A Coronary Artery Growth Method and System Based on CTA Images

Abstract

The present invention provides a coronary artery growth method and system based on CTA images. For seed point generation, a regression algorithm is used to obtain the probability that a point in the initial mask is a seed point, and the first screening is performed according to the distance from the left and right coronary artery ostium points. Then, the second screening is performed according to the probability size and interval sampling. For the coronary artery growth model, most bright and evenly distributed Hu vessels are segmented. For the calcified growth model, the centerline is traced and identified within the vascular calcification lesions to improve the stability of the centerline. For the stenosis and small vessel growth patterns, the vascular stenosis lesions and vessels under low Hu are traced and identified. The edge of the vascular segmentation is also optimized by graph cut.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image processing, and particularly to a coronary artery growth method and system based on CTA images. Background Art

[0002] A CTA (Computed Tomography Angiography) system is a technical system that helps physicians diagnose and evaluate cardiovascular diseases and conditions from CT images after injecting a contrast agent (enhancer) into the blood vessels of the body, such as aneurysms, arterial occlusions, etc.

[0003] Cardiac CTA is one of the important means for early screening of the heart. Accurately segmenting coronary arteries from CTA data can not only provide a quantitative description of the vascular structure, but also observe and compare the structural changes of blood vessels, which is of great significance for the diagnosis and treatment of diseases. Existing coronary artery segmentation methods include segmentation algorithms based on traditional image processing, segmentation algorithms based on machine learning segmentation networks such as Unet, and growth methods based on seed points. Unet is slow in direct segmentation, with stenosis and small blood vessel growth fractures or non-growth, resulting in low accuracy and a large demand for training data. Using the traditional region growing method, the accuracy is not high. The existing method using seed point growth has unstable termination conditions; since the proportion of calcified cases in the total case samples is low, and the proportion of calcified regions in calcified cases is also low, the deep learning model cannot learn too much useful information, resulting in poor handling of calcified blood vessels, and small blood vessels or vascular stenosis are prone to fractures or non-growth.

[0004] Moreover, in practical applications, the accuracy and speed of the algorithm are often inversely proportional. When quickly segmenting coronary arteries from a large amount of CTA data, the accuracy usually needs to be reduced. Summary of the Invention

[0005] In order to overcome the above technical defects, the purpose of the present invention is to provide a coronary artery growth method and system based on CTA images with high segmentation accuracy and stable segmentation process.

[0006] The present invention discloses a coronary artery growth method based on CTA images, including the following steps: obtaining an initial mask of the coronary artery according to CTA data; obtaining the probability of all points in the initial mask as seed points through a regression algorithm, and obtaining a pre-seed point set and a rejection point set according to a preset distance and a first preset probability threshold; performing secondary screening in the rejection point set to obtain several points with the maximum probability within the preset distance, and jointly forming a seed point set with the pre-seed point set; using the seed point set to perform the growth process of the coronary artery, thereby obtaining a coronary artery growth map.

[0007] Preferably, the growth process includes: a base image of size n*n*n cut with the current seed point as the center; generating a number of direction probabilities and an initial radius through the base image; selecting the direction vector with the maximum probability among the number of direction probabilities; drawing a base sphere with the current seed point as the center and the initial radius as the radius, and new seed points are formed at the intersections of the surface of the base sphere and the direction vector; then taking the new seed points as the center to cut a base image of size n*n*n, and repeating the above process.

[0008] Preferably, the selecting the direction vector with the maximum probability among the number of direction probabilities includes: selecting the direction vector with the maximum probability among the number of direction probabilities as the first direction vector; and a number of direction vectors with an angle of α or β with the first direction vector, and selecting the one with the maximum probability as the second direction vector; the angle α is greater than 90 degrees and less than 180 degrees, and β is greater than 0 degrees and less than 90 degrees; the forming new seed points at the intersections of the surface of the base sphere and the direction vector includes: forming new seed points at the intersections of the surface of the base sphere and the second direction vector.

[0009] Preferably, the growth process further includes: calculating the average value of Hu of the current seed point and a number of seed points that have undergone the above growth process before the current seed point, and stopping the growth when the average value is less than the preset CT average value; obtaining the preset CT average value according to the minimum Hu of all coronary angiography blood vessels.

[0010] Preferably, the growth process further includes: calculating the length of each coronary artery grown from the initial seed point, and stopping the growth when the length of each coronary artery grown from the initial seed point is greater than the preset length; obtaining the preset length according to the length of the longest single blood vessel among all coronary arteries.

[0011] Preferably, the growth process further includes: calculating the distance between the current seed point and the center point of the existing coronary artery or the opening points of two coronary arteries; when the distance is less than the preset distance, stopping the growth, and adding the center line and radius grown from the initial seed point to the coronary artery list.

[0012] Preferably, the growth process further includes: when the average value of Hu is less than the Hu threshold of the preset stenotic blood vessels and / or small blood vessels, stopping the growth and entering the growth process of stenotic blood vessels and small blood vessels; obtaining the Hu threshold of the preset stenotic blood vessels and / or small blood vessels according to all stenotic blood vessels and small blood vessels.

[0013] Preferably, the growth process further includes: calculating the average Hu value of the points in the n*n*n image that are greater than the first preset Hu; when the average Hu value of the points in the n*n*n image that are greater than the first preset Hu is higher than a preset value of the numerical value of the seed points that have undergone the above growth process, and there are points with Hu greater than the second preset Hu, it is determined that calcification exists here, and the coronary artery growth is stopped and the calcification growth process is entered; the second preset Hu is greater than the first preset Hu.

[0014] Preferably, after obtaining the coronary artery growth map by using the seed point set for the growth process of the coronary artery, the method further includes: generating a coronary artery mask according to the center line and radius of the seed points, and then inputting the area of the coronary artery mask as a known area into the graph cutting algorithm to adjust the coronary artery mask so as to obtain a finer mask at the vascular calcification site.

[0015] The present invention also discloses a coronary artery growth system based on CTA images, including a deep learning regression model, a coronary artery growth model, a narrow blood vessel and small blood vessel growth model, and a calcification growth model; the coronary artery growth model includes a growth model and a stop model; obtaining an initial mask of the coronary artery based on CTA data; calculating the probability that all points in the initial mask are seed points through the deep learning regression model, and obtaining a pre-seed point set and a rejection point set according to a preset distance and a first preset probability threshold; performing secondary screening on the rejection point set to obtain several points with the maximum probability within the preset distance, and jointly forming a seed point set with the pre-seed point set; using the seed point set to perform the growth process of the coronary artery through the growth model and the stop model, growing narrow blood vessels and small blood vessels through the narrow blood vessel and small blood vessel growth model, and performing calcification growth through the calcification growth model, so as to obtain a coronary artery growth map.

[0016] After adopting the above technical solutions, compared with the prior art, the following beneficial effects are achieved:

[0017] 1. By performing secondary screening on the seed points, the main coronary arteries grow preferentially, thereby improving the accuracy and efficiency of coronary artery segmentation; by setting multiple growth stop judgment conditions and providing separate growth models for calcification, stenosis, and small blood vessels, the segmentation accuracy is higher. Description of the Drawings

[0018] Figure 1 It is a flowchart of the coronary artery growth method provided by the present invention;

[0019] Figure 2 It is a diagram of the coronary artery growth stop judgment conditions provided by the present invention;

[0020] Figure 3Coronary artery growth blood vessel diagram of the coronary artery growth method provided by the present invention;

[0021] Figure 4 Calcification growth diagram of the coronary artery growth method provided by the present invention;

[0022] Figure 5 Narrow blood vessel and small blood vessel growth diagram of the coronary artery growth method provided by the present invention;

[0023] Figure 6 Coronary artery growth effect diagram of the coronary artery growth method provided by the present invention. Detailed implementation manners

[0024] The advantages of the present invention are further elaborated below in conjunction with the accompanying drawings and specific embodiments.

[0025] Exemplary embodiments will be described in detail herein, which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present disclosure as detailed in the appended claims.

[0026] The terms used in the present disclosure are for the purpose of describing specific embodiments only and are not intended to limit the present disclosure. The singular forms "a", "said" and "the" used in the present disclosure and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term "and / or" as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items.

[0027] It should be understood that although the terms first, second, third, etc. may be used in the present disclosure to describe various information, such information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of the present disclosure, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "when" or "while" or "in response to determining".

[0028] In the description of the present invention, it should be understood that the orientation or positional relationship indicated by the terms "longitudinal", "transverse", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. is based on the orientation or positional relationship shown in the drawings. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation to the present invention.

[0029] In the description of the present invention, unless otherwise specified and limited, it should be noted that the terms "installation", "connection", and "coupling" should be understood in a broad sense. For example, it can be a mechanical connection or an electrical connection, or it can be the communication inside two elements. It can be directly connected or indirectly connected through an intermediate medium. For those of ordinary skill in the art, the specific meanings of the above terms can be understood according to specific circumstances.

[0030] In the subsequent description, the use of suffixes such as "module", "component", or "unit" for representing elements is only for the convenience of explaining the present invention, and it has no specific meaning by itself. Therefore, "module" and "component" can be used interchangeably.

[0031] See the appendix Figure 1 , for the segmentation process of coronary artery vessels, the present invention proposes a classification and tracking method, that is, the coronary artery growth method based on CTA images of the present invention, and different network models are provided for calculation under different circumstances, so as to improve the stability of the algorithm.

[0032] As support for the above method, the present invention first provides a coronary artery growth system based on CTA images, including a deep learning regression model, a coronary artery growth model, a stenosis vessel and small vessel growth model, and a calcification growth model. In order to accurately grasp the stopping growth condition, the coronary artery growth model includes a growth model and a stopping model, which are calculated separately to ensure the accuracy and stability of the segmentation.

[0033] First, obtain the initial mask of the coronary artery according to the CTA data, obtain the probability of all points in the initial mask as seed points through the deep learning regression model, and obtain the pre-seed point set and the rejection point set according to the preset distance and the first preset probability threshold; perform secondary screening on the rejection point set to obtain the points with the maximum probability within a certain preset distance, and jointly form the seed point set with the pre-seed point set. Then, use the seed point set to perform the growth process of the coronary artery through the growth model and the stopping model, grow the stenosis vessels and small vessels through the stenosis vessel and small vessel growth model, and perform calcification growth through the calcification growth model, so as to obtain the coronary artery growth map. Finally, fine-tune the coronary artery growth map to optimize the boundary of the coronary artery.

[0034] The mask in the image processing process uses a selected image, graphic, or object to block the processed image (entirely or partially) to control the area or process of image processing. The specific image or object used for covering is called a mask or template. In optical image processing, the mask can be a film, a filter, etc. This invention is mainly used for digital image processing. In digital image processing, the mask is a two-dimensional matrix array, and sometimes a multi-valued image is also used.

[0035] Supported by the above-mentioned coronary artery growth system based on CTA images, this invention provides a coronary artery growth method based on CTA images, including a seed point generation process, a coronary artery growth process, a stenotic vessel and small vessel growth process, a calcification growth process, and an image cutting optimization process.

[0036] For the seed point generation process, first, the blood vessels in the CTA data are subjected to Gaussian filtering at multiple scales, and then the second-order derivatives of each pixel point are calculated to construct a Hessian matrix. Through this matrix, three eigenvalues of each point on the image can be obtained. Here, the larger the spatial curvature eigenvalue that can characterize the blood vessel change, the greater the spatial curvature of the point in the corresponding eigenvector direction. And as a tubular object, the maximum eigenvalue of the blood vessel will be much larger than the other two eigenvalues. Therefore, this characteristic can be used to determine whether the current pixel point belongs to the blood vessel structure. After performing Hessian enhancement on the image, the image is segmented using a preset reasonable region threshold to try to extract all regions with blood vessel structures as much as possible. This region serves as the initial mask of the blood vessel.

[0037] After obtaining the initial mask, seed points are obtained in the initial mask. Specifically, first obtain the left and right coronary artery opening points, calculate the probability that all points in the initial mask are seed points through the regression algorithm of a pre-trained deep learning regression model, sort the seed points according to the probability size, and delete the seed points with a probability less than the first preset probability threshold within a preset area (the area with a preset distance from the two coronary artery opening points). The deleted points are called the culling point set, and the remaining points are called the pre-seed point set. In order to give priority to the growth of the main coronary arteries to improve accuracy and efficiency, the culling point set after deletion is secondarily screened, that is, according to the preset distance, the seed point with the largest probability in each area is selected. The seed point screening of the present invention first screens out the seed points with the largest probability within the preset area according to the distance from the two coronary artery opening points and the first preset probability threshold, and then according to the preset distance, selects the seed point with the largest probability in each area, which together with the pre-seed point set forms the seed point set. For example, after screening according to the first preset probability threshold, 1000 seed points are obtained. Some of these 1000 seed points may be very concentrated in one area. In order to make the seed points as scattered as possible (which can reduce the calculation amount and also make the coronary artery accuracy higher), the point with the largest probability within the preset distance is retained at each opening point, and other points within this preset distance are discarded. Then, using the seed points obtained above, the point with the largest probability within the area is obtained according to the preset distance, and other seed points within the area are discarded. This secondary screening is continuously repeated in this way.

[0038] For the coronary artery growth process, the seed point set is used for the growth process of the coronary artery to obtain the coronary artery growth map.

[0039] Specifically, a base image with a size of n*n*n cut with the current seed point as the center is used as the input and sent to the growth model. The growth model then outputs a number of direction probabilities and an initial radius. The number of direction probabilities is not limited and can be 500 or 1000.

[0040] Then, the direction vector with the largest probability is selected from the number of direction probabilities as the first direction vector, and a number of direction vectors with an angle of α with the first direction vector are selected, and the one with the largest probability is selected from them as the second direction vector.

[0041] A base sphere is drawn with the current seed point as the center and the initial radius as the radius. A new seed point is formed at the intersection of the surface of the base sphere and the second direction vector.

[0042] Then, a base image with a size of n*n*n cut with the new seed point as the center is used, and the above process is repeated. See the appendix Figure 3 。

[0043] It should be noted that if the included angle between the selected second direction vector and the first direction vector was α in the previous step, then in the current step, the included angle with the first direction vector is β. The angle α is any angle greater than 90 degrees and less than 180 degrees, and β is any angle greater than 0 degrees and less than 90 degrees.

[0044] In addition to the growth process, the stopping process also needs to be controlled. See the appendix Figure 2 For this, the present invention provides five stopping judgment conditions:

[0045] 1. Calculate the distance between the current seed point and the center point of the existing coronary artery or the opening points of the two coronary arteries; when this distance is less than the preset distance, stop growing, and add the center line and radius grown from the initial seed point to the coronary artery list.

[0046] 2. Calculate the average value of Hu of the current seed point and the previous several (seed points that have undergone the above growth process) seed points. When the average value is less than the preset CT average value, stop growing. The preset CT average value is obtained based on the minimum Hu of all coronary angiography blood vessels.

[0047] 3. Calculate the length of each coronary artery grown from the initial seed point. When the length of each coronary artery grown from the initial seed point is greater than the preset length, stop growing; the preset length is obtained based on the length of the longest single blood vessel among all coronary arteries.

[0048] 4. Calculate the probability that this seed point is not a coronary artery according to the probability distribution P of the direction vector (i.e., several direction probabilities output by the growth model). Specifically, according to the formula And calculate the average probability P0 together with the previous several (seed points that have undergone the above growth process) seed points; different preset P0 thresholds are set for different Hus of the seed points. When the average probability P0 is less than the preset P0 threshold, stop growing.

[0049] 5. When the distance from the already grown coronary artery is less than the preset value, stop growing.

[0050] In addition to setting the stopping growth judgment conditions, the present invention also provides a growth method for stenotic blood vessels and small blood vessels, specifically: preset the Hu threshold of the preset stenotic blood vessels and / or small blood vessels according to all stenotic blood vessels and small blood vessels; during the growth process, when the average Hu value is less than the Hu threshold of the preset stenotic blood vessels and / or small blood vessels, stop growing and enter the growth process of stenotic blood vessels and small blood vessels.

[0051] The present invention also provides centerline tracking for vascular stenosis diseases and low Hu, specifically including tracking model design and data generation and augmentation. For model design: the input is an n*n*n-sized image cut centered on the seed point, and the final output is a number of direction probabilities and a radius. The growth method is the same as that of coronary artery growth, as Figure 4 shown. For data generation and augmentation, the data includes a part of stenosis data, a part of the tail blood vessels, and a part of small blood vessels. Data augmentation can be performed by operations such as rotation, cropping, and gamma transformation.

[0052] The present invention also provides a calcification growth method, specifically: statistically calculate the average Hu value of the points with Hu greater than the first preset Hu in the n*n*n image; when the average Hu value of the points with Hu greater than the first preset Hu in the n*n*n image is higher than a preset value of the numerical value of the seed point that has undergone the above growth process, and there are points with Hu greater than the second preset Hu, it is determined that there is calcification here, and the coronary artery growth is stopped and the calcification growth process is entered. Here, the second preset Hu is greater than the first preset Hu. For example, the first preset Hu is 100 and the second preset Hu is 800.

[0053] The present invention also provides centerline tracking for under calcification, specifically including tracking model design and data generation and augmentation. For model design: the input is an n*n*n-sized image cut centered on the seed point, and the final output is a number of direction probabilities and a radius. The growth method is the same as that of coronary artery growth, as Figure 5 shown. For data generation and augmentation, the data includes a part of actual calcification data and a part of simulated calcification data. Data augmentation can be performed by operations such as rotation, cropping, and gamma transformation.

[0054] After obtaining the coronary artery growth map, the coronary artery boundary is further optimized, specifically: generating a coronary artery mask according to the centerline and radius of the seed point, and then using the area of the coronary artery mask as the known area to input the graph cutting algorithm to adjust the coronary artery mask so as to obtain a more refined mask at the vascular calcification site.

[0055] Under the same network structure, the present invention: by adding training for special cases, conducting a detailed analysis of special disease locations, and setting more reasonable training data, achieves excellent results with a smaller network structure and fewer parameters, as Figure 6 shown. Thereby improving the stability and segmentation speed of the overall coronary artery segmentation.

[0056] It should be noted that the embodiments of the present invention have better implementability and do not impose any form of limitation on the present invention. Any person skilled in the art may use the technical content disclosed above to modify or transform it into equivalent effective embodiments. However, as long as the content does not depart from the technical solution of the present invention, any modification, equivalent change, or modification made to the above embodiments based on the technical essence of the present invention still falls within the scope of the technical solution of the present invention.

Claims

1. A coronary artery growth method based on CTA images, characterized in that, It includes the following steps: Obtain the initial mask of the coronary artery based on CTA data; Calculate the probability that all points in the initial mask are seed points through a regression algorithm, and obtain a pre-seed point set and a rejection point set according to a preset distance and a first preset probability threshold; Perform secondary screening on the rejection point set, obtain several points with the maximum probability within the preset distance, and jointly form a seed point set with the pre-seed point set; Use the seed point set to perform the growth process of the coronary artery, thereby obtaining a coronary artery growth map; Among them, the growth process includes: A base image with a size of n*n*n cut centered on the current seed point; Generate several direction probabilities and an initial radius through the base image; Select the direction vector with the maximum probability among the several direction probabilities; Draw a base sphere with the current seed point as the center and the initial radius as the radius, and new seed points are formed at the intersection of the surface of the base sphere and the direction vector; Then cut a base image with a size of n*n*n centered on the new seed point, and repeat the above process; Among them, the selection of the direction vector with the maximum probability among the several direction probabilities includes: Select the direction vector with the maximum probability among the several direction probabilities as the first direction vector; And several direction vectors with an angle of α or β with the first direction vector, and select the one with the maximum probability as the second direction vector; the α angle is greater than 90 degrees and less than 180 degrees, and β is greater than 0 degrees and less than 90 degrees; The formation of new seed points at the intersection of the surface of the base sphere and the direction vector includes: new seed points are formed at the intersection of the surface of the base sphere and the second direction vector.

2. The coronary artery growth method according to claim 1, characterized in that The growth process further includes: Calculate the average value of Hu of the current seed point and several seed points that have undergone the above growth process before the current seed point. When the average value is less than the preset CT average value, stop growing; Obtain the preset CT average value according to the minimum Hu of all coronary angiography blood vessels.

3. The coronary artery growth method according to claim 2, characterized in that, The growth process further includes: Calculate the length of each coronary artery grown from the initial seed point. When the length of each coronary artery grown from the initial seed point is greater than the preset length, stop growing; Obtain the preset length according to the length of the longest single blood vessel among all coronary arteries.

4. The coronary artery growth method according to claim 3, wherein The growth process further includes: Calculate the distance between the current seed point and the center point of the existing coronary artery or the opening points of two coronary arteries; when the distance is less than the preset distance, stop growing, and add the center line and radius grown from the initial seed point to the coronary artery list.

5. The coronary artery growth method according to claim 1, characterized in that, The growth process further includes: When the average Hu value is less than the Hu threshold of the preset stenotic blood vessels and / or small blood vessels, stop growing and enter the growth process of stenotic blood vessels and small blood vessels; Obtain the Hu threshold of the preset stenotic blood vessels and / or small blood vessels according to all stenotic blood vessels and small blood vessels.

6. The coronary artery growth method according to claim 1, characterized in that, The growth process further includes: Statistically calculate the average value of Hu of the points in the n*n*n image that are greater than the first preset Hu; When the average value of Hu of the points greater than the first preset Hu in the n*n*n image is higher than a preset value of the numerical value of the seed points that have undergone the above growth process, and there are points with Hu greater than the second preset Hu, it is determined that calcification exists here, and the coronary artery growth stops and enters the calcification growth process; The second preset Hu is greater than the first preset Hu.

7. The coronary artery growth method according to claim 1, characterized in that, After obtaining the coronary artery growth map by using the seed point set to perform the growth process of the coronary artery, it further includes: Generating a coronary artery mask according to the center line and radius of the seed points, and then inputting the area of the coronary artery mask as a known area into the graph cutting algorithm to adjust the coronary artery mask so as to obtain a finer mask at the vascular calcification site.

8. A coronary artery growth system based on CTA images, characterized in that, It includes a deep learning regression model, a coronary artery growth model, a narrow vessel and small vessel growth model, and a calcification growth model; the coronary artery growth model includes a growth model and a stop model; Obtaining an initial mask of the coronary artery based on the CTA data; Calculating the probability that all points in the initial mask are seed points through the deep learning regression model, and obtaining a pre-seed point set and a rejection point set according to a preset distance and a first preset probability threshold; Performing secondary screening in the rejection point set to obtain several points with the maximum probability within the preset distance, and jointly forming a seed point set with the pre-seed point set; Using the seed point set to perform the growth process of the coronary artery through the growth model and the stop model, growing narrow vessels and small vessels through the narrow vessel and small vessel growth model, and performing calcification growth through the calcification growth model, so as to obtain a coronary artery growth map; Among them, the growth process includes: A base image with a size of n*n*n cut with the current seed point as the center; Generating a number of direction probabilities and an initial radius through the base image; Selecting the direction vector with the maximum probability among the number of direction probabilities; Drawing a base sphere with the current seed point as the center and the initial radius as the radius, and new seed points are formed at the intersection of the surface of the base sphere and the direction vector; Then using the new seed points as the center to cut a base image with a size of n*n*n, and repeating the above process; Among them, the selecting the direction vector with the maximum probability among the number of direction probabilities includes: Selecting the direction vector with the maximum probability among the number of direction probabilities as the first direction vector; And a number of direction vectors with an angle of α or β with the first direction vector, and selecting the one with the maximum probability as the second direction vector; the α angle is greater than 90 degrees and less than 180 degrees, and β is greater than 0 degrees and less than 90 degrees; The formation of new seed points at the intersection of the surface of the base sphere and the direction vector includes: the formation of new seed points at the intersection of the surface of the base sphere and the second direction vector.

Images