WO2023280148A1

WO2023280148A1 - Blood vessel segmentation method and apparatus, and electronic device and readable medium

Info

Publication number: WO2023280148A1
Application number: PCT/CN2022/103838
Authority: WO
Inventors: 梁孔明; 潘成伟; 俞益洲; 李一鸣; 乔昕
Original assignee: 杭州深睿博联科技有限公司; 北京深睿博联科技有限责任公司
Priority date: 2021-07-07
Filing date: 2022-07-05
Publication date: 2023-01-12
Also published as: CN113554665A

Abstract

The present invention provides a blood vessel segmentation method and apparatus, and an electronic device and a readable medium. The method comprises: inputting a CTA image into a first convolutional network, and performing blood vessel pre-segmentation based on multi-scale feature extraction; constructing a blood vessel distribution graph; inputting the CTA image into a second convolutional network, multiplying an output of the second convolutional network by the input CTA image, and then inputting a multiplied result into a third convolutional network; the third convolutional network and a graph convolutional network performing feature interaction on the basis of bidirectional mapping, performing global feature modeling on the basis of the blood vessel distribution graph to capture multi-scale local and global features, and implementing blood vessel region prediction; and fusing the multi-scale features to predict a blood vessel segmentation result. According to the present invention, blood vessel segmentation is performed on the basis of cross-network multi-scale feature fusion and fusion of local features and global features, and compared with blood vessel segmentation based on local features in the prior art, the present invention can effectively prevent a model from predicting a segmentation region which does not conform to a blood vessel structure, thereby improving the accuracy of blood vessel segmentation.

Description

A blood vessel segmentation method, device, electronic device and readable medium

The present invention claims the priority of the application proposed by the applicant, the application date is July 7, 2021, the application number is CN2021107686178, and the application name is "a blood vessel segmentation method and device". The entire content of the above application is hereby incorporated by reference in its entirety.

technical field

The invention belongs to the technical field of medical image blood vessel segmentation, and in particular relates to a blood vessel segmentation method, device, electronic equipment and readable medium.

Background of the invention

Common angiographic techniques have been widely used in clinical diagnosis and treatment. The segmentation algorithm can realize automatic reconstruction of blood vessels (such as head and neck vessels, coronary arteries, etc.), which greatly improves the operating efficiency of the hospital while reducing the work pressure of technicians. In the actual scene, some external factors (such as artifacts, noise, shooting technology, etc.) will affect the quality of blood vessel imaging, so that the results of the segmentation algorithm may be missed or broken.

The existing blood vessel segmentation method mainly predicts the blood vessel area through the extraction of local features, and cannot model the overall structure of the blood vessel; the existing method mainly uses multi-scale local features for blood vessel prediction, and cannot predict the global structure of multiple scales. Features are learned and combined.

Contents of the invention

In order to solve the above problems in the prior art, the present invention provides a blood vessel segmentation method, device, electronic equipment and readable medium.

In order to achieve the above object, the present invention adopts the following technical solutions.

In a first aspect, the present invention provides a blood vessel segmentation method, including:

Input the CTA image to the first convolutional network for pre-segmentation of blood vessels based on multi-scale feature extraction;

The pre-segmentation results and segmentation labels are used as input to construct a blood vessel distribution map. The nodes of the graph are pixel areas with a high probability of blood vessels and consistent gray levels. The shape of the nodes is consistent with the direction of the blood vessels; the edges of the graph represent the correlation of connected nodes. The length of the edge is less than the set threshold;

The CTA image is input to the second convolutional network with the same network structure and weight parameters as the first convolutional network, and the output of the second convolutional network is multiplied by the input CTA image and then input to the third convolutional network;

The third convolutional network and the graph convolutional network perform feature interaction based on bidirectional mapping, and perform global feature modeling based on the vascular distribution map to capture multi-scale local and global features, and realize vascular region prediction;

Through the fusion of multi-scale features, the result of blood vessel segmentation is predicted.

Further, the method also includes the step of normalizing the input CTA image, normalizing the grayscale of the pixel to [0,255], and the calculation formula is as follows:

In the formula,

is the gray value x of any pixel after normalization, and x _min and x _max are the minimum and maximum values of the gray value of the pixel before normalization, respectively.

Further, the blood vessel pre-segmentation method specifically includes:

Input the CTA sequence into the encoding layer of the first convolutional network;

The encoding layer performs multi-scale feature extraction through convolution and pooling operations, the space size of the i-th scale is 1/2 ⁱ of the original space size, i=1,2,...,N, N is the number of scales; the Features include shape and spatial location information of vessel segmentation;

The multi-scale features extracted by the encoding layer are input to the decoding layer, and the decoding layer maps the global features of each scale back to the original feature size through upsampling or deconvolution, and combines them with the encoding features at the corresponding scale to obtain the segmentation result.

Further, the method for predicting the vascular region specifically includes:

The third convolutional network projects multiple pixels of encoded features into a single node in the graph convolutional network through forward mapping;

The graph convolutional network propagates the features on the blood vessel distribution map for global feature modeling;

The graph convolutional network maps the node features back to the feature space of the third convolutional network through reverse mapping, and propagates the global features to each part for feature enhancement;

Combine the features of the third convolutional network and the graph convolutional network, capture the local features of the image through the third convolutional network, capture the global features of blood vessels through the graph convolutional network, and use feature interaction to combine local features and global features , to complete the prediction of the area where the blood vessel is located.

Further, the method uses the Dice coefficient to measure the similarity between the blood vessel segmentation result and the gold standard marked by the doctor, and the formula is as follows:

In the formula, A and B are the blood vessel pixel point set of the gold standard and the pixel point set of the blood vessel segmentation result respectively, and |A|, |B| and |A∩B| are A, B and the intersection of A and B respectively The number of pixels.

In a second aspect, the present invention provides a blood vessel segmentation device, comprising:

The pre-segmentation module is used to input the CTA image to the first convolutional network for pre-segmentation of blood vessels based on multi-scale feature extraction;

The distribution map building module is used to construct a blood vessel distribution map with pre-segmentation results and segmentation labels as input. The nodes of the map are pixel areas with a high probability of blood vessels and consistent gray levels. The shape of the nodes is consistent with the direction of the blood vessels; the edges of the graph represent The relevance of the connected nodes, the length of the edge is less than the set threshold;

The first-stage segmentation module is used to input the CTA image to the second convolutional network with the same network structure and weight parameters as the first convolutional network, and to multiply the output of the second convolutional network with the input CTA image and then input it to The third convolutional network;

The second stage segmentation module is used for the third convolutional network and the graph convolutional network to perform feature interaction based on two-way mapping, and to perform global feature modeling based on the blood vessel distribution map to capture multi-scale local and global features, and realize blood vessel region prediction;

The feature fusion module is used to predict the blood vessel segmentation result by fusing multi-scale features.

Further, the device also includes a normalization module for preprocessing the input CTA image, which is used to normalize the grayscale of the pixels to [0, 255], and the calculation formula is as follows:

In the formula,

Further, the blood vessel pre-segmentation method specifically includes:

Further, the method for predicting the vascular region specifically includes:

In a third aspect, the present application provides a readable medium, including execution instructions, and when a processor of an electronic device executes the execution instructions, the electronic device executes the method described in any one of the first aspects.

In a fourth aspect, the present application provides an electronic device, including a processor and a memory storing execution instructions, and when the processor executes the execution instructions stored in the memory, the processor executes the method described in the first aspect any of the methods described.

Compared with the prior art, the present invention has the following beneficial effects:

The present invention models the global characteristics of blood vessels based on the construction of a blood vessel distribution network, uses the overall structure of blood vessels to enhance the features of segmented regions, and taps the potential direction of blood vessels. Compared with the prior art, artificially setting rules to connect fractured regions, It is more applicable to complex scenes in practice; the present invention performs vessel segmentation based on cross-network multi-scale feature fusion and fusion of local features and global features. Compared with the prior art for vessel segmentation based on local features, feature representation with stronger information can be obtained , to effectively prevent the model from predicting segmentation regions that do not conform to the vascular structure, thereby improving the accuracy of vascular segmentation.

Brief description of the drawings

In order to more clearly illustrate the embodiments of the present application or existing technical solutions, the following will briefly introduce the accompanying drawings that need to be used in the description of the embodiments or prior art. Obviously, the accompanying drawings in the following description are only For some embodiments described in the application, those of ordinary skill in the art can also obtain other drawings based on these drawings without any creative effort.

FIG. 1 is a flowchart of a blood vessel segmentation method according to an embodiment of the present invention.

FIG. 2 is a schematic diagram of a network structure according to an embodiment of the present invention.

Fig. 3 is a block diagram of a blood vessel segmentation device according to an embodiment of the present invention.

FIG. 4 is a schematic structural diagram of an electronic device provided by an embodiment of the present invention.

Modes of Carrying Out the Invention

In order to make the technical objectives, technical solutions and beneficial effects of the present invention clearer, the specific embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. The described specific embodiments are only part of the embodiments of the present invention, not all of them. Based on the specific implementation manners of the present invention, other embodiments obtained by those skilled in the art without creative efforts all belong to the protection scope of the present invention.

Fig. 1 is a flowchart of a blood vessel segmentation method according to an embodiment of the present invention, including the following steps:

Step 101, input the CTA image to the first convolutional network, and perform blood vessel pre-segmentation based on multi-scale feature extraction;

Step 102, using the pre-segmentation result and the segmentation label as input to construct a blood vessel distribution map, the nodes of the map are pixel point areas with a high probability of blood vessels and consistent gray levels, and the shape of the nodes is consistent with the direction of the blood vessels; the edges of the graph represent the connected nodes. Relevance, the length of the edge is less than the set threshold;

Step 103, the CTA image is input to the second convolutional network with the same network structure and weight parameters as the first convolutional network, and the output of the second convolutional network is multiplied by the input CTA image and then input to the third convolutional network ;

Step 104, the third convolutional network and the graph convolutional network perform feature interaction based on bidirectional mapping, and perform global feature modeling based on the blood vessel distribution map to capture multi-scale local and global features, and realize blood vessel region prediction;

In step 105, the vessel segmentation result is predicted by fusing the multi-scale features.

In this embodiment, step 101 is mainly used for blood vessel pre-segmentation. The blood vessel segmentation algorithm in this embodiment is mainly implemented by a convolutional neural network (CNN), including three convolutional networks (first to third convolutional networks) and a graph convolutional network, as shown in FIG. 2 . CNN is a feedforward neural network, but unlike the general fully connected feedforward neural network, its convolutional layer has the characteristics of local connection and weight sharing, so it can greatly reduce the number of weight parameters, thereby reducing Model complexity and increased running speed. A typical CNN is composed of convolutional layers, pooling layers (or pooling layers, downsampling layers), and fully connected layers cross-stacked. The function of the convolution layer is to extract the features of a local area through the convolution operation of the convolution kernel and the input image. Different convolution kernels are equivalent to different feature extractors. The role of the pooling layer is to perform feature selection and reduce the number of features, thereby further reducing the number of parameters. Generally, the maximum pooling method and the average pooling method are used. The fully connected layer is used to fuse the different features obtained. In this embodiment, the CTA image is input to the first convolution network, and multiple convolution kernels of different sizes are used for multi-scale feature extraction, so as to pre-segment blood vessels.

In this embodiment, step 102 is mainly used to construct a blood vessel distribution map. In this embodiment, the blood vessel pre-segmentation result obtained in the previous step (the probability that a pixel belongs to a blood vessel) and the segmentation label (the label marked on the image by a senior doctor is the gold standard) are used as input to construct a blood vessel distribution map. The distribution graph consists of nodes and edges, as shown in Figure 2. The nodes in the figure have the following characteristics: first, the probability of blood vessels in the pixels in the nodes is high; second, the gray levels of the pixels in the nodes are consistent; third, the shape of the nodes conforms to the direction of blood vessels. An edge in the graph is connected between two nodes, indicating the relevance of the connected nodes, and the length of the edge is less than the set threshold. In this embodiment, a pre-segmentation model is used to construct a blood vessel distribution map, which can model or fit the real structure of blood vessels, and is beneficial to improve the accuracy of blood vessel segmentation.

In this embodiment, step 103 is mainly used for the first stage of blood vessel segmentation. In this embodiment, blood vessel segmentation is performed based on cross-network multi-scale feature fusion. As shown in Figure 2, blood vessel segmentation consists of two stages: the first stage uses the second convolutional network (UNET-2); the second stage uses the third convolutional network (UNET-3) and graph convolutional network (UNET- G). The network structure of the second convolutional network is the same as that of the pre-segmented first convolutional network, and is initialized with the weights of the pre-segmented model of the first convolutional network. Input the CTA image to the second convolutional network to obtain the prediction result of the first stage. After the prediction result of the first stage is multiplied by the original image, it is input to the third convolutional network of the second stage. In this embodiment, the prediction result of the first stage is not directly used as the input of the third convolutional network, but multiplied by the original image and then used as the input of the third convolutional network, in order to prevent loss of information of the original image.

In this embodiment, step 104 is mainly used for the second-stage blood vessel segmentation of the cross-network model. As shown in Figure 2, the second stage includes a third convolutional network and a graph convolutional network. The graph convolutional network takes the blood vessel distribution map constructed in step 102 as input to perform global feature modeling, and performs two-way propagation and fusion of local features and global features through bidirectional (forward and reverse) mapping with the third convolutional network, In this way, the prediction of blood vessel area is realized.

In this embodiment, step 105 is mainly used to predict the blood vessel segmentation result. The network structure and weight parameters of the second convolutional network in the first stage are the same as those of the pre-segmented first convolutional network, so the second convolutional network also performs multi-scale feature extraction. Therefore, after the fusion of local features and global features, multi-scale features are fused to obtain the final vessel segmentation.

As an optional embodiment, the method also includes the step of normalizing the input CTA image, normalizing the grayscale of the pixel to [0,255], the calculation formula is as follows:

In the formula,

This embodiment provides an image preprocessing method. In this embodiment, normalization processing is performed on the CTA image before it is input, and the grayscale of the pixel is normalized to [0, 255], and the normalization formula is as the above formula. Obviously, when x=x _min ,

When x=x _max ,

Therefore, the above formula can normalize the gray value of all pixels in the original image to [0,255].

As an optional embodiment, the blood vessel pre-segmentation method specifically includes:

This embodiment provides a specific technical solution for blood vessel pre-segmentation. As mentioned above, blood vessel pre-segmentation is achieved by the first convolutional network. Specifically, the first convolutional network (the same for the other two convolutional networks and the graph convolutional network) includes an encoding layer and a decoding layer. The CTA sequence is first input to the encoding layer, and the encoding layer uses different convolution kernels to convolve the image. Multi-scale feature extraction is realized after operation and pooling operation. The size of extracted feature spaces of different scales constitutes a geometric series with a common ratio of 1/2, the space size of the i-th scale is 1/2 ⁱ of the original space size, and the number N of different scales is generally 4-5. The extracted features contain the shape and spatial location information of vessel segmentation. Then the multi-scale features extracted by the encoding layer are input to the decoding layer, and the decoding layer maps the global features of each scale back to the original feature size, and combines them with the encoding features at the corresponding scales (feature cascading or splicing) to obtain segmentation results. The decoding layer maps the global features back to the original feature size through upsampling (interpolation) or deconvolution.

As an optional embodiment, the method for predicting a blood vessel region specifically includes:

This embodiment provides a technical solution for prediction of cross-network blood vessel regions. Cross-network mainly refers to feature interaction between the third convolutional network and the graph convolutional network through bidirectional mapping. First, the third convolutional network performs forward mapping, and projects multiple pixels of encoded features into a single node in the graph convolutional network; then, the graph convolutional network performs global feature modeling based on the blood vessel distribution map, and performs inverse To map the node features back to the feature space of the third convolutional network, propagate the global features to each local feature for feature enhancement; finally, capture the local features of the image through the third convolutional network, and use the graph convolution The network captures the global features of blood vessels, and uses feature interaction to combine local features and global features (feature cascade) to realize the prediction of the area where blood vessels are located.

As an optional embodiment, the method uses the Dice coefficient to measure the similarity between the blood vessel segmentation result and the gold standard marked by the doctor, and the formula is as follows:

This embodiment provides a technical solution for quantitatively evaluating blood vessel segmentation results. In this embodiment, the Dice coefficient is used to represent the similarity between the blood vessel segmentation result and the gold standard marked by doctors, and the formula is as above. The numerator in the formula is twice the number of pixels where the segmentation result coincides with the gold standard, and the denominator is the sum of the pixel points of the segmentation result and the gold standard. According to the formula, the more pixels that overlap the segmentation result and the gold standard, the greater the similarity; when the two completely overlap, the maximum similarity is 1.

Fig. 3 is a schematic diagram of the composition of a blood vessel segmentation device according to an embodiment of the present invention, and the device includes:

The pre-segmentation module 11 is used to input the CTA image to the first convolutional network for pre-segmentation of blood vessels based on multi-scale feature extraction;

The distribution map construction module 12 is used to construct a blood vessel distribution map with pre-segmentation results and segmentation labels as input. The nodes of the map are pixel areas with high blood vessel probability and consistent gray scale, and the shape of the nodes is consistent with the direction of the blood vessels; Indicates the relevance of the connected nodes, and the length of the edge is less than the set threshold;

The first stage segmentation module 13 is used to input the CTA image to the second convolutional network with the same network structure and weight parameters as the first convolutional network, and input the output after multiplying the output of the second convolutional network with the input CTA image to the third convolutional network;

The second-stage segmentation module 14 is used for feature interaction between the third convolutional network and the graph convolutional network based on bidirectional mapping, and global feature modeling based on the blood vessel distribution map to capture multi-scale local and global features and realize blood vessel region prediction;

The feature fusion module 15 is configured to predict the blood vessel segmentation result by fusing multi-scale features.

The device of this embodiment can be used to implement the technical solution of the method embodiment shown in FIG. 1 , and its implementation principle and technical effect are similar, and will not be repeated here. The same is true for the following embodiments, which will not be further described.

As an optional embodiment, the device also includes a normalization module for preprocessing the input CTA image, which is used to normalize the grayscale of the pixels to [0,255], and the calculation formula is as follows:

In the formula,

FIG. 4 is a schematic structural diagram of an electronic device provided by an embodiment of the present application. At the hardware level, the electronic device includes a processor, and optionally also includes an internal bus, a network interface, and a memory. Wherein, the memory may include a memory, such as a high-speed random-access memory (Random-Access Memory, RAM), and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory. Of course, the electronic device may also include hardware required by other services.

The processor, the network interface and the memory can be connected to each other through an internal bus, which can be an ISA (Industry Standard Architecture, industry standard architecture) bus, a PCI (Peripheral Component Interconnect, peripheral component interconnection standard) bus or an EISA (Extended Industry Standard Architecture, extended industry standard architecture) bus, etc. The bus can be divided into address bus, data bus, control bus and so on. For ease of representation, only one double-headed arrow is used in FIG. 4 , but it does not mean that there is only one bus or one type of bus.

Memory for storing execution instructions. Specifically, a computer program that can be executed by executing instructions. The memory, which can include internal memory and non-volatile memory, provides instructions and data to the processor for execution.

In a possible implementation, the processor reads the corresponding execution instructions from the non-volatile memory into the memory and then runs them. It can also obtain the corresponding execution instructions from other devices to form blood vessel segmentation at the logical level. device. The processor executes the execution instructions stored in the memory, so as to implement the blood vessel segmentation method provided in any embodiment of the present application through the executed execution instructions.

The above-mentioned method performed by the blood vessel segmentation apparatus provided in the embodiment shown in FIG. 1 of the present application may be applied to a processor or implemented by the processor. A processor may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by an integrated logic circuit of hardware in a processor or an instruction in the form of software. The above-mentioned processor can be a general-purpose processor, including a central processing unit (Central Processing Unit, CPU), a network processor (Network Processor, NP), etc.; it can also be a digital signal processor (Digital Signal Processor, DSP), a dedicated integrated Circuit (Application Specific Integrated Circuit, ASIC), Field-Programmable Gate Array (Field-Programmable Gate Array, FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. Various methods, steps, and logic block diagrams disclosed in the embodiments of the present application may be implemented or executed. A general-purpose processor may be a microprocessor, or the processor may be any conventional processor, or the like.

The steps of the method disclosed in connection with the embodiments of the present application may be directly implemented by a hardware decoding processor, or implemented by a combination of hardware and software modules in the decoding processor. The software module can be located in a mature storage medium in the field such as random access memory, flash memory, read-only memory, programmable read-only memory or electrically erasable programmable memory, register. The storage medium is located in the memory, and the processor reads the information in the memory, and completes the steps of the above method in combination with its hardware.

The embodiment of the present application also proposes a readable medium, the readable storage medium stores execution instructions, and when the stored execution instructions are executed by the processor of the electronic device, the electronic device can execute the electronic device provided in any embodiment of the present application. The blood vessel segmentation method, and is specifically used to implement the above blood vessel segmentation method.

The electronic equipment described in the foregoing embodiments may be a computer.

Those skilled in the art should understand that the embodiments of the present application may be provided as methods or computer program products. Therefore, the present application may adopt an entirely hardware embodiment, an entirely software embodiment, or a combination of software and hardware.

Each embodiment in the present application is described in a progressive manner, the same and similar parts of each embodiment can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, as for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and for relevant parts, please refer to part of the description of the method embodiment.

It should also be noted that the term "comprises", "comprises" or any other variation thereof is intended to cover a non-exclusive inclusion such that a process, method, article, or apparatus comprising a set of elements includes not only those elements, but also includes Other elements not expressly listed, or elements inherent in the process, method, commodity, or apparatus are also included. Without further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article or apparatus comprising said element.

The above descriptions are only examples of the present application, and are not intended to limit the present application. For those skilled in the art, various modifications and changes may occur in this application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included within the scope of the claims of the present application.

Industrial Applicability

A blood vessel segmentation method, device, electronic equipment, and readable medium provided by the present invention use image recognition technology to build a blood vessel distribution network to model the global characteristics of blood vessels, use the overall structure of blood vessels to enhance the features of the segmented area, and mine blood vessels The potential direction of the structure makes full use of the processing basis of program automation in computer technology, which greatly improves the accuracy of the prediction and segmentation of vessel segmentation regions under the complex imaging quality environment of angiographic images. The formed products can be mass-produced and quickly applied to systems or scenarios with high demand for vascular imaging.

Claims

A blood vessel segmentation method, characterized in that, comprising the following steps:

Input the CTA image to the first convolutional network for pre-segmentation of blood vessels based on multi-scale feature extraction;

The pre-segmentation results and segmentation labels are used as input to construct a blood vessel distribution map. The nodes of the graph are pixel areas with high probability of blood vessels and consistent gray levels. The shape of the nodes is consistent with the direction of the blood vessels; the edges of the graph represent the correlation of connected nodes. The length of the edge is less than the set threshold;

The CTA image is input to the second convolutional network with the same network structure and weight parameters as the first convolutional network, and the output of the second convolutional network is multiplied by the input CTA image and then input to the third convolutional network;

The third convolutional network and the graph convolutional network perform feature interaction based on bidirectional mapping, and perform global feature modeling based on the vascular distribution map to capture multi-scale local and global features, and realize vascular region prediction;

Through the fusion of multi-scale features, the result of blood vessel segmentation is predicted.
The blood vessel segmentation method according to claim 1, characterized in that the method also includes the step of normalizing the input CTA image, normalizing the grayscale of the pixels to [0,255], and the calculation formula is as follows:

In the formula,
is the gray value x of any pixel after normalization, and x min and x max are the minimum and maximum values of the gray value of the pixel before normalization, respectively.
The blood vessel segmentation method according to claim 1, wherein the blood vessel pre-segmentation method specifically comprises:

Input the CTA sequence into the encoding layer of the first convolutional network;

The encoding layer performs multi-scale feature extraction through convolution and pooling operations, the space size of the i-th scale is 1/2 i of the original space size, i=1,2,...,N, N is the number of scales; the Features include shape and spatial location information of vessel segmentation;

The multi-scale features extracted by the encoding layer are input to the decoding layer, and the decoding layer maps the global features of each scale back to the original feature size through upsampling or deconvolution, and combines them with the encoding features at the corresponding scale to obtain the segmentation result.
The blood vessel segmentation method according to claim 1, wherein the method for predicting the blood vessel region specifically comprises:

The third convolutional network projects multiple pixels of encoded features into a single node in the graph convolutional network through forward mapping;

The graph convolutional network propagates the features on the blood vessel distribution map for global feature modeling;

The graph convolutional network maps the node features back to the feature space of the third convolutional network through reverse mapping, and propagates the global features to each part for feature enhancement;

Combine the features of the third convolutional network and the graph convolutional network, capture the local features of the image through the third convolutional network, capture the global features of blood vessels through the graph convolutional network, and use feature interaction to combine local features and global features , to complete the prediction of the area where the blood vessel is located.
The blood vessel segmentation method according to claim 1, wherein the method uses the Dice coefficient to measure the similarity between the blood vessel segmentation result and the gold standard marked by a doctor, and the formula is as follows:

In the formula, A and B are the blood vessel pixel point set of the gold standard and the pixel point set of the blood vessel segmentation result respectively, and |A|, |B| and |A∩B| are A, B and the intersection of A and B respectively The number of pixels.
A blood vessel segmentation device, characterized in that it comprises:

The pre-segmentation module is used to input the CTA image to the first convolutional network for pre-segmentation of blood vessels based on multi-scale feature extraction;

The distribution map building module is used to construct a blood vessel distribution map with pre-segmentation results and segmentation labels as input. The nodes of the map are pixel areas with a high probability of blood vessels and consistent gray levels. The shape of the nodes is consistent with the direction of the blood vessels; the edges of the graph represent The relevance of the connected nodes, the length of the edge is less than the set threshold;

The first-stage segmentation module is used to input the CTA image to the second convolutional network with the same network structure and weight parameters as the first convolutional network, and to multiply the output of the second convolutional network with the input CTA image and then input it to The third convolutional network;

The second stage segmentation module is used for the third convolutional network and the graph convolutional network to perform feature interaction based on two-way mapping, and to perform global feature modeling based on the blood vessel distribution map to capture multi-scale local and global features, and realize blood vessel region prediction;

The feature fusion module is used to predict the blood vessel segmentation result by fusing multi-scale features.
The blood vessel segmentation device according to claim 6, wherein the device further comprises a normalization module for preprocessing the input CTA image, which is used to normalize the grayscale of the pixel to [0, 255], and calculate The formula is as follows:

In the formula,
is the gray value x of any pixel after normalization, and x min and x max are the minimum and maximum values of the gray value of the pixel before normalization, respectively.
The blood vessel segmentation device according to claim 6, wherein the blood vessel pre-segmentation method specifically comprises:

Input the CTA sequence into the encoding layer of the first convolutional network;

The encoding layer performs multi-scale feature extraction through convolution and pooling operations, the space size of the i-th scale is 1/2 i of the original space size, i=1,2,...,N, N is the number of scales; the Features include shape and spatial location information of vessel segmentation;

The multi-scale features extracted by the encoding layer are input to the decoding layer, and the decoding layer maps the global features of each scale back to the original feature size through upsampling or deconvolution, and combines them with the encoding features at the corresponding scale to obtain the segmentation result.
The blood vessel segmentation device according to claim 6, wherein the method for predicting the blood vessel region specifically comprises:

The third convolutional network projects multiple pixels of encoded features into a single node in the graph convolutional network through forward mapping;

The graph convolutional network propagates the features on the blood vessel distribution map for global feature modeling;

The graph convolutional network maps the node features back to the feature space of the third convolutional network through reverse mapping, and propagates the global features to each part for feature enhancement;

Combine the features of the third convolutional network and the graph convolutional network, capture the local features of the image through the third convolutional network, capture the global features of blood vessels through the graph convolutional network, and use feature interaction to combine local features and global features , to complete the prediction of the area where the blood vessel is located.
The blood vessel segmentation device according to claim 6, wherein the method uses the Dice coefficient to measure the similarity between the blood vessel segmentation result and the gold standard marked by the doctor, and the formula is as follows:

In the formula, A and B are the blood vessel pixel point set of the gold standard and the pixel point set of the blood vessel segmentation result respectively, and |A|, |B| and |A∩B| are A, B and the intersection of A and B respectively The number of pixels.
A readable medium, characterized in that the readable medium includes execution instructions, and when the processor of the electronic device executes the execution instructions, the electronic device executes the blood vessel according to any one of claims 1-5. split method.
An electronic device, characterized in that the electronic device includes a processor and a memory storing execution instructions, and when the processor executes the execution instructions stored in the memory, the processor executes the method according to claim 1. - the blood vessel segmentation method described in any one of 5.