CN104021539A - System used for automatically detecting tumour in ultrasonic image - Google Patents

Abstract

A system used for automatically detecting a tumour in an ultrasonic image comprises the flowing structures: a candidate area detector used for detecting at least one candidate area comprising the tumour in the ultrasonic image according to a deformable part model (DPM); a tumour area positioning device used for determining a tumour area from at least one detected candidate area; a tumour contour isolator used for isolating a tumour contour according to the determined tumour area so as to detect the tumour. By employing the system, the area comprising the tumour can be effectively detected from the ultrasonic image without man participation; the tumour contour can be relatively isolated on the base so as to serve as one evidence for tumour diagnosis.

Description

For automatically detect the system of tumour at ultrasonoscopy

Technical field

The present invention relates to Medical Image Processing, relate in particular to a kind of system that detects tumor section from ultrasonoscopy.

Background technology

In modern medicine, ultrasonic imaging is for example, important means for diagnosing various tumours (, the thoracic cavity tumours such as tumor of breast), because ultrasonic examination is relatively convenient, does not invade human body and cost is lower.But the doctor of image department need to mark for every width ultrasonoscopy artificially, to reflect the image characteristics of corresponding organ, as the image basis that judges tumour.But, a large amount of patients' result of ultrasonography to be carried out to manual mark one by one and need to expend a large amount of time, human cost is higher.

Therefore, people pay close attention to the scheme for automatically detect tumour from ultrasonoscopy.For example, Drukker, K. wait people at " Computerized lesion detection on breastultrasound ", Med.Phys., 29 (7): in 1438-46 (2002), propose diseased region almost darker than background area to some extent, and generate tumor region based on this strict hypothesis.

But, in the automatic detection scheme of existing tumour described above for example, because the picture quality of ultrasonic testing results itself is poor, and, the form of tumour is very complicated, mostly not only is irregular figure, and usually follows calcification, therefore, be difficult to effectively in ultrasonoscopy, detect the region at tumour place.

In addition,, because whether relevant range is that tumour is also woven with substantial connection with the particular group at its place, therefore, in existing lesion detection technology, be also easy to occur erroneous judgement.

Summary of the invention

The object of the present invention is to provide a kind of system that can effectively automatically detect tumour from ultrasonoscopy.

According to an aspect of the present invention, provide a kind of for automatically detect the system of tumour at ultrasonoscopy, comprising: candidate region pick-up unit, for detect at least one candidate region that comprises tumour from ultrasonoscopy based on deformable component model (DPM); Tumor region locating device, for determining tumor region from described at least one candidate region detecting; Tumor's profiles tripping device, the profile that separates tumour for the tumor region by based on definite detects tumour.

Candidate region pick-up unit can detect at least one candidate region that comprises tumour by the ratio of width to height that changes the root template in DMP from ultrasonoscopy.

Tumor region locating device can use the two-value sorter based on support vector machine (SVM) to determine tumor region from described at least one candidate region detecting.

The proper vector of described two-value sorter can be based on candidate region contextual feature.

Described proper vector can comprise at least one in following: the DPM of candidate region detects all parts in the position of score value, candidate region and size, candidate region with respect to the part that coexists between the intensity difference between prospect and background in the side-play amount of root, candidate region, candidate region and the highest candidate region of DPM detection score value.

Tumor region locating device can use Multiple Kernel Learning (MKL) method the kernel function of two-value sorter to be defined as to the linear combination of multiple basic kernel functions.

Tumor region locating device can carry out linear combination by the polynomial kernel function of the RBF kernel function of three kinds of bandwidth and three dimensions, thereby train for each characteristic component in proper vector and proper vector entirety, to obtain the kernel function of two-value sorter.

Tumor's profiles tripping device can usage level diversity method separates the profile of tumour, and wherein, the contour curve of tumour carries out iteration using the border of tumor region as initial curve.

Tumor's profiles tripping device can be by making the distance between prospect and the background of tumor image maximize to build the energy function adopting in Level Set Method.

According to a further aspect in the invention, provide a kind of for the system at the automatic detected object of ultrasonoscopy, comprising: candidate region pick-up unit, for detect at least one candidate region that comprises object from ultrasonoscopy based on deformable component model (DPM); Subject area locating device, for determining subject area from described at least one candidate region detecting; Object outline tripping device, carrys out detected object for the profile of the subject area separate object by based on definite.

Candidate region pick-up unit can detect at least one candidate region that comprises object by the ratio of width to height that changes the root template in DMP from ultrasonoscopy.

Subject area locating device can use the two-value sorter based on support vector machine (SVM) to determine subject area from described at least one candidate region detecting.

The proper vector of described two-value sorter can be based on candidate region contextual feature.

Subject area locating device can use Multiple Kernel Learning (MKL) method the kernel function of two-value sorter to be defined as to the linear combination of multiple basic kernel functions.

According to a further aspect in the invention, provide a kind of in the ultrasonoscopy method of detected object automatically, comprising: detect at least one candidate region that comprises object from ultrasonoscopy based on deformable component model (DPM); Determine subject area from described at least one candidate region detecting; Profile by the subject area separate object based on definite carrys out detected object.

According to above-mentioned exemplary embodiment, can, in the situation that not needing artificial participation, effectively from ultrasonoscopy, detect the region that comprises tumour, and isolate relatively accurately on this basis the profile of tumour, using make a definite diagnosis as tumour according to one of.

Brief description of the drawings

By the detailed description of exemplary embodiment being carried out below in conjunction with accompanying drawing, above and other objects of the present invention and feature will become apparent, wherein:

Fig. 1 illustrates the block diagram of lesion detection system according to an exemplary embodiment of the present invention;

Fig. 2 illustrates the treatment scheme that is detected according to an exemplary embodiment of the present invention tumour by lesion detection system;

Fig. 3 illustrates the example of the tumour candidate region detecting according to an exemplary embodiment of the present;

Fig. 4 illustrates the example of definite according to an exemplary embodiment of the present tumor region;

Fig. 5 illustrates the example of the tumor's profiles separating according to an exemplary embodiment of the present;

Fig. 6 illustrates the block diagram of object detection systems according to an exemplary embodiment of the present invention; And

Fig. 7 illustrates the treatment scheme of being carried out according to an exemplary embodiment of the present invention detected object by object detection systems.

Embodiment

Now will describe embodiments of the invention in detail, the example of described embodiment is shown in the drawings, and wherein, identical label refers to identical parts all the time.Below will be by described embodiment is described with reference to accompanying drawing, to explain the present invention.

Fig. 1 illustrates the block diagram of lesion detection system according to an exemplary embodiment of the present invention.As shown in Figure 1, lesion detection system comprises according to an exemplary embodiment of the present invention: candidate region pick-up unit 10, for detect at least one candidate region that comprises tumour from ultrasonoscopy based on deformable component model (DPM); Tumor region locating device 20, for determining tumor region from described at least one candidate region detecting; Tumor's profiles tripping device 30, the profile that separates tumour for the tumor region by based on definite detects tumour.Here,, as example, ultrasonoscopy can the thoracic ultrasound result of pointer to tumor of breast, but should note: the present invention is not limited to this, and lesion detection system can be applicable to the lesion detection to other organ according to an exemplary embodiment of the present invention.

In above-mentioned lesion detection system, candidate region pick-up unit 10 utilizes deformable component model (DPM) method, can effectively detect the candidate region that may occur tumour, on this basis, further determine tumor region by tumor region locating device 20 again, and extract tumor's profiles by tumor's profiles tripping device 30, thereby can effectively complete the automatic detection to the tumour in ultrasonoscopy.

Below, in connection with Fig. 2, the example of carrying out according to an exemplary embodiment of the present lesion detection is described.

Fig. 2 illustrates the treatment scheme that is detected according to an exemplary embodiment of the present invention tumour by lesion detection system.

With reference to Fig. 2, at step S100, detected at least one candidate region that comprises tumour from ultrasonoscopy based on deformable component model (DPM) by candidate region pick-up unit 10.

Particularly, as is known to the person skilled in the art, in DPM method, there is root wave filter and a comparatively meticulous parts wave filter of multiple precision that precision is comparatively rough, correspondingly, have a root and multiple parts, wherein, p ₀instruction is intended to approximate for example covering, by the root of detected whole object (, the whole region at tumour place), and p ₁, p ₂..., p _ninstruction is intended to cover subtly each the different parts by detected object, and correspondingly, candidate region pick-up unit 10 can calculate by equation (1) the DPM detection score value of each candidate region:

$\mathrm{score}(p_0,p_1,...,p_n)=\underset{i=0}{\overset{n}{\mathrm{\Σ}}}{F}_i\left(p_i\right)\·G_i\left(p_i\right)-\underset{i=1}{\overset{n}{\mathrm{\Σ}}}\mathrm{score}({\mathrm{dx}}_i,{\mathrm{dy}}_i)+b---\left(1\right)$

Can find out from equation (1), the DPM of candidate region detects score value score (p ₀, p ₁..., p _n) can mainly be represented as root wave filter/parts wave filter at the outward appearance mark of position separately (wherein, F _i(p _i) be the filter response of root and all parts, G _i(p _i) be the feature of root and all parts) and deformatter (for representing that all parts departs from the position deviation cost of its anchor station) (wherein, x _iand y _ithe location of pixels of instruction all parts) between difference, on this basis, can be by by real-valued offset term b(wherein, b can come to determine by experiment) add described difference to and obtain final DPM detection score value.

As optimal way, can improve above-mentioned DPM method, more to meet the growth characteristics of tumour itself.That is to say, due to tumour itself come in every shape and aspect ratio change larger, and in DPM with the corresponding root template number of root wave filter less (being generally 2 or 3), therefore, be easy to can't detect the tumour of different the ratio of width to height.For above-mentioned situation, candidate region pick-up unit 10 can detect at least one candidate region that comprises tumour by the ratio of width to height that changes the root template in DMP from ultrasonoscopy according to an exemplary embodiment of the present invention.

Particularly, for root p ₀calculate F ₀(p ₀) and G ₀(p ₀) dot product time, can described dot product result be defined as to the maximal value in specific range of deflection according to equation (2):

$F_0\left(p_0\right)\·G_0\left(p_0\right)=\underset{p_0-{p^{\′}}_0\≤\mathrm{\τ}}{\max }(F_0\left(p_0\right)\·G_0\left(p_0^{\′}\right))---\left(2\right)$

Can find out from equation (2), by by F ₀(p ₀) and G ₀(p ₀) dot product be defined as within the scope of predetermined migration (wherein, τ indicates range of deflection, can determine the concrete value of τ for different application according to experiment) maximum dot product value, make the root template in DMP no longer be limited to fixed value, and can in preset range, carry out adjustment to a certain degree, thereby contribute to effectively to detect for the tumour of various the ratio of width to height.

As optimal way, after mode is obtained each candidate region that DMP detection score value is higher as described above, can further screen the candidate region obtaining, to obtain more reliable testing result, for example, the DPM that can remove the particular candidate region overlapping less than 50% higher with DPM detection score value detects the candidate region that score value is lower, thereby realizes the Effective selection to candidate region.

Fig. 3 illustrates the example of the tumour candidate region detecting according to an exemplary embodiment of the present.As shown in Figure 3, with respect to the region at actual tumour and place thereof, some tumours candidate region can be detected, wherein, each rectangular area is corresponding to the candidate region detecting in step S100.

Then,, at step S200, determine tumor region by tumor region locating device 20 from least one candidate region detecting.

Particularly, tumor region locating device 20 can use the two-value sorter based on support vector machine (SVM) to determine tumor region from least one candidate region detecting.

Can adopt various suitable modes to obtain the proper vector for described two-value sorter, for example, can extract described proper vector from position and the size of root template/component model of exporting corresponding to root wave filter/parts wave filter.

In addition, in order further to improve the accuracy of tumor region location, the contextual feature that the proper vector of the described two-value sorter based on support vector machine (SVM) can be based on candidate region.For example, following equation (3) can represent the proper vector f of described two-value sorter:

f=(s,r,offset _i,I,S _MAX,r _MAX) （3）

It is following that above-mentioned proper vector f relates to: the DPM of candidate region detects score value s, represent the position of candidate region and the vectorial r of size, in candidate region all parts with respect to the side-play amount offset of root _i, intensity difference I, candidate region and DPM in candidate region between prospect and background detect the part (S that coexists between the candidate region that score value is the highest _mAX, r _mAX).By above-mentioned according to an exemplary embodiment of the present invention proper vector f, can utilize the contextual feature of candidate region to classify, thereby the position of not only having considered candidate region is with size, has also considered the perienchyma at place, candidate region.In addition,, owing to conventionally can there is not more than three tumours in same width ultrasonoscopy, therefore, consider that current candidate region and DPM detect coexisting between the candidate region that score value is the highest and partly classify and contribute to further to strengthen the accuracy of classification.

But, it will be understood by those skilled in the art that: proper vector f is not limited to listed above every, for example, proper vector f can only comprise at least one or more in every listed above, and needn't comprise whole that in equation (3), list.In addition, the associated vector of the contextual feature of any embodiment candidate region all can be applicable to proper vector f.

Based on as above definite proper vector f, tumor region locating device 20 can use Multiple Kernel Learning (MKL) method the kernel function of two-value sorter to be defined as to the linear combination of multiple basic kernel functions.Particularly, in Multiple Kernel Learning MKL method, can learn the parameter of described multiple basic kernel functions and the weight of each basic kernel function by training data.For example, tumor region locating device 20 can be by the RBF(radial basis of three kinds of bandwidth) the polynomial kernel function of kernel function and three dimensions carries out linear combination, thereby train for each characteristic component in proper vector f and proper vector f entirety, to obtain the kernel function of two-value sorter.Here can determine according to experiment the concrete bandwidth of RBF kernel function.

After having determined the proper vector and kernel function of two-value sorter, tumor region locating device 20 can use corresponding two-value sorter to determine tumor region from least one candidate region detecting,, in candidate region, there is the candidate region of tumour maximum probability.

Fig. 4 illustrates the example of definite according to an exemplary embodiment of the present tumor region.As shown in Figure 4, the irregular figure being included in upper left rectangular area is indicated actual tumour, and bottom-right rectangular area is that tumor region locating device 20 is at the definite tumor region of step S200.

Then,, at step S300, the profile that separates tumour by the tumor region based on definite by tumor's profiles tripping device 30 detects tumour.

Particularly, by adopting Level Set Method, the contour curve of tumour can be expressed as to the zero level collection of more high-dimensional function (being called level set function).Here, can make level set function develop according to its satisfied evolution equation or iteration (wherein, to carry out iteration on the border of the definite tumor region of step S200 as initial curve), because level set function constantly develops, so corresponding zero level collection is also in continuous variation, in the time that level set movements tends to be steady, evolution stops, and obtains the final contour curve of tumour.

According to exemplary embodiment of the present invention, can be by making the represented energy function of equation (4) minimize to obtain the contour curve of tumour:

$F(c_1,c_2,C)={\mathrm{\λ}}_1\underset{\mathrm{inside}\left(C\right)}{\∫}{|u_0(X,t)-c_1|}^2\mathrm{dXdt}+{\mathrm{\λ}}_2\underset{\mathrm{outside}\left(C\right)}{\∫}{|u_0(X,t)-c_2|}^2\mathrm{dXdt}+\mathrm{\μ}\·\mathrm{Length}\left(C\right)+v\·\mathrm{Area}\left(\mathrm{inside}\left(C\right)\right)---\left(4\right)$

In equation (4), the contour curve of C instruction tumour, wherein, the contour curve C of tumour carries out iteration, c using the border of tumor region as initial curve ₁indicate the mean intensity of the pixel of the interior zone of the contour curve C of tumour in the t time iteration, c ₂indicate the mean intensity of the pixel of the perimeter of the contour curve C of tumour in the t time iteration, the interior zone of the contour curve C of inside (C) instruction tumour, the perimeter of the contour curve C of outside (C) instruction tumour, u ₀(X, t) indicate the intensity that is positioned at the pixel of X position in the t time iteration, the length of the contour curve C of Length (C) instruction tumour, the area of the interior zone of the contour curve C of Area (inside (C)) instruction tumour, in addition λ, ₁, λ ₂, μ, v be more than or equal to 0 preset parameter.Can find out from equation (4), the energy function adopting in Level Set Method can be expressed as following several and: the pixel intensity otherness in the interior zone of the contour curve C of tumour pixel intensity otherness in the perimeter of the contour curve C of tumour the area factor vArea (inside (C)) of the interior zone of the length factor μ Length (C) of the contour curve C of tumour, the contour curve C of tumour.

In above-mentioned equation (4), energy function F (c ₁, c ₂, C) mainly reflect that the contour curve that carries out iteration is in inside and outside global image statistics and characteristic.As example, can be by parameter lambda ₁, λ ₂be set to respectively 1, and parameter v is set to 0.

On this basis, in order further to improve the accuracy that separates tumour, for the feature of tumor image, for example, often causes the intensity of image inhomogeneous because of calcification in the middle of tumour, tumor's profiles tripping device 30 can be preferably by making the distance between prospect and the background of tumor image maximize to build the energy function adopting in Level Set Method.

As example, tumor's profiles tripping device 30 can by the basis of the energy function at equation (4), add according to Ba Ta just in the regular terms of refined distance (Bhattacharyya distance) distance between prospect and the background of tumor image is maximized, the corresponding energy function obtaining is optimised for as shown in following equation (5):

E(C)=βF(c ₁,c ₂,C)+(1-β)B(C) （5）

In equation (5), represent the probability density function p of contour curve C _in(X) with probability density function p _out(X) Ba Ta between just in refined distance, wherein, p _in(X) represent to be positioned at the probability density function that the pixel of X position (that is, is positioned at the prospect of tumor image) in contour curve C inside, p _out(X) represent to be positioned at the pixel of X position at the probability density function of contour curve C outside (, being positioned at the background of tumor image).In addition, β ∈ [0,1] for control Ba Ta just in the degree of participation of refined B (C).

On this basis, in order to meet the attribute of symbolic distance function phi and avoid reinitializing, can consider increases additional penalty R to the energy function of equation (5) _p(φ), thus obtain the energy function as shown in equation (6):

E(C)=βF(c ₁,c ₂,C)+(1-β)B(C)+αR _p(φ) （6）

In equation (6), α >0, and, penalty wherein, Ω represents entire image.

As optimal way, can within the scope of arrowband, carry out above-mentioned processing, to improve the speed of iterative processing.By the way, tumor's profiles tripping device 30 separable go out the profile of tumour, using the tumour as detecting.

Fig. 5 illustrates the example of the tumor's profiles separating according to an exemplary embodiment of the present.As shown in Figure 5, at the isolated tumor's profiles of step S300, compared with actual tumour, degree of approximation is higher.

It should be noted that, except the detection system shown in Fig. 1, the method shown in Fig. 2 also can realize by computer programming, the proprietary processor that also can form via specific hardware is carried out.Those skilled in the art can adopt any be known in the art maybe can with means carry out the method shown in execution graph 2.

For example, by (designing special experiment, for from 1941 patients (wherein, 913 benign tumour patients, 1028 malignant tumor patients) 2758 width ultrasonoscopys), can find out: compare with tumour separate mode with existing tumor region locator meams, tumor region locator meams not only obviously increases for the hit rate of tumour according to an exemplary embodiment of the present invention, and omit or locate wrong situation and significantly reduce, and tumour separation method can obtain than the better accuracy rate of existing mode according to an exemplary embodiment of the present invention.

As can be seen here, according to the system of automatic detection tumour described above, can, in the situation that not needing artificial participation, effectively from ultrasonoscopy, detect the region that comprises tumour, and isolate relatively accurately on this basis the profile of tumour, using make a definite diagnosis as tumour according to one of.

For lesion detection, automatic checkout system and method are according to an exemplary embodiment of the present invention described above, but, should understand, above-mentioned exemplary embodiment also can be applicable to other object detection in ultrasonoscopy, particularly, ultrasonoscopy can be applicable to numerous areas as a kind of nondestructiving detecting means, as, can be used for detecting machinery, metallurgical, as the industry such as empty space flight, railway, coal, non-ferrous metal, building foundry goods, forging, sheet material, compound substance, tubing, bar, section bar, weldment, machined piece and use in above-mentioned workpiece sensing.Therefore, tumour automatic checkout system described above and method can be applied to for other object detection in ultrasonoscopy, and will not be restricted to tumour by detected object.

Describe object detection systems according to an exemplary embodiment of the present invention and carried out the treatment scheme of detected object by object detection systems hereinafter with reference to Fig. 6 and Fig. 7.

Fig. 6 illustrates the block diagram of object detection systems according to an exemplary embodiment of the present invention.As shown in Figure 6, object detection systems comprises according to an exemplary embodiment of the present invention: candidate region pick-up unit 101, for detect at least one candidate region that comprises object from ultrasonoscopy based on deformable component model (DPM); Subject area locating device 201, for determining subject area from described at least one candidate region detecting; Object outline tripping device 301, carrys out detected object for the profile of the subject area separate object by based on definite.

Fig. 7 illustrates the treatment scheme of being carried out according to an exemplary embodiment of the present invention detected object by object detection systems.

With reference to Fig. 7, at step S101, detected at least one candidate region that comprises object from ultrasonoscopy based on deformable component model (DPM) by candidate region pick-up unit 101.

Then,, at operation S201, determine subject area by subject area locating device 201 from least one candidate region detecting.

Finally, at operation S301, carry out detected object by object outline tripping device 301 by the profile of the subject area separate object based on definite.

It should be noted that each device in the detection system shown in Fig. 6 is in the time of the operating process shown in execution graph 7, the similar fashion shown in employing and Fig. 1 and Fig. 2, difference is only the parameters about tumour and processes the parameter and the processing that change to for any object.

Although specifically shown with reference to its exemplary embodiment and described the present invention, but it should be appreciated by those skilled in the art, in the case of not departing from the spirit and scope of the present invention that claim limits, can carry out the various changes in form and details to it.

Claims (15)

1. for automatically detect a system for tumour at ultrasonoscopy, comprising:

Candidate region pick-up unit, for detecting at least one candidate region that comprises tumour from ultrasonoscopy based on deformable component model (DPM);

Tumor region locating device, for determining tumor region from described at least one candidate region detecting;

Tumor's profiles tripping device, the profile that separates tumour for the tumor region by based on definite detects tumour.

2. the system as claimed in claim 1, wherein, candidate region pick-up unit detects by the ratio of width to height that changes the root template in DMP at least one candidate region that comprises tumour from ultrasonoscopy.

3. the system as claimed in claim 1, wherein, tumor region locating device uses the two-value sorter based on support vector machine (SVM) to determine tumor region from described at least one candidate region detecting.

4. system as claimed in claim 3, wherein, the contextual feature of the proper vector of described two-value sorter based on candidate region.

5. system as claimed in claim 4, wherein, described proper vector comprises at least one in following: the DPM of candidate region detects all parts in the position of score value, candidate region and size, candidate region with respect to the part that coexists between the intensity difference between prospect and background in the side-play amount of root, candidate region, candidate region and the highest candidate region of DPM detection score value.

6. system as claimed in claim 5, wherein, tumor region locating device uses Multiple Kernel Learning (MKL) method the kernel function of two-value sorter to be defined as to the linear combination of multiple basic kernel functions.

7. system as claimed in claim 6, wherein, the polynomial kernel function of the RBF kernel function of three kinds of bandwidth and three dimensions is carried out linear combination by tumor region locating device, thereby train for each characteristic component in proper vector and proper vector entirety, to obtain the kernel function of two-value sorter.

8. the system as claimed in claim 1, wherein, tumor's profiles tripping device usage level diversity method separates the profile of tumour, and wherein, the contour curve of tumour carries out iteration using the border of tumor region as initial curve.

9. system as claimed in claim 8, wherein, tumor's profiles tripping device is by making the distance between prospect and the background of tumor image maximize to build the energy function adopting in Level Set Method.

10. for the system at the automatic detected object of ultrasonoscopy, comprising:

Candidate region pick-up unit, for detecting at least one candidate region that comprises object from ultrasonoscopy based on deformable component model (DPM);

Subject area locating device, for determining subject area from described at least one candidate region detecting;

Object outline tripping device, carrys out detected object for the profile of the subject area separate object by based on definite.

11. systems as claimed in claim 10, wherein, candidate region pick-up unit detects by the ratio of width to height that changes the root template in DMP at least one candidate region that comprises object from ultrasonoscopy.

12. systems as claimed in claim 10, wherein, subject area locating device uses the two-value sorter based on support vector machine (SVM) to determine subject area from described at least one candidate region detecting.

13. systems as claimed in claim 12, wherein, the contextual feature of the proper vector of described two-value sorter based on candidate region.

14. systems as claimed in claim 13, wherein, subject area locating device uses Multiple Kernel Learning (MKL) method the kernel function of two-value sorter to be defined as to the linear combination of multiple basic kernel functions.

15. 1 kinds in the automatically method of detected object of ultrasonoscopy, comprising:

Detect at least one candidate region that comprises object from ultrasonoscopy based on deformable component model (DPM);

Determine subject area from described at least one candidate region detecting;

Profile by the subject area separate object based on definite carrys out detected object.