Methods, devices, and storage media for training a detection model to detect vascular adhesions

Abstract

The application discloses a method and device for training a detection model for detecting vascular adhesion and a storage medium. The method comprises: collecting multi-modal image data, wherein the multi-modal image data comprises digital subtraction angiography data, CT angiography data and magnetic resonance angiography data; performing vascular segmentation based on the multi-modal image data to obtain a graph structure of the blood vessels; extracting a graph structure feature of a single blood vessel and a target blood vessel feature according to the graph structure, and labeling whether the blood vessels have passed through adhesion blood vessels; constructing training data based on the graph structure feature of the single blood vessel, the target blood vessel feature and the labeling result; inputting the training data into a detection model to perform vascular adhesion detection to obtain a detection result, so as to train the detection model for detecting vascular adhesion. The scheme disclosed by the application can improve the accurate positioning and efficient detection of vascular adhesion.

Description

Technical Field

[0001] This application generally relates to the field of vascular adhesion detection technology. More specifically, this application relates to a method for training a detection model for detecting vascular adhesion, a method for detecting vascular adhesion, an apparatus, and a computer-readable storage medium. Background Technology

[0002] Digital subtraction angiography (DSA) is the gold standard for diagnosing vascular diseases, and its images show intracranial vessels as a tree-like structure. However, due to volume effects and limitations in imaging technology, vessel adhesion often occurs after segmentation, leading to the formation of ring-like structures. Current techniques typically use skeletonization algorithms to extract centerlines and construct vascular maps after segmentation, but the ring-like structures formed by adhesion significantly affect the efficiency and stability of subsequent processing such as path tracking, and can even cause algorithm crashes. Traditional methods rely on manual annotation or simple threshold segmentation, which struggles to accurately detect adhesion in small targets, and also suffer from high model complexity and significant training and deployment costs.

[0003] Therefore, there is an urgent need to provide a scheme for training a detection model for vascular adhesions in order to improve the accurate localization and efficient detection of vascular adhesions. Summary of the Invention

[0004] In order to at least address one or more of the technical problems mentioned above, this application proposes a scheme for training a detection model for detecting vascular adhesions in several aspects.

[0005] In a first aspect, this application provides a method for training a detection model for detecting vascular adhesions, comprising: acquiring multimodal image data, wherein the multimodal image data includes digital subtraction angiography data, CT angiography data, and magnetic resonance angiography data; performing vascular segmentation based on the multimodal image data to obtain the graphical structure of the vascular vessels; extracting the graphical structural features of a single vascular vessel and the features of a target vascular vessel based on the graphical structure, and marking whether it passes through adherent vascular vessels; constructing training data based on the graphical structural features of the single vascular vessel, the features of the target vascular vessel, and the marking results; and inputting the training data into a detection model to detect vascular adhesions to obtain detection results, thereby training the detection model for detecting vascular adhesions.

[0006] In some embodiments, the graph structure features include at least node information, edge information, and blood vessel information at the nodes, wherein the blood vessel information at the nodes includes at least one or more of the following: the radius at the node, the average gray value or standard deviation within the radius region.

[0007] In other embodiments, the target blood vessel features are extracted by setting the target node in the single blood vessel to the target pixel value and setting the remaining nodes to 0.

[0008] In some other embodiments, the method further includes: determining whether a target node passes through an adhesive blood vessel based on node information and edge information; and marking whether the target node passes through an adhesive blood vessel or not based on the determination result to obtain the marking result.

[0009] In some other embodiments, the method further includes: removing the distal vessels of the single vessel; and constructing the training data based on the graph structure features of the single vessel after removing the distal vessels, the target vessel features, and the annotation results.

[0010] In some other embodiments, the detection model includes an encoder and a classifier, and the training data is input into the detection model to perform vascular adhesion detection to obtain detection results, thereby training the vascular adhesion detection model: based on the training data, the encoder is used to perform an encoding operation to obtain an encoding result; the classifier is used to perform a classification operation on the encoding result to obtain the vascular adhesion detection result, thereby training the vascular adhesion detection model.

[0011] In some other embodiments, the detection model includes a graph network model.

[0012] In a second aspect, this application provides a method for detecting vascular adhesion, comprising: acquiring multimodal image data, wherein the multimodal image data includes digital subtraction angiography data, CT angiography data, and magnetic resonance angiography data; performing vascular segmentation based on the multimodal image data to obtain the graphical structure of the vascular vessels; extracting the graphical structural features of a single vascular vessel and the features of a target vascular vessel based on the graphical structure; and inputting the graphical structural features of the single vascular vessel and the features of the target vascular vessel into a detection model trained by one or more embodiments of the first aspect to perform vascular adhesion detection, thereby obtaining a detection result.

[0013] In a third aspect, this application provides an apparatus comprising: a processor; and a memory having computer instructions stored thereon for training a detection model for detecting vascular adhesions, wherein when the computer instructions are executed by the processor, the plurality of embodiments of the first aspect are implemented; or having computer instructions for detecting vascular adhesions stored thereon, wherein when the computer instructions are executed by the processor, the embodiment of the second aspect is implemented.

[0014] In a fourth aspect, this application provides a computer-readable storage medium having stored thereon computer program instructions for training a detection model for detecting vascular adhesions, wherein when executed by one or more processors, the computer program instructions cause to implement the plurality of embodiments of the first aspect described above; or having stored computer instructions for detecting vascular adhesions, wherein when executed by a processor, the computer instructions cause to implement one embodiment of the second aspect described above.

[0015] Using the training scheme for the detection model of vascular adhesion provided above, this embodiment of the application improves the model's adaptability to complex scenes by collecting multimodal image data covering vascular structure information with different imaging principles; the graph structure generated based on vascular segmentation effectively preserves the location, connection relationship and attribute features of vascular nodes, providing a structured data foundation for adhesion analysis; by extracting the graph structure features of a single blood vessel and the features of the target blood vessel and labeling adhesions, multi-dimensional feature fusion and accurate label generation are achieved, ensuring that the model learns the geometric and gray-scale anomaly patterns of the adhesion region; the constructed training data, combined with graph structure features and labeling results, optimizes the model parameters through end-to-end training, ultimately achieving accurate localization and efficient detection of vascular adhesions. Attached Figure Description

[0016] The above and other objects, features, and advantages of exemplary embodiments of this application will become readily understood by reading the following detailed description with reference to the accompanying drawings. In the drawings, several embodiments of this application are illustrated by way of example and not limitation, and the same or corresponding reference numerals denote the same or corresponding parts, wherein:

[0017] Figure 1 This is an exemplary flowchart illustrating a method for training a detection model for detecting vascular adhesions according to an embodiment of this application;

[0018] Figure 2 This is an exemplary flowchart illustrating the overall training of a detection model for detecting vascular adhesions according to an embodiment of this application;

[0019] Figure 3 This is an exemplary schematic diagram illustrating a portion of a blood vessel segment according to an embodiment of this application;

[0020] Figure 4 This is an exemplary flowchart illustrating a method for detecting vascular adhesions according to an embodiment of this application;

[0021] Figure 5 This is an exemplary structural block diagram illustrating a device according to an embodiment of this application. Detailed Implementation

[0022] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0023] It should be understood that the terms "comprising" and "including" used in the specification and claims of this application indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0024] It should also be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the application. As used in this specification and claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this specification and claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes such combinations.

[0025] As used in this specification and claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if [described condition or event] is detected" may be interpreted, depending on the context, as "once determined," "in response to determination," "once [described condition or event] is detected," or "in response to detection of [described condition or event]."

[0026] The specific embodiments of this application will now be described in detail with reference to the accompanying drawings.

[0027] Figure 1 This is an exemplary flowchart illustrating a method 100 for training a detection model for detecting vascular adhesions according to an embodiment of this application. Figure 1 As shown, the method 100 includes: step S101: acquiring multimodal image data, wherein the multimodal image data includes digital subtraction angiography data, CT angiography data, and magnetic resonance angiography data; step S102: performing vessel segmentation based on the multimodal image data to obtain the graphical structure of the vessels; step S103: extracting the graphical structure features of a single vessel and the target vessel features based on the graphical structure, and marking whether it passes through adherent vessels; step S104: constructing training data based on the graphical structure features of a single vessel, the target vessel features, and the marking results; and step S105: inputting the training data into a detection model to perform vessel adhesion detection to obtain detection results, thereby training the detection model for vessel adhesion detection.

[0028] First, in step S101, multimodal image data is acquired. This multimodal image data may include digital subtraction angiography (DSA) data, CT angiography data, and magnetic resonance angiography (MRI) data. DSA data is the gold standard for screening and diagnosing various vascular diseases, providing detailed information about the vascular lumen. CT angiography and MRI data can also be used for vascular disease screening and follow-up, respectively supplementing the spatial relationship between blood vessels and surrounding tissues and displaying hemodynamic characteristics. This embodiment of the application uses three modalities of data to train the detection model, covering vascular structural information based on different imaging principles and improving the model's adaptability to complex scenarios.

[0029] Next, in step S102, blood vessel segmentation is performed based on multimodal image data to obtain the graph structure of the blood vessels. In some embodiments, image processing methods or deep learning methods can be used for blood vessel segmentation. The aforementioned image processing methods may include, but are not limited to, image edge detection and image morphology methods. The aforementioned deep learning methods can be based on existing pre-trained segmentation models for blood vessel segmentation, and this application does not impose any limitations on this. Using the segmented blood vessels, a network consisting of connected nodes and edges can be generated using, for example, the sknw library, thereby obtaining the graph structure of the blood vessels.

[0030] Based on the obtained vascular graph structure, in step S103, the graph structure features of a single vascular vessel and the target vascular features are extracted according to the graph structure, and it is marked whether it passes through adherent vessels. In some embodiments, the aforementioned graph structure features may include, but are not limited to, node information, edge information, and vascular information at the nodes. The vascular information at the nodes may include, but is not limited to, one or more of the following: the radius at the node, the average grayscale value within the radius region, or the standard deviation. The aforementioned node information may include, for example, the coordinate position of the vascular center point in the image space; edge information may include the connection relationship between nodes, reflecting the edge length of the vascular segment, and directional information that helps determine the direction of the vascular vessel.

[0031] In some embodiments, target blood vessel features can be extracted by setting the target node in a single blood vessel to a target pixel value (e.g., 1), and setting the remaining nodes to 0, to extract the target blood vessel features. That is, extracting a binary mask of the target blood vessel. In some embodiments, it can be determined whether the target node passes through adherent blood vessels based on node and edge information, and the target node is labeled as having passed through or not passing through adherent blood vessels based on the determination result to obtain the labeling result. That is, two types of labels are formed: one type passes through adherent blood vessels, whose value can be set to 1; the other type does not pass through adherent blood vessels, whose value can be set to 0. In some implementation scenarios, by traversing all paths of a single blood vessel through a graph structure (nodes, edges, attributes) generated based on blood vessel segmentation, if a node has multiple paths connecting to the same endpoint, it is determined to be a loop structure, i.e., there are adherent paths.

[0032] Further, in step S104, training data is constructed based on the graph structure features of a single blood vessel, the target blood vessel features, and the annotation results. As an example, suppose the graph structure of a single blood vessel is denoted as G, the graph structure features as feature1, the target blood vessel features as feature2, and the annotation result is "passing through adherent blood vessels," with a value of 1. In this scenario, the training data [[G, feature1, feature2], 1] can be obtained. Similarly, if the annotation result is "not passing through adherent blood vessels," with a value of 0, then the training data [[G, feature1, feature2], 0] can be obtained.

[0033] In some implementation scenarios, distal vessels of a single blood vessel can be removed, and training data can be constructed based on the graph structure features of the single blood vessel after removing distal vessels, the target blood vessel features, and the annotation results. Based on this, blood vessel segments in the neighborhood of the target node can be retained, and redundant branches at more distant points can be removed to construct a simplified subgraph, thereby improving training efficiency.

[0034] It can be understood that a distal vessel in a single blood vessel refers to a segment of blood vessel that is more than three edges away from the target node. As an example, let's denote a single blood vessel after removing distal vessels as G_box, and the corresponding graph structure features and target vessel features as feature1_box and feature2_box, respectively. The annotation result is "passing through adherent vessels," with a value of 1. In this scenario, we can obtain training data [[G_box,feature1_box,feature2_box],1]. Alternatively, the annotation result is "passing through adherent vessels," with a value of 0, resulting in training data [[G_box,feature1_box,feature2_box],0].

[0035] Finally, in step S105, the training data is input into the detection model to perform vascular adhesion detection to obtain detection results, thereby training the vascular adhesion detection model. In some embodiments, the detection model may include a graph network model. In other embodiments, the detection model may include an encoder and a classifier. Specifically, based on the training data, an encoder is used to perform an encoding operation to obtain an encoding result, and then a classifier is used to perform a classification operation on the encoding result to obtain the vascular adhesion detection result, thereby training the vascular adhesion detection model.

[0036] In an exemplary scenario, if the encoder is denoted as `encoder` and the classifier as `clf`, the encoding result `feature = encoder(G, feature1, feature2)` is first obtained through encoding operations using the encoder. For the scenario of removing distal blood vessels, `feature = encoder(G_box, feature1_box, feature2_box)`. Next, the encoding result is classified using the classifier, `pred = clf(feature)`, and a loss function such as `Loss = CrossEntroy(label, pred)` is calculated for backpropagation to update the parameters of the encoder `encoder` and the classifier `clf`, thereby training the detection model for blood vessel adhesion detection. Here, `label` represents the label value.

[0037] As described above, the multimodal image data acquisition (digital subtraction angiography data, CT angiography data, and magnetic resonance angiography data) of this application covers vascular information based on different imaging principles, improving the model's adaptability to complex scenarios. By using, for example, the sknw library to generate vascular maps containing node positions, radii, edge connections, and grayscale attributes, a structured data foundation is provided for subsequent analysis. Graph structural features (e.g., node information, edge information, vascular attributes) and target vascular features (e.g., binary masks) are extracted and adhesions are labeled. Multi-dimensional feature fusion accurately captures geometric and grayscale anomalies in adhesion regions, and accurate label generation improves data consistency. Furthermore, by constructing training data or simplifying training data (e.g., removing distal vessels), and using, for example, cross-entropy loss to optimize the graph network model, efficient end-to-end feature learning is achieved, ensuring detection accuracy while reducing model complexity and supporting real-time clinical applications.

[0038] Figure 2 This is an exemplary flowchart illustrating the overall training of a detection model for detecting vascular adhesions according to an embodiment of this application. Figure 2As shown, in step S201, multimodal image data is acquired, including digital subtraction angiography data, CT angiography data, and magnetic resonance angiography data. Next, in step S202, vessel segmentation is performed based on the multimodal image data to obtain the graphical structure of the vessels. Specifically, vessel segmentation can be performed using image processing methods or deep learning methods, and the graphical structure of the vessels can be generated using, for example, the SKNW library. In step S203, distal vessels are removed, and in step S204, the graphical structure features of the single vessel after removing distal vessels, the target vessel features, and the annotation results are extracted to obtain training data. Further, the training data is input into the detection model, and in step S205, an encoder is used to perform encoding operations to obtain the encoding results. In step S206, a classifier is used to classify the encoding results, and in step S207, a loss function (e.g., cross-entropy loss) is calculated to train the detection model for vessel adhesion detection, so that in step S208, the trained detection model is...

[0039] Figure 3 This is an exemplary schematic diagram illustrating a portion of a blood vessel segment according to an embodiment of this application. For example... Figure 3 The image shown is a partial vascular segment diagram X. When training a detection model for detecting vascular adhesions, this partial vascular segment diagram X is first segmented, and then a graph structure of the blood vessels can be generated using a library such as sknw. The image only shows nodes in the graph structure, such as points 1, 2, 3, 4, 5, and 6. By connecting the corresponding nodes according to the graph structure, the corresponding single vascular segment can be represented, such as vascular segment 1-3-6; vascular segment 5-2-1.

[0040] Next, graph structure features and target vessel features of a single vessel are extracted based on the graph structure, and whether it passes through adherent vessels is marked. Taking vessel segment 1-3-6 as an example, the node information, edge information, radius at the node, and average or standard deviation of grayscale values ​​within the radius region are extracted from its graph structure G as graph structure feature feature1. Further, target vessel feature feature2 is extracted, where the value of the corresponding node in the target vessel feature is 1, and the value of other nodes is 0. In addition, since vessel segment 1-3-6 is an adherent vessel, its labeling result is 1, and the training data [[G,feature1,feature2],1] is obtained. Similarly, for vessel segment 5-2-1, which does not pass through adherent vessels, the value is 0, and the training data [[G',feature1',feature2'],0] is obtained. As mentioned above, training data [[G_box,feature1_box,feature2_box],1] can be constructed by removing distal blood vessels. This data is then input into the detection model, where an encoder performs encoding operations to obtain the encoding results. A classifier is then used to classify the encoding results to obtain the detection results for blood vessel adhesion detection, thereby training the blood vessel adhesion detection model.

[0041] Figure 4 This is an exemplary flowchart illustrating a method 400 for detecting vascular adhesions according to an embodiment of this application. Figure 4 As shown, the method 400 includes: Step S401: acquiring multimodal image data, wherein the multimodal image data includes digital subtraction angiography data, CT angiography data, and magnetic resonance angiography data; Step S402: performing vessel segmentation based on the multimodal image data to obtain the graph structure of the vessels; Step S403: extracting the graph structure features of a single vessel and the target vessel features based on the graph structure; Step S404: inputting the graph structure features of the single vessel and the target vessel features into a trained detection model to perform vessel adhesion detection, thereby obtaining the detection results. Further details regarding the extraction of graph structure features of a single vessel and the target vessel features, as well as the training of the detection model, can be found above. Figure 1 , Figure 2 The description of the process is not repeated here. According to the detection results, if the classification operation pred is 1, it indicates that there is adhesion at the intermediate endpoint; other values ​​indicate that there is no adhesion.

[0042] In some embodiments, in detecting vascular adhesions, the distal portion of a single vessel can be removed. The graphical structural features of the single vessel after removing the distal portion, along with the target vessel features, are then input into a trained detection model to perform vascular adhesion detection, thereby obtaining the detection results. Similarly, based on the detection results, if the classification operation `pred` is 1, it indicates that adhesion exists at the intermediate endpoint; other values ​​indicate that adhesion does not exist.

[0043] Figure 5 This is an exemplary structural block diagram illustrating a device 500 according to an embodiment of this application. It is understood that the device 500 may include the apparatus of the embodiments of this application, and the device implementing the solution of this application may be a single device (e.g., a computing device) or a multifunctional device including various peripheral devices.

[0044] like Figure 5 As shown, the device of this application may further include a central processing unit (“CPU”) 511, which may be a general-purpose CPU, a special-purpose CPU, or other information processing and program execution unit. Further, the device 500 may also include a mass storage device 512 and a read-only memory (“ROM”) 513, wherein the mass storage device 512 may be configured to store various types of data, including various multimodal data, graph structure features, target blood vessel features and annotation results, algorithm data, intermediate results, and various programs required to run the device 500. The ROM 513 may be configured to store power-on self-tests for the device 500, initialization of various functional modules in the system, drivers for the system's basic input / output, and data and instructions required to boot the operating system.

[0045] Optionally, device 500 may also include other hardware platforms or components, such as the tensor processing unit (“TPU”) 514, graphics processing unit (“GPU”) 515, field-programmable gate array (“FPGA”) 516, and machine learning unit (“MLU”) 517 shown. It is understood that although various hardware platforms or components are shown in device 500, they are merely exemplary and not limiting, and those skilled in the art can add or remove appropriate hardware as needed. For example, device 500 may include only a CPU, associated storage devices, and interface devices to implement the method of training a detection model for detecting vascular adhesions or the method for detecting vascular adhesions described in this application.

[0046] In some embodiments, to facilitate data transmission and interaction with external networks, the device 500 of this application further includes a communication interface 518, through which it can connect to a local area network / wireless local area network (“LAN / WLAN”) 505, and further through the LAN / WLAN to connect to a local server 506 or to the Internet (“Internet”) 507. Alternatively or additionally, the device 500 of this application can also directly connect to the Internet or cellular network via the communication interface 518 based on wireless communication technology, such as wireless communication technology based on 3G (“3G”), 4G (“4G”), or 5G (“5G”). In some application scenarios, the device 500 of this application can also access the server 508 and database 509 of an external network as needed to obtain various known algorithms, data, and modules, and can remotely store various data, such as various data or instructions for presenting, for example, multimodal data, graph structure features, target vascular features, and annotation results.

[0047] The peripheral devices of device 500 may include a display device 502, an input device 503, and a data transmission interface 504. In one embodiment, the display device 502 may include, for example, one or more speakers and / or one or more visual displays, configured to train the detection model for detecting vascular adhesions of this application or to provide voice prompts and / or image / video display for detecting vascular adhesions. The input device 503 may include, for example, a keyboard, mouse, microphone, posture capture camera, or other input buttons or controls, configured to receive audio data input and / or user commands. The data transmission interface 504 may include, for example, a serial interface, parallel interface, or Universal Serial Bus interface (“USB”), Small Computer System Interface (“SCSI”), Serial ATA, FireWire (“FireWire”), PCI Express, and High Definition Multimedia Interface (“HDMI”), configured for data transmission and interaction with other devices or systems. According to the scheme of this application, the data transmission interface 504 can receive multimodal data acquired from CT or MRI equipment and transmit multimodal data or various other types of data or results to device 500.

[0048] The CPU 511, mass storage 512, ROM 513, TPU 514, GPU 515, FPGA 516, MLU 517, and communication interface 518 of the device 500 of this application can be interconnected via bus 519, and can interact with peripheral devices through this bus. In one embodiment, the CPU 511 can control other hardware components in the device 500 and its peripheral devices through this bus 519.

[0049] The above combination Figure 5A device that can be used to execute this application is described. It should be understood that the device structure or architecture described herein is merely exemplary, and the implementation method and implementation entity of this application are not limited thereto, but can be changed without departing from the spirit of this application.

[0050] Based on the foregoing description in conjunction with the accompanying drawings, those skilled in the art will understand that the embodiments of this application can also be implemented by software programs. Therefore, this application also provides a computer-readable storage medium storing computer-readable instructions for training a detection model for detecting vascular adhesions or for detecting vascular adhesions. When executed by one or more processors, these computer-readable instructions can be used to implement the embodiments of this application in conjunction with the accompanying drawings. Figure 1 The method described for training a detection model for detecting vascular adhesions; or the implementation of this application in conjunction with the appendix. Figure 4 The method described is for detecting vascular adhesions.

[0051] It should be noted that although the operations of the method of this application are described in a specific order in the accompanying drawings, this does not require or imply that these operations must be performed in that specific order, or that all the operations shown must be performed to achieve the desired result. On the contrary, the steps depicted in the flowchart can be performed in a different order. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into one step, and / or one step may be broken down into multiple steps.

[0052] It should be understood that when the terms "first," "second," "third," and "fourth," etc., are used in the claims, specification, and drawings of this application, they are used only to distinguish different objects and not to describe a specific order. The terms "comprising" and "including" as used in the specification and claims of this application indicate the presence of the described features, integrals, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or collections thereof.

[0053] It should also be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the application. As used in this specification and claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this specification and claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes such combinations.

[0054] Although the embodiments of this application are described above, the content is merely an example adopted for the purpose of facilitating understanding of this application and is not intended to limit the scope and application scenarios of this application. Any person skilled in the art described in this application may make any modifications and changes in the form and details of the implementation without departing from the spirit and scope disclosed in this application, but the scope of patent protection of this application shall still be determined by the scope defined in the appended claims.

[0055] Furthermore, the collection and acquisition of various data in this application comply with relevant laws and regulations and are authorized by the data providers. Any organization or individual that needs to obtain external data shall obtain authorization in accordance with the law and ensure data security, and shall not illegally collect, use, process, or transmit unauthorized or unprotected data, nor shall it illegally buy, sell, provide, or disclose unauthorized or unprotected data.

Claims

1. A method for training a detection model for detecting vascular adhesion, comprising: collecting multi-modal image data, wherein the multi-modal image data comprises digital subtraction angiography data, CT angiography data, and magnetic resonance angiography data; performing vascular segmentation based on the multi-modal image data to obtain a graph structure of the blood vessels; extracting graph structure features of a single vessel and target vessel features from the graph structure, and labeling whether the target vessel passes through an adhesion vessel, wherein the graph structure features comprise at least node information, edge information, and vessel information at the nodes, and the vessel information at the nodes comprises at least one or more of a radius at the nodes, a mean gray value, or a standard deviation in the radius area; constructing training data based on the graph structure features of the single vessel, the target vessel features, and the labeling results; inputting the training data into a detection model to perform vascular adhesion detection to obtain a detection result, thereby training the detection model for detecting vascular adhesion, wherein the target vessel features are extracted by: setting a target node in the single vessel to a target pixel value and setting the remaining nodes to 0 to extract the target vessel features; and wherein the method further comprises: determining whether the target node passes through an adhesion vessel based on the node information and the edge information; and labeling the target node as passing through an adhesion vessel or not passing through an adhesion vessel based on the determination result to obtain the labeling results, wherein the node information comprises coordinate positions of vessel center points in an image space, and the edge information comprises connection relationships between the nodes.

2. The method of claim 1, further comprising: removing distal vessels of the single vessel; and constructing the training data based on the graph structure features of the single vessel after removing the distal vessels, the target vessel features, and the labeling results.

3. The method of claim 1 or 2, wherein the detection model comprises an encoder and a classifier, and the training of the detection model for detecting vascular adhesion is performed by: using the encoder to perform an encoding operation based on the training data to obtain an encoding result; and using the classifier to perform a classification operation on the encoding result to obtain a detection result of the detection model for detecting vascular adhesion.

4. The method of claim 3, wherein the detection model comprises a graph network model.

5. A method for detecting vascular adhesion, comprising: obtaining multi-modal image data, wherein the multi-modal image data comprises digital subtraction angiography data, CT angiography data, and magnetic resonance angiography data; performing vascular segmentation based on the multi-modal image data to obtain a graph structure of the blood vessels; extracting graph structure features of a single vessel and target vessel features from the graph structure; and inputting the graph structure features of the single vessel and the target vessel features into a detection model trained according to the method of any one of claims 1-4 to perform vascular adhesion detection to obtain a detection result.

6. An apparatus, comprising: a processor; and ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ a memory having stored thereon computer instructions for training a detection model for detecting vessel adhesion, which when executed by a processor, cause implementation of the method according to any one of claims 1-4; or computer instructions for detecting vessel adhesion, which when executed by a processor, cause implementation of the method according to claim 5.

7. A computer readable storage medium having stored thereon computer program instructions for training a detection model for detecting vessel adhesion, which when executed by one or more processors, cause implementation of the method according to any one of claims 1-4; or computer program instructions for detecting vessel adhesion, which when executed by one or more processors, cause implementation of the method according to claim 5.

Images