Method and device for adjusting blood vessel classification model, electronic equipment and storage medium

Abstract

Embodiments of the present application disclose a blood vessel classification model adjustment method and device, electronic equipment and a storage medium. The method comprises: obtaining a blood vessel point cloud image, a blood vessel classification model and a trained feature extraction model, wherein the blood vessel point cloud image comprises at least one blood vessel point; inputting the blood vessel point cloud image into the feature extraction model, and obtaining point features corresponding to the at least one blood vessel point according to an output result of the feature extraction model; for each blood vessel point in the at least one blood vessel point, determining a direction vector of the blood vessel point according to the point feature corresponding to the blood vessel point; determining an attention sequence based on the at least one blood vessel point and the direction vectors corresponding to the at least one blood vessel point, and adjusting the blood vessel classification model according to the attention sequence. The technical solution of the embodiments of the present application can improve the classification accuracy of the blood vessel classification model.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to a method, apparatus, electronic device and storage medium for adjusting a blood vessel classification model. Background Technology

[0002] In the medical field, accurate classification of blood vessels is of paramount importance.

[0003] Currently, blood vessel classification is usually done using models, but existing blood vessel classification models have low accuracy, which needs to be addressed. Summary of the Invention

[0004] This invention provides a method, apparatus, electronic device, and storage medium for adjusting a vascular classification model to improve the classification accuracy of the vascular classification model.

[0005] According to one aspect of the present invention, a method for adjusting a blood vessel classification model is provided, which may include:

[0006] Acquire a blood vessel point cloud image, a blood vessel classification model, and a trained feature extraction model, wherein the blood vessel point cloud image includes at least one blood vessel point;

[0007] The blood vessel point cloud image is input into the feature extraction model, and the point features corresponding to at least one blood vessel point are obtained based on the output of the feature extraction model.

[0008] For each of at least one blood vessel point, determine the direction vector of the blood vessel point based on the point features corresponding to the blood vessel point;

[0009] An attention sequence is determined based on at least one blood vessel point and the direction vector corresponding to at least one blood vessel point, and the blood vessel classification model is adjusted according to the attention sequence.

[0010] According to another aspect of the present invention, an adjustment device for a blood vessel classification model is provided, which may include:

[0011] The feature extraction model acquisition module is used to acquire blood vessel point cloud images, blood vessel classification models, and trained feature extraction models, wherein the blood vessel point cloud images include at least one blood vessel point.

[0012] The point feature acquisition module is used to input the blood vessel point cloud image into the feature extraction model, and obtain the point features corresponding to at least one blood vessel point based on the output of the feature extraction model.

[0013] The direction vector determination module is used to determine the direction vector of each blood vessel point in at least one blood vessel point based on the point features corresponding to the blood vessel point.

[0014] The blood vessel classification model adjustment module is used to determine the attention sequence based on at least one blood vessel point and the direction vector corresponding to at least one blood vessel point, and adjust the blood vessel classification model according to the attention sequence.

[0015] According to another aspect of the present invention, an electronic device is provided, which may include:

[0016] At least one processor; and

[0017] A memory that is communicatively connected to at least one processor; wherein,

[0018] The memory stores a computer program that can be executed by at least one processor, such that when the at least one processor executes the program, it implements the method for adjusting the blood vessel classification model provided in any embodiment of the present invention.

[0019] According to another aspect of the present invention, a computer-readable storage medium is provided having computer instructions stored thereon for causing a processor to execute and implement the method for adjusting a blood vessel classification model provided in any embodiment of the present invention.

[0020] The technical solution of this invention involves acquiring a vascular point cloud image, a vascular classification model, and a trained feature extraction model. The vascular point cloud image includes at least one vascular point. The vascular point cloud image is input into the feature extraction model, and point features corresponding to each of the at least one vascular point are obtained based on the model's output. For each of the at least one vascular point, a direction vector is determined based on its corresponding point features. An attention sequence is determined based on the at least one vascular point and its corresponding direction vector, and the vascular classification model is adjusted according to this attention sequence. This technical solution, by adjusting the vascular classification model based on an attention sequence focusing on the vascular direction, can introduce attention to the vascular direction into the vascular classification model, thereby improving the classification accuracy of the vascular classification model.

[0021] It should be understood that the description in this section is not intended to identify key or important features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a flowchart of a method for adjusting a blood vessel classification model according to an embodiment of the present invention;

[0024] Figure 2 This is a schematic diagram of the attention sequence in an adjustment method for a blood vessel classification model provided by an embodiment of the present invention;

[0025] Figure 3 This is a flowchart of another method for adjusting a blood vessel classification model according to an embodiment of the present invention;

[0026] Figure 4 This is a flowchart of another method for adjusting a blood vessel classification model according to an embodiment of the present invention;

[0027] Figure 5 This is a flowchart of another method for adjusting a blood vessel classification model according to an embodiment of the present invention;

[0028] Figure 6 This is a flowchart of an optional example of an adjustment method for a blood vessel classification model provided in an embodiment of the present invention;

[0029] Figure 7 This is a structural block diagram of an adjustment device for a blood vessel classification model according to an embodiment of the present invention;

[0030] Figure 8 This is a schematic diagram of the structure of an electronic device that implements the method for adjusting the blood vessel classification model according to an embodiment of the present invention. Detailed Implementation

[0031] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0032] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. The same applies to "target," "original," etc., and will not be repeated here. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0033] Figure 1 This is a flowchart illustrating a method for adjusting a blood vessel classification model provided in this embodiment of the invention. This embodiment is applicable to situations where a blood vessel classification model needs adjustment. The method can be executed by the blood vessel classification model adjustment device provided in this embodiment of the invention. This device can be implemented in software and / or hardware, and can be integrated into an electronic device, which can be various user terminals or servers.

[0034] See Figure 1 The method of this invention specifically includes the following steps:

[0035] S101. Obtain a blood vessel point cloud image, a blood vessel classification model, and a trained feature extraction model, wherein the blood vessel point cloud image includes at least one blood vessel point.

[0036] Among them, blood vessel points are points in the blood vessel point cloud image.

[0037] In this embodiment of the invention, the vascular point cloud image is a point cloud image of blood vessels obtained by acquiring blood vessels. The vascular point cloud image can be, for example, a point cloud image including blood vessels acquired through techniques such as digital radiography (DR), computed tomography (CT), CT angiography (CTA), and / or CT angiography. Image segmentation is performed on the point cloud image including blood vessels to obtain a vascular point cloud image containing only vascular points on the blood vessels. In this embodiment of the invention, the source of the vascular point cloud image is not specifically limited. The number of vascular point cloud images can be one or more. The vascular point cloud image can be a head vascular point cloud image, a leg vascular point cloud image, or an elbow vascular point cloud image, etc. In this embodiment of the invention, the type of vascular point cloud image and the location where the vascular point cloud image is acquired are not specifically limited.

[0038] In this embodiment of the invention, the vascular classification model is a model capable of classifying blood vessels. The vascular classification model can be used to classify blood vessels represented by at least one vascular point, or it can be used to classify each vascular point on a blood vessel. The vascular classification model can be a pre-trained vascular classification model or an untrained vascular classification model; if the vascular classification model is untrained, it can be trained after being adjusted according to an attention sequence. The categories that the vascular classification model can classify include, for example, arteries or veins, or for example, cardiovascular vessels, head vessels, or leg vessels, or for example, main vessels, branch vessels, or capillaries, etc. In this embodiment of the invention, the categories that the vascular classification model can classify are not specifically limited.

[0039] In this embodiment of the invention, the feature extraction model can be a pre-trained model capable of extracting features from a blood vessel point cloud image; for example, the feature extraction model can extract features from each blood vessel point in at least one blood vessel point in the blood vessel point cloud image. In this embodiment of the invention, the type of feature extraction model is not specifically limited.

[0040] Optionally, the feature extraction model is implemented based on the Unet network, which includes at least sparse convolutional layers. Specifically, the convolutional layers in the Unet network can be replaced with sparse convolutional layers, and the Unet network with these sparse convolutional layers can be used as the backbone network of the feature extraction model. In this embodiment of the invention, traditional convolutional layers have a large computational cost because each convolutional kernel is calculated with every pixel of the input image. In sparse convolutional layers, the input image is divided into multiple sub-images, each containing only a small portion of the pixels in the input image. Each convolutional kernel is calculated with only a portion of the pixels in the sub-image, while the remaining pixels are ignored. Therefore, including at least sparse convolutional layers in the Unet network can significantly reduce the computational cost, improve the extraction efficiency and accuracy of the feature extraction model, and allow the Unet network to have higher resolution, making it more suitable for blood vessel classification scenarios. It can also improve the generalization ability and robustness of the feature extraction model, avoiding overfitting and underfitting problems. Furthermore, the output of the Unet network is organized in a point-like manner, making it particularly suitable for determining attention sequences.

[0041] It should be noted that the technical solution of this invention can be applied not only to adjusting blood vessel classification models, but also to acquiring pipe point cloud images, pipe classification models, and trained feature extraction models to adjust pipe classification models. The aforementioned pipes can be tubular structures such as sewer pipes; in this invention, the specific type of pipe is not specifically limited. The pipe point cloud image can be an image including at least one pipe point.

[0042] S102. Input the blood vessel point cloud image into the feature extraction model, and obtain the point features corresponding to at least one blood vessel point based on the output of the feature extraction model.

[0043] In this embodiment of the invention, a blood vessel point cloud image can be input into a feature extraction model. The feature extraction model can extract features for each blood vessel point in at least one blood vessel point, and obtain the output result of the feature extraction model. Since the output result can characterize the features corresponding to each blood vessel point, the point features corresponding to at least one blood vessel point can be obtained based on the output result of the feature extraction model. The point features are the features corresponding to the blood vessel points.

[0044] S103. For each blood vessel point in at least one blood vessel point, determine the direction vector of the blood vessel point based on the point features corresponding to the blood vessel point.

[0045] In this embodiment of the invention, for each of the at least one blood vessel points, the point features corresponding to the blood vessel point can be processed. For example, a filtering algorithm can be used to process the point features to determine the direction vector of the blood vessel point. The direction vector of the blood vessel point can represent the direction of the blood vessel at the blood vessel point. The direction vector can represent the direction of the tangent of the blood vessel at the blood vessel point. Through the direction vector, the direction of origin and destination of the blood vessel at the blood vessel point can be determined.

[0046] S104. Based on at least one blood vessel point and the direction vector corresponding to at least one blood vessel point, determine the attention sequence, and adjust the blood vessel classification model according to the attention sequence.

[0047] In this embodiment of the invention, the attention sequence can be determined based on at least one blood vessel point and the direction vector corresponding to each blood vessel point. See [link to relevant documentation]. Figure 2 The attention sequence is an attention sequence that can characterize the relationship and direction between at least one blood vessel point, and the blood vessel classification model is adjusted based on the attention sequence.

[0048] It should be noted that the way to adjust the blood vessel classification model according to the attention sequence can be by adding modules related to the attention sequence to the blood vessel classification model, or by adjusting the parameters in the blood vessel classification model according to the attention sequence, etc. In this embodiment of the invention, no specific limitation is made on the way to adjust the blood vessel classification model according to the attention sequence.

[0049] The technical solution of this invention involves acquiring a vascular point cloud image, a vascular classification model, and a trained feature extraction model. The vascular point cloud image includes at least one vascular point. The vascular point cloud image is input into the feature extraction model, and point features corresponding to each of the at least one vascular point are obtained based on the model's output. For each of the at least one vascular point, a direction vector is determined based on its corresponding point features. An attention sequence is determined based on the at least one vascular point and its corresponding direction vector, and the vascular classification model is adjusted according to this attention sequence. This technical solution, by adjusting the vascular classification model based on an attention sequence focusing on the vascular direction, can introduce attention to the vascular direction into the vascular classification model, thereby improving the classification accuracy of the vascular classification model.

[0050] An optional technical solution involves determining the direction vector of a blood vessel point based on its corresponding point features, including: processing the point features of the blood vessel point using a target filtering algorithm to obtain at least one feature value, and feature vectors corresponding to each of the at least one feature value; and using the feature vector corresponding to the feature value with the smallest value among the at least one feature values ​​as the direction vector of the blood vessel point.

[0051] In this embodiment of the invention, no specific limitation is made on the specific type of target filtering algorithm.

[0052] For example, the target filtering algorithm can be the Hessian matrix filtering algorithm in the Frangi filtering algorithm. Specifically, it can be that the point features corresponding to the blood vessel points are processed by Gaussian filtering at multiple scales; based on the result of Gaussian filtering, the second derivative of each blood vessel point is calculated to construct a Hessian matrix. In this Hessian matrix constructed by the second derivative, the first derivative represents the change in gray level of the blood vessel point, and the second derivative represents the change in gray level of the blood vessel point, i.e., the gradient of gray level of the blood vessel point; three eigenvalues ​​are calculated based on the Hessian matrix constructed by the second derivative; for each of the three eigenvalues, the eigenvector of the Hessian matrix constructed by the second derivative is solved based on the eigenvalue, thus obtaining the eigenvectors corresponding to the three eigenvalues ​​respectively. These eigenvectors can represent the direction corresponding to the eigenvalue; the eigenvector corresponding to the eigenvalue with the smallest value among the three eigenvalues ​​is taken as the direction vector of the blood vessel point. It is important to note that the gray-level gradients are large in the two perpendicular directions of the blood vessel (i.e., large gradients in the two orthogonal directions on the plane cutting across the vessel), but small gradients along the vessel direction. Since eigenvalues ​​characterize the gray-level gradients of the vessel in its corresponding directions, two of the three eigenvalues ​​obtained above are very large, and one is approximately 0. Utilizing this characteristic of blood vessels, the eigenvector corresponding to the smallest eigenvalue can be used as the direction vector of the blood vessel point. The above technical solution can use the direction vector calculated by the Frangi filter algorithm to reduce the GPU memory occupied by the attention sequence, thereby reducing the computational burden and improving computational efficiency. Furthermore, the Frangi filter algorithm naturally reflects the blood vessel flow direction. Incorporating this directional prior into blood vessel classification can assist in the training of the blood vessel classification model, improving its accuracy and stability. The combination of direction vector determination based on the Frangi filter algorithm and blood vessel classification can effectively improve the accuracy and robustness of blood vessel classification.

[0053] In this embodiment of the invention, the feature value that can characterize the gradient of gray level of the blood vessel point along each direction can be obtained by processing the point features corresponding to the blood vessel point with the target filtering algorithm, thereby characterizing the direction vector of the blood vessel point and realizing the determination of the direction vector at the blood vessel point, which facilitates the subsequent attention sequence.

[0054] Another optional technical solution includes a blood vessel classification model comprising a blood vessel classification module and a blood vessel staining module. After adjusting the blood vessel classification model according to the attention sequence, the solution further includes: updating the blood vessel classification model based on the adjustment results; acquiring a target blood vessel image and inputting the target blood vessel image into the blood vessel classification module; and inputting the output of the blood vessel classification module into the blood vessel staining module to obtain the blood vessel staining results for the target blood vessel image.

[0055] It is important to note that the blood vessel classification model can include a blood vessel classification module for classifying blood vessels and a blood vessel staining module for staining blood vessels. That is, the blood vessel classification model can not only classify blood vessels but also stain them. If only blood vessel classification is required, the blood vessel classification model can be adjusted based on the attention sequence, and then updated according to the adjustment result. A target blood vessel image is acquired and input into the blood vessel classification module. The blood vessel classification result for the target blood vessel image is obtained based on the output of the blood vessel classification module. If blood vessel staining is required, the blood vessel classification model can be adjusted based on the attention sequence, and then updated according to the adjustment result. A target blood vessel image is acquired and input into the blood vessel classification module. The output of the blood vessel classification module is then input into the blood vessel staining module to obtain the blood vessel staining result for the target blood vessel image.

[0056] The technical solution of this invention can introduce the focus on the direction of blood vessels into a blood vessel classification model that can perform blood vessel staining, thereby enabling blood vessel staining without manual intervention or the use of complex algorithms, and improving the efficiency and accuracy of blood vessel classification models that can perform blood vessel staining.

[0057] Figure 3 This is a flowchart of another method for adjusting a blood vessel classification model provided in this embodiment of the invention. This embodiment is based on and optimized from the above-described technical solutions. In this embodiment, optionally, determining an attention sequence based on at least one blood vessel point and the direction vector corresponding to each blood vessel point includes: for each of the at least one blood vessel point, determining a corresponding point from the at least one blood vessel point based on the at least one blood vessel point and the direction vector corresponding to each blood vessel point; and determining an attention sequence based on the corresponding points corresponding to each of the at least one blood vessel point. The explanations of terms that are the same as or corresponding to those in the above embodiments are not repeated here.

[0058] See Figure 3 The method in this embodiment may specifically include the following steps:

[0059] S201. Obtain a blood vessel point cloud image, a blood vessel classification model, and a trained feature extraction model, wherein the blood vessel point cloud image includes at least one blood vessel point.

[0060] S202. Input the blood vessel point cloud image into the feature extraction model, and obtain the point features corresponding to at least one blood vessel point based on the output of the feature extraction model.

[0061] S203. For each blood vessel point in at least one blood vessel point, determine the direction vector of the blood vessel point based on the point features corresponding to the blood vessel point.

[0062] S204. For each of the at least one blood vessel points, based on the at least one blood vessel point and the direction vector corresponding to the at least one blood vessel point, determine the corresponding point from the at least one blood vessel point.

[0063] In this embodiment of the invention, for each of the at least one blood vessel points, a corresponding point that can characterize the front-to-back relationship with the blood vessel point can be determined from the at least one blood vessel point based on the at least one blood vessel point and the direction vector corresponding to the at least one blood vessel point.

[0064] S205. Determine the attention sequence based on the corresponding points of at least one blood vessel point.

[0065] In this embodiment of the invention, a blood vessel point sequence corresponding to each of the at least one blood vessel point can be determined based on the corresponding points corresponding to the at least one blood vessel point. The blood vessel point sequence is a sequence of blood vessel points that represent the sequential relationship between blood vessel points. An attention sequence is determined based on the blood vessel point sequence corresponding to the at least one blood vessel point.

[0066] S206. Adjust the blood vessel classification model based on the attention sequence.

[0067] The technical solution of this invention, for each of at least one blood vessel points, determines a corresponding point from the at least one blood vessel point based on the at least one blood vessel point and its corresponding direction vector; and determines an attention sequence based on the corresponding points of the at least one blood vessel point. By determining the corresponding points, this technical solution allows for the determination of an attention sequence that characterizes the correlation between blood vessel points, thereby facilitating subsequent adjustments to the blood vessel classification model based on the attention sequence to obtain a more accurate blood vessel classification model.

[0068] Figure 4This is a flowchart of another method for adjusting a blood vessel classification model provided in this embodiment of the invention. This embodiment is based on and optimized from the above-mentioned technical solutions. In this embodiment, optionally, the method for adjusting the blood vessel classification model further includes: for each of the at least one blood vessel points, normalizing the direction vector of the blood vessel point to obtain a unit vector; constructing a target K-dimensional tree based on the at least one blood vessel point and the unit vector corresponding to the at least one blood vessel point; for each of the at least one blood vessel points, determining the corresponding point from the at least one blood vessel point based on the at least one blood vessel point and the direction vector corresponding to the at least one blood vessel point, including: for each of the at least one blood vessel points, determining the corresponding point from the at least one blood vessel point according to the target K-dimensional tree. The explanations of terms that are the same as or corresponding to those in the above embodiments are not repeated here.

[0069] See Figure 4 The method in this embodiment may specifically include the following steps:

[0070] S301. Obtain a blood vessel point cloud image, a blood vessel classification model, and a trained feature extraction model, wherein the blood vessel point cloud image includes at least one blood vessel point.

[0071] S302. Input the blood vessel point cloud image into the feature extraction model, and obtain the point features corresponding to at least one blood vessel point based on the output of the feature extraction model.

[0072] S303. For each blood vessel point in at least one blood vessel point, determine the direction vector of the blood vessel point based on the point features corresponding to the blood vessel point.

[0073] S304. Normalize the direction vector of the blood vessel point to obtain a unit vector.

[0074] In this embodiment of the invention, for each of the at least one blood vessel points, the direction vector of the blood vessel point, including offset information such as the direction offset angle, can be normalized to obtain a unit vector that better represents the direction of the blood vessel at the blood vessel point. For example, the direction vector of the blood vessel point can represent the direction of the blood vessel at the blood vessel point as north-south with a clockwise offset of 5 degrees. After normalizing the direction vector, a unit vector representing the direction of the blood vessel at the blood vessel point as north-south can be obtained.

[0075] S305. Construct the target K-dimensional tree based on at least one blood vessel point and the unit vector corresponding to each blood vessel point.

[0076] In this embodiment of the invention, a target K-dimensional tree can be constructed by mapping at least one relatively discrete blood vessel point and the unit vector corresponding to at least one blood vessel point.

[0077] S306. For each blood vessel point in at least one blood vessel point, determine the corresponding point from the at least one blood vessel point according to the target K-dimensional tree.

[0078] In this embodiment of the invention, for each of the at least one blood vessel points, a blood vessel point that is associated with the blood vessel point can be found as a corresponding point in at least one blood vessel point in the target K-dimensional tree.

[0079] S307. Determine the attention sequence based on the corresponding points of at least one blood vessel point.

[0080] S308. Adjust the blood vessel classification model based on the attention sequence.

[0081] The technical solution of this invention involves normalizing the direction vector of each of the at least one blood vessel points to obtain a unit vector; constructing a target K-dimensional tree based on the at least one blood vessel point and its corresponding unit vector; and determining the corresponding point from the at least one blood vessel point based on the target K-dimensional tree. This technical solution eliminates the influence of dimensions on subsequent adjustments to the blood vessel classification model by obtaining the normalized unit vector, and the construction of the target K-dimensional tree accelerates the determination of the corresponding point, thereby speeding up the determination of the attention sequence and saving time spent adjusting the blood vessel classification model.

[0082] Figure 5 This is a flowchart of another method for adjusting a blood vessel classification model provided in this embodiment of the invention. This embodiment is based on and optimized from the above-mentioned technical solutions. In this embodiment, optionally, determining the corresponding point of a blood vessel point from at least one blood vessel point according to the target K-dimensional tree includes: taking the blood vessel point as the current point; starting from the current point, moving along the direction represented by the unit vector of the current point by a preset step size, and determining target candidate points from at least one blood vessel point in the target K-dimensional tree according to the obtained movement result, and updating the target candidate points as the current point; repeating the step of moving along the direction represented by the unit vector of the current point by a preset step size starting from the current point; if the number of determined target candidate points is greater than or equal to a preset number, taking each determined target candidate point as the corresponding point of the blood vessel point. The explanations of terms that are the same as or corresponding to those in the above embodiments are not repeated here.

[0083] See Figure 5 The method in this embodiment may specifically include the following steps:

[0084] S401. Obtain a blood vessel point cloud image, a blood vessel classification model, and a trained feature extraction model, wherein the blood vessel point cloud image includes at least one blood vessel point.

[0085] S402. Input the blood vessel point cloud image into the feature extraction model, and obtain the point features corresponding to at least one blood vessel point based on the output of the feature extraction model.

[0086] S403. For each blood vessel point in at least one blood vessel point, determine the direction vector of the blood vessel point based on the point features corresponding to the blood vessel point.

[0087] S404. Normalize the direction vector of the blood vessel point to obtain a unit vector.

[0088] S405. Construct the target K-dimensional tree based on at least one blood vessel point and the unit vector corresponding to each blood vessel point.

[0089] S406. For each of the at least one blood vessel points, take the blood vessel point as the current point.

[0090] Here, the current point is a blood vessel point among at least one blood vessel point selected as a target candidate point based on the current demand. The target candidate point is a blood vessel point among at least one blood vessel point selected as a corresponding point.

[0091] S407. Starting from the current point, move a preset step length along the direction represented by the unit vector of the current point, and based on the obtained movement result, determine the target candidate point from at least one blood vessel point in the target K-dimensional tree, and update the target candidate point to the current point.

[0092] The preset step size is the pre-set movement step size of the current point; the preset step size can be set according to the requirements, for example, it can be set according to the classification accuracy of the blood vessel classification model.

[0093] It should be noted that the movement result obtained by moving a preset step length along the direction represented by the unit vector of the current point from the current point cannot directly point to the target candidate point. Since the blood vessel point cloud exists discretely, the position pointed to by the movement result may or may not have blood vessel points. Therefore, in the embodiments of the present invention, the obtained movement result is not directly used as the target candidate point. Instead, the target candidate point is determined from at least one blood vessel point in the target K-dimensional tree based on the obtained movement result, and the target candidate point is updated to the current point.

[0094] S408. Return to execute S407, and if the number of determined target candidate points is greater than or equal to the preset number, use each determined target candidate point as the corresponding point of the blood vessel point.

[0095] The preset quantity is the number of corresponding points corresponding to the blood vessel points required by the preset demand; the preset quantity can be set according to the demand, for example, it can be set according to the classification accuracy of the blood vessel classification model required.

[0096] In this embodiment of the invention, if the number of target candidate points is greater than or equal to the preset number, it means that all the corresponding points corresponding to the blood vessel points to be determined have been determined, and each of the determined target candidate points can be used as the corresponding points corresponding to the blood vessel points.

[0097] S409. Determine the attention sequence based on the corresponding points of at least one blood vessel point.

[0098] S410. Adjust the blood vessel classification model based on the attention sequence.

[0099] The technical solution of this invention involves taking a blood vessel point as the current point; starting from the current point, moving it along the direction represented by the unit vector of the current point by a preset step size; and determining target candidate points from at least one blood vessel point in the target K-dimensional tree based on the obtained movement result, updating the target candidate points as the current point; repeating the step of moving the target candidate points from the current point along the direction represented by the unit vector of the current point by a preset step size; and when the number of determined target candidate points is greater than or equal to a preset number, using each determined target candidate point as the corresponding point of the blood vessel point. This technical solution of the present invention can determine target candidate points from at least one blood vessel point in the target K-dimensional tree based on the movement result obtained by moving the target candidate points along the direction represented by the unit vector of the current point by a preset step size, thereby more accurately determining the corresponding points of the blood vessel points.

[0100] An optional technical solution, after determining target candidate points from at least one blood vessel point in the target K-dimensional tree, further includes: determining candidate information of the target candidate points from the target K-dimensional tree based on the obtained movement results; after using each determined target candidate point as a corresponding point of the blood vessel point, further includes: determining the corresponding information of the corresponding point of the blood vessel point based on the candidate information of each target candidate point; determining an attention sequence based on the corresponding points of at least one blood vessel point, including: determining the attention sequence based on the corresponding points of at least one blood vessel point and the corresponding information of the corresponding points of at least one blood vessel point.

[0101] It is important to note that since the target K-dimensional tree includes not only at least one blood vessel point, but also relevant information about each blood vessel point, etc., to improve the accuracy of the subsequent attention sequence determination and to represent more features, candidate information for target candidate points can be determined from the target K-dimensional tree based on the obtained movement results. This candidate information includes relevant information about the target candidate points themselves. Furthermore, based on the candidate information of each target candidate point, corresponding information for the corresponding points of the blood vessel points can be determined. This corresponding information includes relevant information about the corresponding points themselves. The attention sequence is then determined based on the corresponding information of each blood vessel point and the corresponding information of each corresponding point of at least one blood vessel point. The relevant information about the blood vessel point itself may include, for example, a unit vector, adjacent blood vessel points based on the unit vector, a label and / or position set for the blood vessel point, etc. In this embodiment of the invention, the content of the relevant information about the blood vessel point itself is not specifically limited.

[0102] In this embodiment of the invention, by determining the attention sequence based on the corresponding information of the corresponding points of at least one blood vessel point and the corresponding points of at least one blood vessel point, the determined attention sequence can represent more blood vessel point features, such as the directional relationship between blood vessel points, the front-back correspondence relationship between blood vessel points, etc., thereby improving the classification accuracy of the blood vessel classification model adjusted according to the attention sequence.

[0103] Another optional technical solution uses the blood vessel point as the current point, including: using the blood vessel point as the starting point and the starting point as the current point; starting from the current point, moving along the direction represented by the unit vector of the current point by a preset step size, and determining a target candidate point from at least one blood vessel point in the target K-dimensional tree based on the obtained movement result, including: when the current point is the starting point, determining the positive direction based on the unit vector of the current point, and taking the direction opposite to the positive direction as the negative direction; starting from the current point, moving along the positive direction by a preset step size, and determining a first candidate point from at least one blood vessel point in the target K-dimensional tree based on the obtained movement result; from the current point... Starting from the current point, move a preset step length in the negative direction, and determine a second candidate point from at least one blood vessel point in the target K-dimensional tree based on the movement result; use the first and second candidate points as the target candidate point; if the current point is not the starting point, determine the current direction of the current point based on the current direction of the historical point and the unit vector of the current point, where the historical point is the blood vessel point that was most recently used as the current point before the current point; starting from the current point, move a preset step length in the current direction of the current point, and determine a third candidate point from at least one blood vessel point in the target K-dimensional tree based on the movement result; use the third candidate point as the target candidate point.

[0104] The starting point is the blood vessel point from which the target candidate points are determined.

[0105] It is important to note that a unit vector may only represent the direction of a blood vessel at a given point, but it may not necessarily point to a specific direction. For example, a unit vector representing the direction of a blood vessel at a given point as north-south does not necessarily mean the blood vessel is flowing south. In other words, a unit vector cannot represent the direction of blood flow within a blood vessel. Therefore, when the current point is the starting point, there may be two directions represented by the unit vector to determine the target candidate point. Based on the characteristics of the unit vector, in this embodiment of the invention, when the current point is the starting point, the positive direction can be determined based on the unit vector of the current point. There is no specific limitation on how the positive direction is determined based on the unit vector of the current point. A fixed direction can be used as the positive direction, or one of the two directions represented by the unit vector can be chosen as the positive direction, and the opposite direction can be used as the negative direction. For example, when the unit vector represents the direction of the blood vessel at a given point as north-south, the positive direction could be south and the negative direction could be north, or the positive direction could be north and the negative direction could be south. That is, the positive and negative directions can respectively correspond to the two directions represented by the unit vector.

[0106] In this embodiment of the invention, starting from the current point, a preset step length can be moved in the positive direction. Based on the obtained movement result, for example, the positive direction can be multiplied by the preset step length. Starting from the current point, the movement can be performed based on the multiplication result. A first candidate point is determined from at least one blood vessel point in the target K-dimensional tree. This first candidate point is the target candidate point determined based on the positive direction. Starting from the current point, a preset step length can be moved in the negative direction. For example, the negative direction can be multiplied by the preset step length. Starting from the current point, the movement can be performed based on the multiplication result. Based on the obtained movement result, a second candidate point is determined from at least one blood vessel point in the target K-dimensional tree. This second candidate point is the target candidate point determined based on the negative direction. The first candidate point and the second candidate point are used as the target candidate point.

[0107] It is important to note that when the current point is not the starting point, meaning the current point already has a roughly determined direction, using the same method to determine the target candidate point as the starting point could lead to duplicate or incorrect determination of the target candidate point. Therefore, in this embodiment of the invention, when the current point is not the starting point, the current direction of the current point can be determined based on the current direction of the most recent historical point that served as the current point and the unit vector of the current point. For example, if the historical point is the starting point, and the unit vector of the historical point indicates that the direction of the blood vessel at the blood vessel point is north-south, and the current point is a target candidate point determined based on the negative direction of the historical point indicating south, if the unit vector of the current point also indicates that the direction of the blood vessel at the blood vessel point is north-south, then the current direction of the current point can be determined as south based on the current direction of the historical point (which is south) and the unit vector of the current point. Here, the current direction is the direction in which the current point needs to move.

[0108] In this embodiment of the invention, starting from the current point, a preset step size is moved along the current direction of the current point. For example, the current direction can be multiplied by the preset step size. Starting from the current point, the movement is performed according to the multiplication result. Based on the obtained movement result, a third candidate point is determined from at least one blood vessel point in the target K-dimensional tree. The third candidate point is the target candidate point determined based on the current direction. The third candidate point is used as the target candidate point.

[0109] In this embodiment of the invention, setting different target candidate points for the current point that is the starting point and for the current point that is not the starting point can avoid the situation of repeated or incorrect determination of target candidate points, thereby improving the accuracy of the attention sequence obtained in the subsequent determination.

[0110] Another optional technical solution involves moving a preset step length along the direction represented by the unit vector of the current point, starting from the current point, and determining target candidate points from at least one blood vessel point in the target K-dimensional tree based on the obtained movement result. This includes: moving a preset step length along the direction represented by the unit vector of the current point, and determining the expected information corresponding to the current point based on the obtained movement result; and using the blood vessel point in the target K-dimensional tree that matches the expected information as the target candidate point.

[0111] It is important to note that since the blood vessel point cloud is discrete, the movement result obtained by moving a preset step length along the direction represented by the unit vector of the current point from the current point cannot directly point to the target candidate point. It may point to a location where no blood vessel points exist, or even to an area outside the blood vessel point cloud. However, the location pointed to by the movement result is the expected position of the target candidate point under ideal conditions. Therefore, a preset step length can be moved along the direction represented by the unit vector of the current point from the current point, and the expected information corresponding to the current point can be determined based on the obtained movement result. This expected information can represent the relevant information of the target candidate point under ideal conditions. For example, the expected information may include the position information of the target candidate point under ideal conditions, the orientation and / or distance of the target candidate point relative to the current point under ideal conditions, the gray level of the target candidate point under ideal conditions, and / or the feature value with the smallest value obtained by the filtering algorithm under ideal conditions, etc. In the embodiments of this invention, the content of the expected information is not specifically limited.

[0112] In this embodiment of the invention, a blood vessel point matching the expected information can be determined from at least one blood vessel point in the target K-dimensional tree as a target candidate point based on the expected information. For example, the blood vessel point with the highest matching degree with the expected information can be determined from at least one blood vessel point in the target K-dimensional tree as a target candidate point based on the expected information. The criteria for judging the matching degree between the expected information and the blood vessel point can be, for example, the matching degree between the position represented by the expected information and the position of the blood vessel point, or the matching degree between the gray level represented by the expected information and the gray level of the blood vessel point, etc. In this embodiment of the invention, the criteria for judging the matching degree between the expected information and the blood vessel point are not specifically limited.

[0113] In this embodiment of the invention, target candidate points can be determined based on expected information, which can achieve the determination of target candidate points that are more in line with the requirements and have higher accuracy.

[0114] To better understand the technical solutions of the above embodiments of the present invention, an optional example is provided herein. For example, see... Figure 6 The process involves acquiring a blood vessel point cloud image and a trained feature extraction model; inputting the blood vessel point cloud image into the feature extraction model, and obtaining point features corresponding to at least one blood vessel point based on the output of the feature extraction model; inputting the point features corresponding to at least one blood vessel point into a filtering module using the Frangi filtering algorithm to obtain the direction vector corresponding to at least one blood vessel point; and determining an attention sequence based on the direction vector corresponding to at least one blood vessel point.

[0115] Figure 7This is a structural block diagram of a device for adjusting a vascular classification model provided in an embodiment of the present invention. This device is used to execute the method for adjusting a vascular classification model provided in any of the above embodiments. This device and the method for adjusting a vascular classification model in the above embodiments belong to the same inventive concept. Details not described in detail in the embodiments of the device for adjusting a vascular classification model can be found in the embodiments of the method for adjusting a vascular classification model described above. See also... Figure 7 The device may specifically include: a feature extraction model acquisition module 510, a point feature acquisition module 520, a direction vector determination module 530, and a blood vessel classification model adjustment module 540.

[0116] The feature extraction model acquisition module 510 is used to acquire a blood vessel point cloud image, a blood vessel classification model, and a trained feature extraction model. The blood vessel point cloud image includes at least one blood vessel point.

[0117] The point feature acquisition module 520 is used to input the blood vessel point cloud image into the feature extraction model, and obtain the point features corresponding to at least one blood vessel point based on the output of the feature extraction model.

[0118] The direction vector determination module 530 is used to determine the direction vector of each blood vessel point based on the point features corresponding to the blood vessel point for each blood vessel point in at least one blood vessel point.

[0119] The blood vessel classification model adjustment module 540 is used to determine the attention sequence based on at least one blood vessel point and the direction vector corresponding to at least one blood vessel point, and adjust the blood vessel classification model according to the attention sequence.

[0120] Optionally, the blood vessel classification model adjustment module 540 may include:

[0121] The corresponding point determination submodule is used to determine the corresponding point of each of the at least one blood vessel points based on the at least one blood vessel point and the direction vector corresponding to the at least one blood vessel point.

[0122] The attention sequence determination submodule is used to determine the attention sequence based on the corresponding points corresponding to at least one blood vessel point.

[0123] Optionally, based on the above-described apparatus, the apparatus may further include:

[0124] The unit vector acquisition module is used to normalize the direction vector of each blood vessel point in at least one blood vessel point to obtain a unit vector;

[0125] The target K-dimensional tree construction module is used to construct a target K-dimensional tree based on at least one blood vessel point and the unit vector corresponding to at least one blood vessel point.

[0126] The corresponding point determination submodule may include:

[0127] The corresponding point determination unit is used to determine the corresponding point of each blood vessel point from at least one blood vessel point according to the target K-dimensional tree.

[0128] Optionally, based on the above-described device, the corresponding point determination unit may include:

[0129] The current point is used as a sub-unit to select the blood vessel point as the current point;

[0130] The current point update sub-unit is used to move a preset step size from the current point along the direction represented by the unit vector of the current point, and based on the obtained movement result, determine the target candidate point from at least one blood vessel point in the target K-dimensional tree, and update the target candidate point to the current point;

[0131] The repeat execution sub-unit is used to repeatedly execute the step of moving a preset step size from the current point along the direction represented by the unit vector of the current point;

[0132] The corresponding point is used as a sub-unit to determine each target candidate point as the corresponding point of the blood vessel point when the number of determined target candidate points is greater than or equal to the preset number.

[0133] Optionally, based on the above-described apparatus, the apparatus may further include:

[0134] The alternative information determination module is used to determine the alternative information of the target alternative point from the target K-dimensional tree based on the obtained movement result after determining the target alternative point from at least one blood vessel point in the target K-dimensional tree.

[0135] The corresponding information determination module is used to determine the corresponding information of the corresponding point of the blood vessel point based on the candidate information of each target candidate point after determining each target candidate point as the corresponding point of the blood vessel point.

[0136] The attention sequence determination submodule may include:

[0137] The attention sequence determination unit is used to determine the attention sequence based on the corresponding information of at least one blood vessel point and the corresponding information of the corresponding points of at least one blood vessel point.

[0138] Optionally, based on the above device, the current point, as a sub-unit, can be specifically used for:

[0139] Use the blood vessel point as the starting point and the starting point as the current point;

[0140] The current point updates the sub-cell, which can be specifically used for:

[0141] If the current point is the starting point, determine the positive direction based on the unit vector of the current point, and take the direction opposite to the positive direction as the negative direction;

[0142] Starting from the current point, move a preset step length in the positive direction, and determine the first candidate point from at least one blood vessel point in the target K-dimensional tree based on the movement result.

[0143] Starting from the current point, move a preset step length in the negative direction, and determine a second candidate point from at least one blood vessel point in the target K-dimensional tree based on the movement result.

[0144] The first and second alternative points are selected as target alternative points.

[0145] If the current point is not the starting point, the current direction of the current point is determined based on the current direction of the historical point and the unit vector of the current point. The historical point is the blood vessel point that is closest to the current point before the current point.

[0146] Starting from the current point, move a preset step length along the current direction of the current point, and determine a third candidate point from at least one blood vessel point in the target K-dimensional tree based on the movement result.

[0147] The third alternative point will be selected as the target alternative point.

[0148] Optionally, based on the above device, the current point updates the sub-unit, which can be specifically used for:

[0149] Starting from the current point, move a preset step size along the direction represented by the unit vector of the current point, and determine the expected information corresponding to the current point based on the movement result.

[0150] The blood vessel points that match the expected information and are identified from the target K-dimensional tree will be used as target candidate points.

[0151] Optionally, the direction vector determination module 530 may include:

[0152] The feature vector acquisition submodule is used to process the point features corresponding to the blood vessel points based on the target filtering algorithm to obtain at least one feature value, and the feature vectors corresponding to the at least one feature value.

[0153] The direction vector is used as a submodule to take the eigenvector corresponding to the eigenvalue with the smallest value among at least one eigenvalues ​​as the direction vector of the blood vessel point.

[0154] Optionally, the blood vessel classification model includes a blood vessel classification module and a blood vessel staining module;

[0155] The device may also include:

[0156] The blood vessel classification model update module is used to update the blood vessel classification model based on the adjustment results after adjusting the blood vessel classification model according to the attention sequence.

[0157] The target blood vessel image input module is used to acquire the target blood vessel image and input the target blood vessel image into the blood vessel classification module;

[0158] The blood vessel staining result acquisition module is used to input the output results of the blood vessel classification module into the blood vessel staining module to obtain the blood vessel staining results for the target blood vessel image.

[0159] Optionally, the feature extraction model is implemented based on the Unet network, which includes at least sparse convolutional layers.

[0160] The blood vessel classification model adjustment device provided in this embodiment of the invention includes a feature extraction model acquisition module for acquiring a blood vessel point cloud image, a blood vessel classification model, and a trained feature extraction model. The blood vessel point cloud image includes at least one blood vessel point. A point feature acquisition module inputs the blood vessel point cloud image into the feature extraction model and obtains point features corresponding to each of the at least one blood vessel point based on the output of the feature extraction model. A direction vector determination module determines the direction vector of each blood vessel point based on its corresponding point features. A blood vessel classification model adjustment module determines an attention sequence based on the at least one blood vessel point and its corresponding direction vector, and adjusts the blood vessel classification model according to the attention sequence. This device, by adjusting the blood vessel classification model based on an attention sequence focusing on the direction of blood vessels, can introduce attention to the direction of blood vessels into the blood vessel classification model, thereby improving the classification accuracy of the blood vessel classification model.

[0161] The vascular classification model adjustment device provided in this embodiment of the invention can execute the vascular classification model adjustment method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method.

[0162] It is worth noting that in the embodiments of the above-mentioned blood vessel classification model adjustment device, the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy differentiation and are not used to limit the scope of protection of the present invention.

[0163] Figure 8A schematic diagram of an electronic device 10 that can be used to implement embodiments of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0164] like Figure 8 As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded into the RAM 13 from storage unit 18. The RAM 13 can also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0165] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0166] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as the adjustment method for a blood vessel classification model.

[0167] In some embodiments, the method for adjusting the vascular classification model can be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the method for adjusting the vascular classification model described above can be performed. Alternatively, in other embodiments, processor 11 can be configured to perform the method for adjusting the vascular classification model by any other suitable means (e.g., by means of firmware).

[0168] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0169] Computer programs used to implement the methods of the present invention can be written in any combination of one or more programming languages. These computer programs can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The computer programs can be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0170] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0171] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0172] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or computing systems that include middleware components (e.g., application servers), or computing systems that include frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0173] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0174] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0175] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A method for adjusting a blood vessel classification model, characterized in that, include: The process involves acquiring a blood vessel point cloud image, a blood vessel classification model, and a trained feature extraction model, wherein the blood vessel point cloud image includes at least one blood vessel point. The blood vessel point cloud image is input into the feature extraction model, and the point features corresponding to the at least one blood vessel point are obtained according to the output of the feature extraction model. For each of the at least one blood vessel point, a direction vector of the blood vessel point is determined based on the point features corresponding to the blood vessel point, wherein the direction vector represents the direction of the blood vessel at the blood vessel point. Based on the at least one blood vessel point and the direction vector corresponding to the at least one blood vessel point, an attention sequence is determined, and the blood vessel classification model is adjusted according to the attention sequence, wherein the attention sequence is an attention sequence that characterizes the association and direction relationship between the at least one blood vessel point; The step of determining the attention sequence based on the at least one blood vessel point and the direction vector corresponding to the at least one blood vessel point includes: For each of the at least one blood vessel points, a corresponding point is determined from the at least one blood vessel point based on the at least one blood vessel point and the direction vector corresponding to the at least one blood vessel point. An attention sequence is determined based on the corresponding points of each of the at least one blood vessel point; The method further includes: For each of the at least one blood vessel points, the direction vector of the blood vessel point is normalized to obtain a unit vector; Construct a target K-dimensional tree based on the at least one blood vessel point and the unit vector corresponding to the at least one blood vessel point; For each of the at least one blood vessel points, determining the corresponding point based on the at least one blood vessel point and its corresponding direction vector includes: For each of the at least one blood vessel points, a corresponding point is determined from the at least one blood vessel point according to the target K-dimensional tree.

2. The method according to claim 1, characterized in that, The step of determining the corresponding point of the blood vessel point from the at least one blood vessel point according to the target K-dimensional tree includes: Use the aforementioned blood vessel point as the current point; Starting from the current point, move a preset step length along the direction represented by the unit vector of the current point, and based on the obtained movement result, determine a target candidate point from at least one blood vessel point in the target K-dimensional tree, and update the target candidate point to the current point; Repeat the step of moving a preset step size from the current point along the direction represented by the unit vector of the current point; If the number of identified target candidate points is greater than or equal to the preset number, each identified target candidate point will be used as the corresponding point of the blood vessel point.

3. The method according to claim 2, characterized in that, After determining the target candidate point from at least one vessel point in the target K-dimensional tree, the method further includes: Based on the obtained movement results, the candidate information of the target candidate point is determined from the target K-dimensional tree; After determining each of the selected target candidate points as the corresponding points for the blood vessel points, the method further includes: Based on the candidate information of each target candidate point, the corresponding information of the corresponding point corresponding to the blood vessel point is determined; Determining the attention sequence based on the corresponding points of the at least one blood vessel point includes: An attention sequence is determined based on the corresponding points corresponding to the at least one blood vessel point and the corresponding information of the corresponding points corresponding to the at least one blood vessel point.

4. The method according to claim 2, characterized in that, The step of using the blood vessel point as the current point includes: The blood vessel point is taken as the starting point, and the starting point is taken as the current point; The step of moving a preset step size from the current point along the direction represented by the unit vector of the current point, and determining a target candidate point from at least one blood vessel point in the target K-dimensional tree based on the obtained movement result, includes: When the current point is the starting point, the positive direction is determined based on the unit vector of the current point, and the direction opposite to the positive direction is taken as the negative direction. Starting from the current point, move a preset step length along the positive direction, and determine a first candidate point from at least one blood vessel point in the target K-dimensional tree based on the obtained movement result; Starting from the current point, move a preset step length along the negative direction, and determine a second candidate point from at least one blood vessel point in the target K-dimensional tree based on the obtained movement result; The first candidate point and the second candidate point are selected as target candidate points; If the current point is not the starting point, the current direction of the current point is determined based on the current direction of the historical point and the unit vector of the current point, wherein the historical point is the blood vessel point that was the most recent point before the current point; Starting from the current point, move a preset step length along the current direction of the current point, and determine a third candidate point from at least one blood vessel point in the target K-dimensional tree based on the obtained movement result; The third alternative point is selected as the target alternative point.

5. The method according to claim 2, characterized in that, The step of moving a preset step size from the current point along the direction represented by the unit vector of the current point, and determining a target candidate point from at least one blood vessel point in the target K-dimensional tree based on the obtained movement result, includes: Starting from the current point, move a preset step length along the direction represented by the unit vector of the current point, and determine the expected information corresponding to the current point based on the obtained movement result; The blood vessel points that match the expected information and are determined from the target K-dimensional tree are used as target candidate points.

6. The method according to claim 1, characterized in that, The step of determining the direction vector of the blood vessel point based on the point features corresponding to the blood vessel point includes: Based on the target filtering algorithm, the point features corresponding to the blood vessel points are processed to obtain at least one feature value, and the feature vectors corresponding to the at least one feature value respectively; The feature vector corresponding to the feature value with the smallest value among the at least one feature values ​​is used as the direction vector of the blood vessel point.

7. The method according to claim 1, characterized in that, The blood vessel classification model includes a blood vessel classification module and a blood vessel staining module; After adjusting the blood vessel classification model according to the attention sequence, the method further includes: Update the blood vessel classification model based on the obtained adjustment results; Acquire a target blood vessel image and input the target blood vessel image into the blood vessel classification module; The output of the blood vessel classification module is input into the blood vessel staining module to obtain the blood vessel staining result for the target blood vessel image.

8. The method according to any one of claims 1-7, characterized in that, The feature extraction model is implemented based on the Unet network, which includes at least sparse convolutional layers.

9. An adjustment device for a blood vessel classification model, characterized in that, include: The feature extraction model acquisition module is used to acquire a blood vessel point cloud image, a blood vessel classification model, and a trained feature extraction model, wherein the blood vessel point cloud image includes at least one blood vessel point. The point feature acquisition module is used to input the blood vessel point cloud image into the feature extraction model, and obtain the point features corresponding to the at least one blood vessel point according to the output of the feature extraction model. A direction vector determination module is used to determine the direction vector of each blood vessel point in the at least one blood vessel point according to the point features corresponding to the blood vessel point, wherein the direction vector represents the direction of the blood vessel at the blood vessel point. The blood vessel classification model adjustment module is used to determine an attention sequence based on the at least one blood vessel point and the direction vector corresponding to the at least one blood vessel point, and to adjust the blood vessel classification model according to the attention sequence, wherein the attention sequence is an attention sequence that characterizes the association and direction relationship between the at least one blood vessel point; The blood vessel classification model adjustment module includes: The corresponding point determination submodule is used to determine, for each of the at least one blood vessel points, a corresponding point corresponding to the blood vessel point based on the at least one blood vessel point and the direction vector corresponding to the at least one blood vessel point respectively; The attention sequence determination submodule is used to determine the attention sequence based on the corresponding points corresponding to the at least one blood vessel point. The device further includes: The unit vector acquisition module is used to normalize the direction vector of each of the at least one blood vessel points to obtain a unit vector. The target K-dimensional tree construction module is used to construct a target K-dimensional tree based on the at least one blood vessel point and the unit vector corresponding to the at least one blood vessel point. The corresponding point determination submodule includes: The corresponding point determination unit is used to determine the corresponding point of each blood vessel point from the at least one blood vessel point according to the target K-dimensional tree.

10. An electronic device, characterized in that, include: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor to cause the at least one processor to perform the adjustment method of the blood vessel classification model as described in any one of claims 1-8.

11. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause a processor to execute the adjustment method of the blood vessel classification model as described in any one of claims 1-8.

Images