segmentation, methods described comprise the steps：
In step S301, the cartilage detection zone of extraction of magnetic resonance image；The step is realized by extraction module 10；
In step S302, laminated structure characteristic function in the cartilage detection zone is calculated；The step is by computing module 20
Realize；
In step S303, grey level enhancement is carried out to the laminated structure in the cartilage detection zone；The step is by strengthening mould
Block 30 is realized；
In step S304, carry out Threshold-connected to extract cartilage；The step is realized by connectivity module 40.
In this embodiment, first skeleton is extracted and is automatically determined cartilaginous areas, then directly image is carried out
Regional processing and detection, need not collect sample and training study；Directly articular cartilage is carried out from the angle of tissue profile feature
Extract, time loss can be greatly reduced with detection before detection, and provide multiple adjustable parameters to adapt to different quality
Data, make different pieces of information quality cartilage split all reach degree of precision.
In the seventh embodiment of the present invention, step S301 includes：
The gray threshold scope of joint bony areas in A1, setting magnetic resonance image (MRI), count whole magnetic resonance image (MRI) each
The gray value of each point of the pixel in radius neighborhood R；The step is realized by statistic submodule 11；
If A2, each gray value put in above-mentioned pixel radius neighborhood, will be upper all in skeleton threshold range
State pixel and be set to 1；Remaining situation is then set to 0, acquires image X；The step is realized by the first acquisition submodule 12；
Described image X is done n times dilation operation according to cartilage thickness, obtains image Y by A3, estimation cartilage maximum gauge；Should
Step is realized by the second acquisition submodule 13；
A4, a pixel on selection joint skeleton in described image Y, extracting has connected relation with this pixel
Region, be labeled as region Z, the region Z is the detection zone of the articular cartilage；The step is real by extracting sub-module 14
It is existing.
In this embodiment, joint bony areas extraction is carried out to MRI image, and expansion area makes which just surround covering
Cartilage on skeleton, in this, as cartilage detection zone.
In the eighth embodiment of the present invention, step S302 includes：
B1, construction Gaussian filter Gσ, σ is Gaussian filter variance, to Gaussian filter GσSecond dervative is sought, G is usedσ's
Each pixel in second dervative and the magnetic resonance image (MRI) region Z carries out convolution and obtains Ixx、Ixy、Ixz、Iyy、Iyz、Izz；Should
Step is realized by submodule 21 is constructed；
B2, composition Hessian matrixesCalculate three eigenvalue λs of H1、λ2、λ3, and by described three
Individual eigenvalue sorts according to order from small to large：λ1＜ λ2＜ λ3；The step is realized by sorting sub-module 22；
B3, construction lamellar architectural feature discriminant function Vs(λ1, λ2, λ3), the Hessian of each pixel value of function pair
Matrix exgenvalue carries out structural determination, and for laminated structure region, the Hessian matrix eigenvalue distributions rule should meet
[λ1＜ 0, λ2≈ 0, λ3≈0]；The step is realized by judging submodule 23；
B4, the laminated structure characteristic function for calculating all pixels point in the region Z, wherein H (λ1) it is to weigh λ1Effect effect
The function of fruit；The step is realized by calculating sub module 24.
In this embodiment, the Hessian matrix exgenvalues of each point in cartilage detection zone are extracted, and uses laminated structure
Feature decision function pair Hessian matrix exgenvalue is judged.
In the ninth embodiment of the present invention, step S303 includes：
C1, laminated structure characteristic function L (λ) to pixel in the region Z carry out consolidation process；The step is by advising
Whole submodule 31 is realized；
C2, pixel in the region Z is carried out into grey level enhancement according to the laminated structure characteristic function of each pixel, made
The region for belonging to laminated structure is strengthened；The step is realized by submodule 32 is strengthened.
In this embodiment, grey level enhancement is carried out to the pixel that judgement belongs to laminated structure region.
In the tenth embodiment of the present invention, step S304 includes：
The substantially gray threshold scope of the articular cartilage after D1, setting grey level enhancement；The step is by setting submodule 41
Realize；
D2, the seed point chosen in the articular cartilage, extract the portion for having Threshold-connected relation with the seed point
Point, the part is the partition data of articular cartilage described in image；The step is realized by submodule 42 is chosen.
In this embodiment, carry out Threshold-connected and extract cartilage, obtain cartilage partition data.
In above-mentioned multiple embodiments, there is provided the method based on the segmentation of image laminated structure enhanced articular cartilage, bag
Included the improved Threshold-connected algorithm for skeleton being extracted in MRI image；According to bony areas and cartilage substantially thickness pair
The determination method of cartilage detection zone；With the algorithmic function V of Hessian matrix detection image laminated structuress(λ1, λ2, λ3) and
Laminated structure grey level enhancement algorithm, on the automatic cutting techniques of MRI cartilages, algorithm research and development elapsed time is longer at present for solution, segmentation
The algorithm problem not strong to different quality data adaptability, propose it is a kind of based on cutting object gradation of image shape information, need not
The method of training study carries out extracting directly to cartilage, and propose can according to the adjustable multiple parameters of different pieces of information quality,
To improve the adaptability that algorithm is split to various qualitative data cartilages.
Referring to Fig. 4, in one embodiment of the invention there is provided based on the enhanced articular cartilage point of image laminated structure
The method cut, is described as follows：
In step S401, MRI image is input into；
In step S402, bony areas are extracted；
In step S403, cartilaginous areas are determined；
In step S404, Hessian matrix lamellar structural characteristic parameters in each region are calculated；
In step S405, laminated structure is extracted；
In step S406, grey level enhancement is carried out to laminated structure；
In step S407, Threshold-connected extracts cartilage partition data.
In this embodiment, the cartilage detection zone of MRI image is extracted first；Specifically, set joint bone in MRI image
The gray threshold scope in bone region, counts the gray value of each point of whole each pixel of MRI image in radius neighborhood R；
If above-mentioned pixel is set to all in skeleton threshold range by the gray value of each point in above-mentioned pixel radius neighborhood
1；Remaining situation is then set to 0, so obtains image X；Estimate cartilage maximum gauge, image X is done by n times expansion according to cartilage thickness
Computing, obtains image Y；A pixel on the skeleton of joint is chosen on image Y, extracting has connected relation with this pixel
Region, be designated as region Z, region Z is the detection zone of articular cartilage；
Then, Hessian matrix lamellar architectural feature functions are calculated；Specifically, construct Gaussian filter Gσ, σ is Gauss
Wave filter variance, to Gaussian filter GσSecond dervative is sought, G is usedσSecond dervative and MRI image region Z in each pixel
Carry out convolution and obtain Ixx、Ixy、Ixz、Iyy、Iyz、Izz；Constitute Hessian matrixesCalculate H three are special
Value indicative λ1、λ2、λ3, and these three eigenvalues are sorted according to order from small to large：λ1＜ λ2＜ λ3；Construction lamellar architectural feature is sentenced
Other function Vs(λ1, λ2, λ3), the function plays the Hessian matrix exgenvalues to each pixel value and carries out the work of structural determination
With for laminated structure region, its Hessian matrix eigenvalue distributions rule should meet [λ1＜ 0, λ2≈ 0, λ3≈0]；Calculate
The laminated structure characteristic function of all pixels point, wherein H (λ in the Z of region1) it is to weigh λ1The function of action effect.
Then, grey level enhancement is carried out to the laminated structure pixel in the Z of region；Specifically, the piece to pixel in the Z of region
Shape architectural feature function L (λ) carries out consolidation process；And pixel in the Z of region is special according to the laminated structure of each pixel
Levying function carries out grey level enhancement, is strengthened the region for belonging to laminated structure.
Finally, Threshold-connected extracts cartilage；The substantially gray threshold scope of specific setting grey level enhancement articular cartilage of immobilized；
A seed point in articular cartilage is chosen, the part for having Threshold-connected relation with this pixel is extracted, the part is as schemed
The partition data of articular cartilage as in.
Referring to Fig. 5 A～5E, in one embodiment of the invention, there is provided knee joint femoral cartilage dividing method, segmentation
Step is as follows：
1st, extract knee cartilage detection zone
1.1st, extract knee joint femoral；
1.2nd, expansion area is obtaining cartilage detection zone
2nd, calculate laminated structure characteristic function
2.1st, Hessian matrix exgenvalues in cartilage detection zone are calculated；
2.2nd, lamellar architectural feature function is constructed according to the Characteristic Distribution of laminated structure；
2.3rd, the output of laminated structure characteristic function is tried to achieve as independent variable using Hessian matrix exgenvalues；
3rd, grey level enhancement is carried out to the laminated structure in femoral cartilage region
3.1st, the laminated structure ash in femoral cartilage detection zone is lifted according to the output result of laminated structure characteristic function
Degree；
4th, Threshold-connected extracts cartilage
4.1st, set femoral cartilage threshold range；
4.2nd, set seed point and extract femoral cartilage.
Referring to Fig. 6 A～6E, in one embodiment of the invention, there is provided Patella Cartilage dividing method, segmentation step is such as
Under：
1st, extract cartilage detection zone
1.1st, extract knee joint patella；
1.2nd, expansion area is obtaining Patella Cartilage detection zone
2nd, calculate laminated structure characteristic function
2.1st, Hessian matrix exgenvalues in cartilage detection zone are calculated；
2.2nd, lamellar architectural feature function is constructed according to the Characteristic Distribution of laminated structure；
2.3rd, the output of laminated structure characteristic function is tried to achieve as independent variable using Hessian matrix exgenvalues；
3rd, grey level enhancement is carried out to the laminated structure in Patella Cartilage region
3.1st, the laminated structure ash in Patella Cartilage detection zone is lifted according to the output result of laminated structure characteristic function
Degree；
4th, Threshold-connected extracts Patella Cartilage
4.1st, set Patella Cartilage threshold range；
4.2nd, set seed point and extract Patella Cartilage.
In sum, cartilage detection zone of the present invention by extraction of magnetic resonance image；Then calculate the cartilage detection
Laminated structure characteristic function in region；Then grey level enhancement is carried out to the laminated structure in the cartilage detection zone；It is most laggard
Row Threshold-connected is extracting cartilage.Thus, solve at present on the automatic cutting techniques of MRI cartilages algorithm research and development elapsed time compared with
Long, the partitioning algorithm problem not strong to different quality data adaptability is proposed a kind of based on cutting object gradation of image shape letter
Breath, the method without the need for training study carry out extracting directly to cartilage, and proposition can be adjustable more according to different pieces of information quality
Individual parameter, to improve the adaptability that algorithm is split to various qualitative data cartilages.
Certainly, the present invention can also have other various embodiments, in the case of without departing substantially from spirit of the invention and its essence, ripe
Know those skilled in the art and work as and various corresponding changes and deformation, but these corresponding changes and change can be made according to the present invention
Shape should all belong to the protection domain of appended claims of the invention.
Claims (8)
1. a kind of method based on the segmentation of image laminated structure enhanced articular cartilage, it is characterised in that methods described include as
Lower step：
The cartilage detection zone of A, extraction of magnetic resonance image；
Laminated structure characteristic function in B, the calculating cartilage detection zone；
C, grey level enhancement is carried out to the laminated structure in the cartilage detection zone；
D, carry out Threshold-connected to extract cartilage；
Step A includes：
In A1, setting magnetic resonance image (MRI), the gray threshold scope of joint bony areas, counts whole each pixel of magnetic resonance image (MRI)
The gray value of each point of the point in radius neighborhood R；
If A2, each gray value put in above-mentioned pixel radius neighborhood are all in skeleton threshold range, by above-mentioned picture
Vegetarian refreshments is set to 1；Remaining situation is then set to 0, acquires image X；
Described image X is done n times dilation operation according to cartilage thickness, obtains image Y by A3, estimation cartilage maximum gauge；
A4, a pixel on selection joint skeleton in described image Y, extract the area for having connected relation with this pixel
Domain, is labeled as region Z, and the region Z is the detection zone of the articular cartilage.
2. method according to claim 1, it is characterised in that step B includes：
B1, construction Gaussian filter Gσ, σ is Gaussian filter variance, to Gaussian filter GσSecond dervative is sought, G is usedσSecond order
Each pixel in derivative and the magnetic resonance image (MRI) region Z carries out convolution and obtains Ixx、Ixy、Ixz、Iyy、Iyz、Izz；
B2, composition Hessian matrixesCalculate three eigenvalue λs of H1、λ2、λ3, and it is special by described three
Value indicative sorts according to order from small to large：λ1＜ λ2＜ λ3；
B3, construction lamellar architectural feature discriminant function Vs(λ1, λ2, λ3), the Hessian matrixes of each pixel value of function pair are special
Value indicative carries out structural determination, and for laminated structure region, the Hessian matrix eigenvalue distributions rule should meet [λ1＜ 0, λ2
≈ 0, λ3≈0]；
B4, the laminated structure characteristic function for calculating all pixels point in the region Z, wherein H (λ1) it is to weigh λ1Action effect
Function.
3. method according to claim 2, it is characterised in that step C includes：
C1, laminated structure characteristic function L (λ) to pixel in the region Z carry out consolidation process；
C2, pixel in the region Z is carried out into grey level enhancement according to the laminated structure characteristic function of each pixel, make to belong to
The region of laminated structure is strengthened.
4. method according to claim 3, it is characterised in that step D includes：
The gray threshold scope of the articular cartilage after D1, setting grey level enhancement；
D2, the seed point chosen in the articular cartilage, extract the part for having Threshold-connected relation with the seed point, should
Partition data of the part for articular cartilage described in image.
5. a kind of system based on the segmentation of image laminated structure enhanced articular cartilage, it is characterised in that the system includes：
Extraction module, for the cartilage detection zone of extraction of magnetic resonance image；
Computing module, for calculating laminated structure characteristic function in the cartilage detection zone；
Strengthen module, for grey level enhancement being carried out to the laminated structure in the cartilage detection zone；
Connectivity module, for carrying out Threshold-connected to extract cartilage；
The extraction module includes：
Statistic submodule, for setting the gray threshold scope of joint bony areas in magnetic resonance image (MRI), counts whole magnetic resonance
The gray value of each point of each pixel of image in radius neighborhood R；
First acquisition submodule, if for each point in the above-mentioned pixel radius neighborhood gray value all in skeleton threshold value model
In enclosing, then above-mentioned pixel is set to into 1；Remaining situation is then set to 0, acquires image X；
Described image X, for estimating cartilage maximum gauge, is done n times dilation operation according to cartilage thickness by the second acquisition submodule,
Obtain image Y；
Extracting sub-module, for a pixel on the skeleton of joint is chosen in described image Y, extraction is had with this pixel
The region of connected relation, is labeled as region Z, and the region Z is the detection zone of the articular cartilage.
6. system according to claim 5, it is characterised in that the computing module includes：
Construction submodule, for constructing Gaussian filter Gσ, σ is Gaussian filter variance, to Gaussian filter GσSecond order is asked to lead
Number, uses GσSecond dervative and the magnetic resonance image (MRI) region Z in each pixel carry out convolution and obtain Ixx、Ixy、Ixz、Iyy、
Iyz、Izz；
Sorting sub-module, for constituting Hessian matrixesCalculate three eigenvalue λs of H1、λ2、λ3,
And three eigenvalues sort according to order from small to large：λ1＜ λ2＜ λ3；
Judging submodule, for constructing lamellar architectural feature discriminant function Vs(λ1, λ2, λ3), each pixel value of function pair
Hessian matrix exgenvalues carry out structural determination, for laminated structure region, the Hessian matrix eigenvalue distributions rule
[λ should be met1＜ 0, λ2≈ 0, λ3≈0]；
Calculating sub module, for calculating the laminated structure characteristic function of all pixels point in the region Z, wherein H (λ1) it is to weigh
λ1The function of action effect.
7. system according to claim 6, it is characterised in that the enhancing module includes：
Regular submodule, for carrying out consolidation process to laminated structure characteristic function L (λ) of pixel in the region Z；
Strengthen submodule, for pixel in the region Z is carried out ash according to the laminated structure characteristic function of each pixel
Degree strengthens, and is strengthened the region for belonging to laminated structure.
8. system according to claim 7, it is characterised in that the connectivity module includes：
Setting submodule, for setting the gray threshold scope of the articular cartilage after grey level enhancement；
Submodule is chosen, for choosing a seed point in the articular cartilage, extraction has Threshold-connected with the seed point
The part of relation, the part are the partition data of articular cartilage described in image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310325364.2A CN103456004B (en) | 2013-07-30 | 2013-07-30 | Articular cartilage partitioning method and articular cartilage partitioning system based on image sheet structure enhancement |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310325364.2A CN103456004B (en) | 2013-07-30 | 2013-07-30 | Articular cartilage partitioning method and articular cartilage partitioning system based on image sheet structure enhancement |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103456004A CN103456004A (en) | 2013-12-18 |
| CN103456004B true CN103456004B (en) | 2017-04-19 |
Family
ID=49738332
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310325364.2A Active CN103456004B (en) | 2013-07-30 | 2013-07-30 | Articular cartilage partitioning method and articular cartilage partitioning system based on image sheet structure enhancement |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103456004B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108784705B (en) * | 2018-05-28 | 2020-12-15 | 上海交通大学 | High-performance analysis method for joint image |
| CN109544673B (en) * | 2018-10-19 | 2023-06-23 | 瑞梦德医药科技(北京)有限公司 | Three-dimensional imaging method and system for cartilage segmentation of medical image |
| CN109741352A (en) * | 2018-12-25 | 2019-05-10 | 深圳市第二人民医院 | Cartilage damage based on multi-modal magnetic resonance repairs image partition method |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6799066B2 (en) * | 2000-09-14 | 2004-09-28 | The Board Of Trustees Of The Leland Stanford Junior University | Technique for manipulating medical images |
| CN102968783A (en) * | 2012-10-15 | 2013-03-13 | 深圳市旭东数字医学影像技术有限公司 | Method and system for automatically segmenting bones from abdomen image data |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP1974313A4 (en) * | 2005-12-30 | 2011-11-16 | Yeda Res & Dev | An integrated segmentation and classification approach applied to medical applications analysis |
-
2013
- 2013-07-30 CN CN201310325364.2A patent/CN103456004B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6799066B2 (en) * | 2000-09-14 | 2004-09-28 | The Board Of Trustees Of The Leland Stanford Junior University | Technique for manipulating medical images |
| CN102968783A (en) * | 2012-10-15 | 2013-03-13 | 深圳市旭东数字医学影像技术有限公司 | Method and system for automatically segmenting bones from abdomen image data |
Non-Patent Citations (6)
| Title |
|---|
| Automatic knee cartilage segmentation from multi-contrast MR images using support vector machine classification with spatial dependencies;Kunlei Zhang et al.;《Magnetic resonance imaging》;20130715;第1731-1743页 * |
| Automatic knee cartilage segmentation from multi-contrast MR images using support vector machine classification with spatial dependencies;Kunlei Zhang et al.;《Magnetic resonance imaging》;20130715;第2.2节，第3.2.1节，图4，附录A * |
| Automatic Segmentation of the Articular Cartilage in Knee MRI Using a Hierarchical Multi-class Classification Scheme;Jenny Folkesson et al.;《MICCAI 2005》;20051231;第327-334页 * |
| Segmenting Articular Cartilage Automatically Using a Voxel Classification Approach;Jenny Folkesson et al.;《IEEE TRANSACTIONS ON MEDICAL IMAGING》;20070131;第26卷(第1期);第106-115页 * |
| 基于 MRI 的股骨头软骨厚度测量方法的研究;曹宇;《中国博士学位论文全文数据库信息科技辑》;20110515(第5期);第1138-42页 * |
| 基于Hessian矩阵的血管图像增强与水平集分割算法研究;陈丽平;《中国优秀硕士学位论文全文数据库信息科技辑》;20130215(第2期);第I138-1616页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103456004A (en) | 2013-12-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103942577B (en) | Based on the personal identification method for establishing sample database and composite character certainly in video monitoring | |
| CN110070935B (en) | Medical image synthesis method, classification method and device based on antagonistic neural network | |
| Jose et al. | Brain tumor segmentation using k-means clustering and fuzzy c-means algorithms and its area calculation | |
| CN105894517B (en) | CT image liver segmentation method and system based on feature learning | |
| CN104809740B (en) | Knee cartilage image automatic segmentation method based on SVM and Hookean region growth | |
| CN104794708A (en) | Atherosclerosis plaque composition dividing method based on multi-feature learning | |
| CN109697718A (en) | A kind of self-closing disease detection method and device based on graph theory | |
| CN108364294A (en) | Abdominal CT images multiple organ dividing method based on super-pixel | |
| CN110969204B (en) | Sample classification system based on fusion of magnetic resonance image and digital pathology image | |
| Hu et al. | Segmentation of brain from computed tomography head images | |
| CN109064476A (en) | A kind of CT rabat lung tissue image partition method based on level set | |
| CN108564561A (en) | Pectoralis major region automatic testing method in a kind of molybdenum target image | |
| CN103699904A (en) | Image computer-aided diagnosis method for multi-sequence nuclear magnetic resonance images | |
| CN107358267A (en) | A kind of breast ultrasound image multivariate classification system and method based on cross-correlation feature | |
| CN103456004B (en) | Articular cartilage partitioning method and articular cartilage partitioning system based on image sheet structure enhancement | |
| Goswami et al. | A hybrid neuro-fuzzy approach for brain abnormality detection using GLCM based feature extraction | |
| Saygili et al. | Knee meniscus segmentation and tear detection from MRI: a review | |
| Anandgaonkar et al. | Brain tumor detection and identification from T1 post contrast MR images using cluster based segmentation | |
| CN103871057A (en) | Magnetic resonance image-based bone segmentation method and system thereof | |
| CN110021019A (en) | A kind of thickness distributional analysis method of the AI auxiliary hair of AGA clinical image | |
| CN110033448A (en) | A kind of male bald Hamilton classification prediction analysis method of AI auxiliary of AGA clinical image | |
| CN113989551A (en) | Alzheimer disease classification method based on improved ResNet network | |
| Johny et al. | Breast cancer detection in mammogram using fuzzy C-means and random forest classifier | |
| Li | Medical image segmentation based on watershed transformation and rough sets | |
| CN110399891A (en) | A kind of efficient sort management method of the medical image based on big data |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |