Ultrasound image-based carotid plaque analysis method and system

Abstract

The application relates to the field of ultrasonic images, in particular to a carotid artery plaque analysis method and system based on ultrasonic images.The method comprises the following steps: acquiring an input ultrasonic image; performing gray-scale classification processing based on the input ultrasonic image to obtain bright areas of at least one plaque; performing plaque region of interest selection on the input ultrasonic image to obtain a target plaque region of interest image; performing deep learning-based classification processing on the target plaque region of interest image to obtain dark areas of the at least one plaque; combining corresponding bright areas and dark areas of the at least one plaque to obtain a boundary contour of the at least one plaque; and determining the type of the at least one plaque according to the boundary contour of the at least one plaque.The application can realize analysis of the type of carotid artery plaque and reduce the workload of doctors in carotid artery plaque examination.

Description

Technical Field

[0001] This application relates to the field of ultrasound imaging, and in particular to a method and system for analyzing carotid plaques based on ultrasound imaging. Background Technology

[0002] Cardiovascular and cerebrovascular diseases are the leading cause of harm to human health. Because the carotid artery is located superficially, ultrasound is sensitive to it. Color Doppler ultrasound of the carotid artery can detect the degree of carotid artery sclerosis, indirectly reflecting the overall state of atherosclerosis. Carotid artery plaques are one of the direct manifestations of atherosclerosis, and early identification and effective intervention of plaques are important methods for preventing acute cardiovascular and cerebrovascular events.

[0003] Currently, the examination of carotid artery plaques generally involves doctors manually determining the plaque area, extracting the plaque boundary, and measuring parameters such as the long and short diameters of the plaque. This results in a large workload, and due to the influence of different doctors' techniques and experience, the probability of missed diagnoses and misdiagnoses is relatively high, the test data is not accurate enough, and the testing efficiency is low. Summary of the Invention

[0004] To enable the analysis of carotid artery plaque types and reduce the workload of doctors in examining carotid artery plaques, this application provides a carotid artery plaque analysis method and system based on ultrasound imaging.

[0005] In a first aspect, this application provides a method for carotid plaque analysis based on ultrasound images, employing the following technical solution: acquiring an input ultrasound image; performing grayscale classification processing on the input ultrasound image to obtain a bright region of at least one plaque; selecting a region of interest (ROI) for the plaque in the input ultrasound image to obtain a target plaque ROI image; performing deep learning-based classification processing on the target plaque ROI image to obtain a dark region of the at least one plaque; combining the bright region of the at least one plaque with the corresponding bright and dark regions in the dark region of the at least one plaque to obtain the boundary contour of the at least one plaque; and determining the type of the at least one plaque based on the boundary contour of the at least one plaque.

[0006] By employing the above technical solution, grayscale classification processing of the input ultrasound image can obtain the bright region of at least one plaque, enabling the screening of bright regions within the plaque. Furthermore, by selecting the region of interest (ROI) of the plaque in the input ultrasound image and performing deep learning-based classification processing, the dark region of at least one plaque can be obtained, enabling the screening of dark regions. Finally, combining the corresponding bright and dark regions within the bright and dark regions of at least one plaque yields a clearer boundary contour of at least one plaque. Based on this boundary contour, the type of at least one plaque can be determined. This enables the analysis of plaque types in the carotid artery based on the input ultrasound image, reducing the workload of doctors in examining carotid artery plaques, lowering the probability of missed and misdiagnosed cases, and improving the accuracy and efficiency of detection data.

[0007] Optionally, the step of performing grayscale classification processing based on the input ultrasound image to obtain the bright area of ​​at least one patch specifically includes: performing image normalization processing on the input ultrasound image to obtain a normalized ultrasound image; configuring the grayscale of the normalized ultrasound image to obtain a target ultrasound image; and performing grayscale classification processing on the target ultrasound image to obtain the bright area of ​​the at least one patch.

[0008] By adopting the above technical solution, the input ultrasound image is standardized to obtain a standardized ultrasound image, and then the standardized ultrasound image is configured with grayscale to obtain the target ultrasound image. This can improve the clarity of the ultrasound image and reduce the occurrence of key features being ignored due to insufficient grayscale ratio, which would affect the accuracy of subsequent detection processes and cause missed diagnoses and misdiagnoses. By performing grayscale classification processing on the target ultrasound image, at least one bright area of ​​the patch can be obtained, which can realize the screening of bright areas and facilitate the subsequent extraction of the boundary contour of the region of interest of the patch.

[0009] Optionally, the step of performing image standardization processing on the input ultrasound image to obtain a standardized ultrasound image specifically includes: comparing the grayscale percentage of the input ultrasound image with a preset grayscale percentage; if the grayscale percentage of the input ultrasound image does not meet the preset grayscale percentage, adjusting the grayscale percentage of the input ultrasound image to the preset grayscale percentage to obtain the standard ultrasound image; if the grayscale percentage of the input ultrasound image meets the preset grayscale percentage, using the input ultrasound image as the standardized ultrasound image.

[0010] By adopting the above technical solution and setting a preset grayscale percentage, the grayscale percentage of the input ultrasound image can be adjusted in a standardized manner, thereby reducing the possibility of inaccurate detection due to ignoring key features during subsequent plaque type analysis, which could lead to missed diagnoses and misdiagnoses.

[0011] Optionally, before performing image standardization processing based on the grayscale percentage of the input ultrasound image and a preset grayscale percentage to obtain a standard ultrasound image, the process includes: determining whether the grayscale percentage in the input ultrasound image meets a preset overly bright grayscale or a preset overly dark grayscale; if the grayscale percentage in the input ultrasound image meets a preset overly bright grayscale or a preset overly dark grayscale, no processing is performed.

[0012] By adopting the above technical solution, and by judging the grayscale percentage before image standardization processing, images that are too bright or too dark can be pre-screened to improve the quality requirements of ultrasound image detection. Furthermore, pre-screening unqualified images can avoid inaccurate detection due to image quality issues, thus preventing misdiagnosis and missed diagnosis.

[0013] Optionally, the step of selecting a region of interest (ROI) for a plaque in the input ultrasound image to obtain a target plaque ROI image specifically includes: identifying a plaque ROI in the input ultrasound image to obtain an initial plaque ROI; determining whether the initial plaque ROI satisfies the plaque ROI confidence level; and if the initial plaque ROI satisfies the plaque ROI confidence level, using the initial plaque ROI as the target plaque ROI image.

[0014] By adopting the above technical solution, and by judging the confidence level of the region of interest of the identified initial patch, the target patch region of interest image with high confidence can be selected, so as to avoid the influence of patch region of interest images with low confidence on the patch type judgment result, thereby ensuring the accuracy of patch type.

[0015] Optionally, determining the type of the at least one plaque based on its boundary contour specifically includes: determining the continuity of the plaque body and the relationship between the plaque body and the blood vessel wall based on the boundary contour of the at least one plaque; and determining the type of the plaque based on the continuity of the plaque body and the relationship between the plaque body and the blood vessel wall.

[0016] By adopting the above technical solution, the continuity of the plaque body and the relationship between the plaque body and the blood vessel wall can be determined based on the boundary contour of at least one plaque. Then, based on the continuity of the plaque body and the relationship between the plaque body and the blood vessel wall, the type of plaque can be determined to provide reference for doctors, reduce the workload of doctors in examining carotid artery plaques, and the plaque type determined by analysis combined with the doctor's judgment makes the plaque type more accurate.

[0017] Optionally, determining the type of plaque based on the continuity of the plaque body and the relationship between the plaque body and the blood vessel wall specifically includes: If the plaque body is determined to be discontinuous, the type of the plaque is determined to be an artifact or an ulcer; If the continuity of the patch body is determined, determine whether the confidence level of the continuity of the patch body meets the preset continuity confidence level; If the preset continuous confidence level is met, the patch body is determined to be a valid boundary contour, and the type of the patch is obtained as a normal patch. If the preset continuous confidence level is not met, a prompt will be given and input information will be received. Given that the plaque is connected to the blood vessel wall on both sides, determine whether the confidence level of the plaque boundary contour meets the preset boundary contour confidence level. If the preset boundary contour confidence level is met, the patch body is determined to be a valid boundary contour; If the preset boundary profile confidence level is not met, a prompt will be given and input information will be received. If it is determined that one side of the plaque is connected to the blood vessel wall, determine whether the plaque meets the criteria of being proximal or distal; If the plaque meets the criteria of being proximal, the type of the plaque is determined to be interlayer. If the plaque is determined to be distal, the type of the plaque is determined to be ulcer or cliff-like based on color Doppler imaging. If the type of plaque is determined to be a cliff-like image, determine whether the boundary contour of the plaque is far from the blood vessel wall; If the boundary contour of a plaque is determined to be far from the vessel wall, the boundary contour of the plaque is determined to be an invalid boundary contour. If the boundary contour of a plaque is determined to be in contact with the vessel wall, the type of plaque is determined to be a common plaque.

[0018] By adopting the above technical solution, and through judgment of different situations, intelligent judgment of plaque type can be achieved, accurately analyzing the type of plaque in the carotid artery, thereby reducing the probability of missed diagnosis and misdiagnosis caused by manual judgment, and improving the accuracy and efficiency of detection data.

[0019] Optionally, determining the type of the plaque as an ulcer or a cliff-like image based on the color blood flow image specifically includes: determining the type of the plaque as an ulcer when the blood flow direction in the color blood flow image satisfies a single flow direction; and determining the type of the plaque as a cliff-like image when the blood flow direction in the color blood flow image does not satisfy a single flow direction.

[0020] By adopting the above technical solution, based on the direction of blood flow in color blood flow images, it is possible to further analyze and determine the plaque types that may exist in the direction of blood flow, so as to assist doctors in quickly judging the plaque condition of patients and improve the accuracy and efficiency of doctors' diagnosis.

[0021] Optionally, determining the type of the at least one patch based on its boundary contour includes generating a patch identification report based on the type of the at least one patch.

[0022] By adopting the above technical solution, and generating reports based on the identified plaque types, doctors can make more intuitive references and judgments, thereby improving the accuracy and efficiency of their diagnoses.

[0023] Secondly, this application also provides a carotid plaque analysis system based on ultrasound imaging, employing the following technical solution: an input ultrasound image acquisition module for acquiring an input ultrasound image; a plaque bright region acquisition module for performing grayscale classification processing based on the input ultrasound image to obtain a bright region of at least one plaque; a target plaque region of interest image acquisition module for selecting a region of interest from the input ultrasound image to obtain a target plaque region of interest image; a plaque dark region acquisition module for performing deep learning-based classification processing on the target plaque region of interest image to obtain a dark region of the at least one plaque; a plaque boundary contour acquisition module for combining the bright region of the at least one plaque with the corresponding bright and dark regions in the dark region of the at least one plaque to obtain the boundary contour of the at least one plaque; and a plaque type determination module for determining the type of the at least one plaque based on the boundary contour of the at least one plaque.

[0024] By adopting the above technical solution, the input ultrasound image acquisition module can acquire the input ultrasound image. The plaque bright area acquisition module can perform grayscale classification processing based on the input ultrasound image to obtain the bright area of ​​at least one plaque, thus achieving the filtering of the bright area of ​​the plaque. The target plaque region of interest image acquisition module can select the region of interest of the plaque from the input ultrasound image to obtain the target plaque region of interest image. The plaque dark area acquisition module can perform deep learning-based classification processing on the target plaque region of interest image to obtain the dark area of ​​at least one plaque, thus achieving the filtering of the dark area of ​​the plaque. The plaque boundary contour acquisition module is used to combine the corresponding bright and dark areas of the bright area of ​​at least one plaque with the corresponding dark area of ​​at least one plaque to obtain the boundary contour of at least one plaque with a clearer contour. The plaque type determination module can determine the type of at least one plaque based on the boundary contour of at least one plaque. In this way, the analysis of carotid artery plaques based on the input ultrasound image can be realized, reducing the workload of doctors in examining carotid artery plaques, while reducing the probability of missed diagnosis and misdiagnosis, and improving the accuracy and efficiency of detection data.

[0025] In summary, this application includes at least one of the following beneficial technical effects: 1. By performing grayscale classification processing on the input ultrasound image, at least one bright region of a plaque can be obtained, enabling the filtering of bright regions within the plaque. Additionally, by selecting the region of interest (ROI) of the plaque in the input ultrasound image and performing deep learning-based classification processing, at least one dark region of a plaque can be obtained, enabling the filtering of dark regions. Finally, by combining the corresponding bright and dark regions of the at least one plaque with the corresponding bright and dark regions of the at least one plaque, a clearer boundary contour of the at least one plaque can be obtained. Based on this boundary contour, the type of at least one plaque can be accurately determined. This enables the analysis of carotid artery plaques based on input ultrasound images, reducing the workload of doctors in examining carotid artery plaques, lowering the probability of missed and misdiagnosed diagnoses, and improving the accuracy and efficiency of detection data.

[0026] 2. By performing image standardization on the input ultrasound image to obtain a standardized ultrasound image, and then configuring the grayscale of the standardized ultrasound image to obtain the target ultrasound image, the clarity of the ultrasound image can be improved, reducing the possibility of ignoring key features due to insufficient grayscale ratio, which could affect the accuracy of subsequent detection processes and cause missed diagnoses and misdiagnoses. By performing grayscale classification processing on the target ultrasound image to obtain at least one bright area of ​​a patch, the bright area can be screened, facilitating the subsequent extraction of the boundary contour of the region of interest of the patch.

[0027] 3. By setting a preset grayscale percentage, the grayscale percentage of the input ultrasound image can be adjusted in a standardized manner, which can reduce the possibility of inaccurate detection due to ignoring key features during subsequent plaque type analysis, resulting in missed diagnoses and misdiagnoses. Attached Figure Description

[0028] Figure 1 This is a schematic flowchart of the carotid plaque analysis method based on ultrasound imaging disclosed in the embodiments of this application; Figure 2 for Figure 1 A schematic diagram of one implementation process of step S20 shown in the figure; Figure 3 for Figure 1 A schematic diagram of one implementation process of step S21 shown in the figure; Figure 4 for Figure 1 A schematic diagram of one implementation process of step S30 shown in the figure; Figure 5 for Figure 1 A schematic diagram of one implementation process of step S60 shown in the figure; Figure 6 This is a schematic flowchart illustrating an example of a carotid plaque analysis method based on ultrasound imaging disclosed in an embodiment of this application. Figure 7 for Figure 6 A flowchart illustrating an example of patch type determination based on fused patch contours; Figure 8 This is a schematic diagram of the modules of the carotid plaque analysis system based on ultrasound imaging disclosed in the embodiments of this application.

[0029] Explanation of reference numerals in the attached figures: 10. Input ultrasound image acquisition module; 20. Patch bright area acquisition module; 30. Target patch region of interest image acquisition module; 40. Patch dark area acquisition module; 50. Patch boundary contour acquisition module; 60. Patch type determination module. Detailed Implementation

[0030] The present application will be further described in detail below with reference to the accompanying drawings.

[0031] To enable the analysis of plaque types in the carotid artery and reduce the workload of physicians in examining carotid artery plaques, one embodiment of this application provides a carotid artery plaque analysis method based on ultrasound imaging. See also... Figure 1 The method includes the following steps: S10: Acquire the input ultrasound image; The acquisition of input ultrasound images mentioned herein can refer to a doctor inputting a carotid artery image obtained from an ultrasound image of a patient into the device used in the method disclosed in this application. The device can be a computer with data processing capabilities. Of course, in this embodiment, the acquisition of input ultrasound images can also be achieved by the device automatically retrieving the input ultrasound images from a database.

[0032] S20: Based on the input ultrasound image, perform grayscale classification processing to obtain at least one bright area of ​​a patch; The grayscale classification process mentioned can be, for example, using cluster analysis to classify grayscale levels. Grayscale can be understood as the grayscale (brightness) levels between black and white. Based on different grayscale levels, ultrasound images correspond to ultrasound waves with different echo intensities. Echo intensities can be divided into strong echo, high echo, isoecho, low echo, and anechoic. The bright areas mentioned can be the overly bright areas presented by strong echo and high echo in ultrasound images.

[0033] In another embodiment, see Figure 2 Step S20 includes: S21: Perform image standardization processing on the input ultrasound image to obtain a standardized ultrasound image; The standardization of input ultrasound images mentioned above can optimize the quality of ultrasound images, improve clarity, and reduce the occurrence of key features being overlooked due to insufficient grayscale ratio, which could affect the accuracy of subsequent detection processes and cause missed diagnoses and misdiagnoses.

[0034] In another embodiment, see Figure 3 Step S21 specifically includes: Based on the grayscale percentage of the input ultrasound image and a preset grayscale percentage, image standardization processing is performed to obtain a standard ultrasound image. This allows for adjusting the grayscale percentage of the input ultrasound image based on the preset grayscale percentage, thereby improving the quality of the ultrasound image. The specific judgment process includes: S211: Compare the grayscale percentage of the input ultrasound image with a preset grayscale percentage; S212: If the grayscale percentage of the input ultrasound image does not meet the preset grayscale percentage, adjust the grayscale percentage of the input ultrasound image to the preset grayscale percentage to obtain the standard ultrasound image. S213: If the grayscale percentage of the input ultrasound image meets the preset grayscale percentage, the input ultrasound image is used as the standardized ultrasound image.

[0035] The preset grayscale percentage mentioned above can be understood as the echo percentage corresponding to each grayscale level. For example, the preset grayscale percentage is as follows: 0-50 corresponds to 15%~20% anechoic, 51-100 corresponds to 15%~20% low echo, 101-150 corresponds to 30%~35% isoechoic, 151-200 corresponds to 15%~20% high echo, and 201-255 corresponds to 0%~5% strong echo. Of course, it should be noted that the preset grayscale percentage is not limited to these and can be manually set according to the actual situation.

[0036] S22: Perform grayscale configuration on the standardized ultrasound image to obtain the target ultrasound image; The grayscale configuration mentioned above refers to adjusting the grayscale levels to improve the quality of ultrasound images.

[0037] S23: Perform grayscale classification processing on the target ultrasound image to obtain the bright area of ​​the at least one patch; The grayscale classification process mentioned above can be used to classify the grayscale of an image according to its levels, and select the bright areas with high grayscale values, such as strong echoes and high echoes, from the classified grayscale values.

[0038] S30: Select the region of interest for the plaque in the input ultrasound image to obtain the region of interest image of the target plaque; In this embodiment, the selection of the region of interest (ROI) for the input ultrasound image can be understood as automatically selecting the ROI without manual intervention from the doctor. Of course, in another embodiment, the selection of the target ROI image can also be determined by the doctor after the device automatically selects the ROI.

[0039] In another embodiment, see Figure 4 Step S30 involves the selection process by the doctor, specifically including: S31: Identify the region of interest (ROI) of the patch in the input ultrasound image to obtain the first patch ROI; The identification of regions of interest in patches in the input ultrasound image can be achieved, for example, by using semantic segmentation or object detection methods. Semantic segmentation methods include PSPNet (Pyramid Scene Parsing Network), Unet network model, and DeeplabV3+ image semantic segmentation. Object detection methods include Faster R-CNN, SSD, etc. Of course, this embodiment is not limited to these methods; any method that can achieve the same effect is acceptable.

[0040] S32: Determine whether the region of interest of the first patch satisfies the region of interest confidence level; The confidence level of the region of interest mentioned here can be understood as the credibility of the region of interest, the probability of truly believing it.

[0041] S33: If at least one of the first patch regions of interest satisfies the patch region of interest confidence level, the at least one first initial patch region of interest is used as the second patch region of interest image.

[0042] S34: Select at least one region of interest image from the second plaque region of interest image to obtain the target plaque region of interest image; here, the selection from the second plaque region of interest image is a process performed manually, i.e., by a doctor.

[0043] S40: Perform deep learning-based classification processing on the region of interest image of the target patch to obtain the dark region of the at least one patch; The deep learning mentioned can be performed through convolutional neural network models such as VGG, GoogleNet, or ResNet. The classification process based on deep learning is, for example, a semantic segmentation method, which is used to select at least one dark region of a patch, where the dark region can be an overly dark region presented by low echo.

[0044] S50: Combine the bright areas of the at least one patch with the corresponding bright and dark areas in the dark areas of the at least one patch to obtain the boundary contour of the at least one patch.

[0045] The combination of bright and dark areas mentioned here can be understood as the fusion of the boundaries between the bright and dark areas that make up each patch, resulting in a clearer boundary outline that is easier to select. The mention of at least one patch comprising one or more patches is illustrated in this embodiment, which uses the first patch as an example for boundary outline fusion. The specific steps are as follows: Extract the boundary contour of the bright area of ​​the first patch as the first boundary contour; Extract the boundary contour of the dark region of the first patch as the second boundary contour; The first boundary contour is combined with the second boundary contour to form the boundary contour of the first patch in the at least one patch.

[0046] S60: Determine the type of the at least one patch based on the boundary contour of the at least one patch; In this embodiment, the type of patch is determined by the boundary contour of the patch. See [link to relevant documentation]. Figure 4 In another embodiment, see Figure 5 Step S60 specifically includes: S61: Based on the boundary contour of the at least one plaque, determine the continuity of each plaque body and the relationship between each plaque body and the blood vessel wall.

[0047] S62: Determine the type of each plaque based on the continuity of each plaque body and the relationship between each plaque body and the blood vessel wall; The continuity of the plaque body mentioned can be understood as the continuity of the plaque body outline; the relationship between the plaque body and the blood vessel wall mentioned can be understood as whether the two are in contact. In this embodiment, the plaque type can be further determined based on the continuity of the plaque body and the relationship between the plaque body and the blood vessel wall, which can improve the accuracy of plaque type analysis.

[0048] In another implementation, the judgment process in step S62 specifically includes: If any one of the patches is determined to be continuous and the continuity confidence level meets the preset continuity confidence level, the patch type is determined to be a normal patch. The preset continuity confidence level can be set to a threshold range of 80% to 100% to filter out patches with a continuity confidence level greater than 80%, thus avoiding the influence of low continuity confidence level on the determination of patch type.

[0049] If any of the plaque bodies is determined to be continuous, and the continuity confidence level does not meet the preset continuity confidence level, the doctor is prompted to input the plaque type. Prompting the doctor to input the plaque type can improve the accuracy of plaque type determination through the doctor's further judgment.

[0050] If any of the plaques is found to be connected to the blood vessel wall on both sides, and the confidence level of the connection between the plaque and the blood vessel wall on both sides meets a preset connection confidence level, the type of the plaque is determined to be a common plaque. The preset connection confidence level can be, for example, a threshold range of 80% to 100%, to filter out plaques with a plaque body connection confidence level greater than 80%, and to avoid plaques with low confidence levels of connection between the plaque and the blood vessel wall on both sides affecting the determination of the plaque type.

[0051] If any one of the plaques is found to be connected to the blood vessel wall on one side and is located proximal to the heart, the type of the plaque is identified as dissection. The proximal end can be understood as the blood flow input end, and due to the impact of the blood flow input on the plaque and blood vessel wall, dissection or ulceration is very likely to occur. Dissection can be understood as a perforation in the inner wall of the blood vessel, through which blood can flow, presenting a "three-shaped" blood flow channel in the ultrasound image. The middle horizontal line of the "three-shaped" shape is similar to the shape of the inner wall of the blood vessel floating in the blood vessel.

[0052] It should be noted that the above-mentioned criteria for determining plaque type are not limited and can be set according to the actual situation.

[0053] To more clearly illustrate the carotid plaque analysis method based on ultrasound imaging disclosed in this application, please refer to [link to relevant documentation]. Figure 6 and Figure 7 As shown, the process it executes in practical applications is as follows: First, see Figure 6 The process involves manually inputting ultrasound images from patients. One workflow is to first standardize and process the ultrasound images to obtain standard ultrasound images. Then, grayscale processing is performed on the standard ultrasound images to obtain higher quality ultrasound images. Based on the grayscale of the high-quality ultrasound images, bright areas are selected to obtain at least one bright area of ​​a plaque. The other workflow is to first select regions of interest (ROIs) for the plaques in the input ultrasound images. Target ROIs can be manually defined and filtered. Based on deep learning of neural networks, dark areas of the plaques are obtained from the target ROIs. The bright and dark areas of the same plaque after the two workflows are fused together. Furthermore, the type of plaque is determined by the fused boundary contours.

[0054] Again, see Figure 7The process involves determining the continuity of each plaque body and its relationship to the vessel wall based on the fused boundary contours. First, one workflow determines the continuity of the plaque body based on the fused boundary contours. If discontinuity is detected, the plaque type is directly identified as a suspected artifact or ulcer, which can then be further assessed by a physician. If continuity is detected, the continuity confidence level is checked. If both are met, the plaque is classified as a normal plaque. If the continuity confidence level is insufficient, the physician is prompted to input the plaque type. The other workflow determines the relationship between each plaque body and the vessel wall based on the fused boundary contours. First, it checks whether the plaque boundaries connect to the vessel wall on both sides. If they do, the process continues... The process involves several steps. First, it checks if the connectivity confidence level is met. If both conditions are met, the plaque is classified as a normal plaque. If the connectivity confidence level is not met, the doctor is prompted to make a further assessment and input the plaque type. Second, if the plaque does not connect to the vessel wall on both sides of its boundary, but does connect to the vessel wall on one side, it is determined whether this is the proximal end of the blood flow input. If so, it is classified as a suspected ulcer; otherwise, it is classified as a suspected cliff-like image. Third, if the plaque does not connect to the vessel wall on one side of its boundary, it is directly classified as a suspected artifact. This is the general procedure for plaque assessment. The assessment criteria are not limited here and can be adjusted according to the actual situation. The suspected type can be reassessed by the doctor to increase the accuracy of the assessment.

[0055] See Figure 1 In this embodiment, after step S60, the method further includes: S70: Generate a patch identification report based on the type of the at least one patch; In step S70, after determining the plaque type, a type report can be generated based on the determined plaque type. In addition, in this embodiment, the size of the determined ordinary plaque can be measured to generate a measurement report, so that doctors can intuitively and quickly make a judgment on the patient's plaque condition, thereby improving the accuracy and efficiency of doctors' diagnosis.

[0056] Furthermore, in another embodiment, prior to step S20, the following may also be included: Determine whether the grayscale percentage in the input ultrasound image meets the preset overly bright or overly dark grayscale. Overly bright grayscale can be understood as the image having a large proportion of high echoes, and the preset overly bright grayscale can be the proportion of high echoes being 80% or more. Overly dark grayscale can be understood as the image having a large proportion of low echoes, and the preset overly dark grayscale can be the proportion of low echoes being 80% or more. Of course, this percentage data is not limited to these and can be manually set according to the actual situation.

[0057] If the grayscale percentage in the input ultrasound image meets the preset overbright or overdark grayscale, the input ultrasound image will not be further processed. In this way, by comparing the grayscale percentage in the input ultrasound image with the preset overbright or overdark grayscale, unqualified ultrasound images that are too bright or too dark can be pre-screened out to reduce the impact on the subsequent determination of the plaque type.

[0058] In summary, the carotid artery plaque analysis method based on ultrasound images disclosed in this application can obtain at least one bright region of a plaque by performing grayscale classification processing on the input ultrasound image, thereby enabling the screening of bright regions within the plaque. Furthermore, by selecting the region of interest (ROI) of the plaque in the input ultrasound image and performing deep learning-based classification processing, at least one dark region of the plaque can be obtained, enabling the screening of dark regions. Finally, by combining the corresponding bright and dark regions of the at least one plaque with the corresponding bright and dark regions of the at least one plaque, a clearer boundary contour of the at least one plaque can be obtained. Based on this boundary contour, the type of at least one plaque can be accurately determined. This enables the analysis of carotid artery plaques based on input ultrasound images, reducing the workload of doctors in examining carotid artery plaques and decreasing the likelihood of missed or misdiagnosed diagnoses. This method improves the probability of detection and enhances the accuracy and efficiency of detection data. By standardizing the input ultrasound image to obtain a standardized ultrasound image, and then configuring the grayscale of the standardized ultrasound image to obtain the target ultrasound image, the clarity of the ultrasound image can be improved. This reduces the possibility of ignoring key features due to insufficient grayscale ratio, which could affect the accuracy of subsequent detection processes and lead to missed diagnoses and misdiagnoses. By performing grayscale classification processing on the target ultrasound image, at least one bright area of ​​a patch can be obtained, which can facilitate the selection of bright areas and the subsequent extraction of the boundary contour of the region of interest of the patch. By setting a preset grayscale percentage, the grayscale percentage of the input ultrasound image can be adjusted in a standardized manner to reduce the possibility of inaccurate detection due to ignoring key features during subsequent patch type analysis, thus preventing missed diagnoses and misdiagnoses.

[0059] Furthermore, the labels for each step in this embodiment are for illustrative purposes only and do not represent a limitation on the execution order of each step. In practical applications, the execution order of each step can be adjusted or performed simultaneously as needed, and such adjustments or substitutions are all within the protection scope of this invention.

[0060] Another embodiment of this application provides a carotid artery plaque analysis system based on ultrasound imaging. (See also...) Figure 8The system includes: an input ultrasound image acquisition module 10, a bright patch region acquisition module 20, a target patch region of interest image acquisition module 30, a dark patch region acquisition module 40, a patch boundary contour acquisition module 50, and a patch type determination module 60. The input ultrasound image acquisition module 10 acquires an input ultrasound image; the bright patch region acquisition module 20 performs grayscale classification processing on the input ultrasound image to obtain a bright region of at least one patch; the target patch region of interest image acquisition module 30 selects a region of interest from the input ultrasound image to obtain a target patch region of interest image; the dark patch region acquisition module 40 performs deep learning-based classification processing on the target patch region of interest image to obtain a dark region of the at least one patch; the patch boundary contour acquisition module 50 combines the bright and dark regions corresponding to the bright and dark regions of the at least one patch to obtain the boundary contour of the at least one patch; and the patch type determination module 60 determines the type of the at least one patch based on its boundary contour.

[0061] It should be noted that the carotid artery plaque analysis system based on ultrasound imaging disclosed in this embodiment implements the carotid artery plaque analysis method based on ultrasound imaging as described in the previous embodiments, and therefore will not be described in detail here. Optionally, each module, unit, and other operation or function in this embodiment is used to implement the method in the previous embodiments.

[0062] The above are all preferred embodiments of this application and are not intended to limit the scope of protection of this application. Any feature disclosed in this specification (including the abstract and drawings) may be replaced by other equivalent or similar features unless specifically stated otherwise. That is, unless specifically stated otherwise, each feature is only one example of a series of equivalent or similar features.

Claims

1. A method for carotid plaque analysis based on ultrasound imaging, characterized in that, include: Acquire input ultrasound images; Based on the input ultrasound image, grayscale classification processing is performed to obtain at least one bright area of ​​a patch. The region of interest (ROI) of the target plaque is selected from the input ultrasound image to obtain the ROI image of the target plaque. The region of interest image of the target patch is subjected to classification processing based on deep learning to obtain the dark region of the at least one patch; The bright areas of the at least one patch are combined with the corresponding bright and dark areas in the dark areas of the at least one patch to obtain the boundary contour of the at least one patch. The type of the at least one patch is determined based on the boundary contour of the at least one patch; Specifically, the step of performing grayscale classification processing based on the input ultrasound image to obtain at least one bright region of a patch includes: The input ultrasound image is subjected to image standardization processing to obtain a standardized ultrasound image; The standardized ultrasound image is configured with grayscale to obtain the target ultrasound image; The target ultrasound image is subjected to grayscale classification processing to obtain the bright area of ​​the at least one patch; The step of selecting the region of interest (ROI) of the input ultrasound image to obtain the target patch ROI image specifically includes: Identify the region of interest (ROI) of the patch in the input ultrasound image to obtain the first patch ROI; Determine whether the region of interest of the first patch satisfies the region of interest confidence level; If at least one of the regions of interest in the first patch satisfies the region of interest confidence level, the at least one region of interest is used as the second region of interest image; At least one region of interest image is selected from the second patch region of interest image to obtain the target patch region of interest image.

2. The method according to claim 1, characterized in that, The step of performing image standardization processing on the input ultrasound image to obtain a standardized ultrasound image specifically includes: The grayscale percentage of the input ultrasound image is compared with the preset grayscale percentage; If the grayscale percentage of the input ultrasound image does not meet the preset grayscale percentage, the grayscale percentage of the input ultrasound image is adjusted to the preset grayscale percentage to obtain the standardized ultrasound image. If the grayscale percentage of the input ultrasound image meets the preset grayscale percentage, the input ultrasound image is used as the standardized ultrasound image.

3. The method according to claim 2, characterized in that, Before performing image normalization processing on the input ultrasound image by comparing the grayscale percentage with a preset grayscale percentage to obtain a normalized ultrasound image, the process includes: Determine whether the grayscale percentage in the input ultrasound image meets the preset overly bright grayscale or preset overly dark grayscale. If the grayscale percentage in the input ultrasound image meets the preset overly bright or overly dark grayscale, no processing is performed.

4. The method according to claim 1, characterized in that, Determining the type of the at least one patch based on its boundary contour specifically includes: Based on the boundary contour of the at least one plaque, the continuity of each plaque body and the relationship between each plaque body and the blood vessel wall are determined. The type of each plaque is determined based on the continuity of the plaque body and the relationship between the plaque body and the blood vessel wall.

5. The method according to claim 4, characterized in that, The determination of the plaque type based on the continuity of the plaque body and the relationship between the plaque body and the blood vessel wall specifically includes: If any one of the patches is determined to be continuous and the continuity confidence level meets the preset continuity confidence level, the type of the patch is determined to be a normal patch. If any of the plaque bodies is determined to be continuous, and the continuity confidence does not meet the preset continuity confidence, the doctor is prompted to enter the plaque type. If it is determined that any one of the plaques is connected to the blood vessel wall on both sides, and the confidence level of the connection between the two sides of the plaque and the blood vessel wall meets the preset connection confidence level, the type of the plaque is obtained as a normal plaque. If it is determined that one side of any of the plaques is connected to the blood vessel wall and is located proximal to the heart, the type of plaque is identified as dissection.

6. The method according to claim 1, characterized in that, The at least one patch includes a first patch; The step of combining the bright areas of the at least one patch with the corresponding bright and dark areas of the dark areas of the at least one patch to obtain the boundary contour of the at least one patch specifically includes: Extract the boundary contour of the bright area of ​​the first patch as the first boundary contour; Extract the boundary contour of the dark region of the first patch as the second boundary contour; The first boundary contour is combined with the second boundary contour to form the boundary contour of the first patch in the at least one patch.

7. The method according to claim 1, characterized in that, The step of determining the type of the at least one patch based on its boundary contour includes: A patch identification report is generated based on the type of the at least one patch.

8. A carotid plaque analysis system based on ultrasound imaging, characterized in that, For performing the method according to any one of claims 1 to 7, comprising: The input ultrasound image acquisition module is used to acquire input ultrasound images; The bright area acquisition module is used to perform grayscale classification processing based on the input ultrasound image to obtain the bright area of ​​at least one patch. The target patch region of interest image acquisition module is used to select the patch region of interest from the input ultrasound image to obtain the target patch region of interest image. The patch dark region acquisition module is used to perform deep learning-based classification processing on the region of interest image of the target patch to obtain the dark region of the at least one patch. The patch boundary contour acquisition module is used to combine the bright area of ​​the at least one patch with the corresponding bright and dark areas in the dark area of ​​the at least one patch to obtain the boundary contour of the at least one patch. A patch type determination module is used to determine the type of the at least one patch based on the boundary contour of the at least one patch.

Images