A three-dimensional teaching model construction system and method of an intracranial arteriovenous fistula

Abstract

This application relates to a three-dimensional teaching model construction system and method for intracranial arteriovenous fistulas, belonging to the field of medical modeling technology. It includes: a multimodal image enhancement module, which enhances a multimodal image set by enhancing the vessel edges to obtain an enhanced multimodal image set; a point cloud skeleton generation module, which generates point clouds and extracts skeletons from the multimodal image set to obtain three-dimensional vessel point clouds and vessel skeletons; a vessel sub-point cloud segmentation module, which segments the three-dimensional vessel point cloud to obtain segmented three-dimensional vessel sub-point clouds; a vessel sub-point cloud restoration and stitching module, which detects occlusion and restores the segmented three-dimensional vessel sub-point clouds, and performs stitching and simulation to obtain a restored three-dimensional vessel model; and a fistula identification and correction module, which tracks blood flow paths and identifies fistulas to correct and obtain the target three-dimensional vessel teaching model. This invention solves the problem that traditional teaching relies on two-dimensional images or simplified models, which cannot reproduce real blood vessels and blood flow, resulting in low teaching efficiency and high clinical risks.

Description

Technical Field

[0001] This application relates to the field of medical modeling, and in particular to a three-dimensional teaching model construction system and method for intracranial arteriovenous fistulas. Background Technology

[0002] With the development of the field of neurointerventional therapy, intracranial arteriovenous fistulas, as complex lesions, require a high level of operator proficiency. High-quality preoperative training has become a key requirement for reducing clinical risks.

[0003] Currently, traditional teaching methods for intracranial arteriovenous fistulas rely on two-dimensional images or simplified models, which cannot accurately reproduce the three-dimensional structure and blood flow characteristics of real blood vessels. This makes it difficult for operators to clearly understand the instrument insertion path and also hinders effective training in microcatheter superselective manipulation. This not only leads to low teaching efficiency but also significantly increases the risks for operators in subsequent clinical practice. Summary of the Invention

[0004] This application provides a three-dimensional teaching model construction system and method for intracranial arteriovenous fistula, which improves the current situation where traditional teaching relies on two-dimensional images or simple models, which cannot reproduce real blood vessels and blood flow, resulting in low teaching efficiency and high clinical risks.

[0005] The embodiments of this application disclose the following technical solutions:

[0006] In a first aspect, embodiments of this application provide a three-dimensional teaching model construction system for intracranial arteriovenous fistulas, the system comprising:

[0007] The multimodal image enhancement module is used to acquire a multimodal image set, perform image preprocessing, and enhance blood vessel edges using the Frangi filtering algorithm to obtain an enhanced multimodal image set.

[0008] The point cloud skeleton generation module is used to generate point clouds and extract skeletons based on the enhanced multimodal image set to obtain three-dimensional blood vessel point clouds and blood vessel skeletons.

[0009] The blood vessel sub-point cloud division module is used to divide the three-dimensional blood vessel point cloud according to the connection points of the blood vessel skeleton to obtain multiple divided three-dimensional blood vessel sub-point clouds.

[0010] The blood vessel sub-point cloud restoration and stitching module is used to perform occlusion detection and restoration on the multiple divided three-dimensional blood vessel sub-point clouds, and to stitch and simulate the obtained multiple restored divided three-dimensional blood vessel sub-point clouds to obtain a restored three-dimensional blood vessel model.

[0011] The fistula identification and correction module is used to perform blood flow path tracking and fistula identification based on the restored three-dimensional vascular model, and to correct the restored three-dimensional vascular model according to the identification results to obtain the target three-dimensional vascular teaching model.

[0012] Secondly, embodiments of this application provide a method for constructing a three-dimensional teaching model of an intracranial arteriovenous fistula, the method comprising:

[0013] A multimodal image set was acquired, and image preprocessing and vascular edge enhancement were performed using the Frangi filtering algorithm to obtain an enhanced multimodal image set.

[0014] Point cloud generation and skeleton extraction are performed based on the enhanced multimodal image set to obtain three-dimensional blood vessel point cloud and blood vessel skeleton;

[0015] The three-dimensional blood vessel point cloud is divided according to the connection points of the blood vessel skeleton to obtain multiple divided three-dimensional blood vessel sub-point clouds;

[0016] Occlusion detection and restoration are performed on the multiple segmented 3D blood vessel sub-point clouds, and the obtained multiple restored segmented 3D blood vessel sub-point clouds are stitched together and simulated to obtain a restored 3D blood vessel model.

[0017] Based on the restored three-dimensional vascular model, blood flow path tracing and fistula identification are performed, and the restored three-dimensional vascular model is corrected according to the identification results to obtain the target three-dimensional vascular teaching model.

[0018] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0019] This application proposes a system and method for constructing a three-dimensional teaching model of intracranial arteriovenous fistulas. The system obtains a multimodal image set through a multimodal image enhancement module, and enhances the vessel edges using preprocessing and the Frangi filtering algorithm to obtain an enhanced multimodal image set. A point cloud skeleton generation module then generates a three-dimensional vessel point cloud based on this set and extracts the vessel skeleton. A vessel sub-point cloud segmentation module divides the three-dimensional vessel sub-point cloud according to the connection points of the vessel skeleton. A vessel sub-point cloud restoration and stitching module performs occlusion detection, restoration, and stitching simulation on the segmented sub-point clouds to obtain a restored three-dimensional vessel model. Finally, a fistula identification and correction module imports the restored three-dimensional vessel model into a fluid simulation platform for mesh generation. Virtual tracer particles are injected according to preset conditions to generate streamlines and flow traces. The fistula location is determined based on the flow field characteristics and time-series identification. The model is then corrected based on the identification results to obtain the target three-dimensional vessel teaching model. Furthermore, the feasibility of the model is verified through multiple random scene tests to ensure that the model can accurately reproduce the blood flow path of the intracranial artery-fistula-draining vein, meeting the needs of medical training and pre-operative clinical simulation.

[0020] The technical solution proposed in this application solves the problem that traditional teaching relies on two-dimensional images or simple models, which cannot reproduce real blood vessels and blood flow, resulting in low teaching efficiency and high clinical risks. It realizes the visualization and hands-on practice of intracranial arteriovenous fistula teaching, provides a reliable simulation training platform for operators to improve their surgical proficiency, and also provides reliable three-dimensional model support for the verification and testing of new instruments and teaching and scientific research demonstrations, thus promoting the intelligent development of neurointerventional therapy teaching. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 A schematic diagram of a three-dimensional teaching model construction system for intracranial arteriovenous fistulas provided in this application embodiment;

[0023] Figure 2 This is a flowchart illustrating a method for constructing a three-dimensional teaching model of an intracranial arteriovenous fistula, as provided in an embodiment of this application.

[0024] The components represented by each number in the attached diagram are explained below:

[0025] Multimodal image enhancement module 01, point cloud skeleton generation module 02, blood vessel sub-point cloud segmentation module 03, blood vessel sub-point cloud restoration and stitching module 04, fistula identification and correction module 05. Detailed Implementation

[0026] This application provides a three-dimensional teaching model construction system and method for intracranial arteriovenous fistulas, which solves the technical problems in the prior art where teaching of intracranial arteriovenous fistulas relies on two-dimensional images or simple models, which cannot restore the real three-dimensional structure and blood flow characteristics of blood vessels, resulting in unclear instrument entry paths for operators, lack of proficiency in microcatheter superselective operation, low teaching efficiency and high clinical operation risks.

[0027] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0028] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0029] In the description of this application, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed in this application.

[0030] Example 1, as shown in the appendix Figure 1 As shown, this application provides a three-dimensional teaching model construction system for intracranial arteriovenous fistulas, the system comprising the following steps:

[0031] The multimodal image enhancement module 01 is used to acquire a multimodal image set, perform image preprocessing, and enhance blood vessel edges using the Frangi filtering algorithm to obtain an enhanced multimodal image set.

[0032] In this embodiment of the application, in the scenario of constructing a three-dimensional teaching model of intracranial arteriovenous fistula, in order to accurately capture the structural details of intracranial blood vessels from medical images and solve the problem that traditional two-dimensional images or simple models cannot clearly present the relationship between the blood vessel edge and the fistula, it is necessary to first acquire multimodal images covering the complete blood vessel area and preprocess them, and then enhance the blood vessel features through filtering algorithms to ensure that the teaching model constructed subsequently can restore the real blood vessel morphology and provide reliable image data support for surgeon training.

[0033] First, we obtained a multimodal image set by consulting medical literature and reviewing clinical imaging databases to provide basic information support for subsequent model construction.

[0034] Specifically, it is necessary to collect various types of medical images that can comprehensively reflect intracranial vascular and arteriovenous fistula lesions, such as CT images containing lesions of the internal carotid artery, middle cerebral artery and arteriovenous fistula, as well as other modal images that can supplement vascular blood flow information, to ensure that the images cover the complete structure from the main vascular trunk to the small branches.

[0035] At the same time, it is also necessary to collect images of the connection area between the arteriovenous fistula and the surrounding blood vessels, such as the area extending from the main carotid artery to the draining vein of the fistula, in order to avoid incomplete model construction due to missing image information.

[0036] Furthermore, the acquired multimodal image set is preprocessed. This step mainly targets interfering factors in the original images that may affect blood vessel recognition.

[0037] For example, noise caused by fluctuations in imaging equipment parameters is removed. This type of noise can cause speckles at the edges of blood vessels, affecting subsequent feature extraction. At the same time, artifacts caused by the overlap of blood vessels with surrounding tissues such as brain tissue and cerebrospinal fluid are eliminated. For example, when the density of blood vessels and brain tissue is similar, the outline of blood vessels is easily blurred in the original image. Preprocessing can enhance the contrast between blood vessels and surrounding tissues, allowing the blood vessel area to initially appear in the image.

[0038] Furthermore, the Frangai filtering algorithm is used to enhance the vessel edges of the preprocessed image. The Frangai filtering algorithm is highly effective at recognizing tubular structures, accurately locating vascular regions in the image and enhancing their edge features.

[0039] For example, in multimodal images, the edges of small-diameter branches of the middle cerebral artery are easily confused with the surrounding tissues under traditional processing methods. However, after processing with the Frangi filtering algorithm, the edge contours of the branch vessels become significantly clearer, and the vessel wall and lumen can be clearly distinguished.

[0040] Ultimately, the obtained enhanced multimodal image set can completely and clearly present the three-dimensional structural features of intracranial blood vessels and the details of arteriovenous fistula lesions. It not only preserves the direction information of the main blood vessel, but also accurately restores the morphology of the fistula and small branches, providing a data foundation for the subsequent point cloud skeleton generation module to extract three-dimensional blood vessel point clouds and construct accurate blood vessel spatial models.

[0041] The point cloud skeleton generation module 02 is used to generate point clouds and extract skeletons based on the enhanced multimodal image set to obtain three-dimensional blood vessel point clouds and blood vessel skeletons.

[0042] In this embodiment of the application, in order to transform two-dimensional multimodal images into three-dimensional data that can reflect the spatial morphology of blood vessels and provide a structured basis for subsequent blood vessel point cloud division and model stitching, it is necessary to first perform fusion processing on the enhanced multimodal image set, and then extract the core blood vessel data and transform it into a three-dimensional morphology to ensure that the teaching model constructed subsequently has accurate spatial structural features.

[0043] Specifically, the enhanced multimodal image set is first registered and resampled. Registration ensures that images from different modalities are spatially consistent, eliminating image misalignment caused by differences in imaging angles and equipment parameters, and ensuring accurate correspondence of the same vascular region across all modalities. Resampling unifies the resolution and pixel size of each modal image to avoid affecting subsequent fusion results due to resolution inconsistencies. These two steps form a multimodal fused image, integrating the advantages of different modalities to more comprehensively present information about blood vessels and lesion areas.

[0044] Furthermore, vascular segmentation is performed based on the multimodal fused images. This process requires separating vascular and non-vascular regions from the fused images, eliminating interference from irrelevant tissues such as brain tissue and cerebrospinal fluid, identifying the vascular range including the internal carotid artery, middle cerebral artery, and arteriovenous fistula lesions, and then extracting voxel data of the vascular region that can reflect the three-dimensional structure of the blood vessels, laying the data foundation for subsequent conversion into three-dimensional point clouds.

[0045] Furthermore, point cloud generation and skeleton extraction are performed based on voxel data of the vascular region. Point cloud generation involves converting voxel data into discrete 3D points, which together constitute the 3D contour of the blood vessel. Skeleton extraction involves extracting core lines from the point cloud data that reflect the direction of the main trunk and the connection relationship between branches of the blood vessel, and smoothing them to finally obtain the 3D blood vessel point cloud and vascular skeleton.

[0046] This module transforms planar image data into three-dimensional vascular information with spatial structure by fusing, segmenting, and 3D converting enhanced multimodal images. This provides a clear basis for the subsequent vascular sub-point cloud segmentation module to segment the point cloud according to the skeleton connection points. At the same time, it ensures that the subsequently constructed 3D teaching model can accurately restore the spatial orientation of intracranial blood vessels and the positional relationship of arteriovenous fistula lesions.

[0047] In the system provided in this application embodiment, the point cloud skeleton generation module 02 includes:

[0048] The enhanced multimodal image set is registered and resampled to obtain a multimodal fused image;

[0049] Based on the multimodal fused image, blood vessel segmentation is performed, and voxel data of the blood vessel region is extracted;

[0050] Point cloud generation and skeleton extraction are performed based on the voxel data of the blood vessel region to obtain the three-dimensional blood vessel point cloud and blood vessel skeleton.

[0051] In this embodiment of the application, in order to avoid the model being unable to restore the real blood vessel direction and branching relationship due to the dispersion of image data and the inaccuracy of blood vessel information extraction, it is necessary to first perform fusion processing on the enhanced multimodal images, then extract the core blood vessel data and convert it into a three-dimensional shape, so as to ensure that the subsequent model construction can meet the needs of surgeons for clear presentation of blood vessel structure and realistic operation simulation in training.

[0052] Specifically, the enhanced multimodal image set is first registered and resampled to lay the foundation for subsequent generation of multimodal fused images and extraction of vascular data.

[0053] The registration process requires eliminating spatial misalignment caused by differences in imaging equipment and shooting angles between different modal images, ensuring that the same vascular region in each modal image can be accurately matched, and avoiding displacement or overlap of vascular structures after subsequent fusion.

[0054] In addition, resampling unifies the resolution and pixel size of images from different modalities. For example, images with different resolutions are adjusted to the same pixel density to ensure that the fused image maintains consistency in detail and that information about small branches of blood vessels is not lost due to low resolution in some areas. Through these two steps, the scattered enhanced multimodal images are integrated into a multimodal fused image that can fully present vascular information.

[0055] Furthermore, blood vessel segmentation is performed based on the obtained multimodal fusion images to extract voxel data of blood vessel regions that can reflect the three-dimensional structure of blood vessels, providing accurate blood vessel data for subsequent point cloud generation and skeleton extraction.

[0056] Specifically, the vascular segmentation process requires accurately separating vascular and non-vascular regions from the fused image, eliminating interference from surrounding tissues such as brain tissue and cerebrospinal fluid, and using segmentation algorithms in existing technologies to pinpoint the complete vascular range including the internal carotid artery, middle cerebral artery, and arteriovenous fistula lesion areas. This ensures that the segmented vascular region does not miss small branches or contain redundant non-vascular tissue, thereby extracting voxel data of the vascular region that reflects the three-dimensional structure of the blood vessel, laying the data foundation for subsequent conversion into a three-dimensional point cloud.

[0057] Furthermore, after acquiring the voxel data of the vascular region, point cloud generation and skeleton extraction are performed based on the voxel data of the vascular region to obtain a three-dimensional vascular point cloud and vascular skeleton.

[0058] The system provided in this application embodiment performs point cloud generation and skeleton extraction based on the voxel data of the vascular region to obtain the three-dimensional vascular point cloud and vascular skeleton, including:

[0059] The voxel data of the vascular region is converted into a three-dimensional point cloud, and then denoised and sparsified to obtain a three-dimensional vascular point cloud.

[0060] The voxel data of the vascular region is subjected to skeleton extraction and smoothing to obtain the vascular skeleton.

[0061] In this embodiment of the application, in order to convert the voxel data of the vascular region into three-dimensional structural data that can intuitively reflect the spatial morphology of intracranial blood vessels, it is necessary to first convert the voxel data of the vascular region into three-dimensional point clouds and optimize them, and then extract and process the vascular skeleton to provide a vascular spatial structure basis for the subsequent construction of a three-dimensional teaching model of intracranial arteriovenous fistula.

[0062] Specifically, the obtained vascular region voxel data is first converted into a three-dimensional point cloud. The conversion process requires converting each effective voxel into a discrete point in three-dimensional space according to the spatial coordinates and density information of the vascular region voxels. These discrete points are distributed according to the actual morphology of the blood vessels, and together they constitute the preliminary three-dimensional outline of the blood vessels. For example, it can show the main trunk of the internal carotid artery, the branch distribution of the middle cerebral artery, and the connection morphology between the arteriovenous fistula lesion area and the surrounding blood vessels, providing an initial three-dimensional structural basis for subsequent optimization processing.

[0063] Furthermore, after obtaining the initial 3D point cloud, it is necessary to perform denoising and sparsification processing to remove noise points caused by voxel data errors and interference.

[0064] Specifically, denoising primarily targets noise generated during the conversion process due to minor errors in voxel data and environmental interference. These noises may deviate from the actual contours of blood vessels, and retaining them could lead to false protrusions or breakpoints in the subsequent model. For example, tiny noises near arteriovenous fistulas might be misidentified as vascular branches, affecting model accuracy. Denoising algorithms (such as Gaussian filtering and median filtering algorithms in existing technologies) can filter and remove these noises to ensure that the point cloud retains only valid points from the real vascular region.

[0065] In addition, sparsification reduces the number of redundant points in the point cloud while ensuring the integrity of the blood vessel morphology. This is to avoid reducing the computational efficiency of the subsequent model due to excessive points, while not losing key information about the small branches of the blood vessel. For example, the number of points is appropriately reduced in the main body of the blood vessel, while sufficient points are retained in detailed areas such as fistulas and branch nodes. This balances the accuracy of the model with the processing efficiency, and ultimately obtains an accurate three-dimensional blood vessel point cloud.

[0066] While processing the 3D point cloud, the voxel data of the blood vessel region is subjected to skeleton extraction and smoothing to obtain the blood vessel skeleton.

[0067] Among them, skeleton extraction requires extracting the central lines that reflect the core direction of the blood vessels from the voxel data of the vascular region, just like building a supporting skeleton for the blood vessels. This process requires accurately obtaining the extension direction of the main trunk of the blood vessels, the connection nodes of the branches, and the morphology of the blood vessels in the fistula area to ensure that the skeleton can completely map the overall structure of the blood vessels. For example, it can accurately present the skeleton path from the internal carotid artery to the middle cerebral artery and then to the fistula, providing a clear structural basis for the subsequent sub-point cloud division of blood vessels.

[0068] Furthermore, after the skeleton is extracted, it needs to be smoothed. Because the initially extracted skeleton may have abrupt turns due to the discreteness of the voxel data, it does not conform to the physiological characteristics of the natural curvature of real blood vessels. For example, there may be sharp angle changes at the branches of blood vessels, while real blood vessel branches are mostly smooth transitions. This difference will affect the simulation effect of the teaching model on the feel of operation.

[0069] Therefore, the skeleton lines need to be adjusted using smoothing algorithms (such as the existing Bézier curve smoothing algorithm and spline interpolation smoothing algorithm) to eliminate abrupt turns and make the skeleton lines conform to the curvature of real blood vessels, so as to obtain a vascular skeleton with natural shape and accurate structure.

[0070] Finally, through the above steps of generating and optimizing the point cloud of voxel data of the vascular region and extracting the skeleton, the resulting three-dimensional vascular point cloud and vascular skeleton can accurately restore the spatial contour and core structure of intracranial blood vessels. This ensures both the clear presentation of vascular details and a natural morphology that conforms to physiological characteristics, providing key three-dimensional structural support for the subsequent construction of a highly realistic three-dimensional teaching model of intracranial arteriovenous fistula.

[0071] The blood vessel sub-point cloud division module 03 is used to divide the three-dimensional blood vessel point cloud according to the connection points of the blood vessel skeleton to obtain multiple divided three-dimensional blood vessel sub-point clouds.

[0072] In this embodiment of the application, in order to decompose the overall three-dimensional blood vessel point cloud into more easily processed local areas and solve the problems of low efficiency and easy omission of local details when performing occlusion detection and recovery on the overall point cloud alone, it is necessary to clearly define the boundaries based on the connection points of the blood vessel skeleton and process the three-dimensional blood vessel point cloud into regions to ensure that each sub-point cloud can accurately correspond to a specific branch or segment of the blood vessel.

[0073] First, it is necessary to clarify the core role of the vascular skeleton connection points. The vascular skeleton serves as the core line reflecting the direction of the main blood vessels and the relationship between their branches. The vascular skeleton connection points are key nodes for vascular branches, such as the junction of the internal carotid artery and the middle cerebral artery, and the connection point between the middle cerebral artery and the arteriovenous fistula. These connection points constitute the natural boundaries for dividing vascular regions, ensuring that the sub-point clouds after division correspond to independent vascular branches or lesion-related areas.

[0074] Specifically, before segmentation, all connection points on the vascular skeleton must be accurately identified. These connection points should cover the intersection of all vascular branches, including branching nodes of normal vessels and special connection points related to arteriovenous fistula lesions, such as branching nodes of the anterior cerebral artery and middle cerebral artery extending from the internal carotid artery, as well as connection points of the middle cerebral artery leading to the arteriovenous fistula opening, to ensure that each vascular segment that needs to be processed independently has a clear starting point and ending point.

[0075] Furthermore, the three-dimensional vascular point cloud is divided using the identified connection points as boundaries. The division process must follow the principle of attribution according to vascular branches, assigning each point in the three-dimensional vascular point cloud to the region defined by the corresponding connection point based on its position in the vascular spatial structure.

[0076] For example, the point cloud of the internal carotid artery segment before the connection node between the internal carotid artery and the middle cerebral artery is divided into a sub-point cloud, and the point cloud of the middle cerebral artery segment after the connection node is divided into another sub-point cloud; then, the local point cloud near the fistula is divided into a separate sub-point cloud, with the connection node between the middle cerebral artery and the arteriovenous fistula as the boundary, ensuring that each sub-point cloud contains only a single vascular branch or a local area related to the lesion.

[0077] Meanwhile, during the segmentation process, it is also necessary to ensure the integrity and independence of the sub-point clouds. Integrity is reflected in the fact that each sub-point cloud must contain the entire point cloud of the corresponding blood vessel segment, without omitting any part of that blood vessel segment; independence is reflected in the clear boundaries between different sub-point clouds, without point cloud overlap or ambiguous classification.

[0078] For example, for a three-dimensional vascular point cloud containing the internal carotid artery, the middle cerebral artery, and the arteriovenous fistula, the overall point cloud can be divided into three sub-point clouds by identifying two core connection points: the internal carotid artery-middle cerebral artery connection point and the middle cerebral artery-fistula connection point.

[0079] Specifically, the first is the point cloud of the main trunk of the internal carotid artery (from the origin of the vessel to the junction of the internal carotid artery and the middle cerebral artery), the second is the point cloud of the main trunk of the middle cerebral artery (from the junction of the internal carotid artery and the middle cerebral artery to the junction of the middle cerebral artery and the fistula), and the third is the point cloud of the fistula connection (starting from the junction of the middle cerebral artery and the fistula, covering the fistula and the surrounding small blood supply branches).

[0080] Each sub-point cloud corresponds to a specific blood vessel region, and targeted occlusion detection and restoration processing can be carried out based on the structural characteristics of different sub-point clouds.

[0081] Ultimately, the resulting multiple 3D vascular sub-point clouds not only accurately reflect the spatial morphology of each local blood vessel, but also offer the convenience of independent processing. The subsequent vascular sub-point cloud restoration and stitching module can perform occlusion detection and detail restoration for each sub-point cloud one by one, effectively improving the accuracy of local vascular structure processing.

[0082] The vessel sub-point cloud restoration and stitching module 04 is used to perform occlusion detection and restoration on the multiple divided three-dimensional vessel sub-point clouds, and to stitch and simulate the obtained multiple restored divided three-dimensional vessel sub-point clouds to obtain a restored three-dimensional vessel model.

[0083] In this embodiment of the application, in order to repair the occlusion and missing information in the divided three-dimensional blood vessel sub-point cloud caused by image acquisition angle and tissue overlap, and to integrate the independent sub-point clouds into a complete blood vessel structure, it is necessary to first carry out occlusion detection and detail restoration for each sub-point cloud, and then restore the overall blood vessel structure through stitching and simulation to ensure that the finally obtained restored three-dimensional blood vessel model can completely and realistically present the spatial morphology of intracranial blood vessels and arteriovenous fistula lesions.

[0084] During multimodal image acquisition, intracranial blood vessels may be obstructed by surrounding brain tissue, bones, or other blood vessels, resulting in missing or incomplete point cloud data in some areas. For example, small blood supply branches near arteriovenous fistulas may be obstructed by the main blood vessel, causing point cloud breaks. Directly using these data for stitching will result in incomplete fistula structures in the model, affecting the effectiveness of subsequent teaching and training. Therefore, it is necessary to perform occlusion detection on each sub-point cloud separately to accurately locate the missing areas.

[0085] Specifically, occlusion detection is performed for each sub-point cloud of a 3D blood vessel. The detection process requires combining the physiological structural characteristics of the blood vessel with the spatial distribution patterns of the sub-point cloud to identify gaps in the point cloud data that do not conform to the natural shape of the blood vessel.

[0086] For example, in the point cloud of the main trunk of the internal carotid artery, if the point cloud density of a certain segment suddenly decreases significantly, and this area should be a continuous tubular structure in physiological terms, it can be determined that there is an occlusion or missing area in that area. Similarly, in the point cloud of the fistula connection, if the point cloud at the junction of the fistula and the draining vein is interrupted, and the surrounding point cloud distribution shows a discontinuous characteristic, it can also be identified as an occluded area. This detection method, which combines structural features and data distribution, ensures that no occlusion or missing area is missed.

[0087] Furthermore, point cloud reconstruction is performed on the detected occluded areas. The reconstruction process requires supplementing the missing point cloud data based on the vascular physiological structure corresponding to the sub-point cloud and the point cloud features of the adjacent complete regions, using methods such as interpolation and fitting.

[0088] For example, for the occluded missing segments in the sub-point cloud of the main trunk of the internal carotid artery, the characteristics of the vessel diameter, curvature, etc. of the complete area before and after the missing segment can be referenced to generate a supplementary point cloud that conforms to the physiological morphology through curve fitting, so as to ensure that the restored sub-point cloud can completely present the tubular structure of the internal carotid artery.

[0089] In addition, for the missing connection between the fistula and the draining vein in the fistula-related sub-point cloud, the point cloud of the connection area can be supplemented by interpolation calculation of adjacent point clouds, combining the morphological characteristics of the fistula and the direction of the draining vein, so that the connection relationship between the fistula and the draining vein can be fully presented, and finally the restored three-dimensional vascular sub-point cloud corresponding to each sub-point cloud can be obtained.

[0090] Furthermore, after completing the occlusion detection and restoration of all sub-point clouds, the multiple restored 3D vascular sub-point clouds need to be stitched together to form a complete vascular structure.

[0091] Specifically, the stitching process requires using the previously identified vascular skeleton connection points as a benchmark to precisely connect different sub-point clouds at the corresponding connection points.

[0092] For example, the terminal connection point of the internal carotid artery trunk sub-point cloud is aligned with the starting connection point of the middle cerebral artery trunk sub-point cloud to ensure that the two blood vessels are seamlessly connected in space, and that the diameter and curvature of the blood vessels at the connection point are consistent, without any obvious structural breaks; then the terminal connection point of the middle cerebral artery trunk sub-point cloud is aligned with the starting connection point of the fistula association sub-point cloud, so that the connection relationship between the fistula and the middle cerebral artery conforms to the real physiological structure.

[0093] Furthermore, after the splicing is completed, simulation is required to verify the rationality and integrity of the overall vascular structure.

[0094] Specifically, the simulation process needs to check whether the spliced ​​blood vessels conform to the natural course and branching relationship of intracranial blood vessels. For example, whether the overall structure of the internal carotid artery, middle cerebral artery and fistula-related area is smooth, and whether the angle and diameter of each branch blood vessel are consistent with physiological characteristics.

[0095] Simultaneously, it is necessary to ensure that there are no duplicate point clouds or missing areas to avoid overlapping blood vessels or structural breaks after splicing. For example, through simulation, it is possible to visually observe whether the blood vessels in the fistula area are complete and present an abnormally connected morphology, and whether the connection of each branch blood vessel is natural. If a slight deviation is found at a certain connection point, the splicing position of the sub-point cloud needs to be adjusted in time until the overall blood vessel structure conforms to the real physiological characteristics.

[0096] Finally, through occlusion detection and restoration, precise stitching, and simulation, a restored three-dimensional vascular model was obtained. This restored three-dimensional vascular model fully presents the entire structure of intracranial vessels from the main trunk to the branches, clearly restoring the morphology of the arteriovenous fistula lesion area and its connection relationship with surrounding vessels, providing a vascular structural basis for the subsequent fistula identification and correction module to carry out blood flow path tracking and accurate fistula identification.

[0097] The fistula identification and correction module 05 is used to perform blood flow path tracking and fistula identification based on the restored three-dimensional vascular model, and to correct the restored three-dimensional vascular model according to the identification results to obtain the target three-dimensional vascular teaching model.

[0098] In this embodiment of the application, in order to solve the problems of fistula detail deviation and inaccurate blood flow path simulation that may exist in the restoration of the three-dimensional vascular model, it is necessary to use fluid simulation technology to perform blood flow analysis and fistula identification on the model, and then correct the model structure based on the identification results, so as to ensure that the final target three-dimensional vascular teaching model can truly reflect the physiological structure and hemodynamic characteristics of intracranial blood vessels and fistulas.

[0099] Specifically, the three-dimensional vascular model is first imported into the fluid simulation platform. The finite volume method is used to perform tetrahedral hybrid mesh generation on the vascular lumen. This mesh type can flexibly adapt to the complex curvature and thickness variations of the vascular lumen, ensuring mesh density in key areas such as small branches and fistulas to improve computational accuracy, while avoiding excessive meshing in the main vascular region, which would waste computational resources. The final result is a mesh generation that can support accurate blood flow simulation.

[0100] Furthermore, according to the preset simulation conditions, virtual tracer particles are injected at the blood vessel entrance corresponding to the mesh division result. The preset simulation conditions need to match the blood flow parameters under human physiological conditions. The dynamic changes of blood flow in the blood vessel are captured by the movement trajectory of the particles, generating a set of streamlines that can reflect the instantaneous blood flow direction and velocity distribution, and a set of flow traces that can record the movement path of the particles over time.

[0101] Furthermore, flow field characteristics are identified temporally based on the generated streamline and traceline sets. This process requires continuous monitoring of flow field changes at different times, and by analyzing abnormal convergence, divergence, or turbulent regions of streamlines, the potential areas where fistulas may exist are identified.

[0102] Simultaneously, by combining the abnormal turning, stagnation, or acceleration of the flow trace in a specific area, the abnormal blood flow characteristics of the suspected area are verified. By combining the analysis results of the two types of data, the specific location, size, and connection relationship with surrounding blood vessels of the fistula are determined, thus forming the fistula identification result.

[0103] Finally, the reconstructed three-dimensional vascular model was corrected by combining the streamline set, flow trace set, and fistula identification results. For the identified fistula areas, the geometry of the fistula in the model was adjusted to match the actual lesion characteristics, and the connection structure between the fistula and the feeding artery and draining vein was optimized to ensure that blood flow can flow in the model according to physiological and pathological laws.

[0104] Simultaneously, based on the abnormal blood flow areas reflected by streamlines and flow traces, the local morphology of the vascular lumen is corrected, structural defects that may lead to deviations in blood flow simulation are eliminated, and the target three-dimensional vascular teaching model is finally obtained.

[0105] This module combines fluid simulation with model correction, which not only enables accurate identification of fistulas, but also optimizes the model structure through blood flow characteristics. This allows the target three-dimensional vascular teaching model to not only restore the spatial structure of blood vessels and fistulas, but also simulate the real hemodynamic state, effectively improving the authenticity and effectiveness of teaching and training.

[0106] In the system provided in this application embodiment, the fistula identification and correction module 05 includes:

[0107] The restored three-dimensional blood vessel model is imported into a fluid simulation platform, and the blood vessel lumen is meshed using the finite volume method to obtain the meshing result, wherein the mesh type is a tetrahedral hybrid mesh.

[0108] According to the preset simulation conditions, virtual tracer particles are injected into the blood vessel entrance of the mesh division result to generate streamline set and flow path set;

[0109] Based on the streamline set and the flow path set, flow field characteristics are identified in a time sequence to obtain fistula identification results;

[0110] The restored three-dimensional vascular model is corrected by combining the streamline set, flow trace set, and fistula identification results to obtain the target three-dimensional vascular teaching model.

[0111] In this embodiment of the application, in order to accurately identify the location of intracranial arteriovenous fistula and correct the vascular model, it is necessary to use a fluid simulation platform to track blood flow paths and identify fistula characteristics, and then correct the model based on the identification results, so as to ensure that the final target three-dimensional vascular teaching model can realistically reproduce the fistula lesion and abnormal blood flow characteristics, and meet the needs of surgeons for fistula location and hemodynamic simulation in training.

[0112] Specifically, the three-dimensional blood vessel model is first imported into the fluid simulation platform, and the blood vessel lumen is divided into tetrahedral mixed meshes using the finite volume method, providing a spatial discretization basis for subsequent blood flow simulation calculations.

[0113] Among them, the finite volume method divides the computational domain of the blood vessel lumen into several non-overlapping control volumes, and solves the conservation equations such as mass and momentum on each control volume by integration to capture the transmission and change patterns of blood flow in the blood vessel.

[0114] Meanwhile, the use of tetrahedral hybrid meshes can flexibly adapt to the complex morphology of the vascular lumen. For example, in key areas such as the internal carotid artery, middle cerebral artery, and fistulas, the mesh density can be guaranteed to improve the calculation accuracy, while avoiding excessive mesh densification in the main vascular trunk area. Ultimately, mesh division results that can support accurate blood flow simulation are obtained.

[0115] Furthermore, according to the preset simulation conditions, virtual tracer particles are injected at the blood vessel inlet corresponding to the mesh generation result. These particles move with the blood flow, generating a set of streamlines that reflect the instantaneous flow direction and local velocity distribution, as well as a set of flow traces that track the dynamic trajectory of the particles over time.

[0116] The streamlines are generated by adjusting the flow rate at a specific time t. i Expanding on this, at that moment, the blood flow velocity vector field is integrated, and the instantaneous motion law of blood flow at that time point is captured through the integral relationship. Finally, static streamlines that can accurately depict the flow direction and local velocity distribution at the current moment are obtained. Each streamline can clearly present the instantaneous flow topology of blood flow at the corresponding moment, such as the flow direction of blood at the branch of the blood vessel at a certain moment, local velocity changes and other details.

[0117] Furthermore, after generating the streamlines, the set of virtual tracer particles injected at the blood vessel inlet is selected as the tracking object. By continuously recording the spatial position changes of these particles throughout the entire time series, the dynamic trajectory (i.e., the streamline) of the particles as they evolve over time is extracted. The dynamic trajectories of all particles together constitute the streamline set.

[0118] The entire tracking process needs to cover the complete path of the particle from entering the blood vessel, through the main trunk and branches of the blood vessel, until it moves to the end of the simulation, so as to ensure that the flow trace can fully reflect the motion state of the particle in different time periods, and thus allow the flow trace set to present the dynamic law of blood flow in the blood vessel over time.

[0119] For example, in a simulated intracranial arterial blood flow scenario, when at time t0, the velocity vector field near the entrance of the internal carotid artery is integrated, and the generated streamlines can clearly show the instantaneous direction of blood flow to branches such as the middle cerebral artery and the anterior cerebral artery at that moment. Moreover, the streamlines are evenly distributed, reflecting the stable blood flow state of normal vascular segments.

[0120] Meanwhile, a particle set consisting of particles p1, p2, p3, etc. injected at the inlet is selected for tracking. The flow line of each particle will completely record its own movement process from t0 to subsequent times t1 and t2. The flow lines of these particles together constitute a flow line set, which can comprehensively present the differences in the movement of particles at different locations in the normal blood vessel segment and the fistula area.

[0121] In the system provided in this application embodiment, the preset simulation conditions include preset viscosity, preset density, preset inlet flow rate, and preset outlet pressure.

[0122] For example, the preset viscosity can be set to a value close to that of human blood, such as 3.5-4.5 mPa*s; the preset density can be referenced to the density of blood, such as 1050 kg / m³. 3 The preset inlet flow velocity is based on the normal blood flow velocity of intracranial arteries, for example, the inlet flow velocity of the internal carotid artery is about 30-50 cm / s; the preset outlet pressure is based on the pressure at the end of a human vein, for example, 5-10 mmHg.

[0123] By setting the above-mentioned physiological parameters, it can be ensured that the movement of virtual tracer particles can truly reflect the blood flow status of intracranial blood vessels and arteriovenous fistula areas, so that the generated streamline set and flow trace set can accurately present the stability of normal blood flow and the abnormal blood flow disturbance in the fistula area.

[0124] Furthermore, flow field characteristics are identified temporally based on the generated streamline set and flow path set to obtain fistula identification results.

[0125] In the system provided in this application embodiment, flow field feature temporal identification is performed based on the streamline set and the flow path line set to obtain fistula identification results, including:

[0126] Based on the streamline set, the abnormal flow field region is determined, and the abnormal flow path set associated with the abnormal flow field region is extracted from the flow path set.

[0127] The set of abnormal flow lines is traversed to perform spatial feature identification, and a set of spatial feature results of abnormal flow lines is obtained.

[0128] The fistula identification result is determined by performing a union analysis on the set of spatial feature results of the abnormal flow lines.

[0129] In this embodiment of the application, in order to accurately identify the location and characteristics of intracranial arteriovenous fistulas, it is necessary to first lock the spatial range of abnormal flow fields from the streamline set, and then combine the dynamic trajectory of the flow trace set for feature analysis to ensure the accuracy of the fistula identification results, and provide a basis for subsequent correction and restoration of the three-dimensional vascular model and construction of a highly realistic teaching model.

[0130] First, the abnormal flow field region is determined based on the generated streamline set, so as to define the spatial range of subsequent flow path analysis.

[0131] Specifically, the blood flow velocity field obtained by transient CFD solution is first divided into frames according to the time step to obtain several time frames, for example, divided according to the time step of 0.001s-0.002s, to ensure that the dynamic changes of blood flow at different times can be fully captured.

[0132] Furthermore, the velocity vector field is extracted within each frame. For each time t i Integrating the velocity field generates a streamline set S i This allows for precise depiction of the instantaneous flow direction and local velocity distribution of blood at that moment.

[0133] The integration process must satisfy the formula ,in It is the spatial displacement vector of the streamlined infinitesimal segment. It is the arc length of the streamlined infinitesimal segment. Spatial location At time t i The velocity vector below allows each streamline to correspond to the instantaneous flow state of blood flow in a real blood vessel.

[0134] Furthermore, after generating the streamline set, the streamline density, direction bifurcation angle, and convergence region are analyzed. If the streamline density in a certain region increases sharply and the directions converge, it indicates the existence of fluid energy accumulation. This region is marked as the flow direction anomaly region, thereby clarifying the spatial distribution range of the abnormal flow field and defining spatial constraints for the subsequent targeted analysis of flow traces.

[0135] For example, when simulating intracranial arterial blood flow, the streamlines of normal blood vessel segments are evenly distributed and steadily directed toward branches such as the middle cerebral artery and anterior cerebral artery. However, the streamlines near the fistula will show a significant increase in density and chaotic convergence of directions. Such areas are identified as abnormal flow field regions.

[0136] Furthermore, after obtaining the abnormal flow field region, the set of abnormal flow lines associated with the abnormal flow field region is extracted from the set of flow lines to clarify the specific analysis object for subsequent spatial feature identification.

[0137] Specifically, a swarm of tracer particles is first injected into the inlet of the blood vessel, and then the particles are integrated over time based on the velocity field. The integration process follows the formula... ,in The spatial displacement vector representing the infinitesimal segment of the flow path. Representing the time element, Let r be the velocity vector at spatial position r at time t. By performing time integration, a complete set of flow path trajectories can be obtained.

[0138] Furthermore, based on the spatial range of the abnormal flow field region, flow traces that are spatially related to the abnormal flow field region are selected from the set of flow traces. For example, flow traces that pass through the abnormal flow field region of the fistula, or whose trajectories show obvious deflection, stagnation, or turbulence and other abnormal changes, are integrated to form a set of abnormal flow traces.

[0139] Furthermore, the obtained set of abnormal flow lines is traversed and spatial feature recognition is performed to obtain a set of spatial feature results for abnormal flow lines.

[0140] Specifically, for each abnormal flow trajectory, combined with public... The spatial morphology and motion patterns of each trajectory are analyzed one by one. During the analysis, three types of spatial characteristics are identified: first, the topological structure of the trajectory, such as whether there is a sudden bifurcation, multi-directional divergence, or convergence; second, the vector change of velocity, such as whether there is continuous reverse flow or sudden increase or decrease in velocity magnitude caused by local turbulent disturbance; and third, the spatial distribution pattern, such as whether multiple flow lines exhibit chaotic clustering in the same area.

[0141] For example, when an abnormal flow line passes through a suspected fistula area, The vector direction changes frequently, and the speed fluctuates with "surge-drop" in a short period of time. These characteristics need to be recorded in detail. In addition, multiple sets of flow lines show chaotic interweaving of trajectories in a certain spatial region, forming obvious flow vortices, which also need to be marked as spatial features.

[0142] At the same time, all identified features are classified and organized according to the dimensions of trajectory morphology, velocity change, and spatial distribution to form a set of spatial feature results of abnormal flow traces.

[0143] Furthermore, a union analysis is performed based on the obtained set of spatial features of abnormal flow lines to determine the fistula identification result.

[0144] Specifically, the first step is to perform space-time matching, that is, to identify the streamline anomaly region S at each moment. i Spatially align the flow path points with the flow path trajectory points to ensure that they are correlated under the same spatial coordinate system. When a flow path particle crosses the same flow path convergence area multiple times, and its velocity direction reverses at least once during the crossing (i.e., the angle between the velocity vector and the normal blood flow direction is >90°), the area is determined to be a dynamic reflux area and is initially marked as a suspected fistula area.

[0145] Furthermore, the confidence level of the fistula is calculated by combining streamline convergence and flow trajectory backflow persistence to quantitatively assess the possibility that the suspected area is a real fistula, thus avoiding misjudgment caused by single feature analysis.

[0146] Specifically, the formula for calculating the confidence level of the fistula is as follows: .in, Indicates the confidence level of the fistula. Represents streamline density set, Indicates the ratio of reflux duration, and The weighting coefficients representing streamline convergence and flow path line backflow persistence are typically set to 1. =0.4、 =0.6.

[0147] Furthermore, when the calculation yields When the value is greater than 0.75, it will be automatically identified as a real fistula area. Then, the fistula boundary will be highlighted in the 3D model, and the temporal information of the flow field parameters of the fistula area at each time will be recorded as the fistula identification result.

[0148] Furthermore, the obtained streamline set, flow trace set, and fistula identification results are combined to correct the restored three-dimensional vascular model in order to obtain the target three-dimensional vascular teaching model.

[0149] Specifically, based on the fistula boundary and flow field parameter time sequence information highlighted in the fistula identification results, the region in the three-dimensional vascular model that deviates from the actual fistula features is located and restored.

[0150] For example, to address the issues of small fistula size in the model and the lack of backflow phenomenon in the flow field simulation, the diameter of the vascular channel at the fistula in the model is adjusted by referring to the streamline convergence density distribution in the fistula region of the streamline set, so that the streamline distribution of the corrected model is consistent with the streamline characteristics of the real lesion area.

[0151] Meanwhile, by combining the temporal and spatial trajectories of particle backflow in the flow trace set, the spatial topology of blood vessels inside the model is optimized, such as adjusting the curvature of the fistula and the arterial and venous connection segments, to ensure that the flow traces can reproduce the real backflow path and duration in the corrected target three-dimensional blood vessel teaching model.

[0152] The system provided in this application embodiment also includes:

[0153] The target 3D blood vessel teaching model was subjected to multiple random scenario tests to obtain multiple scenario test results;

[0154] Feasibility verification is performed based on the test results of the multiple scenarios. If the feasibility verification fails, a first warning message is generated.

[0155] Specifically, the first step is to determine the core dimensions of randomized scenario testing, which need to cover different operational scenarios and lesion differences that may occur in clinical practice.

[0156] For example, test scenarios are set up for different sizes of arteriovenous fistulas in the target three-dimensional vascular teaching model to simulate the operation process of microcatheter superselective embolization under different fistula sizes; or scenarios are set up for fluctuations in blood flow simulation parameters to test the accuracy of the target three-dimensional vascular teaching model in responding to dynamic changes in blood flow.

[0157] In addition, it can simulate different techniques used by operators, such as the speed of microcatheter advancement and the difference in the amount of embolizing agent injected, to observe whether the target three-dimensional vascular teaching model can realistically reflect the operation feel and treatment effect, so as to ensure that the test scenario can fully cover the key variables in actual application and provide data support for subsequent verification.

[0158] Furthermore, after determining the test scenarios, random scenario tests are conducted a predetermined number of times, and detailed scenario test results are recorded for each test. These results must include three parts: the structural stability of the target 3D blood vessel teaching model, the realism of the flow field simulation, and the accuracy of the operational feedback.

[0159] Specifically, regarding structural stability, it is necessary to record whether the target 3D vascular teaching model exhibits vascular deformation or fistula position displacement after repeated interventions with the simulated device; regarding the realism of the flow field simulation, the deviation between the streamline set and flow trace set in the test scenario and the actual lesion flow field data is compared to determine whether the model's flow field simulation conforms to the actual clinical blood flow state; regarding the accuracy of operational feedback, it is necessary to collect the operator's evaluation of the instrument feel during the simulated operation, such as the resistance feedback when the microcatheter passes through the tortuous section of the blood vessel, and whether the intuitive presentation of the solidification process of the embolizing agent in the fistula area matches clinical understanding. This information should be compiled into scenario test results.

[0160] Furthermore, feasibility verification is carried out based on the collected test results from multiple scenarios, and clear qualification standards need to be set for the verification process.

[0161] For example, the structural stability qualification standard is: the deformation of the model blood vessel does not exceed 0.2 mm and the displacement of the fistula position is less than 0.1 mm in all test scenarios; the flow field simulation realism qualification standard is: the deviation rate of key parameters such as streamline density and flow trace backflow duration from the real data is less than 5%; the operation feedback accuracy qualification standard is: more than 80% of the simulated operation feedback is consistent with the actual clinical feel, and the intuitive presentation of the embolization agent solidification process is not obviously distorted.

[0162] Specifically, if all scenario test results meet the qualification standards, the feasibility verification is deemed successful and the target three-dimensional vascular teaching model can be put into subsequent applications; if any test scenario fails to meet the qualification standards, such as the deviation rate between the flow lines of the fistula area and the actual data reaching 8% in a certain test, the feasibility verification is deemed unsuccessful, and the first warning information will be automatically generated at this time.

[0163] The first warning message should clearly indicate the specific problem that failed the verification, such as "In test scenario 1, the deviation rate of the backflow duration of the flow trace in the fistula area is 8.2%, which exceeds the qualified standard". It should also include an analysis of possible causes of the problem to provide direction for subsequent model optimization. This will ensure that after targeted adjustments, the model can meet the actual needs of clinical teaching and simulation operation, and ultimately realize the reliable application of the highly realistic three-dimensional teaching model of intracranial arteriovenous fistula.

[0164] The embodiments of this application, through the specific implementation methods described above, achieve the following technical effects:

[0165] This application proposes a three-dimensional teaching model construction system for intracranial arteriovenous fistulas. First, a multimodal image enhancement module acquires a multimodal image set, which is then preprocessed and enhanced with Frangi filtering to improve the vessel edges, resulting in an enhanced multimodal image set. Next, a point cloud skeleton generation module registers, resamples, and segments the vessel region, extracting voxel data and converting it into a three-dimensional point cloud. Simultaneously, the vessel skeleton is extracted and smoothed to obtain a three-dimensional vessel point cloud and vessel skeleton. Subsequently, a vessel sub-point cloud segmentation module divides the point cloud into multiple sub-point clouds based on skeleton connection points. A vessel sub-point cloud restoration and stitching module detects and restores occlusions in each sub-point cloud, then precisely stitches and simulates the result to form a restored three-dimensional vessel model. Finally, a fistula identification and correction module imports the restored model into a fluid simulation platform, meshes it, and injects virtual tracer particles to generate streamlines and flow traces. Based on flow field characteristics, the fistula is identified and the model is corrected to obtain the target three-dimensional vessel teaching model. Simultaneously, the target model undergoes random scene testing and feasibility verification; if verification fails, a warning message is generated.

[0166] The system provided in this application solves the problem that traditional teaching relies on two-dimensional images or simple models, which cannot restore the real vascular structure and blood flow characteristics, resulting in low teaching efficiency and high clinical risk. It achieves a high-fidelity construction of intracranial arteriovenous fistula teaching models, providing reliable technical support for medical student training, surgeon preoperative rehearsal, and new device testing. The system uses a technical solution of "multimodal image enhancement - three-dimensional point cloud and skeleton generation - vascular sub-point cloud division and restoration stitching - fistula identification and correction - model testing and verification".

[0167] Example 2, as shown in the appendix Figure 2 As shown, based on the inventive concept of a three-dimensional teaching model construction system for intracranial arteriovenous fistulas provided in Embodiment 1, this application also provides a method for constructing a three-dimensional teaching model for intracranial arteriovenous fistulas, specifically including:

[0168] A multimodal image set is acquired, and image preprocessing and vascular edge enhancement are performed using the Frangi filtering algorithm to obtain an enhanced multimodal image set. Point cloud generation and skeleton extraction are performed based on the enhanced multimodal image set to obtain a three-dimensional vascular point cloud and vascular skeleton. The three-dimensional vascular point cloud is divided according to the connection points of the vascular skeleton to obtain multiple sub-point clouds of the divided three-dimensional vascular system. Occlusion detection and restoration are performed on the multiple sub-point clouds of the divided three-dimensional vascular system, and the obtained multiple restored sub-point clouds of the divided three-dimensional vascular system are stitched together and simulated to obtain a restored three-dimensional vascular model. Blood flow path tracking and fistula identification are performed based on the restored three-dimensional vascular model, and the restored three-dimensional vascular model is corrected according to the identification results to obtain a target three-dimensional vascular teaching model.

[0169] In one embodiment, point cloud generation and skeleton extraction are performed based on the enhanced multimodal image set to obtain a three-dimensional blood vessel point cloud and blood vessel skeleton, which is also used for:

[0170] The enhanced multimodal image set is registered and resampled to obtain a multimodal fused image; blood vessels are segmented based on the multimodal fused image to extract voxel data of the blood vessel region; point cloud generation and skeleton extraction are performed based on the voxel data of the blood vessel region to obtain the three-dimensional blood vessel point cloud and blood vessel skeleton.

[0171] Furthermore, the process of generating point clouds and extracting skeletons based on the voxel data of the vascular region to obtain the three-dimensional vascular point cloud and vascular skeleton also includes:

[0172] The voxel data of the vascular region is converted into a three-dimensional point cloud, and then denoised and sparsified to obtain a three-dimensional vascular point cloud; the voxel data of the vascular region is then subjected to skeleton extraction and smoothing to obtain a vascular skeleton.

[0173] In one embodiment, blood flow path tracing and fistula identification are performed based on the restored three-dimensional vascular model, and the restored three-dimensional vascular model is corrected according to the identification results to obtain a target three-dimensional vascular teaching model. This is also used for:

[0174] The restored 3D blood vessel model is imported into a fluid simulation platform. The vascular lumen is meshed using the finite volume method to obtain a meshing result, wherein the mesh type is a tetrahedral hybrid mesh. According to preset simulation conditions, virtual tracer particles are injected into the blood vessel inlet of the meshed result to generate a streamline set and a flow path set. Based on the streamline set and flow path set, the flow field characteristics are identified in a time sequence to obtain the fistula identification result. The restored 3D blood vessel model is corrected by combining the streamline set, flow path set and fistula identification result to obtain the target 3D blood vessel teaching model.

[0175] Furthermore, the preset simulation conditions also include preset viscosity, preset density, preset inlet flow rate, and preset outlet pressure.

[0176] Furthermore, based on the streamline set and the flow path set, the flow field feature time sequence identification is performed to obtain the fistula identification result, which also includes:

[0177] Based on the streamline set, an abnormal flow field region is determined, and an abnormal flow path set associated with the abnormal flow field region is extracted from the flow path set; the abnormal flow path set is traversed to perform spatial feature identification, and an abnormal flow path spatial feature result set is obtained; the abnormal flow path spatial feature result set is performed according to the abnormal flow path spatial feature result set to determine the fistula identification result.

[0178] In one embodiment, a method for constructing a three-dimensional teaching model of an intracranial arteriovenous fistula further includes:

[0179] The target 3D blood vessel teaching model is subjected to multiple random scenario tests to obtain multiple scenario test results; the feasibility is verified based on the multiple scenario test results, and if the feasibility verification fails, a first warning message is generated.

[0180] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0181] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0182] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application intends to include such modifications and variations.

Claims

1. A three-dimensional teaching model construction system for intracranial arteriovenous fistulas, characterized in that, The system includes: The multimodal image enhancement module is used to acquire a multimodal image set, perform image preprocessing, and enhance blood vessel edges using the Frangi filtering algorithm to obtain an enhanced multimodal image set. The point cloud skeleton generation module is used to generate point clouds and extract skeletons based on the enhanced multimodal image set to obtain three-dimensional blood vessel point clouds and blood vessel skeletons. The blood vessel sub-point cloud division module is used to divide the three-dimensional blood vessel point cloud according to the connection points of the blood vessel skeleton to obtain multiple divided three-dimensional blood vessel sub-point clouds. The blood vessel sub-point cloud restoration and stitching module is used to perform occlusion detection and restoration on the multiple divided three-dimensional blood vessel sub-point clouds, and to stitch and simulate the obtained multiple restored divided three-dimensional blood vessel sub-point clouds to obtain a restored three-dimensional blood vessel model. The fistula identification and correction module is used to perform blood flow path tracking and fistula identification based on the restored three-dimensional vascular model, and to correct the restored three-dimensional vascular model according to the identification results to obtain the target three-dimensional vascular teaching model. Point cloud generation and skeleton extraction are performed based on the enhanced multimodal image set to obtain a three-dimensional blood vessel point cloud and blood vessel skeleton, including: The enhanced multimodal image set is registered and resampled to obtain a multimodal fused image; Based on the multimodal fused image, blood vessel segmentation is performed, and voxel data of the blood vessel region is extracted; Point cloud generation and skeleton extraction are performed based on the voxel data of the blood vessel region to obtain the three-dimensional blood vessel point cloud and blood vessel skeleton. Point cloud generation and skeleton extraction are performed based on the voxel data of the vascular region to obtain the three-dimensional vascular point cloud and vascular skeleton, including: The voxel data of the vascular region is converted into a three-dimensional point cloud, and then denoised and sparsified to obtain a three-dimensional vascular point cloud. The voxel data of the vascular region is subjected to skeleton extraction and smoothing to obtain the vascular skeleton.

2. The three-dimensional teaching model construction system for intracranial arteriovenous fistula as described in claim 1, characterized in that, Based on the restored 3D vascular model, blood flow path tracing and fistula identification are performed, and the restored 3D vascular model is corrected according to the identification results to obtain the target 3D vascular teaching model, including: The restored three-dimensional blood vessel model is imported into a fluid simulation platform, and the blood vessel lumen is meshed using the finite volume method to obtain the meshing result, wherein the mesh type is a tetrahedral hybrid mesh. According to the preset simulation conditions, virtual tracer particles are injected into the blood vessel entrance of the mesh division result to generate streamline set and flow path set; Based on the streamline set and the flow path set, flow field characteristics are identified in a time sequence to obtain fistula identification results; The restored three-dimensional vascular model is corrected by combining the streamline set, flow trace set, and fistula identification results to obtain the target three-dimensional vascular teaching model.

3. The three-dimensional teaching model construction system for intracranial arteriovenous fistula as described in claim 2, characterized in that, The preset simulation conditions include preset viscosity, preset density, preset inlet flow rate, and preset outlet pressure.

4. The three-dimensional teaching model construction system for intracranial arteriovenous fistula as described in claim 2, characterized in that, Based on the streamline set and the flow path set, flow field characteristics are identified in a time sequence to obtain fistula identification results, including: Based on the streamline set, the abnormal flow field region is determined, and the abnormal flow path set associated with the abnormal flow field region is extracted from the flow path set. The set of abnormal flow lines is traversed to perform spatial feature identification, and a set of spatial feature results of abnormal flow lines is obtained. The fistula identification result is determined by performing a union analysis on the set of spatial feature results of the abnormal flow lines.

5. The three-dimensional teaching model construction system for intracranial arteriovenous fistula as described in claim 1, characterized in that, include: The target 3D blood vessel teaching model was subjected to multiple random scenario tests to obtain multiple scenario test results; Feasibility verification is performed based on the test results of the multiple scenarios. If the feasibility verification fails, a first warning message is generated.

6. A method for constructing a three-dimensional teaching model of an intracranial arteriovenous fistula, characterized in that, The method is applied to a three-dimensional teaching model construction system for intracranial arteriovenous fistulas as described in any one of claims 1-5, and the method includes: A multimodal image set was acquired, and image preprocessing and vascular edge enhancement were performed using the Frangi filtering algorithm to obtain an enhanced multimodal image set. Point cloud generation and skeleton extraction are performed based on the enhanced multimodal image set to obtain three-dimensional blood vessel point cloud and blood vessel skeleton; The three-dimensional blood vessel point cloud is divided according to the connection points of the blood vessel skeleton to obtain multiple divided three-dimensional blood vessel sub-point clouds; Occlusion detection and restoration are performed on the multiple segmented 3D blood vessel sub-point clouds, and the obtained multiple restored segmented 3D blood vessel sub-point clouds are stitched together and simulated to obtain a restored 3D blood vessel model. Based on the restored three-dimensional vascular model, blood flow path tracing and fistula identification are performed, and the restored three-dimensional vascular model is corrected according to the identification results to obtain the target three-dimensional vascular teaching model.

Images