Method, apparatus, and computer program for generating artificial intelligence model for predicting blood flow data

Abstract

The present invention relates to a method by which a server generates an artificial intelligence model for predicting blood flow data, comprising the steps of: setting a structure of a blood flow prediction model; generating blood vessel data in the form of a point cloud corresponding to a main blood vessel from a three-dimensional medical image, specifying a boundary area of the blood vessel data, obtaining a blood flow value of the boundary area, and preparing, as training data of the blood flow prediction model, the blood flow value of the boundary area and a location of the point cloud in an arbitrary area; and training the blood flow prediction model by setting a blood flow value of an area other than the boundary area as a label or a mask.

Description

Method, device and computer program for generating an artificial intelligence model for predicting blood flow data

[0001] The present invention relates to a method, device and computer program for generating an artificial intelligence model for predicting blood flow data.

[0002] In modern medicine, medical imaging is a crucial tool for effective disease diagnosis and patient treatment. Furthermore, advancements in imaging technology have led to the generation of increasingly sophisticated medical image data. Consequently, the volume of data has grown exponentially, making it increasingly difficult to analyze medical image data solely through human visualization. Therefore, over the past decade, clinical decision support systems and computer-aided interpretation systems have played an essential role in the automated analysis of medical images.

[0003] Conventional clinical decision support systems or computer-aided reading systems perform the function of detecting and displaying lesion areas or providing reading information to medical staff or medical workers (hereinafter referred to as users).

[0004] For example, in the 'Method and device for producing disease diagnosis information based on medical images' disclosed in Korean Patent Publication No. 10-2017-0017614, the method includes detecting a region of interest in which an object to be analyzed is photographed, calculating a coefficient of variation, creating a coefficient of variation image, and comparing it with a reference sample, and mentions the effect of diagnosing the degree of a patient's disease by utilizing medical images acquired through a computed tomography device, an ultrasound imaging device, etc.

[0005] Recently, artificial intelligence (AI) technologies based on machine learning, such as deep learning, have been making rapid progress in the field of medical image analysis and processing. Deep learning in medical imaging is being applied to a wide range of tasks, including image analysis, diagnosis, treatment, and prediction. For example, deep learning-based auxiliary diagnostic systems are being applied to diagnose cerebrovascular diseases, such as cerebral aneurysms.

[0006] The present invention provides a method for generating an artificial intelligence model for predicting blood flow data. Furthermore, the present invention provides a system for predicting blood flow data using a blood vessel graph and the artificial intelligence model, and a method for executing the same. Furthermore, the present invention provides a method for generating a blood vessel graph containing coordinates and structural information for major blood vessels from a three-dimensional blood vessel image.

[0007] The objectives of the present invention are not limited to those mentioned above. Other objectives and advantages of the present invention not mentioned above can be understood through the following description and will be more clearly understood through the embodiments of the present invention. Furthermore, it will be readily apparent that the objectives and advantages of the present invention can be realized by the means and combinations thereof set forth in the claims.

[0008] In a server according to an embodiment of the present invention, a method for generating an artificial intelligence model for predicting blood flow data is characterized by including the steps of: setting a structure of a blood flow prediction model; generating blood vessel data in the form of a point cloud corresponding to a major blood vessel from a three-dimensional medical image, specifying a boundary region of the blood vessel data, obtaining a blood flow value of the boundary region, and providing a location of a point cloud for an arbitrary region and a blood flow value of the boundary region as learning data of the blood flow prediction model; and setting a blood flow value of a region other than the boundary region as a label or mask to learn the blood flow prediction model.

[0009] According to the present invention as described above, an artificial intelligence model for predicting blood flow data can be created. Furthermore, according to the present invention, a blood vessel graph containing coordinates and structural information for major blood vessels can be created from a three-dimensional blood vessel image, and blood flow data can be calculated using the blood vessel graph and a blood flow prediction artificial intelligence model.

[0010] FIG. 1 is a diagram illustrating a blood flow data prediction system according to one embodiment of the present invention.

[0011] FIG. 2 is a schematic diagram illustrating an operation of a blood flow data prediction system according to one embodiment of the present invention.

[0012] FIG. 3 is a diagram for explaining an operation of a blood flow data prediction system according to one embodiment of the present invention.

[0013] FIG. 4 is a flowchart illustrating one embodiment of a three-dimensional medical image processing method according to one embodiment of the present invention.

[0014] FIGS. 5 to 8 are exemplary diagrams for explaining the operation of a three-dimensional medical image processing device according to one embodiment of the present invention.

[0015] FIG. 9 is a flowchart illustrating one embodiment of a method for generating a blood vessel graph according to one embodiment of the present invention.

[0016] Figures 10 to 14 are exemplary diagrams for explaining the operation of a blood vessel graph generation device according to one embodiment of the present invention.

[0017] FIG. 15 is a diagram for explaining a method for generating an artificial intelligence model for blood flow prediction according to one embodiment of the present invention.

[0018] Figures 16 and 17 are drawings for explaining the structure and operation of an artificial intelligence model for blood flow prediction according to one embodiment of the present invention.

[0019] FIG. 18 is a block diagram illustrating the configuration of a server that generates an artificial intelligence model for predicting blood flow according to an embodiment of the present invention.

[0020]

[0021] The above-described objects, features, and advantages will be described in detail below with reference to the accompanying drawings, so that those skilled in the art can easily practice the technical idea of ​​the present invention. In describing the present invention, if it is determined that a detailed description of known technologies related to the present invention may unnecessarily obscure the gist of the present invention, a detailed description thereof will be omitted. Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. In the drawings, the same reference numerals are used to indicate the same or similar components.

[0022]

[0023] FIG. 1 is a drawing for explaining a blood flow prediction system according to one embodiment of the present invention.

[0024] Referring to FIG. 1, the blood flow prediction system includes a 3D medical image processing device (100), a blood vessel graph generation device (200), and a blood flow prediction device (300). Although FIG. 1 illustrates the blood flow prediction system as including multiple devices, this is for convenience of explanation and the present invention cannot be interpreted as being limited thereto. That is, the blood flow prediction system according to an embodiment of the present invention may have modules performing each function executed in a single or multiple devices, or may be executed in a virtualized environment.

[0025] A 3D medical image processing device (100) can generate a 3D blood vessel image of a major blood vessel from a 3D medical image.

[0026] First, when a 3D medical image processing device (100) receives a 3D medical image, it can analyze the pixel values ​​of the 3D medical image to extract blood vessel pixel points and coordinates. For example, when a certain pixel value in the 3D medical image is greater than a preset threshold value, the 3D medical image processing device (100) can determine that the pixel value corresponds to a blood vessel, identify the blood vessel pixel point, and extract coordinate information of the blood vessel pixel point.

[0027] In the above embodiment, the 3D medical image processing device (100) can extract the coordinates of a pixel if the pixel value is greater than a predetermined threshold value and create a blood vessel pixel coordinate table using the pixel coordinates. Thereafter, the 3D medical image processing device (100) can display blood vessel points on a coordinate system based on the blood vessel pixel coordinate table to create a coordinate system of the entire blood vessel.

[0028] A 3D medical image processing device (100) can crop a major blood vessel portion from a blood vessel pixel point displayed on a coordinate system.

[0029] Thereafter, the 3D medical image processing device (100) can perform clustering on blood vessel pixel points and filter out noise and / or microvascular pixels.

[0030] According to an embodiment of the present invention, a three-dimensional medical image processing device (100) can group blood vessel pixel points into clusters by performing clustering based on at least one of distance, density, and / or attribute information of blood vessel pixel points.

[0031] Thereafter, the 3D medical image processing device (100) can perform a first filtering operation to compare the number of blood vessel pixel points included in any cluster with a threshold value and delete clusters whose number of blood vessel points is less than the threshold value. The filtering operation can be performed by removing pixel points corresponding to clusters whose number of blood vessel points is less than the threshold value from the blood vessel coordinate system. This can remove noise around major blood vessels.

[0032] The 3D medical image processing device (100) can group the blood vessel pixel points into clusters by performing clustering again based on at least one of the distance, density, and attribute information of the first filtered blood vessel pixel points.

[0033] Thereafter, the 3D image processing device (100) can perform a second filtering operation that deletes all clusters except for the blood vessel pixel points included in a random cluster. For example, the filtering operation can be performed by removing pixel points corresponding to clusters other than the largest cluster from the blood vessel coordinate system. This can remove microvessels.

[0034] According to another embodiment of the present invention, a three-dimensional medical image processing device (100) can perform clustering and / or filtering of blood vessel pixel points using various algorithms, and the present invention cannot be interpreted by limiting the clustering and / or filtering algorithms.

[0035] For example, clustering can be performed by grouping blood vessel pixel points into K clusters, calculating the center of each cluster, and assigning the blood vessel pixel points to the cluster with the closest center. Alternatively, clusters can be formed using a similarity matrix for blood vessel pixel points. Subsequently, filtering for noise and microvessels around blood vessels can be performed by modifying or removing clusters.

[0036] According to an embodiment of the present invention, when a 3D medical image processing device (100) performs filtering through clustering, blood vessel pixel points with noise and microvessels removed are formed in a blood vessel coordinate system, and through this, the 3D medical image processing device (100) can generate a 3D blood vessel image.

[0037] The blood vessel graph generation device (200) can generate a blood vessel graph based on a three-dimensional blood vessel image generated by a three-dimensional medical image processing device (100).

[0038] The blood vessel graph generation device (200) can divide a three-dimensional blood vessel image into multiple layers and check the coordinate values ​​of blood vessel pixel points for each layer. At this time, it is appropriate to generate the layers in two dimensions.

[0039] Thereafter, the blood vessel graph generation device (200) can group blood vessel pixel points for each layer. The blood vessel pixel point group of each layer may correspond to a cross-section of an actual blood vessel.

[0040] Thereafter, the vascular graph generation device (200) can calculate the center of gravity of each group. This is for generating nodes of a vascular graph, and nodes of a vascular graph according to an embodiment of the present invention can be set as the center of gravity of a group of vascular pixel points for each layer. In a vascular graph according to an embodiment of the present invention, nodes can correspond to the centers of actual blood vessels, and it is appropriate to set the radius of nodes of the vascular graph to the radius of the actual blood vessel.

[0041] Thereafter, the vascular graph generation device (200) can generate connectivity information of vascular graph nodes. More specifically, connection information can be generated for nodes for the first group and nodes for the second group based on whether there is an intersection between a first group of vascular pixel points of an arbitrary layer and a second group of vascular pixel points of another layer adjacent to the layer. The vascular pixel point group of each layer corresponds to a cross-section of an actual blood vessel, and if there is an intersection with an adjacent layer, it means that the actual blood vessel is connected.

[0042] Through this, the vascular graph generation device (200) can generate node and connection information from a 3D vascular image and generate a vascular graph. Furthermore, the vascular graph generation device (200) can calculate the actual vascular radius from the vascular graph and set it as the node radius. Through this, the vascular graph created according to an embodiment of the present invention can reflect characteristics such as branching, connection, size, and location of the blood vessels. A specific method for this will be described later in the description of the attached drawings.

[0043] The blood flow prediction device (300) can predict blood flow at any point based on a blood vessel graph. To this end, according to an embodiment of the present invention, the blood flow prediction device (300) can generate a blood flow prediction model that predicts the pressure at any node using the location and radius information of each node in the blood vessel graph of the region of interest. A method for generating a blood flow prediction model according to an embodiment of the present invention will be described below with reference to the attached drawings.

[0044] Once the blood flow prediction model is prepared, the blood flow prediction device (300) can specify a blood vessel graph of the region of interest. The blood flow prediction device (300) can distinguish between a boundary node located at the end of the blood vessel graph of the region of interest and other nodes located inside, and can obtain the pressure value (p) of the boundary node together with the location (x, y, z) and node radius (r) information of the boundary node and other nodes and input them into the model.

[0045] The blood flow prediction model can predict and output the pressure of an arbitrary node (query node), and the blood flow prediction device (300) can predict the blood flow of an arbitrary point.

[0046]

[0047] FIG. 2 is a schematic diagram illustrating an operation of a blood flow data prediction system according to one embodiment of the present invention.

[0048] A blood flow prediction system according to one embodiment of the present application can predict blood flow data of an arbitrary area using an artificial intelligence model (blood flow prediction model) that predicts blood flow.

[0049] More specifically, the blood flow prediction system can extract a blood vessel graph from a 3D medical image. The blood vessel graph is in the form of a point cloud and can be configured in the form of a 1D network connecting points (or nodes), and can include coordinate values ​​composed of the x-coordinate, y-coordinate, and z-coordinate of the nodes constituting the blood vessel, and information on the radius (r) of the node corresponding to the actual blood vessel radius.

[0050] Furthermore, the blood flow prediction system can input a vascular graph into a trained blood flow prediction model. The training and structure of the blood flow prediction model are described in more detail below with reference to the attached Figures 15 to 17.

[0051] A blood flow prediction system can predict blood flow data, such as blood pressure, at an arbitrary point through the output value of a blood flow prediction model. For example, when an arbitrary region is specified in a blood vessel graph, the blood flow prediction system can identify the boundary of the region, obtain the location (x, y, z) and node radius (r) information of the boundary node and other nodes in the blood vessel graph, and obtain the pressure value (p) of the boundary node through measurement or calculation and input it into the blood flow prediction model. The blood flow prediction model can predict the pressure of an arbitrary node (query node) and output it, and through this, the blood flow prediction system can predict blood flow at an arbitrary point.

[0052]

[0053] FIG. 3 is a diagram for explaining an operation of a blood flow data prediction system according to one embodiment of the present invention.

[0054] A blood flow prediction system according to one embodiment of the present application can acquire a 3D medical image, extract blood vessel pixel points, and generate a 3D blood vessel image. Then, a region of interest can be specified through user input, and a blood vessel graph of the region of interest can be generated.

[0055] A blood flow prediction system according to one embodiment of the present invention can acquire a 3D medical image, extract blood vessel pixel points to create a 3D blood vessel image, and use the 3D blood vessel image to create a blood vessel graph. Then, a region of interest can be specified through user input and extracted from the previously created blood vessel graph.

[0056] The blood vessel graph of the region of interest is in the form of a point cloud and can be configured in the form of a 1D network connecting points (or nodes), and can include coordinate values ​​composed of the x-coordinate, y-coordinate, and z-coordinate of the nodes constituting the blood vessel, and information on the radius (r) of the node corresponding to the actual blood vessel radius.

[0057] Furthermore, the blood flow prediction system can obtain actual blood flow data (boundary value p in Fig. 3) at the boundary of the region of interest. The actual blood flow data at the boundary of the region of interest may be, for example, blood pressure data at the boundary of the region of interest, which may be measured values ​​or predicted values ​​calculated using an algorithm.

[0058] The blood flow prediction system can predict blood flow data for an arbitrary region by inputting a region of interest blood vessel graph and boundary values ​​into a trained blood flow prediction model. For example, the blood flow prediction system can input the boundary value (p) along with the location of the boundary node and other nodes and the node radius (x, y, z, r) in the region of interest blood vessel graph into the blood flow prediction model. If an arbitrary point in the region of interest is specified as a query node, the blood flow prediction model can predict the blood flow data (P) of the query node, and through this, the blood flow prediction system can predict blood flow, such as pressure, blood flow velocity, and flow rate, at an arbitrary point.

[0059]

[0060] FIG. 4 is a flowchart for explaining an embodiment of a three-dimensional medical image processing method according to an embodiment of the present invention, and FIGS. 5 to 8 are exemplary diagrams for explaining the operation of a three-dimensional medical image processing device according to an embodiment of the present invention.

[0061] Referring to FIG. 4, a 3D medical image processing device (100) can analyze a 3D medical image and extract pixel coordinates corresponding to a blood vessel based on pixel values ​​(step S410).

[0062] In one embodiment of step S410, the 3D medical image processing device (100) analyzes the 3D medical image and determines that a blood vessel exists if any pixel value exceeds a predetermined threshold value, thereby specifying a blood vessel pixel point and extracting coordinate information of the blood vessel pixel point. Thereafter, based on the coordinate information, the blood vessel point can be displayed on a coordinate system to generate a coordinate system of the entire blood vessel.

[0063] For example, a 3D medical image processing device (100) can analyze a 3D medical image of (a) of FIG. 5, extract pixel coordinates corresponding to blood vessels based on pixel values ​​as in (b) of FIG. 5, and then generate a pixel coordinate table as in (a) of FIG. 6. A coordinate system of the entire blood vessel can be generated based on the pixel coordinate table as in (b) of FIG. 6. That is, a coordinate system of the entire blood vessel can be generated by forming blood vessel pixel points on the coordinate system.

[0064] The 3D medical image processing device (100) can crop a major blood vessel portion from the blood vessel pixel points displayed on the coordinate system (step S420). For example, the 3D medical image processing device (100) can specify a major blood vessel portion to be the target of image processing, as shown in (b) of FIG. 7, from the pixel points of the entire blood vessel displayed on the coordinate system, as shown in (a) of FIG. 7.

[0065] Furthermore, the 3D medical image processing device (100) can perform clustering and filtering on blood vessel pixel points in the crop area to remove noise and / or microvascular pixels. (Step S430) Through this, a 3D blood vessel image can be generated. (Step S440)

[0066] For example, the 3D medical image processing device (100) may perform clustering based on at least one of distance, density, and / or attribute information of blood vessel pixel points to group blood vessel pixel points into clusters. Thereafter, filtering may be performed to compare the number of blood vessel points included in any cluster with a threshold value and delete clusters having a number of blood vessel points less than or equal to the threshold value. Filtering may be performed by removing pixel points corresponding to clusters having a number of blood vessel points less than or equal to the threshold value from the blood vessel coordinate system. This may remove noise around major blood vessels.

[0067] Furthermore, the 3D medical image processing device (100) can group the vascular pixel points into clusters by re-clustering the primary filtered vascular pixel points based on at least one of their distance, density, and attribute information. Thereafter, filtering can be performed to exclude the vascular pixel points included in any cluster and delete the remaining clusters. For example, the filtering can be performed by removing pixel points corresponding to clusters other than the largest cluster from the vascular coordinate system. This can remove microvessels.

[0068] For example, a 3D medical image processing device (100) can generate a 3D blood vessel image as shown in (b) of FIG. 8 by removing noise and microvascular pixels from a blood vessel that is the target of image processing as shown in (a) of FIG. 8.

[0069] Fig. 9 is a flowchart illustrating one embodiment of a method for generating a blood vessel graph according to the present invention. Figs. 10 to 14 are exemplary diagrams illustrating the execution process of Fig. 9.

[0070] Referring to FIG. 9, the blood vessel graph generation device (200) can divide a three-dimensional blood vessel image into multiple layers and check the coordinate values ​​of blood vessel pixel points for each layer. At this time, it is appropriate to create the layers in two dimensions (step S910). For example, the blood vessel graph generation device (200) can divide a three-dimensional blood vessel image into multiple layers as in (a) of FIG. 10 and check the coordinate values ​​of blood vessel pixel points for each layer as in (b) of FIG. 10.

[0071] Thereafter, the vascular graph generation device (200) can group vascular pixel points for each layer (step S920) and calculate the center of gravity of each group. This is for generating nodes of a vascular graph, and nodes of a vascular graph according to an embodiment of the present invention can be set as the center of gravity of a group of vascular pixel points for each layer (step S930).

[0072] For example, the blood vessel graph generation device (200) can group blood vessel pixel points in an arbitrary layer and form the center of gravity of the group as a node of the blood vessel graph, as shown in FIG. 11. In the example of FIG. 11, the blood vessel pixel point group corresponds to a cross-section of an actual blood vessel, and the center of gravity of the group will correspond to the center of the cross-section of the blood vessel. Therefore, it is appropriate to set the radius of the node of the blood vessel graph to the radius of the actual blood vessel.

[0073] Afterwards, the blood vessel graph generation device (200) can generate connection information (connectivity) of blood vessel graph nodes. (Step S940)

[0074] For example, the blood vessel graph generation device (200) can check the group of blood vessel pixel points of the i-th layer and the group of blood vessel pixel points of the i+1-th layer adjacent to the i-th layer, as shown in FIG. 12, and if there is an intersection between the groups, as shown in FIG. 13, connection information can be generated for the nodes. This is because the blood vessel pixel point group of each layer corresponds to a cross-section of an actual blood vessel, and if there is an intersection with an adjacent layer, it is interpreted that the actual blood vessel is connected.

[0075] Furthermore, the blood vessel graph generation device (200) can calculate the radius of an actual blood vessel in the blood vessel graph and set it as a node radius. (Step S950)

[0076] Referring to Fig. 14, when a 3D blood vessel image is divided into 2D layers, the cross-sectional radius corresponds to the radius (R0) of the blood vessel pixel point group. That is, since it is not the actual radius of the blood vessel but the radius (R0) of the blood vessel pixel point group, it is necessary to calculate the actual radius (r) of the blood vessel. To this end, the radius (R0) of the blood vessel pixel point group can be converted by cosθ, and the actual radius (r) can be calculated by [Mathematical Formula 1] below.

[0077] [Mathematical Formula 1]

[0078] r = R0Х cosθ

[0079] R0: Radius of the blood vessel pixel point group,

[0080]

[0081] *r: actual radius of the blood vessel,

[0082] cosθ: It is calculated by [Mathematical Formula 2], and θ is the direction of connection information of adjacent layers ( ) and the radial direction of the blood vessel pixel point group of that layer ( ) is calculated using.

[0083] [Equation 2]

[0084]

[0085] : Length direction calculated from connection information, : radial direction of the blood vessel cross-section, or

[0086] : Perpendicular to the length direction calculated from the connection information, : Vertical of the blood vessel pixel point group

[0087] Returning to the description of FIG. 9, in step S960, the vascular graph generation device (200) can generate a vascular graph using nodes, connection information, and node radii. This is a visual representation of the vascular structure, and information on characteristics such as branching, connections, size, and location of the vascular is compressed, and parameters such as length, diameter, angle, branching pattern, and shape of the vascular are included.

[0088]

[0089] FIG. 15 is a drawing for explaining a method for generating an artificial intelligence model for blood flow prediction according to one embodiment of the present invention, and FIGS. 16 and 17 are drawings for explaining the structure and operation of an artificial intelligence model for blood flow prediction according to one embodiment of the present invention.

[0090] Referring to FIG. 15, the blood flow prediction system can collect data for training an artificial intelligence model for blood flow prediction. (Step S1501)

[0091] The learning data may be a blood vessel graph for an arbitrary region, as shown in Fig. 16. The blood vessel graph is in the form of a point cloud and may be configured in the form of a 1D network connecting points (or nodes), and may include a boundary node (1601) corresponding to the boundary of the region and other nodes (1602) that are not boundary nodes. Furthermore, the blood vessel graph may include coordinate values ​​composed of the x-coordinate, y-coordinate, and z-coordinate of each node, and information on the radius (r) of the node corresponding to the actual blood vessel radius. The learning data includes the pressure (p) of the boundary node (1601), and the pressure (p) of other nodes may be set as a label or mask.

[0092] Furthermore, the blood flow prediction system can set an input layer of a blood flow prediction model. (Step S1502)

[0093] The input layer of the blood flow prediction model according to the embodiment of the present invention can be set to a first channel, a second channel, and a third channel, as shown in 1701 of Fig. 17. The first channel can input the coordinate value and radius information (x, y, z, r) of the query node in the blood vessel graph, the second channel can input the coordinate value and radius information (x, y, z, r) of the boundary node in the blood vessel graph, and the third channel can input the pressure value (p) of the boundary node.

[0094] Furthermore, the blood flow prediction system can set a feature extraction layer of the blood flow prediction model. (Step S1503)

[0095] The feature extraction layer of the blood flow prediction model according to an embodiment of the present invention can be referred to as 1702 of FIG. 17.

[0096] In particular, the feature extraction layer (1702) can separate the pressure value (p) of the boundary node input from the third channel into four values ​​corresponding to (x, y, z, r) to reflect the correlation between the coordinate value of the node and the radius information (x, y, z, r). Furthermore, the feature extraction layer 1902 can pass the (x, y, z, r) input from each channel or the pressure (p) separated into four to the fully connected layer (FC1, FC2) that shares the weight.

[0097] Through this, the feature extraction layer (1702) can form a local feature (1705) that reflects the characteristics of the query node and a boundary feature (1706) that reflects the characteristics of the boundary node. At this time, it is appropriate to apply a max pooling layer (1707) to remove the input order information of the boundary node from the boundary feature (1706).

[0098] Furthermore, the blood flow prediction system can set a feature stitching layer. (Step S1504)

[0099] The feature stitching layer of the blood flow prediction model according to an embodiment of the present invention may have a structure that combines local features and boundary features and passes them to a fully connected layer (FC1, FC2), as shown in 1703 of FIG. 17. This is to enable the model to understand global features for all channels.

[0100] Furthermore, the blood flow prediction system can set an output layer of the blood flow prediction model. (Step S1505) The output layer of the blood flow prediction model according to an embodiment of the present invention can be set to predict and output the pressure of the query node, as shown in 1704 of FIG. 17.

[0101] Once the learning data for the blood flow prediction model is prepared and the model structure setting is completed, the blood flow prediction system can train the blood flow prediction model. (Step S1506)

[0102] Once the generation of the blood flow prediction model is completed, the blood flow prediction system can specify a blood vessel graph of the region of interest, distinguish a boundary node located at the end of the specified blood vessel graph from other nodes located inside, and obtain the pressure value (p) of the boundary node together with the location (x, y, z) and node radius (r) information of the boundary node and other nodes and input them into the model. The blood flow prediction model can predict the pressure of an arbitrary node (query node) and output it, through which the blood flow prediction system can predict the blood flow at an arbitrary point.

[0103] FIG. 18 is a block diagram illustrating the configuration of a server that generates an artificial intelligence model for predicting blood flow according to an embodiment of the present invention.

[0104] As illustrated in FIG. 18, a server (1800) according to an embodiment of the present invention may include a communication unit (1810), a storage unit (1830), and a control unit (1820), and although not illustrated in FIG. 18, may further include an input unit and a display unit. In FIG. 18, the server (1800) is illustrated as including a communication unit (1810), a storage unit (1830), and a control unit (1820), but each block may be physically separated. For example, the storage unit (1830) may exist in a virtualized data center and be connected to the control unit (1820) of the server (1800) via the communication unit (1810).

[0105] The communication unit (1810) performs data transmission and reception functions for wired and wireless communication of the server (1800), receives data through a wired or wireless channel, outputs the data to the control unit (1820), and transmits the data output from the control unit (1820) through a wireless channel. In particular, the communication unit (1810) according to an embodiment of the present invention can perform a function of receiving a 3D medical image.

[0106] The storage unit (1830) serves to store programs and data required for the operation of the server (1800), and may be divided into a program area and a data area. The data area of ​​the storage unit (1830) according to an embodiment of the present invention may store blood vessel pixel point coordinates, 3D blood vessel images and blood vessel graphs, blood flow values ​​at any point in the blood vessel, for example, pressure values.

[0107] Furthermore, the program area of ​​the storage unit (1830) according to an embodiment of the present invention can store an artificial intelligence model that predicts blood flow at an arbitrary point in a blood vessel image. The artificial intelligence model can perform a function of predicting pressure at an arbitrary node using the location and radius information of each node in the blood vessel graph of the region of interest.

[0108] Furthermore, the program area of ​​the storage unit (1830) according to an embodiment of the present invention can store a computer program that executes a process of generating a blood vessel graph on a server. The computer program can perform a function of analyzing a 3D medical image to extract blood vessel pixel point coordinates corresponding to blood vessels based on pixel values, removing noise around major blood vessels, and generating a 3D blood vessel image, and a function of dividing the 3D blood vessel image into a plurality of layers, setting a blood vessel pixel point group for each layer as a blood vessel node, and generating connection information to generate a blood vessel graph.

[0109] The control unit (1820) controls the overall operation of each component of the server (1800). In particular, the control unit (1820) according to an embodiment of the present invention can set the structure of a blood flow prediction model and generate data for model learning.

[0110] The control unit (1820) can generate blood vessel data in the form of a point cloud corresponding to a major blood vessel from a 3D medical image, specify a boundary area of ​​the blood vessel data, obtain a blood flow value of the boundary area, and generate learning data for a model by specifying the location of the point cloud for an arbitrary area and the blood flow value of the boundary area.

[0111] The control unit (1820) can set the structure of the model, and the model can include an input layer, a feature extraction layer, an output layer, etc. The control unit (1820) can set the input layer of the blood flow prediction model to include a channel into which location and radius information of a boundary node in a blood vessel graph of a region of interest are input, a channel into which location and radius information of other nodes are input, and a channel into which blood flow values ​​of the boundary nodes are input. Furthermore, the feature extraction layer can be set to form a boundary feature that reflects local features that reflect the characteristics of other nodes and the characteristics of boundary nodes, and from which input order information of the boundary nodes is removed.

[0112] When learning data of a blood flow prediction model is prepared and the structure of the model is set, the control unit (1820) can train the blood flow prediction model.

[0113] Furthermore, the control unit (1820) analyzes a 3D medical image to extract blood vessel pixel point coordinates corresponding to blood vessels based on pixel values, removes noise around major blood vessels, generates a 3D blood vessel image, divides the 3D blood vessel image into multiple layers, sets a group of blood vessel pixel points for each layer as blood vessel nodes, and generates connection information to generate a blood vessel graph, and can control the communication unit (1810) to provide the blood vessel graph.

[0114]

[0115] While described with reference to limited embodiments and drawings, the present invention is not limited to the above-described embodiments, and various modifications and variations are possible based on this disclosure by those skilled in the art. Accordingly, the scope of the present invention should be understood solely by the scope of the claims set forth below, and all equivalent or equivalent modifications thereof are deemed to fall within the scope of the present invention.

Claims

1. A method for generating an artificial intelligence model that predicts blood flow data on a server, Step of setting up the structure of the blood flow prediction model; A step of generating blood vessel data in the form of a point cloud corresponding to a major blood vessel from a three-dimensional medical image, specifying a boundary area of ​​the blood vessel data, obtaining a blood flow value of the boundary area, and providing the location of the point cloud for an arbitrary area and the blood flow value of the boundary area as learning data for the blood flow prediction model; and A method for generating an artificial intelligence model, characterized by including a step of training the blood flow prediction model by setting blood flow values ​​in an area other than the above-mentioned boundary area as a label or mask.

2. In the first paragraph, the step of setting the structure of the blood flow prediction model is as follows: A method for predicting blood flow, comprising: setting an input layer of the blood flow prediction model to include a first channel into which location and radius information of a boundary node in a blood vessel graph for an arbitrary region are input, a second channel into which location and radius information of another node in the blood vessel graph are input, and a third channel into which blood flow values ​​of the boundary nodes are input. How to create an artificial intelligence model.

3. In the first paragraph, the step of setting the structure of the blood flow prediction model is as follows: A method characterized by comprising the step of setting a feature extraction layer of the blood flow prediction model to form a local feature reflecting the characteristics of the other node and a boundary feature reflecting the characteristics of the boundary node. How to create an artificial intelligence model.

4. In the third paragraph, the step of setting the feature extraction layer of the blood flow prediction model comprises: characterized by including a step of forming the boundary feature by removing the input order information of the boundary node. How to create an artificial intelligence model.

5. A 3D medical image processing device that analyzes a 3D medical image and extracts blood vessel pixel point coordinates corresponding to the blood vessel based on pixel values ​​to generate a 3D blood vessel image; and A blood vessel graph generation device that generates a blood vessel graph from the three-dimensional blood vessel image by using the blood vessel pixel point group as a node; and A blood flow prediction model is generated by setting a structure of a blood flow prediction model, specifying a boundary region of the blood vessel graph, obtaining a blood flow value of the boundary region, providing a location of an arbitrary region and a blood flow value of the boundary region as learning data of the blood flow prediction model, and setting a blood flow value of a region other than the boundary region as a label or mask. Artificial intelligence model generation system.

6. In a server that generates an artificial intelligence model that predicts blood flow data, the server, A communication unit for receiving three-dimensional medical images; and A blood flow prediction model is set up, and a control unit is included to generate blood flow prediction model by setting up a structure of a blood flow prediction model, generating blood vessel data in the form of a point cloud corresponding to a major blood vessel from the three-dimensional medical image, specifying a boundary area of ​​the blood vessel data, obtaining a blood flow value of the boundary area, providing a location of a point cloud for an arbitrary area and a blood flow value of the boundary area as learning data of the blood flow prediction model, and setting a blood flow value of an area other than the boundary area as a label or mask. Artificial intelligence model generation server.

7. In a computer program stored in a medium to perform processing for generating an artificial intelligence model that predicts blood flow data, Ability to set the structure of the blood flow prediction model; A function of generating vascular data in the form of a point cloud corresponding to a major blood vessel from a 3D medical image, specifying a boundary area of ​​the vascular data, obtaining a blood flow value of the boundary area, and providing the location of the point cloud for an arbitrary area and the blood flow value of the boundary area as learning data for the blood flow prediction model; and A computer program characterized by performing a function of learning the blood flow prediction model by setting blood flow values ​​in an area other than the above boundary area as a label or mask.

Images