CN102800089B - Main carotid artery blood vessel extraction and thickness measuring method based on neck ultrasound images - Google Patents

Abstract

The invention discloses a main carotid artery blood vessel extraction and thickness measuring method based on neck ultrasound images. The method comprises the following specific steps of: reading neck ultrasound three-dimensional body data, and marking a main carotid artery axis based on a main carotid artery branching node; sequentially projecting and segmenting the neck ultrasound three-dimensional data in the three-view drawing direction to obtain two-dimensional cross section, coronal plane and vertical plane sequence images; and preprocessing, dividing and reconstructing the two-dimensional cross section, coronal plane and vertical plane sequence images or counting the thickness of intima-media membrane to obtain the final relevant information of internal and external profiles of the neck ultrasound main carotid artery blood vessel wall and the thickness of the blood vessel wall. By the method, the defects that the calculating complexity in the blood vessel dividing method is high, the thickness of the blood vessel wall cannot be accurately measured and an error is likely to be caused by subjective factors during computer-aided diagnosis are overcome, and the internal and external profiles of the neck ultrasound main carotid artery blood vessel wall and the thickness of the blood vessel wall can be completely, rapidly and accurately obtained. In comparison with the manual dividing method, the method is rapid in operation, and can be used for auxiliary diagnosis and prevention and treatment of neck atherosclerosis and cardiovascular disease.

Description

Main carotid artery vascular based on neck ultrasonoscopy extracts and method for measuring thickness

Technical field

The invention belongs to biomedical engineering and image procossing crossing domain, be specifically related to a kind of main carotid artery vascular based on neck ultrasonoscopy and extract and method for measuring thickness.

Background technology

The display of World Health Organization (WHO) recent statistics data, angiocardiopathy (Cardiovascular Diseases, CVDs) is one of three large fatal disease in the world.Arteries wall thickening, formation patch, and then cause hemadostewnosis, be one of atherosclerotic characteristic signs.Carotid atherosclerosis be a kind of can the cardiovascular and cerebrovascular disease of sick, the apoplexy of cardiac trigger, serious harm human body is healthy.Therefore, to its early prevention, diagnosis, treatment and monitoring important in inhibiting.

On January 20th, 2010, american heart association (American Heart Association, AHA) 10 years healthy strategic objectives of strategic planning Working Committee issue, the i.e. healthy strategy of AHA 2020---" definition and formulation promote cardiovascular health and reduce the national objective of disease ", first " to improve the general level of the health for main target " is proposed, indicate that the prevention battle line of angiocardiopathy CVDs moves forward by american heart association AHA further, not only for people at highest risk and the patient groups with hazards, and the general level of the health of general population to be improved.Research shows, the key of prevention and corntrol angiocardiopathy CVDs is detection morning, early treatment.Therefore, to atherosclerotic early detection and control, there is extremely important clinical meaning to reduction angiocardiopathy CVDs mortality ratio.

The vessel wall thickening degree of main arteria carotis (Common Carotid Artery, CCA) can be used as the important indicator weighing pathology.Ultrasonic imaging technique specific " real-time, economic, reliable, safety " advantage, makes the Internal-media thickness (Intima-media Thickness, IMT) based on this technology become one of common counter of assessment degree of carotid.Ultrasonoscopy vessel extraction and thickness measure become study hotspot in recent years.

First, with regard to vessel extraction, Chinese Patent Application No. be 200910106119.6 and 201010297322.9 two patents propose the method for ultrasonoscopy vessel extraction, the former is for non-sequence single two-dimensional ultrasonic blood vessel gray level image, the latter is applied to intravascular ultrasound sequence image, and both all lack finally detailed vessel information.As: do not complete the work such as segmentation and method evaluation for sequence image targetedly, cannot obtain closest to real blood vessel thickness value, in domestic ultrasonic machine software upgrading is regenerated, be difficult to commercial application according to main arteria carotis CCA internal anatomy.

Secondly, with regard to thickness measure, CULEX (Completely User-independent Layer Extraction) and CALEX (Completely Automatic Layer Extraction) is method for measuring thickness novel, the most intelligent at present.Both all can reach full-automatic thickness measure, but realize that difficulty is large, computation complexity is high.CULEX and CALEX utilizes different characteristics of image, adopts diverse thought---the former utilizes the partial statistics value of image pixel to differentiate Endovascular pixel and tissue pixels, and dividing method is then the combination based on gradient and motility model; And the latter integrates feature extraction, linear fit and classifies.The segmentation effect of contrast two kinds of methods can be found out, the segmentation of CALEX to lumen of vessels-inner membrance LI profile is imperfect, but the segmentation of centering film-adventitia MA profile is better than CULEX; CULEX affects comparatively large by picture noise and image artifacts simultaneously, and the execution efficiency of CALEX is far above CULEX.

Summary of the invention

The object of the present invention is to provide that a kind of energy is comprehensive, multi-angle, fast, accurately and be easy to realize, easy to operate main arteria carotis CCA vessel extraction and method for measuring thickness.

Main carotid artery vascular based on neck ultrasonoscopy extracts and method for measuring thickness, comprises the following steps:

Read neck ultrasonic three-dimensional volume data, mark main carotid artery vascular central shaft based on main carotid bifuracation point;

According to central shaft, by three-view diagram direction projection, cutting neck ultrasonic three-dimensional volume data, obtains two-dimentional transversal section, sagittal plane and coronal-plane sequence image;

In each image of two-dimentional transversal section sequence image, choose each main carotid artery vascular area-of-interest respectively, and to the pre-service of each main carotid artery vascular area-of-interest; In each main carotid artery vascular area-of-interest after the pre-treatment, segmentation obtains the inside and outside profile of each main carotid artery vascular respectively; According to the inside and outside profile of the main carotid artery vascular area-of-interest of each two-dimentional cross-sectional image, by its spatial relation three-dimensional reconstruction, obtain three-dimensional main carotid artery vascular profile;

In each image of two-dimentional sagittal plane sequence image, choose each main carotid artery vascular area-of-interest respectively, and to the pre-service of each main carotid artery vascular area-of-interest; In each main carotid artery vascular area-of-interest after the pre-treatment, segmentation obtains the inside and outside profile of each main carotid artery vascular respectively; Thickness between the inside and outside profile calculating each main carotid artery vascular respectively and Internal-media thickness; Add up the Internal-media thickness average of each two-dimentional sagittal view picture;

In each image of two-dimensional coronal face sequence image, choose each main carotid artery vascular area-of-interest respectively, and to the pre-service of each main carotid artery vascular area-of-interest; In each main carotid artery vascular area-of-interest after the pre-treatment, segmentation obtains the inside and outside profile of each main carotid artery vascular respectively; Thickness between the inside and outside profile calculating each main carotid artery vascular respectively and Internal-media thickness; The Internal-media thickness average of statistical computation each two-dimensional coronal face image;

Further, in the main carotid artery vascular area-of-interest in each image of described two-dimentional transversal section sequence image, segmentation obtains the inside and outside profile of each main carotid artery vascular in the following manner:

A1, in the pretreated main carotid artery vascular area-of-interest of two-dimentional cross-sectional image, extract the Internal periphery of main carotid artery vascular, and do smoothing processing;

A2, carry out ellipse fitting according to the Internal periphery of main carotid artery vascular, after ellipse matching obtained amplifies, as the initial outline of main carotid artery vascular;

A3, initial outline according to main carotid artery vascular, develop and obtain the final outline of main carotid artery vascular.

Described steps A 1 is specially: in the pretreated main carotid artery vascular area-of-interest of two-dimentional cross-sectional image, extract home position and the radius of main carotid artery vascular, extracts Internal periphery according to home position and radius.

In described steps A 2, oval magnification ratio coefficient is 1.02 ~ 1.08.

The evolution method that described steps A 3 adopts be Active Contour Model, GVF-Snake, FastMarching Method, Level Set, Sparse Field Algorithm any one.

Further, in the main carotid artery vascular area-of-interest in each image of described two-dimentional sagittal plane sequence image, segmentation obtains the inside and outside profile of each main carotid artery vascular in the following manner:

B1, the point arranged as each in pretreated main carotid artery vascular area-of-interest two-dimentional sagittal view, carry out the index value of sequence number mark as each row point from top to bottom successively;

The coarse segmentation of B2, area-of-interest:

B21, acquisition lower boundary profile DB: from left to right, order scans each row of the main carotid artery vascular area-of-interest of two-dimentional sagittal view picture, obtains maximum gradation value of each row, minimum gradation value and meets the lower boundary candidate point of threshold condition; In lower boundary candidate point, the maximum point of marked index value is as unique lower boundary point; Connect all lower boundary point in turn and form lower boundary profile DB; Lower boundary candidate point gray-scale value GV _candithe gray threshold condition that should meet is: GV _candidate>=a × Δ GV+GV _min=a × (GV _max-GV _min)+GV _min, the span of weight coefficient a is 0.87 ~ 0.93;

B22, acquisition coboundary profile UB: lower boundary profile DB is upwards moved in parallel 20 ~ 30 pixels as coboundary temporary profile UB_t in the main carotid artery vascular area-of-interest of two-dimentional sagittal view picture; Choose the template that size is X × X, X gets 8 ~ 15 pixels, utilizes template from left to right, travels through from top to bottom, the pixel average EX of calculation template overlay area and variance DX to the region between coboundary temporary profile UB_t and lower boundary profile DB; If meet mean variance threshold condition EX≤b and DX≤c, b span is 0.05 ~ 0.09, c span is 0.11 ~ 0.15, then the center of template overlay area is coboundary candidate point; If do not meet mean variance threshold condition, then corresponding UB_t point is coboundary candidate point; In the coboundary candidate point of each row, marked index value the maximum is as unique coboundary point; Connect all coboundaries point in turn and form coboundary profile UB;

The segmentation of B3, area-of-interest is cut: the region segmentation between lower boundary profile DB and coboundary profile UB cuts out beyond lumen of vessels, inner membrance and middle film, adventitia and adventitia and organizes;

The interface of B4, extraction lumen of vessels and inner membrance is Internal periphery, and in extraction, the interface of film and adventitia is outline.

Adopt in C average, fuzzy C-mean algorithm, Support Vector Machine SVM, AdaBoost algorithm in the thin segmentation step B3 of described area-of-interest any one.

Further, in the main carotid artery vascular area-of-interest in each image of described two-dimensional coronal face sequence image, segmentation obtains the inside and outside profile of each main carotid artery vascular in the following manner:

Initialization Internal periphery in C1, the main carotid artery vascular area-of-interest of employing Mathematical Morphology Method after the Image semantic classification of two-dimensional coronal face, develops according to initialization Internal periphery and generates final Internal periphery;

C2, to be moved down obtain initial outline by initial Internal periphery, developing according to initial outline generates final outline;

The evolution method that described step C1 and C2 adopt is any one in classical S-Shaped Algorithm, GVF-Snake algorithm, level set, MS model, CV model evolution method.

Beneficial effect of the present invention:

Compare with method for measuring thickness with traditional vessel extraction, the main arteria carotis CCA vessel extraction based on neck ultrasonoscopy provided by the invention and method for measuring thickness, have 4 differences:

(1) be different from classic method and carry out vessel extraction and thickness measure on some, the present invention is by neck ultrasonic three-dimensional volume data, by the cutting of three-view diagram direction projection, obtain the sequence image of two-dimentional transversal section, sagittal plane and coronal-plane respectively, and process respectively on each projecting plane; Be different from classic method and carry out vessel extraction and thickness measure on the single-frame images of some, the present invention processes respectively to the sequence image in three faces targetedly, after intravascular segmentation, extraction, measurements and calculations, original Complete three-dimensional data message can be made full use of, obtain omnibearing blood vessel parameter, as: blood vessel thickness, area, volume and Patch size, quantity etc.;

(2) in the main carotid artery vascular area-of-interest in each image of two-dimentional transversal section sequence image, when segmentation obtains the Internal periphery of each main carotid artery vascular, introduce Hough transform loop truss, obtain blood vessel home position and radius parameter, thus the structural motif size in subsequent mathematical morphology can be instructed to change, optimize Internal periphery segmentation result; In main carotid artery vascular area-of-interest in each image of two-dimentional transversal section sequence image, develop before obtaining the outline of each main carotid artery vascular, introduce ellipse fitting strategy as priori, and using the ellipse of matching as initial outline, more meet the physiology shape of blood vessel; In main carotid artery vascular area-of-interest in each image of two-dimentional transversal section sequence image, when evolution obtains the outline of each main carotid artery vascular, introduce active contour model ACM method, to solve the weak boundary of blood vessel outline, to be difficult to the problem of differentiation;

(3), in the main carotid artery vascular area-of-interest in each image of two-dimentional sagittal plane sequence image, " thick-thin " two-layer segmenting structure when segmentation obtains the inside and outside profile of each main carotid artery vascular, is adopted; When " coarse segmentation ", gray threshold condition and mean variance threshold condition is utilized to use restraint; When " segmentation is cut ", first introduce fuzzy C-mean algorithm FCM clustering algorithm; Secondly, vascular tissue is divided into multiclass; Finally merge into three classes respectively again---" lumen of vessels ", " interior middle film " and " adventitia and outer tissue ", obtain each interfacing profiles, to improve the accuracy that sagittal plane Internal-media thickness is measured;

(4), in the main carotid artery vascular area-of-interest in each image of two-dimensional coronal face sequence image, when segmentation obtains the inside and outside profile of each main carotid artery vascular, Mathematical Morphology Method is utilized to obtain its initial inside and outside profile; Final inside and outside profile is obtained again by evolution; In in the image of two-dimensional coronal face in film segmentation, on the basis of Snake algorithm, again introduce GVF-Snake algorithm, solve deep recess problem, to improve the degree of accuracy that coronal-plane Internal-media thickness is measured.

To sum up, the main arteria carotis CCA vessel extraction based on neck ultrasonoscopy provided by the invention and method for measuring thickness, first can tackle the diversity of ultrasonoscopy effectively, reaches accurately, the object of Fast Segmentation blood vessel and detect thickness; Secondly, through quantitative test with compare, quite, its application directly can alleviate the image hand labeled workload of medical personnel's magnanimity for this method and manual segmentation, method error of measuring; Finally, the clinical parameter (as: blood vessel thickness, area, volume and Patch size, quantity etc.) of gained is analyzed based on the inventive method, not only can reflect vascular lesion in visual rationing ground, and more directive significance can be provided for the early prevention and treatment of carotid atherosclerosis.

Accompanying drawing explanation

Fig. 1 is the concrete steps process flow diagram in the present invention;

Fig. 2 is the process flow diagram of two-dimensional sequence image procossing in the present invention;

Fig. 3 is the process flow diagram of concrete steps of the present invention;

Fig. 4 is mark three-dimensional data 3D_Data bifurcation and central axis direction schematic diagram;

Fig. 5 is that main arteria carotis is according to axis Three-view projection cutting schematic diagram;

Fig. 6 is two-dimentional transversal section sequence image (2D_Data_Z);

Fig. 7 is the 13rd original image (2D_Data_Z_k, k=13) in the sequence image of two-dimentional transversal section;

Fig. 8 is two-dimentional sagittal plane sequence image (2D_Data_Y);

Fig. 9 is the 2nd original image (2D_Data_Y, j=2) in two-dimentional sagittal plane sequence image;

Figure 10 is two-dimensional coronal face sequence image (2D_Data_X);

Figure 11 is the 1st original image (2D_Data_X_i, i=1) in two-dimensional coronal face sequence image (2D_Data_X);

Figure 12 is two-dimentional transversal section segmentation of sequence image process flow diagram;

Figure 13 is the region of interest ROI figure of this two-dimentional cross-sectional image;

Figure 14 be this region of interest ROI pre-service progressively result figure: Figure 14-(a) be divided linear strength result figure; Figure 14-(b) is SRAD nonlinear filtering result figure;

Figure 15 is the result figure to pre-service area-of-interest Hough transform loop truss;

Figure 16 is the result figure to pre-service area-of-interest Canny operator edge detection;

Figure 17 is morphology processing result figure;

Figure 18 is the single pixel Internal periphery result figure extracted;

Figure 19 is that Internal periphery launches point set figure;

Figure 20 is the matched curve figure that Internal periphery launches point set;

Figure 21 is that Internal periphery restores point set figure;

Figure 22 is this two-dimentional transversal section Internal periphery segmentation result figure: wherein filled circles form point represents manual segmentation profile; Solid line represents segmentation contour of the present invention;

Figure 23 is Internal periphery ellipse fitting result figure;

Figure 24 is this two-dimentional transversal section outline segmentation result figure: wherein solid diamond point represents manual segmentation profile; Solid square form point represents segmentation contour of the present invention;

Figure 25 is two-dimentional sagittal plane pre-processed results figure: Figure 25-(a) is normalization result figure; Figure 25-(b) is filter result figure;

Figure 26 is two-dimentional sagittal plane region of interest ROI pre-processed results figure;

Figure 27 is two-dimentional sagittal plane region of interest ROI coarse segmentation result figure: Figure 27-(a) is original size; Figure 27-(b) shows diagram for amplification one; Figure 27-(c) is for amplifying twice display figure;

Figure 28 is thin segmentation result figure: Figure 28-(a) of two-dimentional sagittal plane region of interest ROI is original size; Figure 28-(b) shows diagram for amplification one; Figure 28-(c) is for amplifying twice display figure;

Figure 29 is this two-dimensional coronal face extraction region of interest ROI result figure: Figure 29-(a) is original graph; Figure 29-(b) schemes for corresponding ROI;

Figure 30 is this two-dimensional coronal face region of interest ROI pre-processed results figure;

Figure 31 is this area-of-interest thresholding result figure;

Figure 32 is this area-of-interest filling cavity result figure;

Figure 33 is this area-of-interest intimal surface deburring result figure;

Figure 34 is this area-of-interest inner membrance correction result figure;

Figure 35 is the initial profile (dotted line) of inner membrance and final profile (solid line) result figure: Figure 35-(a) is original size; Figure 35-(b) shows diagram for amplification one; Figure 35-(c) is for amplifying twice display figure;

Figure 36 is the initial profile (dotted line) of adventitia and final profile (solid line) result figure: Figure 36-(a) is original size; Figure 36-(b) shows diagram for amplification one; Figure 36-(c) is for amplifying twice display figure;

Figure 37 be inner membrance (on) and adventitia (under) initial profile (dotted line) and final profile (solid line) result figure: Figure 37-(a) be original size; Figure 37-(b) shows diagram for amplification one; Figure 37-(c) is for amplifying twice display figure;

Figure 38 is this two-dimensional coronal face manual measurement Internal-media thickness goldstandard result figure.

Embodiment

Illustrate below in conjunction with concrete exemplifying embodiment and accompanying drawing, the present invention is described in further detail, and give the solution of the present invention checking.

Based on main arteria carotis CCA vessel extraction and the method for measuring thickness of neck ultrasonoscopy, comprise following five steps and plan-validation, as shown in Figure 1; Concrete, to the process flow diagram of each two-dimensional sequence image procossing, as shown in Figure 2; General flow chart, as indicated at 3.

Step (1) reads neck ultrasonic three-dimensional volume data 3D_Data, marks main carotid bifuracation point " BF " and central shaft, as shown in Figure 4.

(1.1) bifurcation " BF " is located

In three-D ultrasonic volume data 3D_Data, identify main arteria carotis CCA, arteria carotis interna ICA and arteria carotis externa ECA, and determine the position (see Fig. 4) of carotid bifuracation point " BF ".

(1.2) central shaft of main arteria carotis CCA is located

With reference to the position of bifurcation " BF ", according to main arteria carotis CCA physiological anatomic architecture, cental axial position according to a preliminary estimate, the central shaft estimated arbitrarily selects 2 points---" point 1 " and " point 2 " (see Fig. 4), the line of " point 1 " and " putting 2 " is as the central shaft of main arteria carotis CCA.

Step (2) is according to central shaft, by three-view diagram direction projection, as shown in Figure 5, cutting neck ultrasonic three-dimensional volume data 3D_Data, obtains two-dimentional transversal section sequence image (2D_Data_Z), two-dimentional sagittal plane sequence image (2D_Data_Y) and two-dimensional coronal face sequence image (2D_Data_X) respectively;

(2.1) two-dimentional transversal section sequence image (2D_Data_Z)

Carry out cutting by perpendicular to arteria carotis central axis direction, obtain two-dimentional transversal section sequence image.General, two-dimentional transversal section sequence image has 13 ~ 15 at the most, as shown in Figure 6; In the present invention, get the 13rd and exemplarily illustrate, as shown in Figure 7.

(2.2) two-dimentional sagittal plane sequence image (2D_Data_Y)

Carrying out cutting by being parallel to arteria carotis central axis direction, from the side projection, obtaining two-dimentional sagittal plane sequence image.General, two-dimentional sagittal plane sequence image has 4 ~ 6 at the most, as shown in Figure 8; In the present invention, get the 2nd and exemplarily illustrate, as shown in Figure 9.

(2.3) two-dimensional coronal face sequence image (2D_Data_X)

By being parallel to arteria carotis central axis direction, carrying out cutting from front projection, obtain two-dimensional coronal face sequence image.General, the sequence image in two-dimensional coronal face has 3 at the most, as shown in Figure 10; In the present invention, get the 1st and exemplarily illustrate, as shown in figure 11.

Step (3), according to two-dimentional transversal section sequence image, extracts three-dimensional main carotid artery vascular profile.The inside and outside vessel profile of this step segmentation each two-dimentional transversal section sequence image, concrete segmentation step process flow diagram, as shown in figure 12.

In this step, one in two-dimentional transversal section sequence image (2D_Data_Z) is selected this step (see Fig. 7) to be described for example, other image is all adopted to use the same method and is processed, Ink vessel transfusing, the outline of final acquisition all two-dimentional transversal section sequence image, thus carry out the calculating of the parameter such as area, volume, clinic diagnosis characteristic evidences is provided, weighs the clinical state of an illness qualitatively.In view of ultrasonic sequence image quality entirety is poor, need be improved via pre-service; The key of vessel extraction is exactly the segmentation of inside and outside profile, in conjunction with vascular morphology in the present invention, introduces circle, ellipse as shape priors, more effectively splits blood vessel; In order to better present segmentation result, finally targeted again aftertreatment is carried out to Internal periphery.

(3.1) in two-dimentional transversal section sequence image (2D_Data_Z), each main carotid artery vascular area-of-interest (2D_Data_Z_k_ROI) is chosen respectively, and to the pre-service of each main carotid artery vascular area-of-interest.Pretreated technical thought is: first grey scale change, then filtering noise reduction.

Original two-dimentional cross-sectional image gray scale is partially dark, contrast is low, noise is comparatively large, therefore needs to do pre-service to it.The present invention is to extract blood vessel, therefore only the subregion comprising blood vessel in two-dimentional cross-sectional image need be carried out pre-service as region of interest ROI, as shown in figure 13, as shown in figure 14, preprocessing process is specific as follows for each pre-treatment step result for the region of interest ROI of Fig. 7:

(3.1.1) greyscale transformation (Figure 14-a)

When greyscale transformation is to improve image procossing, the dynamic range of gray level, to improve the brightness and contrast of image, can adopt the methods such as linear stretch, Nonlinear extension, image enhaucament.This example selects the simplest piecewise linear transform function to be described, and the method is made up of two basic operations:

(3.1.1.1) two flex points of image being carried out to gray scale stretching are determined;

Statistics with histogram is done to two-dimentional cross-sectional image, obtains the image of L gray level.Gray-scale value before and after greyscale transform process is defined respectively with r and s respectively, and the greyscale transformation function representation of the image of L gray level is s=T (r).Suppose P1, P2 is two flex points of piecewise linear transform function T (r), and the gray-scale value before and after its conversion is respectively (r ₁, s ₁) and (r ₂, s ₂).

(3.1.1.2) greyscale transformation

This example has added up 10 anonymous patient's three-dimensional datas, obtains two-dimentional cross-sectional image 300 altogether, after statistical analysis, arranges the grey level range of paid close attention to former figure for [r _min, r _max], then P1, P2 two flex points are transforming function transformation function T(r) be divided into 3 sections: r _min≤ r < r ₁, r ₁≤ r≤r ₂, r ₂< r≤r _max.The expression formula of T (r) is as follows:

$T\left(r\right)=\left\{\begin{array}{c}(r-r_{\min })\&times;\frac{s_1}{r_1},r_{\min }\&le;r\&le;r_1\\ s_1+(r-r_1)\&times;\frac{s_2-s_1}{r_2-r_1},r_1\&le;r\&le;r_2\\ s_2+(r-r_2)\&times;\frac{(L-1)-s_2}{r_{\max }-r_2},r_2<r\&le;r_{\max }\end{array}\right.$

Thus, just gray level by original paid close attention to scope linear stretch to saturation range [0, L-1], L be conversion after gray level.The greyscale transformation result of Figure 13 is as shown in Figure 14-a.

(3.1.2) filtering noise reduction (Figure 14-b)

Filtering noise reduction can select the methods such as non-local mean, gaussian filtering, speckle noise Anisotropic Diffusion Model SRAD.Wherein, SRAD nonlinear filter owing to adopting different coefficient of diffusion on different dispersal directions, therefore has enhancing contrast, retains the advantages such as details.SRAD wave filter smoothly proposes based on local statistics filter and anisotropy, traditional compared to other based on local statistics filter and anisotropy smoothing filter, there is better homogeneous region Lubricity and retain the advantage that details and border strengthen better.The filtering noise reduction result of Figure 14-a is as shown in Figure 14-b.

(3.2), in each main carotid artery vascular area-of-interest after the pre-treatment, segmentation obtains the inside and outside profile of each main carotid artery vascular.

The technical thought of the inside and outside contours extract of main carotid artery vascular is: first extract Internal periphery; Secondly, matching obtains initial outline; Finally, evolution obtains the final outline of main carotid artery vascular.

(3.2.1), in each main carotid artery vascular area-of-interest after the pre-treatment, the Internal periphery of each main carotid artery vascular is extracted, and smoothing processing.

The technical thought extracting the Internal periphery of main carotid artery vascular is: first extract blood vessel parameter information, secondly, according to blood vessel parameter information, extracts blood vessel list pixel Internal periphery; Finally, smoothing processing is done to Internal periphery.

(3.2.1.1), in each main carotid artery vascular area-of-interest after the pre-treatment, blood vessel parameter information is extracted.

This step is parameter information in order to extract blood vessel and home position and radius, and this example adopts Hough Hough transform circle detection method example that (not limiting to the method) is described.O (centre, R can be obtained in Hough transform loop truss _max, R _min), be respectively central coordinate of circle, maximum radius, least radius, thus provide reference frame for the cell sizes size of subsequent step.The essence of Hough transform is that the picture dot in image space with certain relation is carried out cluster, sets up the parameter space that these picture dots can be connected by a certain analytical form, finds accumulative corresponding point.Detect in circle in Hough transform, the frontier point obtained by Canny operator is picture dot, and central coordinate of circle and these three parameters of radius of circle are the parameter of its corresponding analytical form.Separately in this example, in order to reduce program runtime, selected certain radius of circle scope and home position artificially.Figure 15 gives the result of Figure 14-b being carried out to transform circle detection.

(3.2.1.2) according to blood vessel parameter information, blood vessel list pixel Internal periphery is extracted

Based on Mathematical Morphology Method, active contour model, Level Set Method or Fast Marching, all divisible main carotid artery vascular Internal periphery obtaining area-of-interest after pre-service in two-dimentional cross-sectional image.This example adopts the explanation of Mathematical Morphology Method example.

(3.2.1.2.1) extract the edge in area-of-interest, this example adopts Canny operator edge, the edge extracting result of Figure 13-b, as shown in figure 16;

(3.2.1.2.2) structural motif size and shape are set

Wherein in structural motif size SE_si, R _maxr _mbe respectively the maximum radius, the least radius (see Figure 15) that calculate gained circle in (3.2.1.1); And structural motif size SE_size, shape SE shape is adjustable.If: conventional size is 3,5,7 etc.; Conventional shape is circle, cruciform, square etc.In the present invention, structural motif shape SE_shape adopts anistree disk, and size is the integer multiple of 3.

(3.2.1.2.3) mathematical morphology closed operation operation

Structural motif is acted on the enterprising line number closing operation of mathematical morphology operation of Figure 16, obtain the Internal periphery place local closed.The result of Figure 16 after closed operation operation as shown in figure 17.

(3.2.1.2.4) single pixel Ink vessel transfusing profile is extracted

Vessel profile can be described by its a series of point.Therefore, after morphological segment Ink vessel transfusing profile, single pixel point set is extracted to its place local, namely can be used as the Internal periphery of this two-dimentional transversal section blood vessel.Single pixel profile point of Figure 17 as shown in figure 18.

(3.2.1.3) smoothing processing (Figure 19, Figure 20 and Figure 21) of Ink vessel transfusing profile

The smoothing processing of vessel profile comprises internally, the smoothing processing of outline.In the present invention, adopt " bread rolling " to launch, after fairing, fitting of a polynomial, then restore the new single pixel profile point set of acquisition, detailed process is as follows:

(3.2.1.3.1) profile center point coordinate is calculated

Calculate point concentrate the mean value of a little horizontal, ordinate, and remember point (x centered by this point ₀, y ₀);

(3.2.1.3.2) sort point

An optional point, and with this point for starting point, counterclockwise sort successively by profile;

(3.2.1.3.3) relative distance is calculated

Computing center's point (x successively ₀, y ₀) and sequence point (x _i, y _i) relative distance Dis;

${\mathrm{Dis}}_i=\sqrt{{(x_0-x_i)}^2+{(y_0-y_i)}^2}$

(3.2.1.3.4) Internal periphery point set (Figure 19) is launched

With the sequence number that sorts for horizontal ordinate, each corresponding relative distance is ordinate, obtains former profile and launches point set;

(3.2.1.3.5) fitting of a polynomial (Figure 20)

Curve fitting of a polynomial is carried out to above-mentioned point set, removes burr, make profile launch matched curve level and smooth.Wherein, polynomial expression exponent number n ₀adjustable, generally get 5 ~ 10, preferably, n ₀=9;

(3.2.1.3.6) profile point set (Figure 21) is restored

Using the curve after level and smooth as new profile developed curve, and according to sequence number, the new point (x of reverse _j, y _j) coordinate, namely profile point set restores.

$\left\{\begin{array}{c}{x}_j=x_0+{\mathrm{Dis}}_i\&times;\cos \left(\mathrm{angle}\right)\\ y_j=y_0+{\mathrm{Dis}}_i\&times;\sin \left(\mathrm{angle}\right)\end{array}\right.$

Wherein: angle is former profile central point (x ₀, y ₀) and former point (x _i, y _i) inclination angle of line.Finally, as shown in figure 22, wherein filled circles form point represents manual segmentation profile to Internal periphery result; Solid line represents segmentation contour of the present invention.

(3.2.2) according to the Internal periphery of each main carotid artery vascular, ellipse fitting is carried out to it, after the ellipse each matching obtained amplifies, as the initial outline of each main carotid artery vascular;

(3.2.2.1) according to Ink vessel transfusing profile, ellipse fitting is carried out to it;

The Internal periphery list pixel point set that the present invention determines with step (3.2.1.2.4) is given point set, ellipse is fitting function, matching obtains oval parameter E (centre, p, q, θ): represent long, the oval minor semi-axis length of oval central coordinate of circle, oval major semi-axis, transverse and horizontal line angle respectively.

In the present invention, to the result of Internal periphery ellipse fitting as shown in figure 23.

(3.2.2.2) the initial outline of blood vessel

Based on the priori of shape of blood vessel, utilize consistance and the symmetry of extra vascular profile, after ellipse fitting is carried out to the Ink vessel transfusing profile obtained, amplify the initial profile as blood vessel outline.

Amplify the ellipse of (3.2.2.1) matching, as the initial outline of blood vessel.Preferably, oval magnification ratio coefficient factor is between 1.02 ~ 1.08.The implication that the present invention amplifies is oval major semi-axis, minor semi-axis is multiplied by magnification ratio coefficient, the oval center of circle and inclination angle constant.

(3.2.3) according to the initial outline of each main carotid artery vascular, adopt active contour model ACM to develop, obtain (finally) outline of each main carotid artery vascular

Based on the initial outline of (3.2.2) blood vessel, adopt active contour model (ACM) to develop, obtain final blood vessel outline, as shown in figure 24, wherein solid diamond point represents manual segmentation profile; Solid square form point represents segmentation contour of the present invention.

Because the outline of blood vessel of the present invention is obtained by active contour model, Internal periphery adopts mathematical morphology to obtain, and therefore outline has good flatness relative to Internal periphery, so only can make smoothing processing to Internal periphery according to above-mentioned (3.2.1.3) mode.

(3.3) according to the inside and outside profile of each area-of-interest, by its spatial relation three-dimensional reconstruction, three-dimensional main carotid artery vascular profile is obtained.This vessel profile can adjuvant clinical application.

By step (3) successively to after the sequence specific primers-polymerase chain reaction of all two-dimentional transversal section, obtain the Ink vessel transfusing of each, outline.To marked each area-of-interest of main carotid artery vascular, according to spatial relation sequence stack, three-dimensional reconstruction obtains three-dimensional main carotid artery vascular; After three-dimensional reconstruction, the parameter such as area, volume of main arteria carotis CCA can be obtained, thus provide directive significance for doctor's clinical diagnosis, common as drug evaluation; If introducing comparative study, then can be used for surgical effect and evaluate etc.

Step (4), according to two-dimentional sagittal plane sequence image, calculates sagittal Internal-media thickness.

Main arteria carotis CCA vascular wall, by lumen of vessels (Lumen), is made up of trilamellar membrane from inside to outside successively---inner membrance (Intima), middle film (media) and adventitia (adventitia).Ultrasonicly utilize " lumen of vessels (L)-inner membrance (I)-middle film (M)-adventitia (A) " different acoustic impedance interfaces of being formed just, carry out imaging thus the diagnosis state of an illness.

In the present invention, the inner membrance profile (abbreviation Internal periphery) of extraction is positioned at lumen of vessels and inner membrance intersection, is designated as LIB(Lumen Intima Boundary); The epicardium contours (abbreviation outline) extracted is arranged in the intersection of film and adventitia, is designated as MAB(Media Adventitia Boundary); The Internal-media thickness (Intima-media Thickness, IMT) measured is the distance between lumen of vessels-inner membrance interface LIB and middle film-adventitia interface MAB.

The major technique thinking of this step is: the inside and outside film extracting two-dimentional sagittal plane sequence image, measures Internal-media thickness IMT, after calculating the average of sagittal plane Internal-media thickness IMT, provides blood vessel quantitative test measurement result.Get a wherein two-dimentional sagittal view picture (see Fig. 9) to illustrate as this example, other image is all adopted to use the same method and is processed, and finally obtains thickness measurement;

(4.1) in two-dimentional sagittal plane sequence image (2D_Data_Y), each main carotid artery vascular area-of-interest (2D_Data_Y_j_ROI) is chosen respectively, and to the pre-service of each main carotid artery vascular area-of-interest.Pretreated technical thought is: first normalization, then filtering noise reduction.

The two-dimentional sagittal plane sequence image chosen comprises main arteria carotis CCA, bifurcation BF, lumen of vessels, distal vessels wall adventitia, distal vessels wall adventitia are organized outward, therefrom extract and comprise subregion that distal vessels wall adventitia organizes outward as region of interest ROI, and it is normalized and the pre-service of filtering.Normalized function can adopt linear function conversion, logarithmic function conversion, the conversion of arc cotangent function etc.Filtering can adopt low-pass filtering, mean filter, medium filtering, frequency domain filtering, bandpass filtering etc.Be described for linear function normalization and low-pass filtering below, respectively as shown in Figure 25-(a) He Figure 25-(b), pretreated ROI result as shown in figure 26.

(4.1.1) linear function normalization (Figure 25-(a))

Two-dimentional sagittal plane image intensity value is carried out linear function conversion, and make to be normalized in [0,1], expression formula is as follows:

GV_aft=(GV_pre-MinValue)/(MaxValue-MinValue)

Wherein: GV_pre, GV_aft are respectively the forward and backward gray-scale value of conversion, MaxValue, MinValue are respectively maximum gradation value in figure and minimum gradation value.

(4.1.2) low-pass filtering (Figure 25-(b))

Adopt gaussian filtering, its concrete operations are: be 0 by an average, and standard deviation is 10 templates, each pixel in scan image, and the weighted mean of all pixels in the neighborhood of calculation template place, as the gray-scale value of template center's pixel.General, template size is 5,7,9, in the present invention, gets 5 × 5.Finally, pretreated ROI result as shown in figure 26.

(4.2), in each main carotid artery vascular area-of-interest after the pre-treatment, segmentation obtains inner membrance and the adventitia of each main carotid artery vascular, and calculates the Internal-media thickness of each area-of-interest;

According to the Internal-media thickness that step (4) defines, in the area-of-interest of two-dimentional sagittal view picture after the pre-treatment, extract the inside and outside film of blood vessel, and then measure the Internal-media thickness IMT of two-dimentional sagittal plane blood vessel.The technical thought of its vessel extraction is: first coarse segmentation region of interest ROI, film approximate location in location; Secondly, region of interest ROI is cut in segmentation, obtains inner membrance and adventitia respectively; Finally calculate Internal-media thickness.

(4.2.1) coarse segmentation (Figure 27) of region of interest ROI

To the point of each row on the image in ROI region, mark 0,1,2 successively from top to bottom ... sequence number, as the index value of each row point.In ROI search, in order to reduce redundant search, being necessary ROI region potential for interior, middle film to reduce further, therefore coarse segmentation is carried out to region of interest ROI, to obtain lower boundary profile DB and coboundary profile UB.The coarse segmentation result of region of interest ROI, as shown in figure 27.

(4.2.1.1) lower boundary profile DB obtains

First, lower boundary candidate point gray-scale value GV is set _candidatethe threshold condition that should meet: GV _candidate>=a × Δ GV+GV _min=a × (GV _max-GV _min)+GV _min, (general, the scope of weight coefficient a is 0.87 ~ 0.93, preferably a=0.9); Then, from left to right, each row of sequential scanning ROI, and obtain each maximum gradation value arranged, minimum gradation value and meet the lower boundary candidate point of threshold condition.In lower boundary candidate point, the maximum point of marked index value is as unique lower boundary point; Finally, connect all lower boundary point in ROI in turn and form lower boundary profile DB.

(4.2.1.2) profile UB in coboundary obtains;

First, lower boundary profile DB is upwards moved in parallel Δ in the roi _indexindividual pixel distance, Δ _indexgeneral value is 20 ~ 30 pixels, as coboundary temporary profile UB_t, and preferably Δ _index=25; Secondly, get template size X × X, from left to right, from top to bottom, calculate average EX, the variance DX of each pixel place template in the region between coboundary temporary profile UB_t and lower boundary profile DB in turn.If meet threshold condition: EX≤b and DX≤c, then template center is coboundary candidate point; If do not meet threshold condition, then corresponding UB_t point is coboundary candidate point.(general, template size X gets 8 ~ 15; B gets 0.05 ~ 0.09; C gets 0.11 ~ 0.15.Preferably, X=10; B=0.08; C=0.14.) in the coboundary candidate point of each row, conduct unique coboundary point that marked index value is maximum; Finally, connect all coboundaries point in ROI in turn and form coboundary profile UB.

(4.2.2) segmentation of region of interest ROI is cut (Figure 28)

The segmentation of region of interest ROI is cut, and completes in fact between lower boundary profile DB and coboundary profile UB.The image-region of coarse segmentation gained is interior, middle film region.In order to correct marking off beyond lumen of vessels, inner membrance, middle film, adventitia and adventitia between DB and UB be organized, the methods such as C average, fuzzy C-mean algorithm, Support Vector Machine SVM, AdaBoost algorithm can be adopted.In the present invention, adopt based on fuzzy C-means clustering method, be illustrated the extraction of vascular wall inside and outside contour.The thin segmentation result of region of interest ROI, as shown in figure 28.

According to distal vessels wall characteristics, " black, grey, white " three classes intuitively can be divided into vasculature part, can roughly corresponding " lumen of vessels, interior middle film, adventitia and organize in addition ".In the classification results that the present invention is final, set the minimum class pixel of gray-scale value as " lumen of vessels "; The class pixel that gray-scale value is the highest is " adventitia and in addition tissue "; The pixel of other gray-scale values is " interior middle film ".

Adopt fuzzy division in cluster segmentation of the present invention, make each data-oriented point value determine that it belongs to the degree of each classification in the degree of membership that [0,1] is interval.The degree of membership of a data set and always equal 1.If data set X={x ₁, x ₂..., x _n, each sample has s characteristic attribute x _j={ x _j1, x _j2..., x _js(j=1,2 ..., n).In the present invention using grey scale pixel value as unique features attribute, i.e. s=1.Realize segmentation for each row FCM cluster in ROI, we are for a certain row here, if these row have n pixel samples.Concrete steps are as follows:

(4.2.2.1) parameter initialization: given cluster classification number C(we get C=10), setting iteration stopping threshold epsilon (we get ε=1e-5), setting maximum iteration time N(we get N=100);

(4.2.2.2) initialization clustering prototype distribution matrix P ⁰: random to produce a size be the matrix of n × C, and jth row is to should the distribution being subordinate to angle value of a row jth pixel samples, and namely a jth pixel samples is under the jurisdiction of all kinds of possibilities, and satisfied

(4.2.2.3) with following formulae discovery or renewal Matrix dividing U ^(b), for if d _ik≠ 0 has:

${\mathrm{\&mu;}}_{\mathrm{ik}}^{\left(b\right)}=1/{\mathrm{\&Sigma;}}_{r=1}^c{[d_{\mathrm{ik}}^{\left(b\right)}/d_{\mathrm{rk}}^{\left(b\right)}]}^{\frac{2}{m-1}}$

If d _ik=0, then have: and to j ≠ k,

D in its Chinese style _ikrepresent the pixel samples x in the i-th class _kand the distance between the cluster centre of the i-th class, m is weighted index, and b is iterations.

(4.2.2.4) clustering distribution matrix P is upgraded with (formula 1) ^(b+1):

$P_i^{(b+1)}=\frac{{\mathrm{\&Sigma;}}_{k=1}^n{\left({\mathrm{\&mu;}}_{\mathrm{ik}}^{(b+1)}\right)}^m*x_k}{{\mathrm{\&Sigma;}}_{k=1}^n{\left({\mathrm{\&mu;}}_{\mathrm{ik}}^{(b+1)}\right)}^m},i=\mathrm{1,2},...,C$ (formula 1)

If (4.2.2.5) before and after certain iteration, clustering distribution matrix meets | P ^(b)-P ^(b+1)| < ε or b=N, then algorithm stops and exporting Matrix dividing U and clustering distribution matrix P, otherwise makes b=b+1, returns step (4.2.2.3).Finally, the pixel of all " interior middle film " is obtained.

(4.2.3) this two-dimentional sagittal plane ROI Internal-media thickness is calculated

After (4.2.2) cluster segmentation, the inside and outside film manipulative indexing value that ROI in this two-dimentional sagittal plane often arranges can be obtained, calculate that it is poor, and calculate mean value of each row, be the corresponding Internal-media thickness of this two-dimentional sagittal plane.

(4.3) according to the Internal-media thickness of each area-of-interest, its average of statistical computation, is sagittal Internal-media thickness

To each of two-dimentional sagittal plane sequence image, after the step of above-mentioned (4.1) and (4.2), carry out result statistical study, the parameter such as average, variance of acquisition sagittal plane Ink vessel transfusing media thickness IMT, as blood vessel evaluation index.IMT value has become the conventional efficiency index of prediction cardiovascular and cerebrovascular disease.

Step (5), according to two-dimensional coronal face sequence image, calculates the Internal-media thickness of coronal-plane:

Internal-media thickness in this step defines same step (4), and the Internal-media thickness namely measured (Intima-media Thickness, IMT) is the distance between lumen of vessels-inner membrance interface LIB and middle film-adventitia interface MAB.

The major technique thinking of this step is: the inside and outside film extracting two-dimensional coronal face sequence image, measures Internal-media thickness IMT, after calculating the average of coronal-plane Internal-media thickness IMT, provides blood vessel quantitative test measurement result.Get wherein two-dimensional coronal face image (see Figure 11) to illustrate as this example, other image is all adopted to use the same method and is processed, and finally obtains thickness measurement;

(5.1) in two-dimensional coronal face sequence image (2D_Data_X), each main carotid artery vascular area-of-interest (2D_Data_X_i_ROI) is chosen respectively, and to the pre-service of each main carotid artery vascular area-of-interest.Pretreated technical thought is: first strengthen, then filtering.

The two-dimensional coronal face sequence image chosen comprises main arteria carotis CCA, lumen of vessels, distal vessels wall adventitia, distal vessels wall adventitia are organized outward, therefrom extracts and comprises subregion that distal vessels wall adventitia organizes outward as region of interest ROI.Figure 11 extracts region of interest ROI, and result as shown in figure 29.

Image enhaucament and filter preprocessing are carried out to it.Image enhaucament can adopt the regulation, partial statistics method etc. of adaptive histogram equalization, spatial domain; Filtering can adopt low-pass filtering, mean filter, medium filtering etc.Below for adaptive histogram equalization and self-adaptation mean filter, be described.Obtain pretreated area-of-interest, as shown in figure 30.

(5.1.1) image enhaucament

Adaptive histogram equalization (AHE, Adaptive Histogram Equalization) is adopted to carry out image enhaucament.Compared with traditional histogram equalization, AHE focuses on the local feature of image more.AHE decides contrast enhancement process according to the partial statistics characteristic of pixel.The gray-scale value of each pixel is obtained by an equalization transforming function transformation function, and this transforming function transformation function is obtained by the local word image histogram of centered by this pixel, is called Image Enhancement Method of Local Contrast.Formula is:

x′ _i，j＝m _i，j+k×(x _i，j-m _i,j)

Wherein, k is adaptive reference amount, and expression formula is for the gray variance in window W, for the gray variance of entire image, k ' is scale-up factor; x _{i, j}, x ' _{i, j}be respectively the gray-scale value before and after conversion; m _{i, j}for the mean value of pixel grey scale in window W.

(5.1.2) self-adaptation mean filter

Mean filter, also referred to as linear filtering, mainly removes noise by zone levelling.The ultimate principle of linear filtering is each pixel value replaced by average in original image, namely to the gray-scale value x of pending current pixel point _{i, j}, select a template, this template is made up of some pixels of its neighbour, the average of all pixel grey scales in seeking template, then gives current pixel point this average, as image gray scale at that point after process, i.e. and x ' _{i, j}.Finally, pretreated area-of-interest is obtained, as shown in figure 30.

(5.2), in each main carotid artery vascular area-of-interest after the pre-treatment, segmentation obtains inner membrance and the adventitia of each main carotid artery vascular, and calculates the Internal-media thickness of each area-of-interest

According to the Internal-media thickness that step (4) defines, in the area-of-interest of two-dimensional coronal face image after the pre-treatment, extract the inside and outside film of blood vessel, and then measure the Internal-media thickness IMT of two-dimensional coronal face blood vessel.The technical thought of its vessel extraction is: first initialization extract endangium profile; Next initialization also extracts externa profile; Finally calculate Internal-media thickness.

(5.2.1) initialization Internal periphery

In view of the design feature of main arteria carotis CCA, two-dimensional coronal face sequence image is very easily subject to the artifact effects of bifurcation and arteria carotis externa ECA, causes two-dimensional coronal face sequence image to only have about 3 at the most.Thus add the difficulty of two-dimensional coronal face segmentation.In order to obtain close to real inner membrance fast, adopt in the present invention and obtaining initialization Internal periphery based on Mathematical Morphology Method.

(5.2.1.1) image threshold

The method of thresholding mainly contains four classes: based on the global threshold method of point, global threshold method, local threshold method and the multi thresholds method based on region.The embodiment of the present invention adopts Otsu algorithm OTSU(self-adaption thresholding), such that white portion comprises inner membrance, middle film, adventitia, adventitia are organized outward and partial noise, black region is lumen of vessels, as shown in figure 31;

(5.2.1.2) filling cavity

In order to fill the cavity in white portion shown in Figure 27, template matching method, point by point scanning method etc. can be adopted.In view of the feature of bianry image, the present invention adopts closing operation of mathematical morphology, and result as shown in figure 32.

(5.2.1.3) burr is removed

In order to make endangium surface smoothing, meeting physilogical characteristics, adopting mathematical morphology open operator, remove the burr that white portion surface additional noise produces, result as shown in figure 33;

(5.2.1.3) initial inner membrance profile is revised

For the intimal surface marginal point shown in Figure 33, gap extraction is carried out to the coboundary pixel of white portion, the pixel with its spacing 3 ~ 5 pixel distances is extracted downwards with the pixel extracted, a series of inner membrance edge initial coordinate point can be obtained, using serial inner membrance edge initial coordinate point line as inner membrance initial profile, as shown in figure 34;

(5.2.2) Internal periphery (Figure 35) is extracted

According to the gradient information of area-of-interest, classical snakelike (Snake) method is adopted to calculate the inside and outside acting force of inner membrance initial profile.Compared by inside and outside acting force, according to comparative result evolution profile until inside and outside acting force is equal, the profile finally determined is the Internal periphery of extraction, and as shown in figure 35, wherein the initial profile of inner membrance is dotted line; The final profile of inner membrance is solid line.Except classical snakelike (Snake) method, also can adopt the evolution methods such as level set, CV model, GVF-Snak.

(5.2.3) initialization epicardium contours

The inner membrance initial profile (5.2.1) obtained moves in parallel Δ downwards _indexindividual pixel distance, can obtain initial epicardium contours.General Δ _indexbe 15 ~ 30, preferably Δ _indexget 25.

(5.2.4) epicardium contours (Figure 36) is extracted

According to gradient information, the field information of area-of-interest, gradient vector field method GVF-Snake is adopted to calculate the inside and outside acting force of adventitia initial profile.Compared by inside and outside acting force, according to comparative result evolution profile until inside and outside acting force is equal, the profile finally determined is epicardium contours, and as shown in figure 36, wherein the initial profile of adventitia is dotted line; The final profile of adventitia is solid line.Except GVF-Snake algorithm, also can adopt the classical evolution method such as snakelike (Snake), level set, CV model.

(5.2.5) this two-dimensional coronal face ROI Internal-media thickness (Figure 37) is calculated

After (5.2.1)-(5.2.4) segmentation, the total ROI in this two-dimensional coronal face can be obtained and often arrange inside and outside film corresponding pixel points position, calculate the pixel interval between inside and outside film profile corresponding pixel points, and be translated into actual range, and calculate the mean value of each row, be the corresponding Internal-media thickness in this two-dimensional coronal face.As shown in figure 37, the spacing of the final profile of top inner membrance (solid line) and the final profile of below adventitia (solid line) is required Internal-media thickness.

(5.3) according to the Internal-media thickness of each area-of-interest, its average of statistical computation, is the Internal-media thickness of coronal-plane;

To each of two-dimensional coronal face sequence image, after the step of above-mentioned (5.1) and (5.2), carry out result statistical study, the parameter such as average, variance of acquisition coronal-plane Ink vessel transfusing media thickness IMT, as blood vessel evaluation index.IMT value has become the conventional efficiency index of prediction cardiovascular and cerebrovascular disease.

Plan-validation:

By the segmentation result of (3) (4) (5) and manual segmentation results contrast, calculate DSC (Dice Similarity Coefficient), MAD (Mean Absolute Distance) and the MAXD (Maximum Absolute Distance) of three-view diagram segmentation result, carry out proof of algorithm evaluation and application.As shown in Figure 22, Figure 24, Figure 38 and table 1:

A, similarity coefficient DSC:

Similarity coefficient DSC is a kind of standard of measuring partitioning algorithm, for evaluating the similarity of area.

$\mathrm{DSC}=2\frac{|R_M\&cap;R_A|}{\left|R_M\right|+\left|R_A\right|}$

Wherein, R _mand R _arespectively represent manual contours around region and automatic profile around region.Similarity coefficient DSC, more close to 1, illustrates that the degree of closeness of two methods is higher.

B, average absolute distance MAD and maximum absolute distance MAXD

Average absolute distance MAD and maximum absolute distance MAXD, is used to the standard of measuring partitioning algorithm, for evaluating Distance conformability degree.

Suppose that manual contours point is { m _i: i=1, wide point is as the criterion, and corresponding point is automatically { a _i: i=1, point is.Then

$\mathrm{MAD}=\frac{1}{k}\underset{i=1}{\overset{k}{\mathrm{\&Sigma;}}}\left|d(m_i,A)\right|$

$\mathrm{MAXD}=\underset{i\&Element;\{1,k\}}{\max }\left\{\right|d(m_i,A)\left|\right\}$

Wherein, d (m _i, A) for manual contours and automatic profile corresponding point between relative distance.Average absolute distance MAD and maximum absolute distance MAXD is less, illustrates that the degree of closeness of two methods is higher.

Table 1 the present invention and manual segmentation results contrast

	DSC（%）	MAD(mm)	MAXD(mm)
				Two dimension transversal section inner-con-tour extraction blood vessel	94.1±3.6	0.32±0.21	0.77±0.47
Two dimension transversal section Outside contour extraction blood vessel	92.8±2.5	0.37±0.19	0.84±0.39
				Two dimension sagittal plane IMT detect thickness		0.73±0.47	1.01±0.75
Two-dimensional coronal face IMT detect thickness		1.02±0.76	1.89±0.34

In upper table, detect thickness average, standard deviation, unit is mm.

As can be seen from the table, the dividing method adopted in (1) the present invention, suitable with expert's manual segmentation method, demonstrate its feasibility and robustness; (2) in the segmentation of sequence image of two-dimentional transversal section, the inventive method and expert's manual segmentation method have higher similarity, and similarity coefficient DSC is all greater than 90%, and average absolute distance MAD and maximum absolute distance MAXD is all in error range; (3) in the IMT of two-dimentional sagittal plane and coronal-plane measures, two methods all with manual measurement IMT goldstandard quite, average error is all in 1 millimeter; (4) two-dimentional sagittal measurement result is better than two-dimensional coronal face, and the quantity of this and sequence image, quality, concrete dividing method all have certain relation.In addition, compare from the running time, the present invention can complete the measurement of a three-dimensional data at 3 minutes; Veteran doctor then often needs 8 minutes to complete manual measurement.

Listed exemplifying embodiment only have expressed several embodiment of the present invention above, and it describes comparatively concrete and detailed, but can not therefore be construed as limiting the scope of the patent.Should be understood that and it is emphasised that, to the person skilled of this area, under the premise of not departing from the present invention, can also have various deformation and improvement, these all should belong to protection scope of the present invention.Therefore, the protection domain of patent of the present invention should be as the criterion with claims.

Those skilled in the art will readily understand; the foregoing is only preferred embodiment of the present invention; not in order to limit the present invention, all any amendments done within the spirit and principles in the present invention, equivalent replacement and improvement etc., all should be included within protection scope of the present invention.

Claims (6)

1. the main carotid artery vascular based on neck ultrasonoscopy extracts and method for measuring thickness, comprises the following steps:

Read neck ultrasonic three-dimensional volume data, mark main carotid artery vascular central shaft based on main carotid bifuracation point;

According to central shaft, by three-view diagram direction projection, cutting neck ultrasonic three-dimensional volume data, obtains two-dimentional transversal section, sagittal plane and coronal-plane sequence image;

In each image of two-dimentional transversal section sequence image, choose each main carotid artery vascular area-of-interest respectively, and to the pre-service of each main carotid artery vascular area-of-interest; In each main carotid artery vascular area-of-interest after the pre-treatment, segmentation obtains the inside and outside profile of each main carotid artery vascular respectively; According to the inside and outside profile of the main carotid artery vascular area-of-interest of each two-dimentional cross-sectional image, by its spatial relation three-dimensional reconstruction, obtain three-dimensional main carotid artery vascular profile;

In each image of two-dimentional sagittal plane sequence image, choose each main carotid artery vascular area-of-interest respectively, and to the pre-service of each main carotid artery vascular area-of-interest; In each main carotid artery vascular area-of-interest after the pre-treatment, segmentation obtains the inside and outside profile of each main carotid artery vascular respectively; Thickness between the inside and outside profile calculating each main carotid artery vascular respectively and Internal-media thickness; Add up the Internal-media thickness average of each two-dimentional sagittal view picture;

In each image of two-dimensional coronal face sequence image, choose each main carotid artery vascular area-of-interest respectively, and to the pre-service of each main carotid artery vascular area-of-interest; In each main carotid artery vascular area-of-interest after the pre-treatment, segmentation obtains the inside and outside profile of each main carotid artery vascular respectively; Thickness between the inside and outside profile calculating each main carotid artery vascular respectively and Internal-media thickness; Add up the Internal-media thickness average of each two-dimensional coronal face image;

In main carotid artery vascular area-of-interest in each image of described two-dimentional transversal section sequence image, segmentation obtains the inside and outside profile of each main carotid artery vascular in the following manner:

A1, in the pretreated main carotid artery vascular area-of-interest of two-dimentional cross-sectional image, extract the Internal periphery of main carotid artery vascular, and do smoothing processing;

A2, carry out ellipse fitting according to the Internal periphery of main carotid artery vascular, after ellipse matching obtained amplifies, as the initial outline of main carotid artery vascular;

A3, initial outline according to main carotid artery vascular, develop and obtain the final outline of main carotid artery vascular;

In main carotid artery vascular area-of-interest in each image of described two-dimentional sagittal plane sequence image, segmentation obtains the inside and outside profile of each main carotid artery vascular in the following manner:

B1, the point arranged as each in pretreated main carotid artery vascular area-of-interest two-dimentional sagittal view, carry out sequence number mark from top to bottom successively, as the index value of each row point;

The coarse segmentation of B2, area-of-interest:

B21, acquisition lower boundary profile DB: from left to right, order scans each row of the main carotid artery vascular area-of-interest of two-dimentional sagittal view picture, obtains maximum gradation value of each row, minimum gradation value and meets the lower boundary candidate point of gray threshold condition; In lower boundary candidate point, the maximum point of marked index value is as unique lower boundary point; Connect all lower boundary point in turn and form lower boundary profile DB; Wherein, lower boundary candidate point gray-scale value GV _candidatethe gray threshold condition that should meet is: GV _candidate>=a × Δ GV+GV _min=a × (GV _max-GV _min)+GV _min, the span of weight coefficient a is 0.87 ~ 0.93;

B22, acquisition coboundary profile UB: lower boundary profile DB is upwards moved in parallel 20 ~ 30 pixels as coboundary temporary profile UB_t in the main carotid artery vascular area-of-interest of two-dimentional sagittal view picture; Choose the template that size is X × X, X gets 8 ~ 15 pixels, utilizes template from left to right, travels through from top to bottom, the pixel average EX of calculation template overlay area and variance DX to the region between coboundary temporary profile UB_t and lower boundary profile DB; If meet mean variance threshold condition EX≤b and DX≤c, b span is 0.05 ~ 0.09, c span is 0.11 ~ 0.15, then the center of template overlay area is coboundary candidate point; If do not meet mean variance threshold condition, then corresponding UB_t point is coboundary candidate point; In the coboundary candidate point of each row, marked index value the maximum is as unique coboundary point; Connect all coboundaries point in turn and form coboundary profile UB;

The segmentation of B3, area-of-interest is cut: the region segmentation between lower boundary profile DB and coboundary profile UB cuts out beyond lumen of vessels, inner membrance and middle film, adventitia and adventitia and organizes;

The interface of B4, extraction lumen of vessels and inner membrance is Internal periphery, and in extraction, the interface of film and adventitia is outline;

In main carotid artery vascular area-of-interest in each image of described two-dimensional coronal face sequence image, segmentation obtains the inside and outside profile of each main carotid artery vascular in the following manner:

Initialization Internal periphery in C1, the main carotid artery vascular area-of-interest of employing Mathematical Morphology Method after the Image semantic classification of two-dimensional coronal face, develops according to initialization Internal periphery and generates final Internal periphery;

C2, to be moved down obtain initial outline by initial Internal periphery, developing according to initial outline generates final outline.

2. the main carotid artery vascular based on neck ultrasonoscopy according to claim 1 extracts and method for measuring thickness, it is characterized in that, described steps A 1 is specially: in the pretreated main carotid artery vascular area-of-interest of two-dimentional cross-sectional image, extract home position and the radius of main carotid artery vascular, extract Internal periphery according to home position and radius.

3. the main carotid artery vascular based on neck ultrasonoscopy according to claim 1 extracts and method for measuring thickness, and it is characterized in that, in described steps A 2, oval magnification ratio coefficient is 1.02 ~ 1.08.

4. the main carotid artery vascular based on neck ultrasonoscopy according to claim 1 extracts and method for measuring thickness, it is characterized in that, the evolution method that described steps A 3 adopts be Active Contour Model, GVF-Snake, Fast Marching Method, Level Set, Sparse Field Algorithm any one.

5. main carotid artery vascular according to claim 1 extracts and method for measuring thickness, it is characterized in that, adopt in C average, fuzzy C-mean algorithm, Support Vector Machine SVM, BP, self-adaptive BP, AdaBoost algorithm in the thin segmentation step of described area-of-interest any one.

6. main carotid artery vascular according to claim 1 extracts and method for measuring thickness, it is characterized in that, the evolution method that described step C1 and C2 adopt is any one in classical S-Shaped Algorithm, GVF-Snake algorithm, level set, MS model, CV model evolution method.