KR20190084380A - Method for extracting vascular structure in 2d x-ray angiogram, computer readable medium and apparatus for performing the method - Google Patents

Abstract

The present invention relates to a method for extracting a vascular structure from a two-dimensional (2D) X-ray angiogram, which extracts a 2D vascular tree structure from the 2D X-ray angiogram for a blood vessel of an examinee, performs rigid body matching between the 2D vascular tree structure and a 3D vascular structure divided from a 3D CT angiogram of the examinee, calculates similarity between 2D and 3D vascular segments, and uses the similarity to reconfigure the 2D vascular tree structure. Accordingly, loss of depth information of a blood vessel is removed, thereby increasing correctness when a vascular structure is extracted from a 2D X-ray angiogram.

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention [0001] The present invention relates to a method and apparatus for extracting a vein structure of a two-dimensional X-ray contrast image,

The present invention relates to a method for extracting a blood vessel structure of a two-dimensional X-ray image, a recording medium and an apparatus for performing the same, and more particularly, A method for extracting a vein structure of a two-dimensional X-ray contrast image extracting a vein structure from an X-ray image, and a recording medium and apparatus for performing the same.

Recently, percutaneous coronary intervention (PCI) has been performed with the rapid increase of coronary artery disease (CAD) including angina pectoris and myocardial infarction. Percutaneous coronary intervention (PCI) is a non-invasive, non-invasive method of expanding blood vessels using a long, elongated catheter called a catheter. This is based on imaging techniques to extract coronary arteries.

Imaging techniques include X-ray angiography (XA), cardiac CT angiography (CTA), MR angiography (MRA), cardiac PET and SPECT. Among them, X-ray angiography is a medical imaging technique that visually shows the inside of a blood vessel. It extracts a coronary artery and allows physicians to see the vessel stenosis in real time. Therefore, accurate computer-assisted coronary artery analysis in contrast imaging is very important for diagnosis and treatment of cardiologists and cardiologists.

However, images obtained by X-ray imaging have difficulties in extracting and analyzing blood vessels due to background clutter such as organs, bones, and interventional devices. In addition, since the X-ray contrast image is an image projected on a two-dimensional plane, the structure information of the three-dimensional shape is lost, and there is a problem that it is difficult to find a branch point of the blood vessel when the blood vessels overlap or intersect. Therefore, a technique of accurately extracting the structure of blood vessels in X-rayography has been proposed.

Among the conventional techniques for segmenting a blood vessel in an X-ray contrast image, a method of dividing a blood vessel in a contrast image has been proposed by combining Region Growing and Hessian Matrix Multiscale Filtering . This technique uses Hessian matrix multi-scale filtering to enhance blood vessels, and then extracts a tree of coronary artery blood vessels using a Multi Seed Region Growing algorithm. However, this technique has a disadvantage in that a plurality of seed points must be defined through user intervention, and it is very sensitive to noise because it is based on a Hessian matrix.

In addition, in order to infer the regional vascular structure by analyzing the essential features for segmentation of the X-ray image and the prior knowledge about the contrast image, a multi-feature based Fuzzy Recognition algorithm An automated approach for coronary artery segmentation has been proposed. However, the conventional technique according to the automated approach has a slow average calculation time, and can not accurately determine a corresponding two-dimensional projection structure when the three-dimensional blood vessel structures overlap or cross each other. Accordingly, there is a need for a technique for accurately extracting the structure of a blood vessel based on a topological model of a blood vessel in an X-ray image.

Korean Patent No. 10-1485900 Korean Patent No. 10-1768526

One aspect of the present invention is a method for extracting a vein structure of a two-dimensional X-ray contrast image that accurately extracts a vein structure of a two-dimensional X-ray contrast image based on a graph model using blood vessel information of a three-dimensional CT contrast image of a subject, And a recording medium and an apparatus therefor.

The technical problem of the present invention is not limited to the technical problems mentioned above, and other technical problems which are not mentioned can be understood by those skilled in the art from the following description.

A method of extracting a blood vessel structure of a two-dimensional X-ray contrast image according to an embodiment of the present invention includes extracting a two-dimensional blood vessel tree structure reflecting a weight from a two-dimensional X-ray contrast image of a blood vessel of a subject, Dimensional vessel structure of the two-dimensional vascular tree structure and the three-dimensional vascular structure of the three-dimensional vascular structure by performing rigid body matching of the three- Dimensional vessel segments to be matched to any one of the two-dimensional vessel segments by calculating the degree of similarity to the two-dimensional vessel segments, and extracting a matching candidate group of the three-dimensional vessel segments to be matched with any one of the two- And reconstructing the tree structure.

The step of extracting the vascular tree structure may include setting a pixel extracted from the blood vessel enhancement filter as a node in the two-dimensional X-ray image, and generating an edge connecting the node with another node located in the reference pixel area And extracting a tree-shaped vein path having the lowest weight from the starting point of the predetermined vessel using the weight given to each edge.

Wherein the step of extracting the vascular tree structure comprises the steps of deleting and removing the subtree of the vein path according to the length of the vein so as to remove a vein branch having a low probability of being part of the vein path among the vascular branches constituting the vein path. And repeating the merging process to extract the vascular tree structure.

The calculating of the degree of similarity may be performed by generating a local distance map by performing local distance propagation with respect to the neighboring region based on the two-dimensional vessel segment or the three-dimensional vessel segment, and using the local distance map, Wherein the comparison segment is set as the comparison segment, and when the reference segment is set as the two-dimensional segment, the comparison segment is set as the three-dimensional segment, When the 3D segment is set as the 3D segment, the segment to be compared may be set as the 2D segment.

Wherein the degree of similarity includes a bidirectional matching rate of the comparison segment included in the reference segment and the comparison region, an angle difference between the reference segment and a comparison segment included in the comparison region, a comparison included in the reference segment and the comparison region, And an average distance of the target segment.

The calculation of the degree of similarity may include calculating the bidirectional matching rate based on the ratio of the reference segment and the number of pixels constituting the segment to be compared included in the comparison region and extracting the matching candidate group, And extracting a three-dimensional vessel segment having a predetermined threshold value or more as a matching candidate group for the two-dimensional vessel segment.

The calculation of the degree of similarity may be performed by measuring angles of a reference vector composed of pixels constituting both ends of the reference segment and angles of a comparison vector composed of pixels constituting both ends of the comparison region, The extraction may be performed by extracting a three-dimensional vessel segment having a difference value between an angle of the reference vector and an angle of the comparison vector equal to or less than a predetermined threshold value as a matching candidate for the two-dimensional vessel segment.

The calculating of the degree of similarity may include calculating distance information between pixels of the comparison segment closest to the specific pixel of the reference segment and calculating an average value of the measured distance information for all the pixels constituting the reference segment as the average distance And extracting the matching candidate group as a matching candidate group for the two-dimensional blood vessel segment, wherein the three-dimensional blood vessel segment having the average distance equal to or less than a predetermined threshold value is extracted.

The reconstruction of the two-dimensional vascular tree structure may be performed by excluding a two-dimensional vascular segment in which the matching candidate group is not found, but it is confirmed that the connection between the two-dimensional vascular segments identified as the matching candidate group is disconnected And searching for a connected segment connecting the segmented segments based on the two-dimensional vascular tree structure.

Searching for the connected segment may set a matching candidate group for the upper two-dimensional vascular segment and the lower two-dimensional vascular segment connected to the searched connection segment as the matching candidate group for the connected segment.

The two-dimensional vascular tree structure is reconstructed by separating a structure into segments connected without branching to each two-dimensional vessel segment when there is a plurality of two-dimensional vessel segments in which one of the three-dimensional vessel segments is set as a matching candidate group Dimensional vessel segments matched with each structure, and removes the remaining two-dimensional vessel segments excluding any one of the two-dimensional vessel segments selected in accordance with the similarity.

In addition, the present invention may be a computer-readable recording medium on which a computer program is recorded for performing a method of extracting a vein structure of a two-dimensional X-ray image according to the present invention.

In addition, an apparatus for extracting a blood vessel structure of a two-dimensional X-ray contrast image according to an embodiment of the present invention includes a blood vessel structure extracting a two-dimensional blood vessel tree structure reflecting a weight from a two-dimensional X- Dimensional vessel structure, a rigid body matching unit for performing rigid body matching of the three-dimensional blood vessel structure divided from the two-dimensional blood vessel tree structure and the three-dimensional CT contrast image for the subject, A matching candidate group extracting unit for extracting a matching candidate group of a three-dimensional vascular segment to be matched with any one two-dimensional vascular segment by calculating the similarity of the three-dimensional vascular structure to the three-dimensional vascular segment, and a matching candidate group extracting unit And a vascular structure reconstruction unit for reconstructing the two-dimensional vascular tree structure using a 3D vascular segment.

According to an aspect of the present invention, it is possible to improve the accuracy of extraction of a blood vessel structure using a two-dimensional X-ray contrast image by solving a background noise and a depth information loss of a blood vessel in a two-dimensional X- It can be used as an assistive technology which can assist medical staff by performing automatic image processing of medical data.

In addition, it can be used as a technique for assisted navigation based on real-time display during the PCI procedure by using it as a prior art for non-rigid body matching.

1 is a block diagram showing a schematic configuration of a blood vessel structure extraction apparatus for a two-dimensional X-ray contrast image according to an embodiment of the present invention.
FIGS. 2 to 5 are views showing a specific example of extracting a blood vessel structure by the apparatus for extracting a blood vessel structure of a two-dimensional X-ray contrast image according to FIG.
FIG. 6 is a flowchart showing a schematic flow of a method for extracting a blood vessel structure of a two-dimensional X-ray contrast image according to an embodiment of the present invention.
7 is a diagram showing an example of a performance evaluation result according to the present invention.

The following detailed description of the invention refers to the accompanying drawings, which illustrate, by way of illustration, specific embodiments in which the invention may be practiced. These embodiments are described in sufficient detail to enable those skilled in the art to practice the invention. It should be understood that the various embodiments of the present invention are different, but need not be mutually exclusive. For example, certain features, structures, and characteristics described herein may be implemented in other embodiments without departing from the spirit and scope of the invention in connection with an embodiment. It is also to be understood that the position or arrangement of the individual components within each disclosed embodiment may be varied without departing from the spirit and scope of the invention. The following detailed description is, therefore, not to be taken in a limiting sense, and the scope of the present invention is to be limited only by the appended claims, along with the full scope of equivalents to which such claims are entitled, if properly explained. In the drawings, like reference numerals refer to the same or similar functions throughout the several views.

Hereinafter, preferred embodiments of the present invention will be described in more detail with reference to the drawings.

1 is a block diagram showing a schematic configuration of a blood vessel structure extraction apparatus for a two-dimensional X-ray contrast image according to an embodiment of the present invention.

The apparatus 100 for extracting a blood vessel structure of a two-dimensional X-ray contrast image according to the present invention extracts an initial blood vessel tree structure from a two-dimensional X-ray contrast image of a blood vessel of a subject, After the rigid body alignment with the CT image, the initial vascular tree structure can be automatically reconstructed according to the similarity of the two dimensional vascular structure and the three dimensional vascular structure.

Specifically, the apparatus 100 for extracting a blood vessel structure of a two-dimensional X-ray contrast image according to the present invention includes a blood vessel structure extraction unit 110, a rigid body matching unit 120, a matching candidate group extraction unit 130, (140).

The vascular structure extracting unit 110 can extract a two-dimensional vascular tree structure reflecting a weight from a two-dimensional X-ray image of a subject's blood vessel. In this connection, the description will be made with reference to FIG. 2 and FIG. 3 together.

FIG. 2 and FIG. 3 are diagrams illustrating an example of a process of extracting a two-dimensional blood vessel tree structure generated by the blood vessel structure extraction unit 110 of FIG.

The vascular structure extracting unit 110 may generate a blood vessel graph for generating a two-dimensional vascular tree structure. The vascular structure extracting unit 110 may acquire respective nodes constituting a blood vessel graph from a two-dimensional X-ray image using a Vascular Enhancement Filter. The vascular structure extracting unit 110 may extract the vascularity value for each pixel of the two-dimensional X-ray contrast image using the blood vessel enhancement filter. The vascular structure extracting unit 110 thresholds the vascularity value for each pixel according to a predetermined reference value and thinly binarizes the vascularity value as shown to acquire at least one node . Hereinafter, it is assumed that the blood vessel enhancement filter is a multiscale enhancement filter. However, the present invention is not limited thereto and may be replaced with various known image filters.

The vascular structure extracting unit 110 may generate a blood vessel graph by connecting a specific node with another node within a certain distance. For example, the vascular structure extracting unit 110 may connect any one of the obtained nodes to other nodes within a distance corresponding to 15 pixels. Hereinafter, for convenience of description, a line connected between nodes will be defined as an edge. That is, as shown in the drawing, the vascular structure extracting unit 110 can generate a blood vessel graph in which each node is connected to other nodes located within a certain distance by an edge.

At this time, the vascular structure extracting unit 110 may set a weight for each edge. The vascular structure extracting unit 110 may set different weights on the respective edges according to the following equations.

Here, C _{i, j} denotes a weight set to an edge connected to the i-th node and the j-th node. L is the length of the edge, V is the angular value, which is one of the output values of the multi-scale enhancement filter, I is the intensity value of the 2D X-ray image, θ is the angle of the edge, (Pixel Orientation) which is one of the output values. Also,?,?, And? Are parameters indicating the degree of influence on each weight term, and can be set to have a value of 0 or more. Each weight term can be defined as follows.

Here, d _{i, j} is the i th node and the j length, x _i of the edge (e _{i, j)} is connected to the second _{node, j} is e _i, a set of pixels in by the _j, N _{i, j} are x _{i, j} the number of pixels included in, v _x is the pixel to be obtained from the diagram of a pixel x vascular value, i _x is the intensity value of the pixel x, f (i) is the position of the i-th node in the pixel, θ _x is vascular enhancement filter The direction value of x, Φ _{i, j} , means the orientation of e _{i, j} .

The vascular structure extracting unit 110 may extract the two-dimensional vascular tree structure from the vascular graph generated based on the weights set at the respective edges. Specifically, as shown in FIG. 3, the vascular structure extracting unit 110 may extract a two-dimensional vascular tree structure in which a node having a lowest weight is set as a path with a node as a starting point.

To this end, the vascular structure extracting unit 110 can set a path between nodes. For example, in a blood vessel graph consisting of node V and edge E, the path p between any two nodes v, v 'is given by (vi, vi + 1) ∈ E in i∈ [1, k-1] And may be configured in the order of neighboring nodes. This can be expressed as follows.

The weight of the path p can be determined by the sum of the weights of the edges for the neighboring pixels, and the specific formula for calculating the sum of the weights is as follows.

Using the above-described Equation (4), the vascular structure extracting unit 110 can divide the total weight of the path into the weights of subpaths not meeting at any point along p. That is, the vascular structure extracting unit 110 may divide the total weight of the paths between the nodes v and v 'into the total weight of the paths to the nodes v and vi and the total weight of the paths to the nodes vi and v'. This can be expressed as follows.

Using this, the vascular structure extracting unit 110 can determine an edge having the lowest weight between the two nodes v and v 'as a path. This can be expressed as follows.

Here, P (v, v ') denotes a set of all possible paths between two nodes.

)를 검색할 수 있다. 구체적으로, 혈관 구조 추출부(110)는 미리 설정된 시작 노드를 기준으로 다익스트라 알고리즘(Dijkstra Algorithm)에 따른 The vascular structure extracting unit 110 extracts a minimum weight path between the vertex (specific node) and other vertices (nodes) of the blood vessel graph using the paths generated by the above-

Can be searched. Specifically, the vascular structure extracting unit 110 extracts a vascular structure based on a Dijkstra Algorithm

The extreme algorithm is an algorithm that extracts the shortest path considering the length (weight) of edge between nodes. The vascular structure extracting unit 110 can search for the vein path having the lowest weight from the starting point of the vein using the Daikstra algorithm. The starting point may be defined as any one of a plurality of nodes constituting the blood vessel graph selected by the user. The vascular structure extracting unit 110 can extract a two-dimensional vascular tree structure, which is a tree-shaped vascular path having a root node of the starting point of the dextral algorithm. The vascular structure extracting unit 110 may repeatedly perform deletion and merging on the subtree in order to remove a low vascular branch having a probability of being a part of the vascular in the extracted two dimensional vascular tree structure. Accordingly, the vascular structure extracting unit 110 can extract a two-dimensional vascular tree structure from which an unnecessary vascular branch has been removed.

The rigid body matching unit 120 may perform rigid body matching of the two-dimensional blood vessel tree structure and the three-dimensional blood vessel structure extracted by the blood vessel structure extraction unit 110.

When extracting a blood vessel structure using only a two-dimensional X-ray image, there may arise a problem that a background clutter such as an organ, a bone, and an arbitration mechanism overlaps a blood vessel. This is a problem caused by the projection of a three-dimensional blood vessel into a two-dimensional X-ray image, so there is a limit to solving such a problem with a single two-dimensional X-ray image alone. Accordingly, the apparatus 100 for extracting a blood vessel structure of a two-dimensional X-ray contrast image according to the present invention uses a topology structure of a blood vessel extracted from a three-dimensional CT contrast image of a blood vessel of the same subject, The initial two-dimensional vascular tree structure extracted by the vascular tree 110 can be improved.

For this purpose, the rigid body matching unit 120 performs rigid body matching of the two-dimensional blood vessel tree structure extracted from the two-dimensional X-ray image of the subject and the three-dimensional blood vessel structure extracted from the three-dimensional CT image of the same subject .

In order to match the two-dimensional image and the three-dimensional image with respect to the blood vessel structure of the same subject, the rigid body matching unit 120 may perform initial matching to match the matching space. Specifically, the rigid body matching unit 120 performs a two-dimensional X-ray CT imaging using a medical digital image and communication standard (DICOM) information obtained from a two-dimensional X- It is possible to create the same virtual environment as that at the time of acquiring the line image and apply it to the following mathematical expression.

Here, x, y are two-dimensional coordinates of the X- ray imaging image, X, Y, Z coordinates of the three-dimensional CT imaging image, f is the focal length, m _x, m _y is a unit pixel of the physical unit to the pixel / mm As shown in Fig. Also, p _x and p _y denote principal points, and r and t denote variables used for rotation and movement, respectively.

Even if the same environment as when the 2D X-ray image is acquired by using the Dicom information is used, a registration error due to a difference in protocol between imaging apparatuses or other external factors may occur. In order to reduce such an error, the rigid body matching unit 120 can perform precision matching by performing a transformation function using an optimization technique.

Converting function is x-axis direction movement vector (T _x), y-axis direction movement vector (T _y), z-axis motion vector (T _z), x axis rotation vector (R _x), y axis rotation vector (R _y ) and a z-axis center rotation vector (R _z ). The rigid body matching unit 120 moves the center point (C _3D ) of the three-dimensional vessel center line extracted from the three-dimensional CT contrast image to the origin and performs conversion in the order of the rotation vector and the motion vector to obtain the center point C _2D ). This can be expressed by the following equation.

As described above, the rigid body matching unit 120 can perform initial matching and precision matching stepwise to perform rigid body matching of a two-dimensional vascular tree structure and a three-dimensional vascular structure.

The matching candidate group extraction unit 130 can calculate the similarity to the two-dimensional vascular tree structure and the three-dimensional vascular tree structure based on the rigid-body matching result. At this time, the matching candidate group extracting unit 130 extracts the matching candidate group from the two-dimensional voxel tree structure and the three-dimensional voxel tree structure to shorten the calculation time of the distance comparison in the similarity measurement, Process can be performed.

Specifically, the matching candidate group extraction unit 130 can generate a three-dimensional distance map by performing local distance propagation in a neighboring region of the center line of the blood vessel of the extracted two-dimensional X-ray vessel tree structure. As described above, since the global motion displacement correction between the two-dimensional vascular tree structure and the three-dimensional vascular tree structure due to rigid body matching is performed, generation of a distance map for the entire image is unnecessary. Accordingly, the matching candidate group extraction unit 130 may generate the three-dimensional distance map by performing local distance propagation to shorten the distance map generation time. At this time, the distance map calculates the distances from the blood vessel pixels of the segment based on the Euclidean distance, and allocates the minimum value by comparing the distances to the existing distances to a certain distance.

The matching candidate group extraction unit 130 extracts a matching candidate group for each segment (hereinafter, referred to as a two-dimensional vascular segment) constituting the two-dimensional vascular tree structure (hereinafter referred to as a three-dimensional vascular tree structure) Vascular segment) can be extracted as a matching candidate group. In this regard, FIG. 4 will be described together.

FIG. 4 is a diagram illustrating an example of setting a comparison region between a two-dimensional vessel segment and a three-dimensional vessel segment.

The matching candidate group extraction unit 130 can set a two-dimensional vessel segment or a three-dimensional vessel segment as a reference segment, and determine a segment to be compared with respect to the reference segment. Specifically, the matching candidate group extraction unit 130 may set a three-dimensional vessel segment located within a predetermined distance based on any one of the two-dimensional vessel segments as a comparison region using the distance map generated in the preprocessing process. In addition, the matching candidate group extraction unit 130 may set a three-dimensional vessel segment located within a predetermined distance based on any one of the three-dimensional vessel segments as a comparison region.

Then, the matching candidate group extraction unit 130 can calculate the similarity between the reference segment and the comparison target segment included in the comparison region. In this embodiment, the degree of similarity includes a first similarity to a bidirectional matching rate of a reference segment and a comparison segment, a second similarity to a reference segment and a mean distance of the segment to be compared, And a third similarity to the angular difference of the target segment.

)와 3차원 혈관 트리 구조의 j번째 세그먼트(

)에 대한 양방향 매칭률은, 기준 세그먼트의 전체 픽셀 개수(

)에서 비교 영역으로 설정된 비교대상 세그먼트의 픽셀 개수(

)의 비율로 정의될 수 있다. 이를 수학식으로 표현하면 다음과 같다.First, the matching candidate group extraction unit 130 calculates the bidirectional matching rate of the reference segment and the comparison target segment, compares the calculated bidirectional matching ratio with a predetermined threshold value, and determines whether to set the apportionment target segment as a matching candidate group . For example, the i-th segment of the two-dimensional vascular tree structure (

) And the j-th segment of the three-dimensional vascular tree structure (

) Is the total number of pixels of the reference segment (

), The number of pixels of the segment to be compared set as the comparison region (

). ≪ / RTI > This can be expressed as follows.

)이고, 수학식의 두 번째 행은 j번째 3차원 혈관 세그먼트를 기준 세그먼트로 설정했을 때의 i 번째 2차원 혈관 세그먼트와의 매칭률(

)을 의미한다. 매칭 후보군 추출부(130)는 산출된 두 개의 매칭률 중 어느 하나의 매칭률이 미리 정해진 임계값 이상으로 확인되면, j번째 3차원 혈관 세그먼트를 i번째 2차원 혈관 세그먼트에 대한 매칭 후보군으로 설정할 수 있다. 또는, 매칭 후보군 추출부(130)는 산출된 두 개의 매칭률이 모두 임계값(예컨대, 50%) 이상인 경우에 j번째 3차원 혈관 세그먼트를 i번째 2차원 혈관 세그먼트에 대한 매칭 후보군으로 설정할 수 있다.The first row of the above-mentioned equation represents the matching rate with the j-th three-dimensional vessel segment when the i-th two-dimensional vessel segment is set as the reference segment

), And the second row of the equation shows the matching rate with the i-th two-dimensional vessel segment when the j-th three-dimensional vessel segment is set as the reference segment

). The matching candidate group extraction unit 130 may set the j-th three-dimensional vascular segment as a matching candidate group for the i-th two-dimensional vascular segment if any one of the calculated two matching ratios is confirmed to be equal to or greater than a predetermined threshold value have. Alternatively, the matching candidate group extraction unit 130 may set the j-th three-dimensional vascular segment as a matching candidate group for the i-th two-dimensional vascular segment when the calculated two matching rates are all equal to or greater than a threshold value (for example, 50% .

)에 대한 평균값을 이용하여 제2 유사도를 산출할 수 있다. 매칭 후보군 추출부(130)는 각 평균 거리, 다시 말해 2차원 세그먼트에 대한 3차원 세그먼트의 평균 거리(

) 또는 3차원 세그먼트에 대한 2차원 세그먼트의 평균 거리(

)가 임계값(예컨대, 10픽셀)이하인 경우 j번째 3차원 혈관 세그먼트를 i번째 2차원 혈관 세그먼트에 대한 매칭 후보군으로 설정할 수 있다.Next, the matching candidate group extraction unit 130 extracts the distance between the pixels of the comparison target segment that is the closest distance to the pixels of the reference segment (

) Can be used to calculate the second degree of similarity. The matching candidate group extraction unit 130 extracts a matching candidate group from the average distance, that is, the average distance of the three-dimensional segment with respect to the two-

) Or the average distance of a two-dimensional segment to a three-dimensional segment (

Is less than or equal to a threshold value (for example, 10 pixels), the j-th three-dimensional vessel segment can be set as a matching candidate group for the i-th two-dimensional vessel segment.

Also, the matching candidate group extraction unit 130 may calculate the third similarity degree by extracting the angle difference between the reference segment and the comparison target segment. The matching candidate group extraction unit 130 may extract the vector of the reference segment using the start pixel and the end pixel of the reference segment and extract the vector of the segment to be compared using the start pixel and the end pixel of the comparison target segment. The matching candidate group extraction unit 130 can calculate the difference value of the angles of the two vectors by the third degree of similarity. The matching candidate group extraction unit 130 may set the jth three-dimensional vascular segment as a matching candidate group for the i-th two-dimensional vascular segment when the third similarity is less than or equal to a threshold value (e.g., 30).

In this manner, the matching candidate group extraction unit 130 calculates the first to third similarities between the reference segment and the comparison target segment. If one of the similarity values satisfies the threshold value, the jth three-dimensional segment is divided into i & Dimensional vascular segment can be set as the matching candidate group. In some other embodiments, the matching candidate group extraction unit 130 extracts the matching candidate group from the j-th three-dimensional vessel segment only when two or more similarities among the first to third similarities satisfy a threshold value or if all calculated similarities satisfy a threshold value. Can be set as a matching candidate group for the i-th two-dimensional vessel segment.

The vascular structure reconstructing unit 140 can reconstruct the two-dimensional vascular tree structure generated by the vascular structure extracting unit 110 using the two-dimensional vascular segment in which the matching candidate group exists. In this regard, FIG. 5 will be described together.

The vascular structure reconstructing unit 140 can remove the two-dimensional vascular segment in which the matching candidate group is not set from the two-dimensional vascular tree structure. In this process, there is a possibility that the connection between the two-dimensional segments in which the matching candidate group exists is broken. If the vascular structure reconstructing unit 140 determines that there is a section in which the connection between the two-dimensional segments having the matching candidate group exists, the two-dimensional segment (Hereinafter referred to as " connected segment ").

The vascular structure reconstructing unit 140 may set a matching candidate group for the searched connection segment. The vascular structure reconstructing unit 140 may set all the matching candidates for the upper segment and the lower segment connected by the connecting segment as matching candidates for the connecting segment. Accordingly, different two-dimensional vessel segments can have the same three-dimensional vessel segment as a matching candidate group.

However, since one three-dimensional vessel segment has to be matched with one two-dimensional vessel segment without branching, the vessel structure reconstruction unit 140 determines that a plurality of two-dimensional vessel segments having matching three-dimensional vessel segments as matching candidates Dimensional vessel segment having the highest degree of similarity to the corresponding three-dimensional vessel segment.

Specifically, the vascular structure reconstructing unit 140 can separate the structure into segments that can be branched without branching to a plurality of two-dimensional vascular segment sets having a specific three-dimensional vascular segment as a matching candidate group. The vascular structure reconstructing unit 140 can re-measure the similarity of the three-dimensional segments matched with each structure. The detailed procedure for measuring the similarity to the bi-directional matching rate, the average distance and the angle difference has been described above, so the repetitive description will be omitted. The vascular structure reconstructing unit 140 may normalize the first to third similarities and calculate the final similarity by summing all the normalized similarities. Thereafter, the vascular structure reconstructing unit 140 reconstructs the two-dimensional vascular tree structure by removing the remaining two-dimensional vascular segment in addition to any one two-dimensional vascular segment having the highest degree of similarity to a specific three-dimensional vascular segment.

FIG. 6 is a flowchart showing a schematic flow of a method for extracting a blood vessel structure of a two-dimensional X-ray contrast image according to an embodiment of the present invention.

A blood vessel structure extraction apparatus 100 (hereinafter, referred to as a device) of a two-dimensional X-ray contrast image can extract a two-dimensional blood vessel tree structure from a two-dimensional X-ray contrast image (610). The apparatus 100 generates a blood vessel graph by applying an aneuploidy filter to a two-dimensional X-ray contrast image and generates a blood vessel graph based on a weight value set at each edge of the blood vessel graph, It is possible to search for the vein path having the lowest weight from the starting point. The apparatus 100 can extract the two-dimensional vascular tree structure by repeating the process of cutting and merging the respective subtrees of the vascular path tree according to the vascular length. In addition, the apparatus 100 can extract a three-dimensional blood vessel structure from a three-dimensional CT image of the same subject.

The apparatus 100 may perform rigid body matching of a two-dimensional vascular tree structure and a three-dimensional vascular structure (620). The apparatus 100 may perform initial matching and precision matching sequentially to perform a two-dimensional vessel tree structure extracted from a two-dimensional X-ray image and a three-dimensional vessel structure rigid body extracted from a three-dimensional CT image have. The apparatus 100 may perform only the rigid body matching process in order to use the phase information of the three-dimensional CT contrast image, thereby shortening the execution time required for the matching.

The apparatus 100 may set a matching candidate group of a three-dimensional vessel segment to a two-dimensional vessel segment by measuring the similarity between the rigid-body two-dimensional vessel tree structure and the three-dimensional vessel structure (630). The apparatus 100 may generate a three-dimensional distance map as a preprocessing process, and may set a comparison region of the segment to be compared with respect to the reference segment using the distance map. The apparatus 100 may measure the similarity of the reference segment and the segment to be compared and set the matching candidate group for each two-dimensional vessel segment.

The apparatus 100 may reconstruct a two-dimensional vascular tree structure using a two-dimensional vascular segment in which a matching candidate group is set (640). The apparatus 100 solves the depth information loss of the blood vessel by measuring the similarity of the most similar three-dimensional vessel segment based on the initial division structure of the two-dimensional vessel segment and reconstructing the two-dimensional vessel tree structure based on the similarity degree The accuracy of the extraction of the vascular structure of the two-dimensional X-ray contrast image can be improved.

FIG. 7 is a diagram showing a performance evaluation result of a blood vessel structure extraction method of a two-dimensional X-ray contrast image and a blood vessel structure extraction device performing the same according to the present invention.

In order to evaluate the performance according to the present invention, the experiment was implemented in C ++ and executed on a PC having Intel Core i7-2600 CPU (3.4GHz) RAM 8GB and operating system Windows 7 64bit with Visual Studio 2010 installed. In this embodiment, five LCA images and five RCA images of five subjects were used in the experiments. In this embodiment, it is assumed that the parameter according to Equation 1 is set to? = 0.9,? = 0.7,? = 0.3.

7 (a) is an example of a left coronary artery X-ray image of a subject, and Fig. 7 (b) is an image of a blood vessel value obtained by applying an angioplasty filter. FIG. 7C is a result of thinning the image of the blood vessel value, and each pixel of the thinned image can be set as a node. FIG. 7 (d) shows the result of connecting the edges of the graph with other nodes within a certain distance for each node. FIG. 7 (e) shows the optimal vein path obtained by using the extreme algorithm based on the weight value of each edge. FIG. 7 (f) is a result of repeated deletion and merging of the blood vessel subtree according to the condition in order to remove the branch of the blood vessel which is less likely to be a part of the blood vessel. 7 (g) is a result of rigid body matching of divided blood vessels in a CT image of the same subject as a blood vessel divided from a two-dimensional X-ray contrast image. FIG. 7 (h) is a result of reconstructing the two-dimensional vascular tree structure by measuring the similarity between the two-dimensional vascular segment and the three-dimensional vascular segment based on the rigid-body matching result. As shown in the figure, the problem of overlapping of the background noise with the background of the interventional device can be solved by using the phase structure of the blood vessel extracted from the three-dimensional CT image.

The results of analyzing the blood vessel structure extraction method of the two-dimensional X-ray contrast image according to the present invention and the analysis result of the vascular structure extraction device for performing the same on the data of the left coronary artery image and the right coronary artery image Lt; / RTI >

Accuracy was measured by comparing the result of splitting and the center line of the ground truth manually divided by an expert. In this embodiment, when the blood vessel pixel of the divided result is within 2 pixels of the blood vessel pixel at the center of the ground truth, it is judged as True Positive. And, if the pixel detected at the split point in the split result is within 30 pixels of the ground truth point, it is judged as True Positive.

Precision for vessel pixels averaged 0.859449, Recall averaged 0.360199, and F-measure averaged 0.503774. Precision for the vascular branch was 0.911765 on average, Recall was 0.403644 on average, and F-measure was 0.531179 on average. The reason for the low recall value is that only the vessels of the 2D X-ray image similar to the structure of the divided blood vessels in the 3D CT image were divided. Because the blood vessels divided from the 3 - dimensional CT image are clinically important, the vessels divided from the 3D CT image are clinically important. That is, according to the present invention, since only the blood vessels of the two-dimensional X-ray contrast image similar to the structure of the divided blood vessels in the three-dimensional CT contrast image are divided, the value of Recall is low. Also, it can be seen that the execution time of the experiment is 0.726 seconds, and the extraction of the blood vessel structure is performed quickly and accurately.

Such a technique for providing a method of extracting a vein structure of a two-dimensional X-ray image can be implemented in an application or implemented in the form of program instructions that can be executed through various computer components and recorded on a computer readable recording medium have. The computer-readable recording medium may include program commands, data files, data structures, and the like, alone or in combination.

The program instructions recorded on the computer-readable recording medium may be ones that are specially designed and configured for the present invention and are known and available to those skilled in the art of computer software.

Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks and magnetic tape, optical recording media such as CD-ROMs and DVDs, magneto-optical media such as floptical disks, media, and hardware devices specifically configured to store and execute program instructions such as ROM, RAM, flash memory, and the like.

Examples of program instructions include machine language code such as those generated by a compiler, as well as high-level language code that can be executed by a computer using an interpreter or the like. The hardware device may be configured to operate as one or more software modules for performing the processing according to the present invention, and vice versa.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it will be understood by those skilled in the art that various changes and modifications may be made therein without departing from the spirit and scope of the invention as defined in the appended claims. It will be possible.

100: Vascular structure extraction device of 2D X-ray image
110: blood vessel structure extraction unit
120: Rigid body matching portion
130: matching candidate group extracting unit
140: Vascular structure reconstruction part

Claims (13)

Extracting a two-dimensional vascular tree structure reflecting a weight from a two-dimensional X-ray contrast image of a subject's blood vessel;
Performing rigid body matching of the three-dimensional blood vessel structure divided from the two-dimensional blood vessel tree structure and the three-dimensional CT contrast image of the subject;
Calculating a similarity of the two-dimensional vessel segment of the two-dimensional vessel tree structure and the three-dimensional vessel segment of the three-dimensional vessel structure to extract a matching candidate group of the three-dimensional vessel segment to be matched with any one two-dimensional vessel segment; And
And reconstructing the two-dimensional blood vessel tree structure using the two-dimensional blood vessel segment identified as having the matching candidate group.
The method according to claim 1,
Wherein the step of extracting the vascular tree structure comprises:
Setting a pixel extracted from the blood vessel enhancement filter as a node in the two-dimensional X-ray image and generating an edge connecting the node with another node located in a reference pixel region;
Extracting a tree-shaped blood vessel path having a lowest weight from a starting point of a predetermined blood vessel using a weight given to each edge;
3. The method of claim 2,
Wherein the step of extracting the vascular tree structure comprises:
The subtree of the blood vessel path is repeatedly deleted and merged according to the length of the blood vessel so as to remove the blood vessel branch having a low probability of being part of the blood vessel path among the blood vessel branches constituting the blood vessel path, Extracting the X-ray image from the X-ray image.
The method according to claim 1,
The calculation of the degree of similarity may be performed,
And generating a local distance map by performing local distance propagation on a neighboring region based on the two-dimensional vessel segment or the three-dimensional vessel segment, generating a local distance map by using the local distance map, Is set as a comparison area,
When the reference segment is set as the two-dimensional vessel segment, the comparative segment is set as the three-dimensional vessel segment, and when the reference segment is set as the three-dimensional vessel segment, A method for extracting a vascular structure of a 3D X-ray image, the method comprising the steps of:
5. The method of claim 4,
Preferably,
A reference segment and a comparison segment included in the comparison region, a bidirectional matching rate of the comparison segment included in the reference segment and the comparison region, an angle difference between the reference segment and a comparison segment included in the comparison region, Distance from the at least one of the two-dimensional X-ray images.
6. The method of claim 5,
The calculation of the degree of similarity may be performed,
The bi-directional matching rate is calculated based on the ratio of the number of pixels constituting the segment to be compared and the segment included in the reference segment,
To extract the matching candidate group,
And extracting a three-dimensional vessel segment having the bi-directional matching rate equal to or greater than a predetermined threshold value as a matching candidate group for the two-dimensional vessel segment.
6. The method of claim 5,
The calculation of the degree of similarity may be performed,
Measuring an angle of a reference vector composed of pixels constituting both ends of the reference segment and an angle of a comparison vector composed of pixels constituting both ends of the comparison region,
To extract the matching candidate group,
Dimensional vessel segment, wherein a three-dimensional vessel segment whose difference between the angle of the reference vector and the angle of the comparison vector is equal to or less than a predetermined threshold value is extracted as a matching candidate group for the two-dimensional vessel segment. Vascular structure extraction method.
6. The method of claim 5,
The calculation of the degree of similarity may be performed,
Measuring distance information between pixels of the comparison segment closest to a specific pixel of the reference segment and calculating an average value of the measured distance information for all pixels constituting the reference segment as the average distance,
To extract the matching candidate group,
Dimensional vessel segment is extracted as a matching candidate group for the two-dimensional vessel segment, wherein the three-dimensional vessel segment having the average distance equal to or less than a predetermined threshold value is extracted as a matching candidate group for the two-dimensional vessel segment.
The method according to claim 1,
In order to reconstruct the two-dimensional vascular tree structure,
If it is confirmed that the two-dimensional vascular segment that has been confirmed as having no matching candidate group is disconnected and that the connection between the two-dimensional vascular segment that the matching candidate group is found to be disconnected is disconnected, Segmented X-ray image, and the segment of the X-ray image is retrieved.
10. The method of claim 9,
Searching for the concatenated segment may include:
A method for extracting a vein structure of a two-dimensional X-ray contrast image, the method comprising: setting a matching candidate group for a upper two-dimensional vessel segment and a lower two-dimensional vessel segment connected to the searched connection segment as a matching candidate group for the connection segment.
The method according to claim 1,
In order to reconstruct the two-dimensional vascular tree structure,
When there is a plurality of two-dimensional vessel segments in which any one of the three-dimensional vessel segments is set as the matching candidate group, the structure is separated into segments that are connected without branching to each two-dimensional vessel segment, Calculating a degree of similarity, and removing the remaining two-dimensional vessel segments excluding any one of the two-dimensional vessel segments selected according to the degree of similarity.
A computer-readable recording medium on which a computer program is recorded for performing a method of extracting a vein structure of a two-dimensional X-ray contrast image according to any one of claims 1 to 11.
A vascular structure extracting unit for extracting a two-dimensional vascular tree structure reflecting a weight from a two-dimensional X-ray contrast image of a subject's blood vessel;
A matching unit for performing rigid body matching of the three-dimensional blood vessel structure divided from the two-dimensional blood vessel tree structure and the three-dimensional CT image of the subject;
Dimensional vessel segments of the two-dimensional vessel tree structure and the three-dimensional vessel segments of the three-dimensional vessel structure are calculated to extract a candidate group of three-dimensional vessel segments to be matched with any one of the two- part; And
And a vascular structure reconstruction unit for reconstructing the two-dimensional vascular tree structure using the two-dimensional vascular segment identified as having the matching candidate group.

Images