CCTA image classification method and device and storage medium
By performing segmentation mask processing and automatic classification of CCTA images, the problem of inaccurate vascular classification caused by relying on labor in the prior art is solved, and higher accuracy and labor cost savings are achieved.
Patent Information
- Application Number
- CN202411494115.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-24
- Publication Date
- 2025-02-07
AI Technical Summary
In the prior art, the classification and identification of blood vessels in coronary arteries rely on artificial means, which is prone to inaccurate classification problems, and there are blind spots in two-dimensional image recognition, which affects accuracy.
By obtaining coronary artery image computed tomography angiography (CCTA) images, segmentation mask processing is performed, and the mask image corresponding to the blood vessel is generated, and the blood vessel position is marked. Then, the blood vessels are automatically determined based on the CCTA image, mask image and preset analysis strategy. target classification.
It improves the classification accuracy of blood vessels in CCTA images, avoids classification inaccuracy caused by human factors, saves labor costs, and reduces capture blind spots in three-dimensional image processing, and improves the overall recognition effect.
Smart Images
Figure CN119399533A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of data processing technology, and in particular to a CCTA image classification method, device and storage medium. Background Art
[0002] Currently, the classification and identification method for blood vessels in the coronary arteries is usually to perform interventional examination of the blood vessels in the coronary arteries through invasive coronary angiography (ICA) to obtain angiography images, and then manually identify the classification of the blood vessels in the angiography images. This classification and identification method relies on human factors and is more prone to inaccurate classification problems. Summary of the invention
[0003] In order to solve the above technical problems, the present application provides a CCTA image classification method, device and storage medium, which can improve the classification accuracy of CCTA images.
[0004] In a first aspect, the present application provides a CCTA image classification method, comprising: obtaining a coronary artery image computed tomography angiography (CCTA) image; performing segmentation mask processing on the CCTA image to obtain a mask image corresponding to the blood vessel in the CCTA image; the mask image is used to mark the location of the blood vessel in the CCTA image; and determining the target classification of the blood vessel in the CCTA image based on the CCTA image, the mask image and a preset analysis strategy.
[0005] In some embodiments, a CCTA image is segmented and masked to obtain a mask image corresponding to a blood vessel in the CCTA image, including: inputting the CCTA image into a preset segmentation network to obtain a mask image; wherein the segmentation network includes at least a first extraction subnetwork and a mask generation subnetwork; the process of the segmentation network generating a mask image based on the CCTA image includes: the first extraction subnetwork extracts the position features of the blood vessels in the CCTA image according to a first extraction method; the mask generation subnetwork performs mask processing on the CCTA image according to the position features to obtain a mask image; the mask value of the blood vessel area in the mask image is 1, and the mask value of the non-blood vessel area is 0.
[0006] In some embodiments, the target classification of blood vessels in the CCTA image is determined according to the CCTA image, the mask image and the preset analysis strategy, including: inputting the CCTA image and the mask image into a preset classification network to obtain the target classification of blood vessels in the CCTA image; wherein the classification network at least includes a fusion subnetwork, a second extraction subnetwork, and a target classification subnetwork; the classification network identifies the target classification of blood vessels in the CCTA image according to the CCTA image and the mask image, and the process includes: the fusion subnetwork splices the CCTA image and the mask image to obtain a spliced image; the second extraction subnetwork extracts the shape features of the blood vessels in the spliced image according to the second extraction method; the target classification subnetwork determines the target classification of blood vessels in the CCTA image according to the shape features and the target classification method.
[0007] In some embodiments, when the CCTA image is a three-dimensional image, the segmentation network and the classification network are both three-dimensional convolutional neural networks.
[0008] In some embodiments, the target classification of blood vessels in the CCTA image is determined based on the CCTA image, the mask image and the preset analysis strategy, including: determining the spatial information of the blood vessels in the CCTA image based on the CCTA image and the mask image; determining the shape and size information of the blood vessels based on the spatial information; determining the target classification of the CCTA image based on the shape and size information, the reference information and the preset analysis strategy; the reference information is a preset shape and size standard of the blood vessels.
[0009] In some embodiments, the blood vessels in the CCTA image are blood vessels corresponding to the area where the stent is located.
[0010] In a second aspect, the present application provides a CCTA image classification device, comprising: an acquisition module for acquiring a coronary artery image computed tomography angiography CCTA image; a processing module for performing segmentation mask processing on the CCTA image to obtain a mask image corresponding to the blood vessel in the CCTA image; the mask image is used to mark the location of the blood vessel in the CCTA image; and a determination module for determining the target classification of the blood vessel in the CCTA image based on the CCTA image, the mask image and a preset analysis strategy.
[0011] In some embodiments, the processing module is specifically used to: input the CCTA image into a preset segmentation network to obtain a mask image; wherein the segmentation network includes at least a first extraction subnetwork and a mask generation subnetwork; the process of the segmentation network generating a mask image based on the CCTA image includes: the first extraction subnetwork extracts the position features of the blood vessels in the CCTA image according to the first extraction method; the mask generation subnetwork performs mask processing on the CCTA image according to the position features to obtain a mask image; the mask value of the blood vessel area in the mask image is 1, and the mask value of the non-blood vessel area is 0.
[0012] In some embodiments, the determination module is specifically used to: input the CCTA image and the mask image into a preset classification network to obtain the target classification of the blood vessels in the CCTA image; wherein the classification network at least includes a fusion subnetwork, a second extraction subnetwork, and a target classification subnetwork; the classification network identifies the target classification of the blood vessels in the CCTA image according to the CCTA image and the mask image, and the process includes: the fusion subnetwork splices the CCTA image and the mask image to obtain a spliced image; the second extraction subnetwork extracts the shape features of the blood vessels in the spliced image according to the second extraction method; the target classification subnetwork determines the target classification of the blood vessels in the CCTA image according to the shape features and the target classification method.
[0013] In some embodiments, when the CCTA image is a three-dimensional image, the segmentation network and the classification network are both three-dimensional convolutional neural networks.
[0014] In some embodiments, the determination module is specifically used to: determine the spatial information of the blood vessel in the CCTA image based on the CCTA image and the mask image; determine the shape and size information of the blood vessel based on the spatial information; determine the target classification of the CCTA image based on the shape and size information, the benchmark information and a preset analysis strategy; the benchmark information is a preset shape and size standard of the blood vessel.
[0015] In some embodiments, the blood vessels in the CCTA image are blood vessels corresponding to the area where the stent is located.
[0016] In a third aspect, the present application provides an electronic device, comprising: a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein when the computer program is executed by the processor, a CCTA image classification method as in any embodiment of the first aspect is implemented.
[0017] In a fourth aspect, the present application provides a computer-readable storage medium, comprising: a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the classification method of CCTA images in any embodiment of the first aspect is implemented.
[0018] In a fifth aspect, the present application provides a computer program product, comprising: when the computer program product is run on a computer, the computer implements the CCTA image classification method in any embodiment of the first aspect.
[0019] The technical solution provided by the present application has the following advantages over the prior art: First, a coronary artery image computed tomography angiography (CCTA) image is obtained, and the CCTA image is segmented and masked to obtain a mask image corresponding to the blood vessel in the CCTA image. The mask image is used to mark the location of the blood vessel in the CCTA image. Then, the target classification of the blood vessel in the CCTA image is determined based on the CCTA image, the mask image, and the preset analysis strategy. In this way, there is no human involvement in the entire classification process, thereby avoiding the problem of inaccurate classification caused by reliance on human factors, improving the accuracy of classification, and saving labor costs. BRIEF DESCRIPTION OF THE DRAWINGS
[0020] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.
[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.
[0022] Figure 1 One of the application scenario schematic diagrams of the CCTA image classification method provided in the embodiment of the present application;
[0023] Figure 2 One of the flow charts of the CCTA image classification method provided in the embodiment of the present application;
[0024] Figure 3 The second flowchart of the CCTA image classification method provided in the embodiment of the present application;
[0025] Figure 4 A second schematic diagram of an application scenario of the CCTA image classification method provided in an embodiment of the present application;
[0026] Figure 5 The third flowchart of the CCTA image classification method provided in the embodiment of the present application;
[0027] Figure 6 A fourth flowchart of the CCTA image classification method provided in an embodiment of the present application;
[0028] Figure 7 A schematic diagram of the structure of a CCTA image classification device provided in an embodiment of the present application;
[0029] Figure 8 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION
[0030] In order to more clearly understand the above-mentioned purposes, features and advantages of the present application, the scheme of the present application will be further described below. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.
[0031] In the following description, many specific details are set forth to facilitate a full understanding of the present application, but the present application may also be implemented in other ways different from those described herein; obviously, the embodiments in the specification are only part of the embodiments of the present application, rather than all of the embodiments.
[0032] Percutaneous coronary intervention (PCI) is a treatment method that inserts a catheter to implant a stent into the narrowed area of the blood vessel to treat the stenosis or blockage of the coronary artery. Patients who need PCI treatment usually have more stubborn and diverse risk factors for atherosclerosis and more complex and high-risk coronary artery lesions. They are at higher risk of subsequent cardiac events such as myocardial infarction, revascularization, and sudden cardiac death. They belong to the high-risk group for coronary heart disease, and the progression of lesions in this group needs to be paid special attention. Therefore, after PCI treatment, affected by various factors, there is a probability that the blood vessels in the area where the stent is located will become restenotic. Therefore, timely classification and identification of stented blood vessels in the coronary arteries is crucial.
[0033] At present, the classification and recognition methods for stents in coronary arteries mainly include coronary computed tomography angiography (CCTA) and ICA classification and recognition methods. Among them, the ICA classification and recognition method is to perform interventional examination on stents in coronary arteries through ICA to obtain angiography images, and then manually identify the classification of stents in the angiography images. This classification and recognition method relies on human factors and is prone to inaccurate classification. In addition, since ICA images are two-dimensional (2D) video images, when using the ICA classification and recognition method to identify the target classification of blood vessels in ICA images, it is necessary to combine different angles to judge the degree of stenosis of blood vessels, and the input algorithm is relatively complex. In addition, even if images taken at different angles are combined, there may still be blind spots, and it is impossible to capture 360 degrees of information from each angle, resulting in inaccurate classification results. The CCTA classification and recognition method uses a pre-trained model to classify and recognize CCTA images. This classification and recognition method is affected by stent artifacts and motion artifacts, and has poor accuracy when identifying in-stent restenosis (ISR).
[0034] In view of the above problems, an embodiment of the present application provides a CCTA image classification method, including obtaining a coronary artery image computed tomography angiography CCTA image, and performing segmentation mask processing on the CCTA image to obtain a mask image corresponding to the blood vessel in the CCTA image. The mask image is used to mark the position of the blood vessel in the CCTA image. Then, according to the CCTA image, the mask image and the preset analysis strategy, the target classification of the blood vessel in the CCTA image is determined. The blood vessel is the blood vessel corresponding to the area where the stent is located in the CCTA image. In this way, the blood vessels in the CCTA image are directly segmented for the CCTA image, instead of using the historically trained model to segment the blood vessels in the CCTA image, thereby avoiding the influence of an inaccurate model on the blood vessel segmentation process, and at the same time reducing the influence of motion artifacts on the blood vessel segmentation process as much as possible, improving the accuracy of determining the blood vessel position, and indirectly improving the accuracy of blood vessel classification in the CCTA image. In addition, the entire process is automatic classification and recognition, avoiding the problem of inaccurate classification that may occur due to human participation, improving the accuracy of classification, and saving labor costs.
[0035] Figure 1 A schematic diagram of the scene architecture of the CCTA image classification method provided in the embodiment of the present application. Figure 1 As shown, the scenario architecture provided by the embodiment of the present application includes: a server 100 and an electronic device 200.
[0036] The electronic device 200 provided in the embodiment of the present application may have various implementation forms, for example, it may be a mobile phone, a personal computer (PC), a smart TV, a laser projection device, a TV, a monitor, a wearable device, an electronic table, etc.
[0037] In some embodiments, when receiving the classification instruction of the CCTA image, the electronic device 200 may communicate data with the server 100. The electronic device 200 may be allowed to communicate with the server 100 via a local area network (LAN) or a wireless local area network (WLAN).
[0038] The server 100 may be a server that provides various services, such as a server that provides support for CCTA images acquired by the electronic device 200. The server may perform segmentation mask processing, target classification and recognition, etc. on the received CCTA images, and feed back the processing results to the electronic device 200. The server 100 may be a server cluster or multiple server clusters, and may include one or more types of servers.
[0039] It should be noted that the CCTA image classification method provided in the embodiment of the present application can be executed by the server 100, can be executed by the electronic device 200, or can be executed by the server 100 and the electronic device 200 together, and the present application does not limit this.
[0040] The CCTA image classification method provided in the embodiment of the present application can be executed by a CCTA image classification device, which can be hardware or software. When the CCTA image classification device is hardware, it can be various electronic devices with the function of running CCTA image classification, including but not limited to mobile phones, computers, tablet computers, televisions, smart TVs, laser projection equipment, monitors, electronic bulletinboards, electronic desktops, etc. When the CCTA image classification device is software, it can be installed in the electronic devices listed above. It can be implemented as multiple software or software modules, or it can be implemented as a single software or software module. No specific limitation is made here.
[0041] Figure 2 A schematic diagram of the process of the CCTA image classification method provided in the embodiment of the present application is shown in FIG. Figure 2 As shown, the classification method of CCTA images may include the following steps.
[0042] S11. Obtain coronary artery computed tomography angiography (CCTA) images.
[0043] Among them, CCTA is a non-invasive imaging technology that can use advanced computed tomography technology to clearly display the blood flow of the coronary arteries by injecting contrast agents. It can assist in evaluating the structure and function of the coronary arteries. The CCTA image is the image obtained by CCTA scanning, also known as the coronary artery CTA image.
[0044] Specifically, the CCTA image may be obtained by collecting the CCTA image corresponding to the user to be detected through relevant medical equipment, and this application does not limit this.
[0045] S12, performing segmentation mask processing on the CCTA image to obtain a mask image corresponding to the blood vessels in the CCTA image.
[0046] The mask image is used to mark the location of the blood vessels in the CCTA image, and the size of the mask image is the same as that of the CCTA image.
[0047] Specifically, the CCTA image is segmented and masked to obtain a mask image corresponding to the blood vessels in the CCTA image. The blood vessels in the CCTA image may be firstly identified using an image recognition algorithm, and then the CCTA image is masked according to the blood vessels to obtain the mask image.
[0048] The image recognition algorithm has the feature of distinguishing blood vessels from the background, for example, a threshold segmentation algorithm, an edge detection algorithm, a region growing algorithm, etc.
[0049] The method of performing mask processing on the CCTA image according to the blood vessels to obtain the mask image may be to perform mask binarization processing on the CCTA image according to the blood vessels, set the mask value of the position where the blood vessels are located to 1, and set the mask value of the position where the non-blood vessels are located to 0, and obtain the mask image corresponding to the blood vessels in the CCTA image. Alternatively, the CCTA image with the blood vessels identified may be input into a pre-trained mask model to perform mask processing on the CCTA image to obtain the mask image.
[0050] In some embodiments, the blood vessels in the CCTA image are blood vessels corresponding to the area where the stent is located. In this way, the blood vessels at the stent in the CCTA image are directly segmented from the CCTA image, rather than using the historically trained model to segment the blood vessels at the stent in the CCTA image, thereby avoiding the influence of an inaccurate model on the stent blood vessel segmentation process, and at the same time reducing the influence of stent artifacts and motion artifacts on the stent blood vessel segmentation process as much as possible, thereby improving the accuracy of determining the stent blood vessel position, and indirectly improving the accuracy of blood vessel classification in the CCTA image.
[0051] S13. Determine the target classification of blood vessels in the CCTA image according to the CCTA image, the mask image and the preset analysis strategy.
[0052] The preset analysis strategy may be an algorithm or method for classifying blood vessels in CCTA images that is configured in advance according to the classification corresponding to the blood vessels, for example, a quantitative coronary angiography (QCA) algorithm, a threshold classification algorithm (such as when the blood vessel diameter and stenosis ratio are within the target range, the classification of the blood vessels in the CCTA image is determined to be the target classification), a machine learning method, etc. The classification result of the blood vessels in the CCTA image corresponds to the pre-configured preset analysis strategy, for example, the classification result of the blood vessel stenosis ratio less than 50% is determined as the first classification, the classification result of the blood vessel stenosis ratio between 50% and 65% is determined as the second classification, and the classification result of the blood vessel stenosis ratio greater than 65% is determined as the third classification; for another example, the classification result of the QCA result greater than or equal to 50 is determined as the first classification, and the classification result of the QCA result less than 50 is determined as the second classification, etc.
[0053] Specifically, according to the CCTA image, the mask image and the preset analysis strategy, the target classification of the blood vessels in the CCTA image can be determined by calling a preset analysis strategy configured in advance through a preset interface to determine the target classification of the blood vessels in the CCTA image.
[0054] In the above scheme, first, a coronary artery image computed tomography angiography CCTA image is obtained, and the CCTA image is segmented and masked to obtain a mask image corresponding to the blood vessel in the CCTA image. The mask image is used to mark the location of the blood vessel in the CCTA image. Then, based on the CCTA image, the mask image and the preset analysis strategy, the target classification of the blood vessel in the CCTA image is determined. In this way, there is no human involvement in the entire classification process, thereby avoiding the problem of inaccurate classification caused by reliance on human factors, improving the accuracy of classification, and saving labor costs.
[0055] In some embodiments, the CCTA image is segmented and masked to obtain a mask image corresponding to a blood vessel in the CCTA image by inputting the CCTA image into a preset segmentation network to obtain the mask image. The preset segmentation network may be a pre-trained network model or a network composed of corresponding algorithms.
[0056] The segmentation network at least includes a first extraction subnetwork and a mask generation subnetwork. Figure 3 As shown in Figure 1, the process of generating a mask image from a CCTA image by a segmentation network includes the following steps:
[0057] S121. The first extraction subnetwork extracts position features of blood vessels in the CCTA image according to a first extraction method.
[0058] The first extraction method is an algorithm or method for extracting blood vessel position features in CCTA images. When the preset segmentation network is a network composed of algorithms, the first extraction method may include the image recognition algorithm in step S12; when the preset segmentation network is a pre-trained network model, the first extraction method is a pre-trained blood vessel position feature extraction method.
[0059] In some embodiments, when the preset segmentation network is a pre-trained network model, the first extraction subnetwork may include an encoder and a decoder, and these two parts may use any three-dimensional (3D) convolutional neural network (CNN) structure. The encoder is used to extract the features of blood vessels in the CCTA image. When constructing the encoder, different numbers of downsampling structures may be defined as needed; the decoder is used to decode the features, and the decoder includes the same number of upsampling structures as the downsampling structures in the encoder, so that the decoder can effectively restore the details of the position features of the input CCTA image when decoding the features. Each layer of the encoder and the decoder has cross-layer links to facilitate better use of high-dimensional information and low-dimensional information. The position features output by the decoder can be saved in the form of a position feature image, which is consistent in size with the input CCTA image.
[0060] S122, the mask generation sub-network performs mask processing on the CCTA image according to the position features to obtain a mask image.
[0061] The mask value of the blood vessel area in the mask image is 1, and the mask value of the non-blood vessel area is 0.
[0062] Specifically, when the preset segmentation network is a network composed of algorithms, the mask generation subnetwork performs mask processing on the CCTA image according to the position features to obtain the mask image by directly performing mask binarization processing on the CCTA image according to the position features, setting the mask value of the area corresponding to the position features to 1, and setting the mask value of the area corresponding to the non-position features to 0, to obtain the mask image. When the preset segmentation network is a pre-trained network model, the mask generation subnetwork performs mask processing on the CCTA image according to the position features to obtain the mask image by inputting the CCTA image with the position features into the pre-trained mask generation subnetwork to obtain the mask image.
[0063] In some embodiments, when the preset segmentation network is a pre-trained network model, the mask generation subnetwork may include a segmentation head based on a 3D convolution structure, which is used to perform mask processing on the CCTA image according to position features to obtain a mask image.
[0064] In some embodiments, when the preset segmentation network is a pre-trained network model, before the CCTA image is input into the preset segmentation network to obtain the mask image, Figure 4As shown, the classification method of CCTA images also includes training a segmentation network 41. When training the segmentation network 41, diceloss loss is used for optimization, and the loss function is expressed as L1=2*sum(p*gt) / (sum(p)+sum(gt)), where L1 is used to represent the loss function of the segmentation network 41, also known as the segmentation loss; p is used to represent the predicted output of the segmentation network 41; and gt is used to represent the actual labeled mask image.
[0065] In the above scheme, the step of segmenting the blood vessels in the CCTA image can be extracted and processed separately, instead of using the historically trained classification model including the segmentation model to directly identify the classification of the CCTA image, thereby avoiding the influence of the inaccurate classification model on the blood vessel segmentation process, and at the same time reducing the influence of stent artifacts and motion artifacts on the blood vessel segmentation process as much as possible, thereby improving the accuracy of determining the blood vessel position and indirectly improving the accuracy of blood vessel classification in CCTA images.
[0066] In some embodiments, Figure 5 As shown, according to the CCTA image, the mask image and the preset analysis strategy, the method of determining the target classification of the blood vessels in the CCTA image may include the following steps:
[0067] S131 . Determine the spatial information of the blood vessel in the CCTA image according to the CCTA image and the mask image.
[0068] The spatial information may include the position of the blood vessels in the CCTA image, the shape, branching structure, connectivity, texture features, etc. of the blood vessels.
[0069] Specifically, according to the CCTA image and the mask image, the spatial information of the blood vessel in the CCTA image can be determined by first multiplying the CCTA image and the mask image pixel by pixel, then extracting the blood vessel region in the CCTA image from the background, and using image analysis technology to analyze the blood vessel region to obtain the spatial information of the blood vessel in the CCTA image. The image analysis technology is used to determine the spatial information of the blood vessel, for example, the image analysis technology can be a contour detection algorithm, a connected region analysis algorithm, etc.
[0070] S132. Determine shape and size information of the blood vessel according to the spatial information.
[0071] The shape and size information of the blood vessel may include the diameter, circumference, volume, surface area, branching angle, length, curvature, shape factor, etc. of the blood vessel.
[0072] Specifically, the shape and size information of the blood vessel can be determined based on the spatial information by first reconstructing the blood vessel in three dimensions based on the spatial information of the blood vessel to form a three-dimensional model of the blood vessel. Then, the diameter, circumference, branch angle, length and other information of the blood vessel are extracted based on the reconstructed three-dimensional model, and the curvature, volume, surface area, shape factor and other information of the blood vessel are calculated using the coordinate information.
[0073] S133: Determine the target classification of the CCTA image based on the shape and size information, the reference information, and the preset analysis strategy.
[0074] The reference information is a pre-set shape and size standard of a blood vessel. For example, the minimum lumen diameter is set to 2.5 millimeters (mm) in the reference information, and the preset analysis strategy is set to a stenosis ratio threshold of 50%. When the lumen diameter of a certain segment of a blood vessel in the shape and size information is less than 2.5 mm, the segment of the blood vessel in the CCTA image is determined to be stenotic, and when the lumen diameter of another segment of a blood vessel in the shape and size information is greater than or equal to 2.5 mm, the segment of the blood vessel is determined not to be stenotic; when the stenosis ratio of the blood vessel is less than 50%, the CCTA image is determined to belong to the first category, and when the stenosis ratio of the blood vessel is greater than or equal to 50%, the CCTA image is determined to belong to the second category.
[0075] In the above scheme, the spatial information of the blood vessels in the CCTA images is first determined based on the CCTA images and the mask images, and the shape and size information of the blood vessels is determined based on the spatial information. Then, based on the shape and size information, the reference information and the preset analysis strategy, the target classification of the CCTA images is determined. In this way, the entire classification process is free of human involvement, thus avoiding the problem of inaccurate classification caused by reliance on human factors, improving the accuracy of classification and saving labor costs.
[0076] In some embodiments, according to the CCTA image, the mask image and the preset analysis strategy, the target classification of the blood vessels in the CCTA image can be determined by inputting the CCTA image and the mask image into a preset classification network to obtain the target classification of the blood vessels in the CCTA image. The preset classification network can be a pre-trained network model or a network composed of corresponding algorithms.
[0077] The classification network at least includes a fusion sub-network, a second extraction sub-network, and a target classification sub-network. Figure 6 As shown, the classification network identifies the target classification process of blood vessels in CCTA images based on CCTA images and mask images, including the following steps:
[0078] S134, the fusion sub-network stitches the CCTA image and the mask image to obtain a stitched image.
[0079] Specifically, when the preset classification network is a network composed of corresponding algorithms, the CCTA image and the mask image are spliced to obtain the spliced image by splicing the CCTA image and the mask image in the channel dimension to obtain the spliced image, or by multiplying the CCTA image and the mask image pixel by pixel to obtain the spliced image. When the preset classification network is a pre-trained network model, the CCTA image and the mask image are spliced to obtain the spliced image by inputting the CCTA image and the mask image into a pre-trained fusion sub-network to obtain the spliced image.
[0080] S135. The second extraction subnetwork extracts shape features of blood vessels in the spliced image according to the second extraction method.
[0081] Specifically, the second extraction method is an algorithm or method for extracting the blood vessel shape features in the spliced image. When the preset classification network is a network composed of algorithms, the second extraction method can be various feature extraction algorithms, such as edge detection algorithms, region growing algorithms, etc. When the preset classification network is a pre-trained network model, the second extraction method is a pre-trained blood vessel shape feature extraction method.
[0082] In some embodiments, when the preset classification network is a pre-trained network model, the second extraction subnetwork may include a 3D convolutional layer and a residual connection structure to extract shape features of blood vessels in the spliced image.
[0083] S136. The target classification subnetwork determines the target classification of the blood vessels in the CCTA image according to the shape features and the target classification method.
[0084] Wherein, when the preset classification network is a network composed of algorithms, the target classification method can be various classification algorithms, for example, various classification algorithms corresponding to the preset analysis strategy in step S13. When the preset classification network is a pre-trained network model, the target classification method is a pre-trained algorithm for determining the target classification of blood vessels in CCTA images.
[0085] In some embodiments, when the preset classification network is a pre-trained network model, the target classification subnetwork may be a multi-layer perceptron, which can determine the target classification of blood vessels in CCTA images according to shape features and target classification methods.
[0086] In some embodiments, when the preset classification network is a pre-trained network model, before determining the target classification of the blood vessels in the CCTA image according to the CCTA image, the mask image and the preset analysis strategy, such as Figure 4As shown, the classification method of CCTA images also includes training a classification network 42. When training the classification network 42, cross entropy loss is used for optimization, and the loss function is expressed as L2=-(q*logp+(1-q)*log(1-p)). Wherein, L2 is used to represent the loss function of the classification network 42; q is a sample label (set according to the classification method, for example, the sample label with a QCA result greater than or equal to 50 is recorded as 1, and the sample label with a QCA result less than 50 is recorded as 0); p is the probability that the classification network 42 predicts the sample as a target label (such as 1).
[0087] In the above scheme, the spatial attention mechanism is used to determine the vascular area in the CCTA image in advance, so that the classification network can better focus on the vascular area in the CCTA image, reduce the impact of the CCTA image background on the classification result, and improve the accuracy of classification. At the same time, the entire process of identifying the target classification of blood vessels in CCTA images is automatic classification and recognition, avoiding the problem of inaccurate classification that may occur due to human participation, improving the accuracy of classification, and saving labor costs.
[0088] In some embodiments, before performing segmentation mask processing on the CCTA image to obtain the mask image corresponding to the blood vessel in the CCTA image, or determining the target classification of the blood vessel in the CCTA image according to the CCTA image, the mask image and the preset analysis strategy, the classification method of the CCTA image also includes preprocessing the CCTA image according to preset rules. The preset rules are default rules, or rules formulated by relevant personnel according to actual conditions, which are used to limit the specifications of the CCTA image. For example, the preset rules may include adjusting the window frame and window position, denoising, enhancing, and normalizing the CCTA image. In this way, before processing the CCTA image, the specifications of the CCTA image are unified, avoiding the influence of the specifications of the CCTA image on the classification and recognition results, and improving the accuracy of the classification and recognition results.
[0089] In some embodiments, when the CCTA image is a three-dimensional image, the segmentation network and the classification network are both three-dimensional convolutional neural networks. In this way, the three-dimensional CCTA image is classified and identified using the three-dimensional segmentation network and the classification network. On the one hand, since the CCTA image, the segmentation network, and the classification network of the embodiment of the present application are three-dimensional, it is not necessary to collect images from multiple perspectives, thereby avoiding the problem that even if two-dimensional images taken from different angles are combined, there may still be blind spots for capturing, and it is impossible to capture information from all angles at 360 degrees, thereby causing inaccurate classification results, thereby improving the accuracy of the classification results. At the same time, time resources are saved and the complexity of the algorithm is simplified.
[0090] The embodiment of the present application can divide the CCTA image classification device into functional modules according to the above method example. For example, each functional module can be divided according to each function, or two or more functions can be integrated into one processing unit. The above integrated modules can be implemented in the form of hardware or in the form of software functional modules. It should be noted that the division of modules in the embodiment of the present application is schematic and is only a logical function division. There may be other division methods in actual implementation.
[0091] Figure 7 A schematic diagram of the structure of a CCTA image classification device provided in an embodiment of the present application is shown in FIG. Figure 7 As shown, the CCTA image classification device includes an acquisition module 71 , a processing module 72 , and a determination module 73 .
[0092] The acquisition module 71 is used to acquire a coronary artery image computed tomography angiography (CCTA) image; the processing module 72 is used to perform segmentation mask processing on the CCTA image to obtain a mask image corresponding to the blood vessel in the CCTA image; the mask image is used to mark the location of the blood vessel in the CCTA image; the determination module 73 is used to determine the target classification of the blood vessel in the CCTA image based on the CCTA image, the mask image and a preset analysis strategy.
[0093] In some embodiments, the processing module 72 is specifically used to: input the CCTA image into a preset segmentation network to obtain a mask image; wherein the segmentation network includes at least a first extraction subnetwork and a mask generation subnetwork; the process of the segmentation network generating a mask image based on the CCTA image includes: the first extraction subnetwork extracts the position features of the blood vessels in the CCTA image according to the first extraction method; the mask generation subnetwork performs mask processing on the CCTA image according to the position features to obtain a mask image; the mask value of the blood vessel area in the mask image is 1, and the mask value of the non-blood vessel area is 0.
[0094] In some embodiments, the determination module 73 is specifically used to: input the CCTA image and the mask image into a preset classification network to obtain the target classification of the blood vessels in the CCTA image; wherein the classification network at least includes a fusion subnetwork, a second extraction subnetwork, and a target classification subnetwork; the classification network identifies the target classification of the blood vessels in the CCTA image according to the CCTA image and the mask image, and the process includes: the fusion subnetwork splices the CCTA image and the mask image to obtain a spliced image; the second extraction subnetwork extracts the shape features of the blood vessels in the spliced image according to the second extraction method; the target classification subnetwork determines the target classification of the blood vessels in the CCTA image according to the shape features and the target classification method.
[0095] In some embodiments, when the CCTA image is a three-dimensional image, the segmentation network and the classification network are both three-dimensional convolutional neural networks.
[0096] In some embodiments, the determination module 73 is specifically used to: determine the spatial information of the blood vessel in the CCTA image based on the CCTA image and the mask image; determine the shape and size information of the blood vessel based on the spatial information; determine the target classification of the CCTA image based on the shape and size information, the benchmark information and a preset analysis strategy; the benchmark information is a preset shape and size standard of the blood vessel.
[0097] In some embodiments, the blood vessels in the CCTA image are blood vessels corresponding to the area where the stent is located.
[0098] The CCTA image classification device provided in this embodiment can execute the CCTA image classification method provided in the above method embodiment. Its implementation principle and technical effect are similar to those of the above method and will not be described in detail here.
[0099] Figure 8 An electronic device according to an exemplary embodiment may include a processor 702, and the processor 702 is used to execute application code to implement the CCTA image classification method in the present application.
[0100] The processor 702 may be a central processing unit (CPU), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits for controlling the execution of the program of the present application.
[0101] like Figure 8 As shown, the electronic device may further include a memory 703. The memory 703 is used to store application program codes for executing the solution of the present application, and the execution is controlled by the processor 702.
[0102] The memory 703 may be a read-only memory (ROM) or other types of static storage devices that can store static information and instructions, a random access memory (RAM) or other types of dynamic storage devices that can store information and instructions, or an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical disc, laser disc, optical disc, digital versatile disc, Blu-ray disc, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory 703 may exist independently and be connected to the processor 702 via the bus 704. The memory 703 may also be integrated with the processor 702.
[0103] like Figure 8 As shown, the electronic device may further include a communication interface 701, wherein the communication interface 701, the processor 702, and the memory 703 may be coupled to each other, for example, via a bus 704. The communication interface 701 is used to exchange information with other devices, for example, to support information exchange between the electronic device and other devices.
[0104] It should be pointed out that Figure 8 The device structure shown in the figure does not constitute a limitation on the electronic device, except Figure 8 In addition to the components shown, the electronic device may include more or fewer components than shown, or combine certain components, or arrange the components differently. The electronic device provided in this embodiment may execute the CCTA image classification method provided in the above method embodiment, and its implementation principle and technical effect are similar to the above method, which will not be described in detail here.
[0105] An embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, each process of the CCTA image classification method in the above method embodiment is implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.
[0106] The computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.
[0107] An embodiment of the present application provides a computer program product, which stores a computer program. When the computer program is executed by a processor, each process of the CCTA image classification method in the above method embodiment is implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.
[0108] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media that include computer-usable program code.
[0109] In the present application, the processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.
[0110] In this application, memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.
[0111] In this application, computer-readable media includes permanent and non-permanent, removable and non-removable storage media. Storage media can be implemented by any method or technology to store information, and the information can be computer-readable instructions, data structures, modules of programs or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, read-only compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. According to the definition in this article, computer-readable media does not include temporary computer-readable media (transitory media), such as modulated data and carrier waves.
[0112] It should be noted that, in this article, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.
[0113] The above description is only a specific implementation of the present application, so that those skilled in the art can understand or implement the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments described herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A CCTA image classification method, characterized in that: include: Obtain coronary artery images computed tomography angiography CCTA images; Performing segmentation mask processing on the CCTA image to obtain a mask image corresponding to the blood vessel in the CCTA image; the mask image is used to mark the position of the blood vessel in the CCTA image; According to the CCTA image, the mask image and a preset analysis strategy, a target classification of blood vessels in the CCTA image is determined.
2. The classification method according to claim 1, characterized in that: The performing segmentation mask processing on the CCTA image to obtain a mask image corresponding to the blood vessel in the CCTA image includes: Inputting the CCTA image into a preset segmentation network to obtain a mask image; The segmentation network at least includes a first extraction subnetwork and a mask generation subnetwork; and the process of generating a mask image according to the CCTA image by the segmentation network includes: The first extraction subnetwork extracts the position features of the blood vessels in the CCTA image according to a first extraction method; The mask generation subnetwork performs mask processing on the CCTA image according to the position feature to obtain a mask image; the mask value of the blood vessel area in the mask image is 1, and the mask value of the non-blood vessel area is 0.
3. The classification method according to claim 2, characterized in that: Determining target classification of blood vessels in the CCTA image according to the CCTA image, the mask image, and a preset analysis strategy includes: Inputting the CCTA image and the mask image into a preset classification network to obtain target classification of blood vessels in the CCTA image; Wherein, the classification network at least includes a fusion subnetwork, a second extraction subnetwork, and a target classification subnetwork; The process of the classification network identifying the target classification of the blood vessels in the CCTA image according to the CCTA image and the mask image includes: The fusion sub-network splices the CCTA image and the mask image to obtain a spliced image; The second extraction subnetwork extracts shape features of blood vessels in the spliced image according to a second extraction method; The target classification subnetwork determines the target classification of the blood vessels in the CCTA image according to the shape features and the target classification method.
4. The classification method according to claim 3, characterized in that: When the CCTA image is a three-dimensional image, the segmentation network and the classification network are both three-dimensional convolutional neural networks.
5. The classification method according to claim 1, characterized in that: Determining target classification of blood vessels in the CCTA image according to the CCTA image, the mask image, and a preset analysis strategy includes: Determining spatial information of the blood vessel in the CCTA image according to the CCTA image and the mask image; Determine shape and size information of the blood vessel according to the spatial information; Based on the shape and size information, the reference information and the preset analysis strategy, the target classification of the CCTA image is determined; the reference information is a preset shape and size standard of the blood vessel.
6. The classification method according to any one of claims 1 to 5, characterized in that: The blood vessels in the CCTA image are blood vessels corresponding to the area where the stent is located.
7. A CCTA image classification device, characterized in that: include: An acquisition module, used for acquiring a coronary artery image computed tomography angiography CCTA image; A processing module, used for performing segmentation mask processing on the CCTA image to obtain a mask image corresponding to the blood vessel in the CCTA image; the mask image is used for marking the position of the blood vessel in the CCTA image; The determination module is used to determine the target classification of the blood vessels in the CCTA image according to the CCTA image, the mask image and a preset analysis strategy.
8. An electronic device, characterized in that: include: A processor, a memory, and a computer program stored in the memory and executable on the processor, wherein when the computer program is executed by the processor, the CCTA image classification method according to any one of claims 1 to 6 is implemented.
9. A computer-readable storage medium, characterized in that: include: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the CCTA image classification method according to any one of claims 1 to 6 is implemented.
10. A computer program product, characterized in that When the computer program product is executed on a computer, the computer is enabled to implement the CCTA image classification method according to any one of claims 1 to 6.