Intelligent Identification System for Coronary Atherosclerosis Based on DSA Imaging

Abstract

This invention relates to the field of image processing technology, specifically to an intelligent coronary atherosclerosis recognition system based on DSA images. The system includes: constructing an initial topological structure of the patient's coronary arteries using a topology determination module; constructing candidate vascular regions using a lesion recognition module, and determining the actual vascular region based on the similarity between the blood flow propagation path and the edges, as well as the grayscale continuity of the region; identifying stenotic vascular regions, blood flow obstruction regions, and lesion ports within the actual vascular region based on the rate of change in the diameter of sampling points and the grayscale differences between corresponding points on the blood flow propagation path; identifying vascular rupture regions based on the degree of consistency between the connection of the lesion port and the overall vascular orientation; and completing the vascular rupture region using a lesion marking module, and marking the stenotic vascular region, blood flow obstruction region, and vascular rupture region on the obtained actual vascular atlas. This disclosure can solve problems such as incomplete topology and artifact misjudgment, improving the accuracy of lesion detection.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and more specifically to an intelligent identification system for coronary atherosclerosis based on DSA images. Background Technology

[0002] Coronary atherosclerotic heart disease is caused by plaque buildup on the inner walls of blood vessels, leading to narrowing or blockage of the lumen, resulting in insufficient blood supply to the myocardium and even myocardial infarction. Accurate identification of the location, characteristics, and extent of the lesions is a prerequisite for precise diagnosis and treatment. Digital subtraction angiography (DSA) has become a core imaging technique for clinical assessment of coronary artery lesions because it can dynamically and clearly present the state of coronary blood flow and vascular morphology. To accurately locate the lesion, it is necessary to match the patient's DSA image coronary artery tree with a standard coronary atlas to clarify the correspondence between branches and segments.

[0003] Related techniques typically construct coronary artery trees in DSA images using vessel centerline extraction and geometric topology matching. However, due to atherosclerotic lesions, patients' vessels often exhibit morphological abnormalities such as localized stenosis, narrowing of the lumen, high-brightness calcification, or vessel rupture, resulting in incomplete vascular topology. Simultaneously, artifacts in the images are easily generated by plaque calcification, contrast agent retention, and blood flow turbulence, further interfering with the accurate identification of vascular structures. This leads to problems such as broken vessel chains and misconnected branches near the lesion area, significantly reducing the registration accuracy between standard atlases and patient vessels. This makes it difficult to reliably reconstruct the true vascular tree structure, ultimately causing misjudgment or omission of lesion location features, affecting the accuracy of clinical diagnosis. Summary of the Invention

[0004] This invention provides a DSA-based intelligent identification system for coronary atherosclerosis to address the problems in existing technologies, such as vascular chain breaks and branch misconnections near the lesion area. These problems significantly reduce the registration accuracy between standard atlases and patient vessels, making it difficult to reliably reconstruct the real vascular tree structure. Ultimately, this leads to misjudgment or omission of lesion location features, affecting the accuracy of clinical diagnosis.

[0005] The intelligent identification system for coronary atherosclerosis based on DSA imaging of the present invention adopts the following technical solution: One embodiment of the present invention provides a coronary atherosclerosis intelligent identification system based on DSA images, the system comprising the following modules: The topology determination module is used to extract the centerline of the blood vessels from the grayscale image of the patient's coronary artery DSA image and construct the initial topology structure; The lesion identification module is used to construct a blood vessel candidate region with any two edges for local segments in the initial topology where the number of edges is greater than 2, and to determine the real blood vessel region based on the similarity between the blood flow propagation path and the region edge and the region grayscale continuity in the blood vessel candidate region. The lesion identification module is also used to determine the stenotic vascular region based on the rate of change of the diameter between adjacent sampling points in the real vascular region, determine the starting point of blood flow obstruction and obtain the blood flow obstruction region based on the gray difference between corresponding points of sampling points on the blood flow propagation path, and determine the severe mutation region and lesion port based on the starting point of blood flow obstruction and the sampling point with the maximum rate of change of the diameter. The lesion identification module is also used to construct a compensation curve between the current lesion port and its neighboring lesion ports and extend it to obtain a candidate composite path, determine the compensation smoothness of the compensation curve based on the curvature at each sampling point on the candidate composite path, and determine the vascular rupture area based on the compensation smoothness. The lesion marking module is used to complete the ruptured blood vessel region in the initial topology to obtain a real blood vessel map, and to mark the stenotic blood vessel region, the blood flow obstruction region and the ruptured blood vessel region in the real blood vessel map.

[0006] For example, the lesion recognition module determines the real blood vessel region based on the similarity between the blood flow propagation path and the region edge and the region grayscale continuity by executing the following method: uniformly selecting multiple first sampling points on the blood vessel centerline of the blood vessel candidate region; determining the blood flow propagation path based on the grayscale value curve of the first sampling points changing over time; determining the similarity between the blood flow propagation path and two region edges based on a dynamic time warping algorithm and calculating the similarity product, and recording the similarity product as the blood flow trend stability; determining the grayscale mean of the two region edges respectively, and recording the normalized value of the ratio of the absolute value of the difference of the grayscale mean to the larger value of the grayscale mean as the region grayscale continuity; determining the real blood vessel probability of the blood vessel candidate region based on the blood flow trend stability and the region grayscale continuity, and determining the blood vessel candidate region corresponding to the maximum real blood vessel probability as the real blood vessel region.

[0007] For example, the lesion identification module determines the blood flow propagation path based on the grayscale value curve of the first sampling point changing over time by performing the following method: acquiring the grayscale value curve of each first sampling point changing over time within the coronary artery DSA image acquisition cycle; extracting the grayscale peak points in each grayscale value curve and determining the time of occurrence of the grayscale peak points to construct a peak time series; reordering the occurrence times of each peak in the peak time series in chronological order and assigning each peak occurrence time an index number in ascending order to obtain a target index sequence; arranging the first sampling points based on the target index sequence and sequentially connecting the sorted first sampling points to obtain the blood flow propagation path.

[0008] For example, the lesion identification module determines the stenotic vascular region based on the rate of change in diameter between adjacent sampling points in the real vascular region by performing the following method: uniformly selecting multiple second sampling points on the vascular centerline of the real vascular region; obtaining the maximum inscribed circle of the real vascular region at each of the second sampling points, and determining its diameter as the local diameter at the second sampling point; for each second sampling point, determining the smaller value of the local diameter between the current second sampling point and its adjacent sampling points, and calculating the absolute value of the difference between the local diameter of the current second sampling point and its adjacent sampling points, and determining the ratio of the absolute value of the diameter difference to the smaller value of the local diameter as the rate of change in diameter; if the rate of change in diameter between multiple consecutive adjacent second sampling points is greater than a first preset threshold, then the corresponding vascular segment is determined as a stenotic vascular region.

[0009] For example, the lesion recognition module implements the determination of the blood flow obstruction starting point and the acquisition of the blood flow obstruction region based on the gray-level difference between corresponding points of the sampling points on the blood flow propagation path by performing the following method: determining the propagation path point closest to each of the second sampling points on the blood flow propagation path, calculating the absolute value of the gray-level difference between adjacent propagation path points and recording it as the gray-level change value; determining the second sampling point corresponding to the maximum gray-level change value as the blood flow obstruction starting point, and recording the continuous pixels with gray-level values ​​lower than a second preset threshold downstream of the blood flow obstruction starting point as the blood flow obstruction region.

[0010] For example, the lesion identification module determines the severe mutation region and lesion port based on the sampling point of the blood flow obstruction initiation point and the maximum diameter change rate by performing the following method: if the blood flow obstruction initiation point and the second sampling point with the largest diameter change rate are the same point or the sampling interval is less than a third preset threshold, then the blood flow obstruction region corresponding to the blood flow obstruction initiation point is determined as the severe mutation region, and the endpoint of the severe mutation region is determined as the lesion port.

[0011] For example, the lesion identification module implements the construction of a compensation curve between the current lesion port and its neighboring lesion ports and extends it to obtain a candidate composite path by performing the following method: for each lesion port, determining neighboring lesion ports within a target neighborhood of the current lesion port; the target neighborhood is a circular neighborhood centered on the current lesion port and with a radius of 1 / 2 of the length of the blood vessel where the current lesion port is located; constructing the compensation curve between the current lesion port and the neighboring lesion ports and determining its length; extending the length of the compensation curve in the direction of the current lesion port and the neighboring lesion ports respectively to obtain the candidate composite path.

[0012] For example, the lesion recognition module implements the determination of the compensation smoothness of the compensation curve based on the curvature at each sampling point on the candidate composite path by performing the following method: uniformly selecting multiple third sampling points on the candidate composite path, calculating the curvature of the candidate composite path at each of the third sampling points, constructing a first curvature sequence corresponding to the candidate composite path and a second curvature sequence corresponding to the compensation curve; obtaining the curvature at each of the third sampling points remaining after removing the second curvature sequence from the first curvature sequence, constructing a third curvature sequence; and determining the compensation smoothness of the compensation curve based on the first curvature sequence and the third curvature sequence.

[0013] For example, the lesion recognition module achieves the compensation smoothness of determining the compensation curve based on the first curvature sequence and the second curvature sequence by performing the following method, including: calculating the first curvature variance of the first curvature sequence and the second curvature variance of the third curvature sequence; calculating the difference between the first curvature variance and the second curvature variance, denoted as the compensation smoothness.

[0014] For example, the lesion marking module completes the ruptured vascular region in the initial topology to obtain a real vascular atlas by performing the following method, and marks the stenotic vascular region, the blood flow obstruction region, and the ruptured vascular region in the real vascular atlas, including: using the compensation curve corresponding to the maximum compensation smoothness to complete the ruptured vascular region in the initial topology to obtain the real vascular atlas; and visually marking the stenotic vascular region, the blood flow obstruction region, and the ruptured vascular region in the real vascular atlas with different colors.

[0015] The beneficial effects of the technical solution of the present invention are as follows: In the intelligent coronary atherosclerosis recognition system based on DSA images provided by this invention, the topology determination module extracts the vessel centerline from the grayscale image of the DSA image to construct an initial topology structure, providing a basic framework for subsequent vessel region recognition and lesion determination. The lesion recognition module constructs vessel candidate regions for local segments with more than 2 edges in the initial topology structure. Combining the similarity between the blood flow propagation path and the region edges, as well as the grayscale continuity of the region, it filters out real vessel regions, effectively eliminating artifact interference and avoiding misidentification of artifacts as vessels, thus improving the accuracy of real vessel recognition. Based on the real vessel region... The rate of change in vessel diameter at adjacent sampling points identifies narrowed vessel regions. Combined with the grayscale differences at corresponding points along the blood flow propagation path, the obstructed blood flow area and lesion port are determined. Then, the vessel rupture area is identified through the lesion port compensation curve and candidate composite path curvature analysis, achieving precise localization of different types of coronary artery lesions. The lesion marking module completes the vessel rupture area, obtaining a real vascular atlas and marking various lesion areas, intuitively presenting the distribution and extent of lesions. This provides clear and reliable imaging evidence for clinical diagnosis, avoiding misjudgment or missed judgment of lesions due to incomplete vascular topology, and further ensuring the reliability of coronary atherosclerosis lesion detection. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 The structural block diagram of the intelligent identification system for coronary atherosclerosis based on DSA images provided by the present invention;

[0018] Figure 2 This is a schematic diagram of the grayscale value curve of the first sampling point changing over time during the image acquisition cycle in the intelligent identification system for coronary atherosclerosis based on DSA images provided by the present invention. Detailed Implementation

[0019] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of the intelligent identification system for coronary atherosclerosis based on DSA imaging proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0020] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0021] The specific solution of the intelligent identification system for coronary atherosclerosis based on DSA images provided by the present invention will be described in detail below with reference to the accompanying drawings.

[0022] Please see Figure 1 The diagram illustrates a structural block diagram of a DSA-based intelligent identification system for coronary atherosclerosis, according to an embodiment of the present invention. The system includes the following modules: The topology determination module 110 is used to extract the center line of the blood vessels in the grayscale image of the patient's coronary artery DSA image and construct the initial topology structure. The lesion identification module 120 is used to construct a blood vessel candidate region with any two edges for local segments in the initial topology structure that have more than two edges, and to determine the real blood vessel region based on the similarity between the blood flow propagation path in the blood vessel candidate region and the region edge and the region grayscale continuity. The lesion identification module 120 is also used to determine the stenotic vascular region based on the rate of change of the diameter between adjacent sampling points in the real vascular region, determine the starting point of blood flow obstruction and obtain the blood flow obstruction region based on the gray difference between corresponding points of sampling points on the blood flow propagation path, and determine the severe mutation region and lesion port based on the starting point of blood flow obstruction and the sampling point with the maximum rate of change of the diameter. The lesion identification module 120 is also used to construct a compensation curve between the current lesion port and its neighboring lesion ports and extend it to obtain a candidate composite path. Based on the curvature at each sampling point on the candidate composite path, the compensation smoothness of the compensation curve is determined, and the vascular rupture area is determined based on the compensation smoothness. The lesion marking module 130 is used to complete the ruptured vascular region in the initial topology to obtain a real vascular atlas, and to mark the stenotic vascular region, the blood flow obstruction region and the ruptured vascular region in the real vascular atlas.

[0023] The functions of each module of the above-mentioned intelligent identification system for coronary atherosclerosis based on DSA images are described in detail below in one embodiment: The aforementioned topology determination module 110 is used to extract the center line of the blood vessels from the grayscale image of the patient's coronary artery DSA image and construct an initial topology structure.

[0024] In this embodiment, the aforementioned Digital Subtraction Angiography (DSA) is an imaging technique that uses computer processing of digital image information to eliminate background interference from bones, soft tissues, etc., and clearly display the morphology of blood vessels. The aforementioned coronary artery DSA image combines digital image processing technology with X-ray angiography, using an iodine-containing contrast agent to visualize the coronary arteries, followed by computer subtraction processing to obtain a clear image containing only the coronary artery structure. The aforementioned grayscale image is an image containing only grayscale information obtained by grayscale processing of the patient's color coronary artery DSA image.

[0025] In this embodiment of the application, the aforementioned vascular centerline is the central axis along the direction of vascular extension in the grayscale image, which can reflect the direction and topology of the vascular system; the aforementioned initial topology is an initial framework structure built on the vascular centerline to describe the connection relationship between the branches and nodes of the coronary artery.

[0026] Specifically, in this embodiment of the application, DSA images of a patient's coronary arteries can be obtained through the following method: After the patient undergoes basic examinations, the doctor selects the femoral artery or radial artery as the puncture point based on the patient's condition; the artery is punctured using the Seldinger technique, and under X-ray guidance, the angiography catheter is inserted into the coronary artery ostium (left main coronary artery or right coronary artery); the digital subtraction angiography system is activated, and non-vascular tissue is eliminated through background subtraction, retaining only the vascular image; dynamic images from multiple angles are acquired and saved.

[0027] After acquiring DSA images of the patient's coronary arteries using the above method, the topology determination module can construct the initial topology of the patient's coronary arteries using the following methods: The acquired coronary artery DSA color images are converted to grayscale to obtain grayscale images; vascular enhancement is performed on the grayscale images to highlight grayscale differences in the vascular regions and suppress background noise; a skeleton extraction-based method (such as the watershed algorithm or morphological thinning algorithm) is used to process the enhanced vascular regions to obtain the vascular centerline; branch points and endpoints in the vascular centerline are identified, and the vascular centerline is connected according to the actual connection relationships to construct an initial topology describing the branching and node connections of the coronary arteries.

[0028] In DSA images, the correspondence between vascular branches and segments can be accurately identified by determining the vascular atlas of the patient's coronary arteries. However, the location of coronary atherosclerotic lesions may lead to vascular branch breakage or local deviation, and plaque calcification, high-density shadows, contrast agent retention, or turbulent blood flow can easily produce artifacts. Therefore, in this embodiment, the lesion identification module first identifies the true location of each local blood vessel in the DSA image and removes artifacts; then, it combines the continuity analysis of the vessel diameter to determine the possible location of lesions, so that when a lesion is detected, the lesion marking module can repair the vascular breakage and construct a true vascular atlas.

[0029] Normal coronary artery blood flow exhibits uniformity and stability in DSA images. Contrast agent flows unimpeded along the vessel, causing the vessel grayscale to follow a smooth, continuous curve of increase and decrease over time. This is because the contrast agent concentration gradually increases to a peak during blood perfusion before being flushed away. Conversely, in artifact or noise areas, blood flow is unstable, with abnormal and abrupt grayscale changes that do not propagate continuously with the direction of blood flow. Furthermore, artifacts are typically lighter in color, while real blood vessels are darker. Based on this, the aforementioned lesion identification module can determine the stability of blood flow by analyzing local vessel grayscale changes and blood flow trends, thereby identifying real vascular regions and providing a reliable basis for vascular tree reconstruction and lesion localization.

[0030] The process of the lesion recognition module identifying real blood vessel regions is described in detail below in one embodiment of this application: The aforementioned lesion identification module can be used to construct a blood vessel candidate region with any two edges for local segments in the initial topology where the number of edges is greater than two. The real blood vessel region is determined based on the similarity between the blood flow propagation path in the blood vessel candidate region and the region edge, as well as the region grayscale continuity.

[0031] In this embodiment of the application, the aforementioned local segment is an independent vascular segment obtained by dividing the extracted vascular centerline with the vascular tree intersection point as the boundary.

[0032] In this embodiment of the application, the above-mentioned blood vessel candidate region is a region that may correspond to a real blood vessel, formed by arbitrarily selecting two edges from multiple edges of a local segment.

[0033] In one specific implementation of this application embodiment, the above-mentioned construction of the blood vessel candidate region can be achieved as follows: Identify the intersection points (usually corresponding to blood vessel branches or branch confluences) in the initial topology of the blood vessel; using the intersection points as boundaries, divide the blood vessel into multiple local segments; for any current local segment of the blood vessel, determine the blood vessel edge using the Canny Edge Detection (Canny) algorithm; if there are M (M>2) edges, it indicates the presence of artifacts; arbitrarily combine two edges from the M edges. This is recorded as the candidate blood vessel region for the current local segment.

[0034] In this embodiment, the blood flow propagation path is a path obtained by connecting the sampling points based on the gray-scale peak time sorting of the sampling points on the center line of the blood vessel candidate region, which can reflect the true direction of contrast agent flow.

[0035] For example, the above-mentioned determination of the real blood vessel region based on the similarity between the blood flow propagation path and the region edge and the region grayscale continuity in the blood vessel candidate region can be achieved as follows: Multiple first sampling points are uniformly selected along the blood vessel centerline of the blood vessel candidate region; the blood flow propagation path is determined based on the grayscale value curve of the first sampling point changing over time; the similarity between the blood flow propagation path and the two region edges is determined based on the dynamic time warping algorithm, and the similarity product is calculated, with the similarity product recorded as the blood flow trend stability; the grayscale mean of the two region edges is determined, and the normalized value of the ratio of the absolute value of the difference in grayscale mean to the larger of the grayscale mean is recorded as the region grayscale continuity; the probability of the real blood vessel in the blood vessel candidate region is determined based on the blood flow trend stability and the region grayscale continuity, and the blood vessel candidate region corresponding to the maximum probability of the real blood vessel is determined as the real blood vessel region.

[0036] The method for determining the blood flow propagation path based on the grayscale curve of the first sampling point over time can be achieved as follows: Obtain the grayscale curve of each first sampling point over time within the coronary artery DSA image acquisition cycle; extract the grayscale peak points from each grayscale curve and determine the time of occurrence of the grayscale peak points to construct a peak time series; reorder the occurrence times of each peak in the peak time series according to chronological order and assign an index number to each peak occurrence time in ascending order to obtain a target index sequence; arrange the first sampling points based on the target index sequence and connect the sorted first sampling points sequentially to obtain the blood flow propagation path.

[0037] In one specific implementation of this application, the determination of the real blood vessel region based on the similarity between the blood flow propagation path and the region edge and the region grayscale continuity in the blood vessel candidate region can be achieved as follows: For each blood vessel candidate region, n first sampling points are uniformly selected along the blood vessel centerline of the blood vessel candidate region at fixed length intervals; the grayscale value curve of each first sampling point changing with time during the DSA image acquisition cycle is obtained. Because the contrast agent concentration gradually increases to a peak during blood perfusion before being flushed away, taking the first sampling point k as an example, its grayscale curve... like Figure 2 As shown; determine each grayscale curve. The time point corresponding to the maximum peak value (i.e., the grayscale peak point mentioned above). Construct peak time series The elements in the peak time series T are reordered chronologically to obtain the target index sequence described above. The target index sequence represents the order in which contrast agent passes through the blood vessel. The sorted time point sequence reflects the current trend of blood flow, i.e., the dynamic direction of blood flow from proximal to distal. The target index sequence is then mapped to a spatial point sequence (i.e., the first sampling point mentioned above). Connect adjacent spatial points in sequence: This forms the blood flow propagation path S; the dynamic time warping (DWT) algorithm is used to calculate the blood flow propagation path S and the edge of the vessel candidate region. and The similarity is used to calculate the blood flow trend stability using the following formula: in, The value represents the stability of blood flow trend; the higher the value, the more likely the candidate vessel region is to be a real vessel rather than an artifact. The blood flow propagation path S calculated by DWT and the edge of the candidate vessel region. Similarity; The blood flow propagation path S calculated by DWT and the edge of the candidate vessel region. The similarity is calculated by the formula above, which distinguishes real blood vessels from artifacts by quantifying the consistency between the morphology of the blood flow path and the edge of the blood vessel. The edge morphology of real blood vessels will be highly consistent with the blood flow propagation path (blood flow flows along the blood vessel lumen, and the path needs to fit the edge direction), while the edge of the artifact is random noise and has no regular correlation with the blood flow path.

[0038] Meanwhile, since artifacts are usually lighter in gray and have larger gray values, while real blood vessels have smaller gray values, and the gray values ​​on both sides of a real blood vessel are relatively uniform, this embodiment of the application determines the probability of a real blood vessel by combining the gray value continuity of the blood flow trend with the stability of the blood flow trend when identifying the real blood vessel region. Specifically, the above-mentioned probability of a real blood vessel can be calculated by the following formula: in, The probability of a real blood vessel is represented by , and the larger the value, the more likely the candidate blood vessel region is to be a real blood vessel rather than an artifact; w is the stability of the blood flow trend mentioned above, which reflects the morphological similarity between the blood flow propagation path and the edge of the blood vessel. and These represent the two edges of the candidate blood vessel region, respectively. The average gray level; This is the absolute difference between the mean gray values ​​of the two edges of the blood vessel candidate region, reflecting the degree of difference in edge gray values. The smaller the value, the more uniform the gray values ​​of the blood vessel candidate region. The value is the maximum of the mean gray values ​​at both edges of the candidate blood vessel region. The smaller the value, the greater the probability that the candidate blood vessel region is a real blood vessel. This is the normalization function; This refers to the grayscale continuity of the aforementioned region.

[0039] After determining the probability of real blood vessels using the above method, the candidate region of the blood vessel with the highest probability of being a real blood vessel is determined as the real blood vessel region.

[0040] When plaques or calcification occur in a patient's coronary arteries, the lesion area leads to decreased elasticity of the vessel wall, thickening of the intima, and local hardening. Therefore, if a significant change in vessel diameter is detected at the end of a vessel branch, it indicates that the vessel branch may be truncated. In contrast, the change in vessel diameter along the direction of blood flow in a normal vessel should show a smooth, continuous, and slow trend.

[0041] When there is a truncation in a vascular branch region, the contrast agent in the DSA image will propagate downstream of that branch with the blood flow. If the peak time series of the current vascular branch is interrupted or reversed, and the downstream region shows no significant brightness response in the time series, it indicates that there may be a complete occlusion or severe stenosis lesion at that location. Based on this, the aforementioned lesion identification module can identify the location of coronary atherosclerotic lesions by analyzing the rate of change in diameter between adjacent sampling points in the real vascular region and the grayscale differences between corresponding points on the blood flow propagation path.

[0042] The process of the lesion identification module identifying the location of a lesion is described in detail below in one embodiment of this application:

[0043] The aforementioned lesion identification module 120 is also used to determine the stenotic vascular region based on the rate of change of the diameter between adjacent sampling points in the real vascular region, to determine the starting point of blood flow obstruction and obtain the blood flow obstruction region based on the gray difference between corresponding points on the blood flow propagation path of the sampling points, and to determine the severe mutation region and lesion port based on the starting point of blood flow obstruction and the sampling point with the maximum rate of change of the diameter.

[0044] In this embodiment, the aforementioned narrowed vascular region refers to a vascular segment in the real vascular region whose diameter exhibits continuous contraction or abrupt change, corresponding to a luminal stenosis lesion.

[0045] For example, the above-mentioned determination of narrow vascular regions based on the rate of change of tube diameter between adjacent sampling points in a real vascular region can be achieved as follows: Multiple second sampling points are uniformly selected along the vascular centerline of the real vascular region; the maximum inscribed circle of the real vascular region at each second sampling point is obtained, and its diameter is determined as the local tube diameter at the second sampling point; for each second sampling point, the smaller value between the local tube diameter of the current second sampling point and its adjacent sampling points is determined, and the absolute value of the difference between the local tube diameter of the current second sampling point and its adjacent sampling points is calculated; the ratio of the absolute value of the tube diameter difference to the smaller value of the local tube diameter is determined as the rate of change of tube diameter; if the rate of change of tube diameter between multiple consecutive adjacent second sampling points is greater than a first preset threshold, the corresponding vascular segment is determined as a narrow vascular region.

[0046] In one specific implementation of this application embodiment, the process of determining the narrowed blood vessel region described above can be achieved as follows: extract the blood vessel centerline of the actual blood vessel region, uniformly select n second sampling points at fixed intervals along the blood vessel centerline, and record the spatial coordinates of each second sampling point. The local pipe diameter at each second sampling point is determined using the local maximum inscribed circle method, resulting in a local pipe diameter sequence. For adjacent second sampling points The rate of change of pipe diameter is calculated using the following formula: in, For adjacent second sampling points The rate of change of pipe diameter between; For the second sampling point Local pipe diameter; For the second sampling point The local pipe diameter; setting the pipe diameter change rate threshold (i.e., the first preset threshold mentioned above) to 0.3, and filtering out... If adjacent sampling point pairs of 0.3 appear consecutively ( If a pair of sampling points meets the above threshold conditions, it indicates that the diameter of the current sampling point pair has experienced continuous contraction or abrupt change, and the diameter of this segment is marked as a narrow vascular region.

[0047] In this embodiment, the aforementioned blood flow obstruction initiation point is the starting location of the contrast agent propagation obstruction. The upstream region of this initiation point is sufficiently filled with contrast agent, exhibiting high grayscale, while the downstream region lacks contrast agent, exhibiting low grayscale. Therefore, the node with the greatest difference in grayscale variation can be identified as the blood flow obstruction initiation point. A continuous low-grayscale vascular segment is located on one side of this node; the aforementioned blood flow obstruction region is a vascular segment continuously exhibiting a low-grayscale distribution on one side of the blood flow obstruction initiation point.

[0048] For example, the above method of determining the starting point of blood flow obstruction and obtaining the blood flow obstruction area based on the gray-level difference between corresponding points of sampling points on the blood flow propagation path can be achieved as follows: determine the propagation path point closest to each second sampling point on the blood flow propagation path, calculate the absolute value of the gray-level difference between adjacent propagation path points, and record it as the gray-level change value; determine the second sampling point corresponding to the maximum gray-level change value as the starting point of blood flow obstruction, and record the continuous pixels with gray-level values ​​lower than the second preset threshold downstream of the starting point of blood flow obstruction as the blood flow obstruction area.

[0049] In one specific implementation of this application embodiment, the process of determining the starting point of blood flow obstruction and obtaining the blood flow obstruction region can be implemented as follows: for the second sampling point Extract the nearest path point from the established blood flow propagation path. and extract The gray values ​​are used to construct the corresponding gray-level sequence. For adjacent propagation path points Determine its corresponding grayscale change value Find the second sampling point corresponding to the maximum gray value change and record it as the starting point of blood flow obstruction; observe the downstream blood vessel segment of the starting point of blood flow obstruction. If the gray values ​​of multiple consecutive second sampling points are significantly lower than those of their upstream counterparts, then the consecutive segment is identified as the blood flow obstruction area.

[0050] In this embodiment, the severe mutation region is a vascular region where the starting point of blood flow obstruction coincides with the location of the second sampling point with the maximum rate of change in vessel diameter, or where the sampling distance between the two points is less than a third preset threshold; the lesion port is the boundary point of the severe mutation region, which is the starting or ending port of the vascular lesion.

[0051] For example, the process of determining the severe mutation region and lesion port based on the sampling point of the blood flow obstruction initiation point and the maximum tube diameter change rate can be implemented as follows: if the blood flow obstruction initiation point and the second sampling point with the largest tube diameter change rate are the same point or the sampling interval is less than the third preset threshold, then the blood flow obstruction region corresponding to the blood flow obstruction initiation point is determined as the severe mutation region, and the endpoint of the severe mutation region is determined as the lesion port.

[0052] In one specific implementation of this application, the process of identifying the severe mutation region and the lesion port can be implemented as follows: obtain the second sampling point with the largest change rate of the tube diameter, compare the position of the blood flow obstruction start point with the second sampling point with the largest change rate of the tube diameter, if the spatial coordinates of the two points coincide, or the number of sampling points between the two points is <3, then mark the blood flow obstruction area corresponding to the blood flow obstruction start point as the severe mutation region, and take the two boundary points of the severe mutation region as the lesion port.

[0053] When two lesions are found at adjacent vascular ports in DSA images, but the connection between them is not clearly identified, it indicates that there may be a relatively serious vascular lesion in the area. In this case, this embodiment of the application determines the connectivity by analyzing whether the line connecting the adjacent points of the potential connecting vessels between the two lesion ports conforms to the smoothness of the overall vascular orientation, thus identifying the vascular breakage area. If the line connecting the adjacent points is similar to the overall vascular orientation, it reflects that the line is a true vascular connection, so that the lesion marking module can subsequently perform local structural reconstruction and continuity repair of the vascular area based on the overall vascular orientation characteristics, thereby achieving the completion of the vascular topology and the complete reconstruction of the vascular network.

[0054] The process of the lesion recognition module recognizing the ruptured blood vessel region is described in detail below in one embodiment of this application: The aforementioned lesion identification module 120 is also used to construct a compensation curve between the current lesion port and its neighboring lesion ports and extend it to obtain a candidate composite path, determine the compensation smoothness of the compensation curve based on the curvature at each sampling point on the candidate composite path, and determine the vascular rupture area based on the compensation smoothness.

[0055] In this embodiment of the application, the compensation curve is a curve connecting the target lesion port and the adjacent lesion port, used to fill the possible vascular rupture gap between the two ports; the candidate composite path is a complete continuous path formed by splicing the compensation curve and the vascular centerline segment after extending the compensation curve length into the original vascular segment of the target lesion port and the adjacent lesion port.

[0056] For example, the above-mentioned construction of compensation curves between the current lesion port and its neighboring lesion ports and extension to obtain candidate composite paths can be achieved as follows: For each lesion port, determine the neighboring lesion ports within the target neighborhood of the current lesion port. The target neighborhood is a circular neighborhood centered on the current lesion port and with a radius of 1 / 2 of the length of the vessel where the current lesion port is located; construct compensation curves between the current lesion port and its neighboring lesion ports and determine their lengths; extend the lengths of the compensation curves towards the current lesion port and its neighboring lesion ports respectively to obtain candidate composite paths.

[0057] In one specific implementation of this application embodiment, the process of constructing the candidate composite path described above can be implemented as follows: for any lesion port Define a circular neighborhood of N1 (where N1 is half the current vessel length), obtain other lesion ports within this circular neighborhood, and construct a set of lesion ports. ,in, Indicates the current lesion port The node position and the i-th other vascular lesion port The distance; for any other vascular lesion port ,connect and The compensation curve is obtained. Determine its length, denoted as From the lesion port as well as Each of them cuts a length from the inside of its respective blood vessel. The original vascular central line segment, and combined with By splicing together, candidate composite paths are formed. .

[0058] In the embodiments of this application, the above-mentioned compensation smoothness is an index that quantifies the consistency between the compensation curve and the original vascular segment direction. The larger the value, the better the compensation curve matches the original vascular segment direction.

[0059] For example, the above-mentioned determination of the compensation smoothness of the compensation curve based on the curvature at each sampling point on the candidate composite path can be achieved as follows: Multiple third sampling points are uniformly selected on the candidate composite path, and the curvature of the candidate composite path at each third sampling point is calculated to construct a first curvature sequence corresponding to the candidate composite path and a second curvature sequence corresponding to the compensation curve; the curvature at each remaining third sampling point after removing the second curvature sequence from the first curvature sequence is obtained to construct a third curvature sequence; the compensation smoothness of the compensation curve is determined based on the first curvature sequence and the third curvature sequence.

[0060] The above-mentioned determination of the compensation smoothness of the compensation curve based on the first curvature sequence and the third curvature sequence can be achieved as follows: calculate the first curvature variance of the first curvature sequence and the second curvature variance of the third curvature sequence; calculate the difference between the first curvature variance and the second curvature variance, and denote it as the compensation smoothness.

[0061] In one specific implementation of this application embodiment, the process of determining the compensation smoothness described above can be implemented as follows: for candidate composite paths Multiple third sampling points are uniformly selected at fixed pixel intervals m1 to obtain a discrete point sequence. ; Calculate the discrete point sequence The curvature values ​​of each third sampling point are used to obtain the corresponding first curvature sequence. And mark those belonging to The partial second curvature sequence, denoted as The compensation for smoothness is calculated using the following formula: in, For compensation curve Compensation for smoothness; Indicate candidate composite path The variance of the first curvature sequence, Indicate candidate composite path No The variance of the third curvature sequence of a partial vascular portion; the above formula is calculated by... and The difference in curvature variance determines the compensation smoothness. The larger the value, the smoother the curvature fluctuation of the overall path after the compensation curve is added, and the better the compensation curve matches the original blood vessel direction.

[0062] If the vascular lesions are clustered In the current vascular lesion port If the lesion port with the largest compensation smoothness r among the compensation curves is the port where the rupture currently exists, then the compensation curve between the lesion ports is determined to be the vascular rupture area.

[0063] The aforementioned lesion marking module 130 is used to complete the ruptured vascular region in the initial topology to obtain a real vascular atlas, and to mark the stenotic vascular region, the blood flow obstruction region, and the ruptured vascular region in the real vascular atlas.

[0064] After identifying the ruptured blood vessel region through the aforementioned lesion recognition module, the ruptured blood vessel region in the initial topology can be completed using the compensation curve corresponding to the maximum compensation smoothness, thus obtaining a true vascular atlas.

[0065] Furthermore, after obtaining the actual coronary artery atlas, the identified atherosclerotic stenotic vessel areas are visually marked in yellow, blood flow obstruction areas in purple, and vessel rupture areas in red. These lesion features are then mapped back to the actual vascular atlas, thereby achieving spatial visualization and precise localization of coronary atherosclerotic lesions. Among them, stenotic vessel areas indicate that the vessel is in a state of continuous constriction, which indicates that the vessel may be narrowed or blocked; blood flow obstruction areas reflect blood flow obstruction or lumen collapse; and vessel rupture areas are characterized by a break in the continuity of the vascular structure or a complete loss of grayscale, which usually corresponds to severe lesions or complete occlusion.

[0066] It should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A coronary atherosclerosis intelligent identification system based on DSA imaging, characterized in that, The system includes: The topology determination module is used to extract the centerline of the blood vessels from the grayscale image of the patient's coronary artery DSA image and construct the initial topology structure; The lesion identification module is used to construct a blood vessel candidate region with any two edges for local segments in the initial topology where the number of edges is greater than 2, and to determine the real blood vessel region based on the similarity between the blood flow propagation path and the region edge and the region grayscale continuity in the blood vessel candidate region. The lesion identification module is also used to determine the stenotic vascular region based on the rate of change of the diameter between adjacent sampling points in the real vascular region, determine the starting point of blood flow obstruction and obtain the blood flow obstruction region based on the gray difference between corresponding points of sampling points on the blood flow propagation path, and determine the severe mutation region and lesion port based on the starting point of blood flow obstruction and the sampling point with the maximum rate of change of the diameter. The lesion identification module is also used to construct a compensation curve between the current lesion port and its neighboring lesion ports and extend it to obtain a candidate composite path, determine the compensation smoothness of the compensation curve based on the curvature at each sampling point on the candidate composite path, and determine the vascular rupture area based on the compensation smoothness. The lesion marking module is used to complete the ruptured blood vessel region in the initial topology to obtain a real blood vessel map, and to mark the stenotic blood vessel region, the blood flow obstruction region and the ruptured blood vessel region in the real blood vessel map.

2. The intelligent identification system for coronary atherosclerosis based on DSA imaging according to claim 1, characterized in that, The lesion recognition module determines the real blood vessel region based on the similarity between the blood flow propagation path and the region edge, and the region grayscale continuity, by executing the following method: Multiple first sampling points are uniformly selected along the center line of the blood vessel in the blood vessel candidate region, and the blood flow propagation path is determined based on the gray value curve of the first sampling points changing over time. The similarity between the blood flow propagation path and the two edges of the region is determined based on the dynamic time warping algorithm, and the similarity product is calculated. The similarity product is recorded as the blood flow trend stability. The gray-level mean values ​​of the two edges of the region are determined respectively, and the normalized value of the ratio of the absolute value of the difference between the gray-level mean values ​​to the larger value among the gray-level mean values ​​is recorded as the gray-level continuity of the region. The probability of a real blood vessel in the candidate blood vessel region is determined based on the stability of the blood flow trend and the continuity of the gray level of the region, and the candidate blood vessel region corresponding to the maximum probability of a real blood vessel is determined as the real blood vessel region.

3. The intelligent identification system for coronary atherosclerosis based on DSA imaging according to claim 2, characterized in that, The lesion identification module determines the blood flow propagation path based on the grayscale curve of the first sampling point changing over time by executing the following method, including: Obtain the grayscale value curve of each of the first sampling points as a function of time within the coronary artery DSA image acquisition cycle; Extract the grayscale peak points from each of the grayscale value curves, determine the time when the grayscale peak points appear, and construct a peak time series; The occurrence times of each peak in the peak time series are reordered according to chronological order, and each peak occurrence time is assigned an index number in ascending order to obtain the target index sequence. The first sampling points are arranged according to the target index sequence, and the sorted first sampling points are connected in sequence to obtain the blood flow propagation path.

4. The intelligent identification system for coronary atherosclerosis based on DSA imaging according to claim 1, characterized in that, The lesion identification module determines the stenotic vessel region based on the rate of change in vessel diameter between adjacent sampling points in the real vascular region by executing the following method: Multiple second sampling points are uniformly selected along the center line of the blood vessel in the actual blood vessel region; Obtain the maximum inscribed circle of the real blood vessel region at each of the second sampling points, and determine its diameter as the local diameter of the vessel at the second sampling point; For each of the second sampling points, the smaller value of the local pipe diameter between the current second sampling point and its adjacent sampling points is determined, and the absolute value of the difference between the local pipe diameter of the current second sampling point and its adjacent sampling points is calculated. The ratio of the absolute value of the pipe diameter difference to the smaller value of the local pipe diameter is determined as the pipe diameter change rate. If the rate of change of the tube diameter between multiple consecutive adjacent second sampling points is greater than the first preset threshold, then the corresponding blood vessel segment is determined as a narrow blood vessel region.

5. The intelligent identification system for coronary atherosclerosis based on DSA imaging according to claim 4, characterized in that, The lesion identification module determines the starting point of blood flow obstruction and obtains the blood flow obstruction region based on the gray-level difference between corresponding points of the sampling points on the blood flow propagation path by executing the following method: On the blood flow propagation path, determine the propagation path point that is closest to each of the second sampling points, and calculate the absolute value of the grayscale difference between adjacent propagation path points, which is recorded as the grayscale change value. The second sampling point corresponding to the maximum grayscale change value is determined as the starting point of blood flow obstruction, and the continuous pixels with grayscale values ​​lower than the second preset threshold located downstream of the starting point of blood flow obstruction are recorded as the blood flow obstruction area.

6. The intelligent identification system for coronary atherosclerosis based on DSA imaging according to claim 5, characterized in that, The lesion identification module determines the severe mutation region and lesion port based on the sampling points according to the blood flow obstruction initiation point and the maximum diameter change rate by executing the following methods: If the blood flow obstruction starting point is the same point as the second sampling point with the largest tube diameter change rate, or if the sampling interval is less than a third preset threshold, then the blood flow obstruction area corresponding to the blood flow obstruction starting point is determined as the severe mutation area, and the endpoint of the severe mutation area is determined as the lesion port.

7. The intelligent identification system for coronary atherosclerosis based on DSA imaging according to claim 6, characterized in that, The lesion identification module constructs a compensation curve between the current lesion port and its neighboring lesion ports and extends it to obtain a candidate composite path by performing the following methods: For each of the aforementioned lesion ports, determine the neighboring lesion ports within the target neighborhood of the current lesion port; the target neighborhood is a circular neighborhood centered on the current lesion port and with a radius of 1 / 2 of the length of the blood vessel where the current lesion port is located; Construct the compensation curve between the current lesion port and the adjacent lesion port and determine its length. Extend the length of the compensation curve in the direction of the current lesion port and the adjacent lesion port respectively to obtain the candidate composite path.

8. The intelligent identification system for coronary atherosclerosis based on DSA imaging according to claim 7, characterized in that, The lesion identification module achieves the determination of the compensation smoothness of the compensation curve based on the curvature at each sampling point on the candidate composite path by executing the following method: Multiple third sampling points are uniformly selected on the candidate composite path, and the curvature of the candidate composite path at each of the third sampling points is calculated to construct a first curvature sequence corresponding to the candidate composite path and a second curvature sequence corresponding to the compensation curve. Obtain the curvature at each of the third sampling points remaining after removing the second curvature sequence from the first curvature sequence, and construct the third curvature sequence; The compensation smoothness of the compensation curve is determined based on the first curvature sequence and the third curvature sequence.

9. The intelligent identification system for coronary atherosclerosis based on DSA imaging according to claim 8, characterized in that, The lesion recognition module achieves the compensation smoothness by executing the following method to determine the compensation curve based on the first curvature sequence and the second curvature sequence, including: Calculate the first curvature variance of the first curvature sequence and the second curvature variance of the third curvature sequence; The difference between the first curvature variance and the second curvature variance is calculated and denoted as the compensated smoothness.

10. The intelligent identification system for coronary atherosclerosis based on DSA imaging according to claim 1, characterized in that, The lesion marking module completes the vascular rupture region in the initial topology to obtain a real vascular atlas by performing the following method, and marks the stenotic vascular region, the blood flow obstruction region, and the vascular rupture region in the real vascular atlas, including: The blood vessel fracture region in the initial topology is completed using the compensation curve corresponding to the maximum compensation smoothness to obtain the real blood vessel atlas; In the real vascular atlas, the narrowed vascular region, the blood flow obstruction region, and the vascular rupture region are visually marked with different colors.

Images