WO2015101060A1

WO2015101060A1 - Decomposition and estimation method for multiple motion parameters in single-arm x-ray angiographic image

Info

Publication number: WO2015101060A1
Application number: PCT/CN2014/085727
Authority: WO
Inventors: 张天序; 黄正华; 黄怡宁
Original assignee: 华中科技大学
Priority date: 2013-12-30
Filing date: 2014-09-02
Publication date: 2015-07-09
Also published as: CN103810721A

Abstract

A decomposition and estimation method for multiple motion parameters in a single-arm x-ray angiographic image comprises the following steps: (1) automatically selecting stable vascular structure feature points; (2) automatically tracking the selected vascular structure feature points in a whole angiography image sequence; (3) selecting a sequence ŝ(n) with a length n_s=k*N₁ (k>1) in a point tracking sequence (N₁ being the cycle of heart motion); (4) decomposing ŝ(n) into x-direction motion x(n) and y-direction motion y(n), and separately performing EMD on the x(n) and the y(n), so as to obtain independent motion signals after EMD; and (5) correspondingly classifying the independent signals according to priori physiological knowledge. The method has wide applicability and flexibility, higher safety and operability and better reliability and accuracy.

Description

Multi-motion parameter decomposition estimation method for single-arm x-ray angiography images

[Technical field]

The invention belongs to the field of digital signal processing and medical imaging crossover technology, and particularly relates to a multi-motion parameter decomposition estimation method for single-arm x-ray angiography images.

[Background technique]

The thoracic rhythm is enlarged and reduced to complete inhalation and exhalation. This is the breathing movement. The regular pulsation of the heart itself (heart movement), the movement of capillaries (high-frequency movement), the movement of the breathing, and the movement of the person or the shaking of the bed (translational movement) can cause the overall translational movement of the human heart in three-dimensional space. . In an X-ray angiography system, due to the combined effects of the above-described motions, a two-dimensional translational motion occurs on the contrast surface of the coronary arteries. Therefore, the coronary angiography image records on the one hand the projection of the motion of the heart on a two-dimensional plane, and also superimposes the two-dimensional translational motion of the coronary artery on the contrast surface caused by the respiratory motion, high-frequency motion and translational motion of the human body.

To obtain a two-dimensional angiogram that is closer to the real situation and for three-dimensional reconstruction of the vessel, these movements need to be automatically separated. In the prior art, the multiple motions are separately extracted separately, and one method of extracting the respiratory motion is to perform sequence tracking on the human body by extracting the marker points in advance. According to the characteristics of breathing exercise, people will move other organs in the body together when breathing. It is generally believed that these organs will translate in three dimensions with the movement of the lungs, and their movements are synchronized. Therefore, it is assumed that the motion of the heart caused by the respiratory motion and the motion of the organ adjacent thereto are also coincident in the plane of the contrast image, and some feature points on other tissues outside the heart can be found in the contrast map as the marker points. These points are tracked throughout the sequence to obtain the motion of the points, and then the motion of the points is approximated to the respiratory motion on the two-dimensional projection surface. Another approach also utilizes structural feature points that do not move with the heart, except that the latter records the motion of these points while imaging. Therefore, it is required to select and mark each feature point before the contrast. obviously, Both options are flawed. The applicability of the former is very poor, because there is no guarantee that there are markers in this frame that meet this condition (some feature points on other tissues outside the heart), and it is necessary to find such points (requires The human anatomy is relatively well understood). When the above feature points do not exist in the angiogram, the respiratory motion is difficult to extract. The latter implementation requires a lot of experimental control and is not suitable for general clinical applications. In addition, in the above two methods, whether the human body has other movements during physiological activities cannot be effectively expressed, and the analysis and extraction of other movements may cause secondary injury to the patient.

In addition, there is another method which is realized under the condition of double-arm x-ray contrast. The idea of separating the heart motion and the respiratory motion is: taking two contrast images of different projection angles at the same time, and performing corresponding coronary blood vessels. Three-dimensional reconstruction, obtaining the three-dimensional spatial distribution of blood vessels at this moment. Then, after matching and reconstructing all the contrast pairs in one breathing cycle, a set of three-dimensional structural sequences is obtained, and the spatial displacement vector between them is the respiratory motion. Relatively speaking, relatively reliable respiratory motion estimation results can be obtained by this method, but it cannot be widely applied in practice due to the constraint of dual-arm x-ray contrast conditions, and this method cannot extract heart signals and respiratory signals. Motion signals outside.

[Summary of the Invention]

In view of the deficiencies of the prior art, the present invention aims to propose a multi-motion parameter decomposition estimation method for single-arm x-ray angiography images, which forms a data sequence by tracking structural feature points, and automatically extracts by using empirical mode decomposition (EMD) method. The heart, breathing, high frequency and translational movement of the human body.

In order to achieve the above object, the present invention adopts the following technical solutions:

A method for estimating a multi-motion parameter decomposition of a single-arm x-ray angiography image sequence, comprising the following steps:

(1) selecting a vascular structural feature point;

(2) automatically tracking the selected vascular structural feature points throughout the sequence of contrast images;

(3) Select a sequence of length n _s =k*N ₁ (k>1) in the tracking sequence s(n) of the point

(4) will

Decomposed into motion x(n) in the x direction and motion y(n) in the y direction, and then EMD decomposition is performed on x(n) and y(n), respectively, to obtain independent motion signals after EMD decomposition;

(5) Correspondingly classify each independent signal according to prior physiological knowledge.

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

(1) Compared with simple manual tracking, automatically selecting the combination of vascular structural feature points and EMD to automatically extract each periodic motion and aperiodic motion has wider applicability and flexibility, and is applicable to almost all contrast sequences. Figure

(2) At the same time, the present invention has higher safety and operability than the method of directly setting the marker point near the heart and then tracking by the relevant imaging means. This is because the markers added to tissues in the body are generally invasive, causing more or less damage to the human body itself, and the process of adding, imaging, eliminating, and extracting respiratory movements of the markers is complicated. Inevitable troubles and errors in actual operation;

(3) The feature points selected by the method of the present invention involve the blood vessels of the left and right coronary vessels, and comprehensively take into account the motion information of the left and right coronary vessels, thereby having better reliability and accuracy.

[Description of the Drawings]

The invention will be best understood from the following description, taken in conjunction with the drawings. In the drawings, the same parts may be denoted by the same reference numerals.

Figure 1 is a flow chart of a preferred embodiment of the present invention;

2(a) and 2(b) are respectively a angiogram selected in the embodiment of the present invention and a vascular structure diagram corresponding to the angiogram, and the corresponding projection angle is (-26.5°, -20.9°);

3(a), 3(b), 3(c), 3(d), 3(e), 3(f), 3(g), 3(h), and 3(i), 3(j) are plots of the left-sequence original signal, the high-frequency signal, the cardiac signal, the respiratory signal, and the translational signal on the X-axis and the Y-axis, respectively;

4(a) and 4(b) are respectively a contrast image selected in the embodiment of the present invention and the contrast image The corresponding vascular structure diagram, the corresponding projection angle is (42.3 °, 26.8 °);

Figure 5 (a), 5 (b), Figure 5 (c), 5 (d), Figure 5 (e), 5 (f), Figure 5 (g), 5 (h), Figure 5 (i), 5(j) is a plot of the original sequence of the right sequence, the high frequency signal, the heart signal, the respiratory signal, and the translation signal on the X-axis and the Y-axis, respectively.

[detailed description]

The present invention will be further described in detail below with reference to the drawings and exemplary embodiments. It is understood that the exemplary embodiments described herein are merely illustrative of the invention and are not intended to limit the scope of the invention.

The method of the present invention utilizes the EMD method to automatically extract heart, breath, translation, and other motions. As shown in Figure 1, the following steps are included:

(1) Selecting vascular structural feature points

According to the knowledge of physiology and anatomy, the marked feature points need to be able to comprehensively reflect the motion information of the whole blood vessel. Therefore, the selected special points include the starting point and ending point of each blood vessel segment, and each part between the blood vessel segments. Inflection point. Moreover, in the sequence of contrast images at two different projection angles, all the feature points are numbered, and the corresponding feature points have the same number. As shown in Fig. 2 and Fig. 4, in the pair of contrast images with projection angles of (-26.5°, -20.9°) and (42.3°, 26.8°), there are five numbered points (Fig. The white dots in the middle), their relationship is one-to-one correspondence by numbers.

(2) Empirical mode decomposition (EMD) method for separating multiple motions

Both the heart and the human respiratory movement are periodic movements, but the frequency of respiratory movements is much smaller than that of the heart movement. Generally speaking, the frequency of normal heart movement is 60-100 beats/min, and the period is 0.6. -1.0s, while the cycle of breathing is much longer, usually 3-6s, and may be longer when quiet. On the other hand, the heart movement is more intense, and the magnitude of the breathing movement is smaller, that is, the generated displacement is smaller, which is a relatively stable process. For the vibration of the capillaries, it is more intense and shorter than the heart and respiratory movements. It is generally considered that the period is less than 0.6 s and the amplitude variation range is small. The biggest feature of translational motion is that it is aperiodic motion, as opposed to Periodic movements are easy to discern.

According to these characteristics of each of the above-mentioned motion signals in the angiogram, the heart, the breathing, the translation, and the high-frequency motion can be automatically separated by the empirical mode decomposition (EMD) method. The following describes the EMD method.

1) EMD method

The starting point of EMD is to think of the oscillations in the signal as local. In fact, if you want to see the change between the two adjacent extreme points of the evaluation signal x(t) (2 minimum values, at t- and t+, respectively), you need to define a (local) high-frequency component { d(t), t-≤t≤t+} (local detail), this high-frequency component corresponds to the oscillation, oscillates between 2 minimum values and passes the maximum value (definitely appears at 2 minima between). In order to complete this graph, it is also necessary to define a (local) low frequency component m(t) (local trend) such that x(t) = m(t) + d(t), (t - ≤ t ≤ t +). For all seismic components of the entire signal, if a suitable method can be found for such decomposition, this process can be applied to the residual components of all local trends, so that the constituents of a signal can be extracted in an iterative manner.

For a signal x(t) to be decomposed, the effective EMD decomposition steps are as follows:

(1) Find all extreme points of x(t);

(2) using the interpolation method to form the lower envelope emin(t) for the minimum value and the upper envelope emax(t) for the maximum value;

(3) Calculate the mean m(t)=(emin(t)+emax(t))/2;

(4) Extracting the detail signal d(t)=x(t)-m(t);

(5) For the residual m(t), let x(t) = m(t), repeat steps (1)-(5) until the mean of d(t) is 0, or the stopping criterion is satisfied.

In practice, the above process needs to be redefined by a screening process. The first iterative step of the screening process is to repeat steps (1)-(5) for the detail signal d(t) until the mean of d(t) is 0. , or satisfy some sort of stopping criterion to stop iteration.

Once the stop criterion is met, the detail signal d(t) at this time is called the Intrinsic Mode Function (IMF), and the d(t) corresponding residual signal is calculated in the fifth step. Through the above process, the number of extreme points is less and less accompanied by the generation of residual signals, and the entire decomposition process produces a finite number of IMFs, which are the independent signals required.

2) Separation algorithm

Assume that the motion curve of the x-axis coordinate of a point p(x, y) on the coronary vessel in the sequence of the contrast image is x(n) (n is the number of frames of the contrast frame), and the motion curve of the y-axis coordinate is y(n) Let s(n)=(x(n), y(n)) decompose s(n) into the following expression:

s(n)=c(n)+r(n)+h(n)+L(n) where c(n)=(x _c (n), y _c (n)) represents the blood vessels caused by the movement of the heart Point motion; r(n)=(x _r (n), y _r (n)) indicates motion caused by respiratory motion; h(n)=(x _h (n), y _h (n)) indicates The motion produced by tremor or the beating of the blood vessel itself is generally regarded as a high frequency component; L(n) = (x _L (n), y _L (n)) represents the translational motion (including the movement of the body during the angiography) , as well as the movement of contrast equipment, etc.). For convenience of representation, s(n) is used to indicate the coordinate curve of the extracted blood vessel point along the x-axis and the y-axis, and c(n) represents the coordinate curve of the blood vessel point along the x-axis and the y-axis caused by the cardiac motion. r(n) represents the coordinate curve of the blood vessel point along the x-axis and the y-axis caused by the respiratory motion, and h(n) represents the coordinate curve of the blood vessel point along the x-axis and the y-axis caused by the high-frequency motion, and L(n) represents The coordinate curve of the translational motion along the x-axis and the y-axis. Therefore, the operation on s(n) is the operation on x(n) and y(n) respectively, and the operation on c(n) is the operation on x _c (n) and y _c (n) respectively, on r The operation of (n) is the operation of x _r (n) and y _r (n) respectively. The operation of h(n) is the operation of x _h (n) and y _h (n) respectively, for L(n) The operation is to operate on x _L (n) and y _L (n) respectively.

The specific algorithm is as follows:

Step1: Automatically track the vascular structural feature points selected in (1) throughout the angiographic sequence;

Step2: Select a sequence of length n _s =k*N ₁ (k>1) in the tracking sequence s(n) of the point

If the length of the original sequence s (n) is n, then the selected sequence of length n _s = nn% N _1, i.e., so that n _s is an integer multiple of N _1, where N ₁ is the period of cardiac motion,% is taken Yu symbol.

Step3: Will

Step4: According to the prior physiological knowledge, the independent signals are classified accordingly.

By analyzing the motion signals decomposed by the prior physiological knowledge combined with the EMD method, it is possible to determine the components of each motion information easily, as shown in Fig. 3(a)-3(j) and Fig. 5(a)-5(j). , wherein the heart signal curve, the respiratory signal curve and the high frequency signal curve are respectively obtained by tracking the first five feature points, and the broken line of the translation signal indicates the movement of the manually tracked skeletal muscle, and the solid line indicates It is a graph extracted by the EMD method.

It can be seen from the motion signals separated from Fig. 3(a)-3(j) and Fig. 5(a)-5(j) that the cardiac signals and respiratory signals of each feature point have obvious regularity; Signals are uneven due to a variety of factors (such as vascular tremors and human tremors, etc.), but they are outside the scope of physiological knowledge, so they are classified as high-frequency signals; and automatically extracted by EMD method The translational signal is basically consistent with the translational signal extracted by the manual tracking skeletal muscle, and thus has strong practicability.

The above is only the preferred embodiment of the present invention, and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection of the present invention. Within the scope.

Claims

A method for estimating a multi-motion parameter decomposition of a single-arm x-ray angiography image sequence, comprising the following steps:

(1) selecting stable vascular structural feature points;

(2) automatically tracking the selected vascular structural feature points throughout the sequence of contrast images;

(3) Select a sequence of length n s =k*N 1 (k>1) in the tracking sequence s(n) of the point

(4) will
Decomposed into motion x(n) in the x direction and motion y(n) in the y direction, and empirical mode decomposition (EMD) decomposition of x(n) and y(n), respectively, to obtain independent independence after EMD decomposition Motion signal

(5) Corresponding classification of each independent signal.
The method according to claim 1, wherein in the step (1), the structural feature points include a starting point and a ending point of each blood vessel segment, and respective inflection points between the blood vessel segments.
The method according to claim 1, wherein in step (3), if the length of the original sequence s(n) is n, the sequence length is n s = nn%N 1 (N 1 is the period of cardiac motion), That is, let n s be an integer multiple of N 1 , where % is the remainder symbol.
The method according to claim 1, wherein in step (4), for a signal x(n) to be decomposed, the EMD decomposition is specifically:

(4-1) find all extreme points of x(n);

(4-2) using the interpolation method to form the lower envelope emin(n) for the minimum value and the upper envelope emax(n) for the maximum value;

(4-3) Calculate the mean m(n)=(emin(n)+emax(n))/2;

(4-4) Extracting the detail signal d(n)=x(n)-m(n);

(4-5) For the residual m(n), let x(n)=m(n), repeat steps (4-1) to (4-5) until the mean of d(n) is 0, or satisfy Stop the guidelines.