CN104182984B - Method and system for rapidly and automatically collecting blood vessel edge forms in dynamic ultrasonic image - Google Patents

Abstract

The invention discloses a method and system for rapidly and automatically collecting blood vessel edge forms in a dynamic ultrasonic image and belongs to the field of medical electronic information. The method mainly includes the steps of extracting blood vessel edge vector data in the ultrasonic image of a line scanning mode and calculating and correcting blood vessel inner diameters. According to the processing process of the method, complex arithmetical operations such as image preprocessing, edge detecting and vectorizing in traditional edge detecting are converted into a simple detecting method operated based on numerical value addition and subtraction methods. Compared with a traditional method, the processing speed of the method and system for rapidly and automatically collecting the blood vessel edge forms in the dynamic ultrasonic image can be improved by multiple times even ten or more times, and the real-time detecting and processing requirement can be met in dynamic blood vessel ultrasonic detecting of 20 or more frames per second. The data quality of processing results meets the requirement for later-stage analyzing and diagnosis references through finite experiments.

Description

The fast automatic acquisition method of vessel boundary form in dynamic ultrasound image and system

Technical field

The invention belongs to medical electronic message area, it is related to Edge extraction technology in Digital Image Processing, specially one Plant the fast automatic acquisition method of vessel boundary form in dynamic ultrasound image and system.

Background technology

Cardiovascular disease is one of Health Killer of this century human maximum, and it is the most frequently used at present that ultrasonic non-destructive triage is surveyed Reliable means.With people's quality of life and improving it is desirable to check the patient of cardiovascular status to own health attention rate Quantity increases considerably so that some key indexs of blood vessel image adopt this inefficient operation of hand dipping more and more heavier.

List of references " exploitation of intelligent-tracking blood vessel image analysis system and preliminary operational research (development and Application of auto-tracing vessel image analysis system), the Academic Journal of Kunming Medical College 2010, (12): 24 28), be the system using conventional art exploitation for inventor's early stage of present patent application, with the edge of general pattern Detection is the same with automatic analysis method, and what each edge detection operator will reach go out figure effect it is necessary to through the following steps:

A, filtering: rim detection is based primarily upon derivative calculations, but affected by noise, and wave filter is while reducing noise Lead to the loss of edge strength.

B, enhancing: strengthen algorithm and the point that gray scale in neighborhood has significant change is highlighted, typically pass through to calculate gradient width Value completes.

C, detection: but in some images gradient magnitude larger be not marginal point, simplest rim detection be ladder Degree amplitude thresholds judge.

D, vector quantization: actual measurement can be used for and calculate.

Wherein topmost link rim detection we adopt robe outward flange operator.Roberts edge detection operator is A kind of utilization local difference operator finds the operator at edge, and rudimentary algorithm formula is:

$g(x,y)={\{{\[\sqrt{f(x,y)}-\sqrt{f(x+1,y)}\]}^2+{\[\sqrt{f(x+1,y)}-\sqrt{f(x,y+1)}\]}^2\}}^{\frac{1}{2}}$

Wherein f (x, y) is the input picture with integer pixel coordinate, and square root calculation makes this process regard similar to the mankind The process occurring in feel system.

In image enhaucament, the first step is histogram equalization processing.For the image of m × n dot matrix, use following formula meter Calculate:

$n=t\left(m\right)=\frac{n_k-n_0}{mn}\underset{i=m_0}{\overset{m}{\σ}}i\left(s\right)\mathrm{\δ}s+n_0$

The grey level range of wherein input picture is [m₀, m_k]；The grey level range of output image is [n₀, n_k].

The relatively simple median filtering method of image filtering, the image to a=m × n:

y_{I, j}=averge [x_{I+k, j+l,}(k, l) ∈ a]

By realizing above-mentioned steps, it has been accurately detected vessel boundary relevant parameter in an experiment.Constantly survey in the later stage Problem is encountered that the inefficient of system in examination and application, per second can only process 3～6 two field pictures.Higher for prescription In the case of the real-time processing of collection per second 20～30 frames, complicated Processing Algorithm encounters difficulty.

The algorithm of Image Edge-Detection has roberts, sobel, prewitt, canny, marr etc., and wherein roberts is imitated Really minimum, marr is best.Canny operator image clearly, but very sensitive to noise, and marr can effective detection edge, and energy Suppression influence of noise, is often considered a kind of more satisfactory edge detection operator well.

Marr-hildreth operator is to realize on the basis of Laplace operator, and it has benefited from the vision machine to people The research of reason, has certain biology and physiologic meaning.Because Laplace operator is more sensitive to noise ratio, make an uproar to reduce Sound shadow is rung, and first figure to be detected can be carried out smooth and then detect edge with Laplace operator again.The representational computational chart of marr Reaching formula is:

$\begin{array}{c}{\▿}^{2}g={\▿}^2\[g(x,y)=h(x,y)*f(x,y)\]=\left(\frac{r^2-{\mathrm{\σ}}^2}{{\mathrm{\σ}}^2}\right)\exp (-\frac{r^2}{2{\mathrm{\σ}}^2})*f(x,y)\\ ={\▿}^{2}h*f(x,y)\end{array}$

In formula, σ is variance.

Image border is typically all by carrying out what gradient algorithm was realized to image.Image gradient can be regarded image as Two-dimensional discrete function, image gradient is the derivation of this two-dimensional discrete function.Image gradient:

G (x, y)=dx_i+dy_j

Dx (i, j)=v (i+1, j)-v (i, j)

Dy (i, j)=v (i, j+1)-v (i, j)

Wherein, v is the value of image pixel, and (i, j) is the coordinate of pixel.

Image gradient can also use intermediate value difference sometimes:

Dx (i, j)=[v (i+1, j)-v (i, j)]/2

Dy (i, j)=[v (i, j)+1)-v (i, j)]/2

The size of gradient represents the intensity at edge, and gradient direction moves towards vertical with edge.For detected edge points, choose suitably Thresholding t, binaryzation is carried out to gradient image, then has as g (x, y) >=t, make g (x, y)=1, then for step-like marginal point. Otherwise g (x, y)=0.So form width edge binary images.Gradient operator only uses the gray count of nearest neighbor pixels.

About the limb recognition technology of digital picture, ultrasonic blood vessel image, many articles are delivered, and also much disclose Patent application and authorize patent of invention.Wherein there is a kind of digital picture side that Chinese patent notification number is cn100458847c Edge detection method, Chinese patent notification number are the rim detection in the ultrasonoscopy of cn101357067b, Chinese Patent Application No. Ultrasonic blood vessel border detection System and method for for 201210584625.8, Chinese Patent Application No. are 201210584625.8 It is cn based on the method for acquisition initial profile, Chinese patent notification number in the Ultrasound Image Segmentation of active contour model The use ultrasonoscopy of 101527047b detects method and apparatus of organizational boundary etc..

The existing article delivered and the patent document announced nearly all have the data operation of complexity and substantial amounts of processing procedure, Give prominence to the key points in the effect realized, and we actually encounter is efficiency when great amount of images is processed, this is also one and can not neglect Depending on problem, especially in the stronger applied environment of real-time.

Content of the invention

The purpose of the present invention and technical problem to be solved: provide the vessel boundary form in a kind of dynamic ultrasound image fast Fast automatic acquiring method and system, the required key property possessing is that have compared with high practicability, real-time, reliability.

The fast automatic acquisition method of vessel boundary form in dynamic ultrasound image of the present invention at least comprises the following steps:

(1) open dicom3.0 form or the image of avi form, or received from Ultrasound Instrument by dicom3.0 host-host protocol The image sending on device, draws a wire tag and analyzes and processes scope；

(2) line sweep limitss and direction controlling Vector operation are entered by following equation and method:

$hx=\frac{x_{pt20}-x_{pt10}}{\sqrt{{(x_{pt20}-x_{pt10})}^2+{(y_{pt20}-y_{pt10})}^2}},$

$hy=\frac{y_{pt20}-y_{pt10}}{\sqrt{{(x_{pt20}-x_{pt10})}^2+{(y_{pt20}-y_{pt10})}^2}},$

V_vx=v_hy, v_vy=v_hx；V_wx=-v_hy, v_wy=v_hx,

$scan\_width=\sqrt{{(pt20.x-pt10.x)}^2-{(pt20.y-pt10.y)}^2},$

Wherein pt10 is scan start point, and pt20 is terminating point, and pt10.x, pt20.x, pt10.y, pt20.y are respectively a little The x direction of pt10 and pt20 and the component in y direction；Scan_width represents sweep length, hx and hy is respectively x and y direction Unit vector mould, v_hx and v_hy is the unit vector in x and y direction, v_vx and v_vy is the list in the x and y direction of vector s431 Bit vector, v_wx and v_wy is the unit vector in the x and y direction of vector s432.

Wherein direction controlling parameter, when scanning up be:

N1x=x+v_vx,

N1y=y+v_vy,

N2x=n1x+v_vx,

N2y=n1y+v_vy；

When scanning downwards it is:

N1x=x+v_wx,

N1y=y+v_wy,

N2x=n1x+v_wx,

N2y=n1y+v-wy；

Wherein (x, y) processes location of pixels coordinate for present image, the next pixel coordinate in (n1x, n1y) scanning direction, (n2x, n2y) is next one pixel coordinate of scanning direction.

(3) carry out image gradient calculating, first derivation or second order derivation to process:

Dv [i, j]=v [n1x, n1y]-v [x, y],

$dv{\[i,j\]}^{\′}=\frac{(v\[n2x,n2y\]+v\[n1x,n1y\])}{2}-v\[x,y\],$

$dv{\[i,j\]}^{\′\′}=\frac{1}{3}\sqrt{v{\[n2x,n2y\]}^2-v{\[x,y\]}^2}+\frac{2}{3}\sqrt{v{\[n1x,n1y\]}^2-v{\[x,y\]}^2};$

Wherein v is image pixel grey decision-making, and scope is 0～255.Range of operation i controls for base line, is worth for [0.. scans Terminate], j is the scope [0..scan_width-1] of a line；Y initial value is pt20.y, each base line corresponding, and x initial value is pt20.x；

During line vessel scanning limb recognition, sweep limitss will exceed the ρ pixel at the blood vessel external mold edge of farthest, ρ =10～50, determination methods are the maximums looked in sweep limitss, and processing method is continuous rising and then keeps, rises, more again Keep；First rising edge is determined as vessel boundary, finds bigger graded, just with new discovery in sweep limitss again Edge be defined；

(4) find the upper and lower vessel boundary point in sweep limitss by above-mentioned steps, then to the noise in ultrasonoscopy Processed etc. the edge erroneous judgement causing: set using before and after vessel boundary direction each σ point as reference, σ=3-10；If setting When the continuous difference in vessel boundary direction is more than τ, τ=10～20, be judged to error dot, by discarding, a small amount of error dot abandons Afterwards, front and rear edges point is directly connected to；

To line scanning collection to upper and lower marginal point carry out pair relationhip correction, the Data Position of blood vessel upper wall collection keeps Constant, i=[0..scan_width-1], each upper wall edge point looks for lower wall edge match point, and pair principle is with a top edge Point is with reference to pairing basic point, and the paired point of lower limb is nearest point in all positions of lower limb and with reference to pairing basic point, away from From comparing formula it isFor i-th upper wall edge point, For Searching point before and after i-th lower wall marginal point；Optional formula is:j Span is about line segment l704/5, and the condition that another terminates the minimum distance point search in a direction is to set edge ε Point distance is all big than a upper point when comparing, ε=3～5；

The vessel boundary position that the form specified by table 1 below and table 2, output collect from each frame of dynamic ultrasonic image Data and each position cross section line number evidence, are shown with image synthesis on a computer display simultaneously, and show maximum Internal diameter, minimum diameter and dynamic changing curve；

Table 1: the master data sheet format of result output

Table 2: the detailed content form definition of result output data

The technical characteristic that the inventive method optimizes further is: ρ=30.σ=3, τ=10.ε=3.

The fast automatic acquisition system of vessel boundary form in dynamic ultrasound image of the present invention includes:

Dicom and avi image reading module is pressed step s10 and is executed, and completes wherein relevant treatment；Processing and control module is used for The operation of modules in management and coordination system, controls human-computer interaction module, image-receptive or loading, and vessel boundary shape The execution of state fast automatic collection whole flow process；Image shows module is used for showing ultrasonoscopy, controls dynamic play, and display is handed over Interoperability state and control labelling, display line scans the result of marginal information collection；Man-machine interactive operation module is used for accepting Mouse action event, generates and display target scope identifier figure, starts, stops vessel boundary data acquisition, preserves data；Line is swept Retouch MARG extraction module execution line scanning imaging system, extract vessel boundary data, detection abnormal data simultaneously abandons；Internal diameter corrects After module is used for the line scanning acquisition upper and lower MARG of blood vessel, pair correlation is carried out to upper and lower wall marginal position point, is allowed to reach To each pairing the distance between the shortest；Data storage and interface module are used for for above-mentioned detection process result data being saved in magnetic In disk file, use for further data analysis system, also support that shared drive mode provides other data analysiss systems simultaneously System directly uses the result of present system.

Beneficial effects of the present invention: eliminate the preprocessing process such as filtering in traditional images rim detection, enhancing, and The complicated Mathematical Calculations such as rim detection later stage dot array data vector quantization, are imaged back edge according to tissue and have edge Property feature it is only necessary to the data having local edge to line scanning result filters, exclude possible outlier, with less Computing reach same purpose.With resolution for 1000 × 1000, as a example the dynamic ultrasound image of 100 frames, each point is carried out Addition and subtraction is exactly 100,000,000 computings, and time-consuming bigger than simple signed magnitude arithmetic(al) one of complicated mathematical operation is arrived several quantity Level, the present invention has obvious odds for effectiveness on dynamic ultrasound view synthesis.

Brief description

Fig. 1 is the vessel boundary form fast automatic collection basic flow sheet in embodiment dynamic ultrasound image, is also this The main process chart of provided system is provided；

Fig. 2 is that embodiment processes target zone identification method and labelling schematic diagram, mouse pressing and holding at pt10, Move to pt20 release, system automatically generates labelling shown in Fig. 2；

In Fig. 3, the left side is the artwork of embodiment actual dynamic ultrasound image the 1st frame, extracts MARG, synthesis display effect Fruit is schemed；In Fig. 3 the right with Fig. 3 left side be the same dynamic image data of embodiment, the artwork of the 30th frame, extract MARG, Synthesis display renderings；In Fig. 3, the change that can see vessel position form is compared on the right with the left side；

Fig. 4 is embodiment line vessel scanning edge extracting direction vector schematic diagram；

Fig. 5 top is embodiment x wire scan process schematic diagram, and bottom is that embodiment longitudinally show by upper and lower line scan process It is intended to；

Fig. 6 simplifies process schematic diagram further for x wire scanning；

Fig. 7 calculates (inwall both sides data pair) for embodiment vessel diameter and corrects schematic diagram, show also line scanning The difference processing between line segment and actual vessel internal diameter obtaining after limb recognition；

Fig. 8 is that the vessel boundary form fast automatic acquisition system basic structure in embodiment dynamic ultrasound image is illustrated Figure.

Specific embodiment

Method to the fast automatic collection of vessel boundary form in the dynamic ultrasound image of the present invention below in conjunction with the accompanying drawings And the enforcement technical scheme of system carries out detailed, complete elaboration.

As shown in figure 1, the fast automatic acquisition method of vessel boundary form in dynamic ultrasound image comprises the following steps:

Step s10, loads the dynamic multiple image of Vascular Ultrasonography, obtains the total frame of image from the file of dicom or avi form Number imageframe_n, reads each two field picture.Image handled by the inventive method only uses luma data, and therefore image can With discoloration, retain GTG.If each pixel digit bits allocated > 8 of dicom image, bit stored > 8, The grey decision-making of each image slices vegetarian refreshments is gone to 8,256 grades of GTGs.Being more than 256 grades of GTGs from treatment effect cannot be more preferably Effect, on the contrary using internal memory using being doubled and redoubled.

It is alternatively possible to gather image from ultrasonic instrument, directly in the way of dicom host-host protocol, sent by network And reception, subsequently into subsequent processes, obtain higher efficiency.

Initialization frame under process currentframe_i=0.

Step s20 draws a reference straight line in angiosomeses to be analyzed near middle position, as shown in Figure 2.Operation side Method is that mouse is pressing and holding at pt10, moves to pt20 release, system automatically generates labelling shown in Fig. 2.Point pt10 is , with reference to original position central point, point pt20 is vessel segment reference end position central point to be analyzed for vessel segment to be analyzed, 2 points it Between line segment be vessel segment to be analyzed.Manual operations, must not the selection of centering position be accurate, the mistake about within 30% Difference does not affect result.Because the multiformity of blood vessel, in many cases, only pt10 to be put and point pt20 does not choose outside blood vessel Portion, line segment is not all effective operation in extravascular.

Step s30 is to all picture frame execution step s40～s70.

The direction that step s40 is pressed with reference to straight line executes line scanning imaging system, obtains the data near lower edges data, and Edge approximate center position.With the image continuity features of scanning direction and vertical direction, outlier is carried out to data acquired Filter, exclusion is because of the improper MARG of noise or the generation of other factorses image.

The vascular cross-section diameter correction process of step s50.Normal direction due to scanning direction vector and vessel boundary curve tangent line Vector seldom identical it is therefore desirable to be corrected, as shown in Figure 7.Need in vessels analysis to know the maximum of each position or The data such as minimum blood flow handling capacity, vessel wall thickness, elasticity, a point to blood vessel wall side, need to find opposite side Corresponding point, basic rule is that distance is the shortest between the two.

Step s60 blood vessel wall image featureization is processed, and vessel boundary includes inner membrance, middle film, adventitia, its image has difference Not, but inconspicuous.The maximum inner diameter of calculating blood vessel, minimum diameter, wide-to-narrow ratio, border movement curve, and it is shown in system interface On.Alternatively, add vessel boundary data enhancement process, be easy to output data post analysis process when be easily found inner membrance, Middle film, the border of adventitia.

Step s70 stores the i-th two field picture result data, by Tables 1 and 2 form, preserves and processes output data, this number Process software according to being supplied to the later stage, for showing, analyzing, count.If also picture frame is untreated, go to step s30.

Step s80 initializes to system after completing once analyzing and processing process, prepares to process next time or extract to work as The data of secondary analysis.

Through the process of above step, such as Fig. 3 illustrates the embodiment of the present invention to a dynamic ultrasound image actual treatment As a result, wherein Fig. 3 left side is dynamic ultrasound image the 1st frame, and s310 shows for artwork and pending blood vessel segment limit selects by hand Select labelling, s311 represents the vessel boundary data extracted, s312 represents that testing result data synthesizes display renderings with artwork；

In Fig. 3 the right with the left side be same dynamic image data, present position be the 30th frame, s301 be artwork displaying and Pending blood vessel segment limit craft selected marker, s301 represents the vessel boundary data extracted, and s302 represents testing result number Synthesize display renderings according to artwork.

Process to a dynamic ultrasound image, pending blood vessel segment limit selects only to operate once by hand, dynamic image In each frame process and have a unified start reference line.

The left side in relatively Fig. 3 and the right are it can be seen that same blood vessel is in the position of different time (different cardiac cycle) All there occurs obvious change with form, i.e. the extension of paradoxical expansion arteries, relaxing period arterial vasoconstriction draining blood. Exactly apply this characteristic, by automatically extracting the data of vasomotion process, analyze vascular performance: elasticity, thickness etc., for diagnosis Reference is provided；Arterial blood output etc. can be calculated simultaneously.

As shown in figure 4, line scanogram edge feature data acquisition is an emphasis of the present invention, concrete operation step is such as Under:

1st, basic orientation control Vector operation:

As Fig. 4, vector s430v, after be designated asVector s431v, after be designated asVector s432v, after be designated asMeter Calculation method is:

$\stackrel{\→}{v}=\stackrel{\→}{hx}-\stackrel{\→}{hy}$

$\stackrel{\→}{w}=-\stackrel{\→}{hx}+\stackrel{\→}{hy}$

Unit vector is:

Wherein x direction unit vector:Note variable is v_hx,

Y direction unit vector:Note variable is v_hy

EquallyWithX and y direction unit vector remember that variable is respectively: (v_vx, v_vy), (v_wx, v_wy), wherein Simplification relation is: v_vx=v_hy, v_vy=v_hx；V_wx=-v_hy, v_wy=v_hx.

One dynamic image pro cess process direction vector only needs computing once.Because computer picture shows in actual operation During process, the upper left corner is (0,0), and vertical direction is on the occasion of downward, contrary with numercal custom.

2nd, line vessel scanning MARG collection

As shown in Fig. 5 top, along the line scan process embodiment of blood flow direction or inverse blood flow direction, as shown in the lower part of Figure 5, Along the line scan process embodiment in vascular cross-section direction, calculating with processing method is:

$scan\_width=\sqrt{{(pt20.x-pt10.x)}^2-{(pt20.y-pt10.y)}^2}$

Dv [i, j]=v [n1x, n1y]-v [x, y] (1)

$dv\[i,j\]=\frac{(v\[n2x,n2y\]+v\[n1x,n1y\])}{2}-v\[x,y\]---\left(2\right)$

$dv\[i,j\]=\frac{1}{3}\sqrt{v{\[n2x,n2y\]}^2-v{\[x,y\]}^2}+\frac{2}{3}\sqrt{v{\[n1x,n1y\]}^2-v{\[x,y\]}^2}---\left(3\right)$

Wherein direction controlling parameter, when scanning up:

$\begin{array}{c}n1x=x+v\_vx\\ n1y=y+v\_vy\\ n2x=n1x+v\_vx\\ n2x=n1y+v\_vy\end{array}---\left(4\right)$

When scanning downwards:

$\begin{array}{c}n1x=x+v\_wx\\ n1y=y+v\_wy\\ n2x=n1x+v\_wx\\ n2x=n1y+v\_wy\end{array}---\left(5\right)$

$a\[i\]=\frac{{\σ}_{i=0}^{scan\_width}dv\[i,j\]}{scan\_width}---\left(6\right)$

Wherein v is image pixel grey decision-making, and scope is 0～255.Pt10 is scan start point, and pt20 is terminating point, Pt10.x, pt20.x, pt10.y, pt20.y are respectively the x direction of point pt10 and pt20 and the component in y direction, and (x, y) is current Image procossing location of pixels coordinate, (n1x, n1y) scanning direction the next one pixel coordinate, (n2x, n2y) be scanning direction again under One pixel coordinate；Scan_width is the distance between pt10 to pt20；Range of operation i controls for base line, is worth for [0.. The end of scan], j is the scope [0..scan_width-1] of a line；Y initial value is pt20.y, each base line corresponding, and x is initial It is worth for pt20.x.

Vessel borders identification ultimate principle calculates for image gradient, first derivation or second order derivation, edge continuity judge Deng composition, this example demonstrates that based on the concrete execution step of computer programing.According to above-mentioned main executable expressions, under The concrete implementation procedure of face explanation:

Simplify most in the present embodiment and select execution formula (1), preferably can replace executing formula (2) or formula (3), program In execution, simple division is switched to subtraction execution.System to determine selection executive mode by configuring and arranging.According to limited The practical application test of amount, formula (2) is slightly better than formula (1), and formula (3) speed declines substantially, but effect promoting is inconspicuous.Few Run into formula (1) or formula (2) execution in the case of number when effect is not ideal enough, temporarily select formula (3).

As shown in figure 4, line vessel scanning MARG collection by scan up and downwards scan two parts, upwards for S431v direction, is downwards s432v direction, and i starts scanning until the end of scan from 0 row respectively.Direction controlling ginseng when scanning up Number selects formula (4), and when scanning downwards, direction controlling parameter selects formula (5), scans up and scans in this process downwards Middle other step is identical.

Formula (4) and formula (5) illustrate source and ultimate principle from the preceding paragraph falls, and transport vector during actual execution Calculation be converted to plus, subtraction to be executing.

Through above-mentioned calculating it can be determined that the vessel boundary of line scanning direction.Ultimate principle is the size representative edge of gradient The intensity of edge, gradient direction most preferably judges that direction is that edge trend is vertical.During embodiment of the present invention line scan process, the ladder obtaining Degree direction rarely reaches vertical relation with vessel boundary direction.Other algorithms are also the same, before vessel boundary does not find, gradient side To being unknown, can only attempt.As shown in Figure 4 and Figure 5, the angle at scanning direction s431v, s432v and edge can be at 45 degree Between～135 degree, reach and judge edge concerns mandate.

Generally edge starting point judges to need to select suitable threshold values t, and recognition methodss of the present invention are slightly different.Line of the present invention During vessel scanning limb recognition, sweep limitss will exceed the individual pixel of ρ (ρ=10～50) at the blood vessel external mold edge of farthest. Basic determination methods are the maximums looked in sweep limitss, and more specifically processing method is continuous rising and then keeps, goes up Rise, keep again；First rising edge is determined as vessel boundary.Certainly find bigger graded in sweep limitss again, Just it is defined by newfound edge.Basic basis for estimation is the feature of ultrasonoscopy medium vessels image: blood flow image is close, blood flow There is larger difference with vessel boundary, the image that internal model, middle mold, external mold and other are organized is close but has difference.

Present invention omits the preprocessing process such as basic filtering in the identification of image border, enhancing, and ultrasonoscopy is made an uproar Sound problem is prominent, and the different images producing under same acquisition parameter also has a lot of difference, therefore in single-point gradiometer Calculation mode nearly all has the situation of erroneous judgement.The present invention solves the situation of error with the continuous judgement at edge, in theory with seriality Equation is foundation, actual simplify be embodied as method be using before and after vessel boundary direction the individual point of each σ (σ=3-10) as reference, The continuous difference in vessel boundary direction is more than τ (such as τ=10), is judged to error dot.For error dot treating method be by Discarding, after a small amount of erroneous judgement point abandons, front and rear edges point is directly connected to.Always it is intended to during actual treatment abandon some points, backmost Edge curve is equally the curve continuously referring to.

The edge data of discretization can't be used for automatically calculating, it is desirable to have vector quantization process.Line scanning of the present invention When had the position vector of each data, general Sexual behavior mode is that MARG is carried out some curve matchings, obtains edge beautiful The vessel boundary curve seen；But in the case of not carrying out curve fitting, application demand can be reached, the therefore present invention is implemented Example does not include this step；When needing in practical application to show, can show after being fitted by the data that system exports.

As shown in table 1, vessel boundary detection master data comprises both sides of the edge coordinate points, withWithTo represent, wherein frame_i represents the i-th two field picture, n represents n-th group vessel boundary data.Secondly also wrap Include detailed data pointer, withDetailed data content and form are shown in Table 2, comprising: the lattice calibration will of 4 bytes “spim”；The length of 4 bytes, comprises the whole section of detailed data length identifying；The unused dtd--data type definition of 2 bytes, at present All deposited with integer numerical value form；On the vessel boundary of each two bytes, the length of segment data and lower segment data, short for no symbol Integer, both are worth identical, are typically worth for 32；32 byte reserved data positions；Vessel boundary data when subsequently.

Vessel boundary detailed data is start recording when gradient continuously rises is detected.

Point on blood vessel upper wall s412 (s701)With the point on blood vessel lower wall s411 (s702) It is beeline relation between 2 points, that is, both are mutually the beeline points that offside can find.But line scanning of the present invention The lower edges match point finding is frequently not both beelines, is not vessel diameter.

3rd, vessel boundary point pair correlation is processed

For example, shown in Fig. 7, when line scanning is found respectively, during i=x (certain value), upper wall edge point pt70 and lower wall side are found Edge point pt72；After upper wall edge matching, the tangent line in pt70 point is l703, and the line segment of its normal direction to lower wall edge is l705, Lower wall intersection point is pt71.Obviously pt71 and pt72 is not same point, and line segment l704 and l705 is unequal, therefore line scanning result number Need to correct according to pair relationhip.Specific implementation method is as follows:

The Data Position of blood vessel upper wall collection keeps constant, i=[0..scan_width-1], and each upper wall edge point looks for lower wall Edge match point., the step finding pt71 is taking pt70 as a example: start to search for the near distance spot respectively forwardly and backward from pt72, Simplifiedly distance compares formula and is For i-th upper wall edge point, For Searching point before and after i-th lower wall marginal point.Optional formula is:Real Can be achieved the goal from simplification formula when border compares to the distance of continuity point, and processing speed is fast.J span is about line Section l704/5, the condition that another terminates the minimum distance point search in a direction is all than upper when 3, edge point distance compares One point is big.

Another embodiment of the present invention, as shown in Fig. 4 and Fig. 5 top, line scanning direction is the direction of vector s430v, its Its step ibid embodiment.This embodiment line scan method is as follows: range of operation i is that base line controls, and is worth and is [0..scan_width-1], j is the scope [the 0.. end of scan] of a line；Y initial value is pt20.y, each base line corresponding, x Initial value is pt20.x.

After first a few row scanning discovery vessel boundaries, marginal feature is had based on vessel boundary, subsequently scanning only with Above some keeps scanning in continuous scope, reduces scanning amount of calculation further.As shown in fig. 6, s440 shows for sweep limitss It is intended to.

The present invention provides the fast automatic acquisition system of vessel boundary form in a kind of dynamic ultrasound image, shown in 8, bag Include: dicom and avi image reading module is executed by step s10, completes wherein relevant treatment；Processing and control module is used for managing With the operation of modules in coordination system, control human-computer interaction module, image-receptive or loading, and vessel boundary form is fast The execution of fast automatic data collection whole flow process；Image shows module is used for showing ultrasonoscopy, controls dynamic play, display interaction behaviour Make state and control labelling, display line scans the result of marginal information collection；Man-machine interactive operation module is used for accepting mouse Action Events, generate and show target zone mark figure as shown in Figure 2, start, stop vessel boundary data acquisition, preservation number According to；Line scanning MARG extraction module execution line scanning imaging system, extracts vessel boundary data, detection abnormal data simultaneously abandons； After internal diameter correction module is used for the line scanning acquisition upper and lower MARG of blood vessel, upper and lower wall marginal position point is carried out match school Just, make up to the distance between each pairing the shortest, as shown in Figure 7；Data storage is used for above-mentioned detection with interface module Result data is saved in disk file, uses for further data analysis system, also supports shared drive side simultaneously Formula provides other data analysis systems directly using the result of present system.

The specific implementation process of invention system: activation system, open the dynamic ultrasound blood vessel shadow of dicom3.0 or avi form As file, or view data is received by dicom agreement, by the operational approach selection target treatment region of step s20 explanation, open Dynamic detection start button, system is automatically held line scan, internal diameter correction etc. and is processed, and alternatively shows side over the display in real time Edge collection result, calculating maximum inner diameter, minimum diameter, wide-to-narrow ratio, border movement curve etc. are simultaneously shown on system interface, automatically Store gathered data in internal memory, and be saved in disk file, alternatively, to the data using the system gathered data Analysis software sends the sendmessage of windows, and the information such as transmission sharing memory pointer, after this sample system process terminates Start the analysis software of application-dependent data at once, show the data having diagnosis reference value.

Hereinafter we are applied the situation before and after the present invention to illustrate:

Vessel boundary form automatic data collection and the analysis system beginning one's study before 5 years in development behavior ultrasonic image, by biography The route of system: gray scale histogram equalization, medium filtering, canny edge detection algorithm, method of least square is to the blood vessel side extracted Edge carries out curve fitting process, finally gives vessel boundary data, enters the calculating analysis about blood vessel characteristic index.To 512 × 512 dynamic ultrasound image, on intel i7 high-performance computer, process per second about 4～5 frames, comprise blood vessel feature below Index analysis.After the inventive method replaces said method, same computer, in the case of identical image, per second process figure As 30 frames can be reached, fully meet the process of real-time monitoring ultrasonoscopy medium vessels motion change.

The foregoing is only embodiments of the invention, the present invention, all thought in the present invention and side can not be limited with this In the range of method, any modification of being made, improvement, should be included within the scope of protection of the invention.

Claims (4)

1. the fast automatic acquisition method of vessel boundary form in a kind of dynamic ultrasound image is it is characterised in that at least include following Step:

(1) open dicom3.0 form or the image of avi form, or received from ultrasonic instrument by dicom3.0 host-host protocol The image sending, draws a wire tag and analyzes and processes scope；

(2) line sweep limitss and direction controlling Vector operation are entered by following equation and method:

$hx=\frac{x_{pt20}-x_{pt10}}{\sqrt{{(x_{pt20}-x_{pt10})}^2+{(y_{pt20}-y_{pt10})}^2}},$

$hy=\frac{y_{pt20}-y_{pt10}}{\sqrt{{(x_{pt20}-x_{pt10})}^2+{(y_{pt20}-y_{pt10})}^2}},$

V_vx=v_hy, v_vy=v_hx；V_wx=-v_hy, v_wy=v_hx,

$scan\_width=\sqrt{{(pt20.x-pt10.x)}^2-{(pt20.y-pt10.y)}^2},$

Wherein pt10 is scan start point, and pt20 is terminating point, and pt10.x, pt20.x, pt10.y, pt20.y are respectively point pt10 With the x direction of pt20 and the component in y direction, xpt10, xpt20, ypt10, ypt20 be respectively point pt10 and pt20 x direction and The component in y direction, scan_width represents sweep length, hx and hy is respectively the unit vector mould in x and y direction, v_hx and v_ Hy is the unit vector in x and y direction, v_vx and v_vy processes the Unit Vector in the x and y direction of direction vector upwards for line scanning Amount, v_wx and v_wy processes downwards the unit vector in the x and y direction of direction vector for line scanning；

Wherein direction controlling parameter, when scanning up be:

N1x=x+v_vx,

N1y=y+v_vy,

N2x=n1x+v_vx,

N2y=n1y+v_vy；

When scanning downwards it is:

N1x=x+v_wx,

N1y=y+v_wy,

N2x=n1x+v_wx,

N2y=n1y+v_wy；

(3) carry out image gradient calculating, first derivation or second order derivation to process:

Dv [i, j]=v [n1x, n1y]-v [x, y],

$dv{\[i,j\]}^{\′}=\frac{(v\[n2x,n2y\]+v\[n1x,n1y\])}{2}-v\[x,y\],$

$dv{\[i,j\]}^{\′\′}=\frac{1}{3}\sqrt{v{\[n2x,n2y\]}^2-v{\[x,y\]}^2}+\frac{2}{3}\sqrt{v{\[n1x,n1y\]}^2-v{\[x,y\]}^2};$

Wherein v is image pixel grey decision-making, and scope is 0～255；Range of operation i controls for base line, is worth for [0..scan_ Untiledge], j is the scope [0..scan_width-1] of a line；Y initial value is pt20.y, each base line corresponding, at the beginning of x Initial value is pt20.x；Scan_untiledge scans the generic pixel length along line scanning direction reaching marginal point for line；

During line vessel scanning limb recognition, sweep limitss will exceed the ρ pixel at the blood vessel external mold edge of farthest, ρ=10 ～50, determination methods are the maximums looked in sweep limitss, and processing method is continuous rising and then keeps, rise, protect Hold；First rising edge is determined as vessel boundary, finds bigger graded, just with newfound in sweep limitss again Edge is defined；

(4) find the upper and lower vessel boundary point in sweep limitss by above-mentioned steps, then the noise in ultrasonoscopy is caused Edge erroneous judgement processed: set using before and after vessel boundary direction each σ point as reference, σ=3-10；If setting blood vessel side When the continuous difference in edge direction is more than τ, τ=10～20, be judged to error dot, by discarding, after a small amount of error dot abandons, in front and back Marginal point is directly connected to；

To line scanning collection to upper and lower marginal point carry out pair relationhip correction, the Data Position of blood vessel upper wall collection keeps constant, I=[0..scan_width-1], each upper wall edge point looks for lower wall edge match point, and pair principle is to be with a up contour point With reference to pairing basic point, the paired point of lower limb is in all positions of lower limb and with reference to the pairing nearest point of basic point, distance than Compared with formula it is For i-th upper wall edge point,For i-th Searching point before and after individual lower wall marginal point, formula is:J takes Value scope is line segment l704/5, the condition that another terminates the minimum distance point search in a direction be set ε, edge point away from From all big than a upper point when comparing, ε=3～5, l704 scans the length in y direction for line；

By the vessel boundary position data being collected from each frame of dynamic ultrasonic image with specified form, output and each position Cross section line number evidence, simultaneously on a computer display with image synthesis show, and show maximum inner diameter, minimum diameter with And dynamic changing curve；

The master data sheet format of result output is:

The detailed content form of result output data is defined as:

2. the fast automatic acquisition method of vessel boundary form in dynamic ultrasound image as claimed in claim 1, its feature exists In ρ=30.

3. the fast automatic acquisition method of vessel boundary form in dynamic ultrasound image as claimed in claim 1, its feature exists In σ=3, τ=10.

4. the fast automatic acquisition method of vessel boundary form in dynamic ultrasound image as claimed in claim 1, its feature exists In ε=3.