Medical image segmentation method based on area-preserving convex curve evolution model

Abstract

The invention discloses a medical image segmentation method based on an area-preserving convex curve evolution model, and belongs to the field of left ventricle medical image segmentation, and the method comprises the steps: obtaining a heart magnetic resonance image, and obtaining a to-be-segmented image; segmenting the to-be-segmented image by using a region scalable fitting image segmentation method to obtain a preliminary segmentation result of the left ventricle; determining an initial level set function of the area-preserving convexity curve evolution model based on the preliminary segmentation result of the left ventricle; and driving the initial level set function to move towards the contour of the left ventricle through the area-preserving convex curve evolution model so as to determine a final segmentation result of the left ventricle of the current layer. And segmenting the next layer of left ventricle by using the final segmentation result of the current layer of left ventricle until the left ventricle regions in all layers of images of the to-be-segmented image are completely segmented. According to the method, the problems that the left ventricle is difficult to completely segment by a common segmentation method and the left ventricle structure is often lost are solved.

Description

Technical Field

[0001] The invention belongs to the field of left ventricle medical image segmentation, and in particular relates to a medical image segmentation method based on an area-preserving convexity curve evolution model. Background Art

[0002] In the field of cardiac medicine, CMR can provide important information about cardiac function, such as the size, shape and function of the ventricles and atria. Left ventricle segmentation in CMR images plays a vital role in medical image analysis. It can accurately outline the structure of the patient's left ventricle and provide important information for doctors to gain a deeper understanding of their heart health. This technology not only helps to accurately diagnose heart diseases such as myocardial infarction or myocardial abnormalities, but is also indispensable in formulating treatment plans and monitoring treatment efficacy.

[0003] According to the anatomical structure of the left ventricle, the left ventricle should be convex, including blood cavity, trabeculae and papillary muscles. The papillary muscles originate from the endocardium and are connected to the chordae tendineae. According to clinical standards, the final segmentation result of the left ventricle should include trabeculae and papillary muscles. However, the grayscale values ​​of trabeculae and papillary muscles are similar to those of myocardium, and simple segmentation methods often have difficulty in accurately distinguishing these structures.

[0004] From the perspective of anatomy, the endocardium of the left ventricle should be smooth, closed, and approximately circular in shape. Therefore, we need to design an algorithm based on the shape characteristics of the left ventricle to ensure that the segmentation results are consistent with the geometric characteristics of the cardiac anatomical structure. Summary of the invention

[0005] In view of the above-mentioned deficiencies in the prior art, the present invention provides a medical image segmentation method based on an area-preserving convexity curve evolution model, which solves the problem that ordinary segmentation methods are difficult to completely segment the left ventricle and often miss the left ventricle structure.

[0006] In order to achieve the above-mentioned invention object, the technical solution adopted by the present invention is: a medical image segmentation method based on an area-preserving convexity curve evolution model, comprising the following steps:

[0007] S1, acquiring a cardiac magnetic resonance image to obtain an image to be segmented;

[0008] S2. Segment the image to be segmented using a regional scalable fitting image segmentation method to obtain a preliminary segmentation result of the left ventricle;

[0009] S3, determining the initial level set function of the area-preserving convexity curve evolution model based on the preliminary segmentation result of the left ventricle;

[0010] S4, driving the initial level set function to move toward the contour of the left ventricle through the area-preserving convex curve evolution model, thereby determining the final segmentation result of the left ventricle of the current layer;

[0011] S5. Use the final segmentation result of the left ventricle in the current layer to segment the left ventricle in the next layer, until the left ventricle regions in all layers of the image to be segmented are completely segmented.

[0012] Furthermore, the step S2 specifically includes: using the regional scalable fitting image segmentation method to perform full-image segmentation on the image to be segmented, and randomly selecting a point to extract the segmentation result of the left ventricle as the preliminary segmentation result of the left ventricle.

[0013] Furthermore, the expression of the initial level set function of the area-preserving convexity curve evolution model in step S3 is:

[0014]

[0015] in, is the initial level set function; x is the pixel point.

[0016] Furthermore, the expression of the area-preserving convexity curve evolution model is:

[0017]

[0018] in, To preserve the area convexity curve evolution model; is a data item; is the curvature convergence term; is the distance regularization term; μ is the distance regularization term The weight coefficient of is the initial level set function; β is the curvature convergence term The weight coefficient of .

[0019] Furthermore, the evolution function of the area-preserving convexity curve evolution model is:

[0020]

[0021] in, is the initial level set function; t is the time; is the symbol of partial derivative; To preserve the area convexity curve evolution model; is the distance regularization term; μ is the distance regularization term The weight coefficient of is a data item; is the curvature convergence term; β is the curvature convergence term The weight coefficient; div() is the divergence calculation function; d p is the correlation function of the regularization term; |·| is the absolute value; is the gradient sign; ν is a constant parameter greater than zero; δ() is the Dirac function; λ1 and λ2 are both constants greater than 0; e1 is the fitting error of the foreground area; e2 is the fitting error of the background area; κ is the curvature of the initial level set function; is the mean curvature of the initial level set function; K σ () is the Gaussian kernel function; y and x are both pixel points; I() is the image information; f1() is the grayscale fitting function of the foreground area; H() is the unit step function; f2() is the fitting function of the background area; ∫ Ω ·dy and ∫ Ω ·dx is integrated over the image area.

[0022] Furthermore, the step S5 is specifically as follows:

[0023] S501, calculating the centroid coordinates of the left ventricle of the current layer according to the final segmentation result of the left ventricle of the current layer;

[0024] S502, after corroding the final segmentation result of the left ventricle of the current layer, a new image to be segmented is obtained, the image to be segmented is updated to the new image to be segmented, and the process returns to step S2, and based on the centroid coordinates of the left ventricle of the current layer, a preliminary segmentation result of the left ventricle is extracted;

[0025] S503 , repeating steps S2 - S5 until all left ventricle regions in all layer images of the image to be segmented are segmented.

[0026] The beneficial effects of the present invention are: the average curvature is introduced to ensure that the segmentation result is convex, so that the segmentation result is a complete left ventricular structure; the curvature convergence term makes the area of ​​the curve remain unchanged during the evolution process, so that even if the grayscale values ​​of the trabeculae and papillary muscles are similar to those of the myocardium, the curve will not leak when it evolves. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0028] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0029] like Figure 1 As shown, in one embodiment of the present invention, a medical image segmentation method based on an area-preserving convexity curve evolution model comprises the following steps:

[0030] S1, acquiring a cardiac magnetic resonance image to obtain an image to be segmented;

[0031] S2. Segment the image to be segmented using a regional scalable fitting image segmentation method to obtain a preliminary segmentation result of the left ventricle;

[0032] S3, determining the initial level set function of the area-preserving convexity curve evolution model based on the preliminary segmentation result of the left ventricle;

[0033] S4, driving the initial level set function to move toward the contour of the left ventricle through the area-preserving convex curve evolution model, thereby determining the final segmentation result of the left ventricle of the current layer;

[0034] S5. Use the final segmentation result of the left ventricle in the current layer to segment the left ventricle in the next layer, until the left ventricle regions in all layers of the image to be segmented are completely segmented.

[0035] In this embodiment, the left ventricle is first segmented using a region-based method (RSF) to obtain a preliminary segmentation result.

[0036] The step S2 specifically includes: using the regional scalable fitting image segmentation method to perform full image segmentation on the image to be segmented, and randomly selecting a point to extract the segmentation result of the left ventricle as the preliminary segmentation result of the left ventricle.

[0037] The expression of the initial level set function of the area-preserving convexity curve evolution model in step S3 is:

[0038]

[0039] in, is the initial level set function; x is the pixel point.

[0040] The expression of the area-preserving convexity curve evolution model is:

[0041]

[0042] in, To preserve the area convexity curve evolution model; is a data item; is the curvature convergence term; is the distance regularization term; μ is the distance regularization term The weight coefficient of is the initial level set function; β is the curvature convergence term The weight coefficient of .

[0043] The evolution function of the area-preserving convexity curve evolution model is:

[0044]

[0045]

[0046] in, is the initial level set function; t is the time; is the symbol of partial derivative; To preserve the area convexity curve evolution model; is the distance regularization term; μ is the distance regularization term The weight coefficient of is a data item; is the curvature convergence term; β is the curvature convergence term The weight coefficient; div() is the divergence calculation function; d p is the correlation function of the regularization term; |·| is the absolute value; is the gradient sign; ν is a constant parameter greater than zero; δ() is the Dirac function; λ1 and λ2 are both constants greater than 0; e1 is the fitting error of the foreground area; e2 is the fitting error of the background area; κ is the curvature of the initial level set function; is the mean curvature of the initial level set function; K σ () is the Gaussian kernel function; y and x are both pixel points; I() is the image information; f1() is the grayscale fitting function of the foreground area; H() is the unit step function; f2() is the fitting function of the background area; ∫ Ω ·dy and ∫ Ω dx is integrated over the image area.

[0047] In the example of the present application, the RSF method is first used to segment the entire CMR image, and then a point is selected in the left ventricle region to separate the left ventricle segmented by the RSF method, thereby obtaining a preliminary segmentation result of the left ventricle. Based on the preliminary result, the initial level set function of the area-preserving convexification curve evolution model is determined.

[0048] The initial level set function is 0-level set, which represents the left ventricular contour:

[0049]

[0050] The area-preserving convexity curve evolution model is a model proposed in this application, which is used to evolve the left ventricular contour. In the embodiment of this application, the area-preserving convexity curve evolution model is specifically shown in the following formula:

[0051]

[0052] In the formula, is a data item, is the curvature convergence term, μ is the distance regularity The weight coefficient of is the initial level set function.

[0053] Therefore, the segmentation problem of the left ventricle in the cardiac magnetic resonance image is transformed into the problem of solving the energy functional minimization problem in the image area. The energy minimization process is implemented using the steepest descent method, and the evolution function of the area-preserving convex curve evolution model is obtained:

[0054]

[0055] Where μ, ν, and β are constant parameters greater than zero. e1 and e1 are defined as follows:

[0056]

[0057] Where K σ represents the Gaussian kernel function, H represents the unit step function, * represents convolution, and I represents the image information.

[0058] In addition, in the evolution equation, the Greek letters κ and are the curvature and mean curvature of the initial level set function, respectively. κ and The specific definition is as follows:

[0059]

[0060]

[0061] In the formula, ∫ Ω ·dx represents the integral over the image area, δ is the Dirac function, and div is used to calculate the divergence. Used to calculate the gradient of the level set function. It is worth noting that the total curvature of the curve is greater than zero, and the average curvature of the curve is Also greater than zero.

[0062] As the curve evolves, the curvature convergence term The evolution result is guaranteed to be convex. Specifically, when the curvature of a point (denoted by κ) is greater than the average curvature When the level set function of this point The value of increases with time, while κ gradually decreases. On the contrary, when the curvature at a point is less than the average curvature When the level set function of this point The value of decreases with time, and κ gradually increases. Due to the above mechanism, the curvature of the curve will converge to its mean curvature, and the final segmentation result of the zero level set is convex, which is in line with expectations.

[0063] In addition, mathematically, it can be proved that the curvature convergence term It has the property of preserving the area, which prevents the curve from leaking out during evolution because the grayscale values ​​of trabeculae and papillary muscles are similar to those of the myocardium.

[0064] The step S5 is specifically as follows:

[0065] S501, calculating the centroid coordinates of the left ventricle of the current layer according to the final segmentation result of the left ventricle of the current layer;

[0066] S502, after corroding the final segmentation result of the left ventricle of the current layer, a new image to be segmented is obtained, the image to be segmented is updated to the new image to be segmented, and the process returns to step S2, and based on the centroid coordinates of the left ventricle of the current layer, a preliminary segmentation result of the left ventricle is extracted;

[0067] S503 , repeating steps S2 - S5 until all left ventricle regions in all layer images of the image to be segmented are segmented.

[0068] The initial level set function is driven to move toward the contour of the left ventricle by the area-preserving convexity curve evolution model, thereby determining the final segmentation result of the left ventricle.

[0069] The segmentation result of the left ventricle layer is used to perform segmentation of the left ventricle layer of the next layer until the segmentation of the left ventricle regions in all layer images of the image to be segmented is completed.

[0070] Finally, the left ventricle segmentation result of the image to be segmented is output.

Claims

1. A medical image segmentation method based on an area-preserving convexity curve evolution model, characterized in that: The following steps are involved: S1, acquiring a cardiac magnetic resonance image to obtain an image to be segmented; S2. Segment the image to be segmented using a regional scalable fitting image segmentation method to obtain a preliminary segmentation result of the left ventricle; S3, determining the initial level set function of the area-preserving convexity curve evolution model based on the preliminary segmentation result of the left ventricle; S4, driving the initial level set function to move toward the contour of the left ventricle through the area-preserving convex curve evolution model, thereby determining the final segmentation result of the left ventricle of the current layer; S5. Use the final segmentation result of the left ventricle in the current layer to segment the left ventricle in the next layer, until the left ventricle regions in all layers of the image to be segmented are completely segmented.

2. The medical image segmentation method based on the area-preserving convexity curve evolution model according to claim 1, characterized in that: The step S2 specifically includes: using the regional scalable fitting image segmentation method to perform full image segmentation on the image to be segmented, and randomly selecting a point to extract the segmentation result of the left ventricle as the preliminary segmentation result of the left ventricle.

3. The medical image segmentation method based on the area-preserving convexity curve evolution model according to claim 1, characterized in that: The expression of the initial level set function of the area-preserving convexity curve evolution model in step S3 is: in, is the initial level set function; x is the pixel point.

4. The medical image segmentation method based on the area-preserving convexity curve evolution model according to claim 3 is characterized in that: The expression of the area-preserving convexity curve evolution model is: in, To preserve the area convexity curve evolution model; is a data item; is the curvature convergence term; is the distance regularization term; μ is the distance regularization term The weight coefficient of is the initial level set function; β is the curvature convergence term The weight coefficient of .

5. The medical image segmentation method based on the area-preserving convexity curve evolution model according to claim 3 is characterized in that: The evolution function of the area-preserving convexity curve evolution model is: in, is the initial level set function; t is the time; is the symbol of partial derivative; To preserve the area convexity curve evolution model; is the distance regularization term; μ is the distance regularization term The weight coefficient of is a data item; is the curvature convergence term; β is the curvature convergence term The weight coefficient; div() is the divergence calculation function; d p is the correlation function of the regularization term; |·| is the absolute value; is the gradient sign; ν is a constant parameter greater than zero; δ() is the Dirac function; λ1 and λ2 are both constants greater than 0; e1 is the fitting error of the foreground area; e2 is the fitting error of the background area; κ is the curvature of the initial level set function; is the average curvature of the initial level set function; K σ () is the Gaussian kernel function; y and x are both pixel points; I() is the image information; f1() is the grayscale fitting function of the foreground area; H() is the unit step function; f2() is the fitting function of the background area; and All are integrated over the image area.

6. The medical image segmentation method based on the area-preserving convexity curve evolution model according to claim 1, characterized in that: The step S5 is specifically as follows: S501, calculating the centroid coordinates of the left ventricle of the current layer according to the final segmentation result of the left ventricle of the current layer; S502, after corroding the final segmentation result of the left ventricle of the current layer, a new image to be segmented is obtained, the image to be segmented is updated to the new image to be segmented, and the process returns to step S2, and based on the centroid coordinates of the left ventricle of the current layer, a preliminary segmentation result of the left ventricle is extracted; S503 , repeating steps S2 - S5 until all left ventricle regions in all layer images of the image to be segmented are segmented.

Images