Blood vessel recognition method, system and device based on image features and medium

Abstract

The invention discloses a blood vessel recognition method, system and device based on image features and a medium, and relates to the technical field of data transmission processing, and the method comprises the steps: obtaining a target image, and dividing the target image into a plurality of sub-pixel regions; obtaining a gray gradient amplitude of each sub-pixel region, and obtaining a dynamic extension threshold of two adjacent sub-pixel regions; acquiring gray variation of two adjacent sub-pixel areas, and connecting the two adjacent sub-pixel areas with the gray variation smaller than a dynamic extension threshold to form a to-be-extended section; obtaining a preset angle threshold value, obtaining the number of the to-be-extended sections formed by each sub-pixel region, and obtaining a target angle threshold value according to the number of the to-be-extended sections formed by each sub-pixel region and the preset angle threshold value; if the included angle between the adjacent to-be-extended sections is larger than or equal to the target angle threshold value, the adjacent to-be-extended sections are connected, and a to-be-blood-vessel section is formed. The method has the advantages of image complexity perception, bifurcation and extension recognition and adaptive threshold adjustment.

Description

Technical Field

[0001] This invention relates to the field of data transmission processing technology, and specifically to a method, system, device, and medium for blood vessel recognition based on image features. Background Technology

[0002] Traditional vascular identification methods have many shortcomings when processing medical images (such as CTA and MRA) due to the topological complexity of vascular structures and the variability of image grayscale features.

[0003] Existing vascular identification methods, which employ fixed angle thresholds and static extension thresholds, cannot adapt to the dynamic characteristics of vascular structures in medical images. For example, in bifurcation regions (such as coronary artery bifurcation points or cerebral vascular rings), fixed angle thresholds (such as 135°) can cause acute-angle branches (commonly 30° to 80°) to break, resulting in missed diagnoses of bifurcation lesions. In calcified or low-contrast areas, static extension thresholds can easily confuse real vascular boundaries with artifacts, leading to misdiagnosis of vessels or misconnection of non-vascular tissues. Furthermore, due to the rigidity of global parameters, it is impossible to simultaneously meet the contradictory requirements of straight vessel segments (which require large angles to suppress misconnection) and bifurcation regions (which require small angles to avoid breakage). This results in a high rate of peripheral vessel breakage and a proliferation of pseudo-branches in bifurcation regions, ultimately destroying the three-dimensional topological integrity of vessels and severely affecting the accuracy of hemodynamic simulation and the reliability of clinical surgical planning. Summary of the Invention

[0004] To address the problems in existing technologies, such as the lack of causal relationships between features and dynamic parameters, as well as the lack of extended recognition capabilities including image complexity perception, bifurcation structure identification, and inverse adaptation of angle thresholds, this invention provides a method, system, device, and medium for blood vessel recognition based on image features.

[0005] A blood vessel recognition method based on image features includes: acquiring a target image of a target region and dividing the target image into multiple sub-pixel regions according to a preset window size; obtaining the gray-level gradient magnitude of each sub-pixel region based on the gray-level values ​​of each pixel in each sub-pixel region, and obtaining a dynamic extension threshold of two adjacent sub-pixel regions based on the gray-level gradient magnitudes of two adjacent sub-pixel regions; obtaining the gray-level change amount of two adjacent sub-pixel regions, and connecting two adjacent sub-pixel regions whose gray-level change amount is less than the dynamic extension threshold to form a proposed extension segment; obtaining a preset angle threshold, and obtaining the number of sub-pixel regions participating in the formation of the proposed extension segment, and obtaining a target angle threshold of each sub-pixel region based on the number of sub-pixel regions participating in the formation of the proposed extension segment and the preset angle threshold; if the included angle between adjacent proposed extension segments is greater than or equal to the target angle threshold of the overlapping sub-pixel region, then connecting the adjacent proposed extension segments, and forming a proposed blood vessel segment from multiple consecutive connected proposed extension segments.

[0006] Optionally, obtaining the grayscale change of adjacent sub-pixel regions includes: obtaining the average grayscale value of each sub-pixel region based on the grayscale value of each pixel in each sub-pixel region; obtaining the difference between the average grayscale values ​​of adjacent sub-pixel regions and using it as the grayscale change.

[0007] Optionally, obtaining the number of sub-pixel regions participating in the formation of the proposed extension segment includes: obtaining multiple grayscale changes between the i-th sub-pixel region and multiple sub-pixel regions adjacent to the i-th sub-pixel region and using them as multiple unprocessed changes of the i-th sub-pixel region; obtaining the number of multiple unprocessed changes of the i-th sub-pixel region that are less than the dynamic extension threshold and using them as the number of the i-th sub-pixel regions participating in the formation of the proposed extension segment.

[0008] Optionally, obtaining the grayscale gradient magnitude of each sub-pixel region based on the grayscale value of each pixel within each sub-pixel region includes: calculating the grayscale change between each pixel and all its neighboring pixels within each sub-pixel region, and taking the absolute value of the largest grayscale change as the gradient intensity of that pixel; and summing the gradient intensities of all pixels within each sub-pixel region to obtain the grayscale gradient magnitude of each sub-pixel region.

[0009] Optionally, obtaining the dynamic extension threshold of two adjacent sub-pixel regions based on the gray-level gradient magnitude of two adjacent sub-pixel regions includes: obtaining the absolute difference between the gray-level gradient magnitudes of two adjacent sub-pixel regions, obtaining the larger value between the gray-level gradient magnitudes of two adjacent sub-pixel regions, and obtaining the correction parameters of the two adjacent sub-pixel regions based on the absolute difference and the larger value; obtaining the standard extension threshold of two adjacent sub-pixel regions, and obtaining the dynamic extension threshold of the two adjacent sub-pixel regions based on the standard extension threshold and the correction parameters of the two adjacent sub-pixel regions.

[0010] Optionally, obtaining the target angle threshold for each sub-pixel region based on the number of sub-pixel regions participating in the formation of the proposed extension segment and the preset angle threshold includes: obtaining the attenuation angle value mapped by the maximum adjustment number and the unit number; obtaining the target angle threshold for each sub-pixel region based on the number of sub-pixel regions participating in the formation of the proposed extension segment, the preset angle threshold, the maximum adjustment number, and the attenuation angle value mapped by the unit number.

[0011] A blood vessel recognition system based on image features is also provided. The system includes: a data acquisition module for acquiring a target image of a target region and dividing the target image into multiple sub-pixel regions according to a preset window size; a first data processing module for obtaining the gray-level gradient amplitude of each sub-pixel region based on the gray-level values ​​of each pixel in each sub-pixel region, and obtaining a dynamic extension threshold of two adjacent sub-pixel regions based on the gray-level gradient amplitude of two adjacent sub-pixel regions; a second data processing module for obtaining the gray-level change of two adjacent sub-pixel regions, connecting two adjacent sub-pixel regions whose gray-level change is less than the dynamic extension threshold to form a pseudo-extension segment; a third data processing module for obtaining a preset angle threshold, obtaining the number of each sub-pixel region participating in the formation of the pseudo-extension segment, and obtaining a target angle threshold of each sub-pixel region based on the number of each sub-pixel region participating in the formation of the pseudo-extension segment and the preset angle threshold; and a recognition module for connecting adjacent pseudo-extension segments if the included angle between adjacent pseudo-extension segments is greater than or equal to the target angle threshold of the overlapping sub-pixel region, and forming a pseudo-blood vessel segment from multiple consecutive connected pseudo-extension segments.

[0012] Optionally, the second data processing module is further configured to: obtain the average gray value of each sub-pixel region based on the gray value of each pixel in each sub-pixel region; and obtain the difference between the average gray values ​​of adjacent sub-pixel regions and use it as the gray value change.

[0013] An electronic device is also provided, comprising: a memory storing a computer program thereon; and a processor for executing the computer program in the memory to implement the above-described image feature-based blood vessel recognition method.

[0014] A non-transitory computer-readable storage medium is also provided, on which a computer program is stored, which, when executed by a processor, implements the above-described image feature-based blood vessel recognition method.

[0015] The beneficial effects of this invention are reflected in: In the entire image feature-based vessel identification method, the accuracy and robustness of vessel identification are improved by establishing the correlation between image features and dynamic parameters. First, the dynamic extension threshold generated based on the gray-scale gradient amplitude of the sub-region can adaptively distinguish between calcification artifacts and real vessel boundaries. Furthermore, by adjusting the angle threshold inversely by the number of connections, the contradiction between the breakage of acute-angle branches at bifurcation points (such as the preservation of coronary artery branches) and the proliferation of pseudo-branches in straight segments is simultaneously resolved, ultimately achieving the preservation of the integrity of the vessel topology. Attached Figure Description

[0016] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0017] Figure 1 This is a schematic diagram of a portion of the blood vessel recognition method based on image features of the present invention; Figure 2 This is a schematic diagram of another part of the blood vessel recognition method based on image features of the present invention; Figure 3 This is a schematic diagram illustrating the steps of the image feature-based blood vessel recognition method of the present invention; Figure 4 This is a schematic diagram of a portion of step S3 in the image feature-based blood vessel recognition method of the present invention; Figure 5 This is a schematic diagram of a portion of step S4 in the image feature-based blood vessel recognition method of the present invention; Figure 6 This is a schematic diagram of a portion of step S2 in the image feature-based blood vessel recognition method of the present invention; Figure 7 This is a schematic diagram of another part of step S2 in the image feature-based blood vessel recognition method of the present invention; Figure 8 This is a schematic diagram of another part of step S4 in the image feature-based blood vessel recognition method of the present invention; Figure 9 This is a block diagram illustrating an electronic device according to an embodiment of the present invention.

[0018] Figure label: 700 - Electronic device; 701 - Processor; 702 - Memory; 703 - Multimedia component; 704 - I / O interface; 705 - Communication component. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0020] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0021] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, the terms "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0022] like Figure 1 , Figure 2 and Figure 3 As shown, a blood vessel recognition method based on image features is provided. In one embodiment, the method includes: S1. Obtain the target image of the target area and divide the target image into multiple sub-pixel areas according to the preset window size; S2. Obtain the grayscale gradient magnitude of each sub-pixel region based on the grayscale value of each pixel in each sub-pixel region, and obtain the dynamic extension threshold of two adjacent sub-pixel regions based on the grayscale gradient magnitude of two adjacent sub-pixel regions. S3. Obtain the grayscale change of two adjacent sub-pixel regions, connect two adjacent sub-pixel regions whose grayscale change is less than the dynamic extension threshold, and form a proposed extension segment. S4. Obtain a preset angle threshold and the number of sub-pixel regions participating in the formation of the proposed extension segment. Obtain the target angle threshold of each sub-pixel region based on the number of sub-pixel regions participating in the formation of the proposed extension segment and the preset angle threshold. S5. If the included angle between adjacent proposed extension segments is greater than or equal to the target angle threshold of the sub-pixel region at the overlap, then the adjacent proposed extension segments are connected together, and multiple consecutive connected proposed extension segments are combined into a proposed blood vessel segment.

[0023] In this embodiment, it should be noted that in S1, vascular image data of the target region is first acquired. This target image typically originates from medical imaging techniques (such as CTA or MRA scans in the prior art), covering the entire anatomical structure region of interest, such as the coronary arteries, cerebral vascular networks, or other densely vascularized areas. The image data is presented in the form of a two-dimensional pixel matrix, with each pixel storing grayscale value information. Next, by applying a preset window size (e.g., a fixed-size square grid), the entire target image is divided into multiple continuous small sub-pixel regions. This partitioning strategy aims to decompose the complex global image into locally manageable units to adapt to the spatial heterogeneity of vascular structures and avoid insufficient flexibility caused by uniform parameters during overall processing. It lays the foundation for subsequent steps (such as local gradient analysis and dynamic threshold calculation), ensuring that the method can capture dynamic changes in vascular morphology within small regions, thereby improving sensitivity to complex regions such as branching or bifurcation.

[0024] Furthermore, assuming the target image is a cerebral vascular MRA scan instance, the preset window size is defined as a uniform square region (e.g., covering a specific number of pixels in width and height), and its size is set based on the average width of the blood vessels and the image resolution. For example, in coronary artery image processing, the entire image is divided into multiple non-overlapping or partially overlapping sub-pixel regions, each representing an independent local unit, similar to a checkerboard layout. After division, each sub-pixel region contains adjacent sub-pixel regions, and its boundary corresponds to the sliding position of the window on the image, ensuring that all pixels are covered without omission. This division method enables the method to identify local details (such as small blood vessels or calcified areas at branches) and adaptively handle low-contrast variations in blood vessels through local parameter calculations (such as gray-level gradients within sub-regions). Through this operation, the original image is transformed into a set of operable sub-regions, providing structured input for subsequent dynamic thresholding and angle adjustments.

[0025] In step S2, the structural complexity of each sub-pixel region needs to be quantified first: based on the sub-regions defined in step S1, each pixel within the region is traversed, and the grayscale difference between it and its neighboring pixels (e.g., in the top, bottom, left, right, or diagonal directions) is calculated. By extracting the maximum local grayscale abrupt change value (i.e., gradient intensity) of each pixel, the edge strength or the degree of structural change at that location is reflected—for example, a sudden change in grayscale at the boundary of a blood vessel will produce a high intensity value, while calcified areas may exhibit abnormally high values ​​due to grayscale chaos, and low-contrast blood vessels will show weak values. Subsequently, the gradient intensities of all pixels within the sub-region are summed to form the grayscale gradient amplitude. This value serves as a key feature: a higher amplitude indicates a more complex structure within the region (e.g., bifurcation points or blood vessel-calcification junctions), while a lower amplitude may correspond to homogeneous tissue or low-contrast blood vessel segments, providing a basis for subsequent dynamic threshold adjustment.

[0026] Furthermore, a dynamic extension threshold is generated by comparing the gradient amplitude relationships of adjacent regions: the gray-level gradient amplitudes of two adjacent sub-regions calculated in the previous step are used as inputs to analyze the degree of difference between them. If the amplitude difference between the two regions is significant (e.g., one is a high-value calcification area and the other is a low-value vascular segment), it indicates that there is tissue discontinuity at the boundary, and the extension threshold needs to be tightened to avoid false connections of artifacts; if the amplitudes are similar (e.g., both are low-contrast vascular segments), the threshold is relaxed to ensure vascular continuity. In practice, the preset standard extension threshold (based on empirical values ​​of image resolution and window size) is corrected by using the ratio of the absolute difference in amplitude between adjacent regions to the larger amplitude, thus generating a dynamic extension threshold. For example, when processing cerebral vascular rings, if two adjacent sub-regions at the bifurcation point simultaneously present medium gradient amplitudes (possibly corresponding to a smooth transition of branch vessels), a lower dynamic extension threshold will be generated, allowing sub-regions with small gray-level differences to connect; conversely, near calcification areas, large amplitude differences will trigger a higher dynamic threshold, strictly blocking erroneous fusion with calcified plaques, thereby solving the problem that traditional static thresholds cannot distinguish between real vascular boundaries and interfering structures.

[0027] In S3, the connectivity of adjacent sub-regions is determined by a dynamic extension threshold. First, based on the dynamic extension threshold generated in S2, the grayscale change between adjacent sub-pixel regions is calculated: for each sub-region, the average grayscale of all pixels within it is calculated, and then the absolute difference between the average grayscale values ​​of adjacent sub-regions is calculated as the grayscale change. This design smooths individual pixel fluctuations (such as noise interference) through the regional mean, focusing on macroscopic grayscale trend differences—for example, in low-contrast vascular segments, the mean difference between adjacent sub-regions is slight; while at the boundary between blood vessels and calcified plaques, the mean difference is significantly amplified due to abrupt changes in tissue density. Subsequently, the calculated grayscale change is compared with the dynamic extension threshold: if the change is less than the threshold, it indicates that the grayscale transition between the two regions is smooth and the tissue continuity is high, and the two are determined to belong to the same vascular structure and connected to form a proposed extension segment (i.e., preliminary vascular segmentation); otherwise, the connection is broken to avoid erroneous fusion of non-vascular tissues.

[0028] Furthermore, in calcified regions, due to the large differences in gradient amplitude between adjacent sub-regions (e.g., a significant difference in grayscale gradient amplitude between calcified plaques and blood vessels), the dynamic extension threshold is lowered. In this case, even if the absolute value of the grayscale change is small (e.g., subtle density fluctuations), the lower dynamic extension threshold may still exceed the threshold and be judged as a break, thus preventing erroneous connections between calcification artifacts and blood vessels. Conversely, in low-contrast vascular regions (e.g., small cerebral venules), the gradient amplitudes of adjacent sub-regions are similar and relatively low, and the dynamic threshold adaptively increases (e.g., decreasing to a fraction of the standard threshold), allowing larger grayscale changes to be judged through connection. For example, when processing fuzzy branches of cerebral vascular rings, although there are slight grayscale differences between adjacent sub-regions (possibly due to partial volume effects or noise), the relaxed dynamic threshold allows them to be connected as a proto-extended segment, effectively avoiding the vascular breakage problem caused by fixed thresholds in traditional methods, and providing a complete local fragment basis for subsequent topology reconstruction.

[0029] In S4, the bifurcation structure of blood vessels is identified through local connectivity features, and the angle threshold is dynamically adjusted. First, based on the proposed extension segment results formed in S3, the number of neighboring regions participating in the connection is counted for each sub-pixel region. Specifically, for a given sub-region, it is checked whether the grayscale change between it and all directly adjacent sub-regions is less than the corresponding dynamic extension threshold: if the condition is met, it is counted as a valid connection. The number of connections successfully established by the sub-region is accumulated (e.g., if a sub-region may be connected to the upper and left sub-regions at the same time, the number is 2). This value directly reflects the complexity of the local topology—for example, in the core area of ​​blood vessel bifurcation (such as the coronary artery bifurcation), a single sub-region is often connected to 3-4 directions at the same time, indicating that there are multiple branches converging here; while in straight blood vessel segments, sub-regions are usually only connected to 2 directions (anterior and posterior extension).

[0030] Furthermore, the angle threshold is adapted inversely based on the number of connections: statistical values ​​are combined with a preset threshold (empirically set, such as 135°), and a reverse adjustment logic is implemented where the more connections there are, the more strictly the angle threshold decreases. For example, when the number of connections in a sub-region increases significantly (e.g., more than 2), the location is automatically identified as a bifurcation node (e.g., a bifurcation in a cerebral vascular ring), and the angle threshold is significantly reduced (e.g., from 135° to 45°, i.e., from oblique upward to oblique downward), allowing smaller acute-angle connections between adjacent intended extension segments, thus avoiding the erroneous truncation of acute-angle branches; conversely, when only 2 connections are detected in a straight vessel segment, a larger angle threshold is maintained (e.g., 135° is retained), forcing the extension segments to be arranged in a near-straight line, suppressing the generation of false branches (e.g., calcification artifacts or noise interference). This adaptive threshold adjustment mechanism, for the first time, achieves a causal relationship between bifurcation structural features and angle determination thresholds, and to some extent solves the problems of numerous breaks in bifurcation areas and false connections in straight vessel areas in existing methods.

[0031] In S5, local proximate extension segments are integrated using dynamic angle thresholds to construct a complete vascular topology. Based on the proximate extension segments (i.e., local vascular fragments) generated in the previous steps and the dynamically adjusted target angle threshold, the spatial angle formed by spatially adjacent proximate extension segments in overlapping sub-pixel regions (i.e., vascular bifurcation or confluence points) is first detected. This angle reflects the degree of geometric deviation between branches; for example, straight vascular segments have an angle close to 180°, while acute bifurcation can be as small as 30°-50°. The key decision lies in comparing the angle with the target angle threshold at the overlapping position (dynamically set in step S4 based on the number of connections): if the measured angle is ≥ the target angle threshold, the two segments are determined to have anatomical continuity and a connection operation is performed; otherwise, they remain disconnected to prevent abnormal branches from being mistakenly connected. This mechanism precisely balances the contradictory needs of straight vascular regions and bifurcation regions—maintaining a large angle threshold (e.g., 135°) in straight vascular segments ensures that only near-parallel extension segments can be connected, eliminating pseudo-branch interference; while a small angle threshold (e.g., 45°) is used in bifurcation regions, allowing acute-angled branches to merge naturally.

[0032] Furthermore, the dynamic nature of the target angle threshold is crucial in this step: when dealing with coronary artery bifurcation lesions, if the number of connections in the sub-pixel region of the bifurcation core area reaches four (indicating a complex branch), the target angle threshold drops to an extremely low level (e.g., 40°). At this point, even if adjacent proto-extended segments form a 45° acute angle (far exceeding the traditional 135° limit), their validity can still be identified and the connection completed, avoiding the risk of missed diagnoses due to acute-angle branch breakage. Conversely, in straight vessel regions such as the aorta, the target angle threshold remains high, forcing adjacent extended segments to form a gentle angle close to 180° (excluding abnormal 90° angle connections caused by calcification artifacts). Ultimately, multiple relatively complete pseudo-vascular segments are formed, maintaining the physiological bifurcation characteristics of the vessel and providing a reliable anatomical basis for hemodynamic simulation.

[0033] In summary, the accuracy and robustness of vessel identification are improved by establishing a correlation between image features and dynamic parameters in the entire image feature-based vessel identification method. First, the dynamic extension threshold generated based on the gray-scale gradient amplitude of the sub-region can adaptively distinguish between calcification artifacts and real vessel boundaries. Furthermore, by adjusting the angle threshold inversely by the number of connections, the contradiction between the breakage of acute-angle branches at bifurcation points (such as the preservation of coronary artery branches) and the proliferation of pseudo-branches in straight segments is simultaneously resolved, ultimately achieving the preservation of the integrity of the vessel topology.

[0034] like Figure 4 As shown, in one embodiment, obtaining the grayscale change of adjacent sub-pixel regions in S3 includes: S31. Obtain the average gray value of each sub-pixel region based on the gray value of each pixel in each sub-pixel region. S32. Obtain the difference in the average grayscale values ​​of adjacent sub-pixel regions and use it as the grayscale change.

[0035] In this embodiment, it should be noted that in S31, the overall grayscale features of the sub-pixel region are extracted. For each sub-pixel region generated by S1 (such as a 3×3 pixel square in a cerebral vascular image), the grayscale values ​​of all pixels within the region are traversed, and the arithmetic mean of the region is calculated to obtain the grayscale mean of the region. This operation suppresses local noise interference through a statistical smoothing mechanism—for example, within calcified regions, although individual pixels may exhibit extremely high grayscale values ​​due to calcium salt deposition, the mean calculation integrates surrounding transitional pixels, preventing individual outliers from dominating the region's features; while in low-contrast small vascular regions (such as cortical perforating arteries), the regional mean can stably reflect the weak density difference between the vascular lumen and surrounding tissues. This design aggregates discrete pixels into a regional grayscale representation, providing a noise-resistant basis for subsequent continuous analysis.

[0036] In S32, based on the results of S31, tissue continuity is quantified by the difference in the gray-scale mean values ​​of adjacent sub-regions. Using two spatially adjacent sub-pixel regions (such as horizontally or vertically adjacent windows) as units, the absolute difference in their gray-scale mean values ​​is calculated as the gray-scale change. This difference reflects the transitional characteristics of boundary tissues: for example, in the extension of a straight vessel, adjacent sub-regions have similar gray-scale mean values ​​because they belong to the same vessel lumen, and the difference approaches zero; while in the vessel-calcification interface, the mean difference significantly increases due to the sudden increase in the density of calcified plaques (e.g., calcified area mean 150 HU vs. vessel area mean 30 HU). Notably, the adjacency relationship covers omnidirectional connectivity (including the diagonal direction), capturing obliquely extending vascular branches at coronary artery bifurcations. This change eliminates the complexity of pixel-by-pixel comparison in traditional methods, efficiently identifying continuity anomalies at the regional scale, and synergizes with the dynamic extension threshold generated in S2—the latter adaptively adjusting based on local structural complexity—to jointly constitute a dual determination mechanism for vascular connectivity.

[0037] like Figure 5 As shown, in one embodiment, obtaining the number of each sub-pixel region participating in the formation of the proposed extension segment in S4 includes: S41. Obtain multiple grayscale changes between the i-th sub-pixel region and multiple sub-pixel regions adjacent to the i-th sub-pixel region, and use them as multiple changes to be processed in the i-th sub-pixel region. S42. Obtain the number of multiple unprocessed changes in the i-th sub-pixel region that are less than the dynamic extension threshold, and use them as the number of i-th sub-pixel regions participating in the formation of the proposed extension segment.

[0038] In this embodiment, it should be noted that in S41, for the i-th sub-pixel region, all its adjacent sub-regions (covering 8 neighboring directions) are traversed, and the grayscale change between them is calculated one by one. This value adopts the method of S31-S32, that is, the absolute difference of the grayscale mean of adjacent regions. For example, when processing the cerebral vascular ring, the grayscale change of a sub-pixel block in a bifurcation core region needs to be calculated with its 8 adjacent blocks in the east, south, west, north and four diagonal directions, forming a set of changes to be processed. These differences quantify the grayscale transition characteristics between the current region and the surrounding environment: in continuous vascular regions (such as straight vascular segments), the changes in each direction are close to zero; while in complex bifurcation regions (such as the carotid sinus), the differences in different directions show obvious differentiation (small difference in the main trunk direction, medium difference in the direction of new branches).

[0039] In S42, valid connections are filtered using a dynamic threshold. The set of unprocessed changes generated in the previous step and the corresponding dynamic extension threshold calculated in S2 are used as inputs to count the number of changes below the dynamic threshold. For example, in a vascular segment surrounded by calcification artifacts, although there are eight directions of unprocessed changes, only one or two vascular extension directions meet the criteria because the threshold for the calcification area is increased (e.g., 3 times the standard value). Conversely, at low-contrast bifurcation points (e.g., branches of the basilar artery), changes in multiple directions are included in the statistics because the threshold is reduced (e.g., 0.5 times the standard value), reaching 3-4 in number. This number represents the number of segments participating in the formation of the proposed extension, essentially revealing the local topological structure—a value of 1-2 represents straight vessels or terminals, while ≥3 marks bifurcation nodes (e.g., 3 for coronary artery trifles, and up to 4 for cerebral vascular crosses).

[0040] like Figure 6 As shown, in one embodiment, S2, obtaining the grayscale gradient magnitude of each sub-pixel region based on the grayscale value of each pixel within each sub-pixel region includes: S21. For each pixel in each sub-pixel region, calculate the gray level change between it and all its neighboring pixels, and take the absolute value of the largest gray level change as the gradient intensity of that pixel. S22. Accumulate the gradient intensity of all pixels in each sub-pixel region and obtain the grayscale gradient magnitude of each sub-pixel region.

[0041] In this embodiment, it should be noted that in S21, the abrupt change features of the local structure are captured by calculating the absolute value of the maximum grayscale change between each pixel and its neighboring pixels within the sub-pixel region. For example, at the edge of a blood vessel, the grayscale difference between the pixel and the surrounding tissue is drastic (such as the junction of the blood vessel lumen and calcification), resulting in a high gradient intensity value; while in homogeneous regions (such as the blood flow region in the center of a blood vessel), the gradient intensity is low.

[0042] In S22, the gradient intensities of all pixels within a sub-region are summed (not averaged) to form a region-level complex quantification index. In high-amplitude regions, the cumulative value increases significantly at bifurcation points due to the superposition of edges in multiple directions (such as the intersection of three vessel walls at a coronary artery bifurcation) or chaotic abrupt changes in grayscale within calcified plaques; in low-amplitude regions, only the lateral edges of straight vessel segments contribute, resulting in a lower cumulative value. This amplitude serves as the physical basis for subsequent dynamic thresholding, addressing the problem of insufficient sensitivity to local structures in traditional methods.

[0043] like Figure 7 As shown, in one embodiment, obtaining the dynamic extension threshold of two adjacent sub-pixel regions based on the grayscale gradient magnitude of two adjacent sub-pixel regions in S2 includes: S23. Obtain the absolute difference between the gray-level gradient magnitudes of two adjacent sub-pixel regions, and obtain the larger value between the gray-level gradient magnitudes of two adjacent sub-pixel regions. Obtain the correction parameters of the two adjacent sub-pixel regions based on the absolute difference and the larger value. S24. Obtain the standard extension threshold of two adjacent sub-pixel regions, and obtain the dynamic extension threshold of two adjacent sub-pixel regions based on the standard extension threshold and the correction parameters of the two adjacent sub-pixel regions.

[0044] In this embodiment, it should be noted that in S23, the continuity of the organization is quantified based on the amplitude difference between adjacent regions. First, the absolute difference between the gray-level gradient amplitudes of adjacent sub-regions is calculated. A large difference indicates that there is a structural fault at the boundary. Second, the larger value of the amplitudes of the two regions is taken as the benchmark for standardization to eliminate absolute intensity interference. Finally, a correction factor is generated through the ratio between the difference and the larger value. The larger the ratio (significant difference), the higher the output value, driving the subsequent threshold tightening.

[0045] In S24, the standard extension threshold (related to image resolution and window size) is used as a benchmark, and the correction factor output from S23 is used to finally obtain the dynamic extension threshold. The standard extension threshold is directly affected by the spatial resolution of the medical image; high-resolution images (slice thickness ≤ 0.625 mm) can detect subtle grayscale differences, so the standard threshold is lower (e.g., 30-50 HU); low-resolution images (slice thickness > 1 mm) need to tolerate greater noise, so the standard threshold is higher (e.g., 60-80 HU). Simultaneously, the standard extension threshold is also directly affected by the preset window size; small windows (3×3 pixels) are suitable for capillaries, where the grayscale mean is easily affected by noise, requiring a higher standard threshold (+20%); large windows (7×7 pixels) are suitable for the aorta, where the mean is more stable, allowing for a lower standard threshold (-20%).

[0046] It should also be noted that the dynamic extension threshold of two adjacent sub-pixel regions obtained in S2 based on the grayscale gradient magnitude of two adjacent sub-pixel regions is expressed as follows: ;in, The dynamic extension threshold for two adjacent sub-pixel regions. For standard extended threshold, This represents the grayscale gradient magnitude of the first sub-pixel region in two adjacent sub-pixel regions. It represents the grayscale gradient magnitude of the second sub-pixel region among two adjacent sub-pixel regions.

[0047] It should also be noted that throughout the entire expression, Convert absolute differences to relative proportions (e.g., calcified areas). , The ratio (100 / 120≈0.83) eliminates the influence of absolute grayscale differences between different devices. The ratio of vascular calcification boundary >0.6 (strong discontinuity), and the ratio of vascular soft tissue boundary <0.3 (weak discontinuity).

[0048] Furthermore, Precise control of threshold sensitivity was achieved. To achieve an output from 0 to 1. In this formula, the larger the relative proportion, the better. The smaller the value, the smaller the relative proportion. The larger; at the same time, since the relative proportion is between 0 and 1, Between arrive In between, that is It belongs to (0.37, 1).

[0049] The entire expression uses a nonlinear mapping of the relative difference in grayscale gradient magnitudes to ensure that the extended threshold is strictly matched with the discontinuity of local tissues, thus solving to some extent the problem that a fixed threshold cannot simultaneously meet the contradictory needs of calcified areas and small blood vessel areas.

[0050] like Figure 8 As shown, in one embodiment, obtaining the target angle threshold of each sub-pixel region in S4 based on the number of sub-pixel regions participating in the formation of the proposed extension segment and a preset angle threshold includes: S43. Obtain the attenuation angle value mapped to the maximum adjustment quantity and the unit quantity; S44. Obtain the target angle threshold of each sub-pixel region based on the number of sub-pixel regions participating in the formation of the proposed extension segment, the preset angle threshold, the maximum adjustment number, and the attenuation angle value mapped per unit number.

[0051] In this embodiment, it should be noted that in S43, two key parameters are preset: the maximum number of adjustments represents the upper limit of the bifurcation complexity (corresponding to the convergence of four branches in a typical anatomy), and the unit attenuation angle value defines the angle widening step size when the number of connections increases. For example, in the coronary artery bifurcation model, the maximum number of adjustments is set to 3 (to adapt to common tri-branch structures), and the attenuation angle value mapped by the unit number is 30°, which means that the angle threshold decreases by 30° for each additional effective connection. This ensures that the adjustment strength covers physiological characteristics (such as human cerebral blood vessel bifurcation usually ≤4 branches) and prevents excessive attenuation caused by noise interference (such as a sudden drop in angle triggered by a single noise point).

[0052] In S42, the target angle threshold is generated by taking the number of connections, the preset angle threshold and the attenuation parameter obtained in S42 as input.

[0053] It should also be noted that the target angle threshold of each sub-pixel region obtained in S42 or S4 based on the number of sub-pixel regions participating in the formation of the proposed extension segment and the preset angle threshold is expressed as follows: ;in, Let the target angle threshold be the i-th sub-pixel region. For the preset angle threshold, The number of sub-pixel regions participating in the formation of the proposed extension segment. To maximize the adjustment quantity, The attenuation angle value mapped per unit quantity.

[0054] It should also be noted that the entire expression is achieved through... Achieve triple control; when When, the adjustment amount is Bifurcation complexity is positively correlated with threshold relaxation, such as at coronary artery bifurcation points. ,like Then it will eventually decay by 60°; when When, the adjustment amount is To avoid excessive attenuation due to noise interference, such as pseudo-bifaction in calcified areas (noise). ),according to Upper limit decay suppresses pseudo-connectivity; threshold inverse generation, straight vessel segment ( ), (Maintaining a high threshold to suppress false connections), bifurcation core region ( ), (Very low threshold captures acute-angled branches).

[0055] In summary, by dynamically adjusting parameters driven by anatomical features, the accuracy of blood vessel identification is improved in bifurcation lesion areas, while the pseudo-connection rate of straight segments is greatly reduced, thus solving the topological integrity problem to a certain extent.

[0056] A blood vessel recognition system based on image features is also provided, including: The data acquisition module is used to acquire the target image of the target area and divide the target image into multiple sub-pixel regions according to a preset window size; The first data processing module is used to obtain the gray gradient magnitude of each sub-pixel region based on the gray value of each pixel in each sub-pixel region, and to obtain the dynamic extension threshold of two adjacent sub-pixel regions based on the gray gradient magnitude of two adjacent sub-pixel regions. The second data processing module is used to obtain the grayscale change of two adjacent sub-pixel regions, connect two adjacent sub-pixel regions whose grayscale change is less than the dynamic extension threshold, and form a proposed extension segment. The third data processing module is used to obtain a preset angle threshold, obtain the number of sub-pixel regions participating in the formation of the proposed extension segment, and obtain the target angle threshold of each sub-pixel region based on the number of sub-pixel regions participating in the formation of the proposed extension segment and the preset angle threshold. The identification module is used to connect adjacent proposed extension segments if the included angle between them is greater than or equal to the target angle threshold of the sub-pixel region at the overlap, and to form a proposed blood vessel segment from multiple consecutive connected proposed extension segments.

[0057] In one embodiment, the second data processing module is further configured to: obtain the average gray value of each sub-pixel region based on the gray value of each pixel in each sub-pixel region; and obtain the difference between the average gray values ​​of adjacent sub-pixel regions and use it as the gray value change.

[0058] In this embodiment, it should be noted that the specific method of performing the operation of the above-mentioned image feature-based blood vessel recognition system has been described in detail in the embodiments of the image feature-based blood vessel recognition method, and will not be elaborated here.

[0059] Figure 9 This is a block diagram of an electronic device illustrating a blood vessel recognition method based on image features, according to an exemplary embodiment. Figure 9 As shown, the electronic device 700 may include: a processor 701 and a memory 702. The electronic device 700 may also include one or more of a multimedia component 703, an I / O interface 704 (input / output interface), and a communication component 705.

[0060] The processor 701 controls the overall operation of the electronic device 700 to complete all or part of the steps in the image feature-based blood vessel recognition method described above. The memory 702 stores various types of data to support the operation of the electronic device 700. This data may include, for example, instructions for any application or method operating on the electronic device 700, and application-related data such as contact data, sent and received messages, pictures, audio, video, etc. The memory 702 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The multimedia component 703 may include a screen and audio components. The screen may be, for example, a touchscreen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signals may be further stored in memory 702 or transmitted via communication component 705. The audio component also includes at least one speaker for outputting audio signals. I / O interface 704 provides an interface between processor 701 and other interface modules, such as a keyboard, mouse, buttons, etc. These buttons may be virtual or physical buttons. Communication component 705 is used for wired or wireless communication between the electronic device 700 and other devices. Wireless communication, such as Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, 4G, NB-IoT, eMTC, or other 5G technologies, or a combination thereof, is not limited here. Therefore, the corresponding communication component 705 may include: a Wi-Fi module, a Bluetooth module, an NFC module, etc.

[0061] In an exemplary embodiment, the electronic device 700 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above-described image feature-based blood vessel recognition method.

[0062] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided, which, when executed by a processor, implement the steps of the image feature-based blood vessel recognition method described above. For example, the computer-readable storage medium may be the memory 702 including the program instructions described above, which may be executed by the processor 701 of the electronic device 700 to complete the image feature-based blood vessel recognition method described above.

[0063] In another exemplary embodiment, a computer program product is also provided, the computer program product comprising a computer program executable by a programmable device, the computer program having a code portion for performing the above-described image feature-based blood vessel recognition method when executed by the programmable device.

[0064] The preferred embodiments of the present disclosure have been described in detail above with reference to the accompanying drawings. However, the present disclosure is not limited to the specific details of the above embodiments. Within the scope of the technical concept of the present disclosure, various simple modifications can be made to the technical solutions of the present disclosure, and these simple modifications all fall within the protection scope of the present disclosure.

[0065] It should also be noted that the various specific technical features described in the above embodiments can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, this disclosure will not describe the various possible combinations separately.

[0066] Furthermore, various different embodiments of this disclosure can be combined in any way, as long as they do not violate the spirit of this disclosure, they should also be regarded as the content disclosed in this disclosure.

[0067] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the claims and specification of the present invention.

Claims

1. A method for blood vessel identification based on image features, characterized in that, include: Acquire the target image of the target area and divide the target image into multiple sub-pixel regions according to a preset window size; The grayscale gradient magnitude of each sub-pixel region is obtained based on the grayscale value of each pixel in each sub-pixel region, and the dynamic extension threshold of two adjacent sub-pixel regions is obtained based on the grayscale gradient magnitude of two adjacent sub-pixel regions. Obtain the grayscale change of two adjacent sub-pixel regions, and connect two adjacent sub-pixel regions whose grayscale change is less than the dynamic extension threshold to form a proposed extension segment. Obtain a preset angle threshold, and obtain the number of sub-pixel regions participating in the formation of the proposed extension segment. Based on the number of sub-pixel regions participating in the formation of the proposed extension segment and the preset angle threshold, obtain the target angle threshold of each sub-pixel region. If the included angle between adjacent proposed extension segments is greater than or equal to the target angle threshold of the sub-pixel region at the overlap, then the adjacent proposed extension segments are connected, and multiple consecutive connected proposed extension segments are combined to form a proposed blood vessel segment.

2. The blood vessel recognition method based on image features according to claim 1, characterized in that, The acquisition of grayscale changes in adjacent sub-pixel regions includes: The average gray value of each sub-pixel region is obtained based on the gray value of each pixel within each sub-pixel region. The difference in the mean grayscale values ​​of adjacent sub-pixel regions is obtained and used as the grayscale change.

3. The blood vessel recognition method based on image features according to claim 1, characterized in that, The number of sub-pixel regions participating in the formation of the proposed extension segment includes: Obtain multiple grayscale changes between the i-th sub-pixel region and multiple sub-pixel regions adjacent to the i-th sub-pixel region, and use them as multiple changes to be processed in the i-th sub-pixel region; The number of multiple unprocessed changes in the i-th sub-pixel region that are less than the dynamic extension threshold is obtained and used as the number of i-th sub-pixel regions to participate in the formation of the proposed extension segment.

4. The blood vessel identification method based on image features according to claim 1, characterized in that, The step of obtaining the grayscale gradient magnitude of each sub-pixel region based on the grayscale value of each pixel within each sub-pixel region includes: For each pixel within each sub-pixel region, calculate the grayscale change between it and all its neighboring pixels, and take the absolute value of the largest grayscale change as the gradient intensity of that pixel. The gradient intensities of all pixels within each sub-pixel region are summed to obtain the grayscale gradient magnitude of each sub-pixel region.

5. The blood vessel recognition method based on image features according to claim 1, characterized in that, The step of obtaining the dynamic extension threshold of two adjacent sub-pixel regions based on the grayscale gradient magnitude of two adjacent sub-pixel regions includes: Obtain the absolute difference between the gray-level gradient magnitudes of two adjacent sub-pixel regions, and obtain the larger value between the gray-level gradient magnitudes of two adjacent sub-pixel regions. Then, obtain the correction parameters for the two adjacent sub-pixel regions based on the absolute difference and the larger value. Obtain the standard extension threshold of two adjacent sub-pixel regions, and obtain the dynamic extension threshold of two adjacent sub-pixel regions based on the standard extension threshold and the correction parameters of the two adjacent sub-pixel regions.

6. The blood vessel identification method based on image features according to claim 1, characterized in that, The step of obtaining the target angle threshold for each sub-pixel region based on the number of sub-pixel regions participating in the formation of the proposed extension segment and a preset angle threshold includes: Obtain the attenuation angle value mapped to the maximum adjustment number and the number of units; The target angle threshold for each sub-pixel region is obtained based on the number of sub-pixel regions participating in the formation of the proposed extension segment, the preset angle threshold, the maximum adjustment number, and the attenuation angle value mapped per unit number.

7. A blood vessel recognition system based on image features, characterized in that, The system includes: The data acquisition module is used to acquire the target image of the target area and divide the target image into multiple sub-pixel regions according to a preset window size; The first data processing module is used to obtain the gray gradient magnitude of each sub-pixel region based on the gray value of each pixel in each sub-pixel region, and to obtain the dynamic extension threshold of two adjacent sub-pixel regions based on the gray gradient magnitude of two adjacent sub-pixel regions. The second data processing module is used to obtain the grayscale change of two adjacent sub-pixel regions, connect two adjacent sub-pixel regions whose grayscale change is less than the dynamic extension threshold, and form a proposed extension segment. The third data processing module is used to obtain a preset angle threshold, obtain the number of sub-pixel regions participating in the formation of the proposed extension segment, and obtain the target angle threshold of each sub-pixel region based on the number of sub-pixel regions participating in the formation of the proposed extension segment and the preset angle threshold. The identification module is used to connect adjacent proposed extension segments if the included angle between them is greater than or equal to the target angle threshold of the sub-pixel region at the overlap, and to form a proposed blood vessel segment from multiple consecutive connected proposed extension segments.

8. The blood vessel recognition system based on image features according to claim 7, characterized in that, The second data processing module is also used for: The average gray value of each sub-pixel region is obtained based on the gray value of each pixel within each sub-pixel region. The difference in the mean grayscale values ​​of adjacent sub-pixel regions is obtained and used as the grayscale change.

9. An electronic device, characterized in that, include: A memory on which computer programs are stored; A processor for executing the computer program in the memory to implement the image feature-based blood vessel identification method according to any one of claims 1 to 6.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the image feature-based blood vessel identification method according to any one of claims 1 to 6.

Images