A method for extracting a blood vessel contour from a coronary angiogram image

Abstract

The application provides a coronary angiography image blood vessel contour extraction method, comprising the following steps: (a) performing cavity outside direct shooting on a to-be-measured coronary blood vessel through a coronary angiography system to obtain a coronary angiography image of the to-be-measured coronary blood vessel; (b) performing cavity inside scanning on the to-be-measured coronary blood vessel through an optical coherence tomography system to obtain a cavity inside two-dimensional cross-section image of the to-be-measured coronary blood vessel; (c) performing noise reduction on the coronary angiography image; (d) performing enhancement on the noise-reduced coronary angiography image; (e) performing segmentation on the enhanced coronary angiography image; (f) extracting a coronary blood vessel contour image from the segmented coronary angiography image; and (g) registering and fusing the extracted coronary blood vessel contour image with the cavity inside two-dimensional cross-section image. The coronary angiography image blood vessel contour extraction method combines the blood vessel contour in the coronary angiography image with the cavity inside two-dimensional cross-section image obtained by the optical coherence tomography system, so that the morphology of the same point of the blood vessel inside and outside the blood vessel can be observed quickly and clearly.

Description

Technical Field

[0001] This invention belongs to the field of medical imaging technology, specifically relating to a method for extracting vascular contours from coronary angiography images. Background Technology

[0002] The heart is one of the most important organs in the human body. Through its pumping action, the heart transports blood to all tissues and organs to maintain their normal life functions. The coronary circulation plays a vital role in supplying nutrients to the heart. It is mainly composed of coronary arteries and coronary veins. When the coronary arteries are narrowed or even blocked, and collateral circulation cannot be established in time, myocardial infarction will occur, which can even be life-threatening.

[0003] Extracorporeal imaging techniques mainly include coronary angiography (CAG), coronary CT angiography (CTA), and coronary magnetic resonance imaging (MRI). Among them, coronary angiography is the most commonly used imaging modality in the catheterization lab. It can visualize the overall anatomical structure of the coronary arteries and observe lesions and collateral circulation. It is one of the most important imaging techniques for the diagnosis and interventional procedures of coronary heart disease and is considered the "gold standard" for diagnosing coronary heart disease.

[0004] Coronary angiography uses contrast agents and X-rays to image the coronary arteries, achieving a spatial resolution of approximately 0.1 mm. It reveals the anatomical structure and overall morphology of the coronary arteries, allowing observation of the heart's blood supply function and the presence of stenosis lesions within the vessels. Therefore, coronary angiography can determine the location and extent of coronary anatomical abnormalities, observe anatomical characteristics of lesions such as diameter stenosis rate and stenosis length, and is clinically used to diagnose coronary heart disease and coronary artery malformations, as well as guide interventional surgical treatment. However, because the vessels in coronary angiography images are actually projections of the vessel lumen after contrast agent filling, it has some inherent limitations: coronary angiography images are difficult to accurately extract the vessel contours of specific locations, and can only provide a two-dimensional assessment of lumen size, making it difficult to clearly display internal lesions. Summary of the Invention

[0005] The purpose of this invention is to provide a method for extracting vascular contours from coronary angiography images, which aims to solve the problem that in the prior art, it is difficult to accurately extract the vascular contours of specific locations from coronary angiography images, and only two-dimensional assessment of the lumen size is possible.

[0006] This invention is achieved through the following technical solution:

[0007] A method for extracting vessel contours from coronary angiography images, comprising:

[0008] (a) The coronary angiography image of the coronary vessel under test is obtained by direct extraluminal imaging of the coronary vessel under test using a coronary angiography system;

[0009] (b) The coronary artery under test is scanned intraluminally using an optical coherence tomography system to obtain a two-dimensional cross-sectional image of the coronary artery under test.

[0010] (c) Noise reduction of coronary angiography images;

[0011] (d) Enhance the noise-reduced coronary angiography images;

[0012] (e) Segmenting the enhanced coronary angiography image;

[0013] (f) Extract the coronary artery contour map from the segmented coronary angiography image;

[0014] (g) The extracted coronary artery contour map is registered and fused with the intraluminal two-dimensional cross-sectional image.

[0015] Preferably, step (a) specifically includes: directing an X-ray beam generated by an X-ray tube through a filter and a collimator toward the coronary artery to be tested, and the X-rays passing through the coronary artery to be tested and reaching an image detector to generate a coronary angiography image.

[0016] Preferably, step (b) specifically includes: inserting a fiber-optic catheter into the coronary artery, rotating the inserted catheter within a protective sheath to obtain a two-dimensional cross-sectional image of the coronary artery, and simultaneously retracting the catheter to generate multiple frames of images to obtain a two-dimensional cross-sectional image of the entire coronary artery.

[0017] Furthermore, step (g) specifically includes: step (g) includes: using coronary angiography images and their frame rates, intraluminal two-dimensional cross-sectional images and their frame rates, and catheter retraction speed and retraction distance to achieve fusion of intraluminal two-dimensional cross-sectional images and coronary angiography images.

[0018] Preferably, step (c) involves denoising the coronary angiography image using median filtering.

[0019] Furthermore, step (c) specifically includes:

[0020] (1) Arrange the pixel values ​​of the pixels in the coronary angiography image from smallest to largest;

[0021] (2) Extract the odd number of data points within each pixel in the coronary angiography image and sort them in ascending order;

[0022] (3) Take the median value after sorting and replace the value of the pixel to obtain the noise-reduced coronary angiography image;

[0023] (4) Output the noise-reduced coronary angiography image.

[0024] Preferably, step (d) specifically includes:

[0025] (1) Input the denoised coronary angiography image and generate pixel matrix I;

[0026] (2) Initialize the scale parameter β = β1;

[0027] (3) Traverse all pixels I(x,y) in the pixel matrix I, calculate the convolution of pixel I(x,y) with the second derivative of the Gaussian function according to the scale parameter β, and obtain the convolution result; calculate the Hessian matrix H and its eigenvalues ​​according to the convolution result; calculate the output value V of the filter enhancement at the current scale according to the Hessian matrix H and its eigenvalues.

[0028] (4) Update the scaling parameter β→β+s;

[0029] (5) Check whether the scale parameter β exceeds the preset scale range. If it does, proceed to step (6); if it does not exceed the preset scale range, return to step (3). min ,β max The iteration step size is s;

[0030] (6) After the scale parameter iteration ends, calculate the maximum value V of the filtered enhancement output value of all pixels under each scale parameter. max This refers to the enhanced output of each pixel;

[0031] (7) The enhanced output of each pixel is multiplied by its corresponding scale parameter to obtain the output result of each pixel;

[0032] (8) Output the enhanced coronary angiography image based on the output results of each pixel.

[0033] Preferably, step (e) specifically includes:

[0034] (1) Count the number of pixels corresponding to gray value T in the enhanced coronary angiography image, where t1≤T≤t2. Based on the statistical results, find the gray value T0 with the largest number of pixels other than t1 and t2.

[0035] (2) Divide the coronary angiography image into target and background based on the gray value T0, calculate the probabilities p1 and p2 of target pixels and background pixels respectively, and calculate the overall median m0, background median m1, and target median m2 of the coronary angiography image. Replace the overall mean u0, background mean u1, and target mean u2 in the Otsu algorithm with the overall median m0, background median m1, and target median m2 respectively. Then the inter-class variance is:

[0036] σ 2 =p1(m1-m0) 2 +p2(m2-m0) 2

[0037] (3) Iterate through all gray values ​​in t1-t2, calculate the inter-class variance according to the method in step (2), and obtain the maximum inter-class variance and the corresponding T1; T1 = arg Maxσ 2 (t1≤T≤t2);

[0038] (4) Compare T1 with T0. If T1 > T0, use the median-based maximum inter-class variance to obtain the lower threshold in the gray value region [t1, T0-1] and the median-based maximum inter-class variance to obtain the upper threshold in the gray value region [T0+1, T1-1]. If T1 < T0, use the median-based maximum inter-class variance to obtain the lower threshold in the gray value region [T1+1, T0-1] and the median-based maximum inter-class variance to obtain the upper threshold in the gray value region [T0+1, t2]. Assign all pixels between the upper and lower thresholds to the target and the remaining pixels to the background.

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] This invention provides a method for extracting vascular contours from coronary angiography images. This method combines the vascular contours from coronary angiography images with a two-dimensional cross-sectional image of the lumen obtained by an optical coherence tomography (OCT) system, allowing for rapid and clear observation of the morphology of the same point within and outside the vessel. This invention extracts the vascular contours from coronary angiography images using a simple algorithm. This method is simple to operate and easy to implement. Attached Figure Description

[0041] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.

[0042] Figure 1 This is a schematic diagram of the coronary angiography imaging principle in this invention;

[0043] Figure 2 This is a schematic diagram of the steps in the coronary angiography image segmentation algorithm of the present invention. Detailed Implementation

[0044] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0045] This invention provides a method for extracting vessel contours from coronary angiography images, comprising:

[0046] (a) The coronary angiography image of the coronary vessel under test is obtained by direct extraluminal imaging of the coronary vessel under test using a coronary angiography system;

[0047] (b) Intraluminal scanning of the coronary artery under test was performed using an optical coherence tomography (OTC) scanning system to obtain a two-dimensional cross-sectional image of the coronary artery under test.

[0048] (c) Noise reduction of coronary angiography images of coronary vessels;

[0049] (d) Enhance the noise-reduced coronary angiography images;

[0050] (e) Segmenting the enhanced coronary angiography image;

[0051] (f) Extracting coronary artery contour maps from the segmented coronary angiography images;

[0052] (g) The extracted coronary artery contour map is registered and combined with the intraluminal two-dimensional cross-sectional image.

[0053] The aforementioned coronary angiography system includes an X-ray generator, an image detector, a display, a catheter, and a support system. The X-ray generator includes a high-voltage transformer, an X-ray tube, and a filter.

[0054] The aforementioned optical coherence tomography (OCT) system includes an interferometric imaging system and a catheter system. The catheter system includes a rotary connector and a catheter. The rotary connector controls the rotation and retraction of the catheter within the blood vessel and connects the catheter system and the interferometric imaging system. The catheter includes a fiber optic probe and a protective sheath for probing the coronary artery wall.

[0055] In one embodiment of the present invention, such as Figure 1 As shown, step (a) includes: an X-ray beam generated by an X-ray tube is directed toward the patient through a filter and a collimator. Upon contact with the patient, most of the X-rays are reflected or absorbed, while the remaining X-rays pass through completely and reach the image detector to generate a coronary angiography image. The image is then transmitted to a central processing unit for processing and displayed on a monitor. The coronary angiography image is stored in DICOM file format, which allows information exchange between systems from different manufacturers.

[0056] In an embodiment of the present invention, step (b) includes: using an optical fiber-based catheter to transmit a focused laser beam to the vessel wall; the laser scans the surface of the lumen by rotating the catheter to generate a cross-sectional view of the vessel. The inserted catheter is rotated within a protective sheath to obtain a two-dimensional cross-sectional image of the coronary artery within the lumen, and simultaneously retracted to generate multiple frames, resulting in the original two-dimensional cross-sectional image of the entire vessel segment. The original two-dimensional cross-sectional image reconstructed based on the detector's electrical signal is an image in a polar coordinate system. The longitudinal direction represents tissue signals at different depths, and the transverse direction represents the scanned image of the entire cross-section obtained by rotating the imaging catheter one revolution. Through coordinate system transformation and interpolation calculation, the original two-dimensional cross-sectional image is transformed from polar coordinates to a Cartesian coordinate system. The converted two-dimensional cross-sectional image in the Cartesian coordinate system better conforms to the actual anatomical structure of the vessel. From the segmented image of the main branch vessel, information on the lumen of the main branch vessel and the centerline of the main branch vessel can be obtained, and the lumen diameter of the vessel cross-section at each point on the centerline can also be calculated.

[0057] In an embodiment of the present invention, step (c) performs noise reduction using a median filtering method, specifically including:

[0058] (1) Arrange the pixels of a certain pixel in the coronary angiography image from smallest to largest;

[0059] (2) Extract the odd number of data points within a certain pixel in the coronary angiography image and sort them in ascending order;

[0060] (3) Take the median value after sorting and replace the value of the pixel to obtain the noise-reduced coronary angiography image;

[0061] (4) Output the noise-reduced coronary angiography image.

[0062] In an embodiment of the present invention, step (d) includes:

[0063] (1) Input the denoised coronary angiography image and generate pixel matrix I;

[0064] (2) Initialize the scale parameter β = β1;

[0065] (3) Traverse all pixels I(x,y) in the pixel matrix I, calculate the convolution of pixel I(x,y) with the second derivative of the Gaussian function according to the scale parameter β, and obtain the convolution result; calculate the Hessian matrix H and its eigenvalues ​​according to the convolution result; calculate the output value V of the filter enhancement at the current scale according to the Hessian matrix H and its eigenvalues.

[0066] (4) Update the scaling parameter β→β+s;

[0067] (5) Check whether the scale parameter β exceeds the preset scale range. If it does, proceed to step (6); otherwise, return to step (3). The preset scale range β[β...] min ,β max The iteration step size is s;

[0068] (6) After the scale parameter iteration ends, calculate the maximum value V of the filtered enhancement output value of all pixels under each scale parameter. max This refers to the enhanced output of each pixel;

[0069] (7) The enhanced output of each pixel is multiplied by its corresponding scale parameter to obtain the output result of each pixel;

[0070] (8) Output the enhanced coronary angiography image based on the output results of each pixel.

[0071] In one embodiment of the present invention, such as Figure 2 As shown, step (e) includes:

[0072] (1) Count the number of pixels corresponding to gray value T in the enhanced coronary angiography image, where t1≤T≤t2. Based on the statistical results, find the gray value T0 with the largest number of pixels other than t1 and t2.

[0073] (2) Based on the grayscale value T0, the coronary angiography image is divided into target and background. The probabilities p1 and p2 of target pixels and background pixels are calculated respectively. The overall median m0, background median m1, and target (foreground) median m2 of the coronary angiography image are calculated. The overall median m0, background median m1, and target median m2 are used to replace the overall mean u0, background mean u1, and target mean u2 in the Otsu algorithm, respectively. Then the inter-class variance is:

[0074] σ 2 =p1(m1-m0) 2 +p2(m2-m0) 2

[0075] (3) Iterate through all gray values ​​in t1-t2, calculate the inter-class variance according to the method in step (2), and obtain the maximum inter-class variance and the corresponding T1; T1 = arg Maxσ 2 (t1≤T≤t2).

[0076] (4) Compare T1 with T0. If T1 > T0, use the median-based maximum inter-class variance (MAV) to obtain the lower threshold in the grayscale region [t1, T0-1] and the median-based MAV to obtain the upper threshold in the grayscale region [T0+1, T1-1]. If T1 < T0, use the median-based MAV to obtain the lower threshold in the grayscale region [T1+1, T0-1] and the median-based MAV to obtain the upper threshold in the grayscale region [T0+1, t2]. This allows us to obtain the upper and lower thresholds for the grayscale values ​​of the coronary angiography image. Pixels falling between the upper and lower thresholds are classified as foreground, and the remaining pixels are classified as background.

[0077] In one embodiment of the present invention, step (f) includes:

[0078] (1) Image I can be viewed as a two-dimensional surface composed of pixel two-dimensional coordinates and their corresponding gray values, and can be represented as:

[0079] C = {(u,v,I)I = I(u,v)}

[0080] Where (u,v) are the position coordinates of a pixel in image I, and I(u,v) is the gray value of the pixel at coordinates (u,v).

[0081] Point P is located at (u,v). The curvature of the surface at point P can be represented by a Hessian matrix, where I uu (p), I uv (p), I vu (p) and I vv (p) is the second derivative of image I(p).

[0082] (2) Obtained by convolving the image with the second derivative of a Gaussian filter (G(P,s)) of a certain scale:

[0083] I x，y (P, α) = α 2 *G x，y (P, α)*I(P);

[0084] G(u, v, α) = {exp[-(u...v, α)} 2 +v 2 / 2α 2 )]} / (2πα 2 ) 1 / 2 ;

[0085] x and y represent u or v; the obtained second derivative is related to the scale α, and convolution can reduce the influence of noise.

[0086] (3) The blood vessel similarity function is constructed as follows:

[0087]

[0088] Here, λ1 and λ2 are two eigenvalues ​​of the Hessian matrix at point X0(x,y) in the image, and satisfy |λ1(x,y)|≥|λ2(x,y)|. The value of the Z(x,y,λ1) function determines the probability that pixel point X0(x,y) belongs to a blood vessel point. The larger the value of Z(x,y,λ1), the greater the probability that the pixel point belongs to a blood vessel point.

[0089] (4) Construct the multi-scale similarity measure function as shown below;

[0090]

[0091] (5) Process the image using the Hessian matrix and output the coronary artery contour map.

[0092] In one embodiment of the present invention, step (g) includes: using coronary angiography images and their frame rates, intraluminal two-dimensional cross-sectional images of the coronary artery obtained by optical coherence tomography system and their frame rates, and the speed and retraction distance of catheter retraction imaging to achieve fusion of intraluminal two-dimensional cross-sectional images with coronary angiography images.

Claims

1. A method for extracting vessel contours from coronary angiography images, characterized in that, include: (a) The coronary angiography image of the coronary vessel under test is obtained by direct extraluminal imaging of the coronary vessel under test using a coronary angiography system; (b) Intraluminal scanning of the coronary artery under test is performed using an optical coherence tomography system to obtain a two-dimensional cross-sectional image of the coronary artery under test. (c) Noise reduction of coronary angiography images; (d) Enhance the noise-reduced coronary angiography images; (e) Segmenting the enhanced coronary angiography image; Specifically, it includes: (1) Count the number of pixels corresponding to gray value T in the enhanced coronary angiography image, where t1≤T≤t2. Based on the statistical results, find the gray value T0 with the largest number of pixels other than t1 and t2. (2) Divide the coronary angiography image into target and background based on the gray value T0, calculate the probabilities p1 and p2 of target pixels and background pixels respectively, and calculate the overall median m0, background median m1, and target median m2 of the coronary angiography image. Replace the overall mean u0, background mean u1, and target mean u2 in the Otsu algorithm with the overall median m0, background median m1, and target median m2 respectively. Then the inter-class variance is: σ 2 =p1(m1-m0) 2 +p2(m2-m0) 2 (3) Iterate through all gray values ​​in t1-t2, calculate the inter-class variance according to the method in step (2), and obtain the maximum inter-class variance and the corresponding T1; T1=arg Maxσ 2 ; (4) Compare T1 with T0. If T1 > T0, use the median-based maximum inter-class variance to obtain the lower threshold in the gray value region [t1, T0-1] and the median-based maximum inter-class variance to obtain the upper threshold in the gray value region [T0+1, T1-1]. If T1 < T0, use the median-based maximum inter-class variance to obtain the lower threshold in the gray value region [T1+1, T0-1] and the median-based maximum inter-class variance to obtain the upper threshold in the gray value region [T0+1, t2]. Assign all pixels between the upper and lower thresholds to the target and the remaining pixels to the background. (f) Extract the coronary artery contour map from the segmented coronary angiography image; (g) Using coronary angiography images and their frame rates, intraluminal two-dimensional cross-sectional images and their frame rates, and the speed and distance of catheter retraction, the intraluminal two-dimensional cross-sectional images and coronary vessel contour maps are registered and fused.

2. The method for extracting vascular contours from coronary angiography images according to claim 1, characterized in that, Step (a) specifically includes: directing the X-ray beam generated by the X-ray tube through a filter and a collimator toward the coronary artery to be tested, and the X-rays passing through the coronary artery to be tested and reaching the image detector to generate a coronary angiography image.

3. The method for extracting vascular contours from coronary angiography images according to claim 1, characterized in that, Step (b) specifically includes: inserting a fiber-optic catheter into the coronary artery, rotating the inserted catheter within a protective sheath to obtain a two-dimensional cross-sectional image of the coronary artery, and simultaneously retracting the catheter to generate multiple frames of images to obtain a two-dimensional cross-sectional image of the entire coronary artery.

4. The method for extracting vascular contours from coronary angiography images according to claim 1, characterized in that, Step (c) involves denoising the coronary angiography images using median filtering.

5. The method for extracting vascular contours from coronary angiography images according to claim 4, characterized in that, Step (c) specifically includes: (1) Arrange the pixel values ​​of the pixels in the coronary angiography image from smallest to largest; (2) Extract the odd number of data points within the pixels of the coronary angiography image and sort them in ascending order; (3) Replace the value of the pixel with the median value after sorting to obtain the noise-reduced coronary angiography image; (4) Output the noise-reduced coronary angiography image.

6. The method for extracting vascular contours from coronary angiography images according to claim 1, characterized in that, Step (d) specifically includes: (1) Input the denoised coronary angiography image and generate pixel matrix I; (2) Initialize the scale parameter β = β1; (3) Traverse all pixels I(x,y) in the pixel matrix I, calculate the convolution of pixel I(x,y) with the second derivative of the Gaussian function according to the scale parameter β, and obtain the convolution result; calculate the Hessian matrix H and its eigenvalues ​​according to the convolution result; calculate the output value V of the filter enhancement at the current scale according to the Hessian matrix H and its eigenvalues. (4) Update the scale parameter β→β+s; (5) Check whether the scale parameter β exceeds the preset scale range. If it does, proceed to step (6); if it does not exceed the preset scale range, return to step (3). min ,β max The iteration step size is s; (6) After the scale parameter iteration ends, calculate the maximum value V of the filtered enhancement output value of all pixels under each scale parameter. max This refers to the enhanced output of each pixel; (7) The enhanced output of each pixel is multiplied by its corresponding scale parameter to obtain the output result of each pixel; (8) Output the enhanced coronary angiography image based on the output results of each pixel.

Images