A method for detecting vascular protrusions and related devices

Abstract

The present application discloses a method for detecting vascular protrusions and related devices. The method includes: after obtaining a CTA image collected for a target body part, first using the CTA image to construct three-dimensional image data of the target body part, and extracting spherical distribution characterization data from the CTA image, so that the spherical distribution characterization data can as accurately as possible represent the distribution state of spherical-like objects in the target body part; then using a pre-constructed vascular protrusion detection model to determine the vascular protrusion detection result of the target body part from the three-dimensional image data and the spherical distribution characterization data, so that the vascular protrusion detection result can represent the distribution of protrusions on the blood vessel wall in the target body part, so that doctors can refer to the vascular protrusion detection result and other information for diagnosis and treatment related processing in the future, thereby effectively improving the recognition effect of protrusions on the blood vessel wall.

Description

Technical Field

[0001] This application relates to the technical field of medical data processing, and particularly relates to a method for detecting vascular bulges and related devices thereof. Background Art

[0002] Blood vessels refer to a series of tubes through which blood flows; and except for places such as the cornea, hair, fingernails (toenails), dentin, and epithelium, blood vessels are distributed throughout the human body. It can be seen that vascular health is very important.

[0003] In fact, under the influence of some adverse factors (such as trauma, infection, atherosclerosis, bad living habits, etc.), spherical bulges may appear on the blood vessel wall (such as bulges similar to unruptured intracranial aneurysms (UIAs) formed on the cerebral artery wall, etc.), and once such a bulge ruptures, it will cause the blood vessel to be unable to transport blood normally, thus endangering physical health.

[0004] Currently, for the spherical-like bulges that appear on the blood vessel wall, the common bulge inspection methods are as follows: after a doctor obtains a computed tomography angiography (CTA) image, the doctor manually inspects the CTA image based on his rich imaging inspection experience (such as bulge inspection experience, etc.).

[0005] However, due to the defects of the above-mentioned manual inspection method, the recognition effect for the bulges on the blood vessel wall is relatively poor. Summary of the Invention

[0006] In view of this, the embodiments of this application provide a method for detecting vascular bulges and related devices thereof, which can improve the recognition effect for the bulges on the blood vessel wall.

[0007] To solve the above problems, the technical solutions provided by the embodiments of this application are as follows:

[0008] The embodiments of this application provide a method for detecting vascular bulges, and the method includes:

[0009] After obtaining a CTA image of a target body part, use the CTA image to construct three-dimensional image data of the target body part;

[0010] Extract spherical distribution characterization data from the CTA image;

[0011] Determine the vascular bulge detection result of the target body part according to the three-dimensional image data, the spherical distribution characterization data, and a pre-constructed vascular bulge detection model.

[0012] An embodiment of the present application further provides a vascular bulge detection device, and the device includes:

[0013] An image construction unit, configured to, after acquiring a CTA angiography image collected for a target body part, use the CTA image to construct three-dimensional image data of the target body part;

[0014] A spherical extraction unit, configured to extract spherical distribution characterization data from the CTA image;

[0015] A bulge detection unit, configured to determine a vascular bulge detection result of the target body part according to the three-dimensional image data, the spherical distribution characterization data, and a pre-constructed vascular bulge detection model.

[0016] An embodiment of the present application further provides a vascular bulge detection device, including: a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, any implementation manner of the vascular bulge detection method provided in the embodiment of the present application is implemented.

[0017] An embodiment of the present application further provides a computer-readable storage medium, in which instructions are stored. When the instructions run on a terminal device, the terminal device is enabled to execute any implementation manner of the vascular bulge detection method provided in the embodiment of the present application.

[0018] An embodiment of the present application further provides a computer program product, characterized in that when the computer program product runs on a terminal device, the terminal device is enabled to execute any implementation manner of the vascular bulge detection method provided in the embodiment of the present application.

[0019] Thus, the embodiment of the present application has the following beneficial effects:

[0020] In the technical solution provided by the embodiments of the present application, after obtaining the CTA image collected for the target body part (for example, the brain), the three-dimensional image data of the target body part is first constructed by using the CTA image, so that the three-dimensional image data can better represent the target body part in a three-dimensional manner, and the spherical distribution characterization data is extracted from the CTA image, so that the spherical distribution characterization data can as accurately as possible represent the distribution state of the spherical objects in the target body part; then, by using the pre-constructed blood vessel bulge detection model, the blood vessel bulge detection result of the target body part is determined from the three-dimensional image data and the spherical distribution characterization data, so that the blood vessel bulge detection result can represent the distribution of the bulges on the blood vessel wall in the target body part, so that in the future, doctors can refer to the blood vessel bulge detection result and other information (such as specific clinical manifestations) for diagnosis and treatment-related processing (for example, physical health status assessment or disease diagnosis, etc.). In this way, the defects existing in the above artificial inspection method can be effectively overcome, and thus the recognition effect of the bulges on the blood vessel wall can be effectively improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a flowchart of a blood vessel bulge detection method provided by an embodiment of the present application;

[0022] Figure 2 It is a schematic diagram of the blood vessel bulge detection result of a target body part provided by an embodiment of the present application;

[0023] Figure 3 It is a schematic diagram of the construction of three-dimensional image data provided by an embodiment of the present application;

[0024] Figure 4 It is a schematic diagram of the structure of a blood vessel bulge detection model provided by an embodiment of the present application;

[0025] Figure 5 It is a schematic diagram of the working principle of a blood vessel bulge detection model provided by an embodiment of the present application;

[0026] Figure 6 It is a schematic diagram of the structure of a blood vessel bulge detection device provided by an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0027] To make the above objects, features, and advantages of the present application more obvious and understandable, the embodiments of the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0028] The inventors found in the research on the above artificial inspection method that the artificial inspection method has at least the following defects shown in ①-③:

[0029] ①In this manual inspection method, doctors need to conduct a detailed analysis of each part of the CTA image, resulting in a relatively low recognition efficiency for the protrusions on the blood vessel wall.

[0030] ②In some cases, a doctor needs to inspect a large number of CTA images for protrusions in a short period of time, which easily causes the doctor's eyes to become fatigued, leading to an incorrect protrusion inspection result given by the doctor, and further resulting in a relatively low recognition accuracy for the protrusions on the blood vessel wall.

[0031] ③This manual inspection method usually requires rich imaging inspection experience, resulting in relatively high requirements for doctors' imaging inspection experience in this manual inspection method. Consequently, there are relatively few doctors in each hospital who can inspect CTA images for protrusions, and further resulting in relatively poor recognition effects for the protrusions on the blood vessel wall in this hospital (for example, low recognition efficiency and low recognition accuracy, etc.).

[0032] Based on the above findings, to solve the technical problems shown in the background art section, the embodiments of the present application provide a method for detecting blood vessel protrusions. The method includes: after obtaining the CTA image collected for a target body part (such as the brain, etc.), first using the CTA image to construct three-dimensional image data of the target body part, so that the three-dimensional image data can better represent the target body part in a three-dimensional stereoscopic manner, and extracting spherical distribution characterization data from the CTA image, so that the spherical distribution characterization data can as accurately as possible represent the distribution state of the spherical-like objects in the target body part; then using a pre-constructed blood vessel protrusion detection model to determine the blood vessel protrusion detection result of the target body part from the three-dimensional image data and the spherical distribution characterization data, so that the blood vessel protrusion detection result can represent the distribution of the protrusions on the blood vessel wall in the target body part, so that doctors can refer to the blood vessel protrusion detection result and other information (such as specific clinical manifestations) for diagnosis and treatment-related processing in the future (such as physical health status assessment or disease diagnosis, etc.). In this way, the defects existing in the above manual inspection method can be effectively overcome, and further the recognition effect for the protrusions on the blood vessel wall can be effectively improved.

[0033] It can be understood that the vascular bulge detection method provided in the embodiments of the present application is a method that can perform object detection on the CTA image of the target body part to obtain the object detection result of the CTA image (that is, the detection result of the spherical-like object on the blood vessel wall in the target body part). It can be seen from this that the vascular bulge detection result determined based on the vascular bulge detection method provided in the embodiments of the present application is only an image processing result (especially an image processing result that can prominently display the spherical-like object on the blood vessel wall in the target body part), and is not a diagnosis result, so that doctors can refer to the vascular bulge detection result and other information (such as specific clinical manifestations) for medical treatment-related processing in the future (for example, physical health status assessment or disease diagnosis, etc.).

[0034] In addition, the embodiments of the present application do not limit the execution entity of the vascular bulge detection method. For example, the vascular bulge detection method provided in the embodiments of the present application can be applied to data processing devices such as terminal devices or servers. Among them, the terminal device can be a smart phone, a computer, a personal digital assistant (Personal Digital Assitant, PDA), or a tablet computer, etc. The server can be an independent server, a cluster server, or a cloud server.

[0035] To facilitate the understanding of the present application, the vascular bulge detection method provided in the embodiments of the present application will be described below with reference to the accompanying drawings.

[0036] See Figure 1 , which is a flowchart of a vascular bulge detection method provided in the embodiments of the present application. The vascular bulge detection method may include S1 - S3:

[0037] S1: After obtaining the CTA image collected for the target body part, use the CTA image to construct the three-dimensional image data of the target body part.

[0038] Among them, the target body part refers to the body part that needs to be subjected to vascular bulge detection processing; and the embodiments of the present application do not limit the target body part. For example, it can be any body part (such as the brain or the abdomen) of a certain physical examination object (such as a person or an animal, etc.).

[0039] The CTA image refers to the image data obtained by performing computed tomography (CT) on the complete target body part. For example, when the target body part is the brain, the CTA image may include the image obtained by performing computed tomography on the complete brain.

[0040] In addition, the embodiments of the present application do not limit the CTA images. For example, it may include all the slice images obtained by scanning a target body part, so that these slice images can completely represent the state of the target body part.

[0041] Furthermore, the embodiments of the present application do not limit the vascular information carried by the CTA images. For example, in some medical application scenarios, the vascular information carried by the CTA images can clearly represent the vascular morphology of the arteries in the target body part. Also, in some medical application scenarios, the vascular information carried by the CTA images can clearly represent the vascular morphology of the veins in the target body part. Additionally, in some medical application scenarios, the vascular information carried by the CTA images can clearly represent the vascular morphology of both the arteries and veins in the target body part.

[0042] Moreover, the embodiments of the present application do not limit the representation method of the CTA images. For example, it can be implemented in the format of *.dicom. That is, the CTA images can be volume images represented in the dicom format.

[0043] The above-mentioned "three-dimensional image data" is used to represent the target body part in a three-dimensional stereoscopic mode; and the embodiments of the present application do not limit the construction process of this "three-dimensional image data". For example, any existing or future method capable of constructing a three-dimensional stereoscopic image based on CTA images can be used for implementation. Also, in order to further improve the description effect of the above "three-dimensional image data" for the target body part, any of the following implementation methods for constructing three-dimensional image data can be used for implementation.

[0044] Based on the relevant content of S1 above, for the vascular bulge detection device provided by the embodiments of the present application (that is, the electronic device for executing the vascular bulge detection method provided by the embodiments of the present application), after the vascular bulge detection device obtains the CTA images collected from the target body part of a certain physical examination object, the vascular bulge detection device uses the CTA images to construct the three-dimensional image data of the target body part, so that the three-dimensional image data can better represent the target body part in a three-dimensional stereoscopic manner, so as to subsequently analyze whether there are spherical-like bulges on the blood vessel walls distributed in the target body part of the physical examination object, the positions where these spherical-like bulges are located, and the shapes presented by these spherical-like bulges from the three-dimensional image data.

[0045] S2: Extract spherical distribution characterization data from the CTA images.

[0046] Among them, the spherical distribution characterization data is used to represent the distribution state of spherical-like objects (such as spherical-like bulges) in the target body part.

[0047] In addition, the embodiments of the present application do not limit the extraction process of the spherical distribution characterization data. For example, any existing or future method capable of identifying and processing the spherical-like objects in the CTA image can be used for implementation. In addition, in order to further improve the recognition effect of the spherical-like objects, any implementation manner of S2 shown below, or any implementation manner of extracting the spherical distribution characterization data shown below can also be used for implementation.

[0048] Based on the relevant content of S2 above, for the vascular bulge detection device provided by the embodiments of the present application, after the vascular bulge detection device obtains the CTA image collected for the target body part, the vascular bulge detection device can perform spherical-like object recognition processing on the CTA image to obtain spherical distribution characterization data, so that the spherical distribution characterization data can represent the distribution state of the spherical-like objects in the target body part, so as to be able to refer to the spherical distribution characterization data subsequently to determine the spherical-like objects (such as spherical-like bulges, etc.) distributed on the blood vessel wall in the target body part.

[0049] It should be noted that the embodiments of the present application do not limit the execution order between S2 and the step "using the CTA image to construct the three-dimensional image data of the target body part" in S1 above.

[0050] S3: Determine the vascular bulge detection result of the target body part according to the three-dimensional image data of the target body part, the spherical distribution characterization data, and the pre-constructed vascular bulge detection model.

[0051] Among them, the vascular bulge detection model is used to perform bulge recognition processing on the blood vessel wall for the input data of the vascular bulge detection model.

[0052] In addition, the embodiments of the present application do not limit the vascular bulge detection model. For example, the vascular bulge detection model can be a physical device with the function of bulge recognition processing on the blood vessel wall. Another example is that the vascular bulge detection model can be a machine learning model with the function of bulge recognition processing on the blood vessel wall; and this machine learning model can be trained using a large number of three-dimensional image samples collected for the target body part, the spherical distribution characterization data extracted from each three-dimensional image sample, and the actual blood vessel bulge description information of each three-dimensional image sample.

[0053] It should be noted that the above "actual blood vessel bulge description information" is used to describe the spherical bulges distributed on the blood vessel wall in the target body part presented by a three-dimensional image sample; and this "actual blood vessel bulge description information" can be used as label information to guide the training process of the machine learning model. In addition, the embodiments of the present application do not limit the training process of the machine learning model. Moreover, the embodiments of the present application do not limit the collection method of the three-dimensional image sample. For example, specifically, it can be: calling a large number of stored CTA images representing the target body part from the hospital database as a large number of three-dimensional image samples.

[0054] The above "detection result of blood vessel bulge in the target body part" is used to represent the distribution state of the spherical bulges on the blood vessel wall in the target body part, so that doctors can refer to the spherical bulges on the blood vessel wall represented by the "detection result of blood vessel bulge in the target body part" and other information (such as specific clinical manifestations) for diagnosis and treatment-related processing (for example, determining whether UIAs appear on the blood vessel wall and the morphology of UIAs, etc.).

[0055] In addition, the embodiments of the present application do not limit the representation method of the above "detection result of blood vessel bulge in the target body part". For example, it can be represented in the Figure 2 way shown. It should be noted that for Figure 2 , the spherical object circled by the circle is the predicted spherical bulge that appears on the blood vessel wall in the target body part.

[0056] Moreover, the embodiments of the present application do not limit the determination process of the blood vessel bulge detection result. For example, in some application scenarios, the three-dimensional image data and the spherical distribution characterization data can be directly input into a pre-constructed blood vessel bulge detection model, so that the blood vessel bulge detection model performs blood vessel bulge detection processing on the three-dimensional image data and the spherical distribution characterization data, and obtains and outputs the blood vessel bulge detection result of the target body part. Another example is that in some other application scenarios, in order to improve the recognition effect of the bulges on the blood vessel wall, any implementation manner of S3 shown below can be adopted for implementation.

[0057] Based on the relevant content of S1 to S3 above, for the vascular bulge detection method provided in the embodiments of the present application, after obtaining the CTA image collected for the target body part (for example, the brain), first use the CTA image to construct the three-dimensional image data of the target body part, so that the three-dimensional image data can better represent the target body part in a three-dimensional stereoscopic manner, and extract the spherical distribution characterization data from the CTA image, so that the spherical distribution characterization data can as accurately as possible represent the distribution state of the spherical-like objects in the target body part; then use the pre-constructed vascular bulge detection model to determine the vascular bulge detection result of the target body part from the three-dimensional image data and the spherical distribution characterization data, so that the vascular bulge detection result can represent the bulge distribution on the blood vessel wall in the target body part, so that doctors can refer to the vascular bulge detection result and other information (such as specific clinical manifestations) for diagnosis and treatment related processing in the future (for example, physical health status assessment or disease diagnosis, etc.), so as to effectively overcome the defects of the above-mentioned manual inspection method, and thus effectively improve the recognition effect of the bulge on the blood vessel wall.

[0058] In a possible implementation manner, in order to further improve the expression effect of the above-mentioned "three-dimensional image data" on the target body part, the embodiments of the present application also provide a possible implementation manner for constructing the three-dimensional image data, which may specifically include steps 11 - step 12:

[0059] Step 11: Perform isotropic resampling on the CTA image to obtain the image data to be used, so that the pixel sampling parameters (for example, pixel pitch) in each direction of the image data to be used are kept consistent.

[0060] The embodiments of the present application do not limit the implementation manner of step 11. For example, when the above-mentioned "CTA image" is a three-dimensional image, step 11 may specifically be: directly perform isotropic resampling on the CTA image to obtain the image data to be used, so that the pixel sampling parameters in each direction of the image data to be used are kept consistent.

[0061] In fact, in order to adapt to the data storage requirements, CTA images are usually stored in the form of image sequences. Based on this, it can be known that when the above-mentioned "CTA image" is an image sequence, step 11 may specifically include steps 111 - step 112:

[0062] Step 111: Convert the CTA image into an initial three-dimensional stereogram, so that the image information expressed by the initial three-dimensional stereogram is the same as the image information expressed by the CTA image.

[0063] It should be noted that the embodiments of the present application do not limit the implementation manner of step 111. For example, it can be implemented by any existing or future method that can convert an image sequence into a three-dimensional stereoscopic image.

[0064] Step 112: Perform isotropic resampling on the initial three-dimensional stereoscopic image according to preset pixel parameters to obtain the image data to be used, so that the pixel sampling parameters in all directions of the image data to be used conform to the preset pixel parameters.

[0065] Among them, the preset pixel parameters can be set in advance; and the embodiments of the present application do not limit the preset pixel parameters. For example, it can be: the pixel pitch is 0.5 mm.

[0066] It can be seen that when the preset pixel parameter is 0.5 mm, step 112 can specifically be: according to the requirement that the pixel pitch = 0.5 mm, perform linear interpolation on all pixels in the X-axis direction of the initial three-dimensional stereoscopic image, perform linear interpolation on all pixels in the Y-axis direction of the initial three-dimensional stereoscopic image, and perform linear interpolation on all pixels in the Z-axis direction of the initial three-dimensional stereoscopic image to obtain the image data to be used, so that the pixel pitches in all axis directions of the image data to be used are kept consistent, so as to achieve the purpose of isotropic resampling of the initial three-dimensional stereoscopic image.

[0067] Based on the relevant content of step 11 above, for the vascular bulge detection device provided by the embodiments of the present application, after the vascular bulge detection device obtains the CTA image collected for the target body part, the vascular bulge detection device can perform isotropic resampling on the CTA image to obtain the image data to be used, so that the pixel sampling parameters (such as pixel pitch) in all directions of the image data to be used are kept consistent, so as to effectively avoid the adverse effects caused by inconsistent pixel sampling parameters in different directions (for example, for the brain, if the pixel pitches in different axis directions are inconsistent, it will cause the three-dimensional stereoscopic image of the brain to be presented in an elliptical form instead of a spherical form, etc.), thereby effectively improving the three-dimensional stereoscopic description effect of the image data to be used for the target body part.

[0068] Step 12: Determine the three-dimensional image data of the target body part according to the image data to be used.

[0069] The embodiments of the present application do not limit the implementation manner of step 12. For example, it can specifically be: directly determine the image data to be used as the three-dimensional image data of the target body part.

[0070] In addition, in order to exclude the interference of pixels in other regions of the image data to be used as much as possible except for the region where the target body part is located, an exemplary embodiment of the present application further provides a possible implementation of step 12, which may specifically be: extracting the minimum bounding rectangle of the target body part from the image data to be used (as shown in Figure 3 ), as the three-dimensional image data of the target body part; or, cutting off other regions in the image data to be used except for the minimum bounding rectangle of the target body part (that is, the diagonal region shown in Figure 3 ) to obtain the three-dimensional image data of the target body part.

[0071] Based on the relevant content of the above steps 11 to 12, for the vascular bulge detection device provided by the exemplary embodiment of the present application, after the vascular bulge detection device obtains the CTA image collected for the target body part, the vascular bulge detection device may first use the CTA image to construct the image data to be used, so that the pixel sampling parameters (for example, pixel pitch) in each direction of the image data to be used are kept consistent, so that the three-dimensional shape described by the image data to be used for the target body part is closer to the actual three-dimensional shape of the target body part; then extracting the three-dimensional image data of the target body part from the image data to be used, so as to effectively exclude the interference of pixels in other regions of the image data to be used except for the region where the target body part is located, so that the three-dimensional image data can represent the three-dimensional state of the target body part as accurately as possible, which is beneficial to improving the expression effect of the three-dimensional image data on the target body part.

[0072] In a possible implementation, in order to further improve the expression effect of the above "three-dimensional image data" on the target body part, an exemplary embodiment of the present application further provides another possible implementation of constructing the three-dimensional image data, which may specifically include steps 21-23:

[0073] Step 21: Extract the part description data for describing the target body part from the CTA image.

[0074] Among them, the part description data is used to record the CT values in the above "CT image" for describing the target body part; and the exemplary embodiment of the present application does not limit the part description data. For example, it may include the CT values in the CTA image for describing the target body part.

[0075] In addition, the exemplary embodiment of the present application does not limit the extraction process of the part description data. For example, it may specifically be: directly determining the CT values in the CTA image for describing the target body part as the part description data.

[0076] In addition, to further improve the description effect of the body part description data for the target body part, another possible implementation manner of the extraction process of the body part description data (that is, step 21) is provided in the embodiments of the present application, which may specifically include steps 211 to 212:

[0077] Step 211: Construct a three-dimensional image to be processed according to the CTA image and a preset numerical threshold, so that there is no CT value for describing air in the three-dimensional image to be processed.

[0078] Among them, the preset numerical threshold can be set in advance; and the preset numerical threshold can be determined according to the air density. For example, the preset numerical threshold can be -200 Hu, so that the preset numerical threshold can be used to remove air.

[0079] In addition, the embodiments of the present application do not limit the implementation manner of step 211. For example, when the above "CTA image" is a three-dimensional image, step 211 can specifically be: updating the values of each pixel in the CTA image whose CT value does not exceed the preset numerical threshold to a first value (for example, 0).

[0080] In fact, to meet the data storage requirements, the CTA image is usually stored in the form of an image sequence. Based on this, it can be known that when the above "CTA image" is an image sequence, step 211 can specifically include steps 2111 to 2112:

[0081] Step 2111: Convert the CTA image into an initial three-dimensional stereogram, so that the image information expressed by the initial three-dimensional stereogram is the same as the image information expressed by the CTA image.

[0082] It should be noted that for the relevant content of step 2111, please refer to the relevant content of step 111 above.

[0083] Step 2112: Update the values of each pixel in the initial three-dimensional stereogram whose CT value does not exceed the preset numerical threshold to a first value (for example, 0) to obtain a three-dimensional image to be processed.

[0084] In the embodiments of the present application, after obtaining the initial three-dimensional stereogram, the values of each pixel in the initial three-dimensional stereogram whose CT value does not exceed the preset numerical threshold (for example, -200 Hu) can be updated to a first value (for example, 0) to obtain a three-dimensional image to be processed, so that there are no pixels in the three-dimensional image to be processed whose CT value does not exceed the preset numerical threshold, thus achieving the purpose of removing air interference in the initial three-dimensional stereogram by using the preset numerical threshold.

[0085] Based on the relevant content of the above step 211, for the vascular bulge detection device provided in the embodiments of the present application, after the vascular bulge detection device obtains the CTA image collected for the target body part, the vascular bulge detection device can remove the CT value for describing air in the CTA image according to a preset numerical threshold to obtain a three-dimensional image to be processed, so that there is no CT value for describing air in the three-dimensional image to be processed, thereby ensuring that the bulge recognition process on the blood vessel wall will not be interfered by air, which is beneficial to improving the bulge recognition effect on the blood vessel wall.

[0086] Step 212: Determine the part description data for describing the target body part according to the largest connected region in the three-dimensional image to be processed.

[0087] The embodiments of the present application do not limit the implementation manner of step 212. For example, specifically, it may be: directly determine the largest connected region in the three-dimensional image to be processed as the part description data for describing the target body part.

[0088] In fact, for the convenience of data storage and subsequent data use, the above "part description data" can be represented in matrix form. Therefore, in order to better adapt to this matrix form, the embodiments of the present application also provide another possible implementation manner of step 212, which may specifically include steps 2121-2122:

[0089] Step 2121: Determine a data extraction mask according to the largest connected region in the three-dimensional image to be processed.

[0090] Among them, the data extraction mask is used to represent the position of the target body part in the three-dimensional image to be processed (that is, the CTA image).

[0091] In addition, the embodiments of the present application do not limit the determination process of the data extraction mask. For example, specifically, it may be: set each pixel in the largest connected region of the three-dimensional image to be processed to a second value (for example, 1), and set each pixel in other regions of the three-dimensional image to be processed except the largest connected region to a third value (for example, 0) to obtain a data extraction mask, so that the data extraction mask can represent the position of the largest connected region (that is, the target body part) in the CTA image.

[0092] Step 2122: Perform mask processing on the three-dimensional image to be processed by using the data extraction mask to obtain the part description data for describing the target body part

[0093] In the embodiments of the present application, after obtaining the data extraction mask, the data extraction mask can be used to perform mask processing on the three-dimensional image to be processed, so as to obtain part description data for describing the target body part, so that the part description data only records as much as possible the respective CT values for describing the target body part, thus effectively avoiding interference from other CT values (such as air density, etc.) unrelated to the target body part to the subsequent bulge recognition process.

[0094] Based on the relevant content of the above steps 211 to 212, for the vascular bulge detection device provided in the embodiments of the present application, after the vascular bulge detection device obtains the CTA image collected for the target body part, the vascular bulge detection device can first remove air interference from the CTA image according to a preset numerical threshold to obtain a three-dimensional image to be processed; then determine the largest connected region in the three-dimensional image to be processed as the part description data for describing the target body part, so that the part description data only records as much as possible the respective CT values for describing the target body part, thus effectively avoiding interference from other CT values (such as air density, etc.) unrelated to the target body part to the subsequent bulge recognition process.

[0095] Actually, in some cases (for example, when detecting bulges on the blood vessel wall of the brain), not only can air interference be removed, but also human bone (such as the skull, etc.) interference can be removed. Based on this, the embodiments of the present application also provide another possible implementation manner of step 21, which may specifically include steps 213 - 215:

[0096] Step 213: Construct a three-dimensional image to be processed according to the CTA image and a preset numerical threshold, so that there is no CT value for describing air in the three-dimensional image to be processed.

[0097] It should be noted that for the relevant content of step 213, please refer to step 211 above.

[0098] Step 214: Determine object description data according to the largest connected region in the three-dimensional image to be processed, so that the object description data can represent the object (such as the skull + brain) characterized by the largest connected region.

[0099] It should be noted that the implementation manner of step 214 is similar to the implementation manner of step 212 above. For the sake of brevity, it will not be elaborated here.

[0100] Step 215: Use a pre-set bone removal algorithm to perform bone removal processing on the object description data to obtain part description data for describing the target body part.

[0101] Among them, the bone removal algorithm is used to remove bones (e.g., skulls) in the image; moreover, the embodiments of the present application do not limit this bone removal algorithm. For example, it can be implemented using any existing or future algorithm that can remove bones (e.g., skulls) in the image (e.g., active contour method, etc.).

[0102] Based on the relevant content of the above steps 213 to 215, for the vascular bulge detection device provided in the embodiments of the present application, after the vascular bulge detection device obtains the CTA image collected for the target body part, the vascular bulge detection device can first remove air interference from the CTA image according to a preset numerical threshold to obtain a to-be-processed three-dimensional image; then, determine the largest connected region in the to-be-processed three-dimensional image as the object description data, so that the object description data can represent the object characterized by the largest connected region (e.g., skull + brain); finally, use the pre-set bone removal algorithm to perform bone removal processing on the object description data to obtain the part description data for describing the target body part, so that the part description data only records as much as possible the respective CT values for describing the target body part, thus effectively avoiding interference from other CT values (e.g., air density, skull, etc.) unrelated to the target body part to the subsequent bulge recognition process.

[0103] Based on the relevant content of the above step 21, for the vascular bulge detection device provided in the embodiments of the present application, after the vascular bulge detection device obtains the CTA image collected for the target body part, the vascular bulge detection device can extract the part description data for describing the target body part from the CTA image, so that the part description data can more accurately represent the three-dimensional state of the target body part, thus effectively avoiding interference from the pixels in other regions except the region where the target body part is located in the CTA image, which is beneficial to improving the subsequent bulge recognition effect.

[0104] Step 22: Perform isotropic resampling processing on the part description data to obtain the image data to be used.

[0105] As an example, when the part description data includes at least one first pixel in the first direction, at least one second pixel in the second direction, and at least one third pixel in the third direction, step 22 may specifically include steps 221 - 224:

[0106] Step 221: Perform linear interpolation processing on at least one first pixel according to a preset pixel pitch to obtain a first pixel sequence.

[0107] Among them, the preset pixel pitch can be preset; and the embodiments of the present application do not limit the preset pixel pitch. For example, it can be 0.5 millimeters.

[0108] The above-mentioned "at least one first pixel" is used to represent the pixels arranged in the first direction in the part description data. Among them, the first direction is used to represent an axis direction; and the embodiments of the present application do not limit the first direction. For example, it can specifically be the X-axis direction.

[0109] The above-mentioned "first pixel sequence" refers to the processing result of linearly interpolating at least one first pixel according to the preset pixel pitch; and the distance between any two adjacent pixels in the first direction in the "first pixel sequence" is the preset pixel pitch.

[0110] Step 222: Linearly interpolate at least one second pixel according to the preset pixel pitch to obtain a second pixel sequence.

[0111] The above-mentioned "at least one second pixel" is used to represent the pixels arranged in the second direction in the part description data. Among them, the second direction is used to represent an axis direction; and the embodiments of the present application do not limit the second direction. For example, it can specifically be the Y-axis direction.

[0112] The above-mentioned "second pixel sequence" refers to the processing result of linearly interpolating at least one second pixel according to the preset pixel pitch; and the distance between any two adjacent pixels in the second direction in the "second pixel sequence" is the preset pixel pitch.

[0113] Step 223: Linearly interpolate at least one third pixel according to the preset pixel pitch to obtain a third pixel sequence.

[0114] The above-mentioned "at least one third pixel" is used to represent the pixels arranged in the third direction in the part description data. Among them, the third direction is used to represent an axis direction; and the embodiments of the present application do not limit the third direction. For example, it can specifically be the Z-axis direction.

[0115] The above-mentioned "third pixel sequence" refers to the processing result of linearly interpolating at least one third pixel according to the preset pixel pitch; and the distance between any two adjacent pixels in the third direction in the "third pixel sequence" is the preset pixel pitch.

[0116] Step 224: Use the first pixel sequence, the second pixel sequence, and the third pixel sequence to construct the image data to be used.

[0117] In the embodiments of the present application, after obtaining the first pixel sequence, the second pixel sequence, and the third pixel sequence, the first pixel sequence, the second pixel sequence, and the third pixel sequence can be spliced to obtain the image data to be used, so that each pixel in the first direction of the image data to be used comes from the first pixel sequence, each pixel in the second direction of the image data to be used comes from the second pixel sequence, and each pixel in the third direction of the image data to be used comes from the third pixel sequence. In this way, the purpose of isotropic resampling processing for the part description data can be achieved.

[0118] Based on the relevant content of step 22 above, for the vascular bulge detection device provided in the embodiments of the present application, after the vascular bulge detection device obtains the part description data for describing the target body part, the vascular bulge detection device can perform isotropic resampling processing on the part description data to obtain the image data to be used, so that the pixel sampling parameters (e.g., pixel pitch) in each direction of the image data to be used are kept consistent. In this way, the adverse effects caused by inconsistent pixel sampling parameters in different directions can be effectively avoided, so that the three-dimensional shape described by the image data to be used for the target body part is closer to the actual three-dimensional shape of the target body part, and thus the three-dimensional description effect of the image data to be used for the target body part can be effectively improved.

[0119] Step 23: Determine the three-dimensional image data of the target body part according to the image data to be used.

[0120] It should be noted that for the relevant content of step 23, please refer to step 12 above.

[0121] Based on the relevant content of steps 21 to 23 above, for the vascular bulge detection device provided in the embodiments of the present application, after the vascular bulge detection device obtains the CTA image collected for the target body part, the vascular bulge detection device can first extract the part description data from the CTA image so that the part description data can represent the target body part as accurately as possible; then perform isotropic resampling processing on the part description data to obtain the image data to be used so that the pixel sampling parameters (e.g., pixel pitch) in each direction of the image data to be used are kept consistent; finally, extract the three-dimensional image data of the target body part from the image data to be used so that the three-dimensional image data can represent the three-dimensional state of the target body part as accurately as possible. In this way, it is beneficial to improve the expression effect of the three-dimensional image data on the target body part.

[0122] It can be seen that since the above-mentioned "part description data" only carries the CT values for describing the target body part in the CTA image, the "part description data" can more purely represent the target body part, thus effectively avoiding the interference caused by other objects in the CTA image except the target body part, which is beneficial to improving the expression effect of the subsequent determined three-dimensional image data on the target body part.

[0123] In fact, the protrusions appearing on the blood vessel wall have a spherical-like structure, making the shape structures of these protrusions significantly different from the shape structure of the blood vessel. Based on this, in order to further improve the expression effect of the spherical distribution characterization data on the spherical-like objects in the target body part, the above S2 can be implemented by means of a spherical-like three-dimensional shape feature (3D Shape Index, DSI) extraction method. Based on this, an embodiment of the present application provides a possible implementation manner of S2, which may specifically include S21 - S24:

[0124] S21: Determine the Hessian matrix corresponding to each pixel in the CTA image (as shown in formulas (1)-(2)), so that the Hessian matrix can represent the image gray change rate of each pixel in the CTA image.

[0125]

[0126]

[0127] In the formula, H x,y,z represents the Hessian matrix corresponding to the pixel at the coordinate position (x, y, z) in the CTA image; I(x, y, z) represents the CT value of the pixel at the coordinate position (x, y, z) in the CTA image; G(x, y, z) represents a Gaussian function with δ as the scale parameter, and the G(x, y, z) has second-order partial derivatives; δ represents the scale parameter, and δ can be preset.

[0128] It should be noted that the embodiment of the present application does not limit δ. For example, it can be set according to the blood vessel diameter and the maximum diameter of the spherical-like protrusions on the blood vessel wall. For the sake of understanding, the following is illustrated with examples.

[0129] As an example, first, δ = 2.5 can be set with reference to the blood vessel diameter, and δ = 12.5 can be set with reference to the maximum diameter of the spherical-like protrusions on the blood vessel wall (for example, the maximum diameter of the spherical-like protrusions on the blood vessel wall can reach 40 mm); then δ = 7.5 can be deduced based on δ = 2.5 and δ = 12.5, so as to subsequently determine the Hessian matrix corresponding to each pixel in the CTA image in the multi-scale space based on the three scale parameters of δ = 2.5, δ = 7.5, and δ = 12.5.

[0130] S22: Calculate the eigenvalues of the Hessian matrix corresponding to each pixel in the CTA image (as shown in Equation (3)) to obtain the matrix eigenvalues corresponding to each pixel in the CTA image, so that the matrix eigenvalues can represent the degree of image gray-scale change of each pixel in the CTA image.

[0131]

[0132] In the formula, represents the matrix eigenvalue corresponding to the pixel at the coordinate position (x, y, z) in the CTA image; H x,y,z represents the Hessian matrix corresponding to the pixel at the coordinate position (x, y, z) in the CTA image; F(·) represents the eigenvalue calculation processing function, and the embodiments of the present application do not limit F(·). For example, it can be implemented using the QR (orthogonal triangular) decomposition method.

[0133] It should be noted that for the three eigenvalues λ 1 , λ 2 , and λ 3 among them, these three eigenvalues satisfy the condition |λ 1 | ≤ |λ 2 | ≤ |λ 3 |.

[0134] S23: Obtain the DSI value of each pixel in the CTA image according to the matrix eigenvalue corresponding to each pixel in the CTA image (as shown in Equation (4)).

[0135]

[0136] In the formula, DSI (x,y,z) represents the DSI value of the pixel at the coordinate position (x, y, z) in the CTA image; represents the matrix eigenvalue corresponding to the pixel at the coordinate position (x, y, z) in the CTA image.

[0137] It should be noted that the value range of DSI (x,y,z) is [-1, 1].

[0138] S24: Use the DSI values of all pixels in the CTA image to construct spherical distribution characterization data, so that the spherical distribution characterization data can represent the DSI values of each pixel in the CTA image.

[0139] As an example, when δ is preset as the first scale parameter value (e.g., 2.5), the second scale parameter value (e.g., 7.5), and the third scale parameter value (e.g., 12.5), S24 may specifically include S241 - S244:

[0140] S241: Arrange the DSI values calculated for each pixel in the CTA image at the first scale parameter value according to the image coordinates of each pixel in the CTA image to obtain the spherical distribution image data corresponding to the first scale parameter value.

[0141] S242: Arrange the DSI values calculated for each pixel in the CTA image at the second scale parameter value according to the image coordinates of each pixel in the CTA image to obtain the spherical distribution image data corresponding to the second scale parameter value.

[0142] S243: Arrange the DSI values calculated for each pixel in the CTA image at the third scale parameter value according to the image coordinates of each pixel in the CTA image to obtain the spherical distribution image data corresponding to the third scale parameter value.

[0143] S244: Perform set processing (or splicing processing) on the spherical distribution image data corresponding to the first scale parameter value, the spherical distribution image data corresponding to the second scale parameter value, and the spherical distribution image data corresponding to the third scale parameter value to obtain spherical distribution characterization data, so that the spherical distribution characterization data includes the spherical distribution image data corresponding to the first scale parameter value, the spherical distribution image data corresponding to the second scale parameter value, and the spherical distribution image data corresponding to the third scale parameter value.

[0144] Based on the relevant content of S21 to S24 above, for the vascular bulge detection device provided in the embodiments of the present application, after the vascular bulge detection device acquires the CTA image collected for the target body part, the vascular bulge detection device can perform spherical feature extraction processing on the CTA image by means of the three-dimensional shape index for a quasi-spherical shape, to obtain spherical distribution image data, so that the spherical distribution image data can prominently represent the quasi-spherical objects in the CTA image, so as to further determine the quasi-spherical bulges on the blood vessel wall in the target body part from the spherical distribution image data subsequently, which is beneficial to improving the detection accuracy of the quasi-spherical bulges on the blood vessel wall.

[0145] In addition, based on the relevant content of the above "three-dimensional image data", it can be known that in some cases, the "three-dimensional image data" (e.g., Figure 3The data size of the three-dimensional image data) shown may not be consistent with the data size of the CTA image. Therefore, in order to better perform the convex recognition process, the data size of the above-mentioned "spherical distribution characterization data" can be made consistent with the data size of the "three-dimensional image data".

[0146] Based on this, the embodiments of the present application provide two possible implementation manners for extracting the above-mentioned "spherical distribution characterization data", which are introduced separately below.

[0147] In the first possible implementation manner, in order to simplify the detection process of the spherical-like convexities on the blood vessel wall as much as possible, the extraction process of the above-mentioned "spherical distribution characterization data" can specifically be: extracting the spherical distribution characterization data from the three-dimensional image data.

[0148] It can be seen that for the blood vessel convexity detection device provided by the embodiments of the present application, after the blood vessel convexity detection device constructs the three-dimensional image data of the target body part by using the CTA image, the blood vessel convexity detection device can directly extract the spherical distribution characterization data from the three-dimensional image data, so that the data size of the spherical distribution characterization data is consistent with the data size of the three-dimensional image data. In this way, the adverse effects caused by the inconsistent data sizes of the two can be effectively avoided, thereby being conducive to improving the recognition effect of the convexities on the blood vessel wall.

[0149] It should be noted that the specific meaning of "the data size of the spherical distribution characterization data is consistent with the data size of the three-dimensional image data" can be as follows: when the above-mentioned "spherical distribution characterization data" includes the spherical distribution image data corresponding to the first scale parameter value, the spherical distribution image data corresponding to the second scale parameter value, and the spherical distribution image data corresponding to the third scale parameter value, the data size of the "spherical distribution image data corresponding to the first scale parameter value" is consistent with the data size of the three-dimensional image data, the data size of the "spherical distribution image data corresponding to the second scale parameter value" is consistent with the data size of the three-dimensional image data, and the data size of the "spherical distribution image data corresponding to the third scale parameter value" is consistent with the data size of the three-dimensional image data.

[0150] It should also be noted that the embodiments of the present application do not limit the implementation manner of the above step "extracting the spherical distribution characterization data from the three-dimensional image data". For example, it is similar to the implementation manner shown in S21-S24 above. For the sake of brevity, it will not be elaborated here.

[0151] In the second possible implementation manner, in order to make more full use of the image information carried by the CTA image, the extraction process of the above-mentioned "spherical distribution characterization data" can specifically include steps 31 - step 32:

[0152] Step 31: Obtain the mask of the body part to be used.

[0153] Among them, the mask of the body part to be used is used to represent the position of the target body part in the three-dimensional image data; and the data size of the mask of the body part to be used is consistent with the data size of the three-dimensional image data.

[0154] In addition, the embodiment of the present application does not limit the acquisition method of the mask of the body part to be used. For example, specifically, it can be: setting each pixel in the largest connected region in the three-dimensional image data to a fourth value (for example, 1), and setting each pixel in other regions except the largest connected region in the three-dimensional image data to a fifth value (for example, 0), to obtain the mask of the body part to be used, so that the mask of the body part to be used can represent the position of the target body part in the three-dimensional image data.

[0155] In fact, the mask of the body part to be used can also be determined during the construction process of the above-mentioned "three-dimensional image data". Based on this, the embodiment of the present application also provides a second possible implementation manner for obtaining the mask of the body part to be used. For the convenience of understanding, two examples are described below.

[0156] Example 1, when the determination process of the above-mentioned "part description data" includes Step 211-Step 212, and Step 212 includes Step 2121-Step 2122, the acquisition process of the mask of the body part to be used can specifically be: performing data processing on the above-mentioned "data extraction mask" according to the data conversion operation from the part description data to the three-dimensional image data (for example, Step 22-Step 23), to obtain the mask of the body part to be used.

[0157] Example 2, when the determination process of the above-mentioned "part description data" includes Step 213-Step 215, and Step 214 is implemented in a manner similar to that shown in Step 2121-Step 2122, the acquisition process of the mask of the body part to be used can specifically be: performing data processing on the above-mentioned "data extraction mask" according to the data conversion operation from the object description data to the three-dimensional image data (for example, Step 215, and Step 22-Step 23), to obtain the mask of the body part to be used.

[0158] Based on the relevant content of the second possible implementation manner for obtaining the mask of the body part to be used described above, for the vascular bulge detection device provided by the embodiment of the present application, after the vascular bulge detection device acquires the CTA image collected for the target body part, the vascular bulge detection device can also synchronously construct the mask of the body part to be used when constructing the three-dimensional image data of the target body part by using the CTA image, so that the mask of the body part to be used can represent the position of the target body part in the three-dimensional image data.

[0159] Based on the relevant content of the above step 31, for the vascular bulge detection device provided in the embodiments of the present application, the vascular bulge detection device not only needs to construct three-dimensional image data of the target body part, but also needs to construct a mask of the part to be used, so that the mask of the part to be used can represent the position of the target body part in the three-dimensional image data, so that subsequently, with the help of the mask of the part to be used, the data sizes of the spherical distribution characterization data and the three-dimensional image data can be unified.

[0160] Step 32: Determine the spherical distribution characterization data according to the CTA image and the mask of the part to be used, so that the data size of the spherical distribution characterization data is consistent with the data size of the three-dimensional image data.

[0161] As an example, step 32 may specifically include steps 321-322:

[0162] Step 321: Extract the spherical distribution data to be processed from the CTA image.

[0163] Among them, the spherical distribution data to be processed is used to describe the distribution state of the spherical-like objects (for example, spherical-like bulges) in the CTA image.

[0164] In addition, the embodiments of the present application do not limit the implementation manner of step 321. For example, it can be implemented by using S21-S23 above, and only needs to replace "spherical distribution characterization data" in S21-S23 above with "spherical distribution data to be processed".

[0165] In addition, the data size of the spherical distribution data to be processed is consistent with the data size of the CTA image. For example, when the above "spherical distribution data to be processed" includes spherical distribution image data corresponding to the first scale parameter value, spherical distribution image data corresponding to the second scale parameter value, and spherical distribution image data corresponding to the third scale parameter value, the data size of the "spherical distribution image data corresponding to the first scale parameter value" is consistent with the data size of the CTA image, the data size of the "spherical distribution image data corresponding to the second scale parameter value" is consistent with the data size of the CTA image, and the data size of the "spherical distribution image data corresponding to the third scale parameter value" is consistent with the data size of the CTA image.

[0166] Step 322: Determine the spherical distribution characterization data according to the mask of the part to be used and the spherical distribution data to be processed.

[0167] To facilitate the understanding of step 322, the following will be described in combination with two cases.

[0168] Case 1, when the data size of the CTA image is the same as the data size of the three-dimensional image data (that is, the data size of the spherical distribution data to be processed is the same as the data size of the mask of the part to be used), step 322 can specifically be: directly using the mask of the part to be used to perform masking processing on the spherical distribution data to be processed, obtaining spherical distribution characterization data, so that the data size of the spherical distribution characterization data is the same as the data size of the three-dimensional image data, thereby ensuring that the spherical distribution characterization data records as much as possible only the distribution state of the spherical-like objects (such as spherical-like protrusions) within the target body part, so as to effectively avoid interference from spherical-like objects in other image regions outside the area where the target body part is located during the subsequent protrusion recognition process.

[0169] Case 2, when the data size of the CTA image is different from the data size of the three-dimensional image data (that is, the data size of the spherical distribution data to be processed is different from the data size of the mask of the part to be used), the construction process of the three-dimensional image data includes the above steps 11 - step 12 (or, steps 21 - step 23), and the three-dimensional image data refers to the minimum circumscribed rectangle of the target body part in the image data to be used, step 322 can specifically include steps 3221 - step 3222:

[0170] Step 3221: Perform isotropic resampling processing on the spherical distribution data to be processed, obtaining the spherical distribution data to be used, so that the pixel sampling parameters in all directions of the spherical distribution data to be used are consistent.

[0171] It should be noted that the implementation manner of step 3221 is similar to the implementation manner of the above step 11. For the sake of brevity, it will not be elaborated here.

[0172] Based on the relevant content of the above step 3221, after extracting the spherical distribution data to be processed from the CTA image, perform isotropic resampling processing on the spherical distribution data to be processed, obtaining the spherical distribution data to be used, so that the pixel sampling parameters in all directions of the spherical distribution data to be used are consistent, thereby making the data size of the spherical distribution data to be used the same as the data size of the "image data to be used" above.

[0173] Step 3222: Determine the area of the spherical distribution data to be used that is at the position of the image to be used as the spherical distribution characterization data, so that the spherical distribution characterization data includes the minimum circumscribed rectangle of the target body part in the spherical distribution data to be used.

[0174] Among them, the position of the image to be used is used to represent the position of the three-dimensional image data in the image data to be used (for example, Figure 3the position of the minimum bounding rectangle of the target body part shown in the image data to be used), so that the position of the image to be used can represent the position of the minimum bounding rectangle of the target body part in the image data to be used in the image data to be used.

[0175] It can be seen that since the data size of the spherical distribution data to be used is consistent with the data size of the "image data to be used" above, the position of the image to be used can not only represent the position of the minimum bounding rectangle of the target body part in the image data to be used in the image data to be used, but also represent the position of the minimum bounding rectangle of the target body part in the spherical distribution data to be used in the spherical distribution data to be used. Therefore, data extraction can be directly performed on the spherical distribution data to be used according to the position of the image to be used to obtain spherical distribution characterization data, so that the spherical distribution characterization data can represent the distribution state of the spherical object (for example, spherical protrusion) within the minimum bounding rectangle of the target body part in the spherical distribution data to be used, so that the spherical distribution characterization data can represent the distribution state of the spherical object (for example, spherical protrusion) in the above "three-dimensional image data".

[0176] Based on the relevant content of the above steps 3221 to 3222, it can be known that if there are some processing operations that can affect the image data size (such as isotropic resampling processing, extraction of the minimum bounding rectangle of the target body part, etc.) during the process of constructing three-dimensional image data using CTA images, then after extracting the spherical distribution data to be processed from the CTA image, these processing operations that can affect the image data size (such as isotropic resampling processing, extraction of the minimum bounding rectangle of the target body part, etc.) can also be performed on the spherical distribution data to be processed, so that the data size of the spherical distribution data to be processed is consistent with the data size of the "three-dimensional image data" above, so that the spherical distribution characterization data can more accurately represent the distribution state of the spherical object in the three-dimensional image data, so as to effectively ensure that the spherical distribution characterization data and the three-dimensional image data are both used to describe the same region, which is beneficial to improving the subsequent protrusion recognition effect.

[0177] Based on the relevant content of the above steps 31 to 32, for the vascular bulge detection device provided in the embodiments of the present application, after the vascular bulge detection device obtains the CTA image collected for the target body part, the vascular bulge detection device not only needs to construct the three-dimensional image data of the target body part, but also needs to construct a mask for the part to be used, so that the mask for the part to be used can represent the position of the target body part in the three-dimensional image data, so that the vascular bulge detection device can subsequently determine the spherical distribution characterization data according to the CTA image and the mask for the part to be used, so that the data size of the spherical distribution characterization data is consistent with the data size of the three-dimensional image data. In this way, the adverse effects caused by the inconsistent data sizes of the two can be effectively avoided, thereby facilitating the improvement of the recognition effect of the bulge on the blood vessel wall.

[0178] In addition, since the above "spherical distribution characterization data" is extracted from the CTA image, the image information carried by the CTA image can be fully utilized in the extraction process of the "spherical distribution characterization data", so that the "spherical distribution characterization data" can more accurately represent the distribution state of the spherical objects (for example, spherical bulges) in the target body part. In this way, it is beneficial to further improve the recognition accuracy of the bulge on the blood vessel wall.

[0179] In fact, in order to further improve the recognition effect of the bulge on the blood vessel wall, the embodiments of the present application also provide a possible implementation manner of the above S3, which may specifically include S31-S34:

[0180] S31: Determine at least one data block to be used from the three-dimensional image data.

[0181] Among them, the nth data block to be used is extracted from the nth region in the three-dimensional image data, so that the nth data block to be used is used to represent the image information carried by the nth region in the three-dimensional image data. n is a positive integer, n ≤ N, and N represents the number of data blocks to be used.

[0182] In addition, the embodiments of the present application do not limit the data size of the above nth data block to be used. For example, it can be specifically implemented with a data size of 80×80×80.

[0183] In addition, the embodiments of the present application do not limit the determination process of the above "at least one data block to be used". For example, it can be specifically: directly divide the three-dimensional image data according to a preset data block size (for example, 80×80×80, etc.) to obtain at least one data block to be used.

[0184] In fact, in order to avoid the interference caused by tissues other than blood vessels in the target body part (for example, a large amount of skull and gray matter in the brain, etc.), an embodiment of the present application also provides a possible implementation manner for determining the above-mentioned "at least one data block to be used" (that is, S31), which may specifically include S311 - S312:

[0185] S311: Extract vascular characterization data from the three-dimensional image data.

[0186] Among them, the vascular characterization data is used to describe the vascular distribution state recorded for the target body part in the three-dimensional image data, so that the vascular characterization data can represent a large number of blood vessels in the target body part.

[0187] In fact, the densities of different tissues in the target body part are different (for example, the densities of the skull, blood vessels, and gray matter are all different). Based on this, an embodiment of the present application also provides a possible implementation manner for S311, which may specifically be: updating each pixel in the three-dimensional image data whose CT value does not belong to the preset density range to a sixth value to obtain the vascular characterization data.

[0188] Among them, the preset density range is used to represent the range where the blood vessel density is located; and this preset density range can be preset in advance. For example, it can be an interval range of [100Hu, 700Hu].

[0189] The sixth value can be preset in advance; and an embodiment of the present application does not limit this sixth value. For example, this sixth value can be 0.

[0190] In addition, an embodiment of the present application does not limit the implementation manner of the above step "updating each pixel in the three-dimensional image data whose CT value does not belong to the preset density range to a sixth value to obtain the vascular characterization data". For example, it can be implemented using formula (5).

[0191]

[0192] In the formula, v′ (x,y,z) represents the value of the pixel at the coordinate position (x, y, z) in the vascular characterization data; v (x,y,z) represents the value of the pixel at the coordinate position (x, y, z) in the three-dimensional image data; V max represents the upper limit value of the preset density range (for example, 700Hu); V min represents the lower limit value of the preset density range (for example, 100Hu).

[0193] Based on the relevant content of S311 above, for the vascular bulge detection device provided in the embodiments of the present application, after the vascular bulge detection device obtains the three-dimensional image data of the target body part, the vascular bulge detection device can extract vascular characterization data from the three-dimensional image data so that the vascular characterization data can represent the vascular distribution state within the target body part, thus effectively avoiding the interference caused by other tissues within the target body part (for example, a large number of skulls and cerebral gray and white matter in the brain) to the recognition process of the spherical-like bulges on the blood vessel wall.

[0194] S312: Determine at least one data block to be used from the vascular characterization data.

[0195] In the embodiments of the present application, for the vascular bulge detection device provided in the embodiments of the present application, after the vascular bulge detection device obtains the vascular characterization data, the vascular bulge detection device can use a sliding window method to cut out each data block to be used from the vascular characterization data so that each data block to be used respectively records different segments of blood vessels within the target body part, thereby enabling each data block to be used to represent the states of different segments of blood vessels within the target body part respectively.

[0196] In addition, in order to enable the above "at least one data block to be used" to be better used by the vascular bulge detection model, the embodiments of the present application also provide another possible implementation manner for determining the "at least one data block to be used" (that is, S312), which may specifically include S3121 - S3122:

[0197] S3121: Perform normalization processing on the vascular characterization data to obtain vascular normalized data (as shown in formula (6)) so that the values of each pixel in the vascular normalized data belong to the value range of [0, 1].

[0198]

[0199] In the formula, v″ (x,y,z) represents the value of the pixel at the coordinate position (x, y, z) in the vascular normalized data; v′ (x,y,z) represents the value of the pixel at the coordinate position (x, y, z) in the vascular characterization data; V max represents the upper limit value of the preset density range (for example, 700 Hu); V min represents the lower limit value of the preset density range (for example, 100 Hu).

[0200] S3122: Determine at least one data block to be used from the vascular normalized data.

[0201] In the embodiments of the present application, for the vascular bulge detection device provided in the embodiments of the present application, after the vascular bulge detection device obtains the vascular standardized data, the vascular bulge detection device may adopt a sliding window method to cut out each data block to be used from the vascular standardized data, so that each data block to be used respectively records different segments of blood vessels in the target body part, thereby enabling each data block to be used to respectively represent the states of different segments of blood vessels in the target body part.

[0202] Based on the relevant content of S31 above, for the vascular bulge detection device provided in the embodiments of the present application, after the vascular bulge detection device obtains the three-dimensional image data of the target body part, the vascular bulge detection device may determine at least one data block to be used from the three-dimensional image data, so that these data blocks to be used can respectively represent a certain area (for example, a certain segment of blood vessel) in the three-dimensional image data.

[0203] S32: Determine the spherical distribution data blocks corresponding to each data block to be used from the spherical distribution characterization data.

[0204] Among them, the spherical distribution data block corresponding to the nth data block to be used is used to represent the distribution state of the spherical object (for example, spherical bulge) in the nth area in the three-dimensional image data. n is a positive integer, n ≤ N, and N represents the number of data blocks to be used.

[0205] In addition, the embodiments of the present application do not limit the above-mentioned "spherical distribution data block corresponding to the nth data block to be used". For example, when the above-mentioned "spherical distribution characterization data" includes spherical distribution image data corresponding to a first scale parameter value (for example, δ = 2.5), spherical distribution image data corresponding to a second scale parameter value (for example, δ = 7.5), and spherical distribution image data corresponding to a third scale parameter value (for example, δ = 12.5), the "spherical distribution data block corresponding to the nth data block to be used" may include the nth data sub-block corresponding to the first scale parameter value, the nth data sub-block corresponding to the second scale parameter value, and the nth data sub-block corresponding to the third scale parameter value.

[0206] The above-mentioned "nth data sub-block corresponding to the first scale parameter value" is used to represent the distribution state of the spherical object (for example, spherical bulge) in the nth area in the spherical distribution image data corresponding to the first scale parameter value.

[0207] The above-mentioned "nth data sub-block corresponding to the second scale parameter value" is used to represent the distribution state of the spherical object (for example, spherical bulge) in the nth area in the spherical distribution image data corresponding to the second scale parameter value.

[0208] The above-mentioned "nth data sub-block corresponding to the third scale parameter value" is used to represent the distribution state of the spherical object (e.g., spherical protrusion) in the nth region of the spherical distribution image data corresponding to the third scale parameter value.

[0209] In addition, the embodiment of the present application does not limit the data size of the above-mentioned "spherical distribution data block corresponding to the nth data block to be used". For example, the data size of the above-mentioned "spherical distribution data block corresponding to the nth data block to be used" can be the same as the data size of the nth data block to be used. For example, when the above-mentioned "spherical distribution data block corresponding to the nth data block to be used" includes the nth data sub-block corresponding to the first scale parameter value, the nth data sub-block corresponding to the second scale parameter value, and the nth data sub-block corresponding to the third scale parameter value, the data size of the "nth data sub-block corresponding to the first scale parameter value" is the same as the data size of the nth data block to be used, the data size of the "nth data sub-block corresponding to the second scale parameter value" is the same as the data size of the nth data block to be used, and the data size of the "nth data sub-block corresponding to the third scale parameter value" is the same as the data size of the nth data block to be used.

[0210] In fact, in order to enable the above-mentioned "spherical distribution data blocks corresponding to each data block to be used" to be better used by the blood vessel protrusion detection model, the embodiment of the present application also provides another possible implementation manner for determining the "spherical distribution data blocks corresponding to each data block to be used" (that is, S32), for example, it may specifically include S321-S322:

[0211] S321: Perform normalization processing on the spherical distribution characterization data to obtain spherical distribution normalized data, so that the values of each pixel in the spherical distribution normalized data belong to the value range of [0, 1].

[0212] As an example, S321 may specifically include S3211-S3212:

[0213] S3211: Compare the values of each pixel in the spherical distribution characterization data to obtain the maximum value and the minimum value.

[0214] Among them, the maximum value refers to the maximum value of the values of all pixels in the spherical distribution characterization data.

[0215] The minimum value refers to the minimum value of the values of all pixels in the spherical distribution characterization data.

[0216] S3212: Refer to the maximum value and the minimum value, and perform normalization processing on the spherical distribution characterization data to obtain spherical distribution normalized data, so that the values of each pixel in the spherical distribution normalized data belong to the value range of [0, 1].

[0217] It should be noted that the embodiments of the present application do not limit the implementation manner of S3212. For example, it can be implemented by any existing or future standardized method.

[0218] Based on the relevant content of S321 above, for the vascular bulge detection device provided in the embodiments of the present application, after the vascular bulge detection device obtains the spherical distribution characterization data, the vascular bulge detection device can perform standardization processing on the spherical distribution characterization data to obtain spherical distribution standardized data, so that the values of each pixel in the spherical distribution standardized data belong to the value range of [0, 1].

[0219] S322: Determine the spherical distribution data blocks corresponding to each data block to be used from the spherical distribution standardized data.

[0220] It should be noted that the embodiments of the present application do not limit the implementation manner of S322. For example, specifically, it can be: according to the position of the nth data block to be used in the three-dimensional image data, extract the spherical distribution data block corresponding to the nth data block to be used from the spherical distribution standardized data, so that the "spherical distribution data block corresponding to the nth data block to be used" can represent the distribution state of the spherical object in the nth region in the three-dimensional image data, so that the "spherical distribution data block corresponding to the nth data block to be used" and the nth data block to be used can describe the relevant information of the same region.

[0221] Based on the relevant content of S32 above, for the vascular bulge detection device provided in the embodiments of the present application, after the vascular bulge detection device obtains the spherical distribution characterization data, the vascular bulge detection device can segment the spherical distribution data blocks corresponding to each data block to be used from the spherical distribution characterization data according to the position of each data block to be used in the three-dimensional image data.

[0222] S33: Determine the vascular bulge detection results of each data block to be used according to each data block to be used, the spherical distribution data blocks corresponding to each data block to be used, and the pre-constructed vascular bulge detection model.

[0223] Among them, the vascular bulge detection result of the nth data block to be used is used to represent the distribution state of the spherical object in the nth data block to be used (for example, the distribution state of the spherical bulge on the blood vessel in the nth data block to be used). n is a positive integer, n ≤ N, and N represents the number of data blocks to be used.

[0224] The vascular bulge detection model is used to perform bulge recognition processing on the input data of the vascular bulge detection model; and the embodiments of the present application do not limit the above-mentioned vascular bulge detection model. For example,Figure 4 As shown in the figure, the blood vessel bulge detection model 400 may specifically include a feature extraction module 401, a segmentation processing module 402, a feature splicing module 403, and a data prediction module 404. Among them, the input data of the feature splicing module 403 includes the output data of the feature extraction module 401 and the output data of the segmentation processing module 402. The input data of the data prediction module 404 includes the output data of the feature splicing module 403.

[0225] To facilitate the understanding of the working principle of the blood vessel bulge detection model 400, the determination process of the "blood vessel bulge detection result of the nth data block to be used" will be used as an example for description below.

[0226] As an example, the determination process of the above "blood vessel bulge detection result of the nth data block to be used" may specifically include S331-S334:

[0227] S331: Determine the feature extraction result of the nth data block to be used according to the nth data block to be used and the feature extraction module 401.

[0228] Among them, the feature extraction module 401 is used to perform feature extraction processing on the input data of the feature extraction module 401; and the embodiments of the present application do not limit the feature extraction module 401. For example, it may be implemented by using any existing or future feature extraction method (such as Dual Attention Network (DA), etc.).

[0229] The above "feature extraction result of the nth data block to be used" is used to represent the blood vessel distribution state in the nth data block to be used; and the embodiments of the present application do not limit the determination process of the "feature extraction result of the nth data block to be used". For example, specifically, the nth data block to be used is input into the feature extraction module 401 to obtain the feature extraction result of the nth data block output by the feature extraction module 401.

[0230] In addition, the embodiments of the present application do not limit the data size of the above "feature extraction result of the nth data block to be used". For example, when the nth data block to be used has a data size of 80×80×80, the "feature extraction result of the nth data block to be used" also has a data size of 80×80×80.

[0231] Based on the relevant content of the above S331, for the vascular bulge detection model 400, after the feature extraction module 401 in the vascular bulge detection model 400 obtains the nth data block to be used, the feature extraction module 401 can directly perform feature extraction processing on the nth data block to be used, obtain and output the feature extraction result of the nth data block to be used, so that the "feature extraction result of the nth data block to be used" can represent the vascular distribution state in the nth data block to be used.

[0232] S332: Input the spherical distribution data block corresponding to the nth data block to be used into the segmentation processing module 402, and obtain the spherical region segmentation result corresponding to the nth data block to be used output by the segmentation processing module 402.

[0233] Among them, the segmentation processing module 402 is used to perform spherical bulge segmentation processing on the input data of the segmentation processing module 402; moreover, the embodiments of the present application do not limit the implementation manner of the segmentation processing module 402. For example, it can adopt any existing or future image segmentation method (for example, Figure 5 the segmentation network based on the dual attention mechanism (Dual attention network for scene segmentation, DAResUNet) shown, etc.) for implementation.

[0234] It should be noted that the embodiments of the present application do not limit Figure 5 the implementation manner of the convolution processing involved. For example, it can adopt any existing or future convolution processing method for implementation (for example, implement it by using a convolution layer + a normalization layer + an activation layer).

[0235] The above "spherical region segmentation result corresponding to the nth data block to be used" refers to the segmentation result of the spherical bulge in the nth data block to be used.

[0236] In addition, the embodiments of the present application do not limit the data size of the above "spherical region segmentation result corresponding to the nth data block to be used". For example, when the nth data block to be used has a data size of 80×80×80, the "spherical region segmentation result corresponding to the nth data block to be used" also has a data size of 80×80×80.

[0237] Based on the relevant content of S332 above, for the vascular bulge detection model 400, after the segmentation processing module 402 in the vascular bulge detection model 400 obtains the spherical distribution data block corresponding to the nth data block to be used, the segmentation processing module 402 can perform segmentation processing on the spherical distribution data block to obtain the spherical region segmentation result corresponding to the nth data block to be used, so that the "spherical region segmentation result corresponding to the nth data block to be used" can represent the distribution state of the quasi-spherical object in the nth data block to be used.

[0238] S333: Input the feature extraction result of the nth data block to be used and the spherical region segmentation result corresponding to the nth data block to be used into the feature splicing module 403 to obtain the to-be-processed spliced feature output by the feature splicing module 403.

[0239] Among them, the feature splicing module 403 is used to perform splicing processing on the input data of the feature splicing module 403; moreover, the embodiments of the present application do not limit the implementation manner of the feature splicing module 403. For example, any existing or future method capable of splicing two features can be used for implementation.

[0240] The above "to-be-processed spliced feature" is obtained by splicing the feature extraction result of the nth data block to be used and the spherical region segmentation result corresponding to the nth data block to be used, so that the "to-be-processed spliced feature" includes the feature extraction result of the nth data block to be used and the spherical region segmentation result corresponding to the nth data block to be used.

[0241] In addition, the embodiments of the present application do not limit the data size of the above "to-be-processed spliced feature". For example, when the above "feature extraction result of the nth data block to be used" has a data size of 80×80×80, and the above "spherical region segmentation result corresponding to the nth data block to be used" also has a data size of 80×80×80, the "to-be-processed spliced feature" has a data size of 80×80×160.

[0242] Based on the relevant content of S333 above, for the vascular bulge detection model 400, after the feature splicing module 403 in the vascular bulge detection model 400 obtains the feature extraction result of the nth data block to be used and the spherical region segmentation result corresponding to the nth data block to be used, the feature splicing module 403 can splice these two features into one feature to obtain the to-be-processed spliced feature, so that the to-be-processed spliced feature includes the feature extraction result of the nth data block to be used and the spherical region segmentation result corresponding to the nth data block to be used.

[0243] S334: Input the splicing feature data to be processed into the data prediction module 404 to obtain the blood vessel bulge detection result of the nth data block to be used output by the data prediction module 404.

[0244] Among them, the data prediction module 404 is used to perform prediction processing on the spherical bulge on the blood vessel wall for the input data of the data prediction module 404. Moreover, the embodiments of the present application do not limit the implementation manner of the data prediction module 404. For example, any existing or future prediction processing network (such as a softmax layer, a sigmoid layer, etc.) can be used for implementation. Another example is that the data prediction module 404 may also include a convolution processing network and a prediction processing network (such as a softmax layer, a sigmoid layer, etc.); and the input data of the prediction processing network includes the output data of the convolution processing network. Among them, the convolution kernel size of the convolution processing network is 1; and the convolution processing network is used to perform convolution processing on the input data of the convolution processing network.

[0245] It should be noted that the embodiments of the present application do not limit the above-mentioned "convolution processing network". For example, it can be implemented by using a combination of a convolution layer + a normalization layer + an activation layer.

[0246] Based on the relevant content of the above S33, for the blood vessel bulge detection device provided by the embodiments of the present application, after the blood vessel bulge detection device obtains the nth data block to be used and the spherical distribution data block corresponding to the nth data block to be used, the blood vessel bulge detection device can use the pre-constructed blood vessel bulge detection model to perform blood vessel spherical object detection processing on the nth data block to be used and the spherical distribution data block corresponding to the nth data block to be used, to obtain the blood vessel bulge detection result of the nth data block to be used, so that the blood vessel bulge detection result can represent the distribution state of the spherical objects in the nth data block to be used (in particular, the distribution state of the spherical bulges on the blood vessel wall in the nth data block to be used).

[0247] S34: Determine the blood vessel bulge detection result of the target body part according to the blood vessel bulge detection results of at least one data block to be used.

[0248] In the embodiments of the present application, after obtaining the blood vessel bulge detection results of all data blocks to be used, the blood vessel bulge detection results of these data blocks to be used can be aggregated (or spliced) to obtain the blood vessel bulge detection result of the target body part, so that the "blood vessel bulge detection result of the target body part" can represent the distribution state of the spherical objects on the blood vessel wall in the target body part.

[0249] Based on the relevant contents of S31 to S34 above, it can be known that for the vascular protrusion detection device provided in the embodiment of the present application, after the vascular protrusion detection device obtains the three-dimensional image data of the target body part, the vascular protrusion detection device can use the pre-constructed vascular protrusion detection model and the spherical distribution characterization data to perform protrusion detection processing on the vascular wall for the spherical object in the three-dimensional image data, and obtain the vascular protrusion detection result of the target body part, so that the "vascular protrusion detection result of the target body part" can represent the distribution state of the protrusions on the vascular wall in the target body part, which is conducive to improving the recognition effect of the protrusions on the vascular wall.

[0250] In fact, for the target body part, the blood vessels have neighborhood connectivity, and the spherical protrusions on the blood vessel wall exist in isolation. It can be seen that in order to avoid false detection as much as possible (for example, identifying the severed blood vessels in a data block as spherical objects), the severed blood vessels and the spherical protrusions on the blood vessel wall can be more effectively distinguished with the help of neighbor information.

[0251] Based on this, the present application embodiment also provides another possible implementation of the above S3, which may specifically include steps 41 to 44:

[0252] Step 41: Determine at least one data block to be used and at least one neighboring data block of each data block to be used from the three-dimensional image data.

[0253] Among them, at least one neighboring data block of the nth data block to be used is used to represent the surrounding information recorded for the nth area in the three-dimensional image data, so that the "at least one neighboring data block of the nth data block to be used" can perform image semantic supplement for the nth data block to be used (in particular, image semantic supplement can only be performed for the severed blood vessels in the nth data block to be used).

[0254] In addition, the embodiment of the present application does not limit the data size of the above-mentioned "neighboring data block". For example, the data size of the "neighboring data block" is consistent with the data size of the nth data block to be used. It can be seen that when the nth data block to be used has a data size of 80×80×80, each neighboring data block of the nth data block to be used has a data size of 80×80×80.

[0255] In addition, the embodiment of the present application does not limit the number of the above-mentioned "at least one neighboring data block of the nth data block to be used". For example, it may include 6 neighboring data blocks, so that the 6 neighboring data blocks can represent the surrounding information in the three-dimensional space recorded for the nth area in the three-dimensional image data.

[0256] It should be noted that if the nth data block to be used is located at the edge position in the three-dimensional image data, it will result in the inability to collect 6 neighboring data blocks of the nth data block to be used from the three-dimensional image data. Therefore, pure zero data blocks can be used for supplementation.

[0257] In addition, the embodiment of the present application does not limit the determination process of the above "at least one data block to be used and at least one neighboring data block of each data block to be used" (that is, the implementation manner of step 41). For example, any existing or future method capable of extracting data blocks and their neighboring data blocks from a three-dimensional image can be used for implementation.

[0258] Actually, in order to avoid the interference caused by other tissues in the target body part except blood vessels (for example, a large amount of skull and gray and white matter in the brain), the embodiment of the present application also provides a possible implementation manner for determining "at least one data block to be used and at least one neighboring data block of each data block to be used" (that is, step 41), which may specifically include steps 411 - 412:

[0259] Step 411: Extract vascular characterization data from the three-dimensional image data.

[0260] It should be noted that for the relevant content of step 411, please refer to the relevant content of S311 above.

[0261] Step 412: Determine at least one data block to be used and at least one neighboring data block of each data block to be used from the vascular characterization data.

[0262] In the embodiment of the present application, for the vascular bulge detection device provided by the embodiment of the present application, after the vascular bulge detection device obtains the vascular characterization data, the vascular bulge detection device can use a sliding window method to cut out each data block to be used from the vascular characterization data, so that each data block to be used can respectively represent the states of different segments of blood vessels in the target body part; and at least one neighboring data block (for example, 6 neighboring data blocks) of each data block to be used can also be cut out from the vascular characterization data, so that at least one neighboring data block of each data block to be used can represent the peripheral information of different segments of blood vessels in the target body part.

[0263] In addition, in order to enable the above "at least one data block to be used and at least one neighboring data block of each data block to be used" to be better used by the vascular bulge detection model, the embodiment of the present application also provides another possible implementation manner for determining the "at least one data block to be used and at least one neighboring data block of each data block to be used" (that is, step 412), which may specifically include steps 4121 - 4122:

[0264] Step 4121: Standardize the vascular characterization data to obtain standardized vascular data (as shown in formula (6)) so that the values of each pixel in the standardized vascular data belong to the value range of [0, 1].

[0265] It should be noted that for the relevant content of step 4121, please refer to the relevant content of S3121 above.

[0266] Step 4122: Determine at least one data block to be used and at least one neighboring data block for each data block to be used from the standardized vascular data.

[0267] In the embodiments of the present application, for the vascular bulge detection device provided in the embodiments of the present application, after the vascular bulge detection device obtains the standardized vascular data, the vascular bulge detection device can use a sliding window method to cut out each data block to be used from the standardized vascular data so that each data block to be used can respectively represent the states of different segments of blood vessels in the target body part; and it can also cut out at least one neighboring data block (for example, 6 neighboring data blocks) for each data block to be used from the standardized vascular data so that at least one neighboring data block for each data block to be used can represent the peripheral information of different segments of blood vessels in the target body part.

[0268] Based on the relevant content of the above step 41, it can be known that for the vascular bulge detection device provided in the embodiments of the present application, after the vascular bulge detection device obtains the three-dimensional image data of the target body part, the vascular bulge detection device can determine at least one data block to be used and at least one neighboring data block thereof from the three-dimensional image data so that these data blocks to be used and at least one neighboring data block thereof can represent a certain area (for example, a certain segment of blood vessel) in the three-dimensional image data and its peripheral information, so that these data blocks to be used and at least one neighboring data block thereof can better describe the blood vessel distribution information and spherical object distribution information carried by the data block to be used.

[0269] Step 42: Determine the spherical distribution data block corresponding to each data block to be used from the spherical distribution characterization data.

[0270] It should be noted that for the relevant content of step 42, please refer to the relevant content of S32 above.

[0271] Step 43: Determine the vascular bulge detection result of each data block to be used according to each data block to be used, at least one neighboring data block of each data block to be used, the spherical distribution data block corresponding to each data block to be used, and the pre-constructed vascular bulge detection model.

[0272] In fact, in order to further improve the recognition effect for the spherical protrusions on the blood vessel wall, all neighboring data blocks of the nth data block to be used can be utilized to supplement the image semantic information for the nth data block to be used. Based on this, another possible implementation manner of the blood vessel bulge detection model is provided in the embodiments of the present application. For the convenience of understanding, the following is described in conjunction with examples and Figure 4-5 for illustration.

[0273] As an example, when the blood vessel bulge detection model 400 includes a feature extraction module 401, a segmentation processing module 402, a feature splicing module 403, and a data prediction module 404, step 43 may specifically include steps 431 to 434:

[0274] Step 431: Determine the feature extraction result of the nth data block to be used according to the nth data block to be used, at least one neighboring data block of the nth data block to be used, and the feature extraction module 401.

[0275] In fact, in order to better fuse the image semantic information carried by all neighboring data blocks of the nth data block to be used into the nth data block to be used, another possible implementation manner of the feature extraction module 401 is provided in the embodiments of the present application. It may include a spatial structure feature extraction sub-module and a neighborhood feature fusion sub-module; and the input data of the neighborhood feature fusion sub-module includes the output data of the spatial structure feature extraction sub-module. For the convenience of understanding its working principle, the following is described in conjunction with examples.

[0276] As an example, when the feature extraction module 401 includes a spatial structure feature extraction sub-module and a neighborhood feature fusion sub-module, step 431 may specifically include steps 4311 to 4313:

[0277] Step 4311: Input the nth data block to be used into the spatial structure feature extraction sub-module to obtain the spatial feature to be used output by the spatial structure feature extraction sub-module.

[0278] Among them, the spatial feature to be used is used to represent the spatial distribution state of the blood vessels carried by the nth data block to be used.

[0279] The spatial structure feature extraction sub-module is used to perform spatial structure feature extraction processing on the input data of the spatial structure feature extraction sub-module; and the embodiments of the present application do not limit the implementation manner of the spatial structure feature extraction sub-module. For example, it may adopt any existing or future method with spatial structure feature extraction function (for example, Figure 5 DA shown) for implementation.

[0280] Based on the relevant content of the above step 4311, for the feature extraction module 401, after the feature extraction module 401 obtains the nth data block to be used, the spatial structure feature extraction sub-module in the feature extraction module 401 can extract spatial structure information from the nth data block to be used, obtain and output the spatial feature to be used, so that the spatial feature to be used can represent the spatial distribution state of the blood vessels carried by the nth data block to be used.

[0281] Step 4312: Use the spatial structure feature extraction sub-module to extract at least one neighboring reference spatial feature corresponding to the spatial feature to be used from at least one neighboring data block of the nth data block to be used.

[0282] Among them, the hth neighboring reference spatial feature corresponding to the spatial feature to be used is used to represent the spatial distribution state of the blood vessels carried in the hth neighborhood of the nth data block to be used. h is a positive integer, h ≤ H n , H n is a positive integer, H n represents the number of neighboring data blocks in the above "at least one neighboring data block of the nth data block to be used".

[0283] In addition, the specific determination process of the above "hth neighboring reference spatial feature corresponding to the spatial feature to be used" can be: input the hth neighboring data block of the nth data block to be used into the spatial structure feature extraction sub-module, and obtain the "hth neighboring reference spatial feature corresponding to the spatial feature to be used" output by the spatial structure feature extraction sub-module.

[0284] Based on the relevant content of the above step 4312, for the feature extraction module 401, after the feature extraction module 401 obtains each neighboring data block of the nth data block to be used, the spatial structure feature extraction sub-module in the feature extraction module 401 can respectively extract spatial structure information from each neighboring data block of the nth data block to be used, and obtain and output each neighboring reference spatial feature corresponding to the spatial feature to be used.

[0285] Step 4313: Input the spatial feature to be used and at least one neighboring reference spatial feature corresponding to the spatial feature to be used into the neighborhood feature fusion sub-module, and obtain the feature extraction result of the nth data block to be used output by the neighborhood feature fusion sub-module.

[0286] Among them, the neighborhood feature fusion sub-module is used to perform neighborhood feature fusion processing on the input data of the neighborhood feature fusion sub-module; and the embodiments of the present application do not limit the implementation manner of the "neighborhood feature fusion sub-module". For example, any existing or future method capable of performing neighborhood feature fusion processing can be used (for example,Figure 5 It is implemented using a bidirectional long short-term memory network (such as Bi-directional Long Short-Term Memory, BiLSTM), etc.

[0287] It can be seen that for the feature extraction module 401, after the neighborhood feature fusion sub-module in the feature extraction module 401 obtains the to-be-used spatial feature and at least one neighboring reference spatial feature corresponding to the to-be-used spatial feature, the neighborhood feature fusion sub-module can refer to the neighborhood information carried by these neighboring reference spatial features to perform image semantic supplementation processing on the to-be-used spatial feature, and obtain the feature extraction result of the nth to-be-used data block, so that the "feature extraction result of the nth to-be-used data block" can conform to the characteristic of blood vessel neighborhood connectivity as much as possible, thereby enabling the "feature extraction result of the nth to-be-used data block" to more accurately represent the blood vessels in the nth to-be-used data block.

[0288] Based on the relevant content of step 431 above, for the blood vessel bulge detection model 400, after the feature extraction module 401 in the blood vessel bulge detection model 400 obtains the nth to-be-used data block and at least one neighboring data block thereof, the feature extraction module 401 can use the image semantic information carried by these neighboring data blocks to perform feature extraction processing on the nth to-be-used data block, and obtain the feature extraction result of the nth to-be-used data block, so that the "feature extraction result of the nth to-be-used data block" can conform to the characteristic of blood vessel neighborhood connectivity as much as possible, thereby enabling the "feature extraction result of the nth to-be-used data block" to more accurately represent the blood vessels in the nth to-be-used data block, and further effectively avoiding the misidentification phenomenon that the blood vessels are misidentified as spherical objects after being divided into multiple small segments. In this way, it is beneficial to improve the recognition accuracy of the blood vessel bulge detection model 400 for spherical bulges on the blood vessel wall.

[0289] Step 432: Input the spherical distribution data block corresponding to the nth to-be-used data block into the segmentation processing module 402, and obtain the spherical region segmentation result corresponding to the nth to-be-used data block output by the segmentation processing module 402.

[0290] It should be noted that for the relevant content of step 432, please refer to the relevant content of S332 above.

[0291] Step 433: Input the feature extraction result of the nth to-be-used data block and the spherical region segmentation result corresponding to the nth to-be-used data block into the feature splicing module 403, and obtain the to-be-processed splicing feature output by the feature splicing module 403.

[0292] It should be noted that for the relevant content of step 433, please refer to the relevant content of S333 above.

[0293] Step 434: Input the splicing feature to be processed into the data prediction module 404, and obtain the blood vessel bulge detection result of the nth data block to be used output by the data prediction module 404.

[0294] It should be noted that for the relevant content of step 434, please refer to the relevant content of S334 above.

[0295] Based on the relevant content of the above step 43, for the blood vessel bulge detection device provided in the embodiment of the present application, after the blood vessel bulge detection device obtains the nth data block to be used, at least one neighboring data block of the nth data block to be used, and the spherical distribution data block corresponding to the nth data block to be used, the blood vessel bulge detection device can, by means of a pre-constructed blood vessel bulge detection model, perform blood vessel-like spherical object detection processing on the nth data block to be used, at least one neighboring data block of the nth data block to be used, and the spherical distribution data block corresponding to the nth data block to be used, so as to obtain the blood vessel bulge detection result of the nth data block to be used, so that the blood vessel bulge detection result can represent the distribution state of the spherical-like objects in the nth data block to be used (in particular, the distribution state of the spherical-like bulges on the blood vessel wall in the nth data block to be used).

[0296] Step 44: Determine the blood vessel bulge detection result of the target body part according to the blood vessel bulge detection results of at least one data block to be used.

[0297] It should be noted that for the relevant content of step 44, please refer to the relevant content of S34 above.

[0298] Based on the relevant content of the above steps 41 to 44, for the blood vessel bulge detection device provided in the embodiment of the present application, after the blood vessel bulge detection device obtains the three-dimensional image data of the target body part, it can use spherical distribution characterization data and neighborhood information to perform detection processing on the spherical-like objects in the three-dimensional image data, so as to obtain the blood vessel bulge detection result of the target body part, so that the "blood vessel bulge detection result of the target body part" meets the detection requirements that blood vessels have neighborhood connectivity and the bulges on the blood vessel wall are isolated bulges, so that the "blood vessel bulge detection result of the target body part" can better represent the distribution state of the bulges on the blood vessel wall in the target body part, which is beneficial to improving the recognition effect of the bulges on the blood vessel wall.

[0299] Based on the relevant content of the above blood vessel bulge detection method, the embodiment of the present application also provides a blood vessel bulge detection device. For the convenience of understanding, the following will explain and illustrate the blood vessel bulge detection device with reference to the accompanying drawings.

[0300] See Figure 6 , which is a schematic structural diagram of a vascular bulge detection device provided by an embodiment of the present application.

[0301] The vascular bulge detection device 600 provided by the embodiment of the present application includes:

[0302] An image construction unit 601, configured to construct three-dimensional image data of the target body part by using the CTA image after acquiring a CTA angiography image of the target body part.

[0303] A spherical extraction unit 602, configured to extract spherical distribution characterization data from the CTA image.

[0304] A bulge detection unit 603, configured to determine a vascular bulge detection result of the target body part according to the three-dimensional image data, the spherical distribution characterization data, and a pre-constructed vascular bulge detection model.

[0305] In a possible implementation manner, the image construction unit 601 includes:

[0306] A resampling subunit, configured to perform isotropic resampling processing on the CTA image to obtain image data to be used;

[0307] An image determination subunit, configured to determine three-dimensional image data of the target body part according to the image data to be used.

[0308] In a possible implementation manner, the image construction unit 601 further includes:

[0309] An image extraction subunit, configured to extract part description data for describing the target body part from the CTA image;

[0310] The resampling subunit is specifically configured to: perform isotropic resampling processing on the part description data to obtain the image data to be used.

[0311] In a possible implementation manner, the part description data includes at least one first pixel in a first direction, at least one second pixel in a second direction, and at least one third pixel in a third direction;

[0312] The re-sampling unit is specifically configured to: perform linear interpolation processing on the at least one first pixel according to a preset pixel pitch to obtain a first pixel sequence; perform linear interpolation processing on the at least one second pixel according to the preset pixel pitch to obtain a second pixel sequence; perform linear interpolation processing on the at least one third pixel according to the preset pixel pitch to obtain a third pixel sequence; and construct the image data to be used by using the first pixel sequence, the second pixel sequence, and the third pixel sequence.

[0313] In a possible implementation manner, the vascular bulge detection device 600 further includes:

[0314] A mask acquisition unit, configured to acquire a mask of a part to be used; wherein, the mask of the part to be used is used to represent the position of the target body part in the three-dimensional image data.

[0315] The spherical extraction unit 602 is specifically configured to: determine the spherical distribution characterization data according to the CTA image and the mask of the part to be used.

[0316] In a possible implementation manner, the spherical extraction unit 602 is specifically configured to: extract the spherical distribution data to be processed from the CTA image; and determine the spherical distribution characterization data according to the mask of the part to be used and the spherical distribution data to be processed.

[0317] In a possible implementation manner, the spherical extraction unit 602 is specifically configured to: extract the spherical distribution characterization data from the three-dimensional image data.

[0318] In a possible implementation manner, the bulge detection unit 603 includes:

[0319] A first cutting subunit, configured to determine at least one data block to be used from the three-dimensional image data.

[0320] A second cutting subunit, configured to determine the spherical distribution data blocks corresponding to the data blocks to be used from the spherical distribution characterization data.

[0321] A bulge detection subunit, configured to determine the vascular bulge detection results of the data blocks to be used according to the data blocks to be used, the spherical distribution data blocks corresponding to the data blocks to be used, and a pre-constructed vascular bulge detection model.

[0322] A result integration subunit, configured to determine the vascular bulge detection result of the target body part according to the vascular bulge detection results of the at least one data block to be used.

[0323] In a possible implementation manner, the bulge detection unit 603 further includes:

[0324] A third cutting subunit, configured to determine at least one neighboring data block of each of the to-be-used data blocks from the three-dimensional image data;

[0325] The protrusion detection subunit is specifically configured to: determine the blood vessel protrusion detection results of each of the to-be-used data blocks according to each of the to-be-used data blocks, at least one neighboring data block of each of the to-be-used data blocks, the spherical distribution data block corresponding to each of the to-be-used data blocks, and a pre-constructed blood vessel protrusion detection model.

[0326] In a possible implementation manner, the protrusion detection unit 603 further includes:

[0327] A blood vessel characterization subunit, configured to extract blood vessel characterization data from the three-dimensional image data;

[0328] The first cutting subunit is specifically configured to: determine at least one to-be-used data block from the blood vessel characterization data.

[0329] In a possible implementation manner, the protrusion detection unit 603 further includes:

[0330] A standard processing subunit, configured to perform standardization processing on the blood vessel characterization data and the spherical distribution characterization data respectively to obtain blood vessel standardized data and spherical distribution standardized data;

[0331] The first cutting subunit is specifically configured to: determine at least one to-be-used data block from the blood vessel standardized data;

[0332] The second cutting subunit is specifically configured to: determine the spherical distribution data block corresponding to each of the to-be-used data blocks from the spherical distribution standardized data.

[0333] In a possible implementation manner, the blood vessel protrusion detection model includes: a feature extraction module, a segmentation processing module, a feature splicing module, and a data prediction module; the number of the to-be-used data blocks is N;

[0334] The convexity detection subunit is specifically configured to: determine the feature extraction result of the nth data block to be used according to the nth data block to be used and the feature extraction module; input the spherical distribution data block corresponding to the nth data block to be used into the segmentation processing module to obtain the spherical region segmentation result of the nth data block to be used output by the segmentation processing module; input the feature extraction result of the nth data block to be used and the spherical region segmentation result of the nth data block to be used into the feature splicing module to obtain the to-be-processed splicing feature output by the feature splicing module; input the to-be-processed splicing feature into the data prediction module to obtain the blood vessel convexity detection result of the nth data block to be used output by the data prediction module; where n is a positive integer, n ≤ N, and N is a positive integer.

[0335] Based on the relevant content of the above blood vessel convexity detection device 600, for the blood vessel convexity detection device 600 provided in the embodiment of the present application, after obtaining the CTA image collected for the target body part (for example, the brain), first use the CTA image to construct the three-dimensional image data of the target body part, so that the three-dimensional image data can better represent the target body part in a three-dimensional stereoscopic manner, and extract the spherical distribution characterization data from the CTA image, so that the spherical distribution characterization data can as accurately as possible represent the distribution state of the spherical objects in the target body part; then use the pre-constructed blood vessel convexity detection model to determine the blood vessel convexity detection result of the target body part from the three-dimensional image data and the spherical distribution characterization data, so that the blood vessel convexity detection result can represent the convexity distribution on the blood vessel wall in the target body part, so that in the future, doctors can refer to the blood vessel convexity detection result and other information (such as specific clinical manifestations) for diagnosis and treatment-related processing (such as physical health status assessment or disease diagnosis, etc.). In this way, the defects existing in the above artificial inspection method can be effectively overcome, and thus the recognition effect for the convexities on the blood vessel wall can be effectively improved.

[0336] In addition, an embodiment of the present application further provides a device, including: a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, any implementation manner of the blood vessel convexity detection method provided in the embodiment of the present application is implemented.

[0337] In addition, an embodiment of the present application further provides a computer-readable storage medium, in which instructions are stored. When the instructions are run on a terminal device, the terminal device is caused to execute any implementation manner of the blood vessel convexity detection method provided in the embodiment of the present application.

[0338] In addition, an embodiment of the present application further provides a computer program product. When the computer program product runs on a terminal device, the terminal device is caused to execute any implementation manner of the blood vessel bulge detection method provided by the embodiment of the present application.

[0339] It should be noted that the various embodiments in this specification are described in a progressive manner. The key point of each embodiment is to illustrate the differences from other embodiments. The same or similar parts among the various embodiments can be referred to each other. For the systems or devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple, and the relevant parts can be referred to the descriptions in the method part.

[0340] It should be understood that in the present application, "at least one (item)" means one or more, and "a plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships can exist. For example, "A and / or B" can mean: only A exists, only B exists, and both A and B exist at the same time. Among them, A and B can be singular or plural. The character " / " generally means that the associated objects before and after are in an "or" relationship. "At least one (one) of the following" or its similar expression means any combination of these items, including any combination of single item (one) or plural items (ones). For example, at least one (one) of a, b, or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0341] It should also be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. Without further limitations, the element defined by the statement "including a..." does not exclude the existence of additional identical elements in the process, method, article or device including the said element.

[0342] The steps of the methods or algorithms described in connection with the embodiments disclosed herein may be implemented directly in hardware, in software modules executed by a processor, or in a combination thereof. The software modules may be disposed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0343] The foregoing description of the disclosed embodiments enables those skilled in the art to make or use the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Thus, the present application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for detecting blood vessel protrusions, characterized in that, the method includes: After obtaining the CTA image of the angiography collected for the target body part, using the CTA image to construct the three-dimensional image data of the target body part; Extract spherical distribution characterization data from the CTA image. The extraction process of the spherical distribution characterization data includes: for any pixel in the CTA image, determine the Hessian matrix corresponding to the pixel. The Hessian matrix is used to represent the image gray change rate of the pixel, perform eigenvalue calculation processing on the Hessian matrix to obtain the matrix eigenvalue corresponding to the pixel. The matrix eigenvalue is used to represent the image gray change degree of the pixel. According to the matrix eigenvalue, determine the three-dimensional shape feature DSI value of the pixel; use the DSI values of all pixels in the CTA image to construct the spherical distribution characterization data; According to the three-dimensional image data, the spherical distribution characterization data, and a pre-constructed blood vessel protrusion detection model, determine the blood vessel protrusion detection result of the target body part; The determination process of the blood vessel protrusion detection result includes: determining at least one data block to be used and at least one neighboring data block of each data block to be used from the three-dimensional image data. For any data block to be used, at least one neighboring data block of the data block to be used is used to perform image semantic supplementation for the blood vessels cut off in the data block to be used; determine the spherical distribution data blocks corresponding to each data block to be used from the spherical distribution characterization data; according to each data block to be used, at least one neighboring data block of each data block to be used, the spherical distribution data blocks corresponding to each data block to be used, and a pre-constructed blood vessel protrusion detection model, determine the blood vessel protrusion detection result of each data block to be used; according to the blood vessel protrusion detection results of the at least one data block to be used, determine the blood vessel protrusion detection result of the target body part; The determination process of the three-dimensional image data of the target body part includes: according to the CTA image and a preset numerical threshold, construct a three-dimensional image to be processed. The preset numerical threshold is determined based on the air density, and there is no CT value for describing the air in the three-dimensional image to be processed; according to the largest connected region in the three-dimensional image to be processed, determine the part description data for describing the target body part; perform isotropic resampling processing on the part description data to obtain the image data to be used; according to the image data to be used, determine the three-dimensional image data of the target body part.

2. The method according to claim 1, characterized in that, the part description data includes at least one first pixel in the first direction, at least one second pixel in the second direction, and at least one third pixel in the third direction; The performing isotropic resampling processing on the part description data to obtain the image data to be used includes: Performing linear interpolation processing on the at least one first pixel according to a preset pixel pitch to obtain a first pixel sequence; Perform linear interpolation processing on the at least one second pixel according to a preset pixel pitch to obtain a second pixel sequence; Perform linear interpolation processing on the at least one third pixel according to a preset pixel pitch to obtain a third pixel sequence; Construct the image data to be used by using the first pixel sequence, the second pixel sequence, and the third pixel sequence.

3. The method according to claim 1, wherein, the method further includes: Obtain a mask of the part to be used; wherein, the mask of the part to be used is used to represent the position of the target body part in the three-dimensional image data; The extracting the spherical distribution characterization data from the CTA image includes: Determine the spherical distribution characterization data according to the CTA image and the mask of the part to be used.

4. The method according to claim 3, wherein, The determining the spherical distribution characterization data according to the CTA image and the mask of the part to be used includes: Extract the spherical distribution data to be processed from the CTA image; Determine the spherical distribution characterization data according to the mask of the part to be used and the spherical distribution data to be processed.

5. The method according to claim 1, wherein, The extracting the spherical distribution characterization data from the CTA image includes: Extract the spherical distribution characterization data from the three-dimensional image data.

6. The method according to claim 1, wherein, the method further includes: Extract vascular characterization data from the three-dimensional image data; The determining at least one data block to be used from the three-dimensional image data includes: Determine at least one data block to be used from the vascular characterization data.

7. The method according to claim 6, wherein, the method further includes: Perform normalization processing on the vascular characterization data and the spherical distribution characterization data respectively to obtain vascular normalization data and spherical distribution normalization data; The determining at least one data block to be used from the vascular characterization data includes: Determine at least one data block to be used from the vascular normalization data; The determining the spherical distribution data blocks corresponding to the respective data blocks to be used from the spherical distribution characterization data includes: Determine the spherical distribution data blocks corresponding to the respective data blocks to be used from the spherical distribution normalization data.

8. The method according to claim 1, wherein, the vascular bulge detection model includes: a feature extraction module, a segmentation processing module, a feature splicing module, and a data prediction module; The number of the data blocks to be used is N; The process of determining the vascular bulge detection result of the nth data block to be used includes: Determine the feature extraction result of the nth data block to be used according to the nth data block to be used and the feature extraction module; wherein, n is a positive integer, n ≤ N, and N is a positive integer; Input the spherical distribution data block corresponding to the nth data block to be used into the segmentation processing module to obtain the spherical region segmentation result corresponding to the nth data block to be used output by the segmentation processing module; Input the feature extraction result of the nth data block to be used and the spherical region segmentation result corresponding to the nth data block to be used into the feature splicing module, and obtain the splicing feature to be processed output by the feature splicing module; Input the splicing feature to be processed into the data prediction module, and obtain the blood vessel bulge detection result of the nth data block to be used output by the data prediction module.

9. A blood vessel bulge detection device Characterized in that The device includes: An image construction unit, configured to, after acquiring an angiography CTA image collected for a target body part, use the CTA image to construct three-dimensional image data of the target body part; A spherical extraction unit, configured to extract spherical distribution characterization data from the CTA image; A bulge detection unit, configured to determine the blood vessel bulge detection result of the target body part according to the three-dimensional image data, the spherical distribution characterization data, and a pre-constructed blood vessel bulge detection model; The spherical extraction unit is specifically configured to: for any pixel in the CTA image, determine the Hessian matrix corresponding to the pixel, where the Hessian matrix is used to represent the image gray change rate of the pixel, perform eigenvalue calculation processing on the Hessian matrix to obtain the matrix eigenvalue corresponding to the pixel, where the matrix eigenvalue is used to represent the image gray change degree of the pixel, and determine the three-dimensional shape feature DSI value of the pixel according to the matrix eigenvalue; use the DSI values of all pixels in the CTA image to construct the spherical distribution characterization data; The bulge detection unit is specifically configured to: determine at least one data block to be used and at least one neighboring data block of each data block to be used from the three-dimensional image data, and for any data block to be used, at least one neighboring data block of the data block to be used is used to perform image semantic supplementation for the blood vessels cut off in the data block to be used; determine the spherical distribution data blocks corresponding to each data block to be used from the spherical distribution characterization data; determine the blood vessel bulge detection results of each data block to be used according to each data block to be used, at least one neighboring data block of each data block to be used, the spherical distribution data blocks corresponding to each data block to be used, and a pre-constructed blood vessel bulge detection model; determine the blood vessel bulge detection result of the target body part according to the blood vessel bulge detection results of the at least one data block to be used; The image construction unit is specifically configured to: construct a three-dimensional image to be processed according to the CTA image and a preset numerical threshold, where the preset numerical threshold is determined according to air density, and the CT value for describing the air does not exist in the three-dimensional image to be processed; determine part description data for describing the target body part according to the largest connected region in the three-dimensional image to be processed; perform isotropic resampling processing on the part description data to obtain image data to be used; determine the three-dimensional image data of the target body part according to the image data to be used.

10. A blood vessel bulge detection device Characterized in that Includes: A memory, a processor, and a computer program stored on the memory and executable on the processor, wherein when the processor executes the computer program, the method for detecting blood vessel bulges according to any one of claims 1-8 is implemented.

11. A computer-readable storage medium, characterized in that, instructions are stored in the computer-readable storage medium, and when the instructions are run on a terminal device, the terminal device is caused to execute the method for detecting blood vessel bulges according to any one of claims 1-8.

12. A computer program product, characterized in that, when the computer program product is run on a terminal device, the terminal device is caused to execute the method for detecting blood vessel bulges according to any one of claims 1-8.

Images