A method and system for identifying a vessel in a living organism

Abstract

The embodiment of the specification provides a kind of recognition method and system of organism vessel, the method includes obtaining the three-dimensional image data of organism and its first segmentation result, second segmentation result, at least one organism vessel in second segmentation result is type pending vessel, fusion first segmentation result and second segmentation result, obtain fusion result;Skeletonization processing is carried out to first segmentation result, obtain first vessel skeleton set, wherein first vessel skeleton set includes at least one type determined vessel skeleton;Skeletonization processing is carried out to fusion result, obtain the vessel skeleton of type pending vessel;The type determined vessel skeleton in first vessel skeleton set is obtained as reference vessel skeleton set with the minimum spatial distance of the vessel skeleton of type pending vessel less than preset value D;Determine the nearest point group of the vessel skeleton of type pending vessel and reference vessel skeleton in reference vessel skeleton set;Determine the type of type pending vessel based on nearest point group.

Description

TECHNICAL FIELD

[0001] The present specification relates to the field of medical image processing, and in particular, to a method and system for identifying vessels in a living body. BACKGROUND

[0002] In the biomedical field, the segmentation of human vessels has been widely studied and is of great significance for detecting diseases and surgery. However, automatic segmentation of multi-class vessels in images under some scenarios is difficult, and the contrast of the fine branch region at the end of the vessel is low, which is also a big challenge for automatic segmentation and classification. In the past 20 years, the technology for automatic segmentation of multi-class vessels has been continuously improved and optimized, but the correct segmentation of five to six levels of vessels required by the clinic for some organ tissues, such as CT pulmonary arteries and veins, is still a great challenge. Some methods can accurately distinguish different classes of vessels, but the richness cannot reach five to six levels of vessels, while another part of the method can divide five to six levels of vessels or even seven levels of vessels, but there are a large number of vessel class errors. Most methods use a single model to segment vessels, and the accuracy and richness cannot be achieved.

[0003] Therefore, there is an urgent need for an automatic multi-class vessel segmentation method that is both accurate and rich. SUMMARY

[0004] One or more embodiments of the present specification provide a method for identifying vessels in a living body. The method for identifying vessels in a living body includes: obtaining three-dimensional image data of a living body; obtaining a first segmentation result of the three-dimensional image data based on a first segmentation model, performing skeletonization processing on the first segmentation result, and obtaining a first vessel skeleton set, wherein the first vessel skeleton set includes at least one type-determined vessel skeleton; obtaining a second segmentation result of the three-dimensional image data based on a second segmentation model, at least one vessel in the living body in the second segmentation result being a type-pending vessel; fusing the first segmentation result and the second segmentation result to obtain a fusion result; performing skeletonization processing on the fusion result to obtain a vessel skeleton of the type-pending vessel; obtaining a reference vessel skeleton set from the first vessel skeleton set, the reference vessel skeleton set being a type-determined vessel skeleton with a minimum spatial distance less than a preset value D from the vessel skeleton of the type-pending vessel; determining a nearest point set formed by two points with the smallest spatial distance between the vessel skeleton of the type-pending vessel and a reference vessel skeleton in the reference vessel skeleton set; and determining the type of the type-pending vessel based on the nearest point set.

[0005] The one or more embodiments of the specification provide a system for identifying a vessel in a living body. The system for identifying the vessel in the living body comprises a display device configured to display at least one of a first segmentation result, a second segmentation result, a fusion result, a nearest point group, and a type of a type-undetermined vessel; and a control system comprising one or more processors and a memory, the memory comprising programming instructions adapted to cause the one or more processors to perform operations comprising: obtaining three-dimensional image data of a living body; obtaining a first segmentation result of the three-dimensional image data based on a first segmentation model, skeletonizing the first segmentation result, and obtaining a first vessel skeleton set, wherein the first vessel skeleton set comprises at least one type-determined vessel skeleton; obtaining a second segmentation result of the three-dimensional image data based on a second segmentation model, wherein at least one vessel in the living body in the second segmentation result is a type-undetermined vessel; fusing the first segmentation result and the second segmentation result to obtain a fusion result; skeletonizing the fusion result to obtain a vessel skeleton of the type-undetermined vessel; obtaining, as a reference vessel skeleton set, type-determined vessel skeletons in the first vessel skeleton set that have a minimum spatial distance less than a preset value D from the vessel skeleton of the type-undetermined vessel; determining a nearest point group formed by two points in the vessel skeleton of the type-undetermined vessel and reference vessel skeletons in the reference vessel skeleton set that have a minimum spatial distance; and determining a type of the type-undetermined vessel based on the nearest point group.

[0006] The one or more embodiments of the specification provide a device for identifying a vessel in a living body, comprising a processor configured to execute a method for identifying a vessel in a living body.

[0007] The one or more embodiments of the specification provide a computer-readable storage medium storing computer instructions, wherein when a computer reads the computer instructions in the storage medium, the computer executes a method for identifying a vessel in a living body.

[0008] The embodiments of the specification first train a low-accuracy but high-precision model and a high-accuracy but low-precision model, then use a post-processing algorithm to grow vessels on the results of the low-accuracy but high-precision model using the results of the high-accuracy but low-precision model, fuse the two models, and finally accurately and effectively obtain a high-accuracy and high-precision multi-class vessel segmentation result. BRIEF DESCRIPTION OF DRAWINGS

[0009] The specification will be further described in the manner of exemplary embodiments, which will be described in detail with reference to the accompanying drawings. These embodiments are not limiting, and in these embodiments, the same reference numbers represent the same structures, wherein:

[0010] Figure 1is a schematic diagram of an application scenario of an exemplary biological vessel recognition system according to some embodiments of the present specification;

[0011] Figure 2 is an exemplary flowchart of a biological vessel recognition method according to some embodiments of the present specification;

[0012] Figures 3a-3f is a schematic diagram of an exemplary biological vessel recognition method according to some embodiments of the present specification;

[0013] Figure 4 is an exemplary flowchart of a method for determining the type of a type-undetermined vessel by a vessel type determination module based on connectivity, spatial relationship, etc. according to some embodiments of the present specification;

[0014] Figure 5 is an exemplary flowchart of a method for determining the type of a type-undetermined vessel based on the relationship between the vessel skeleton of the type-undetermined vessel and a reference vessel skeleton set according to some embodiments of the present specification;

[0015] Figures 6a-6i is a schematic diagram of an exemplary method for determining the type of a type-undetermined vessel according to some embodiments of the present specification;

[0016] Figures 7a-7b is a schematic diagram of an exemplary method for determining the type of a type-undetermined vessel according to some other embodiments of the present specification;

[0017] Figure 8 is a schematic diagram of exemplary model training according to some embodiments of the present specification. DETAILED DESCRIPTION

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present specification, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are only some examples or embodiments of the present specification, and for those skilled in the art, the present specification can be applied to other similar scenarios without creative labor. Unless it is obvious from the language environment or otherwise stated, the same reference numbers in the drawings represent the same structures or operations.

[0019] It should be understood that the "system", "device", "unit" and / or "module" used herein is a method for distinguishing different components, elements, parts, portions or assemblies at different levels. However, if other words can achieve the same purpose, the words can be replaced by other expressions.

[0020] As shown in the specification and claims, the words "comprise", "comprising", "include", "including" and / or "contain", "containing", etc. do not exclude the presence of other elements or steps than those listed in the specification and claims. As used herein, these words are used in their "open" sense, that is, they are not intended to be construed as limiting the present application to the elements or steps present in any other application.

[0021] Flow diagrams are used in the specification to illustrate the operation of systems in accordance with embodiments of the specification. It will be understood that the acts shown in the figures can not necessarily be performed in the order shown. Rather, various acts can be performed in different orders or concurrently with each other. Additionally, various acts can be performed in different ways or omitted altogether, such that the ordering presented in the figures is an example only.

[0022] The present application is a method and system that can be used to determine the type of a vessel in an animal. For the sake of convenience, the embodiments of the present application will be described primarily in the context of determining the type of a blood vessel in a human. However, one of ordinary skill in the art will readily appreciate that the present application can be applied to other similar contexts without the need for undue experimentation, such as determining the type of a vessel other than a blood vessel in a human, or determining the type of a vessel in an animal other than a human, such as a dog or a cat.

[0023] Animals generally have various vessels, such as blood vessels, airways, bile ducts, or ureters. A living organism often has multiple vessels. The same type of vessel can be further divided into multiple types based on structure and function. For example, blood vessels include at least two main types, arteries and veins. In some embodiments, the types of vessels in a living organism can include sub-types of vessels, such as pulmonary veins, pulmonary arteries, hepatic veins, hepatic portal veins, hepatic arteries, etc.

[0024] Figure 1 is a schematic diagram of an example application scenario of a system for identifying vessels in a living organism according to some embodiments of the present application. As shown in Figure 1 , the system for identifying vessels in a living organism 100 can include at least a medical device 110, a processing device 120, a terminal device 130, a storage device 140, and a network 150.

[0025] The medical device 110 can scan a target object in a detection region or a scanning region to obtain scanning data of the target object. In some embodiments, the target object can include a living object and / or a non-living object. For example, the target object can be a living or non-living organic and / or inorganic substance.

[0026] In some embodiments, the medical device 110 can include a single modality scanner and / or a multi-modality scanner. The single modality scanner can include, for example, an ultrasound scanner, an X-ray scanner, a computed tomography (CT) scanner, a magnetic resonance imaging (MRI) scanner, an ultrasonograph, a positron emission computed tomography (PET) scanner, an optical coherence tomography (OCT) scanner, an ultrasound (US) scanner, an intravascular ultrasound (IVUS) scanner, a near-infrared spectroscopy (NIRS) scanner, a far-infrared (FIR) scanner, or any combination thereof. The multi-modality scanner can include, for example, an X-ray imaging-magnetic resonance imaging (X-ray-MRI) scanner, a positron emission tomography-X-ray imaging (PET-X-ray) scanner, a single photon emission computed tomography-magnetic resonance imaging (SPECT-MRI) scanner, a positron emission tomography-computed tomography (PET-CT) scanner, a digital subtraction angiography-magnetic resonance imaging (DSA-MRI) scanner, or any combination thereof. The above descriptions of the medical devices are for illustrative purposes only and are not intended to limit the scope of the present specification.

[0027] The processing device 120 can process data and / or information acquired from the medical device 110, the terminal device 130, the storage device 140, and / or other components of the identification system 100 of the vessel in a living organism. For example, the processing device 120 can acquire image data (e.g., tomographic images, PET scan images, etc.) of a target object from the medical device 110 and perform analysis processing thereon. For another example, the processing device 120 can determine a type of the vessel in a living organism based on the image data.

[0028] In some embodiments, the processing device 120 can be a single server or a group of servers. The group of servers can be centralized or distributed. In some embodiments, the processing device 120 can be local or remote. For example, the processing device 120 can access information and / or data from the medical device 110, the terminal device 130, and / or the storage device 140 through the network 150. For another example, the processing device 120 can be directly connected to the medical device 110, the terminal device 130, and / or the storage device 140 to access information and / or data. In some embodiments, the processing device 120 can be implemented on a cloud platform. For example, the cloud platform can include a private cloud, a public cloud, a hybrid cloud, a community cloud, a distributed cloud, an inter-cloud, a multi-cloud, or any combination thereof.

[0029] In some embodiments, the processing device 120 and the medical device 110 can be integrated. In some embodiments, the processing device 120 and the medical device 110 can be directly or indirectly connected to work together to implement the methods and / or functions described in the present specification.

[0030] The terminal device 130 can communicate and / or connect with the medical device 110, the processing device 120, and / or the storage device 140. In some embodiments, the interaction with the user can be implemented through the terminal device 130. In some embodiments, the terminal device 130 (or all or part of its functions) can be integrated in the processing device 120.

[0031] The storage device 140 can store data, instructions, and / or any other information. In some embodiments, the storage device 140 can store data acquired from the medical device 110, the processing device 120, the terminal device 130, and / or (e.g., image data, etc.). In some embodiments, the storage device 140 can store data and / or instructions used by the processing device 120 to perform or use to complete the exemplary methods described in this specification.

[0032] In some embodiments, the storage device 140 can include one or more storage components, each of which can be a standalone device or a part of other devices. In some embodiments, the storage device 140 can be implemented on a cloud platform. In some embodiments, the storage device 140 can be a part of the medical device 110, the processing device 120, and / or the terminal device 130.

[0033] The network 150 can include any suitable network capable of facilitating the exchange of information and / or data. In some embodiments, at least one component of the identification system of vessels within a living organism 100 (e.g., the medical device 110, the processing device 120, the terminal device 130, the storage device 140) can exchange information and / or data with at least one other component of the identification system of vessels within a living organism 100 through the network 150. For example, the processing device 120 can acquire image data of a target object from the medical device 110 through the network 150.

[0034] It should be noted that the above description of the identification system of vessels within a living organism 100 is provided only for the purpose of illustration and is not intended to limit the scope of the present specification. Numerous modifications or changes can be made to the identification system of vessels within a living organism 100 according to the description of the present specification by those of ordinary skill in the art. For example, the identification system of vessels within a living organism 100 can implement similar or different functions on other devices. However, these changes and modifications will not depart from the scope of the present specification.

[0035] Figure 2is an exemplary flowchart of a method of identifying a vessel in a living body according to some embodiments of the present disclosure. In some embodiments, flow 200 can be performed by processing device 120 or the system 100 of identifying a vessel in a living body. For example, flow 200 can be stored in a storage device (e.g., storage device 140, storage unit of processing device 120) in the form of a program or instructions, and when the processing device 120 executes the program or instructions, flow 200 can be implemented. In some embodiments, flow 200 can utilize one or more additional operations not described below, and / or be completed without one or more of the operations discussed below. Additionally, the order of the operations as shown is not limiting. Figure 2 The order of the operations as shown is not limiting.

[0036] At step 210, three-dimensional image data of a living body is obtained.

[0037] The living body can include the entirety or a portion of a biological object involved in a scanning process. For example, a head, ear-nose, mouth, neck, chest, abdomen, liver-gallbladder-pancreas-spleen, kidney, spine, etc. of a biological object.

[0038] In some embodiments, the three-dimensional image data of the living body can be a three-dimensional image that shows internal tissue and organ structures and densities of the living body by interacting with the living body through a certain medium (e.g., X-rays, electromagnetic fields, ultrasound waves, etc.). For example, computed tomography (CT) scan image, ultrasound scan image, X-ray scan image, magnetic resonance imaging (MRI) scan image, positron emission computed tomography (PET) scan image, optical coherence tomography (OCT) scan image, near-infrared spectroscopy (NIRS) scan image, far-infrared (FIR) scan image, X-ray imaging-magnetic resonance imaging (X-ray-MRI) scan image, positron emission tomography-X-ray imaging (PET-X-ray) scan image, single photon emission computed tomography-magnetic resonance imaging (SPECT-MRI) scan image, positron emission tomography-computed tomography (PET-CT) scan image, digital subtraction angiography-magnetic resonance imaging (DSA-MRI) scan image, etc. or any combination thereof. In some embodiments, the three-dimensional image data is image data that converts a sequence of two-dimensional tomographic data obtained by a medical imaging device (e.g., medical device 110) into three-dimensional data to visually and stereoscopically show three-dimensional shapes, spatial information, etc. of the living body.

[0039] In some embodiments, the three-dimensional image data of the living body can be obtained from the medical device 110, the storage device 140, the storage unit of the processing device 120, etc. In some embodiments, the processing device 120 can obtain the three-dimensional image data of the living body by reading from a storage device, a database, calling a data interface, etc.

[0040] At step 220, a first segmentation result of the three-dimensional image data is obtained based on the first segmentation model, wherein the first segmentation result comprises segmented images of vessels in the specific organism, and wherein types of at least one vessel are determined, wherein the types include a first type and a second type.

[0041] The first segmentation model can segment vessels in the organism more accurately and determine types of some of the vessels. Using the first segmentation model, accurate and / or fine-grained types of vessels in the organism can be obtained, such as pulmonary veins, pulmonary arteries, hepatic veins, and hepatic portal veins. In some embodiments, the first segmentation model is a multi-class segmentation model. The first segmentation model has a relatively high classification accuracy. To ensure a high accuracy, the first segmentation model can sacrifice a certain level of richness. The first segmentation model can classify all or part of the segmented vessels. In some embodiments, the first segmentation model focuses on classifying segmented vessels within a certain level range, and can also classify part of the segmented vessels outside the certain level range. In some embodiments, the first segmentation model focuses on classifying vessels within a certain level range.

[0042] The types of vessels include at least two types, which are referred to as the first type and the second type below. They are different types of vessels that appear simultaneously in the three-dimensional image. The first type of vessels and the second type of vessels in the three-dimensional image data often have similar or similar characteristics (e.g., contours, gray values, etc.). For example, the first type and the second type can be veins and arteries, respectively. For another example, under CT images, the first type and the second type are (renal vein, ureter), (portal vein, abdominal artery), etc.

[0043] In some embodiments, the types of vessels can include multiple types. For example, the three-dimensional image of the abdominal or liver region can include types of vessels in the organism such as the hepatic portal vein, the hepatic vein, and the hepatic artery, and the type of the type-determined vessel can be any of the above types. In this case, more types can be classified, not limited to only the first type and the second type, but also the third type and even more. For example, under MR images, the first type, the second type, and the third type are (hepatic artery, hepatic vein, and hepatic portal vein), etc.

[0044] In some embodiments, the first segmentation model can be obtained by training. The first segmentation model can be a machine learning model, which can include but is not limited to one or more combinations of neural network models, support vector machine models, k-nearest neighbor models, decision tree models, etc. Among them, the neural network model can include one or more combinations of CNN, LeNet, GoogLeNeT, ImageNet, AlexNet, VGG, ResNet, etc.

[0045] In some embodiments, the first segmentation model employs a CNN model. The processing device 120 can use training techniques to ensure the accuracy of blood vessel classification within a set range within the organism. Training techniques can include increasing the network's receptive field, increasing the network depth, etc. Increasing the network's receptive field can be achieved using methods such as dilated convolution.

[0046] For the training method of the first segmentation model, please refer to this manual. Figure 8 Explanation.

[0047] In some embodiments, the input to the first segmentation model is three-dimensional image data of an organism, and the output is a first segmentation result. The first segmentation result includes segmented images of vessels (e.g., human blood vessels) within a specific organism. In some embodiments, the first segmentation result may include segmented images of the pulmonary artery and pulmonary vein, or segmented images of the hepatic artery and portal vein, etc. At least one type of vessel within the organism is determined in the first segmentation result. Different types of vessels within the organism in the first segmentation result can be distinguished by methods such as separate coloring or different grayscale values. In some embodiments, such as... Figure 3a Example: Figure 3a The pixels (or voxels) of the arteries in the image are uniformly set to a darker grayscale. Figure 3a The pixels (or voxels) of veins in the image are uniformly set to a lighter grayscale.

[0048] Step 230: Obtain a second segmentation result of the three-dimensional image data based on the second segmentation model. The second segmentation result includes segmented images of blood vessels in the organism. At least one blood vessel in the organism in the second segmentation result is a blood vessel of undetermined type, and at least one blood vessel in the organism in the second segmentation result is not included in the first segmentation result.

[0049] The second segmentation model can segment blood vessels within a living organism in a relatively rich manner. Using the second segmentation model, images including deep branches and / or small blood vessels can be obtained. In some embodiments, the second segmentation model can segment images including blood vessels of grades 1-6 or even smaller, and images including blood vessels of grades 1-6 or even smaller. In some embodiments, the second segmentation model is a single-class segmentation model. The second segmentation model has a high degree of richness. The second segmentation model can segment all or part of the segmented blood vessels. In some embodiments, the second segmentation model can segment the smallest blood vessels possible.

[0050] In some embodiments, the second segmentation model can be obtained through training.

[0051] The second segmentation model can be a machine learning model, which can include, but is not limited to, one or more combinations of a neural network model, a support vector machine model, a k-nearest neighbor model, a decision tree model, etc. The neural network model can include one or more combinations of a CNN, LeNet, GoogLeNeT, ImageNet, AlexNet, VGG, ResNet, etc.

[0052] In some embodiments, the second segmentation model adopts a CNN model, and when constructing the second segmentation model, the number of down-sampling can be reduced to avoid losing details due to excessive down-sampling, so that the second segmentation model can identify more detailed vessels.

[0053] The training method of the second segmentation model is described in the specification. Figure 8

[0054] In some embodiments, the input of the second segmentation model is three-dimensional image data of a biological body, and the output is a second segmentation result. The second segmentation result includes a segmented image of vessels, for example, the edges of the vessels in the second segmentation result are marked, and the vessels in the image are uniformly colored. For example, Figure 3b the edges of the vessels in the image are marked, and the pixels (or voxels) of the vessels in the image are filled with the same gray value.

[0055] The vessels in the segmented image output by the second segmentation model are at least partially of an uncertain type, referred to as type-undetermined vessels. For example, the vessels in the lung are temporarily unable to be determined as venous or arterial vessels, the vessels in the kidney are temporarily unable to be determined as renal veins or ureteral vessels, the vessels in the liver are temporarily unable to be determined as hepatic veins, hepatic portal veins or hepatic arteries, etc. In some embodiments, the types of all vessels in the segmented image output by the second segmentation model are uncertain, i.e., all vessels are type-undetermined vessels.

[0056] ​The second segmentation model can be used to obtain deep branches and / or small vessels. Compared with the first segmentation model, the second segmentation model has higher richness. In some embodiments, the range of the first segmentation level of the first segmentation model is smaller than the range of the second segmentation level of the second segmentation model. The second segmentation model can segment a larger range of vessels than the first segmentation model. In some embodiments, the range of the second segmentation level of the second segmentation model and the range of the first segmentation level of the first segmentation model have an intersection, but the second segmentation model can segment finer vessels than the first segmentation model. In some embodiments, the range of the first segmentation level of the first segmentation model can overlap with the range of the second segmentation level of the second segmentation model, but based on the model design and training, the second segmentation model has higher richness and / or recognition than the first segmentation model when segmenting finer vessels. For example, the first segmentation result includes vessels of levels 1-4, and the second segmentation result includes vessels of levels 1-6 or even smaller vessels. The vessels of levels 5-6 or even smaller vessels in the second segmentation result can not be included in the first segmentation result. In some embodiments, the vessels not included in the first segmentation result in the second segmentation result are type-undefined vessels.

[0057] At step 240, the first segmentation result and the second segmentation result are fused to obtain a fusion result.

[0058] In some embodiments, the processing device 120 can synthesize the information of the first segmentation result and the second segmentation result to obtain vessels with certain richness and sufficient accuracy and the category information of some of the vessels. In some embodiments, the fusion result can be a union of the first segmentation result and the second segmentation result after processing, and then the first segmentation result set is removed from the union, and the difference set is taken as the fusion result.

[0059] As described above, at least part of the vessels in the first segmentation result have a determined type, at least one vessel in the second segmentation result is a type-undefined vessel, and at least one vessel is not included in the first segmentation result. In some embodiments, the fusion result (the difference set described above) is a set of type-undefined vessels remaining after removing the vessels with a labeled type in the first segmentation result from the second segmentation result. For example, the first segmentation result labels the categories of vessels of levels 1-4, and the second segmentation result includes vessels of levels 1-6 or even smaller vessels, and the fusion result is a set of vessels of levels 5-6 or even smaller vessels whose types are not yet clear.

[0060] In some embodiments, the processing device 120 can fuse the first segmentation result and the second segmentation result based on a plurality of fusion methods to obtain the fusion result. In some embodiments, the fusion methods can include a principal component transformation fusion method, a product transformation fusion method, a wavelet transformation fusion method, a Laplace transformation fusion method, or any combination thereof.

[0061] The second segmentation result contains more vessels than the first segmentation result. The fusion of the first segmentation result and the second segmentation result is equivalent to a blood vessel growing process, which is beneficial to obtaining a high-abundance and high-accuracy multi-class vessel segmentation result.

[0062] At step 250, the type of the type-pending vessel is determined based on the fusion result.

[0063] Based on the fusion result, the vessel type determination module can determine the type of the type-pending vessel based on the connectivity relationship, the spatial relationship, etc. Details are described in the embodiments below. In particular, reference is made to Figure 4 and Figure 5 and related embodiments.

[0064] Figures 3a-3f is a schematic diagram of an example method for identifying vessels in a living body according to some embodiments of the present specification.

[0065] As shown in Figure 3a , the first segmentation result Figure 3a contains vessels whose types have been determined. Specifically, the dark vessels are arteries, and the medium gray vessels are veins. The second segmentation result Figure 3b indicates the vessels, but does not distinguish the specific types of the vessels, and a large number of small vessels are not included in the first segmentation result.

[0066] By fusing the first segmentation result Figure 3a and the second segmentation result Figure 3b , the types of more small vessels can be identified. For example, as shown in Figure 3d and the local enlargement thereof Figure 3c , the fusion result adds an artery (light-colored vessel). For another example, as shown in Figure 3f and the local enlargement thereof Figure 3e , the fusion result adds a vein (light-colored vessel).

[0067] Figure 4 is an example flowchart of a method for determining the type of a type-pending vessel by a vessel type determination module according to some embodiments of the present specification.

[0068] At step 410, the first segmentation result is skeletonized to obtain a first vessel skeleton set, wherein the first vessel skeleton set includes at least one vessel skeleton whose type has been determined.

[0069] Skeletonization is a process of simplifying a vessel image or video into a centerline of unit width (e.g., unit pixel width, unit voxel width). Skeletonization can preserve the centerline, endpoints of lines, intersections, and the like of the original image or video, thereby preserving the connectivity of the original image. Skeletonization can reduce redundant information and only keep useful information for topology analysis, shape analysis, and the like. Skeletonization can enable an object to be represented by a simpler data structure, to simplify data analysis, reduce data storage, and requirements on transmission equipment. Skeletonization of the first segmentation result can facilitate subsequent calculation and improve the efficiency of the recognition method.

[0070] The method of skeletonization can be a parallel fast thinning algorithm, a K3M algorithm, or the like.

[0071] The types of at least some of the vessels in the first segmentation result are determined. Thus, the first skeleton set is obtained by skeletonizing the first segmentation result, and the skeletons in the first skeleton set correspond to the vessels whose types have been determined, i.e., the types of at least some of the skeletons in the first skeleton set have been determined.

[0072] At step 420, the fusion result is skeletonized to obtain the vessel skeleton of the vessel whose type is to be determined (hereinafter referred to as the skeleton to be determined).

[0073] In some embodiments, the fusion result is a set of vessels whose types are to be determined. The skeleton to be determined can be obtained by skeletonizing the fusion result.

[0074] At step 430, the type of the vessel whose type is to be determined is determined based on the connectivity relationship between the vessel skeleton of the vessel whose type is to be determined and the vessel skeletons in the first skeleton set.

[0075] In some embodiments, if there is a vessel skeleton (K2) in the first skeleton set that is connected to the vessel skeleton (K1) of the vessel whose type is to be determined, the type of the vessel skeleton K1 of the vessel whose type is to be determined is the same as the type of the vessel skeleton K2. Thus, the type of the vessel corresponding to the vessel skeleton can be determined. For example, if a vein skeleton in the first skeleton set is connected to a skeleton in the skeleton to be determined, the blood vessel corresponding to the skeleton to be determined is also a vein.

[0076] At step 440, the vessel skeletons in the first skeleton set whose types have been determined and whose minimum spatial distance from the vessel skeleton of the vessel whose type is to be determined is less than a preset value D are obtained as a reference skeleton set.

[0077] In the vessel skeleton of the type-pending vessel, for a segment of the type-pending vessel skeleton, the vessel skeletons in the first vessel skeleton set with a spatial distance less than a preset value D from the segment of the type-pending vessel skeleton can be obtained, so as to obtain a reference vessel skeleton set. The vessels in the reference vessel skeleton set are the vessels most closely related to the segment of the type-pending vessel. The reference vessel skeleton set is the basis for subsequent calculation. The preset value D determines the range of the reference vessel skeleton and is an important parameter. The selection of the preset value D has an important influence on the final recognition effect.

[0078] Based on different methods of spatial distance calculation, the preset value D as a comparison parameter of the spatial distance can be different physical quantities. For example, when the actual spatial distance is taken as the basis for distance measurement, the preset value D can be a physical quantity representing a length, such as 10 mm. In some embodiments, the calculation of the spatial distance can be based on the conversion of the voxel points in the image information, and then the actual distance value can be converted into the number of voxel points in the image, and the number of voxel points can be used to represent the preset value D. For example, the actual distance value is converted into 5 voxel points, and the preset value D is 5. In some embodiments, when the projection angles of the three-dimensional images are consistent, the actual distance value can be converted into the number of pixel points, and the number of pixel points can be used to represent the preset value D. For example, the actual distance value is converted into 5 pixel points, and the preset value D is 5.

[0079] In some embodiments, the preset value D can be obtained according to experience or demand. In some embodiments, the preset value D can be customized by the user. In some embodiments, the preset value D can be different based on the different levels of the type-pending vessel.

[0080] In some embodiments, the preset value D is obtained by a machine learning method. For example, by constructing a machine learning model, for the training data of the parts of different organisms, the optimized preset value D corresponding to the parts of the organism is obtained by machine learning. In actual application, when the parts are identified, the preset value D obtained after optimization training corresponding thereto is used. The machine learning model can include but is not limited to one or more combinations of neural network model, support vector machine model, k-nearest neighbor model, decision tree model, etc.

[0081] In some embodiments, the machine learning method of the preset value D is based on the images and type judgment results of the corresponding parts of the same type of organism. For example, the images of the corresponding parts of the same type of organism can be taken as samples, and the type judgment results can be taken as labels, and the preset value D of the same type of organism can be obtained by training.

[0082] In some embodiments, the machine training can obtain preset values D related to the parameters of gender, age, region, race, etc. by training, taking at least one of the gender, age, region, race of the organism as a parameter. For example, the preset value D for a female over 50 years old is 5, and the preset value D for a female under 50 years old is 6.

[0083] At step 450, the type of the type-pending vessel is determined based on the relationship between the vessel skeleton of the type-pending vessel and the reference vessel skeleton set. For example, the type of the type-pending vessel can be determined based on the spatial relationship, topological relationship, etc. between the type-pending skeleton and the vessel skeleton in the reference vessel skeleton set (hereinafter referred to as the reference skeleton). In some embodiments, the processing device 120 can determine the type of the type-pending vessel based on the distance and angle between the vessel skeleton of the type-pending vessel and the vessel skeleton in the reference vessel skeleton set.

[0084] For embodiments of determining the type of the type-pending vessel based on the relationship between the vessel skeleton of the type-pending vessel and the reference vessel skeleton set, see Figure 5 , Figures 6a-6i , Figures 7a-7b and the description thereof.

[0085] It should be noted that the above description of the flow 400 is only for example and illustration, and does not limit the scope of the present specification. Various modifications and changes can be made to the flow 400 by those skilled in the art under the guidance of the present specification. However, these modifications and changes are still within the scope of the present specification.

[0086] Figure 5 is an exemplary flowchart of determining the type of the type-pending vessel based on the relationship between the vessel skeleton of the type-pending vessel and the reference vessel skeleton set according to some embodiments of the present specification.

[0087] As shown in Figure 5 , different determination logic is used based on the number of vessel skeletons in the reference vessel skeleton set. When the reference vessel skeleton set includes only one vessel skeleton, the processing device 120 performs step 510; when the reference vessel skeleton set includes more than one vessel skeleton, the processing device 120 performs step 520.

[0088] At step 510, when the reference vessel skeleton set includes only one vessel skeleton, the closest point set consisting of the two points with the smallest spatial distance between the vessel skeleton of the type-pending vessel (the type-pending skeleton) and the reference vessel skeleton in the reference vessel skeleton set is determined; the type of the type-pending vessel is determined based on at least the position of the closest point set.

[0089] The closest point set refers to the point set consisting of the two points with the smallest spatial distance between the vessel skeleton of the type-pending vessel (the type-pending skeleton) and the reference vessel skeleton. For example, as shown in Figure 6aAs shown, the two points (AAA, CCC) in the pending skeleton 610 with the smallest spatial distance from the reference vessel skeleton 620 are the closest point set of the two skeletons.

[0090] In some embodiments, the processing device 120 can determine the type of the type-pending vessel based on the position of the closest point set. For example, the type of the type-pending vessel can be determined based on the positional relationship between the closest point set and the end points of the skeleton.

[0091] The end point of the skeleton refers to a point having only one adjacent point on the skeleton.

[0092] In some embodiments, if there is a point MMM in the closest point set, and the closest distance between the point MMM and any end point of the skeleton where the point MMM is located is less than a preset value n1, it is considered that the type-pending vessel and the reference vessel skeleton belong to the same category of vessels.

[0093] Based on different methods of spatial distance calculation, the preset value n1 as a comparison parameter of spatial distance can be different physical quantities. For example, when the actual spatial distance is taken as the basis for distance calculation, the preset value n1 can be a physical quantity representing a specific length, such as 5 mm. In some embodiments, the calculation of spatial distance can be based on the conversion of voxel points in image information, and then the actual distance value can be converted into the number of voxel points, and the preset value n1 can be represented by the number of voxel points, for example, the actual distance value is converted into 5 voxel points, and the preset value n1 is 5. In some embodiments, when the projection angles of three-dimensional images are consistent, the actual distance value can be converted into the number of pixel points, and the preset value n1 can be represented by the number of pixel points, for example, the actual distance value is converted into 5 pixel points, and the preset value n1 is 5.

[0094] In some embodiments, the preset value n1 can be obtained according to experience or demand. In some embodiments, the preset value n1 can be customized by the user. In some embodiments, the preset value n1 can be different based on the different levels of the type-pending vessel. For example, the smaller the preset value n1, the finer the end of the vessel classification, and the larger the preset value n1, the closer to the trunk of the vessel classification. In some embodiments, the preset value n1 is related to the thickness of the type-pending vessel. For example, the thinner the vessel, the smaller the preset value n1; the thicker the vessel, the larger the preset value n1.

[0095] In some embodiments, the preset value n1 is obtained through machine learning methods. For example, by constructing a machine learning model and training data on different parts of an organism, an optimized preset value n1 corresponding to that part of the organism is obtained through machine learning. In practical applications, when identifying the part, the preset value n1 obtained after optimization training is used. The machine learning model may include, but is not limited to, one or more combinations of neural network models, support vector machine models, k-nearest neighbor models, decision tree models, etc.

[0096] In some embodiments, the machine learning method for the preset value n1 is obtained based on images of corresponding parts of the same type of organism and type judgment results. For example, images of corresponding parts of the same type of organism can be used as samples, and type judgment results can be used as labels. Through training, the preset value n1 for that type of organism can be obtained.

[0097] like Figures 6a-6b As shown, Figure 6a It is a reconstructed local 3D image. Figure 6b for Figure 6a The corresponding skeleton simulation diagram. Figure 6a The two blood vessels are on the same plane in space (the same applies to blood vessels that are not on the same plane), and the minimum spatial distance between the two blood vessels is less than the preset value D. Figure 6b The solid line represents the skeleton, and the dashed line represents the shortest distance. In the nearest point group (AAA, CCC), the skeleton containing AAA is the skeleton of the dark-colored vessel, and the skeleton containing CCC is the skeleton of the light-colored vessel. Since the distance between AAA and the endpoint of its skeleton in the nearest point group is 0 pixels, and the distance between CCC and the endpoint of its skeleton is also 0 pixels, and the distance is also within n1 pixels, the light-colored vessel and the dark-colored vessel are considered to be of the same category.

[0098] like Figures 6c-6e As shown, Figure 6c It is a reconstructed, top-down view of a local 3D image. Figure 6d for Figure 6c A skeleton simulation diagram with the same corresponding perspective. Figure 6e for Figure 6c Simulated diagram of the vascular skeleton from the corresponding side view angle. Figure 6c Two blood vessels are located on different planes in space (the same applies to blood vessels on the same plane), and the minimum spatial distance between these two blood vessels is less than a preset value D. For example... Figure 6d As shown, in the nearest point group (AAA', CCC'), the skeleton containing AAA' is a dark vascular skeleton, and the skeleton containing CCC' is a light vascular skeleton. AAA' obscures CCC', and the line connecting CCC' and AAA' is perpendicular to the plane of the paper. Figure 6eAs shown, the dotted line is the distance from AAA' to CCC'. In the nearest point set (AAA', CCC'), AAA' is within n1 pixels from the end point of the skeleton it belongs to, and CCC' is within n1 pixels from the end point of the skeleton it belongs to. If this is the case, the light-colored vessel and the dark-colored vessel are considered to be the same type of vessel.

[0099] As shown, Figures 6f-6i , Figure 6f is a local three-dimensional image of the top view angle after reconstruction, Figure 6g is Figure 6f a skeleton simulation diagram corresponding to the same view angle, Figure 6h is Figure 6f a local three-dimensional image of the side view angle, Figure 6i is Figure 6h a skeleton simulation diagram corresponding to the same view angle. Figure 6f , Figure 6h In the two vessels in the space are in different planes (also applicable to vessels in the same plane). Figure 6g In the two vessels in the space are in different planes (also applicable to vessels in the same plane). Figure 6i In the two vessels in the space are in different planes (also applicable to vessels in the same plane).

[0100] Step 520, when the reference vessel skeleton set includes more than one vessel skeleton, the type of the type-to-be-determined vessel is determined based on the spatial relationship between the vessel skeleton in the reference vessel skeleton set (hereinafter referred to as the reference skeleton) and the vessel skeleton of the type-to-be-determined vessel (hereinafter referred to as the to-be-determined skeleton).

[0101] In some embodiments, when the reference vascular skeleton set includes more than one reference skeleton, the nearest point group between the reference vascular skeletons and the vascular skeleton of type undetermined is determined. Referring to the method in step 510, it is determined whether each reference vascular skeleton is a vascular vessel of the suspected same category as the vascular skeleton of type undetermined. For example, if there is a point MMM in the nearest point group between the reference vascular skeleton and the vascular skeleton of type undetermined, and the nearest distance between MMM and any endpoint of its containing skeleton is less than a preset value n1, then the vascular vessel of type undetermined is considered to be of the suspected same category as the reference vascular skeleton. Within the reference vascular skeleton set, only reference vascular skeletons that are suspected to be of the same category as the vascular skeleton of type undetermined are retained. If only one reference vascular skeleton in the reference vascular skeleton set is suspected to be of the same category as the vascular skeleton of type undetermined, then the type of this reference vascular is the type of the vascular vessel of type undetermined. If multiple reference vascular skeletons in the reference vascular skeleton set are suspected to be of the same category as the vascular skeleton of type undetermined, and these reference vascular skeletons are all of the same type, then the type of these reference vascular skeletons is the type of the vascular vessel of type undetermined. If multiple reference vascular skeletons in the reference vascular skeleton set are suspected to be of the same category as the vascular skeleton of type undetermined, but these reference vascular skeletons are not all of the same type, then it is necessary to determine the generalized distance between the vascular skeleton of type undetermined and the vascular skeletons in these reference vascular skeleton sets; and then, based on the generalized distance, determine the type of the vascular skeleton of type undetermined.

[0102] Generalized distance refers to a physical quantity that reflects the degree of proximity between skeletons (e.g., proximity in distance, proximity in direction).

[0103] In some embodiments, the generalized distance is obtained based on the minimum spatial distance and the generalized angle. The generalized angle is a physical quantity that reflects the directional proximity between skeletons. For example, Figure 7b α and β in the equation.

[0104] In some embodiments, the generalized included angle is obtained based on the generalized angle of the nearest point group of the blood vessels.

[0105] In some embodiments, tangents are drawn to the skeleton using points in the nearest point group as tangents, and the angle between these tangents is taken as the generalized angle. For example... Figure 7b As shown, the nearest point groups are (AAA1, CCC) and (AAA2, CCC). Tangents are drawn to the undetermined skeleton 710 with CCC as the tangent point, to the reference skeleton 720 with AAA1 as the tangent point, and to the reference skeleton 730 with AAA2 as the tangent point. The angles between the tangents (e.g., α, β) are taken as generalized angles.

[0106] If a point in the nearest point group is at a bifurcation point of the skeleton, then the bifurcation point can be used as the tangent point to draw the tangent lines of each skeleton branch, find the median of each tangent line, and use the median as the tangent line of the skeleton at the bifurcation point.

[0107] In some embodiments, the generalized angle can be obtained in other ways. For example, a fitting straight line can be made for each skeleton, and the angle between each fitting straight line can be taken as the generalized angle.

[0108] Figures 7a-7b A method of obtaining the distance based on the spatial distance and the generalized angle is shown, Figure 7a is the reconstructed local three-dimensional image, Figure 7b is Figure 7a The corresponding skeleton simulation diagram. For convenience of description, Figure 7a The three vessels in the middle are in the same plane in space (also applicable to vessels in different planes), and there are two reference skeletons (reference skeleton 720, reference skeleton 730) suspected to be of the same category as the to-be-determined skeleton 710. The nearest point groups of the two reference skeletons and the to-be-determined skeleton 710 are (AAA1, CCC) and (AAA2, CCC) respectively. Let the distance weight be f1 and the angle weight be f2 (for example, f1 = 0.4 and f2 = 0.6), then the score of the reference skeleton 710 is S1 = f1 x distance(AAA1, CCC) + f2 x β, and the score of the reference skeleton 720 is S2 = f1 x distance(AAA2, CCC) + f2 x α. The one with the smallest score is of the same category as the to-be-determined skeleton 710, for example, if S1 is smaller, then the category of the to-be-determined skeleton 710 is consistent with that of the reference skeleton 720.

[0109] It should be noted that the above description of the process 500 is only for example and illustration, and does not limit the scope of the present specification. Various modifications and changes can be made to the process 500 under the guidance of the present specification. However, these modifications and changes are still within the scope of the present specification.

[0110] Figure 8 is a schematic diagram of model training according to some embodiments of the present specification.

[0111] As Figure 8 shown, in some embodiments, the initial model 810 can be trained based on a large number of training samples with labels to update the parameters of the initial model to obtain the trained model 820. The initial model 810 can include an initial first segmentation model and an initial second segmentation model, and the trained model 820 can include a first segmentation model and a second segmentation model.

[0112] In some embodiments, the initial first segmentation model can be trained based on a large number of first training samples to update the parameters of the initial first segmentation model to obtain the first segmentation model. In some embodiments, the first training samples can be input into the initial first segmentation model, and the parameters of the initial first segmentation model can be updated through training iterations.

[0113] The first training sample refers to three-dimensional image data used for training the first segmentation model. The three-dimensional image data can include historical three-dimensional image data.

[0114] In some embodiments, each first training sample can include sample three-dimensional image data as input of the training model, and a vessel type of a vessel in the three-dimensional image data as a label. The vessel type can include at least a first type and a second type, and can further include a third type or more. For example, the vessel type includes the abdominal portal vein and the abdominal artery. For another example, the vessel type includes the hepatic portal vein, the hepatic vein, and the hepatic artery. In some embodiments, the first type of vessel in the three-dimensional image data can be marked with a first gray value, the second type of vessel can be marked with a second gray value, the third type of vessel can be marked with a third gray value, and so on. It is worth noting that the above label only includes the vessel type of the vessel in the three-dimensional image data, and does not include the level of the vessel.

[0115] In some embodiments, for example, in the embodiment of the three-dimensional image data being CT image data, the sample CT image data can be processed by adjusting the window width (the range of CT values displayed on the CT image) and the window level (the center value of the CT value), so as to increase the gray difference between the structures in the image, so as to make the sample annotation result more accurate. The label can be added by manual addition or automatic addition, or can be added by other means, which is not limited in the embodiment.

[0116] In some embodiments, the processing device can obtain the plurality of first training samples and the corresponding labels by reading or calling the data interface from the database or the storage device.

[0117] In some embodiments, the first training sample only labels the type of the vessel meeting the preset condition.

[0118] The preset condition can include a preset range of contrast of the vessel in the image, a preset range of the level of the vessel, or any combination thereof. In some embodiments, the preset condition can be set according to experience or demand. For example, different types of organisms, different parts, organs, tissues, and the like can correspond to different preset conditions. In some embodiments, the preset condition can be set by a user.

[0119] In some embodiments, the preset condition is that the level of the vessel is less than a set level.

[0120] The level of the vessel refers to the relative relationship between the vessel and the main vessel, for example, the fewer branches the vessel passes through from the main vessel, the smaller the level of the vessel.

[0121] For the thoracic artery, the thoracic aorta is a 1st order vessel, the main trunk of the pulmonary artery is a 2nd order vessel, the pulmonary lobar artery is a 3rd order vessel, the pulmonary segmental artery is a 4th order vessel, the pulmonary subsegmental artery is a 5th order vessel, and the pulmonary sub-subsegmental artery is a 6th order vessel, etc.

[0122] For the hepatic portal vein, the main trunk of the hepatic portal vein is a 1st order vessel, the left / right branch of the hepatic portal vein is a 2nd order vessel, the hepatic lobar portal vein is a 3rd order vessel, the hepatic segmental portal vein is a 4th order vessel, the hepatic subsegmental portal vein is a 5th order vessel, and the hepatic sub-subsegmental portal vein is a 6th order vessel.

[0123] For the hepatic vein, the main trunk of the hepatic vein is a 1st order vessel, the left / right branch of the hepatic vein is a 2nd order vessel, the hepatic lobar vein is a 3rd order vessel, the hepatic segmental vein is a 4th order vessel, the hepatic subsegmental vein is a 5th order vessel, and the hepatic sub-subsegmental vein is a 6th order vessel.

[0124] For the hepatic artery, the main trunk of the hepatic artery is a 1st order vessel, the left / right branch of the hepatic artery is a 2nd order vessel, the hepatic lobar artery is a 3rd order vessel, and the hepatic segmental artery is a 4th order vessel.

[0125] In some embodiments, the order of the vessel can reflect the richness of the image or detection result. For example, the larger the order number, the richer the richness. For example, the detection result containing the vessel with the maximum order of 6 is richer than the detection result containing the vessel with the maximum order of 4.

[0126] The set order can be a pre-set order of the vessel, for example, 5th order. The set order can be used to guide which vessels need to be labeled (e.g., vessels less than 5th order) and which vessels do not need to be labeled (e.g., vessels greater than or equal to 5th order). The set order can be set according to needs and / or experience. In some embodiments, the set order can be set by a user.

[0127] Only labeling the vessels with an order less than the set order can help the first segmentation model focus on the segmentation classification of the main trunk vessel, and improve the accuracy of segmentation.

[0128] In some embodiments, the processing device can input the sample three-dimensional image data of the first training sample into the first segmentation model, and output the prediction result of the vessel in the sample three-dimensional image data from the first segmentation model.

[0129] In some embodiments, the processing device can construct a loss function based on the prediction result and the label of the first training sample. The loss function can reflect the difference between the prediction result and the label. The processing device can adjust the parameters of the first segmentation model based on the loss function to reduce the difference between the prediction result and the label. For example, by continuously adjusting the parameters of the first segmentation model, the loss function value is reduced or minimized.

[0130] In some embodiments, the first segmentation model can also be obtained according to other training methods, such as setting an initial learning rate (e.g., 0.1) and a learning rate decay strategy for the training process. This application does not impose any limitations on this.

[0131] In some embodiments, an initial second segmentation model can be trained based on a large number of second training samples to update the parameters of the initial second segmentation model, thereby obtaining the second segmentation model. In some embodiments, the second training samples can be input into the initial second segmentation model, and the parameters of the initial second segmentation model can be updated through training iterations.

[0132] The second training sample refers to the 3D image data used to train the second segmentation model. This 3D image data may include historical 3D image data.

[0133] In some embodiments, each second training sample may include sample 3D image data as input to the training model, with vessels in the 3D image data serving as labels, for example, outlining the vessels in the 3D image data. It is worth noting that the labels mentioned above only include vessels (e.g., blood vessels), and do not include vessel types (e.g., portal vein, hepatic vein, hepatic artery, etc.).

[0134] In some embodiments, such as those where the 3D image data is CT image data, the sample CT image data can be processed by adjusting window width and window level to enhance the contrast of small blood vessels, so as to cover as many small blood vessels as possible, allowing the second training sample to cover more levels of blood vessels. Labels can be added manually or automatically, or in other ways; this embodiment does not limit this.

[0135] As mentioned earlier, in some embodiments, the first training sample only identifies the types of vessels that meet preset conditions. In some embodiments, the second training sample identifies at least one in vivo vessel that does not meet the preset conditions. In other words, compared to the first training sample, the second training sample identifies more vessels (with deeper branches and finer details). For example, if the preset condition is that the level of in vivo vessels is less than 5, the first training sample only identifies types of vessels of levels 1-4, while the second training sample can identify vessels of levels 1-6 or even finer details.

[0136] Covering as many small blood vessels as possible, as well as those not covered in the first training sample, helps the second segmentation model learn the features of small blood vessels and improves the richness of the segmentation.

[0137] In some embodiments, the processing device may obtain multiple second training samples, including their corresponding labels, by reading from a database or storage device or by calling a data interface.

[0138] In some embodiments, the processing device can input the sample three-dimensional image data of the second training sample into the second segmentation model, and output a prediction result of the vessels in the sample three-dimensional image data from the second segmentation model.

[0139] In some embodiments, the processing device can construct a loss function based on the prediction result and the label of the second training sample. The loss function can reflect the difference between the prediction result and the label. The processing device can adjust the parameters of the second segmentation model based on the loss function to reduce the difference between the prediction result and the label. For example, by continuously adjusting the parameters of the second segmentation model, the loss function value is reduced or minimized.

[0140] In some embodiments, the second segmentation model can also be obtained according to other training methods, for example, setting a corresponding initial learning rate (for example, 0.1) and a learning rate decay strategy for the training process. The present application does not limit this.

[0141] It should be noted that the above description of the process 800 is only for example and illustration, and does not limit the scope of the present specification. Those skilled in the art can make various modifications and changes to the process 800 under the guidance of the present specification. However, these modifications and changes are still within the scope of the present specification.

[0142] In some embodiments of the present specification, (1) obtaining the preset value D by multiple methods can reduce manual operation and enable application to multiple scenarios, improving the universality; (2) determining the type of vessels in the biological body through the connection relationship, the nearest point group and the generalized distance can improve the recognition accuracy; (3) fusing the output results of the high-accuracy model and the high-abundance model, and processing the fusion results can identify two or more kinds of vessels with close gray values that are prone to misclassification, and obtain the biological vessel recognition result with both accuracy and abundance. For example, the embodiments of the present specification can identify the hepatic portal vein, hepatic vein, hepatic artery, etc. of 5-6 levels.

[0143] The above has described the basic concepts, and it is obvious that the above detailed disclosure is only for example and does not limit the present specification. Although it is not explicitly stated here, those skilled in the art can make various modifications, improvements and corrections to the present specification. Such modifications, improvements and corrections are suggested in the present specification, so such modifications, improvements and corrections are still within the spirit and scope of the exemplary embodiments of the present specification.

[0144] Furthermore, this specification uses specific terms to describe embodiments thereof. For example, "an embodiment," "one embodiment," and / or "some embodiments" refer to a particular feature, structure, or characteristic associated with at least one embodiment of this specification. Therefore, it should be emphasized and noted that references to "an embodiment," "one embodiment," or "an alternative embodiment" in different locations throughout this specification do not necessarily refer to the same embodiment. Moreover, certain features, structures, or characteristics in one or more embodiments of this specification can be appropriately combined.

[0145] Furthermore, unless expressly stated in the claims, the order of processing elements and sequences, the use of numbers and letters, or other names described in this specification are not intended to limit the order of the processes and methods described herein. Although various examples have been discussed in the foregoing disclosure of some embodiments of the invention that are currently considered useful, it should be understood that such details are for illustrative purposes only, and the appended claims are not limited to the disclosed embodiments; rather, the claims are intended to cover all modifications and equivalent combinations that conform to the spirit and scope of the embodiments described herein. For example, while the system components described above can be implemented using hardware devices, they can also be implemented solely using software solutions, such as installing the described system on existing servers or mobile devices.

[0146] Similarly, it should be noted that, in order to simplify the description disclosed herein and thus aid in the understanding of one or more embodiments of the invention, the foregoing description of embodiments in this specification may sometimes combine multiple features into a single embodiment, drawing, or description thereof. However, this method of disclosure does not imply that the subject matter of this specification requires more features than those mentioned in the claims. In fact, the embodiments contain fewer features than all the features of a single embodiment disclosed above.

[0147] In some embodiments, numbers describing the quantity of components and attributes are used. It should be understood that such numbers used in the description of embodiments are modified in some examples with the terms "approximately," "approximately," or "generally." Unless otherwise stated, "approximately," "approximately," or "generally" indicates that the numbers are allowed to vary by ±20%. Accordingly, in some embodiments, the numerical parameters used in the specification and claims are approximate values, which may be changed depending on the characteristics required by individual embodiments. In some embodiments, numerical parameters should take into account specified significant digits and employ a general method of digit reservation. Although the numerical ranges and parameters used to confirm their breadth of range in some embodiments of this specification are approximate values, in specific embodiments, such values ​​are set as precisely as feasible.

[0148] Each patent, patent application, patent publication, and other material cited in this specification is hereby incorporated by reference in its entirety herein for the teachings relevant to the sentence and / or paragraph in which the reference is presented. Document histories, to the extent not inconsistent with the pertinent U.S. patent application file history, are also incorporated by reference herein. To the extent that material incorporated by reference contradicts or contradicts any portion of this specification, including definition, the portion of the material incorporated by reference prevails. Note, however, that in the event of inconsistencies between any such material and the present specification, including definitions, the present specification, including definitions, will control.

[0149] Finally, it should be understood that the embodiments described herein are merely exemplary of the principles of the present description. Other embodiments can be devised without departing from the scope of the present description. Accordingly, the embodiments described herein are not intended to limit the scope of the present description, but rather are intended to be exemplary thereof.

Claims

1. A method of identifying a vessel in a living organism, characterized by, The method comprises: acquiring three-dimensional image data of a living body; acquiring a first segmentation result of the three-dimensional image data based on a first segmentation model, and performing skeletonization processing on the first segmentation result to acquire a first vessel skeleton set, wherein the first vessel skeleton set comprises at least one vessel skeleton of a determined type; acquiring a second segmentation result of the three-dimensional image data based on a second segmentation model, wherein at least one vessel in the second segmentation result is a vessel of an undetermined type; fusing the first segmentation result and the second segmentation result to acquire a fusion result; performing skeletonization processing on the fusion result to acquire a vessel skeleton of the vessel of the undetermined type; acquiring, as a reference vessel skeleton set, a vessel skeleton of a determined type in the first vessel skeleton set that has a minimum spatial distance less than a preset value D from the vessel skeleton of the vessel of the undetermined type; determining a nearest point set formed by two points of the vessel skeleton of the vessel of the undetermined type and a reference vessel skeleton in the reference vessel skeleton set that have the minimum spatial distance; determining the type of the vessel of the undetermined type based on the position of the nearest point set, which further comprises: determining the type of the vessel of the undetermined type based at least on the positional relationship between the point in the nearest point set and the end point of the skeleton in which the point is located.

2. The method of claim 1, wherein: the first segmentation result comprises segmented images of vessels in a specific living body, and at least one of the vessels in the living body has a determined type, wherein the type comprises a first type and a second type; the second segmentation result comprises segmented images of vessels in the living body, and at least one of the vessels in the second segmentation result is not included in the first segmentation result.

3. The method of claim 1, wherein, The method further comprises: determining the type of the vessel of the undetermined type based on the connectivity relationship between the vessel skeleton of the vessel of the undetermined type and the vessel skeletons in the first vessel skeleton set.

4. The method of claim 1, wherein, The determination of the type of the vessel of the undetermined type based at least on the positional relationship between the point in the nearest point set and the end point of the skeleton in which the point is located comprises: when the nearest distance between the point in the nearest point set and the end point of the skeleton in which the point is located is less than a preset value, determining that the vessel of the undetermined type and the reference vessel skeleton are of the same category of vessels.

5. The method of claim 1, wherein, The determination of the type of the vessel of the undetermined type based on the nearest point set further comprises: when the reference vessel skeleton set comprises more than one vessel skeleton, determining the type of the vessel of the undetermined type based at least on the spatial relationship between the vessel skeletons in the reference vessel skeleton set and the vessel skeleton of the vessel of the undetermined type.

6. The method of claim 5, wherein, The determination of the type of the vessel of the undetermined type based on the spatial relationship between the vessel skeletons in the reference vessel skeleton set and the vessel skeleton of the vessel of the undetermined type comprises: determining the generalized distance between the vessel skeleton of the vessel of the undetermined type and the vessel skeletons in the reference vessel skeleton set; determining the type of the vessel of the undetermined type based on the generalized distance.

7. The method of claim 6, wherein, The generalized distance is obtained based on the minimum spatial distance and a generalized included angle.

8. The method of claim 7, wherein, The generalized included angle is obtained based on a generalized angle of the nearest point set of the vessel.

9. The method of claim 1, wherein, The preset value D is obtained by one or more of the following: The preset value D is obtained based on at least a part of the organism corresponding to the three-dimensional image data; The preset value D is obtained by a machine learning method; The machine learning method of the preset value D is based on images and type judgment results of parts of the same type of organism.

10. A system for identifying a vessel in a living organism, comprising: The system comprises: A display device displays at least one of the first segmentation result, the second segmentation result, the fusion result, the nearest point group, and the type of the type-undetermined vessel; and A control system, comprising one or more processors and a memory, the memory comprising programming instructions adapted to cause the one or more processors to perform operations comprising: Obtaining three-dimensional image data of an organism; Obtaining a first segmentation result of the three-dimensional image data based on a first segmentation model, skeletonizing the first segmentation result, and obtaining a first vessel skeleton set, wherein the first vessel skeleton set comprises at least one type-determined vessel skeleton; Obtaining a second segmentation result of the three-dimensional image data based on a second segmentation model, at least one vessel in the second segmentation result being a type-undetermined vessel; Fusing the first segmentation result and the second segmentation result to obtain a fusion result; Skeletonizing the fusion result to obtain a vessel skeleton of the type-undetermined vessel; Obtaining a reference vessel skeleton set from the type-determined vessel skeletons in the first vessel skeleton set, whose minimum spatial distance to the vessel skeleton of the type-undetermined vessel is less than a preset value D; Determining a nearest point group formed by two points with the minimum spatial distance between the vessel skeleton of the type-undetermined vessel and the reference vessel skeleton set; Determining the type of the type-undetermined vessel based on the position of the nearest point group, further comprising: determining the type of the type-undetermined vessel based on at least the position relationship between the points in the nearest point group and the end points of the skeletons.

11. A device for identifying blood vessels in a living organism, characterized in that, The device comprises: At least one storage medium storing computer instructions; At least one processor executing the computer instructions to implement the method of any one of claims 1-9.

12. A computer-readable storage medium storing computer instructions, when the computer reads the computer instructions, the computer executes the method of any one of claims 1-9.

Images