Three-dimensional cerebrovascular image segmentation method and device, electronic equipment and storage medium
Through downsampling segmentation model and window sampling technology, combined with local partial separating model, the problem of low accuracy of three-dimensional cerebrovascular image segmentation is solved, the integration of global semantics and local semantics is achieved, and the accuracy of image segmentation is improved.
Patent Information
- Application Number
- CN202410793408.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-06-19
- Publication Date
- 2024-11-01
- Estimated Expiration
- 2044-06-19
AI Technical Summary
In the prior art, the segmentation accuracy of three-dimensional cerebrovascular images is not high because only local area information is used, the global semantics of the entire volume is ignored, and it is difficult to process the entire intracranial artery volume under memory limitations.
Global segmentation is performed through downsampling and segmentation model, and the cropping window collection is obtained in combination with window sampling, and the three-dimensional cerebrovascular images are cropped to obtain image subblocks containing global semantics, and local semantics are used for local semantics to realize global semantics and Fusion of local semantics.
It improves the segmentation accuracy of three-dimensional cerebrovascular images, can better capture the global morphology and topological structure of intracranial arteries, and improves the accuracy of image segmentation.
Smart Images

Figure CN118887397A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of medical image segmentation, and in particular to a three-dimensional cerebral vascular image segmentation method, device, electronic equipment and storage medium. Background Art
[0002] During intracranial artery interventional surgery, three-dimensional digital silhouette angiography can achieve accurate segmentation and morphological analysis of the intracranial arteries and provide accurate anatomical information of the intracranial artery region, which is of great significance for preoperative planning and intraoperative navigation.
[0003] In the prior art, due to the very large size of the 3D cerebrovascular images of the intracranial arteries, it is difficult to put the entire volume of the intracranial arteries into the 3D segmentation model due to the actual memory limitations. Therefore, the conventional method is to apply the 3D segmentation model to the 3D sub-blocks cropped from the original volume. Although this method can capture the dependencies of the images in three dimensions, it only uses the information of the local area for segmentation, while ignoring the global semantics of the entire volume, resulting in low segmentation accuracy of the 3D cerebrovascular images.
[0004] Therefore, how to better achieve the segmentation of three-dimensional cerebral vascular images has become a technical problem that needs to be solved urgently in the industry.
[0005] Summary of the invention / Summary of the invention
[0006] The present invention provides a three-dimensional cerebral vascular image segmentation method, device, electronic equipment and storage medium, which are used to better achieve the segmentation of three-dimensional cerebral vascular images.
[0007] The present invention provides a three-dimensional cerebral vascular image segmentation method, comprising: Inputting the downsampled image of the three-dimensional cerebrovascular image to be segmented into the downsampled segmentation model to obtain a global segmented image of the downsampled image output by the downsampled segmentation model; the downsampled segmentation model is trained based on the three-dimensional cerebrovascular image samples and their corresponding segmented image labels; Performing window sampling on the global segmented image, determining a clipping window set of the global segmented image, and clipping the three-dimensional cerebral vascular image based on the clipping window set to obtain a plurality of image sub-blocks; Inputting the multiple image sub-blocks into a local segmentation model to obtain a local segmentation image of each image sub-block output by the local segmentation model; the local segmentation model is trained based on the three-dimensional cerebrovascular image samples and their corresponding segmentation image labels using a trained downsampling segmentation model; The local segmented images are spliced to obtain the segmentation result of the three-dimensional cerebral vascular image.
[0008] According to a three-dimensional cerebrovascular image segmentation method provided by the present invention, the downsampling segmentation model includes a feature extraction network; after the three-dimensional cerebrovascular image is cropped based on the cropping window set to obtain a plurality of image sub-blocks, the method further includes: Obtaining a target feature image output by the last layer of the feature extraction network; Inputting the target feature image and the global segmentation image into a spatial fusion module to obtain a global feature image of the three-dimensional cerebral vascular image output by the spatial fusion module; Splicing the global feature image with the multiple image sub-blocks respectively to obtain multiple spliced image sub-blocks; Each of the spliced image sub-blocks is locally segmented and spliced to obtain a segmentation result of the three-dimensional cerebral vascular image.
[0009] According to a three-dimensional cerebrovascular image segmentation method provided by the present invention, the spatial fusion module is specifically used to: crop the global segmented image based on the cropping window set to obtain a plurality of cropping image sub-blocks, and crop the target feature image based on the cropping window set to obtain a plurality of feature image sub-blocks; Determine a cerebral vascular tissue topology map based on the multiple cropped image sub-blocks and the multiple feature image sub-blocks; Feature extraction is performed on the cerebral vascular tissue topology map to obtain the global feature image.
[0010] According to a three-dimensional cerebrovascular image segmentation method provided by the present invention, the cerebrovascular tissue topology map is determined based on the multiple cropping image sub-blocks and the multiple feature image sub-blocks, including: For any cropped image sub-block among the plurality of cropped image sub-blocks, determining a first number of cerebrovascular voxels in the any cropped image sub-block, a second number of cerebrovascular voxels in any other cropped image sub-block among the plurality of cropped image sub-blocks, and a third number of identical vascular voxels in the any cropped image sub-block and any other cropped image sub-block; Determine the adjacency between each of the cropped image sub-blocks based on the first number, the second number, and the third number, and determine an adjacency matrix according to each of the adjacency; Performing global average pooling processing on each of the feature map sub-blocks to obtain a feature vector corresponding to each of the feature map sub-blocks; establishing the cerebral vascular tissue topology map with the feature vectors as nodes and the adjacency matrix as edges.
[0011] According to a three-dimensional cerebrovascular image segmentation method provided by the present invention, the feature extraction of the cerebrovascular tissue topology map to obtain the global feature image includes: Inputting the cerebrovascular tissue topology map into a graph neural network to obtain a node vector group output by the graph neural network; the graph neural network is used to encode and fuse information of adjacent topological nodes in the cerebrovascular tissue topology map; Each node vector in the node vector group is used as a weight vector of the corresponding cropping window in the cropping window set, and each weight vector is weightedly summed with the feature map sub-block of the corresponding cropping window to obtain a global feature of the blood vessel; the global feature of the blood vessel is upsampled to obtain the global feature image.
[0012] According to a three-dimensional cerebral vascular image segmentation method provided by the present invention, the window sampling of the global segmented image and determining the clipping window set of the global segmented image include: Converting the global segmented image into point cloud data; Performing farthest point sampling on the point cloud data to obtain sparse point cloud data corresponding to the global segmented image; A cropping window set of the global segmented image is generated based on a preset window size and with each point in the sparse point cloud data as the center.
[0013] According to a three-dimensional cerebrovascular image segmentation method provided by the present invention, before inputting the downsampled image of the three-dimensional cerebrovascular image to be segmented into the downsampled segmentation model to obtain the global segmentation image of the downsampled image output by the downsampled segmentation model, the method further includes: Using the acquired three-dimensional cerebral vascular image samples and their corresponding segmented image labels as a group of training samples, and acquiring multiple groups of the training samples; Using the multiple groups of training samples to train the downsampling segmentation model to obtain a trained downsampling segmentation model; Inputting the downsampled image of the three-dimensional cerebrovascular image sample into the trained downsampled segmentation model to obtain a global segmentation image of the three-dimensional cerebrovascular image sample, and determining a cropping window set corresponding to the three-dimensional cerebrovascular image sample based on the global segmentation image of the three-dimensional cerebrovascular image sample; Based on the cropping window set corresponding to the three-dimensional cerebrovascular image sample, the three-dimensional cerebrovascular image sample and its corresponding segmented image label are cropped respectively to obtain a plurality of sub-block segmentation masks and a label sub-block corresponding to each sub-block segmentation mask; The local segmentation model is trained using the multiple sub-block segmentation masks and the label sub-block corresponding to each of the sub-block segmentation masks to obtain a trained local segmentation model.
[0014] The present invention also provides a three-dimensional cerebral vascular image segmentation device, comprising: A first segmentation module is used to input the downsampled image of the three-dimensional cerebrovascular image to be segmented into a downsampled segmentation model to obtain a global segmented image of the downsampled image output by the downsampled segmentation model; the downsampled segmentation model is trained based on the three-dimensional cerebrovascular image samples and their corresponding segmented image labels; A first cropping module is used to perform window sampling on the global segmented image, determine a cropping window set of the global segmented image, and perform cropping processing on the three-dimensional cerebrovascular image based on the cropping window set to obtain a plurality of image sub-blocks; a second segmentation module is used to input the plurality of image sub-blocks into a local segmentation model to obtain a local segmentation image of each image sub-block output by the local segmentation model; the local segmentation model is trained based on the three-dimensional cerebrovascular image samples and their corresponding segmentation image labels using a trained downsampling segmentation model; The first splicing module is used to splice the local segmented images to obtain the segmentation result of the three-dimensional cerebral vascular image.
[0015] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, any one of the three-dimensional cerebral vascular image segmentation methods described above is implemented.
[0016] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the three-dimensional cerebral vascular image segmentation method as described above is implemented.
[0017] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the computer program implements any one of the above-mentioned three-dimensional cerebral vascular image segmentation methods.
[0018] The three-dimensional cerebrovascular image segmentation method, device, electronic device and storage medium provided by the present invention obtain a downsampling segmentation model by training with three-dimensional cerebrovascular image samples and their corresponding segmentation image labels, and further train a local segmentation model based on the three-dimensional cerebrovascular image samples and their corresponding segmentation image labels using the trained downsampling segmentation model, so that the downsampling segmentation model performs global segmentation on the downsampled image of the three-dimensional cerebrovascular image, and obtains a cropping window set of adaptive cerebrovascular three-dimensional morphology through window sampling, and then obtains image sub-blocks containing global semantics based on cropping processing of the original three-dimensional cerebrovascular image, and then locally segments each image sub-block through the local segmentation model, so as to splice each local segmented image, and finally obtain the segmentation result of the three-dimensional cerebrovascular image, thereby realizing an image segmentation method that integrates global semantics and local semantics, and can effectively improve the segmentation accuracy of the three-dimensional cerebrovascular image. BRIEF DESCRIPTION OF THE DRAWINGS
[0019] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0020] Figure 1 It is a flow chart of the three-dimensional cerebral vascular image segmentation method provided by the present invention.
[0021] Figure 2 This is one of the principle flow diagrams of the three-dimensional cerebral vascular image segmentation method provided by the present invention.
[0022] Figure 3 This is the second schematic diagram of the principle flow of the three-dimensional cerebral vascular image segmentation method provided by the present invention.
[0023] Figure 4 It is a principle flow chart of the space fusion module in the three-dimensional cerebral vascular image segmentation method provided by the present invention.
[0024] Figure 5 It is a structural schematic diagram of the three-dimensional cerebral vascular image segmentation device provided by the present invention.
[0025] Figure 6 It is a schematic diagram of the physical structure of the electronic device provided by the present invention. DETAILED DESCRIPTION
[0026] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0027] In the description of the invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.
[0028] Combine the following Figure 1-Figure 6 The present invention describes a three-dimensional cerebral vascular image segmentation method, device, electronic device and storage medium.
[0029] Figure 1 FIG. 1 is a flow chart of a three-dimensional cerebral vascular image segmentation method provided by the present invention. Figure 1 As shown, it includes: step 110, step 120, step 130 and step 140.
[0030] Step 110, input the downsampled image of the three-dimensional cerebrovascular image to be segmented into a downsampled segmentation model to obtain a global segmentation image of the downsampled image output by the downsampled segmentation model; the downsampled segmentation model is trained based on three-dimensional cerebrovascular image samples and their corresponding segmented image labels.
[0031] Step 120 , window sampling is performed on the global segmented image, a cropping window set of the global segmented image is determined, and a cropping process is performed on the three-dimensional cerebral vascular image based on the cropping window set to obtain a plurality of image sub-blocks.
[0032] Step 130, input multiple image sub-blocks into a local segmentation model to obtain a local segmentation image of each image sub-block output by the local segmentation model; the local segmentation model is trained using a trained downsampling segmentation model based on three-dimensional cerebrovascular image samples and their corresponding segmentation image labels.
[0033] Step 140, stitching the local segmented images to obtain a segmentation result of a three-dimensional cerebral vascular image.
[0034] Specifically, the three-dimensional cerebrovascular image described in the embodiment of the present invention refers to a three-dimensional cerebrovascular image to be segmented, which can be a cerebrovascular image of any three-dimensional modality, such as a common CT scan image, a magnetic resonance imaging (MRI) scan image, etc.
[0035] The downsampled image described in the embodiment of the present invention refers to an image obtained by downsampling a three-dimensional cerebral vascular image.
[0036] The downsampling segmentation model described in the embodiment of the present invention is obtained by training a deep neural network model based on three-dimensional cerebrovascular image samples and their corresponding segmented image labels, and is used to perform global segmentation on the input downsampled image. The deep neural network model can specifically adopt a three-dimensional U-Net network model, a three-dimensional segmentation model SwinUNETR, or other neural network models that can perform three-dimensional segmentation, and the present invention does not specifically limit this.
[0037] The clipping window set described in the embodiment of the present invention refers to a window set that characterizes and covers the three-dimensional morphology of cerebral blood vessels, that is, the combination of each window in the set at the corresponding position can cover the three-dimensional morphology of the entire intracranial artery and cerebral blood vessels.
[0038] The local segmentation model described in the embodiment of the present invention is obtained by training a deep neural network model based on three-dimensional cerebrovascular image samples and their corresponding segmented image labels using a trained downsampling segmentation model, and is used to further locally segment the input three-dimensional image sub-blocks to obtain fine segmentation results. Among them, the deep neural network model can also specifically adopt a three-dimensional U-Net network model, a three-dimensional segmentation model SwinUNETR, or other neural network models that can perform three-dimensional segmentation, and the present invention does not specifically limit this.
[0039] In an embodiment of the present invention, the downsampling segmentation model can be trained in advance by collecting a large number of three-dimensional cerebrovascular image samples and their corresponding segmentation image labels to obtain a trained downsampling segmentation model. Then, the three-dimensional cerebrovascular image samples and their corresponding segmentation image labels are used again in combination with the trained downsampling segmentation model to train the local segmentation model, and finally a trained local segmentation model is obtained.
[0040] In an embodiment of the present invention, in step 110, the three-dimensional cerebral vascular image to be segmented is first downsampled to obtain a corresponding downsampled image, and the downsampled image is input into a trained downsampled segmentation model for global segmentation processing, so that a global segmentation image of the downsampled cerebral vascular image can be obtained, thereby realizing global segmentation of the downsampled image.
[0041] In the embodiment of the present invention, in step 120, a dynamic window sampling method can be used to perform window sampling on the global segmented image obtained in step 110 to obtain a cropping window set of the global segmented image, thereby obtaining a window set of adaptive vascular three-dimensional morphology. Furthermore, the cropping window set can be used to crop the original-size three-dimensional cerebral vascular image to obtain multiple image sub-blocks.
[0042] Based on the content of the above embodiment, as an optional embodiment, performing window sampling on the global segmented image and determining a clipping window set of the global segmented image includes: Convert the global segmented image into point cloud data; The farthest point sampling is performed on the point cloud data to obtain the sparse point cloud data corresponding to the global segmentation image; Taking each point in the sparse point cloud data as the center and according to the preset window size, a set of cropped windows for the global segmentation image is generated.
[0043] Specifically, the preset window size described in the embodiment of the present invention refers to a pre-set three-dimensional window size for three-dimensional image segmentation, which can be specifically set according to actual design requirements.
[0044] In an embodiment of the present invention, a dynamic window sampling method is adopted to perform window sampling on the global segmented image to determine a clipping window set of the global segmented image.
[0045] More specifically, the global segmentation image output by the downsampling segmentation model is projected into three-dimensional space and converted into corresponding point cloud data. Then, in the spatial point cloud, multiple rounds of farthest point sampling are performed on the point cloud data to collect point cloud data used to characterize the three-dimensional spatial morphology of the cerebral blood vessels, and the covered point cloud coordinates are removed in each round of sampling. This iteration is repeated until all point clouds are covered, and finally sparse point cloud data corresponding to the global segmentation image can be obtained. It can be understood that the sparse point cloud data can maintain the spatial morphology of the cerebral blood vessels.
[0046] Furthermore, in an embodiment of the present invention, a window centered on each point in the sparse point cloud data is generated according to a preset window size, thereby forming a cropping window set for the global segmentation image.
[0047] The method of the embodiment of the present invention converts the global segmented image into a point cloud, utilizes dynamic window sampling to continuously iteratively sample the farthest point cloud, obtains a sparse point cloud that can characterize the spatial morphology of cerebral blood vessels, and then generates a three-dimensional window set with each point in the sparse point cloud as the center, thereby obtaining a window set containing the global semantics of the image, which is beneficial to fully utilizing the global semantics of the entire intracranial artery and cerebral blood vessel volume and improving the accuracy of three-dimensional cerebral vascular image segmentation.
[0048] Furthermore, in an embodiment of the present invention, in step 130, the multiple image sub-blocks obtained in step 120 are input into a trained local segmentation model, and a local segmentation image of each image sub-block output by the local segmentation model can be obtained, thereby achieving fine segmentation of the image sub-blocks.
[0049] Furthermore, in the embodiment of the present invention, in step 140, the local segmented images are sequentially spliced according to the position of the aforementioned cropping window, and finally a segmentation result of a three-dimensional cerebral vascular image can be obtained.
[0050] Figure 2 FIG. 1 is one of the principle flow charts of the three-dimensional cerebral vascular image segmentation method provided by the present invention. Figure 2As shown, in an embodiment of the present invention, after the three-dimensional cerebrovascular image to be segmented is downsampled to obtain a downsampled image, it is input into the downsampled segmentation model to obtain a global segmented image of the downsampled image. Then, the global segmented image is window sampled using a dynamic window sampling method to generate a cropping window set of the global segmented image, and the original three-dimensional cerebrovascular image is cropped using the cropping window set to obtain multiple image sub-blocks. Furthermore, these multiple image sub-blocks are input into the local segmentation model for local segmentation to obtain a local segmented image of each image sub-block. Finally, each local segmented image is spliced to obtain a segmentation result of the three-dimensional cerebrovascular image.
[0051] The three-dimensional cerebrovascular image segmentation method of the embodiment of the present invention obtains a downsampling segmentation model by training with three-dimensional cerebrovascular image samples and their corresponding segmented image labels, and further trains a local segmentation model based on the three-dimensional cerebrovascular image samples and their corresponding segmented image labels using the trained downsampling segmentation model, so that the downsampling segmentation model performs global segmentation on the downsampled image of the three-dimensional cerebrovascular image, and obtains a cropping window set of the adaptive cerebrovascular three-dimensional morphology through window sampling, and then obtains an image sub-block containing global semantics based on cropping processing of the original three-dimensional cerebrovascular image, and then locally segments each image sub-block through the local segmentation model, so as to splice each local segmented image, and finally obtain the segmentation result of the three-dimensional cerebrovascular image, thereby realizing an image segmentation method that integrates global semantics and local semantics, and can effectively improve the segmentation accuracy of the three-dimensional cerebrovascular image.
[0052] Based on the content of the above embodiment, as an optional embodiment, the downsampling segmentation model includes a feature extraction network; after the three-dimensional cerebral vascular image is cropped based on the cropping window set to obtain multiple image sub-blocks, the method further includes: Get the target feature image output by the last layer of the feature extraction network; Input the target feature image and the global segmentation image into the spatial fusion module to obtain the global feature image of the three-dimensional cerebral vascular image output by the spatial fusion module; Splicing the global feature image with the multiple image sub-blocks respectively to obtain multiple spliced image sub-blocks; Each spliced image sub-block is locally segmented and spliced to obtain the segmentation result of the three-dimensional cerebral vascular image.
[0053] It should be noted that the image segmentation model includes a feature extraction part and a segmentation head part. The segmentation head part refers to the part of the model specifically used to perform image segmentation tasks. It is usually located at the end of the model, immediately after the feature extraction part, and is responsible for converting the feature map extracted by the previous network into pixel-level classification predictions.
[0054] Specifically, in the embodiment of the present invention, the downsampling segmentation model includes a feature extraction network, which belongs to the feature extraction part. It can be understood that the downsampling segmentation model also includes a segmentation head part.
[0055] The target feature image described in the embodiment of the present invention refers to the feature image output by the last layer of the feature extraction network in the model after the downsampled image of the three-dimensional cerebral vascular image is input into the downsampled segmentation model.
[0056] It can be understood that the global segmentation image is a segmentation image that is finally output by the downsampling segmentation model after the downsampling image of the three-dimensional cerebral vascular image is input into the downsampling segmentation model.
[0057] The spatial fusion module described in the embodiment of the present invention is used to extract the global features of the three-dimensional cerebral vascular image based on the input target feature image and the global segmentation image to obtain the global feature image.
[0058] Figure 3 FIG. 2 is a schematic diagram of the principle flow of the three-dimensional cerebral vascular image segmentation method provided by the present invention. Figure 3 As shown, in an embodiment of the present invention, the 3D cerebral vascular image to be segmented is downsampled to obtain a corresponding downsampled image, which is then input into a downsampled segmentation model to obtain a global segmented image of the downsampled image. The global segmented image is then subjected to dynamic window sampling to obtain a cropping window set of the global segmented image, and the original 3D cerebral vascular image is cropped using the cropping window set to obtain multiple image sub-blocks.
[0059] Furthermore, the target feature image output by the last layer of the feature extraction network in the downsampling segmentation model is obtained, and the target feature image and the global segmentation image are input into the spatial fusion module, and the target feature image and the global segmentation image are cropped using a cropping window set, and the global feature image of the three-dimensional cerebrovascular image is extracted by combining the fusion processing of the image features.
[0060] Furthermore, in an embodiment of the present invention, the global feature image extracted by the spatial fusion module is concatenated with the multiple image sub-blocks obtained by cropping the original three-dimensional cerebrovascular image to obtain multiple spliced image sub-blocks. Furthermore, each spliced image sub-block can be input into a local segmentation model for local segmentation, and the local segmentation results output by the local segmentation model are spliced according to the position of the cropping window set, and finally the segmentation result of the three-dimensional cerebrovascular image is obtained.
[0061] The method of the embodiment of the present invention introduces a spatial fusion module to perform feature segmentation and fusion on the feature image output by the feature extraction network in the downsampling segmentation model and the global segmentation image output by the model, thereby extracting the global features of the three-dimensional cerebrovascular image, which is beneficial to further improve the segmentation accuracy of the three-dimensional cerebrovascular image.
[0062] Figure 4 is a principle flow chart of the spatial fusion module in the three-dimensional cerebrovascular image segmentation method provided by the present invention, such as Figure 4 As shown, in an embodiment of the present invention, the spatial fusion module is specifically used for: The global segmented image is cropped based on the cropping window set to obtain a plurality of cropped image sub-blocks, and the target feature image is cropped based on the cropping window set to obtain a plurality of feature image sub-blocks; Determine a cerebral vascular tissue topology map based on the plurality of cropped image sub-blocks and the plurality of feature image sub-blocks; Feature extraction is performed on the cerebral vascular tissue topology map to obtain a global feature image.
[0063] Specifically, the cropped image sub-block described in the embodiment of the present invention refers to an image sub-block obtained by cropping the global segmented image using a cropping window set.
[0064] The feature image sub-block described in the embodiment of the present invention refers to an image sub-block obtained by cropping the target feature image using a cropping window set.
[0065] The cerebrovascular tissue topology map described in the embodiment of the present invention is a characteristic image used to characterize the topological regularity of cerebrovascular tissue.
[0066] In the embodiment of the present invention, the specific implementation of the spatial fusion module is as follows: After the target feature image and the global segmentation image are input into the spatial fusion module, the two types of input images are cropped based on the aforementioned cropping window set, and the global segmentation image is cropped to obtain multiple cropped image sub-blocks; at the same time, the target feature image is cropped to obtain multiple feature image sub-blocks.
[0067] Furthermore, in an embodiment of the present invention, a cerebral vascular tissue topology map is determined based on a plurality of cropped image sub-blocks and a plurality of feature image sub-blocks.
[0068] Based on the content of the above embodiment, as an optional embodiment, determining a cerebral vascular tissue topology map based on multiple cropping image sub-blocks and multiple feature image sub-blocks includes: For any cropped image subblock in the plurality of cropped image subblocks, determining a first number of cerebrovascular voxels in the cropped image subblock, a second number of cerebrovascular voxels in any other cropped image subblock in the plurality of cropped image subblocks, and a third number of vascular voxels that are the same between the cropped image subblock and any other cropped image subblock; Determine the adjacency between each cropped image sub-block based on the first number, the second number, and the third number, and determine an adjacency matrix based on each adjacency; Perform global average pooling (GAP) on each feature map sub-block to obtain the feature vector corresponding to each feature map sub-block; The cerebrovascular tissue topology map was established with eigenvectors as nodes and adjacency matrices as edges.
[0069] Specifically, the first number described in the embodiment of the present invention refers to the number of cerebrovascular voxels contained in any cropping image sub-block selected from a plurality of cropping image sub-blocks.
[0070] Any other cropped image sub-block described in the embodiment of the present invention refers to any cropped image sub-block among the multiple cropped image sub-blocks except the selected any cropped image sub-block.
[0071] The second number described in the embodiment of the present invention refers to the number of cerebrovascular voxels contained in any other cropped image sub-block.
[0072] The third number described in the embodiment of the present invention refers to the number of the same blood vessel voxels contained in any cropped image sub-block and any other cropped image sub-block.
[0073] Continue to refer to Figure 4 In an embodiment of the present invention, for any cropped image sub-block among the multiple cropped image sub-blocks, the number of cerebrovascular voxels in the cropped image sub-block, i.e., the first number, can be determined. At the same time, the number of cerebrovascular voxels in any other cropped image sub-block among the multiple cropped image sub-blocks except the above-selected any cropped image sub-block, i.e., the second number, can be determined; at the same time, the number of the same vascular voxels contained in the any cropped image sub-block and any other cropped image sub-block, i.e., the third number, can also be determined.
[0074] Further, in this embodiment, based on the first number, the second number, and the third number, a value of a preset multiple of the third number (i.e., the number of voxels contained together), such as twice, and the ratio of the sum of the first number plus the second number can be calculated to determine the adjacency (CA) between each cropped image sub-block, which is used as the adjacency between two window sub-blocks in the cerebrovascular tissue topology map. Furthermore, an adjacency matrix can be generated based on the adjacency of each cropped image sub-block relative to other cropped image sub-blocks.
[0075] At the same time, in this embodiment, by performing GAP processing on each feature map sub-block, a feature vector corresponding to each feature map sub-block can be obtained.
[0076] Furthermore, in an embodiment of the present invention, a cerebral vascular tissue topology map is established by taking the feature vector corresponding to each feature map sub-block as a node and the generated adjacency matrix as an edge.
[0077] The method of the embodiment of the present invention considers the number of cerebrovascular voxels in each cropped image sub-block, mines the adjacency between each cropped image sub-block, constructs an adjacency matrix, and performs global average pooling on each feature image sub-block to obtain the corresponding feature vector, so that a cerebrovascular tissue topology map can be established with the feature vector as the node and the adjacency matrix as the edge, which is conducive to the subsequent more accurate extraction of the global feature image of the three-dimensional cerebrovascular image, and further improves the segmentation accuracy of the three-dimensional cerebrovascular image.
[0078] Furthermore, in an embodiment of the present invention, feature extraction is performed on the cerebral vascular tissue topology map to obtain a global feature image.
[0079] Based on the content of the above embodiment, as an optional embodiment, feature extraction is performed on the cerebral vascular tissue topology map to obtain a global feature image, including: The cerebrovascular tissue topology map is input into the graph neural network to obtain a node vector group output by the graph neural network; the graph neural network is used to encode and fuse the information of adjacent topological nodes in the cerebrovascular tissue topology map; Each node vector in the node vector group is used as a weight vector of a corresponding cropping window in the cropping window set, and each weight vector is weighted summed with a feature map sub-block of the corresponding cropping window to obtain a global feature of the blood vessel; The global features of the blood vessels are upsampled to obtain a global feature image.
[0080] Specifically, the Graph Neural Network (GNN) described in the embodiment of the present invention is a type of neural network model for processing graph structure data, which uses a neural network to learn the graph structure data in the cerebrovascular tissue topology map, thereby extracting and exploring the features and patterns in the graph structure data. It is specifically used to encode and fuse the data information of adjacent topological nodes in the cerebrovascular tissue topology map, and extract node feature vectors.
[0081] Continue to refer to Figure 4 In an embodiment of the present invention, the cerebrovascular tissue topology map established above is input into the graph neural network for processing. After graph convolution processing, the information of adjacent topological nodes therein is encoded and fused to obtain a set of new node vectors, that is, a node vector group.
[0082] Furthermore, in an embodiment of the present invention, each node vector in the node vector group is used as a weight vector of a corresponding cropping window in the cropping window set, and the feature map sub-block obtained by cropping the target feature image is combined, and each weight vector is weighted and summed with the feature map sub-block of the corresponding cropping window to obtain an encoded global feature image of the blood vessel. Then, the global feature image of the blood vessel is upsampled to obtain a global feature image of a three-dimensional cerebral vascular image.
[0083] In an embodiment of the present invention, by utilizing graph neural network processing to establish a cerebrovascular tissue topology map, the features of the graph structure data in the topology map can be fully extracted and explored, and by fusing the features with the corresponding cropping window feature map sub-blocks, the accuracy of extracting the global features of the three-dimensional cerebrovascular image can be further improved.
[0084] The method of the embodiment of the present invention, by introducing a spatial fusion module, crops and blocks the global segmented image and the target feature image based on a cropping window set, uses local block information to establish a cerebrovascular tissue topology map, and performs feature extraction on it. It can accurately and effectively extract the global features of the three-dimensional cerebrovascular image, which is beneficial to improving the segmentation accuracy of the subsequent three-dimensional cerebrovascular image.
[0085] Based on the content of the above embodiment, as an optional embodiment, before inputting the downsampled image of the three-dimensional cerebral vascular image to be segmented into the downsampled segmentation model to obtain the global segmented image of the downsampled image output by the downsampled segmentation model, the method further includes: The acquired three-dimensional cerebral vascular image samples and their corresponding segmented image labels are used as a group of training samples to obtain multiple groups of training samples; Using multiple groups of training samples to train the downsampling segmentation model, to obtain a trained downsampling segmentation model; Inputting the downsampled image of the three-dimensional cerebrovascular image sample into the trained downsampled segmentation model to obtain a global segmentation image of the three-dimensional cerebrovascular image sample, and determining a cropping window set corresponding to the three-dimensional cerebrovascular image sample based on the global segmentation image of the three-dimensional cerebrovascular image sample; Based on the cropping window set corresponding to the three-dimensional cerebrovascular image samples, the three-dimensional cerebrovascular image samples and their corresponding segmented image labels are cropped respectively to obtain multiple sub-block segmentation masks and label sub-blocks corresponding to each sub-block segmentation mask; the local segmentation model is trained using the multiple sub-block segmentation masks and the label sub-blocks corresponding to each sub-block segmentation mask to obtain a trained local segmentation model.
[0086] Specifically, in an embodiment of the present invention, before the downsampled image of the three-dimensional cerebral vascular image to be segmented is input into the downsampled segmentation model to obtain the global segmentation image of the downsampled image output by the downsampled segmentation model, the downsampled segmentation model and the local segmentation model need to be trained.
[0087] In an embodiment of the present invention, a large number of 3D cerebrovascular image sample sets are collected, and each 3D cerebrovascular image sample and its corresponding segmented image label are used as a set of training samples, thereby obtaining multiple sets of training samples. By using multiple sets of training samples to train the downsampling segmentation model, a trained downsampling segmentation model is obtained.
[0088] Specifically, in an embodiment of the present invention, for any training sample, the three-dimensional cerebrovascular image sample and its corresponding segmented image label are downsampled to obtain the downsampled image and the corresponding downsampled label corresponding to the three-dimensional cerebrovascular image sample. Then, the downsampled image is input into the downsampled segmentation model of the above-mentioned dynamic window inference method to generate a downsampled segmentation mask.
[0089] Then, in an embodiment of the present invention, based on the downsampled segmentation mask and the downsampled label, the loss function L of the downsampled segmentation model is calculated. d , and train the network using the stochastic gradient descent algorithm. Among them, the loss function L d It can be expressed as: Among them, Y d represents the aforementioned downsampled segmentation mask, represents the aforementioned downsampling label,
[0090] Then, in an embodiment of the present invention, after the downsampling segmentation model training is completed, the downsampling image of the three-dimensional cerebrovascular image sample is input into the trained downsampling segmentation model to obtain a global segmentation image corresponding to the sample, and based on dynamic window sampling, the global segmentation image is window sampled to obtain a set of cropping windows corresponding to the three-dimensional cerebrovascular image sample.
[0091] Furthermore, in an embodiment of the present invention, based on the cropping window set, the three-dimensional cerebral vascular image samples and their corresponding segmented image labels are cropped to obtain multiple sub-block segmentation masks and label sub-blocks corresponding to each sub-block segmentation mask. Then, based on the sub-block segmentation masks and the corresponding label sub-blocks, the loss function L can be calculated. s , the local segmentation model is trained by the stochastic gradient descent algorithm, where the loss function L s It can be expressed as: Among them, Y s represents the aforementioned sub-block segmentation mask, represents the aforementioned label sub-block,
[0092] In an embodiment of the present invention, the local segmentation model is iteratively trained by using the sub-block segmentation mask and the label sub-block corresponding to each sample, and finally a trained local segmentation model can be obtained.
[0093] Optionally, in an embodiment of the present invention, for a technical scenario in which a spatial fusion module is introduced, a specific implementation method of model training is described as follows.
[0094] First, the downsampling segmentation model training can be completed according to the aforementioned training method. Then, the model parameters of the downsampling segmentation model are fixed, the downsampling images of the three-dimensional cerebrovascular image samples are input into the downsampling segmentation model, and the feature map of the last layer of the feature extraction network in the output model is output.
[0095] Then, in this embodiment, the above-mentioned cropping window set is used to sample the three-dimensional cerebral vascular image samples and their labels, and the above-mentioned feature map in the vascular area to obtain image sub-blocks, label sub-blocks, and feature map sub-blocks, respectively. Then, the label sub-block and the feature map sub-block can be input into the spatial fusion module to generate fusion features, i.e., global features. Then, after splicing the image sub-block and the fusion feature, they are input into the local segmentation model of the above-mentioned dynamic window reasoning method to generate the corresponding sub-block segmentation mask; Further, in this embodiment, referring to the above method, based on the sub-block segmentation mask and the label sub-block, the loss function L is calculated. s The spatial fusion module and the local segmentation model are trained by the stochastic gradient descent algorithm until the training is completed.
[0096] The method of the embodiment of the present invention uses multiple groups of training samples to train the segmentation model, controls the loss value of each model within the convergence range, improves the model training accuracy, and further improves the model's accuracy in three-dimensional cerebrovascular image segmentation.
[0097] The three-dimensional cerebral vascular image segmentation device provided by the present invention is described below. The three-dimensional cerebral vascular image segmentation device described below and the three-dimensional cerebral vascular image segmentation method described above can be referred to each other.
[0098] Figure 5 Schematic diagram of the structure of the three-dimensional cerebral vascular image segmentation device provided by the present invention. Figure 5 As shown, it includes: a first segmentation module 510, a first cropping module 520, a second segmentation module 530 and a first splicing module 540 connected in sequence.
[0099] Among them, the first segmentation module 510 is used to input the downsampled image of the three-dimensional cerebrovascular image to be segmented into the downsampled segmentation model to obtain a global segmentation image of the downsampled image output by the downsampled segmentation model; the downsampled segmentation model is trained based on the three-dimensional cerebrovascular image samples and their corresponding segmented image labels.
[0100] The first cropping module 520 is used to perform window sampling on the global segmented image, determine a cropping window set of the global segmented image, and perform cropping processing on the three-dimensional cerebral vascular image based on the cropping window set to obtain multiple image sub-blocks.
[0101] The second segmentation module 530 is used to input multiple image sub-blocks into the local segmentation model to obtain a local segmentation image of each image sub-block output by the local segmentation model; the local segmentation model is trained using a trained downsampling segmentation model based on three-dimensional cerebrovascular image samples and their corresponding segmentation image labels.
[0102] The first stitching module 540 is used to stitch the local segmented images to obtain a segmentation result of a three-dimensional cerebral vascular image.
[0103] The three-dimensional cerebrovascular image segmentation device described in this embodiment can be used to execute the three-dimensional cerebrovascular image segmentation method embodiment described above. Its principles and technical effects are similar and will not be repeated here.
[0104] The three-dimensional cerebrovascular image segmentation device of the embodiment of the present invention obtains a downsampling segmentation model by training with three-dimensional cerebrovascular image samples and their corresponding segmented image labels, and further trains a local segmentation model based on the three-dimensional cerebrovascular image samples and their corresponding segmented image labels using the trained downsampling segmentation model, so that the downsampling segmentation model performs global segmentation on the downsampled image of the three-dimensional cerebrovascular image, and obtains a cropping window set of the adaptive three-dimensional cerebrovascular morphology through window sampling, and then obtains an image sub-block containing global semantics based on cropping processing of the original three-dimensional cerebrovascular image, and then locally segments each image sub-block through the local segmentation model, so as to splice each local segmented image, and finally obtain the segmentation result of the three-dimensional cerebrovascular image, thereby realizing an image segmentation method that integrates global semantics and local semantics, and can effectively improve the segmentation accuracy of the three-dimensional cerebrovascular image.
[0105] Figure 6 is a schematic diagram of the physical structure of the electronic device provided by the present invention, such as Figure 6 As shown, the electronic device may include: a processor (processor) 610, a communication interface (Communications Interface) 620, a memory (memory) 630 and a communication bus 640, wherein the processor 610, the communication interface 620, and the memory 630 communicate with each other through the communication bus 640. The processor 610 can call the logic instructions in the memory 630 to execute the three-dimensional cerebrovascular image segmentation method provided by the above methods, which includes: inputting the downsampled image of the three-dimensional cerebrovascular image to be segmented into the downsampled segmentation model to obtain the global segmented image of the downsampled image output by the downsampled segmentation model; the downsampled segmentation model is trained according to the three-dimensional cerebrovascular image samples and their corresponding segmented image labels; performing window sampling on the global segmented image to determine the cropping window set of the global segmented image, and cropping the three-dimensional cerebrovascular image based on the cropping window set to obtain multiple image sub-blocks; inputting the multiple image sub-blocks into the local segmentation model to obtain the local segmented image of each of the image sub-blocks output by the local segmentation model; the local segmentation model is trained based on the three-dimensional cerebrovascular image samples and their corresponding segmented image labels using the trained downsampled segmentation model; and splicing the local segmented images to obtain the segmentation result of the three-dimensional cerebrovascular image.
[0106] In addition, the logic instructions in the above-mentioned memory 630 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.
[0107] On the other hand, the present invention also provides a computer program product, which includes a computer program, which can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the three-dimensional cerebrovascular image segmentation method provided by the above methods, the method comprising: inputting the downsampled image of the three-dimensional cerebrovascular image to be segmented into a downsampled segmentation model to obtain a global segmented image of the downsampled image output by the downsampled segmentation model; the downsampled segmentation model is trained based on three-dimensional cerebrovascular image samples and their corresponding segmented image labels; performing window sampling on the global segmented image, determining a cropping window set of the global segmented image, and cropping the three-dimensional cerebrovascular image based on the cropping window set to obtain a plurality of image sub-blocks; inputting the plurality of image sub-blocks into a local segmentation model to obtain a local segmented image of each of the image sub-blocks output by the local segmentation model; the local segmentation model is trained based on the three-dimensional cerebrovascular image samples and their corresponding segmented image labels using a trained downsampled segmentation model; splicing each of the local segmented images to obtain a segmentation result of the three-dimensional cerebrovascular image.
[0108] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which is implemented by a processor to execute the three-dimensional cerebrovascular image segmentation method provided by the above-mentioned methods when the computer program is executed, the method comprising: inputting the downsampled image of the three-dimensional cerebrovascular image to be segmented into a downsampled segmentation model to obtain a global segmented image of the downsampled image output by the downsampled segmentation model; the downsampled segmentation model is trained based on three-dimensional cerebrovascular image samples and their corresponding segmented image labels; performing window sampling on the global segmented image, determining a cropping window set of the global segmented image, and cropping the three-dimensional cerebrovascular image based on the cropping window set to obtain a plurality of image sub-blocks; inputting the plurality of image sub-blocks into a local segmentation model to obtain a local segmented image of each of the image sub-blocks output by the local segmentation model; the local segmentation model is trained based on the three-dimensional cerebrovascular image samples and their corresponding segmented image labels using a trained downsampled segmentation model; and splicing each of the local segmented images to obtain a segmentation result of the three-dimensional cerebrovascular image.
[0109] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.
[0110] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.
[0111] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A three-dimensional cerebral vascular image segmentation method, characterized in that: include: Inputting the downsampled image of the three-dimensional cerebral vascular image to be segmented into the downsampled segmentation model to obtain a global segmented image of the downsampled image output by the downsampled segmentation model; The downsampling segmentation model is trained based on three-dimensional cerebrovascular image samples and their corresponding segmented image labels; Performing window sampling on the global segmented image, determining a clipping window set of the global segmented image, and clipping the three-dimensional cerebral vascular image based on the clipping window set to obtain a plurality of image sub-blocks; Inputting the multiple image sub-blocks into a local segmentation model to obtain a local segmentation image of each image sub-block output by the local segmentation model; the local segmentation model is trained based on the three-dimensional cerebrovascular image samples and their corresponding segmentation image labels using a trained downsampling segmentation model; The local segmented images are spliced to obtain the segmentation result of the three-dimensional cerebral vascular image.
2. The three-dimensional cerebral vascular image segmentation method according to claim 1, characterized in that: The downsampling segmentation model includes a feature extraction network; After the three-dimensional cerebral vascular image is cropped based on the cropping window set to obtain a plurality of image sub-blocks, the method further includes: Obtaining a target feature image output by the last layer of the feature extraction network; Inputting the target feature image and the global segmentation image into a spatial fusion module to obtain a global feature image of the three-dimensional cerebral vascular image output by the spatial fusion module; Splicing the global feature image with the multiple image sub-blocks respectively to obtain multiple spliced image sub-blocks; Each of the spliced image sub-blocks is locally segmented and spliced to obtain a segmentation result of the three-dimensional cerebral vascular image.
3. The three-dimensional cerebral vascular image segmentation method according to claim 2, characterized in that: The spatial fusion module is specifically used for: Cropping the global segmented image based on the cropping window set to obtain a plurality of cropped image sub-blocks, and cropping the target feature image based on the cropping window set to obtain a plurality of feature image sub-blocks; Determine a cerebral vascular tissue topology map based on the multiple cropped image sub-blocks and the multiple feature image sub-blocks; Feature extraction is performed on the cerebral vascular tissue topology map to obtain the global feature image.
4. The three-dimensional cerebral vascular image segmentation method according to claim 3, characterized in that: The determining of the cerebral vascular tissue topology map based on the multiple cropped image sub-blocks and the multiple feature image sub-blocks includes: For any cropped image sub-block among the plurality of cropped image sub-blocks, determine a first number of cerebrovascular voxels in the any cropped image sub-block, a second number of cerebrovascular voxels in any other cropped image sub-block among the plurality of cropped image sub-blocks, and a third number of identical vascular voxels in the any cropped image sub-block and any other cropped image sub-block; determine an adjacency between each of the cropped image sub-blocks based on the first number, the second number, and the third number, and determine an adjacency matrix based on each of the adjacency degrees; Performing global average pooling processing on each of the feature map sub-blocks to obtain a feature vector corresponding to each of the feature map sub-blocks; establishing the cerebral vascular tissue topology map with the feature vectors as nodes and the adjacency matrix as edges.
5. The three-dimensional cerebral vascular image segmentation method according to claim 3, characterized in that: The step of extracting features from the cerebral vascular tissue topology map to obtain the global feature image includes: Inputting the cerebrovascular tissue topology map into a graph neural network to obtain a node vector group output by the graph neural network; the graph neural network is used to encode and fuse information of adjacent topological nodes in the cerebrovascular tissue topology map; Each node vector in the node vector group is used as a weight vector of the corresponding cropping window in the cropping window set, and each weight vector is weightedly summed with the feature map sub-block of the corresponding cropping window to obtain a global feature of the blood vessel; the global feature of the blood vessel is upsampled to obtain the global feature image.
6. The three-dimensional cerebral vascular image segmentation method according to claim 1, characterized in that: The performing window sampling on the global segmented image to determine a clipping window set of the global segmented image includes: Converting the global segmented image into point cloud data; Performing farthest point sampling on the point cloud data to obtain sparse point cloud data corresponding to the global segmented image; A cropping window set of the global segmented image is generated based on a preset window size and with each point in the sparse point cloud data as the center.
7. The three-dimensional cerebral vascular image segmentation method according to any one of claims 1 to 6, characterized in that: Before inputting the downsampled image of the three-dimensional cerebral vascular image to be segmented into the downsampled segmentation model to obtain the global segmented image of the downsampled image output by the downsampled segmentation model, the method further includes: Using the acquired three-dimensional cerebral vascular image samples and their corresponding segmented image labels as a group of training samples, and acquiring multiple groups of the training samples; Using multiple groups of training samples to train the downsampling segmentation model to obtain a trained downsampling segmentation model; Inputting the downsampled image of the three-dimensional cerebrovascular image sample into the trained downsampled segmentation model to obtain a global segmentation image of the three-dimensional cerebrovascular image sample, and determining a cropping window set corresponding to the three-dimensional cerebrovascular image sample based on the global segmentation image of the three-dimensional cerebrovascular image sample; Based on the cropping window set corresponding to the three-dimensional cerebrovascular image sample, the three-dimensional cerebrovascular image sample and its corresponding segmented image label are cropped respectively to obtain a plurality of sub-block segmentation masks and a label sub-block corresponding to each sub-block segmentation mask; The local segmentation model is trained using the multiple sub-block segmentation masks and the label sub-block corresponding to each of the sub-block segmentation masks to obtain a trained local segmentation model.
8. A three-dimensional cerebral vascular image segmentation device, characterized in that: include: A first segmentation module is used to input the downsampled image of the three-dimensional cerebrovascular image to be segmented into a downsampled segmentation model to obtain a global segmented image of the downsampled image output by the downsampled segmentation model; the downsampled segmentation model is trained based on the three-dimensional cerebrovascular image samples and their corresponding segmented image labels; A first cropping module is used to perform window sampling on the global segmented image, determine a cropping window set of the global segmented image, and perform cropping processing on the three-dimensional cerebral vascular image based on the cropping window set to obtain a plurality of image sub-blocks; A second segmentation module is used to input the multiple image sub-blocks into a local segmentation model to obtain a local segmentation image of each image sub-block output by the local segmentation model; the local segmentation model is trained based on the three-dimensional cerebrovascular image samples and their corresponding segmentation image labels using a trained downsampling segmentation model; The first splicing module is used to splice the local segmented images to obtain the segmentation result of the three-dimensional cerebral vascular image.
9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the three-dimensional cerebrovascular image segmentation method as described in any one of claims 1 to 7 is implemented.
10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the three-dimensional cerebral vascular image segmentation method according to any one of claims 1 to 7 is implemented.
Citation Information
Patent Citations
Blood vessel segmentation method, device and computer readable medium
CN114037663A
Blood vessel segmentation method and device based on feature enhancement
CN114155193A
Blood vessel segmentation method and device, storage medium and electronic equipment
CN114612404A
A computer-implemented method for generating a 3-dimensional wireframe model of an object comprising a plurality of parts
US20220245896A1
Global and local feature reconstruction network-based medical image segmentation method
US20230274531A1
