Head and neck blood vessel completion method and system based on particle swarm optimization algorithm

Abstract

The invention relates to the technical field of blood vessel completion, and provides a head and neck blood vessel completion method and system based on a particle swarm optimization algorithm, and the method comprises the steps: obtaining a head and neck blood vessel angiography image, extracting an initial segmentation blood vessel and a mask of a skeleton region, and carrying out the segmentation of the initial segmentation blood vessel through a point cloud segmentation model of each branch of the head and neck; dividing the branch blood vessel regions into a plurality of to-be-complemented blood vessel groups in a plurality of division modes; for each to-be-complemented blood vessel in each to-be-complemented blood vessel group, searching the nearest to-be-complemented blood vessel, taking the adjacent points of the two to-be-complemented blood vessels as starting and ending point coordinates, and taking a plurality of points uniformly distributed between the starting and ending points as initial control points; and by taking avoiding of a skeleton and other blood vessels, passing through the maximum Henz unit value region and curve smoothness as targets, optimizing the control points by adopting a particle swarm algorithm to obtain a complemented connecting line. The accuracy and authenticity of blood vessel complementing are ensured, and the physiological characteristics of blood vessels are met.

Description

Technical Field

[0001] The present invention belongs to the technical field of blood vessel completion, and in particular relates to a head and neck blood vessel completion method and system based on a particle swarm optimization algorithm. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] Vascular diseases of the head and neck, such as stroke and carotid artery stenosis, are one of the important causes of death and disability. Accurate assessment of the condition of the head and neck blood vessels is crucial for early diagnosis and treatment. In recent years, with the development of Computed Tomography Angiography (CTA) technology, imaging of head and neck blood vessels has become more efficient and safer.

[0004] Although CTA technology has made significant progress in head and neck vascular imaging, there are still some challenges in practical application. Specifically, CTA imaging methods may have problems with broken or missing blood vessels in certain situations. These problems may be caused by a variety of factors, including stenosis of the blood vessels themselves, motion artifacts of the patient, noise interference during imaging, and artifacts caused by metal implants. These defects not only affect the integrity of the imaging, but may also mislead doctors' judgment of the location of the lesion, thereby affecting subsequent diagnosis and treatment decisions.

[0005] The vascular completion method can fill in blood vessel breaks caused by various reasons (such as vascular stenosis, motion artifacts, noise, metal artifacts, etc.). The generated complete vascular structure can be used in a variety of clinical applications, such as the diagnosis of vascular diseases, surgical planning, treatment effect evaluation, etc., improving the accuracy and efficiency of clinical diagnosis and treatment.

[0006] However, traditional vascular completion methods do not take into account the anatomical characteristics and imaging features of blood vessels, and the completed blood vessels are difficult to conform to the physiological characteristics of blood vessels. Summary of the invention

[0007] In order to solve the technical problems existing in the above-mentioned background technology, the present invention provides a head and neck blood vessel completion method and system based on a particle swarm optimization algorithm, which combines vascular anatomical knowledge, vascular connectivity domain analysis, HU value of the image, adjacent blood vessels and adjacent bone masks and other information, and generates connecting lines between broken blood vessels through a particle swarm algorithm and a B-spline curve, fully considering the anatomical characteristics and imaging characteristics of the blood vessels, ensuring the accuracy and authenticity of the completed blood vessels, and conforming to the physiological characteristics of the blood vessels.

[0008] In order to achieve the above object, the present invention adopts the following technical solution:

[0009] The first aspect of the present invention provides a head and neck blood vessel completion method based on a particle swarm optimization algorithm, comprising:

[0010] Obtain angiography images of the head and neck, extract the masks of the initial segmented blood vessels and the bone area through the blood vessel bone segmentation model, and then obtain the blood vessel areas of each branch of the head and neck through the point cloud segmentation model of each branch of the head and neck for the initial segmented blood vessels;

[0011] Dividing the branch blood vessel regions of the head and neck into multiple blood vessel groups to be completed by multiple division methods;

[0012] For each vessel to be completed in each vessel group, the nearest neighbor vessel to be completed is found, and the adjacent points of the two vessels to be completed are used as the starting and ending point coordinates. After using several points evenly distributed between the starting and ending points as initial control points, the particle swarm algorithm is used to optimize the control points with the goal of avoiding the skeleton and other vessels and passing through the maximum sum of curve smoothness in the Heinrich unit value area. The optimized control points are used to generate a B-spline curve, and the completion connection line is obtained by interpolation;

[0013] Based on the completed connection lines, the completed blood vessels are uniformly expanded into the specified radius.

[0014] Furthermore, the division method includes: division according to different segments of the same branch blood vessel, combination according to connected branch blood vessels, and combination according to possibly connected branch blood vessels.

[0015] Furthermore, the number of the control points is determined according to the distance between the starting and ending points and the curvature of the blood vessel.

[0016] Furthermore, before searching for the nearest neighbor blood vessels to be completed, the method also includes: for each group of blood vessels to be completed, removing blood vessels that are noise-prone or segmented incorrectly by the blood vessel skeleton segmentation model; removing blood vessels that are segmented incorrectly by the point cloud segmentation model of each branch of the head and neck; and extracting the coordinates of the blood vessel surface layer.

[0017] Further, the avoiding skeleton and other blood vessels is expressed as:

[0018]

[0019] Among them, p c are the coordinates of the skeleton and other blood vessels, d(B(t),p c )) is the distance from point B(t) to p on the B-spline curve c distance, R is a threshold.

[0020] Further, the passing Hounsfield unit value region is expressed as:

[0021]

[0022] Where HU(B(t)) is the value of the Hounsfield unit at the point B(t) on the B-spline curve.

[0023] Furthermore, the curve smoothness is expressed as:

[0024]

[0025] Where B′(t) is the first-order derivative of the B-spline curve.

[0026] The second aspect of the present invention provides a head and neck blood vessel completion system based on a particle swarm optimization algorithm, comprising:

[0027] The segmentation unit is configured to: acquire a head and neck angiography image, extract masks of the initial segmented blood vessels and the bone region through a blood vessel skeleton segmentation model, and obtain the head and neck branch blood vessel regions through a head and neck branch point cloud segmentation model for the initial segmented blood vessels;

[0028] A blood vessel division unit is configured to: divide the branch blood vessel regions of the head and neck into a plurality of blood vessel groups to be completed by using a plurality of division methods;

[0029] The computing unit is configured to: for each blood vessel to be completed in each blood vessel group to be completed, find the nearest neighbor blood vessel to be completed, and use the adjacent points of the two blood vessels to be completed as the starting and ending point coordinates, and use a number of points evenly distributed between the starting and ending points as initial control points, and then use the particle swarm algorithm to optimize the control points with the goal of avoiding the skeleton and other blood vessels and passing through the maximum sum of curve smoothness in the Heinrich unit value area, and use the optimized control points to generate a B-spline curve, and interpolate to obtain the completed connecting line;

[0030] The optimization unit is configured to: based on the completed connection line, uniformly expand the completed blood vessel into a specified radius.

[0031] A third aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the head and neck blood vessel completion method based on a particle swarm optimization algorithm as described above.

[0032] The fourth aspect of the present invention provides a computer device, comprising a computer-readable storage medium, a processor, and a computer program stored on the computer-readable storage medium and executable on the processor, wherein when the processor executes the program, the steps in the head and neck blood vessel completion method based on the particle swarm optimization algorithm as described above are implemented.

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] The present invention combines vascular anatomical knowledge, vascular connectivity domain analysis, HU value of the image, adjacent blood vessels and adjacent bone masks and other information, and generates connecting lines between broken blood vessels through particle swarm optimization and B-spline curves, fully considering the anatomical characteristics and imaging characteristics of the blood vessels, ensuring the accuracy and authenticity of the completed blood vessels, and conforming to the physiological characteristics of the blood vessels.

[0035] The present invention can generate reasonable connection lines in complex vascular structures, avoiding the limitations of traditional methods in processing complex vascular structures. The generated complete vascular structure can provide more accurate diagnostic information, help doctors better evaluate vascular lesions, and improve the accuracy and reliability of diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0037] Figure 1 is a flow chart of a head and neck blood vessel completion method based on a particle swarm optimization algorithm according to a first embodiment of the present invention;

[0038] Figure 2 is a comparison diagram of a part of the head and neck blood vessels before and after completion in Example 1 of the present invention;

[0039] Figure 3 It is a module diagram of each unit of the second embodiment of the present invention;

[0040] Figure 4 It is a structural diagram of a computer device according to a fourth embodiment of the present invention. DETAILED DESCRIPTION

[0041] To make the purpose, technical solution and advantages of the embodiments of the present invention more clear, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0042] It should be noted that the following detailed descriptions are all illustrative and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0043] Embodiment 1

[0044] This embodiment provides a head and neck blood vessel completion method based on a particle swarm optimization algorithm.

[0045] The present embodiment provides a head and neck blood vessel completion method based on a particle swarm optimization algorithm, which can be applied to a computer in a CTA imaging system.

[0046] This embodiment provides a head and neck blood vessel completion method based on a particle swarm optimization algorithm, such as Figure 1 As shown, the method includes the following steps: first, inputting a CTA image of head and neck vessels, and extracting masks of the initial segmented vessels and bone areas based on a pre-trained vascular bone segmentation model; then, naming each branch of the initial segmented vessels based on the point cloud segmentation model of each branch of the head and neck; according to vascular information (including vascular anatomical knowledge, vascular connectivity domain analysis, HU value of CTA images, adjacent vessels and adjacent bone masks), generating connecting lines between broken vessels through particle swarm optimization (PSO) and B-spline curves to ensure the accuracy and rationality of the connecting lines; finally, the completed connecting lines are uniformly expanded into completed vessels of a specified radius, and combined with the initial segmented vessels to obtain the final complete vascular structure.

[0047] The present embodiment provides a head and neck vessel completion method based on a particle swarm optimization algorithm, which can find the optimal connection path in a complex vascular structure, avoiding the problems of mis-segmentation and missed segmentation that are prone to occur in traditional methods. The method is applicable to CTA images of head and neck vessels of different individuals, has strong generalization capabilities, and can process vascular structures of different complexities. The generated complete vascular structure can be used in a variety of clinical applications, such as the diagnosis of vascular diseases, surgical planning, and treatment effect evaluation, thereby improving the accuracy and efficiency of clinical diagnosis and treatment.

[0048] Step 1: Input the head and neck vascular CTA image, and extract the initial segmented blood vessels (the extracted blood vessels are called initial segmented blood vessels to distinguish them from the blood vessels that are subsequently completed) and the mask of the bone area from the head and neck vascular CTA image based on the pre-trained blood vessel bone segmentation model. The specific steps include: loading the head and neck vascular CTA image, inputting it into the blood vessel bone segmentation model, and outputting the mask of the initial segmented blood vessels and the bone area as the input for the next step.

[0049] Step 2: Based on the pre-trained point cloud segmentation model of the head and neck branches, extract the branch vessels from the initial segmented vessels.

[0050] Step 3: Divide each branch vessel based on vascular anatomy knowledge to obtain multiple groups of vessels to be completed.

[0051] Step 4: Based on the vascular information (including vascular anatomical knowledge, vascular connectivity analysis, HU values ​​of CTA images, adjacent blood vessels and adjacent bone masks), a connection line is generated for each blood vessel to be completed through a particle swarm algorithm and a B-spline curve.

[0052] Step 5: Based on the completed connection lines, the completed blood vessels are uniformly expanded into the specified radius, combined with the initial segmented blood vessels to obtain the final complete blood vessel structure.

[0053] In this embodiment, the vascular skeleton segmentation model and the head and neck branch point cloud segmentation model can be trained in advance, so as to extract the initial segmented blood vessels and skeleton regions from the head and neck vascular CTA image using the trained vascular skeleton segmentation model, and extract the head and neck branch blood vessel regions from the initial segmented blood vessels using the trained head and neck branch point cloud segmentation model. The vascular skeleton segmentation model can be a commonly used segmentation network such as nnUnet (a deep learning framework that automatically adapts to different medical image segmentation tasks), and the head and neck branch point cloud segmentation model can be a commonly used point cloud segmentation network such as PointNet (a deep learning model based on three-dimensional point cloud data).

[0054] Among them, the trained vascular bone segmentation model is obtained by the following steps: first, the head and neck CTA image with manually labeled vascular and bone areas is obtained, and the head and neck CTA image with manually labeled vascular and bone areas is input into the pre-built vascular bone segmentation model, and then the relevant parameters in the vascular bone segmentation model are adjusted to obtain a vascular bone segmentation model with better performance, and the segmented vascular and bone areas are used as the input of the next level.

[0055] Among them, the pre-trained head and neck branch point cloud segmentation model is obtained by the following steps:

[0056] (1) Obtain manually annotated head and neck vascular regions of each branch. The annotated head and neck vascular branches mainly include: aortic arch, left subclavian artery, left common carotid artery, left external carotid artery, left internal carotid artery C1 cervical segment, left internal carotid artery C2 petrous segment, left internal carotid artery C3 foramen segment, left internal carotid artery C4 cavernous sinus segment, left internal carotid artery C5 clinoid segment, left internal carotid artery C6 ocular segment, left internal carotid artery C7 communicating segment, left vertebral artery V1 extraosseous segment, left vertebral artery V2 transverse segment, left vertebral artery V3 transverse segment, left vertebral artery V ... transverse segment, left vertebral artery V5 transverse segment, left vertebral artery V6 transverse segment, left vertebral artery V7 transverse segment, left vertebral artery V7 transverse segment, left vertebral artery V7 transverse segment, left vertebral artery V7 transverse segment, left vertebral artery V7 transverse segment, left vertebral artery V7 transverse segment, left vertebral artery V7 transverse segment, left vertebral artery V7 transverse segment, left vertebral artery V7 transverse segment, left vertebral artery V7 transverse segment, left vertebral artery V7 transverse segment, left vertebral artery V7 transverse segment, left vertebral artery V7 transverse segment, left vertebral artery V Vertebral artery V3 extravertebral-1, left vertebral artery V3 extravertebral-2, left vertebral artery V4 intradural-1, left vertebral artery V4 intradural-2, left posterior communicating artery, left anterior cerebral artery A1, left anterior cerebral artery A2, left anterior cerebral artery A3, left anterior cerebral artery A4-5, left middle cerebral artery M1, left middle cerebral artery M2, left middle cerebral artery M3, left middle cerebral artery M4-5, left posterior cerebral artery P1, left posterior cerebral artery P2, left posterior cerebral artery P3, left posterior cerebral artery P4, superior cerebral artery, inferior cerebral artery, right subclavian artery, right common carotid artery, right external carotid artery, right internal carotid artery C1 cervical segment, right internal carotid artery C2 petrous segment, right internal carotid artery C3 foramen segment, right internal carotid artery C4 cavernous sinus segment, right internal carotid artery C5 clinoid segment, right internal carotid artery C6 ophthalmic segment, right internal carotid artery C7 communicating segment, right vertebral artery V1 extraosseous segment, right vertebral artery V2 transverse segment, right vertebral artery V3 extravertebral-1, right vertebral artery V3 extravertebral-2, right vertebral artery Artery V4 intradural-1, right vertebral artery V4 intradural-2, right posterior communicating artery, right anterior cerebral artery A1, right anterior cerebral artery A2, right anterior cerebral artery A3, right anterior cerebral artery A4-5, right middle cerebral artery M1, right middle cerebral artery M2, right middle cerebral artery M3, right middle cerebral artery M4-5, right posterior cerebral artery P1, right posterior cerebral artery P2, right posterior cerebral artery P3, right posterior cerebral artery P4, basilar artery, anterior communicating artery, a total of 63 detailed vascular branches;

[0057] (2) The manually labeled head and neck branch blood vessel regions are input into the pre-built head and neck branch point cloud segmentation model, and then the relevant parameters in the head and neck branch point cloud segmentation model are adjusted to obtain a head and neck branch point cloud segmentation model with better performance. The segmented head and neck branch blood vessels are used as the input of the next level.

[0058] In this embodiment, after obtaining the trained vascular skeleton segmentation model, the head and neck CTA image of a patient can be identified based on the vascular skeleton segmentation model to obtain the vascular and skeleton areas; further, the initial segmented vascular area is input into the trained head and neck branch point cloud segmentation model to extract the head and neck branch vascular areas.

[0059] In step 3, the vascular regions of each branch of the head and neck are divided based on the vascular anatomy knowledge to obtain multiple vascular groups to be completed, so as to facilitate the rapid completion of the broken blood vessels. The specific steps include: based on the vascular regions of each branch of the head and neck and the vascular anatomy knowledge, each branch of the blood vessel is divided into multiple vascular groups to be completed in a variety of ways:

[0060] ① A group of different segmented blood vessels of the same branch vessel: for example, the left internal carotid artery includes: the left internal carotid artery C1 cervical segment, the left internal carotid artery C2 petrous segment, the left internal carotid artery C3 rupture foramen segment, the left internal carotid artery C4 cavernous sinus segment, the left internal carotid artery C5 clinoid segment, the left internal carotid artery C6 ocular segment, and the left internal carotid artery C7 communicating segment;

[0061] ② A group of connected branch vessels: for example, left common carotid artery + left internal carotid artery;

[0062] ③ According to vascular anatomy, the possible connected branch vessels are grouped into: left internal carotid artery C7 communicating segment + basilar artery + left posterior communicating artery + left middle cerebral artery M1.

[0063] In step 4, based on the vascular information (including vascular anatomical knowledge, vascular connectivity analysis, HU values ​​of CTA images, adjacent blood vessels and adjacent bone masks), the particle swarm algorithm and B-spline curve are used to generate connection lines for each blood vessel to be completed. The specific steps include:

[0064] Step 401: Analyze each vessel group to be completed obtained in step 3 one by one, and separate the unconnected vessel segments in the vessel group through connected domain analysis (the connected vessel segments indicate that the vessels are not interrupted and do not need to be completed), and separate them into separate regions. Based on the volume threshold, remove those regions with too small volume, which are usually caused by noise or segmentation errors of the vessel skeleton segmentation model. For example, set a volume threshold V threshold , if the volume V of a region is less than V threshold, it is considered as noise and removed. Based on the vascular anatomy knowledge and spatial relative position, the blood vessels that are incorrectly segmented in step 1 (i.e., the blood vessels that are incorrectly segmented by the point cloud segmentation model of each branch of the head and neck) are identified and removed. The specific method may include: (1) checking whether the spatial distribution of the blood vessel segment conforms to the anatomical law; (2) comparing the spatial relative position of the blood vessel segment with other known blood vessels, and excluding those blood vessel segments with abnormal positions. Then, the surface layer of each blood vessel segment is extracted based on the gradient (the absolute value of the gradient of the blood vessel surface is greater than 0), while reducing the amount of calculation and retaining the characteristics of the blood vessel as much as possible. Based on the extracted blood vessel surface layer coordinates, the Fast Approximate Nearest Neighbor Search (FANNS) algorithm is used to calculate the relationship between each blood vessel segment and its nearest blood vessel, as well as the coordinates of the two nearest points between the two blood vessels. The coordinates of these two nearest points will be used as the starting point and the end point of the connection line to be completed. By analyzing the distance d and direction between the two points, the number of points n = βdk required to complete the connection line can be inferred, thereby achieving accurate completion of the blood vessel model.

[0065] Where d represents the Euclidean distance between the starting and ending points of the blood vessel to be completed. This distance is an important indicator for measuring the length of the blood vessel segment, and intuitively reflects the extension range of the blood vessel part to be completed in space. A longer distance means that more control points are needed to accurately depict the morphology of the blood vessel, because the curve needs to make reasonable transitions and bends in a longer interval.

[0066] Among them, k is the curvature coefficient of the blood vessel. It is a parameter used to quantify the curvature of the blood vessel. The larger the k value, the more curved the blood vessel. The determination of the curvature coefficient is based on anatomical knowledge and is obtained through research and statistical analysis of the curvature characteristics of different types of blood vessels under normal physiological conditions. For example, some blood vessels have a large curvature in anatomy, such as the C4 cavernous sinus segment of the left internal carotid artery, which has a relatively high value; while some relatively straight blood vessel segments, such as the right common carotid artery, have a lower value.

[0067] Among them, β is a proportional coefficient and an empirical value. Its main purpose is to comprehensively consider the distance d and the curvature coefficient k to calculate the number of points required to complete the connecting line.

[0068] Step 402, based on the two blood vessel segments to be completed and the starting and ending points of the connecting line obtained in step 401, combined with the HU value of the head and neck CTA image, the particle swarm algorithm and the B-spline curve are used to make the connecting line pass through the area with high HU value (the HU value of the blood vessel is high) as much as possible when generating the connecting line, and avoid the bones (the HU value of the bones is also high) and other blood vessels.

[0069] In an implementation scenario, suppose the two blood vessels to be completed are a and b, and the blood vessel centerlines of blood vessels a and b are extracted respectively. According to the nearest neighbor principle, the two points closest to the centerlines of the two blood vessels are found, which are defined as the starting point p and b respectively. a , the end point is p b ; According to the distance between the starting and ending points to be completed, determine the number of control points of the B-spline curve. As the distance of the completed connecting line increases, the number of control points should also increase; the initial control points can be evenly distributed on p a and p b between:

[0070]

[0071] Among them, c i represents the ith control point. n represents the number of control points of the B-spline curve. Control points play a key role in the generation of B-spline curves, and they determine the shape and direction of the curve. More control points can provide more flexible curve fitting capabilities, thus better adapting to the complex shape of curved blood vessels.

[0072] Step 403, then, the objective function should contain three main parts: a penalty term for avoiding the skeleton and other blood vessels, a reward term for passing through high HU value areas, and a penalty term for curve smoothness.

[0073] Avoid the skeleton and other blood vessels:

[0074]

[0075] Among them, p c are the coordinates of the skeleton and other blood vessels, d(B(t),p c )) is the distance from point B(t) to p on the B-spline curve c distance, R is a threshold, indicating the minimum distance allowed.

[0076] Passing through high HU (Hounsfield Unit) value areas:

[0077]

[0078] Wherein, HU(B(t)) is the HU value at point B(t) on the B-spline curve.

[0079] Curve smoothness:

[0080]

[0081] Where B′(t) is the first-order derivative of the B-spline curve. When the curve is not smooth and has many bends and twists, the first-order derivative of the curve will change more, and the corresponding sum of squares of the modulus will increase. Therefore, E3 The value of is negatively correlated with the smoothness of the curve, that is, the less smooth the curve is, the larger the value is. This feature is used to encourage the algorithm to find a smoother curve, so that the completed blood vessels are more consistent with physiological characteristics.

[0082] Comprehensive objective function:

[0083] E=λ 1 E 1 +λ 2 E 2 +λ 3 E 3

[0084] Among them, λ 1 , 2 and λ 3 is a weight parameter used to balance various objectives.

[0085] Step 404, particle swarm optimization is used to adjust the coordinates of the control points to minimize the objective function E. During the optimization process, it is also necessary to ensure that the starting point and end point of the B-spline curve are fixed at P a and P b Each particle represents the coordinates of a set of control points, namely x i =[c i0 ,c i1 ,…,c in ], where c ij is the coordinate of the jth control point. Each particle has a velocity v i , indicating the direction and magnitude of its position update.

[0086] Step 405: Initialize the velocity of each particle to zero using the initial control point obtained in step 402. Calculate the objective function value E of each particle. Record the historical best position p of each particle. i and the global optimal position g.

[0087] Step 406, continuously iterate and update, for each particle, update its speed and position:

[0088]

[0089] Among them, w is the inertia weight, which controls the inertia of the particle's velocity update; c 1 and c 2 is the acceleration constant, which controls the speed at which the particle moves to the historical best position and the global best position; r 1 and r 2 is a random number between [0,1].

[0090] Step 407, update the optimal position and terminate, calculate the new objective function value E of each particle. If the new objective function value is less than the current historical optimal value E(p i ), then update p i If the new objective function value is less than the current global optimal value E(g), g is updated. When the maximum number of iterations is reached or the objective function value meets the predetermined threshold, the iteration is stopped.

[0091] Step 408: After generating the final B-spline curve based on the optimized control points, the connecting line is obtained by interpolation. The interpolation step includes:

[0092] Choose a range of parameter values ​​t between [0,1]. For example, you can use equally spaced parameter values:

[0093]

[0094] Where N is the number of points generated.

[0095] Calculate the coordinates of the resulting connecting line, for each parameter value t i , calculate the x, y, and z coordinates respectively:

[0096]

[0097]

[0098] Among them, (c jx ,c jy ,c jz ) is the coordinate of the control point, (x i ,y i ,z i ) are the interpolation coordinates, is the k-order B-spline basis function, which is a piecewise polynomial function that represents the parameter value t i The contribution weight of the jth control point to the corresponding point on the curve is combined into a three-dimensional coordinate p i =(x i ,y i ,z i ) to complete the missing connection line.

[0099] Step 409, the completed connection lines are uniformly expanded into the completed blood vessels with the specified radius and the initially segmented blood vessels to form a complete blood vessel structure, such as Figure 2 shown.

[0100] Combined with the above description, it can be known that the embodiment of the present application uses the head and neck vascular CTA image to extract the mask of the initial segmented blood vessels and the bone area based on the vascular skeleton segmentation model, and names the branch blood vessels of the initial segmented blood vessels based on the point cloud segmentation model of each branch of the head and neck. According to the vascular information, the particle swarm algorithm and B-spline curve are used to generate the connecting lines between the broken blood vessels, and finally the completed connecting lines are uniformly expanded into the completed blood vessels of the specified radius, combined with the initial segmented blood vessels, to obtain the final complete vascular structure. The use of the present application scheme can generate reasonable connecting lines in complex vascular structures, avoiding the limitations of traditional methods in dealing with complex vascular structures. The generated complete vascular structure can provide more accurate diagnostic information, help doctors better evaluate vascular lesions, and improve the accuracy and reliability of diagnosis.

[0101] This embodiment provides a head and neck vascular completion method based on a particle swarm optimization algorithm, which uses vascular anatomical knowledge, vascular connectivity analysis, image HU values, adjacent blood vessels, and adjacent bone masks to generate connecting lines between broken blood vessels through a particle swarm algorithm, thereby generating a complete vascular structure. The complete vascular structure generated by the present application solution not only fills the broken part, but also provides a higher quality vascular image, which helps doctors analyze vascular lesions more carefully and improves the quality of diagnosis.

[0102] Embodiment 2

[0103] This embodiment provides a head and neck blood vessel completion system based on a particle swarm optimization algorithm. Figure 3 As shown, it specifically includes:

[0104] Segmentation unit: This unit first loads the head and neck vascular CTA image, and extracts the mask of the initially segmented blood vessels and bone areas based on the pre-trained vascular bone segmentation model. Subsequently, the pre-trained head and neck branch point cloud segmentation model is used to further extract the branch vessels from the initially segmented blood vessels. These extracted branch vessels will serve as input data for subsequent processing. In addition, the segmentation unit is also responsible for obtaining detailed head and neck vascular area information for each branch based on the head and neck branch segmentation model.

[0105] Vessel division unit: This unit divides each branch vessel into multiple groups of vessels to be completed in a variety of ways based on vascular anatomy knowledge and regional information of each branch head and neck vessel. The division methods include but are not limited to: division according to different segments of the same branch vessel, combination according to connected branch vessels, and preset combination according to potentially connected branch vessels.

[0106] Computational unit: This unit is responsible for generating connection lines for each group of blood vessels to be completed based on vascular information (including vascular anatomical knowledge, vascular connectivity domain analysis, HU values ​​of CTA images, adjacent blood vessels and adjacent bone masks, etc.) through particle swarm algorithm and B-spline curve technology.

[0107] Optimization unit: This unit is based on the completed connection lines generated by the calculation unit. It expands them uniformly into completed blood vessels with a specified radius and combines them with the initial segmented blood vessels to finally obtain a complete blood vessel structure.

[0108] It should be noted here that each module in this embodiment corresponds to each step in Example 1 one by one, and the specific implementation process is the same, which will not be repeated here.

[0109] Embodiment 3

[0110] This embodiment provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps in the head and neck blood vessel completion method based on a particle swarm optimization algorithm as described in the first embodiment above are implemented.

[0111] Embodiment 4

[0112] This embodiment provides a computer device, such as Figure 4 As shown, it includes a display device, an input device, a computer-readable storage medium (volatile memory and non-volatile storage medium), a processor, a communication interface (i.e., a network interface), and a computer program stored on the computer-readable storage medium and executable on the processor, wherein the processor, the communication interface, and the computer-readable storage medium can be connected via a bus or other means. The communication interface is used to receive and send data, and when the processor executes the program, the steps in the head and neck blood vessel completion method based on the particle swarm optimization algorithm as described in the first embodiment above are implemented.

[0113] Among them, any reference to memory, storage, database or other media provided by the present application and used in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual-speed data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0114] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0115] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0116] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0117] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A head and neck blood vessel completion method based on particle swarm optimization algorithm, characterized in that: include: Obtain angiography images of the head and neck, extract the masks of the initial segmented blood vessels and the bone area through the blood vessel bone segmentation model, and then obtain the blood vessel areas of each branch of the head and neck through the point cloud segmentation model of each branch of the head and neck for the initial segmented blood vessels; Dividing the branch blood vessel regions of the head and neck into multiple blood vessel groups to be completed by multiple division methods; For each vessel to be completed in each vessel group, the nearest neighbor vessel to be completed is found, and the adjacent points of the two vessels to be completed are used as the starting and ending point coordinates. After using several points evenly distributed between the starting and ending points as initial control points, the particle swarm algorithm is used to optimize the control points with the goal of avoiding the skeleton and other vessels and passing through the maximum sum of curve smoothness in the Heinrich unit value area. The optimized control points are used to generate a B-spline curve, and the completion connection line is obtained by interpolation; Based on the completed connection lines, the completed blood vessels are uniformly expanded into the specified radius.

2. A head and neck blood vessel completion method based on particle swarm optimization algorithm as claimed in claim 1, characterized in that: The division methods include: division according to different segments of the same branch blood vessel, combination according to connected branch blood vessels, and combination according to possibly connected branch blood vessels.

3. The head and neck blood vessel completion method based on particle swarm optimization algorithm as claimed in claim 1, characterized in that: The number of the control points is determined according to the distance between the starting and ending points and the curvature of the blood vessel.

4. The head and neck blood vessel completion method based on particle swarm optimization algorithm as claimed in claim 1, characterized in that: Before searching for the nearest neighbor blood vessels to be completed, the method also includes: for each group of blood vessels to be completed, removing blood vessels that are noise-prone or segmented incorrectly by the blood vessel skeleton segmentation model; removing blood vessels that are segmented incorrectly by the point cloud segmentation model of each branch of the head and neck; and extracting the coordinates of the blood vessel surface layer.

5. The head and neck blood vessel completion method based on particle swarm optimization algorithm as claimed in claim 1, characterized in that: The avoiding skeleton and other blood vessels is expressed as: Among them, p c are the coordinates of the skeleton and other blood vessels, d(B(t),p c )) is the distance from point B(t) to p on the B-spline curve c distance, R is a threshold.

6. The head and neck blood vessel completion method based on particle swarm optimization algorithm as claimed in claim 1, characterized in that: The area of ​​the Hounsfield Unit value is expressed as: Where HU(B(t)) is the value of the Hounsfield unit at the point B(t) on the B-spline curve.

7. The head and neck blood vessel completion method based on particle swarm optimization algorithm as claimed in claim 1, characterized in that: The curve smoothness is expressed as: Where B′(t) is the first-order derivative of the B-spline curve.

8. A head and neck vascular completion system based on particle swarm optimization algorithm, characterized in that: include: The segmentation unit is configured to: acquire a head and neck angiography image, extract masks of the initial segmented blood vessels and the bone region through a blood vessel skeleton segmentation model, and obtain the head and neck branch blood vessel regions through a head and neck branch point cloud segmentation model for the initial segmented blood vessels; A blood vessel division unit is configured to: divide the branch blood vessel regions of the head and neck into a plurality of blood vessel groups to be completed by using a plurality of division methods; The computing unit is configured to: for each blood vessel to be completed in each blood vessel group to be completed, find the nearest neighbor blood vessel to be completed, and use the adjacent points of the two blood vessels to be completed as the starting and ending point coordinates, and use a number of points evenly distributed between the starting and ending points as initial control points, and then use the particle swarm algorithm to optimize the control points with the goal of avoiding the skeleton and other blood vessels and passing through the maximum sum of curve smoothness in the Heinrich unit value area, and use the optimized control points to generate a B-spline curve, and interpolate to obtain the completed connecting line; The optimization unit is configured to: based on the completed connection line, uniformly expand the completed blood vessel into a specified radius.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps in a head and neck blood vessel completion method based on a particle swarm optimization algorithm as described in any one of claims 1 to 7 are implemented.

10. A computer device comprising a computer-readable storage medium, a processor, and a computer program stored in the computer-readable storage medium and executable on the processor, characterized in that: When the processor executes the program, the steps in the head and neck blood vessel completion method based on particle swarm optimization algorithm as described in any one of claims 1 to 7 are implemented.

Images