Method and apparatus for processing vascular images based on user input - Patents.com

Abstract

To provide means that identifies and corrects an error part in an automatic extraction result on the basis of a user input.SOLUTION: A method for processing a blood vessel video from the blood vessel video performed by a processor includes the steps of: extracting a target blood vessel from the blood vessel video; determining a region of interest (ROI) with respect to an extraction result of the target blood vessel on the basis of a first input received from a user; and identifying an error part of the extraction result on the basis of a second input received from the user in the determined region of interest and correcting the identified error part.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] Below, techniques relating to a method for processing vascular images are provided. [Background technology]

[0002] Angiography images are used to observe major blood vessels and identify problem areas within the blood vessels. It is widely used to diagnose and administer necessary treatments and measures. In order to improve the accuracy and quantify the diagnosis results, an image processing engine has been developed that automatically identifies major blood vessels from vascular images. For example, Pi Medical's Caas QCA engine is used to The main blood vessels can be found and displayed in the vascular imaging image. The miscellaneous blood vessels are not the main blood vessels that must actually be extracted. dentification) and part is disconnected (disconnection n), etc. Therefore, in clinical practice, the engine is the primary The extracted classification results cannot be used as is, and manual effort is required to analyze the angiography images. After checking, the errors are manually corrected and stored. Summary of the Invention [Means for solving the problem]

[0003] According to one embodiment, a blood vessel image processing method performed by a processor includes: extracting a target vessel based on a first input received from a user; Determine the Region of Interest (ROI) for the extraction results and determining based on a second input received from the user within the determined region of interest. identifying an error portion in the extraction result and correcting the identified error portion. Includes:

[0004] In one embodiment of the blood vessel image processing method, the step of determining the region of interest includes: determining a morphology of the region of interest based on the points at which the force is detected. do.

[0005] In the blood vessel image processing method according to an embodiment, the step of correcting the error portion includes the steps of: in response to inputs being detected from a plurality of points within the region of interest, A step of correcting the error location based on the multiple points can be included.

[0006] Correcting an error location based on the plurality of points in the blood vessel image processing method according to an embodiment of the present invention The step of moving the point corresponding to the second input along the region extracted from the target blood vessel. detecting whether there is a discontinuity at a point passed by said second input while said second input is moving. In response to the disconnection portion being detected, and b. linking the moieties.

[0007] According to an embodiment of the present invention, a step of correcting an error portion based on the movement trajectory of the blood vessel image processing method is provided. The step is that the point corresponding to the second input is a branch point within a region extracted from the target blood vessel. In response to the target blood vessel being moved to a region separated from the branch point, Among the blood vessel branches, blood vessel branches located outside the region corresponding to the movement trajectory of the second input are erroneously detected. determining the target blood vessel as a recognized portion; and determining the target blood vessel as a recognized portion instructed by the second input after removing the target blood vessel. A step of replacing the extracted vascular branch with a new one and correcting the remaining extraction results based on the replaced vascular branch. The method may include the steps.

[0008] In an embodiment of the blood vessel image processing method, the step of correcting the error portion includes: In response to the determination of the region, the error portion is identified from the extraction result of the target blood vessel. and providing a result of the correction of the error portion to a user. The steps may include:

[0009] In one embodiment of the blood vessel image processing method, the step of providing the correction result to a user includes: In response to the error portion being a misidentified portion, one or more candidate branches are provided to the user. and receiving a point from the user for any one of the one or more candidate branches. In response to receiving a pointing input, a branch corresponding to the error site is selected. and alternating with branching.

[0010] In one embodiment of the blood vessel image processing method, the step of providing the correction result to a user includes: In response to a case where a user input is detected at a plurality of points on the correction result, A region corresponding to the blood vessel branch corresponding to the starting point of the force and a region corresponding to the end point of the user input The method may include the step of connecting regions corresponding to branching vessels.

[0011] In an embodiment of the blood vessel image processing method, the step of correcting the error portion includes: In response to the region being determined, vascular structure data regarding the target vessel; At least one of the curvature information, the diameter information of the target blood vessel, and the brightness information of the target blood vessel The method may include identifying the error site based on one.

[0012] In the blood vessel image processing method according to an embodiment, the step of correcting the error portion includes: responsive to a location being identified, vascular structure data relating to the target blood vessel; At least one of the curvature information of the vessel, the diameter information of the target vessel, and the brightness information of the target vessel. The method may further include correcting the error location based on one of the plurality of errors.

[0013] In the blood vessel image processing method according to an embodiment, the step of correcting the error portion includes: generating a plurality of new extraction results for the target vessel in response to the location being identified; and selecting said selected one of said plurality of new extraction results in response to a user selection input. and selecting one of the extraction results indicated by the selection input.

[0014] The blood vessel image processing device according to one embodiment includes an image receiving unit for receiving a blood vessel image, Extracting a target vessel from the image and extracting the target vessel based on a first input received from a user. Determine the Region of Interest (ROI) for the results and a second input received from the user within the determined region of interest, A processor for identifying an error location and correcting the identified error location. Effect of the Invention

[0015] According to an embodiment of the present invention, a method for processing a blood vessel image performed by a blood vessel image processing device includes: It is possible to extract target blood vessels corresponding to the main blood vessels from the blood vessel image into which the agent has been injected. The blood vessel image processing device according to the embodiment receives a user's input and selects a region of interest (REGION OF INTEREST). f Interest, ROI) and based on input received from the user within the region of interest. Based on this, the location of the error can be identified and corrected. [Brief description of the drawings]

[0016] [Figure 1] 1 is a flowchart illustrating a method for processing a vascular image according to an embodiment. [Figure 2A] 1 is a diagram illustrating a blood vessel image on which a target blood vessel and a region of interest (ROI) are displayed according to an embodiment; [Figure 2B] 1 is a diagram illustrating a blood vessel image on which a target blood vessel and a region of interest (ROI) are displayed according to an embodiment; [Figure 2C] 1 is a diagram illustrating a blood vessel image on which a target blood vessel and a region of interest (ROI) are displayed according to an embodiment; [Diagram 3] 13 illustrates an example in which an error portion occurs in a target blood vessel extraction result according to an embodiment. [Figure 4] A method is shown for identifying locations of errors in the extraction result based on second input received from a user within the region of interest and correcting the identified locations of errors. [Diagram 5] A method is shown for identifying locations of errors in the extraction result based on second input received from a user within the region of interest and correcting the identified locations of errors. [Figure 6] A method is shown for identifying locations of errors in the extraction result based on second input received from a user within the region of interest and correcting the identified locations of errors. [Figure 7] 1 illustrates a method for correcting an error location according to an embodiment. [Figure 8] 1 illustrates a method for correcting an error location according to an embodiment. [Figure 9] A method is presented for identifying errors in a target vessel using a topology of vascular structure data based on nodes corresponding to vessel branches. [Figure 10] 1 illustrates a method for correcting an error location according to an embodiment. [Figure 11]1 illustrates a method for correcting an error location according to an embodiment. [Figure 12] A method is presented for identifying and correcting errors in a target vessel using curvature information of the target vessel. [Figure 13] A method is shown for identifying and correcting errors in a target vessel using diameter information of the target vessel. [Figure 14] We present a method for identifying and correcting errors in a target vessel using brightness information of the target vessel. [Figure 15] 1 is a block diagram illustrating a blood vessel image processing apparatus according to an embodiment of the present invention; DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0017] Specific structural or functional descriptions of the embodiments are disclosed for illustrative purposes only. The embodiments are given in the following description and may be modified in various ways. The present invention is not limited to the specific disclosed form, and the scope of the present specification includes modifications within the technical spirit. , including equivalents or alternatives.

[0018] Terms such as first and second may be used to describe various components. Such terms shall not be construed as being used solely to distinguish one element from another. For example, a first component can be named a second component, and similarly, a second A component may also be named the first component.

[0019] Whenever any element is referred to as being "connected" to another element, It is directly connected or connected to another component, but there is another component in between. It must be understood as possible.

[0020] The singular expression includes the plural expression unless the context clearly indicates otherwise. In the specification, the terms "comprise" and "have" are used to mean any feature, number, step or feature set described in the specification. indicates the presence of a step, action, component, part, or combination thereof. may refer to one or more other features, numbers, steps, operations, components, parts, or combinations thereof. It must be understood that the possibility of the existence or addition of combinations, etc. is not precluded. Must be.

[0021] All terms used herein, including technical or scientific terms, unless otherwise defined. All terms are as commonly understood by those having ordinary skill in the art to which the present embodiment pertains. Terms that are commonly used and predefined have the same meaning as those used in the relevant art. shall be construed to have a meaning consistent with the meaning it has in the context and Unless expressly defined, it is not to be construed as having an ideal or overly formal meaning. Hereinafter, the embodiments will be described in detail with reference to the accompanying drawings. Reference numerals indicate identical elements.

[0022] FIG. 1 is a flow chart illustrating a method for processing vascular images according to one embodiment.

[0023] First, in step 110, the blood vessel image processing device generates a blood vessel image (blood vessel image Extract the target vessel from the sel image. The vessel may be referred to as a main vessel. According to one embodiment, the image receiving device of the vascular image processing device The unit receives blood vessel images captured by the blood vessel image capturing device. Angiography is an image of blood vessels in the living body. Computed agography (CAG) images and / or magnetic resonance images (Magnet may be generated using MRI (Microscopy Imaging) For example, blood vessel images are obtained by taking X-rays of a living body into which a contrast agent has been injected. The image may be a captured image.

[0024] According to an embodiment, the vascular image processing device is configured to: Based on the RNing model, the target blood vessels can be extracted from the blood vessel images. The machine learning model is designed to extract target blood vessels from a vascular image in response to the input of the vascular image. One or more models having a designed machine learning structure, e.g., a neural network The blood vessel image processing device may perform the above-described machine learning on the received blood vessel image. By performing calculations using the model, the extraction results of the target blood vessels can be calculated. For example, The output data of the machine learning model is multiple pixels in a vascular image, with each pixel indicating a target blood vessel. The vascular image processing device outputs the output data including a score corresponding to the likelihood (e.g., probability) of the vascular image being detected. The pixels with a score above the threshold are determined as the target blood vessels, and the target blood vessels are extracted. In another example, the output data of the machine learning model can generate a vascular A target blood vessel region is segmented from the image by dividing a number of pixels in the blood vessel image. The extraction result of the target blood vessel can be expressed as follows: For example, a set of pixels extracted as a target blood vessel from among pixels of a blood vessel image, and / or is an image corresponding to a target vascular region segmented from a vascular image (e.g., a target vascular image) obtain.

[0025] For reference, the neural network is a deep neural network (DNN, DNN includes fully connected networks ( fully connected network), deep convolutional network Deep convolutional network and recurrent Neural networks (recurrent neural networks) Neural networks are based on deep learning to learn nonlinear relationships between inputs. By mapping the data and the output data to each other, object classification, object Deep learning is a technology that can recognize objects and objects by using a single bit of information. It is a machine learning method for solving problems such as object recognition from a set of images. The input data and the output data are then mapped to each other through supervised or unsupervised learning. In the case of supervised learning, the machine learning model described above can be trained. training input (e.g., vascular image for training) and mapping to the corresponding training input Training output (e.g., experts on vascular images for training) A set of images containing pairs of target vessels segmented by the ground truth image For example, a machine learning model can be trained based on training data. The training input may be trained to output a training output. The machine learning model (hereafter, the “ad hoc model”) of the algorithm responds to training inputs. A temporary output can be generated, and a correlation between the temporary output and the training output (e.g., true value) can be calculated. During the training process, the machine learning model is trained to minimize the loss. Parameters (e.g., weights between nodes / layers in a neural network) are lost. It can be updated accordingly.

[0026] However, an example was given in which the machine learning model immediately extracts a target blood vessel from a blood vessel image. However, the machine learning model is not limited to this. For example, the machine learning model may be a whole blood vessel extraction model and The whole blood vessel extraction model extracts the whole blood vessel area from the blood vessel image. The target vessel extraction model may be a model designed to indicate the entire vessel region. A model designed to extract the target vascular region from a target image (e.g., a whole blood vessel image). Alternatively, instead of the whole blood vessel extraction model, the blood vessel image processing device may By detecting the boundaries based on the grayscale level difference between the pixels in the image and the surrounding pixels, In this case, the entire blood vessel region can be extracted. The gradient value of the grayscale levels of the pixel and the surrounding pixels is the threshold gradient value. If the pixel is larger than , the pixel can be detected as a blood vessel boundary. The image processing device can detect the area where the grayscale level changes suddenly as a boundary. The blood vessel image processing device can extract the whole image based on the gradient value of the gray scale level. A target blood vessel image may be extracted from the blood vessel image using a target blood vessel extraction model.

[0027] Furthermore, the blood vessel image processing device has a plurality of functions according to the morphology, type and / or region of the blood vessel. Among the machine learning models, a machine learning model used for target blood vessel extraction may be selectively used. According to one embodiment, the vascular image processing device performs a vascular image processing for each type of blood vessel (e.g., left main coronary artery (LMA)). eft Main Coronary Artery (LM), left anterior descending artery (Left Anterior Descending Artery (LAD), left circumflex artery (L eft Circumflex Artery, LCX), right coronary artery (Right C oronary artery (RCA)) and / or by vascular territory (e.g., proximal territory ( proximal region, mid region, and distal region It can store multiple machine learning models for the distal region. In theory, a vascular region is a region distant from the vascular point where a catheter is inserted. Depending on the distance, the proximal region, the middle region, The endothelial region is divided into the endothelial region, the distal region, and the distal portion. However, the present invention is not limited to this. The vascular region is a region where a contrast agent is injected into a vascular insertion portion to obtain a vascular image. The ratio of the distance away from the point of injection and the distance away from the end of the blood vessel where the contrast agent can be injected For example, the blood vessel image processing device may classify blood vessels according to the type of blood vessel to be extracted. to load the machine learning model that corresponds to the identified vessel type. The vascular image processing device detects blood vessels selected from the vascular image using the loaded machine learning model. A result of extraction of a target blood vessel corresponding to the type of vessel can be generated. The treatment device supports multiple cardiovascular types (e.g., one right coronary artery, two left coronary arteries) Machine learning models for multiple cardiovascular types can be stored. Each of the Dell devices is trained based on the training data corresponding to the cardiovascular type. The trained parameters of the machine learning model for each cardiovascular type may be may be different from each other, and may further be different from each other in terms of machine learning structures (e.g., convolutional neural For ease of explanation, the types of blood vessels are Describe machine learning models with different parameters and / or machine learning structures for each class. However, the present invention is not limited to this. The blood vessel image processing device is capable of detecting the morphology, type, and / or stores multiple machine learning models that are differentiated from each other for each vascular region, and selects the required vascular model. It may be selectively loaded and used for target vessel extraction.

[0028] Then, in step 120, the blood vessel image processing apparatus receives a first input from the user. A region of interest (ROI) for the target blood vessel extraction result is calculated based on the force. The region of interest (ROI) is a region of interest that the user selects within the vascular image for vascular image analysis. The blood vessel image processing device determines a region of interest by: Processing of the vascular image may be focused on the region of interest. For a detailed description of the region of interest, see FIG. This will be described later with reference.

[0029] In step 130, the vascular image processing device performs the image processing for the user within the determined region of interest. and detecting an error port of the extracted result based on a second input received from the user. The error site is identified as a discontinuity site (d ion) and the identified error site is corrected. iscontinuity portion and / or misidentification portion The severance portion includes the severance portion extracted as the target blood vessel. At least one region and / or pixel indicating the target blood vessel region and / or the target blood vessel is selected. or at least one pixel is separated or isolated from other regions and / or other pixels. The misidentified area is not the target blood vessel that should actually be extracted from the vascular image. Regions and / or pixels corresponding to other vessels that are not included in the target vessel were incorrectly extracted as the target vessel. Indicates the location.

[0030] For example, the blood vessel image processing device may detect a location where the second input is detected within the region of interest. based on a movement trajectory of the second input in response to the As another example, the second input can be The blood vessel image processing device detects a starting point corresponding to the pointing input. Errors may be identified and corrected based on the area between the point and the end point. The blood vessel image processing device detects that the second input is detected at a point in the region of interest for a period of time equal to or longer than the threshold value. and in response to the second input being held down, identifying the point at which the second input was held as an error site. The correction based on the second input will be described with reference to FIGS. 4 to 6 below. do.

[0031] 2A to 2C are diagrams illustrating a target blood vessel and a region of interest (ROI) according to an embodiment. FIG. 13 shows a blood vessel image in which the ROI (region of interest) is displayed.

[0032] The blood vessel image processing device captures blood vessel images 201, 202, and 203. The blood vessel image processing device receives blood vessel images 201, 202, and 203 from the blood vessel image photographing device. The apparatus extracts target blood vessels 211, 212, and 213 from the received blood vessel images 201, 202, and 203. The vascular image processing device extracts target blood vessels 211, 212, The blood vessel image processing device extracts 213 based on a first input 231 received from a user. The region of interest 221 for the extraction result of the target blood vessel can be determined by the blood vessel image processing device. The device may determine a morphology of the region of interest based on the point at which the first input is detected. The tube image processing device receives a first input 231 from a user and a start point 232 of the first input 231. The first input 231 is detected at the start point 232 and then the end point 233 of the first input 231 is detected. It may be held until the end point 233, at which point it may be released.

[0033] Referring to FIG. 2A, the blood vessel image processing apparatus receives a signal from a user from a start point 231 to an end point 232. Up to 33 drag inputs can be received as the first input 231. To enter a drag, click (c) at any point on the display (for example, the starting point 232). A lick or touch occurs and the click or touch is held. The input point moves to another point (e.g., end point 233) where the click or touch is released. The blood vessel image processing device also detects the start point 232 and the end point 233. A region of interest 221 can be determined in a predefined form between the two. 21 is arranged so that a start point 232 and an end point 233 are located on the boundary of the region of interest 221. As shown in FIG. 2A, the predefined shape of the region of interest 221 may be a rectangle. However, the shape is not limited to this and may be polygonal or circular. The processing device detects the current position of the first input 231 from the point where the detection of the first input 231 started. A temporary region of interest 220 between the points is output via a display and presented to the user. The blood vessel image processing apparatus provides the user with a temporary region of interest 220. Before deciding on the 21, it is possible to provide feedback to the user. When the detection of the first input 231 is completed, the blood vessel image processing device between the start point 232 and the end point 233 of the first input 231 in a predefined form. The region of interest 221 can be determined and output to the user via a display. do.

[0034] Referring to FIG. 2B, the blood vessel image processing device receives a first input including a plurality of inputs from a user. The first input includes a plurality of inputs for any point on the display. The vascular image processing device receives a first input from a user. The points 241, 242, and 243 on the display corresponding to multiple inputs are displayed based on the user's input. The area inside the polygon formed by continuously connecting the lines is determined as the area of ​​interest 222. The blood vessel image processing device detects the current position of the first input from the point where the detection of the first input started. The user may be provided with the form of a portion of a polygon that will be formed between the points.

[0035] Referring to FIG. 2C, the blood vessel image processing apparatus receives a start point 251 from a user. A drag input to the end point 251, which is the same point, can be received as the first input. The blood vessel image processing device automatically detects the start and end points of the drag inputs that correspond to the same drag input. The area inside the free curve is determined as the region of interest 223. The processing device detects the first input from the point where the detection of the first input started to the point where the first input is currently detected. A partial form of a freeform shape consisting of a free curve to be formed can be provided to the user.

[0036] According to an embodiment, the blood vessel image processing device determines based on a first input received from a user. If a vascular branch extracted from the target blood vessel exists within the region of interest, the branch is included in the region of interest. Based on one branching point, the upper branch and the lower branch adjacent to each other are The branch can be fixed to the correct region branch. Among the target blood vessels extracted by the method, those that are not identified as error sites or that the user needs to correct are The blood vessel image processing device detects the upper branch fixed to the correct region branch. As a criterion, error sites can be identified and corrected only for the lower branches.

[0037] According to another embodiment of the present invention, a blood vessel image processing device includes: The determined region of interest may be modified according to a predetermined condition. The condition is that the region of interest is a part of the boundary of the region of interest and is extracted by the vascular image processing device. At least one correct portion of the image vessel that does not correspond to an error portion or that the user does not want to correct Illustratively, the blood vessel image processing apparatus receives the solution region branch and A part of the boundary of the region of interest determined based on the first input is extracted by the vascular image processing device. Among the target blood vessels, there is no error area, or the user does not want to correct The region of interest is expanded or moved so that it intersects with at least one correct region branch at a time. The vascular image processing device determines the correct region branch among the vascular branches in the region of interest. The region of interest may be modified so that it intersects the upper boundary of the region of interest. The region of interest may be determined so as to intersect with the target blood vessel at any position on the boundary of the region of interest. The blood vessel image processing device may also modify the region based on a first input received from a user. If the extracted target blood vessel is not included in the region of interest determined by the user, The first input may again be received from the

[0038] The blood vessel image processing device detects the error location based on the branch point. A branch point is a point where two or more Shows where blood vessel branches join and / or split into multiple blood vessel branches. The vascular region captured at 0 may have multiple branching points.

[0039] The blood vessel image processing apparatus according to the embodiment finally generates a target blood vessel 210 as shown in FIG. In order to determine the region of interest, the target blood vessel 210 is extracted, the region of interest 220 is determined, and the determined region of interest 220 is processed. Errors in the extraction results can be identified and corrected within 20.

[0040] FIG. 3 illustrates an example in which an error portion occurs in the extraction result of a target blood vessel according to an embodiment.

[0041] According to an embodiment of the present invention, a blood vessel image processing apparatus is provided for identifying branching points and branches from a blood vessel image. For example, the blood vessel image processing device can extract the entire blood vessel region and The branching point and the branch can be identified from the region. The vascular branch can be identified based on the anchor point. The blood vessel region can be extracted from the whole body. Based on the extraction model, the entire blood vessel region can be extracted from the blood vessel image. As described above, the vascular image is divided into vascular regions and the remaining non-vascular regions. The model may be trained to generate output data indicative of

[0042] FIG. 3 shows an example of an area of ​​interest of the blood vessel image 200 shown in FIG. 2 among the entire blood vessel area extracted. 220. The vascular image processing device detects branching points within the region of interest 220. For reference, in this specification, the upper vascular branch is The branch is shown before the branching point according to the direction of blood flow or the direction of contrast medium flow, and the lower blood A vessel branch refers to a branch beyond a branching point. For example, a vessel branch identified by a vascular image processing device. As a result, the first to seventh branches 301 to 307 and branching points 311 and 312 are shown. Based on the first branch point 311, the first branch 300 is a superior vascular branch, and the second branch 300 is a superior vascular branch. The second branch 302 and the third branch 303 are lower vascular branches. The relationship of the canal branches is relative and may vary from branch point to branch point. For example, the second branch 3 02 is the superior vascular branch for the fourth branch 304 and the fifth branch 305. The processing device generates and indexes nodes corresponding to each vascular branch, Structure data can be generated. The vascular structure data is described below with reference to FIG. 9 below. .

[0043] In the example shown in FIG. 3, the vascular image processing device detects a first branch as a vascular branch belonging to a target blood vessel. It is possible to extract a branch 301, a third branch 303, and a fifth branch 305. However, in FIG. The vascular pathway 320 corresponding to the illustrated target blood vessel 220 includes a first branch 301, a second branch 302, and a , and the fifth branch 305. Therefore, in the exemplary target vessel extraction result, The first branch point 311 from the first branch 301 to the third branch 303 is an error site. The detection and correction of error sites is described below.

[0044] 4-6 show an example of an error correction method for the extraction result based on a second input received from a user within the region of interest. The present invention shows how to identify error sites and correct the identified error sites.

[0045] FIG. 4 is a block diagram showing a blood vessel image processing apparatus according to an embodiment of the present invention. The blood vessel image processing device includes a second in response to the input being detected from a plurality of points within the region of interest, Based on this, the error location can be corrected.

[0046] The second input received by the blood vessel image processing device according to an embodiment from a user is a drag (dr ag) may indicate the input. The vascular image processing device may indicate that the point where the second input is detected is the region of interest. In response to a movement within the second input 491, The system can identify and correct error locations based on the vascular imaging device. While the point corresponding to the second input 491 moves along the region extracted from the target blood vessel, Then, it is detected whether there is a disconnection at the point where the second input 491 passes. The image processing device extracts the blood vessel branches 401 and 405 as the target blood vessels, and the second input 491 of the user is input. If the point corresponding to moves along the region 420 extracted from the target blood vessel, the blood vessel image processing The processing device detects whether there is a discontinuity at the point where the second input 491 passes. The blood vessel image processing device then extracts the point corresponding to the second input 491 from the target blood vessel. Even if the point corresponding to the second input 491 is moved away from the target blood vessel for a while, For example, the blood vessel image processing device can determine that the second input 49 moves. 1 is a branch point within the region 420 extracted from the target blood vessel. and then move from the point where the target object left the area 420 to the point where the target object left the area 420 before the threshold time has elapsed. When the second input 491 is restored to the region 420, it is determined that the second input 491 is held within the region 420. The blood vessel image processing device receives the second input 491 and extracts the branch points in the region 420 from the target blood vessel. In response to a move from the point to another branch that has left the region 420, a second input 491 is It can be determined that the user has left the area 420.

[0047] For example, the vascular image processing device may include a step of: The second input corresponds to a drag input between the graphic points corresponding to the two included points. 491. The drag input may be received when the user drags the target blood vessel within the target blood vessel area. Click or touch the graphic point that corresponds to one of the two points shown, and then click Click or touch the graphic point that corresponds to the remaining point. The blood vessel image processing device receives an input from a user. In response to a second input 491 being received, a blood vessel branch corresponding to the starting point of the second input 491 and a second input Detect whether there is a disconnection at each point between the end point of the force 491 and the corresponding vascular branch. Therefore, the blood vessel image processing device receives the second input 491 as a region extracted from the target blood vessel. While moving along the angiogram, the detection of the disconnection can be continuously attempted. The image processing device determines whether the start point and end point corresponding to the second input 491 received from the user are The points are included in the areas corresponding to the blood vessel branches 401 and 405, respectively, and correspond to the second input 491. When the target blood vessel moves along the region 420 extracted from the target blood vessel, the blood vessel image processing device The blood vessel image processing device detects whether there is a disconnection between the branches 401 and 405. The regions corresponding to the vascular branches 401 and 405 of the extracted target blood vessel are mutually If the blood vessels are separated or separated from each other, it can be determined that the area is disconnected. The processing device detects the corresponding second input 491 while the second input 491 passes through a point between the blood vessel branches 401 and 405. It is determined whether the point belongs to the region 420 extracted from the target blood vessel. In response to a case where the point does not belong to the region 420, the point is determined to be a disconnection portion. In accordance with an embodiment, the blood vessel image processing apparatus detects a discontinuity by receiving a second input. It is possible to connect disconnected parts along the movement trajectory of 491. For example, blood vessel image processing The device may be coupled to the contour and / or border of the branch vessel determined to be the site of disruption. For example, the point corresponding to the second input 491 of the user is located in the region 4 extracted from the target blood vessel. When moving along 20, the blood vessel image processing device detects the blood vessel branch 4 determined as a disconnection site. 01,405 along the movement trajectory of the second input 491 received from the user. The blood vessel image processing device may receive the blood vessel segment having a cutoff point corresponding to the second input 490. When moving along the region 402 between the branches 401 and 405, the vascular image processing device The region 402 between the vascular branches 401 and 405 can be extracted as the target blood vessel. The image processor may add region 402 to the extraction result of the target vessel.

[0048] The second input received by the blood vessel image processing device according to an embodiment from the user is a pointing Pointing input can be indicated by a As an input for indicating a point, for example, a click input and / or a tap input by a mouse operation is used. The input may include an input by a touch operation on a touch display (for example, a touch input). However, the pointing input is not limited to this, and various pointing inputs can be used depending on the design. touch devices (e.g., trackball mice, touchpads, and trackpads) The vascular image processing device may include a position instruction input by a user. A pointing input to an end point can be received as a second input. However, Without being limited thereto, the blood vessel image processing device may receive clicks from a user to two or more points. The blood vessel image processing device may receive a pointing input of a click or a touch. The pointing input identifies the error area based on the area between the start and end points of the pointing input. In one embodiment, the blood vessel image processing device detects blood vessel branches 401, 402, and the blood vessel branches 403, 404. 5 is extracted as the target blood vessel, and the blood vessel between points 492 and 493 corresponding to the second input of the user is extracted. It is possible to detect whether there are any breaks in the region.

[0049] For example, the vascular image processing device may include a target blood vessel region extracted from a region of interest by a user. Pointing including click or touch input to two points 492, 493 included in the The vascular image processing device may receive a second input corresponding to a gating input from a user. In response to the second input being received, a blood vessel branch corresponding to the starting point 492 of the second input and a second input Detect whether there is a disconnection at each point between the end point of the force and the blood vessel branch corresponding to the end point of the force, The disconnection can be corrected in the same way as when a drag input is received via two inputs.

[0050] FIG. 5 is a diagram illustrating a blood vessel image processing apparatus according to an embodiment of the present invention. A method for identifying and correcting erroneous portions of the extraction result based on the above will be described. The point corresponding to the second input 590 is the branch point in the region extracted from the target blood vessel. Extraction from the target vessel connected to the branch point in response to moving the vessel to the detached region Among the vascular branches detected, those located outside the area corresponding to the movement trajectory of the second input are selected. The blood vessel image processing device can determine the branch point as a misidentified part. If the region of interest is determined as a recognition site, the region of interest based on the first input is the entire lower part based on the branch point. The blood vessel image processing may be performed by expanding the blood vessel branch so that it is included within the region of interest. The device extracts the blood vessel branches 501, 502, and 505 as the target blood vessels, and receives the second input 59 from the user. The point corresponding to 0 is the branch point 513 in the region 520 extracted from the target blood vessel. When the user moves to the area 530 from which the user left the original position, the part related to the branch point 513 is judged as a misidentified part. For example, the blood vessel image processing device may divide the target blood vessel based on the branching point 513. The second extraction is performed on the region corresponding to the lower vascular branch 504 different from the lower vascular branch 505 extracted in the first step. When the second input 590 moves, the point corresponding to the second input 590 received from the user is It can be determined that the image of the blood vessel has moved to the area where the image of the blood vessel has been removed. At least one branch 513 of the blood vessel branches extracted from the target blood vessel adjacent to the branch point 513 is 05 can be determined to be a misidentified area based on the extraction results for the target blood vessel.

[0051] The vascular image processing device targets at least one of the vascular branches identified as the erroneous portion. The blood vessel is replaced by a blood vessel branch indicated by the second input 590, and the replaced blood vessel branch is The remaining extraction results can be modified based on the branch. For example, the vascular image processing device 2. When the point corresponding to the input 590 moves away from the region corresponding to the extracted target blood vessel, In this case, the primary extracted sub-vascular branch 505 is divided into the target blood vessel based on the branch point 513. The blood vessel image processing device can then perform the following based on the replaced blood vessel branch 504: For example, the blood vessel image processing device may correct the extraction result of the target blood vessel based on the extracted blood vessel. The blood vessel branches subsequent to the primary extracted lower blood vessel branch 505 are excluded from the result, and the replaced blood vessel The blood vessel branches subsequent to the vessel branch 504 may be added to the extraction result of the target blood vessel.

[0052] FIG. 6 illustrates a blood vessel image processing apparatus according to an embodiment of the present invention. Based on the results, correction candidates are provided to the user, and the error portion is identified and corrected according to the user's selection input. The blood vessel image processing device performs the following at a point in the region corresponding to the extraction result of the target blood vessel: If the second input 690 is detected for a period of time equal to or longer than the first threshold time, the corresponding point is determined as an error portion. The vascular image processing device can perform correction processing corresponding to the point determined as the error portion. The correction candidate is provided to the user. The correction candidate is that the point where the second input 690 is held is the error portion. When the vascular branch is judged as a set of candidate vascular branches having a form capable of filling the disconnection site, or a misidentified part, A set of candidate vascular branches with alternating morphologies and structures can be shown. For example, The blood vessel image processing device receives blood vessel structure data, blood vessel image data, and blood vessel image data, which will be described later with reference to FIGS. Determining correction candidates based on vessel curvature information, vessel diameter information, brightness information, etc. For example, the first revision candidate 661 is a candidate branch determined based on the vascular structure data. , the second revision candidate 662 is a candidate branch determined based on the curvature information, and the third revision candidate 663 is The candidate branch determined based on the diameter information, and the fourth revised candidate 664 is determined based on the brightness information. However, this is merely an example, and the determination of the candidate branch may be The setting may vary depending on the design. Furthermore, the point at which the second input 690 is held may be set as an error. In the above description, the blood vessel image processing device determines the location of the blood vessel as the second input 690. For the points where blood vessels are held, blood vessel structure data and blood vessels are stored as described below with reference to FIGS. Even if the error location is determined based on the curvature information of the vessel, the diameter information of the blood vessel, and the brightness information, good.

[0053] The vascular image processing device, based on a third input for selecting a correction candidate received from a user, One of the correction candidates is selected. The vascular image processing device selects one of the correction candidates included in the selected correction candidate. Vascular branches may be added to the extraction result of the target vessel.

[0054] For example, the blood vessel image processing device first extracts blood vessel branches 601, 602, and 605 as target blood vessels. The point at which the vascular image processing device detects a second input 690 received from a user. is included in the area corresponding to the branching point 611, the blood vessel image processing device The user selects one of the suggestions 661, 662, 663, or 664 corresponding to 11. A graphic object 660 that can be provided to a user can be output. ,The third input allows the user to select one of the correction suggestions 661, 662, 663, 664 Receive the vascular branch (e.g., vascular branch 602, 60 4) can be extracted to the target blood vessel.

[0055] In the blood vessel image processing apparatus according to an embodiment, a point corresponding to the second input 690 is extracted. The point corresponds to the boundary of the image vessel (e.g., the vessel edge), and the second input 69 If the time when 0 is detected is equal to or greater than the second threshold, the correction candidate corresponding to the boundary is displayed to the user. For example, the blood vessel image processing device may provide a second input received from the user. The boundary of the target blood vessel corresponding to the force 690 is smoothed and flattened (smooth correction suggestions for highlighting, and correction suggestions for enhancing contrast with the background Correct candidate or alternative option to select a boundary that is different from the existing boundary The user can then be provided with a

[0056] The blood vessel image processing device detects an error portion from the extraction result of the target blood vessel in response to the determination of the region of interest. The blood vessel image processing device can identify the error portion and correct the identified error portion. The result of the correction of the part can be provided to the user. Provide the correct result to the user, receive input from the user corresponding to the correction result, and correct the error part. Then, the error location is corrected based on a second input 690 received from the user within the modified region of interest. In other words, the blood vessel image processing device can identify and correct the error portion by using the above-mentioned method. before receiving second input 690 from the user within the region of interest to identify and correct error locations. Then, the processor automatically identifies and corrects the error, and provides the user with the corrected results. A pretreatment process can be carried out.

[0057] For reference, in FIG. 4 to FIG. 6, the second inputs 490, 590, and 69 0 is shown to be touch input, but is not limited to this, click input, etc. It may also include other forms of input such as pointing to a location on the display.

[0058] 7 and 8 show a method for correcting an error portion according to an embodiment.

[0059] Figure 7 shows that an error area is identified from the extraction result of the target blood vessel, and the identified error area is a misidentification. In the case where the misidentified portion is determined to be a part, a method for providing the user with a correction result for the misidentified portion will be described. According to the embodiment, when an error portion is identified in the target blood vessel, the blood vessel image processing device Correcting identified errors based on input (e.g., touch input) The blood vessel image processing device can perform one of the following in response to the error portion being a misidentified portion: The blood vessel image processing device provides the user with one or more candidate branches. In response to receiving a pointing input for any one of the branches, an error The branch corresponding to the region is replaced with the selected branch. When the branches 701 and 702 are extracted as the target blood vessels, vascular structure data regarding the target blood vessels At least one of the curvature information of the target blood vessel, the diameter information of the target blood vessel, and the brightness information of the target blood vessel. Based on the above, at least one of the blood vessel branches 701 and 702 is determined to be relative to the target blood vessel. The blood vessel image processing device can determine the error area based on the extracted result. When an error site is a misidentified site, it may contain one or more candidate vessel segments that are not extracted from the target vessel. The branches 711, 712, and 713 can be provided to the user. The interface allows the user to select candidate vascular branches 711, 712, and 713 via a display. It outputs graphic objects 771, 772, 773 that can provide the user with a choice of The graphic object may be in the form of a shape that surrounds the area corresponding to the candidate vascular branch. In FIG. 7, the graphic objects 771, 772, and 773 are shown as rectangular shapes. However, the form is not limited to this. The user simply wants to For a point corresponding to one of the graphic objects corresponding to the candidate vascular branch, By pointing input such as clicking or touching The candidate blood vessel branch can be selected. That is, the blood vessel image processing device uses the graphic object When detecting user input to a point corresponding to the object, It is determined that the corresponding candidate vascular branch has been selected, and the vascular branch corresponding to the error site is selected. For example, the vascular image processing device may receive a graphic image from the user. When detecting a user input to a point corresponding to the graphic object 771, A candidate vascular branch 711 corresponding to the Fick object 771 is extracted as a target blood vessel. That is, the blood vessel image processing device can extract the blood vessel branch 702 from the target blood vessel extraction result. It can be replaced by branch 711.

[0060] Figure 8 shows the error area identified from the target blood vessel extraction results, and the identified error area is disconnected. If the disconnection is detected, a method for providing the user with the correction result of the disconnection will be described. According to the present invention, the blood vessel image processing device detects an error portion in the target blood vessel when the error portion is a disconnection portion. In response, the blood vessel region including the disconnection can be provided to the user. In response to detecting a user input at multiple points on the modified result, and a region corresponding to the end point of the user input Regions corresponding to vascular branches can be linked.

[0061] The blood vessel image processing apparatus according to an embodiment of the present invention receives drag and drop signals from a user for a correction result. In response to the drag input 891, The region and the region corresponding to the blood vessel branch corresponding to the end point of the drag input 891 are connected. For example, when the vascular image processing device primarily extracts vascular branches 801 and 802 as target blood vessels, The blood vessel image processing device detects whether the regions corresponding to the blood vessel branches 801 and 802 of the extracted target blood vessel are mutually different. The input / output interface of the vascular image processing device is A graphic representation of the connection between the regions corresponding to the vascular branches 801, 802 is displayed on the display. The graphic object 870 can be output. The blood vessels 801 and 802 are connected to each other by a severing portion. In the embodiment shown in FIG. 1, blood vessel branches 801 and 802 can be connected by the user. The tube image processor responds to the drag input 891 received from the user by The regions corresponding to the vascular branches 801 and 802 selected by 891 are connected, and the disconnected vascular A blood vessel branch 803 between the branches 801 and 802 is extracted as a target blood vessel. The branch selected by 1 is the blood vessel branch corresponding to the drag start point and the drag end point. In other words, the blood vessel image processing device extracts the target blood vessel as follows: A branch vessel 803 may be included.

[0062] According to another embodiment of the present invention, a blood vessel image processing apparatus is provided that receives two correction results from a user. In response to a pointing input corresponding to the above points, A region corresponding to a blood vessel branch corresponding to a start point 892 and an end point 893 corresponding to a ping input. That is, the blood vessel image processing apparatus can receive a start point and an end point from the user. The blood vessel image processing device can receive a click or touch input to the user. In response to a pointing input received from The area corresponding to the corresponding blood vessel branch 801 and the end point 893 of the pointing input are In this case, the regions corresponding to the blood vessel branches 802 can be connected.

[0063] FIG. 9 shows the topology of the vascular structure data based on nodes corresponding to vascular branches. 1 shows how to identify errors in elephant vessels.

[0064] According to one embodiment, the vascular image processing device calculates vascular structure data 900 relating to the target blood vessel. Based on this, the error portion can be identified. The data 900 is topology data indicative of the structure of a blood vessel, for example, for blood vessel branches. The corresponding nodes may be indexed tree-structure data. The processing device identifies blood vessel branches based on blood vessel branching points for the entire blood vessel region, and A node corresponding to each of the detected vascular branches is generated, and the generated nodes are indexed. The vascular structure data 900 can be generated based on the

[0065] The vascular image processing device divides a superior vascular branch into a plurality of inferior vascular branches based on the branching point. When a blood vessel is branched, a node corresponding to a higher blood vessel branch is connected to a node corresponding to a lower blood vessel branch. The blood vessel image processing device can detect the branching points and blood vessel branches identified from the blood vessel image. By repeating the above-mentioned node connection, vascular structure data 900 for the entire blood vessels is obtained. For example, in the blood vessel structure data 900 shown in FIG. The first to seventh nodes N1 to N7 are successively connected to the first to seventh branches 30 shown in FIG. 1 to 307. The first node N1 is the top node, and the second node N2 and the third node N3 are Node N3 is a subordinate node of the first node N1. The remaining nodes have a similar hierarchy. do.

[0066] The blood vessel image processing device according to an embodiment of the present invention is configured to extract a blood vessel segment as a target blood vessel from a blood vessel image. Node connectivity data 910 between nodes corresponding to branches can be calculated. The image processing device extracts the target blood vessels shown in FIG. Node connection data 910 in which the third node N3 and the fifth node N5 are connected can be calculated.

[0067] The blood vessel image processing device compares the node connection data 910 with the blood vessel structure data 900. The vascular image processing device can identify an error portion of the target blood vessel by the generated vascular image. Detecting connection errors between nodes of a target blood vessel based on the structure data 900. The device detects a vascular branch and / or a branch point corresponding to the node where the connection error is detected. The blood vessel image processing device determines the blood vessel structure data 910 as a part. The error location can be determined based on the nodes that do not match 00. For example, The node link data 910 extracted in FIG. 9 includes a first node N1, a third node N3, and The fifth node N5 indicates the nodes connected in order. The node N3 is not connected to the fifth node N5. In the node connection data 910, the connection between the third node N3 and the fifth node N5 is a vascular structure. It can be determined that the data does not match the data 900. The connection between node N5 and node N6 is an error. The vascular branch corresponding to at least one of the nodes is determined as an error site. For example, the blood vessel image processing device in FIG. It is possible to determine that a vascular branch corresponding to at least one of the nodes N5 is an error site. In the above example, the blood vessel branch corresponding to the third node N3 or the fifth node N5 is the target. This is a misidentified area because it was mistakenly extracted as belonging to a blood vessel.

[0068] In another example, the blood vessel image processing device may generate a node connection data based on the blood vessel structure data 900. A blood vessel branch corresponding to a node that has fallen off the blood vessel data 920 may be detected as a disconnection site. The tube image processing device is a node-connected device directly connected to the first node N1 to the fifth node N5. The blood vessel image processing device can calculate the blood vessel structure data 920 based on the blood vessel structure data 900. It is found that the second node N2 is missing from the node connection data 920. In this case, the blood vessel image processing device can output a blood vessel branch corresponding to the second node N2. It can be detected as a position.

[0069] 10 and 11 show a method for correcting an error location according to one embodiment.

[0070] FIG. 10 illustrates the modification when the node link data 910 is calculated in FIG. According to one embodiment, the blood vessel image processing device detects a misidentified area. In response to a case where the "on portion" is identified, the node corresponding to the misidentified area is assigned to the vascular The node can be replaced with a node that matches the structure data. In the node connection data 910, it is determined that the fifth node N5 does not match the blood vessel structure data. The blood vessel image processing device can detect the fifth node N 5 is excluded from the node connection data, and the sixth node N6 that matches the vascular structure data is selected. 10. Adding the modified node link data to the node link data 1010. Therefore, the blood vessel image processing device can detect the blood vessel corresponding to the replaced sixth node N6. A branch (eg, the sixth branch 306 in FIG. 3) can be extracted as a target vessel.

[0071] FIG. 11 illustrates the modification when the node link data 920 is calculated in FIG. According to one embodiment, the vascular image processing device is configured to, in response to identifying a discontinuity, A connectable node can be inserted between nodes corresponding to regions. The image processing device detects that the first node N1 and the fifth node N5 are disconnected from each other in the node connection data 920. The blood vessel image processing device has a second node N2 between the nodes N1 and N5. The insertion can generate modified node connection data 1110. The image processor detects a blood vessel branch corresponding to the second node N2 (e.g., the second branch 302 in FIG. 3). It can be extracted as a target blood vessel.

[0072] As described above, the vascular image processing is performed by correcting the misidentified part by node replacement and node addition. The processing device can remove node connectivity errors based on the topology of the vascular structure data. can.

[0073] FIG. 12 illustrates a method for identifying and correcting errors in a target vessel using curvature information of the target vessel. vinegar.

[0074] According to an embodiment, the vascular image processing device detects a bend between adjacent vascular branches in the target vessel. For example, the vascular image processing device can use the ratio to identify and correct the erroneous part. In response to a curvature between adjacent branches of a blood vessel exceeding a threshold curvature, For example, the blood vessel image processing device determines the blood vessel branch 1201 as a misidentified part. When the target blood vessel 1201 and 1202 are extracted as the first blood vessel, the curvature between the blood vessel branches 1201 and 1202 is The vascular image processing device, in response to a case where the calculated curvature exceeds a threshold curvature, At least one of the above-mentioned vascular branches 1201 and 1202 is extracted from the target blood vessel. In the blood vessel image, it is necessary to actually extract the area. The adjacent vascular branches belonging to the target vessel that should be treated have a relatively small curvature and are within the threshold curvature. Adjacent vessel branches that exhibit curvatures that exceed the rate are potential error sites.

[0075] The vascular image processing device responds when an error portion is identified in the target blood vessel using a threshold curvature. The vascular image processing device can correct the identified error portion by For each lamellar site, one of the adjacent branches is aligned so that the curvature between the remaining branches is equal to the threshold. The candidate branch can be replaced with a candidate branch that has a value less than or equal to the curvature. Among multiple vascular branches connected to one vascular branch as a single vascular branch, those that have not been extracted from the target blood vessel It shows one or more remaining branches.

[0076] For example, the blood vessel image processing device may extract a blood vessel branch 1203 that is not extracted from the target blood vessel and If the curvature between the distal branch 1201 and the distal branch 1202 is equal to or less than the threshold curvature, In addition, the blood vessel branch 1203 can be extracted as a target blood vessel. From the extraction result of the image vessel, the blood vessel branch 1202 can be replaced with a blood vessel branch 1203. Furthermore, the blood vessel image processing device corrects the identified error portion by taking into consideration the type of blood vessel. (e.g., Left Main Coronary Artery, LM ), Left Anterior Descending Artery , LAD), Left Circumflex Artery (LCX), Right Coronary Artery (RCA) or vascular area Different threshold curvatures can be applied depending on the region (e.g., proximal, middle, and distal regions). Cut.

[0077] The vascular image processing device performs a principal direction component analysis (e.g., principal direction component analysis) of the vascular branch to analyze the curvature difference. Perform component analysis (PCA, principal component analysis) The blood vessel image processing device may perform the following steps for each of the blood vessel branches divided based on the branching points: The main direction component of the blood vessel branch can be analyzed. For example, the vascular image processing device may calculate the main direction component for each vascular branch. Using the segmental analysis, the main direction vector can be obtained for each blood vessel branch. The tor is a vector having a directional component representing the direction in which the blood vessel branch extends from the branching point. For example, the main direction vector for any branch vessel is is the direction of the principal component of the vector component from the branch point to the point corresponding to the blood vessel branch. Illustratively, the main direction vectors of the blood vessel branches 1201, 1202, and 1203 are respectively The vascular image processing device corresponds to vectors 1221, 1222, and 1223. In response to a dot product value of the main direction vectors of adjacent blood vessel branches exceeding a threshold value, The branch adjacent to can be judged as a misidentified site.

[0078] For example, the vascular image processing device may include a vascular branch 1201, 1202, and a vascular image processing device may include a vascular branch 1203. If the inner product value of the two main direction vectors 1221 and 1222 exceeds a threshold, the two main direction vectors are adjacent to each other. The vascular branches 1201 and 1202 were misidentified. The main direction vector of each blood vessel branch can be determined as the unit vector Since the magnitude of the main direction vectors is In the end, the main direction vector corresponding to the upper branch and the main direction vector corresponding to the lower branch are Since the two main direction vectors are pointing in opposite directions, the dot product of the two main direction vectors has a negative value. Therefore, if the inner product value of the main direction vector is greater than or equal to the threshold, the curvature between the blood vessel branches is large. This means that the target blood vessels that should actually be extracted from the vascular image are one blood vessel. ,There is generally no large change in curvature between adjacent vascular branches, so The fact that the inner product value of the main direction vector between the branches exceeds the threshold value is actually detected by the blood vessel image processing device. The following figure shows the extraction of blood vessels that are not the target blood vessels. In correcting the blood vessel image, the blood vessel image processing device may Different thresholds can be applied.

[0079] The blood vessel image processing device divides the extracted target blood vessel into blood vessels based on branching points. The main direction vector can be obtained for each vascular branch, but the blood vessel is divided into a predetermined length. Then, the main direction vector is calculated for each piece divided into a predetermined length. The blood vessel image processing device may compare the main direction of adjacent fragments with respect to the extracted target blood vessel. By doing so, the system finds the part where the curvature is larger than the threshold value and detects the error part in the target blood vessel. In other words, the vascular image processing device may identify the extracted target blood vessel by dividing the extracted target blood vessel by a threshold length. The blood vessel image processing device can be divided into the following: In response to a case where the inner product value of the main direction vectors of the divided fragments exceeds a threshold value, The blood vessel image processing device can determine the target blood vessel by dividing it into a threshold and a fragment that is in contact with the target blood vessel. The curvature between adjacent segments can be calculated by dividing the branching points into smaller segments. The error site of the target vessel is identified by the divided vascular branch as a reference. It is possible to accurately identify the error site in the target blood vessel.

[0080] When an error portion is identified using a main direction vector in a target blood vessel, the blood vessel image processing device In response to the identification, the identified error location can be corrected. For each error site, one of the adjacent branches is separated from the remaining branches. The candidate branch may be replaced with a candidate branch whose inner product value of the main direction vector is equal to or less than a threshold value. The vascular image processing device extracts the vascular branch 1203 that has not been extracted from the target blood vessel and the upper vascular branch 1 If the inner product value of the main direction vectors 1223 and 1221 between the blood vessel Instead of the branch 1202, the blood vessel branch 1203 can be extracted as the target blood vessel. The vascular image processing device replaces the vascular branch 1202 with the vascular branch 1203 based on the extraction result of the target blood vessel. It is possible.

[0081] FIG. 13 illustrates a method for identifying and correcting errors in a target vessel using diameter information of the target vessel. vinegar.

[0082] According to one embodiment, the vascular image processing device detects a direct line between adjacent vascular branches in the target blood vessel. The difference in diameter can be calculated, and the erroneous part can be identified and corrected. responds when the difference in diameter between adjacent branches of the target vessel is equal to or greater than the threshold diameter difference The blood vessel image processing device can determine adjacent branches as erroneous regions. Diameter information is obtained for each identified branch of the blood vessel based on the branching point. For example, The diameters of the blood vessel branches 1301, 1302, and 1303 shown in FIG. 13 are r1, r2, and r3, respectively. 2, r3. For example, the blood vessel image processing device may When the blood vessel is primarily extracted as an image vessel, the difference in diameter between the blood vessel branches 1301 and 1302 is calculated. The blood vessel image processing device responds by detecting a difference in the calculated diameters that is equal to or greater than a threshold diameter difference. At least one of the blood vessel branches 1301 and 1302 described above is connected to the target blood vessel. Based on the extraction results, it is judged as a false positive. The diameter changes between adjacent branches of the target vessels are relatively small, and the threshold Adjacent vessel branches that exhibit a diameter difference greater than or equal to the diameter difference are likely to be error sites.

[0083] The vascular image processing device calculates the difference in diameter between adjacent vascular branches for the extracted target blood vessel. However, the diameter information of the blood vessel branches is smoothed to calculate the difference in diameter. In the blood vessel image, adjacent blood vessels that belong to the target vessels that should actually be extracted may be extracted. It is common for blood vessel branches to exhibit relatively small diameter changes. If the target vessel contains a diseased branch, the diseased branch and the adjacent The change in diameter between adjacent vascular branches is highly evident. , multiple blood vessel branches are grouped, and the diameter information of the blood vessel branches is smoothed to calculate the diameter difference. This makes it possible to identify erroneous areas from the extracted target blood vessels.

[0084] The vascular image processing device responds when an error portion is identified using diameter information in the target blood vessel. The identified error sites can then be corrected.

[0085] The vascular image processing device detects one of the adjacent branches for the identified erroneous portion. One branch may be replaced with a candidate branch whose diameter difference with the remaining branches is less than a threshold diameter difference. Exemplarily, the vascular image processing device may include a vascular branch 130 that is not extracted from the target vessel. If the difference in diameter between the blood vessel branch 1301 and the superior blood vessel branch 1302 is less than the threshold diameter difference, Instead of 302, a blood vessel branch 1303 can be extracted as the target blood vessel. The processing device replaces the blood vessel branch 1302 with a blood vessel branch 1303 based on the extraction result of the target blood vessel. Furthermore, the blood vessel image processing device can correct the identified error portion. Depending on the type of blood vessel (e.g., LM, LAD, LCX, RCA) or the blood vessel region, Different threshold diameter differences can be applied to

[0086] FIG. 14 illustrates a method for identifying and correcting errors in a target blood vessel using brightness information of the target blood vessel. show.

[0087] According to one embodiment, the vascular image processing device detects a clear area between adjacent vascular branches in the target blood vessel. By calculating the brightness difference, it is possible to identify and correct the misidentified areas. For example, the blood vessel image processing device detects whether the brightness difference between adjacent branches of the target blood vessel is equal to or smaller than a threshold brightness difference. In response to the above, adjacent branches can be judged as misidentified sites. The blood vessel image processing device performs brightness calculation for each of the blood vessel branches identified based on the branching point. For example, the brightness of blood vessel branches can be obtained by projecting The image processing device 1400 may include a blood vessel branch 1401, 1402, and a blood vessel branch 1403. When 402 is extracted as the target blood vessel, the brightness difference between the blood vessel branches 1401 and 1402 is calculated. The blood vessel image processing device outputs a signal when the calculated brightness difference is equal to or greater than the brightness difference threshold. In response to the above, at least one of the blood vessel branches 1401 and 1402 is It can be determined that the target blood vessel is a misidentified part in the extraction results. In this case, the relative distance between adjacent vascular branches belonging to the target vessel that must actually be extracted is The difference in brightness between the adjacent pixels is relatively small, but the difference in brightness between the adjacent pixels is greater than the threshold. Branch vessels are potential sites of errors.

[0088] The blood vessel image processing device responds when an error portion is identified using brightness information in the target blood vessel. In response to the identified error portion, the blood vessel image processing device can correct the identified error portion. For the misidentified area, the difference in brightness between one of the adjacent branches and the remaining branch was Candidate branches that have a brightness difference less than a threshold value can be replaced. The bright area between the blood vessel branch 1403 that is not extracted as the target blood vessel and the upper blood vessel branch 1401 is If the brightness difference is less than the threshold brightness difference, the blood vessel branch 140 is selected instead of the blood vessel branch 1402. 3 can be extracted as the target blood vessel. The blood vessel image processing device extracts the target blood vessel from the extraction result. Therefore, the blood vessel branch 1402 can be replaced with the blood vessel branch 1403. The device may be adapted to correct the identified error sites by using a blood vessel type (e.g., LMLM). LLADLCLCXRCRCA) or apply different threshold diameter differences depending on the vessel region That is, the blood vessel image processing device can perform the primary processing based on the method of FIG. Identify error areas in the extracted target blood vessels and provide correction results for the identified error areas to the user. A pre-processing operation can be performed to provide

[0089] In accordance with an embodiment, the blood vessel image processing apparatus performs a previous Based on a model ensemble that is distinct from the model ensemble applied in the extraction results of By using the 3D model, a new extraction result can be generated for the target vessel. In this case, the vascular image processing device does not directly correct the error portion, but uses multiple stored machine learning models. One model from the model ensemble that was applied in the previous extraction results is distinct from the model ensemble. Based on the Leu ensemble, new extraction results can be generated for the target vessel.

[0090] In addition, the vascular image processing device selects a previously extracted result from among a plurality of stored machine learning models. Multiple model ensembles based on model ensembles that are distinct from the applied model ensemble For each of the ensembles, multiple new extraction results for the target vessel are generated. The vascular image processing device can extract the target blood vessels generated by the model ensemble. For each result, vascular structure data, curvature information, and diameter information of the target blood vessel are included. and identifying an error portion based on at least one of the brightness information of the target blood vessel. The vascular image processing device selects the actual blood vessels from among the generated extraction results for the plurality of target blood vessels. The extraction result closest to the target vessel that needs to be extracted is automatically selected as the target vessel. For example, the blood vessel image processing device may provide the identified error portion to the user. The extraction result for the target vessel with the smallest number of positions is compared with the target vessel that should actually be extracted. However, the extraction result may be selected as the closest to the target, but is not limited to this. The processing device extracts the selected extraction result into the target blood vessel in response to the selection input by the user. Furthermore, the blood vessel image processing device can store information on the extraction results for the plurality of target blood vessels. The vascular image processing device can provide the user with information on a plurality of target blood vessels. One of the extraction results for the target blood is input, and the extraction result designated by the input is used as the target blood. Therefore, the vascular image processing device can use different machine learning models to A pre-processing step can be performed to provide the user with the results of the correction of the target vessel used.

[0091] In one embodiment, an error portion is identified in the extraction result of a target blood vessel. In response, a connectivity score for each candidate vascular branch is calculated. Based on the y score, one of the candidate vascular branches is extracted as the target blood vessel. When an error portion is identified from the extraction result of the target blood vessel, the blood vessel image processing device Then, one of the candidate vascular branches capable of correcting the error portion is selected. The device calculates a connectivity score for each candidate vascular branch. In the blood vessel extraction result, between the blood vessel branch adjacent to any candidate blood vessel branch and the corresponding candidate blood vessel branch As a score indicating connectivity, for example, the vascular structure between the candidate vascular branch and the adjacent vascular branches may be used. The calculation is based on the degree of matching to the structural data, curvature difference, diameter difference, and brightness difference. The connectivity score may be modified by adding a candidate vessel branch to the target vessel. This corresponds to the degree to which the vessel resembles the target vessel that must actually be extracted.

[0092] For example, when a candidate vascular branch is extracted from a target blood vessel, the candidate vascular branch and the candidate vascular branch are vascular structure data for a target vessel between the branch and an adjacent connected vascular branch; At least one of the curvature information of the blood vessel, the diameter information of the target blood vessel, and the brightness information of the target blood vessel. The connectivity score can be calculated based on any one of the following, but is not limited to this. Instead, the connectivity score for each candidate vascular branch may be calculated in a variety of ways. The image processor selects one of the candidate vascular branches based on the connectivity score. For example, the vascular branch with the highest connectivity score among the candidate vascular branches can be selected as the endovascular branch. The blood vessel branch that is judged to have the smallest error is selected. The vessel branches can be extracted to the target vessel.

[0093] For example, in FIG. 9, the blood vessel image processing device extracts a blood vessel from the first node N1 , the third node N3, and the fifth node N5 are connected to each other. In this case, the blood vessel image processing device 100 determines whether the third node N3 and the fifth node N4 are connected to each other in the node connection data 910. It can be determined that the connection between the nodes N5 does not match the vascular structure data 900. The blood vessel image processing device detects at least one of the nodes N3 and N5 corresponding to the connection error. The blood vessel branch corresponding to the node is determined as an error portion. If the vascular branch corresponding to node N3 is judged as an error, it is not matched to the vascular structure data. The third node N3, which is not matched, is excluded from the node connection data and matched to the blood vessel structure data. The blood vessel image processing device adds the second node N2 corresponding to the fifth node N1 to the node connection data. If the vascular branch corresponding to node N5 is judged as an error, the vascular structure data is matched. The fifth node N3 that does not match the blood vessel structure data is excluded from the node connection data. The sixth node N6 or the seventh node N7 is added to the node connection data. The device detects candidates corresponding to the second node, the sixth node, and the seventh node that can be nodes of the target blood vessel. A connectivity score for each of the complementary vascular branches can be calculated. Based on the connectivity score for each candidate branch, the candidate branch with the smallest error is determined. Therefore, the blood vessel branch that is detected can be selected and extracted as the target blood vessel.

[0094] As a further example, in FIG. 12, a vascular image processing device is configured to process vascular branches 1201, 1202. When the target blood vessel is extracted for the first time, the curvature between the blood vessel branches 1201 and 1202 exceeds the threshold curvature. The blood vessel image processing device may determine that the adjacent branches are in an error region. Among them, one of the branches is replaced with a candidate branch whose curvature with the remaining branches is equal to or less than the threshold curvature. Unlike FIG. 12, when there are a plurality of candidate vascular branches, the vascular image processing device may A connectivity score may be calculated for each candidate vascular branch. Based on the connectivity score for each of the complementary vascular branches, the error is determined to be the smallest. It is possible to select and extract the vascular branch as the target vascular branch. The image processing device selects one of the candidate vascular branches as a target vessel in the same manner as described above. It can be extracted.

[0095] FIG. 15 is a block diagram illustrating a blood vessel image processing apparatus according to an embodiment.

[0096] The system 1500 for processing blood vessel images according to an embodiment includes a blood vessel image processing device 1510. and a blood vessel image photographing device 1520. The blood vessel image processing device 1510 includes an image receiving unit 151 1, a processor 1512, and an input / output interface 1513.

[0097] The image receiving unit 1511 receives a blood vessel image captured by a blood vessel image capturing device 1520. The image receiving unit 1511 receives an image from the blood vessel image capturing device 1520 via wired or wireless data communication. However, the present invention is not limited to this, and the image receiving unit 1511 receives the blood vessel image. The imaging device 1520 may be integrated.

[0098] The processor 1512 uses the machine learning model to calculate the blood vessel image received from the image receiving unit. A target blood vessel is extracted, and a calculation is performed on the extracted result of the target blood vessel based on a first input received from a user. and determining a region of interest within the determined region of interest based on a second input received from the user. It is possible to identify errors in the extraction results and correct the identified errors. The operation of the processor 1512 is not limited to this, and the processor 1512 may be any of the processors shown in FIGS. The above-mentioned operations can be performed with reference to the above.

[0099] The input / output interface 1513 receives input from the user and communicates it to the processor. For example, the input / output interface 1513 can handle mouse operations and touch operations. The input / output interface 1513 also receives inputs from visual feedback. For example, the input / output interface 1513 may provide feedback to the user. The results of blood vessel extraction are output on a display for each step.

[0100] The national research and development projects that supported this invention are as follows:

[0101] [Project unique number]1415166912 [Subject number] 20001638 [Department name] Ministry of Trade, Industry and Energy - Ministry of Science, ICT and Communication - Ministry of Health and Welfare - Food and Drug Administration [Name of issue management (specialized) organization] Korea Institute for Industrial Technology Evaluation and Management [Research project name] Artificial Intelligence Bio-Robot Medical Fusion Project [Research title] Cardiac intervention treatment assistant artificial intelligence and semi-automatic treatment tool based on cardiovascular big data Development of a wheel control robot system [Supervising Institution] Seoul Asan Hospital [Research period] May 1, 2018 to December 31, 2020 [Project unique number]1711117085 [Subject number] 2020-O-00159 [Department name] Ministry of Science, ICT and Technology [Name of issue management (specialized) organization] Information and Communications Planning and Evaluation Agency [Research Project Name] Artificial Intelligence Fusion Pioneer Project [Research title] Artificial intelligence-based automated cardiovascular disease diagnosis assistance and treatment tool recommendation system Stem Development [Supervising agency] Medipixel Inc. [Research period] April 1, 2020 to December 31, 2021 [Project unique number]1425133196 [Project number] S2758883 [Department name] Small and Medium-sized Venture Business Department [Name of issue management (specialized) organization] Small and Medium Enterprise Technology Information Promotion Agency [Research Project Name] Start-up and Growth Technology Development (R&D) [Research title] Development of artificial intelligence-based support system for recanalization of completely occluded lesions [Supervising agency] Medipixel Inc. [Research period] June 1, 2019 ~ May 31, 2020 [Project unique number]1711106903 [Project number] 2020R1C1C1010470 [Department name] Ministry of Science, ICT and Technology [Name of issue management (specialized) organization] Korea Research Foundation [Research project name] Emerging researcher training support project [Research title] Cardiovascular imaging using deep learning-based intravascular ultrasound image segmentation technology Diagnostics Development [Supervising Institution] University of Ulsan [Research period] March 1, 2020 ~ February 28, 2021

[0102] The above-described embodiments may be implemented using hardware components, software components, or The present embodiment is implemented as a combination of hardware and software components. The devices and components described in the embodiments may include, for example, a processor, a controller, an ALU (ar ithmetic logic unit, digital signal processor signal processor), microcomputer, FPA (field programmable array), PLU(programmable log ic unit), microprocessor, or instruction execution One or more general-purpose or special-purpose computers, such as different devices that respond to The processor is implemented using an operating system (OS) and One or more software applications that run on the operating system. In addition, the processing unit may access, store, manipulate, process, and / or modify data in response to execution of the software. For ease of understanding, the processing device will be described as being used in the following. Although there are cases where a processor is a processor that includes multiple processing elements, those skilled in the art will appreciate that It is understood that the processing element and / or multiple types of processing elements may be included. For example, a processing unit may include multiple processors or one processor and one controller. Also includes other processors such as parallel processors. The processing configuration is also possible.

[0103] Software means any computer program, code, instructions, or one or more of these. The term "processor" includes any combination of the above, and may be used to configure a processor to operate as desired, or to process independently or in combination. The software and / or data can be interpreted by the processor and can instruct the processor. Any type of machine, component, or other device that can be used to interpret or provide instructions or data to a processor Any element, physical device, virtual device, computer storage medium or device, or transmitted signal wave that is permanent The software may be embodied in a network-linked computer. The software may be distributed across multiple computer systems and stored and executed in a distributed fashion. The software and data may be stored on one or more computer readable storage media.

[0104] The method according to the present embodiment is implemented by program instructions implemented via various computer means. The present invention is embodied in the form of a computer readable recording medium. This includes, alone or in combination, program instructions, data files, data structures, etc. The computer and program instructions may be those specially designed and constructed for the purposes of the present invention. Those skilled in the art of computer software technology will recognize and use the following: Examples of the computer-readable recording medium include a hard disk. , magnetic media such as floppy disks and magnetic tapes, CD-ROMs, D Optical recording media such as VDs, magneto-optical media such as floptical disks, and ROMs; RAM, flash memory, etc., specially designed to store and execute program instructions Examples of program instructions include hardware devices configured by a compiler. Not only machine code generated by a computer, but also code written by a computer using an interpreter, etc. The hardware devices described above include the high level language code that is executed by the configured to operate as one or more software modules to perform the operations of and vice versa.

[0105] Although the embodiments have been described above with reference to the drawings, the present invention is not limited to the above embodiments. Those with ordinary skill in the art can apply various technical modifications and variations based on the above description. For example, the techniques described may be performed in a different order than described, and and / or that the components of the described system, structure, device, circuit, etc., differ from the manner described. may be combined or combined in any form and may be replaced by other components or equivalents. or may be substituted and still achieve suitable results.

Claims

1. A blood vessel image processing method performed by a processor, comprising: extracting a target blood vessel from the blood vessel image; determining a region of interest (ROI) for the target blood vessel extraction result; identifying a first error portion from the extraction result of the target blood vessel in the determined region of interest, and providing a correction result of the first error portion to a user; responsive to input received from the user being detected from a plurality of points within the region of interest, identifying a region of interest in the extraction result based on the plurality of points; identifying the target region as a second error site in response to a point corresponding to the input leaving the target vessel from a branch point within the region extracted from the target vessel; correcting the second error portion of the extraction result using a blood vessel branch selected based on the target region from among blood vessel branches connected to the branch point; A blood vessel image processing method comprising:

2. The blood vessel image processing method according to claim 1 , wherein the step of determining the region of interest includes the step of determining a shape of the region of interest based on a point at which another input is detected.

3. The blood vessel image processing method includes: detecting whether a discontinuity exists at a point through which the input passes while the point corresponding to the input moves along a region extracted from the target vessel; in response to the discontinuity being detected, connecting the discontinuities based on the identified regions of interest; The blood vessel image processing method according to claim 1 , further comprising:

4. The step of identifying the first error portion and providing a correction result of the first error portion to a user includes: in response to the first error location being a misidentified location, providing one or more candidate branches to a user; replacing the branch corresponding to the first error site with a selected branch in response to receiving a pointing input from the user to any one of the one or more candidate branches; The blood vessel image processing method according to claim 1 , further comprising:

5. 2. The vascular image processing method of claim 1, wherein the step of identifying the first error portion and providing the correction result of the first error portion to the user includes a step of connecting a region corresponding to a vascular branch corresponding to a start point of the user input and a region corresponding to a vascular branch corresponding to an end point of the user input in response to a case where user input is detected at a plurality of points for the correction result.

6. 2. The vascular image processing method of claim 1, wherein the step of identifying the first error portion and providing a user with a correction result for the first error portion includes a step of identifying the first error portion based on at least one of vascular structure data regarding the target blood vessel, curvature information of the target blood vessel, diameter information of the target blood vessel, and brightness information of the target blood vessel in response to the region of interest being determined.

7. 2. The vascular image processing method of claim 1, wherein the step of identifying the first error portion and providing the user with a correction result of the first error portion includes a step of providing the user with a correction result of correcting the first error portion based on at least one of vascular structure data regarding the target blood vessel, curvature information of the target blood vessel, diameter information of the target blood vessel, and brightness information of the target blood vessel in response to the first error portion being identified.

8. The step of identifying the first error portion and providing a correction result of the first error portion to a user includes: generating a plurality of new extraction results for the target vessel in response to the first error location being identified; in response to a user selection input, selecting one of the plurality of new extraction results indicated by the selection input; The blood vessel image processing method according to claim 1 , comprising:

9. A blood vessel image processing device, an image receiving unit for receiving a blood vessel image; a processor for extracting a target vessel from the vascular image, determining a region of interest (ROI) for an extraction result of the target vessel, identifying a first error portion from the extraction result of the target vessel in the determined region of interest, and providing a correction result of the first error portion to a user, identifying a target region in the extraction result based on a plurality of points in response to an input received from the user being detected from the plurality of points in the ROI, identifying the target region in response to a point corresponding to the input leaving the target vessel from a branch point in the region extracted from the target vessel, and correcting the second error portion in the extraction result using a vascular branch selected based on the target region from vascular branches connected to the branch point; A blood vessel image processing device comprising:

Images