Method, device and equipment for automatically identifying thrombus proportion

Abstract

The invention discloses a method, a device and equipment for automatically identifying a thrombus proportion. The method comprises the following steps: performing blood vessel segmentation and thrombus segmentation on a medical image to obtain a blood vessel segmentation image and a thrombus segmentation image; extracting a blood vessel center line based on the blood vessel segmentation image, and obtaining spatial position information of a point set on the blood vessel center line; clustering based on the spatial position information of each thrombus voxel in the thrombus segmentation image to form a thrombus clustering center set; and respectively determining a to-be-analyzed point set on a corresponding blood vessel center line for the thrombus clustering center. And calculating the thrombus proportion on the blood vessel section where each point included in each to-be-analyzed point set is located. Automatic segmentation of the blood vessel and the thrombus in the medical image is achieved, the core position of the thrombus is aimed at, a plurality of center points related to the position are determined from the center line of the blood vessel according to the core position, the blood vessel section is conveniently determined, and then the thrombus proportion is obtained. A large amount of time and manpower are not needed to identify, judge and mark the thrombus, and the difficulty of thrombus severity typing is effectively reduced.

Description

Technical Field

[0001] The present application relates to the field of medical image processing technology, and in particular to a method, device and equipment for automatically identifying thrombus proportion. Background Art

[0002] Thrombi are small pieces of blood that form on the surface of peeling or repaired areas of cardiovascular blood vessels. Thrombosis is generally a multifactorial process caused by the interaction and mutual influence of genetic and environmental factors. Thrombi are generally composed of insoluble fibrin, deposited platelets, accumulated white blood cells, and trapped red blood cells, and are harmful to the human body. Portal vein thrombosis (PVT), for example, refers to thrombosis of the main portal vein and / or its left and right branches, with or without thrombosis of the mesenteric and splenic veins. Acute PVT can easily lead to serious adverse consequences such as mesenteric ischemia and even intestinal necrosis; chronic PVT can lead to portal vein occlusion or cavernous degeneration of the portal vein, followed by portal hypertension. Figure 1 Schematic diagram of PVT, showing PVT occurring on the main trunk of the portal vein.

[0003] Currently, the severity of thrombosis can be classified with the help of imaging examinations. The main imaging examination methods include Doppler ultrasound, enhanced computed tomography (CT), magnetic resonance imaging (MRI), and CT angiography (CTA). Figure 2 This is a schematic diagram of PVT in CTA images. The dark area indicated by the arrow represents PVT, which manifests as a filling defect within the portal vein lumen. In CTA images, dark areas represent low-density shadows. A filling defect within the portal vein lumen is an imaging finding caused by the presence of thrombus or tumor within the portal vein system. Figure 2 The low-density shadow in the area pointed by the middle arrow is most likely a local area of decreased density formed due to obstruction of blood flow in the portal venous system.

[0004] However, despite the assistance of the above-mentioned imaging examination methods, the current clinical classification of the severity of thrombus is still generally achieved by visual inspection. Specifically, the prerequisite for the classification of the severity of thrombus is that the imaging department first accurately identifies the thrombus. In actual work, due to the complex abdominal anatomical structure, the large amount of CTA image data, and the fact that thrombus is a very small target, the recognition rate of thrombus by doctors in the general imaging department through visual inspection is only about 60%, and the missed diagnosis and misdiagnosis rate is high. Therefore, the clinical department needs to conduct a second detailed review of the image to confirm whether the imaging department's identification is incorrect, and then use computer technology to manually mark the thrombus and measure its size, and finally classify the severity of the thrombus. The above-mentioned complex and detailed analysis and labeling requirements increase the workload of clinical doctors and bring great trouble to the clinical department in making thrombolysis decisions. At present, this field urgently needs a technology that can reduce the manual burden and reduce the difficulty of thrombus severity classification. Summary of the Invention

[0005] Based on the above problems, the present application provides a method, device and equipment for automatically identifying thrombus proportion, with the aim of reducing the burden of manual identification and labeling and reducing the difficulty of thrombus classification.

[0006] The embodiments of this application disclose the following technical solutions:

[0007] In a first aspect, the present application provides a method for automatically identifying thrombus proportion, the method comprising:

[0008] Performing blood vessel segmentation and thrombus segmentation on the medical image to obtain a blood vessel segmentation image corresponding to the medical image and a thrombus segmentation image corresponding to the medical image, respectively;

[0009] extracting a blood vessel centerline based on the blood vessel segmentation image, and obtaining spatial position information of a point set on the blood vessel centerline;

[0010] Clustering the spatial position information of each thrombus voxel in the thrombus segmentation image to form a thrombus clustering center set; the thrombus clustering center set includes one or more thrombus clustering centers;

[0011] Based on the spatial positional relationship between the point set on the blood vessel centerline and the thrombus cluster center, determining a corresponding set of points to be analyzed on the blood vessel centerline for each thrombus cluster center; the set of points to be analyzed includes a plurality of points in the point set on the blood vessel centerline that are adjacent to the corresponding thrombus cluster center;

[0012] The thrombus ratio on the blood vessel cross section at each point included in each set of points to be analyzed is calculated.

[0013] In an optional implementation, after calculating the thrombus proportion on the blood vessel cross section at each point included in each set of points to be analyzed, the method further includes:

[0014] determining the maximum value of the thrombus ratio on the blood vessel cross section at each point included in each set of points to be analyzed as the maximum thrombus ratio of the medical image;

[0015] The severity of the thrombus in the medical image is classified according to the maximum thrombus ratio.

[0016] In an optional implementation, the thrombus in the medical image is a portal vein thrombosis; and classifying the severity of the thrombus in the medical image according to the maximum thrombus ratio includes:

[0017] If the maximum value of the thrombus ratio is less than 0.5, a first classification result is output; the first classification result indicates that the severity of the portal vein thrombosis is mural thrombosis;

[0018] If the maximum value of the thrombus ratio is greater than 0.5 and less than 1, a second classification result is output; the second classification result indicates that the severity of the portal vein thrombosis is partial thrombosis;

[0019] If the maximum thrombus ratio is equal to 1 and the grayscale mean of the thrombus in the medical image is less than 50HU, a third classification result is output; the third classification result indicates that the severity of the portal vein thrombosis is obstructive thrombosis;

[0020] If the maximum thrombus ratio is equal to 1 and the average thrombus grayscale value is greater than 100HU, a fourth classification result is output; the fourth classification result indicates that the severity of the portal vein thrombosis is corded thrombosis.

[0021] In an optional implementation, after extracting the blood vessel centerline based on the blood vessel segmentation image and obtaining spatial position information of a point set on the blood vessel centerline, the method further includes:

[0022] Linking the points in the point set on the center line of the blood vessel according to the blood flow direction of the blood vessel to form a linked center point set;

[0023] The step of determining a corresponding set of points to be analyzed on the blood vessel centerline for each thrombus cluster center based on the spatial positional relationship between the point set on the blood vessel centerline and the thrombus cluster centers includes:

[0024] Based on the spatial position information of each point in the linked center point set and the spatial position information of each thrombus cluster center, multiple adjacent center points are determined for each thrombus cluster center from the linked center point set and added to the set of points to be analyzed on the blood vessel centerline corresponding to the thrombus cluster center.

[0025] In an optional implementation, a target thrombus cluster center belongs to the thrombus cluster center set; based on the spatial position information of each point in the linked center point set and the spatial position information of the target thrombus cluster center, a plurality of adjacent center points are determined for the target thrombus cluster center from the linked center point set, and added to the set of points to be analyzed on the blood vessel centerline corresponding to the target thrombus cluster center, including:

[0026] Determining a center point with the shortest spatial distance to the target thrombus cluster center from the linked center points as the target center point;

[0027] The vascular region from the Mth center point before the target center point to the Nth center point after the linking center point set is determined as the vascular sub-segment whose thrombus proportion is to be analyzed, and the center points contained in the vascular sub-segment are added to the set of points to be analyzed on the vascular centerline corresponding to the target thrombus cluster center; M and N are both positive integers.

[0028] In an optional implementation, calculating the thrombus ratio on the blood vessel cross section at each point included in the set of points to be analyzed on the blood vessel centerline corresponding to the target thrombus cluster center includes:

[0029] Determine M+N+1 vascular cross sections within the vascular sub-segment; the M+N+1 vascular cross sections correspond one-to-one to the center points included in the vascular sub-segment, and each of the M+N+1 vascular cross sections passes through the corresponding center point;

[0030] For each of the blood vessel cross sections, the ratio of the number of thrombus voxels on the blood vessel cross section to the number of all voxels on the blood vessel cross section is calculated as the thrombus ratio on the blood vessel cross section.

[0031] In an optional implementation, the blood vessel cross section is determined as follows:

[0032] Determining a section normal vector based on spatial position information of a center point and adjacent center points corresponding to the blood vessel cross section;

[0033] Substituting the section normal vector and the spatial position information of the center point corresponding to the blood vessel cross section into the point normal equation to obtain an expression for the blood vessel section where the center point corresponding to the blood vessel cross section is located;

[0034] determining a maximum inscribed sphere in the blood vessel with the center point corresponding to the blood vessel cross section as the sphere center;

[0035] The intersection of the blood vessel section and the largest inscribed sphere in the blood vessel is determined as the blood vessel cross section.

[0036] A second aspect of the present application provides a device for automatically identifying thrombus proportion, the device comprising:

[0037] a segmentation module, configured to perform blood vessel segmentation and thrombus segmentation on the medical image, and obtain a blood vessel segmentation image corresponding to the medical image and a thrombus segmentation image corresponding to the medical image, respectively;

[0038] an acquisition module, configured to extract a blood vessel centerline based on the blood vessel segmentation image and obtain spatial position information of a point set on the blood vessel centerline;

[0039] a clustering module, configured to cluster the spatial position information of each thrombus voxel in the thrombus segmentation image to form a thrombus clustering center set; the thrombus clustering center set includes one or more thrombus clustering centers;

[0040] a first determining module configured to determine, for each thrombus cluster center, a corresponding set of points to be analyzed on the vascular centerline based on a spatial positional relationship between the set of points on the vascular centerline and the thrombus cluster centers; the set of points to be analyzed comprising a plurality of points in the set of points on the vascular centerline that are adjacent to the corresponding thrombus cluster center;

[0041] The calculation module is used to calculate the thrombus ratio on the blood vessel cross section where each point included in the set of points to be analyzed is located.

[0042] In an optional implementation, the device for automatically identifying the proportion of thrombus further includes:

[0043] a second determining module, configured to determine the maximum value of the thrombus ratios on the cross section of the blood vessel at each point included in the set of points to be analyzed as the maximum thrombus ratio of the medical image;

[0044] The classification module is used to classify the severity of the thrombus in the medical image according to the maximum thrombus ratio.

[0045] A third aspect of the present application provides a device for automatically identifying thrombus proportion, the device comprising: a processor and a memory communicatively connected to each other;

[0046] The memory stores a computer program;

[0047] The processor is used to run the computer program to implement the method for automatically identifying the thrombus ratio as described in any implementation of the first aspect.

[0048] Compared with the existing technology, this application has the following beneficial effects:

[0049] The present application provides a method, device and equipment for automatically identifying the proportion of thrombus. The method, device and equipment are not limited to PVT, but can also be applied to the automatic identification of the proportion of thrombus in other parts. First, the medical image is segmented into blood vessels and thrombus, and the blood vessel segmentation image and the thrombus segmentation image corresponding to the medical image are obtained respectively. Then, the blood vessel centerline is extracted based on the blood vessel segmentation image to obtain the spatial position information of the point set on the blood vessel centerline; based on the spatial position information of each thrombus voxel in the thrombus segmentation image, a thrombus clustering center set is formed; the thrombus clustering center set includes one or more thrombus clustering centers. Thereafter, based on the spatial position relationship between the point set on the blood vessel centerline and the thrombus clustering center, the corresponding point set to be analyzed on the blood vessel centerline is determined for each thrombus clustering center. Finally, the thrombus proportion on the blood vessel cross section where each point contained in each point set to be analyzed is calculated. Through the technical solution of the present application, the automatic segmentation of blood vessels and thrombi in medical images is achieved. On this basis, clustering is also used to target the core location of the thrombus and, based on this, determine multiple center points related to the location on the centerline of the vessel. In this way, the core location of the thrombus is associated with several center points on the centerline of the vessel. This facilitates the determination of the vessel cross-section and the subsequent thrombus proportion. The above solution can simply, conveniently, quickly, and intelligently complete the automatic identification of the thrombus proportion on the vessel cross-section, without investing a lot of time and manpower to identify, judge, and label thrombi, effectively reducing the difficulty of thrombus severity classification and reducing the manual burden. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0051] Figure 1 It is a PVT schematic diagram;

[0052] Figure 2 Schematic diagram of PVT manifestation in CTA images;

[0053] Figure 3 A flow chart of a method for automatically identifying thrombus proportion provided in an embodiment of the present application;

[0054] Figure 4 Schematic diagram of the cross section of segmented blood vessels;

[0055] Figure 5Schematic diagram of 3D image of segmented blood vessels;

[0056] Figure 6 This is a schematic diagram of a three-dimensional image of a segmented thrombus;

[0057] Figure 7 Schematic diagram of the relative positions of segmented blood vessels and thrombus in a three-dimensional image;

[0058] Figure 8 A schematic diagram of the blood vessel centerline and blood flow direction in a three-dimensional image;

[0059] Figure 9 is a schematic diagram of the linked center point set;

[0060] Figure 10 Schematic diagram of multiple thrombus cluster centers formed;

[0061] Figure 11 Schematic diagram of the set of points to be analyzed on the blood vessel centerline for determining the thrombus cluster center;

[0062] Figure 12 An example flow chart for calculating thrombus ratio provided in an embodiment of the present application;

[0063] Figure 13 A schematic diagram of forming a cross section of a blood vessel;

[0064] Figure 14 Schematic diagram for calculating the proportion of thrombus in multiple vascular sections;

[0065] Figure 15 A flow chart of another method for automatically identifying thrombus proportion provided in an embodiment of the present application;

[0066] Figure 16 A schematic diagram of the structure of a device for automatically identifying thrombus proportion provided in an embodiment of the present application;

[0067] Figure 17 A schematic structural diagram of another device for automatically identifying thrombus proportion provided in an embodiment of the present application. DETAILED DESCRIPTION

[0068] Currently, although imaging examinations can classify the severity of thrombi, because imaging doctors identify thrombi manually, the rate of missed or misdiagnosis is high. Therefore, clinical departments still need to re-examine the images provided by the imaging department and manually annotate and measure the thrombi using computer technology. Therefore, in order to classify the severity of thrombi, clinical doctors have to perform complex and detailed analysis and annotation work, which is a heavy labor burden and makes thrombus classification difficult.

[0069] In view of the above problems, the inventors have proposed a method, device, and equipment that can automatically identify thrombus proportions. The entire process of the entire method (such as image segmentation, vessel centerline extraction, voxel clustering, point set determination, and ratio calculation) can be automatically implemented by a computer, reducing the reliance on professional doctors. This allows for simple, convenient, rapid, and intelligent automatic identification of thrombus proportions on vascular cross-sections. This eliminates the need to invest a large amount of time and manpower in identifying, judging, and labeling thrombi, effectively reducing the difficulty of thrombus severity classification and reducing the workload of doctors.

[0070] In order to help those skilled in the art better understand the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this application.

[0071] See also Figure 3 , which is a flow chart of a method for automatically identifying thrombus proportion provided by an embodiment of the present application. Figure 3 As shown, the method for automatically identifying thrombus proportion includes:

[0072] S301 , performing blood vessel segmentation and thrombus segmentation on a medical image to obtain a blood vessel segmentation image and a thrombus segmentation image corresponding to the medical image, respectively.

[0073] Segmenting blood vessels and thrombi in medical images can be accomplished using established segmentation algorithms. Alternatively, pre-trained segmentation models can be used to segment blood vessels, thrombi, and other targets. The following provides an alternative implementation method.

[0074] In the field of medical image segmentation, the nnUNet open-source framework is a leading UNet variant. Implementations can utilize the nnUNet open-source framework and collected datasets for training multi-target segmentation of blood vessels and thrombi.

[0075] The dataset can be a public dataset, such as the 3D-IRCADb organ segmentation dataset, which contains 20 sets of CT images.

[0076] Alternatively, private datasets can be constructed by obtaining authorized CTA images from medical institutions. For example, a total of 256 sets of abdominal venous phase CTA images from a hospital were obtained from March 2015 to June 2024. Two attending physicians annotated the blood vessels and thrombi in these images, and a deputy chief physician reviewed the images. The images were pre-processed using morphological opening and closing operations to construct a private dataset. The private dataset includes data with 1mm and 5mm slice thicknesses, with a 2:1 ratio.

[0077] In practical applications, both the aforementioned public and private datasets can be used to train the nnUNet model. Deep supervision, the 3d_ffullres mode, and 5-fold cross-validation are used during training. Testing has shown that the trained nnUNet model achieves Dice coefficients of 0.849 and 0.6 for segmenting blood vessels and thrombi in medical images, respectively. This demonstrates that the model performs well in segmenting blood vessels and thrombi in medical images, meeting segmentation requirements.

[0078] Figure 4 This is a schematic diagram of the cross-section of the segmented blood vessels. The light green area in the cross-section shown in the figure is the segmented blood vessel. Figure 5 This is a schematic diagram of a 3D image of segmented blood vessels, which also uses light green to show the blood vessel structure in the image. Figure 5 It is more convenient to observe the three-dimensional distribution of blood vessels in the image, so as to have a more accurate and vivid understanding of information such as the extension direction and size of blood vessels.

[0079] Figure 6 This is a schematic diagram of a 3D image of a segmented thrombus. The brightest area in this figure represents the segmented thrombus. To facilitate the observation of the distribution of thrombi in blood vessels, please refer to Figure 7 , This figure is a schematic diagram of the relative positions of segmented blood vessels and thrombi in a three-dimensional image. Figure 7 In the figure, white thrombi are distributed regionally on the light green blood vessels. Figure 7 The colors of each component are only used to distinguish different types of segmented targets, rather than the actual physiological colors of the segmented targets.

[0080] Combine Figures 4 to 7 As described above, in the embodiment of the present application, step S301 can be used to segment both blood vessels and thrombi in the same medical image. Specifically, the same model can be used to segment these multiple targets. Alternatively, two different segmentation models can be used, one for blood vessel segmentation and the other for thrombus segmentation.

[0081] The main purpose of segmenting blood vessels and thrombi is to clearly locate the distribution of vascular structures and thrombi in the three-dimensional space of medical images. This facilitates more accurate calculation of thrombus proportions and assists in the precise classification of thrombus severity.

[0082] In the vascular segmentation image corresponding to the medical image, each voxel in the vascular region is labeled with a vessel label. Similarly, in the thrombus segmentation image corresponding to the medical image, each voxel in the thrombus region is labeled with a thrombus label. The vessel label and thrombus label respectively distinguish between blood vessels and non-blood vessels, as well as thrombi and non-thrombi in the image, facilitating the accurate execution of subsequent image processing and calculation steps.

[0083] S302 : Extracting a blood vessel centerline based on the blood vessel segmentation image, and obtaining spatial position information of a point set on the blood vessel centerline.

[0084] Based on step S301, a blood vessel segmentation image corresponding to the medical image has been obtained. Since blood vessels are columnar in shape, their centerlines can characterize the direction of the vessels and calibrate the spatial distribution of the vessels. Therefore, this application proposes extracting the blood vessel centerlines in S302.

[0085] In one possible implementation, this step can use the skeletonize3d algorithm in the skimage toolkit to extract the vessel centerlines in the segmented image. skimage, short for scikit-image, is a Python-based image processing library. The skimage toolkit can be used to perform image preprocessing, feature extraction, analysis, and enhancement. The skeletonize3d algorithm's key idea is to remove redundant information from the image while retaining the primary structure and features of the object or shape. This algorithm effectively reduces data size and simplifies image representation while preserving important topological relationships and shape features.

[0086] In another possible implementation, this step can also use the vmtkcenterlines function in the VMTK toolkit to extract vessel centerlines from segmented vessel images. VMTK, short for Vascular Modeling Toolkit, is an open-source Python-based software package. As a tool specifically designed for vascular morphological analysis and modeling, VMTK can effectively meet the accuracy requirements of vascular modeling in the medical imaging field. VMTK covers a wide range of functions, from image preprocessing and segmentation to centerline extraction, surface reconstruction, and fluid dynamics simulation. The vmtkcenterlines function used in this tool is specifically designed for extracting vessel centerlines. Its algorithm is based on principles of image processing and computational geometry: morphological operations are applied to the segmented vessels using opening and closing operations to reduce image noise and smooth vessel edges. Connected components are then used to analyze the identified vascular branches. The vessel centerline is calculated for each connected component, typically using a distance transform to determine the distance between the voxels on the line and the vessel wall. Finally, the final vessel centerline is constructed by iteratively solving for the local shortest path.

[0087] It can be understood that the vascular centerline, as a virtual line, itself contains many points. For the convenience of description, the set of these points is called the point set on the vascular centerline. Since each point in the point set is distributed on the above-mentioned vascular centerline, it is called a center point. Figure 8 It is a schematic diagram of the blood vessel centerline and blood flow direction in the three-dimensional image. Figure 8 As you can see in the figure, the dark green blood vessels are formed by a number of orange-red center points forming the blood vessel centerline. The white arrows in the figure indicate the direction of blood flow. Figure 8 The blood vessel shown in the diagram is the portal vein. The blood flow direction of the portal vein is from bottom to top, from the trunk to the branches. Figure 8 The center point of the lowest blood vessel is called the starting point. Starting from the starting point, the center points on the blood vessel centerline are linked in sequence along the direction of blood flow from bottom to top, from the trunk to the branches, and finally form the following Figure 9 The set of linked center points is shown.

[0088] Compared with the point set on the center line of the blood vessel, the linked center point set further shows the relationship between the trunk or branch to which the adjacent center points that are linked to each other belong, as well as the relationship relative to the direction of blood flow. Figure 9 The distribution of the linked center point set is shown in three-dimensional space, with three mutually perpendicular coordinate axes and Figure 8 The scales of the coordinate axes of the image space shown correspond to Figure 9 In the diagram, information such as the initial point, endpoint, and bifurcation point are further displayed.

[0089] It is understood that, based on the extracted vascular centerline, the spatial position information of each center point on the vascular centerline can be extracted. In addition, the link relationship of the linked center point set introduced above can also be established after the spatial position information of each center point is extracted.

[0090] S303 , clustering based on the spatial position information of each thrombus voxel in the thrombus segmentation image to form a thrombus cluster center set.

[0091] Based on the completion of thrombus segmentation in the medical image in step S301, the present application proposes to further cluster each thrombus voxel based on the impact of thrombus segmentation. Through clustering, the thrombus voxels that may appear dispersed are "focused", thereby facilitating the precise positioning of the thrombus and reducing the difficulty and complexity of calculating the thrombus ratio.

[0092] In specific implementations, a density-based spatial clustering algorithm (DBSCAN) can be used to cluster the known spatial locations of each thrombus voxel to obtain the cluster center of the thrombus mass. Since the DBSCAN algorithm is a relatively mature algorithm, its principles are not detailed here. Of course, in specific applications, the clustering method is not limited to the aforementioned DBSCAN algorithm; for example, K-means can also be used. The clustering method used here is not limited.

[0093] In practical applications, the thrombus corresponding to an individual in a medical image may be single or multiple, so the thrombus clustering centers formed by clustering may be one or multiple. In this application, the thrombus clustering centers formed by clustering are added to a set to facilitate data management of the thrombus clustering centers. Figure 10 Schematic diagram of multiple thrombus cluster centers formed. Figure 10 In this example, four thrombus cluster centers are formed, indicated by arrows. For ease of description, this set is referred to as the thrombus cluster center set, which includes the one or more thrombus cluster centers formed in this step. This step effectively focuses the dispersed thrombus voxels into a limited number of one or more thrombus cluster centers, simplifying the analysis and calculation process.

[0094] In the above-described steps S302 and S303, the execution order can be S302 first and then S303, or S303 first and then S302, or S302 and S303 can be executed simultaneously. Since the execution order of the two steps is not necessarily related, the execution order of the two steps is not limited here. Figure 3 Shown is just an example sequence.

[0095] S304 : Based on the spatial positional relationship between the point set on the blood vessel centerline and the thrombus clustering centers, determine a corresponding point set to be analyzed on the blood vessel centerline for each thrombus clustering center.

[0096] In an embodiment of the present application, in order to better assist in the classification of the severity of thrombus, it is proposed to use a slicing method to calculate the proportion of thrombus on the cross section of the blood vessel. Obviously, not every cross section of the blood vessel has thrombus. If the thrombus proportion is blindly calculated for each cross section of the blood vessel, a lot of computing power will be wasted and the efficiency of analysis and calculation will be affected. In this application, it is proposed to use the point set on the center line of the blood vessel and the thrombus cluster center formed by clustering in S303 to lock part of the area on the blood vessel, and then further take the blood vessel cross section of the locked part of the area, and then calculate the thrombus proportion on the blood vessel cross section in S305. In this way, computing power resources can be greatly saved, the efficiency of calculating the thrombus proportion can be improved, and the efficiency of the classification of the severity of thrombus can be improved.

[0097] An optional implementation of step S304 is introduced below. As mentioned above, after extracting the vascular centerline based on the vascular segmentation image and obtaining the spatial position information of the point set on the vascular centerline, each point in the point set on the vascular centerline is linked together according to the blood flow direction of the blood vessel in which it is located to form a linked center point set. When it is necessary to execute step S304, based on the spatial position information of each point in the linked center point set and the spatial position information of each thrombus cluster center, multiple adjacent center points can be determined for each thrombus cluster center from the linked center point set, and added to the set of points to be analyzed on the vascular centerline corresponding to the thrombus cluster center. For example, multiple center points adjacent to a thrombus cluster center are determined from the linked center point set, and added to the set of points to be analyzed on the vascular centerline corresponding to the thrombus cluster center. Figure 10 For example, in Figure 10 In this example, four thrombus cluster centers are pre-clustered. For each of these four thrombus cluster centers, a corresponding set of points to be analyzed can be constructed. The points in the set are taken from the linked center point set. Since the points in the linked center point set are also distributed along the vessel centerline, it can also be understood that the points in each set are taken from the point set on the vessel centerline.

[0098] For ease of understanding, we'll use a target thrombus cluster center as an example to describe the process of generating the set of points to be analyzed on the corresponding vascular centerline. The target thrombus cluster center can be any thrombus cluster center in the set of thrombus cluster centers; it's referred to here simply for convenience.

[0099] Based on the spatial position information of each point in the linked center point set and the spatial position information of the target thrombus cluster center, multiple adjacent center points are determined for the target thrombus cluster center from the linked center point set and added to the set of points to be analyzed on the vascular centerline corresponding to the target thrombus cluster center, including:

[0100] First, a center point with the shortest spatial distance to the target thrombus cluster center is determined from the linked center point set as the target center point. Next, the vascular region between the Mth center point before and the Nth center point after the target center point in the linked center point set is determined as the vascular subsegment for thrombus proportion analysis. The center points contained in the vascular subsegment are added to the set of points to be analyzed on the vascular centerline corresponding to the target thrombus cluster center.

[0101] In the introduction of the above implementation method, by determining the center point with the shortest spatial distance to the target thrombus cluster center as the target center point, the effect of aiming the target thrombus cluster center to the point set on the blood vessel centerline or the linked center point set is achieved by relying on spatial position information. It can be understood that the target center point is the point in the entire linked center point set that is most representative of the spatial position of the target thrombus cluster center compared to other center points. The reason is that the target center point is the center point in the entire linked center point set that is closest to the spatial position of the target thrombus cluster center and has the shortest spatial distance. Obviously, it is very valuable and meaningful to make a blood vessel section at the target center point to form a blood vessel cross section and then calculate the thrombus proportion on the blood vessel cross section.

[0102] In actual applications, although the spatial position of the target center point is closest to the target thrombus cluster center, it does not mean that the thrombus ratio calculated on the cross-section of the blood vessel where the target center point is located can accurately reflect the severity of the thrombus. This is mainly because the thrombus is not distributed and exists as an isolated point. In the vicinity of the target center point, there are very likely other intervals that can more accurately reflect the severity of the thrombus. To this end, the embodiment of the present application proposes to determine the blood vessel area where the Mth center point before the target center point to the Nth center point after the linked center point set is located as the blood vessel sub-segment of the thrombus ratio to be analyzed, and add the center points contained in the blood vessel sub-segment to the point set to be analyzed on the blood vessel centerline corresponding to the target thrombus cluster center. Here, the purpose of taking the Mth center point before the target center point to the Nth center point after the target center point is to lock the segment where the thrombus may be distributed near the target center point, and expand from a point of interest (target center point) to a segment of interest, that is, the above-mentioned blood vessel sub-segment of the thrombus ratio to be analyzed.

[0103] Both M and N are positive integers. The values of M and N can be set according to the density of the center points in the center point set after linking, the requirement for the analysis granularity of the segment of interest, or the requirement for the amount of computation. For example, if the density of the center points in the center point set after linking is high, the values of M and N can be larger; if the requirement for the analysis granularity of the segment of interest is finer, the values of M and N can also be larger; if the amount of computation needs to be reduced, the values of M and N can be smaller. In practical applications, the values of M and N can be the same or different. For example, M = 5, N = 4; or M = N = 5.

[0104] For ease of understanding, the following Figure 11 For display. Figure 11 Schematic diagram of the set of points to be analyzed on the blood vessel centerline for determining the thrombus cluster center. Four key points are marked in the figure, among which P center represents a target thrombus cluster center, P center (x,y,z) represents P center Spatial location information of P skelen Represents a central point after a link is concentrated with P center The center point with the shortest spatial distance is called the target center point in this application, P skelen (x,y,z) represents P skelen The spatial location information of Figure 11 In the example, M=N=5, Figure 11 Medium P -5 Represents the first five center points of a target center point, P -5 (x,y,z) represents P -5 Spatial location information of P +5 Represents the fifth center point after a target center point, P +5 (x,y,z) represents P +5 spatial location information.

[0105] Combine Figure 11 As mentioned above, Figure 11 The target thrombus cluster center P is shown center , the corresponding set of points to be analyzed on the centerline of the blood vessel includes the following points: P -5 、P -4 、P -3 、P -2 、P -1 、P skelen 、P +1 、P +2 , P +3 、P +4 、P +5 , where P -4 、P -3 、P-2 、P -1 Located in P -5 With P skelen between them, not shown one by one by arrows in the figure; P +1 、P +2 , P +3 、P +4 Located in P skelen With P +5 The two components are not shown one by one with arrows in the figure.

[0106] S305 , calculating the thrombus ratio on the blood vessel cross section at each point included in each set of points to be analyzed.

[0107] For ease of understanding, the calculation of the thrombus ratio on the vascular cross section at each point is introduced using the set of points to be analyzed on the vascular centerline corresponding to the target thrombus cluster center as an example. Figure 12 , which is an example flow chart of calculating the thrombus ratio provided by the embodiment of the present application. Figure 12 As shown, the steps for calculating the thrombus ratio include:

[0108] S3051. Determine M+N+1 blood vessel cross sections within the blood vessel sub-segment.

[0109] As mentioned above, the vascular area from the first M center points to the last N center points of the target center point in the linked center point set is determined as the vascular sub-segment of the thrombus proportion to be analyzed, and the center points contained therein are added to the set of points to be analyzed on the vascular centerline corresponding to the target thrombus cluster center. This means that the set of points to be analyzed determined for the target thrombus cluster center should include M+N+1 center points. In the embodiment of the present application, it is proposed that for these M+N+1 center points, a vascular cross-section needs to be determined respectively, and finally M+N+1 vascular cross-sections are obtained. The M+N+1 vascular cross-sections correspond one-to-one to the center points contained in the vascular sub-segment, and each of the M+N+1 vascular cross-sections passes through the corresponding center point.

[0110] Regarding the vascular cross-section mentioned above, the following describes an implementation method for determining the vascular cross-section. Optionally, the method for determining the vascular cross-section includes: determining a section normal vector based on the spatial position information of the center point corresponding to the vascular cross-section and adjacent center points; substituting the section normal vector and the spatial position information of the center point corresponding to the vascular cross-section into the point normal equation to obtain an expression for the vascular cross-section at which the center point corresponding to the vascular cross-section is located; determining the maximum inscribed sphere in the blood vessel with the center point corresponding to the vascular cross-section as the sphere center; and determining the intersection of the vascular cross-section and the maximum inscribed sphere in the blood vessel as the vascular cross-section.

[0111] Simply put, in the implementation method for determining the blood vessel cross section corresponding to a certain center point provided in the embodiments of this application, it is necessary to first determine a tangent plane normal vector passing through the center point. The blood vessel cross section is constructed using this tangent plane normal vector and the position of the center point. The blood vessel cross section should be perpendicular to the tangent plane normal vector. However, the range of the cross section is infinite, and this application only focuses on the thrombus situation inside the blood vessel. Therefore, a maximum inscribed sphere is constructed within the blood vessel with the center point as the sphere center. The blood vessel range near the center point is circled with the maximum inscribed sphere within the blood vessel. Finally, the intersection of the blood vessel cross section and the maximum inscribed sphere within the blood vessel is calculated as the blood vessel cross section corresponding to the center point.

[0112] It can be understood that, as mentioned above, M+N+1 vascular cross-sections need to be determined, which means that for M+N+1 center points, each center point can use the implementation method of the above example to determine its corresponding vascular section and the maximum inscribed sphere in the blood vessel, and finally determine its corresponding vascular cross-section.

[0113] The tangent plane normal vector mentioned above can be determined by the center point and the previous or next center point. As an example, we need to determine the center point P s The corresponding blood vessel cross section. s and the next center point P s+1 The spatial position information of the tangent plane is used to construct the normal vector, which is expressed as follows:

[0114]

[0115] Among them, (x s+1 ,y s+1 ,z s+1 ) is P s+1 The spatial location information, (x s ,y s ,z s ) is P s spatial location information.

[0116] Based on P s The spatial position information of the three-dimensional space point can be formed by the three-dimensional space point formula equation through the center point P s The section equation is as follows, which is used to characterize the vascular section:

[0117] A(xx s )+B(yy s )+C(zz s )=0

[0118] The parameters A, B, and C in the above equation can be determined based on the above-mentioned tangent plane normal vector, as shown below:

[0119]

[0120] Substituting the above relationship between A, B, and C into the equation for the vascular section, we obtain:

[0121] (x s+1 -x s )(xx s )+(y s+1 -y s )(yy s )+(z s+1 -z s )(zz s )=0

[0122] All points whose spatial position information (x, y, z) satisfies the above equation are considered to be at the center point P s On the cross section of the blood vessel.

[0123] In order to determine the center point P s is the maximum inscribed sphere in the blood vessel at the center of the sphere. In practical applications, the distance_transform_edt method in the Python package monai.transform can be used to obtain the blood vessel radius at this point, thereby obtaining P s The specific process is: calculate P s The spatial position (x s ,y s ,z s ) to the nearest non-vascular area, and then take R as the radius to obtain the center of the sphere at (x s ,y s ,z s ) is expressed as follows:

[0124]

[0125] The point that satisfies the above formula is considered to be located at P s It is the largest inscribed point in the blood vessel at the center of the sphere or on the surface of the sphere.

[0126] Since (x s ,y s ,z s ) is a known quantity, and R can also be calculated through a certain calculation method. Therefore, the set of points that satisfy the above expression is searched in space, and the set is P s The largest inscribed sphere in the blood vessel at the center of the sphere.

[0127] In the known P s The largest inscribed sphere in the blood vessel at the center of the sphere and P s On the premise of the vascular section, the intersection of the surface and the sphere is obtained, that is, a circular or quasi-circular area is formed, which is called the vascular cross section. Figure 13 Schematic diagram of forming a blood vessel cross section.

[0128] S3052. For each blood vessel cross section, calculate the ratio of the number of thrombus voxels on the blood vessel cross section to the number of all voxels on the blood vessel cross section as the thrombus ratio on the blood vessel cross section.

[0129] As mentioned above, in a specific implementation, since there are multiple center points in the set of points to be analyzed, multiple blood vessel cross sections are also formed. Figure 14 Schematic diagram for calculating the proportion of thrombus on multiple blood vessel sections. Figure 14 As shown, five graphs are stacked together. The red and green circular regions represent the vessel cross-sections formed by the intersection. The red region represents a thrombus. Since the number of all voxels in the vessel cross-section is quantifiable, and the number of thrombus voxels in the red region is also quantifiable, the ratio can be calculated to obtain the proportion of thrombus in the corresponding vessel cross-section. Figure 13 The middle and right sides show the proportion of thrombus in the blood vessel cross-section at two different locations, which are 49% and 35% respectively.

[0130] Through the introduction of the above embodiments, it can be seen that the present application provides a method for automatically identifying the proportion of thrombus. The method, device and equipment are not limited to PVT, but can also be applied to the automatic identification of the proportion of thrombus in other parts. In the method, the medical image is firstly subjected to vascular segmentation and thrombus segmentation, and the vascular segmentation image corresponding to the medical image and the thrombus segmentation image corresponding to the medical image are obtained respectively. Then, the vascular centerline is extracted based on the vascular segmentation image to obtain the spatial position information of the point set on the vascular centerline; based on the spatial position information of each thrombus voxel in the thrombus segmentation image, a thrombus clustering center set is formed; the thrombus clustering center set includes one or more thrombus clustering centers. Thereafter, based on the spatial position relationship between the point set on the vascular centerline and the thrombus clustering center, the corresponding set of points to be analyzed on the vascular centerline is determined for each thrombus clustering center. Finally, the thrombus proportion on the vascular cross section where each point contained in each set of points to be analyzed is calculated.

[0131] Through the technical solution of this application, the automatic segmentation of blood vessels and thrombi in medical images is realized. On this basis, the core position of the thrombus is targeted by clustering, and multiple center points related to the position are determined from the center line of the blood vessel accordingly, so that the core position of the thrombus is associated with several center points on the center line of the blood vessel. Thereby, it is convenient to determine the cross section of the blood vessel and then obtain the thrombus ratio. Through the above solution, the automatic identification of the thrombus ratio on the blood vessel cross section can be completed simply, conveniently, quickly and intelligently without investing a lot of time and manpower to identify, judge and mark the thrombus, which effectively reduces the difficulty of thrombus severity classification and reduces the manual burden.

[0132] The following will further introduce the technical implementation of thrombosis severity classification based on the embodiment introduced above.

[0133] See Figure 15 , which is a flow chart of another method for automatically identifying thrombus proportion provided by an embodiment of the present application. Figure 15 As shown, the method includes:

[0134] S301 to S305: The specific implementation of S301 to S305 has been introduced in the previous embodiment and will not be repeated here.

[0135] S306 , determining the maximum value of the thrombus ratios on the blood vessel cross section at each point included in each set of points to be analyzed as the maximum thrombus ratio of the medical image.

[0136] As mentioned earlier, a thrombus cluster center set may contain multiple thrombus cluster centers. Therefore, when multiple thrombus cluster centers are included, a separate set of points to be analyzed must be determined for each thrombus cluster center. This means that each thrombus cluster center can identify a vessel subsegment and obtain the thrombus proportions on multiple vessel sections within the vessel subsegment that pass through each center point.

[0137] To more accurately determine the severity of an individual's thrombus, it is necessary to analyze the thrombus proportions of the thrombus sections associated with these different thrombus cluster centers one by one. Specifically, this can be accomplished by determining the maximum thrombus proportion within each vascular subsegment, then re-ranking the maximum thrombus proportions within the different vascular subsegments. The maximum thrombus proportion after ranking is determined as the maximum thrombus proportion for the medical image. Alternatively, the thrombus proportions of all acquired thrombus sections can be uniformly ranked, ultimately determining the maximum thrombus proportion as the maximum thrombus proportion for the medical image.

[0138] S307. Classify the severity of the thrombus in the medical image according to the maximum thrombus ratio.

[0139] The technical solution provided in the embodiments of this application, when implemented, classifies thrombus severity based on the maximum thrombus ratio determined in the previous step S306. Of course, in specific implementations, the severity classification methods for thrombi in different locations and with different mechanisms may vary. Here, PVT is used as an example to illustrate the implementation of PVT severity classification.

[0140] In an optional implementation, the thrombus in the medical image is a portal vein thrombosis; and the severity of the thrombus in the medical image is classified according to the maximum thrombus ratio, including:

[0141] If the maximum thrombus ratio is less than 0.5, the first classification result is output; the first classification result indicates that the severity of the portal vein thrombosis is mural thrombosis;

[0142] If the maximum thrombus ratio is greater than 0.5 and less than 1, a second classification result is output; the second classification result indicates that the severity of the portal vein thrombosis is partial thrombosis;

[0143] If the maximum thrombus ratio is equal to 1 and the grayscale mean value of the thrombus in the medical image is less than 50HU, the third classification result is output; the third classification result indicates that the severity of the portal vein thrombosis is obstructive thrombosis;

[0144] If the maximum thrombus ratio is equal to 1 and the average thrombus grayscale value is greater than 100HU, the fourth classification result is output; the fourth classification result indicates that the severity of the portal vein thrombosis is a corded thrombosis.

[0145] Currently, some Chinese researchers categorize the severity of PVT into mural, partial, obstructive, and corded. Compared to the Yerdel classification, this classification is simpler, more practical, and helpful for treatment selection and prognosis assessment. Mural PVT refers to thrombus occupying less than 50% of the portal vein lumen; obstructive PVT refers to thrombus completely or nearly completely occupying the portal vein lumen; partial PVT refers to thrombus severity between mural and obstructive; and corded PVT refers to thrombus organization due to long-term obstruction of the portal vein, making the portal vein lumen indecipherable on imaging.

[0146] Through the above embodiments, it can be seen that the technical solution of the present application can utilize the automatically identified thrombus ratio to achieve the classification of the severity of thrombus in medical images in a simple, convenient, efficient and intelligent manner, thereby reducing the workload of manual identification and classification and improving the work efficiency of doctors.

[0147] Based on the method for automatically identifying the proportion of thrombus described in the aforementioned method embodiment, this application also provides a device embodiment to introduce the technical implementation of the device for automatically identifying the proportion of thrombus. Figure 16 , which is a schematic diagram of the structure of a device for automatically identifying thrombus proportion provided by an embodiment of the present application. Figure 16 As shown, in an embodiment of the present application, a device for automatically identifying thrombus proportion includes:

[0148] a segmentation module 161 for performing blood vessel segmentation and thrombus segmentation on the medical image, and obtaining a blood vessel segmentation image corresponding to the medical image and a thrombus segmentation image corresponding to the medical image, respectively;

[0149] An acquisition module 162 is configured to extract a blood vessel centerline based on the blood vessel segmentation image and obtain spatial position information of a point set on the blood vessel centerline;

[0150] A clustering module 163 is configured to cluster the thrombus voxels based on the spatial position information of the thrombus voxels in the thrombus segmentation image to form a thrombus cluster center set; the thrombus cluster center set includes one or more thrombus cluster centers;

[0151] A first determining module 164 is configured to determine, for each thrombus cluster center, a corresponding set of points to be analyzed on the vascular center line based on a spatial positional relationship between the set of points on the vascular center line and the thrombus cluster centers; the set of points to be analyzed includes a plurality of points in the set of points on the vascular center line that are adjacent to the corresponding thrombus cluster center;

[0152] The calculation module 165 is used to calculate the thrombus ratio on the blood vessel cross section where each point included in each set of points to be analyzed is located.

[0153] Combine Figure 16 The device structure shown realizes the automatic segmentation of blood vessels and thrombi in medical images through the technical solution provided by the embodiment of the device of this application. On this basis, the core position of the thrombus is targeted by clustering, and multiple center points related to the position are determined from the center line of the blood vessel accordingly, so that the core position of the thrombus is associated with several center points on the center line of the blood vessel. Thereby, it is convenient to determine the cross section of the blood vessel and then obtain the thrombus ratio. Through the above scheme, the automatic identification of the thrombus ratio on the cross section of the blood vessel can be completed simply, conveniently, quickly and intelligently without investing a lot of time and manpower to identify, judge and mark the thrombus, which effectively reduces the difficulty of thrombus severity classification and reduces the manual burden.

[0154] The thrombus ratio calculated in this application can be further used to achieve the classification of thrombus severity. Figure 17 The schematic diagram of another device for automatically identifying the proportion of thrombus is shown in FIG. Figure 17 As shown, different from Figure 16 ,exist Figure 17 The device structure shown further includes:

[0155] A second determining module 171 is configured to determine the maximum value of the thrombus ratio on the blood vessel cross section at each point included in the set of points to be analyzed as the maximum thrombus ratio of the medical image;

[0156] The classification module 172 is configured to classify the severity of the thrombus in the medical image according to the maximum thrombus ratio.

[0157] In an optional implementation, the thrombus in the medical image is a portal vein thrombosis; the classification module 172 is specifically configured to:

[0158] If the maximum value of the thrombus ratio is less than 0.5, a first classification result is output; the first classification result indicates that the severity of the portal vein thrombosis is mural thrombosis;

[0159] If the maximum value of the thrombus ratio is greater than 0.5 and less than 1, a second classification result is output; the second classification result indicates that the severity of the portal vein thrombosis is partial thrombosis;

[0160] If the maximum thrombus ratio is equal to 1 and the grayscale mean of the thrombus in the medical image is less than 50HU, a third classification result is output; the third classification result indicates that the severity of the portal vein thrombosis is obstructive thrombosis;

[0161] If the maximum thrombus ratio is equal to 1 and the average thrombus grayscale value is greater than 100HU, a fourth classification result is output; the fourth classification result indicates that the severity of the portal vein thrombosis is corded thrombosis.

[0162] In an optional implementation, the device further includes:

[0163] a center point linking module, configured to link the points in the point set on the center line of the blood vessel according to the blood flow direction of the blood vessel to form a linked center point set;

[0164] The first determining module 164 is specifically configured to:

[0165] Based on the spatial position information of each point in the linked center point set and the spatial position information of each thrombus cluster center, multiple adjacent center points are determined for each thrombus cluster center from the linked center point set and added to the set of points to be analyzed on the blood vessel centerline corresponding to the thrombus cluster center.

[0166] In an optional implementation, the target thrombus cluster center belongs to the thrombus cluster center set;

[0167] For the target thrombus cluster center, the first determining module 164 is specifically configured to:

[0168] Determining a center point with the shortest spatial distance to the target thrombus cluster center from the linked center points as the target center point;

[0169] The vascular region from the Mth center point before the target center point to the Nth center point after the linking center point set is determined as the vascular sub-segment whose thrombus proportion is to be analyzed, and the center points contained in the vascular sub-segment are added to the set of points to be analyzed on the vascular centerline corresponding to the target thrombus cluster center; M and N are both positive integers.

[0170] In an optional implementation, for the target thrombus cluster center, the calculation module 165 is specifically configured to:

[0171] Determine M+N+1 vascular cross sections within the vascular sub-segment; the M+N+1 vascular cross sections correspond one-to-one to the center points included in the vascular sub-segment, and each of the M+N+1 vascular cross sections passes through the corresponding center point;

[0172] For each of the blood vessel cross sections, the ratio of the number of thrombus voxels on the blood vessel cross section to the number of all voxels on the blood vessel cross section is calculated as the thrombus ratio on the blood vessel cross section.

[0173] In an optional implementation, the calculation module 165 is specifically configured to determine the blood vessel cross section by:

[0174] Determining a section normal vector based on spatial position information of a center point and adjacent center points corresponding to the blood vessel cross section;

[0175] Substituting the section normal vector and the spatial position information of the center point corresponding to the blood vessel cross section into the point normal equation to obtain an expression for the blood vessel section where the center point corresponding to the blood vessel cross section is located;

[0176] determining a maximum inscribed sphere in the blood vessel with the center point corresponding to the blood vessel cross section as the sphere center;

[0177] The intersection of the blood vessel section and the largest inscribed sphere in the blood vessel is determined as the blood vessel cross section.

[0178] Based on the above method and device embodiments, the present application also provides a device for automatically identifying thrombus proportions. The device mainly includes: a processor and a memory that are communicatively connected to each other;

[0179] The memory stores a computer program;

[0180] The processor is used to run the computer program to implement some or all steps of the method for automatically identifying thrombus proportion as described in any implementation manner in the method embodiment.

[0181] It should be noted that the various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device and equipment embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiments. The device and equipment embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components indicated as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. A person of ordinary skill in the art can understand and implement it without expending creative work.

[0182] The above is merely one specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A method for automatically identifying thrombus proportion, characterized in that: include: Performing blood vessel segmentation and thrombus segmentation on the medical image to obtain a blood vessel segmentation image corresponding to the medical image and a thrombus segmentation image corresponding to the medical image, respectively; extracting a blood vessel centerline based on the blood vessel segmentation image, and obtaining spatial position information of a point set on the blood vessel centerline; Clustering the spatial position information of each thrombus voxel in the thrombus segmentation image to form a thrombus clustering center set; the thrombus clustering center set includes one or more thrombus clustering centers; Based on the spatial positional relationship between the point set on the blood vessel centerline and the thrombus cluster center, determining a corresponding set of points to be analyzed on the blood vessel centerline for each thrombus cluster center; the set of points to be analyzed includes a plurality of points in the point set on the blood vessel centerline that are adjacent to the corresponding thrombus cluster center; The thrombus ratio on the blood vessel cross section at each point included in each set of points to be analyzed is calculated.

2. The method according to claim 1, characterized in that After calculating the thrombus proportion on the blood vessel cross section at each point included in each set of points to be analyzed, the method further includes: determining the maximum value of the thrombus ratio on the blood vessel cross section at each point included in each set of points to be analyzed as the maximum thrombus ratio of the medical image; The severity of the thrombus in the medical image is classified according to the maximum thrombus ratio.

3. The method according to claim 2, characterized in that The thrombus in the medical image is a portal vein thrombosis; and the classification of the severity of the thrombus in the medical image according to the maximum thrombus ratio includes: If the maximum value of the thrombus ratio is less than 0.5, a first classification result is output; the first classification result indicates that the severity of the portal vein thrombosis is mural thrombosis; If the maximum value of the thrombus ratio is greater than 0.5 and less than 1, a second classification result is output; the second classification result indicates that the severity of the portal vein thrombosis is partial thrombosis; If the maximum thrombus ratio is equal to 1 and the grayscale mean of the thrombus in the medical image is less than 50HU, a third classification result is output; the third classification result indicates that the severity of the portal vein thrombosis is obstructive thrombosis; If the maximum thrombus ratio is equal to 1 and the average thrombus grayscale value is greater than 100HU, a fourth classification result is output; the fourth classification result indicates that the severity of the portal vein thrombosis is corded thrombosis.

4. The method according to any one of claims 1 to 3, characterized in that After extracting the blood vessel centerline based on the blood vessel segmentation image and obtaining spatial position information of a point set on the blood vessel centerline, the method further includes: Linking the points in the point set on the center line of the blood vessel according to the blood flow direction of the blood vessel to form a linked center point set; The step of determining a corresponding set of points to be analyzed on the blood vessel centerline for each thrombus cluster center based on the spatial positional relationship between the point set on the blood vessel centerline and the thrombus cluster centers includes: Based on the spatial position information of each point in the linked center point set and the spatial position information of each thrombus cluster center, multiple adjacent center points are determined for each thrombus cluster center from the linked center point set and added to the set of points to be analyzed on the blood vessel centerline corresponding to the thrombus cluster center.

5. The method according to claim 4, characterized in that The target thrombus cluster center belongs to the thrombus cluster center set; Based on the spatial position information of each point in the linked central point set and the spatial position information of the target thrombus cluster center, a plurality of adjacent central points are determined from the linked central point set for the target thrombus cluster center, and added to the set of points to be analyzed on the blood vessel centerline corresponding to the target thrombus cluster center, including: Determining a center point with the shortest spatial distance to the target thrombus cluster center from the linked center points as the target center point; The vascular region from the Mth center point before the target center point to the Nth center point after the linking center point set is determined as the vascular sub-segment whose thrombus proportion is to be analyzed, and the center points contained in the vascular sub-segment are added to the set of points to be analyzed on the vascular centerline corresponding to the target thrombus cluster center; M and N are both positive integers.

6. The method according to claim 5, characterized in that Calculating the thrombus ratio on the blood vessel cross section at each point in the set of points to be analyzed on the blood vessel centerline corresponding to the target thrombus cluster center includes: Determine M+N+1 vascular cross sections within the vascular sub-segment; the M+N+1 vascular cross sections correspond one-to-one to the center points included in the vascular sub-segment, and each of the M+N+1 vascular cross sections passes through the corresponding center point; For each of the blood vessel cross sections, the ratio of the number of thrombus voxels on the blood vessel cross section to the number of all voxels on the blood vessel cross section is calculated as the thrombus ratio on the blood vessel cross section.

7. The method according to claim 6, characterized in that The blood vessel cross section is determined as follows: Determining a section normal vector based on spatial position information of a center point and adjacent center points corresponding to the blood vessel cross section; Substituting the section normal vector and the spatial position information of the center point corresponding to the blood vessel cross section into the point normal equation to obtain an expression for the blood vessel section where the center point corresponding to the blood vessel cross section is located; determining a maximum inscribed sphere in the blood vessel with the center point corresponding to the blood vessel cross section as the sphere center; The intersection of the blood vessel section and the largest inscribed sphere in the blood vessel is determined as the blood vessel cross section.

8. A device for automatically identifying thrombus proportion, characterized in that: include: a segmentation module, configured to perform blood vessel segmentation and thrombus segmentation on the medical image, and obtain a blood vessel segmentation image corresponding to the medical image and a thrombus segmentation image corresponding to the medical image, respectively; an acquisition module, configured to extract a blood vessel centerline based on the blood vessel segmentation image and obtain spatial position information of a point set on the blood vessel centerline; a clustering module, configured to cluster the spatial position information of each thrombus voxel in the thrombus segmentation image to form a thrombus clustering center set; the thrombus clustering center set includes one or more thrombus clustering centers; a first determining module configured to determine, for each thrombus cluster center, a corresponding set of points to be analyzed on the vascular centerline based on a spatial positional relationship between the set of points on the vascular centerline and the thrombus cluster centers; the set of points to be analyzed comprising a plurality of points in the set of points on the vascular centerline that are adjacent to the corresponding thrombus cluster center; The calculation module is used to calculate the thrombus ratio on the blood vessel cross section where each point included in the set of points to be analyzed is located.

9. The device according to claim 8, characterized in that Also includes: a second determining module, configured to determine the maximum value of the thrombus ratios on the cross section of the blood vessel at each point included in the set of points to be analyzed as the maximum thrombus ratio of the medical image; The classification module is used to classify the severity of the thrombus in the medical image according to the maximum thrombus ratio.

10. A device for automatically identifying thrombus proportion, characterized in that: include: a processor and memory communicatively connected to each other; The memory stores a computer program; The processor is configured to run the computer program to implement the method for automatically identifying thrombus proportion according to any one of claims 1 to 7.

Images