Automatic measurement method for abdominal aorta parameters

Abstract

The application discloses an automatic abdominal aorta parameter measuring method, and belongs to the technical field of automatic abdominal aorta parameter measuring. The method comprises the following steps: acquiring an abdominal aorta aneurysm voxel model, and pre-processing the voxel model; generating a center line of the voxel model based on a thinning algorithm; generating a pixel-level node list of the blood vessel through a region growing algorithm, storing the pixel-level node by using a tree structure to obtain a blood vessel tree pixel-level node structure; recursively accessing the blood vessel tree pixel-level node structure to obtain a blood vessel segment, and performing down-sampling on the obtained blood vessel segment to obtain a blood vessel down-sampling structure; recursively accessing the blood vessel down-sampling structure to obtain a blood vessel segment, and performing interpolation up-sampling on the obtained blood vessel segment to generate a blood vessel up-sampling structure; establishing a segmented list of the center line through the blood vessel up-sampling structure, and constructing a spatial cross section corresponding to each segment; and calculating geometric morphology parameters of the abdominal aorta according to the center line and the spatial cross section.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of automatic measurement of abdominal aortic parameters, and particularly relates to an automatic measurement method of abdominal aortic parameters. BACKGROUND

[0002] Abdominal aortic aneurysm (AAA) is a kind of aneurysmal dilatation disease of the aorta, which is common in the elderly. The mortality rate after the rupture of abdominal aortic aneurysm is more than 70%, so the risk assessment of the aneurysm is very important for the detection and prognosis evaluation of abdominal aortic aneurysm, especially the establishment of an automatic abdominal aortic segmentation system for automatic risk assessment of the aneurysm.

[0003] At present, the shape change of the abdominal aorta is usually detected by using a manual segmentation modeling method in the clinic, and the main monitoring index is the aneurysm diameter of the abdominal aortic aneurysm, and the volume parameter is used as an evaluation index. However, the manual modeling method consumes a long time and cannot quickly and accurately obtain a high-quality model that can be used for blood flow dynamics calculation. At the same time, the difference in the model extraction by different doctors leads to the difficulty in accurately reflecting the prognosis by using a single evaluation index, and the abdominal aortic aneurysm has no special clinical symptoms, and once it ruptures, it has a great risk, which seriously threatens the life and health of the patient. In the existing clinical recommendations, whether the maximum diameter of the abdominal aortic aneurysm is more than 5.5 cm and whether the diameter expansion rate is more than 1 cm / year are used as risk thresholds. However, studies have shown that the above criteria cannot accurately determine the risk of abdominal aortic aneurysm rupture and the growth of the aneurysm. Some scholars have used blood flow dynamics parameters and geometric morphological parameters for the measurement and evaluation of abdominal aortic aneurysm. Among them, the morphological analysis of the abdominal aorta mainly includes length dimension parameter analysis based on the center line, area dimension parameter analysis based on the space section, and volume dimension parameter analysis based on the space region division. However, the generation of the center line and the calculation of the space section in the limited three-dimensional closed geometry are easily affected by factors such as complex modeling boundary and local curvature. It can be seen that the measurement and analysis of the morphological parameters of the abdominal aortic aneurysm are still extremely complex and difficult work.

[0004] Therefore, a fast and reliable automatic segmentation and calculation evaluation system of abdominal aortic aneurysm is urgently needed in the clinic. SUMMARY

[0005] The purpose of the present application is to overcome one or more of the deficiencies of the prior art and provide an automatic measurement method of abdominal aortic parameters.

[0006] The purpose of the present application is achieved by the following technical solutions:

[0007] An automatic measurement method of abdominal aortic parameters comprises:

[0008] Obtaining an abdominal aortic aneurysm voxel model, and preprocessing the voxel model;

[0009] Generating a centerline of the voxel model based on a thinning algorithm;

[0010] Selecting a seed node from the centerline, and generating a pixel-level node list of the blood vessel based on the seed node and a preset adjacency determination criterion through a region growing algorithm, and storing the pixel-level node using a tree structure to obtain a blood vessel tree pixel-level node structure;

[0011] Recursively accessing the blood vessel tree pixel-level node structure to obtain a blood vessel segment, and performing down-sampling on the obtained blood vessel segment to obtain a blood vessel down-sampling structure;

[0012] Recursively accessing the blood vessel down-sampling structure to obtain a blood vessel segment, and performing interpolation up-sampling on the obtained blood vessel segment to generate a blood vessel up-sampling structure;

[0013] Taking each interpolation point formed during the interpolation up-sampling as a segmentation point, establishing a segmented list of the centerline of the blood vessel up-sampling structure, and constructing a spatial cross section corresponding to each segment, the spatial cross section being a cross section obtained by intersecting a tangent plane of the segment with the voxel model, and a normal vector of the tangent plane being a tangent vector of an interpolation point on the segment;

[0014] Calculating a geometric morphology parameter of the abdominal aorta according to the centerline and the spatial cross section.

[0015] Preferably, the method further comprises:

[0016] Establishing a blood vessel tree logical structure after recursively accessing the nodes of the blood vessel up-sampling structure;

[0017] Obtaining depths of the nodes of the blood vessel tree after recursively scanning the blood vessel tree logical structure;

[0018] Performing branch naming according to the blood vessel tree logical structure and the depths of the nodes.

[0019] Preferably, after the segmented list of the centerline of the blood vessel up-sampling structure is established and the spatial cross section corresponding to each segment is constructed, the method further comprises:

[0020] Initializing a search vector direction of the corresponding segment using a normal vector of the spatial cross section, and taking finding a minimum spatial cross section corresponding to the segment as a search target, and performing a spatial quadtree search for optimizing the spatial cross section;

[0021] The search vector direction reserved after the spatial quadtree search converges is recorded as an optimized vector direction, and the optimized vector direction is used as a normal vector to construct a cross plane intersecting the voxel model, and the cross plane is recorded as an optimized spatial cross section.

[0022] Preferably, the method further comprises: visualizing and outputting the calculated geometric morphology parameters in a report form.

[0023] Preferably, the method further comprises: obtaining an abdominal aortic aneurysm voxel model, and preprocessing the voxel model, specifically:

[0024] reading in the binarized abdominal aortic aneurysm voxel model;

[0025] performing connectivity determination on the binarized image, and intercepting a binarized image of a largest connected region in volume as a target binarized image;

[0026] performing edge smoothing on the target binarized image by performing image closing and image opening operations;

[0027] generating a centerline of the voxel model based on a thinning algorithm, specifically:

[0028] extracting a centerline of the target binarized image after edge smoothing by a thinning algorithm pre-installed in a medical image processing software package to obtain the centerline of the voxel model.

[0029] Preferably, the method further comprises: selecting a seed node from the centerline, and generating a pixel-level node list of the blood vessel based on the seed node and a preset adjacency determination criterion, storing each pixel-level node in the pixel-level node list by using a tree structure to obtain a blood vessel tree pixel-level node structure, specifically:

[0030] scanning the centerline from top to bottom, recording a first non-zero coordinate, and taking a node corresponding to the first non-zero coordinate as a blood vessel tree origin, and taking the blood vessel tree origin as the seed node;

[0031] performing adjacency determination according to the preset adjacency determination criterion;

[0032] establishing a queue structure Q, the blood vessel tree origin being an element of the queue structure Q by being added to the queue structure Q, performing adjacency determination on each element in the queue structure Q, adding adjacency nodes of the element determined to the queue structure Q, removing the element on which adjacency determination has been performed from the queue structure Q and adding the element to a blood vessel pixel-level node list, and cyclically analyzing the queue structure Q until the length of the queue structure Q is 0;

[0033] storing each pixel-level node in the blood vessel pixel-level node list by using a tree structure to obtain a blood vessel tree pixel-level node structure,

[0034] the original node of the blood vessel tree pixel-level node structure corresponding to the blood vessel tree origin;

[0035] The preset adjacency determination criterion is that two pixels are adjacent if there is a same binarized pixel in a 26-neighborhood of the pixel.

[0036] Preferably, the recursive access to the pixel-level node structure of the blood vessel tree is used to obtain blood vessel segments, and the obtained blood vessel segments are then processed...

[0037] After downsampling, the blood vessel downsampling structure is obtained, specifically as follows:

[0038] Starting from the original node of the pixel-level node structure of the vascular tree, we recursively visit the nodes and analyze the number of child nodes of each node in the vascular tree. If the number of child nodes of a node is greater than 1, then the node is recorded as a branch node and the path between two adjacent branch nodes is recorded as a vascular segment.

[0039] Skip sampling is performed in each blood vessel segment;

[0040] The skip sampling process is as follows: Let the number of skip intervals be n. Starting from the front end of the blood vessel segment, points are taken at intervals of n-1, and the process ends at the rear end of the blood vessel segment. The front end, the rear end, and all the points taken constitute the blood vessel downsampling structure.

[0041] Preferably, the recursive access to the blood vessel downsampling structure to obtain blood vessel segments, and the interpolation upsampling of the obtained blood vessel segments to generate the blood vessel upsampling structure, specifically involves:

[0042] The blood vessel sampling structure is recursively accessed, and the number of child nodes of each node in the blood vessel tree is analyzed. If the number of child nodes of a node is greater than 1, the node is recorded as a branch node, and the path between two adjacent branch nodes is recorded as a blood vessel segment.

[0043] For each vascular segment, spline curve interpolation is performed, and the interpolated vascular segments constitute the vascular sampling structure.

[0044] Preferably, the step of establishing a segmented list of the centerline of the blood vessel upsampling structure using each interpolation point formed during interpolation upsampling as a segmentation point, and constructing a spatial cross-section corresponding to each segment, wherein the spatial cross-section is the cross-section obtained by intersecting the cross-section of the segment with the voxel model, and the normal vector of the cross-section is the tangent vector of the interpolation point on the segment, specifically:

[0045] Using the interpolation points formed during interpolation upsampling as segmentation points, a segmentation list of the center line of the blood vessel upsampling structure is established.

[0046] Let the interpolation point on each segment be the sampling point and denoted as P. j Take the interpolation points before and after the sampling point and denote them as P respectively. j-1 With P j+1 , let P j+1 Point to P j-1 vector V j For P j The original tangent vector at the point is used as the surface normal vector to construct the tangent plane f. j Let f be the cross section. jThe cross section obtained by intersecting with the voxel model is P. j The spatial cross section at the point;

[0047] The search vector direction of the corresponding segment is initialized using the normal vector of the spatial section, and the search objective is to find the minimum spatial section corresponding to that segment. A spatial quadtree search is then performed to optimize the spatial section, specifically as follows:

[0048] The original vector V j The length is standardized to 1, with V j Construct an end plane for the normal vector, with V on the end plane. j Draw an initial circle with a radius of 0.5 centered at the endpoint of the circle. Divide the circle into four equal parts, and denote the four division points as E1 to E4. Let P be the endpoint of the circle. j The spatial vector to E1 is the update vector, and the area S1 of the intersection plane obtained by the plane constructed with the update vector as the normal vector and the voxel model is calculated. Similarly, the areas S2 to S4 are obtained by corresponding E2 to E4 one by one.

[0049] Compare the P j The area S0 of the spatial cross section at the point, and the areas S1 to S4, if S0 is the minimum value, then the current round converges, and the circle radius is updated to half of the initial circle radius for the next round of search. If S0 is not the minimum value, then the update vector corresponding to the minimum value among S1 to S4 in the current round is taken as the original vector for the current round of search, and the calculation is repeated until the current round converges. The search is performed in a total of three rounds.

[0050] The search vector direction retained after the spatial quadtree search converges is denoted as the optimized vector direction, specifically:

[0051] The original vector direction that is finally retained after convergence is denoted as the optimized vector direction.

[0052] Preferably, the step of establishing a blood vessel tree logical structure after performing node recursion on the blood vessel sampling structure specifically involves:

[0053] Recursively traversing from the original node of the blood vessel sampling structure, extracting branch nodes separately and preserving the connection relationship between branch nodes, while recording the path length between blood vessel branch nodes, and establishing a blood vessel tree logical structure based on branch nodes, the connection relationship between branch nodes and the path length, wherein the branch node is a node with more than 1 child node;

[0054] The formula for calculating the path length is:

[0055] ;

[0056] in, L Let n be the path length, and n be the number of nodes in the path between blood vessel branch nodes. x coordinate of a branch node in the path , x coordinate of a branch node adjacent to the branch node in the path , y coordinate of the branch node , y coordinate of the branch node , z coordinate of the branch node , z coordinate of the branch node ;

[0057] the depth of each node on the blood vessel tree logical structure obtained after the recursive scanning of the blood vessel tree logical structure, specifically:

[0058] recursion is performed on the blood vessel tree logical structure, a depth value is introduced for each branch node, and the depth value of the branch node is updated layer by layer through the outermost node upward recursion, and the maximum depth value of the branch node is defined as the branch node depth of the node;

[0059] branch naming is performed according to the blood vessel tree logical structure and the depth of each node, specifically:

[0060] the branch nodes on the blood vessel tree logical structure are traversed, branch nodes with a number of sub-branches greater than two are selected for analysis of whether a preset condition is met, two child nodes on the branch node are taken, and the sum of the branch node depth of one of the child nodes itself and the center line length from the child node to the branch node is recorded as D1, and the sum of the branch node depth of the other child node itself and the center line length from the child node to the branch node is recorded as D2, the preset condition includes f1 condition, f2 condition and f3 condition;

[0061] firstly, the branch node meeting the f1 condition is recorded as an abdominal aorta analysis starting point, then abdominal aorta main branch node searching is performed, the branch node meeting the f2 condition is recorded as an abdominal aorta main branch node, and finally abdominal aorta secondary branch node searching is performed, and the branch node meeting the f3 condition is recorded as an abdominal aorta secondary branch node;

[0062] wherein, the f1 condition is: ;

[0063] the f2 condition is: ;

[0064] the f3 condition is: .

[0065] The present application has the beneficial effects of:

[0066] (1) The up-sampling process and the down-sampling process of the center line are utilized to optimize the curvature distribution while the spatial morphology of the center line is reserved, and thus the abdominal aorta parameter automatic measurement method realized by the application has the characteristics of smooth and continuous center line, uniform and non-mutated parameter distribution, and no need for manual intervention, and the measurement of the abdominal aorta parameter geometric morphology parameter is realized accurately, reliably and efficiently compared with the traditional parameter measurement method;

[0067] (2) The application proposes an optimized definition of the spatial cross section, i.e., searching for the minimum cross section passing through a fixed point in a closed figure, which avoids the leakage of the spatial cross section and the interference of the bifurcation system in the blood vessel on the parameter calculation, and improves the measurement accuracy of various dimensional parameters based on the spatial cross section; further, a spatial quadtree search method is proposed to search and calculate the minimum cross section, which realizes the optimized analysis of the spatial cross section under the time complexity of log(n) and improves the efficiency of the parameter measurement;

[0068] (3) The center line is described by establishing the logical structure of the blood vessel tree, and automatic branch naming is realized without manual intervention;

[0069] (4) The parameters are output in the form of a report for visual output, which is convenient for systematic export and reading of the parameters. BRIEF DESCRIPTION OF DRAWINGS

[0070] Figure 1 Fig. 1 is a flowchart of the abdominal aorta parameter automatic measurement method;

[0071] Figure 2 Fig. 3 is a schematic diagram of the center line optimization effect;

[0072] Figure 3 Fig. 5 is a schematic diagram of the leakage of the spatial cross section;

[0073] Figure 4 Fig. 7 is a schematic diagram of the optimized spatial cross section;

[0074] Figure 5 Fig. 9 is a comparison diagram of the perimeter measurement before and after the optimization of the spatial cross section;

[0075] Figure 6 Fig. 11 is a schematic diagram of the logical structure of the blood vessel tree and the branch naming. DETAILED DESCRIPTION

[0076] The technical solutions of the application will be described clearly and completely in combination with the embodiments, obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments. Based on the embodiments in the application, all other embodiments obtained by the person skilled in the art without creative labor belong to the protection scope of the application.

[0077] Reference Figures 1-6The embodiment provides an abdominal aorta parameter automatic measurement method, which is used for measuring geometric morphological parameters of an abdominal aorta in a clinic.

[0078] An abdominal aorta parameter automatic measurement method, specifically comprising the following steps:

[0079] S100. Obtain an abdominal aorta aneurysm voxel model, and pre-process the voxel model.

[0080] S200. Generate a center line of the voxel model based on a thinning method.

[0081] S300. Select a seed node from the center line, and generate a pixel-level node list of a blood vessel based on the seed node and a preset adjacency criterion through a region growing algorithm, and store the pixel-level node in a tree structure to obtain a blood vessel tree pixel-level node structure.

[0082] S400. Recursively access the blood vessel tree pixel-level node structure to obtain a blood vessel segment, and obtain a blood vessel down-sampling structure after down-sampling the obtained blood vessel segment.

[0083] S500. Recursively access the blood vessel down-sampling structure to obtain a blood vessel segment, and generate a blood vessel up-sampling structure after up-sampling the obtained blood vessel segment through interpolation.

[0084] S600. Take each interpolation point formed in the interpolation up-sampling as a segmentation point, establish a segmented list of a center line of the blood vessel up-sampling structure, and construct a spatial cross section corresponding to each segment, wherein the spatial cross section is a cross section obtained by intersecting a tangent plane of the segment and the voxel model, and a normal vector of the tangent plane is a tangent vector of an interpolation point on the segment.

[0085] S700. Calculate geometric morphological parameters of the abdominal aorta according to the center line and the spatial cross section. The geometric morphological parameters include one or more of a length dimensional parameter based on the center line, an area dimensional parameter based on the spatial cross section, and a volume dimensional parameter.

[0086] A specific implementation process of step S100 is as follows:

[0087] S101. Read in a binary abdominal aorta aneurysm voxel model.

[0088] S102. Perform connectivity determination on the binary image, and intercept a binary image of a maximum volume connected region as a target binary image. In some embodiments, the connectivity determination on the binary image is performed by using a medical image processing software package ITK to eliminate the influence of artifacts.

[0089] S103. Edge smoothing is performed on the target binary image by performing image closing operation and image opening operation. In some embodiments, the image closing operation and the image opening operation are performed twice in succession.

[0090] wherein the image closing operation refers to performing image dilation first and then performing image erosion. The image opening operation refers to performing image erosion first and then performing image dilation. The image erosion refers to: wherein A is the image to be eroded, B is the erosion operator, and Z is the erosion result. The image dilation refers to: wherein A is the image to be dilated, B is the dilation operator, and Z is the dilation result.

[0091] A specific implementation process of step S200 is as follows:

[0092] S201. The centerline of the target binary image after edge smoothing is extracted by using a thinning method preset in a medical image processing software package ITK to obtain the centerline of the voxel model.

[0093] A specific implementation process of step S300 is as follows:

[0094] S301. The binary image of the centerline is scanned from top to bottom (z direction), the first non-zero coordinate is recorded, and the node corresponding to the first non-zero coordinate is taken as the original point of the blood vessel tree. The original point of the blood vessel tree is taken as the seed node.

[0095] S302. Adjacency determination is performed according to a preset adjacency determination criterion.

[0096] S303. A queue structure Q is established, the original point of the blood vessel tree is taken as an element of the queue structure Q by being added to the queue structure Q, adjacency determination is performed on each element in the queue structure Q, the adjacent nodes of the element obtained by the adjacency determination are also added to the queue structure Q, the element on which the adjacency determination has been performed is removed from the queue structure Q and added to the blood vessel pixel-level node list, and the queue structure Q is analyzed in a loop until the length of the queue structure Q is 0.

[0097] S304. Each pixel-level node in the blood vessel pixel-level node list is stored by using a tree structure to obtain a blood vessel tree pixel-level node structure, and the original node of the blood vessel tree pixel-level node structure corresponds to the original point of the blood vessel tree.

[0098] wherein the preset adjacency determination criterion is that two pixels are adjacent if there is a same binary pixel in the neighborhood of the pixel 26.

[0099] A specific implementation process of step S400 is as follows:

[0100] S401. Recursively visit from the original node of the pixel-level node structure of the vessel tree, analyze the number of child nodes of each node on the vessel tree, if the number of child nodes of the node is greater than 1, the node is recorded as a branch node, and the path between two adjacent branch nodes is recorded as a vessel segment;

[0101] S402. Jump sampling is performed in each vessel segment. In the jump sampling process, the number of jump intervals is n, starting from the front end point of the vessel segment, taking points at intervals of n-1 to obtain various taking points, and ending at the rear end point of the vessel segment, the front end point, the rear end point and all the taking points constitute the down-sampling structure of the vessel.

[0102] A specific implementation process of step S500 is as follows:

[0103] S501. Recursively visit the down-sampling structure of the vessel, analyze the number of child nodes of each node on the vessel tree, if the number of child nodes of the node is greater than 1, the node is recorded as a branch node, and the path between two adjacent branch nodes is recorded as a vessel segment.

[0104] S502. Spline curve interpolation is performed on each vessel segment, and the vessel up-sampling structure is constituted by all the interpolated vessel segments. In some embodiments, the b-spline curve interpolation is performed by using the preset vtkParametricSpline method class in the medical imaging library VTK, and the number of interpolation points formed by the b-spline curve interpolation is twice the number of original taking points on the corresponding vessel segment.

[0105] Figure 2 The difference between the center line of the voxel model obtained based on the direct extraction method and the center line of the voxel model obtained based on the above down-sampling and up-sampling method is shown. It can be seen that the center line obtained by the down-sampling and up-sampling process is continuous and smooth, and there is no obvious mutation.

[0106] As an improvement, step S600 further includes the following content:

[0107] The normal vector of the spatial section is used to initialize the search vector direction of the corresponding segment, and the spatial quadtree search for optimizing the spatial section is performed by taking finding the minimum spatial section corresponding to the segment as the search target;

[0108] The search vector direction reserved after the convergence of the spatial quadtree search is recorded as the optimized vector direction, and the optimized vector direction is used as the normal vector to construct a section plane intersecting the voxel model, and the section plane is recorded as the optimized spatial section.

[0109] A specific implementation process of step S600 is as follows:

[0110] S601. The interpolation points formed during the interpolation up-sampling are used as the segmentation points to establish a segment list of the center line of the vessel up-sampling structure.

[0111] S602. Take the interpolation point located on each segment as the sampling point and denote it as P. j Take the interpolation points before and after the sampling point and denote them as P respectively. j-1 With P j+1 , let P j+1 Point to P j-1 vector V j For P j The original tangent vector at the point is used as the surface normal vector to construct the tangent plane f. j Let f be the cross section. j The cross section obtained by intersecting with the voxel model is P. j The spatial cross-section at a point, and the spatial cross-sections of each segment, form a sequence of spatial cross-sections along the centerline.

[0112] S603. Transfer the original vector V j The length is standardized to 1, with V j Construct an end plane for the normal vector, with V on the end plane. j Draw an initial circle with a radius of 0.5 centered at the endpoint of the circle. Divide the circle into four equal parts, and denote the four division points as E1, E2, E3, and E4. 2、 E 3、 E4, let P j The spatial vector to E1 is used as the update vector, and the area S1 of the intersection plane obtained by taking the plane constructed with the update vector as the normal vector and the voxel model is calculated. Similarly, areas S2 to S4 are obtained by corresponding to E2 to E4. Specifically, let P... j The spatial vector to E2 is used as the update vector, and the area S of the intersection plane obtained by constructing the intersection plane with the voxel model using the update vector as the normal vector is calculated. j , note P j The spatial vector to E3 is the update vector, and the area S3 of the intersection plane obtained by constructing the cross-plane with the update vector as the normal vector and the voxel model is calculated. Let P be the area of ​​the cross-plane. j The spatial vector to E4 is the update vector, and the area S4 of the cross-section obtained by intersecting the cross-section constructed with the update vector as the normal vector and the voxel model is calculated.

[0113] S604. Compare P j The area S0, area S1, and area S of the spatial cross section at point S. 2、 Area S 3、 The area is S4; if S0 is the minimum value, the current round converges, and the circle radius is updated to half of the initial circle radius for the next round of search; if S0 is not the minimum value, the update vector corresponding to the minimum value among S1 to S4 in this round is taken as the original vector for this round of search, and the calculation is repeated until the current round converges. The search is performed in a total of three rounds;

[0114] S605. The original vector direction that is finally retained after convergence is denoted as the optimized vector direction. The optimized vector direction is used as the normal vector to construct the intersection plane with the voxel model. The intersection plane is denoted as the optimized spatial section.

[0115] Figure 3 This illustrates a potential leakage scenario in the spatial cross-section before optimization, and the result after optimization. Figure 4 The spatial cross-section shown. Leakage in the spatial cross-section can cause errors in the measurement of perimeter parameters, such as... Figure 5 As shown, spatial cross-section leakage leads to an incorrect estimate of the peak perimeter. After spatial cross-section optimization, the estimate of the peak perimeter is made closer to the true value.

[0116] As an improvement, the automatic measurement method for abdominal aortic parameters implemented in this embodiment further includes the following steps:

[0117] S800. After performing node recursion on the blood vessel sampling structure, a blood vessel tree logical structure is established.

[0118] S900. After recursively scanning the logical structure of the vascular tree, the depth of each node in the vascular tree is obtained.

[0119] S1000. Name the branches according to the logical structure of the vascular tree and the depth of each node.

[0120] One specific implementation of step S800 is as follows:

[0121] S801. Starting from the original node of the blood vessel sampling structure, perform recursive visits, extract branch nodes separately and retain the connection relationship between branch nodes, and record the path length between blood vessel branch nodes. Based on the branch nodes, the connection relationship between branch nodes and the path length, establish the logical structure of the blood vessel tree.

[0122] The formula for calculating the path length is:

[0123] ;

[0124] In the formula, L Let n be the path length, and n be the number of nodes in the path between blood vessel branch nodes. A branch node in the path x-coordinate, For the branch nodes in the path Adjacent branch nodes x-coordinate, branch node y-coordinate, branch node y-coordinate, branch node z-coordinate, branch node The z-coordinate. For example, there are 3 nodes in the path between blood vessel branch nodes p0 and p4, denoted as p1, p2 and p3 respectively. Calculate the distance between p0 and p1, the distance between p1 and p2, the distance between p2 and p3 and the distance between p3 and p4 respectively. Summing up all the distances gives the path length between blood vessel branch nodes p0 and p4.

[0125] One specific implementation of step S900 is as follows:

[0126] S901. Recursively process the blood vessel tree logical structure, introduce a depth value for each branch node, and recursively update the depth value of the branch node by layer through the outermost node. Define the maximum depth value of the branch node as the branch node depth of that node.

[0127] One specific implementation of step S1000 is as follows:

[0128] S1001. Traverse the branch nodes in the blood vessel tree logical structure, select the branch nodes with more than two sub-branches and analyze whether they meet the preset conditions. Take the two child nodes of the branch node, and record the sum of the branch node depth of one child node and the length of the center line from the child node to the branch node as D1. Record the sum of the branch node depth of the other child node and the length of the center line from the child node to the branch node as D2. The preset conditions include condition f1, condition f2 and condition f3.

[0129] S1002. First, the branch nodes that satisfy condition f1 are recorded as the starting point of the abdominal aorta analysis. Then, the main branch nodes of the abdominal aorta are searched, and the branch nodes that satisfy condition f2 are recorded as the main branch nodes of the abdominal aorta. Finally, the secondary branch nodes of the abdominal aorta are searched, and the branch nodes that satisfy condition f3 are recorded as the secondary branch nodes of the abdominal aorta.

[0130] The condition for f1 is: ;

[0131] The condition for f2 is: ;

[0132] The condition for f3 is: .

[0133] Figure 6 This embodiment illustrates a blood vessel tree logic structure and branch naming.

[0134] As an improvement, the automatic measurement method for abdominal aortic parameters implemented in this embodiment further includes the following steps:

[0135] S1100. The calculated geometric morphology parameters are output in the form of a report. Specifically, a report output model is established through a pyQT5 visualization program interface, and then a visualization interface provided by a medical imaging library VTK is used to preview and display the model parameters. The naming and definition of the parameters in the report implemented in this embodiment are shown in Tables 1-9. In addition to length-dimension parameters, area-dimension parameters, and volume-dimension parameters, the parameter types also include angle parameters and combined parameters. It is known that different report output models can be established to achieve different parameter report display effects. In addition, the types of parameters calculated are not limited to those shown in Tables 1-9.

[0136] Table 1

[0137]

[0138] Table 2

[0139]

[0140] Table 3

[0141]

[0142] Table 4

[0143]

[0144] Table 5

[0145]

[0146] Table 6

[0147]

[0148] Table 7

[0149]

[0150] Table 8

[0151]

[0152] Table 9

[0153]

[0154] The foregoing is considered as illustrative only of the principles of the application. Further, since numerous modifications and changes will readily occur to those skilled in the art, it is not desired to limit the application to the exact construction and operation described. Accordingly, all such variations are intended to be included within the scope of the present application as defined in the claims below and their equivalents.

Claims

1. An automatic measurement method for abdominal aortic parameters, characterized in that, include: Obtain a voxel model of an abdominal aortic aneurysm and preprocess the voxel model; Centerlines of voxel models are generated based on a refinement algorithm; Seed nodes are selected from the centerline, and a pixel-level node list of blood vessels is generated by the region growing algorithm based on the seed nodes and the preset adjacency determination criteria. The pixel-level nodes are stored in a tree structure to obtain the pixel-level node structure of the blood vessel tree. The blood vessel segments are obtained by recursively accessing the pixel-level node structure of the blood vessel tree. The obtained blood vessel segments are then downsampled to obtain the blood vessel downsampled structure. The blood vessel downsampling structure is recursively accessed to obtain blood vessel segments, and the obtained blood vessel segments are then upsampled by interpolation to generate the blood vessel upsampling structure. Using each interpolation point formed during interpolation upsampling as a segmentation point, a segmented list of the center line of the blood vessel upsampling structure is established, and a spatial cross section corresponding to each segment is constructed. The spatial cross section is the cross section obtained by intersecting the cross section of the segment with the voxel model, and the normal vector of the cross section is the tangent vector of the interpolation point on the segment. The process of establishing a segmented list of the centerline using a blood vessel sampling structure and constructing a spatial cross-section corresponding to each segment further includes: The search vector direction of the corresponding segment is initialized using the normal vector of the spatial section, and a spatial quadtree search is performed to optimize the spatial section with the search objective of finding the minimum spatial section corresponding to the segment. The search vector direction retained after the spatial quadtree search converges is denoted as the optimized vector direction. The optimized vector direction is used as the normal vector to construct the intersection plane with the voxel model. This intersection plane is denoted as the optimized spatial section. The geometric morphological parameters of the abdominal aorta were calculated based on the centerline and the optimized spatial cross-section.

2. The automatic measurement method for abdominal aortic parameters according to claim 1, characterized in that, The method further includes: After performing node recursion on the blood vessel sampling structure, a logical structure of the blood vessel tree is established. The depth of each node in the vascular tree is obtained by recursively scanning the logical structure of the vascular tree. Branches are named according to the logical structure of the vascular tree and the depth of each node.

3. The automatic measurement method for abdominal aortic parameters according to claim 1, characterized in that, The method further includes: visualizing the calculated geometric morphological parameters in the form of a report.

4. The automatic measurement method for abdominal aortic parameters according to claim 1, characterized in that, The process of obtaining a voxel model of an abdominal aortic aneurysm and preprocessing the voxel model specifically involves: Read in the binarized abdominal aortic aneurysm protoxin model; Perform connectivity determination on the binarized image and extract the binarized image of the largest connected region as the target binarized image; Edge smoothing is performed on the target binarized image by performing image closing and image opening operations; The centerline of the voxel model generated based on the refinement algorithm is specifically as follows: The centerline of the target binarized image after edge smoothing is extracted by the thinning algorithm in the medical image processing software package, and the centerline of the voxel model is obtained.

5. The automatic measurement method for abdominal aortic parameters according to claim 1, characterized in that, The process involves selecting a seed node from the centerline, and based on the seed node and a preset adjacency criterion, generating a pixel-level node list for the blood vessel using a region growing algorithm. The pixel-level nodes are then stored in a tree structure to obtain the blood vessel tree pixel-level node structure. Specifically: Scan the center line from top to bottom, record the first non-zero coordinate, and take the node corresponding to the first non-zero coordinate as the origin of the blood vessel tree, and take the origin of the blood vessel tree as the seed node. Adjacency determination is performed according to the preset adjacency determination criteria; Establish a queue structure Q. The origin of the blood vessel tree becomes an element of the queue structure Q by adding it to the queue structure Q. Perform adjacency determination on each element in the queue structure Q, and add the adjacent nodes of the determined element to the queue structure Q. Remove the elements that have undergone adjacency determination from the queue structure Q and add them to the blood vessel pixel-level node list. Analyze the queue structure Q in a loop until the length of the queue structure Q is 0. By using a tree structure to store each pixel-level node in the list of pixel-level nodes for blood vessels, we obtain the pixel-level node structure of the blood vessel tree. The original node of the pixel-level node structure of the vascular tree corresponds to the origin of the vascular tree. The preset adjacency determination criterion is: if there are identical binary pixels in the 26-neighborhood of pixel 2, then the two pixels are adjacent.

6. The automatic measurement method for abdominal aortic parameters according to claim 1, characterized in that, The process of recursively accessing the pixel-level node structure of the blood vessel tree to obtain blood vessel segments, and then downsampling the obtained blood vessel segments to obtain the blood vessel downsampling structure, is as follows: Starting from the original node of the pixel-level node structure of the vascular tree, we recursively visit the nodes and analyze the number of child nodes of each node in the vascular tree. If the number of child nodes of a node is greater than 1, then the node is recorded as a branch node and the path between two adjacent branch nodes is recorded as a vascular segment. Skip sampling is performed in each blood vessel segment; The skip sampling process is as follows: Let the number of skip intervals be n. Starting from the front end of the blood vessel segment, points are taken at intervals of n-1, and the process ends at the rear end of the blood vessel segment. The front end, the rear end, and all the points taken constitute the blood vessel downsampling structure.

7. The automatic measurement method for abdominal aortic parameters according to claim 1, characterized in that, The process of recursively accessing the blood vessel downsampling structure to obtain blood vessel segments, and then performing interpolation upsampling on the obtained blood vessel segments to generate the blood vessel upsampling structure, specifically involves: The blood vessel sampling structure is recursively accessed, and the number of child nodes of each node in the blood vessel tree is analyzed. If the number of child nodes of a node is greater than 1, the node is recorded as a branch node, and the path between two adjacent branch nodes is recorded as a blood vessel segment. For each vascular segment, spline curve interpolation is performed, and the interpolated vascular segments constitute the vascular sampling structure.

8. The automatic measurement method for abdominal aortic parameters according to claim 1, characterized in that, The method involves using each interpolation point formed during upsampling as a segmentation point to establish a segmented list of the centerline of the blood vessel upsampling structure, and constructing a spatial cross-section corresponding to each segment. The spatial cross-section is the cross-section obtained by intersecting the cross-section of that segment with the voxel model, and the normal vector of the cross-section is the tangent vector of the interpolation point on that segment. Specifically: Using the interpolation points formed during interpolation upsampling as segmentation points, a segmentation list of the center line of the blood vessel upsampling structure is established. Let the interpolation point on each segment be the sampling point and denoted as P. j Take the interpolation points before and after the sampling point and denote them as P respectively. j-1 With P j+1 , let P j+1 Point to P j-1 vector V j For P j The original tangent vector at the point is used as the surface normal vector to construct the tangent plane f. j Let f be the cross section. j The cross section obtained by intersecting with the voxel model is P. j The spatial cross section at the point; The search vector direction of the corresponding segment is initialized using the normal vector of the spatial section, and the search objective is to find the minimum spatial section corresponding to that segment. A spatial quadtree search is then performed to optimize the spatial section, specifically as follows: The original vector V j The length is standardized to 1, with V j Construct an end plane for the normal vector, with V on the end plane. j Draw an initial circle with a radius of 0.5 centered at the endpoint of the circle. Divide the circle into four equal parts, and denote the four division points as E1 to E4. Let P be the endpoint of the circle. j The spatial vector to E1 is the update vector, and the area S1 of the intersection plane obtained by the plane constructed with the update vector as the normal vector and the voxel model is calculated. Similarly, the areas S2 to S4 are obtained by corresponding E2 to E4 one by one. Compare the P j The area S0 of the spatial cross section at the point, and the areas S1 to S4, if S0 is the minimum value, then the current round converges, and the circle radius is updated to half of the initial circle radius for the next round of search. If S0 is not the minimum value, then the update vector corresponding to the minimum value among S1 to S4 in the current round is taken as the original vector for the current round of search, and the calculation is repeated until the current round converges. The search is performed in a total of three rounds. The search vector direction retained after the spatial quadtree search converges is denoted as the optimized vector direction, specifically: The original vector direction that is finally retained after convergence is denoted as the optimized vector direction.

9. The automatic measurement method for abdominal aortic parameters according to claim 2, characterized in that, The process of establishing a blood vessel tree logical structure after recursively processing the nodes of the blood vessel sampling structure is as follows: Recursively traversing from the original node of the blood vessel sampling structure, extracting branch nodes separately and preserving the connection relationship between branch nodes, while recording the path length between blood vessel branch nodes, and establishing a blood vessel tree logical structure based on branch nodes, the connection relationship between branch nodes and the path length, wherein the branch node is a node with more than 1 child node; The formula for calculating the path length is: ； in, L Let n be the path length, and n be the number of nodes in the path between blood vessel branch nodes. A branch node in the path x-coordinate, For the branch nodes in the path Adjacent branch nodes x-coordinate, branch node y-coordinate, branch node y-coordinate, branch node z-coordinate, branch node The z-coordinate; The recursive scanning of the vascular tree logical structure yields the depth of each node in the vascular tree, specifically as follows: The logical structure of the blood vessel tree is recursively processed, a depth value is introduced for each branch node, and the depth value of the branch node is updated layer by layer by recursively upward through the outermost node. The maximum depth value of the branch node is defined as the branch node depth of that node. Branches are named according to the logical structure of the blood vessel tree and the depth of each node, specifically as follows: Traverse the branch nodes in the blood vessel tree logical structure, select the branch nodes with more than two sub-branches and analyze whether they meet the preset conditions. Take the two child nodes of the branch node, and record the sum of the branch node depth of one child node and the length of the center line from the child node to the branch node as D1. Record the sum of the branch node depth of the other child node and the length of the center line from the child node to the branch node as D2. The preset conditions include condition f1, condition f2 and condition f3. First, the branch nodes that satisfy condition f1 are recorded as the starting point of the abdominal aorta analysis. Then, the main branch nodes of the abdominal aorta are searched, and the branch nodes that satisfy condition f2 are recorded as the main branch nodes of the abdominal aorta. Finally, the secondary branch nodes of the abdominal aorta are searched, and the branch nodes that satisfy condition f3 are recorded as the secondary branch nodes of the abdominal aorta. The condition for f1 is: ; The condition for f2 is: ; The condition for f3 is: .

Images