Ultrasound-guided nerve block image processing method based on artificial intelligence

Abstract

The application relates to the technical field of image processing, in particular to an ultrasound-guided nerve blockage image processing method based on artificial intelligence, which comprises the following steps: obtaining the weight of other pixel points in the neighborhood of each pixel point according to the gradient amplitude of each pixel point, the distance between other pixel points in the neighborhood and the center point of the neighborhood, and the neighborhood radius; obtaining the weighted weight of each pixel point according to the weight of other pixel points in the neighborhood of each pixel point; obtaining the search window size according to the weighted weight of each pixel point and a window adjustment parameter, and obtaining a search window with each pixel point as the center according to the search window size; obtaining the iterative convergence condition according to the pixel point density difference in the search window and the distance of the clustering center; performing mean shift clustering according to the convergence condition and an initial clustering center to obtain a plurality of clustering clusters; and performing segmentation and visualization of nerve tissue regions according to the obtained clustering clusters.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image processing, in particular to an ultrasound-guided nerve blockage image processing method based on artificial intelligence. BACKGROUND

[0002] Ultrasound-guided nerve blockage is a method of using ultrasound imaging technology for real-time guidance and monitoring during nerve blockage, commonly used in anesthesia and pain treatment, and is a safe, fast and effective technology. This method can clearly determine the location and depth of nerves and surrounding blood vessels, avoid blood vessels and other important tissue structures, and improve the success rate and safety of nerve tissue. However, due to the lack of contrast in ultrasound imaging data, it may not be possible to identify or accurately extract nerve tissue, which can affect the success rate of nerve blockage, leading to errors or misjudgments by doctors when analyzing image features. Therefore, with the improvement of medical imaging technology, there is a need to improve the feature extraction accuracy of ultrasound-guided nerve blockage images to improve the success rate and safety of nerve blockage.

[0003] The mean shift clustering algorithm is an algorithm that iteratively calculates local optimal points to form clustering centers based on pixel density, and can handle clusters of any shape. However, the ultrasound image may have a problem of insufficient contrast, which can lead to inaccurate clustering results. SUMMARY

[0004] To solve the problem of inaccurate ultrasound tissue extraction results caused by insufficient contrast in ultrasound images, the present application provides an ultrasound-guided nerve blockage image processing method based on artificial intelligence, which adopts the following technical solutions:

[0005] One embodiment of the present application provides an ultrasound-guided nerve blockage image processing method based on artificial intelligence, which includes the following steps:

[0006] Collecting ultrasound-guided nerve blockage images, using a density clustering algorithm to obtain initial clustering centers based on the parameter information of the pixel points in the ultrasound image;

[0007] Taking a neighborhood of a predetermined radius centered on each pixel point, determining the weight of the pixel points in the neighborhood based on the distance between the parameter information of the pixel points in the neighborhood and the center point and the gradient amplitude of the center point;

[0008] Determining the window adjustment parameters of each pixel point based on the distribution of the local density of the pixel points in the neighborhood centered on each pixel point and the distribution of the weight;

[0009] Adaptively determining the size of the search window when moving to each pixel point as the center of the window based on the window adjustment parameters of each pixel point and the search window of the initial clustering center;

[0010] The method using mean shift obtains the segmentation result of the neural tissue based on the size of the search window and a convergence condition, and visualizes the segmentation result of the neural tissue.

[0011] Preferably, the method for obtaining the initial clustering center comprises the following specific steps:

[0012] The coordinate information and the gray value of each pixel point jointly constitute the parameter information of each pixel point;

[0013] The parameter information of all pixel points is taken as input, the Euclidean distance between the parameter information of two pixel points is taken as the measurement distance during clustering, and the density clustering algorithm is used to divide all pixel points into a plurality of clustering clusters;

[0014] The initial clustering center is determined according to the local density of the pixel points in all clustering clusters.

[0015] Preferably, the method for determining the initial clustering center according to the local density of the pixel points in all clustering clusters comprises the following specific steps:

[0016] The average value of the local density of all pixel points in each clustering cluster is calculated respectively, the clustering cluster corresponding to the maximum value among all the average values is taken as the clustering cluster with the highest local density, and a pixel point in the clustering cluster with the highest local density is randomly selected as the initial clustering center.

[0017] Preferably, the method for determining the weight of the pixel point in the neighborhood comprises the following specific steps:

[0018] ;

[0019] In the formula, represents the weight of the qth pixel point in the neighborhood with the mth pixel point as the center, is an exponential function with a natural constant as the base; represents the gradient amplitude of the mth pixel point; represents the parameter distance between the mth pixel point and the qth pixel point, and h is the radius of the neighborhood.

[0020] Preferably, the method for determining the window adjustment parameter of each pixel point comprises the following specific steps:

[0021] All pixel points in the neighborhood constructed with each pixel point as the center are divided into two types of first-type pixel points and second-type pixel points by using threshold segmentation;

[0022] The absolute value of the difference between the variance of the local density of the first-type pixel points and the variance of the local density of the second-type pixel points is calculated;

[0023] ​The absolute value of the difference, the variance of the weight of all pixel points in the neighborhood constructed with each pixel point as the center, and a constant parameter are summed as the denominator;

[0024] The absolute value of the difference between the number of the first type of pixel points and the second type of pixel points in the neighborhood and the denominator are taken as the window adjustment coefficient of each pixel point.

[0025] Preferably, the method of dividing the pixel points into the first type of pixel points and the second type of pixel points comprises:

[0026] The local density of all pixel points in the window constructed with each pixel point as the center is taken as the input, the threshold value is obtained by using the Otsu threshold segmentation method, the pixel points with the local density greater than the threshold value are marked as the first type of pixel points, and the pixel points with the local density less than or equal to the threshold value are marked as the second type of pixel points.

[0027] Preferably, the method of adaptively determining the size of the search window when moving to each pixel point as the center of the window comprises:

[0028] The normalized result of the ratio of the arithmetic square root of the number of all pixel points to the weighted weight of each pixel point is summed with the normalized result of the window adjustment coefficient of each pixel point.

[0029] The integral result of the product of the search window of the initial clustering center and the sum value is taken as the size of the search window when moving to each pixel point as the center of the window in the mean shift clustering process.

[0030] Preferably, the method of obtaining the segmentation result of the neural tissue comprises:

[0031] The convergence condition is determined according to the change of the local density of the pixel points before and after iteration and the change of the parameter information of the clustering center.

[0032] The parameter information of all pixel points is taken as the input, and the pixel points are divided into a plurality of clustering clusters based on the convergence condition by using the mean shift algorithm.

[0033] Preferably, the method of determining the convergence condition comprises:

[0034] The absolute value of the difference between the mean value of the local density of all pixel points in the same clustering cluster in each iteration and the result of the adjacent previous iteration is calculated.

[0035] The sum of the absolute value of the difference and the Euclidean distance between the parameter information of the clustering center in the same clustering cluster in each iteration and the result of the adjacent previous iteration is less than a preset convergence threshold value, as the convergence condition.

[0036] Preferably, the method of visualizing the segmentation result of the neural tissue comprises:

[0037] statistical gray level variance of each cluster and form a gray level variance interval, divide the obtained gray level variance interval by 0.1 as a step;

[0038] According to the gray level variance from small to large, give light to dark color, get different colors of each cluster in the ultrasound image, and display in the ultrasound image.

[0039] The technical scheme of the present application has the beneficial effects that: the initial clustering center is obtained by the density distribution of the ultrasound-guided nerve tissue image, which is more accurate than random selection, the weight of other pixel points in the neighborhood of each pixel point is obtained according to the gradient amplitude of the pixel point, the distance between other pixel points in the neighborhood of each pixel point and the neighborhood center point, and the neighborhood radius of each pixel point; the weighted weight of each pixel point is obtained according to the weight of other pixel points in the neighborhood of each pixel point; the search window size is obtained according to the weighted weight of each pixel point, and the search window centered on each pixel point is obtained according to the search window size; the iterative convergence condition is obtained according to the density difference of the pixel points in the search window and the distance of the clustering center; the mean shift clustering is performed according to the convergence condition and the initial clustering center, and a plurality of cluster clusters are obtained; the nerve tissue region is determined according to the gray level features of the obtained cluster clusters; and the nerve tissue region is segmented according to the obtained cluster clusters. The present application can adaptively determine the search window size in the mean shift clustering process according to the density distribution of the pixel points, and determine the convergence condition by combining threshold processing, which not only avoids the problem of too few or too many clustering centers caused by too large or too small search window, but also improves the operation efficiency, and the threshold is combined to determine the convergence condition in the iteration process according to the clustering result, thereby further improving the segmentation accuracy of the nerve tissue. BRIEF DESCRIPTION OF DRAWINGS

[0040] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0041] Figure 1 The implementation flowchart of the artificial intelligence-based ultrasound-guided nerve block image processing method provided by an embodiment of the present application. DETAILED DESCRIPTION

[0042] To further illustrate the technical means and effects adopted by this application to achieve the intended inventive objective, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of the ultrasound-guided nerve block image processing method based on artificial intelligence proposed in this application. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0043] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0044] The following description, in conjunction with the accompanying drawings, details the specific scheme of the ultrasound-guided nerve block image processing method based on artificial intelligence provided in this application.

[0045] Please see Figure 1 The diagram illustrates an implementation flowchart of an artificial intelligence-based ultrasound-guided nerve block image processing method according to an embodiment of this application. The method includes the following steps:

[0046] S001, Acquire ultrasound-guided nerve block images to obtain initial cluster centers for clustering nerve tissue.

[0047] Set the parameters of the ultrasound equipment, select an appropriate ultrasound probe according to the application site, and set the scanning depth and frequency to obtain a clear ultrasound-guided nerve block image. To facilitate subsequent processing, the obtained image is converted to grayscale, and the resulting grayscale image is recorded as the ultrasound-guided nerve block image.

[0048] It should be noted that when selecting the initial cluster center for clustering neural tissue in ultrasound-guided nerve block images, the probability of selecting the final cluster center is higher when choosing the initial cluster center in a region with high pixel density. Therefore, this method uses density clustering to find regions with high pixel density in ultrasound images and randomly selects a pixel in that region as the initial cluster center.

[0049] Furthermore, in the embodiments of this application, the parameter information of each pixel in the nerve-guided nerve block image is composed of coordinate information and grayscale value. Taking the first pixel in the ultrasound-guided nerve tissue image as an example... Taking a pixel as an example, the parameter information of the m-th pixel is represented as follows: ,in, , Indicates the first The x-coordinate and y-coordinate of each pixel in the plane Indicates the first The grayscale value of each pixel.

[0050] Specifically, the specific process of obtaining the initial clustering center by the parameter information of all the pixel points through the density clustering is: taking the parameter information of all the pixel points as input, taking the Euclidean distance between the parameter information of two pixel points as the measurement distance in clustering, and dividing all the pixel points into a plurality of clustering clusters by using the density clustering algorithm.

[0051] It should be noted that the density clustering is a known technology in the field of image processing, and the specific process will not be repeated. Common density clustering algorithms include but are not limited to DBSCAN (Density-Based Spatial Clustering Applications with Noise), density peak clustering DPC, and OPTICS (Ordering Points To Identify the Clustering Structure). Preferably, in the present embodiment, the pixel points in the ultrasound image are divided into clustering clusters by using the density peak clustering DPC algorithm.

[0052] Further, the average value of the local density of all the pixel points in each clustering cluster is calculated, the clustering cluster corresponding to the maximum value among all the average values is taken as the clustering cluster with the highest local density, and a pixel point in the clustering cluster with the highest local density is randomly selected as the initial clustering center.

[0053] S002, obtain the pixel point density of the ultrasound-guided neural tissue image, and obtain the search window size centered on each pixel point.

[0054] It should be noted that when the mean shift clustering algorithm is used to cluster and segment the neural tissue in the ultrasound image, too large or too small search window will increase the calculation amount and calculation complexity. Therefore, adaptively determining the window size in the drift clustering process according to the pixel density can appropriately reduce the calculation amount. Therefore, in the present embodiment, the pixel density in the ultrasound-guided nerve block image is calculated by using the Gaussian kernel density estimation method.

[0055] It should be noted that when the Gaussian kernel density estimation method is used to calculate the pixel density, the weighting weight of the smooth part of the density change is relatively accurate, but the density change is relatively complex or affected by noise points, which may reduce the accuracy of the weighting weight. Therefore, in the present embodiment, the pixel gradient is used to set the weight to enhance the density calculation of the local complex area to improve the calculation accuracy.

[0056] First, the gradient image of the ultrasound image is obtained. When the pixel gradient is larger, it indicates that the pixel is more likely to be located at the edge of neural tissue, so the pixel is given a larger weight. Then, the Gaussian kernel density estimation function and the weighting function are combined. By reducing the transformation rate of the weights, the accuracy of the density estimation is improved. This increases the weight of pixels with large gradients in areas with large gradients, so as to retain as much image detail as possible.

[0057] For each pixel in the ultrasound image, taking the m-th pixel as an example, a matrix is ​​constructed centered on the m-th pixel. A circular neighborhood with radius , where The first step is to calculate the mean Euclidean distance between the parameter information of all pixels. Next, the weights of all pixels surrounding the center pixel are calculated. These weights are determined by the Gaussian kernel function and adjusted based on the gradient magnitude. The calculation format is as follows:

[0058] ;

[0059] In the formula, This represents the weight of the q-th pixel within the neighborhood centered at the m-th pixel. It is an exponential function with the natural constant as its base; Indicates the first Gradient magnitude of each pixel; This represents the parameter distance between the m-th pixel and the q-th pixel, and its size is equal to the square of the Euclidean distance between the parameter information of the two pixels.

[0060] It should be noted that, using Adjust the rate of change of pixel weights: when The larger the value, the more likely the m-th pixel is to be located at the edge of neural tissue, so it needs to be assigned a larger weight. Conversely, the lower the probability that the m-th pixel is at the edge of neural tissue, the less effective it is in preserving the detailed information of neural tissue. Reduce the rate of change of pixel weights.

[0061] Furthermore, after adjusting the weights of each pixel in the image, the weights of all pixels are averaged according to the calculation steps of the existing Gaussian kernel density estimation algorithm to obtain the weighted weight of each pixel. Specifically, the average weight of all pixels in the neighborhood constructed with each pixel as the center is used as the weighted weight of each pixel.

[0062] It should be noted that in Gaussian kernel density estimation, the weight function is a measure of the contribution weight of each pixel, which determines the importance of each pixel in the estimation process. Therefore, this method adjusts the weight function by gradient, increasing the weight of pixels with large gradients. This not only reduces the impact of outliers such as isolated noise points on the results, but also enhances the robustness of kernel density estimation.

[0063] Furthermore, to adaptively determine the window size during drift clustering, for regions with relatively flat pixel density distribution, appropriately increasing the window size can reduce the over-smoothing effect during clustering, thereby preserving more image details; for regions with complex pixel density distribution, the search window size needs to be reduced to ensure that the drift amount converges to the global extremum within a small range. Therefore, this embodiment constructs a relationship between the search window size and pixel density to adaptively obtain the size of the search window during clustering, and performs mean drift clustering to obtain accurate neural tissue morphology and location.

[0064] Specifically, the window constraint coefficient of each pixel is determined based on the local image features of each pixel and the pixels in its constructed circular neighborhood, which is then used to adjust the size of the search window.

[0065] First, for pixels belonging to the corresponding region of neural tissue, within a neighborhood constructed centered on the edge pixels of the neural tissue region, there are pixels both within and outside the neural tissue region. Pixels within the neural tissue region have similar grayscale values ​​and high local density, while pixels outside the region correspond to different tissues, resulting in differences in grayscale values ​​and lower local density. Therefore, if a pixel exhibits a local density distribution within its neighborhood that is relatively consistent on one side and significantly different on the other, the probability that the pixel belongs to the edge of neural tissue is higher, more image details need to be preserved, and a smaller window is required. If the neighborhood constructed centered on the pixel does not have the above characteristics, a larger window is needed.

[0066] Secondly, for each pixel, the local density of all pixels within a window centered on each pixel is used as input, and the Otsu thresholding method is used to obtain the segmentation threshold. Pixels with local density greater than the segmentation threshold are marked as first-class pixels; pixels with local density less than or equal to the segmentation threshold are marked as second-class pixels. The Otsu thresholding method is a well-known technique, and the specific process will not be described in detail.

[0067] Based on the above analysis, the window adjustment parameter for each pixel is calculated to characterize the required window size for each pixel. The window adjustment parameter for the m-th pixel is expressed as follows: :

[0068] ;

[0069] In the formula, Let represent the variances of the local densities of pixels categorized as class 1 and class 2 within the neighborhood centered on the m-th pixel. It represents the difference in the number of pixels in the first category and the second category, and it represents the variance of the weight distribution of all pixels in the neighborhood constructed with the m-th pixel as the center. This is a constant parameter, its purpose is to prevent the denominator from being zero. The value range is greater than 0 and less than 1. Preferably, here... The size is taken as an empirical value of 0.1.

[0070] The greater the probability that the m-th pixel belongs to the edge of neural tissue, the more image details need to be preserved; the smaller the value of the adjustment parameter of the m-th pixel, the smaller the required window. Furthermore, in this embodiment, based on the correlation between the obtained pixel density and the search window, a functional relationship is constructed to obtain the window size for each pixel of each search value:

[0071] First, determine the search window for the initial cluster centers. Then, during the clustering process, the window size is adaptively adjusted based on the pixel density, and the calculation method is as follows:

[0072] ;

[0073] in, Indicates moving to the first Search window size when the cluster center is 100 pixels. This indicates that the first value obtained using Gaussian density estimation is... The weighted average of each pixel, where M is the number of pixels in the ultrasound image. This is the rounding function. Represents the normalization function. It is the normalized result of the window adjustment parameter of the m-th pixel within the range of window adjustment parameters of all pixels.

[0074] It should be noted that by using the steps described above to calculate the search window size for pixels, the window size can be automatically adjusted according to changes in pixel density, thereby preserving more image details.

[0075] S003, determine the clustering iteration stopping condition based on density value changes and clustering effect evaluation, and obtain the location and morphology of neural tissue.

[0076] It should be noted that the traditional mean shift clustering algorithm mainly stops iteration by fixing the number of iterations or when the length of the drift vector is less than a certain threshold. If the iteration stopping condition is not set well, the clustering result will be inaccurate. Therefore, the method constructs a clustering evaluation index system combined with threshold processing, and stops iteration when the clustering effect reaches a certain index. At this time, the corresponding clustering result is better.

[0077] It should be noted that in the process of repeated search and movement of the cluster center, when the iteration stopping condition converges, the iteration is stopped. When the difference in pixel point density change is small enough, it means that the difference in the same cluster is smaller, and the similarity between the pixel points in the same cluster is higher. At the same time, in the process of moving the cluster center point, the distance between the two moving cluster center points can dynamically stop the iteration threshold. When the distance between the two moving cluster center points is small enough, it means that the clustering effect is good at this time. Therefore, in order to improve the accuracy of clustering, the embodiments of the present application combine the two to construct the following convergence condition:

[0078]

[0079] Wherein, The difference between the mean value of the local density of all pixel points in the same cluster in the a-th iteration and the a-1-th iteration process. The Euclidean distance between the parameter information of the cluster center in the a-th iteration and the a-1-th iteration process, The convergence threshold. Wherein, .

[0080] It should be noted that the smaller the value of the convergence threshold, the higher the accuracy requirement for image clustering and segmentation. However, too small convergence threshold will cause iteration to fail to converge. The value range is usually 0.0001-0.01, and the preferred value of the convergence threshold is 0.001.

[0081] It should be noted that the clustering process is a repeated calculation process. The steps of repeated mean shift clustering are compared with the iteration stopping condition until the above convergence condition is met and the number of clustering clusters is obtained.

[0082] S004, visualizing the ultrasonic image after clustering and segmentation.

[0083] It should be noted that after the improved mean shift clustering algorithm is used to cluster and segment the ultrasonic guided nerve block image, the number of clustering clusters is obtained. Because the nerve tissue region is usually displayed as light and dark interlaced, it is contrasted with the surrounding region, so the embodiments obtain the number of clustering clusters The gray variance value is given according to the different gray variance values​ different colors to the different clusters so that the doctor can more intuitively obtain the accurate nerve tissue position and shape. The operation process of assigning different colors to different clusters is as follows:

[0084] First, the gray variance of each cluster of the image is counted and a gray variance interval is formed. The obtained gray variance interval is divided by 0.1 as a step, and then based on the gray variance size distributed in the multiple gray variance intervals obtained after dividing the gray variance interval, the color from light to deep is assigned from small to large gray variance, and the different colors of each cluster are obtained in each cluster, and then the segmentation image of the ultrasound guided nerve block image with stronger visibility is obtained.

[0085] It should be noted that the above-described embodiments are only used to illustrate the technical solutions of the present application, but not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement to some technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

[0086] Each of the embodiments in the specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other. Each embodiment mainly describes the difference from other embodiments.

Claims

1. An ultrasound-guided nerve block image processing method based on artificial intelligence, characterized in that, The method comprises the following steps: The method comprises the following steps: An ultrasound-guided nerve block image is collected, and an initial clustering center is obtained from parameter information of pixel points in the ultrasound image by using a density clustering algorithm; A neighborhood of a preset radius is taken as a center of each pixel point, and a weight of a pixel point in the neighborhood is determined based on a distance between parameter information of the pixel point and the center and a gradient amplitude of the center; A window adjustment parameter of each pixel point is determined according to a distribution of local densities of the pixel points in the neighborhood taken as the center and a distribution of the weights; A size of a search window when the search window is moved to each pixel point as a window center is adaptively determined based on the window adjustment parameter of each pixel point and a search window of the initial clustering center; A segmentation result of the nerve tissue is obtained based on the size of the search window and a convergence condition by using a mean shift method, and the segmentation result of the nerve tissue is visualized; The method for determining the window adjustment parameter of each pixel point comprises the following steps: All pixel points in the neighborhood constructed as a center of each pixel point are divided into two types of first-type pixel points and second-type pixel points by using a threshold segmentation method; An absolute value of a difference between a variance of the local density of the first-type pixel points and a variance of the local density of the second-type pixel points is calculated; A sum of the absolute value, a variance of the weights of all pixel points in the neighborhood constructed as the center of each pixel point, and a constant parameter is taken as a denominator; 2. The ultrasound-guided nerve block image processing method based on artificial intelligence according to claim 1, characterized in that, A ratio of an absolute value of a difference between the number of the first-type pixel points and the number of the second-type pixel points in the neighborhood and the denominator is taken as a window adjustment coefficient of each pixel point. The method for obtaining the initial clustering center comprises the following steps: The parameter information of each pixel point comprises coordinate information and a gray value; All pixel points are divided into a plurality of clustering clusters by using the density clustering algorithm, with the parameter information of all pixel points as input and an Euclidean distance between parameter information of two pixel points as a measurement distance in clustering; 3. The ultrasound-guided nerve block image processing method based on artificial intelligence according to claim 2, characterized in that, An initial clustering center is determined according to local densities of pixel points in all clustering clusters. The method for determining the initial clustering center according to the local densities of the pixel points in all clustering clusters comprises the following steps:

4. The ultrasound-guided nerve block image processing method based on artificial intelligence according to claim 1, characterized in that, An average value of the local densities of all pixel points in each clustering cluster is calculated, a clustering cluster corresponding to a maximum value among all the average values is taken as a clustering cluster with the highest local density, and a pixel point in the clustering cluster with the highest local density is randomly selected as the initial clustering center. ； wherein, represents the weight of the qth pixel point in the neighborhood centered on the mth pixel point, is an exponential function with a natural constant as the base; represents the gradient amplitude of the mth pixel point; represents the gradient amplitude of the mth pixel point; represents the parametric distance between the mth pixel point and the qth pixel point, and h is the radius of the neighborhood.

5. The ultrasound-guided nerve block image processing method based on artificial intelligence according to claim 1, characterized in that, The method for determining the weight of the pixel point in the neighborhood comprises the following steps: The method for dividing the pixel points into two types of first-type pixel points and second-type pixel points comprises the following steps:

6. The ultrasound-guided nerve block image processing method based on artificial intelligence according to claim 1, characterized in that, The local density of all pixel points in a window constructed as a center of each pixel point is taken as input, a threshold value is obtained by using the Otsu threshold segmentation method, pixel points with a local density greater than the threshold value are marked as first-type pixel points, and pixel points with a local density less than or equal to the threshold value are marked as second-type pixel points. The method for adaptively determining the size of the search window when the search window is moved to each pixel point as the window center comprises the following steps: A sum of a normalized result of a ratio of an arithmetic square root of the number of all pixel points and a weighted weight of each pixel point and a normalized result of a window adjustment coefficient of each pixel point is calculated. The search window of the initial clustering center is multiplied by the integer result of the sum value, and the product is taken as the size of the search window when moving to each pixel point as the window center in the mean shift clustering process.

7. The ultrasound-guided nerve block image processing method based on artificial intelligence according to claim 1, characterized in that, The method for obtaining the segmentation result of the neural tissue is: A convergence condition is determined according to the change of the local density of the pixel points in the clustering cluster before and after iteration and the change of the parameter information of the clustering center; The parameter information of all the pixel points is taken as input, and the pixel points are divided into a plurality of clustering clusters based on the convergence condition by using the mean shift algorithm.

8. The ultrasound-guided nerve block image processing method based on artificial intelligence according to claim 7, characterized in that, The method for determining the convergence condition is: The absolute value of the difference between the mean value of the local density of all the pixel points in the same clustering cluster in each iteration and the result of the adjacent previous iteration is calculated; The sum value of the absolute value of the difference and the Euclidean distance between the parameter information of the clustering center in the same clustering cluster in each iteration and the result of the adjacent previous iteration is less than a preset convergence threshold value, and the sum value is taken as the convergence condition.

9. The ultrasound-guided nerve block image processing method based on artificial intelligence according to claim 1, characterized in that, The method for visualizing the segmentation result of the neural tissue is: The gray variance of each clustering cluster is counted and a gray variance interval is formed, the obtained gray variance interval is divided with a step of 0.1; According to the color from light to dark in the order of the gray variance from small to large, different colors of each clustering cluster are obtained, and the different colors of each clustering cluster are displayed in the ultrasonic image.

Images