A blood vessel segmentation method, device and computer readable medium

Abstract

The embodiment of the present invention discloses a blood vessel segmentation method, device and computer-readable medium, the method comprising: extracting image features from a CT image to be tested to obtain a multi-scale feature image; marking the blood vessels in the CT image to be tested with structural points to obtain a structural point map, wherein the structural points are used to indicate the endpoints of the blood vessels and the intersections of the blood vessel branches; then extracting structural point features from the structural point map to obtain a structural point feature map, wherein the structural point feature map is used to indicate the positional features of the structural points; finally, based on the multi-scale feature map and the structural point feature map, segmenting the blood vessels to obtain a blood vessel segmentation result. Thus, by respectively extracting image features and structural point features from the CT image to be tested, slender and criss-crossing blood vessels can be accurately identified, which is beneficial to the segmentation of blood vessels and improves the integrity of blood vessel segmentation.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a blood vessel segmentation method, device and computer-readable medium. Background Art

[0002] The living body is a very complex system. The distribution of blood vessels in each organ is also intricate. Blood vessel detection is the basis for the diagnosis of organ diseases. Therefore, it is particularly important to accurately segment the blood vessels in the organs.

[0003] Traditional vascular segmentation methods mainly use the trained segmentation model to predict the CT image to be tested, thereby obtaining the vascular segmentation results; however, since some blood vessels in the internal organs are not only slender blood vessels but also crisscross with other blood vessels in distribution, the existing vascular segmentation methods often cause incomplete segmentation due to misidentification of blood vessels. Therefore, for slender blood vessels with complex distribution, it is urgent to provide an effective vascular segmentation method to improve the integrity and accuracy of vascular segmentation. Summary of the invention

[0004] In order to solve the above problems in the prior art, the embodiments of the present invention provide a blood vessel segmentation method, device and computer-readable medium, which can accurately identify blood vessels, thereby improving the integrity of blood vessel segmentation.

[0005] To achieve the above-mentioned purpose, according to a first aspect of an embodiment of the present invention, a blood vessel segmentation method is provided, the method comprising: performing image feature extraction on a computerized tomography (CT) image to be tested to obtain a multi-scale feature map; performing structural point marking on the blood vessels in the CT image to be tested to obtain a structural point map, wherein the structural points are used to indicate the endpoints of the blood vessels and the intersections of the blood vessel branches; performing structural point feature extraction on the structural point map to obtain a structural point feature map, wherein the structural point feature map is used to indicate the positional features of the structural points; and performing segmentation processing on the blood vessels based on the multi-scale feature map and the structural point feature map to obtain a blood vessel segmentation result.

[0006] Optionally, the extracting structure point features from the structure point map to obtain a structure point feature map includes: for any structure point in the structure point map: performing position encoding on the structure point to obtain a structure point position map; performing association encoding on the structure point to obtain a structure point association map, wherein the structure point association map is used to indicate the positional relationship between adjacent structure points; and determining the structure point position map and the structure point association map as the structure point feature map.

[0007] Optionally, the segmentation processing of the blood vessels based on the multi-scale feature map and the structure point feature map to obtain the blood vessel segmentation result includes: extracting the structure point feature map to obtain a blood vessel region, wherein the blood vessel region includes structure points; registering the structure points in the blood vessel region using a blood vessel structure point template map to obtain a registration function; performing affine transformation processing on the multi-scale feature map and the structure point feature map respectively using the registration function; and segmenting the blood vessels based on the transformed multi-scale feature map and structure point feature map to obtain the blood vessel segmentation result.

[0008] Optionally, performing structure point feature extraction on the structure point graph to obtain a structure point feature graph includes: classifying all structure points in the structure point graph to obtain structure points of several levels; and performing feature extraction on structure points of the same level to obtain a structure point feature graph.

[0009] Optionally, the blood vessel is the hepatic vein and / or the portal vein.

[0010] Optionally, the blood vessel segmentation processing is performed based on the multi-scale feature map and the structure point feature map to obtain a blood vessel segmentation result, including: acquiring an original CT image; performing image feature extraction and blood vessel structure point feature extraction on the original CT image, respectively, to obtain an original multi-scale feature map and an original structure point feature map; using the original multi-scale feature map and the original structure point feature map together as training samples, and using multiple of the training samples to perform model training to obtain a segmentation model; using the segmentation model to perform segmentation processing on the multi-scale feature map and the structure point feature map to obtain a blood vessel segmentation result.

[0011] To achieve the above-mentioned purpose, according to a second aspect of an embodiment of the present invention, a blood vessel segmentation device is further provided, the device comprising: an image extraction module, used to extract image features from a CT image to be tested, and obtain a multi-scale feature map; a structure point marking module, used to mark structure points of blood vessels in the CT image to be tested, and obtain a structure point map, wherein the structure points are used to indicate the endpoints of the blood vessels and the intersections of the blood vessel branches; a structure point extraction module, used to extract structure point features from the structure point map, and obtain a structure point feature map, wherein the structure point feature map is used to indicate the position features of the structure points; and a segmentation processing module, used to perform segmentation processing on the blood vessels based on the multi-scale feature map and the structure point feature map, and obtain a blood vessel segmentation result.

[0012] Optionally, the structure point extraction module includes: an encoding unit, used to: for any structure point in the structure point map: perform position encoding on the structure point to obtain a structure point position map; perform association encoding on the structure point to obtain a structure point association map, wherein the structure point association map is used to indicate the positional relationship between adjacent structure points; and a determination unit, used to determine the structure point position map and the structure point association map as a structure point feature map.

[0013] To achieve the above-mentioned purpose, according to the third aspect of an embodiment of the present invention, an electronic device is also provided, which includes one or more processors; a memory for storing one or more programs, and when the one or more programs are executed by the one or more processors, the one or more processors implement the blood vessel segmentation method as described in the first aspect.

[0014] To achieve the above objective, according to a fourth aspect of an embodiment of the present invention, there is further provided a computer-readable medium having a computer program stored thereon, wherein when the program is executed by a processor, the blood vessel segmentation method as described in the first aspect is implemented.

[0015] Compared with the prior art, the embodiment of the present invention provides a blood vessel segmentation method, device and computer-readable medium. First, image feature extraction is performed on the CT image to be tested to obtain a multi-scale feature image; and structural point marking is performed on the blood vessels in the CT image to be tested to obtain a structural point map, wherein the structural points are used to indicate the end points of the blood vessels and the intersections of the blood vessel branches; then, structural point feature extraction is performed on the structural point map to obtain a structural point feature map, and the structural point feature map is used to indicate the positional features of the structural points; finally, based on the multi-scale feature map and the structural point feature map, the blood vessels are segmented to obtain a blood vessel segmentation result. Therefore, by performing image feature extraction and structural point feature extraction on the CT image to be tested, respectively, slender and criss-crossing blood vessels can be accurately identified, which is beneficial to the segmentation of blood vessels, improves the integrity of blood vessel segmentation, and effectively solves the problem of incomplete segmentation due to the complex distribution of slender blood vessels in the prior art.

[0016] It should be understood that the teachings of the present invention are not required to achieve all of the beneficial effects described above, but specific technical solutions can achieve specific technical effects, and other embodiments of the present invention can also achieve beneficial effects not mentioned above. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The accompanying drawings are used to better understand the present invention and do not constitute an improper limitation of the present invention. In the accompanying drawings, the same or corresponding reference numerals represent the same or corresponding parts.

[0018] Figure 1 is a schematic flow chart of a blood vessel segmentation method according to an embodiment of the present invention;

[0019] Figure 2 is a schematic flow chart of a blood vessel segmentation method according to another embodiment of the present invention;

[0020] Figure 3 is a schematic flow chart of a blood vessel segmentation method according to yet another embodiment of the present invention;

[0021] Figure 4 It is a flowchart of a blood vessel segmentation method according to an application example of the present invention;

[0022] Figure 5 It is a structural point diagram in the application embodiment of the present invention;

[0023] Figure 6 FIG. 4 is a schematic block diagram of a blood vessel segmentation device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0024] The following is a description of exemplary embodiments of the present invention in conjunction with the accompanying drawings, including various details of the embodiments of the present invention to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present invention. Similarly, for clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0025] like Figure 1 As shown, a schematic flow chart of a blood vessel segmentation method according to an embodiment of the present invention. A blood vessel segmentation method includes the following operation flow: S101, extracting image features from a CT image to be tested to obtain a multi-scale feature map; S102, marking structure points of blood vessels in the CT image to be tested to obtain a structure point map, wherein the structure points are used to indicate the endpoints of the blood vessels and the intersections of the blood vessel branches; S103, extracting structure point features from the structure point map to obtain a structure point feature map, wherein the structure point feature map is used to indicate the positional features of the structure points; S104, segmenting the blood vessels based on the multi-scale feature map and the structure point feature map to obtain a blood vessel segmentation result.

[0026] In S101, specifically, the trained image model is used to extract image features from the CT image to be tested, and a multi-scale feature image is obtained. The CT image to be tested refers to a CT image of a specific part, for example, the CT image to be tested may be a CT image of the liver, a CT image of the stomach, a CT image of the brain, etc. Here, the CT image to be tested is not specifically limited.

[0027] For example, multiple liver CT images and liver feature images are obtained; the liver CT images are used as training samples, and the multiple training samples and liver feature images are input into the convolutional neural network to perform model training to obtain training results, and the model parameters are adjusted based on the training results to obtain an image model. The trained image model is then used to extract image features from the liver CT image to be tested to obtain a multi-scale liver feature image. Here, the multi-scale in the multi-scale liver feature image is used to indicate multiple liver feature images with different resolutions. The liver feature image includes not only the vascular features of the liver, but also other tissue features of the liver.

[0028] In S102, specifically, a designated blood vessel in the CT image to be tested is determined, and the end points and the intersection points of the branches of the designated blood vessel are marked with structural points to obtain a structural point map. Thus, by marking the end points and the intersection points of the branches of the blood vessel with structural points, it is helpful to determine the trend of the blood vessel and improve the integrity of the blood vessel segmentation.

[0029] In S103, the structure point feature extraction may be performed for each structure point in the structure point map to obtain a structure point feature map, in which case one structure point corresponds to one structure point feature map. The structure point feature extraction may also be performed for multiple structure points in the structure point map at the same time to obtain a structure point feature map; for example, all structure points in the structure map are graded to obtain structure points of several grades; feature extraction is performed for multiple structure points of the same grade to obtain a structure point feature map, in which case structure points of the same grade correspond to one structure point feature map.

[0030] It should be noted that after performing structural point feature extraction on all structural points in the structural point graph, multiple structural point feature graphs are obtained.

[0031] In S104, an original CT image is obtained; image features are extracted from the original CT image using an image model to obtain an original multi-scale feature map; structural point marking is performed on the intersection points of the terminal end points and branches of the designated blood vessels in the original CT image to obtain an original structural point map; structural point features are extracted from the original structural point map to obtain an original structural point feature map; the original multi-scale feature map and the original structural point feature map are used as training samples, and the training samples are input into the segmentation network for model training to obtain training results, and the parameters of the model are adjusted based on multiple training results to obtain a segmentation model. Therefore, on the basis of learning the characteristics of the vascular image, combined with the learning of the location characteristics of the vascular structural points, the segmentation model can be accurately trained, which is beneficial to the segmentation of the vascular image and improves the accuracy of vascular segmentation.

[0032] Afterwards, the multi-scale feature map and the structure point feature map are input into the trained segmentation model, and the trained segmentation model is used to predict the blood vessels to obtain a blood vessel segmentation image; thus, the structure point feature map and the multi-scale feature map are used in combination for blood vessel segmentation, thereby improving the accuracy of blood vessel segmentation.

[0033] Here, the segmentation network can be a fully convolutional neural network consisting of 10 cascaded residual units.

[0034] It should be noted that the blood vessel segmentation method of this embodiment is applicable to the segmentation of any blood vessel, even to thin and long blood vessels that crisscross with other blood vessels, such as the segmentation of the portal vein and the hepatic vein in the liver.

[0035] The embodiment of the present invention marks the structure points of the CT image to be tested, and extracts the structure point features of the marked structure point map to obtain a structure point feature map; then, the blood vessels are segmented based on the structure point feature map and the multi-scale feature map to obtain a blood vessel segmentation image. In this way, the blood vessels can be accurately segmented, solving the problem of incomplete segmentation in the prior art due to the complex distribution of slender blood vessels.

[0036] like Figure 2 As shown, a schematic flow chart of a blood vessel segmentation method according to another embodiment of the present invention. This embodiment is obtained by further optimizing the above-mentioned embodiment. A blood vessel segmentation method, the method at least includes the following operation flow: S201, image feature extraction of the CT image to be tested is performed to obtain a multi-scale feature map; S202, structural point marking of the blood vessels in the CT image to be tested is performed to obtain a structural point map, and the structural points are used to indicate the endpoints of the blood vessels and the intersections of the blood vessel branches; S203, for any structural point in the structural point map: position encoding of the structural points is performed to obtain a structural point position map; correlation encoding of the structural points is performed to obtain a structural point association map, and the structural point association map is used to indicate the positional relationship between adjacent structural points; S204, the structural point position map and the structural point association map are determined as structural point feature maps; S205, based on the multi-scale feature map and the structural point feature map, the blood vessels are segmented to obtain a blood vessel segmentation result.

[0037] Among them, the specific implementation process of S201, S202 and S205 is the same as Figure 1 The specific implementation processes of S101, S102 and S104 in the illustrated embodiment are similar and will not be described in detail here.

[0038] In S203 and S204, the structure point location map is used to represent the location information of the structure point. The structure location map may include the location information of one structure point or may include the location information of multiple structure points. The structure point location map may be a Gaussian distribution heat map centered on the structure point or a position coordinate map. The Gaussian distribution heat map centered on the structure point is used as an example for explanation. The pixel value at the corresponding position of the structure point is 1, and the pixel values ​​around the structure point are 1. The regular decreasing pattern is used to represent the location information of the structural points through probability distribution.

[0039] The structural point association diagram is used to represent the positional relationship between adjacent structural points. The structural point association diagram may include the positional relationship for one structural point or the positional relationship for multiple structural points. Specifically, the structural point association diagram may be a unit vector diagram of adjacent structural points or an oblique variance matrix diagram. The unit vector diagram of adjacent structural points includes an X-axis coordinate diagram, a Y-axis coordinate diagram, and a Z-axis coordinate diagram.

[0040] This embodiment obtains a structure point position map and a structure point association map by extracting structure point features; thereby, the position and trend of the structure points can be accurately located, which is beneficial to the segmentation of blood vessel images and improves the accuracy of blood vessel image segmentation.

[0041] like Figure 3 As shown, a schematic flow chart of a blood vessel segmentation method according to another embodiment of the present invention. This embodiment is obtained by further optimizing the aforementioned embodiment. A blood vessel segmentation method, the method comprising the following operation procedures: S301, extracting image features from a CT image to be tested to obtain a multi-scale feature map; S302, marking structural points of blood vessels in the CT image to be tested to obtain a structural point map, wherein the structural points are used to indicate the endpoints of the blood vessels and the intersections of the blood vessel branches; S303, classifying all structural points in the structural point map to obtain structural points of several levels; S304, for structural points of the same level in the structural point map: position encoding all structural points of the same level to obtain a structural point position map; correlation encoding all structural points of the same level to obtain a structural point map. Point association map, the structure point association map is used to indicate the positional relationship between adjacent structure points; S305, the structure point position map and the structure point association map are determined as the structure point feature map; S306, the structure point feature map is extracted and processed to obtain the blood vessel area, and the blood vessel area includes the structure points; S307, the structure points in the blood vessel area are registered using the blood vessel structure point template map to obtain the registration function; S308, the multi-scale feature map and the structure point feature map are affine transformed using the registration function; S309, the blood vessel is segmented based on the transformed multi-scale feature map and the structure point feature map to obtain the blood vessel segmentation result.

[0042] The specific implementation process of S301 and S302 is the same as Figure 1 The specific implementation processes of S101 and S102 in the illustrated embodiment are similar and will not be described in detail here.

[0043] In S303 to S304, the structural point at the root of the blood vessel is determined as the root node; for any structural point on a blood vessel branch: determine the distance between the structural point and the root node, and sort the multiple distances from small to large, and determine the level of the structural point according to the sorting result, for example, the structural point corresponding to the minimum distance is a level 1 structural point, and so on, so as to obtain structural points of different levels. On the same branch, the closer the distance of the structural point to the root node, the lower the level of the structural point, and the farther the distance of the structural point from the root node, the higher the level of the structural point. Then, all structural points of the same level on different branches of the blood vessel are position-coded to obtain a structural point position map; at this time, the structural point position map includes the position information of multiple structural points of the same level. All structural points of the same level on different branches of the blood vessel are associated with each other to obtain a structural point association map, and at this time, the structural point association map includes the position relationship information corresponding to each structural point of the same level. Therefore, by classifying all the structural points in the structural point map and extracting structural point features for the structural points of the same level at the same time, it is possible to extract features of all the structural points in the structural point map in an orderly manner, avoiding repeated extraction and feature omission, improving the speed of structural point feature extraction, and facilitating the later diagnosis of blood vessels by doctors.

[0044] In S306 to S308, the structure point template map is obtained in advance, for example, CT images of multiple persons are obtained, and the structure points of the designated blood vessels in each person's CT image are calibrated; and the calibrated structure points are clustered to obtain multiple structure point template maps. Then, the blood vessel region is extracted based on the structure points in the structure point feature map; the structure point template map with the highest similarity to the blood vessel region is selected from the multiple structure point template maps; and the structure points in the blood vessel region are registered using the selected structure point template map to obtain a registration function. Finally, the multi-scale feature map, the structure point position map, and the structure point association map are affine transformed using the registration function to obtain the transformed scale feature map, the structure point position map, and the structure point association map. Therefore, by registering the structure points in the blood vessel region and using the registration function to perform affine transformation on the multi-scale feature map, the structure point position map, and the structure point association map, it is beneficial to blood vessel segmentation and improves the accuracy of blood vessel segmentation.

[0045] In S309 , the trained segmentation model is used to perform blood vessel segmentation processing on the transformed multi-scale feature map and the structural feature map corresponding to the CT image to be tested, so as to obtain a blood vessel segmentation image.

[0046] The segmentation model is obtained by the following method: obtaining the original CT image; using the image model to extract the image features of the original CT image to obtain the original multi-scale feature map; marking the intersection points of the end points and branches of the blood vessels in the original CT image respectively to obtain the original structure point map; extracting the structure point features of the original structure point map to obtain the original structure point feature map. All the structure points in the original structure point map are divided into levels to obtain structure points of several levels; position encoding and association encoding are performed on the structure points of the same level in the structure point map respectively to obtain the original structure point position map and the original structure point association map. The original structure point feature map is extracted to obtain the original blood vessel region with structure points; the structure points in the original blood vessel region are registered by selecting the blood vessel structure point template map to obtain the registration function; the original multi-scale feature map, the original structure point position map and the original structure point association map are affine transformed by the registration function respectively; the transformed original multi-scale feature map, the original structure point position map and the original structure point association map are input into the segmentation network to perform model training to obtain the segmentation result, and the model parameters are adjusted based on multiple segmentation results to obtain the segmentation model. The greater the difference between blood vessels of different patients, the more difficult it is to train the segmentation model. Therefore, registration can reduce the difference between different blood vessels and improve the accuracy of segmentation model training.

[0047] It should be noted that in order to enable the segmentation network to effectively learn the image features and structure point features, during model training, the original multi-scale feature map, the original structure point position map, and the original structure point association map are all adjusted to the same size as the original CT image.

[0048] This embodiment uses a vascular structure point template map to align the structure points in the vascular area to obtain a registration function; then uses the registration function to perform affine transformation processing on the multi-scale feature map, the structure point position map and the structure point association map, thereby aligning the vascular image to a unified direction according to the vascular structure points, improving the accuracy of vascular segmentation, and thus providing a basis for tumor diagnosis and surgical planning.

[0049] It should be understood that in the embodiments of the present invention, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0050] The blood vessel segmentation method of the above embodiment will be described in detail below in conjunction with specific application scenarios.

[0051] Since both the hepatic vein and the portal vein are long and thin blood vessels, and they are distributed in a crisscross pattern in the liver, in order to effectively segment the hepatic vein and the portal vein, the blood vessel segmentation method of this embodiment is used to segment the hepatic vein and the portal vein. Figure 4 As described below.

[0052] like Figure 4 As shown, the liver CT image to be tested is input into the image model for image feature extraction, and a multi-scale feature map is output; the liver CT image to be tested is marked with structural points to obtain a hepatic vein structural point map and a portal vein structural point map. The hepatic vein structural point map is positionally encoded and associated with each other to obtain a structural point position map and a structural point association map corresponding to the hepatic vein; the portal vein structural point map is positionally encoded and associated with each other to obtain a structural point position map and a structural point association map corresponding to the portal vein. Afterwards, the multi-scale feature map, the structural point position map and the structural point association map corresponding to the hepatic vein, and the structural point position map and the structural point association map corresponding to the portal vein are input into the segmentation model for vascular segmentation processing to obtain a vascular segmentation result. Furthermore, after the structural point feature extraction, structural point registration processing is required to obtain the multi-scale feature map, structural point position map and structural point association map corresponding to the transformed hepatic vein, and the multi-scale feature map, structural point position map and structural point association map corresponding to the transformed portal vein; then the multi-scale feature map, structural point position map and structural point association map corresponding to the transformed hepatic vein, and the multi-scale feature map, structural point position map and structural point association map corresponding to the transformed portal vein are input into the trained segmentation model, and the segmentation processing is performed to obtain the blood vessel segmentation result. Here, the blood vessel segmentation result is used to indicate the hepatic vein segmentation image and the portal vein segmentation image.

[0053] like Figure 5 As shown, through Figure 5 The portal vein structure point map and the hepatic vein structure point map can be clearly displayed, wherein the gray points represent the hepatic vein structure points and the black points represent the portal vein structure points.

[0054] Segmentation model training includes two stages: image preprocessing and model training.

[0055] Image preprocessing includes image feature extraction, structural point marking, structural point feature extraction, and structural registration. Specifically: obtain a liver CT image, use the trained image model to extract image features from the liver CT image, and obtain a multi-scale feature map. Mark the structural points of the terminal endpoints and branch intersections of the hepatic veins in the liver CT image to obtain a hepatic vein structural point map. Obtain the root node in the hepatic vein structural point map; determine the structural points adjacent to the root node as level 1 structural points, determine the structural points next adjacent to the root node as level 2 structural points, and so on to determine the level of the hepatic vein structural points. By Figure 5It can be seen that there are 2 level 1 structural points, 3 level 2 structural points, and 7 level 3 structural points in the hepatic vein. Afterwards, the structural points of the same level in the hepatic vein are positionally encoded and associated; the structural point position map and structural point association map of the hepatic vein are obtained. Finally, the hepatic vein region including the hepatic vein structural points is extracted from the hepatic vein structural point map, and a hepatic vein structural point template map similar to the hepatic vein region is selected from multiple hepatic vein structural point template maps, and the structural points in the hepatic vein region are registered using the selected hepatic vein structural point template map to obtain the registration function; then, the multi-scale feature map, the hepatic vein structural point position map, and the hepatic vein structural point association map are affine transformed using the registration function to obtain the transformed multi-scale feature map, the hepatic vein structural point position map, and the hepatic vein structural point association map. Similarly, the portal vein in the liver CT image is preprocessed in the same way to obtain the transformed multi-scale feature map, the portal vein structural point position map, and the portal vein structural point association map.

[0056] The specific process of model training is as follows: the transformed multi-scale feature map corresponding to the hepatic vein, the hepatic vein structure point position map and the hepatic vein structure point association map are used as the first training sample, and the transformed multi-scale feature map corresponding to the portal vein, the portal vein structure point position map and the portal vein structure point association map are used as the second training sample; then the first training sample is input into the segmentation network, and the model is trained to obtain the first training result; the second catenary sample is input into the segmentation network, and the model is trained to obtain the second training result; the model parameters are adjusted based on multiple first training results and multiple second training results to obtain a segmentation model.

[0057] Therefore, the blood vessel segmentation method of this embodiment can realize automatic identification of the hepatic vein and the portal vein, thereby improving the accuracy of the identification of the hepatic vein and the portal vein, so that the hepatic vein and the portal vein can be completely segmented, and the problem of misidentification and missegmentation caused by the complex distribution of slender blood vessels in the prior art is solved.

[0058] like Figure 6 The figure is a schematic block diagram of a blood vessel segmentation device according to an embodiment of the present invention. A blood vessel segmentation device, the device 600 comprises: an image extraction module 601, used to extract image features from a CT image to be tested, and obtain a multi-scale feature map; a structure point marking module 602, used to mark structure points of blood vessels in the CT image to be tested, and obtain a structure point map, wherein the structure points are used to indicate the end points of the blood vessels and the intersection points of the blood vessel branches; a structure point extraction module 603, used to extract structure point features from the structure point map, and obtain a structure point feature map, wherein the structure point feature map is used to indicate the position features of the structure points; a segmentation processing module 604, used to segment the blood vessels based on the multi-scale feature map and the structure point feature map, and obtain a blood vessel segmentation result.

[0059] In an optional embodiment, the structure point extraction module 602 includes: a coding unit, which is used to: for any structure point in the structure point map: perform position coding on the structure point to obtain a structure point position map; perform association coding on the structure point to obtain a structure point association map, and the structure point association map is used to indicate the positional relationship between adjacent structure points; and a determination unit, which is used to determine the structure point position map and the structure point association map as the structure point feature map.

[0060] In an optional embodiment, the segmentation processing module 604 includes: an extraction processing unit, which is used to extract the structure point feature map to obtain a blood vessel area, and the blood vessel area includes structure points; a registration unit, which is used to use the blood vessel structure point template map to register the structure points in the blood vessel area to obtain a registration function; a transformation unit, which is used to use the registration function to perform affine transformation processing on the multi-scale feature map and the structure point feature map respectively; a segmentation processing unit, which is used to segment the blood vessel based on the transformed multi-scale feature map and the structure point feature map to obtain a blood vessel segmentation result.

[0061] In an optional embodiment, the structure point extraction module 603 includes: a level division unit, used to divide all structure points in the structure point map into levels to obtain structure points of several levels; a feature extraction unit, used to extract features for structure points of the same level to obtain a structure point feature map.

[0062] In alternative embodiments, the blood vessel is the hepatic vein and / or the portal vein.

[0063] In an optional embodiment, the segmentation processing module 604 includes: an acquisition unit, used to acquire the original CT image; a feature extraction unit, used to perform image feature extraction and vascular structure point feature extraction on the original CT image, respectively, to obtain an original multi-scale feature map and an original structure point feature map; a model training unit, used to use the original multi-scale feature map and the original structure point feature map as training samples, and use multiple training samples to perform model training to obtain a segmentation model; a segmentation processing unit, used to use the segmentation model to perform segmentation processing on the multi-scale feature map and the structure point feature map to obtain a vascular segmentation result.

[0064] The above device can execute the blood vessel segmentation method provided in an embodiment of the present invention, and has the corresponding functional modules and beneficial effects of executing the blood vessel segmentation method. For technical details not described in detail in this embodiment, please refer to the blood vessel segmentation method provided in an embodiment of the present invention.

[0065] According to yet another embodiment of the present invention, an electronic device is provided, comprising: one or more processors; a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the blood vessel segmentation method provided by the above embodiment of the present invention.

[0066] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0067] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are only schematic. For example, the division of units is only a logical function division. There may be other division methods in actual implementation, such as: multiple units or components can be combined, or can be integrated into another device, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of devices or units can be electrical, mechanical or other forms.

[0068] The units described above as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units; they may be located in one place or distributed on multiple network units; some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0069] In addition, all functional units in the embodiments of the present application may be integrated into one processing unit, or each unit may be a separate unit, or two or more units may be integrated into one unit; the above-mentioned integrated units may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0070] A person of ordinary skill in the art can understand that all or part of the steps of implementing the above method embodiment can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above method embodiment; and the aforementioned storage medium includes: mobile storage medium, read-only memory (ROM), magnetic disk or optical disk, etc., various media that can store program codes.

[0071] Alternatively, if the above-mentioned integrated unit of the present application is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this embodiment of the present application can be essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a storage medium and includes a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform the following operation steps: S101, image feature extraction is performed on the CT image to be tested to obtain a multi-scale feature map; S102, structural point marking is performed on the blood vessels in the CT image to be tested to obtain a structural point map, and the structural points are used to indicate the endpoints of the blood vessels and the intersections of the blood vessel branches; S103, structural point feature extraction is performed on the structural point map to obtain a structural point feature map, and the structural point feature map is used to indicate the positional features of the structural points; S104, based on the multi-scale feature map and the structural point feature map, the blood vessels are segmented to obtain a blood vessel segmentation result.

[0072] The aforementioned storage media include: mobile storage media, ROM, magnetic disks, optical disks, and other media that can store program codes.

[0073] The above are only specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A blood vessel segmentation method, characterized in that: include: Performing image feature extraction on the electronic computer tomography (CT) image to be tested to obtain a multi-scale feature map; Marking the blood vessels in the CT image to be tested with structural points to obtain a structural point map, wherein the structural points are used to indicate the endpoints of the blood vessels and the intersections of the blood vessel branches; Extracting structure point features from the structure point graph to obtain a structure point feature graph, wherein the structure point feature graph is used to indicate position features of the structure points; Based on the multi-scale feature map and the structure point feature map, segment the blood vessel to obtain a blood vessel segmentation result; The step of extracting structure point features from the structure point graph to obtain a structure point feature graph includes: obtaining a structure point position graph and a structure point association graph for any structure point in the structure point graph, wherein the structure point position graph is used to represent the position information of the structure point, and the structure point association graph is used to indicate the position relationship between adjacent structure points; The structure point position map and the structure point association map are determined as a structure point feature map.

2. The method according to claim 1, characterized in that The step of obtaining a structure point location diagram and a structure point association diagram for any structure point in the structure point diagram includes: The structural points are position-coded to obtain a structural point position map; and the structural points are relevance-coded to obtain a structural point relevance map.

3. The method according to claim 1 or 2, characterized in that: The segmenting of the blood vessels based on the multi-scale feature map and the structural point feature map to obtain a blood vessel segmentation result includes: Extracting the structure point feature map to obtain a blood vessel region, wherein the blood vessel region includes the structure point; Using the blood vessel structure point template image to register the structure points in the blood vessel region to obtain a registration function; Using the registration function to perform affine transformation processing on the multi-scale feature map and the structure point feature map respectively; Based on the transformed multi-scale feature map and structure point feature map, the blood vessels are segmented to obtain the blood vessel segmentation result.

4. The method according to claim 1, characterized in that: The step of extracting structure point features from the structure point graph to obtain a structure point feature graph includes: Classifying all the structural points in the structural point graph into different levels to obtain structural points of several levels; Feature extraction is performed on structural points of the same level to obtain a structural point feature map.

5. The method according to claim 1, characterized in that: The blood vessel is the hepatic vein and / or the portal vein.

6. The method according to claim 1, characterized in that The segmenting of the blood vessels based on the multi-scale feature map and the structural point feature map to obtain a blood vessel segmentation result includes: Obtaining original CT images; Performing image feature extraction and vascular structure point feature extraction on the original CT image to obtain an original multi-scale feature map and an original structure point feature map; The original multi-scale feature map and the original structure point feature map are used as training samples, and a plurality of the training samples are used to perform model training to obtain a segmentation model; The multi-scale feature map and the structure point feature map are segmented using a segmentation model to obtain a blood vessel segmentation result.

7. A blood vessel segmentation device, characterized in that: include: An image extraction module is used to extract image features from the CT image to be tested and obtain a multi-scale feature map; A structure point marking module is used to mark the structure points of the blood vessels in the CT image to be tested to obtain a structure point map, wherein the structure points are used to indicate the end points of the blood vessels and the intersection points of the blood vessel branches; A structure point extraction module is used to extract structure point features from the structure point map to obtain a structure point feature map, where the structure point feature map is used to indicate position features of structure points; A segmentation processing module, used to perform segmentation processing on the blood vessel based on the multi-scale feature map and the structural point feature map to obtain a blood vessel segmentation result; The structure point extraction module is further used to obtain a structure point position map and a structure point association map according to any structure point in the structure point map, wherein the structure point position map is used to represent the position information of the structure point, and the structure point association map is used to indicate the position relationship between adjacent structure points; The structure point extraction module is further used to determine the structure point position map and the structure point association map as a structure point feature map.

8. The device according to claim 7, characterized in that The structure point extraction module comprises: The encoding unit is used to perform position encoding on the structure points to obtain a structure point position map; perform relevance encoding on the structure points to obtain a structure point relevance map; A determining unit is used to determine the structure point position map and the structure point association map as a structure point feature map.

9. An electronic device, characterized in that: include: one or more processors; a memory for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors are enabled to implement the method according to any one of claims 1 to 6.

10. A computer readable medium having a computer program stored thereon, wherein when the program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images