CN112017152B - Processing method of two-dimensional image of atrial impression - Google Patents

Abstract

The invention discloses a processing method of an atrial impression two-dimensional image, which comprises the following steps: selecting one frame from the two-dimensional image of the left atrium positioned on the horizontal section of the root of the arterial sinus as an interested frame, wherein the aortic valve in the interested frame is clearly visible and the atrial impression is obvious; extracting a foreground boundary of a left atrium and a central point of the left atrium in the interested frame, and identifying a key inflection point in the foreground boundary; locating an inflection point around the left atrium based on the location of the critical inflection point in the foreground boundary and the center point of the left atrium; and extracting the left atrium boundary according to the inflection points around the left atrium to obtain the impression curve of the left atrium. Compared with the prior art, the method for processing the complex image of the three-dimensional structure image based on the left atrium obtains the left atrium indentation curve through the two-dimensional image of the left atrium, is simple and easy to implement, can estimate the pressure in the left atrium by quantifying the indentation curve of the front wall of the left atrium, and realizes the analysis of the pressure in the left atrium by a non-invasive method under the atrial fibrillation state.

Description

Processing method of two-dimensional image of atrial impression

technical field

The invention relates to a method for processing two-dimensional images of the heart, and belongs to the field of cardiac medical image processing.

Background technique

When atrial fibrillation occurs in the heart, the pressure in the left atrium is difficult to assess by non-invasive methods such as ultrasound due to irregular atrial contractions. Since the left atrial anterior wall indentation is determined by the intra-atrial pressure, the left atrial pressure can be estimated by analyzing and processing the structural image of the left atrium, quantifying the left atrial anterior wall indentation, and, in the state of atrial fibrillation, the available indentation The curve analyzes the pressure in the left atrium, and the curvature of the indentation curve can be further used to quantify the key structures of the left atrium.

At present, the institutions that study the morphological structure of the left atrium are represented by the Institute of Computational Science and Imaging of the University of Utah in the United States, but it analyzes the three-dimensional structural images of the left atrium, and the analysis and image processing methods are relatively complex and cannot be widely promoted.

SUMMARY OF THE INVENTION

Aiming at the defects of the prior art, the technical problem to be solved by the present invention is to provide a simple and easy processing method for a two-dimensional image of atrial indentation, so as to obtain the indentation curve of the left atrium.

The technical means adopted by the present invention to solve its technical problems are: the processing method of the two-dimensional image of atrial impression in the present invention comprises the following steps:

(1) Select a frame of interest from a two-dimensional image of the left atrium in the horizontal section of the root of the arterial sinus, in which the aortic valve is clearly visible and the atrial indentation is obvious;

(2) extracting the foreground boundary of the left atrium and the center point of the left atrium in the frame of interest, and identifying key inflection points in the foreground boundary;

(3) According to the position of the key inflection point in the foreground boundary and the center point of the left atrium, locate the inflection point around the left atrium;

(4) Extract the boundary of the left atrium according to the inflection point around the left atrium, and obtain the indentation curve of the left atrium.

Further, in the step (2) of the present invention, a subpixel-based boundary extraction method is used to identify the weak boundary between the left atrium and other cardiac chambers, and obtain the foreground boundary of the left atrium including the weak boundary.

Further, in the present invention, before extracting the foreground boundary of the left atrium in the frame of interest, the sobel operator is used to extract the boundary distribution of the left atrium in the frame of interest.

Further, in the step (2) of the present invention, the CPDA method is used to identify the key inflection points in the foreground boundary.

Further, in the step (2) of the present invention, an iterative algorithm is used to extract the center point of the left atrium.

Further, in the step (4) of the present invention, the spline interpolation method is used to complete the missing boundary of the left atrium.

Further, in the step (4) of the present invention, the boundary of the left atrium is smoothed.

Further, the present invention smoothes the indentation curve.

Further, the method for smoothing the indentation curve in the present invention is polynomial interpolation.

Further, the present invention calculates the curvature of the smoothed indentation curve.

Compared with the prior art, the beneficial effects of the present invention are:

The prior art is based on a three-dimensional structural image of the left atrium, and the image processing method is relatively complicated, which is difficult to be widely applied in practice. The present invention obtains the indentation curve of the left atrium by analyzing and processing the two-dimensional image of the left atrium, and the image processing method is simple and easy to implement, and can be widely used; Atrial indentation curve, whereby the indentation curve is used to estimate the left atrial pressure by quantifying the left atrial anterior wall indentation, and the non-invasive method to analyze the left atrial pressure in the state of atrial fibrillation is realized; The curvature of the indentation curve of the atrium can further be used to quantify key structures of the left atrium.

Description of drawings

FIG. 1 is a schematic flowchart of a method for processing a two-dimensional image of atrial impression in the present invention.

Fig. 2 is a state diagram when the two-dimensional image of the left atrium is processed by the processing method of the present invention, wherein A is the result of extracting the foreground boundary of the left atrium based on the sub-pixel boundary extraction method; B is the key inflection point in the foreground boundary C is the positioning result of the inflection point around the left atrium; D is the extracted left atrium boundary; E is a segment of the indentation curve intercepted for calculating the curvature; F is the indentation smoothed by polynomial interpolation Curve segment.

Figure 3 shows three typical graphs of left atrial indentation. From Figure A to Figure C, the curvature of the indentation curve of the left atrium gradually decreases.

Detailed ways

The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but the present invention is not limited to these embodiments.

As shown in FIG. 1 to FIG. 3 , the processing method of the two-dimensional image of atrial impression in the present invention includes the following steps:

1. From the two-dimensional image of the left atrium in the horizontal section of the root of the arterial sinus, select a frame of interest, in which the aortic valve is clearly visible and the atrial indentation is obvious. The specific instructions are as follows:

In the present invention, CT or MRI images of the left atrium are preferably used, and CT images are used as an example for description below. The acquired image is a two-dimensional image of the horizontal section of the root of the sinus of the left atrium. The images of the same ECG cycle are preferred during acquisition, and one of them is selected as the frame of interest. In the selected frame of interest, the aortic valve should be clearly visible and the atrial indentation is obvious. The clearer the aortic valve and the indentation are, the better.

2. Extract the foreground boundary of the left atrium and the center point of the left atrium in the frame of interest, and identify the inflection point in the foreground boundary. The specific instructions are as follows:

(1) Extract the foreground boundary of the left atrium in the frame of interest

In order to objectively reflect the morphological and structural characteristics of the left atrium, compared with the prior art, the present invention introduces the boundary extraction technology for the first time. Before extracting the left atrial foreground boundary in the frame of interest, the background of the CT tomographic image can be removed by the binarization method, and the frame of interest can be preprocessed, which includes two steps: (1) First, use the snake model to evolve, Remove the more obvious part of the sternum; (2) Then take the foreground center as the seed point, and use the region growing model to remove the noise region that is not adhered to the foreground. The preprocessed image includes the left atrium, left ventricle, right atrium and right ventricular outflow tract, and heart chambers such as the aortic root, collectively referred to as the foreground.

Since the grayscale and texture between the left atrium and other cardiac chambers are similar, and the boundary is relatively blurred, the present invention preferably adopts a subpixel-based boundary extraction method to identify the weak boundary between the left atrium and other cardiac chambers, and obtains the weak boundary between the left atrium and other cardiac chambers. Inside the foreground border of the left atrium. details as follows:

The subpixel can diagnose the boundary position within a single pixel. Assuming that the boundary line traverses a single pixel, the grayscale F _i,j of the pixel can be expressed as formula (1), where A and B represent the pixel grayscales on both sides of the boundary, respectively. value, and S _A and S _B respectively represent the area corresponding to the pixel gray level on both sides of the boundary.

As a preferred solution of the present invention, before using the sub-pixel algorithm to extract the foreground boundary of the left atrium in the frame of interest, the sobel operator may be used to extract the boundary distribution of the left atrium in the frame of interest. Based on this distribution, it is assumed that a curve C(x ₁ , x ₂ ... x _N ) traverses the pixel (i, j), which is the estimated curve parameter. To this end, a mask with the center point at pixel (i, j) is constructed, and the sum of the gray levels of each column of pixels in the mask is calculated (as shown in formula (2)). Among them, the subscripts L, M, and R of the mask S in the formula (2) represent the left, middle, and right columns in the mask, respectively, and r represents the mask radius.

When formula (1) is introduced into formula (2), only variables A, B and S _A and S _B are included in formula (2). In addition, the values of A and B can be generated by the approximation of the surrounding pixels, and S _A and S _B are related to the slope of the curve within the pixel, which can be expressed by curve parameters. Therefore, the curve parameters can be solved by formula (2). Using the above mask to traverse the image, the position of the foreground boundary of the left atrium including the weak boundary can be obtained (as shown in Figure 2A).

(2) Identify the key inflection points in the foreground boundary

The borders of the left atrium and adhesions of other tissues were included in the foreground borders diagnosed by the subpixel algorithm. In order to distinguish the boundary positions of different tissues, the present invention can adopt an effective inflection point diagnosis algorithm, called CPDA method. Through this method, the position of the inflection point is detected, and then the key inflection point in the foreground boundary is identified. Thus, the problem of extracting the boundary position of the left atrium can be transformed into the problem of extracting the inflection point near the boundary of the left atrium.

The full English name of CPDA is chord-to-point distance accumulation. It calculates the vertical distance between each point on the curve and its corresponding arc segment. The arc segment moves along the boundary to obtain the point-to-arc distance accumulation, and then based on the distance The cumulant calculates the curvature value and determines the position of the inflection point. Compared with the traditional inflection point localization method based on the curvature scale space, the CPDA method does not directly calculate the derivative of any point on the curve, so it is more robust to local changes of the curve and noise points. The identification results of key inflection points in the foreground boundary are shown in Fig. 2B.

(3) The present invention can use an iterative algorithm to find the center point of the left atrium.

3. Locate the inflection point around the left atrium according to the position of the key inflection point in the foreground boundary and the center point of the left atrium

Taking the found center point of the left atrium as the origin, the Euclidean distance between each key inflection point in the foreground boundary and the origin is counted. Assuming that the Euclidean distances of all key inflection points obey the normal distribution of (μ, σ), then discard the inflection points outside the (μ-σ, μ+σ) interval to form a new inflection point set. A new image center is obtained with a new set of inflection points to form a new origin, and a new round of iterative process begins. The above iteration is performed twice to obtain key points near the left atrium (ie, the inflection point around the left atrium) (as shown in Figure 2C).

4. Extract the boundary of the left atrium according to the inflection point around the left atrium, and obtain the impression curve of the left atrium

In a clockwise direction, identify the boundary points around the inflection point of the left atrium. For the missing boundary position in a small range, the spline interpolation method can be used to complete it. Finally, morphological operations were used to smooth the boundary of the left atrium, and the position of the smoothed boundary is shown in Figure 2D. If there is a deviation in the smoothed boundary position, the boundary position can be fine-tuned by means of human-computer interaction. Thus, the indentation curve of the left atrium is obtained.

V. Further, the curvature of the left atrial indentation curve can be obtained by the following method

The indentation curve of the left atrium was intercepted, and its start and end points could be marked manually (as shown in Figure 2E). The curvature is calculated for the intercepted indentation curve. Curvature refers to the rotation rate of the tangent direction angle to a point on the curve to the arc length, and its calculation method is shown in the following formula (3):

In formula (3), k represents the curvature of the indentation curve, and y' and y" represent the first-order and second-order derivatives of the indentation curve, respectively.

It can be seen from formula (3) that the absolute value of the curvature will increase with the increase of the degree of curvature of the curve. In addition, the numerical sign of the curvature depends on the properties of the curve: convex curves correspond to positive curvatures, and concave curves correspond to negative curvatures.

Since the pixel points are discretely distributed on the indentation curve, the direct calculation of the curvature of the indentation curve will have large fluctuations. Preferably, before the curvature is formally calculated, the intercepted indentation curve can be smoothed by means of polynomial interpolation. The calculation efficiency of polynomial interpolation is high, and it can fully express the distribution of the indentation curve. Its expression is shown in formula (4). The indentation curve segments before and after smoothing are shown in Figure 2F.

y=p ₁ x ⁿ +p ₂ x ^n-1 +...+p _n x ¹ +p _n+1 , where n=2 (4)

In formula (4), y is the fitting curve, x is the fitting independent variable, and p represents the coefficient of each order polynomial.

The CT images of 169 cases of left atrium were processed by using the two-dimensional image processing method of atrial indentation of the present invention, and 169 cases of left atrial indentation images were obtained correspondingly. The pressure in the left atrium is usually within 10mmHg, and if it exceeds this value, the pressure in the left atrium increases, which can cause the pressure trace on the anterior wall of the left atrium to become shallow or even disappear. According to the commonness of the impression curves of the 169 cases of left atrial indentation, they can be divided into three types, A, B, and C of Figure 3 are typical diagrams of the three types of left atrial indentation curves, respectively. From A to C of FIG. 3 , it is shown that the pressure in the left atrium changes from normal to elevated, with the atrial impression curve flattening and even bulging forward.

After smoothing the indentation curves intercepted from the 169 left atrial indentation images using polynomial interpolation, the average curvatures of the three types of left atrial indentation curves of A, B, and C in Figure 3 were calculated respectively. The results are shown in Figure 3 The average curvatures of the three types of indentation curves shown from A to C gradually become smaller, respectively -2.6(-3.55,-1.95)×10 ^-3 , -1.40(-2.13,-0.43)×10 ^-3 , 0.51 (0.02,1.40)×10 ^-3 , thus, the quantification of the left atrial indentation is achieved, which can be used for subsequent quantitative analysis of the key structures of the left atrium.

Claims (28)

1. A processing method of two-dimensional images of atrial footprints is characterized by comprising the following steps:

(1) selecting one frame from a two-dimensional image of the left atrium positioned on a horizontal section at the root of an arterial sinus as an interested frame, wherein in the interested frame, an aortic valve is clearly visible and an atrial impression is obvious;

(2) extracting a foreground boundary of a left atrium and a central point of the left atrium in the interesting frame, and identifying a key inflection point in the foreground boundary;

(3) taking the central point of the left atrium as an origin, counting Euclidean distances between each key inflection point in the foreground boundary and the origin, and selecting the Euclidean distance from the origin to obey

The normally distributed key inflection points form a new inflection point set, and the key inflection points are abandoned

Critical inflection points outside the interval; obtaining a new image center by using the new inflection point set, forming a new origin, and starting a new iteration process; taking a new inflection point set obtained in a new iteration as an inflection point around the left atrium;

(4) and extracting the left atrium boundary according to the inflection point around the left atrium to obtain the left atrium pressure trace curve.

2. The method of processing two-dimensional image of atrial impression of claim 1, wherein: in the step (2), a weak boundary between the left atrium and other heart chambers is identified by adopting a boundary extraction method based on sub-pixels, and a foreground boundary of the left atrium including the weak boundary is obtained.

3. The method for processing the two-dimensional image of the atrial impression of the human body according to claim 1 or 2, wherein: before extracting the foreground boundary of the left atrium in the interested frame, extracting the boundary distribution of the left atrium in the interested frame by adopting a sobel operator.

4. The method for processing the two-dimensional image of the atrial impression of the human body according to claim 1 or 2, wherein: in the step (2), a CPDA method is adopted to identify a critical inflection point in the foreground boundary.

5. The method of processing two-dimensional image of atrial impression of claim 3, wherein: in the step (2), a CPDA method is adopted to identify a critical inflection point in the foreground boundary.

6. The method of processing two-dimensional image of atrial impression of the claims 1, 2 or 5, wherein: in the step (2), an iterative algorithm is adopted to extract the central point of the left atrium.

7. The method of processing two-dimensional image of atrial impression of claim 3, wherein: in the step (2), the central point of the left atrium is extracted by adopting an iterative algorithm.

8. The method of processing two-dimensional image of atrial impression of claim 4, wherein: in the step (2), an iterative algorithm is adopted to extract the central point of the left atrium.

9. The method for processing the two-dimensional image of the atrial impression of the claims 1, 2, 5, 7 or 8, wherein: in the step (4), the boundary of the left atrial defect is completed by spline interpolation.

10. The method of processing two-dimensional image of atrial impression of claim 3, wherein: in the step (4), the boundary of the left atrial defect is completed by spline interpolation.

11. The method of processing two-dimensional image of atrial impression of claim 4, wherein: in the step (4), the boundary of the left atrial defect is completed by spline interpolation.

12. The method of processing two-dimensional images of atrial footprints as in claim 6, further comprising: in step (4), the left atrial defect boundary is completed using spline interpolation.

13. The method for processing the two-dimensional image of the atrial impression of claim 1, 2, 5, 7, 8, 10, 11 or 12, wherein: in step (4), smoothing the left atrial boundary.

14. The method of processing two-dimensional image of atrial impression of claim 3, wherein: in the step (4), the left atrial boundary is smoothed.

15. The method of processing two-dimensional image of atrial impression of claim 4, wherein: in the step (4), the left atrial boundary is smoothed.

16. The method of processing two-dimensional images of atrial footprints as in claim 6, further comprising: in the step (4), the left atrial boundary is smoothed.

17. The method for processing the two-dimensional image of the atrial impression of claim 9, wherein: in the step (4), the left atrial boundary is smoothed.

18. The method for processing a two-dimensional image of an atrial impression of the heart of claim 1, 2, 5, 7, 8, 10, 11, 12, 14, 15, 16 or 17, wherein: and smoothing the impression curve.

19. The method of processing two-dimensional image of atrial impression of claim 3, wherein: and smoothing the impression curve.

20. The method of processing two-dimensional image of atrial impression of claim 4, wherein: and smoothing the impression curve.

21. The method of processing two-dimensional images of atrial footprints as in claim 6, further comprising: and smoothing the impression curve.

22. The method of processing two-dimensional images of atrial footprints as in claim 9, further comprising: and smoothing the impression curve.

23. The method for processing the two-dimensional image of the atrial impression of claim 13, wherein: and smoothing the impression curve.

24. The method for processing the atrial impression two-dimensional image according to claim 18, wherein: the way of smoothing the impression curve is polynomial interpolation.

25. The method for processing the two-dimensional image of the atrial impression as claimed in any one of claims 19 to 23, wherein: the way of smoothing the impression curve is polynomial interpolation.

26. The method for processing the atrial impression two-dimensional image according to claim 18, wherein: and calculating the curvature of the smoothed impression curve.

27. The method for processing the two-dimensional image of the atrial impression as claimed in any one of claims 19 to 24, wherein: and calculating the curvature of the smoothed impression curve.

28. The method for processing the atrial impression two-dimensional image according to claim 25, wherein: and calculating the curvature of the smoothed impression curve.

Images