CN104008545A - Method for achieving human egg cell segmentation in optical microinjection process - Google Patents

Abstract

The invention discloses a method for achieving human egg cell segmentation in an optical microinjection process. The method includes the following steps that (1) an interested target ROI is extracted through a framework morphology operator according to shape differences of segmentation targets; (2) rough classification is conducted on preprocessed images with fuzzy boundaries and large noise according to the local grey level and the local variance characteristics by the adoption of an improved fuzzy clustering method; (3) the level set initial position and evolution control parameters are set based on the fuzzy clustering, and segmentation of cell plasma, a polar body and an entry needle is achieved; (4) for partial entry needle images in the cell plasma, the complete entry needle image is marked through direct fitting. In a single sperm microinjection process in ICSI, namely, egg cytoplasm, through the method, the segmentation of the cell plasma of a human egg cell, the polar body of the human egg cell and the entry needle can be effectively achieved, automation and increasing of the success rate are facilitated, and the requirement for high-difficulty manual operation and dependence on manual operation are reduced.

Description

One human egg in optical microphotograph injection process is cut apart implementation method

Technical field

The present invention relates to Medical Image Processing and applied technical field, refer in particular to one human egg in optical microphotograph injection process and cut apart implementation method.

Background technology

Cell microinjection is the important technical in biological gene engineering.The present invention mainly studies in ICSI (Intra Cytoplasmic Sperm Injection) is ooecium slurry in monosperm microinjection process, the segmentation problem of human egg's tenuigenin, polar body and entry needle.From egg mother cell, be partitioned into tenuigenin and polar body, not only can help analysis expert oocyte maturation degree (as the thickness of diameter, granularity, oolemma, polar body diameter, all gap sizes of ovum etc.), and contribute to the raising of ICSI robotization realization and success ratio, and reduce manually-operated highly difficult requirement and dependence.

The sight of ICSI injection as shown in Figure 6.Mainly be divided into egg cell, entry needle and three objects of absorption pin.Wherein, egg cell periphery fuzzy region is corona radiata, and black curve band is oolemma inside corona radiata, and the material that oolemma surrounds is tenuigenin, the polar body that cytoplasmic upper rounded particulate material is cell; The absorption pin in left side is used for adsorbing and fixing egg cell; The entry needle of the long fine strip shape in the right is realized transport and the injection of sperm.In ICSI process, need to guarantee that entry needle correctly thrusts the position of cell, do not damage the polar body of cell, and judge whether really to pierce through tenuigenin simultaneously.So, realize above-mentioned cut apart most important.

Through the literature search of prior art is found, the robotization that realizes in recent years cell microinjection is current primary study direction.In the research and development of experimental provision, existing scholar launches the people such as basic research work Ladjal H at paper " Micro-to-Nano biomechanical modeling for assisted biological cell injection " (IEEE Transactions on Biomedical Engineering both at home and abroad, 2013, (60) 9:2461-2471), and the medium people in osmanthus, field is at paper " cell pose regulation technology and experimental study in microinjection " (China Mechanical Engineering, 2009, 20 (4): 500-503) in, introduce respectively the injection platform based on optical microphotograph of building.

And aspect the selection of egg mother cell, the people such as Manna are at paper " Artificial intelligence techniques for embryo and oocyte classification " (Reproductive biomedicine online, 2013,26 (1): 42-49) in, use textural characteristics to realize classification and the selection to egg mother cell, prepare for next step ICSI improves success ratio.The people such as Basile are at paper " A texture-based image processing approach for the description of human oocyte cytoplasm " (Instrumentation and Measurement, IEEE Transactions on, 2010, 59 (10): 2591-2601) in, hew out a rectangular area from tenuigenin center by image pre-service, after Har wavelet transformation, extract multiple Statistic Textures of egg mother cell, use again FCM (Fuzzy C-Means) cluster to realize the egg mother cell classification of granularity in various degree.The people such as Caponetti are respectively at paper " Multiresolution texture analysis for human oocyte cytoplasm description " (Medical Measurements and Applications, 2009.MeMeA2009.IEEE International Workshop on.IEEE, 2009:150-155) and in paper " Fuzzy mathematical morphology for biological image segmentation " (Applied Intelligence.2014:1-11), first realize and cutting apart with fuzzy mathematics method, and combined with texture feature realizes egg mother cell classification.But the pre-service of Basile and Caponetti all uses Hough conversion to realize the pre-segmentation of round cell matter, cannot be applied to the situation of cell deformation, and not be partitioned into the polar body of cell. deng people at paper " Elevated active contour with global image energy based on electrostatic force " (Zeszyty Naukowe Politechniki informatyka, 2010:5-21) in the power-actuated overall active contour snake model of a kind of basic exterior static has been proposed, although be partitioned into cytoplasmic profile, manual initialization evolution profile, and do not realize cutting apart of cell polar body.

For the motion control of entry needle, the people such as Zhang are at paper " Controlled aspiration and positioning of biological cells in a micropipette " (IEEE Transactions on Biomedical Engineering, 2012,59 (4): 1032-1040) in, study the closed loop control method that entry needle is moved based on micro-vision, realized location and the tracking of the sperm motility in entry needle simultaneously.

In sum, the various experimental provisions of ICSI and the signature analysis of egg mother cell have attracted scholar's concern, but for cutting apart and rare research of orientation problem of egg mother cell in cell microinjection process, and this is cell microinjection robotization key point.Therefore in primary study ICSI process, tenuigenin, entry needle and egg cell polar body are cut apart herein.

Summary of the invention

The object of the invention is to overcome the deficiencies in the prior art and defect, provide one effectively, human egg is cut apart implementation method in optical microphotograph injection process reliably, can under common hardware condition, realize human egg and cut apart.

For achieving the above object, technical scheme provided by the present invention is: one human egg in optical microphotograph injection process is cut apart implementation method, comprises the following steps:

1), according to the segmentation object shape difference opposite sex, adopt matrix morphology operator extraction interesting target ROI;

2) adopt improved fuzzy clustering method, according to local gray level and local Variance feature, obscurity boundary and the larger pretreatment image of noise are carried out to rough sort;

3) on the basis of fuzzy clustering, set level set initialized location and Evolution Control parameter, realize cutting apart of tenuigenin, polar body and entry needle;

4) for intracytoplasmic part entry needle image, adopt direct matching, mark complete entry needle image.

In step 1) in, combining form is learned operator and skeletal extraction is carried out pre-segmentation to cytoplasm image, its process is: first, obtain complete bianry image by OTSU method, then do dilation operation, again through two-value skeletonizing, intactly extract the polar body of tenuigenin, egg mother cell and the profile of absorption pin, its profile width is a pixel, absorption pin skeleton is turned to single pixel straight line simultaneously, the image of skeletonizing only has tenuigenin to be communicated with, and is easy to fill and mark; Secondly,, by area filling and mark, then service regeulations wave filter divides the tenuigenin of egg mother cell to leave from the object background of periphery; Finally, extract interesting target ROI.

In step 2) in, described improved fuzzy clustering method is to add local gray level and variance information, the classification of current central point pixel is determined jointly by the key words sorting of the pixel in its neighbour territory, thereby realize the rough sort of tenuigenin, polar body and endocellular injection pin, its process is as follows:

Suppose that the real image y that collects is made up of true picture x and biasing territory b, and b slowly changes, image y has N pixel altogether, has:

$y_k=x_k+b_k,\∀k\∈\{\mathrm{1,2},...,N\}---\left(1\right)$

So, as follows to the FCM objective cost function rectification of standard:

$J_m=\underset{i=1}{\overset{c}{\mathrm{\Σ}}}\underset{k=1}{\overset{N}{\mathrm{\Σ}}}{u}_{\mathrm{ik}}^p{\left|\right|x_k-v_i\left|\right|}^2+\frac{\mathrm{\α}}{N_R}\underset{i=1}{\overset{c}{\mathrm{\Σ}}}\underset{k=1}{\overset{N}{\mathrm{\Σ}}}{u}_{\mathrm{ik}}^p\left(\underset{x\∈N_k}{\mathrm{\Σ}}{\left|\right|x_r-v_i\left|\right|}^2\right),\underset{j=1}{\overset{c}{\mathrm{\Σ}}}{u}_{\mathrm{jk}}=1,\∀k---\left(2\right)$

Wherein, p is fuzzy membership weighted index, the fog-level of the final classification of impact; N _krepresent with x _kcentered by neighbour territory; N _rrepresent N _kthe number of middle pixel; α is the control parameter in neighbour territory, and value is inversely proportional to the signal to noise ratio (S/N ratio) of image, and above formula (1) substitution is had:

$J_m=\underset{i=1}{\overset{c}{\mathrm{\Σ}}}\underset{k=1}{\overset{N}{\mathrm{\Σ}}}{u}_{\mathrm{ik}}^p{\left|\right|y_k-b_k-v_i\left|\right|}^2+\frac{\mathrm{\α}}{N_R}\underset{i=1}{\overset{c}{\mathrm{\Σ}}}\underset{k=1}{\overset{N}{\mathrm{\Σ}}}{u}_{\mathrm{ik}}^p\left(\underset{y_r\∈N_k}{\mathrm{\Σ}}{\left|\right|y_r-b_r-v_i\left|\right|}^2\right)---\left(3\right)$

Use Lagrange multiplier to ask the optimal problem of above formula:

$F_m=\underset{i=1}{\overset{c}{\mathrm{\Σ}}}\underset{k=1}{\overset{N}{\mathrm{\Σ}}}(u_{\mathrm{ik}}^p{D}_{\mathrm{ik}}+\frac{\mathrm{\α}}{N_R}{u}_{\mathrm{ik}}^p{\mathrm{\γ}}_i)+\mathrm{\λ}(1-\underset{i=1}{\overset{c}{\mathrm{\Σ}}}{u}_{\mathrm{ik}}^p)---\left(4\right)$

Wherein, D _ik=|| y _k-b _k-v _i|| ²,

Respectively to u _ik, v _ikasking single order partial derivative and making result is 0, can obtain degree of membership and cluster centre in like manner, to biasing territory b _kestimate, asking single order partial derivative and making its result is 0, can be in the hope of:

$u_{\mathrm{ik}}^{*}=\underset{j=1}{\overset{c}{\mathrm{\Σ}}}{\left(\frac{D_{\mathrm{ik}}+\frac{\mathrm{\α}}{N_R}{\mathrm{\γ}}_i}{(D_{\mathrm{jk}}+\frac{\mathrm{\α}}{N_R}{\mathrm{\γ}}_j)}\right)}^{-1/(p-1)}---\left(5\right)$

$v_{\mathrm{ik}}^{*}=\frac{\underset{k=1}{\overset{N}{\mathrm{\Σ}}}{u}_{\mathrm{ik}}^p((y_k-b_k)+\frac{\mathrm{\α}}{N_R}\underset{y_r\∈N_k}{\mathrm{\Σ}}(y_r-b_r))}{(1+\mathrm{\α})\underset{k=1}{\overset{N}{\mathrm{\Σ}}}{u}_{\mathrm{ik}}^p}---\left(6\right)$

$b_k^{*}=y_k-\frac{\underset{i=1}{\overset{c}{\mathrm{\Σ}}}{u}_{\mathrm{ik}}^p{v}_i}{\underset{i=1}{\overset{c}{\mathrm{\Σ}}}{u}_{\mathrm{ik}}^p}---\left(7\right).$

In step 3) in, in conjunction with fuzzy clustering result, realize the initialization of level set position and level set movements parameter is set, the result of fuzzy clustering is used as to the direction and the speed parameter that develop without initialized variation level set, make can automatically determine on profile border evolution direction, accurately lock-on boundary position, thereby correctly the cutting apart of realize target; Variation equation is as follows:

$E_0\left(\mathrm{\φ}\right)=\mathrm{\μ}\underset{\mathrm{\Ω}}{\∫\∫}\frac{1}{2}{\left(\right|\▿\mathrm{\φ}|-1)}^2\mathrm{dxdy}+\mathrm{\λ}\underset{\mathrm{\Ω}}{\∫\∫}\mathrm{g\δ}\left(\mathrm{\φ}\right)|\▿\mathrm{\φ}|\mathrm{dxdy}+v\underset{\mathrm{\Ω}}{\∫\∫}\mathrm{gH}(-\mathrm{\φ})\mathrm{dxdy}---\left(8\right)$

Wherein, φ is level set function, and three, the right is respectively the area item that penalty term, zero level collection length of curve regular terms and acceleration curve develop; Coefficient is respectively μ, λ and ν; G is edge indicator function, and δ (φ) is single argument Dirac function, and H is odd symmetry Heaviside function; By the Euler-Lagrange variational method, in the gradient descent method solving model of use standard, energy function minimizes; The Evolution Equation of obtaining level set is:

$\frac{\∂\mathrm{\φ}}{\∂t}=\mathrm{\μ}[\mathrm{\Δ\φ}-\mathrm{div}\left(\frac{\▿\mathrm{\φ}}{|\▿\mathrm{\φ}|}\right)]+\mathrm{\δ}\left(\mathrm{\φ}\right)[\mathrm{\λ}\mathrm{div}\left(g\frac{\▿\mathrm{\φ}}{|\▿\mathrm{\φ}|}\right)+\mathrm{vg}]---\left(9\right)$

Wherein, the positive and negative decision initial profile of coefficient ν value does contraction or expansion with respect to target to be split and develops, and the speed developing, and for actual segmentation problem, while expecting wide border, Evolution Rates is fast, otherwise, develop slow; In the time that profile strides across object boundary, can automatically change evolution direction, but not unalterable; Therefore, adaptive change coefficient ν value is more suitable for the actual demand of cutting apart, and the image that makes fuzzy clustering degree of membership is that R has so:

ν _k＝G(R _k)＝1-2R _k,R _k∈(0,1) (10)

The Evolution Equation of level set is updated to:

$\frac{\∂\mathrm{\φ}}{\∂t}=\mathrm{\μ}[\mathrm{\Δ\φ}-\mathrm{div}\left(\frac{\▿\mathrm{\φ}}{|\▿\mathrm{\φ}|}\right)]+\mathrm{\δ}\left(\mathrm{\φ}\right)[\mathrm{\λ}\mathrm{div}\left(g\frac{\▿\mathrm{\φ}}{|\▿\mathrm{\φ}|}\right)+\mathrm{Gg}]---\left(11\right)$

Choosing of other parameter:

μ＝0.2/ζ，λ＝0.1ζ， $\mathrm{\ζ}=\frac{\∫\mathrm{\δ}\left({\mathrm{\φ}}_0\right)\mathrm{dxdy}}{\∫H\left({\mathrm{\φ}}_0\right)\mathrm{dxdy}}$

Level set initialization and evolution thereof are as follows:

The image of supposing selected fuzzy clustering degree of membership is R, sets an adjustable threshold value b ₀∈ (0,1), obtains a bianry image B _k:

$B_k=\left\{\begin{array}{c}1,R_k\&GreaterEqual;b_0\\ 0,R_k<b_0\end{array}\right.---\left(12\right)$

So, level set initialization function reach for:

φ ₀＝-4ε(0.5-B _k),ε＝1.5 (13)。

Compared with prior art, tool has the following advantages and beneficial effect in the present invention:

1, in ICSI is ooecium slurry in monosperm microinjection process, can effectively realize the segmentation problem of human egg's tenuigenin, polar body and entry needle by the inventive method, contribute to the raising of robotization realization and success ratio, and reduce manually-operated highly difficult requirement and dependence;

2, adopt improved fuzzy clustering algorithm to improve the problem of traditional F CM to noise-sensitive, add local gray level and variance information, the classification of current central point pixel is determined jointly by the key words sorting of the pixel in its neighbour territory, thereby realize the rough sort of tenuigenin, polar body and endocellular injection pin, inside is substantially level and smooth, little particle and empty bubble, borderline phase is to clear;

3, the result that fuzzy clustering obtains can be used for the initial position that lead-out level collection develops, reduce curve in-depth operand, improve segmentation precision, the result of cluster is used as to the direction and the speed parameter that develop without initialized variation level set, make can automatically determine on profile border evolution direction, accurately lock-on boundary position.

Brief description of the drawings

Fig. 1 is the process flow diagram that human egg of the present invention is cut apart implementation method.

Fig. 2 a is the original image before image pre-service.

Skeletal extraction figure when Fig. 2 b is image pre-service.

Fig. 2 c is the pretreated pre-segmentation result of image figure.

Fig. 3 a is that ROI image adopts the pre-segmentation image before improved fuzzy clustering method.

Fig. 3 b is the cluster result figure of ooplasm and polar body.

Fig. 3 c is the image of part entry needle in ROI image.

Fig. 4 a is the image of cutting apart of the tenuigenin of level set while cutting apart and polar body.

Fig. 4 b is level set movements 3 d effect graph.

Fig. 4 c is the segmentation result figure of the inner one section of entry needle of egg mother cell matter.

Entry needle when Fig. 5 a is entry needle matching is cut apart figure.

Rectilinear in cell when Fig. 5 b is entry needle matching.

Fig. 5 c is the complete pin image after entry needle matching.

Fig. 6 is the ICSI injection sight figure in background technology;

In figure: 1 is absorption pin, and 2 is tenuigenin, and 3 is polar body, 4 is cell week gap, and 5 is oolemma, and 6 is entry needle.

Embodiment

Below in conjunction with specific embodiment, the invention will be further described.

As shown in Figure 1, the human egg in optical microphotograph injection process described in the present embodiment is cut apart implementation method, comprises the following steps:

1) according to the segmentation object shape difference opposite sex, adopt matrix morphology operator extraction interesting target ROI, be specially extraction and comprise polar body, one section of entry needle and cytoplasmic pretreatment image;

2) consider that pretreatment image intensity is uneven and be subject to noise pollution, adopt improved fuzzy clustering method, according to local gray level and local Variance feature, obscurity boundary and the larger pretreatment image of noise are carried out to rough sort, realize the rough sort of tenuigenin, polar body and endocellular injection pin;

3) on the basis of fuzzy clustering, set level set initialized location and Evolution Control parameter, realize correctly cutting apart of tenuigenin, polar body and entry needle;

4) for intracytoplasmic part entry needle image, adopt direct matching, simulate the position of entry needle, mark complete entry needle image.

In step 1) in, combining form is learned operator and skeletal extraction is carried out pre-segmentation to cytoplasm image, considers that tenuigenin is transparent in two pixel wide, and combining form of the present invention is learned the knowledge of operator and skeletal extraction, the coarse segmentation of realize target.Bianry image skeletonizing can be reduced to lines by object, but does not revise the basic structure of image, retains well the elementary contour of image simultaneously.

First, obtain complete bianry image by OTSU method, then do dilation operation, again through two-value skeletonizing, intactly extract the polar body of tenuigenin, egg mother cell and the profile of absorption pin, its profile width is a pixel, absorption pin skeleton is turned to single pixel straight line simultaneously, the image of skeletonizing only has tenuigenin to be communicated with, and is easy to fill and mark; Secondly,, by area filling and mark, then service regeulations wave filter divides the tenuigenin of egg mother cell to leave from the object background of periphery; Finally, extract interesting target ROI.

In step 2) in, adopting improved fuzzy clustering method is to add local gray level and variance information, the classification of current central point pixel is determined jointly by the key words sorting of the pixel in its neighbour territory, thereby realize the rough sort of tenuigenin, polar body and endocellular injection pin.

The ROI image that pre-service is extracted for image, has three cutting objects: one section of entry needle in tenuigenin, polar body and tenuigenin.Because ICSI view data collects from optical microscope, have uneven illumination and noise pollution, and traditional FCM is a kind of cost function of overall mean square deviation minimum, in the time there is abnormity point, there is poor cluster result.And noise unavoidably occurs in real Biomedical Image.Conventionally solving thinking is to consider to add local characteristics, and the classification of current central point pixel is determined jointly by the key words sorting of the pixel in its neighbour territory, and then improves the problem of traditional F CM to noise-sensitive.

Suppose that the real image y that collects is made up of true picture x and biasing territory b, and b slowly changes, image y has N pixel altogether, has:

$y_k=x_k+b_k,\&ForAll;k\&Element;\{\mathrm{1,2},...,N\}---\left(1\right)$

So, as follows to the FCM objective cost function rectification of standard:

$J_m=\underset{i=1}{\overset{c}{\mathrm{\&Sigma;}}}\underset{k=1}{\overset{N}{\mathrm{\&Sigma;}}}{u}_{\mathrm{ik}}^p{\left|\right|x_k-v_i\left|\right|}^2+\frac{\mathrm{\&alpha;}}{N_R}\underset{i=1}{\overset{c}{\mathrm{\&Sigma;}}}\underset{k=1}{\overset{N}{\mathrm{\&Sigma;}}}{u}_{\mathrm{ik}}^p\left(\underset{x\&Element;N_k}{\mathrm{\&Sigma;}}{\left|\right|x_r-v_i\left|\right|}^2\right),\underset{j=1}{\overset{c}{\mathrm{\&Sigma;}}}{u}_{\mathrm{jk}}=1,\&ForAll;k---\left(2\right)$

Wherein, p is fuzzy membership weighted index, the fog-level of the final classification of impact; N _krepresent with x _kcentered by neighbour territory; N _rrepresent N _kthe number of middle pixel; α is the control parameter in neighbour territory, and value is inversely proportional to the signal to noise ratio (S/N ratio) of image, and above formula (1) substitution is had:

$J_m=\underset{i=1}{\overset{c}{\mathrm{\&Sigma;}}}\underset{k=1}{\overset{N}{\mathrm{\&Sigma;}}}{u}_{\mathrm{ik}}^p{\left|\right|y_k-b_k-v_i\left|\right|}^2+\frac{\mathrm{\&alpha;}}{N_R}\underset{i=1}{\overset{c}{\mathrm{\&Sigma;}}}\underset{k=1}{\overset{N}{\mathrm{\&Sigma;}}}{u}_{\mathrm{ik}}^p\left(\underset{y_r\&Element;N_k}{\mathrm{\&Sigma;}}{\left|\right|y_r-b_r-v_i\left|\right|}^2\right)---\left(3\right)$

Use Lagrange multiplier to ask the optimal problem of above formula:

$F_m=\underset{i=1}{\overset{c}{\mathrm{\&Sigma;}}}\underset{k=1}{\overset{N}{\mathrm{\&Sigma;}}}(u_{\mathrm{ik}}^p{D}_{\mathrm{ik}}+\frac{\mathrm{\&alpha;}}{N_R}{u}_{\mathrm{ik}}^p{\mathrm{\&gamma;}}_i)+\mathrm{\&lambda;}(1-\underset{i=1}{\overset{c}{\mathrm{\&Sigma;}}}{u}_{\mathrm{ik}}^p)---\left(4\right)$

Wherein, D _ik=|| y _k-b _k-v _i|| ²,

Respectively to u _ik, v _ikasking single order partial derivative and making result is 0, can obtain degree of membership and cluster centre in like manner, to biasing territory b _kestimate, asking single order partial derivative and making its result is 0, can be in the hope of:

$u_{\mathrm{ik}}^{*}=\underset{j=1}{\overset{c}{\mathrm{\&Sigma;}}}{\left(\frac{D_{\mathrm{ik}}+\frac{\mathrm{\&alpha;}}{N_R}{\mathrm{\&gamma;}}_i}{(D_{\mathrm{jk}}+\frac{\mathrm{\&alpha;}}{N_R}{\mathrm{\&gamma;}}_j)}\right)}^{-1/(p-1)}---\left(5\right)$

$v_{\mathrm{ik}}^{*}=\frac{\underset{k=1}{\overset{N}{\mathrm{\&Sigma;}}}{u}_{\mathrm{ik}}^p((y_k-b_k)+\frac{\mathrm{\&alpha;}}{N_R}\underset{y_r\&Element;N_k}{\mathrm{\&Sigma;}}(y_r-b_r))}{(1+\mathrm{\&alpha;})\underset{k=1}{\overset{N}{\mathrm{\&Sigma;}}}{u}_{\mathrm{ik}}^p}---\left(6\right)$

$b_k^{*}=y_k-\frac{\underset{i=1}{\overset{c}{\mathrm{\&Sigma;}}}{u}_{\mathrm{ik}}^p{v}_i}{\underset{i=1}{\overset{c}{\mathrm{\&Sigma;}}}{u}_{\mathrm{ik}}^p}---\left(7\right).$

In step 3) in, in conjunction with fuzzy clustering result, realize the initialization of level set position and level set movements parameter is set, the result of fuzzy clustering is used as to the direction and the speed parameter that develop without initialized variation level set, make can automatically determine on profile border evolution direction, accurately lock-on boundary position, thereby correctly the cutting apart of realize target.

The result that fuzzy clustering obtains is a rough ideal sort, the initial position that available cluster result lead-out level collection develops, thus reduce curve in-depth operand, improve segmentation precision.Level Set Method is easy to guide or drive in conjunction with priori the evolution of profile.The priori such as Li propose without initialized variation level diversity method, variation equation is as follows:

$E_0\left(\mathrm{\&phi;}\right)=\mathrm{\&mu;}\underset{\mathrm{\&Omega;}}{\&Integral;\&Integral;}\frac{1}{2}{\left(\right|\&dtri;\mathrm{\&phi;}|-1)}^2\mathrm{dxdy}+\mathrm{\&lambda;}\underset{\mathrm{\&Omega;}}{\&Integral;\&Integral;}\mathrm{g\&delta;}\left(\mathrm{\&phi;}\right)|\&dtri;\mathrm{\&phi;}|\mathrm{dxdy}+v\underset{\mathrm{\&Omega;}}{\&Integral;\&Integral;}\mathrm{gH}(-\mathrm{\&phi;})\mathrm{dxdy}---\left(8\right)$

Wherein, φ is level set function, and three, the right is respectively the area item that penalty term, zero level collection length of curve regular terms and acceleration curve develop; Coefficient is respectively μ, λ and ν; G is edge indicator function, and δ (φ) is single argument Dirac function, and H is odd symmetry Heaviside function; By the Euler-Lagrange variational method, in the gradient descent method solving model of use standard, energy function minimizes; The Evolution Equation of obtaining level set is:

$\frac{\&PartialD;\mathrm{\&phi;}}{\&PartialD;t}=\mathrm{\&mu;}[\mathrm{\&Delta;\&phi;}-\mathrm{div}\left(\frac{\&dtri;\mathrm{\&phi;}}{|\&dtri;\mathrm{\&phi;}|}\right)]+\mathrm{\&delta;}\left(\mathrm{\&phi;}\right)[\mathrm{\&lambda;}\mathrm{div}\left(g\frac{\&dtri;\mathrm{\&phi;}}{|\&dtri;\mathrm{\&phi;}|}\right)+\mathrm{vg}]---\left(9\right)$

Wherein, the positive and negative decision initial profile of coefficient ν value does contraction or expansion with respect to target to be split and develops, and the speed developing, and for actual segmentation problem, while expecting wide border, Evolution Rates is fast, otherwise, develop slow; In the time that profile strides across object boundary, can automatically change evolution direction, but not unalterable; Therefore, adaptive change coefficient ν value is more suitable for the actual demand of cutting apart, and the image that makes fuzzy clustering degree of membership is that R has so:

ν _k＝G(R _k)＝1-2R _k,R _k∈(0,1) (10)

The Evolution Equation of level set is updated to:

$\frac{\&PartialD;\mathrm{\&phi;}}{\&PartialD;t}=\mathrm{\&mu;}[\mathrm{\&Delta;\&phi;}-\mathrm{div}\left(\frac{\&dtri;\mathrm{\&phi;}}{|\&dtri;\mathrm{\&phi;}|}\right)]+\mathrm{\&delta;}\left(\mathrm{\&phi;}\right)[\mathrm{\&lambda;}\mathrm{div}\left(g\frac{\&dtri;\mathrm{\&phi;}}{|\&dtri;\mathrm{\&phi;}|}\right)+\mathrm{Gg}]---\left(11\right)$

Choosing of other parameter:

μ＝0.2/ζ，λ＝0.1ζ， $\mathrm{\&zeta;}=\frac{\&Integral;\mathrm{\&delta;}\left({\mathrm{\&phi;}}_0\right)\mathrm{dxdy}}{\&Integral;H\left({\mathrm{\&phi;}}_0\right)\mathrm{dxdy}}$

Level set initialization and evolution thereof are as follows:

Fuzzy clustering obtains a coarse classification results, the initial position that can develop from cluster result lead-out level collection, thus reduce operand.The image of supposing selected fuzzy clustering degree of membership is R, sets an adjustable threshold value b ₀∈ (0,1), obtains a bianry image B _k:

$B_k=\left\{\begin{array}{c}1,R_k\&GreaterEqual;b_0\\ 0,R_k<b_0\end{array}\right.---\left(12\right)$

So, level set initialization function reach for:

φ ₀＝-4ε(0.5-B _k),ε＝1.5 (13)。

Below in conjunction with Fig. 2 a to Fig. 5 c, human egg of the present invention to be cut apart to implementation method and be specifically described, its situation is as follows:

Experiment porch is PentiumIV3.0G, 1G RAM, and video card is NVIDIA Quadro FX1400.

Original image as shown in Figure 2 a, obtains complete bianry image by OTSU method, extracts skeleton, as shown in Figure 2 b after expansive working.Skeletonizing has intactly retained the profile of tenuigenin, polar body, absorption pin and entry needle, and width is single pixel, and entry needle skeleton turns to single pixel straight line, and in the image of skeletonizing, tenuigenin is communicated with; Secondly,, by area filling and mark, then service regeulations wave filter divides the tenuigenin of egg mother cell to leave from the object background of periphery; Finally, extract interesting target ROI, as shown in Figure 2 c.

Cluster feature is chosen each pixel local mean value and local variance, according to FCM sorting procedure, fuzzy clustering result is as shown in Fig. 3 a to Fig. 3 c, Fig. 3 a is the pre-service ROI image that must put in place, be divided into two classes with improved fuzzy clustering algorithm, Fig. 3 b is the cluster result of ooplasm and polar body, and tenuigenin inside is substantially level and smooth, borderline phase is to clear, and the Main Boundaries region of polar body is able to separate with tenuigenin; Fig. 3 c is the image of part entry needle in ROI image, and the cell outline of periphery can be removed in subsequent operation.

Utilize method of finite difference to realize quantizing and solve, time partial derivative value adopts forward difference, and quantizing of space local derviation adopts central difference.On the basis of dendrogram picture, the cluster result of the cluster result based on egg mother cell matter and entry needle respectively, realizes and cutting apart in conjunction with level set equation.

Fig. 4 a is the image of cutting apart of tenuigenin and polar body, smoother and develop exactly and boundary profile; Fig. 4 b is level set movements 3 d effect graph, the position that middle empty region is entry needle, and bottom right independently Yi Ge community is polar body; Fig. 4 c is the segmentation result of the inner one section of entry needle of egg mother cell matter.Consider the linear feature of entry needle, on the basis of binaryzation, and in conjunction with Fig. 4 c, realize straight-line detection.Fig. 5 a to Fig. 5 c has shown the complete cutting procedure of entry needle.

The above examples of implementation, only for preferred embodiment of the present invention, not limits practical range of the present invention with this, therefore the variation that all shapes according to the present invention, principle are done all should be encompassed in protection scope of the present invention.

Claims (4)

1. in optical microphotograph injection process, human egg is cut apart an implementation method, it is characterized in that, comprises the following steps:

1), according to the segmentation object shape difference opposite sex, adopt matrix morphology operator extraction interesting target ROI;

2) adopt improved fuzzy clustering method, according to local gray level and local Variance feature, obscurity boundary and the larger pretreatment image of noise are carried out to rough sort;

3) on the basis of fuzzy clustering, set level set initialized location and Evolution Control parameter, realize cutting apart of tenuigenin, polar body and entry needle;

4) for intracytoplasmic part entry needle image, adopt direct matching, mark complete entry needle image.

2. one according to claim 1 human egg in optical microphotograph injection process is cut apart implementation method, it is characterized in that: in step 1) in, combining form is learned operator and skeletal extraction is carried out pre-segmentation to cytoplasm image, its process is: first, obtain complete bianry image by OTSU method, then do dilation operation, again through two-value skeletonizing, intactly extract tenuigenin, the profile of the polar body of egg mother cell and absorption pin, its profile width is a pixel, absorption pin skeleton is turned to single pixel straight line simultaneously, the image of skeletonizing only has tenuigenin to be communicated with, be easy to fill and mark, secondly,, by area filling and mark, then service regeulations wave filter divides the tenuigenin of egg mother cell to leave from the object background of periphery, finally, extract interesting target ROI.

3. one according to claim 1 human egg in optical microphotograph injection process is cut apart implementation method, it is characterized in that: in step 2) in, described improved fuzzy clustering method is to add local gray level and variance information, the classification of current central point pixel is determined jointly by the key words sorting of the pixel in its neighbour territory, thereby the rough sort that realizes tenuigenin, polar body and endocellular injection pin, its process is as follows:

Suppose that the real image y that collects is made up of true picture x and biasing territory b, and b slowly changes, image y has N pixel altogether, has:

$y_k=x_k+b_k,\&ForAll;k\&Element;\{\mathrm{1,2},...,N\}---\left(1\right)$

So, as follows to the FCM objective cost function rectification of standard:

$J_m=\underset{i=1}{\overset{c}{\mathrm{\&Sigma;}}}\underset{k=1}{\overset{N}{\mathrm{\&Sigma;}}}{u}_{\mathrm{ik}}^p{\left|\right|x_k-v_i\left|\right|}^2+\frac{\mathrm{\&alpha;}}{N_R}\underset{i=1}{\overset{c}{\mathrm{\&Sigma;}}}\underset{k=1}{\overset{N}{\mathrm{\&Sigma;}}}{u}_{\mathrm{ik}}^p\left(\underset{x\&Element;N_k}{\mathrm{\&Sigma;}}{\left|\right|x_r-v_i\left|\right|}^2\right),\underset{j=1}{\overset{c}{\mathrm{\&Sigma;}}}{u}_{\mathrm{jk}}=1,\&ForAll;k---\left(2\right)$

Wherein, p is fuzzy membership weighted index, the fog-level of the final classification of impact; N _krepresent with x _kcentered by neighbour territory; N _rrepresent N _kthe number of middle pixel; α is the control parameter in neighbour territory, and value is inversely proportional to the signal to noise ratio (S/N ratio) of image, and above formula (1) substitution is had:

$J_m=\underset{i=1}{\overset{c}{\mathrm{\&Sigma;}}}\underset{k=1}{\overset{N}{\mathrm{\&Sigma;}}}{u}_{\mathrm{ik}}^p{\left|\right|y_k-b_k-v_i\left|\right|}^2+\frac{\mathrm{\&alpha;}}{N_R}\underset{i=1}{\overset{c}{\mathrm{\&Sigma;}}}\underset{k=1}{\overset{N}{\mathrm{\&Sigma;}}}{u}_{\mathrm{ik}}^p\left(\underset{y_r\&Element;N_k}{\mathrm{\&Sigma;}}{\left|\right|y_r-b_r-v_i\left|\right|}^2\right)---\left(3\right)$

Use Lagrange multiplier to ask the optimal problem of above formula:

$F_m=\underset{i=1}{\overset{c}{\mathrm{\&Sigma;}}}\underset{k=1}{\overset{N}{\mathrm{\&Sigma;}}}(u_{\mathrm{ik}}^p{D}_{\mathrm{ik}}+\frac{\mathrm{\&alpha;}}{N_R}{u}_{\mathrm{ik}}^p{\mathrm{\&gamma;}}_i)+\mathrm{\&lambda;}(1-\underset{i=1}{\overset{c}{\mathrm{\&Sigma;}}}{u}_{\mathrm{ik}}^p)---\left(4\right)$

Wherein, D _ik=|| y _k-b _k-v _i|| ²,

Respectively to u _ik, v _ikasking single order partial derivative and making result is 0, can obtain degree of membership and cluster centre in like manner, to biasing territory b _kestimate, asking single order partial derivative and making its result is 0, can be in the hope of:

$u_{\mathrm{ik}}^{*}=\underset{j=1}{\overset{c}{\mathrm{\&Sigma;}}}{\left(\frac{D_{\mathrm{ik}}+\frac{\mathrm{\&alpha;}}{N_R}{\mathrm{\&gamma;}}_i}{(D_{\mathrm{jk}}+\frac{\mathrm{\&alpha;}}{N_R}{\mathrm{\&gamma;}}_j)}\right)}^{-1/(p-1)}---\left(5\right)$

$v_{\mathrm{ik}}^{*}=\frac{\underset{k=1}{\overset{N}{\mathrm{\&Sigma;}}}{u}_{\mathrm{ik}}^p((y_k-b_k)+\frac{\mathrm{\&alpha;}}{N_R}\underset{y_r\&Element;N_k}{\mathrm{\&Sigma;}}(y_r-b_r))}{(1+\mathrm{\&alpha;})\underset{k=1}{\overset{N}{\mathrm{\&Sigma;}}}{u}_{\mathrm{ik}}^p}---\left(6\right)$

$b_k^{*}=y_k-\frac{\underset{i=1}{\overset{c}{\mathrm{\&Sigma;}}}{u}_{\mathrm{ik}}^p{v}_i}{\underset{i=1}{\overset{c}{\mathrm{\&Sigma;}}}{u}_{\mathrm{ik}}^p}---\left(7\right).$

4. one according to claim 1 human egg in optical microphotograph injection process is cut apart implementation method, it is characterized in that: in step 3) in, in conjunction with fuzzy clustering result, realize the initialization of level set position and level set movements parameter is set, the result of fuzzy clustering is used as to the direction and the speed parameter that develop without initialized variation level set, make can automatically determine on profile border evolution direction, accurately lock-on boundary position, thereby correctly the cutting apart of realize target; Variation equation is as follows:

$E_0\left(\mathrm{\&phi;}\right)=\mathrm{\&mu;}\underset{\mathrm{\&Omega;}}{\&Integral;\&Integral;}\frac{1}{2}{\left(\right|\&dtri;\mathrm{\&phi;}|-1)}^2\mathrm{dxdy}+\mathrm{\&lambda;}\underset{\mathrm{\&Omega;}}{\&Integral;\&Integral;}\mathrm{g\&delta;}\left(\mathrm{\&phi;}\right)|\&dtri;\mathrm{\&phi;}|\mathrm{dxdy}+v\underset{\mathrm{\&Omega;}}{\&Integral;\&Integral;}\mathrm{gH}(-\mathrm{\&phi;})\mathrm{dxdy}---\left(8\right)$

Wherein, φ is level set function, and three, the right is respectively the area item that penalty term, zero level collection length of curve regular terms and acceleration curve develop; Coefficient is respectively μ, λ and ν; G is edge indicator function, and δ (φ) is single argument Dirac function, and H is odd symmetry Heaviside function; By the Euler-Lagrange variational method, in the gradient descent method solving model of use standard, energy function minimizes; The Evolution Equation of obtaining level set is:

$\frac{\&PartialD;\mathrm{\&phi;}}{\&PartialD;t}=\mathrm{\&mu;}[\mathrm{\&Delta;\&phi;}-\mathrm{div}\left(\frac{\&dtri;\mathrm{\&phi;}}{|\&dtri;\mathrm{\&phi;}|}\right)]+\mathrm{\&delta;}\left(\mathrm{\&phi;}\right)[\mathrm{\&lambda;}\mathrm{div}\left(g\frac{\&dtri;\mathrm{\&phi;}}{|\&dtri;\mathrm{\&phi;}|}\right)+\mathrm{vg}]---\left(9\right)$

Wherein, the positive and negative decision initial profile of coefficient ν value does contraction or expansion with respect to target to be split and develops, and the speed developing, and for actual segmentation problem, while expecting wide border, Evolution Rates is fast, otherwise, develop slow; In the time that profile strides across object boundary, can automatically change evolution direction, but not unalterable; Therefore, adaptive change coefficient ν value is more suitable for the actual demand of cutting apart, and the image that makes fuzzy clustering degree of membership is that R has so:

ν _k＝G(R _k)＝1-2R _k,R _k∈(0,1) (10)

The Evolution Equation of level set is updated to:

$\frac{\&PartialD;\mathrm{\&phi;}}{\&PartialD;t}=\mathrm{\&mu;}[\mathrm{\&Delta;\&phi;}-\mathrm{div}\left(\frac{\&dtri;\mathrm{\&phi;}}{|\&dtri;\mathrm{\&phi;}|}\right)]+\mathrm{\&delta;}\left(\mathrm{\&phi;}\right)[\mathrm{\&lambda;}\mathrm{div}\left(g\frac{\&dtri;\mathrm{\&phi;}}{|\&dtri;\mathrm{\&phi;}|}\right)+\mathrm{Gg}]---\left(11\right)$

Choosing of other parameter:

μ＝0.2/ζ，λ＝0.1ζ， $\mathrm{\&zeta;}=\frac{\&Integral;\mathrm{\&delta;}\left({\mathrm{\&phi;}}_0\right)\mathrm{dxdy}}{\&Integral;H\left({\mathrm{\&phi;}}_0\right)\mathrm{dxdy}}$

Level set initialization and evolution thereof are as follows:

The image of supposing selected fuzzy clustering degree of membership is R, sets an adjustable threshold value b ₀∈ (0,1), obtains a bianry image B _k:

$B_k=\left\{\begin{array}{c}1,R_k\&GreaterEqual;b_0\\ 0,R_k<b_0\end{array}\right.---\left(12\right)$

So, level set initialization function reach for:

φ ₀＝-4ε(0.5-B _k),ε＝1.5 (13)。