Carotid plaque detection and evaluation method and system based on multi-modal image feature analysis

Abstract

The application relates to a carotid artery plaque detection and evaluation method and system based on multi-modal image feature analysis, wherein the method comprises the following steps: inputting a pretreated carotid artery ultrasound image into a target positioning model to obtain a carotid artery ROI region and a plaque ROI region; inputting the carotid artery ROI region and the plaque ROI region into corresponding structure segmentation models respectively to obtain carotid artery segmentation results and plaque segmentation results, and performing post-processing on the carotid artery segmentation results and the plaque segmentation results respectively to obtain a carotid artery binary mask image and a plaque binary mask image; extracting all contours from the carotid artery binary mask image and the plaque binary mask image by using a contour extraction algorithm, and selecting the largest contour as a carotid artery main contour and a plaque main contour; calculating plaque quantitative features according to the plaque main contour; and determining plaque smoothness according to the plaque quantitative features and the carotid artery main contour. The application can improve detection accuracy.

Description

Technical Field

[0001] This invention relates to the field of medical image processing technology, and in particular to a method and system for detecting and evaluating carotid plaques based on multimodal image feature analysis. Background Technology

[0002] The carotid artery, a vital blood supply artery in the human body, directly impacts the safety of blood supply to the brain. The formation of atherosclerotic plaques in the carotid arteries is a significant contributing factor to serious cardiovascular and cerebrovascular diseases such as stroke. Therefore, accurate detection and morphological assessment of these plaques are crucial for early disease screening, risk stratification, and clinical intervention.

[0003] Currently, ultrasound imaging has become the primary method for carotid artery plaque screening due to its advantages such as being non-invasive, real-time, and low-cost. However, existing technologies still have significant shortcomings in practical applications, mainly in the following aspects:

[0004] First, most existing plaque detection methods only focus on identifying whether a plaque exists, lacking in-depth analysis of plaque morphological characteristics. In particular, they lack effective means to determine key clinical indicators such as whether a plaque is "smooth" or "non-smooth," making it difficult to meet clinical needs for plaque stability assessment.

[0005] Secondly, existing methods generally lack the ability to accurately determine the spatial location of plaques and cannot effectively distinguish whether a plaque is located on the anterior, posterior, or lateral wall of the carotid artery, thus limiting further analysis and classification of plaque risk.

[0006] Furthermore, in the image processing workflow, some methods directly perform inference calculations on the entire ultrasound image without effectively constraining the region of interest, resulting in a large amount of computation, low processing efficiency, and high CPU resource consumption, making it difficult to meet the needs of real-time or high-concurrency applications.

[0007] Furthermore, in terms of plaque type discrimination (such as soft plaques and hard plaques), existing technologies often lack an adaptive analysis mechanism for local gray-scale distribution features and cannot combine brightness reference information near the blood vessel wall to construct dynamic thresholds, resulting in insufficient stability of classification results.

[0008] In summary, current technologies have not yet formed a complete plaque analysis system that balances detection accuracy, computational efficiency, and clinical interpretability, making it difficult to provide structured and quantifiable auxiliary diagnostic evidence for clinical practice. Summary of the Invention

[0009] The technical problem to be solved by the present invention is to provide a method and system for carotid plaque detection and evaluation based on multimodal image feature analysis, which can significantly reduce computational overhead while improving detection accuracy.

[0010] The technical solution adopted by this invention to solve its technical problem is: to provide a method for carotid plaque detection and evaluation based on multimodal image feature analysis, comprising:

[0011] The carotid ultrasound image was preprocessed and then input into the target localization model to obtain the carotid ROI region and the plaque ROI region.

[0012] The carotid artery ROI region and plaque ROI region are respectively input into the corresponding structural segmentation model to obtain carotid artery segmentation results and plaque segmentation results. The carotid artery segmentation results and plaque segmentation results are then post-processed to obtain carotid artery binary mask images and plaque binary mask images.

[0013] A contour extraction algorithm was used to extract all contours from the carotid artery binary mask image and the plaque binary mask image, and the contour with the largest area was selected as the main contour of the carotid artery and the main contour of the plaque.

[0014] Quantitative features of the patch are calculated based on the main outline of the patch.

[0015] The smoothness of the plaque is determined based on the plaque quantification features and the main contour of the carotid artery.

[0016] The post-processing of the carotid artery segmentation results and plaque segmentation results to obtain the carotid artery binary mask image and plaque binary mask image specifically includes:

[0017] Set a confidence threshold, and use the confidence threshold to perform binarization processing on the carotid artery segmentation result and the plaque segmentation result respectively;

[0018] Connectivity detection was performed on the target pixels in the binarized carotid artery segmentation results and plaque segmentation results to identify all interconnected target pixel regions and retain the target pixel region with the largest area, thus obtaining the carotid artery binary mask image and the plaque binary mask image.

[0019] The step of calculating patch quantification features based on the main contour of the patch specifically includes:

[0020] pass Calculate the aspect ratio of the main outline region of the patch. Aspect ratio, To determine the length of the main outline region of the patch and width The maximum value in, To determine the length of the main outline region of the patch and width The minimum value in;

[0021] pass Calculate the boundary complexity of the main contour region of the patch. For boundary complexity, This represents the perimeter of the main outline of the patch. This represents the area of ​​the main outline of the patch.

[0022] The determination of plaque smoothness based on the plaque quantification features and the main contour of the carotid artery specifically includes:

[0023] Determine whether the aspect ratio of the main outline region of the patch exceeds a preset value, whether the length of the main outline region of the patch exceeds a length threshold, and whether the width of the main outline region of the patch exceeds a width threshold.

[0024] If the aspect ratio of the main outline region of the patch exceeds the preset value, and the length of the main outline region of the patch exceeds the length threshold, and the width of the main outline region of the patch exceeds the width threshold, then the patch is directly determined to be unsmooth.

[0025] If the aspect ratio of the main contour region of the plaque does not exceed the preset value, or the length of the main contour region of the plaque does not exceed the length threshold, or the width of the main contour region of the plaque does not exceed the width threshold, then an adaptive threshold is determined based on the main contour of the carotid artery, and the nature of the plaque is determined based on the adaptive threshold.

[0026] When the patch is hard, the smoothness of the patch is determined by the boundary complexity and the curvature of each point on the main contour of the patch.

[0027] When the patch is a soft patch, the smoothness of the patch is determined by the aspect ratio of the main contour region, the boundary complexity, and the curvature of each point on the main contour of the patch.

[0028] The step of determining an adaptive threshold based on the main contour of the carotid artery and determining the nature of the plaque based on the adaptive threshold specifically includes:

[0029] The main contour of the carotid artery is expanded outward by M pixels, and the area within the main contour of the carotid artery is removed to obtain an annular reference area;

[0030] Extract the top 30% of the brightest pixels in the annular reference region, calculate the average pixel value and subtract the preset reference bias value to obtain the adaptive threshold;

[0031] An adaptive threshold is used to binarize the region within the main contour of the patch, and the number of connected components is counted.

[0032] When the number of connected components exceeds a certain threshold, the patch is a soft patch; when the number of connected components does not exceed the threshold, the patch is a hard patch.

[0033] When the patch is a hard patch, whether the patch is smooth is determined based on the boundary complexity and the curvature of each point on the main contour of the patch. Specifically:

[0034] Calculate the curvature of each point on the main contour of the patch and take the average value to obtain the mean curvature.

[0035] Determine whether the boundary complexity exceeds a complexity threshold and whether the mean curvature is less than a curvature threshold;

[0036] If the boundary complexity exceeds the complexity threshold and the mean curvature is less than the curvature threshold, the patch is determined to be smooth.

[0037] If the boundary complexity does not exceed the complexity threshold, or the mean curvature is not less than the curvature threshold, the patch is determined to be unsmooth.

[0038] When the patch is a soft patch, the smoothness of the patch is determined based on the aspect ratio of the main contour region, the boundary complexity, and the curvature of each point on the main contour of the patch. Specifically:

[0039] Determine whether the aspect ratio of the main outline region of the patch exceeds a preset value;

[0040] If the aspect ratio of the main outline region of the patch exceeds a preset value, the patch is determined to be smooth.

[0041] If the aspect ratio of the main contour region of the patch does not exceed the preset value, the curvature of each point on the main contour of the patch is calculated and the average value is obtained.

[0042] Determine whether the boundary complexity exceeds a complexity threshold and whether the mean curvature is less than a curvature threshold;

[0043] If the boundary complexity exceeds the complexity threshold and the mean curvature is less than the curvature threshold, the patch is determined to be smooth.

[0044] If the boundary complexity does not exceed the complexity threshold, or the mean curvature is not less than the curvature threshold, the patch is determined to be unsmooth.

[0045] The carotid plaque detection and evaluation method based on multimodal image feature analysis further includes: determining the location of the plaque based on the carotid ROI region and the main outline of the plaque, specifically including:

[0046] The geometric center of the main contour of the patch is calculated using the image spatial distance. The calculation method is as follows: ,in, These are the x and y coordinates of the geometric center of the main contour of the patch, respectively. The area of ​​the main outline region of the patch. The 10th spatial distance of the image within the main contour region of the patch. The 0-1 spatial distance of the image within the main contour region of the patch. Within the main outline area of ​​the patch Pixel value at;

[0047] The location of the midline of the carotid artery ROI region is calculated as follows: ,in, This indicates the location of the midline of the carotid artery ROI region. The vertical coordinate is the upper left corner of the carotid artery ROI region. The height of the carotid artery ROI region;

[0048] When the ordinate of the geometric center of the main contour of the plaque is greater than the midline of the carotid ROI region minus the tolerance value, the plaque is located on the anterior wall of the carotid artery.

[0049] When the ordinate of the geometric center of the main contour of the plaque is less than the location of the midline of the carotid ROI region plus the tolerance value, the plaque is located on the posterior wall of the carotid artery.

[0050] In other cases, the plaque is located on the side wall of the carotid artery.

[0051] The technical solution adopted by this invention to solve its technical problem is: to provide a carotid plaque detection and evaluation system based on multimodal image feature analysis, comprising:

[0052] The ROI detection module is used to preprocess the carotid ultrasound image and input the preprocessed carotid ultrasound image into the target localization model to obtain the carotid ROI region and plaque ROI region;

[0053] The segmentation processing module is used to input the carotid artery ROI region and the plaque ROI region into the corresponding structural segmentation model respectively to obtain the carotid artery binary mask image and the plaque binary mask image;

[0054] The target contour extraction module is used to extract all outer contours from the carotid artery binary mask image and the plaque binary mask image using a contour extraction algorithm, and select the outer contour with the largest area as the main contour of the carotid artery and the main contour of the plaque.

[0055] The patch feature calculation module is used to calculate patch quantization features based on the main outline of the patch.

[0056] The plaque smoothness determination module is used to determine the smoothness of plaques based on the plaque quantification features and the main contour of the carotid artery.

[0057] The technical solution adopted by the present invention to solve its technical problem is: to provide a computer-readable storage medium on which a computer program is stored, wherein when the computer program is executed by a processor, the steps of the above-mentioned carotid plaque detection and evaluation method based on multimodal image feature analysis are implemented.

[0058] Beneficial effects

[0059] By employing the aforementioned technical solutions, this invention offers the following advantages and positive effects compared to existing technologies: It accurately acquires the ROI region of the carotid artery and plaque through a target localization model, avoiding the high overhead of whole-image calculation and significantly improving detection efficiency. Simultaneously, it avoids errors from manual selection, achieving automated and standardized region localization. Relying on a structural segmentation model and post-processing to obtain a precise binary mask image, combined with a contour extraction algorithm to select the main contour, it can realistically restore the actual morphology of blood vessels and plaques, laying a precise structural foundation for feature calculation. Quantitative features are calculated through the main contour of the plaque, and plaque smoothness is determined by combining it with the main contour of the carotid artery. Multi-dimensional analysis is achieved by integrating geometric and structural features, and the judgment logic aligns with clinical experience, improving diagnostic reliability. The overall method achieves fully automated analysis of carotid artery plaques from localization and segmentation to feature calculation and smoothness determination. It boasts high detection accuracy, excellent processing efficiency, and results with strong clinical interpretability, providing reliable quantitative evidence for early screening and risk grading of carotid artery plaques, and assisting in precise clinical intervention. Attached Figure Description

[0060] Figure 1 This is a flowchart of the carotid plaque detection and evaluation method according to the first embodiment of the present invention;

[0061] Figure 2 This is a flowchart of the process for determining the smoothness of a patch in the first embodiment of the present invention. Detailed Implementation

[0062] The present invention will be further illustrated below with reference to specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. Furthermore, it should be understood that after reading the teachings of this invention, those skilled in the art can make various alterations or modifications to the invention, and these equivalent forms also fall within the scope defined by the appended claims.

[0063] The first embodiment of this invention relates to a carotid plaque detection and evaluation method based on multimodal image feature analysis. This method, by introducing a region localization mechanism, contour feature modeling, and an adaptive grayscale analysis strategy, achieves a comprehensive evaluation of plaque structural characteristics and tissue attributes. This significantly reduces computational overhead while improving detection accuracy and enhancing the clinical interpretability and practical value of the results. Figure 1 As shown, this method for detecting and evaluating carotid plaques includes the following steps:

[0064] Step 1: Preprocess the carotid ultrasound image and input the preprocessed carotid ultrasound image into the target localization model to obtain the carotid ROI region and plaque ROI region.

[0065] The preprocessing in this step includes size normalization and grayscale standardization. The target localization model in this step is based on a multi-scale feature extraction mechanism, which enables stable detection of targets with different sizes and structures. This processing step reduces the scope of image processing and decreases the computational load of subsequent calculations.

[0066] Step 2: Input the carotid artery ROI region and plaque ROI region into the corresponding structural segmentation model to obtain the carotid artery segmentation result and plaque segmentation result, and perform post-processing on the carotid artery segmentation result and plaque segmentation result to obtain the carotid artery binary mask image and plaque binary mask image respectively.

[0067] This step, during post-processing, specifically includes:

[0068] Set a confidence threshold, and use the confidence threshold to perform binarization processing on the carotid artery segmentation result and the plaque segmentation result respectively;

[0069] Connectivity detection was performed on the target pixels in the binarized carotid artery segmentation results and plaque segmentation results to identify all interconnected target pixel regions and retain the target pixel region with the largest area, thus obtaining the carotid artery binary mask image and the plaque binary mask image.

[0070] To facilitate subsequent calculations, this step can restore the carotid artery binary mask image and the plaque binary mask image to the original image coordinate space, thereby obtaining an accurate segmentation result aligned with the original image.

[0071] Step 3: A contour extraction algorithm is used to extract all contours from the carotid artery binary mask image and the plaque binary mask image, and the contour with the largest area is selected as the main contour of the carotid artery and the main contour of the plaque. This implementation relies on the structural segmentation model and post-processing to obtain accurate binary mask images, and then combines the contour extraction algorithm to select the main contours, which can realistically restore the actual shape of blood vessels and plaques, laying an accurate structural foundation for feature calculation.

[0072] Step 4: Calculate the patch quantification features based on the main contour of the patch. This step specifically includes:

[0073] pass Calculate the aspect ratio of the main outline region of the patch. Aspect ratio, To determine the length of the main outline region of the patch and width The maximum value in, To determine the length of the main outline region of the patch and width The minimum value in;

[0074] pass Calculate the boundary complexity of the main contour region of the patch. For boundary complexity, This represents the perimeter of the main outline of the patch. This represents the area of ​​the main outline of the patch.

[0075] It is worth mentioning that this step can also calculate the mean and variance of gray levels in the main contour region of the patch. The mean gray level is calculated as follows:

[0076] ;

[0077] The formula for calculating the variance of gray levels is:

[0078] ;

[0079] in, This is the average grayscale value. For grayscale variance, Main outline region of the patch The total number of pixels within, Main outline region of the patch within The pixel value at that location.

[0080] Step 5: Determine the smoothness of the plaque based on the plaque quantification features and the main contour of the carotid artery. For example... Figure 2 As shown, this step specifically includes:

[0081] The system determines whether the aspect ratio of the main outline region of the patch exceeds a preset value, whether the length of the main outline region of the patch exceeds a length threshold, and whether the width of the main outline region of the patch exceeds a width threshold. Whether it is true or not, among which, For length threshold, 5 is the width threshold, and 5 is the preset value;

[0082] If the aspect ratio of the main outline region of the patch exceeds the preset value, and the length of the main outline region of the patch exceeds the length threshold, and the width of the main outline region of the patch exceeds the width threshold, then the patch is directly determined to be unsmooth.

[0083] If the aspect ratio of the main contour region of the plaque does not exceed the preset value, or the length of the main contour region of the plaque does not exceed the length threshold, or the width of the main contour region of the plaque does not exceed the width threshold, then an adaptive threshold is further determined based on the main contour of the carotid artery, and the nature of the plaque is determined based on the adaptive threshold.

[0084] When determining the nature of a plaque, firstly, the main contour of the carotid artery is expanded outward by M pixels, and the region within the main contour of the carotid artery is removed to obtain an annular reference region, where M is a positive integer from 3 to 5. Next, the top 30% of the brightest pixels in the annular reference region are extracted, the average pixel value is calculated, and a preset reference bias value is subtracted to obtain an adaptive threshold. Then, the region within the main contour of the plaque is binarized using the adaptive threshold, and the number of connected components is counted. When the number of connected components exceeds the threshold, the plaque is a soft plaque; when the number of connected components does not exceed the threshold, the plaque is a hard plaque.

[0085] When a patch is hard, its smoothness is determined based on the boundary complexity and the curvature of points on the main contour of the patch. Specifically:

[0086] Calculate the curvature of each point on the main contour of the patch and take the average value to obtain the mean curvature, i.e. ,in, The mean curvature, The first on the main outline of the patch The curvature at a point, The number of points on the main outline of the patch;

[0087] Determine whether the boundary complexity exceeds a complexity threshold and whether the mean curvature is less than a curvature threshold. Whether it is satisfied, among which, As a complexity threshold, The curvature threshold is used as the boundary complexity threshold. If the boundary complexity exceeds the complexity threshold and the mean curvature is less than the curvature threshold, the patch is determined to be smooth. If the boundary complexity does not exceed the complexity threshold or the mean curvature is not less than the curvature threshold, the patch is determined to be unsmooth.

[0088] When a patch is a soft patch, its smoothness is determined based on the aspect ratio of the main contour region, the complexity of its boundaries, and the curvature of each point on the main contour. The specific determination method is as follows:

[0089] Determine whether the aspect ratio of the main contour region of the patch exceeds a preset value, i.e., determine. Whether it is true or not; if the aspect ratio of the main contour region of the patch exceeds a preset value, the patch is determined to be smooth; if the aspect ratio of the main contour region of the patch does not exceed the preset value, the curvature of each point on the main contour of the patch is further calculated, and the average value is obtained to obtain the mean curvature value, which is calculated as follows: ,in, The mean curvature, The first on the main outline of the patch The curvature at a point, The number of points on the main contour of the patch is used as the basis for determination. Then, it is determined whether the boundary complexity exceeds a complexity threshold and whether the mean curvature is less than a curvature threshold. Whether it is true or not; if the boundary complexity exceeds the complexity threshold and the mean curvature is less than the curvature threshold, the patch is determined to be smooth; if the boundary complexity does not exceed the complexity threshold, or the mean curvature is not less than the curvature threshold, the patch is determined to be unsmooth.

[0090] It is easy to see that this invention calculates quantitative features by plaque principal contour calculation, combines them with carotid artery principal contour to determine plaque smoothness, integrates geometric and structural features to achieve multi-dimensional analysis, and the judgment logic is in line with clinical experience, thus improving diagnostic reliability.

[0091] It is worth mentioning that the carotid plaque detection and evaluation method based on multimodal image feature analysis in this embodiment may further include: determining the location of the plaque based on the carotid ROI region and the main outline of the plaque, which specifically includes:

[0092] First, the geometric center of the main contour of the patch is calculated using the image spatial distance. The calculation method is as follows:

[0093] ;

[0094] in, These are the x and y coordinates of the geometric center of the main contour of the patch, respectively. The area of ​​the main outline region of the patch. The 10th spatial distance of the image within the main contour region of the patch. The 0-1 spatial distance of the image within the main contour region of the patch. Within the main outline area of ​​the patch Pixel value at;

[0095] Next, the location of the midline of the carotid artery ROI region is calculated as follows:

[0096] ;

[0097] in, This indicates the location of the midline of the carotid artery ROI region. The vertical coordinate is the upper left corner of the carotid artery ROI region. The height of the carotid artery ROI region;

[0098] Next, determine the ordinate of the geometric center of the main contour of the patch. Location of the midline of the carotid artery ROI region Subtract tolerance value The relationship between the height and the proportional coefficient of the carotid artery ROI region. The product of, i.e. When the ordinate of the geometric center of the plaque's principal contour is greater than the location of the midline of the carotid artery ROI region minus the tolerance value, i.e. The plaque is located on the anterior wall of the carotid artery; when the ordinate of the geometric center of the plaque's principal contour is less than the location of the midline of the carotid artery ROI region plus a tolerance value, i.e. In some cases, the plaque is located on the posterior wall of the carotid artery; in others, the plaque is located on the lateral wall of the carotid artery.

[0099] Therefore, this invention can accurately determine whether a plaque is located on the anterior, posterior, or lateral wall of the carotid artery based on the spatial geometric relationship between the plaque and the carotid artery contour. This fills the gap in the existing technology regarding the ambiguity in determining the spatial location of plaques and provides key locational basis for plaque risk classification and depth analysis. By calculating the geometric center of the plaque, the carotid artery midline, and the tolerance value, it achieves stable classification of areas with ambiguous boundaries, effectively avoids location determination bias, and improves the accuracy and robustness of the determination results.

[0100] A second embodiment of the present invention relates to a carotid plaque detection and evaluation system based on multimodal image feature analysis, comprising:

[0101] The ROI detection module is used to preprocess the carotid ultrasound image and input the preprocessed carotid ultrasound image into the target localization model to obtain the carotid ROI region and plaque ROI region;

[0102] The segmentation processing module is used to input the carotid artery ROI region and the plaque ROI region into the corresponding structural segmentation model respectively to obtain the carotid artery binary mask image and the plaque binary mask image;

[0103] The target contour extraction module is used to extract all outer contours from the carotid artery binary mask image and the plaque binary mask image using a contour extraction algorithm, and select the outer contour with the largest area as the main contour of the carotid artery and the main contour of the plaque.

[0104] The patch feature calculation module is used to calculate patch quantization features based on the main outline of the patch.

[0105] The plaque smoothness determination module is used to determine the smoothness of plaques based on the plaque quantification features and the main contour of the carotid artery.

[0106] The segmentation processing module includes:

[0107] The binarization processing unit is used to set a confidence threshold and perform binarization processing on the carotid artery segmentation result and the plaque segmentation result respectively using the confidence threshold;

[0108] The identification and retention unit is used to perform connected component detection on the target pixels in the binarized carotid artery segmentation result and plaque segmentation result, respectively, to identify all interconnected target pixel regions and retain the target pixel region with the largest area, thus obtaining the carotid artery binary mask image and the plaque binary mask image.

[0109] The patch feature calculation module includes:

[0110] Aspect ratio calculation unit, used to calculate aspect ratio Calculate the aspect ratio of the main outline region of the patch. Aspect ratio, To determine the length of the main outline region of the patch and width The maximum value in, To determine the length of the main outline region of the patch and width The minimum value in;

[0111] Boundary complexity calculation unit, used to calculate the boundary complexity through Calculate the boundary complexity of the main contour region of the patch. For boundary complexity, This represents the perimeter of the main outline of the patch. This represents the area of ​​the main outline of the patch.

[0112] The patch smoothness determination module includes:

[0113] The first judgment unit is used to determine whether the aspect ratio of the main outline region of the patch exceeds a preset value, whether the length of the main outline region of the patch exceeds a length threshold, and whether the width of the main outline region of the patch exceeds a width threshold.

[0114] The first determination unit is used to directly determine that the patch is not smooth when the aspect ratio of the main outline region of the patch exceeds a preset value, the length of the main outline region of the patch exceeds a length threshold, and the width of the main outline region of the patch exceeds a width threshold.

[0115] The plaque nature determination unit is used to determine an adaptive threshold based on the main outline of the carotid artery when the aspect ratio of the main outline region of the plaque does not exceed a preset value, or the length of the main outline region of the plaque does not exceed a length threshold, or the width of the main outline region of the plaque does not exceed a width threshold, and to determine the nature of the plaque based on the adaptive threshold.

[0116] The second determination unit is used to determine whether the patch is smooth based on the boundary complexity and the curvature of each point on the main contour of the patch when the patch is hard.

[0117] The third determination unit is used to determine whether a patch is smooth based on the aspect ratio of the main contour region of the patch, the boundary complexity, and the curvature of each point on the main contour of the patch when the patch is a soft patch.

[0118] The patch nature determination unit includes:

[0119] An annular reference region determination subunit is used to expand the main contour of the carotid artery by M pixels and remove the area within the main contour of the carotid artery to obtain the annular reference region.

[0120] An extraction calculation subunit is used to extract the top 30% of the pixels in brightness in the annular reference region, calculate the average pixel value and subtract a preset reference bias value to obtain an adaptive threshold.

[0121] The binarization subunit is used to binarize the region within the main contour of the patch using an adaptive threshold and to count the number of connected components obtained.

[0122] The determination subunit is used to determine that the patch is a soft patch when the number of connected components exceeds a threshold, and to determine that the patch is a hard patch when the number of connected components does not exceed the threshold.

[0123] The second determination unit includes:

[0124] The first curvature mean calculation subunit is used to calculate the curvature of each point on the main contour of the patch and to obtain the average value of the curvature.

[0125] The first judgment subunit is used to determine whether the boundary complexity exceeds the complexity threshold and whether the mean curvature is less than the curvature threshold.

[0126] The first determination subunit is used to determine that the patch is smooth when the boundary complexity exceeds the complexity threshold and the mean curvature is less than the curvature threshold; and to determine that the patch is not smooth when the boundary complexity does not exceed the complexity threshold or the mean curvature is not less than the curvature threshold.

[0127] The third determination unit includes:

[0128] The second judgment subunit is used to determine whether the aspect ratio of the main outline region of the patch exceeds a preset value;

[0129] The second determination subunit is used to determine that the patch is smooth when the aspect ratio of the main outline region of the patch exceeds a preset value.

[0130] The second curvature mean calculation subunit is used to calculate the curvature of each point on the main contour of the patch when the aspect ratio of the main contour region of the patch does not exceed the preset value, and to obtain the average value of the curvature.

[0131] The third judgment subunit is used to determine whether the boundary complexity exceeds the complexity threshold and whether the mean curvature is less than the curvature threshold.

[0132] The third determination subunit is used to determine that the patch is smooth when the boundary complexity exceeds the complexity threshold and the mean curvature is less than the curvature threshold; and to determine that the patch is not smooth when the boundary complexity does not exceed the complexity threshold or the mean curvature is not less than the curvature threshold.

[0133] The carotid plaque detection and evaluation system based on multimodal image feature analysis further includes: a plaque spatial location determination module, used to determine the location of the plaque based on the carotid ROI region and the main outline of the plaque; the plaque spatial location determination module includes:

[0134] The geometric center calculation unit is used to calculate the geometric center of the main contour of the patch using the image spatial distance. The calculation method is as follows: ,in, These are the x and y coordinates of the geometric center of the main contour of the patch, respectively. The area of ​​the main outline region of the patch. The 10th spatial distance of the image within the main contour region of the patch. The 0-1 spatial distance of the image within the main contour region of the patch. Within the main outline area of ​​the patch Pixel value at;

[0135] The midline calculation unit is used to calculate the location of the midline in the carotid artery ROI region. The calculation method is as follows: ,in, This indicates the location of the midline of the carotid artery ROI region. The vertical coordinate is the upper left corner of the carotid artery ROI region. The height of the carotid artery ROI region;

[0136] The location determination unit is used to determine that the plaque is located on the anterior wall of the carotid artery when the ordinate of the geometric center of the main contour of the plaque is greater than the position of the midline of the carotid ROI region minus the tolerance value; when the ordinate of the geometric center of the main contour of the plaque is less than the position of the midline of the carotid ROI region plus the tolerance value, the plaque is located on the posterior wall of the carotid artery; otherwise, the plaque is located on the lateral wall of the carotid artery.

[0137] The tolerance value is the product of the height of the carotid artery ROI region and the scaling factor.

[0138] The third embodiment of the present invention relates to a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the carotid plaque detection and evaluation method based on multimodal image feature analysis of the first embodiment.

[0139] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage) containing computer-usable program code.

[0140] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0141] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction methods implemented in a process. Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0142] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0143] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for detecting and evaluating carotid plaques based on multimodal image feature analysis, characterized in that, include: The carotid ultrasound image was preprocessed and then input into the target localization model to obtain the carotid ROI region and the plaque ROI region. The carotid artery ROI region and plaque ROI region are respectively input into the corresponding structural segmentation model to obtain carotid artery segmentation results and plaque segmentation results. The carotid artery segmentation results and plaque segmentation results are then post-processed to obtain carotid artery binary mask images and plaque binary mask images. A contour extraction algorithm was used to extract all contours from the carotid artery binary mask image and the plaque binary mask image, and the contour with the largest area was selected as the main contour of the carotid artery and the main contour of the plaque. Quantitative features of the patch are calculated based on the main outline of the patch. The smoothness of the plaque is determined based on the plaque quantification features and the main contour of the carotid artery.

2. The carotid plaque detection and evaluation method based on multimodal image feature analysis according to claim 1, characterized in that, The post-processing of the carotid artery segmentation results and plaque segmentation results to obtain the carotid artery binary mask image and plaque binary mask image specifically includes: Set a confidence threshold, and use the confidence threshold to perform binarization processing on the carotid artery segmentation result and the plaque segmentation result respectively; Connectivity detection was performed on the target pixels in the binarized carotid artery segmentation results and plaque segmentation results to identify all interconnected target pixel regions and retain the target pixel region with the largest area, thus obtaining the carotid artery binary mask image and the plaque binary mask image.

3. The carotid plaque detection and evaluation method based on multimodal image feature analysis according to claim 1, characterized in that, The step of calculating patch quantification features based on the main contour of the patch specifically includes: pass Calculate the aspect ratio of the main outline region of the patch. Aspect ratio, To determine the length of the main outline region of the patch and width The maximum value in, To determine the length of the main outline region of the patch and width The minimum value in; pass Calculate the boundary complexity of the main contour region of the patch. For boundary complexity, This represents the perimeter of the main outline of the patch. This represents the area of ​​the main outline of the patch.

4. The carotid plaque detection and evaluation method based on multimodal image feature analysis according to claim 3, characterized in that, The determination of plaque smoothness based on the plaque quantification features and the main contour of the carotid artery specifically includes: Determine whether the aspect ratio of the main outline region of the patch exceeds a preset value, whether the length of the main outline region of the patch exceeds a length threshold, and whether the width of the main outline region of the patch exceeds a width threshold. If the aspect ratio of the main outline region of the patch exceeds the preset value, and the length of the main outline region of the patch exceeds the length threshold, and the width of the main outline region of the patch exceeds the width threshold, then the patch is directly determined to be unsmooth. If the aspect ratio of the main contour region of the plaque does not exceed the preset value, or the length of the main contour region of the plaque does not exceed the length threshold, or the width of the main contour region of the plaque does not exceed the width threshold, then an adaptive threshold is determined based on the main contour of the carotid artery, and the nature of the plaque is determined based on the adaptive threshold. When the patch is hard, the smoothness of the patch is determined by the boundary complexity and the curvature of each point on the main contour of the patch. When the patch is a soft patch, the smoothness of the patch is determined by the aspect ratio of the main contour region, the boundary complexity, and the curvature of each point on the main contour of the patch.

5. The carotid plaque detection and evaluation method based on multimodal image feature analysis according to claim 4, characterized in that, The step of determining an adaptive threshold based on the main contour of the carotid artery and determining the nature of the plaque based on the adaptive threshold specifically includes: The main contour of the carotid artery is expanded outward by M pixels, and the area within the main contour of the carotid artery is removed to obtain an annular reference area; Extract the top 30% of the brightest pixels in the annular reference region, calculate the average pixel value and subtract the preset reference bias value to obtain the adaptive threshold; An adaptive threshold is used to binarize the region within the main contour of the patch, and the number of connected components is counted. When the number of connected components exceeds a certain threshold, the patch is a soft patch; when the number of connected components does not exceed the threshold, the patch is a hard patch.

6. The carotid plaque detection and evaluation method based on multimodal image feature analysis according to claim 4, characterized in that, When the patch is a hard patch, whether the patch is smooth is determined based on the boundary complexity and the curvature of each point on the main contour of the patch. Specifically: Calculate the curvature of each point on the main contour of the patch and take the average value to obtain the mean curvature. Determine whether the boundary complexity exceeds a complexity threshold and whether the mean curvature is less than a curvature threshold; If the boundary complexity exceeds the complexity threshold and the mean curvature is less than the curvature threshold, the patch is determined to be smooth. If the boundary complexity does not exceed the complexity threshold, or the mean curvature is not less than the curvature threshold, the patch is determined to be unsmooth.

7. The carotid plaque detection and evaluation method based on multimodal image feature analysis according to claim 4, characterized in that, When the patch is a soft patch, the smoothness of the patch is determined based on the aspect ratio of the main contour region, the boundary complexity, and the curvature of each point on the main contour of the patch. Specifically: Determine whether the aspect ratio of the main outline region of the patch exceeds a preset value; If the aspect ratio of the main outline region of the patch exceeds a preset value, the patch is determined to be smooth. If the aspect ratio of the main contour region of the patch does not exceed the preset value, the curvature of each point on the main contour of the patch is calculated and the average value is obtained. Determine whether the boundary complexity exceeds a complexity threshold and whether the mean curvature is less than a curvature threshold; If the boundary complexity exceeds the complexity threshold and the mean curvature is less than the curvature threshold, the patch is determined to be smooth. If the boundary complexity does not exceed the complexity threshold, or the mean curvature is not less than the curvature threshold, the patch is determined to be unsmooth.

8. The carotid plaque detection and evaluation method based on multimodal image feature analysis according to claim 1, characterized in that, Also includes: The location of the plaque is determined based on the carotid artery ROI region and the main outline of the plaque, specifically including: The geometric center of the main contour of the patch is calculated using the image spatial distance. The calculation method is as follows: ,in, These are the x and y coordinates of the geometric center of the main contour of the patch, respectively. The area of ​​the main outline region of the patch. The 10th spatial distance of the image within the main contour region of the patch. The 0-1 spatial distance of the image within the main contour region of the patch. Within the main outline area of ​​the patch Pixel value at; The location of the midline of the carotid artery ROI region is calculated as follows: ,in, This indicates the location of the midline of the carotid artery ROI region. The vertical coordinate is the upper left corner of the carotid artery ROI region. The height of the carotid artery ROI region; When the ordinate of the geometric center of the main contour of the plaque is greater than the midline of the carotid ROI region minus the tolerance value, the plaque is located on the anterior wall of the carotid artery. When the ordinate of the geometric center of the main contour of the plaque is less than the location of the midline of the carotid ROI region plus the tolerance value, the plaque is located on the posterior wall of the carotid artery. In other cases, the plaque is located on the lateral wall of the carotid artery; The tolerance value is the product of the height of the carotid artery ROI region and the scaling factor.

9. A carotid plaque detection and evaluation system based on multimodal image feature analysis, characterized in that, include: The ROI detection module is used to preprocess the carotid ultrasound image and input the preprocessed carotid ultrasound image into the target localization model to obtain the carotid ROI region and plaque ROI region; The segmentation processing module is used to input the carotid artery ROI region and the plaque ROI region into the corresponding structural segmentation model respectively to obtain the carotid artery binary mask image and the plaque binary mask image; The target contour extraction module is used to extract all outer contours from the carotid artery binary mask image and the plaque binary mask image using a contour extraction algorithm, and select the outer contour with the largest area as the main contour of the carotid artery and the main contour of the plaque. The patch feature calculation module is used to calculate patch quantization features based on the main outline of the patch. The plaque smoothness determination module is used to determine the smoothness of plaques based on the plaque quantification features and the main contour of the carotid artery.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the carotid plaque detection and evaluation method based on multimodal image feature analysis as described in any one of claims 1-8.

Images