A method and system for quantification and classification of musculoskeletal ultrasound ultramicro blood flow based on optical flow analysis

Abstract

The present invention discloses a method and system for quantitatively grading musculoskeletal ultrasound ultramicro blood flow based on optical flow analysis, the method comprising: obtaining an original ultramicro blood flow image sequence of an object to be detected; preprocessing the original ultramicro blood flow image sequence; extracting a region of interest, i.e., a ROI region, from the ultramicro blood flow image sequence after preprocessing; extracting ultramicro blood vessel morphology from the ROI region; dynamically analyzing the ultramicro blood flow in the ROI region, the dynamic analysis comprising: ultramicro blood flow calculation, motion optical flow field estimation, and fusion of microblood flow morphological features; generating two indicators, motion density of ultramicro blood flow and elastic density of ultramicro blood vessels, according to the dynamic analysis results; and grading the blood flow signal of the region of interest according to the motion density and elastic density. The method of the present invention improves the accuracy of quantitative grading of musculoskeletal ultrasound ultramicro blood flow.

Description

Technical Field

[0001] The invention belongs to the field of biomedical imaging, and in particular relates to a method and system for quantifying and grading muscle-bone ultrasonic ultramicro blood flow based on optical flow analysis. Background Art

[0002] Superb Microvascular Imaging (SMI) is a new ultrasound imaging technology that effectively reduces image interference caused by weak blood flow signals and motion artifacts by optimizing the detection of tissue movement, allowing microblood flow to be clearly visualized in ultrasound images. Compared with traditional blood flow imaging technologies (such as color Doppler and power Doppler), SMI has significantly improved its ability to detect low-speed blood flow, especially when showing blood flow activity in tiny blood vessels, with higher spatial and temporal resolution. Specifically, SMI technology shows better sensitivity and accuracy than conventional ultrasound imaging methods in evaluating low-grade blood flow signals in synovial inflammation of joints.

[0003] The blood flow signal of the synovial membrane is an important indicator in the diagnosis of joint inflammation, and ultramicroscopic blood flow analysis based on SMI can provide more accurate quantitative evaluation than traditional technologies. Studies have shown that SMI can detect low-grade inflammation of the synovial membrane better than energy Doppler, and has a significant correlation with X-ray imaging features and MRI test results. Especially in the diagnosis of rheumatoid arthritis, osteoarthritis and other inflammatory joint diseases, SMI has shown a high clinical value.

[0004] However, traditional ultramicro blood flow imaging technology still faces some challenges in its application, mainly reflected in the lack of quantitative analysis and automated grading of blood flow signals. Most of the current ultramicro blood flow quantification methods rely on semi-quantitative rating systems, such as the Szkudlarek semi-quantitative grading system, which divides ultramicro blood flow signals into 0 to 3 levels. This system mainly relies on the physician's experience and judgment, which is highly subjective and affects the consistency and repeatability of the results. In addition, the dynamic changes of ultramicro blood flow in joints are often accompanied by arteriovenous pulsation, and the analysis of static images tends to ignore this change, thus affecting the accuracy of blood flow quantification.

[0005] In order to overcome these problems, dynamic blood flow quantification methods based on motion estimation and optical flow analysis have begun to attract attention in recent years. As a motion estimation method based on image sequences, the optical flow model can accurately capture the dynamic changes of blood flow at different time points. By calculating the optical flow field in the image, the motion information of the blood flow can be extracted, thereby realizing the real-time dynamic quantification of ultra-micro blood flow signals. Optical flow analysis can not only improve the accuracy of blood flow quantification, but also effectively reduce the impact of noise and artifacts on image quality, especially in the case of low signal-to-noise ratio, showing excellent anti-interference ability.

[0006] In the field of automatic grading of joint osteoarthritis, recent research has focused on methods based on machine learning and deep learning, which have made some progress by combining ultrasound image features and clinical data. However, despite the significant progress made by deep learning methods in image analysis, traditional optical flow models still play an important role in motion estimation and dynamic blood flow analysis. Optical flow models can accurately capture the dynamic changes of ultramicro blood flow and provide quantitative indicators for inflammation grading. Compared with machine learning methods, traditional optical flow models have significant advantages in noise suppression, processing efficiency, stability and real-time performance in clinical environments. Summary of the invention

[0007] In view of this, the object of the present invention is to provide a method and system for quantification and grading of muscle-bone ultrasonic ultramicro blood flow based on optical flow analysis.

[0008] The objective of the present invention is achieved through the following technical solutions:

[0009] A method for quantitatively grading muscle-bone ultrasound ultramicro blood flow based on optical flow analysis, comprising:

[0010] Acquire an original ultramicro blood flow image sequence of the object to be detected;

[0011] Preprocessing is performed from the original ultramicroscopic blood flow image sequence;

[0012] Extracting the region of interest, i.e., the ROI region, from the preprocessed ultramicroscopic blood flow image sequence;

[0013] Extract ultramicrovascular morphology in the ROI area;

[0014] Performing dynamic analysis on the ultramicro blood flow in the ROI area, the dynamic analysis including: ultramicro blood flow calculation and motion optical flow field estimation and fusion of microblood flow morphological features;

[0015] According to the dynamic analysis results, two indicators are generated: the motion density of ultramicro blood flow and the elastic density of ultramicro blood vessels. The motion density of ultramicro blood flow refers to the ratio of the motion optical flow of the optical flow component parallel to the blood flow signal skeleton at all pixel points in two adjacent frames to the ultramicro blood flow. The elastic density of ultramicro blood vessels refers to the ratio of the motion optical flow of the optical flow component perpendicular to the blood flow signal skeleton at all pixel points in two adjacent frames to the ultramicro blood flow.

[0016] The blood flow signals in the region of interest were graded according to the motion density and elastic density.

[0017] Furthermore, the ultra-microvascular morphology extraction calculation includes:

[0018] Performing grayscale conversion and binarization processing on each frame of the original ultramicro blood flow image sequence, setting the pixel values ​​of the pixels whose grayscale values ​​are greater than the preset grayscale threshold to 1, and the others to 0, thereby generating a binary image;

[0019] Extract the ROI region from each binary image to retain the target blood flow signal;

[0020] Segment and cluster the blood flow area for each binary image;

[0021] Perform edge detection and skeleton extraction on the blood flow area obtained by segmentation and clustering to generate a skeleton image;

[0022] The skeleton in the skeleton image is subjected to curve fitting to obtain microvascular parameters.

[0023] Furthermore, segmenting the blood flow area of ​​the binary image includes:

[0024] Step 11: Initialize the queue;

[0025] Step 12: Traverse each pixel in the ROI area and mark the pixel with a pixel value of 1 as the target pixel;

[0026] Step 13: Add the target pixel to the queue;

[0027] Step 14: When the queue is not empty, remove the first element in the queue and use the removed element as the current pixel, then execute step 15;

[0028] Step 15: Obtain neighboring pixels within a preset range of the current pixel. For each neighboring pixel, if the neighboring pixel value is 1 and has not been marked, mark the neighboring pixel as a target pixel and add the neighboring pixel to a queue until all neighboring pixels are checked;

[0029] Step 16: When the queue is empty, the connected target pixels are grouped into an independent blood flow area;

[0030] Step 17: Repeat steps 12-16 until all target pixels are processed.

[0031] Furthermore, blood flow region clustering includes:

[0032] Step 21: Obtain a blood flow area set;

[0033] Step 22: Calculate the distance between every two blood flow areas in the blood flow area set;

[0034] Step 23: When there are two or more blood flow regions in the blood flow region set, clustering and merging can be performed, and steps 24-25 are executed;

[0035] Step 24: If the distance between the blood flow areas is less than a preset threshold, the two blood flow areas are merged into one blood flow area;

[0036] Step 25: Update the blood flow area set using the merged blood flow area, and return to step 23;

[0037] Step 26: When there are no more blood flow regions that can be merged, a clustered blood flow region set is obtained.

[0038] Furthermore, edge detection and skeleton extraction include:

[0039] For each pixel point in the blood flow area obtained by segmentation and clustering, if the pixel value of the pixel point is 1, and the pixel values ​​of the pixels in the neighborhood within the preset range of the pixel point are all 1, it means that the pixel point is completely surrounded, and the pixel value of the pixel point is changed to 0, that is, the pixel is corroded;

[0040] After processing each pixel in the binary image, a skeleton image is obtained.

[0041] Furthermore, the motion optical flow field estimation includes:

[0042] After the blood flow area of ​​each binary image is determined, the two consecutive binary images are divided into a group, and the optical flow is calculated for each group of binary images to obtain the optical flow corresponding to each group of binary images;

[0043] The motion density of the ultramicro blood flow and the elastic density of the ultramicro blood vessels corresponding to each set of binary images are determined, thereby obtaining a set of motion densities of the ultramicro blood flow and a set of elastic densities of the ultramicro blood vessels.

[0044] Furthermore, the blood flow signal of the region of interest is graded according to the motion density and elastic density, including:

[0045] According to the movement density of ultramicro blood flow, the average value and the maximum value of the movement density of ultramicro blood flow are obtained;

[0046] According to the elastic density set of ultramicrovessels, the average value and maximum value of the elastic density of ultramicrovessels are obtained;

[0047] The blood flow signals were graded according to the average and maximum values ​​of the motion density of the ultramicro blood flow and the average and maximum values ​​of the elastic density of the ultramicrovessels.

[0048] The present invention also provides a muscle-bone ultrasound ultra-micro blood flow quantification and grading system based on optical flow analysis, comprising:

[0049] An image acquisition module, used to acquire an original ultramicro blood flow image sequence of the object to be detected;

[0050] A preprocessing module, which preprocesses the original ultramicro blood flow image sequence;

[0051] The ROI region extraction module is used to extract the region of interest, i.e., the ROI region, from the ultramicro blood flow image sequence after preprocessing;

[0052] Ultra-microvascular morphology extraction module, used to extract ultra-microvascular morphology from ROI area;

[0053] A dynamic analysis module is used to dynamically analyze the ultramicro blood flow in the ROI area, wherein the dynamic analysis includes ultramicro blood flow calculation, motion optical flow field estimation, and fusion of microblood flow morphological features;

[0054] The index determination module is used to generate two indexes, namely, the motion density of ultramicro blood flow and the elastic density of ultramicro blood vessels, according to the dynamic analysis results. The motion density of ultramicro blood flow refers to the ratio of the motion optical flow of the optical flow component parallel to the blood flow signal skeleton at all pixel points in two adjacent frames to the ultramicro blood flow, and the elastic density of ultramicro blood vessels refers to the ratio of the motion optical flow of the optical flow component perpendicular to the blood flow signal skeleton at all pixel points in two adjacent frames to the ultramicro blood flow.

[0055] The classification module is used to classify the blood flow signal of the region of interest according to the motion density and elastic density.

[0056] Furthermore, the ultramicrovascular morphology extraction module is also used for:

[0057] Performing grayscale conversion and binarization processing on each frame of the original ultramicro blood flow image sequence, setting the pixel values ​​of the pixels whose grayscale values ​​are greater than the preset grayscale threshold to 1, and the others to 0, thereby generating a binary image;

[0058] Extract the ROI region from each binary image to retain the target blood flow signal;

[0059] Segment and cluster the blood flow area for each binary image;

[0060] Perform edge detection and skeleton extraction on the blood flow area obtained by segmentation and clustering to generate a skeleton image;

[0061] The skeleton in the skeleton image is subjected to curve fitting to obtain microvascular parameters.

[0062] Furthermore, the grading module grades the blood flow signal of the region of interest according to the motion density and elastic density, including:

[0063] Determine the average and maximum values ​​of the motion density of ultramicroscopic blood flow;

[0064] Determine the mean and maximum values ​​of the elastic density of ultramicrovessels;

[0065] The blood flow signals are graded according to the average and maximum values ​​of the motion density of the ultramicro blood flow and the average and maximum values ​​of the elastic density of the ultramicro blood vessels.

[0066] The beneficial effects of the present invention are:

[0067] The present invention is based on the dynamic analysis of optical flow, with motion density (SMI-MD) and elastic density (SMI-ED) as core indicators, which improves the accuracy of blood flow quantification, and realizes multi-dimensional accurate evaluation and grading of ultramicro blood flow in joints by quantifying the dynamic characteristics of blood flow signals. It also effectively reduces the subjective influence of operators and readers, provides a reliable and objective tool for the non-invasive diagnosis of joint osteoarthritis, and provides more comprehensive support for the early diagnosis of joint inflammation, efficacy evaluation and disease management.

[0068] Other advantages, objectives and features of the present invention will be described in the following description to some extent, and to some extent, will be obvious to those skilled in the art based on the following examination and study, or can be taught from the practice of the present invention. The objectives and other advantages of the present invention can be realized and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be further described in detail below with reference to the accompanying drawings, in which:

[0070] Figure 1 is a schematic flow chart of the method for quantitative grading of musculoskeletal ultrasound ultramicro blood flow based on optical flow analysis;

[0071] Figure 2 It is a schematic diagram of ultramicrovascular morphology extraction;

[0072] Figure 3 It is a schematic diagram of neighborhood segmentation search;

[0073] Figure 4 It is the classification of blood flow signals in patients with different symptoms;

[0074] Figure 5 is the fluctuation curve of SMI-MD over time;

[0075] Figure 6 is the fluctuation curve of SMI-ED over time;

[0076] Figure 7 Schematic diagram of the musculoskeletal ultrasound ultramicroblood flow quantification grading system based on optical flow analysis. DETAILED DESCRIPTION

[0077] The preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the preferred embodiments are only for illustrating the present invention, rather than for limiting the protection scope of the present invention.

[0078] The present application proposes a method for quantifying and grading muscle-bone ultrasonic ultramicro blood flow based on optical flow analysis, which can be used for quantifying and evaluating ultramicro blood flow signals in joints. Figure 1 This is a schematic flow chart of the method for quantitative grading of musculoskeletal ultrasound ultramicro blood flow based on optical flow analysis. Figure 1 As shown, the method includes:

[0079] Acquire the original ultramicro blood flow image sequence of the object to be detected, wherein the collected ultrasonic ultramicro blood flow image can be used as the original data, decoded by the recording frame rate, and generate the original ultramicro blood flow image (or "image") sequence, which includes multiple continuous images;

[0080] Preprocessing is performed on the original ultramicro blood flow image sequence, which may include image enhancement (such as over-enhancement, appropriate enhancement, under-enhancement, etc.), image denoising, etc. In the image enhancement process, the blood flow signal in the original image is separated into a red blood flow signal channel and a blue blood flow signal channel. After the respective enhancement processing, the images of the red blood flow signal channel and the blue blood flow signal channel can be fused by superposition (for example, superposition of the overlapping area of ​​the red channel and the blue channel), thereby further enhancing the visualization effect of the blood flow signal, improving the image detail performance, and providing clearer basic data for subsequent dynamic analysis and feature extraction. The fused image sequence will be gray-scale adjusted to optimize the visual effect, adjust the dynamic range of the signal, make the blood flow signal more prominent in the grayscale image, and provide high-quality input data for subsequent analysis and processing.

[0081] Next, a region of interest, i.e., a ROI region, is extracted from the preprocessed ultramicroscopic blood flow image sequence, wherein the ROI region refers to a region that may contain blood flow signals;

[0082] Extract ultramicrovascular morphology in the ROI area;

[0083] The ultramicro blood flow in the ROI area is dynamically analyzed, and the dynamic analysis includes: ultramicro blood flow calculation, motion optical flow field estimation, and fusion of microblood flow morphological features, wherein the optical flow field obtained after the optical flow field estimation includes an optical flow component along the blood flow motion direction (i.e., along the blood vessel skeleton direction) and an optical flow component perpendicular to the blood flow motion direction (i.e., perpendicular to the blood vessel skeleton direction), and the microblood flow morphological features are determined according to the extracted ultramicrovascular morphology, and the microblood flow morphological features include motion features along the blood flow motion direction and elastic features perpendicular to the blood flow motion direction;

[0084] Next, based on the dynamic analysis results, two indicators are generated: the motion density of ultramicro blood flow and the elastic density of ultramicro blood vessels. The motion density of ultramicro blood flow refers to the ratio of the motion optical flow of the optical flow component parallel to the blood flow signal skeleton at all pixel points in two adjacent frames to the ultramicro blood flow. The elastic density of ultramicro blood vessels refers to the ratio of the motion optical flow of the optical flow component perpendicular to the blood flow signal skeleton at all pixel points in two adjacent frames to the ultramicro blood flow.

[0085] Finally, the blood flow signals in the region of interest were graded according to the motion density and elastic density.

[0086] Figure 7 This is a schematic diagram of the quantification and grading system of muscle-bone ultrasound ultra-micro blood flow based on optical flow analysis. Figure 7 As shown, the grading system includes an image acquisition module, a preprocessing module, a ROI region extraction module, an ultramicrovascular morphology extraction module, a dynamic analysis module and a grading module.

[0087] An image acquisition module, used to acquire an original ultramicro blood flow image sequence of the object to be detected;

[0088] A preprocessing module, which preprocesses the original ultramicro blood flow image sequence;

[0089] ROI region extraction module, used to extract the region of interest, i.e., ROI region, from the ultramicro blood flow image sequence after preprocessing;

[0090] Ultra-microvascular morphology extraction module, used to extract ultra-microvascular morphology from ROI area;

[0091] A dynamic analysis module is used to dynamically analyze the ultramicro blood flow in the ROI area, and the dynamic analysis includes: ultramicro blood flow calculation, motion optical flow field estimation, and fusion of microblood flow morphological features, wherein the optical flow field obtained after the optical flow field estimation includes an optical flow component along the blood flow motion direction (i.e., along the blood vessel skeleton direction) and an optical flow component perpendicular to the blood flow motion direction (i.e., perpendicular to the blood vessel skeleton direction), and the microblood flow morphological features are determined according to the extracted ultramicrovascular morphology, and the microblood flow morphological features include motion features along the blood flow motion direction and elastic features perpendicular to the blood flow motion direction;

[0092] The index determination module is used to generate two indexes, namely, the motion density of ultramicro blood flow and the elastic density of ultramicro blood vessels, according to the dynamic analysis results. The motion density of ultramicro blood flow refers to the ratio of the motion optical flow of the optical flow component parallel to the blood flow signal skeleton at all pixel points in two adjacent frames to the ultramicro blood flow, and the elastic density of ultramicro blood vessels refers to the ratio of the motion optical flow of the optical flow component perpendicular to the blood flow signal skeleton at all pixel points in two adjacent frames to the ultramicro blood flow.

[0093] The classification module is used to classify the blood flow signal of the region of interest according to the motion density and elastic density.

[0094] Figure 2 is a schematic diagram of ultramicrovascular morphology extraction, such as Figure 2 As shown, the extraction of ultramicrovascular morphology may include the following steps:

[0095] Each frame of the original ultramicro blood flow image sequence is converted into grayscale and binarized to generate a binary image, such as Figure 2 As shown in Figure A in the figure. For a binary image, the pixel values ​​of pixels whose grayscale values ​​exceed the preset grayscale threshold are set to 1, and the pixel values ​​of the remaining pixels are set to 0;

[0096] Extract the ROI region from each binary image to retain the target blood flow signal;

[0097] Segment and cluster the blood flow area for each binary image, such as Figure 2 As shown in Figure B;

[0098] Perform edge detection and skeleton extraction on the blood flow area obtained by segmentation and clustering (the blood flow area is still a binary image) to generate a skeleton image; Figure 2 As shown in Figure C;

[0099] The skeleton in the skeleton image is subjected to curve fitting (e.g., fitting by the least square method) to obtain microvascular parameters, such as Figure 2 As shown in Figure D.

[0100] Among them, the ultra-microvascular morphology includes parameters such as microvascular blood flow direction and microvascular width. The microvascular blood flow direction refers to the tangent direction of the blood flow signal per pixel; the microvascular width refers to the difference between the upper and lower limits of the blood flow signal per pixel perpendicular to the tangent direction, where the upper and lower limits correspond to the upper and lower walls of the blood flow signal, respectively.

[0101] According to the extracted skeleton and the curve fitting of the skeleton, the slope of each pixel point and the slope of each position of the blood vessel can be obtained, and the slope is the micro-blood flow morphological feature. Among them, the micro-blood flow morphological feature can be divided into morphological motion feature and morphological elastic feature. The morphological motion feature reflects the dynamic behavior of blood flow, and the morphological elastic feature can characterize the local deformation ability of blood flow. Therefore, the component of the slope parallel to the skeleton (i.e., the direction of blood flow) obtained by curve fitting according to the skeleton is used as the morphological elastic feature, and the component of the slope perpendicular to the skeleton (i.e., the direction of blood flow) obtained by curve fitting according to the skeleton is used as the morphological elastic feature. Further, according to the slope, the optical flow can be decomposed into the optical flow component along the blood flow direction and the optical flow component perpendicular to the blood flow direction, and the optical flow along the blood flow direction and the optical flow perpendicular to the blood flow direction are recorded as motion features and elastic features respectively. Then, the motion density and elastic density can be determined using the motion features and elastic features.

[0102] The segmentation and clustering of blood flow regions are described in detail below.

[0103] The specific method of the blood flow area can refer to Algorithm 1 in Table 1. The specific method is: traverse each pixel in the image, identify the target pixel with a value of 1, and add it to the queue. Then, use the breadth-first search algorithm (BFS) to analyze the neighborhood pixels (e.g., 8 neighborhood pixels) within a preset range of the target pixel. If the neighborhood pixel value is 1 and has not been marked, it will be added to the queue. The neighborhood segmentation search is as follows: Figure 3 This process continues until the queue is empty, and all connected pixel sets are stored as a separate blood flow region. This segmentation operation is repeated throughout the image until all target pixels are labeled and classified into different blood flow regions.

[0104] Table 1 Pseudocode for blood flow region segmentation

[0105]

[0106] Specifically, the segmentation of the blood flow area of ​​the binary image can be summarized into the following steps:

[0107] Step 11: Initialize the queue;

[0108] Step 12: Traverse each pixel in the ROI area and mark the pixel with a pixel value of 1 as the target pixel;

[0109] Step 13: Add the target pixel to the queue;

[0110] Step 14: When the queue is not empty, remove the first element in the queue and use the removed element as the current pixel, then execute step 15;

[0111] Step 15: Obtain neighboring pixels within a preset range of the current pixel. For each neighboring pixel, if the neighboring pixel value is 1 and has not been marked, mark the neighboring pixel as a target pixel and add the neighboring pixel to a queue until all neighboring pixels are checked;

[0112] Step 16: When the queue is empty (i.e. connected), the connected target pixels are combined into an independent blood flow area;

[0113] Step 17: Repeat steps 12-16 until all target pixels are processed.

[0114] Next, the separated blood flow regions are clustered using Algorithm 2. First, the distance between each pair of regions is calculated, and then the distance threshold is used to determine whether to merge adjacent regions. If the distance meets the conditions, the two regions are merged into a new cluster. This operation is repeated until there are no more regions that can be merged. Finally, the clustered blood flow region set is returned for subsequent analysis. For details, please refer to the algorithm in Table 2.

[0115] Table 2 Pseudocode for blood flow region clustering

[0116]

[0117] Specifically, clustering blood flow areas may include the following steps:

[0118] Step 21: Obtain a blood flow area set;

[0119] Step 22: Calculate the distance between every two blood flow areas in the blood flow area set;

[0120] Step 23: When there are two or more blood flow regions in the blood flow region set, clustering and merging can be performed, and steps 24-25 are executed;

[0121] Step 24: If the distance between the blood flow areas is less than a preset threshold, the two blood flow areas are merged into one blood flow area;

[0122] Step 25: Update the blood flow area set using the merged blood flow area, and return to step 23;

[0123] Step 26: When there are no more blood flow regions that can be merged, a clustered blood flow region set is obtained.

[0124] After the blood vessel region segmentation and clustering, it is necessary to further perform edge detection and skeleton extraction on the blood vessel region, such as Figure 2As shown in Figure C. In some embodiments of the present invention, the skeleton extraction of the image can be achieved through edge detection and corrosion operations to extract the core skeleton features that can be used for curve fitting. For each pixel in the binary image (i.e., the result obtained based on vascular region segmentation and clustering), if the pixel value of the pixel is 1, and the pixel values ​​of the pixels in the neighborhood within the preset range of the pixel are all 1, it means that the pixel is completely surrounded, and the pixel value of the pixel is changed to 0, that is, the pixel is corroded; after processing each pixel in the binary image (i.e., the result obtained based on vascular region segmentation and clustering), a skeleton image is obtained.

[0125] For example, initialize the input image I , and get its size m × n , and then traverse the image pixel by pixel, and when the pixel value is detected to be 1, initialize the variable sum = 0, then check the neighboring pixels of the pixel (for example, the pixel values ​​of the 8-neighborhood). Add up all the pixel values ​​in the 8-neighborhood. If sum =8, which means that the pixel is completely surrounded, then it is set to 0 (that is, the pixel is corroded). Through multiple rounds of iterations, the image is gradually corroded until only the skeleton structure remains. Finally, the extracted skeleton image is returned I’ ,This skeleton retains the main features of blood flow morphology and can be used for subsequent curve fitting analysis. The specific algorithm is shown in Table 3.

[0126] Table 3 Pseudocode for edge detection and erosion skeleton extraction

[0127]

[0128] Taking into account the low quality of ultrasound images and the characteristics of ultra-micro blood flow changes, the present invention uses the optical flow method for motion estimation to extract the dynamic characteristics of blood vessels, so as to facilitate high-resolution motion field calculation. The optical flow method is a visual analysis method, which refers to the relative motion speed of pixels of a moving object on the observation imaging surface. The relationship between the motion and structure of the object in the scene is characterized by the change of the image grayscale on the time axis. In order to overcome the noise of ultrasound images and improve the accuracy of motion estimation, the relevant dynamic change assumptions of grayscale, gradient, and optical flow field and multi-resolution analysis technology are integrated into the motion field reconstruction model. According to the assumptions of constant grayscale, constant gradient and smooth optical flow field, the corresponding deviation penalty function is obtained, and the energy function for calculating the optical flow can be directly obtained. Among them, the energy functions of grayscale and gradient are related to the image data itself, and are classified as data items in the total energy function, while the smooth optical flow field assumption establishes corresponding smooth items in the energy function.

[0129] In this way, the total energy function consists of the data term and the optical flow field smoothing term, as shown in formula (1).

[0130] (1)

[0131] in α is the regularization parameter, E Data and E Smooth are data term and smoothing term respectively, as shown in formula (2).

[0132] (2)

[0133] in, I represents an image sequence; represents the coordinate vector; express The displacement vector at , u represents the horizontal displacement, and v represents the vertical displacement; represents the spatial gradient operator; Represents the weight between the grayscale hypothesis and the gradient hypothesis; is a concave function, which strengthens the penalty for isolated points through quadratic square, thereby improving the robustness of the algorithm. ε is a constant in numerical operations. According to the coverage domain of the binary image and the optical flow method, the influence of the noise background is removed and the optical flow rate of the blood flow is calculated. The coverage domain of the optical flow method refers to the pixel area that can be effectively calculated and tracked when using optical flow technology to detect motion in a video or image sequence. This technology reflects its changes over time by estimating the motion vector of the pixel point. The size and accuracy of the coverage domain depend on factors such as image resolution, camera motion, algorithm implementation, and pixel matching efficiency. Under ideal conditions, the coverage domain can cover most of the moving areas in the image. By performing element-by-element dot multiplication of the binary image matrix obtained by cluster segmentation and the optical flow coverage domain matrix, the area with a pixel value of 1 indicates the presence of a reliable, non-noise blood flow signal.

[0134] Therefore, after the blood flow area of ​​each binary image is determined, two consecutive binary images are grouped together, and the optical flow is calculated for each group of binary images to obtain the optical flow corresponding to each group of binary images.

[0135] The ultramicro blood flow dynamic analysis (SMI Dynamic Parameter, SMI-DP) indexes involved in the present invention include blood flow motion density index and vascular elasticity density index.

[0136] The ratio of the motion optical flow of all pixels of the optical flow component parallel to the blood flow signal skeleton in two adjacent frames to the ultra-micro blood flow (i.e., the total number of pixels in the vascular area) is used as the ultra-micro blood flow motion density (SMI Motion Density, SMI-MD) of a certain frame, and the fluctuation curve of SMI-MD over time is drawn. The SMI-MD index reflects the overall dynamic characteristics of blood flow by measuring the motion distribution density of the vascular skeleton in adjacent frames. It comprehensively considers the blood flow velocity, direction, and consistency of motion in the blood flow area. It is a quantitative index used to analyze the motion density of micro blood flow, as shown in formula (3).

[0137] (3)

[0138] in, Represents the pixel points along the vascular skeleton in adjacent frames The optical flow component in the direction of motion. It represents the ultramicro blood flow, i.e. the total number of pixels in the vascular area, and is used to normalize the accumulated dynamic intensity.

[0139] Considering that the intraluminal area of ​​the microvascular cross section is very small, the present invention uses the width variation characteristics of the micro blood flow in the cross section to characterize the elastic characteristics of the blood vessel. Therefore, the ratio of the movement of the optical flow component perpendicular to the blood flow signal skeleton at all pixel points in two adjacent frames to the total amount of ultramicro blood flow (i.e., the total number of pixels in the blood vessel area) can be defined as the ultramicrovascular elastic density (SMI Elastic Density, SMI-ED) of a certain frame. Based on this, a fluctuation curve of SMI-ED over time can be drawn. This indicator is used to characterize the distribution characteristics of vascular elastic density during the dynamic change of blood flow, so as to quantitatively describe the dynamic elastic characteristics of the blood flow signal, as shown in formula (4).

[0140] (4)

[0141] in Indicates the pixel points perpendicular to the vascular skeleton in adjacent frames The optical flow component in the direction of motion. These parameters provide important basis for subsequent pathological diagnosis and joint inflammation assessment.

[0142] The present invention proposes an automatic classification method for joint ultramicro blood flow signals based on musculoskeletal ultrasonic ultramicro blood flow dynamic analysis (SMI-DP). With motion density (SMI-MD) and elastic density (SMI-ED) as core indicators, the dynamic characteristics of blood flow signals are quantified to achieve multi-dimensional accurate evaluation and classification of ultramicro blood flow in joints.

[0143] Specifically, the motion density of the ultramicro blood flow and the elastic density of the ultramicro blood vessels corresponding to each set of binary images can be determined according to the optical flow corresponding to each set of binary images, thereby obtaining a set of motion densities of the ultramicro blood flow and a set of elastic densities of the ultramicro blood vessels;

[0144] Then, the average value V of the ultramicro blood flow motion density is obtained based on the ultramicro blood flow motion density set. MD ave and maximum value V MD max ;

[0145] Next, the average value V of the elastic density of the ultramicrovessels is obtained based on the elastic density set of the ultramicrovessels. ED ave and maximum value V ED max ; That is, V MD ave represents the average value of the SMI-MD fluctuation curve, V MD max Indicates the maximum value of the SMI-MD fluctuation curve, V ED ave represents the average value of the SMI-ED fluctuation curve, V ED max It represents the maximum value of the SMI-ED fluctuation curve;

[0146] Finally, the blood flow signals were graded according to the average and maximum values ​​of the motion density of the ultramicro blood flow and the average and maximum values ​​of the elastic density of the ultramicrovessels.

[0147] In the automatic grading process, the above two indicators must meet the standards of the corresponding levels at the same time (strict matching principle). The specific grading standards are as follows: Level 0 means that there is no obvious fluctuation in the ultrafine blood flow signal, and the motion and elastic density are both low; Level 1 means that the blood flow signal shows mild fluctuations, and the motion and elastic density are slightly increased; Level 2 means that the blood flow signal fluctuates significantly, and the motion and elastic density are at a medium level; Level 3 means that the blood flow signal fluctuates extremely significantly, and the motion and elastic density reach a high level. In addition, when the levels of the two indicators are inconsistent, the higher level is used as the final grading result to ensure high sensitivity to abnormal blood flow signals. The specific grading standards are detailed in Table 4.

[0148] Table 4 Automatic classification standard of joint ultramicroscopic blood flow signals based on SMI-DP

[0149]

[0150] It should be noted that the setting of the thresholds of the various indicators in Table 4 is only exemplary and is not intended to limit the present invention.

[0151] The advantages of the present invention are described below through specific experiments.

[0152] This experiment collected ultrasound ultramicroscopic blood flow imaging data of patients with joint pain who visited the hospital from June to September 2024. Inclusion criteria: joint pain for one day or more in the past month, and blood flow signals visible in ultrasound ultramicroscopic blood flow mode. Exclusion criteria: patients who had undergone joint replacement surgery before. A total of 47 subjects were included, including 20 males and 27 females.

[0153] Canon Aplio i800 is a high-end ultrasound imaging device that provides high-resolution images and real-time dynamic image functions, and is particularly suitable for microblood flow and vascular imaging. The SMI technology of this device can significantly improve the detection ability of microblood flow signals and is suitable for the evaluation of joint inflammation and synovial lesions. In this experiment, all patients were examined using Canon Aplioi800, and the L15-3WU linear array probe with a frequency of 15 MHz was selected. During the examination, the patient was either seated or placed on a sponge pad according to the location of the pain, so that the examination site was fully exposed. Routine ultrasound scanning of the painful joint was performed to observe the general characteristics of the painful joint. If there were multiple lesions, the largest lesion was selected for detailed examination. When the CDFI mode showed the section with the richest blood flow signals in the joint inflammation, switch to the color ultramicro blood flow (cUMA) mode, with a blood flow velocity scale of 3 cm / s, a color frequency of 4 MHz, and a frame rate of 30 Hz. The gain was adjusted to the best imaging, and the images and dynamic videos were stored. During the measurement, the region of interest (ROI) was minimized while covering the area to be examined to improve the sensitivity of blood flow measurement. All imaging data were archived for later dynamic evaluation of ultramicro blood flow. In order to ensure the consistency of the basic flow during the quantification of ultramicro blood flow indicators, this experiment selected ROI areas of equal area for research.

[0154] Figure 4 Five ultrasound images of patients with different symptoms are shown in the Figure 4 Figure A in the middle shows the ultrasound image of a patient with ultramicrovascular grading of grade 0. Figure 4 Figure B shows the ultrasound image of a patient with ultramicrovascular grade 1. Figure 4 Figure C shows the ultrasound image of a patient with ultramicrovascular grade 2. Figure 4 D Figure and Figure 4 Figure E shows the ultrasound image of a patient with grade 3 ultramicrovascular flow. Figure 4 In the figure, the green rectangle marks the target area, which reflects the blood flow distribution and pathological manifestations of different anatomical parts. These images are graded according to the distribution density of blood flow signals, synovial characteristics, joint effusion and the severity of soft tissue lesions, showing an increasing trend from mild to severe (A is 0, B is 1, C is 2, D and E are 3).

[0155] Figure 4 Figure A shows effusion in the suprapatellar bursa and medial and lateral joint spaces of the left knee (2.33×0.48 cm, 1.66×0.38 cm, 1.36×0.38 cm), accompanied by synovial thickening and hypoechoicity, indicating mild inflammatory changes in the synovium. No obvious blood flow signal was seen, indicating that local vascular proliferation was not significant and the degree of osteoarthritis was mild. At the same time, the lateral collateral ligament of the left knee was thickened and the fiber structure was blurred, indicating the possibility of ligament injury. The comprehensive judgment was grade 0.

[0156] Figure 4 Figure B shows effusion in the suprapatellar bursa (0.61×0.17 cm), medial (0.84×0.30 cm) and lateral joint spaces (0.58×0.34 cm) of the left knee, and thickened synovium with uneven echoes, indicating inflammatory lesions. A small amount of blood flow signals indicates that the inflammation is initially active. Figure 4 The blood flow signal in Figure A has increased, which is judged as grade 1.

[0157] Figure 4 Figure C shows a hypoechoic area at the proximal attachment of the right popliteal longus muscle, accompanied by abundant blood flow signals, indicating local tear injury. The fuzzy fiber structure indicates mechanical injury with local inflammatory response, and the richness of blood flow signals reflects the high dynamic activity of inflammation. Although the lesion is limited in scope, the blood flow signal is significantly enhanced and classified as grade 2.

[0158] Figure 4 Figure D shows effusion in the left knee joint cavity and the medial and lateral joint spaces (0.98×0.23 cm, 0.81×0.38 cm, 2.86×0.59 cm), accompanied by significant thickening and hypoechoic synovium, indicating osteoarthritis. Soft tissue edema at the lower edge of the patella, rich blood flow signals and blurred structure, indicating significant changes in synovial and soft tissue blood supply. The lesion is extensive and the inflammatory reaction is active, which is judged to be grade 3.

[0159] Figure 4 Figure E in the figure shows thickening of the prepatellar fat layer and a sheet-like hypoechoic area (3.03×0.88 cm) of the right knee, accompanied by abundant blood flow signals; effusion in the suprapatellar bursa and medial and lateral joint spaces (1.76×0.32 cm, 1.23×0.28 cm, 0.74×0.28 cm). The dense distribution of blood flow signals indicates significant inflammatory changes in the synovium and soft tissues and joint cavity effusion, and the activity is related to Figure 4 It is equivalent to Figure D in and is classified as level 3.

[0160] Figure 5 It shows Figure 4The fluctuation curves of supermicro blood flow motion density (SMI-MD) corresponding to the ultrasound images of five patients with different symptoms over time are divided into five sub-graphs (AE), each of which contains four curves at different grayscale thresholds (0.75, 0.80, 0.85 and 0.90). It should be noted that the grayscale threshold set when the image is binarized can be any suitable value, for example, it can be any value between 0 and 100. The several grayscale thresholds shown here are only exemplary and do not constitute a limitation of the present invention. The colors corresponding to each curve are: gray (0.75 threshold), blue (0.80 threshold), pink (0.85 threshold) and cyan (0.90 threshold), and the legend in the lower right corner marks each threshold. These curves reflect the dynamic change characteristics of SMI-MD over time at different thresholds. Under a lower threshold (such as 0.75), the curve fluctuation amplitude is larger, showing more significant changes; while under a higher threshold (such as 0.90), the SMI-MD curve fluctuation amplitude is relatively small, and the performance is more stable. Each sub-graph shows a different fluctuation pattern, revealing the differences in ultramicroscopic blood flow density under different time and threshold conditions.

[0161] Figure 6 It shows Figure 4 Five sub-graphs (AE) of super microvascular elastic density (SMI-ED) corresponding to ultrasound images of five patients with different symptoms. Each sub-graph shows the fluctuation curve of SMI-ED over time, which is drawn under four different grayscale thresholds (0.75, 0.80, 0.85 and 0.90). The colors of each curve are gray (0.75 threshold), blue (0.80 threshold), pink (0.85 threshold) and cyan (0.90 threshold), and the legend in the lower right corner marks each threshold. In these curves, the dynamic trend of SMI-ED over time can be observed. At a lower threshold (0.75), the SMI-ED curve fluctuates more, reflecting a more sensitive response, while the SMI-ED curve at a higher threshold (0.90) is relatively stable and has a relatively small fluctuation. The curves with different thresholds show different fluctuation patterns over time, revealing the differences in super microvascular elastic density at different time points and threshold conditions.

[0162] Figure 4 All patients showed osteoarthritis features such as synovial thickening, hypoechoic and joint effusion, and varying degrees of blood flow signal enhancement indicated changes in inflammatory activity. Soft tissue lesions were also involved. Figure 4 The ligament thickening in Figure A Figure 4 The tear injury in Figure C Figure 4The edema of the lower edge of the patella in the middle D image reflects the association between osteoarthritis and soft tissue lesions. The difference is mainly reflected in the distribution and density of blood flow signals. Figure 4 Middle A) No blood flow signal, indicating mild inflammatory reaction; Grade 1 ( Figure 4 Middle B) A small amount of blood flow signal, inflammation is initially active; Grade 2 ( Figure 4 Middle C) The blood flow signal is rich but limited; Grade 3 ( Figure 4 Medium D picture, Figure 4 Middle (E) The blood flow signal is highly dense, indicating widespread and active inflammation.

[0163] from Figure 4 A picture to Figure 4 In Figure E, the severity of the lesion gradually escalates, with the range of synovial thickening from limited to extensive, the blood flow signal density from sparse to highly dense, and the inflammatory activity from mild to severe. Grade 0 reflects static synovial inflammation; Grades 1 and 2 indicate gradually worsening inflammation; and Grade 3 represents the most active state of inflammation, with significant changes in blood supply and tissue. This grading system clearly demonstrates the dynamic development of osteoarthritis and provides a reliable basis for blood flow quantification and automatic grading research based on the optical flow model.

[0164] It is worth noting that as the level of blood flow signals increases, the curve trends of the ultramicro blood flow motion density and ultramicrovascular elastic density indicators become clearer and more regular.

[0165] The present invention adopts SPSS Statistical analysis was performed using version 26.0. Descriptive statistics were calculated using the Chi-square test. p Value. Mainly used Kappa The agreement test evaluates the agreement between the automatic classification of microvascular flow and the physician rating. The specific method includes calculating the observed agreement and expected agreement. The observed agreement is calculated by the diagonal sum of the confusion matrix, and the expected agreement is estimated based on the marginal sums of rows and columns. Kappa The specific calculation formula of the value is shown in formula (5):

[0166] (5)

[0167] in, represents the observed consistency, Represents expected consistency. Kappa The value ranges from -1 to 1, and the closer the value is to 1, the better the consistency. Kappa Values ​​greater than 0.6 indicate good rating agreement, while values ​​less than 0 indicate significant inconsistency.p Values ​​were also calculated to further verify the statistical significance of the agreement between ratings.

[0168] According to the confusion matrix (see Table 5), a total of 47 subjects were included in this experiment, including 20 males and 27 females ( p =0.307>0.05, χ 2 =1.042), with no statistically significant difference in gender distribution. The age range of patients was 30 to 65 years old, with an average age of 45.3 years, covering the main high-risk population for osteoarthritis. The observed consistency was 70.2%, indicating significant consistency between the automatic rating and the physician rating. The Kappa value was 0.627, indicating good rating consistency (Kappa>0.6); when the Kappa value was less than 0, it indicated significant inconsistency between the ratings. In addition, the observed consistency was significantly higher than the expected consistency (51.2%). The significance test showed p<0.001, further supporting the statistical significance of the consistency.

[0169] Table 5 Comparison of joint microblood flow automatic classification and physician rating (N=47)

[0170]

[0171] Further analysis found that when the physician rating was 2, the automatic rating system was more inclined to give a higher grade result (i.e., as shown in Table 5, when the physician rating was 2, through automatic rating, 1 case was rated as 0, 0 cases were rated as 1, that is, 7 were rated as 2, and 5 were rated as 3. Compared with the number of cases rated as 0 and 1, there were more cases rated as 3, that is, higher grades), which may be related to the noise in the blood flow signal or the threshold setting of the blood flow signal grading algorithm for the intermediate grade. This phenomenon suggests that the automatic rating system needs to be optimized to further improve the accuracy in the identification of moderate osteoarthritis. Overall, the research results verified the effectiveness of the automatic rating system in the grading of osteoarthritis, and pointed out the necessity of optimizing the system algorithm in specific situations in the future.

[0172] This paper realizes the automatic classification of ultrafine blood flow signals of synovial joints through the dynamic analysis method of optical flow model, and explores the application value of dynamic changes of blood flow signals in the quantitative evaluation of osteoarthritis. The results show that there is good consistency between the automatic rating system and the physician rating (Kappa=0.627, p<0.001), and the observed consistency is significantly higher than the expected consistency (70.2% vs 51.2%). These results support the effectiveness of this method in the quantitative analysis of ultrafine blood flow signals.

[0173] Compared with traditional osteoarthritis rating methods, the method of the present invention has significant advantages in analysis dimension and objectivity of results. Traditional methods are mostly based on subjective grading based on static ultrasound images, such as the semi-quantitative grading method proposed by Szkudlarek et al., and the results are often affected by operator experience, equipment parameters and image quality, and the rating consistency is low. In recent years, some studies have explored automated grading methods based on deep learning, which have improved the accuracy and consistency of grading through feature extraction of static images, but these methods have failed to fully capture the dynamic characteristics of blood flow signals. The present invention uses dynamic analysis of optical flow models to achieve the quantification of dynamic changes in ultra-micro blood flow signals for the first time, thereby improving the objectivity and clinical applicability of grading results.

[0174] In summary, the dynamic analysis based on the optical flow model proposed in the present invention realizes the automatic classification of ultra-micro blood flow signals in joints, and verifies the effectiveness of this method in the quantitative assessment of osteoarthritis. Moreover, the experimental results show that the automatic rating system has good consistency with the physician rating, which is significantly higher than the expected consistency, reflecting the advantages of the method of the present invention in reducing subjective influences and improving the objectivity of ratings. At the same time, the introduction of dynamic analysis enables the motion characteristics of blood flow signals to be quantified, overcoming the limitations of traditional static image-based grading methods. In general, the present invention provides a new technical means for the non-invasive diagnosis and dynamic monitoring of osteoarthritis, and also points out the direction for further optimizing the algorithm performance and expanding its clinical application scenarios in the future.

[0175] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solution, which should be included in the scope of the claims of the present invention.

Claims

1. A method for quantification and classification of muscle-bone ultrasound ultramicro blood flow based on optical flow analysis, characterized in that: include: Acquire an original ultramicro blood flow image sequence of the object to be detected; Preprocessing is performed from the original ultramicroscopic blood flow image sequence; Extracting the region of interest, i.e., the ROI region, from the preprocessed ultramicroscopic blood flow image sequence; Extract ultramicrovascular morphology in the ROI area; Performing dynamic analysis on the ultramicro blood flow in the ROI area, the dynamic analysis including: ultramicro blood flow calculation, motion optical flow field estimation and fusion of microblood flow morphological features; According to the dynamic analysis results, two indicators are generated: the motion density of ultramicro blood flow and the elastic density of ultramicro blood vessels. The motion density of ultramicro blood flow refers to the ratio of the motion optical flow of the optical flow component parallel to the blood flow signal skeleton at all pixel points in two adjacent frames to the ultramicro blood flow. The elastic density of ultramicro blood vessels refers to the ratio of the motion optical flow of the optical flow component perpendicular to the blood flow signal skeleton at all pixel points in two adjacent frames to the ultramicro blood flow. The blood flow signal in the region of interest is graded according to the motion density and elastic density; Ultramicrovascular morphology extraction calculations include: Performing grayscale conversion and binarization processing on each frame of the original ultramicro blood flow image sequence, setting the pixel values ​​of the pixels whose grayscale values ​​are greater than the preset grayscale threshold to 1, and the others to 0, thereby generating a binary image; Extract the ROI region from each binary image to retain the target blood flow signal; Segment and cluster the blood flow area for each binary image; Perform edge detection and skeleton extraction on the blood flow area obtained by segmentation and clustering to generate a skeleton image; Perform curve fitting on the skeleton in the skeleton image to obtain microvascular parameters; Motion optical flow estimation includes: After the blood flow area of ​​each binary image is determined, the two consecutive binary images are divided into a group, and the optical flow is calculated for each group of binary images to obtain the optical flow corresponding to each group of binary images; The motion density of the ultramicro blood flow and the elastic density of the ultramicro blood vessels corresponding to each set of binary images are determined, thereby obtaining a set of motion densities of the ultramicro blood flow and a set of elastic densities of the ultramicro blood vessels.

2. The method for quantification and grading of muscle-bone ultrasound ultramicro blood flow based on optical flow analysis according to claim 1 is characterized in that: The segmentation of the blood flow area of ​​the binary image includes: Step 11: Initialize the queue; Step 12: Traverse each pixel in the ROI area and mark the pixel with a pixel value of 1 as the target pixel; Step 13: Add the target pixel to the queue; Step 14: When the queue is not empty, remove the first element in the queue and use the removed element as the current pixel, then execute step 15; Step 15: Obtain neighboring pixels within a preset range of the current pixel. For each neighboring pixel, if the neighboring pixel value is 1 and has not been marked, mark the neighboring pixel as a target pixel and add the neighboring pixel to a queue until all neighboring pixels are checked; Step 16: When the queue is empty, the connected target pixels are grouped into an independent blood flow area; Step 17: Repeat steps 12-16 until all target pixels are processed.

3. The method for quantification and grading of muscle-bone ultrasound ultramicro blood flow based on optical flow analysis according to claim 1, characterized in that: Blood flow region clustering includes: Step 21: Obtain a blood flow area set; Step 22: Calculate the distance between every two blood flow areas in the blood flow area set; Step 23: When there are two or more blood flow regions in the blood flow region set, clustering and merging can be performed, and steps 24-25 are executed; Step 24: If the distance between the blood flow areas is less than a preset threshold, the two blood flow areas are merged into one blood flow area; Step 25: Update the blood flow area set using the merged blood flow area, and return to step 23; Step 26: When there are no more blood flow regions that can be merged, a clustered blood flow region set is obtained.

4. The method for quantification and grading of muscle-bone ultrasound ultramicro blood flow based on optical flow analysis according to claim 1, characterized in that: Edge detection and skeleton extraction include: For each pixel point in the blood flow area obtained by segmentation and clustering, if the pixel value of the pixel point is 1, and the pixel values ​​of the pixels in the neighborhood within the preset range of the pixel point are all 1, it means that the pixel point is completely surrounded, and the pixel value of the pixel point is changed to 0, that is, the pixel is corroded; After processing each pixel in the binary image, a skeleton image is obtained.

5. The method for quantification and grading of muscle-bone ultrasound ultramicro blood flow based on optical flow analysis according to claim 1, characterized in that: The blood flow signal of the region of interest is graded according to the motion density and elastic density, including: According to the movement density of ultramicro blood flow, the average value and the maximum value of the movement density of ultramicro blood flow are obtained; According to the elastic density set of ultramicrovessels, the average value and maximum value of the elastic density of ultramicrovessels are obtained; The blood flow signals were graded according to the average and maximum values ​​of the motion density of the ultramicro blood flow and the average and maximum values ​​of the elastic density of the ultramicrovessels.

6. A muscle-bone ultrasound ultra-micro blood flow quantification grading system based on optical flow analysis, used to implement a muscle-bone ultrasound ultra-micro blood flow quantification grading method based on optical flow analysis as described in any one of claims 1 to 5, characterized in that: include: An image acquisition module, used to acquire an original ultramicro blood flow image sequence of the object to be detected; A preprocessing module, which preprocesses the original ultramicro blood flow image sequence; The ROI region extraction module is used to extract the region of interest, i.e., the ROI region, from the ultramicro blood flow image sequence after preprocessing; Ultra-microvascular morphology extraction module, used to extract ultra-microvascular morphology from ROI area; A dynamic analysis module is used to dynamically analyze the ultramicro blood flow in the ROI area, wherein the dynamic analysis includes ultramicro blood flow calculation, motion optical flow field estimation, and fusion of microblood flow morphological features; The index determination module is used to generate two indexes, namely, the motion density of ultramicro blood flow and the elastic density of ultramicro blood vessels, according to the dynamic analysis results. The motion density of ultramicro blood flow refers to the ratio of the motion optical flow of the optical flow component parallel to the blood flow signal skeleton at all pixel points in two adjacent frames to the ultramicro blood flow, and the elastic density of ultramicro blood vessels refers to the ratio of the motion optical flow of the optical flow component perpendicular to the blood flow signal skeleton at all pixel points in two adjacent frames to the ultramicro blood flow. The classification module is used to classify the blood flow signal of the region of interest according to the motion density and elastic density.

7. The musculoskeletal ultrasound ultra-micro blood flow quantification and grading system based on optical flow analysis according to claim 6 is characterized in that: The Ultramicrovascular Morphology Extraction Module is also used to: Performing grayscale conversion and binarization processing on each frame of the original ultramicro blood flow image sequence, setting the pixel values ​​of the pixels whose grayscale values ​​are greater than the preset grayscale threshold to 1, and the others to 0, thereby generating a binary image; Extract the ROI region from each binary image to retain the target blood flow signal; Segment and cluster the blood flow area for each binary image; Perform edge detection and skeleton extraction on the blood flow area obtained by segmentation and clustering to generate a skeleton image; The skeleton in the skeleton image is subjected to curve fitting to obtain microvascular parameters.

8. The musculoskeletal ultrasound ultra-micro blood flow quantification and grading system based on optical flow analysis according to claim 7 is characterized in that: The grading module grades the blood flow signals in the region of interest according to the motion density and elastic density, including: Determine the average and maximum values ​​of the motion density of ultramicroscopic blood flow; Determine the mean and maximum values ​​of the elastic density of ultramicrovessels; The blood flow signals are graded according to the average and maximum values ​​of the motion density of the ultramicro blood flow and the average and maximum values ​​of the elastic density of the ultramicro blood vessels.

Images