Blood vessel segmentation method, device, system, equipment, medium and product

Abstract

The invention discloses a blood vessel segmentation method, device and equipment, a medium and a product. The invention relates to the technical field of image processing. The method comprises the following steps: acquiring a to-be-segmented image; the to-be-segmented image is input to a deformation partitioning module, a partitioned image aligned with the blood vessel form is obtained, and the deformation partitioning module is used for dividing the to-be-segmented image into a plurality of areas aligned with the blood vessel form; the partition image aligned with the blood vessel form is input into a blood vessel segmentation model, a blood vessel segmentation result is obtained, the blood vessel segmentation model comprises at least one space and semantic converter module, and the space and semantic converter module is used for capturing the space relation of the area and the semantic relation of the area. According to the technical scheme, self-adaptive image partitioning aligned with the blood vessel form is achieved, the spatial relation and the semantic relation of all the partitions are captured, therefore, the topological structures of slender blood vessels and bent blood vessels can be accurately captured, and the blood vessel segmentation precision is improved.

Description

Technical Field

[0001] The present disclosure relates to the field of image processing technologies, and in particular, to a blood vessel segmentation method, apparatus, device, medium, and product. Background Art

[0002] With the rapid development of deep learning technologies, the application of deep learning technologies in image processing research has become increasingly widespread, and has become an important means for segmenting complex blood vessel structures from images.

[0003] In current blood vessel segmentation technologies, Transformer or convolutional neural networks are often used for segmentation, and there are problems with the accuracy of blood vessel segmentation. Summary of the Invention

[0004] The present disclosure provides a blood vessel segmentation method, apparatus, device, medium, and product, which improves the accuracy of blood vessel segmentation.

[0005] According to one aspect of the present disclosure, there is provided a blood vessel segmentation method, including:

[0006] Obtain an image to be segmented;

[0007] Input the image to be segmented into a deformation partitioning module to obtain a partitioned image aligned with the blood vessel morphology, where the deformation partitioning module is used to divide the image to be segmented into multiple regions aligned with the blood vessel morphology;

[0008] Input the partitioned image aligned with the blood vessel morphology into a blood vessel segmentation model to obtain a blood vessel segmentation result, where the blood vessel segmentation model includes at least one spatial and semantic converter module, and the spatial and semantic converter module is used to capture the spatial relationship and semantic relationship of the regions.

[0009] According to another aspect of the present disclosure, there is provided a blood vessel segmentation apparatus, including:

[0010] An image to be segmented acquisition module, configured to obtain an image to be segmented;

[0011] An image deformation partitioning module, configured to input the image to be segmented into a deformation partitioning module to obtain a partitioned image aligned with the blood vessel morphology, where the deformation partitioning module is used to divide the image to be segmented into multiple regions aligned with the blood vessel morphology;

[0012] An image blood vessel segmentation module, configured to input the partitioned image aligned with the blood vessel morphology into a blood vessel segmentation model to obtain a blood vessel segmentation result, where the blood vessel segmentation model includes at least one spatial and semantic converter module, and the spatial and semantic converter module is used to capture the spatial relationship and semantic relationship of the regions.

[0013] According to another aspect of the present disclosure, there is provided an electronic device, which includes:

[0014] at least one processor;

[0015] and a memory communicatively connected to the at least one processor;

[0016] wherein, the memory stores a computer program executable by the at least one processor, and when the computer program is executed by the at least one processor, the at least one processor is enabled to execute the blood vessel segmentation method according to any embodiment of the present disclosure.

[0017] According to another aspect of the present disclosure, there is provided a computer-readable storage medium storing computer instructions for implementing the blood vessel segmentation method according to any embodiment of the present disclosure when executed by a processor.

[0018] According to another aspect of the present disclosure, there is provided a computer program product including a computer program, which implements the blood vessel segmentation method according to any one of the embodiments of the present disclosure when executed by a processor.

[0019] In the technical solution of the embodiments of the present disclosure, by acquiring an image to be segmented, and then inputting the image to be segmented into a deformation partitioning module to obtain a partitioned image aligned with the blood vessel morphology, wherein the deformation partitioning module is used to divide the image to be segmented into multiple regions aligned with the blood vessel morphology, and then inputting the partitioned image aligned with the blood vessel morphology into a blood vessel segmentation model to obtain a blood vessel segmentation result, wherein the blood vessel segmentation model includes at least one spatial and semantic converter module, and the spatial and semantic converter module is used to capture the spatial relationship and semantic relationship of the regions. In the above technical solution, an adaptive image partitioning aligned with the blood vessel morphology is achieved, and the spatial relationship and semantic relationship of each partition are captured. Compared with a fixed image partitioning scheme, the situation of blood vessel breakage or mis-segmentation can be avoided, especially for slender blood vessels and curved blood vessels, so that the topological structure of slender blood vessels and curved blood vessels can be accurately captured, and the blood vessel segmentation accuracy can be improved. In addition, by capturing the spatial relationship and semantic relationship of the regions through the spatial and semantic converter module, the understanding ability of the blood vessel segmentation model for the blood vessel structure is improved, and further the segmentation accuracy of the blood vessel segmentation model is improved.

[0020] It should be understood that the content described in this part is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it used to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] To more clearly illustrate the technical solutions in the embodiments of the present disclosure, the following will briefly introduce the accompanying drawings required for the description of the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present disclosure. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can be obtained based on these drawings.

[0022] Figure 1 is a flowchart of a blood vessel segmentation method provided according to an embodiment of the present disclosure;

[0023] Figure 2 is a schematic diagram for comparing the results of a deformed partition processing provided according to an embodiment of the present disclosure;

[0024] Figure 3 is a flowchart of another blood vessel segmentation method provided according to an embodiment of the present disclosure;

[0025] Figure 4 is a schematic structural diagram of a deformed partition module provided according to an embodiment of the present disclosure;

[0026] Figure 5 is a flowchart of another blood vessel segmentation method provided according to an embodiment of the present disclosure;

[0027] Figure 6 is a schematic structural diagram of a space and semantic converter module provided according to an embodiment of the present disclosure;

[0028] Figure 7 is a schematic structural diagram of a space and semantic attention unit provided according to an embodiment of the present disclosure;

[0029] Figure 8 is a flowchart of another blood vessel segmentation method provided according to an embodiment of the present disclosure;

[0030] Figure 9 is a schematic structural diagram of a deformed partition module and a blood vessel segmentation model provided according to an embodiment of the present disclosure;

[0031] Figure 10 is a flowchart of a blood vessel segmentation method provided according to an embodiment of the present disclosure;

[0032] Figure 11 is a comparison chart of the experimental results of blood vessel segmentation provided according to an embodiment of the present disclosure;

[0033] Figure 12 is a comparison chart of the experimental results of another blood vessel segmentation provided according to an embodiment of the present disclosure;

[0034] Figure 13 is a schematic structural diagram of a blood vessel segmentation device provided according to an embodiment of the present disclosure;

[0035] Figure 14 It is a schematic structural diagram of an electronic device for implementing the blood vessel segmentation method of the embodiments of the present disclosure. Detailed implementation manners

[0036] In order to enable those skilled in the art to better understand the solutions of the present disclosure, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only a part of the embodiments of the present disclosure, rather than all the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present disclosure.

[0037] It should be noted that the terms "first", "second", etc. in the specification and claims of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily need to be used to describe a specific order or sequence. It should be understood that such used data can be interchanged under appropriate circumstances so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device including a series of steps or units does not necessarily need to be limited to those clearly listed steps or units, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices. The acquisition, storage, use, processing, etc. of data in the technical solutions of the present disclosure all comply with the relevant regulations of national laws and regulations.

[0038] The technical field and background technology of the embodiments of the present disclosure will be described below.

[0039] In the existing blood vessel segmentation technology, the fixed partitioning or blocking of traditional Transformers can lead to blood vessel breakage or mis-segmentation, especially for slender or curved blood vessels, which destroys the semantic integrity of blood vessels. Some blood vessel segmentation technologies introduce deformable convolutional kernels to capture morphological features, but rely on rectangular sampling regions and have limited effects on the segmentation of slender blood vessels. There are also some blood vessel segmentation technologies that combine window attention mechanisms and pyramid structures, but the fixed window design restricts the dynamic aggregation of multi-scale semantic features, seriously affecting the accuracy of blood vessel segmentation. For this reason, the embodiments of the present disclosure provide a blood vessel segmentation method, device, equipment, medium and product, which can effectively solve this problem. The blood vessel segmentation method, device, equipment, medium and product provided by the embodiments of the present disclosure will be further described in detail below.

[0040] Figure 1The flowchart of a vascular segmentation method provided by an embodiment of the present disclosure. This embodiment is applicable to the situation of automatically segmenting vascular structures in an image. This method can be executed by a vascular segmentation device, which can be implemented in the form of hardware and / or software, and can be configured in electronic devices such as terminals and servers. As Figure 1 shown, the method includes:

[0041] S110. Obtain the image to be segmented.

[0042] The image to be segmented refers to the image to be subjected to vascular segmentation operation, which can be a two-dimensional Computed Tomography (CT) or a three-dimensional Computed Tomography Angiography (CTA) and other images containing vascular structures. The blood vessels can be the aorta, pulmonary artery, or left and right coronary arteries, etc., which are not specifically limited here.

[0043] Exemplarily, the image to be segmented can be read from a preset storage path of the electronic device, or can also be obtained from a device communicatively connected to the electronic device or the cloud, which is not specifically limited here.

[0044] S120. Input the image to be segmented into a deformation partitioning module to obtain a partitioned image aligned with the vascular morphology, where the deformation partitioning module is used to divide the image to be segmented into multiple regions aligned with the vascular morphology.

[0045] The deformation partitioning module is a module with the function of image partitioning or blocking. It can divide the image to be segmented into multiple regions aligned with the vascular morphology. In other words, the shape of any region will be automatically adjusted according to the structures such as the bending or branching of the blood vessels, so that the edges of the region are aligned with the vascular morphology. For example, at the bending of the blood vessels, the edges of the region maintain the same bending angle as the bending of the blood vessels. The partitioned image aligned with the vascular morphology is an image containing multiple regions aligned with the vascular morphology, and each region can be a part or local region of the image to be segmented.

[0046] It should be noted that dividing the image to be segmented into multiple regions aligned with the vascular morphology enables the divided regions to adapt to the complex and irregular shapes of the blood vessels, thereby improving the ability to model the geometric features of the blood vessels. Compared with the fixed partitioning scheme, deformation partitioning can avoid the occurrence of blood vessel breakage or mis-segmentation, especially for slender blood vessels and curved blood vessels, so as to accurately capture the topological structures of slender blood vessels and curved blood vessels and improve the accuracy of vascular segmentation.

[0047] Exemplarily, Figure 2 is a schematic diagram for comparing the processing results of deformation partitioning according to an embodiment of the present disclosure. AsFigure 2 As shown, A represents a schematic diagram of the fixed partition of the traditional Transformer, and B represents a schematic diagram of the deformable partition by the deformable partition module. In other words, B is the partition image aligned with the blood vessel morphology.

[0048] Specifically, the deformable partition module can predict the velocity field according to the image to be segmented, obtain the velocity field corresponding to the image to be segmented, and then can predict the deformation field according to the velocity field corresponding to the image to be segmented, obtain the deformation field corresponding to the velocity field, and then can divide the image to be segmented according to the deformation field corresponding to the velocity field to obtain the partition image aligned with the blood vessel morphology.

[0049] S130. Input the partition image aligned with the blood vessel morphology into the blood vessel segmentation model to obtain a blood vessel segmentation result, where the blood vessel segmentation model includes at least one spatial and semantic transformer module, and the spatial and semantic transformer module is used to capture the spatial relationship of the region and the semantic relationship of the region.

[0050] Among them, the blood vessel segmentation model refers to a trained deep learning model for blood vessel segmentation of images. Its network architecture can adopt an encoder-decoder structure or other structures, which are not specifically limited here. The spatial and semantic transformer module is a module in the blood vessel segmentation model and can be used to capture the spatial relationship of the region and the semantic relationship of the region. The spatial relationship of the region can include the spatial relationship between regions and / or the spatial relationship within the region, and the semantic relationship of the region can include the semantic relationship between regions and / or the semantic relationship within the region. The blood vessel segmentation result output by the blood vessel segmentation model is a blood vessel segmentation image, which can be a blood vessel segmentation mask image or other blood vessel segmentation images, which are not specifically limited here.

[0051] It should be noted that through the spatial and semantic transformer module, the spatial relationship of the region and the semantic relationship of the region can be captured, thereby improving the understanding ability of the blood vessel segmentation model for the blood vessel structure, and further improving the segmentation accuracy of the blood vessel segmentation model.

[0052] In the embodiments of the present disclosure, the deformation partitioning module and the blood vessel segmentation model can be trained in an end-to-end manner or separately, which is not specifically limited herein. Specifically, the training steps of the deformation partitioning module and the blood vessel segmentation model may include: inputting a plurality of sample images to be segmented into the deformation partitioning module to be trained. For any sample image to be segmented, the deformation partitioning module to be trained outputs a partitioned image corresponding to the sample image to be segmented and aligned with the blood vessel morphology. Then, the partitioned image corresponding to the sample image to be segmented and aligned with the blood vessel morphology is input into the blood vessel segmentation model to be trained. The blood vessel segmentation model to be trained outputs a predicted blood vessel segmentation result. Based on the predicted blood vessel segmentation result and the true blood vessel segmentation result, the model loss is calculated. The Dice coefficient can be used as the loss function to calculate the model loss. Then, based on the model loss, the model parameters of the deformation partitioning module and the blood vessel segmentation model are updated until the training stop condition is met, and the trained deformation partitioning module and blood vessel segmentation model are obtained.

[0053] In the technical solution of the embodiments of the present disclosure, by obtaining an image to be segmented, and then inputting the image to be segmented into the deformation partitioning module, a partitioned image aligned with the blood vessel morphology is obtained. Among them, the deformation partitioning module is used to divide the image to be segmented into multiple regions aligned with the blood vessel morphology. Then, the partitioned image aligned with the blood vessel morphology is input into the blood vessel segmentation model to obtain a blood vessel segmentation result. Among them, the blood vessel segmentation model includes at least one spatial and semantic converter module, and the spatial and semantic converter module is used to capture the spatial relationship and semantic relationship of the regions. In the above technical solution, an adaptive image partition aligned with the blood vessel morphology is realized, and the spatial relationship and semantic relationship of each partition are captured. Compared with the fixed image partition scheme, the situation of blood vessel breakage or mis-segmentation can be avoided, especially for slender blood vessels and curved blood vessels, so that the topological structure of slender blood vessels and curved blood vessels can be accurately captured, and the blood vessel segmentation accuracy can be improved. In addition, by capturing the spatial relationship and semantic relationship of the regions through the spatial and semantic converter module, the understanding ability of the blood vessel segmentation model for the blood vessel structure is improved, and further the segmentation accuracy of the blood vessel segmentation model is improved.

[0054] Figure 3 The flowchart of another blood vessel segmentation method provided by the embodiments of the present disclosure. The method of this embodiment can be combined with each optional solution in the blood vessel segmentation method provided in the above embodiments. On the basis of the above embodiments, this embodiment further refines the deformation partitioning module.

[0055] As Figure 3 shown, the method includes:

[0056] S210. Obtain an image to be segmented.

[0057] S220. Input the image to be segmented into the velocity field prediction unit to obtain the velocity field corresponding to the image to be segmented.

[0058] In the embodiments of the present disclosure, the velocity field prediction unit is a functional unit for predicting the velocity field. The velocity field represents the direction and amplitude of the deformation required for a region to adapt to the complex shape of blood vessels.

[0059] The velocity field prediction unit can be a convolutional neural network or a neural network with other structures, which is not specifically limited herein.

[0060] Exemplarily, the size of the image to be segmented can be 128×128×128 voxels. Input the image to be segmented into the convolutional neural network, and the convolutional neural network outputs the velocity field corresponding to the image to be segmented. The velocity field is also a three-dimensional vector field with the same size as the input image to be segmented. It can be understood that the velocity vector at each voxel position p=(x, y, z) can be v(p)=(vx, vy, vz), representing the instantaneous displacement rate of this position during the deformation process. Specific example: Assume that a certain blood vessel region in the image to be segmented needs to bend slightly upward and to the right. The velocity field predicted by the convolutional neural network may be as follows: At the bend of the blood vessel, the velocity vector v(p)=(0.2, 0.1, 0.0), indicating that the displacement rate in the x direction is 0.2, the displacement rate in the y direction is 0.1, and the z direction remains unchanged.

[0061] S230. Input the velocity field corresponding to the image to be segmented into the scaling and squaring unit to obtain the deformation field corresponding to the velocity field.

[0062] In the embodiments of the present disclosure, the scaling and squaring unit is a functional unit that converts the velocity field into a diffeomorphic deformation field by using the scaling and squaring method.

[0063] Exemplarily, the scaling and squaring unit obtains the deformation field corresponding to the velocity field by integrating the velocity field over time t=[0, 1]. The ordinary differential equation can be:

[0064] ;

[0065] where Id represents the deformation state at the initial time (t = 0), at which time no deformation has been applied and all coordinates remain in their original positions. represents the deformation field at time t. The solution process includes:

[0066] Starting from the initial time step calculate the initial deformation field:

[0067] ;

[0068] Among them, represents the Composition Operation of the deformation field, that is, the sequential application of two deformation fields. Exemplarily, represents applying to transform the coordinates first, and then applying to further transform the coordinates.

[0069] Furthermore, use the recurrence relation: to iterate until t = 1 to obtain the final deformation field . It should be noted that is the small deformation field generated in the previous step. By composing the small deformation field with itself, the time step is "doubled". For example, if the initial step size is Δt = 1 / 8, after composition, it is equivalent to a time step of Δt = 1 / 4. By repeatedly composing with itself until the total time t = 1 is reached, the complete deformation field can be obtained. This iterative process can ensure smooth and reversible regional deformation and maintain the topological integrity of the vascular structure.

[0070] S240. Interpolate the image to be segmented based on the deformation field corresponding to the velocity field to obtain a partitioned image aligned with the vascular morphology.

[0071] Among them, the interpolation can be bilinear interpolation, trilinear interpolation or other interpolation methods, which are not specifically limited here.

[0072] Specifically, the deformed coordinates of each voxel position in the image to be segmented can be calculated through the deformation field , and then is applied to the image to be segmented through bilinear interpolation or trilinear interpolation to obtain a partitioned image aligned with the vascular morphology.

[0073] Exemplarily, Figure 4 is a schematic structural diagram of a deformation partitioning module provided according to an embodiment of the present disclosure. Specifically, the image to be segmented is input into a convolutional neural network, the convolutional neural network outputs the velocity field corresponding to the image to be segmented, and then the velocity field is input into a scaling and squaring unit. The scaling and squaring unit integrates the velocity field to generate a deformation field, and then the deformation field is interpolated with the image to be segmented to obtain a partitioned image aligned with the vascular morphology.

[0074] S250. Input the partitioned image aligned with the vascular morphology into a vascular segmentation model to obtain a vascular segmentation result, where the vascular segmentation model includes at least one spatial and semantic converter module, and the spatial and semantic converter module is used to capture the spatial relationship and semantic relationship of the region.

[0075] In the technical solution of the embodiment of the present disclosure, by inputting the image to be segmented into the velocity field prediction unit, the velocity field corresponding to the image to be segmented is obtained. Then, the velocity field corresponding to the image to be segmented is input into the scaling and squaring unit to obtain the deformation field corresponding to the velocity field. Further, interpolation is performed on the image to be segmented based on the deformation field corresponding to the velocity field to obtain a partitioned image aligned with the blood vessel morphology, realizing adaptive image partitioning aligned with the blood vessel morphology, and capturing the spatial relationship and semantic relationship of each partition. Compared with the fixed image partitioning scheme, it can avoid the occurrence of blood vessel breakage or mis-segmentation, especially for slender blood vessels and curved blood vessels. Therefore, the topological structure of slender blood vessels and curved blood vessels can be accurately captured, and the blood vessel segmentation accuracy can be improved.

[0076] Figure 5 FIG. is a flowchart of another blood vessel segmentation method provided by an embodiment of the present disclosure. The method of this embodiment can be combined with each optional solution in the blood vessel segmentation method provided in the above embodiment. On the basis of the above embodiments, this embodiment further refines the spatial and semantic converter module.

[0077] As Figure 5 shown, the method includes:

[0078] S310. Obtain the image to be segmented.

[0079] S320. Input the image to be segmented into the deformation partitioning module to obtain a partitioned image aligned with the blood vessel morphology, where the deformation partitioning module is used to divide the image to be segmented into multiple regions aligned with the blood vessel morphology.

[0080] S330. Input the partitioned image aligned with the blood vessel morphology into the blood vessel segmentation model to obtain a blood vessel segmentation result, where the blood vessel segmentation model includes at least one spatial and semantic converter module, and the spatial and semantic converter module includes a deformation partitioning unit, at least one spatial and semantic attention unit, and a window offset unit; the deformation partitioning unit is used to adjust the window shape to be aligned with the blood vessel morphology; the spatial and semantic attention unit is used to capture the spatial relationship and semantic relationship of the region; the window offset unit is used to offset the position of the window.

[0081] In window attention, the window is a design of a local attention range, which is used to limit the voxels in the image to only focus on a local area around them when calculating attention, rather than all elements globally. This design aims to reduce the computational complexity and capture local features, and at the same time, the global information can be gradually fused by stacking or moving the window.

[0082] In the embodiments of the present disclosure, the deformation partition unit has the same structure as the deformation partition module, including a velocity field prediction unit, a scaling and squaring unit, interpolation, etc. It should be noted that the window shape is adjusted to align with the blood vessel morphology through the deformation partition unit, so that the window adapts to the blood vessel morphology, thereby improving the efficiency of the attention mechanism.

[0083] Exemplarily, Figure 6 FIG. is a schematic structural diagram of a spatial and semantic transformer module provided according to an embodiment of the present disclosure. The spatial and semantic transformer module may include a deformation partition unit, a first spatial and semantic attention unit, a first multi-layer perceptron, a window offset unit, a second spatial and semantic attention unit, and a second multi-layer perceptron. The specific connection relationship is as Figure 6 shown.

[0084] Based on the above embodiments, optionally, the spatial and semantic attention unit includes a window attention sub-unit and a semantic clustering attention sub-unit; the window attention sub-unit is used to capture the spatial relationship within the region and the spatial relationship between regions; the semantic clustering attention sub-unit is used to aggregate regions with similar semantic features and capture the semantic relationship between regions.

[0085] Specifically, the window attention sub-unit can capture the spatial relationship within the region and the spatial relationship between regions based on window attention and a window shifting strategy. The semantic clustering attention sub-unit can extract the semantic center features of multiple regions; determine semantic attention based on the semantic center features of multiple regions; and aggregate regions with similar semantic features based on semantic attention.

[0086] Exemplarily, Figure 7 FIG. is a schematic structural diagram of a spatial and semantic attention unit provided according to an embodiment of the present disclosure. As Figure 7 shown, the spatial and semantic attention unit includes a window attention sub-unit and a semantic clustering attention sub-unit. The window attention sub-unit can capture the spatial relationship within the region and the spatial relationship between regions based on the window attention and shifted window strategy of Swin Transformer. The semantic clustering attention sub-unit can extract the semantic center features of multiple regions through the Soft K-means algorithm. The calculation process of the semantic center features is as follows:

[0087] ;

[0088] ;

[0089] Wherein, represents the feature vector of the i-th region in the region feature set, and the region feature set is , m represents the number of regions. represents the s-th semantic center in the initial semantic center set, and each semantic center represents a semantic category or a clustering center. The initial semantic center set is , n represents the number of semantic centers. is a smooth differentiable assignment function. β represents a parameter for adjusting the smoothness and is used to control the attenuation speed. . represents the feature of the s-th updated semantic center.

[0090] Furthermore, based on calculate the semantic attention, is a set or matrix composed of , , represents the semantic center features of multiple regions. The formula for calculating the semantic attention is as follows:

[0091] ;

[0092] where represents the semantic attention, Q represents the query vector, K represents the key vector, V represents the value vector, represents a learnable weight matrix, and d represents the feature dimension. Furthermore, according to the semantic attention, regions with similar semantic features are dynamically aggregated to improve the ability to segment blood vessels.

[0093] In the technical solution of the embodiments of the present disclosure, the window shape is adjusted to align with the blood vessel morphology through the deformation partition unit, so that the window adapts to the blood vessel morphology, thereby improving the efficiency of the attention mechanism. The spatial relationship within the region and the spatial relationship between regions are captured through the window attention sub-unit, and regions with similar semantic features are aggregated through the semantic clustering attention sub-unit to capture the semantic relationship between regions. Compared with the fixed window scheme, the situation of blood vessel breakage or mis-segmentation can be avoided, especially for slender blood vessels and curved blood vessels, so that the topological structure of slender blood vessels and curved blood vessels can be accurately captured, and the blood vessel segmentation accuracy can be improved.

[0094] Figure 8 is a flowchart of another blood vessel segmentation method provided by the embodiments of the present disclosure. The method of this embodiment can be combined with each optional solution in the blood vessel segmentation method provided in the above embodiments. On the basis of the above embodiments, the blood vessel segmentation model is further refined in this embodiment.

[0095] As Figure 8 shown, the method includes:

[0096] S410. Obtain the image to be segmented.

[0097] S420. Input the image to be segmented into the deformation partitioning module to obtain a partitioned image aligned with the blood vessel morphology, where the deformation partitioning module is used to divide the image to be segmented into multiple regions aligned with the blood vessel morphology.

[0098] S430. Input the partitioned image aligned with the blood vessel morphology into the first spatial and semantic converter module to obtain first spatial and semantic features; input the first spatial and semantic features into the first downsampling module to obtain first downsampled features.

[0099] S440. Input the first downsampled features into the second spatial and semantic converter module to obtain second spatial and semantic features; input the second spatial and semantic features into the second downsampling module to obtain second downsampled features.

[0100] S450. Input the second downsampled features into the third spatial and semantic converter module to obtain third spatial and semantic features; input the third spatial and semantic features into the third downsampling module to obtain third downsampled features.

[0101] S460. Input the third downsampled features into the fourth spatial and semantic converter module to obtain fourth spatial and semantic features; input the fourth spatial and semantic features into the fifth spatial and semantic converter module to obtain fifth spatial and semantic features; input the fifth spatial and semantic features and the third downsampled features into the first upsampling module to obtain first upsampled features.

[0102] S470. Input the first upsampled features into the sixth spatial and semantic converter module to obtain sixth spatial and semantic features; input the sixth spatial and semantic features and the second downsampled features into the second upsampling module to obtain second upsampled features.

[0103] S480. Input the second upsampled features into the seventh spatial and semantic converter module to obtain seventh spatial and semantic features; input the seventh spatial and semantic features and the first downsampled features into the third upsampling module to obtain third upsampled features.

[0104] S490. Input the third upsampled features into the eighth spatial and semantic converter module to obtain eighth spatial and semantic features; input the eighth spatial and semantic features into the image expansion module to obtain the blood vessel segmentation result.

[0105] In the embodiments of the present disclosure, the vascular segmentation model may include a first spatio-semantic transformer module, a first downsampling module, a second spatio-semantic transformer module, a second downsampling module, a third spatio-semantic transformer module, a third downsampling module, a fourth spatio-semantic transformer module, a fifth spatio-semantic transformer module, a first upsampling module, a sixth spatio-semantic transformer module, a second upsampling module, a seventh spatio-semantic transformer module, a third upsampling module, an eighth spatio-semantic transformer module, and an image expansion module; the first downsampling module is residually connected to the third upsampling module, the second downsampling module is residually connected to the second upsampling module, and the third downsampling module is residually connected to the first upsampling module.

[0106] Among them, the vascular segmentation model may be a three-dimensional UNet-like structure, that is, the vascular segmentation model may adopt an encoder-decoder structure, combined with residual connections, to achieve multi-scale feature extraction and fusion, so as to ensure that the vascular segmentation model captures both the global semantic information and local spatial details of the vascular structure, thereby improving the vascular segmentation accuracy. The downsampling module can reduce the spatial dimension of the features, capture higher-level abstract features, and understand the global context of the vascular structure. The downsampling module can restore the features to the original resolution, combined with the residual connections in the encoder path, to fuse high-level semantic information and low-level spatial details. Residual connections are used to ensure the retention and transmission of fine-grained spatial information of blood vessels.

[0107] Exemplarily, Figure 9 FIG. is a schematic structural diagram of a deformation partitioning module and a vascular segmentation model provided according to an embodiment of the present disclosure. As Figure 9 shown, the input is the image to be segmented, and the output is the segmentation mask, which is used to identify the vascular structure in the image to be segmented.

[0108] Figure 10 FIG. is a flowchart of a vascular segmentation method provided according to an embodiment of the present disclosure. The method of this embodiment is a preferred example of the above embodiment. The method includes:

[0109] 1. Use the nnU-Net framework to perform normalization processing on the 3D CTA image. The normalization processing includes:

[0110] Normalization: Adjust the pixel intensity of the image to be segmented, scale the pixel intensity value to [0, 1] or normalize it to a mean of 0 and a variance of 1 to reduce the differences between scanning devices.

[0111] Resampling: Adjust the spatial resolution of the image to be segmented so that all images have a consistent voxel spacing.

[0112] Cropping: Crop the image to be segmented into a unified size of 128×128×128 voxels to ensure input consistency and reduce computational complexity.

[0113] 2. Input the 3D CTA image after normalization processing into the deformation partitioning module to obtain a partitioned image aligned with the blood vessel morphology, and then input it into a blood vessel segmentation model based on an encoder-decoder structure to obtain a 3D segmentation mask of the aortic blood vessels. The size of the 3D segmentation mask of the aortic blood vessels is 128×128×128 voxels. Among them, the blood vessel segmentation model includes a first spatial and semantic converter module, a first downsampling module, a second spatial and semantic converter module, a second downsampling module, a third spatial and semantic converter module, a third downsampling module, a fourth spatial and semantic converter module, a fifth spatial and semantic converter module, a first upsampling module, a sixth spatial and semantic converter module, a second upsampling module, a seventh spatial and semantic converter module, a third upsampling module, an eighth spatial and semantic converter module, and an image expansion module; the first downsampling module is residually connected to the third upsampling module, the second downsampling module is residually connected to the second upsampling module, and the third downsampling module is residually connected to the first upsampling module.

[0114] The technical solution of the embodiment of the present disclosure realizes adaptive image partitioning aligned with the blood vessel morphology through the deformation partitioning module, captures the spatial relationship and semantic relationship of each partition in the 3D CTA image, and compared with the fixed image partitioning scheme, it can avoid the occurrence of blood vessel breakage or missegmentation, especially for slender blood vessels and curved blood vessels, so as to accurately capture the topological structure of slender blood vessels and curved blood vessels and improve the blood vessel segmentation accuracy. Through the spatial and semantic converter module, the semantic relationship and spatial relationship of regions in the 3D CTA image can be effectively extracted, thereby improving the blood vessel segmentation accuracy.

[0115] Figure 11 It is a comparison chart of the experimental results of blood vessel segmentation provided by the embodiment of the present disclosure. Figure 11 The result in the last row is the experimental result of aortic segmentation of the blood vessel segmentation method proposed in this embodiment on the SegA dataset. The corresponding blood vessel segmentation models in other rows are: MedNeXt, 3DUXNet, UNETR, SwinUNETR, TransBTS, MPTUNETR, nnFOrmer models. As Figure 11 shown, the blood vessel segmentation method proposed in this embodiment has obtained the highest Dice score (0.856±0.106), mIoU (0.762±0.150), and sensitivity (0.880±0.088).

[0116] Figure 12 It is another comparison chart of the experimental results of blood vessel segmentation provided by the embodiment of the present disclosure. Figure 12 In it, MP represents the deformation partitioning module, and SCA represents the spatial and semantic converter module. Specifically, Figure 12The comparison graph of the experimental results of blood vessel segmentation is the ablation experiment comparison graph, that is, the experimental results of three cases are compared: without using MP and SCA, using MP and not using SCA, and using MP and using SCA. As can be seen from Figure 12 it, when the deformation partition module and the spatial and semantic converter module are used simultaneously, the blood vessel segmentation result is the best.

[0117] Figure 13 The following is a schematic structural diagram of a blood vessel segmentation device provided by an embodiment of the present disclosure. As shown in Figure 13 it, the device includes:

[0118] An image to be segmented acquisition module 510, configured to acquire an image to be segmented;

[0119] An image deformation partition module 520, configured to input the image to be segmented into the deformation partition module to obtain a partitioned image aligned with the blood vessel morphology, wherein the deformation partition module is configured to divide the image to be segmented into multiple regions aligned with the blood vessel morphology;

[0120] An image blood vessel segmentation module 530, configured to input the partitioned image aligned with the blood vessel morphology into a blood vessel segmentation model to obtain a blood vessel segmentation result, wherein the blood vessel segmentation model includes at least one spatial and semantic converter module, and the spatial and semantic converter module is configured to capture the spatial relationship and semantic relationship of the regions.

[0121] The technical solution of the embodiment of the present disclosure is to acquire an image to be segmented, and then input the image to be segmented into the deformation partition module to obtain a partitioned image aligned with the blood vessel morphology. Among them, the deformation partition module is configured to divide the image to be segmented into multiple regions aligned with the blood vessel morphology, and then input the partitioned image aligned with the blood vessel morphology into the blood vessel segmentation model to obtain a blood vessel segmentation result. Among them, the blood vessel segmentation model includes at least one spatial and semantic converter module, and the spatial and semantic converter module is configured to capture the spatial relationship and semantic relationship of the regions. In the above technical solution, adaptive image partitioning aligned with the blood vessel morphology is realized, and the spatial relationship and semantic relationship of each partition are captured. Compared with the fixed image partitioning scheme, the situation of blood vessel breakage or mis-segmentation can be avoided, especially for slender blood vessels and curved blood vessels, so that the topological structure of slender blood vessels and curved blood vessels can be accurately captured, and the blood vessel segmentation accuracy can be improved. In addition, the spatial relationship and semantic relationship of the regions are captured through the spatial and semantic converter module, thereby improving the understanding ability of the blood vessel segmentation model for the blood vessel structure, and further improving the segmentation accuracy of the blood vessel segmentation model.

[0122] Based on any optional technical solution in the embodiment of the present disclosure, optionally, the deformation partition module includes a velocity field prediction unit and a scaling and squaring unit;

[0123] Correspondingly, the image deformation partition module 520 is specifically configured to:

[0124] Input the image to be segmented into the velocity field prediction unit to obtain the velocity field corresponding to the image to be segmented;

[0125] Input the velocity field corresponding to the image to be segmented into the scaling and squaring unit to obtain the deformation field corresponding to the velocity field;

[0126] Interpolate the image to be segmented based on the deformation field corresponding to the velocity field to obtain a partitioned image aligned with the blood vessel morphology.

[0127] Based on any optional technical solution in the embodiments of the present disclosure, optionally, the space and semantic converter module includes a deformation partition unit, at least one space and semantic attention unit, and a window offset unit;

[0128] The deformation partition unit is used to adjust the window shape to align with the blood vessel morphology;

[0129] The space and semantic attention unit is used to capture the spatial relationship of regions and the semantic relationship of regions;

[0130] The window offset unit is used to offset the position of the window.

[0131] Based on any optional technical solution in the embodiments of the present disclosure, optionally, the space and semantic attention unit includes a window attention subunit and a semantic clustering attention subunit;

[0132] The window attention subunit is used to capture the spatial relationship within a region and the spatial relationship between regions;

[0133] The semantic clustering attention subunit is used to aggregate regions with similar semantic features and capture the semantic relationship between regions.

[0134] Based on any optional technical solution in the embodiments of the present disclosure, optionally, the semantic clustering attention subunit is specifically configured to:

[0135] Extract the semantic center features of multiple regions;

[0136] Determine semantic attention based on the semantic center features of the multiple regions;

[0137] Aggregate regions with similar semantic features based on the semantic attention.

[0138] Based on any optional technical solution in the embodiments of the present disclosure, optionally, the blood vessel segmentation model includes a first spatio-semantic converter module, a first downsampling module, a second spatio-semantic converter module, a second downsampling module, a third spatio-semantic converter module, a third downsampling module, a fourth spatio-semantic converter module, a fifth spatio-semantic converter module, a first upsampling module, a sixth spatio-semantic converter module, a second upsampling module, a seventh spatio-semantic converter module, a third upsampling module, an eighth spatio-semantic converter module, and an image expansion module; the first downsampling module is residually connected to the third upsampling module, the second downsampling module is residually connected to the second upsampling module, and the third downsampling module is residually connected to the first upsampling module;

[0139] Correspondingly, the image blood vessel segmentation module 530 is specifically configured to:

[0140] Input the partition image aligned with the blood vessel morphology into the first spatio-semantic converter module to obtain first spatio-semantic features;

[0141] Input the first spatio-semantic features into the first downsampling module to obtain first downsampled features;

[0142] Input the first downsampled features into the second spatio-semantic converter module to obtain second spatio-semantic features;

[0143] Input the second spatio-semantic features into the second downsampling module to obtain second downsampled features;

[0144] Input the second downsampled features into the third spatio-semantic converter module to obtain third spatio-semantic features;

[0145] Input the third spatio-semantic features into the third downsampling module to obtain third downsampled features;

[0146] Input the third downsampled features into the fourth spatio-semantic converter module to obtain fourth spatio-semantic features;

[0147] Input the fourth spatio-semantic features into the fifth spatio-semantic converter module to obtain fifth spatio-semantic features;

[0148] Input the fifth spatio-semantic features and the third downsampled features into the first upsampling module to obtain first upsampled features;

[0149] Input the first upsampled features into the sixth spatio-semantic converter module to obtain sixth spatio-semantic features;

[0150] Input the sixth space, semantic features, and the second downsampled features into the second upsampling module to obtain second upsampled features;

[0151] Input the second upsampled features into the seventh space and semantic converter module to obtain seventh space and semantic features;

[0152] Input the seventh space and semantic features and the first downsampled features into the third upsampling module to obtain third upsampled features;

[0153] Input the third upsampled features into the eighth space and semantic converter module to obtain eighth space and semantic features;

[0154] Input the eighth space and semantic features into the image expansion module to obtain the blood vessel segmentation result.

[0155] The blood vessel segmentation device provided by the embodiments of the present disclosure can execute the blood vessel segmentation method provided by any embodiment of the present disclosure, and has the corresponding functional modules and beneficial effects for executing the method.

[0156] Figure 14 FIG. shows a schematic structural diagram of an electronic device 10 that can be used to implement the embodiments of the present disclosure. The electronic device is intended to represent various forms of digital computers, such as, laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as, personal digital assistants, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0157] As Figure 14 shown, the electronic device 10 includes at least one processor 11, and a memory communicatively connected to at least one processor 11, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc. Among them, the memory stores a computer program executable by at least one processor. The processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 into the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other through a bus 14. The I / O interface 15 is also connected to the bus 14.

[0158] Multiple components in the electronic device 10 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a disk, an optical disc, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0159] The processor 11 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The processor 11 executes the various methods and processes described above, such as the blood vessel segmentation method, which includes:

[0160] Obtain the image to be segmented;

[0161] Input the image to be segmented into the deformation partitioning module to obtain a partitioned image aligned with the blood vessel morphology, where the deformation partitioning module is used to divide the image to be segmented into multiple regions aligned with the blood vessel morphology;

[0162] Input the partitioned image aligned with the blood vessel morphology into the blood vessel segmentation model to obtain a blood vessel segmentation result, where the blood vessel segmentation model includes at least one spatial and semantic converter module, and the spatial and semantic converter module is used to capture the spatial relationship and semantic relationship of the regions.

[0163] In some embodiments, the blood vessel segmentation method can be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as the storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed onto the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the blood vessel segmentation method described above can be executed. Alternatively, in other embodiments, the processor 11 can be configured to execute the blood vessel segmentation method in any other suitable manner (e.g., by means of firmware).

[0164] The various embodiments of the systems and techniques described above in this specification can be implemented in digital electronic circuitry, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), system on a chip (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include: being implemented in one or more computer programs that are executable and / or interpretable on a programmable system including at least one programmable processor, which can be a special-purpose or a general-purpose programmable processor that receives data and instructions from, and transmits data and instructions to, a storage system, at least one input device, and at least one output device.

[0165] The computer programs for implementing the methods of the present disclosure can be written in any combination of one or more programming languages. These computer programs can be provided to a processor of a general purpose computer, a special purpose computer, or other programmable data processing apparatus, such that the computer programs, when executed by the processor, cause the functions / operations specified in the flowchart and / or block diagram to be implemented. The computer programs can be executed entirely on the machine, partly on the machine, as a stand-alone software package partly on the machine and partly on a remote machine or entirely on the remote machine or server.

[0166] In the context of the present disclosure, a computer-readable storage medium can be a tangible medium that can contain or store a computer program for use by or in connection with an instruction execution system, apparatus, or device. The computer-readable storage medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. Alternatively, the computer-readable storage medium can be a machine-readable signal medium. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0167] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or a trackball) through which the user can provide input to the electronic device. Other kinds of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and the input received from the user can be in any form (including acoustic input, voice input, or tactile input).

[0168] The systems and techniques described herein can be implemented in a computing system including backend components (e.g., as a data server), or a computing system including middleware components (e.g., an application server), or a computing system including frontend components (e.g., a user computer having a graphical user interface or a web browser through which the user can interact with an implementation of the systems and techniques described herein), or a computing system including any combination of such backend components, middleware components, or frontend components. The components of the system can be interconnected to each other by digital data communication in any form or medium (e.g., a communication network). Examples of communication networks include: local area network (LAN), wide area network (WAN), blockchain network, and the Internet.

[0169] A computing system can include a client and a server. The client and the server are generally far from each other and usually interact through a communication network. The relationship between the client and the server is created by computer programs running on respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or a cloud host, which is a host product in the cloud computing service system and solves the defects of difficult management and weak business scalability existing in traditional physical hosts and VPS services.

[0170] It should be understood that various forms of the processes shown above can be used, with steps reordered, added, or deleted. For example, the steps recited in this disclosure can be executed in parallel, sequentially, or in a different order, as long as the desired results of the technical solution of this disclosure can be achieved, and this is not limited herein.

[0171] The embodiments of the present disclosure also provide a computer program product, including a computer program which, when executed by a processor, implements the blood vessel segmentation method provided in any embodiment of the present disclosure.

[0172] In the process of implementing the computer program product, computer program code for performing the operations of the present disclosure can be written in one or more programming languages or combinations thereof. The programming languages include object-oriented programming languages such as Java, Smalltalk, C++, and also include conventional procedural programming languages such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, executed as an independent software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any type of network - including a local area network (LAN) or a wide area network (WAN) - or, alternatively, can be connected to an external computer (e.g., by connecting through the Internet using an Internet service provider).

[0173] In particular, according to an embodiment of the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present invention includes a computer program product that includes a computer program carried on a non-transitory computer-readable medium, and the computer program includes program code for performing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication unit 19, or installed from the storage unit 18, or installed from the ROM 12. When the computer program is executed by the processor 11, the above-mentioned functions defined in the method of the embodiment of the present invention are performed.

[0174] The above specific embodiments do not constitute a limitation on the protection scope of the present disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present disclosure should be included within the protection scope of the present disclosure.

Claims

1. A blood vessel segmentation method, characterized in that: include: Obtain the image to be segmented; Inputting the image to be segmented into a deformation partitioning module to obtain a partitioned image aligned with the blood vessel morphology, wherein the deformation partitioning module is used to divide the image to be segmented into a plurality of regions aligned with the blood vessel morphology; The partitioned image aligned with the vascular morphology is input into a vascular segmentation model to obtain a vascular segmentation result, wherein the vascular segmentation model includes at least one spatial and semantic converter module, and the spatial and semantic converter module is used to capture the spatial relationship of the region and the semantic relationship of the region.

2. The method according to claim 1, characterized in that The deformation partition module includes a velocity field prediction unit and a scaling and squaring unit; Accordingly, the step of inputting the image to be segmented into a deformation partitioning module to obtain a partitioned image aligned with the vascular morphology includes: Inputting the image to be segmented into the velocity field prediction unit to obtain a velocity field corresponding to the image to be segmented; Inputting the velocity field corresponding to the image to be segmented into the scaling and squaring unit to obtain a deformation field corresponding to the velocity field; The image to be segmented is interpolated based on the deformation field corresponding to the velocity field to obtain a partitioned image aligned with the blood vessel morphology.

3. The method according to claim 1, characterized in that The spatial and semantic converter module includes a deformation partition unit, at least one spatial and semantic attention unit and a window offset unit; The deformation partition unit is used to adjust the window shape to align with the blood vessel morphology; The spatial and semantic attention unit is used to capture the spatial relationship of regions and the semantic relationship of regions; The window offset unit is used to offset the position of the window.

4. The method according to claim 3, characterized in that The spatial and semantic attention unit includes a window attention subunit and a semantic clustering attention subunit; The window attention subunit is used to capture the spatial relationship within the region and the spatial relationship between regions; The semantic clustering attention subunit is used to aggregate regions with similar semantic features and capture the semantic relationship between regions.

5. The method according to claim 4, characterized in that The semantic clustering attention subunit is specifically used for: Extract semantic center features of multiple regions; determining semantic attention based on semantic center features of the plurality of regions; Regions with similar semantic features are aggregated based on the semantic attention.

6. The method according to any one of claims 1 to 5, characterized in that: The blood vessel segmentation model includes a first space and semantic converter module, a first down-sampling module, a second space and semantic converter module, a second down-sampling module, a third space and semantic converter module, a third down-sampling module, a fourth space and semantic converter module, a fifth space and semantic converter module, a first up-sampling module, a sixth space and semantic converter module, a second up-sampling module, a seventh space and semantic converter module, a third up-sampling module, an eighth space and semantic converter module and an image expansion module; the first down-sampling module is residually connected to the third up-sampling module, the second down-sampling module is residually connected to the second up-sampling module, and the third down-sampling module is residually connected to the first up-sampling module; Accordingly, the step of inputting the partitioned image aligned with the blood vessel morphology into the blood vessel segmentation model to obtain the blood vessel segmentation result includes: Inputting the partitioned image aligned with the vascular morphology into the first spatial and semantic converter module to obtain first spatial and semantic features; Inputting the first spatial and semantic features into the first downsampling module to obtain first downsampling features; Inputting the first down-sampled features into the second spatial and semantic converter module to obtain second spatial and semantic features; Inputting the second spatial and semantic features into the second downsampling module to obtain second downsampling features; Inputting the second down-sampled features into the third space and semantic converter module to obtain third space and semantic features; Inputting the third spatial and semantic features into the third downsampling module to obtain third downsampling features; Inputting the third down-sampled features into the fourth spatial and semantic converter module to obtain fourth spatial and semantic features; Inputting the fourth space and semantic features into the fifth space and semantic converter module to obtain the fifth space and semantic features; Inputting the fifth spatial and semantic features and the third down-sampled features into the first up-sampling module to obtain a first up-sampled feature; Inputting the first up-sampled features into the sixth spatial and semantic converter module to obtain sixth spatial and semantic features; Inputting the sixth spatial and semantic features and the second down-sampled features into the second up-sampling module to obtain second up-sampled features; Inputting the second up-sampled features into the seventh spatial and semantic converter module to obtain seventh spatial and semantic features; Inputting the seventh spatial and semantic features and the first down-sampled features into the third up-sampling module to obtain a third up-sampled feature; Inputting the third up-sampled feature into the eighth spatial and semantic converter module to obtain an eighth spatial and semantic feature; The eighth space and semantic features are input into the image expansion module to obtain a blood vessel segmentation result.

7. A blood vessel segmentation device, characterized in that: include: The image to be segmented acquisition module is used to acquire the image to be segmented; An image deformation partitioning module, used for inputting the image to be segmented into the deformation partitioning module to obtain a partitioned image aligned with the blood vessel morphology, wherein the deformation partitioning module is used for dividing the image to be segmented into a plurality of regions aligned with the blood vessel morphology; The image vascular segmentation module is used to input the partitioned image aligned with the vascular morphology into a vascular segmentation model to obtain a vascular segmentation result, wherein the vascular segmentation model includes at least one spatial and semantic converter module, and the spatial and semantic converter module is used to capture the spatial relationship of the region and the semantic relationship of the region.

8. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively coupled to the at least one processor; The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the blood vessel segmentation method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the blood vessel segmentation method according to any one of claims 1 to 6 when executed.

10. A computer program product, characterized in that The computer program product comprises a computer program, which, when executed by a processor, implements the blood vessel segmentation method according to any one of claims 1 to 6.

Images