Aortic vessel segmentation method and device, storage medium and electronic equipment

Abstract

The present disclosure relates to an aorta blood vessel segmentation method, device, storage medium and electronic equipment. The method comprises: acquiring an aorta CTA image; inputting the CTA image into a trained aorta trunk blood vessel extraction model to obtain an aorta trunk blood vessel image, the aorta trunk blood vessel image being a binary image; performing image enhancement processing on the CTA image to obtain an aorta branch blood vessel enhancement image, the aorta branch blood vessel enhancement image being a binary image; and segmenting an aorta blood vessel image from the CTA image according to the aorta trunk blood vessel image and the aorta branch blood vessel enhancement image. In this way, the complete aorta trunk blood vessel and branch blood vessel can be effectively segmented.

Description

Technical Field

[0001] This disclosure relates to the field of image processing technology, and more specifically, to a method, apparatus, storage medium, and electronic device for aortic vessel segmentation. Background Technology

[0002] In related technologies, with the development of deep learning technology, medical image segmentation schemes based on deep learning have been widely used. Currently, deep learning technology can be used to segment aortic vessel images. However, while deep learning-based aortic vessel image segmentation methods can effectively extract the main aortic trunk, their extraction of aortic branch vessels is very limited, especially the extraction or segmentation of the fine aortic branch vessels in the human abdomen. Therefore, there is an urgent need for a method to extract the complete main aortic trunk and branch vessels. Summary of the Invention

[0003] To address the problems existing in related technologies, this disclosure provides a method, apparatus, storage medium, and electronic device for aortic vessel segmentation.

[0004] To achieve the above objectives, a first aspect of this disclosure provides a method for aortic vessel segmentation, the method comprising:

[0005] Acquire aortic CTA images;

[0006] The CTA image is input into the trained aortic trunk vessel extraction model to obtain an aortic trunk vessel image, which is a binary image.

[0007] The CTA image is subjected to image enhancement processing to obtain an enhanced image of the aortic branch vessels, which is a binary image;

[0008] Based on the enhanced images of the main aortic vessels and the aortic branch vessels, the aortic vessel image is segmented from the CTA image.

[0009] Optionally, the step of performing image enhancement processing on the CTA image to obtain an enhanced image of the aortic branch vessels includes:

[0010] The Hessian image corresponding to the CTA image is determined based on the Hessian recursive Gaussian function.

[0011] The Hessian image is processed using the Hessian 3D vessel metric function to obtain an enhanced image of the tubular structure.

[0012] The enhanced image of the tubular object is subjected to binary threshold segmentation based on a preset first tubular object metric range to obtain the enhanced image of the aortic branch vessels.

[0013] Optionally, determining the Hessian image corresponding to the CTA image based on the Hessian recursive Gaussian function includes:

[0014] The CTA image is processed using a Hessian recursive Gaussian function with a sigma value of a first preset value to obtain a first Hessian image;

[0015] The CTA image is processed using a Hessian recursive Gaussian function with a sigma value of a second preset value to obtain a second Hessian image;

[0016] The Hessian image includes the first Hessian image and the second Hessian image.

[0017] Optionally, the Hessian image includes a first Hessian image and a second Hessian image; correspondingly, the tubular enhanced image includes a first tubular enhanced image and a second tubular enhanced image. The step of performing binary threshold segmentation processing on the tubular enhanced image according to a preset first tubular object metric range to obtain the aortic branch vessel enhanced image includes:

[0018] The enhanced image of the first tubular object is subjected to binary threshold segmentation based on the measurement range of the first tubular object to obtain the enhanced image of the first branch blood vessel.

[0019] The enhanced image of the second tubular object is subjected to binary threshold segmentation based on the measurement range of the first tubular object to obtain the enhanced image of the second branch blood vessel.

[0020] The enhanced images of aortic branch vessels include a first enhanced image of a branch vessel and a second enhanced image of a branch vessel. The vessel thickness in the first enhanced image of a branch vessel is positively correlated with the sigma value used when determining the first Hessian image. Correspondingly, the vessel thickness in the second enhanced image of a branch vessel is positively correlated with the sigma value used when determining the second Hessian image.

[0021] Optionally, the enhanced image of the aortic branch vessels includes enhanced images of the first branch vessels and the second branch vessels. The step of segmenting the aortic vessel image from the CTA image based on the aortic trunk vessel image and the enhanced images of the aortic branch vessels includes:

[0022] Image registration processing is performed on the main aortic trunk image, the enhanced image of the first branch vessel, and the enhanced image of the second branch vessel.

[0023] The union of the registered aortic trunk image, the enhanced image of the first branch vessel, and the enhanced image of the second branch vessel is determined as the first aortic vessel image;

[0024] The first aortic vessel image is denoised to obtain the second aortic vessel image.

[0025] The second aortic vascular image is processed to complete the vascular image, thus obtaining the aortic vascular image.

[0026] Optionally, the step of denoising the first aortic vessel image to obtain the second aortic vessel image includes:

[0027] The first aortic vessel image is subjected to erosion processing to obtain an eroded first aortic vessel image;

[0028] Connectivity component analysis was performed on the corroded first aortic vessel image to determine the first target connectivity component with the largest number of voxels.

[0029] The binary image corresponding to the first target connected component is dilated to obtain the second aortic blood vessel image.

[0030] Optionally, the enhanced tubular image includes a first enhanced tubular image and a second enhanced tubular image, and the step of performing vascular completion processing on the second aortic vascular image to obtain the aortic vascular image includes:

[0031] The target tubular enhanced image is subjected to binary threshold segmentation based on a preset second tubular object measurement value range to obtain a third branch vessel enhanced image. The target tubular enhanced image is one of the first tubular enhanced image and the second tubular enhanced image. The lower limit of the second tubular object measurement value range is greater than the lower limit of the first tubular object measurement value range.

[0032] Image registration processing is performed on the aortic trunk vessel image and the enhanced image of the third branch vessel, and the union of the registered aortic trunk vessel image and the enhanced image of the third branch vessel is determined as the third aortic vessel image;

[0033] Connectivity component analysis was performed on the third aortic vascular image to identify the second target connectivity component that includes the largest number of voxels;

[0034] The union of the binary image corresponding to the second target connected component and the second aortic vascular image is determined as the aortic vascular image.

[0035] A second aspect of this disclosure provides an aortic vessel segmentation device, the device comprising:

[0036] The acquisition module is used to acquire aortic CTA images;

[0037] The input module is used to input the CTA image into the trained aortic trunk vessel extraction model to obtain an aortic trunk vessel image, wherein the aortic trunk vessel image is a binary image.

[0038] The execution module is used to perform image enhancement processing on the CTA image to obtain an enhanced image of the aortic branch vessels, wherein the enhanced image of the aortic branch vessels is a binary image;

[0039] The segmentation module is used to segment the aortic vessel image from the CTA image based on the aortic trunk vessel image and the enhanced aortic branch vessel image.

[0040] Optionally, the execution module includes:

[0041] The first determining submodule is used to determine the Hessian image corresponding to the CTA image based on the Hessian recursive Gaussian function;

[0042] The second determining submodule is used to process the Hessian image according to the Hessian3D blood vessel measurement function to obtain an enhanced image of the tubular structure;

[0043] The first segmentation submodule is used to perform binary threshold segmentation processing on the enhanced image of the tubular object according to a preset first tubular object metric value range to obtain the enhanced image of the aortic branch vessels.

[0044] Optionally, the first determining submodule includes:

[0045] The first processing submodule is used to process the CTA image according to a Hessian recursive Gaussian function with a sigma value of a first preset value to obtain a first Hessian image.

[0046] The second processing submodule is used to process the CTA image according to a Hessian recursive Gaussian function with a sigma value of a second preset value to obtain a second Hessian image; the Hessian image includes the first Hessian image and the second Hessian image.

[0047] Optionally, the Hessian image includes a first Hessian image and a second Hessian image; correspondingly, the tubular enhanced image includes a first tubular enhanced image and a second tubular enhanced image; the first segmentation submodule includes:

[0048] The second segmentation submodule is used to perform binary threshold segmentation processing on the enhanced image of the first tubular object based on the measurement value range of the first tubular object to obtain the enhanced image of the first branch blood vessel.

[0049] The third segmentation submodule is used to perform binary threshold segmentation processing on the second tubular enhanced image based on the first tubular object metric range to obtain the second branch vessel enhanced image; wherein, the aortic branch vessel enhanced image includes the first branch vessel enhanced image and the second branch vessel enhanced image, the vessel thickness in the first branch vessel enhanced image is positively correlated with the sigma value used when determining the first Hessian image, and correspondingly, the vessel thickness in the second branch vessel enhanced image is positively correlated with the sigma value used when determining the second Hessian image.

[0050] Optionally, the enhanced image of the aortic branch vessels includes enhanced images of the first branch vessels and the second branch vessels, and the segmentation module includes:

[0051] The first registration submodule is used to perform image registration processing on the aortic trunk image, the enhanced image of the first branch vessel, and the enhanced image of the second branch vessel.

[0052] The first union operation submodule is used to determine the first aortic vessel image by the union of the registered aortic trunk vessel image, the first branch vessel enhanced image, and the second branch vessel enhanced image.

[0053] The noise reduction submodule is used to perform noise reduction processing on the first aortic vessel image to obtain the second aortic vessel image.

[0054] The completion submodule is used to perform vascular completion processing on the second aortic vascular image to obtain the aortic vascular image.

[0055] Optionally, the noise reduction submodule is used for:

[0056] The first aortic vessel image is subjected to erosion processing to obtain an eroded first aortic vessel image; connected component analysis is performed on the eroded first aortic vessel image to determine the first target connected component with the largest number of voxels; the binary image corresponding to the first target connected component is subjected to dilation processing to obtain the second aortic vessel image.

[0057] Optionally, the enhanced tubular image includes a first enhanced tubular image and a second enhanced tubular image, and the completion submodule is used for:

[0058] The target tubular enhanced image is subjected to binary threshold segmentation based on a preset second tubular object measurement value range to obtain a third branch vessel enhanced image. The target tubular enhanced image is one of the first tubular enhanced image and the second tubular enhanced image. The lower limit of the second tubular object measurement value range is greater than the lower limit of the first tubular object measurement value range.

[0059] Image registration processing is performed on the aortic trunk vessel image and the enhanced image of the third branch vessel, and the union of the registered aortic trunk vessel image and the enhanced image of the third branch vessel is determined as the third aortic vessel image;

[0060] Connectivity component analysis was performed on the third aortic vascular image to identify the second target connectivity component that includes the largest number of voxels;

[0061] The union of the binary image corresponding to the second target connected component and the second aortic vascular image is determined as the aortic vascular image.

[0062] A third aspect of this disclosure provides a non-transitory computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the steps of the aortic vessel segmentation method described in the first aspect.

[0063] A fourth aspect of this disclosure provides an electronic device, including:

[0064] A memory on which computer programs are stored;

[0065] A processor is configured to execute the computer program in the memory to implement the steps of the aortic vessel segmentation method described in the first aspect.

[0066] By adopting the above technical solution, at least the following beneficial technical effects can be achieved:

[0067] Aortic CTA images are acquired. These CTA images are then input into a trained aortic trunk vessel extraction model to obtain an aortic trunk vessel image. Image enhancement processing is performed on the CTA images to obtain enhanced aortic branch vessel images. Based on these two binary images—the aortic trunk vessel image and the enhanced aortic branch vessel image—the aortic vessels can be segmented from the CTA image. This method, utilizing a deep learning-based aortic trunk vessel extraction model, effectively extracts the aortic trunk vessels, while image enhancement processing of the CTA image highlights the aortic branch vessels, making their segmentation easier. Therefore, this method effectively extracts the complete aortic trunk and branch vessels.

[0068] Other features and advantages of this disclosure will be described in detail in the following detailed description section. Attached Figure Description

[0069] The accompanying drawings are provided to further illustrate the present disclosure and form part of the specification. They are used together with the following detailed description to explain the present disclosure, but do not constitute a limitation thereof. In the drawings:

[0070] Figure 1 This is a flowchart illustrating an aortic vessel segmentation method according to an exemplary embodiment.

[0071] Figure 2 This is a grayscale image of an aortic CTA image according to an exemplary embodiment.

[0072] Figure 3 This is a grayscale image of the main aortic trunk, as illustrated in an exemplary embodiment.

[0073] Figure 4 This is a grayscale image of the main aortic trunk and its branches, as illustrated in an exemplary embodiment.

[0074] Figure 5 This is a block diagram illustrating an aortic vessel segmentation device according to an exemplary embodiment.

[0075] Figure 6 This is a block diagram illustrating an electronic device for implementing aortic vessel segmentation according to an exemplary embodiment. Detailed Implementation

[0076] The specific embodiments of this disclosure will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit this disclosure.

[0077] It should be noted that all actions involving the acquisition of signals, information, or data in this disclosure are carried out in compliance with the relevant data protection laws and policies of the country where the location is situated, and with authorization from the owner of the relevant device.

[0078] In healthy individuals, the aorta is intact and has a uniform blood density, making it easy to identify and extract the aortic vessels from aortic angiography images. However, in cases of aortic disease, such as aortic dissection, the aortic vessels become difficult to identify and extract from aortic angiography images due to the lower density of the false lumen and the blurring of the aortic vessels with surrounding tissues.

[0079] In related technologies, with the development of deep learning technology, medical image segmentation schemes based on deep learning have been widely used. Currently, deep learning technology can be used to segment aortic vessel images. However, while deep learning-based aortic vessel image segmentation methods can effectively extract the main aortic trunk, their extraction of aortic branch vessels is very limited, especially in extracting or segmenting the small aortic branch vessels in the human abdomen. This is because the volume difference between the main aortic trunk and aortic branch vessels is significant, making it difficult for the model to simultaneously learn the features of both. Furthermore, the aortic branch vessels are too small to be labeled, resulting in a lack of effective labeled data for training the model, making it difficult to obtain a highly accurate model for segmenting aortic branch vessels.

[0080] In view of this, embodiments of the present disclosure provide a method, apparatus, storage medium and electronic device for aortic vessel segmentation, so as to effectively segment the complete main aortic vessels and branch vessels.

[0081] Figure 1 This is a flowchart illustrating an aortic vessel segmentation method according to an exemplary embodiment. Figure 1 As shown, this aortic vessel segmentation method includes the following steps:

[0082] S11. Obtain aortic CTA images.

[0083] Aortic CTA (Computed Tomography Angiography) images refer to the images obtained after a patient undergoes an aortic CTA examination; they can also be called aortic CT angiography images. Aortic CTA images are 3D images. CT (Computed Tomography) is a technique that uses precisely collimated X-ray beams, gamma rays, ultrasound waves, etc., along with highly sensitive detectors, to perform a series of cross-sectional scans around a specific part of the body, thereby obtaining three-dimensional CT images.

[0084] Because the aorta is the largest artery in the human body, it originates from the left ventricle of the heart, ascends, then curves slightly downwards in an arc shape, and continues down the spine, branching into many smaller arteries within the thoracic and abdominal cavities. Therefore, aortic CTA examinations primarily focus on the thoracic and abdominal cavities. Aortic CTA images include 3D images of the aorta, which encompass the main aortic trunk and its branches.

[0085] Aortic CTA images typically consist of more than 500 2D layers, each with a high resolution of up to 512*512. The axial spacing between layers is generally between 0.5 and 2 mm. The axial direction refers to the Z-axis of the CT scan, which is typically the direction of the human spine.

[0086] S12. Input the CTA image into the trained aortic trunk vessel extraction model to obtain an aortic trunk vessel image, wherein the aortic trunk vessel image is a binary image.

[0087] The aortic trunk vessel extraction model disclosed herein can be a model employing the nnUNet deep learning framework from related technologies. Inputting a CTA image into the trained aortic trunk vessel extraction model yields an aortic trunk vessel image output by the model, which is a binary image. In the binary image, voxels with a value of 1 represent aortic trunk vessels, and voxels with a value of 0 represent non-aortic trunk vessels. It should be noted that since the training method of the aortic trunk vessel extraction model is similar to that of the 2D UNet model in related technologies, the training method will not be elaborated here. A voxel, in CT image processing, is a cube of equal volume divided into several slices; each cube is called a voxel. That is, a voxel is the smallest unit of three-dimensional spatial segmentation. Voxels are used in three-dimensional imaging, scientific data, and medical imaging. Conceptually, a voxel is similar to a pixel, the smallest unit in two-dimensional space, except that a pixel is used in two-dimensional computer image data.

[0088] In some implementations, when the architecture of the aortic trunk vessel extraction model is the nnUNet deep learning framework in related technologies, the aortic trunk vessel extraction model includes multiple downsampling network layers. The loss function of each downsampling network layer may include a cross-entropy loss function and a Dice loss function. The weighted sum of the loss values ​​corresponding to all downsampling network layers is the total loss of the aortic trunk vessel extraction model. For example, assuming the number of downsampling network layers is 5, the total loss is L = w1·L1 + w2·L2 + w3·L3 + w4·L4 + w5·L5, where w represents the weight and L represents the loss of the downsampling network layer. The weight w is halved as the resolution of the downsampling network layer decreases, i.e., w2 = 1 / 2·w1; w3 = 1 / 4·w1; w4 = 1 / 8·w1; w5 = 1 / 16·w1, and w1 + w2 + w3 + w4 + w5 = 1.

[0089] S13. Perform image enhancement processing on the CTA image to obtain an enhanced image of the aortic branch vessels, wherein the enhanced image of the aortic branch vessels is a binary image.

[0090] Image enhancement processing refers to enhancing the useful information in an image. It can be a distortion process, but its purpose is to improve the visual effect of the image. This includes intentionally emphasizing the overall or local characteristics of an image, making an originally blurry image clearer, highlighting certain features of interest, amplifying the differences between features of different objects in the image, and suppressing features of little interest. These methods improve image quality, enrich information content, and enhance image interpretation and recognition to meet the needs of image analysis.

[0091] In this embodiment of the disclosure, an enhanced image of the aortic branch vessels can be obtained by performing image enhancement processing on the CTA image to highlight the aortic branch vessels in the CTA image.

[0092] S14. Based on the aortic trunk image and the enhanced aortic branch image, segment the aortic vessel image from the CTA image.

[0093] Since both the main aortic vessel image and the enhanced images of the aortic branches are binary images, the aortic vessel image, including the main aortic vessel and the aortic branches, can be segmented from the CTA image based on these images. For example, the target coordinates of voxels with a value of 1 in the binary image are determined. Then, the voxels corresponding to the target coordinates in the CTA image are extracted, and these extracted voxels constitute the aortic vessel image.

[0094] The above method involves acquiring aortic CTA images. These images are then input into a trained aortic trunk vessel extraction model to obtain aortic trunk vessel images. Image enhancement processing is performed on the CTA images to obtain enhanced images of aortic branch vessels. Based on the aortic trunk vessel images and the enhanced aortic branch vessel images, the aortic vessels are segmented from the CTA images. This method, utilizing a deep learning-based aortic trunk vessel extraction model, effectively extracts the aortic trunk vessels, while image enhancement processing of the CTA images highlights the aortic branch vessels, making their segmentation easier. Therefore, this method effectively extracts the complete aortic trunk and branch vessels.

[0095] Optionally, the step of performing image enhancement processing on the CTA image to obtain an enhanced image of the aortic branch vessels includes:

[0096] The Hessian image corresponding to the CTA image is determined based on the Hessian recursive Gaussian function; the Hessian image is processed according to the Hessian 3D vessel metric function to obtain the tubular enhanced image; the tubular enhanced image is subjected to binary threshold segmentation processing according to the preset first tubular object metric value interval to obtain the aortic branch vessel enhanced image.

[0097] The Hessian recursive Gaussian function refers to the `hessian_recursive_gaussian` function in the ITK image processing tool, which is used to solve for the Hessian image of the input image. The principle and processing procedure of the `hessian_recursive_gaussian` function can be found in the descriptions in related technologies or in the source code of the `hessian_recursive_gaussian` function; this disclosure will not explain them further.

[0098] The Hessian3D vessel measurement function refers to the Hessian3DtoVesselnessMeasure function in the ITK image processing tool, which is used in related technologies. This Hessian3DtoVesselnessMeasure function is used to obtain measurements of tubular objects in an image to generate a tubular-enhanced image. The principle and processing procedure of the Hessian3DtoVesselnessMeasure function can be found in the descriptions in related technologies or in the source code of the Hessian3DtoVesselnessMeasure function; this disclosure will not explain them further.

[0099] In practical implementation, the ITK image processing tool can be used to solve for the Hessian image corresponding to the CTA image using the Hessian recursive Gaussian function. The Hessian image is then processed using the Hessian3D vessel metric function to obtain tubular object metrics, generating an enhanced tubular image. Since the Hessian image is processed using the Hessian3D vessel metric function, the non-tubular object metrics in the generated enhanced tubular image are generally less than 2, while the tubular object metrics are generally no more than 500. Therefore, the first tubular object metric value range can be set to [2, 500]. Thus, binary thresholding can be performed on the enhanced tubular image based on the first tubular object metric value range [2, 500] to obtain an enhanced image of the aortic branch vessels.

[0100] The algorithm for implementing binary threshold segmentation can be the BinaryThreshold function in the ITK image processing tool, which is a related technology. The process of performing binary threshold segmentation on the tubular enhanced image based on the first tubular object measurement value interval [2, 500] is roughly as follows: the voxel values ​​in the tubular enhanced image that are in the interval [2, 500] are reset to 1, and the values ​​of the remaining voxels are reset to 0. The resulting binary image is the aortic branch vessel enhanced image.

[0101] Optionally, determining the Hessian image corresponding to the CTA image based on the Hessian recursive Gaussian function includes:

[0102] The CTA image is processed using a Hessian recursive Gaussian function with a sigma value of a first preset value to obtain a first Hessian image; the CTA image is then processed using a Hessian recursive Gaussian function with a sigma value of a second preset value to obtain a second Hessian image; the Hessian image includes the first Hessian image and the second Hessian image.

[0103] An important parameter of the hessian_recursive_gaussian function in the ITK image processing tool is sigma, which is the amount of smoothing used during Hessian image estimation. It can be simply understood as the variance of the Gaussian function component.

[0104] Considering the uneven thickness of aortic branch vessels, this embodiment proposes setting multiple different values ​​for the sigma parameter to better extract the aortic branch fine and large vessels. For example, the sigma parameter is set to a first preset value of 2.5 to better extract the aortic branch large vessels. Alternatively, the sigma parameter is set to a second preset value of 1.0 to better extract the aortic branch fine vessels. Thus, by processing the CTA image using a Hessian recursive function with a sigma value of 2.5, a first Hessian image corresponding to the aortic branch large vessels can be obtained. By processing the CTA image using a Hessian recursive function with a sigma value of 1.0, a second Hessian image corresponding to the aortic branch fine vessels can be obtained.

[0105] When the Hessian image includes a first Hessian image and a second Hessian image, the tubular enhancement image correspondingly includes both a first tubular enhancement image and a second tubular enhancement image. For example, the first Hessian image is processed according to the Hessian3D vessel metric function to obtain a first tubular enhancement image. The second Hessian image is processed according to the Hessian3D vessel metric function to obtain a second tubular enhancement image.

[0106] Optionally, the step of performing binary threshold segmentation on the enhanced image of the tubular object according to a preset first tubular object metric range to obtain the enhanced image of the aortic branch vessels includes:

[0107] The first tubular enhanced image is segmented using binary thresholding based on the first tubular object metric range to obtain a first branch vessel enhanced image; the second tubular enhanced image is segmented using binary thresholding based on the first tubular object metric range to obtain a second branch vessel enhanced image; wherein the aortic branch vessel enhanced image includes the first branch vessel enhanced image and the second branch vessel enhanced image, the vessel thickness in the first branch vessel enhanced image is positively correlated with the sigma value used when determining the first Hessian image, and correspondingly, the vessel thickness in the second branch vessel enhanced image is positively correlated with the sigma value used when determining the second Hessian image.

[0108] For example, assuming the first tubular object's metric value range is [2, 500], performing binary thresholding on the first tubular object enhanced image based on this range yields a first branch vessel enhanced image. Performing binary thresholding on the second tubular object enhanced image based on the same range yields a second branch vessel enhanced image. The vessel thickness in the first branch vessel enhanced image is positively correlated with the sigma value used to determine the first Hessian image, and similarly, the vessel thickness in the second branch vessel enhanced image is positively correlated with the sigma value used to determine the second Hessian image. For instance, if the sigma value for the first Hessian image is 2.5 and the sigma value for the second Hessian image is 1.0, then the vessels in the first branch vessel enhanced image are thicker than those in the second branch vessel enhanced image.

[0109] Optionally, the enhanced image of the aortic branch vessels includes enhanced images of the first branch vessels and the second branch vessels. The step of segmenting the aortic vessel image from the CTA image based on the aortic trunk vessel image and the enhanced images of the aortic branch vessels includes:

[0110] Image registration processing is performed on the aortic trunk image, the enhanced image of the first branch vessel, and the enhanced image of the second branch vessel; the union of the registered aortic trunk image, the enhanced image of the first branch vessel, and the enhanced image of the second branch vessel is determined as the first aortic vessel image; noise reduction processing is performed on the first aortic vessel image to obtain the second aortic vessel image; vessel completion processing is performed on the second aortic vessel image to obtain the aortic vessel image.

[0111] Registration refers to the matching of geographic coordinates between different images of the same region obtained using different imaging methods. It includes geometric correction, projection transformation, and scaling. In this disclosure, image registration refers to the process of matching and overlaying two or more images acquired at different times, with different sensors (imaging devices), or under different conditions (weather, illumination, camera position and angle, etc.). The registration process generally involves: extracting feature points from multiple images; finding matching feature point pairs between points in one image and points in another image through similarity measurement; obtaining image spatial coordinate transformation parameters from the matched feature point pairs; and finally, performing image registration using the coordinate transformation parameters.

[0112] For example, after obtaining enhanced images of the first and second branch vessels, representing the fine vessels of the aortic branches and the thick vessels of the aortic branches respectively, image registration processing can be performed on the aortic trunk vessel image, the first branch vessel enhanced image, and the second branch vessel enhanced image. The union of the registered aortic trunk vessel image, the first branch vessel enhanced image, and the second branch vessel enhanced image is determined as the first aortic vessel image. That is, the voxels with a value of 1 in the first aortic vessel image are the union of the voxels with a value of 1 in the registered aortic trunk vessel image, the first branch vessel enhanced image, and the second branch vessel enhanced image. This first aortic vessel image includes the aortic trunk vessels, the fine vessels of the aortic branches, and the thick vessels of the aortic branches.

[0113] Inevitably, the enhanced images of the first and second branch vessels may contain interfering information generated from the enhancement of other tubular structures, such as bones and marginal regions. Therefore, after obtaining the first aortic vessel image from the enhanced images of the first and second branch vessels, it is necessary to perform noise reduction processing (i.e., interference removal processing) on ​​the first aortic vessel image to obtain the second aortic vessel image. However, during the noise reduction process of the first aortic vessel image, small vessels at the tail end are often treated as interference and eliminated. Therefore, further vessel completion processing is required on the second aortic vessel image to obtain a complete aortic vessel image.

[0114] Optionally, the step of denoising the first aortic vessel image to obtain the second aortic vessel image includes:

[0115] The first aortic vessel image is subjected to erosion processing to obtain an eroded first aortic vessel image; connected component analysis is performed on the eroded first aortic vessel image to determine the first target connected component with the largest number of voxels; the binary image corresponding to the first target connected component is subjected to dilation processing to obtain the second aortic vessel image.

[0116] For example, the BinaryErodeImageFilter function in the ITK image processing tool is used to erode the first aortic vessel image to separate the virtually connected tubular structures, thus obtaining an eroded first aortic vessel image. Based on the biological characteristics of the aorta, which occupies the largest proportion in the image, connected component analysis can be performed on the eroded first aortic vessel image to identify the first target connected component containing the most voxels. The first target connected component represents the aortic vessel. The BinaryDilateImageFilter function in the ITK image processing tool is then used to dilate the binary image corresponding to the first target connected component to restore it from the eroded state to its pre-erosion state, thus obtaining a second aortic vessel image. It should be explained that a connected component represents a 3D image region composed of foreground voxels with the same voxel value (or satisfying a specific similarity criterion) and adjacent positions. Connected component analysis identifies all connected components.

[0117] Optionally, the enhanced tubular image includes a first enhanced tubular image and a second enhanced tubular image, and the step of performing vascular completion processing on the second aortic vascular image to obtain the aortic vascular image includes:

[0118] The target tubular enhanced image is segmented using binary thresholding based on a preset second tubular object metric range to obtain a third branch vessel enhanced image. This target tubular enhanced image is either the first or second tubular enhanced image, where the lower limit of the second tubular object metric range is greater than the lower limit of the first tubular object metric range. Image registration is performed on the aortic trunk vessel image and the third branch vessel enhanced image. The union of the registered aortic trunk vessel image and the third branch vessel enhanced image is then determined as the third aortic vessel image. Connectivity component analysis is performed on the third aortic vessel image to identify the second target connected component with the largest number of voxels. The union of the binary image corresponding to the second target connected component and the second aortic vessel image is then determined as the aortic vessel image.

[0119] Since the Hessian image is processed using the Hessian3D vessel metric function, the metric values ​​of non-tubular objects in the generated tubular enhanced image are generally less than 2, while the metric values ​​of tubular objects are generally no more than 500. Therefore, the tubular enhanced image can be processed by binary threshold segmentation based on the first tubular object metric value interval [2, 500] to obtain the aortic branch vessel enhanced image.

[0120] The inventors of this disclosure have discovered that increasing the lower limit of the measurement range of the first tubular object can also achieve tubular separation. However, increasing the lower limit of the measurement range of the first tubular object will cause partial loss at the junction of aortic branch vessels, resulting in incomplete connections (i.e., gaps resembling serrations) or discontinuity. However, the branch vessels themselves are unaffected (e.g., the tail end of the branch vessel is unaffected), and the branch vessels remain intact. Therefore, embodiments of this disclosure propose utilizing this characteristic to recover the tail end portion of branch vessels lost due to corrosion. Specific examples are as follows:

[0121] Assuming the first tubular object's metric range is [2, 500], the second tubular object's metric range can be set to [5, 500]. Binary thresholding is performed on the enhanced image of the target tubular object based on the second tubular object's metric range [5, 500] to obtain the enhanced image of the third branch vessel. Image registration is performed on the aortic trunk vessel image and the enhanced image of the third branch vessel. The union of the registered aortic trunk vessel image and the enhanced image of the third branch vessel is determined as the third aortic vessel image. Connectivity component analysis is performed on the third aortic vessel image to determine the second target connected component with the largest number of voxels. The second aortic vessel image is completed based on the binary image corresponding to the second target connected component; for example, the union of the binary image corresponding to the second target connected component and the second aortic vessel image is determined as the aortic vessel image.

[0122] Using the method described above, an aortic CTA image is acquired, and the grayscale image of the aortic CTA image is as follows: Figure 2 As shown. Inputting this CTA image into the trained aortic trunk vessel extraction model yields an aortic trunk vessel image, the corresponding grayscale image of which is shown below. Figure 3 As shown. Image enhancement processing is performed on this CTA image to obtain enhanced images of the aortic branch vessels. Based on the aortic trunk vessel image and the enhanced aortic branch vessel image, the aortic vessel image is segmented from the CTA image. The grayscale image of this aortic vessel image is shown below. Figure 4 As shown. In other words, the aortic vessel segmentation method proposed in this disclosure integrates deep learning technology with traditional image processing techniques. Image vessel enhancement processing can highlight tubular structures, achieving effective vessel extraction, which is very effective for extracting aortic branch vessels. However, due to the excessive thickness of the aortic trunk and interference from the dissecting region, the extraction effect of the aortic trunk using image vessel enhancement processing is not ideal. Using deep learning technology, through model training, better extraction of the aortic trunk can be achieved. This invention combines the advantages of both methods, achieving effective extraction of the complete aortic trunk and branch vessels.

[0123] In some implementations, since the spatial spacing between voxels in a CTA image is not necessarily the same, and this discrepancy can affect the accuracy of the output of the aortic trunk vessel extraction model, in this embodiment, voxel spacing preprocessing can be performed on the CTA image before inputting it into the trained aortic trunk vessel extraction model. For example, resampling can be performed on the CTA image before inputting it into the trained aortic trunk vessel extraction model to ensure that the spatial spacing between voxels in the CTA image is consistent. Resampling can be implemented by determining the mean spatial spacing ΔX of each voxel along the X-axis in the CTA image, and then performing image transformation using a cubic spline interpolation algorithm or other interpolation algorithms in the related art, so that the spatial spacing between each voxel along the X-axis in the transformed image is ΔX.

[0124] Similarly, the mean spatial interval ΔY of each voxel along the Y-axis and the mean spatial interval ΔZ of each voxel along the Z-axis in the CTA image can be determined. Image transformation is performed using cubic spline interpolation or other interpolation algorithms to ensure that the spatial interval between voxels along the Y-axis and the spatial interval between voxels along the Z-axis in the transformed image are both ΔY and ΔZ.

[0125] On the other hand, due to the large number of layers in CTA images and the high resolution of each layer, the file size of CTA images is large. A larger CTA image file size increases the workload of the aortic trunk vessel extraction model and reduces its computational efficiency. To reduce the workload of the aortic trunk vessel extraction model and improve its computational efficiency, this disclosure proposes pre-segmenting the CTA image before inputting it into the trained aortic trunk vessel extraction model. For example, the region of interest (ROI) image containing the aortic vessels can be identified and segmented from the CTA image. Then, the ROI image is input into the trained aortic trunk vessel extraction model to obtain the aortic trunk vessel image.

[0126] One implementation method for identifying and segmenting the region of interest (ROI) containing the aortic vessels from a CTA image may involve extracting a skeleton image from the CTA image and determining the central axis of the skeleton image. A cube image is then cut from the CTA image, centered on this central axis and with preset X-axis and Y-axis lengths. This cube image is then identified as the ROI containing the 3D aortic vessels.

[0127] In detail, given the physiological characteristics of the aorta being contained within the sternal skeleton and the high density of the skeleton image, which makes it easy to extract, this disclosure proposes that the image region where the skeleton is located can be determined from the CTA image first, and then the complete 3D vascular image of the aorta can be further extracted from the image region where the skeleton is located.

[0128] For example, a skeleton image is extracted from a CTA image, and the central axis of the skeleton image is determined. A cube image centered on this central axis, with preset X-axis and Y-axis lengths, is cut out from the CTA image. A 3D aortic vessel image is extracted from the cube image. Here, the central axis is the Z-axis. The preset X-axis and Y-axis lengths are empirical values. Specifically, image processing tools from related technologies can be used to analyze the shape characteristics of the skeleton image and determine the initial vertex coordinates (X0, Y0, Z0) of the circumscribed cube of the skeleton. Based on the preset X-axis length a and preset Y-axis length b, the center point coordinates (X0, Y0, Z0) of each layer are calculated. c Y c ), where X c =X0+a / 2, Y c =Y0+b / 2. Next, with the CTA image pixel size at 512*512 and the number of layers at 512, according to the region [X... c -128, X c +128]、[Y c -128, Y c[+128] and [0, 512] are used to cut out cube images from CTA images. In this way, a high-resolution CTA image (512*512*512) can be cut into cube images containing the aorta with a resolution of (256*256*512). This not only allows for the extraction of a complete 3D aortic image from the lower-resolution cube image but also reduces computational load.

[0129] Alternatively, the skeleton image can be extracted from the CTA image in the following ways:

[0130] Image thresholding is performed on the CTA image based on the bone grayscale threshold to obtain the corresponding bone region image. The grayscale value of each voxel in the bone region image is greater than the bone grayscale threshold. Connected component analysis is performed on the binary image of the bone region image to obtain at least one connected component of the bone region image. The connected component with the most voxels in the at least one connected component of the bone region image is taken as the bone component, and the image corresponding to the bone component is taken as the skeleton image.

[0131] The bone grayscale threshold can be set to 200 HU. Image thresholding is performed on the CTA image based on the bone grayscale threshold of 200 to obtain the corresponding bone region image. The grayscale value of each voxel in this bone region image is greater than (or equal to) the bone grayscale threshold of 200. The bone grayscale threshold of 200 does not need to be an exact bone extraction value because this embodiment is only for determining the bone region image where the bone is located; it does not require that every voxel in the bone region image represent a bone. The bone region image may include other voxels representing non-bones structures.

[0132] After obtaining the image of the skeletal region containing the skeleton, connected component analysis can be performed on the binary image of the skeletal region to obtain at least one connected component of the skeletal region image. The connected component with the largest number of voxels among the at least one connected component of the skeletal region image is taken as the skeletal component, and the image corresponding to the skeletal component is taken as the skeleton image.

[0133] Another possible scenario is that even if the region of interest (ROI) containing the aorta is identified and segmented from the CTA image, the ROI image may still be quite large. The larger the ROI image, the more difficult it becomes for the GPU memory and computing power of the electronic device supporting the aortic trunk extraction model to meet the demands. In this case, it is impossible to further segment the ROI image into smaller segments. Therefore, in some implementations, the ROI image (or the unprocessed CTA image) input to the aortic trunk extraction model can be patched to obtain multiple patches (which can be simply understood as image blocks). Then, the ROI image (or the unprocessed CTA image) is input into the aortic trunk extraction model in patch units.

[0134] The principle of patch processing can be found in related technologies. This disclosure only describes the patch size used in this disclosure. In the embodiments of this disclosure, the patch size can be a multiple of 2 to the power of N, where N is the number of downsampling network layers in the aortic trunk vascular extraction model, and 2 is the size of the pooling kernel. After the image patch is processed by N downsampling network layers, the size of the resulting feature map needs to be about 4 to 10. For example, the image patch size that can be used in this disclosure is (288, 97, 80), and the number of its downsampling network layers is [5, 4, 4] (the number of downsampling network layers in each axis can be different), and the corresponding downsampling kernels are [2, 2, 2], [2, 2, 2], [2, 2, 2], [2, 2, 2], [2, 1, 1].

[0135] In some implementations, the CTA images may be enhanced before being input into the trained aortic trunk vessel extraction model.

[0136] For example, CTA images are randomly segmented by patch size, and image enhancement processing is performed on each patch. The image enhancement processing includes at least one of the following enhancement processes:

[0137] Elastic Transform Processing: The value range of the elastic transform processing parameter alpha can be set to [0, 900], the value range of the parameter sigma can be set to [9, 13], and the application probability can be set to 0.2. The application probability refers to randomly generating a value between 0 and 1. If the generated value is less than or equal to 0.2, then elastic transform processing is used to enhance the patch image. If the generated value is greater than 0.2, then the step of using elastic transform processing to enhance the patch image is skipped.

[0138] Scaling transformation processing: The scaling ratio can be set to [0.65, 1.6], and the application probability can be set to 0.3;

[0139] Random rotation processing: The rotation amplitude can be set to [-15, 15] (in degrees), and the application probability can be set to 0.2;

[0140] Gamma transformation processing: The transformation range can be set to [0.7, 0.5], and the application probability can be set to 0.2;

[0141] Contrast enhancement processing: Saturation factor can be set to [0.1, 1.9], and application probability can be set to 0.2;

[0142] Cutout enhancement processing: can be set to fill rectangle, whose width and height ratio coefficient to the image width and height can be set to [0, 0.4], the fill pixel value can be set to the image average value, and the application probability can be set to 0.2.

[0143] The meanings of the parameters in the above image enhancement processing methods can be found in related technologies, and this disclosure will not provide a detailed explanation of them.

[0144] Figure 5 This is a block diagram illustrating an aortic vessel segmentation device according to an exemplary embodiment of the present disclosure, such as... Figure 5 As shown, the aortic vessel segmentation device 500 includes:

[0145] Acquisition module 510 is used to acquire aortic CTA images;

[0146] Input module 520 is used to input the CTA image into the trained aortic trunk vessel extraction model to obtain an aortic trunk vessel image, wherein the aortic trunk vessel image is a binary image.

[0147] Execution module 530 is used to perform image enhancement processing on the CTA image to obtain an enhanced image of the aortic branch vessels, wherein the enhanced image of the aortic branch vessels is a binary image;

[0148] The segmentation module 540 is used to segment the aortic vessel image from the CTA image based on the aortic trunk vessel image and the enhanced aortic branch vessel image.

[0149] Using the aforementioned device 500, an aortic CTA image is acquired. This CTA image is then input into a trained aortic trunk vessel extraction model to obtain an aortic trunk vessel image. Image enhancement processing is performed on the CTA image to obtain enhanced images of the aortic branch vessels. Based on these two binary images—the aortic trunk vessel image and the enhanced aortic branch vessel image—the aortic vessel image can be segmented from the CTA image. This method, utilizing a deep learning-based aortic trunk vessel extraction model, effectively extracts the aortic trunk vessel, while image enhancement processing of the CTA image highlights the aortic branch vessels, making their segmentation easier. Therefore, this method effectively extracts the complete aortic trunk vessel and its branch vessels.

[0150] Optionally, the execution module 530 includes:

[0151] The first determining submodule is used to determine the Hessian image corresponding to the CTA image based on the Hessian recursive Gaussian function;

[0152] The second determining submodule is used to process the Hessian image according to the Hessian3D blood vessel measurement function to obtain an enhanced image of the tubular structure;

[0153] The first segmentation submodule is used to perform binary threshold segmentation processing on the enhanced image of the tubular object according to a preset first tubular object metric value range to obtain the enhanced image of the aortic branch vessels.

[0154] Optionally, the first determining submodule includes:

[0155] The first processing submodule is used to process the CTA image according to a Hessian recursive Gaussian function with a sigma value of a first preset value to obtain a first Hessian image.

[0156] The second processing submodule is used to process the CTA image according to a Hessian recursive Gaussian function with a sigma value of a second preset value to obtain a second Hessian image; the Hessian image includes the first Hessian image and the second Hessian image.

[0157] Optionally, the Hessian image includes a first Hessian image and a second Hessian image; correspondingly, the tubular enhanced image includes a first tubular enhanced image and a second tubular enhanced image; the first segmentation submodule includes:

[0158] The second segmentation submodule is used to perform binary threshold segmentation processing on the enhanced image of the first tubular object based on the measurement value range of the first tubular object to obtain the enhanced image of the first branch blood vessel.

[0159] The third segmentation submodule is used to perform binary threshold segmentation processing on the second tubular enhanced image based on the first tubular object metric range to obtain the second branch vessel enhanced image; wherein, the aortic branch vessel enhanced image includes the first branch vessel enhanced image and the second branch vessel enhanced image, the vessel thickness in the first branch vessel enhanced image is positively correlated with the sigma value used when determining the first Hessian image, and correspondingly, the vessel thickness in the second branch vessel enhanced image is positively correlated with the sigma value used when determining the second Hessian image.

[0160] Optionally, the enhanced image of the aortic branch vessels includes an enhanced image of the first branch vessels and an enhanced image of the second branch vessels, and the segmentation module 540 includes:

[0161] The first registration submodule is used to perform image registration processing on the aortic trunk image, the enhanced image of the first branch vessel, and the enhanced image of the second branch vessel.

[0162] The first union operation submodule is used to determine the first aortic vessel image by the union of the registered aortic trunk vessel image, the first branch vessel enhanced image, and the second branch vessel enhanced image.

[0163] The noise reduction submodule is used to perform noise reduction processing on the first aortic vessel image to obtain the second aortic vessel image.

[0164] The completion submodule is used to perform vascular completion processing on the second aortic vascular image to obtain the aortic vascular image.

[0165] Optionally, the noise reduction submodule is used for:

[0166] The first aortic vessel image is subjected to erosion processing to obtain an eroded first aortic vessel image; connected component analysis is performed on the eroded first aortic vessel image to determine the first target connected component with the largest number of voxels; the binary image corresponding to the first target connected component is subjected to dilation processing to obtain the second aortic vessel image.

[0167] Optionally, the enhanced tubular image includes a first enhanced tubular image and a second enhanced tubular image, and the completion submodule is used for:

[0168] The target tubular enhanced image is subjected to binary threshold segmentation based on a preset second tubular object measurement value range to obtain a third branch vessel enhanced image. The target tubular enhanced image is one of the first tubular enhanced image and the second tubular enhanced image. The lower limit of the second tubular object measurement value range is greater than the lower limit of the first tubular object measurement value range.

[0169] Image registration processing is performed on the aortic trunk vessel image and the enhanced image of the third branch vessel, and the union of the registered aortic trunk vessel image and the enhanced image of the third branch vessel is determined as the third aortic vessel image;

[0170] Connectivity component analysis was performed on the third aortic vascular image to identify the second target connectivity component that includes the largest number of voxels;

[0171] The union of the binary image corresponding to the second target connected component and the second aortic vascular image is determined as the aortic vascular image.

[0172] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0173] Figure 6 This is a block diagram illustrating an electronic device 700 according to an exemplary embodiment. Figure 6 As shown, the electronic device 700 may include a processor 701 and a memory 702. The electronic device 700 may also include one or more of a multimedia component 703, an input / output (I / O) interface 704, and a communication component 705.

[0174] The processor 701 controls the overall operation of the electronic device 700 to complete all or part of the steps in the aortic vessel segmentation method described above. The memory 702 stores various types of data to support the operation of the electronic device 700. This data may include, for example, instructions for any application or method operating on the electronic device 700, and application-related data such as contact data, sent and received messages, pictures, audio, video, etc. The memory 702 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. Multimedia component 703 may include a screen and an audio component. The screen may be, for example, a touchscreen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signals may be further stored in memory 702 or transmitted via communication component 705. The audio component also includes at least one speaker for outputting audio signals. I / O interface 704 provides an interface between processor 701 and other interface modules, such as a keyboard, mouse, buttons, etc. These buttons may be virtual or physical buttons. Communication component 705 is used for wired or wireless communication between the electronic device 700 and other devices. Wireless communication, such as Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, 4G, NB-IoT, eMTC, or other 5G technologies, or combinations thereof, is not limited here. Therefore, the corresponding communication component 705 may include: a Wi-Fi module, a Bluetooth module, an NFC module, etc.

[0175] In an exemplary embodiment, the electronic device 700 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the aortic vessel segmentation method described above.

[0176] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided, which, when executed by a processor, implement the steps of the aortic vessel segmentation method described above. For example, the computer-readable storage medium may be the memory 702 including program instructions, which may be executed by the processor 701 of the electronic device 700 to complete the aortic vessel segmentation method described above.

[0177] In another exemplary embodiment, a computer program product is also provided, comprising a computer program executable by a programmable device, the computer program having a code portion for performing the above-described aortic vessel segmentation method when executed by the programmable device.

[0178] The preferred embodiments of this disclosure have been described in detail above with reference to the accompanying drawings. However, this disclosure is not limited to the specific details of the above embodiments. Within the scope of the technical concept of this disclosure, various simple modifications can be made to the technical solutions of this disclosure, and these simple modifications all fall within the protection scope of this disclosure.

[0179] It should also be noted that the various specific technical features described in the above embodiments can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, this disclosure will not describe the various possible combinations separately.

[0180] Furthermore, various different embodiments of this disclosure can be combined in any way, as long as they do not violate the spirit of this disclosure, they should also be regarded as the content disclosed in this disclosure.

Claims

1. An aorta blood vessel segmentation method, characterized by, The method includes: Acquire aortic CTA images; The CTA image is input into the trained aortic trunk vessel extraction model to obtain an aortic trunk vessel image, which is a binary image. The CTA image is subjected to image enhancement processing to obtain an enhanced image of the aortic branch vessels, which is a binary image; Based on the enhanced images of the main aortic vessels and the aortic branch vessels, the aortic vessel image is segmented from the CTA image; The enhanced images of aortic branch vessels include enhanced images of a first branch vessel and a second branch vessel. The step of segmenting the aortic vessel image from the CTA image based on the aortic trunk vessel image and the enhanced images of aortic branch vessels includes: Image registration processing is performed on the main aortic trunk image, the enhanced image of the first branch vessel, and the enhanced image of the second branch vessel. The union of the registered aortic trunk image, the enhanced image of the first branch vessel, and the enhanced image of the second branch vessel is determined as the first aortic vessel image; The first aortic vessel image is denoised to obtain the second aortic vessel image. The second aortic vessel image is subjected to vessel completion processing to obtain the aortic vessel image; The step of performing vascular completion processing on the second aortic vascular image to obtain the aortic vascular image includes: The target tubular enhanced image is subjected to binary threshold segmentation processing according to the preset second tubular object measurement value interval to obtain the third branch blood vessel enhanced image. The target tubular enhanced image is one of the first tubular enhanced image and the second tubular enhanced image. The lower limit of the second tubular object measurement value interval is greater than the lower limit of the first tubular object measurement value interval. Image registration processing is performed on the aortic trunk vessel image and the enhanced image of the third branch vessel, and the union of the registered aortic trunk vessel image and the enhanced image of the third branch vessel is determined as the third aortic vessel image; Connectivity component analysis was performed on the third aortic vascular image to identify the second target connectivity component that includes the largest number of voxels; The union of the binary image corresponding to the second target connected component and the second aortic vascular image is determined as the aortic vascular image.

2. The method of claim 1, wherein, The step of performing image enhancement processing on the CTA image to obtain enhanced images of aortic branch vessels includes: The Hessian image corresponding to the CTA image is determined based on the Hessian recursive Gaussian function. The Hessian image is processed using the Hessian 3D vessel measurement function to obtain an enhanced image of the tubular structure. The enhanced image of the tubular object is subjected to binary threshold segmentation based on a preset first tubular object metric range to obtain the enhanced image of the aortic branch vessels.

3. The method of claim 2, wherein, The process of determining the Hessian image corresponding to the CTA image based on the Hessian recursive Gaussian function includes: The CTA image is processed using a Hessian recursive Gaussian function with a sigma value of a first preset value to obtain a first Hessian image; The CTA image is processed using a Hessian recursive Gaussian function with a sigma value of a second preset value to obtain a second Hessian image; The Hessian image includes the first Hessian image and the second Hessian image.

4. The method of claim 2, wherein, The Hessian image includes a first Hessian image and a second Hessian image. Correspondingly, the tubular enhanced image includes a first tubular enhanced image and a second tubular enhanced image. The step of performing binary threshold segmentation processing on the tubular enhanced image according to a preset first tubular object metric range to obtain the aortic branch vessel enhanced image includes: The enhanced image of the first tubular object is subjected to binary threshold segmentation based on the measurement range of the first tubular object to obtain the enhanced image of the first branch blood vessel. The enhanced image of the second tubular object is subjected to binary threshold segmentation based on the measurement range of the first tubular object to obtain the enhanced image of the second branch blood vessel. The enhanced images of aortic branch vessels include a first enhanced image of a branch vessel and a second enhanced image of a branch vessel. The vessel thickness in the first enhanced image of a branch vessel is positively correlated with the sigma value used when determining the first Hessian image. Correspondingly, the vessel thickness in the second enhanced image of a branch vessel is positively correlated with the sigma value used when determining the second Hessian image.

5. The method of claim 1, wherein, The step of denoising the first aortic vascular image to obtain the second aortic vascular image includes: The first aortic vessel image is subjected to erosion processing to obtain an eroded first aortic vessel image; Connectivity component analysis was performed on the corroded first aortic vessel image to determine the first target connectivity component with the largest number of voxels. The binary image corresponding to the first target connected component is dilated to obtain the second aortic blood vessel image.

6. An aortic vessel segmentation device, characterized in that, The device includes: The acquisition module is used to acquire aortic CTA images; The input module is used to input the CTA image into the trained aortic trunk vessel extraction model to obtain an aortic trunk vessel image, wherein the aortic trunk vessel image is a binary image. The execution module is used to perform image enhancement processing on the CTA image to obtain an enhanced image of the aortic branch vessels, wherein the enhanced image of the aortic branch vessels is a binary image; The segmentation module is used to segment the aortic vessel image from the CTA image based on the aortic trunk vessel image and the enhanced aortic branch vessel image; The enhanced images of the aortic branch vessels include enhanced images of the first branch vessels and enhanced images of the second branch vessels. The segmentation module includes: The first registration submodule is used to perform image registration processing on the aortic trunk image, the enhanced image of the first branch vessel, and the enhanced image of the second branch vessel. The first union operation submodule is used to determine the first aortic vessel image by the union of the registered aortic trunk vessel image, the first branch vessel enhanced image, and the second branch vessel enhanced image. The noise reduction submodule is used to perform noise reduction processing on the first aortic vessel image to obtain the second aortic vessel image. The completion submodule is used to perform vascular completion processing on the second aortic vascular image to obtain the aortic vascular image; The completion submodule is used for: The target tubular enhanced image is subjected to binary threshold segmentation processing according to the preset second tubular object measurement value interval to obtain the third branch blood vessel enhanced image. The target tubular enhanced image is one of the first tubular enhanced image and the second tubular enhanced image. The lower limit of the second tubular object measurement value interval is greater than the lower limit of the first tubular object measurement value interval. Image registration processing is performed on the aortic trunk vessel image and the enhanced image of the third branch vessel, and the union of the registered aortic trunk vessel image and the enhanced image of the third branch vessel is determined as the third aortic vessel image; Connectivity component analysis was performed on the third aortic vascular image to identify the second target connectivity component that includes the largest number of voxels; The union of the binary image corresponding to the second target connected component and the second aortic vascular image is determined as the aortic vascular image.

7. A non-transitory computer-readable storage medium having stored thereon a computer program, characterized in that, When executed by a processor, the program implements the steps of the method described in any one of claims 1-5.

8. An electronic device, comprising: include: A memory on which computer programs are stored; A processor for executing the computer program in the memory to implement the steps of the method according to any one of claims 1-5.

Images