CN104867147A - SYNTAX automatic scoring method based on coronary angiogram image segmentation - Google Patents

Abstract

An SYNTAX automatic scoring method based on coronary angiogram image segmentation segments and extracts a coronary arterial tree and central line based on coronary angiogram images of a patient and employs a blood vessel side branch separation, reconnection and K main curve method, and generates a binary image of the coronary arterial tree The blood vessel segment numbers are automatically matched based on the SYNTAX scoring standard by employing a central line direction tracking method, and coronary artery pathological changes and adverse characteristics are identified and scored by selecting a coronary left (right) advantageous type in combination with the width attenuation degree of the cross section of the blood vessel and the curvature change of the center line, and the scores are calculated to give a reasonable score result to allow a doctor to employ optimal operation scheme. The method overcomes the problems of great computation amount, tedious calculating steps, complex scoring standard and long calculating time of artificial calculating SYNTAX scoring in conventional clinics.

Description

SYNTAX automatic scoring method based on coronary angiography image segmentation

Technical Field

The invention relates to segmentation and extraction of a coronary angiography image vessel tree, and belongs to the technical field of clinical medical image processing.

Background

Cardiovascular disease, one of the highly fatal diseases that endanger human health, is usually caused by stenosis and blockage of coronary arteries. Coronary angiography is an important method for diagnosing heart diseases, and is the "gold standard" for clinical diagnosis of vascular diseases. Although the brightness of blood vessels in the contrast image is higher than that of other tissues such as the heart due to the injection of the contrast agent, the gray value of the image is inconsistent due to the uneven distribution of the contrast agent and the influence of environmental noise in the imaging process; in addition, the blood vessel branches are numerous, the shape is complex, and the gray scale contrast of the far-end blood vessel is very low, so that the accuracy of manual diagnosis in clinic is greatly influenced. In order to improve the clinical diagnosis level, it is necessary to automatically extract and segment the coronary artery vessel tree from the contrast image. Meanwhile, due to individual difference between people and complexity of human anatomy tissue structures, researchers at home and abroad propose a plurality of medical image segmentation methods, each method has unique segmentation advantages, but the application of a single segmentation method cannot meet the segmentation requirements of coronary angiography images. The method combines a plurality of image processing and segmentation methods, and firstly adopts a Frangi method based on a Hessian matrix to enhance a coronary angiography image; then, according to the pixel connectivity of the blood vessel intersection, the intersection is removed, and the main trunk and each branch of the coronary blood vessel are separated; then, according to the clustering principle, the separated blood vessel trunk and each branch are reconnected; and finally, extracting the centerline of the coronary artery based on a K principal curve method (KPC). The method can effectively segment and extract the coronary artery tree and the central line, retain the detailed characteristics of the blood vessel to the maximum extent, and simultaneously effectively remove noise fragments.

Coronary heart disease is one of the most serious high-grade cardiovascular diseases, and angiogenesis is an important means for treating coronary heart disease, and comprises two types of surgery, namely Percutaneous Coronary Intervention (PCI) and Coronary Artery Bypass Grafting (CABG). The problem of PCI and CABG selection has been debated in the angioplasty-type selection of a segment of coronary lesions. Therefore, how to correctly select the operation formula by the clinician, the success rate is improved, and the complications are reduced, so that a method for guiding is urgently needed by the internal and external departments. The narrow coronary Study (SYNTAX) of cardiac surgery and intervention published in the european cardiology society (ESC)2008 was selected and designed to address the above issues. The important contribution of the SYNTAX research is to provide a concept of SYNTAX score and establish a SYNTAX scoring system, which is a new scoring system for risk stratification according to anatomical characteristics of coronary artery lesion, quantitatively evaluates the complexity of coronary artery lesion according to anatomical characteristics such as lesion position, severity, bifurcation, calcification and the like, and is used as a preliminary judgment means for surgical mode selection. The SYNTAX scoring method was developed based on the following coronary lesion scoring and grading system: improved American Heart Association (AHA) coronary vessel segment Classification in Artificial Revascularization Therapy Study (ARTS) randomised triple (ARTS study); leaman score; 3. american cardiovascular association/american heart association (ACC/AHA) lesion grading system; 4. a total occlusion typing system; duke and Institute cardio pulmonary disease split (ICPS) bifurcation lesion typing system; 6. and (5) expert opinion. The SYNTAX scoring system scores the lesions with the diameter of more than or equal to 1.5mm and the stenosis degree of more than or equal to 50 percent by adopting a coronary artery tree 16 segmentation method and combining the advantage distribution, the lesion part, the stenosis degree and the lesion characteristics of the coronary artery. The scoring system comprises the advantages, disease variables, involved segments and lesion characteristics (complete occlusion, three-bifurcation lesion, two-bifurcation lesion, aortic opening lesion, severe tortuosity, lesion length larger than 20mm, severe calcification, thrombus and diffuse lesion/small vessel lesion), and the total score after scoring each lesion is the SYNTAX score. Wherein, for a plurality of lesions in one segment, if the distance between adjacent lesions is less than 3 times of the reference diameter, the lesion is taken as a lesion score, and if the distance between adjacent lesions is more than or equal to 3 times of the reference diameter, the lesion score is 2.

At present, a clinician manually calculates the SYNTAX score based on a coronary angiography image, and the scoring program is complex and the calculation amount is large, so that the workload of the clinician is increased, and therefore the SYNTAX score is not widely applied to clinic. In order to better popularize the SYNTAX scoring to the clinic and solve the problems of complex program, large calculation amount and the like caused by manual scoring, it is necessary to develop an automatic SYNTAX scoring method.

In recent years, patent proposals have been made on methods for segmenting and extracting blood vessels from coronary angiography images, but none of these methods have been combined with SYNTAX scores, and cannot provide a clinically direct score of the degree of coronary lesions with reference. Meanwhile, related patents of a SYNTAX scoring system and patents combining coronary angiography image segmentation and SYNTAX scoring do not exist in China. Therefore, the establishment of the grading process and the grading method integrating the automatic segmentation of the coronary angiography image vessel tree and the automatic grading of SYNTAX has very important clinical significance.

Disclosure of Invention

Aiming at the problems of complex scoring program, large calculation amount, long time consumption, no combination with a coronary angiography image segmentation technology and the like existing in the current clinical SYNTAX manual scoring, the invention provides a SYNTAX automatic scoring method based on coronary angiography image segmentation, which comprises the following steps: the system comprises a coronary angiography image reading and displaying module, a coronary angiography image vessel segmentation module (comprising a coronary angiography image vessel enhancement module, a coronary angiography image vessel and branch identification module, a vessel center line extraction module and a vessel diameter calculation module), a SYNTAX automatic scoring module (comprising a coronary tree segment marking module, a lesion part and lesion feature identification module and a SYNTAX automatic scoring module).

The invention adopts the following technical scheme to achieve the aim.

In the method of the invention, the coronary angiography image reading and displaying module reads and displays the coronary angiography images of the liver and the right oblique position. At present, the projection positions of clinical coronary angiography include a left position, a right position, a left anterior oblique position, a right anterior oblique position, a head position, a foot position, a right shoulder position (right anterior oblique position + head position), a left shoulder position (left anterior oblique position + head position), a liver position (right anterior oblique position + foot position), a spider position (left anterior oblique position + foot position), and the like, and different projection positions are used for observing different coronary artery branches. Wherein, the coronary angiography image projected by the liver (right oblique position + foot position) can directly observe the main right coronary vessels of right coronary artery, such as the arterial cone branch, the sinoatrial node branch, the right ventricular branch, the posterior branch, the acute branch, the posterior descending branch, and the like; the right oblique coronary angiography image can directly observe main left coronary vessels of the left main trunk, anterior descending branch, circumflex branch, diagonal branch, septal branch, blunt edge branch and the like of the left coronary artery. The coronary angiography images projected at the liver position and the right oblique position respectively greatly satisfy

The information required by the SYNTAX score on the main artery and the branch vessels of the left and the right coronary arteries is obtained, so that the coronary angiography images projected by the two body positions are selected as target images. In the coronary angiography image reading and displaying module, the angiography images of the two body positions are read simultaneously, and the two body positions are manually selected as the target angiography images by adopting a manual interactive interface, so that the target blood vessel information is more intuitively and quickly positioned, and the calculation amount of the blood vessel segmentation step is reduced.

The coronary angiography image blood vessel segmentation module in the method comprises a coronary angiography image blood vessel enhancement module, a coronary angiography image blood vessel and branch identification module thereof, a blood vessel central line extraction module and a blood vessel diameter calculation module. The coronary angiography image blood vessel enhancement module adopts a Frangi method based on a Hessian matrix. The method is a multiscale blood vessel enhancement method proposed by Frangi on the basis of Sato and Lorenz research results, and the probability of local blood vessel occurrence is calculated by using the characteristic value of a Hessian matrix. Unlike the case of the Lorenz method, which considers only two eigenvalues, all eigenvalues are fully considered in the Frangi method and an intuitive geometric interpretation of the vessel characteristic measure is given. Based on the coronary angiography image after the blood vessel enhancement, the coronary angiography image blood vessel and branch identification module thereof carries out image preprocessing, blood vessel branch and trunk separation, blood vessel branch and trunk connection and the like on the enhanced angiography image in sequence. The image preprocessing adopts a mean filtering method to smooth the edges of the enhanced blood vessels, removes some artifacts, and avoids the spikes generated by the edges of the blood vessels in the next blood vessel refining process in advance. After preprocessing, obtaining the outline of the central line of the blood vessel by threshold segmentation and morphological operation of a binary image obtained by blood vessel enhancement, removing the intersection points by utilizing the pixel connectivity of the intersection points of the blood vessel, separating the main trunk and each branch of the coronary blood vessel, and marking. On the basis, in order to more accurately connect the separated blood vessel trunk and each branch to form a complete coronary tree, according to the principle of a clustering method, a similarity function is introduced by extracting the characteristics of pixel positions, directions, end point blood vessel widths and the like of the end points of the separated blood vessel segments, and the separated blood vessel trunk and each branch are reconnected on the basis of the factors of continuity of the blood vessel directions, the distance between the end points of the blood vessel segments, the difference of the blood vessel widths between the end points of the blood vessel segments and the like. And the blood vessel central line extraction module adopts a principal curve theory improved by Kegl to carry out coronary artery skeleton extraction on the image which is obtained after the steps and retains the blood vessel details, namely the blood vessel central line. The blood vessel diameters of all the pixel positions calculated in the steps are arranged in sequence by the blood vessel diameter calculating module, and preparation is made for identifying the coronary lesion characteristics.

The SYNTAX scoring system is mainly used for evaluating the structures of the left dominant coronary artery tree and the right dominant coronary artery tree and is ineffective for the balanced coronary artery tree. The coronary artery tree in the SYNTAX scoring system adopts a 16-segment method, as shown in fig. 1 and fig. 2, the left crown and the right crown of the coronary artery tree of the left dominant type and the right dominant type are respectively numbered and named, and the serial numbers and names of the segments of the coronary artery are as follows: 1. the proximal right coronary artery, 2. the middle right coronary artery, 3. the distal right coronary artery, 4. the posterior right coronary descending branch, 16a. the first branch of the right coronary-posterior side branch, 16b. the second branch of the right coronary-posterior side branch, 16c. the third branch of the right coronary-posterior side branch, 5. the left trunk, 6. the proximal anterior descending branch, 7. the middle anterior descending branch, 8. the apical section of the anterior descending branch, 9. the first diagonal branch, 9a. the first diagonal branch a, 10. the second diagonal branch, 10a. the second diagonal branch a, 11. the proximal convoluted branch, 12. the middle branch, 12a. the first blunt edge branch, 12b. the second blunt edge branch, 13. the distal convoluted branch, 14. the posterior left side branch, 14a. the posterior left side branch a, 14b. the posterior left side branch b, 15 convoluted branch-posterior descending branch. Wherein, the difference of the left and right dominant types of the coronary artery tree is as follows: the left corona presents with circumflex branches-posterior descending branches, while the right corona does not; the right corona presents posterior descending and collateral branches, while the left corona does not. Based on the SYNTAX scoring criteria, different segments of different coronary tree dominance types have different weighting factors in the scoring. The SYNTAX scoring system is mainly used for evaluating the pathological change adverse characteristics of coronary arteries, and the blood vessels with the diameter of more than or equal to 1.5mm and the stenosis degree of more than or equal to 50% are scored by combining the dominant distribution, pathological change positions and the stenosis degree of the coronary arteries. Adverse lesion characterization of coronary arteries in the SYNTAX scoring system was: vessel stenosis (including total and 50-99% stenosis), total occlusion (including greater than 3 months or less detailed occlusion u blunt stumps, bridge side branches, first visible segment after occlusion, side and side branches less than 1.5mm), trifurcate lesions (including 1 lesion segment, 2 lesion segments, 3 lesion segments, and 4 lesion segments), bifurcation lesions (including A, B, C lesions, E, D, F, G lesions, and an angle less than 70 °), open lesions, severe tortuosity, length greater than 20mm, severe calcification, thrombosis, and diffuse lesions or small vessel lesions, and coronary artery lesions with different poor feature scores. The coronary 16 segments and the ill lesion characteristics are obtained based on a SYNTAX scoring system.

In a SYNTAX automatic scoring module, based on a binary image of a coronary artery vessel tree obtained by segmentation, according to a coronary artery tree 16 segmentation method of a SYNTAX scoring system, in combination with end points and directions of marked vessel segments in the separation steps of a coronary artery anatomical structure and a vessel branch, a central line tracking method is adopted to mark two contrast image segments of a liver position and a right oblique position respectively, wherein the liver position contrast image provides right coronary artery segment information, and the right oblique position contrast image provides left coronary artery segment information; based on the pathological change characteristics of the coronary artery, the pathological change characteristics of the blood vessel with the diameter of more than or equal to 1.5mm and the stenosis degree of more than or equal to 50 percent are identified by combining the change of the diameter of the blood vessel and the curvature of the central line of the blood vessel caused by the factors such as the dominant distribution, the pathological change part, the stenosis degree and the pathological change characteristics of the coronary artery. Based on the SYNTAX scoring standard, the weighting factor of the lesion part and the scoring of the lesion characteristic are combined, a reasonable scoring result is given to the coronary artery radiography image so as to evaluate the serious condition of coronary artery stenosis and assist a doctor to determine the optimal operation mode.

Compared with the prior art, the invention has the following beneficial effects:

(1) the SYNTAX automatic scoring method based on coronary artery angiography image segmentation provided by the invention innovatively combines the blood vessel segmentation technology of the coronary artery angiography image with the SYNTAX scoring method, and the integrated innovation not only greatly reduces the workload and time of a clinician, but also provides reference for assisting the clinician to determine the optimal operation mode.

(2) The coronary artery tree 16-segment automatic marking method and the lesion feature identification method not only avoid the time and labor waste of the existing clinical manual identification of the coronary artery segments and the lesion features, but also innovatively realize the conversion of the segmented coronary artery angiography image to the SYNTAX scoring module.

(3) The SYNTAX automatic scoring method provided by the invention avoids the problems of large calculation amount, complex calculation steps, complex scoring standard, long calculation time and the like of the existing SYNTAX scoring manually calculated clinically, and better promotes the SYNTAX scoring to be clinically applied.

Drawings

FIG. 1 is a left superior view of a coronary artery according to the present invention;

fig. 2 is a right-hand side view of the coronary artery of the present invention.

Detailed Description

A SYNTAX automatic scoring method based on coronary artery angiography image segmentation comprises reading and displaying of coronary artery angiography images in two positions of a liver position and a right oblique position, enhancement of blood vessels of the coronary artery angiography images, identification of coronary artery angiography blood vessels and branches thereof, extraction of blood vessel center lines, calculation of blood vessel diameters, marking of coronary artery tree sections, identification of lesion parts and lesion characteristics and SYNTAX automatic scoring. The detailed implementation flow is as follows:

step 1, simultaneously reading and displaying coronary artery angiography images of two body positions, namely a liver position and a right oblique position according to the two body positions of coronary artery angiography projection.

And 2, simultaneously performing blood vessel enhancement on the target coronary angiography images of the two selected body positions.

And 3, separating the main body and the branch of the enhanced blood vessel, and marking each separated blood vessel segment.

And 4, reconnecting the separated main vessel and the branch to generate a complete coronary artery tree binary image.

And 5, extracting the center line of the blood vessel (namely the blood vessel skeleton) based on the complete coronary artery tree binary image.

And 6, calculating and storing the diameters of the blood vessels at all pixel positions of the image based on the center lines of the blood vessels.

And 7, identifying and tracking the marked vessel segments according to the coronary 16-segment method in the SYNTAX score based on the marked vessel segments in the step 3 and the vessel direction calculated in the step 5.

And 8, based on the marked coronary tree image, combining the blood vessel diameters of all the pixel positions stored in the step 6, and identifying the corresponding blood vessel lesion according to the lesion characteristics and the diameter change of the lesion part in the SYNTAX score.

And 9, automatically giving a reasonable SYNTAX score according to the weight factor occupied by each segment of the coronary artery tree in the SYNTAX score standard and the score of each lesion feature, thereby assisting a doctor to determine an optimal operation mode.

The method for reading and displaying the coronary angiography image with the right oblique position and the liver position in the step 1 comprises the following steps:

a user autonomously selects a target contrast image for guiding the right oblique position and the liver position through a manual interactive interface. The coronary angiography image of the liver position is used for extracting the blood vessel information of the right coronary artery, and the coronary angiography image of the right oblique position is used for extracting the blood vessel information of the left coronary artery.

The method for performing the blood vessel enhancement on the selected target coronary angiography image in the step 2 comprises the following steps:

the Frangi vascular filtering method based on the Hessian matrix effectively enhances the target coronary tree in the contrast image. The method utilizes the eigenvalue of the Hessian matrix to calculate the probability of local blood vessel appearance. All characteristic values are fully considered in the Frangi method and an intuitive geometric interpretation is given to the blood vessel characteristic measure. In the method of Frangi, vessel enhancement is considered as a filtering process that looks for tubular geometry. Since the vessels have different diameters, a measuring scale varying within a certain range is introduced. Let image f (x, y):

$H=\left[\begin{array}{cc}{f}_{xx}& f_{xy}\\ f_{yx}& f_{yy}\end{array}\right]---\left(1\right)$

wherein f (x, y) represents the gray value of a pixel point (x, y) in the image; f. of_xx、f_xy、f_yx、f_yyAnd respectively representing four second order partial differentials of the two-dimensional image f (x, y), namely the gray gradient difference between the pixel point (x, y) and eight adjacent pixel points. And constructing a Hessian matrix H (x, y) by using the second-order gradient value of the gray value of each pixel point (x, y) in the image.

Second partial differential in the X direction:

<math> <mrow> <msub> <mi>f</mi> <mrow> <mi>x</mi> <mi>x</mi> </mrow> </msub> <mo>=</mo> <mfrac> <mrow> <msup> <mo>&part;</mo> <mn>2</mn> </msup> <mi>f</mi> </mrow> <mrow> <mo>&part;</mo> <msup> <mi>x</mi> <mn>2</mn> </msup> </mrow> </mfrac> <mo>=</mo> <mi>f</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>-</mo> <mn>1</mn> <mo>,</mo> <mi>y</mi> <mo>)</mo> </mrow> <mo>+</mo> <mi>f</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>+</mo> <mn>1</mn> <mo>,</mo> <mi>y</mi> <mo>)</mo> </mrow> <mo>-</mo> <mn>2</mn> <mi>f</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>)</mo> </mrow> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>2</mn> <mo>)</mo> </mrow> </mrow> </math>

second partial differential in Y direction:

<math> <mrow> <msub> <mi>f</mi> <mrow> <mi>y</mi> <mi>y</mi> </mrow> </msub> <mo>=</mo> <mfrac> <mrow> <msup> <mo>&part;</mo> <mn>2</mn> </msup> <mi>f</mi> </mrow> <mrow> <mo>&part;</mo> <msup> <mi>y</mi> <mn>2</mn> </msup> </mrow> </mfrac> <mo>=</mo> <mi>f</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>-</mo> <mn>1</mn> <mo>)</mo> </mrow> <mo>+</mo> <mi>f</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>+</mo> <mn>1</mn> <mo>)</mo> </mrow> <mo>-</mo> <mn>2</mn> <mi>f</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>)</mo> </mrow> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>3</mn> <mo>)</mo> </mrow> </mrow> </math>

mixed partial differential in X, Y direction:

<math> <mrow> <msub> <mi>f</mi> <mrow> <mi>x</mi> <mi>y</mi> </mrow> </msub> <mo>=</mo> <msub> <mi>f</mi> <mrow> <mi>y</mi> <mi>x</mi> </mrow> </msub> <mo>=</mo> <mfrac> <mrow> <msup> <mo>&part;</mo> <mn>2</mn> </msup> <mi>f</mi> </mrow> <mrow> <mo>&part;</mo> <mi>x</mi> <mo>&part;</mo> <mi>y</mi> </mrow> </mfrac> <mo>=</mo> <mi>f</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>+</mo> <mn>1</mn> <mo>,</mo> <mi>y</mi> <mo>+</mo> <mn>1</mn> <mo>)</mo> </mrow> <mo>+</mo> <mi>f</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>)</mo> </mrow> <mo>-</mo> <mi>f</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>+</mo> <mn>1</mn> <mo>,</mo> <mi>y</mi> <mo>)</mo> </mrow> <mo>-</mo> <mi>f</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>+</mo> <mn>1</mn> <mo>)</mo> </mrow> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>4</mn> <mo>)</mo> </mrow> </mrow> </math>

due to f_xy＝f_yxH is a real symmetric matrix, so two eigenvalues λ can be used₁、λ₂To construct the enhancement filter. Two eigenvalues λ of the Hessian matrix in two-dimensional image space₁、λ₂Can be calculated from the following equation:

<math> <mrow> <msub> <mi>&lambda;</mi> <mn>1</mn> </msub> <mo>=</mo> <mi>K</mi> <mo>-</mo> <msqrt> <mrow> <msup> <mi>K</mi> <mn>2</mn> </msup> <mo>-</mo> <msup> <mi>Q</mi> <mn>2</mn> </msup> </mrow> </msqrt> <mo>,</mo> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>5</mn> <mo>)</mo> </mrow> </mrow> </math>

<math> <mrow> <msub> <mi>&lambda;</mi> <mn>2</mn> </msub> <mo>=</mo> <mi>K</mi> <mo>+</mo> <msqrt> <mrow> <msup> <mi>K</mi> <mn>2</mn> </msup> <mo>-</mo> <msup> <mi>Q</mi> <mn>2</mn> </msup> </mrow> </msqrt> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>6</mn> <mo>)</mo> </mrow> </mrow> </math>

wherein, K ═ f_xx+f_yy)/2， $Q=\sqrt{f_{xx}{f}_{yy}-f_{xy}{f}_{yx}}.$

Eigenvalue λ of Hessian matrix₁And λ₂The method is used for judging whether a point on the image is an angular point, wherein the angular point refers to a point with severe density change in the image. Definition of | λ₁|≤|λ₂I, the vessel direction is defined by the eigenvalue λ with the smallest absolute value₁Corresponding feature vectorIt is given. Since the blood vessel is highlighted in the contrast image, the local gray scale variation along the blood vessel direction is ideally zero, while the gray scale variation along the cross-sectional direction of the blood vessel is severe. Therefore, the following condition should be satisfied for a two-dimensional vascular structure pixel:

|λ₁|≈0,|λ₁|＜＜|λ₂| (7)

in the coronary artery angiography shooting process, the problems of different blood flow in each blood vessel caused by different resistance positions of the coronary artery vessel wall, non-uniform staining of an angiography image caused by blood vessel overlapping and projection imaging of ribs and vertebrae and the like are extremely obvious in small blood vessel branches at the far end of the blood vessel in the angiography image. Also, due to the variation in the diameter of the coronary arteries, it is not appropriate to use a single scale of enhancement. In order to enhance the contrast of the target small blood vessel and eliminate the non-uniform dyeing generated by projection, a Gaussian function is introduced to construct a multi-scale filter based on the change of the diameter of the blood vessel, and different scales are adopted for enhancing filtering to compensate the pixel chromaticity difference among the blood vessels.

And combining the differential operation of the Hessian matrix with a Gaussian function, and changing the standard offset of the Gaussian function to obtain linear enhancement filtering under different scales of sigma. According to the convolution property of the gaussian function, the inverse scale space g is obtained by the convolution of the input coronary angiogram image with the second derivative of the gaussian filter:

<math> <mrow> <mi>g</mi> <mo>=</mo> <mi>I</mi> <mo>&CircleTimes;</mo> <mfrac> <mrow> <msup> <mo>&part;</mo> <mn>2</mn> </msup> <mi>G</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>;</mo> <mi>&sigma;</mi> <mo>)</mo> </mrow> </mrow> <mrow> <mo>&part;</mo> <mi>a</mi> <mo>&part;</mo> <mi>b</mi> </mrow> </mfrac> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>8</mn> <mo>)</mo> </mrow> </mrow> </math>

where I is a given coronary angiography image, G is a Gaussian function in the two-dimensional image, and the expression of the Gaussian function G is:

<math> <mrow> <mi>G</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>;</mo> <mi>&sigma;</mi> <mo>)</mo> </mrow> <mo>=</mo> <mfrac> <mn>1</mn> <mrow> <mn>2</mn> <msup> <mi>&pi;&sigma;</mi> <mn>2</mn> </msup> </mrow> </mfrac> <mi>exp</mi> <mrow> <mo>(</mo> <mo>-</mo> <mfrac> <mrow> <msup> <mi>x</mi> <mn>2</mn> </msup> <mo>+</mo> <msup> <mi>y</mi> <mn>2</mn> </msup> </mrow> <mrow> <mn>2</mn> <msup> <mi>&sigma;</mi> <mn>2</mn> </msup> </mrow> </mfrac> <mo>)</mo> </mrow> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>9</mn> <mo>)</mo> </mrow> </mrow> </math>

wherein x is the distance on the horizontal axis from the initial point; y is the distance on the vertical axis from the initial point; σ is the standard deviation of the gaussian distribution, which is a spatial scale factor.

On the premise that equation (7) is satisfied, the formula of the franli two-dimensional tubular filter at a certain scale σ is as follows:

<math> <mrow> <msub> <mi>R</mi> <mi>B</mi> </msub> <mo>=</mo> <mfrac> <msub> <mi>&lambda;</mi> <mn>2</mn> </msub> <msub> <mi>&lambda;</mi> <mn>1</mn> </msub> </mfrac> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>11</mn> <mo>)</mo> </mrow> </mrow> </math>

<math> <mrow> <mi>S</mi> <mo>=</mo> <mo>|</mo> <mo>|</mo> <mi>H</mi> <mo>|</mo> <msub> <mo>|</mo> <mi>F</mi> </msub> <mo>=</mo> <msqrt> <mrow> <msubsup> <mi>&lambda;</mi> <mn>1</mn> <mn>2</mn> </msubsup> <mo>+</mo> <msubsup> <mi>&lambda;</mi> <mn>2</mn> <mn>2</mn> </msubsup> </mrow> </msqrt> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>12</mn> <mo>)</mo> </mrow> </mrow> </math>

where σ, β and c are constant factors for regulating R_BAnd sensitivity of S such that voxels satisfying equation (7) have the greatest feedback to the filter. In two-dimensional tubular filtering, Frangi introduces two parameters: r_BAnd S. Wherein R is_BFor identifying spherical structures: when the voxel is in a spherical structure, R_BThe maximum value is taken. Meanwhile, since the background pixels have the characteristic that the derivative value thereof is very small, in order to distinguish the background pixels, the measure S is introduced to remove the inherent noise.

The coronary angiography image enhancement steps are as follows:

(1) inputting a gray level image to generate a pixel matrix I;

(2) for each element of I (pixel I)_ij) Executing (3) - (9);

(3) initializing a spatial scale sigma;

(4) if the sigma meets the stop condition, jumping to (9);

(5) at the current scale, element I is calculated_ijConvolution with a second order differential of a gaussian function;

(6) generating Hessian matrix H and calculating characteristic value lambda₁And λ₂；

(7) Substituting the characteristic value into Frangi filtering formula (4) and calculating;

(8) after the calculation of all elements under the current scale is finished, carrying out sigma iteration (step), skipping (4) and calculating the next scale;

(9) after traversing all scales, the scale iteration is finished, and the maximum value of each element under each scale is recorded to obtain a final result;

(10) and finishing the traversal of the image pixels and outputting the enhanced image.

The method for separating the main vessel from the branch of the enhanced vessel in the step 3 is as follows:

firstly, preprocessing an enhanced target blood vessel, smoothing the edges of the blood vessel by adopting a mean filtering method, and eliminating spikes possibly generated at the edges of the blood vessel in the next blood vessel segment connection process;

secondly, a threshold value is set for filtering background information in the coronary binary image obtained after image enhancement because the background intensity is lower than the blood vessel intensity. The pixels of all signals (i.e. the target coronary tree) are set to white and the other parts are black.

And thirdly, obtaining a rough central line according to morphological operation. Where the intersection of the vessels is considered to have four or more connected adjacent pixels, the pixels that are the intersections are deleted, resulting in vessel segments without any branches. The separated blood vessel segment can almost completely reserve the related information of the tiny blood vessels, and lays a foundation for obtaining a more complete coronary artery tree.

Finally, each independent and unbranched vessel segment is labeled sequentially. Based on the binary image composed of only "1" pixels (foreground points, i.e. target blood vessels) and "0" pixels (background points) obtained in the first three steps, the mutually adjacent "1" value pixels (target blood vessel pixels) are combined into a region and are mutually connected, and each connected region (i.e. blood vessel segment) is described by boundary information. Two scans were performed on the binary image: scanning pixels row by row and column by column for the first time, judging the adjacent relation between the pixels, and endowing the pixels belonging to the same connected region with the same connected label; the second scan is performed to eliminate the sub-regions of the repeatedly marked connected regions, merging sub-regions belonging to the same connected region but having different marks. Thereby enabling sequential labeling of each independent and unbranched vessel segment (i.e., connected region of "1" pixels).

The method for connecting the separated coronary vessel segments in the step 4 comprises the following steps:

the connection of the vessel segment end points without branches is regarded as a clustering problem to be solved.

First, three features of the centerline endpoint of each vessel segment need to be extracted:

(1) position x ═ x, y: pixel coordinates of the end points of the vessel segments;

(2) direction of vessel endpoints: determining the direction of the blood vessel segment by adopting a Gabor filter, wherein because the filtering process of each marked blood vessel segment is relatively independent, the interference possibly generated at the intersection point can be effectively inhibited;

<math> <mrow> <mi>g</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>;</mo> <mi>&theta;</mi> <mo>,</mo> <mi>&psi;</mi> <mo>,</mo> <mi>&sigma;</mi> <mo>,</mo> <mi>&gamma;</mi> <mo>)</mo> </mrow> <mo>=</mo> <mi>exp</mi> <mrow> <mo>(</mo> <mo>-</mo> <mfrac> <mrow> <msup> <msup> <mi>x</mi> <mo>&prime;</mo> </msup> <mn>2</mn> </msup> <mo>+</mo> <msup> <mi>&gamma;</mi> <mn>2</mn> </msup> <msup> <msup> <mi>y</mi> <mo>&prime;</mo> </msup> <mn>2</mn> </msup> </mrow> <mrow> <mn>2</mn> <msup> <mi>&delta;</mi> <mn>2</mn> </msup> </mrow> </mfrac> <mo>)</mo> </mrow> <mi>exp</mi> <mrow> <mo>(</mo> <mi>i</mi> <mo>(</mo> <mn>2</mn> <mi>&pi;</mi> <mfrac> <msup> <mi>x</mi> <mo>&prime;</mo> </msup> <mi>&lambda;</mi> </mfrac> <mo>+</mo> <mi>&psi;</mi> <mo>)</mo> </mrow> <mo>)</mo> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>13</mn> <mo>)</mo> </mrow> </mrow> </math>

x′＝x cosθ+y sinθ,y′＝-x sinθ+y cosθ

wherein λ is the wavelength of the sine factor; θ is the direction of the parallel stripes on the Gabor function normal; ψ is a phase offset; σ is the standard deviation of the Gaussian function; γ is the spatial aspect ratio, which is the aspect ratio of the Gabor kernel gaussian function.

(3) Vessel width of the marked vessel segment end points, i.e. end point diameter: the normal direction of the end point determines the width of the blood vessel, and the sum of the pixels along the two normal directions from the end point is the width.

Secondly, the similarity between the two measured blood vessel segments is utilized, and a similarity function is introduced to determine the continuity of the blood vessel at the tail end of each blood vessel segment. It contains three parts of information:

(1) continuity in direction between the endpoints of adjacent vessel segments: the adjacent end points of the blood vessel segments have directional continuity, which is a key factor for judging whether the blood vessel segments are the same branch blood vessel, and the entering and the outgoing parts of one branch blood vessel can be matched properly according to the directional continuity. Meanwhile, due to the existence of partial difference, the directions of the two parts are difficult to be completely and accurately matched, a range needs to be set, and the directions of the blood vessels are considered to be consistent in the variation range;

(2) the distance between the endpoints; the distance between the end points of the vessel segments is another relatively important reference factor. Introducing energy distribution fields of the distance and the azimuth angle of the end points of the vessel segments. That is, a sector is designed based on the angular range of the Gabor filter, and its radius is the distance between the endpoints. Thus, the following attenuation function is defined:

<math> <mrow> <mi>D</mi> <mi>K</mi> <mrow> <mo>(</mo> <msub> <mi>x</mi> <mi>i</mi> </msub> <mo>,</mo> <msub> <mi>y</mi> <mi>j</mi> </msub> <mo>)</mo> </mrow> <mo>=</mo> <msup> <mi>exp</mi> <mrow> <mo>-</mo> <mrow> <mo>(</mo> <mo>(</mo> <msup> <mi>d</mi> <mn>2</mn> </msup> <mo>(</mo> <msub> <mi>x</mi> <mi>i</mi> </msub> <mo>,</mo> <msub> <mi>y</mi> <mi>j</mi> </msub> <mo>)</mo> </mrow> <mo>)</mo> <mo>/</mo> <msup> <mi>&delta;</mi> <mn>2</mn> </msup> <mo>)</mo> </mrow> </msup> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>14</mn> <mo>)</mo> </mrow> </mrow> </math>

in the formula (d)²(x_i,y_i) Is the euclidean distance between two points and is a scale parameter for controlling the decay rate as the distance increases.

(3) Difference in vessel width between adjacent vessel segment: within a short distance, the rate of change of the width of the blood vessel is small. The difference in vessel width between the two vessel segments is another factor in determining whether they belong to the same vessel.

However, the three factors have different importance, and the importance is the direction, the distance and the width in turn from large to small. Setting a weighted value, i.e. setting p and q to represent two points in the three-dimensional feature space of the end points, has a similarity function as follows:

<math> <mrow> <mi>S</mi> <mrow> <mo>(</mo> <mrow> <msub> <mi>x</mi> <mi>i</mi> </msub> <mo>,</mo> <msub> <mi>y</mi> <mi>j</mi> </msub> <mo>,</mo> <mi>&epsiv;</mi> </mrow> <mo>)</mo> </mrow> <mo>=</mo> <mfenced open='{' close=''> <mtable> <mtr> <mtd> <mrow> <mfrac> <msub> <mi>&omega;</mi> <mn>1</mn> </msub> <mrow> <mi>&Delta;</mi> <mi>&theta;</mi> <mrow> <mo>(</mo> <mrow> <msub> <mi>x</mi> <mi>i</mi> </msub> <mo>,</mo> <msub> <mi>y</mi> <mi>i</mi> </msub> </mrow> <mo>)</mo> </mrow> </mrow> </mfrac> <mo>+</mo> <msub> <mi>&omega;</mi> <mn>2</mn> </msub> <mi>D</mi> <mi>K</mi> <mrow> <mo>(</mo> <mrow> <msub> <mi>x</mi> <mi>i</mi> </msub> <mo>,</mo> <msub> <mi>y</mi> <mi>j</mi> </msub> </mrow> <mo>)</mo> </mrow> <mo>+</mo> <mfrac> <msub> <mi>&omega;</mi> <mn>3</mn> </msub> <mrow> <mi>D</mi> <mi>W</mi> <mrow> <mo>(</mo> <mrow> <msub> <mi>x</mi> <mi>i</mi> </msub> <mo>,</mo> <msub> <mi>y</mi> <mi>j</mi> </msub> </mrow> <mo>)</mo> </mrow> </mrow> </mfrac> </mrow> </mtd> <mtd> <mrow> <mi>&epsiv;</mi> <mi>i</mi> <mi>n</mi> <mi>r</mi> <mi>a</mi> <mi>n</mi> <mi>g</mi> <mi>e</mi> </mrow> </mtd> </mtr> <mtr> <mtd> <mn>0</mn> </mtd> <mtd> <mrow> <mi>o</mi> <mi>t</mi> <mi>h</mi> <mi>e</mi> <mi>r</mi> <mi>w</mi> <mi>i</mi> <mi>s</mi> <mi>e</mi> </mrow> </mtd> </mtr> </mtable> </mfenced> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>15</mn> <mo>)</mo> </mrow> </mrow> </math>

in the formula, ω₁，ω₂And ω₃The importance degree for adjusting the direction, distance and width difference; x is the number of_iIs the ith endpoint; is the angle reasonable variation range of the control direction; DK is the definition of the distance decay function; DW is the difference in vessel width. Therefore, the function can well handle small angles, high curvatures or long distances.

The step 5 is a method for extracting a blood vessel central line (namely a blood vessel skeleton) based on the complete coronary artery tree binary image as follows:

and (4) on the basis of the work in the steps 2, 3 and 4, continuously thinning the extracted rough center line by adopting a K main curve method. First, for a set of data points K, curve f^*A main curve called length L, if on all the clusters of curves of length less than or equal to L, f^*The distance function Δ (f) is minimized.

Wherein,

<math> <mrow> <mtable> <mtr> <mtd> <mrow> <mi>&Delta;</mi> <mrow> <mo>(</mo> <mi>f</mi> <mo>)</mo> </mrow> <mo>=</mo> <mi>E</mi> <mo>=</mo> <mo>&lsqb;</mo> <mi>&Delta;</mi> <mrow> <mo>(</mo> <mi>X</mi> <mo>,</mo> <mi>f</mi> <mo>)</mo> </mrow> <mo>&rsqb;</mo> </mrow> </mtd> </mtr> <mtr> <mtd> <mrow> <mo>=</mo> <mi>E</mi> <mo>&lsqb;</mo> <munder> <mi>inf</mi> <mi>t</mi> </munder> <mo>|</mo> <mo>|</mo> <mi>X</mi> <mo>-</mo> <mi>f</mi> <mrow> <mo>(</mo> <mi>t</mi> <mo>)</mo> </mrow> <mo>|</mo> <msup> <mo>|</mo> <mn>2</mn> </msup> <mo>&rsqb;</mo> <mo>=</mo> <mi>E</mi> <mo>&lsqb;</mo> <mo>|</mo> <mo>|</mo> <mi>X</mi> <mo>-</mo> <mi>f</mi> <mrow> <mo>(</mo> <msub> <mi>t</mi> <mi>f</mi> </msub> <mo>(</mo> <mi>x</mi> <mo>)</mo> </mrow> <mo>)</mo> <mo>|</mo> <msup> <mo>|</mo> <mn>2</mn> </msup> <mo>&rsqb;</mo> </mrow> </mtd> </mtr> </mtable> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>16</mn> <mo>)</mo> </mrow> </mrow> </math>

wherein t is a projection index; f (-) represents the m-dimensional vector function between the projection index and the data point, is a description with a nonlinear relation with the internal distribution variable, and the K main curve always exists as long as the data distribution satisfies the finite second moment.

Next, the shortest first principal component straight line is used as the initial curve f_1,nThen the algorithm is iterated, each iteration taking the curve f obtained in the previous time_i-1,nA new vertex is added to the line to increase the number of segments of the curve. After adding a new vertex, the previous vertex position is updated in an inner loop so that the penalty distance function is minimized, thereby forming a new curve f_1,n. The algorithm stops when k exceeds a threshold. The inner loop consists of a projection step and an optimization step.

Finally, in the projection step, the data points are divided into different regions according to the vertices or line segments of their projections. In the optimization step, the new position of each vertex is determined by the minimum mean square distance, where the mean square distance function is penalized by the local curvature. And (3) repeatedly executing the projection step and the optimization step until convergence, generating a curve, and finally obtaining the centerline of the coronary artery tree (namely the blood vessel skeleton).

The method for identifying the left (right) dominant type of the coronary artery tree and tracking and marking the vessel 16 segment in the step 7 is as follows:

first, based on the labeled blood vessel segments in step 3 and the blood vessel direction calculated in step 5, the labeled "1" pixel connected region (blood vessel segment) is scanned according to the coronary 16 segment method in SYNTAX score. Scanning a coronary angiography image of the liver position from a pixel coordinate (seed point) of the end point of the right coronary vessel segment with the largest diameter value, tracking the vessels which are judged to be the same branch in the step 4 along the direction of the central line from the end point of the vessel segment, sequentially labeling the labeled connected regions according to the serial numbers of the 16 segments of the coronary vessels in the SYNTAX score, and judging whether the coronary tree has a posterior descending branch and a posterior lateral branch (comprising 4. right coronary-posterior descending branch, 16a. right coronary-posterior lateral branch first branch, 16b. right coronary-posterior lateral branch second branch and 16c. right coronary-posterior lateral branch third branch), if so, the coronary tree is a right dominant coronary tree; otherwise, the left dominant coronary artery tree is obtained. Tracking the right oblique coronary angiography image by the method, and judging whether the coronary tree has 15. circumflex branches-posterior descending branches, if so, the coronary tree is a left dominant coronary tree; otherwise, it is the right dominant coronary artery tree.

The method for identifying the vascular lesion in the step 8 is as follows:

in the SYNTAX coronary artery scoring system, vascular stenosis, total occlusion, trigeminal lesion, bifurcation lesion, opening lesion, severe distortion, vascular lesion length, calcification, thrombus, diffuse lesion and other vascular lesions are scored.

First, the degree of vascular stenosis is determined and scored. And (4) judging whether the blood vessel is narrow or not according to the blood vessel width attenuation function with the diameter of more than or equal to 1.5mm obtained in the step (4). If the width of the cross section of the blood vessel segment at a certain position suddenly attenuates to 0 from more than or equal to 1.5mm, namely the cross section of the next adjacent blood vessel along the direction of the central line of the cross section of the blood vessel segment more than or equal to 1.5mm is 0, the blood vessel at the pixel position is judged to be completely occluded, the position is marked, and meanwhile, a popup window is arranged to inquire whether the completely occluded time is more than 3 months or not, and scoring is carried out. If the width attenuation rate of the cross section of the adjacent blood vessel along the blood vessel direction of the cross section width of the blood vessel segment at a certain position is more than or equal to 50 percent, the blood vessel at the pixel position is judged to be narrow, the position is marked, and the score is given.

Second, bifurcation lesions (including trifurcate lesions and bifurcate lesions) are identified and scored. The judgment of the three-fork lesion comprises 3 steps: step 1, knowing the endpoint pixel coordinates of the blood vessel segments obtained in step 4, judging whether the pixel positions of the endpoints of the blood vessel segments with the diameters of more than or equal to 1.5mm are in a circle which is designed in step 4 and takes the radius of a Gabor filter as a threshold value, wherein the circle comprises a main blood vessel segment and 2 side branch blood vessel segments, and judging whether the angles formed by the endpoint directions of the 3 blood vessel segments in pairs are less than 70 degrees; and 2, if the end points of the three blood vessel segments meeting the requirements are in the same circle with the radius as the threshold, continuously judging whether the numbering of the blood vessel side branch segment is as follows: 3/4/16/16a, 5/6/11/12, 11/12a/12b/13, 6/7/9/9a, and 7/8/10/10 a; and 3, if the side branch blood vessel accords with the segment number, continuously judging whether the marked position of the blood vessel stenosis is in the circle with the radius threshold, if so, counting the number of the stenosis (if the distance between adjacent lesions is less than 3 reference diameters, the lesion is taken as a lesion score, otherwise, the lesion is taken as two lesions), and integrating the 1 st step and the 2 nd step to score.

Bifurcate lesions include 3 steps: step 1, knowing the end point pixel coordinates of the blood vessel segments obtained in step 4, judging whether the pixel positions of the end points of the blood vessel segments with the diameters of more than or equal to 1.5mm are in a radius threshold circle based on a Gabor filter designed in step 4, wherein the radius threshold circle comprises a main blood vessel segment and 1 side branch blood vessel segment, and judging whether the angle formed by the end point directions of the 2 blood vessel segments is less than 70 degrees; and 2, if 2 blood vessel segments meeting the requirements are in the same circle with the radius as the threshold, continuously judging whether the numbering of the blood vessel side branch segment is the following side branch: 5/6/11, 6/7/9, 7/8/10, 11/13/12a, 13/14/14a, 3/4/16, and 13/14/15; and 3, if the side branch blood vessel accords with the segment number, continuously judging whether the marked position of the blood vessel stenosis is in the circle with the radius threshold, if so, counting the number of the stenosis (if the distance between adjacent lesions is less than 3 reference diameters, the lesion is taken as a lesion score, otherwise, the lesion is taken as two lesions), and integrating the 1 st step and the 2 nd step to score.

Third, open lesions are identified and scored. Whether more than 50% of angiostenosis exists between the proximal ends of the blood vessel segments of the 1 st segment and the 5 th, 6 th and 11 th segments of the right oblique coronary angiography respectively, namely the cross section of the segment end point to the cross section of the pixel point along the central line direction by 3 mm. Aortic ostial lesions of the coronary artery if segments 1 and 5 are present; if segments 6 and 11 are present, then a circumflex double ostial lesion is present. And scored according to the SYNTAX scoring criteria.

Fourth, vessel distortion is identified and scored. Firstly, based on the blood vessel center line extracted in the step 5, for the pixel points on the center line of the blood vessel segment of each label, the curvature of all the pixel points on the center line is extracted by taking the ratio of the distance between the front point and the rear point of the point and the distance between the point and the whole connecting line as approximate curvature. Secondly, by calculating the curvature of each point on the contour, respective local maximum value points are found in each blood vessel segment and are used as characteristic points. Finally, if a certain segment of the blood vessel has 1 or more characteristic points which are more than or equal to 90 degrees or 3 or more characteristic points which are 45-90 degrees, the segment has blood vessel distortion and is scored.

Fifth, vascular calcification is identified and scored. Firstly, the marked blood vessel segment pixel positions in the two contrast binary images after the liver position and the right oblique position are respectively corresponding to the original gray level image and marked with the same mark. Secondly, by adopting a piecewise stretching function, the total gray level number of the original image is set as M, most gray levels are distributed in the [ a, b ] interval, and only a small part of gray levels are outside the interval, linear transformation can be performed in the [ a, b ] interval as follows

<math> <mrow> <mi>g</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>)</mo> </mrow> <mo>=</mo> <mfenced open = '{' close = ''> <mtable> <mtr> <mtd> <mi>d</mi> </mtd> <mtd> <mrow> <mn>0</mn> <mo>&le;</mo> <mi>f</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>)</mo> </mrow> <mo>&lt;</mo> <mi>a</mi> </mrow> </mtd> </mtr> <mtr> <mtd> <mrow> <mfrac> <mrow> <mi>c</mi> <mo>-</mo> <mi>d</mi> </mrow> <mrow> <mi>b</mi> <mo>-</mo> <mi>a</mi> </mrow> </mfrac> <mo>&lsqb;</mo> <mi>f</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>)</mo> </mrow> <mo>-</mo> <mi>a</mi> <mo>&rsqb;</mo> <mo>+</mo> <mi>d</mi> </mrow> </mtd> <mtd> <mrow> <mi>a</mi> <mo>&le;</mo> <mi>f</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>)</mo> </mrow> <mo>&lt;</mo> <mi>b</mi> </mrow> </mtd> </mtr> <mtr> <mtd> <mi>c</mi> </mtd> <mtd> <mrow> <mi>b</mi> <mo>&lt;</mo> <mi>f</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>)</mo> </mrow> <mo>&lt;</mo> <mi>M</mi> </mrow> </mtd> </mtr> </mtable> </mfenced> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>17</mn> <mo>)</mo> </mrow> </mrow> </math>

Where g (x, y) is the transformed gray value and [ d, c ] is the gray value after stretching the [ a, b ] segment. The gray scale change range of the coronary artery part can be obtained according to the average gray scale histogram of coronary artery angiography, the rest is the background part, the gray scale value is 0, the gray scale stretching is carried out on the original image, and then the calcifications can be obtained by adopting the Otsu method for segmentation. And finally, if the blood vessel segment region is background except the calcifications, judging that the blood vessel segment is calcified, and grading according to the calcifications in the divided binary image.

Sixth, long lesions greater than 20mm are identified and scored. Judging whether a stenosis with the cross section width attenuation of more than or equal to 50 percent exists in a certain blood vessel segment, judging whether the length of the stenosis along the central line direction is more than or equal to 20mm, if so, judging that the lesion is a long lesion with the length of more than or equal to 20mm, and scoring according to a SYNTAX scoring standard.

Seventh, small vessel lesions are identified and scored. Firstly, judging whether a segment area with the cross section width between 1.5mm and 2mm exists in each blood vessel segment; secondly, if the region exists, judging whether the ratio of the central line pixel in the region to the sum of the long central line pixels of the whole blood vessel segment is more than or equal to 75 percent, if so, generating small blood vessel lesion; finally, scoring was performed according to the SYNTAX scoring criteria.

In step 9, the method for automatically scoring the SYNTAX is as follows:

and (4) automatically superposing or multiplying the grading results of the vascular lesion characteristics such as vascular stenosis, total occlusion, trigeminal lesion, bifurcation lesion, opening lesion, severe distortion, vascular lesion length, calcification, thrombus, diffuse lesion and the like in the step (8) according to the weight factor occupied by each section of the coronary artery tree in the SYNTAX grading standard and the lesion degree, and finally giving a SYNTAX grading result so as to assist a doctor to determine the optimal operation mode.

Table 1 shows the weighting factors of the 16 segments of the coronary artery according to the SYNTAX scoring criteria of the present invention

Table 2 shows the pathological adverse feature score of the SYNTAX score criteria of the present invention

Claims (1)

1. The SYNTAX automatic scoring method based on coronary angiography image segmentation is characterized in that: the method comprises the steps of reading and displaying coronary angiography images of a liver position and a right oblique position, enhancing blood vessels of the coronary angiography images, identifying coronary angiography blood vessels and branches thereof, extracting central lines of the blood vessels, calculating the diameters of the blood vessels, marking segments of a coronary tree, identifying lesion parts and lesion characteristics, and automatically scoring SYNTAX; the detailed implementation flow is as follows:

step 1, simultaneously reading and displaying coronary angiography images of a liver position and a right oblique position according to two positions of coronary angiography projection;

step 2, performing blood vessel enhancement on the target coronary angiography images of the two selected body positions at the same time;

step 3, separating the main body and the branch of the enhanced blood vessel, and marking each separated blood vessel segment;

step 4, reconnecting the separated main trunk and branches of the blood vessel to generate a complete coronary artery tree binary image;

step 5, extracting the center line of the blood vessel, namely the blood vessel skeleton, based on the complete coronary artery tree binary image;

step 6, calculating and storing the diameters of the blood vessels at all pixel positions of the image based on the center line of the blood vessels;

step 7, based on the marked blood vessel segments in the step 3 and the blood vessel direction calculated in the step 5, identifying and tracking the marked blood vessel segments according to the coronary 16-segment method in the SYNTAX score;

step 8, based on the marked coronary tree image, combining the blood vessel diameters of all the pixel positions stored in the step 6, and identifying the corresponding blood vessel lesion according to the lesion features and the diameter change of the lesion part in the SYNTAX score;

step 9, automatically giving a reasonable SYNTAX score according to the weight factor occupied by each segment of the coronary artery tree in the SYNTAX score standard and the score of each lesion feature, thereby assisting a doctor to determine an optimal operation mode;

the method for reading and displaying the coronary angiography image of the right oblique position and the liver position in the step 1 is as follows,

a user autonomously selects a target contrast image for introducing a right oblique position and a liver position through a manual interactive interface; wherein, the coronary angiography image of the liver position is used for extracting the blood vessel information of the right coronary artery, and the coronary angiography image of the right oblique position is used for extracting the blood vessel information of the left coronary artery;

the method for performing the blood vessel enhancement on the selected target coronary angiography image in the step 2 comprises the following steps:

the Frangi vessel filtering method based on the Hessian matrix effectively enhances a target coronary tree in a contrast image; the method utilizes the eigenvalue of the Hessian matrix to calculate the probability of local blood vessel appearance; fully considering all characteristic values in the Frangi method, and giving visual geometric explanation to the blood vessel characteristic measure; in the method of Frangi, vessel enhancement is considered as a filtering process that looks for tubular geometry; because the blood vessels have different diameters, a measuring scale which changes within a certain range is introduced; let image f (x, y):

$H=\left[\begin{array}{cc}{f}_{\mathrm{xx}}& f_{\mathrm{xy}}\\ f_{\mathrm{yx}}& f_{\mathrm{yy}}\end{array}\right]---\left(1\right)$

wherein f (x, y) represents the gray value of a pixel point (x, y) in the image; f. of_xx、f_xy、f_yx、f_yyRespectively representing four second order partial differentials of the two-dimensional image f (x, y), namely the gray gradient difference value of the pixel point (x, y) and eight adjacent pixel points; constructing a Hessian matrix H (x, y) by using the second-order gradient value of the gray value of each pixel point (x, y) in the image;

second partial differential in the X direction:

<math> <mrow> <msub> <mi>f</mi> <mi>xx</mi> </msub> <mo>=</mo> <mfrac> <mrow> <msup> <mo>&PartialD;</mo> <mn>2</mn> </msup> <mi>f</mi> </mrow> <msup> <mrow> <mo>&PartialD;</mo> <mi>x</mi> </mrow> <mn>2</mn> </msup> </mfrac> <mo>=</mo> <mi>f</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>-</mo> <mn>1</mn> <mo>,</mo> <mi>y</mi> <mo>)</mo> </mrow> <mo>+</mo> <mi>f</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>+</mo> <mn>1</mn> <mo>,</mo> <mi>y</mi> <mo>)</mo> </mrow> <mo>-</mo> <mn>2</mn> <mi>f</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>)</mo> </mrow> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>2</mn> <mo>)</mo> </mrow> </mrow> </math>

second partial differential in Y direction:

<math> <mrow> <msub> <mi>f</mi> <mi>yy</mi> </msub> <mo>=</mo> <mfrac> <mrow> <msup> <mo>&PartialD;</mo> <mn>2</mn> </msup> <mi>f</mi> </mrow> <msup> <mrow> <mo>&PartialD;</mo> <mi>y</mi> </mrow> <mn>2</mn> </msup> </mfrac> <mo>=</mo> <mi>f</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>-</mo> <mn>1</mn> <mo>)</mo> </mrow> <mo>+</mo> <mi>f</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>+</mo> <mn>1</mn> <mo>)</mo> </mrow> <mo>-</mo> <mn>2</mn> <mi>f</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>)</mo> </mrow> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>3</mn> <mo>)</mo> </mrow> </mrow> </math>

mixed partial differential in X, Y direction:

<math> <mrow> <msub> <mi>f</mi> <mi>xy</mi> </msub> <mo>=</mo> <msub> <mi>f</mi> <mi>yx</mi> </msub> <mo>=</mo> <mfrac> <mrow> <msup> <mo>&PartialD;</mo> <mn>2</mn> </msup> <mi>f</mi> </mrow> <mrow> <mo>&PartialD;</mo> <mi>x</mi> <mo>&PartialD;</mo> <mi>y</mi> </mrow> </mfrac> <mo>=</mo> <mi>f</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>+</mo> <mn>1</mn> <mo>,</mo> <mi>y</mi> <mo>+</mo> <mn>1</mn> <mo>)</mo> </mrow> <mo>+</mo> <mi>f</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>)</mo> </mrow> <mo>-</mo> <mi>f</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>+</mo> <mn>1</mn> <mo>,</mo> <mi>y</mi> <mo>)</mo> </mrow> <mo>-</mo> <mi>f</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>+</mo> <mn>1</mn> <mo>)</mo> </mrow> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>4</mn> <mo>)</mo> </mrow> </mrow> </math>

due to f_xy＝f_yxH is a real symmetric matrix, so two eigenvalues λ can be used₁、λ₂To construct an enhancement filter; two eigenvalues λ of the Hessian matrix in two-dimensional image space₁、λ₂Can be calculated from the following equation:

<math> <mrow> <msub> <mi>&lambda;</mi> <mn>1</mn> </msub> <mo>=</mo> <mi>K</mi> <mo>-</mo> <msqrt> <msup> <mi>K</mi> <mn>2</mn> </msup> <mo>-</mo> <msup> <mi>Q</mi> <mn>2</mn> </msup> </msqrt> <mo>,</mo> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>5</mn> <mo>)</mo> </mrow> </mrow> </math>

<math> <mrow> <msub> <mi>&lambda;</mi> <mn>2</mn> </msub> <mo>=</mo> <mi>K</mi> <mo>-</mo> <msqrt> <msup> <mi>K</mi> <mn>2</mn> </msup> <mo>-</mo> <msup> <mi>Q</mi> <mn>2</mn> </msup> </msqrt> <mo>,</mo> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>6</mn> <mo>)</mo> </mrow> </mrow> </math>

wherein, K ═ f_xx+f_yy)/2， $Q=\sqrt{f_{\mathrm{xx}}{f}_{\mathrm{yy}}-f_{\mathrm{xy}}{f}_{\mathrm{yx}}};$

Eigenvalue λ of Hessian matrix₁And λ₂The method is used for judging whether a point on the image is an angular point, wherein the angular point refers to a point with violent density change in the image; definition of | λ₁|≤|λ₂I, the vessel direction is defined by the eigenvalue λ with the smallest absolute value₁Corresponding feature vectorGiving out; because the blood vessel is highlighted in the contrast image, the local gray scale change along the blood vessel direction is zero under the ideal condition, and the gray scale change along the section direction of the blood vessel is severe; therefore, the following condition should be satisfied for a two-dimensional vascular structure pixel:

|λ₁|≈0,|λ₁|＜＜|λ₂| (7)

in the coronary artery angiography shooting process, the problems of different blood flow in each blood vessel caused by different resistance positions of the coronary artery vessel wall, nonuniform staining of an angiography image caused by blood vessel overlapping and projection imaging of ribs and vertebrae and the like are very obvious in tiny blood vessel branches at the far end of the blood vessel in the angiography image; meanwhile, the diameter of the coronary artery is changed, so that the single-scale enhancement effect is not suitable for being used; in order to enhance the contrast of the target small blood vessel and eliminate the non-uniform dyeing generated by projection, a Gaussian function is introduced to construct a multi-scale filter based on the change of the diameter of the blood vessel, and different scales are adopted for enhancing filtering to compensate the pixel chromaticity difference among the blood vessels;

combining the differential operation of the Hessian matrix with a Gaussian function, and obtaining linear enhancement filtering under different scales sigma by changing the standard offset of the Gaussian function; according to the convolution property of the gaussian function, the inverse scale space g is obtained by the convolution of the input coronary angiogram image with the second derivative of the gaussian filter:

<math> <mrow> <mi>g</mi> <mo>=</mo> <mi>I</mi> <mo>&CircleTimes;</mo> <mfrac> <mrow> <msup> <mo>&PartialD;</mo> <mn>2</mn> </msup> <mi>G</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>;</mo> <mi>&sigma;</mi> <mo>)</mo> </mrow> </mrow> <mrow> <mo>&PartialD;</mo> <mi>a</mi> <mo>&PartialD;</mo> <mi>b</mi> </mrow> </mfrac> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>8</mn> <mo>)</mo> </mrow> </mrow> </math>

where I is a given coronary angiography image, G is a Gaussian function in the two-dimensional image, and the expression of the Gaussian function G is:

<math> <mrow> <mi>G</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>;</mo> <mi>&sigma;</mi> <mo>)</mo> </mrow> <mo>=</mo> <mfrac> <mn>1</mn> <mrow> <mn>2</mn> <msup> <mi>&pi;&sigma;</mi> <mn>2</mn> </msup> </mrow> </mfrac> <mi>exp</mi> <mrow> <mo>(</mo> <mo>-</mo> <mfrac> <mrow> <msup> <mi>x</mi> <mn>2</mn> </msup> <mo>+</mo> <msup> <mi>y</mi> <mn>2</mn> </msup> </mrow> <mrow> <mn>2</mn> <msup> <mi>&sigma;</mi> <mn>2</mn> </msup> </mrow> </mfrac> <mo>)</mo> </mrow> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>9</mn> <mo>)</mo> </mrow> </mrow> </math>

wherein x is the distance on the horizontal axis from the initial point; y is the distance on the vertical axis from the initial point; σ is the standard deviation of the gaussian distribution, which is the spatial scale factor;

on the premise that equation (7) is satisfied, the formula of the franli two-dimensional tubular filter at a certain scale σ is as follows:

<math> <mrow> <msub> <mi>R</mi> <mi>B</mi> </msub> <mo>=</mo> <mfrac> <msub> <mi>&lambda;</mi> <mn>2</mn> </msub> <msub> <mi>&lambda;</mi> <mn>1</mn> </msub> </mfrac> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>11</mn> <mo>)</mo> </mrow> </mrow> </math>

<math> <mrow> <mi>S</mi> <mo>=</mo> <msub> <mrow> <mo>|</mo> <mo>|</mo> <mi>H</mi> <mo>|</mo> <mo>|</mo> </mrow> <mi>F</mi> </msub> <mo>=</mo> <msqrt> <msubsup> <mi>&lambda;</mi> <mn>1</mn> <mn>2</mn> </msubsup> <mo>+</mo> <msubsup> <mi>&lambda;</mi> <mn>2</mn> <mn>2</mn> </msubsup> </msqrt> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>12</mn> <mo>)</mo> </mrow> </mrow> </math>

where σ, β and c are constant factors for regulating R_BAnd S, such that the feedback of voxels satisfying equation (7) to the filter is maximized; in two-dimensional tubular filtering, Frangi introduces two parameters: r_BAnd S; wherein R is_BFor identifying spherical structures: when the voxel is in a spherical structure, R_BObtaining a maximum value; meanwhile, as the background pixel has the characteristic of very small derivative value, in order to distinguish the background pixel, the measure S is introduced to remove the inherent noise;

the coronary angiography image enhancement steps are as follows:

(1) inputting a gray level image to generate a pixel matrix I;

(2) for each element of I (pixel I)_ij) Executing (3) - (9);

(3) initializing a spatial scale sigma;

(4) if the sigma meets the stop condition, jumping to (9);

(5) at the current scale, element I is calculated_ijConvolution with a second order differential of a gaussian function;

(6) generating Hessian matrix H and calculating characteristic value lambda₁And λ₂；

(7) Substituting the characteristic value into Frangi filtering formula (4) and calculating;

(8) after the calculation of all elements under the current scale is finished, carrying out sigma iteration (step), skipping (4) and calculating the next scale;

(9) after traversing all scales, the scale iteration is finished, and the maximum value of each element under each scale is recorded to obtain a final result;

(10) outputting an enhanced image after the traversal of the image pixels is finished;

the method for separating the main vessel from the branch of the enhanced vessel in the step 3 is as follows:

firstly, preprocessing an enhanced target blood vessel, smoothing the edges of the blood vessel by adopting a mean filtering method, and eliminating spikes possibly generated at the edges of the blood vessel in the next blood vessel segment connection process;

secondly, setting a threshold value to filter background information in a coronary binary image obtained after image enhancement because the background intensity of the coronary binary image is lower than the blood vessel intensity; all the pixels of the signal, namely the target coronary tree, are set to be white, and the other parts are black;

thirdly, obtaining a rough center line according to morphological operation; wherein the intersection of the blood vessels is considered to have four or more connected adjacent pixels, and the pixels as the intersection are deleted to obtain a blood vessel segment without any branch; the separated blood vessel segment can almost completely reserve the related information of the tiny blood vessels, and lays a foundation for obtaining a more complete coronary artery tree;

finally, each independent and unbranched vessel segment is sequentially labeled; based on the binary image which is obtained in the first three steps and only consists of '1' pixels and '0' pixels, the mutually adjacent '1' value pixels and the target blood vessel pixels are combined into regions which are mutually communicated, and boundary information is used for describing each communicated region, namely a blood vessel segment; two scans were performed on the binary image: scanning pixels row by row and column by column for the first time, judging the adjacent relation between the pixels, and endowing the pixels belonging to the same connected region with the same connected label; performing scanning for the second time to eliminate sub-areas of the repeatedly marked connected areas, and combining the sub-areas which belong to the same connected area and have different mark numbers; thereby enabling sequential labeling of each independent and unbranched vessel segment;

the method for connecting the separated coronary vessel segments in the step 4 comprises the following steps:

the connection of the end points of each non-branched blood vessel segment is regarded as a clustering problem to be solved;

first, three features of the centerline endpoint of each vessel segment need to be extracted:

(1) position x ═ x, y: pixel coordinates of the end points of the vessel segments;

(2) direction of vessel endpoints: determining the direction of the blood vessel segment by adopting a Gabor filter, wherein because the filtering process of each marked blood vessel segment is relatively independent, the interference possibly generated at the intersection point can be effectively inhibited;

<math> <mrow> <mi>g</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>;</mo> <mi>&theta;</mi> <mo>,</mo> <mi>&psi;</mi> <mo>,</mo> <mi>&sigma;</mi> <mo>,</mo> <mi>&gamma;</mi> <mo>)</mo> </mrow> <mo>=</mo> <mi>exp</mi> <mrow> <mo>(</mo> <mo>-</mo> <mfrac> <mrow> <msup> <mi>x</mi> <mrow> <mo>&prime;</mo> <mn>2</mn> </mrow> </msup> <mo>+</mo> <msup> <mi>&gamma;</mi> <mn>2</mn> </msup> <msup> <mi>y</mi> <mrow> <mo>&prime;</mo> <mn>2</mn> </mrow> </msup> </mrow> <mrow> <mn>2</mn> <msup> <mi>&delta;</mi> <mn>2</mn> </msup> </mrow> </mfrac> <mo>)</mo> </mrow> <mi>exp</mi> <mrow> <mo>(</mo> <mi>i</mi> <mrow> <mo>(</mo> <mn>2</mn> <mi>&pi;</mi> <mfrac> <msup> <mi>x</mi> <mo>&prime;</mo> </msup> <mi>&lambda;</mi> </mfrac> <mo>+</mo> <mi>&psi;</mi> <mo>)</mo> </mrow> <mo>)</mo> </mrow> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>13</mn> <mo>)</mo> </mrow> </mrow> </math>

x′＝xcosθ+ysinθ,y′＝-xsinθ+ycosθ

wherein λ is the wavelength of the sine factor; θ is the direction of the parallel stripes on the Gabor function normal; ψ is a phase offset; σ is the standard deviation of the Gaussian function; γ is the spatial aspect ratio, which is the aspect ratio of the Gabor kernel gaussian function;

(3) vessel width of the marked vessel segment end points, i.e. end point diameter: the normal direction of the end point determines the width of the blood vessel, and the sum of the pixels along the two normal directions from the end point is the width of the blood vessel;

secondly, determining the continuity of the terminal blood vessel of each blood vessel segment by introducing a similarity function according to the measured similarity between the two blood vessel segments; it contains three parts of information:

(1) continuity in direction between the endpoints of adjacent vessel segments: the adjacent end points of the blood vessel segments have directional continuity, which is a key factor for judging whether the adjacent end points are the same branch blood vessel, and the entering and the outgoing parts of one branch blood vessel can be matched properly according to the directional continuity; meanwhile, due to the existence of partial difference, the directions of the two parts are difficult to be completely and accurately matched, a range needs to be set, and the directions of the blood vessels are considered to be consistent in the variation range;

(2) the distance between the endpoints; the distance between the end points of the vessel segments is another relatively important reference factor; introducing energy distribution fields of the distance and the azimuth angle of the end points of the blood vessel segments; namely, a sector is designed based on the angular range of the Gabor filter, and the radius of the sector is the distance between endpoints; thus, the following attenuation function is defined:

<math> <mrow> <mi>DK</mi> <mrow> <mo>(</mo> <msub> <mi>x</mi> <mi>i</mi> </msub> <mo>,</mo> <msub> <mi>y</mi> <mi>j</mi> </msub> <mo>)</mo> </mrow> <mo>=</mo> <msup> <mi>exp</mi> <mrow> <mo>-</mo> <mrow> <mo>(</mo> <mrow> <mo>(</mo> <msup> <mi>d</mi> <mn>2</mn> </msup> <mrow> <mo>(</mo> <msub> <mi>x</mi> <mi>i</mi> </msub> <mo>,</mo> <msub> <mi>y</mi> <mi>j</mi> </msub> <mo>)</mo> </mrow> <mo>)</mo> </mrow> <mo>/</mo> <msup> <mi>&delta;</mi> <mn>2</mn> </msup> <mo>)</mo> </mrow> </mrow> </msup> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>14</mn> <mo>)</mo> </mrow> </mrow> </math>

in the formula (d)²(x_i,y_i) Is the Euclidean distance between two points, is a scale parameter and is used for controlling the attenuation speed increasing along with the distance;

(3) difference in vessel width between adjacent vessel segment: within a short distance, the width change rate of the blood vessel is very small; the difference of the vessel widths of the two vessel segments is another factor for judging whether the two vessel segments belong to the same vessel;

however, the importance of the above three factors is different, and the importance is the direction, the distance and the width from large to small; setting a weighted value, i.e. setting p and q to represent two points in the three-dimensional feature space of the end points, has a similarity function as follows:

<math> <mrow> <mi>S</mi> <mrow> <mo>(</mo> <msub> <mi>x</mi> <mi>i</mi> </msub> <mo>,</mo> <msub> <mi>y</mi> <mi>j</mi> </msub> <mo>,</mo> <mi>&epsiv;</mi> <mo>)</mo> </mrow> <mo>=</mo> <mfenced open='{' close=''> <mtable> <mtr> <mtd> <mfrac> <msub> <mi>&omega;</mi> <mn>1</mn> </msub> <mrow> <mi>&Delta;&theta;</mi> <mrow> <mo>(</mo> <msub> <mi>x</mi> <mi>i</mi> </msub> <mo>,</mo> <msub> <mi>y</mi> <mi>j</mi> </msub> <mo>)</mo> </mrow> </mrow> </mfrac> <mo>+</mo> <msub> <mi>&omega;</mi> <mn>2</mn> </msub> <mi>DK</mi> <mrow> <mo>(</mo> <msub> <mi>x</mi> <mi>i</mi> </msub> <mo>,</mo> <msub> <mi>y</mi> <mi>j</mi> </msub> <mo>)</mo> </mrow> <mo>+</mo> <mfrac> <msub> <mi>&omega;</mi> <mn>3</mn> </msub> <mrow> <mi>DW</mi> <mrow> <mo>(</mo> <msub> <mi>x</mi> <mi>i</mi> </msub> <mo>,</mo> <msub> <mi>y</mi> <mi>j</mi> </msub> <mo>)</mo> </mrow> </mrow> </mfrac> </mtd> <mtd> <mi>&epsiv;in range</mi> </mtd> </mtr> <mtr> <mtd> <mn>0</mn> </mtd> <mtd> <mi>otherwise</mi> </mtd> </mtr> </mtable> </mfenced> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>15</mn> <mo>)</mo> </mrow> </mrow> </math>

in the formula, ω₁，ω₂And ω₃The importance degree for adjusting the direction, distance and width difference; x is the number of_iIs the ith endpoint; is the angle reasonable variation range of the control direction; DK is the definition of the distance decay function; DW is the difference in vessel width; therefore, the function can well handle the conditions of small angle, high curvature or long distance;

the step 5 is based on the complete coronary artery tree binary image, and the method for extracting the center line of the blood vessel, namely the blood vessel skeleton, comprises the following steps:

on the basis of the work in the steps 2, 3 and 4, the extracted rough center line can be continuously thinned by adopting a K main curve method; first, for a set of data points K, curve f^*A main curve called length L, if on all the clusters of curves of length less than or equal to L, f^*Minimizing a distance function Δ (f);

wherein,

<math> <mrow> <mfenced open='' close=''> <mtable> <mtr> <mtd> <mi>&Delta;</mi> <mrow> <mo>(</mo> <mi>f</mi> <mo>)</mo> </mrow> <mo>=</mo> <mi>E</mi> <mo>[</mo> <mi>&Delta;</mi> <mrow> <mo>(</mo> <mi>X</mi> <mo>,</mo> <mi>f</mi> <mo>)</mo> </mrow> <mo>]</mo> </mtd> </mtr> <mtr> <mtd> <mo>=</mo> <mi>E</mi> <mo>[</mo> <munder> <mrow> <mi>inf</mi> <msup> <mrow> <mo>|</mo> <mo>|</mo> <mi>X</mi> <mo>-</mo> <mi>f</mi> <mrow> <mo>(</mo> <mi>t</mi> <mo>)</mo> </mrow> <mo>|</mo> <mo>|</mo> </mrow> <mn>2</mn> </msup> </mrow> <mi>t</mi> </munder> <mo>]</mo> <mo>=</mo> <mi>E</mi> <mo>[</mo> <msup> <mrow> <mo>|</mo> <mo>|</mo> <mi>X</mi> <mo>-</mo> <mi>f</mi> <mrow> <mo>(</mo> <msub> <mi>t</mi> <mi>f</mi> </msub> <mrow> <mo>(</mo> <mi>X</mi> <mo>)</mo> </mrow> <mo>)</mo> </mrow> <mo>|</mo> <mo>|</mo> </mrow> <mn>2</mn> </msup> <mo>]</mo> </mtd> </mtr> </mtable> </mfenced> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>16</mn> <mo>)</mo> </mrow> </mrow> </math>

wherein t is a projection index; f (·) represents an m-dimensional vector function between the projection index and the data point, is a description with a nonlinear relation with an internal distribution variable, and the K main curve always exists as long as the data distribution meets a finite second moment;

next, the shortest first principal component straight line is used as the initial curve f_1,nThen the algorithm is iterated, each iteration taking the curve f obtained in the previous time_i-1,nAdding a new vertex on the line to increase the number of sections of the curve; after adding a new vertex, the previous vertex position is updated in an inner loop so that the penalty distance function is minimized, thereby forming a new curve f_1,n(ii) a The algorithm stops when k exceeds a threshold; the inner loop consists of a projection step and an optimization step;

finally, in the projection step, the data points are divided into different areas according to the projected vertexes or line segments; in the optimization step, determining a new position of each vertex by a minimum mean square distance, wherein a mean square distance function is penalized by a local curvature; repeatedly executing the projection step and the optimization step until convergence, generating a curve, and finally obtaining a centerline of the coronary artery tree, namely the blood vessel skeleton;

the method for identifying the left or right dominant type of the coronary artery tree and tracking and marking the vessel 16 segment in the step 7 is as follows:

firstly, based on each marked blood vessel segment in the step 3 and the blood vessel direction calculated in the step 5, scanning a marked '1' pixel connected region according to a coronary 16-segment method in SYNTAX scoring; scanning a coronary angiography image of a liver position from a pixel coordinate of an end point of a right coronary vessel segment with the largest diameter value, tracking the vessel which is judged to be the same branch in the step 4 along the direction of a central line from the end point of the vessel segment, sequentially labeling the labeled connected regions according to the serial numbers of 16 segments of coronary vessels in the SYNTAX score, and judging whether the coronary tree has a posterior descending branch and a posterior side branch, wherein the method comprises the following steps: 4. right coronal-posterior descending branch, 16. right coronal-posterior descending branch, 16a. right coronal-posterior collateral first branch, 16b. right coronal-posterior collateral second branch, and 16c. right coronal-posterior collateral third branch, if present, right dominant coronary tree; otherwise, the coronary artery tree is the left dominant type coronary artery tree; tracking the right oblique coronary angiography image by the method, and judging whether the coronary tree has 15. circumflex branches-posterior descending branches, if so, the coronary tree is a left dominant coronary tree; otherwise, the coronary artery tree is the right dominant type coronary artery tree;

the method for identifying the vascular lesion in the step 8 is as follows:

grading the vascular lesions such as vascular stenosis, total occlusion, trigeminal lesion, bifurcation lesion, opening lesion, severe distortion, vascular lesion length, calcification, thrombus, diffuse lesion and the like in a SYNTAX coronary artery grading system;

firstly, judging the stenosis degree of a blood vessel and grading; judging whether the blood vessel is narrow or not according to the blood vessel width attenuation function with the diameter of more than or equal to 1.5mm obtained in the step 4; if the width of the cross section of the blood vessel segment at a certain position suddenly attenuates to 0 from more than or equal to 1.5mm, namely the cross section of the next adjacent blood vessel along the direction of the central line of the cross section of the blood vessel segment more than or equal to 1.5mm is 0, judging that the blood vessel at the pixel position is completely occluded, marking the position, simultaneously setting a popup window to inquire whether the completely occluded time is more than 3 months, and scoring; if the width attenuation rate of the cross section of the adjacent blood vessel along the blood vessel direction of the cross section width of the blood vessel segment at a certain position is more than or equal to 50 percent, the blood vessel at the pixel position is judged to be narrow, the position is marked, and scoring is carried out;

secondly, identifying bifurcation lesions comprising trisection bifurcation lesions and double bifurcation lesions, and scoring; the judgment of the three-fork lesion comprises 3 steps: step 1, knowing the endpoint pixel coordinates of the blood vessel segments obtained in step 4, judging whether the pixel positions of the endpoints of the blood vessel segments with the diameters of more than or equal to 1.5mm are in a circle which is designed in step 4 and takes the radius of a Gabor filter as a threshold value, wherein the circle comprises a main blood vessel segment and 2 side branch blood vessel segments, and judging whether the angles formed by the endpoint directions of the 3 blood vessel segments in pairs are less than 70 degrees; and 2, if the end points of the three blood vessel segments meeting the requirements are in the same circle with the radius as the threshold, continuously judging whether the numbering of the blood vessel side branch segment is as follows: 3/4/16/16a, 5/6/11/12, 11/12a/12b/13, 6/7/9/9a, and 7/8/10/10 a; step 3, if the side branch blood vessel accords with the segment number, continuously judging whether the marked position of the blood vessel stenosis is in the circle of the radius threshold, if the marked position of the blood vessel stenosis exists, counting the number of the stenosis, if the distance between adjacent lesions is less than 3 reference diameters, using the stenosis as a lesion score, otherwise, considering the lesion as two lesions, and integrating the 1 st step and the 2 nd step to score;

bifurcate lesions include 3 steps: step 1, knowing the end point pixel coordinates of the blood vessel segments obtained in step 4, judging whether the pixel positions of the end points of the blood vessel segments with the diameters of more than or equal to 1.5mm are in a radius threshold circle based on a Gabor filter designed in step 4, wherein the radius threshold circle comprises a main blood vessel segment and 1 side branch blood vessel segment, and judging whether the angle formed by the end point directions of the 2 blood vessel segments is less than 70 degrees; and 2, if 2 blood vessel segments meeting the requirements are in the same circle with the radius as the threshold, continuously judging whether the numbering of the blood vessel side branch segment is the following side branch: 5/6/11, 6/7/9, 7/8/10, 11/13/12a, 13/14/14a, 3/4/16, and 13/14/15; step 3, if the side branch blood vessel accords with the segment number, continuously judging whether the marked position of the blood vessel stenosis is in the circle of the radius threshold value, if so, counting the number of the stenosis (if the distance between adjacent lesions is less than 3 reference diameters, the lesion is taken as a lesion score, otherwise, the lesion is considered as two lesions), and integrating the 1 st step and the 2 nd step to score;

thirdly, identifying and scoring open lesions; respectively judging whether more than 50% of angiostenosis exists between the proximal ends of the blood vessel segments of the 1 st segment and the 5 th, 6 th and 11 th segments of the right oblique coronary angiography, namely the cross section of the segment end point and the cross section of the pixel point at the position of 3mm along the central line direction; aortic ostial lesions of the coronary artery if segments 1 and 5 are present; a circumflex double ostial lesion if segments 6 and 11 are present; and, scoring it according to the SYNTAX scoring criteria;

fourthly, identifying blood vessel distortion and scoring; firstly, based on the blood vessel central line extracted in the step 5, for the pixel points on the central line of the blood vessel segment of each label, the curvature of all the pixel points on the central line is extracted by taking the ratio of the distance between the front point and the rear point of the pixel point to the distance between the point and the whole connecting line as approximate curvature; secondly, respectively finding out respective local maximum value points in each blood vessel segment by calculating the curvature of each point on the contour, and taking the local maximum value points as characteristic points; finally, if 1 or more characteristic points which are more than or equal to 90 degrees or 3 or more characteristic points which are 45-90 degrees exist in a certain section of blood vessel, the section has blood vessel distortion and is scored;

fifthly, identifying vascular calcification points and scoring; firstly, respectively corresponding the marked blood vessel segment pixel positions in the two contrast binary images after the liver position and the right oblique position are processed to the original gray level image, and marking the blood vessel segment pixel positions as the same marks; secondly, by adopting a piecewise stretching function, the total gray level number of the original image is set as M, most gray levels are distributed in the [ a, b ] interval, and only a small part of gray levels are outside the interval, linear transformation can be performed in the [ a, b ] interval as follows

<math> <mrow> <mi>g</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>)</mo> </mrow> <mo>=</mo> <mfenced open='{' close=''> <mtable> <mtr> <mtd> <mi>d</mi> </mtd> <mtd> <mn>0</mn> <mo>&le;</mo> <mi>f</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>)</mo> </mrow> <mo>&lt;</mo> <mi>a</mi> </mtd> </mtr> <mtr> <mtd> <mfrac> <mrow> <mi>c</mi> <mo>-</mo> <mi>d</mi> </mrow> <mrow> <mi>b</mi> <mo>-</mo> <mi>a</mi> </mrow> </mfrac> <mo>[</mo> <mi>f</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>)</mo> </mrow> <mo>-</mo> <mi>a</mi> <mo>]</mo> <mo>+</mo> <mi>d</mi> </mtd> <mtd> <mi>a</mi> <mo>&le;</mo> <mi>f</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>)</mo> </mrow> <mo>&lt;</mo> <mi>b</mi> </mtd> </mtr> <mtr> <mtd> <mi>c</mi> </mtd> <mtd> <mi>b</mi> <mo>&lt;</mo> <mi>f</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>)</mo> </mrow> <mo>&lt;</mo> <mi>M</mi> </mtd> </mtr> </mtable> </mfenced> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>17</mn> <mo>)</mo> </mrow> </mrow> </math>

Wherein g (x, y) is the transformed gray value, and [ d, c ] is the gray value after stretching the [ a, b ] segment gray value; obtaining the gray scale change range of the coronary artery part according to the average gray scale histogram of coronary artery angiography, stretching the gray scale of the original image by using the background part and the gray scale value of 0 as the rest, and then segmenting by adopting an Otsu method to obtain calcifications; finally, if the area of the blood vessel segment has a background except the calcifications, the blood vessel segment can be judged to be calcified and is correspondingly graded in the divided binary image according to the calcifications;

sixth, long lesions greater than 20mm are identified and scored; judging whether a stenosis with the cross section width attenuation being more than or equal to 50% exists in a certain blood vessel segment, judging whether the length of the stenosis along the central line direction is more than or equal to 20mm, if so, judging that the lesion is a long lesion more than or equal to 20mm, and scoring according to a SYNTAX scoring standard;

seventhly, identifying small vessel lesions and scoring; firstly, judging whether a segment area with the cross section width between 1.5mm and 2mm exists in each blood vessel segment; secondly, if the region exists, judging whether the ratio of the central line pixel in the region to the sum of the long central line pixels of the whole blood vessel segment is more than or equal to 75 percent, if so, generating small blood vessel lesion; finally, scoring according to the SYNTAX scoring criteria;

in step 9, the method for automatically scoring the SYNTAX is as follows:

and (4) automatically superposing or multiplying the grading results of the vascular lesion characteristics such as vascular stenosis, total occlusion, trigeminal lesion, bifurcation lesion, opening lesion, severe distortion, vascular lesion length, calcification, thrombus, diffuse lesion and the like in the step (8) according to the weight factor occupied by each section of the coronary artery tree in the SYNTAX grading standard and the lesion degree, and finally giving a SYNTAX grading result so as to assist a doctor to determine the optimal operation mode.

TABLE 1 weight factors for segments of the coronary 16 segment for the SYNTAX scoring criteria

Adverse characteristics of coronary artery disease Scoring Stenosis of blood vessel Total occlusion *5 -severe lesions (50-99%) *2 Total occlusion (T0) Time greater than 3 months or unknown +1 -blunt stump +1

Lateral branch of self-bridge +1 Total occlusion lesion and first visible segment + 1/each invisible part -side branch less than 1.5mm +1 Pathological changes of the trigeminal nerve -1 lesion segment +3 -2 lesion segments +4 -3 lesion segments +5 -4 lesion segments +6 Bifurcation lesion -A, B, C type lesion +1 -E, D, F, G type lesion +2 -angle less than 70 degrees +1 Open lesion +1 Severe distortion +2 The length is more than 20mm +1 Severe calcification +2 Thrombus +1 Diffuse lesion/small vessel lesion + 1/segment number

TABLE 2 lesion adverse feature score for the SYNTAX score criteria

Adverse characteristics of coronary artery disease Scoring Stenosis of blood vessel Total occlusion *5 -severe lesions (50-99%) *2 Total occlusion (T0) Time greater than 3 months or unknown +1 -blunt stump +1

Lateral branch of self-bridge +1 Total occlusion lesion and first visible segment + 1/each invisible part -side branch less than 1.5mm +1 Pathological changes of the trigeminal nerve -1 lesion segment +3 -2 lesion segments +4 -3 lesion segments +5 -4 lesion segments +6 Bifurcation lesion -A, B, C type lesion +1 -E, D, F, G type lesion +2 -angle less than 70 degrees +1 Open lesion +1 Severe distortion +2 The length is more than 20mm +1 Severe calcification +2 Thrombus +1 Diffuse lesion/small vessel lesion + 1/segment number