Blood vessel three-dimensional reconstruction method and device based on multi-view angiography image, computer device and storage medium

Abstract

The application discloses a blood vessel three-dimensional reconstruction method and device based on multi-view angiography images, computer equipment and a storage medium, relates to the field of blood vessel simulation, and comprises the following steps: firstly, key frames are selected from a multi-angle angiography sequence, high-precision blood vessel tree segmentation is carried out, and a reference image is determined; secondly, a three-dimensional projection cone is constructed by extracting projection parameters, an intersection thereof is taken and is discretized, the reference image and other view information are iteratively screened, and a blood vessel tree space domain is accurately defined; finally, a blood vessel center line is extracted, a method plane is generated along the center line, a blood vessel cross section is obtained, and three-dimensional model reconstruction is carried out based on a topological structure. The application is innovative in that the space discretization and traversal projection thought are introduced, so that the increase of the number of views does not significantly improve the calculation complexity, thereby breaking through the view number limitation of traditional multi-view reconstruction, and the application has the advantages of high precision and strong expandability, and can meet the harsh demands of high-end clinical applications such as blood flow dynamics simulation on the precision of the blood vessel model.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of blood vessel simulation, in particular to a blood vessel three-dimensional reconstruction method and device based on multi-view angiography images, a computer device and a storage medium. BACKGROUND

[0002] In recent years, hemodynamic simulation technology has been increasingly applied in the medical field, especially in the diagnosis of blood vessel diseases, including intracardiac blood vessels, intracranial blood vessels and systemic blood vessels. In the implementation process of such prior art, blood vessel reconstruction is often the most critical step, and the accuracy of the reconstructed three-dimensional blood vessels greatly affects the simulation results of hemodynamics.

[0003] In the prior art, blood vessel reconstruction methods based on angiography images are mainly divided into two categories. The first category is a blood vessel reconstruction method based on a single angiography image. This method does not directly reconstruct a three-dimensional blood vessel model, but only obtains the three-dimensional spatial parameters of the blood vessel through angiography parameters or catheter calibration. However, this method does not consider the spatial distortion and projection reduction of the blood vessel, and cannot accurately obtain the length information of the blood vessel axis. The second category is a blood vessel reconstruction method based on two or more angiography images. This method often obtains a three-dimensional blood vessel model through spatial registration. Although this method considers spatial distortion and projection reduction, it often uses a circular cross-section or an elliptical cross-section to replace the actual blood vessel cross-section, which often has a large error when dealing with eccentric stenosis or complex plaque morphology. Therefore, the blood vessel reconstruction methods based on angiography images in the prior art have certain limitations. SUMMARY

[0004] The present application aims to provide a blood vessel three-dimensional reconstruction method and device based on multi-view angiography images, a computer device and a storage medium, which takes different view angiography images of the same blood vessel as input, fully considers the morphological characteristics of the two-dimensional projection of the blood vessel under different views, and finally obtains a spatial three-dimensional model of the blood vessel using a spatial traversal method, to solve the problems raised in the background art.

[0005] To achieve the above-mentioned purpose, the present application provides the following technical solution: a blood vessel three-dimensional reconstruction method based on multi-view angiography images, comprising the following steps:

[0006] Blood vessel tree segmentation: import multi-angle angiography images labeled with at least one key frame image at each angle, obtain the corresponding blood vessel tree two-dimensional region of each key frame image through segmentation strategy, and mark the reference two-dimensional region and the corresponding reference image based on the analysis of the blood vessel tree two-dimensional region;

[0007] Space domain initialization: extract the projection parameters of each key frame image to generate the corresponding projection area, obtain the spatial intersection of all projection areas to construct the target space domain, and discretize the target space domain into multiple space grids. Starting from the key frame image marked with the reference image, calculate the projection point of each space grid on the corresponding key frame image one by one, and remove the space grids whose projection points are not within the corresponding blood vessel tree two-dimensional area to obtain the blood vessel tree space domain;

[0008] Centerline extraction: based on the blood vessel tree space domain, obtain a blood vessel tree centerline composed of multiple discrete center points;

[0009] Three-dimensional reconstruction of blood vessels: generate a tangent plane through each center point of the blood vessel tree centerline, and construct an initial blood vessel cross-section with the blood vessel tree space domain. Extract the boundary line of each initial blood vessel cross-section to generate a closed curve and construct a blood vessel boundary line for the initial blood vessel cross-section. Based on the blood vessel boundary line, construct a three-dimensional model of the blood vessel tree.

[0010] As a preferred, the segmentation strategy includes segmenting the blood vessel tree on each key frame image based on threshold method or deep neural network method to obtain the blood vessel tree segmentation result corresponding to each key frame image. The blood vessel tree segmentation result on each key frame image is traversed for error and error correction to obtain the blood vessel tree two-dimensional area on each key frame image.

[0011] As a preferred, the analysis of the blood vessel tree two-dimensional area includes calculating the area of each blood vessel tree two-dimensional area, and marking the blood vessel tree two-dimensional area with the smallest area as the reference two-dimensional area, and the corresponding key frame image as the reference image.

[0012] As a preferred, the generation method of the projection area includes the following steps:

[0013] Determine the point light source coordinates based on the projection parameters of the key frame image and the endpoint coordinates of the key frame image;

[0014] Connect the point light source and the endpoint respectively to generate a projection ray, and the area surrounded by the projection ray constitutes the projection area corresponding to the key frame image.

[0015] As a preferred, the extraction method of the blood vessel tree centerline specifically includes the following steps:

[0016] Based on the three-dimensional erosion algorithm, the blood vessel tree space domain is eroded to obtain a blood vessel tree centerline grid;

[0017] Downsample the blood vessel tree centerline grid to generate a blood vessel tree centerline composed of multiple discrete center points.

[0018] As a preferred, the method for generating a closed curve from a boundary line includes the following steps:

[0019] extracting vertices of the boundary line, selecting at least one point on the edge between two adjacent vertices as a blood vessel boundary control point;

[0020] connecting the blood vessel boundary control points in sequence to generate the closed curve, the closed curve being tangent to each edge between two adjacent vertices.

[0021] Preferably, the method for constructing the closed curve from the blood vessel boundary line comprises dynamically adjusting the position of the blood vessel boundary control point until the closed curve reaches a convergence condition of minimum area, and the closed curve after reaching the convergence condition is the blood vessel boundary line on the initial blood vessel cross section.

[0022] Preferably, the method for constructing the blood vessel tree three-dimensional model from the blood vessel boundary line comprises establishing a topology based on the blood vessel tree center line, and connecting all the blood vessel boundary lines on the initial blood vessel cross sections in sequence according to the topology.

[0023] Preferably, the discretization of the spatial grid comprises setting a spatial step length based on the blood vessel size, and discretizing the target spatial domain according to the spatial step length.

[0024] Preferably, the down-sampling is based on a multiple of the spatial step length to generate a smooth blood vessel tree center line.

[0025] To solve the above technical problems, the application further provides a blood vessel three-dimensional reconstruction device based on multi-view angiography images, comprising:

[0026] a blood vessel tree segmentation module for importing multi-angle angiography images marked with at least one key frame image at each angle, obtaining a blood vessel tree two-dimensional region corresponding to each key frame image through a segmentation strategy, and marking a reference two-dimensional region and a corresponding reference image based on analysis of the blood vessel tree two-dimensional region;

[0027] a spatial domain setting module for extracting projection parameters of each key frame image to generate a corresponding projection region, obtaining a spatial intersection of all projection regions to construct a target spatial domain, discretizing the target spatial domain into a plurality of spatial grids based on a spatial step length set according to the blood vessel size, and starting from the key frame image marked with the reference image, calculating the projection point of each spatial grid on the corresponding key frame image one by one, and removing the spatial grid whose projection point is not within the corresponding blood vessel tree two-dimensional region to obtain a blood vessel tree spatial domain;

[0028] A blood vessel three-dimensional reconstruction module is used for obtaining a blood vessel centerline composed of a plurality of discrete center points based on the blood vessel tree spatial domain, generating a normal plane through each center point of the blood vessel centerline, and constructing an initial blood vessel section with the blood vessel tree spatial domain, extracting a boundary line of each initial blood vessel section to generate a closed curve as a blood vessel boundary line of the initial blood vessel section, and constructing a blood vessel tree three-dimensional model based on the blood vessel boundary line.

[0029] The application further provides a computer device, comprising:

[0030] a memory for storing a computer program;

[0031] a processor for executing the computer program, which, when executed by the processor, implements the following steps:

[0032] blood vessel tree segmentation: importing multi-angle contrast imaging marked with at least one key frame image at each angle, obtaining a blood vessel tree two-dimensional region corresponding to each key frame image through a segmentation strategy, and marking a reference two-dimensional region and a corresponding reference image based on analysis of the blood vessel tree two-dimensional region;

[0033] spatial domain preliminary determination: extracting a projection parameter of each key frame image to generate a corresponding projection region, obtaining a spatial intersection of all projection regions to construct a target spatial domain, and discretizing the target spatial domain into a plurality of spatial grids, starting from the key frame image marked with the reference image, calculating a projection point of each spatial grid on the corresponding key frame image one by one, and removing the spatial grid whose projection point is not in the corresponding blood vessel tree two-dimensional region to obtain a blood vessel tree spatial domain;

[0034] center line extraction: obtaining a blood vessel centerline composed of a plurality of discrete center points based on the blood vessel tree spatial domain;

[0035] blood vessel three-dimensional reconstruction: generating a normal plane through each center point of the blood vessel centerline, and constructing an initial blood vessel section with the blood vessel tree spatial domain, extracting a boundary line of each initial blood vessel section to generate a closed curve as a blood vessel boundary line of the initial blood vessel section, and constructing a blood vessel tree three-dimensional model based on the blood vessel boundary line.

[0036] The application further provides a readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the following steps:

[0037] blood vessel tree segmentation: importing multi-angle contrast imaging marked with at least one key frame image at each angle, obtaining a blood vessel tree two-dimensional region corresponding to each key frame image through a segmentation strategy, and marking a reference two-dimensional region and a corresponding reference image based on analysis of the blood vessel tree two-dimensional region;

[0038] Space domain initial determination: extract the projection parameters of each key frame image to generate the corresponding projection area, obtain the spatial intersection of all projection areas to construct a target space domain, and discretize the target space domain into a plurality of space grids, starting from the key frame image marked with the reference image, calculate the projection point of each space grid on the corresponding key frame image one by one, and remove the space grid whose projection point is not in the corresponding blood vessel tree two-dimensional area to obtain a blood vessel tree space domain;

[0039] Center line extraction: based on the blood vessel tree space domain, a blood vessel tree center line composed of a plurality of discrete center points is obtained;

[0040] Vessel three-dimensional reconstruction: a tangent plane is generated through each center point of the blood vessel tree center line, and an initial blood vessel cross section is constructed with the blood vessel tree space domain, a boundary line of each initial blood vessel cross section is extracted to generate a closed curve to construct a blood vessel boundary line of the initial blood vessel cross section, and a blood vessel tree three-dimensional model is constructed based on the blood vessel boundary line.

[0041] In summary, the present application has the following advantages:

[0042] Compared with the single-view blood vessel reconstruction method in the prior art, the present application introduces multiple-view angiography images, which overcomes the technical defects of blood vessel distortion and projection reduction. The traditional multi-view blood vessel reconstruction method in the prior art needs to consider the spatial matching problem under different views. With more imported angiography images, the technical complexity increases geometrically. Therefore, such methods are often limited to double-view or triple-view blood vessel reconstruction, which limits the accuracy of blood vessel cross section. In the present application, the method of spatial discretization and spatial traversal projection is creatively used. Importing more different view angiography images does not significantly increase the computational complexity, has a higher precision limit and scalability, and can meet the different clinical needs of blood flow dynamics simulation for blood vessel reconstruction accuracy.

[0043] The present application also provides a blood vessel three-dimensional reconstruction device based on multi-view angiography images, a computer device and a storage medium, which have the above-mentioned advantages, which will not be described here. BRIEF DESCRIPTION OF DRAWINGS

[0044] In order to more clearly illustrate the technical solutions of the embodiments of the application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiment or prior art description. Obviously, the drawings in the following description are only some embodiments of the application, and for those skilled in the art, other drawings can also be obtained without creative labor.

[0045] Figure 1 The present application provides a blood vessel three-dimensional reconstruction method based on multi-view angiography images.

[0046] Figure 2 Fig. 1 is a schematic diagram of a spatial domain initial flow framework of a blood vessel three-dimensional reconstruction method based on multi-view angiography images according to the present application.

[0047] Figure 3 Fig. 2 is a schematic diagram of a blood vessel three-dimensional reconstruction flow framework of a blood vessel three-dimensional reconstruction method based on multi-view angiography images according to the present application.

[0048] Figure 4 Fig. 3 is a structural block diagram of a blood vessel three-dimensional reconstruction device based on multi-view angiography images according to the present application.

[0049] Figure 5 Fig. 4 is an internal structure diagram of a computer device according to an embodiment of the present application. DETAILED DESCRIPTION

[0050] The present application will now be described in further detail by reference to the drawings, wherein:

[0051] For the purpose of clarity, the description will be made with reference to the accompanying drawings. The application can be implemented in many different ways and is not limited to the embodiments described herein. Rather, the embodiments are provided as examples of the application to make the disclosure more thorough and complete.

[0052] All features disclosed in this specification, and / or all steps of any methods disclosed, may be combined in any combination, provided that the features and / or steps are not mutually inconsistent.

[0053] Any feature disclosed in this specification, unless otherwise indicated, can be replaced by an alternative feature serving the same or a similar purpose.

[0054] The application will now be described in more detail with reference to the accompanying drawings. Figure 1 - Figure 5 The application will now be described in more detail with reference to the accompanying drawings. Figures 1-3 The method comprises four work flows, i.e. blood vessel tree segmentation, spatial domain initial flow, center line extraction and blood vessel three-dimensional reconstruction.

[0055] First step: blood vessel tree segmentation

[0056] Import multiple angiography images of the target vessel tree, each image corresponding to a different angiography angle, and select a key frame in each image with the most complete blood vessel display.

[0057] Specifically, this step is the data basis and quality assurance of the entire blood vessel three-dimensional reconstruction. For example, input a set of two-dimensional X-ray sequences collected during angiography surgery. Each sequence is a series of dynamic images taken from a specific angle during the rotation of the C-arm X-ray machine around the patient, recording the filling-peak-decay process of the contrast agent in the blood vessels. These sequences must come from different angiography angles, such as left anterior oblique 30°, right anterior oblique 45°, and frontal 0°. Multiple angles are a prerequisite for subsequent three-dimensional spatial positioning.

[0058] In each angiography sequence at different angles, the static image with the most clear and complete blood vessel tree display is selected as the key frame. This is usually the moment when the contrast agent just fills the target blood vessel tree and reaches peak concentration. The quality of the key frame directly determines the difficulty and effectiveness of segmentation.

[0059] Segment the blood vessel tree in each key frame image using thresholding or deep neural network methods to obtain the blood vessel tree segmentation result corresponding to each key frame image. Thresholding works well when the contrast between blood vessels and background is extremely high, but it is easily disturbed by factors such as bone and noise. Deep neural network methods can understand the morphological characteristics of blood vessels, such as tubular and connectivity, thus more accurately distinguishing blood vessels from background areas with similar brightness and having stronger anti-interference ability.

[0060] Check the blood vessel tree segmentation results on each key frame image one by one, correct parts with large errors or segmentation errors, and finally obtain the blood vessel tree two-dimensional area on each key frame image.

[0061] Calculate the area of each blood vessel tree two-dimensional area, and select the one with the smallest area as the reference two-dimensional area, and the corresponding key frame image as the reference image.

[0062] Area calculation: For each corrected binary mask, that is, the blood vessel tree two-dimensional area, count the total number of pixels marked as blood vessels. This value represents the projected area of the blood vessels at this angle. Select the key frame image with the smallest projected area at this angle as the reference.

[0063] This is because when a long object is not directly opposite the viewing direction, its projection will be shortened, the smallest area of view, most likely closest to the direction perpendicular to the main stem of the blood vessels, at this viewing angle, the projection of the blood vessels is the least deformed, the structure is relatively most true, so in the subsequent spatial domain determination, first use this reference viewing angle to eliminate a large number of spatial grids that definitely do not belong to the blood vessels, use a deformed minimum, most compact projection as the first screening standard, which can most efficiently and accurately narrow down the search range and avoid prematurely misremoving the real blood vessels.

[0064] Second step: spatial domain determination

[0065] 1. Extract the projection parameters from each key frame image, determine the point light source coordinates, connect the point light source with the four end points of the key frame image to generate four projection rays, and the area surrounded by the four projection rays constitutes the projection area corresponding to the key frame image. For each two-dimensional viewing angle, define an area that can be seen in three-dimensional space.

[0066] Specifically, in an angiography system, each image corresponds to a unique pose of the C-arm in space, and the key projection parameters include:

[0067] Point light source coordinates: the focal point position of the X-ray emitter, which is the origin of all projection rays.

[0068] Detector plane orientation: the spatial position and angle of the flat panel detector that receives X-rays.

[0069] Image center point and pixel spacing: used to convert pixel coordinates on the image into directions in three-dimensional space.

[0070] These parameters are usually encapsulated in a projection matrix and can be directly read from the DICOM file header information of the angiography device or calculated.

[0071] 2. Take the spatial intersection of the projection areas corresponding to all key frame images to form the target spatial domain, and discretize the target spatial domain into multiple spatial grids according to a preset spatial step, thereby finding a spatial range that is recognized by all viewing angles as possibly containing blood vessels and dividing it into small units for easy processing.

[0072] Specifically, perform intersection operation on all the projection areas corresponding to the key frame images in three-dimensional space. This intersection area is the maximum possible space where blood vessels may exist, forming the target spatial domain, which effectively excludes a large number of impossible spaces. Divide this target spatial domain into small cubes according to the set spatial step to form a spatial grid.

[0073] The spatial step determines the size of the cube, for example, 0.5mm square, the smaller the step, the higher the subsequent reconstruction accuracy, but the larger the amount of calculation.

[0074] 3. Extract the projection parameters from the reference image, calculate the corresponding projection matrix, and calculate the projection points of each spatial grid in the target spatial domain on the reference image one by one. If the projection point is located outside the reference two-dimensional region, the corresponding spatial grid is removed from the target spatial domain.

[0075] The angle with the smallest projection area is selected as the reference, and the blood vessel profile at this angle is the most compact, meaning that the side of the blood vessel is the narrowest from this angle. Using it for the first round of screening can most conservatively remove grids and maximize the avoidance of false deletions of real blood vessel parts. After this step, the target spatial domain is greatly reduced.

[0076] 4. Next, process the remaining key frame images in a certain order.

[0077] Extract the projection parameters from the next key frame image, calculate the corresponding projection matrix, and calculate the projection points of each spatial grid in the updated target spatial domain on the key frame image one by one. If the projection point is located outside the corresponding blood vessel tree two-dimensional region, the corresponding spatial grid is removed from the target spatial domain.

[0078] Repeat the previous step until all remaining key frame images are processed. When all key frame images are processed, the remaining spatial grid set is the blood vessel tree spatial domain that meets all projection constraints from different angles.

[0079] Using the reference image and other key frame images, determine whether the projection point of each spatial grid falls within the corresponding blood vessel two-dimensional region one by one, and perform iterative removal to accurately select the real range from the possible range using real blood vessel profile information.

[0080] Step 3: Centerline extraction

[0081] Use a three-dimensional erosion algorithm to erode the blood vessel tree spatial domain to obtain the blood vessel tree centerline grid. Downsample the blood vessel tree centerline grid by 6 times the spatial step to generate the final blood vessel tree centerline. The blood vessel tree centerline is composed of multiple discrete center points. Slim the thick and solid three-dimensional blood vessel model until only its most core skeleton remains. This skeleton is the centerline of the blood vessel.

[0082] Among them, erosion is a basic operation in image morphology. In three-dimensional space, it uses a preset structure element, usually a small cube such as 3x3x3, to scan the entire blood vessel tree spatial domain. If there is any non-blood vessel in the neighborhood of a given spatial grid, the grid will be eroded, i.e. removed from the blood vessel domain.

[0083] Next, the generated vessel tree centerline mesh is simplified and smoothed by using a 6-fold spatial step to down-sample the vessel tree centerline mesh.

[0084] Along each centerline path of the vessel tree centerline mesh, points are sampled at fixed distance intervals, where the fixed distance interval is set to 6 times the original spatial step d. Starting from the path starting point, the cumulative path distance is accumulated along the path. When the cumulative distance reaches 6 * d, a point is selected at the current position as a new center point. This process is repeated until the path ending point, so that the number of points is greatly reduced and the path is simplified.

[0085] Step 4: Vessel 3D reconstruction

[0086] 1. Generate a plane for each center point that constitutes the vessel tree centerline, and the area of the plane in the vessel tree spatial domain constitutes the initial vessel cross section. At each point of the vessel centerline, determine the local cross-sectional shape and size of the vessel.

[0087] From the obtained vessel tree centerline, the centerline is composed of a series of ordered center points P_i. For each point P_i, we can calculate the tangent direction T_i at that point by its adjacent points before and after. The plane passing through point P_i and perpendicular to the tangent direction T_i is the normal plane at that point, which represents the cross section of the vessel at that position. Perform intersection operation between this normal plane and the previously obtained vessel tree spatial domain. The intersection of the normal plane and the vessel tree spatial domain will form one or more two-dimensional regions, which is the initial vessel cross section.

[0088] 2. Extract the boundary line of each initial vessel cross section one by one, and the boundary line is a polygon, so that the corresponding vertices can be extracted. Find the midpoint of the edge between the two adjacent vertices as the vessel boundary control point, and connect the vessel boundary control points in turn to generate a closed curve tangent to each edge of the polygon. Optimize the rough polygon boundary into a smooth, compact and accurate curve that reflects the true boundary of the vessel.

[0089] 3. Dynamically adjust the position of the control points on each edge until the closed curve reaches the area minimum convergence condition. The closed curve after reaching the convergence condition is the vessel boundary line on the cross section. The area minimum curve will shrink inward as much as possible to fit the inner wall of the vessel, thereby eliminating the boundary error caused by voxel discretization or noise, and obtaining the most compact and possible vessel boundary.

[0090] 4. Establish a topology based on the vessel tree centerline, and connect all the vessel boundary lines on the cross sections in turn according to the topology, to finally generate a vessel tree 3D model.

[0091] Specifically, along the main stem of the centerline, the optimized blood vessel boundary lines corresponding to the adjacent two center points P_i and P{i+1} are found out, and a series of triangular facets are formed by connecting the points on the previous cross-sectional boundary line with the points on the next cross-sectional boundary line one by one using a triangular meshing algorithm such as the shortest diagonal method.

[0092] When the centerline has a branch, special mesh stitching needs to be performed at the branch point. Typically, a cross section is first generated at the end of the main stem and at the start of the branch, and then a smooth transition triangular mesh region is constructed between the multiple cross sections to ensure the geometric continuity at the branch connection.

[0093] After all the adjacent cross sections and branch connection points are covered by triangular facets, a complete three-dimensional surface model of the blood vessel tree composed of countless small triangular facets is obtained.

[0094] In addition, in an embodiment, as shown in Figure 4 a blood vessel three-dimensional reconstruction device based on multi-angle contrast images is provided, which comprises a blood vessel tree segmentation module 100, a spatial domain setting module 200 and a blood vessel three-dimensional reconstruction module 300. Specifically:

[0095] The blood vessel tree segmentation module 100 is used to import multi-angle contrast images marked with at least one key frame image at each angle, obtain a blood vessel tree two-dimensional region corresponding to each key frame image through a segmentation strategy, and mark a reference two-dimensional region and a corresponding reference image based on the analysis of the blood vessel tree two-dimensional region.

[0096] The spatial domain setting module 200 is used to extract the projection parameters of each key frame image to generate a corresponding projection region, obtain the spatial intersection of all projection regions to construct a target spatial domain, and discretize the target spatial domain into multiple spatial grids based on the spatial step set according to the blood vessel size. Starting from the key frame image marked with the reference image, the projection points of the spatial grids on the corresponding key frame image are calculated one by one, and the spatial grids whose projection points are not within the corresponding blood vessel tree two-dimensional region are removed to obtain a blood vessel tree spatial domain.

[0097] The blood vessel three-dimensional reconstruction module 300 is used to obtain a blood vessel tree centerline composed of multiple discrete center points for the blood vessel tree spatial domain, generate a tangent plane through each center point of the blood vessel tree centerline, and construct an initial blood vessel cross section with the blood vessel tree spatial domain. The boundary line of each initial blood vessel cross section is extracted to generate a closed curve to construct a blood vessel boundary line of the initial blood vessel cross section, and a blood vessel tree three-dimensional model is constructed based on the blood vessel boundary line.

[0098] The specific limitations of the device for reconstructing blood vessels in three dimensions based on multi-view angiography images can refer to the limitations of the method for reconstructing blood vessels in three dimensions based on multi-view angiography images described above, which will not be repeated here. Each module in the device for reconstructing blood vessels in three dimensions based on multi-view angiography images described above can be realized by software, hardware, and combinations thereof, in whole or in part. The above-mentioned modules can be embedded in or independent of the processor in the computer device in hardware form, or can be stored in the memory in the computer device in software form, so that the processor calls and executes the operations corresponding to each of the above modules.

[0099] The above describes the embodiments of the method, device, computer device, and storage medium for reconstructing blood vessels in three dimensions based on multi-view angiography images, and the present application also discloses a computer device and a readable storage medium corresponding to the above method.

[0100] The computer device comprises a memory for storing a computer program, and a processor for executing the computer program, wherein the computer program, when executed by the processor, can implement the related steps in the method for reconstructing blood vessels in three dimensions based on multi-view angiography images disclosed in any of the preceding embodiments.

[0101] Figure 5 The computer device 400 shown comprises a memory 401, a processor 402, and a communication interface 403. The memory 401, the processor 402, and the communication interface 403 are communicatively connected to each other. For example, the memory 401, the processor 402, and the communication interface 403 can be communicatively connected in the form of network connection. Alternatively, the computer device 400 described above can further comprise a bus 404. The memory 401, the processor 402, and the communication interface 403 are communicatively connected to each other through the bus 404. Figure 5 The computer device 400 is communicatively connected to each other through the bus 404.

[0102] The memory 401 can be a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 401 can store a program, and when the program stored in the memory 401 is executed by the processor 402, the processor 402 and the communication interface 403 are used to execute the method for reconstructing blood vessels in three dimensions based on multi-view angiography images described above.

[0103] The processor 402 can be a general purpose central processing unit (CPU), a microprocessor, an application specific integrated circuit (ASIC), a graphics processing unit (GPU), or one or more integrated circuits.

[0104] The processor 402 can also be an integrated circuit chip on a circuit board. In implementation, the method according to the present application can be implemented by an integrated logic circuit or an instruction in the form of software in the processor 402. The processor 402 can also be a general purpose processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic device, a discrete gate or transistor logic device, a discrete hardware component. The general purpose processor can be a microprocessor or the processor can also be any conventional processor. The software module can be located in a random access memory, a flash memory, a read only memory, a programmable read only memory or an electrically erasable programmable memory, a register, or other mature storage medium in the art. The storage medium is located in the memory 401, and the processor 402 reads information in the memory 401 and combines the hardware to complete the method for three-dimensional reconstruction of blood vessels based on multi-view angiography images according to the embodiments of the present application.

[0105] The communication interface 403 uses a transceiver module such as but not limited to a transceiver to realize the communication between the computer device 400 and other devices or communication networks.

[0106] When the above computer device 400 includes a bus 404, the bus 404 can include a path for transmitting information between the various components (e.g., the memory 401, the processor 402, the communication interface 403) of the computer device 400.

[0107] In the above embodiments, the method can be implemented by software, hardware, firmware or any combination thereof. When implemented by software, the method can be implemented in the form of a computer program product in whole or in part. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the flow or function of the method for three-dimensional reconstruction of blood vessels based on multi-view angiography images according to the embodiments of the present application is generated in whole or in part.

[0108] The computer can be a general purpose computer, a special purpose computer, a computer network, or other programmable apparatus. The computer instructions can be stored in or transmitted from a computer-readable storage medium, such as from one computer-readable storage medium to another computer-readable storage medium; for example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center, through wired (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (such as infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible by a computer or a data storage device such as a server, data center, etc. integrated with one or more available media sets. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a magnetic tape), an optical medium (for example, a DVD), or a semiconductor medium (for example, a solid state disk (SSD)), etc.

[0109] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described system, device and unit can refer to the corresponding processes in the foregoing method embodiments, which will not be described herein. In addition, each method embodiment and each device embodiment can also be mutually referred to, and the same or corresponding content in different embodiments can be mutually referred to, which will not be described herein.

Claims

1. A method for three-dimensional reconstruction of blood vessels based on multi-view angiographic images, characterized in that: The method comprises the following steps: Vessels Tree segmentation: import multi-angle angiography images marked with at least one key frame image at each angle, obtain the two-dimensional area of each key frame image corresponding to the vessel tree by segmentation strategy, calculate the area of each two-dimensional area of the vessel tree, mark the two-dimensional area with the smallest area as the reference two-dimensional area, and mark the corresponding key frame image as the reference image; Spatial domain initial determination: determine the point light source coordinates and the endpoint coordinates of the key frame image based on the projection parameters of each key frame image, respectively connect the point light source and the endpoint to generate a projection ray, and the area surrounded by the projection ray constitutes the projection area of the corresponding key frame image; obtain the spatial intersection of all projection areas to construct a target spatial domain, and set a spatial step based on the vessel size to discretize the target spatial domain into multiple spatial grids; starting from the key frame image marked with the reference image, calculate the projection point of each spatial grid on the corresponding key frame image one by one, and remove the spatial grid whose projection point is not in the corresponding two-dimensional area of the vessel tree to obtain the vessel tree spatial domain; Center line extraction: based on the vessel tree spatial domain, a vessel tree center line composed of multiple discrete center points is obtained; Vessel three-dimensional reconstruction: a normal plane is generated through each center point of the vessel tree center line, and an initial vessel cross section is constructed with the vessel tree spatial domain; the boundary line of each initial vessel cross section is extracted to generate a closed curve to construct a vessel boundary line of the initial vessel cross section; and a vessel tree three-dimensional model is constructed based on the vessel boundary line.

2. The method of claim 1, wherein: The segmentation strategy includes segmenting the vessel tree on each key frame image based on a threshold method or a deep neural network method to obtain a vessel tree segmentation result corresponding to each key frame image, and traversing the vessel tree segmentation result on each key frame image to correct errors and mistakes to obtain a two-dimensional area of the vessel tree on each key frame image.

3. The method of claim 1, wherein: The extraction method of the vessel tree center line specifically comprises the following steps: Based on a three-dimensional erosion algorithm, the vessel tree spatial domain is eroded to obtain a vessel tree center line grid; The vessel tree center line grid is down-sampled based on a multiple of the spatial step to generate a vessel tree center line composed of multiple discrete center points.

4. The method of claim 1, wherein the method further comprises: The method for generating a closed curve from the boundary line comprises the following steps: Extract the vertices of the boundary line, and select at least one point on the edge line between two adjacent vertices as a vessel boundary control point; The vessel boundary control points are connected in turn to generate the closed curve, and the closed curve is tangent to each edge line between two adjacent vertices.

5. The method of claim 4, wherein: The method for constructing the closed curve as the vessel boundary line comprises dynamically adjusting the position of the vessel boundary control point until the closed curve reaches the convergence condition of the smallest area, and the closed curve after reaching the convergence condition is the vessel boundary line on the initial vessel cross section.

6. A device for three-dimensional reconstruction of blood vessels based on multi-view angiographic images, characterized by: Comprise: The blood vessel tree segmentation module is used for importing the multi-angle contrast image marked with at least one key frame image under each angle, obtaining the blood vessel tree two-dimensional area corresponding to each key frame image through a segmentation strategy, calculating the area of each blood vessel tree two-dimensional area, marking the blood vessel tree two-dimensional area with the smallest area as a reference two-dimensional area, and marking the corresponding key frame image as a reference image; The spatial domain setting module is used for determining the point light source coordinates and the end point coordinates of each key frame image based on the projection parameters of each key frame image, generating a projection ray by connecting the point light source and the end point respectively, and constructing a projection area of the corresponding key frame image surrounded by the projection ray; obtaining the spatial intersection of all projection areas to construct a target spatial domain, and discretizing the target spatial domain into a plurality of spatial grids based on the blood vessel size to set a spatial step; starting from the key frame image marked with the reference image, the projection point of each spatial grid on the corresponding key frame image is calculated one by one, and the spatial grid whose projection point is not in the corresponding blood vessel tree two-dimensional area is removed, to obtain a blood vessel tree spatial domain; The blood vessel three-dimensional reconstruction module is used for obtaining a blood vessel tree center line composed of a plurality of discrete center points from the blood vessel tree spatial domain, generating a tangent plane through each center point of the blood vessel tree center line, and constructing an initial blood vessel cross section with the blood vessel tree spatial domain, extracting the boundary line of each initial blood vessel cross section to generate a closed curve and construct a blood vessel boundary line of the initial blood vessel cross section, and constructing a blood vessel tree three-dimensional model based on the blood vessel boundary line.

7. A computer device, characterized by: It comprises: A memory for storing a computer program; A processor for executing the computer program, which, when executed by the processor, implements the steps of the blood vessel three-dimensional reconstruction method based on multi-view contrast images according to any one of claims 1-5.

8. A readable storage medium, having stored thereon a computer program, characterized in that: The computer program, when executed by the processor, implements the steps of the blood vessel three-dimensional reconstruction method based on multi-view contrast images according to any one of claims 1-5.

Images