A lower extremity venogram image processing method, system, terminal and medium

Abstract

The present application relates to the field of medical image processing, and specifically provides a lower limb venography image processing method, system, terminal and medium. The method comprises: acquiring a KTS patient lower limb venography three-dimensional image, using two groups of threshold values to extract superficial veins, deep veins and collateral veins, after morphological closing operation processing, extracting the blood vessel skeleton through the topological keeping three-dimensional thinning algorithm and pruning; then converting the skeleton into a graph structure and marking the loop nodes and edges, calculating the quantitative parameters such as blood vessel tortuosity, branch and loop complexity according to the above, and finally generating a three-dimensional visual image. The present application can adapt to the KTS blood vessel gray scale distribution and complex topological characteristics, effectively avoid the problems of blood vessel missed detection, noise interference and topological distortion, improve the image processing precision, and ensure the reliability of the parameter calculation result.

Description

Technical Field

[0001] This invention relates to the field of medical image processing, specifically to a method, system, terminal, and medium for processing lower extremity venography images. Background Technology

[0002] Klippel-Trenaunay syndrome (KTS) is a rare congenital vascular malformation characterized by underdeveloped or absent deep veins in the lower extremities, varicose superficial veins, and extensive collateral circulation. The venous network in the lower extremities of patients often presents a typical "tangled" high-density tortuous structure. Clinical assessment of the lower extremity veins in KTS patients primarily relies on computed tomography venography (CTV) or magnetic resonance venography (MRV) to obtain three-dimensional image data for observation and evaluation of the vascular structure.

[0003] In the field of medical image processing, there are already relatively mature technical solutions for the segmentation and three-dimensional reconstruction of vascular structures. For example, threshold segmentation methods are used to extract vascular regions from angiographic images, and three-dimensional thinning algorithms are used to extract the vascular centerline skeleton, thereby constructing a graph structure for topological analysis. These methods are widely used in the quantitative analysis of luminal structures such as cerebral blood vessels and coronary arteries.

[0004] Existing thresholding methods typically extract blood vessels using a single threshold range. However, deep veins in KTS patients often suffer from underdeveloped growth, making it difficult for contrast agents to effectively fill them. Their image grayscale values ​​are significantly lower than those of normally filled superficial veins. When using a single threshold, if the threshold is set too high, poorly visualized deep veins and small collateral vessels are missed; if the threshold is set too low, background noise such as bone and soft tissue hyperplasia is often misidentified as blood vessels. While some studies have attempted to improve segmentation using local or adaptive thresholding, these methods primarily target ordinary vein images with relatively uniform grayscale distributions. For the unique grayscale distribution characteristics of blood vessels in KTS patients, the segmentation results are unsatisfactory. Furthermore, the lower extremity venous system in KTS patients exhibits a tangled, malformed appearance, containing numerous abnormal collateral circulations, communicating branches, and vascular loops. Existing image processing techniques often struggle to remove bone interference while maintaining the complete topological structure of the vascular network, reducing the accuracy of image processing results and consequently affecting the accuracy of parameter calculations. Summary of the Invention

[0005] To address the aforementioned issues, this invention provides a method, system, terminal, and medium for processing lower extremity venography images, thereby improving the accuracy of the overall image processing results and ensuring the reliability of vascular quantification parameter calculations.

[0006] In a first aspect, the technical solution of the present invention provides a method for processing lower extremity venography images, comprising the following steps: Three-dimensional image data of lower extremity venography of KTS patients were acquired, and the three-dimensional image data were binarized and segmented to extract the initial three-dimensional binarized volume data of the veins. Three-dimensional morphological closing operation was performed on the initial three-dimensional binarized volume data. The segmentation process includes extracting superficial veins using a first threshold range and extracting deep veins and collateral veins using a second threshold range. A topology-preserving 3D thinning algorithm is performed on the 3D binary volume data after the closing operation to extract the vascular centerline skeleton with a single pixel width, and a pruning operation is performed on the vascular centerline skeleton based on a preset pruning threshold. The extracted vascular centerline skeleton is converted into a graph structure G=(V,E), and the loop nodes and loop edges are marked, where V represents the set of nodes and E represents the set of edges connecting the nodes. The nodes include vascular bifurcation points and vascular endpoints. Based on the graph structure, vascular quantification parameters are calculated, including tortuosity index, whole graph weighted average tortuosity index, branch complexity index, and loop complexity index. Based on the graph structure, a three-dimensional visualization image is generated.

[0007] Secondly, the technical solution of the present invention provides a lower extremity venography image processing system, comprising: The data acquisition and segmentation module is used to acquire three-dimensional image data of lower extremity venography of KTS patients, perform binarization segmentation on the three-dimensional image data, extract the initial three-dimensional binarized volume data of the veins, and perform three-dimensional morphological closing operation on the initial three-dimensional binarized volume data; wherein, the segmentation process includes extracting superficial veins using a first threshold range and extracting deep veins and collateral veins using a second threshold range; The skeleton extraction and pruning module is used to perform a topology-preserving 3D thinning algorithm on the 3D binary volume data after the closing operation, extract the blood vessel centerline skeleton with a single pixel width, and perform pruning operation on the blood vessel centerline skeleton based on a preset pruning threshold. The graph structure conversion module is used to convert the extracted vascular centerline skeleton into a graph structure G=(V,E) and mark the loop nodes and loop edges in it, where V represents the set of nodes and E represents the set of edges connecting the nodes. The nodes include vascular bifurcation points and vascular endpoints. The quantification parameter calculation module is used to calculate the quantification parameters of blood vessels based on the graph structure. The quantification parameters include tortuosity index, whole graph weighted average tortuosity index, branch complexity index, and loop complexity index. A 3D visualization module is used to generate 3D visualization images based on the graph structure.

[0008] Thirdly, the technical solution of the present invention provides a terminal, comprising: The memory is used to store the lower extremity venography image processing program; A processor is configured to implement the steps of the lower extremity venography image processing method as described above when executing the lower extremity venography image processing program.

[0009] Fourthly, the present invention provides a computer-readable storage medium storing a lower extremity venography image processing program, wherein the lower extremity venography image processing program, when executed by a processor, implements the steps of the lower extremity venography image processing method as described in any of the above claims.

[0010] As can be seen from the above technical solutions, this application has the following advantages: In the segmentation stage, two sets of threshold ranges are distinguished and used to extract superficial veins, deep veins, and collateral veins, respectively. The Otsu algorithm is preferred to expand the two sets of thresholds based on the globally optimal threshold to ensure that low-grayscale pathological vessels are not missed and to avoid large-scale background noise mixing. This adapts to the characteristics of grayscale differentiation of superficial and deep veins in KTS and improves the integrity and accuracy of 3D vessel binarization segmentation, thereby ensuring the accuracy of skeleton extraction and subsequent parameter calculation. At the same time, 3D morphological closing operation is used to process the segmentation results, fill in the small breaks and holes in the vessels, ensure the connectivity of the vascular structure, and obtain complete and continuous 3D binarized data of veins from the source, further ensuring the accuracy of image processing results. First, a topology-preserving 3D thinning algorithm is applied to the segmented volume data, simplifying blood vessels into single-pixel centerline skeletons. The entire process maintains the original connectivity, bifurcation, and loop topological relationships of the blood vessels, without disrupting the complex vascular network characteristic of KTS. Then, pruning operations are used to remove invalid branches from the skeleton and eliminate false structures caused by image noise. The final result is a concise, clean, and topology-preserving vascular centerline skeleton, achieving both lightweight representation of vascular morphology and complete preservation of the structural features of diseased blood vessels, effectively ensuring the accuracy of the image processing results. This method transforms the vascular skeleton into a standard graph structure containing nodes and edges, and labels the nodes and edges related to loops. This converts the complex venous network into a topological model that can be analyzed by a computer. Based on this, parameters such as tortuosity, branch complexity, and loop complexity are quantitatively calculated to achieve numerical evaluation. Furthermore, the accuracy of the numerical calculation results is ensured by leveraging the accuracy of image processing. KTS (Kinetic Traumatic Syndrome) is characterized by vascular loops formed by abnormal collaterals and communicating branches. Individually labeling the nodes and edges related to these loops facilitates the differentiation between ordinary vascular segments and pathological loop structures, avoiding confusion between loop data and conventional vascular data. This allows for independent statistical analysis of indicators reflecting collateral compensation, such as the number and length of loops. Moreover, it enables rapid location of all loop regions during topological retrieval and data filtering, and allows for differentiated display and highlighting of loop structures during 3D visualization. Attached Figure Description

[0011] To more clearly illustrate the technical solution of this application, the accompanying drawings used in the description will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is a schematic flowchart of a lower extremity venography image processing method provided in an embodiment of the present invention.

[0013] Figure 2 This is a display image showing the MPR multiplane reconstruction effect of the original DICOM 3D volume data.

[0014] Figure 3 This is a schematic diagram showing the segmentation effect of superficial veins extracted using the first threshold range.

[0015] Figure 4 This is a schematic diagram showing the segmentation effect of deep veins and collateral veins extracted using the second threshold range.

[0016] Figure 5 This is a schematic diagram illustrating the effect of merging the two segmented structures.

[0017] Figure 6 This is a schematic diagram illustrating the effect after performing a three-dimensional morphological closing operation.

[0018] Figure 7 A schematic diagram illustrating the 3D visualization effect of tortuosity.

[0019] Figure 8 This is a 3D visualization of the loop complexity index.

[0020] Figure 9 This is a schematic block diagram of a lower extremity venography image processing system provided in an embodiment of the present invention.

[0021] Figure 10 This is a schematic diagram of the structure of a terminal provided in an embodiment of the present invention. Detailed Implementation

[0022] To make the purpose, features, and advantages of this application more apparent and understandable, specific embodiments and accompanying drawings will be used to clearly and completely describe the technical solution protected by this application. Obviously, the embodiments described below are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0023] Unless otherwise defined, all technical and scientific terms used in this application have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used in this application and in the specification of this invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention.

[0024] Figure 1 This is a schematic flowchart of a lower extremity venography image processing method provided in an embodiment of the present invention. Figure 1 The executing entity can be a lower extremity venography image processing system. The lower extremity venography image processing method provided in this embodiment of the invention is executed by a computer device, and correspondingly, the lower extremity venography image processing system runs on the computer device. Depending on different needs, the order of the steps in this flowchart can be changed, and some steps can be omitted.

[0025] like Figure 1 As shown, the method includes the following steps.

[0026] S1. Acquire three-dimensional image data of lower extremity venography of KTS patients, perform binarization segmentation on the three-dimensional image data, extract the initial three-dimensional binarized volume data of veins, and perform three-dimensional morphological closing operation on the initial three-dimensional binarized volume data; wherein, the segmentation process includes extracting superficial veins using a first threshold range and extracting deep veins and collateral veins using a second threshold range.

[0027] S2 executes a topology-preserving 3D thinning algorithm on the 3D binary volume data after the closing operation to extract the vascular centerline skeleton with a single pixel width, and performs a pruning operation on the vascular centerline skeleton based on a preset pruning threshold.

[0028] S3. The extracted vascular centerline skeleton is converted into a graph structure G=(V,E), and the loop nodes and loop edges are marked, where V represents the set of nodes and E represents the set of edges connecting the nodes. The nodes include vascular bifurcation points and vascular endpoints.

[0029] S4. Based on the graph structure, calculate the vascular quantification parameters, including the tortuosity index, the whole graph weighted average tortuosity index, the branch complexity index, and the loop complexity index.

[0030] S5. Based on the graph structure, generate a three-dimensional visualization image.

[0031] As a refinement and extension of the specific implementation of the above embodiments, in order to fully explain the specific implementation process of this embodiment, the following will provide possible embodiments to describe the specific implementation of the above steps in a non-limiting manner.

[0032] In this embodiment, dual-source CT lower extremity venography (CTV) is used to acquire three-dimensional image data of the lower extremities of KTS patients. Specifically, the original DICOM three-dimensional volumetric data of the lower extremities of KTS patients is acquired, with each voxel value being the CT value / HU value. This volumetric data is stored in the form of a three-dimensional array. To facilitate a visual representation of the spatial structure of the volumetric data, Figure 2 The image shows the MPR multiplanar reconstruction display effect of the original DICOM 3D volume data. Considering that KTS patients often have lower extremity soft tissue hyperplasia and lymphedema, leading to increased image noise and blurred vascular boundaries, this embodiment first uses adaptive Gaussian filtering to denoise the 3D image data before binarization segmentation. The filter kernel size is adaptively adjusted according to the local image noise level. Simultaneously, in KTS patients with incomplete deep vein development, contrast agents have difficulty filling the underdeveloped lumen, resulting in low grayscale values ​​in the deep vein region of the image. This embodiment uses multi-threshold segmentation, i.e., setting a first threshold range for extracting normally filled superficial veins, and setting a second threshold range for extracting poorly visualized deep veins and collateral veins. Then, the two segmentation results are merged to extract all venous structures of KTS patients as completely as possible.

[0033] Specifically, in step S1 of this embodiment, the three-dimensional image data is binarized and segmented to extract the three-dimensional binarized volume data of the vein, which specifically includes the following steps.

[0034] S1.1 Calculate the filter kernel size based on the local image noise level, and perform Gaussian filtering on the 3D image data based on the filter kernel size.

[0035] This embodiment aims to suppress image noise caused by soft tissue hyperplasia and lymphedema in KTS patients while preserving vascular edge details, based on the original three-dimensional image data. Perform adaptive Gaussian filtering.

[0036] Specifically, for each voxel in the 3D image First, calculate its target neighborhood, such as the 26-neighborhood. Local gray-level variance within) , represented as:

[0037] in, The mean gray value of the voxels in the neighborhood.

[0038] Based on local variance Calculate the adaptive Gaussian kernel parameters , represented as:

[0039] in, Based on the Gaussian kernel standard deviation, For adaptive adjustment coefficient, These are the minimum and maximum values ​​of the local variance of all voxels in the entire image, respectively.

[0040] grayscale value after filtering of voxel p Calculated by the following Gaussian convolution:

[0041] in, This is the normalization coefficient.

[0042] Through the above processing, the filtered image data is output. This serves as the input for the next step.

[0043] S1.2, the region growing algorithm is used to automatically identify and mark the skeletal region, and the skeletal region is excluded from the range to be segmented.

[0044] Patients with KTS often have unilateral limb osteophyte formation, which interferes with venous visualization. This embodiment utilizes the CT value characteristics of bone tissue and the grayscale difference of veins to automatically identify and mark bone regions using a region growing algorithm before binarization segmentation. Then, bone regions are excluded from the segmentation range to avoid missegmenting bones as blood vessels.

[0045] Specifically, the baseline region is determined based on CT values. For example, the CT value of bone tissue is usually above 300 HU, and the base area is set to [300, 1500] HU.

[0046] statistics CT values ​​in Within the range of voxels, spatially continuous voxels located in the central region of the image are selected as candidate regions. Among these candidate regions, the voxel with the highest CT value in the intermediate layer image along the z-axis (human body long axis) is selected as the initial seed point. It should be noted that the seed point must be located in an anatomical skeletal region.

[0047] Using the 26-neighborhood connectivity rule, starting from the seed point Starting from this point, iteratively incorporate neighboring voxels that satisfy the following growth conditions: and

[0048] That is, the CT value of candidate voxel q is in Within the range, and the difference in CT value from the seed point is less than HU, This is an empirical threshold. Repeat the above process until no new voxels are included, generating a binary mask for the skeletal region. .

[0049] In the filtered image, the pixel values ​​of the area covered by the bone mask are forced to be 0 (background):

[0050] After the above processing, the image data after excluding the skeleton is output. This serves as the input for the next step.

[0051] S1.3, a first threshold range and a second threshold range are preset. The lower limit of the first threshold range is higher than the upper limit of the second threshold range. Normally filled superficial veins are extracted using the first threshold range, and poorly visualized deep veins and collateral veins are extracted using the second threshold range. The two segmentation results are merged to generate initial three-dimensional binary volume data. For example... Figure 3 The image shows a schematic diagram of the segmentation effect of superficial veins extracted using the first threshold range. Figure 4 The image shows a schematic diagram illustrating the segmentation effect of deep veins and collateral veins extracted using the second threshold range. Figure 5 The image shows a schematic diagram illustrating the effect of merging the two segmented structures.

[0052] In patients with KTS, the CT values ​​of normally filled superficial veins are high (high contrast agent concentration), while the CT values ​​of underdeveloped deep veins are low (less contrast agent filling and narrower lumen). Traditional single-threshold segmentation uses a uniform CT value range for all vessels, which either misses deep veins or misinterprets noise as vessels. Therefore, this embodiment uses dual-threshold segmentation.

[0053] Specifically, a first threshold range is preset. Second threshold range , Based on the first threshold range Voxels with grayscale values ​​within this range are extracted as normally filled superficial veins, represented as follows:

[0054] Based on the second threshold range Voxels with grayscale values ​​within this range are extracted as poorly visualized deep veins and collateral veins, represented as follows:

[0055] Perform a logical OR operation on the two segmentation results to generate the initial 3D binary volume data, represented as follows:

[0056] That is, as long as a voxel meets either the first threshold range or the second threshold range, it is marked as a blood vessel (value 1).

[0057] After the above processing, the initial 3D binarized volume data is output. This serves as the input for the next step.

[0058] S1.4 performs a 3D morphological closing operation on the initial 3D binary volume data, first performing dilation and then erosion, outputting 3D binary volume data with preserved connectivity. For example... Figure 6 As shown, this is for Figure 5 The result is a schematic diagram of the effect after performing three-dimensional morphological closing operation, which connects the broken segments and enhances the connectivity of the vascular structure.

[0059] Patients with KTS may have irregular stenosis of the blood vessel lumen or discontinuous contrast agent filling, resulting in holes or breaks in the binarized blood vessels. In this embodiment, a three-dimensional morphological closing operation is performed on the binarized volume data. First, an expansion operation is performed, and then an erosion operation is performed. This can fill the tiny holes on the blood vessel wall, connect blood vessel segments that are broken due to poor imaging, and maintain the overall connectivity of the vascular network.

[0060] Specifically, using structural element B... A 3D dilation operation is performed. This operation sets the central voxel to 1 if any voxel within the structuring element's coverage area is set to 1, thereby expanding the vessel region, filling holes, and connecting adjacent fracture segments. The same structuring element B is then used to perform a 3D erosion operation on the dilation result. This operation preserves the central voxel if the structuring element is completely within the vessel region; otherwise, it sets it to 0, thus restoring the over-expanded area from the dilation operation to the original vessel boundary. The combination of dilation and erosion constitutes a closing operation. After these processes, the output is a connectivity-preserving 3D binary volumetric data. Voxels with a value of 1 represent blood vessels, and voxels with a value of 0 represent the background.

[0061] In some alternative implementations, the first threshold range and the second threshold range are determined based on the Otsu algorithm, specifically including the following steps S1.31 to S1.33.

[0062] S1.31, the Otsu algorithm is used to calculate the globally optimal threshold for the 3D image data. .

[0063] The Otsu algorithm, based on gray-level histograms, determines the segmentation threshold by maximizing the inter-class variance, dividing the image into foreground (blood vessels) and background (non-blood vessels). For gray-level ranges of... A 3D image, let the number of voxels at gray level i be... Total prime numbers The probability of gray level i appearing is .

[0064] Let the threshold be T, and divide the image into two categories: (grayscale levels [0,T]) and (Gray level [T+1, L) 1). The probabilities of the two types occurring are as follows:

[0065] The mean gray values ​​of the two classes are as follows:

[0066] The global grayscale mean is:

[0067] The formula for calculating between-class variance is:

[0068] Global optimal threshold To make the inter-class variance Maximum T value: .

[0069] S1.32, using the globally optimal threshold Determine the basic threshold range centered on the target.

[0070] In patients with KTS (Kidney Tissue Syndrome), the deep veins are underdeveloped, resulting in insufficient contrast filling. Their CT values ​​are often lower than those of normally filled vessels, and the image grayscale distribution exhibits a low-grayscale tailing characteristic. Standard Otsu thresholding tends to treat the main peak in the grayscale histogram as the foreground, easily misclassifying the underdeveloped veins in the low-grayscale tailing region as background. Therefore, a globally optimal threshold is used. Based on the baseline, extend downwards by a preset amplitude. This is to increase the detection rate of poorly visualized blood vessels.

[0071] Let the expanded low threshold boundary be The high threshold boundary is The basic threshold range is defined as follows:

[0072]

[0073] in, This is the amount of downward expansion from the low threshold. This represents the upward expansion amount from the high threshold. These are the preset minimum and maximum allowed thresholds, respectively.

[0074] S1.33, Set the first threshold range to The second threshold range is ,in, , This represents the second downward expansion amount.

[0075] In this embodiment, the three-dimensional binary volume data obtained through step S1 is obtained. Subsequently, the "tangled" three-dimensional vascular network of the lower extremity veins of KTS patients was simplified into a single-pixel-width centerline skeleton, while maintaining the original topological structure of the blood vessels, in order to facilitate subsequent graph structure conversion and quantitative analysis.

[0076] This embodiment employs a parallel 3D thinning algorithm to extract the vascular centerline skeleton with a single pixel width. This algorithm iteratively deletes boundary voxels that meet specific conditions in the 3D binary volumetric data. Simultaneously, topological number constraints ensure that each deletion operation does not alter the vascular topology, i.e., it does not produce breaks, holes, or merge independent branches. The lower extremity veins of KTS patients exhibit a "tangled" network structure with numerous collateral circulations and aberrant communicating branches. This embodiment uses topological number constraints and pruning optimization to ensure the correct skeleton can be extracted from the complex vascular network of KTS.

[0077] Specifically, in step S2 of this embodiment, a topology-preserving 3D thinning algorithm is performed on the 3D binary volume data to extract the vascular centerline skeleton with a single pixel width, which specifically includes the following steps S2.1 and S2.2.

[0078] S2.1 Determine if any voxels have been deleted in the current global iteration. If they have been deleted in the first iteration or in the current iteration, proceed to the next step. Otherwise, terminate the iteration and output all remaining foreground voxels as the skeleton of the blood vessel centerline with a single pixel width.

[0079] It should be noted that the input is 3D binarized volume data. Voxels with a value of 1 represent blood vessels (foreground) and voxels with a value of 0 represent background.

[0080] If this is the first round, then execute step S2.2. If it is not the first round, and no voxels are deleted in the global iteration of this round, then terminate the iteration and output all remaining foreground voxels as the skeleton of the blood vessel centerline with a single pixel width. Otherwise, execute step S2.2.

[0081] S2.2, execute sub-iterations in six directions in sequence, the six directions being east, west, south, north, up, and down.

[0082] In a 3D image space, each voxel has three coordinate axes (X, Y, Z) in a Cartesian coordinate system, and each axis has two orientations (positive and negative). Therefore, boundary voxels in 3D space can be divided into six categories according to the orientation of their neighborhood background: east, west, south, north, up, and down.

[0083] The east direction corresponds to the X-axis and represents the positive direction; the west direction corresponds to the X-axis and represents the negative direction; the south direction corresponds to the Y-axis and represents the positive direction; the north direction corresponds to the Y-axis and represents the negative direction; the up direction corresponds to the Z-axis and represents the positive direction; the down direction corresponds to the Z-axis and represents the negative direction.

[0084] The six sub-iterations are executed sequentially, each sub-iteration including the following steps S2.2.1 to S2.2.5.

[0085] S2.2.1, In the current sub-iteration direction, mark all boundary voxels with the current direction neighborhood voxels as background.

[0086] Specifically, in the current sub-iteration direction d, voxels that satisfy the following condition are marked: the neighboring voxels of this voxel in the current direction are background (with a value of 0). For example, in the east sub-iteration, all voxels at position x+1 are marked as background voxels.

[0087] S2.2.2, for each marked boundary voxel p, according to formula Calculate the topological number, where Let be the number of connected components of the foreground voxels in the 26-neighborhood of voxel p under 26 connectivity. denoted as the number of connected components of the background voxel within the 6-neighborhood of voxel p under 6-connectivity.

[0088] in, It is the set of all foreground voxels in the 26-neighborhood of voxel p (excluding p itself). This is the set of all background voxels within the 6-neighborhood of voxel p (excluding p itself). In this embodiment, the foreground (vessels) uses 26 connectivity, and the background (non-vessels) uses 6 connectivity. It should be noted that vessels can travel in any direction in three-dimensional space, and the 26-neighborhood ensures the connectivity of adjacent vessels traveling diagonally or in high-density networks, avoiding false breaks caused by discretization.

[0089] The topological number takes values ​​of 0, 1, and 2. Deleting voxel p does not change the connectivity of the foreground.

[0090] It should be noted that whether the topology of the local region changes after deleting a voxel p depends on two factors: first, the connectivity of the foreground voxels in the neighborhood of p; and second, the connectivity of the foreground voxels in the neighborhood of p. Secondly, the connectivity of background voxels within the neighborhood of p. ).when When p is deleted, the two independent foreground regions are not merged, nor are new background holes created, so the topology remains unchanged.

[0091] Using 26 connectivity for the foreground and 6 connectivity for the background is a choice made by the 3D thinning algorithm to avoid the connectivity paradox. This combination ensures that the calculation of the topological number is consistent with the Euler feature, thereby ensuring that the original topological structure of the blood vessels is not destroyed during the thinning process.

[0092] S2.2.3 Determine whether the boundary voxel simultaneously satisfies the following deleteable conditions: the number of foreground voxels in the 26-neighborhood of the voxel is greater than or equal to 2, the topology number is equal to 1, and no new holes are generated after deletion.

[0093] The method for determining whether a new hole is generated after deletion is as follows: For candidate voxel p, temporarily change its value from 1 to 0, calculate the number of connected components of the background voxels in its 6-neighborhood, and denot it as p. Restore the original value of p, calculate the number of connected components of the background voxels in its 6-neighborhood, and denot it as p. .like If p is deleted, no new hole will be created; if If p is deleted, a new hole will be created, and deletion should be prohibited. Preferably, to reduce computational complexity, a predefined voxel connectivity pattern is typically used for table lookup, eliminating the need for real-time calculation of connected components.

[0094] If the number of foreground voxels in the 26-neighborhood is greater than or equal to 2, it means that the voxel is not a blood vessel terminal, and the blood vessel endpoint can be preserved to avoid skeletal shrinkage.

[0095] A topology number of 1 indicates that deletion does not change local connectivity, ensuring that independent branches are not merged and no breaks occur.

[0096] Deleting a hole will not create a new background cavity, ensuring that no false holes are created.

[0097] S2.2.4 removes boundary voxels that meet all the deletability conditions from the foreground, i.e., changes the voxel value from 1 to 0.

[0098] S2.2.5 If the sub-iterations in the six directions are not completed, the sub-iterations in the next direction are executed; if they are completed, the next round of global iterations is executed.

[0099] Specifically, if the six directional sub-iterations are not completed, return to S2.2.1 and proceed to the next direction; if all six directional sub-iterations are completed, return to S2.1 and begin a new round of global iteration. When no voxels are deleted in a certain round of global iteration, the algorithm terminates, outputting all remaining foreground voxels to form a single-pixel-wide vessel centerline skeleton. .

[0100] Because of the irregularity of the blood vessel walls in KTS patients or the presence of slight noise in image segmentation, there may be extremely short false branches (spicules) on the extracted skeleton. In order to eliminate the interference of these spicules on the subsequent image structure transformation, this embodiment performs a pruning operation after the thinning is completed, specifically including the following steps S2.3 to S2.5.

[0101] S2.3, Set a preset pruning threshold based on the statistical characteristics of lower extremity venous images of KTS patients.

[0102] In lower extremity venography images of KTS patients, two types of "short branch" structures exist on the skeleton extracted by the 3D thinning algorithm: noise spikes and real fine collateral branches. Noise spikes originate from irregular vessel walls and image segmentation noise and should be deleted, while real fine collateral branches are a characteristic pathological change of KTS and should be retained.

[0103] In patients with KTS, the actual collateral vessels, such as the lateral vein, persistent sciatic vein, and ascending lumbar vein, appear as tubular structures of a certain length in CTV images, typically greater than 5 mm. In contrast, spurious structures created by image segmentation noise and local irregularities in the vessel wall are usually shorter.

[0104] This embodiment is based on clinical observation. Through statistical analysis of KTS patient samples, the length distribution of actual collateral vessels was measured, and the 5th percentile of the length distribution was taken as the pruning threshold.

[0105] S2.4, traverse all endpoint voxels in the skeleton, starting from each endpoint voxel and tracing inward along the skeleton path, recording the path length.

[0106] It should be noted that an endpoint voxel is defined as a voxel that has only one skeleton voxel in the target neighborhood.

[0107] From each endpoint voxel Starting from there, trace inward along the skeletal path, recording the voxel sequences encountered. The path continues until it encounters a fork (e.g., a skeleton voxel count ≥ 3 within a 26-neighborhood) or another endpoint. The physical length of this path... for:

[0108] Where m+1 is the total number of voxels on the path, and r is the voxel resolution.

[0109] S2.5, remove all voxels from the skeleton on paths whose path length is less than the preset pruning threshold.

[0110] like If the value is less than the preset pruning threshold, the short branch is identified as a noise spur, and all voxels along that path are removed from the skeleton. Steps S2.4 and S2.5 are repeated until no new short branches are deleted, to avoid missing shorter branches that were originally connected to a short branch after its deletion.

[0111] This embodiment removes noise spikes smaller than a preset pruning threshold, reduces false bifurcation points, improves the accuracy of image structure conversion, and retains real, small collateral vessels longer than the preset pruning threshold. This ensures that the characteristic collateral circulation of KTS is fully presented in the skeleton, guaranteeing the accuracy of subsequent vascular quantification parameters. Simultaneously, the pruning threshold is automatically calculated based on voxel resolution, making it applicable to scanning parameters of different CT devices.

[0112] This embodiment addresses the characteristic of the "tangled" high-density venous network in the lower extremities of KTS patients. Before refinement, morphological closing operation preprocessing is added to eliminate the interference of tiny gaps on the topology calculation. After refinement, a pruning threshold is set based on the disease specificity of the minimum physiological length of collateral vessels in KTS to perform pruning processing and improve topology accuracy.

[0113] In this embodiment, the blood vessel centerline skeleton is based on the single-pixel width after pruning. The "tangled" vascular skeleton of the lower extremity veins in KTS patients is transformed into a computer-computable, searchable, and quantifiable graph structure.

[0114] The graph structure G=(V,E) is a mathematical model used in discrete mathematics to represent network relationships. Here, V (nodes / vertices) represents key locations in the network, including vascular bifurcation points and vascular endpoints, while E (edges / arcs) represents vascular segments connecting two nodes, containing all skeletal points along the path. The lower extremity veins of KTS patients exhibit a "tangled" network structure, containing numerous bifurcation points and loops. For example, collateral convergence points and abnormal communicating branches constitute bifurcation points, while the communication between the ascending lumbar vein and the paravertebral venous plexus forms loops. This embodiment ensures the correctness of the extracted graph structure through node identification rules and loop processing.

[0115] Specifically, in this embodiment, step S3 converts the extracted vascular centerline skeleton into a graph structure G=(V,E), which specifically includes the following steps S3.1 to S3.5.

[0116] S3.1, traverse all voxels in the skeleton and count the number of adjacent skeleton points in the 26-neighborhood for each voxel. .

[0117] Traversing the skeleton All voxels (skeleton points) with a value of 1 are included. For each skeleton point p, the number of its neighboring skeleton points in its 26-neighborhood is counted. :

[0118] in, This represents the 26-neighborhood of voxel p (excluding p itself). .

[0119] S3.2, Identify nodes.

[0120] Will The voxels are labeled as endpoint nodes, and the voxels are marked as endpoint nodes. The voxels are marked as branch point nodes, and the voxels are used as branch point nodes. The voxels are labeled as path voxels.

[0121] Specifically, this step is based on The numerical values ​​are used to classify the skeleton points. The voxel markers are used as endpoint nodes to represent the blood vessel terminals, serving as nodes. The voxels are labeled as path voxels, representing points inside blood vessels; no separate nodes are created for them. The voxel markers are used as bifurcation point nodes, representing the points where blood vessels branch or merge, and are used as nodes.

[0122] Due to the high density of the KTS vascular network, voxels may appear at the bifurcation points. In even higher cases, this embodiment will They are all categorized as branching points, without distinguishing specific numerical values, in order to avoid excessive fragmentation.

[0123] In the dense, tangled network of veins in the lower extremities of patients with KTS (Kidney Tract Syndrome), multiple collateral branches may converge into the main trunk within extremely small spatial areas, resulting in multiple bifurcation voxels with very close spatial locations after skeletal extraction. For example, in areas where lateral veins communicate with multiple pelvic collateral branches, or where the ascending lumbar vein joins the paravertebral venous plexus, the spatial distance between multiple bifurcation points may be only 1-2 voxels. Treating all these spatially adjacent bifurcation points as independent nodes would lead to excessive fragmentation of the graph structure, distortion of the branch complexity index, difficulties in topological analysis, and cluttered visualization.

[0124] Based on this, this embodiment performs multi-point bifurcation merging. When the spatial distance between multiple bifurcation voxels is less than a preset merging threshold, these bifurcation points are merged into a composite node, whose coordinates are taken as the geometric center of the coordinates of each bifurcation point. For example, let the merging threshold be 2 voxels.

[0125] S3.3, construct the node set V.

[0126] Add all identified endpoint voxels and bifurcation voxels to the node set V. Each node records its node index, node type, 3D coordinates, and node degree.

[0127] The degree of a node is defined as the number of edges originating from that node, expressed as:

[0128] in This is an indicator function, taking the value 1 if an edge connecting v and u exists, and 0 otherwise. For endpoint nodes, deg(v) = 1; for branching nodes, deg(v) ≥ 3.

[0129] S3.4, trace all path voxels between adjacent nodes along the skeleton path and construct the edge set E.

[0130] Specifically, starting from each node, trace along the skeleton path in all directions; record all path voxels from the current node to the next node; stop tracing when another node is encountered and establish an edge; record each edge only once to avoid duplication.

[0131] Each edge records the starting node index, ending node index, path voxel list, and actual path length. and European distance .

[0132] in, , m is the number of voxels along the path, and r is the voxel resolution. , respectively, are the voxel coordinates of the starting and ending nodes of edge e in the 3D image space.

[0133] S3.5 uses depth-first search to detect cycles in the graph structure and marks the nodes and edges involved in the cycle as cycle nodes and cycle edges, and finally outputs the graph structure G=(V,E).

[0134] KTS patients have multiple communicating veins in their lower extremities forming loop structures, such as communicating veins between the ascending lumbar vein and the paravertebral venous plexus. In this embodiment, loops in the graph structure are detected by depth-first search (DFS), specifically including: performing DFS traversal on each unvisited node; recording the node sequence on the current traversal path; if a node already on the current path is encountered, a loop is detected; and marking the nodes and edges involved in the loop as loop nodes and loop edges.

[0135] Preferably, the graph structure is first simplified and preprocessed before loop detection, including deleting dangling dendrites and merging short edges. Deleting dangling dendrites involves traversing inwards from the endpoint node (the node with a degree of 1) along the path, deleting all dangling structures that do not form loops. These structures do not participate in loop formation and their deletion will not affect the detection results. Merging short edges involves merging short edges with a length less than a preset threshold, reducing redundant nodes and edges, and lowering the complexity of the graph structure.

[0136] This embodiment automatically calculates quantitative assessment indicators of lower extremity veins in KTS patients based on the node and edge attributes in the graph structure.

[0137] Regarding the tortuosity index, the lower extremity veins of KTS patients often exhibit abnormal tortuosity. For any vascular segment, corresponding to an edge e∈E in the graph structure, the tortuosity index is... Defined as the ratio of the actual path length of the blood vessel segment to the straight-line distance between the start and end points, it is expressed as:

[0138] in, This represents the actual curve length along the centerline of the blood vessel. This represents the three-dimensional Euclidean distance between the start and end points of the blood vessel segment.

[0139] This indicates that the blood vessels are completely straight and without twisting. This indicates that the blood vessels are tortuous; the higher the value, the more severe the tortuosity.

[0140] The lower extremity veins of patients with KTS contain numerous short collateral vessels. The calculation of the tortuosity index of these short vessels is easily affected by measurement errors. Therefore, a length-weighted tortuosity index is calculated and expressed as:

[0141] The weighted average tortuosity index of the entire image uses the length of the vessel segment as the weight to avoid the excessive influence of extreme values ​​of short vessels on the overall assessment.

[0142] Regarding the branch complexity index, the branch complexity index Defined as the number of branching points per unit volume in a graph structure, expressed as:

[0143] in, This represents the number of nodes at the branching point (the number of nodes with a degree greater than or equal to 3 in the graph structure). The total volume of the blood vessels is calculated from the binary volume data output in step S1.

[0144] Regarding the loop complexity index, loops are an important characteristic of KTS side branch cycles. The number, size, and tortuosity of loops can reflect the richness of side branch compensation. Loop complexity index Defined as the total number of loops in a graph structure, expressed as:

[0145] in, The total number of loops detected in the graph structure is output by the DFS algorithm in step S3.

[0146] Alternatively, it can be defined as the proportion of the total length of the loop to the total volume of the blood vessel, expressed as:

[0147] in Let c be the total path length of the loop, and let c be the length of all edges in the loop. the sum of The total vascular volume is represented by the loop complexity index, which quantifies the degree of collateral circulation compensation. A higher number of loops indicates richer collateral compensation and a greater number of abnormal vessels requiring treatment.

[0148] In this embodiment, step S5, based on a graph structure, generates a 3D visualization image, specifically including the following processes. For example... Figure 7 The image shown is a 3D visualization of the tortuosity. Figure 8 The image shown is a 3D visualization of the loop complexity index.

[0149] First, map the edges E in the graph structure to tubular models in 3D space. For each edge e, based on its path voxel list... A three-dimensional tubular mesh is generated with each voxel as the center and a preset radius. This tubular mesh is obtained by connecting a series of spheres or cylinders along the path points and reconstructing the surface. After merging all the tubular meshes, a complete three-dimensional blood vessel model is formed.

[0150] Secondly, visualization rendering is performed. The 3D blood vessel model is rendered using semi-transparent or solid materials, and lighting and viewpoint parameters are set to present a stereoscopic visual effect.

[0151] Next, a visualization mapping of the quantified parameters is performed. The tortuosity index of each vessel segment calculated in step S4 is then used. It is mapped onto the three-dimensional blood vessel model in a color-coded form.

[0152] Finally, the interactive display function is implemented. When a user clicks on any blood vessel segment, the system locates the corresponding edge 'e' in the graph structure, highlights the edge and its adjacent nodes, and displays quantitative parameters such as the tortuosity index and length of the blood vessel segment. After the user selects a target blood vessel segment, the system traverses the graph to find all nodes and edges connected to that blood vessel, hides the connected region from the rendering, or sets its transparency to 0 to achieve a perspective effect, allowing observation of the obscured deep veins or collateral vessels.

[0153] The foregoing has described in detail an embodiment of a lower extremity venography image processing method. Based on the lower extremity venography image processing method described in the above embodiment, this invention also provides a lower extremity venography image processing system corresponding to the method.

[0154] Figure 9This is a schematic block diagram of a lower extremity venography image processing system provided in an embodiment of the present invention. In this embodiment, the lower extremity venography image processing system performs specific functions. The module referred to in this invention is a series of computer program segments that can be executed by at least one processor and perform a fixed function, and is stored in memory.

[0155] The data acquisition and segmentation module is used to acquire three-dimensional image data of lower extremity venography of KTS patients, perform binarization segmentation on the three-dimensional image data, extract the initial three-dimensional binarized volume data of the veins, and perform three-dimensional morphological closing operation on the initial three-dimensional binarized volume data; wherein, the segmentation process includes extracting superficial veins using a first threshold range and extracting deep veins and collateral veins using a second threshold range.

[0156] The skeleton extraction and pruning module is used to perform a topology-preserving 3D thinning algorithm on the 3D binary volume data after the closing operation, extract the vessel centerline skeleton with a single pixel width, and perform a pruning operation on the vessel centerline skeleton based on a preset pruning threshold.

[0157] The graph structure conversion module is used to convert the extracted vascular centerline skeleton into a graph structure G=(V,E) and mark the loop nodes and loop edges in it, where V represents the set of nodes and E represents the set of edges connecting the nodes. The nodes include vascular bifurcation points and vascular endpoints.

[0158] The quantization parameter calculation module is used to calculate the quantization parameters of blood vessels based on the graph structure. The quantization parameters include tortuosity index, whole graph weighted average tortuosity index, branch complexity index, and loop complexity index.

[0159] A 3D visualization module is used to generate 3D visualization images based on the graph structure.

[0160] The lower extremity venography image processing system of this embodiment is used to implement the aforementioned lower extremity venography image processing method. Therefore, the specific implementation of this system can be found in the embodiment section of the lower extremity venography image processing method above. Thus, the specific implementation can be referred to the description of the corresponding embodiments, and will not be elaborated here.

[0161] Furthermore, since the lower extremity venography image processing system of this embodiment is used to implement the aforementioned lower extremity venography image processing method, its function corresponds to the function of the above method, and will not be described again here.

[0162] Figure 10This is a schematic diagram of a terminal 300 provided in an embodiment of the present invention, including: a processor 310, a memory 320, and a communication unit 330. The processor 310 is used to implement the flow steps of the above-described lower extremity venography image processing method embodiment when implementing the lower extremity venography image processing program stored in the memory 320.

[0163] This invention also provides a computer storage medium, which may be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc. The computer storage medium stores a lower extremity venography image processing program. When the lower extremity venography image processing program is executed by a processor, it implements the process steps of the above-described lower extremity venography image processing method embodiment.

[0164] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for processing lower extremity venography images, characterized in that, Includes the following steps: Three-dimensional image data of lower extremity venography from KTS patients were acquired. The three-dimensional image data was then binarized and segmented to extract initial three-dimensional binary volume data of the veins. A three-dimensional morphological closing operation was performed on the initial three-dimensional binary volume data. The segmentation process included extracting superficial veins using a first threshold range and extracting deep veins and collateral veins using a second threshold range. The lower limit of the first threshold range was higher than the upper limit of the second threshold range. The Otsu algorithm was used to calculate the globally optimal threshold for the three-dimensional image data. Using the globally optimal threshold Determine the basic threshold range centered on [the target]. and ,in, This is the amount of downward expansion from the low threshold. This represents the upward expansion amount from the high threshold. These are the preset minimum and maximum allowed thresholds; the first threshold range is set to... The second threshold range is ,in, , This is the amount of downward expansion at two points; A topology-preserving 3D thinning algorithm is performed on the 3D binary volume data after the closing operation to extract the vessel centerline skeleton with a single pixel width. The 3D thinning algorithm then performs pruning on this vessel centerline skeleton based on a preset pruning threshold. The algorithm includes: determining whether any voxels were deleted in the previous global iteration; if so, proceeding to the next step if deletion occurred in the first or previous iteration; otherwise, terminating the iteration and outputting all remaining foreground voxels as the vessel centerline skeleton with a single pixel width; sequentially executing sub-iterations in six directions, including east, west, south, north, up, and down; marking all neighboring voxels as background boundary voxels in the current sub-iteration direction; and for each marked boundary voxel p, applying the formula... Calculate the topological number, where Let be the number of connected components of the foreground voxels in the 26-neighborhood of voxel p under 26 connectivity. Let p be the number of connected components of the background voxels in the 6-neighborhood of voxel p under 6-connectivity; determine whether the boundary voxel simultaneously satisfies the following deleteable conditions: the number of foreground voxels in the 26-neighborhood of the voxel is greater than or equal to 2, the topology number is equal to 1, and no new holes are generated after deletion; delete the boundary voxel that satisfies all deleteable conditions from the foreground, that is, change the voxel value from 1 to 0. The extracted vascular centerline skeleton is converted into a graph structure G=(V,E), and the loop nodes and loop edges are marked, where V represents the set of nodes and E represents the set of edges connecting the nodes. The nodes include vascular bifurcation points and vascular endpoints. Based on the graph structure, vascular quantification parameters are calculated, including tortuosity index, whole graph weighted average tortuosity index, branch complexity index, and loop complexity index. Based on the graph structure, a three-dimensional visualization image is generated; The Otsu algorithm maximizes the inter-class variance. Determine the optimal threshold ,in Let T represent the probabilities of foreground and background voxels, respectively, at threshold T. These are the average gray levels of the foreground and background, respectively. The global grayscale mean; Specifically, the pruning operation of the vascular centerline skeleton is performed based on a preset pruning threshold, which includes: setting a preset pruning threshold based on the statistical characteristics of the lower extremity vein images of KTS patients; traversing all endpoint voxels in the skeleton, starting from each endpoint voxel and tracing inward along the skeleton path, and recording the path length; and deleting all voxels on the path whose path length is less than the preset pruning threshold from the skeleton. Specifically, the extracted vascular centerline skeleton is converted into a graph structure G=(V,E), and loop nodes and loop edges are labeled, including: Iterate through all voxels in the skeleton and count the number of adjacent skeleton points in the 26-neighborhood for each voxel. ; Identify nodes, The voxels are labeled as endpoint nodes, and the voxels are marked as endpoint nodes. The voxels are marked as branch point nodes, and the voxels are used as branch point nodes. The voxels are labeled as path voxels; when the spatial distance between multiple bifurcation voxels is less than the preset merging threshold, the multiple bifurcation points are merged into a composite node, whose coordinates are the geometric center of the coordinates of each bifurcation point. Construct a node set V, recording the node type, 3D coordinates, and degree of each node, where the node degree is... Defined as the number of edges originating from this node, with endpoint nodes having a degree of 1 and branching nodes having a degree greater than or equal to 3; Tracing all path voxels between adjacent nodes along the skeleton path, constructing an edge set E, where each edge records the starting node index, ending node index, path voxel list, and actual path length. and European distance ,in, , m is the number of voxels along the path, and r is the voxel resolution. These are the voxel coordinates of the starting and ending nodes of edge e in the 3D image space, respectively. A depth-first search is used to detect cycles in the graph structure, and the nodes and edges involved in the cycle are marked as cycle nodes and cycle edges. The final output graph structure is G=(V,E).

2. The lower extremity venography image processing method according to claim 1, characterized in that, Binarization segmentation is performed on the 3D image data to extract the 3D binary volumetric data of the veins, specifically including: The filter kernel size is calculated based on the local image noise level, and Gaussian filtering is performed on the 3D image data based on the filter kernel size; A region growing algorithm is used to automatically identify and label skeletal regions, and then exclude skeletal regions from the area to be segmented. A first threshold range and a second threshold range are preset. The lower limit of the first threshold range is higher than the upper limit of the second threshold range. The first threshold range is used to extract normally filled superficial veins, and the second threshold range is used to extract poorly visualized deep veins and collateral veins. The two segmentation results are merged to generate initial three-dimensional binary volume data. Perform a three-dimensional morphological closing operation on the initial three-dimensional binary volume data, first performing a dilation operation and then an erosion operation, to output three-dimensional binary volume data with connectivity preserved.

3. The lower extremity venography image processing method according to claim 1, characterized in that, The calculation of vascular quantification parameters specifically includes: For each edge e∈E in the graph structure, through Calculate the tortuosity index, where, This represents the actual curve length along the centerline of the blood vessel. The three-dimensional Euclidean distance between the start and end points of the blood vessel segment; pass Calculate the weighted average tortuosity index of the entire map; pass Calculate the branch complexity exponent, where, This represents the number of branching nodes. This refers to the total volume of the blood vessels. pass Calculate the number of loops index, and / or by Calculate the loop length density index, where The set of detected loops, This represents the total path length of the loop.

4. A lower extremity venography image processing system, characterized in that, include: The data acquisition and segmentation module is used to acquire three-dimensional image data of lower extremity venography in KTS patients, perform binarization segmentation on the three-dimensional image data, extract initial three-dimensional binary volume data of veins, and perform three-dimensional morphological closing operations on the initial three-dimensional binary volume data. The segmentation process includes extracting superficial veins using a first threshold range and extracting deep veins and collateral veins using a second threshold range; the lower limit of the first threshold range is higher than the upper limit of the second threshold range; and the Otsu algorithm is used to calculate the globally optimal threshold for the three-dimensional image data. Using the globally optimal threshold Determine the basic threshold range centered on [the target]. and ,in, This is the amount of downward expansion from the low threshold. This represents the upward expansion amount from the high threshold. These are the preset minimum and maximum allowed thresholds; the first threshold range is set to... The second threshold range is ,in , This is the amount of downward expansion at two points; The skeleton extraction and pruning module is used to perform a topology-preserving 3D thinning algorithm on the 3D binary volume data after the closing operation, extract the vessel centerline skeleton with a single pixel width, and perform a pruning operation on the vessel centerline skeleton based on a preset pruning threshold. The 3D thinning algorithm includes: determining whether any voxels were deleted in the previous global iteration; if so, proceeding to the next step if deletion occurred in the first iteration or the previous iteration; otherwise, terminating the iteration and outputting all remaining foreground voxels as the vessel centerline skeleton with a single pixel width; sequentially executing sub-iterations in six directions, including east, west, south, north, up, and down; marking all neighboring voxels as background boundary voxels in the current sub-iteration direction; and for each marked boundary voxel p, applying the formula... Calculate the topological number, where Let be the number of connected components of the foreground voxels in the 26-neighborhood of voxel p under 26 connectivity. Let p be the number of connected components of the background voxels in the 6-neighborhood of voxel p under 6-connectivity; determine whether the boundary voxel simultaneously satisfies the following deleteable conditions: the number of foreground voxels in the 26-neighborhood of the voxel is greater than or equal to 2, the topology number is equal to 1, and no new holes are generated after deletion; delete the boundary voxel that satisfies all deleteable conditions from the foreground, that is, change the voxel value from 1 to 0. The graph structure conversion module is used to convert the extracted vascular centerline skeleton into a graph structure G=(V,E) and mark the loop nodes and loop edges in it, where V represents the set of nodes and E represents the set of edges connecting the nodes. The nodes include vascular bifurcation points and vascular endpoints. The quantification parameter calculation module is used to calculate the quantification parameters of blood vessels based on the graph structure. The quantification parameters include tortuosity index, whole graph weighted average tortuosity index, branch complexity index, and loop complexity index. A 3D visualization module is used to generate a 3D visualization image based on the graph structure; The Otsu algorithm maximizes the inter-class variance. Determine the optimal threshold ,in Let T represent the probabilities of foreground and background voxels, respectively, at threshold T. These are the average gray levels of the foreground and background, respectively. The global grayscale mean; Specifically, the pruning operation of the vascular centerline skeleton is performed based on a preset pruning threshold, which includes: setting a preset pruning threshold based on the statistical characteristics of the lower extremity vein images of KTS patients; traversing all endpoint voxels in the skeleton, starting from each endpoint voxel and tracing inward along the skeleton path, and recording the path length; and deleting all voxels on the path whose path length is less than the preset pruning threshold from the skeleton. Specifically, the extracted vascular centerline skeleton is converted into a graph structure G=(V,E), and loop nodes and loop edges are labeled, including: Iterate through all voxels in the skeleton and count the number of adjacent skeleton points in the 26-neighborhood for each voxel. ; Identify nodes, The voxels are labeled as endpoint nodes, and the voxels are marked as endpoint nodes. The voxels are marked as branch point nodes, and the voxels are used as branch point nodes. The voxels are labeled as path voxels; when the spatial distance between multiple bifurcation voxels is less than the preset merging threshold, the multiple bifurcation points are merged into a composite node, whose coordinates are the geometric center of the coordinates of each bifurcation point. Construct a node set V, recording the node type, 3D coordinates, and degree of each node, where the node degree is... Defined as the number of edges originating from this node, with endpoint nodes having a degree of 1 and branching nodes having a degree greater than or equal to 3; Tracing all path voxels between adjacent nodes along the skeleton path, constructing an edge set E, where each edge records the starting node index, ending node index, path voxel list, and actual path length. and European distance ,in, , m is the number of voxels along the path, and r is the voxel resolution. These are the voxel coordinates of the starting and ending nodes of edge e in the 3D image space, respectively. A depth-first search is used to detect cycles in the graph structure, and the nodes and edges involved in the cycle are marked as cycle nodes and cycle edges. The final output graph structure is G=(V,E).

5. A terminal, characterized in that, include: The memory is used to store the lower extremity venography image processing program; A processor is configured to implement the steps of the lower extremity venography image processing method as described in any one of claims 1 to 3 when executing the lower extremity venography image processing program.

6. A computer-readable storage medium, characterized in that, The readable storage medium stores a lower extremity venography image processing program, which, when executed by a processor, implements the steps of the lower extremity venography image processing method as described in any one of claims 1 to 3.

Images