Method, device and electronic device for cerebral blood vessel segmentation based on region growing and improved U-net

Abstract

The present application relates to a cerebrovascular segmentation method, device and electronic device based on region growing and improved U-net, the method comprising preprocessing the original cerebrovascular image to obtain a preprocessed cerebrovascular image, using a U-net neural network model to infer the preprocessed cerebrovascular image to obtain a preliminary segmentation image, automatically selecting seed points, performing region growing on the preliminary segmentation image to obtain a region growing image, and performing image fusion and image processing operations on the preliminary segmentation image and the region growing image to obtain a final fused segmentation image. The method achieves a more widely used cerebrovascular segmentation method by improving the fusion of the U-net deep learning segmentation method and the region growing segmentation method, which not only improves the accuracy of the deep learning segmentation method under general image quality conditions, but also greatly improves the efficiency and accuracy of seed point selection of the region growing segmentation method.

Description

Technical Field

[0001] The present application relates to the technical field of medical image segmentation, and in particular to a method, device and electronic device for cerebrovascular segmentation based on region growing and improved U-net. Background Art

[0002] With the continuous development of medical imaging technology, high-resolution medical imaging equipment has made the acquisition of brain images more convenient, such as computed tomography (CT), magnetic resonance imaging (MRI), etc., so doctors can obtain a large amount of brain image data. In the processing of brain image data, cerebrovascular segmentation method is one of the key technologies. However, general cerebrovascular segmentation methods mainly include threshold segmentation, region growing, morphological operations, etc., which generally have problems such as low segmentation accuracy, weak robustness, and low computational efficiency. How to quickly and accurately extract valuable cerebrovascular information from these complex medical images is a key problem that needs to be solved urgently. Summary of the invention

[0003] In view of this, the present application proposes a cerebral vascular segmentation method based on region growing and improved U-net to solve the problems reflected in the above-mentioned background technology.

[0004] According to one aspect of the present application, a cerebral blood vessel segmentation method based on region growing and improved U-net is provided, comprising the following steps:

[0005] Preprocessing the original cerebral vascular image to obtain a preprocessed cerebral vascular image;

[0006] Using a U-net neural network model to infer the preprocessed cerebral vascular image to obtain a preliminary segmentation image;

[0007] Automatically selecting seed points, performing region growing on the preliminary segmented image, and obtaining a region growing image;

[0008] The final fused segmented image is obtained by performing image fusion and image processing operations on the preliminary segmented image and the region growth image.

[0009] As an optional implementation scheme of the present application, optionally, a U-net neural network model is constructed and trained, and after completion, the original cerebral vascular image is inferred to obtain a preliminary segmented image, including:

[0010] The U-net neural network model is an improved U-net neural network model, which is mainly composed of an encoder and a decoder;

[0011] The U-net neural network model is trained using the preprocessed data, the optimal model in the training process is saved, and the original cerebral vascular image is inferred using the optimal model to obtain a preliminary segmented image.

[0012] As an optional implementation scheme of the present application, optionally, automatically selecting seed points and performing region growing on the preliminary segmented image to obtain a region growing image includes:

[0013] Automatically select the initial seed point and determine the growth criteria;

[0014] The region growing is performed according to the position and pixel value of the seed point to obtain the region growing image.

[0015] As an optional implementation scheme of the present application, optionally, it is characterized in that the final fused segmented image is obtained by image fusion and image processing, including:

[0016] fusing the region growing image and the preliminary segmented image to obtain a fully segmented cerebral vascular image;

[0017] The fully segmented cerebral vascular image is subjected to morphological post-processing to obtain a final fused segmented image.

[0018] As an optional implementation scheme of the present application, optionally, the improved U-net neural network model is a U-net neural network model that is integrated with a residual connection module.

[0019] As an optional implementation of the present application, optionally, the fully segmented cerebral vascular image is subjected to morphological post-processing to obtain a final image, including processing the fully segmented cerebral vascular image using two operations, an opening operation and a closing operation.

[0020] As an optional implementation of the present application, optionally, the original cerebral vascular image is preprocessed, including a processing method for CTA cerebral vascular images.

[0021] According to the second aspect of the present application, a cerebral blood vessel segmentation device based on region growing and improved U-net is provided, comprising the following modules:

[0022] A data preprocessing module is used to preprocess the original cerebral vascular image to obtain preprocessed data;

[0023] Constructing and training a model module, which is used to construct and train a U-net neural network model, and to perform reasoning on the original cerebrovascular image to obtain a preliminary segmented image;

[0024] A region growing module, used for automatically selecting seed points and performing region growing on the preliminary segmented image to obtain a region growing image;

[0025] The image fusion module is used to fuse the preliminary segmented image and the region growth image to obtain a fused image and perform morphological post-processing to obtain a final fused segmented image.

[0026] As an optional implementation scheme of the present application, optionally, constructing and training a model module includes:

[0027] Build a model module, which is used to build a U-net neural network model;

[0028] The training model module is used to train the constructed U-net neural network model to obtain the optimal model;

[0029] The model reasoning and segmentation module is used to use the optimal model to reason about the original cerebral vascular image to obtain a preliminary segmented image.

[0030] According to the third aspect of the present application, an electronic device is proposed, comprising:

[0031] processor;

[0032] a memory for storing processor-executable instructions;

[0033] Wherein, the processor is configured to implement a cerebral blood vessel segmentation method based on region growing and improved U-net and a cerebral blood vessel segmentation device based on region growing and improved U-net as described in any one of claims 1 to 8 when executing the executable instructions.

[0034] Beneficial effects of this application:

[0035] The present invention greatly improves the performance and stability of the cerebrovascular image segmentation results by integrating the U-net deep learning segmentation method and the regional growth segmentation method, especially when the image quality is not high, the present invention is significantly better than the segmentation method based on U-net deep learning alone. At the same time, the present invention also designs a preprocessing process for the original cerebrovascular image data, including image standardization, data enhancement, etc., which further improves the generalization ability of the model. When selecting seed points for regional growth, the present invention adopts the coordinate points of the improved U-net deep learning segmentation image as the seed points of regional growth, greatly improving the effectiveness and accuracy of seed point selection, integrating the segmentation advantages of regional growth at the pixel level, not only improving the accuracy (dice value) of the image segmented by the U-net deep learning segmentation method alone, but also making the cerebrovascular segmentation image still achieve high accuracy (dice value) when the segmentation quality is poor using the U-net deep learning segmentation method.

[0036] Other features and aspects of the present application will become apparent from the following detailed description of exemplary embodiments with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate exemplary embodiments, features, and aspects of the present application and, together with the description, serve to explain the principles of the present application.

[0038] Figure 1 A flowchart showing a cerebral blood vessel segmentation method based on region growing and improved U-net according to an embodiment of the present application is shown;

[0039] Figure 2 A block diagram of a cerebral blood vessel segmentation device based on region growing and improved U-net according to an embodiment of the present application is shown;

[0040] Figure 3 A comparison diagram showing the segmentation effect of the cerebral blood vessel segmentation method based on region growing and improved U-net and the segmentation effect of the U-net deep learning method alone according to the embodiment of the present application is shown;

[0041] Figure 4 The U_net example in the cerebral blood vessel segmentation method based on region growing and improved U-net in the embodiment of the present application is shown;

[0042] Figure 5 A comparison diagram of the cerebral vascular segmentation method based on region growing and improved U-net in an embodiment of the present application and the segmentation imaging using only the U-net deep learning method (two-dimensional slices of three-dimensional cerebral vascular imaging) is shown; DETAILED DESCRIPTION

[0043] Various exemplary embodiments, features and aspects of the present application will be described in detail below with reference to the accompanying drawings. The same reference numerals in the accompanying drawings represent elements with the same or similar functions. Although various aspects of the embodiments are shown in the accompanying drawings, the drawings are not necessarily drawn to scale unless otherwise specified.

[0044] Among them, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise", "axial", "radial", "circumferential" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present application or simplifying the description, and do not indicate or imply that the referred device or element must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as a limitation on the present application.

[0045] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of this application, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0046] The word “exemplary” is used exclusively herein to mean “serving as an example, example, or illustration.” Any embodiment described herein as “exemplary” is not necessarily to be construed as preferred or advantageous over other embodiments.

[0047] In addition, in order to better illustrate the present application, numerous specific details are given in the following specific embodiments. It should be understood by those skilled in the art that the present application can also be implemented without certain specific details. In some examples, methods, means, components and circuits well known to those skilled in the art are not described in detail in order to highlight the subject matter of the present application.

[0048] Example 1

[0049] Figure 1 FIG. 1 is a flowchart of a cerebral blood vessel segmentation method based on region growing and improved U-net according to the first embodiment of the present application. Figure 1 As shown, the flow chart includes:

[0050] S100: preprocessing the original cerebral vascular image to obtain a preprocessed cerebral vascular image.

[0051] In this embodiment, the original cerebral vascular images are subjected to image normalization processing, that is, four 96*96*96 volume blocks are randomly segmented from each image to obtain the preprocessed data.

[0052] S200, using a U-net neural network model to infer the preprocessed cerebral vascular image to obtain a preliminary segmented image.

[0053] In this embodiment, it is first necessary to construct a U-net neural network model, which is mainly composed of an encoder and a decoder. Then, the U-net neural network model is trained using the preprocessed data obtained in S100. After the training is completed, the original cerebrovascular image is inferred by the model obtained after the training to obtain the preliminary segmented image.

[0054] S300, automatically selecting seed points, performing region growing on the preliminary segmented image, and obtaining a region growing image.

[0055] In this embodiment, by finding the binary boundary of the preliminary segmented image after deep learning segmentation of the U-net neural network model, the position and pixel value of the seed point are extracted, and the edge of the preliminary segmented image is extended according to the pixel value, so that the preliminary segmented image continues to extend on the basis of the original image segmentation to obtain the region growth image. At the same time, a higher growth threshold needs to be set for the region growth to prevent excessive growth of the image.

[0056] S400 , obtaining a final fused segmented image by performing image fusion and image processing operations on the preliminary segmented image and the region growing image.

[0057] In this embodiment, a more fully segmented cerebral vascular image is obtained by performing image fusion on the preliminary segmentation image and the region growth image, and then the more fully segmented cerebral vascular image is subjected to morphological post-processing to make the image splicing smoother and remove some small unconnected segmentation results to obtain the final fused segmentation image.

[0058] As an optional implementation scheme of the present application, optionally, in step S200, the preprocessed cerebral vascular image is inferred using a U-net neural network model to obtain a preliminary segmented image, including: the U-net neural network model is an improved U-net neural network model, mainly composed of an encoder and a decoder; the U-net neural network model is trained using the preprocessed data, and the optimal model in the training process is saved; and the original cerebral vascular image is inferred using the optimal model to obtain a preliminary segmented image.

[0059] In this embodiment, the constructed U-net neural network model is an improved U-net neural network model, which further improves the performance of the model while maintaining the advantages of the original U-net model architecture. The improved U-net neural network model is mainly composed of an encoder and a decoder, wherein the encoder part consists of four levels, each level includes two 3D convolutional layers with ReLU activation, followed by a maximum pooling layer with a stride of 2. A Dropout layer is applied after the first convolutional layer. As the encoder goes deeper, the spatial resolution of the feature map is halved step by step, and the number of feature channels is doubled step by step. The number of feature channels at each level is 16, 32, 64, 128 and 256, respectively, and the downsampling stride of each level is 2. The decoder part gradually restores the original resolution of the input image by upsampling the feature map and splicing it with the feature map of the corresponding level of the encoder. Each level includes an upsampling operation, followed by two 3D convolutional layers, the first convolutional layer is followed by a ReLU activation and a Dropout layer, and the second convolutional layer is followed by a ReLU activation. Between the encoder and decoder, the bottom layer of the model includes two 3D convolutional layers without pooling operations. Two residual units are used in each encoder and decoder level. Finally, the model uses a 1x1 convolutional layer to map the feature map to binary predictions (i.e., vascular and non-vascular). The preprocessed data is input into the U-net neural network model to train the model, and the performance of the model is evaluated using the Dice loss function to obtain the optimal model in the training process. After the optimal model is selected through training, the optimal model is used to infer the unsegmented original cerebrovascular image to obtain a preliminary segmented image. The Dice loss function for evaluating the performance of the model is an indicator specifically used to evaluate the performance of the model in image segmentation tasks, which can effectively deal with the problem of class imbalance. The Dice coefficient measures the similarity of two sample sets and is defined as:

[0060]

[0061] In the actual training process, this method uses the continuous value version of the Dice loss function to facilitate back propagation and gradient optimization. Let p and g represent the predicted value and true label of the i-th pixel, respectively, then the continuous value Dice loss function is defined as:

[0062]

[0063] Where N is the total number of pixels in the image. By minimizing the Dice loss function, the model is able to better learn segmentation results that are more consistent with the true labels, especially when dealing with unbalanced data.

[0064] As an optional implementation scheme of the present application, optionally, in step S300, seed points are automatically selected to perform region growing on the preliminary segmented image, including: automatically selecting initial seed points and determining growth criteria; performing region growing according to the positions and pixel values ​​of the seed points to obtain a region growing image.

[0065] In this embodiment, the method of automatically selecting seed points is to select the coordinate points of the boundary points of the preliminary segmentation result (binary image) as the seed points for regional growth, record the pixel values ​​of the coordinate points, merge the same or similar pixels into the seeds, set a higher grayscale threshold to avoid excessive growth of seed points leading to segmentation errors, and obtain a regional growth image after regional growth is completed. The specific operation is: select the boundary points based on the preliminary segmentation results as the initial seed points, and convert the index of the boundary points into a seed point list:

[0066] Seed points={(x1,y1,z1),(x2,y2,z2),...,(x N ,y N ,z N )}

[0067] Where (x N ,y N , z N ) represents the coordinates of the Nth seed point, and N is the total number of seed points. In order to avoid the destruction of the initial segmentation results by region growing, a higher grayscale threshold is set, and region growing is performed on all the seed points found, extending to the neighboring pixels that meet the conditions.

[0068] As an optional implementation scheme of the present application, optionally, in step S400, a final fused segmented image is obtained by performing image fusion and image processing operations on the preliminary segmented image and the region growing image, including: fusing the region growing image and the preliminary segmented image to obtain a fully segmented cerebrovascular image; performing morphological post-processing on the fully segmented cerebrovascular image to obtain a final fused segmented image.

[0069] In this embodiment, the image fusion operation mainly fuses the preliminary segmented image and the region growth image, which can effectively avoid the low cerebrovascular segmentation accuracy when the deep learning method is used to segment the image in the face of images with poor quality or unclear features, and can also avoid the low segmentation accuracy commonly found in the fusion of traditional segmentation methods. After the image fusion operation, the image should continue to be morphologically post-processed. The morphological post-processing mainly uses two operations: opening operation and closing operation. The opening operation is to first perform an erosion operation and then a dilation operation. The closing operation is to first perform a dilation operation and then an erosion operation to fill small holes in the segmentation result, eliminate depressions on the boundary, make the boundary smoother, and eliminate small protruding areas in the segmentation result, and remove small connected components. After using the morphological post-processing operation, such as first performing an erosion operation on the image and then performing a dilation operation, or first dilating and then eroding, the fused segmented image can be spliced ​​more gently, and some small unconnected segmentation results are removed at the same time.

[0070] This method obtains the fusion segmentation result through the fusion segmentation method based on region growing and improved U-net, while the general U-net deep learning method can only obtain the deep learning segmentation result. If the evaluation method is used to evaluate the two results, the evaluation method selects the dice value and AHD (Average Hausdorff Distance). In the context of medical image segmentation, given two point sets A and B, point sets A and B usually represent the boundary or surface of the segmentation result and the reference standard. AHD measures the average distance between the two surfaces and provides a geometric similarity measure between the segmentation result and the reference standard.

[0071]

[0072] The evaluation results are as follows Figure 3 In 99 cases where the U-net deep learning method and the fusion segmentation method were used alone, the overall dice coefficient of the fusion segmentation method was significantly higher than that of the deep learning segmentation method. In extreme cases, even if the deep learning segmentation method obtained a very low dice coefficient, the dice value could still be improved to the average level after the fusion region growing method was used. This fully reflects the stability of the fusion segmentation method.

[0073] As an optional implementation scheme of the present application, optionally, the U-net neural network model is an improved U-net neural network model, including that the improved U-net neural network model is a U-net neural network model after integrating modules such as residual connection.

[0074] In this embodiment, the U-net neural network model is a convolutional neural network (CNN) architecture specially designed for biomedical image segmentation, which can be applied to various biomedical image segmentation tasks. The improved U-net neural network model refers to a model that has been optimized and enhanced to a certain extent on the basis of the original U-net architecture in order to improve performance and efficiency in image segmentation tasks. The main optimization in this method is reflected in the integration of modules such as residual connections. Residual connections are an important technology in deep learning. It adds direct connections (i.e., "shortcuts") between convolutional layers, allowing gradients to flow back to earlier layers more directly during back propagation, thereby helping to alleviate the problem of gradient vanishing. In the improved U-net model, residual connections are integrated into each convolutional block of the encoder and decoder, allowing the model to learn image features more deeply while maintaining the stability of training.

[0075] As an optional implementation of the present application, optionally, the fully segmented cerebral vascular image is subjected to morphological post-processing to obtain a final image, including processing the fully segmented cerebral vascular image using two operations, an opening operation and a closing operation.

[0076] In this embodiment, the two operations of opening and closing are two basic operations in morphological image processing, which are mainly used for tasks such as filtering, denoising, and segmentation of images. Both operations use structural elements (also called kernels or templates) to perform morphological transformations on images. Among them, the opening operation is the process of first performing an erosion operation on the image and then performing an expansion operation. The main function is to remove small bright spots in the image (areas smaller than the structural elements) while keeping the larger bright areas unchanged. It can also be used to disconnect small connections between objects and smooth the boundaries of objects. The closing operation is the process of first performing an expansion operation on the image and then performing an erosion operation. The main function is to fill small black spots in the image (areas smaller than the structural elements) while keeping the larger dark areas unchanged. It can also be used to connect adjacent objects and smooth the boundaries of objects. The erosion operation is to slide the structural element on the image, and for each position covered by the structural element, take the minimum value of the image pixel value at that position as the output pixel value. This step will reduce the bright area in the image and remove small bright spots or noise. The expansion operation is to slide the structural element on the image again, and for each position covered by the structural element, take the maximum value of the image pixel value at that position as the output pixel value. This step will restore bright areas in the image that were reduced by the erosion operation, but will not restore small bright spots or noise that were completely removed.

[0077] As an optional implementation of the present application, optionally, the original cerebral vascular image is preprocessed, including a processing method for CTA cerebral vascular images.

[0078] In this embodiment, since the brightness of bone imaging in CT images is high, which affects the model's extraction of cerebrovascular features, a deep learning method for removing skull bones is provided, that is, a method for processing CTA cerebrovascular images. Before model training, the original cerebrovascular images are first preprocessed, brain tissue is first extracted, and the part that affects the extraction of cerebrovascular features is removed. The U-net neural network model is then used for training, and the regional growing segmentation method and morphological post-processing are integrated.

[0079] Example 2

[0080] Based on the same principle as the above method, a cerebral vascular segmentation device based on region growing and improved U-net is also proposed, see Figure 2 A cerebral blood vessel segmentation device 100 based on region growing and improved U-net according to an embodiment of the present disclosure includes:

[0081] The data preprocessing module 110 is used to preprocess the original cerebral vascular image to obtain preprocessed data;

[0082] A model building and training module 120 is used to build and train a U-net neural network model to infer the original cerebral vascular image to obtain a preliminary segmented image;

[0083] A region growing module 130 is used to automatically select seed points and perform region growing on the preliminary segmented image to obtain a region growing image;

[0084] The image fusion module 140 is used to fuse the preliminary segmented image and the region growing image to obtain a fused image and perform morphological post-processing on the fused image.

[0085] As an optional implementation scheme of the present application, optionally, constructing and training a model module includes:

[0086] Build a model module, which is used to build a U-net neural network model;

[0087] The training model module is used to train the constructed U-net neural network model to obtain the optimal model;

[0088] The model reasoning and segmentation module is used to use the optimal model to reason about the original cerebral vascular image to obtain a preliminary segmented image.

[0089] Obviously, those skilled in the art should understand that all or part of the processes in the above-mentioned embodiment method can be implemented by instructing the relevant hardware through a computer program. The program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned control methods. The modules or steps of the present invention can be implemented by a general-purpose computing device. They can be concentrated on a single computing device or distributed on a network composed of multiple computing devices. Optionally, they can be implemented by program codes executable by computing devices, so that they can be stored in a storage device and executed by the computing device, or they can be made into individual integrated circuit modules respectively, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. In this way, the present invention is not limited to any specific combination of hardware and software.

[0090] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment method can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned control methods. Among them, the storage medium can be a disk, an optical disk, a read-only memory (ROM), a random access memory (RAM), a flash memory (Flash Memory), a hard disk (Hard Disk Drive, abbreviated as: HDD) or a solid-state drive (SSD), etc.; the storage medium can also include a combination of the above-mentioned types of memory.

[0091] Example 3

[0092] Furthermore, an electronic device is proposed, comprising:

[0093] processor;

[0094] a memory for storing processor-executable instructions;

[0095] Wherein, the processor is configured to implement the cerebral blood vessel segmentation method based on region growing and improved U-net described in Example 1 when executing the executable instructions.

[0096] The electronic device of the embodiment of the present disclosure includes a processor and a memory for storing processor executable instructions, wherein the processor is configured to implement any of the above-mentioned cerebral blood vessel segmentation methods based on region growing and improved U-net when executing the executable instructions.

[0097] It should be noted that the number of processors may be one or more. At the same time, the electronic device of the embodiment of the present disclosure may also include an input device and an output device. The processor, memory, input device, and output device may be connected via a bus or in other ways, which are not specifically limited here.

[0098] The memory is a computer-readable storage medium for the cerebral blood vessel segmentation method based on region growing and improved U-net, and can be used to store software programs, computer executable programs, and various modules, such as: the program or module corresponding to the cerebral blood vessel segmentation method based on region growing and improved U-net in the embodiment of the present disclosure. The processor executes various functional applications and data processing of the electronic device by running the software program or module stored in the memory.

[0099] The input device can be used to receive input numbers or signals. The signal can be a key signal related to user settings and function control of the device / terminal / server. The output device can include a display device such as a display screen.

[0100] The embodiments of the present application have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A cerebral vascular segmentation method based on region growing and improved U-net, characterized in that: By integrating the deep learning segmentation method and the region growing segmentation method, the cerebral vascular image is segmented, and the method includes the following steps: Preprocessing the original cerebral vascular image to obtain a preprocessed cerebral vascular image; Using a U-net neural network model to infer the preprocessed cerebral vascular image to obtain a preliminary segmentation image; Automatically selecting seed points, performing region growing on the preliminary segmented image, and obtaining a region growing image; The final fused segmented image is obtained by performing image fusion and image processing operations on the preliminary segmented image and the region growth image.

2. A cerebral blood vessel segmentation method based on region growing and improved U-net as claimed in claim 1, characterized in that: The U-net neural network model is an improved U-net neural network model, which is mainly composed of an encoder and a decoder; The U-net neural network model is trained using the preprocessed data, the optimal model in the training process is saved, and the original cerebral vascular image is inferred using the optimal model to obtain a preliminary segmented image.

3. The cerebral blood vessel segmentation method based on region growing and improved U-net as claimed in claim 1, characterized in that: Automatically selecting seed points and performing region growing on the preliminary segmented image to obtain a region growing image includes: Automatically select the initial seed point and determine the growth criteria; The region growing is performed according to the position and pixel value of the seed point to obtain the region growing image.

4. The cerebral blood vessel segmentation method based on region growing and improved U-net as claimed in claim 1, characterized in that: The final fused segmentation image is obtained through image fusion and image processing, including: fusing the region growing image and the preliminary segmented image to obtain a fully segmented cerebral vascular image; The fully segmented cerebral vascular image is subjected to morphological post-processing to obtain a final fused segmented image.

5. The cerebral blood vessel segmentation method based on region growing and improved U-net as claimed in claim 2, characterized in that: The improved U-net neural network model is a U-net neural network model that incorporates a residual connection module.

6. The cerebral blood vessel segmentation method based on region growing and improved U-net as claimed in claim 4, characterized in that: The fully segmented cerebral vascular image is subjected to morphological post-processing to obtain a final image, including processing the fully segmented cerebral vascular image using two operations, an opening operation and a closing operation.

7. The cerebral blood vessel segmentation method based on region growing and improved U-net as claimed in claim 1, characterized in that: The original cerebral vascular image is preprocessed, including a processing method for CTA cerebral vascular image.

8. A cerebral vascular segmentation device based on region growing and improved U-net, characterized in that: include: A data preprocessing module is used to preprocess the original cerebral vascular image to obtain preprocessed data; Constructing and training a model module, which is used to construct and train a U-net neural network model, and to perform reasoning on the original cerebrovascular image to obtain a preliminary segmented image; A region growing module, used for automatically selecting seed points and performing region growing on the preliminary segmented image to obtain a region growing image; The image fusion module is used to fuse the preliminary segmented image and the region growth image to obtain a fused image and perform morphological post-processing on the fused image to obtain a final fused segmented image.

9. The cerebral blood vessel segmentation device based on region growing and improved U-net as claimed in claim 8, characterized in that: Build and train model modules, including: Build a model module, which is used to build a U-net neural network model; The training model module is used to train the constructed U-net neural network model to obtain the optimal model; The model reasoning and segmentation module is used to use the optimal model to reason about the original cerebral vascular image to obtain a preliminary segmented image.

10. An electronic device, characterized in that: include: processor; a memory for storing processor-executable instructions; Wherein, the processor is configured to implement a cerebral blood vessel segmentation method based on region growing and improved U-net and a cerebral blood vessel segmentation device based on region growing and improved U-net as described in any one of claims 1 to 8 when executing the executable instructions.

Images