Method, device and system for identifying and detecting vulnerable plaque of cardiovascular image
Through grayscale analysis and determination of global segmentation threshold, the segmentation threshold is adjusted in combination with the correlation between the external rectangle of the connecting area and the blood vessel wall, the problem of poor segmentation effect of vulnerable plaque recognition in the prior art is solved, and more accurate and close Real image segmentation effect.
Patent Information
- Application Number
- CN202411406051.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-10
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2044-10-10
AI Technical Summary
When identifying vulnerable plaques in cardiovascular images, the segmentation threshold is inaccurate due to the small difference in grayscale values, which affects the image segmentation effect.
Through grayscale analysis and determination of global segmentation threshold, the iterative initial segmentation threshold is selected, and the segmentation threshold is adjusted based on the correlation between the external rectangle of the connecting area and the blood vessel wall, and iteratively optimizes the segmentation results to improve the identification accuracy of vulnerable plaques.
The image segmentation effect of vulnerable plaques in cardiovascular images is improved, and the accuracy of detection and the authenticity of segmentation results are enhanced.
Smart Images

Figure CN118918136A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of data analysis technology, and in particular to a method, device and system for identifying and detecting vulnerable plaques in cardiovascular images. Background Art
[0002] Vulnerable plaques in cardiovascular imaging refer to those plaques that are prone to rupture and may rapidly progress to lesions, which is one of the main pathogenesis of acute coronary syndrome. Because vulnerable plaques are unstable and prone to rupture, thrombi will form after rupture. These thrombi may block coronary arteries, leading to myocardial ischemia or myocardial infarction, which is one of the main causes of acute coronary syndrome and sudden cardiac death. Even after successful revascularization and optimal drug therapy, patients with vulnerable plaques are at high risk of recurrent ischemic events when angiography shows mild to moderate stenosis. Therefore, the identification and treatment of vulnerable plaques is a challenge in the management of cardiovascular disease, because they may suddenly rupture without obvious symptoms, leading to serious health problems. Therefore, it is important and necessary to identify and detect vulnerable plaques in cardiovascular imaging.
[0003] At present, the identification technology of vulnerable plaques in cardiovascular images mainly adopts the iterative threshold segmentation algorithm, which is a commonly used image segmentation algorithm. It mainly uses the difference between the grayscale value of vulnerable plaques and the grayscale value of normal blood vessels to find the optimal segmentation threshold through iterative threshold, thereby achieving the segmentation of vulnerable plaques.
[0004] However, the problem with the existing technology is that, since the vulnerable plaque is located inside the blood vessel, the difference in grayscale value between the plaque and the normal blood vessel is relatively small, while the difference in grayscale value between different tissues is large. This will cause the stopping threshold obtained by segmentation to not be the optimal threshold for achieving vulnerable plaque identification and segmentation, thereby affecting the final image segmentation effect. Summary of the invention
[0005] The purpose of the embodiments of the present application is to provide a method, device and system for identifying and detecting vulnerable plaques in cardiovascular images, which can adjust the stopping threshold of iterative threshold segmentation based on the possibility that the connected area obtained by segmentation appears to be a vulnerable plaque, thereby improving the image segmentation effect of vulnerable plaques in cardiovascular images.
[0006] The present application also provides a method for detecting vulnerable plaques in cardiovascular imaging, comprising the following steps: S1. Obtain target cardiovascular images; S2, performing grayscale analysis on the target cardiovascular image, and selecting an iterative initial segmentation threshold in a high grayscale region corresponding to the determined global segmentation threshold; S3, performing connected area segmentation based on the initial segmentation threshold, and determining the possibility of vulnerable plaque in each connected area by quantifying the correlation between the long sides of the circumscribed rectangle of the area and the blood vessel wall; S4, comprehensively considering the possibilities of various vulnerable plaques, and determining the segmentation threshold for the next iteration; S5, in two adjacent iterations, based on the difference in the likelihood of vulnerable plaques before and after the iteration and the difference in the segmentation threshold, determining the performance of the vulnerable plaques in the two adjacent iteration regions; S6. In a corresponding iteration, when it is determined that the required expression capability is greater than a preset threshold, the segmentation result of the current iteration is output.
[0007] Furthermore, in step S1, acquiring the target cardiovascular image includes: S11, obtaining an initial cardiovascular image, and performing noise assessment on the initial cardiovascular image; S12, performing denoising processing based on the initial cardiovascular image to obtain a corresponding denoised image; S13, performing grayscale processing based on the denoised image to obtain a corresponding grayscale image; S14. Perform background separation based on the grayscale image to obtain a target cardiovascular image.
[0008] Furthermore, in step S2, the grayscale analysis of the target cardiovascular image is performed, and an iterative initial segmentation threshold is selected in a high grayscale area corresponding to the determined global segmentation threshold, including: S21, determining a grayscale histogram of the target cardiovascular image; S22, based on the target cardiovascular image, using the OSTU algorithm to determine a global segmentation threshold; S23, determining a target position of the global segmentation threshold on the grayscale histogram, and obtaining a maximum peak value on the right side of the target position, and using the maximum peak value as an initial segmentation threshold for iteration.
[0009] Furthermore, in step S3, for each connected region, determining the possibility of vulnerable plaque in each connected region by quantifying the association between the long sides of the circumscribed rectangle of the region and the blood vessel wall includes: S31, constructing the maximum circumscribed rectangle corresponding to the connected area; S32, determining the possibility that each rectangular long side is a vascular wall by a first grayscale difference between the non-vulnerable plaque areas on both sides of the rectangular long side and a second grayscale difference between the non-vulnerable plaque area outside the rectangular long side and the background area; S33, quantifying the compression characteristics of the vulnerable plaque on the blood vessel based on the possibility of the blood vessel wall corresponding to each long side of the maximum circumscribed rectangle; S34, traversing the long side of the rectangle that is represented by the blood vessel wall, and determining a monotonically decreasing trend characteristic of the proportion of the number of pixels of suspected vulnerable plaques; S35. Determine the possibility of vulnerable plaques in the connected area by combining the compression characteristics and the trend characteristics.
[0010] Further, in step S32, the possibility of each rectangular long side being a vascular wall is determined by a first grayscale difference between non-vulnerable plaque areas on both sides of the rectangular long side and a second grayscale difference between the non-vulnerable plaque area outside the rectangular long side and the background area, including: S321, for each long side of the maximum circumscribed rectangle, constructing the neighborhoods inside and outside the corresponding rectangle, and the non-vulnerable patch regions that appear as non-vulnerable patches in each neighborhood; S322, determining a first grayscale difference between the relatively non-vulnerable patch regions and a second grayscale difference between the non-vulnerable patch region outside the corresponding long side and the background region based on the average grayscale level of the region; S323: Determine the possibility that the long side of each rectangle is a blood vessel wall by combining the first grayscale difference and the second grayscale difference.
[0011] Furthermore, in step S34, the one side of the traversed rectangle is represented as the long side of the blood vessel wall, and the monotonically decreasing trend characteristics of the proportion of the number of pixels of suspected vulnerable plaques are determined, including: S341, selecting one side of the rectangle that is the long side of the blood vessel wall, and traversing along this side in a preset order; S342, during the traversal process, for each pixel point, determining whether it is a suspected vulnerable plaque; S43, while traversing, calculating the proportion of suspected vulnerable plaques to normal blood vessels among all pixel points before the current position; S44. Compare the proportion of suspected vulnerable plaques in different locations and analyze the monotonic downward trend of this proportion.
[0012] Furthermore, in step S5, the difference in the likelihood of vulnerable plaques before and after the iteration is determined by the following steps: S51, obtaining the possibility of a first vulnerable patch in a first connected area in the current iteration process; S52, determining the likelihood of all second connected areas and corresponding second vulnerable plaques belonging to the first connected area in the previous iteration process; S53, comprehensively analyzing the obtained second vulnerable plaque probabilities to obtain a comprehensive vulnerable plaque probability; S54. Determine the vulnerable plaque likelihood difference before and after iteration based on the sum of the absolute differences between the first vulnerable plaque likelihood and the comprehensive vulnerable plaque likelihood.
[0013] Furthermore, in step S5, the performance of the two adjacent iteration regions for vulnerable plaques is determined by the following formula: ; in, It represents the performance of two adjacent iteration regions for vulnerable plaques. Indicates The connected regions in the iteration The possibility of vulnerable plaque, Indicates In the iterations, the region All connected areas of The mean probability of vulnerable plaques is Indicates The segmentation threshold is determined by the iteration. Indicates The segmentation threshold is determined by the iteration. is the normalization function, Express expectations.
[0014] The embodiment of the present application also provides a vulnerable plaque identification and detection device for cardiovascular imaging, the device comprising an image acquisition module, a segmentation threshold iteration analysis module, and a detection output module, wherein: The image acquisition module is used to acquire target cardiovascular images; The segmentation threshold iteration analysis module is used to perform grayscale analysis on the target cardiovascular image and select an iterative initial segmentation threshold in a high grayscale area corresponding to the determined global segmentation threshold; The segmentation threshold iteration analysis module is further used to perform connected region segmentation based on the initial segmentation threshold, and determine the possibility of vulnerable plaques in each connected region by quantifying the correlation between the long sides of the circumscribed rectangle of the region and the blood vessel wall; The segmentation threshold iteration analysis module is also used to comprehensively consider the possibilities of various vulnerable plaques and determine the segmentation threshold for the next iteration; The segmentation threshold iteration analysis module is also used in two adjacent iterations to determine the performance of vulnerable plaques in two adjacent iteration regions based on the difference in the likelihood of vulnerable plaques before and after the iteration and the difference in the segmentation threshold; The detection output module is used in the corresponding iteration to output the segmentation result of the current iteration when it is determined that the required expression ability is greater than a preset threshold.
[0015] An embodiment of the present application also provides a vulnerable plaque identification and detection system for cardiovascular imaging, the system comprising a memory and a processor, the memory storing a computer program executed by the processor, and the computer program executing a vulnerable plaque identification and detection method when executed by the processor.
[0016] As can be seen from the above, the embodiments of the present application provide a vulnerable plaque identification and detection method, device and system for cardiovascular images. On the one hand, the present application can preliminarily identify high grayscale areas through grayscale analysis and determination of the global segmentation threshold. These areas are often associated with vulnerable plaques in the cardiovascular system. The connected region segmentation based on the initial segmentation threshold can effectively separate the blood vessels and potential vulnerable plaque areas in the image. The correlation between the long sides of the circumscribed rectangle of the quantified region and the vascular wall can provide a basis for determining the possibility of vulnerable plaques in each connected area, thereby improving the accuracy of the detection. On the other hand, the present application also continuously adjusts the segmentation threshold according to the possibility of vulnerable plaques through iteration, so that the segmentation results are closer to the actual situation. In the iterative process, the changes in the possibility of vulnerable plaques and the differences in the segmentation thresholds are comprehensively considered to ensure that each iteration can improve the performance of vulnerable plaques and improve the image segmentation effect.
[0017] Other features and advantages of the present application will be described in the subsequent description, and in part will become apparent from the description, or will be understood by implementing the embodiments of the present application. BRIEF DESCRIPTION OF THE DRAWINGS
[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments of the present application will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.
[0019] Figure 1 A flowchart of a vulnerable plaque identification and detection method for cardiovascular imaging provided in an embodiment of the present application; Figure 2 Schematic diagram of each neighborhood inside and outside the corresponding maximum circumscribed rectangle; Figure 3 A device structure diagram of a vulnerable plaque identification and detection device for cardiovascular imaging provided in an embodiment of the present application; Figure 4 A system structure diagram of a vulnerable plaque identification and detection system for cardiovascular imaging provided in an embodiment of the present application. DETAILED DESCRIPTION
[0020] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. The components of the embodiments of the present application described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application claimed for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present application.
[0021] It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. At the same time, in the description of this application, the terms "first", "second", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.
[0022] Please refer to Figure 1 , Figure 1 This is a flow chart of a vulnerable plaque identification and detection method for cardiovascular imaging in some embodiments of the present application. The method comprises the following steps: Step S1, acquiring a target cardiovascular image.
[0023] Step S2, performing grayscale analysis on the target cardiovascular image, and selecting an iterative initial segmentation threshold in a high grayscale region corresponding to the determined global segmentation threshold.
[0024] Step S3, performing connected region segmentation based on the initial segmentation threshold, and determining the possibility of vulnerable plaques in each connected region by quantifying the association between the long sides of the circumscribed rectangle of the region and the blood vessel wall.
[0025] Step S4: Comprehensively consider the possibilities of various vulnerable plaques and determine the segmentation threshold for the next iteration.
[0026] Step S5, in two adjacent iterations, based on the difference in the likelihood of vulnerable plaques before and after the iterations and the difference in the segmentation threshold, the representation capabilities of the two adjacent iteration regions for vulnerable plaques are determined.
[0027] Step S6: In a corresponding iteration, when it is determined that the required expression capability is greater than a preset threshold, the segmentation result of the current iteration is output.
[0028] As can be seen from the above, the present application discloses a vulnerable plaque identification and detection method for cardiovascular images. On the one hand, the present application can preliminarily identify high grayscale areas through grayscale analysis and determination of the global segmentation threshold. These areas are often associated with vulnerable plaques in the cardiovascular system. The connected region segmentation based on the initial segmentation threshold can effectively separate the blood vessels and potential vulnerable plaque areas in the image. The correlation between the long sides of the circumscribed rectangle of the quantified region and the vascular wall can provide a basis for determining the possibility of vulnerable plaques in each connected area, thereby improving the accuracy of the detection. On the other hand, the present application also continuously adjusts the segmentation threshold according to the possibility of vulnerable plaques through an iterative method, so that the segmentation results are closer to the actual situation. In the iterative process, the changes in the possibility of vulnerable plaques and the differences in the segmentation thresholds are comprehensively considered to ensure that each iteration can improve the representation of vulnerable plaques and improve the image segmentation effect.
[0029] In one embodiment, in step S1, acquiring the target cardiovascular image includes: Step S11, acquiring an initial cardiovascular image, and performing noise evaluation on the initial cardiovascular image.
[0030] Specifically, the present application uses computed tomography angiography (CTA) to obtain cardiovascular images that will provide detailed structural and functional information of the cardiovascular system.
[0031] Furthermore, noise assessment is the process of determining the noise level of an image. The present application adopts a quantitative noise assessment method, such as calculating the signal-to-noise ratio of an image, to measure the noise level.
[0032] Step S12: performing denoising processing based on the initial cardiovascular image to obtain a corresponding denoised image.
[0033] Specifically, the present application selects a suitable denoising method according to the result of noise evaluation. Commonly used denoising methods include Gaussian filtering, median filtering, bilateral filtering, non-local filtering, etc.
[0034] Step S13, performing grayscale processing based on the denoised image to obtain a corresponding grayscale image.
[0035] Specifically, converting a color cardiovascular image into a grayscale image is common knowledge in the art, and this application does not provide a detailed description thereof. The grayscale image obtained after processing contains only one brightness channel, and the pixel value ranges from 0 to 255.
[0036] Step S14: performing background separation based on the grayscale image to obtain a target cardiovascular image.
[0037] Specifically, a background separation method is selected, including a threshold segmentation method, a region growing method, an edge detection method, etc., and background separation is performed based on the method to obtain a target cardiovascular image. Among them, the threshold segmentation method divides the pixels in the image into the target and the background by setting one or more thresholds. The region growing method starts from a selected seed point and gathers adjacent pixels with similar grayscale values together to form a target area. The edge detection method detects the edges between the target area and the background in the image and separates the background based on the edge information.
[0038] In the above embodiment, a series of preprocessing is performed on the image before connected region segmentation, which can enhance the robustness of the subsequent algorithm. These preprocessing steps can reduce noise and interference factors in the image, making the subsequent algorithm more stable and reliable.
[0039] In one embodiment, in step S2, performing grayscale analysis on the target cardiovascular image and selecting an iterative initial segmentation threshold in a high grayscale region corresponding to the determined global segmentation threshold comprises: Step S21, determining the grayscale histogram of the target cardiovascular image.
[0040] Specifically, the grayscale histogram counts the frequency of each grayscale level in the image. In the grayscale histogram, the peak value usually corresponds to the main structure or area in the image. For cardiovascular images, vulnerable plaques usually have a larger lipid core, and the density of these lipid components is relatively low. Low-density substances absorb less X-rays, making vulnerable plaques appear lower in cardiovascular images. When there are multiple peaks in the grayscale histogram, it means that there are multiple different areas in the image, among which higher peaks correspond to areas with higher grayscale values such as blood vessel walls, and lower peaks correspond to areas with lower grayscale values such as lipid cores.
[0041] Step S22: Based on the target cardiovascular image, an OSTU algorithm is used to determine a global segmentation threshold.
[0042] Specifically, the present application uses the OSTU algorithm to process the target cardiovascular image, and when it is determined that the inter-class variance reaches the maximum value, a global segmentation threshold is obtained. This global segmentation threshold usually corresponds to the optimal segmentation point between the foreground and the background in the image.
[0043] That is to say, after being processed by the OSTU algorithm, the target cardiovascular image is divided into two types of pixels, including: 1) a type with a larger grayscale value (vascular wall or normal vascular area); 2) a type with a smaller grayscale value. Here, it is assumed that the pixel area with a smaller grayscale value corresponds to a vulnerable plaque.
[0044] Step S23, determining the target position of the global segmentation threshold on the grayscale histogram, and obtaining the maximum peak value on the right side of the target position, and using the maximum peak value as the initial segmentation threshold for iteration.
[0045] Specifically, since the grayscale histogram shows the frequency of occurrence of different gray levels in an image, after determining the global segmentation threshold, the present application obtains the position of the grayscale level equal to the global segmentation threshold in the grayscale histogram. For example, if the global segmentation threshold is 120, the column (or bin) with a grayscale level of 120 can be found in the grayscale histogram.
[0046] Once the position of the global segmentation threshold on the grayscale histogram is located, continue searching to the right of the position (i.e., in the direction of higher grayscale values) until the maximum peak on the histogram is found. It should be noted that this peak represents the grayscale of normal blood vessels in the image. Since normal blood vessels are mainly composed of high-density tissues, their grayscale values in the image are larger.
[0047] The reason for choosing a slightly higher initial threshold is to gradually adjust this value through iterative optimization, so that the areas with high grayscale values (such as blood vessel walls) and low grayscale values (such as lipid cores) can be better separated, so as to better identify the lipid core.
[0048] In the above embodiment, by combining global and local information to determine the segmentation threshold, the influence of noise and artifacts on the segmentation result can be reduced, and the accuracy and reliability of subsequent analysis can be improved.
[0049] In one embodiment, in step S3, for each connected region, determining the possibility of vulnerable plaque in each connected region by quantifying the association between the long sides of the circumscribed rectangle of the region and the blood vessel wall includes: Step S31: construct the maximum circumscribed rectangle corresponding to the connected area.
[0050] Specifically, the maximum circumscribed rectangle refers to the smallest rectangle that can completely contain the connected area.
[0051] Step S32: determine the possibility that each rectangular long side is a vascular wall by a first grayscale difference between the non-vulnerable plaque areas on both sides of the rectangular long side and a second grayscale difference between the non-vulnerable plaque area outside the rectangular long side and the background area.
[0052] Specifically, for each non-vulnerable plaque region and background region, the present application uses the mean grayscale value of the region in the target cardiovascular image as an indicator for evaluating the grayscale level, and determines the corresponding grayscale difference based on the difference between the indicators.
[0053] Step S33, based on the possibility of the blood vessel wall corresponding to each long side of the maximum circumscribed rectangle, quantify the compression characteristics of the vulnerable plaque on the blood vessel.
[0054] Specifically, when it is known that the blood vessel wall probabilities corresponding to the long sides of the maximum circumscribed rectangle are parameters b1 and b2 respectively, the present application is based on Quantify the compression characteristics of vulnerable plaques on blood vessels. The larger the value of this feature, the greater the possibility that the connected area is a vulnerable plaque.
[0055] Step S34, traverse the long side of the rectangle that is represented by the blood vessel wall, and determine the monotonically decreasing trend characteristic of the proportion of the number of suspected vulnerable plaque pixels.
[0056] Specifically, when traversing along the long side of a rectangular area, the traversal process usually starts from one end of the long side and gradually moves to the other end, wherein all pixels before the current position are considered in each step.
[0057] During the traversal process, for each position, the application will calculate the proportion of suspected vulnerable plaque pixels in all pixels from the starting position to the current position. This proportion can be obtained by dividing the number of suspected vulnerable plaque pixels by the number of pixels in the normal blood vessel area.
[0058] As the traversal proceeds, the application also checks whether the proportion of pixels suspected of being vulnerable patches shows a monotonically decreasing trend. This usually means that the density or distribution of suspected vulnerable patches decreases as you move along the long edge. To determine this trend, we can compare the proportion of each step with the proportion of the previous step. If the proportion of each step is less than or equal to the proportion of the previous step, then it can be considered that the proportion is monotonically decreasing.
[0059] Step S35, combining the compression feature and the trend feature to determine the possibility of vulnerable plaque in the connected area.
[0060] Specifically, the vulnerable plaque possibility of the connected area may be determined based on the product of the compression feature and the trend feature.
[0061] It should be noted that the reason for multiplying the compression feature and the trend feature is to comprehensively evaluate the possibility that the connected area is a vulnerable plaque. This multiplication operation is usually based on an assumption that both the compression feature and the trend feature have an important influence on the judgment of vulnerable plaques, and there may be some multiplication effect between them. In other words, if an area has both a high compression feature value and an obvious downward trend feature, then it is more likely to be considered a vulnerable plaque.
[0062] In the above embodiment, comprehensively considering the compression characteristics and trend characteristics to determine the possibility of vulnerable plaques in the connected area is a comprehensive and reliable method. This method combines information from multiple aspects and improves the accuracy and reliability of the assessment.
[0063] In one embodiment, in step S32, determining the possibility that each rectangular long side is a vascular wall by a first grayscale difference between non-vulnerable plaque areas on both sides of the rectangular long side and a second grayscale difference between the non-vulnerable plaque area outside the rectangular long side and the background area includes: Step S321 : for each long side of the maximum circumscribed rectangle, constructing the neighborhoods inside and outside the corresponding rectangle and the non-vulnerable patch regions that appear as non-vulnerable patches in each neighborhood.
[0064] For details, please refer to Figure 2 , the length of each neighborhood inside and outside the corresponding rectangle is the same as the length of the maximum circumscribed rectangle. In order to avoid introducing too much irrelevant information and to capture enough information, a compromise selection of the field width can be considered. In this application, the neighborhood width is set to 3 pixels.
[0065] Step S322, based on the average gray level of the area, determine the first gray difference between the relatively non-vulnerable patch areas and the second gray difference between the non-vulnerable patch area outside the corresponding long side and the background area.
[0066] Specifically, given the long side The mean gray value of the non-vulnerable patch area in the original image , and the long side The mean gray value of the non-vulnerable patch area corresponding to the outside in the original image When, based on Determine the first grayscale difference between each relatively non-vulnerable patch area. The greater the grayscale difference between the relatively non-vulnerable patch areas on the inner and outer sides (i.e., the greater the first grayscale difference), the longer side The more likely it is to appear as a blood vessel wall, the greater the possibility is, because the grayscale values on both sides of the blood vessel wall are very different, while the grayscale value of the boundary of the vulnerable plaque is less different from that of the blood vessel wall.
[0067] Specifically, the average gray value of the known background area in the original image z background When, based on Determine the second grayscale difference between the non-vulnerable patch area outside the corresponding rectangle and the background area. The smaller the grayscale difference between the outer non-vulnerable patch area and the background area (that is, the smaller the second grayscale difference), the longer edge The possibility of being a blood vessel wall is also greater, because when the second grayscale difference value is smaller, the proportion of background pixels in the corresponding non-vulnerable plaque area is higher, because the outside of the blood vessel wall is usually the background area, such as surrounding tissue or blank area during scanning. If the proportion of background pixels on the outside of the long side of the rectangular area is high, then this side is likely to be the outside of the blood vessel wall.
[0068] Step S323 , combining the first grayscale difference and the second grayscale difference to determine the possibility that the long side of each rectangle is a blood vessel wall.
[0069] Specifically, this application is based on and The ratio between the first grayscale difference and the second grayscale difference is taken as the parameter Substitution In this functional expression, the possibility that the long side of each rectangle is a blood vessel wall is calculated.
[0070] In the above embodiment, by comprehensively considering the grayscale difference between the inner and outer sides of the long side of the rectangle, the possibility of the blood vessel wall can be more comprehensively evaluated. The grayscale difference on the inner side reflects the grayscale characteristics of the blood vessel wall itself, while the grayscale difference on the outer side reflects the contrast between the blood vessel wall and the surrounding tissue. By comprehensively considering these two aspects of information, the existence and position of the blood vessel wall can be determined more accurately.
[0071] In one embodiment, in step S34, the one side of the traversed rectangle is represented as the long side of the blood vessel wall, and determining the monotonically decreasing trend characteristics of the proportion of the number of suspected vulnerable plaque pixels includes: Step S341, select one side of the rectangle which is the long side of the blood vessel wall, and traverse along this side in a preset order.
[0072] Specifically, after determining the long side of the rectangle, traverse along this side from left to right (or from right to left).
[0073] Step S342: During the traversal process, for each pixel point, determine whether it is a suspected vulnerable plaque.
[0074] Specifically, this may be based on the color of the pixels, wherein for a binary image, the pixels belonging to suspected vulnerable plaques are black, and the pixels belonging to normal blood vessel areas are white.
[0075] Step S43, while traversing, calculate the proportion of suspected vulnerable plaques to normal blood vessels in all pixel points before the current position.
[0076] Specifically, this can be achieved by simple counting, that is, dividing the total number of pixels belonging to suspected vulnerable plaques by the total number of pixels belonging to normal blood vessels.
[0077] Step S44, comparing the proportions of suspected vulnerable plaques at different locations, and analyzing the monotonic downward trend of the proportions.
[0078] Specifically, a monotonically decreasing trend may mean that the suspected vulnerable patch gradually decreases or disappears from a certain edge of the rectangle. The larger the value of this trend is, the greater the possibility that the corresponding connected area is a vulnerable patch.
[0079] Based on the above steps, the value of the monotonically decreasing trend characteristic of the proportion of the number of suspected vulnerable plaque pixels can be further calculated by the following formula: ; in, For the summation symbol, represents the number of traversals from the long side of the rectangle representing the blood vessel wall to the other long side of the rectangle, where The starting value of is 1. The ending value is the rectangle width. Indicates the number of the long sides of a rectangle. The ratio of the number of pixels in suspected vulnerable plaques to the number of pixels in normal vascular areas in the traversal. Represents the characteristic function, used in " "When this condition is met, the function value is 1, otherwise the function value is 0. It means to sum the current and previous function values, and compare multiple adjacent summation results to determine the monotonically decreasing trend characteristics. Indicates the width of the rectangle, which reflects the number of traversals.
[0080] In summary, combined with the compression feature calculation formula shown in step S33, the possibility of vulnerable plaques in each connected area can be determined by the following formula: .
[0081] In one embodiment, in step S5, the difference in the likelihood of vulnerable plaques before and after the iteration is determined by the following steps: Step S51, obtaining the possibility of the first vulnerable patch in the first connected area in the current iteration process.
[0082] Step S52: determining the likelihood of all second connected areas and corresponding second vulnerable plaques belonging to the first connected area in the previous iteration process.
[0083] Specifically, in this step, the results of the previous iteration are reviewed, and all connected areas that belong to the current first connected area in the previous iteration are found (the application refers to them as second connected areas). For each second connected area, the application also obtains the corresponding vulnerable plaque probability (herein referred to as the second vulnerable plaque probability).
[0084] Step S53, synthesizing the obtained second vulnerable plaque probabilities to obtain a comprehensive vulnerable plaque probability.
[0085] Specifically, the present application aggregates or summarizes the obtained second vulnerable plaque possibilities to obtain a corresponding comprehensive vulnerable plaque possibility.
[0086] In one embodiment, the present application specifically adopts the method of finding an average value to obtain the comprehensive vulnerable plaque possibility. Of course, in order to ensure the accuracy of the subsequent vulnerable plaque possibility difference analysis, a method such as finding a weighted average value can also be adopted, and the present application does not specifically limit it.
[0087] Step S54: determining the vulnerable plaque likelihood difference before and after iteration based on the sum of the absolute differences between the first vulnerable plaque likelihood and the comprehensive vulnerable plaque likelihood.
[0088] Specifically, the sum of absolute differences is obtained by taking the absolute value of the sum and difference of the two parameters and adding them together. This value is used to measure the difference between the vulnerable plaque probability in the current iteration and the comprehensive vulnerable plaque probability in the previous iteration. This difference can reflect the degree of change in the vulnerable plaque probability during the iteration. If this value is large, it means that there is a large difference between the vulnerable plaque probability in the current iteration and the comprehensive vulnerable plaque probability in the previous iteration, which also means that more iterations are needed to make the segmentation threshold complete the segmentation of vulnerable plaques and stabilize.
[0089] In one embodiment, in step S5, the performance of the two adjacent iteration regions for vulnerable plaques is determined by the following formula: ; in, It represents the performance of two adjacent iteration regions for vulnerable plaques. Indicates The connected regions in the iteration The possibility of vulnerable plaque, Indicates In the iterations, the region All connected areas of The mean probability of vulnerable plaques is Indicates The segmentation threshold is determined by the iteration. Indicates The segmentation threshold is determined by the iteration. is the normalization function, Express expectations.
[0090] Specifically, the formula It represents the same connected area before and after two adjacent iterations. Ability to express the likelihood of vulnerable plaque. Formula It reflects the difference in the likelihood of vulnerable plaques before and after iteration.
[0091] In the current embodiment, the performance of vulnerable plaques in two adjacent iteration regions is evaluated based on the ratio between the difference in the likelihood of vulnerable plaques before and after iteration and the difference in the segmentation threshold. Specifically, if the difference in the likelihood of vulnerable plaques is large and the difference in the segmentation threshold is small, it means that the recognition ability of the current iteration region for vulnerable plaques has been improved, and the fine-tuning of the segmentation threshold has played a key role. On the contrary, if the difference in the likelihood of vulnerable plaques is small and the difference in the segmentation threshold is large, it may mean that the recognition ability of the current iteration region for vulnerable plaques has not been significantly improved, and the segmentation threshold needs to be further adjusted.
[0092] In summary, based on the difference in the likelihood of vulnerable plaques before and after iteration and the ratio of the difference in segmentation thresholds, the representation ability of two adjacent iterative regions for vulnerable plaques is determined, and the recognition accuracy of vulnerable plaques is improved by continuously optimizing the segmentation strategy.
[0093] Please refer to Figure 3 , which is a device structure diagram of a vulnerable plaque identification and detection device for cardiovascular imaging disclosed in the present application, the device includes an image acquisition module, a segmentation threshold iteration analysis module, and a detection output module, wherein: The image acquisition module is used to acquire target cardiovascular images; The segmentation threshold iteration analysis module is used to perform grayscale analysis on the target cardiovascular image and select an iterative initial segmentation threshold in a high grayscale area corresponding to the determined global segmentation threshold; The segmentation threshold iteration analysis module is further used to perform connected region segmentation based on the initial segmentation threshold, and determine the possibility of vulnerable plaques in each connected region by quantifying the correlation between the long sides of the circumscribed rectangle of the region and the blood vessel wall; The segmentation threshold iteration analysis module is also used to comprehensively consider the possibilities of various vulnerable plaques and determine the segmentation threshold for the next iteration; The segmentation threshold iteration analysis module is also used in two adjacent iterations to determine the performance of vulnerable plaques in two adjacent iteration regions based on the difference in the likelihood of vulnerable plaques before and after the iteration and the difference in the segmentation threshold; The detection output module is used in the corresponding iteration to output the segmentation result of the current iteration when it is determined that the required expression ability is greater than a preset threshold.
[0094] In one embodiment, the above modules are also used to implement the implementation steps as illustrated in the above method embodiments.
[0095] As can be seen from the above, the present application discloses a vulnerable plaque identification and detection device for cardiovascular images. On the one hand, the present application can preliminarily identify high grayscale areas through grayscale analysis and determination of the global segmentation threshold. These areas are often associated with vulnerable plaques in the cardiovascular system. The connected region segmentation based on the initial segmentation threshold can effectively separate the blood vessels and potential vulnerable plaque areas in the image. The correlation between the long sides of the circumscribed rectangle of the quantified region and the vascular wall can provide a basis for determining the possibility of vulnerable plaques in each connected area, thereby improving the accuracy of the detection. On the other hand, the present application also continuously adjusts the segmentation threshold according to the possibility of vulnerable plaques through an iterative method, so that the segmentation results are closer to the actual situation. In the iterative process, the changes in the possibility of vulnerable plaques and the differences in the segmentation thresholds are comprehensively considered to ensure that each iteration can improve the performance of vulnerable plaques and improve the image segmentation effect.
[0096] Please refer to Figure 4 , which is a system structure diagram of a vulnerable plaque identification and detection system for cardiovascular imaging disclosed in the present application, the system includes a memory and a processor, the memory stores a computer program run by the processor, and the computer program executes a vulnerable plaque identification and detection method when run by the processor.
[0097] As can be seen from the above, the present application discloses a vulnerable plaque identification and detection system for cardiovascular images. On the one hand, the present application can preliminarily identify high grayscale areas through grayscale analysis and determination of the global segmentation threshold. These areas are often associated with vulnerable plaques in the cardiovascular system. The connected region segmentation based on the initial segmentation threshold can effectively separate the blood vessels and potential vulnerable plaque areas in the image. The correlation between the long sides of the circumscribed rectangle of the quantified region and the vascular wall can provide a basis for determining the possibility of vulnerable plaques in each connected area, thereby improving the accuracy of the detection. On the other hand, the present application also continuously adjusts the segmentation threshold according to the possibility of vulnerable plaques in an iterative manner, so that the segmentation results are closer to the actual situation. In the iterative process, the changes in the possibility of vulnerable plaques and the differences in the segmentation thresholds are comprehensively considered to ensure that each iteration can improve the representation of vulnerable plaques and improve the image segmentation effect.
[0098] In this document, relational terms such as first and second, etc. are used merely to distinguish one entity or operation from another entity or operation, but do not necessarily require or imply any such actual relationship or order between these entities or operations.
[0099] The above description is only an embodiment of the present application and is not intended to limit the protection scope of the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.
Claims
1. A vulnerable plaque identification and detection method for cardiovascular imaging, characterized in that: The following steps are involved: S1. Obtain target cardiovascular images; S2, performing grayscale analysis on the target cardiovascular image, and selecting an iterative initial segmentation threshold in a high grayscale region corresponding to the determined global segmentation threshold; S3, performing connected area segmentation based on the initial segmentation threshold, and determining the possibility of vulnerable plaque in each connected area by quantifying the correlation between the long sides of the circumscribed rectangle of the area and the blood vessel wall; S4, comprehensively considering the possibilities of various vulnerable plaques, and determining the segmentation threshold for the next iteration; S5, in two adjacent iterations, based on the difference in the likelihood of vulnerable plaques before and after the iteration and the difference in the segmentation threshold, determining the performance of the vulnerable plaques in the two adjacent iteration regions; S6. In a corresponding iteration, when it is determined that the required expression capability is greater than a preset threshold, the segmentation result of the current iteration is output.
2. The vulnerable plaque identification and detection method of cardiovascular imaging according to claim 1, characterized in that: In step S1, obtaining a target cardiovascular image includes: S11, obtaining an initial cardiovascular image, and performing noise assessment on the initial cardiovascular image; S12, performing denoising processing based on the initial cardiovascular image to obtain a corresponding denoised image; S13, performing grayscale processing based on the denoised image to obtain a corresponding grayscale image; S14. Perform background separation based on the grayscale image to obtain a target cardiovascular image.
3. The vulnerable plaque identification and detection method of cardiovascular imaging according to claim 1, characterized in that: In step S2, the grayscale analysis of the target cardiovascular image is performed, and an iterative initial segmentation threshold is selected in a high grayscale area corresponding to the determined global segmentation threshold, including: S21, determining a grayscale histogram of the target cardiovascular image; S22, based on the target cardiovascular image, using the OSTU algorithm to determine a global segmentation threshold; S23, determining a target position of the global segmentation threshold on the grayscale histogram, and obtaining a maximum peak value on the right side of the target position, and using the maximum peak value as an initial segmentation threshold for iteration.
4. The vulnerable plaque identification and detection method of cardiovascular imaging according to claim 1, characterized in that: In step S3, for each connected region, determining the possibility of vulnerable plaque in each connected region by quantifying the association between the long sides of the circumscribed rectangle of the region and the blood vessel wall includes: S31, constructing the maximum circumscribed rectangle corresponding to the connected area; S32, determining the possibility that each rectangular long side is a vascular wall by a first grayscale difference between the non-vulnerable plaque areas on both sides of the rectangular long side and a second grayscale difference between the non-vulnerable plaque area outside the rectangular long side and the background area; S33, quantifying the compression characteristics of the vulnerable plaque on the blood vessel based on the possibility of the blood vessel wall corresponding to each long side of the maximum circumscribed rectangle; S34, traversing the long side of the rectangle that is represented by the blood vessel wall, and determining a monotonically decreasing trend characteristic of the proportion of the number of pixels of suspected vulnerable plaques; S35. Determine the possibility of vulnerable plaques in the connected area by combining the compression characteristics and the trend characteristics.
5. The vulnerable plaque identification and detection method of cardiovascular imaging according to claim 4, characterized in that: In step S32, the possibility of each rectangular long side being a vascular wall is determined by a first grayscale difference between non-vulnerable plaque areas on both sides of the rectangular long side and a second grayscale difference between the non-vulnerable plaque area outside the rectangular long side and the background area, including: S321, for each long side of the maximum circumscribed rectangle, constructing the neighborhoods inside and outside the corresponding rectangle, and the non-vulnerable patch area that appears as a non-vulnerable patch in each neighborhood; S322, determining a first grayscale difference between the relatively non-vulnerable patch regions and a second grayscale difference between the non-vulnerable patch region outside the corresponding long side and the background region based on the average grayscale level of the region; S323: Determine the possibility that the long side of each rectangle is a blood vessel wall by combining the first grayscale difference and the second grayscale difference.
6. The vulnerable plaque identification and detection method of cardiovascular imaging according to claim 4, characterized in that: In step S34, the one side of the traversed rectangle is represented as the long side of the blood vessel wall, and the monotonically decreasing trend characteristics of the proportion of the number of suspected vulnerable plaque pixels are determined, including: S341, selecting one side of the rectangle that is the long side of the blood vessel wall, and traversing along this side in a preset order; S342, during the traversal process, for each pixel point, determine whether it is a suspected vulnerable plaque; S43, while traversing, calculating the proportion of suspected vulnerable plaques to normal blood vessels among all pixel points before the current position; S44. Compare the proportion of suspected vulnerable plaques in different locations and analyze the monotonic downward trend of this proportion.
7. The vulnerable plaque identification and detection method of cardiovascular imaging according to claim 1, characterized in that: In step S5, the difference in the likelihood of vulnerable plaques before and after the iteration is determined by the following steps: S51, obtaining the possibility of a first vulnerable patch in a first connected area in the current iteration process; S52, determining the likelihood of all second connected areas and corresponding second vulnerable plaques belonging to the first connected area in the previous iteration process; S53, comprehensively analyzing the obtained second vulnerable plaque probabilities to obtain a comprehensive vulnerable plaque probability; S54. Determine the vulnerable plaque likelihood difference before and after iteration based on the sum of the absolute differences between the first vulnerable plaque likelihood and the comprehensive vulnerable plaque likelihood.
8. The vulnerable plaque identification and detection method of cardiovascular imaging according to claim 1, characterized in that: In step S5, the performance of the two adjacent iterative regions on the vulnerable plaque is determined by the following formula: ; in, It represents the performance of two adjacent iteration regions for vulnerable plaques. Indicates The connected regions in the iteration The possibility of vulnerable plaque, Indicates In the iterations, the region All connected areas of The mean probability of vulnerable plaques is Indicates The segmentation threshold is determined by the iteration. Indicates The segmentation threshold is determined by the iteration. is the normalization function, Express expectations.
9. A vulnerable plaque recognition and detection device for cardiovascular imaging, characterized in that: The device comprises an image acquisition module, a segmentation threshold iteration analysis module, and a detection output module, wherein: The image acquisition module is used to acquire target cardiovascular images; The segmentation threshold iteration analysis module is used to perform grayscale analysis on the target cardiovascular image and select an iterative initial segmentation threshold in a high grayscale area corresponding to the determined global segmentation threshold; The segmentation threshold iteration analysis module is further used to perform connected region segmentation based on the initial segmentation threshold, and determine the possibility of vulnerable plaques in each connected region by quantifying the correlation between the long sides of the circumscribed rectangle of the region and the blood vessel wall; The segmentation threshold iteration analysis module is also used to comprehensively consider the possibilities of various vulnerable plaques and determine the segmentation threshold for the next iteration; The segmentation threshold iteration analysis module is also used in two adjacent iterations to determine the performance of vulnerable plaques in two adjacent iteration regions based on the difference in the likelihood of vulnerable plaques before and after the iteration and the difference in the segmentation threshold; The detection output module is used in the corresponding iteration to output the segmentation result of the current iteration when it is determined that the required expression ability is greater than a preset threshold.
10. A vulnerable plaque recognition and detection system for cardiovascular imaging, characterized in that: The system comprises a memory and a processor, wherein a computer program executed by the processor is stored in the memory, and when the computer program is executed by the processor, the vulnerable plaque identification and detection method according to any one of claims 1 to 8 is executed.
Citation Information
Patent Citations
Easily-damaged plaque detection method and related equipment
CN115511797A
Pavement crack detection system based on emulsified high-viscosity asphalt
CN116934748A
Photovoltaic steel structure assembly fracture detection method based on image features
CN117437228A
Medical image segmentation method for intracranial arteriosclerosis plaque of atrial fibrillation patient
CN118379313A
Vulnerable plaque identification method, application server thereof, and computer readable medium
US20200202519A1
