A method, device, electronic device and medium for naming blood vessels based on the centerline

Abstract

The present application provides a centerline-based vascular naming method, apparatus, electronic device, and medium. The vascular naming method comprises: obtaining an initial vascular image, performing initial vascular segmentation processing on the initial vascular image to obtain an initial vascular segmentation image; performing skeletonization processing on the initial vascular segmentation image to obtain a vascular centerline image; performing point cloud feature processing on the vascular centerline image based on the vascular centerline image to extract point cloud features of the vascular centerline; inputting the point cloud features of the vascular centerline into a point cloud segmentation model to obtain a vascular centerline point cloud classification result; and naming the vascular centerline based on the vascular centerline point cloud classification result. The technical solution provided by the present application can extract point cloud features on the vascular centerline, predict the classification of the point cloud on the vascular centerline, and name the vascular centerline based on the classification result, thereby reducing interference from image background noise and improving the accuracy and efficiency of vascular naming.

Description

Technical Field

[0001] The present application relates to the field of medical imaging technology, and in particular to a method, device, electronic device, and medium for naming blood vessels based on centerlines. Background Art

[0002] Vascular diseases pose a serious threat to human health and life. Accurately naming individual vessels in images is an essential and challenging task in intelligent vascular disease screening. Accurate vessel naming provides the foundation for subsequent vessel registration and 3D reconstruction.

[0003] Currently, most vascular naming technologies use the point cloud coordinate features of the entire image using binary segmentation as input. This not only introduces background noise, reducing the segmentation accuracy of deep networks; but also requires predicting more points during inference, affecting the efficiency of subsequent vascular naming. Therefore, how to name blood vessels has become an urgent problem to be solved. Summary of the Invention

[0004] In view of this, the purpose of this application is to provide a centerline-based vascular naming method, device, electronic device and medium, which can extract point cloud features on the vascular centerline, predict the classification of the point cloud on the vascular centerline, and name the blood vessels based on the classification results, thereby reducing the interference of image background noise and improving the accuracy and efficiency of vascular naming.

[0005] This application mainly includes the following aspects:

[0006] In a first aspect, an embodiment of the present application provides a centerline-based blood vessel naming method, the blood vessel naming method comprising:

[0007] Acquiring an initial blood vessel image, and performing initial blood vessel segmentation processing on the initial blood vessel image to obtain an initial blood vessel segmentation image;

[0008] performing skeletonization processing on the blood vessel initial segmentation image to obtain a blood vessel centerline image;

[0009] Based on the blood vessel centerline image, performing point cloud feature processing on the blood vessel centerline image to extract point cloud features of the blood vessel centerline;

[0010] Inputting the point cloud features of the blood vessel centerline into a point cloud segmentation model to obtain a point cloud classification result of the blood vessel centerline;

[0011] The blood vessel is named based on the classification result of the centerline point cloud of the blood vessel.

[0012] Furthermore, the step of performing point cloud feature processing on the blood vessel centerline image based on the blood vessel centerline image to extract point cloud features of the blood vessel centerline includes:

[0013] Based on the predicted label value of each point cloud in the blood vessel centerline image, determining, among all point clouds included in the blood vessel centerline image, a point cloud where the predicted label value indicates a foreground point as a target point cloud;

[0014] Extracting the position coordinates and grayscale values of the target point cloud from the blood vessel centerline image, and determining the blood vessel radius corresponding to each target point cloud and the direction vector of each target point cloud;

[0015] For each target point cloud, the position coordinates, grayscale value, corresponding blood vessel radius and direction vector of the target point cloud are spliced to obtain the point cloud features of the target point cloud;

[0016] The combination of the point cloud features of each target point cloud is determined as the point cloud feature of the blood vessel centerline.

[0017] Furthermore, the blood vessel radius corresponding to each target point cloud is determined by the following steps:

[0018] Based on the predicted label value of each point cloud in the initial segmented blood vessel image, the initial segmented blood vessel image is inverted to obtain an inverted image;

[0019] For each target point cloud, with the target point cloud as the center of the circle, based on the preset step size, the expansion area corresponding to the target point cloud is obtained;

[0020] Determine whether the expansion area corresponding to the target point cloud intersects with the inverted image;

[0021] If not, increasing the radius of the expansion area corresponding to the target point cloud by the preset step size, updating the expansion area corresponding to the target point cloud, and determining again whether the expansion area corresponding to the target point cloud intersects with the inverted image;

[0022] If so, the radius of the blood vessel corresponding to the target point cloud is determined based on the area of the expansion region corresponding to the target point cloud.

[0023] Furthermore, the direction vector of each target point cloud is determined by the following steps:

[0024] Determining the position coordinates of the next target point cloud of each target point cloud based on the position coordinates of each target point cloud in the blood vessel centerline image;

[0025] For each target point cloud, a direction vector of the target point cloud is determined based on the position coordinates of the next target point cloud of the target point cloud and the position coordinates of the target point cloud.

[0026] Furthermore, the step of naming the blood vessel based on the classification result of the centerline point cloud of the blood vessel includes:

[0027] Based on the centerline point cloud classification results of the blood vessels, extracting the position coordinates of the target point cloud in each type of centerline point cloud classification results, and obtaining a coordinate list corresponding to each type of centerline point cloud classification results;

[0028] For each type of centerline point cloud classification result corresponding to the coordinate list, determine the intersection of the coordinate list corresponding to the type of centerline point cloud classification result and the coordinate list of each centerline in the predetermined centerline two-dimensional list;

[0029] In the intersection sets, the blood vessel represented by the coordinate list of the center line corresponding to the intersection set with the largest number of intersection points in the intersection sets is named as the classification result of the center line point cloud of this type.

[0030] Furthermore, the two-dimensional list of center lines is determined by the following steps:

[0031] Obtaining the number of target point clouds adjacent to each target point cloud in the blood vessel centerline image;

[0032] Determine the target point cloud having one adjacent target point cloud as an endpoint;

[0033] For each endpoint, determining a target intersection between a preset rectangular area centered at the endpoint and the blood vessel initial segmentation image;

[0034] In the target intersections corresponding to each endpoint, the endpoint corresponding to the target intersection with the largest number of intersection points in the target intersections is determined as the starting point;

[0035] A depth-first traversal is performed starting from the starting point, and a set of coordinate lists consisting of the position coordinates of the target point cloud of each center line is determined as a center line two-dimensional list.

[0036] In a second aspect, an embodiment of the present application further provides a centerline-based blood vessel naming device, the blood vessel naming device comprising:

[0037] an acquisition module, configured to acquire an initial blood vessel image, and perform initial blood vessel segmentation processing on the initial blood vessel image to obtain an initial blood vessel segmentation image;

[0038] a processing module, configured to perform skeletonization processing on the initially segmented blood vessel image to obtain a blood vessel centerline image;

[0039] an extraction module, configured to perform point cloud feature processing on the blood vessel centerline image based on the blood vessel centerline image, and extract point cloud features of the blood vessel centerline;

[0040] a classification module, configured to input the point cloud features of the blood vessel centerline into a point cloud segmentation model to obtain a point cloud classification result of the blood vessel centerline;

[0041] A naming module is used to name the blood vessel based on the classification result of the centerline point cloud of the blood vessel.

[0042] Furthermore, the extraction module is specifically used to:

[0043] Based on the predicted label value of each point cloud in the blood vessel centerline image, determining, among all point clouds included in the blood vessel centerline image, a point cloud where the predicted label value indicates a foreground point as a target point cloud;

[0044] Extracting the position coordinates and grayscale values of the target point cloud from the blood vessel centerline image, and determining the blood vessel radius corresponding to each target point cloud and the direction vector of each target point cloud;

[0045] For each target point cloud, the position coordinates, grayscale value, corresponding blood vessel radius and direction vector of the target point cloud are spliced to obtain the point cloud features of the target point cloud;

[0046] The combination of the point cloud features of each target point cloud is determined as the point cloud feature of the blood vessel centerline.

[0047] In a third aspect, an embodiment of the present application further provides an electronic device comprising: a processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor and the memory communicate via the bus, and when the machine-readable instructions are executed by the processor, the steps of the above-mentioned centerline-based blood vessel naming method are performed.

[0048] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the above-mentioned centerline-based blood vessel naming method are executed.

[0049] Embodiments of the present application provide a centerline-based blood vessel naming method, apparatus, electronic device, and storage medium. The blood vessel naming method includes: acquiring an initial blood vessel image, performing initial blood vessel segmentation processing on the initial blood vessel image to obtain an initial blood vessel segmentation image; performing skeletonization processing on the initial blood vessel segmentation image to obtain a blood vessel centerline image; performing point cloud feature processing on the blood vessel centerline image based on the blood vessel centerline image to extract point cloud features of the blood vessel centerline; inputting the point cloud features of the blood vessel centerline into a point cloud segmentation model to obtain a blood vessel centerline point cloud classification result; and naming the blood vessel based on the blood vessel centerline point cloud classification result.

[0050] In this way, the technical solution provided by this application can extract the point cloud features on the blood vessel centerline, predict the classification of the point cloud on the blood vessel centerline, and name the blood vessel based on the classification results, thereby reducing the interference of image background noise and improving the accuracy and efficiency of blood vessel naming.

[0051] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without creative work.

[0053] Figure 1 A flow chart of a centerline-based blood vessel naming method provided in an embodiment of the present application is shown;

[0054] Figure 2 A flowchart of another centerline-based blood vessel naming method provided in an embodiment of the present application is shown;

[0055] Figure 3 A schematic diagram of determining a blood vessel radius provided by an embodiment of the present application is shown;

[0056] Figure 4 One of the structural diagrams of a centerline-based blood vessel naming device provided in an embodiment of the present application is shown;

[0057] Figure 5 The second structural diagram of a centerline-based blood vessel naming device provided in an embodiment of the present application is shown;

[0058] Figure 6 A schematic structural diagram of an electronic device provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0059] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. It should be understood that the drawings in the present application only serve the purpose of illustration and description and are not used to limit the scope of protection of the present application. In addition, it should be understood that the schematic drawings are not drawn to scale. The flowcharts used in this application illustrate the operations implemented according to some embodiments of the present application. It should be understood that the operations of the flowcharts can be implemented out of sequence, and steps without logical context can be reversed or implemented simultaneously. In addition, those skilled in the art, under the guidance of the contents of this application, can add one or more other operations to the flowchart, or remove one or more operations from the flowchart.

[0060] In addition, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. The components of the embodiments of the present application generally described and shown in the drawings here can be arranged and designed in various configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work are within the scope of protection of the present application.

[0061] In order to enable those skilled in the art to use the contents of this application, the following implementation methods are provided in combination with the specific application scenario of "vascular naming based on centerlines". For those skilled in the art, the general principles defined herein can be applied to other embodiments and application scenarios without departing from the spirit and scope of this application.

[0062] The following methods, devices, electronic devices, or computer-readable storage media of the embodiments of the present application can be applied to any scenario where blood vessels need to be named based on the centerline. The embodiments of the present application are not limited to specific application scenarios. Any solution using the centerline-based blood vessel naming method, device, electronic device, and storage medium provided in the embodiments of the present application is within the scope of protection of this application.

[0063] It is worth noting that vascular diseases pose a serious threat to human health and life. Accurately naming individual vessels in images is an essential and challenging task in intelligent vascular disease screening. Accurate vessel naming provides the foundation for subsequent vessel registration and 3D reconstruction.

[0064] Currently, most vascular naming technologies use the point cloud coordinate features of the entire image using binary segmentation as input. This not only introduces background noise, reducing the segmentation accuracy of deep networks; but also requires predicting more points during inference, affecting the efficiency of subsequent vascular naming. Therefore, how to name blood vessels has become an urgent problem to be solved.

[0065] Based on this, the present application proposes a centerline-based blood vessel naming method, device, electronic device and storage medium. The blood vessel naming method includes: obtaining an initial blood vessel image, performing initial blood vessel segmentation processing on the initial blood vessel image to obtain an initial blood vessel segmentation image; performing skeletonization processing on the initial blood vessel segmentation image to obtain a blood vessel centerline image; based on the blood vessel centerline image, performing point cloud feature processing on the blood vessel centerline image to extract point cloud features of the blood vessel centerline; inputting the point cloud features of the blood vessel centerline into a point cloud segmentation model to obtain a blood vessel centerline point cloud classification result; and naming the blood vessel based on the blood vessel centerline point cloud classification result.

[0066] In this way, the technical solution provided by this application can extract the point cloud features on the blood vessel centerline, predict the classification of the point cloud on the blood vessel centerline, and name the blood vessel based on the classification results, thereby reducing the interference of image background noise and improving the accuracy and efficiency of blood vessel naming.

[0067] To facilitate understanding of the present application, the technical solutions provided in the present application will be described in detail below in conjunction with specific embodiments.

[0068] See also Figure 1 , Figure 1 A flow chart of a method for naming blood vessels based on centerlines provided in an embodiment of the present application is shown in FIG. Figure 1 As shown in , the blood vessel naming method includes:

[0069] S101, acquiring an initial blood vessel image, and performing initial blood vessel segmentation processing on the initial blood vessel image to obtain an initial blood vessel segmentation image;

[0070] In this step, the initial vascular image can be coronary angiography data, and the initial vascular segmentation processing can use a segmentation model. For the training stage of the segmentation model, a large amount of coronary angiography data can be binary-labeled with blood vessels. For example, the labeling result can be that the blood vessels are labeled as 1 and the background is labeled as 0. Based on the labeling results, a 2D segmentation model for vascular segmentation is trained, such as 2D U-net, V-net, etc.; for the inference stage of the segmentation model, the trained 2D segmentation model can be used to predict new coronary angiography data, and a 2D vascular initial segmentation image can be obtained through post-processing operations, that is, an image that only includes blood vessels after removing the background.

[0071] S102, performing skeletonization processing on the initially segmented blood vessel image to obtain a blood vessel centerline image;

[0072] In this step, for the extraction of the vascular centerline, a series of morphological operations (such as corrosion and dilation) can be performed on the 2D vascular initial segmentation image, and a smoothing algorithm (such as Gaussian filtering and median filtering) can be used to smooth the image to obtain a 2D vascular centerline image.

[0073] S103, performing point cloud feature processing on the blood vessel centerline image based on the blood vessel centerline image to extract point cloud features of the blood vessel centerline;

[0074] In this step, the point cloud features of the blood vessel centerline extracted may include position coordinates, grayscale values, corresponding blood vessel radius, and direction vector.

[0075] Please note that Figure 2 , Figure 2 This is a flow chart of another method for naming blood vessels based on centerlines provided in an embodiment of the present application. Figure 2 As shown in , based on the vascular centerline image, the step of performing point cloud feature processing on the vascular centerline image and extracting the point cloud features of the vascular centerline includes:

[0076] S201, based on the predicted label value of each point cloud in the blood vessel centerline image, determining, among all point clouds included in the blood vessel centerline image, a point cloud where the predicted label value indicates a foreground point as a target point cloud;

[0077] In this step, the foreground points of the vascular centerline image are extracted. The predicted label value of each point cloud in the vascular centerline image can be used to extract the point cloud whose predicted label value indicates that the corresponding point cloud is a foreground point, and the extracted point cloud is determined as the target point cloud, that is, the foreground point.

[0078] S202, extracting the position coordinates and grayscale values of the target point cloud from the blood vessel centerline image, and determining the blood vessel radius corresponding to each target point cloud and the direction vector of each target point cloud;

[0079] In this step, the position coordinates [x i ,y i ](i=1, ..., n, n is the number of foreground points in the vascular centerline image) (the processing object here is the vascular centerline); secondly, obtain the position coordinates of the corresponding foreground point [x i ,y i ]The gray value of the contrast image [g i]; Then, the radius of the blood vessel corresponding to the position coordinates of each foreground point is obtained by the blood vessel radius calculation algorithm [r i ]; Then, the blood vessel direction calculation algorithm is used to obtain the direction vector of the position coordinates of each foreground point [d xi ,d yi ].

[0080] It should be noted that the blood vessel radius corresponding to each target point cloud is determined through the following steps:

[0081] 1. Based on the predicted label value of each point cloud in the initial segmented blood vessel image, invert the initial segmented blood vessel image to obtain an inverted image;

[0082] In this step, the initial segmentation image of the blood vessels is inverted to obtain the inverted image. For example, the predicted label value of the point cloud with a value of 1 in the initial segmentation image of the blood vessels can be changed to 0, and the predicted label value of the point cloud with a value of 0 can be changed to 1. The obtained inverted image is recorded as

[0083] 2. For each target point cloud, with the target point cloud as the center of the circle, based on the preset step size, obtain the expansion area corresponding to the target point cloud;

[0084] In this step, a target point cloud on the center line of the blood vessel is used as the center of the circle. The preset step size can be set to one pixel. Generally, a step size greater than one pixel will introduce additional errors, thereby obtaining a circular area corresponding to the target point cloud, namely the expansion area.

[0085] 3. Determine whether the dilated area corresponding to the target point cloud intersects with the inverted image;

[0086] 4. If not, increasing the radius of the dilated area corresponding to the target point cloud by the preset step size, updating the dilated area corresponding to the target point cloud, and again determining whether the dilated area corresponding to the target point cloud intersects with the inverted image;

[0087] In this step, if the expansion area and negate the image If no intersection occurs, The circular area continues to expand, expanding outward according to the preset step size, and updating Until and Produce an intersection.

[0088] 5. If yes, determine the radius of the blood vessel corresponding to the target point cloud based on the area of the expansion region corresponding to the target point cloud.

[0089] In this step, if an intersection is generated, the radius of the blood vessel corresponding to the target point cloud can be calculated by the circular area formula. According to the circular area formula S = πr 2, we can get:

[0090]

[0091] Where S is the expansion area The circular area, where S is the expansion area The sum of the pixel values; r is the radius of the blood vessel corresponding to the target point cloud.

[0092] For example, see Figure 3 , Figure 3 A schematic diagram of determining the radius of a blood vessel is provided in an embodiment of the present application, as shown in FIG. Figure 3 As shown in the figure, the area formed by the two thick arcs is the blood vessel, and the thin line in the middle of the blood vessel is the blood vessel centerline. The initial segmentation image of the blood vessel is subjected to the inversion operation to obtain the inverted image. Starting from a point on the center line of the blood vessel, the expansion algorithm with a step size of 1 (the size of 1 pixel) can be used to expand to the surrounding areas, and the expansion algorithm is continuously iterated until the expansion area (Right now Figure 3 The circular area in If an intersection occurs, the expansion stops, that is, Stop iteration. Expand the region Approximately a circle, the above-mentioned circle area formula is used to calculate the approximate blood vessel radius r of the target point cloud on the center line.

[0093] It should be noted that the direction vector of each target point cloud is determined by the following steps:

[0094] 1. determining the position coordinates of the next target point cloud of each target point cloud based on the position coordinates of each target point cloud in the blood vessel centerline image;

[0095] 2. For each target point cloud, determine the direction vector of the target point cloud based on the position coordinates of the next target point cloud of the target point cloud and the position coordinates of the target point cloud.

[0096] Specifically, a target point cloud [x i ,y i ], and obtain the position coordinates of the subsequent target point cloud of the target point cloud [x i+1 ,y i+1 ], the direction vector of the target point cloud on the center line is obtained by the following mathematical formula

[0097]

[0098] Here, the method for obtaining the position coordinates of the subsequent target point cloud can adopt a depth-first traversal algorithm to save the position coordinates of the target point cloud of each center line into a center line list P; the coordinates of the successor point of point P[j] in the list are P[j+1].

[0099] S203 , for each target point cloud, concatenate the position coordinates, grayscale value, corresponding blood vessel radius, and direction vector of the target point cloud to obtain point cloud features of the target point cloud;

[0100] In this step, all the features of the i-th target point cloud (including the position coordinates [x i ,y i ], grayscale value [g i ], the corresponding vascular radius [r i ] and the direction vector [d xi ,d yi ]) are spliced together to form the point cloud features of the foreground point (i.e., the target point cloud), i.e., [x i ,y i ,g i ,r i ,d xi ,d yi ](i=1,…,n, where n is the number of foreground points in the blood vessel centerline image).

[0101] S204 : Determine the combination of the point cloud features of each target point cloud as the point cloud feature of the blood vessel centerline.

[0102] S104, inputting the point cloud features of the blood vessel centerline into a point cloud segmentation model to obtain a blood vessel centerline point cloud classification result;

[0103] In this step, the point cloud segmentation model can use the common deep learning model PointNet or PointTransformer (PT). Taking Point Transformer as an example, the point cloud segmentation model can be composed of several downsampling (Transition down) and upsampling (Transition up) modules, and a self-attention module (Transformer Block) can be added to each layer. Among them, the self-attention module uses vector self-attention (vector attention), and the mathematical formula is as follows:

[0104]

[0105] Among them, y i is the output feature of the i-th foreground point; X i is the input feature of the i-th foreground point; X jis the input feature of the jth foreground point; χ(i) is the point cloud set of the neighborhood of the i-th foreground point; γ is a mapping function, such as MLP; ρ is a normalized exponential function, such as softmax, used to normalize the features; ψ and α are linear mapping layers; δ is the position encoding function.

[0106] Here, the training process of the point cloud segmentation model is as follows: First, professional doctors annotate the data. The annotation rule can be: small branches are annotated as 0, the main coronary artery (left anterior descending artery LAD / left circumflex artery LCX / right coronary artery RCA) is annotated as 1, and the main branches (diagonal branch D / obtuse marginal branch OM / right posterior descending artery RPDA) is annotated as 2. Let the annotation corresponding to the foreground point be l i , concatenate it with the input point cloud feature, namely [x i ,y i ,g i ,r i ,d xi ,d yi ,l i ] as the input of the point cloud segmentation model; the point cloud segmentation model outputs the prediction result by learning the point cloud features The output prediction results are compared with the annotation gold standard l i Calculate the loss (the loss function can be cross entropy loss), and reduce the loss through continuous iteration until the point cloud segmentation model converges to obtain a trained point cloud segmentation model. Inference process of point cloud segmentation model: Input the point cloud features of the blood vessel centerline [x i ,y i ,g i ,r i ,d xi ,d yi ], output the label (0, 1, 2) to which the corresponding point belongs, and obtain the centerline point cloud classification result of the blood vessel.

[0107] S105 : Name the blood vessel based on the classification result of the centerline point cloud of the blood vessel.

[0108] It should be noted that the steps for naming blood vessels based on the classification results of the centerline point cloud of the blood vessels include:

[0109] S1051. Based on the centerline point cloud classification results of the blood vessel, extract the position coordinates of the target point cloud in each type of centerline point cloud classification results to obtain a coordinate list corresponding to each type of centerline point cloud classification results;

[0110] In this step, each type of blood vessel in the centerline point cloud classification result is extracted and the corresponding coordinate list C is obtained.

[0111] S1052. For each type of centerline point cloud classification result, determine the intersection of the coordinate list corresponding to the centerline point cloud classification result and the coordinate list of each centerline in the predetermined two-dimensional centerline list;

[0112] In this step, the intersection of the coordinate list C corresponding to each category and each center line list in the obtained center line two-dimensional list P is calculated.

[0113] It should be noted that the two-dimensional list of center lines is determined by the following steps:

[0114] 1) Obtaining the number of target point clouds adjacent to each target point cloud in the blood vessel centerline image;

[0115] 2) determining the target point cloud with one adjacent target point cloud as an endpoint;

[0116] 3) For each endpoint, determining a target intersection between a preset rectangular area centered on the endpoint and the initial segmented blood vessel image;

[0117] 4) In the target intersection corresponding to each endpoint, the endpoint corresponding to the target intersection with the largest number of intersection points in the target intersection is determined as the starting point;

[0118] In the above steps 1) to 4), the starting point of the blood vessel can be obtained by using a starting point search algorithm. For example, all points on the center line can be traversed to obtain the number of neighbors of each point. The point with 1 neighbor is determined as the endpoint, and the position coordinates of the endpoint are saved as a list, that is, L = [[x1, y1], [x2, y2], ..., [x m ,y m ]]. For coordinates [x j ,y j ] endpoint j, calculate the k neighborhood of this endpoint [x j ±k,y j ±k] and the initial segmentation image of the blood vessel, and use the above algorithm to traverse all endpoints of the list L to obtain the intersection list Ω = [ω1, ω2, ..., ω m ], take the endpoint corresponding to the maximum value in the intersection list Ω as the starting point of the blood vessel, denoted as P s Here, based on the radius data of all endpoints, the endpoint with the largest radius among the endpoints can be used as the starting point. This can also be accomplished by using a deep learning method, which is not limited here.

[0119] 5) Perform a depth-first traversal starting from the starting point, and determine the set of coordinate lists consisting of the position coordinates of the target point cloud of each center line as a center line two-dimensional list.

[0120] In this step, for the starting point P obtained above, s , using the depth-first traversal algorithm, starting from the starting point, the coordinates of each center line point are saved as a center line two-dimensional list P = [P0, ..., P n ], P0,…,P n is a list of coordinates of each centerline.

[0121] S1053 . Name the blood vessel represented by the coordinate list of the center line corresponding to the intersection with the largest number of intersection points in the intersection as the center line point cloud classification result of this type.

[0122] In this step, the centerline list corresponding to the largest intersection is named as the category. For example, the coordinates of label = 1 in the centerline point cloud classification results are extracted into a coordinate list. The intersection of this coordinate list and each centerline list in the centerline two-dimensional list P is calculated, and the centerline with the largest intersection is named label = 1.

[0123] In summary, this embodiment first acquires an initial vascular image, performs preliminary vascular segmentation on the vascular image, and obtains a corresponding preliminary vascular segmentation image. A skeletonization (morphological correlation) algorithm is then applied to the preliminary vascular segmentation image to obtain the vascular centerline. Point cloud feature processing is then performed on the vascular centerline to obtain point cloud coordinate features, vascular radius features at the location, and point cloud features of the vascular direction. The extracted point cloud features are concatenated and input into a point cloud segmentation network to obtain the corresponding centerline point cloud result. The centerline predicted by the deep network is then post-processed to obtain the vascular name. This point cloud segmentation-based approach uses only the image foreground information as input. The image topology is preserved through coordinate features. The skeletonized binary segmentation result is used as input to the point cloud segmentation network, enabling faster and more direct vascular centerline results. Furthermore, additional point cloud features of vascular radius and direction are introduced on top of the point cloud coordinates, significantly improving the accuracy of the point cloud segmentation network. Furthermore, using the vascular centerline for inference only requires processing approximately one-tenth of the number of point clouds required for the preliminary vascular segmentation, accelerating inference and improving the efficiency of subsequent vascular naming.

[0124] An embodiment of the present application provides a centerline-based blood vessel naming method, comprising: acquiring an initial blood vessel image, performing initial blood vessel segmentation processing on the initial blood vessel image to obtain an initial blood vessel segmentation image; performing skeletonization processing on the initial blood vessel segmentation image to obtain a blood vessel centerline image; performing point cloud feature processing on the blood vessel centerline image based on the blood vessel centerline image to extract point cloud features of the blood vessel centerline; inputting the point cloud features of the blood vessel centerline into a point cloud segmentation model to obtain a blood vessel centerline point cloud classification result; and naming the blood vessel based on the blood vessel centerline point cloud classification result.

[0125] In this way, the technical solution provided by this application can extract the point cloud features on the blood vessel centerline, predict the classification of the point cloud on the blood vessel centerline, and name the blood vessel based on the classification results, thereby reducing the interference of image background noise and improving the accuracy and efficiency of blood vessel naming.

[0126] Based on the same application concept, the embodiments of the present application also provide a centerline-based blood vessel naming device corresponding to the centerline-based blood vessel naming method provided in the above embodiment. Since the principle of solving the problem by the device in the embodiments of the present application is similar to the centerline-based blood vessel naming method in the above embodiment of the present application, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be repeated.

[0127] See also Figure 4 、 Figure 5 , Figure 4 This is one of the structural diagrams of a centerline-based blood vessel naming device provided in an embodiment of the present application. Figure 5 This is the second structural diagram of a centerline-based blood vessel naming device provided in an embodiment of the present application. Figure 4 As shown in , the blood vessel naming device 410 includes:

[0128] An acquisition module 411 is configured to acquire an initial blood vessel image and perform initial blood vessel segmentation processing on the initial blood vessel image to obtain an initial blood vessel segmentation image;

[0129] A processing module 412 is configured to perform skeletonization processing on the initially segmented blood vessel image to obtain a blood vessel centerline image;

[0130] An extraction module 413 is configured to perform point cloud feature processing on the blood vessel centerline image based on the blood vessel centerline image to extract point cloud features of the blood vessel centerline;

[0131] A classification module 414 is configured to input the point cloud features of the blood vessel centerline into a point cloud segmentation model to obtain a blood vessel centerline point cloud classification result;

[0132] The naming module 415 is configured to name the blood vessel based on the classification result of the centerline point cloud of the blood vessel.

[0133] Optionally, the extraction module 413 is specifically configured to:

[0134] Based on the predicted label value of each point cloud in the blood vessel centerline image, determining, among all point clouds included in the blood vessel centerline image, a point cloud where the predicted label value indicates a foreground point as a target point cloud;

[0135] Extracting the position coordinates and grayscale values of the target point cloud from the blood vessel centerline image, and determining the blood vessel radius corresponding to each target point cloud and the direction vector of each target point cloud;

[0136] For each target point cloud, the position coordinates, grayscale value, corresponding blood vessel radius and direction vector of the target point cloud are spliced to obtain the point cloud features of the target point cloud;

[0137] The combination of the point cloud features of each target point cloud is determined as the point cloud feature of the blood vessel centerline.

[0138] Optionally, when the extraction module 413 is used to determine the blood vessel radius corresponding to each target point cloud, the extraction module 413 is specifically configured to:

[0139] Based on the predicted label value of each point cloud in the initial segmented blood vessel image, the initial segmented blood vessel image is inverted to obtain an inverted image;

[0140] For each target point cloud, with the target point cloud as the center of the circle, based on the preset step size, the expansion area corresponding to the target point cloud is obtained;

[0141] Determine whether the expansion area corresponding to the target point cloud intersects with the inverted image;

[0142] If not, increasing the radius of the expansion area corresponding to the target point cloud by the preset step size, updating the expansion area corresponding to the target point cloud, and determining again whether the expansion area corresponding to the target point cloud intersects with the inverted image;

[0143] If so, the radius of the blood vessel corresponding to the target point cloud is determined based on the area of the expansion region corresponding to the target point cloud.

[0144] Optionally, when the extraction module 413 is used to determine the direction vector of each target point cloud, the extraction module 413 is specifically used to:

[0145] Determining the position coordinates of the next target point cloud of each target point cloud based on the position coordinates of each target point cloud in the blood vessel centerline image;

[0146] For each target point cloud, a direction vector of the target point cloud is determined based on the position coordinates of the next target point cloud of the target point cloud and the position coordinates of the target point cloud.

[0147] Optionally, the naming module 415 is specifically configured to:

[0148] Based on the centerline point cloud classification results of the blood vessels, extracting the position coordinates of the target point cloud in each type of centerline point cloud classification results, and obtaining a coordinate list corresponding to each type of centerline point cloud classification results;

[0149] For each type of centerline point cloud classification result corresponding to the coordinate list, determine the intersection of the coordinate list corresponding to the type of centerline point cloud classification result and the coordinate list of each centerline in the predetermined centerline two-dimensional list;

[0150] In the intersection sets, the blood vessel represented by the coordinate list of the center line corresponding to the intersection set with the largest number of intersection points in the intersection sets is named as the classification result of the center line point cloud of this type.

[0151] Optional, such as Figure 5 As shown, the blood vessel naming device 410 further includes a determination module 416, and the determination module 416 is configured to:

[0152] Obtaining the number of target point clouds adjacent to each target point cloud in the blood vessel centerline image;

[0153] Determine the target point cloud having one adjacent target point cloud as an endpoint;

[0154] For each endpoint, determining a target intersection between a preset rectangular area centered at the endpoint and the blood vessel initial segmentation image;

[0155] In the target intersections corresponding to each endpoint, the endpoint corresponding to the target intersection with the largest number of intersection points in the target intersections is determined as the starting point;

[0156] A depth-first traversal is performed starting from the starting point, and a set of coordinate lists consisting of the position coordinates of the target point cloud of each center line is determined as a center line two-dimensional list.

[0157] An embodiment of the present application provides a centerline-based blood vessel naming device, comprising: an acquisition module for acquiring an initial blood vessel image and performing initial blood vessel segmentation processing on the initial blood vessel image to obtain an initial blood vessel segmentation image; a processing module for performing skeletonization processing on the initial blood vessel segmentation image to obtain a blood vessel centerline image; an extraction module for performing point cloud feature processing on the blood vessel centerline image based on the blood vessel centerline image to extract point cloud features of the blood vessel centerline; a classification module for inputting the point cloud features of the blood vessel centerline into a point cloud segmentation model to obtain a blood vessel centerline point cloud classification result; and a naming module for naming the blood vessel based on the blood vessel centerline point cloud classification result.

[0158] In this way, the technical solution provided by this application can extract the point cloud features on the blood vessel centerline, predict the classification of the point cloud on the blood vessel centerline, and name the blood vessel based on the classification results, thereby reducing the interference of image background noise and improving the accuracy and efficiency of blood vessel naming.

[0159] See also Figure 6 , Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. Figure 6 As shown in FIG, the electronic device 600 includes a processor 610 , a memory 620 and a bus 630 .

[0160] The memory 620 stores machine-readable instructions executable by the processor 610. When the electronic device 600 is running, the processor 610 communicates with the memory 620 via the bus 630. When the machine-readable instructions are executed by the processor 610, the above-mentioned Figure 1 as well as Figure 2 The specific implementation of the steps of the centerline-based blood vessel naming method in the illustrated method embodiment can be found in the method embodiment and will not be repeated here.

[0161] The embodiment of the present application also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the computer program can execute the above-mentioned Figure 1 as well as Figure 2 The specific implementation of the steps of the centerline-based blood vessel naming method in the illustrated method embodiment can be found in the method embodiment and will not be repeated here.

[0162] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0163] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. There may be other division methods in actual implementation. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed may be through some communication interface, indirect coupling or communication connection of devices or units, which may be electrical, mechanical or other forms.

[0164] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0165] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0166] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a non-volatile computer-readable storage medium that is executable by a processor. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0167] Finally, it should be noted that the above-described embodiments are only specific implementation methods of the present application, which are used to illustrate the technical solutions of the present application, rather than to limit them. The scope of protection of the present application is not limited thereto. Although the present application has been described in detail with reference to the above-mentioned embodiments, those skilled in the art should understand that any person skilled in the art can modify or easily conceive of changes to the technical solutions described in the above-mentioned embodiments within the technical scope disclosed in the present application, or perform equivalent replacements for some of the technical features thereof. These modifications, changes, or replacements do not deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A method for naming blood vessels based on centerlines, characterized in that: The blood vessel naming method includes: Acquiring an initial blood vessel image, and performing initial blood vessel segmentation processing on the initial blood vessel image to obtain an initial blood vessel segmentation image; performing skeletonization processing on the blood vessel initial segmentation image to obtain a blood vessel centerline image; Based on the blood vessel centerline image, performing point cloud feature processing on the blood vessel centerline image to extract point cloud features of the blood vessel centerline; The step of performing point cloud feature processing on the blood vessel centerline image based on the blood vessel centerline image to extract point cloud features of the blood vessel centerline includes: Based on the predicted label value of each point cloud in the blood vessel centerline image, determining, among all point clouds included in the blood vessel centerline image, a point cloud where the predicted label value indicates a foreground point as a target point cloud; Extracting the position coordinates and grayscale values of the target point cloud from the blood vessel centerline image, and determining the blood vessel radius corresponding to each target point cloud and the direction vector of each target point cloud; wherein the blood vessel radius corresponding to each target point cloud is determined by the following steps: Based on the predicted label value of each point cloud in the initial segmented blood vessel image, the initial segmented blood vessel image is inverted to obtain an inverted image; For each target point cloud, with the target point cloud as the center of the circle, based on the preset step size, the expansion area corresponding to the target point cloud is obtained; Determine whether the expansion area corresponding to the target point cloud intersects with the inverted image; If not, increasing the radius of the expansion area corresponding to the target point cloud by the preset step size, updating the expansion area corresponding to the target point cloud, and determining again whether the expansion area corresponding to the target point cloud intersects with the inverted image; If yes, determining the radius of the blood vessel corresponding to the target point cloud based on the area of the expansion region corresponding to the target point cloud; For each target point cloud, the position coordinates, grayscale value, corresponding blood vessel radius and direction vector of the target point cloud are spliced to obtain the point cloud features of the target point cloud; Determining the combination of point cloud features of each target point cloud as the point cloud feature of the blood vessel centerline; inputting the point cloud feature of the blood vessel centerline into the point cloud segmentation model to obtain a blood vessel centerline point cloud classification result; The blood vessel is named based on the classification result of the centerline point cloud of the blood vessel.

2. The blood vessel naming method according to claim 1, characterized in that: Determine the direction vector for each target point cloud by following these steps: Determining the position coordinates of the next target point cloud of each target point cloud based on the position coordinates of each target point cloud in the blood vessel centerline image; For each target point cloud, a direction vector of the target point cloud is determined based on the position coordinates of the next target point cloud of the target point cloud and the position coordinates of the target point cloud.

3. The blood vessel naming method according to claim 1, characterized in that: The step of naming the blood vessel based on the classification result of the centerline point cloud of the blood vessel includes: Based on the centerline point cloud classification results of the blood vessels, extracting the position coordinates of the target point cloud in each type of centerline point cloud classification results, and obtaining a coordinate list corresponding to each type of centerline point cloud classification results; For each type of centerline point cloud classification result corresponding to the coordinate list, determine the intersection of the coordinate list corresponding to the type of centerline point cloud classification result and the coordinate list of each centerline in the predetermined centerline two-dimensional list; In the intersection sets, the blood vessel represented by the coordinate list of the center line corresponding to the intersection set with the largest number of intersection points in the intersection sets is named as the classification result of the center line point cloud of this type.

4. The blood vessel naming method according to claim 3, characterized in that: Determine the 2D list of centerlines using the following steps: Obtaining the number of target point clouds adjacent to each target point cloud in the blood vessel centerline image; Determine the target point cloud having one adjacent target point cloud as an endpoint; For each endpoint, determining a target intersection between a preset rectangular area centered at the endpoint and the blood vessel initial segmentation image; In the target intersections corresponding to each endpoint, the endpoint corresponding to the target intersection with the largest number of intersection points in the target intersections is determined as the starting point; A depth-first traversal is performed starting from the starting point, and a set of coordinate lists consisting of the position coordinates of the target point cloud of each center line is determined as a center line two-dimensional list.

5. A device for naming blood vessels based on centerlines, characterized in that: The blood vessel naming device comprises: an acquisition module, configured to acquire an initial blood vessel image, and perform initial blood vessel segmentation processing on the initial blood vessel image to obtain an initial blood vessel segmentation image; a processing module, configured to perform skeletonization processing on the initially segmented blood vessel image to obtain a blood vessel centerline image; an extraction module, configured to perform point cloud feature processing on the blood vessel centerline image based on the blood vessel centerline image, and extract point cloud features of the blood vessel centerline; The extraction module is specifically used for: Based on the predicted label value of each point cloud in the blood vessel centerline image, determining, among all point clouds included in the blood vessel centerline image, a point cloud where the predicted label value indicates a foreground point as a target point cloud; Extracting the position coordinates and grayscale values of the target point cloud from the blood vessel centerline image, and determining the blood vessel radius corresponding to each target point cloud and the direction vector of each target point cloud; wherein, when the extraction module is used to determine the blood vessel radius corresponding to each target point cloud, the extraction module is used to: Based on the predicted label value of each point cloud in the initial segmented blood vessel image, the initial segmented blood vessel image is inverted to obtain an inverted image; For each target point cloud, with the target point cloud as the center of the circle, based on the preset step size, the expansion area corresponding to the target point cloud is obtained; Determine whether the expansion area corresponding to the target point cloud intersects with the inverted image; If not, increasing the radius of the expansion area corresponding to the target point cloud by the preset step size, updating the expansion area corresponding to the target point cloud, and determining again whether the expansion area corresponding to the target point cloud intersects with the inverted image; If yes, determining the radius of the blood vessel corresponding to the target point cloud based on the area of the expansion region corresponding to the target point cloud; For each target point cloud, the position coordinates, grayscale value, corresponding blood vessel radius and direction vector of the target point cloud are spliced to obtain the point cloud features of the target point cloud; The combination of the point cloud features of each target point cloud is determined as the point cloud feature of the blood vessel centerline; a classification module, configured to input the point cloud features of the blood vessel centerline into a point cloud segmentation model to obtain a point cloud classification result of the blood vessel centerline; A naming module is used to name the blood vessel based on the classification result of the centerline point cloud of the blood vessel.

6. An electronic device, characterized in that: include: A processor, a memory, and a bus, wherein the memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor and the memory communicate via the bus. When the processor runs the machine-readable instructions, the steps of the centerline-based blood vessel naming method according to any one of claims 1 to 4 are executed.

7. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the centerline-based blood vessel naming method according to any one of claims 1 to 4 are executed.

Images