Skeletal muscle texture feature extraction method for sports injury assessment

Abstract

The present invention relates to the technical field of muscle ultrasound image analysis, and more specifically to a method for extracting skeletal muscle texture features for sports injury assessment. The method first extracts edge lines and their edge directions, as well as texture regions, from an ultrasound image. The method further analyzes the sports injury status of each texture region, weighting the texture directions of all texture regions using a sports injury index to obtain the run direction of the ultrasound image. Furthermore, based on the grayscale values ​​and spatial distribution of foreground pixels in the amplitude spectrum, combined with the grayscale fluctuations of the pixel sequence to be analyzed in the run direction and the run direction, a multi-dimensional index-adjusted grayscale run matrix is ​​constructed to adaptively extract skeletal muscle texture features, thereby improving the accuracy of skeletal muscle texture feature extraction for sports injuries.

Description

Technical Field

[0001] The present invention relates to the technical field of muscle ultrasonic image analysis, and in particular to a skeletal muscle texture feature extraction method for sports injury assessment. Background Art

[0002] The texture characteristics of skeletal muscle can reflect muscle health, injury severity, and recovery. Skeletal muscle is composed of bundles of muscle fibers, which exhibit specific directionality and structural patterns on imaging. Injury or disease may lead to abnormal muscle fiber arrangement, changes in muscle density, or fibrosis, which can be quantified using texture analysis methods. In recent years, with the rapid development of high-frequency ultrasound technology and ultrasound imaging, ultrasound images can accurately display specific parameters such as muscle thickness and echo intensity, making ultrasound an important imaging method for the skeletal muscle system.

[0003] When existing texture extraction methods process the texture of skeletal muscle with sports injuries, they are easily affected by liquid areas such as edema caused by sports injuries or other non-sports-injured skeletal muscle textures, resulting in incomplete or inaccurate texture extraction, affecting the accuracy of subsequent texture feature analysis. Summary of the Invention

[0004] In order to solve the technical problem that existing texture extraction is easily interfered with by other areas or textures, affecting the extraction accuracy of skeletal muscle texture features, the purpose of the present invention is to provide a skeletal muscle texture feature extraction method for sports injury assessment. The technical solution adopted is as follows:

[0005] Acquire an ultrasound image to be analyzed and establish a coordinate system; extract edge lines and edge directions in the ultrasound image, and acquire a texture region of the ultrasound image based on the spatial distribution of the edge lines and the difference in the edge directions;

[0006] Obtaining a motion impairment index for each texture region based on the directional similarity between each texture region and the nearest texture region and combining the grayscale features in the non-texture region between the texture region and its minimum circumscribed circle; and weighting the texture directions of all texture regions using the motion impairment index as a weight to obtain a run direction of the ultrasound image.

[0007] Acquire an amplitude spectrum of the ultrasound image; acquire a skeletal muscle injury index of the ultrasound image based on the grayscale values ​​and spatial distribution of foreground pixels in the amplitude spectrum; divide the pixels according to the distribution characteristics of the pixels in the run direction to obtain a pixel sequence to be analyzed; acquire a muscle fiber significance index of the ultrasound image based on the grayscale fluctuations of the pixel sequence to be analyzed;

[0008] The grayscale run matrix is ​​adjusted according to the run direction, the skeletal muscle damage index and the muscle fiber significance index to extract the skeletal muscle texture features of the ultrasound image.

[0009] Furthermore, the method for obtaining the texture area includes:

[0010] Obtaining a cluster distance based on the distance between the midpoints of the edge lines and the difference in the edge directions;

[0011] All the edge lines are clustered and classified based on the cluster distance, and the convex hull area of ​​each cluster is used as a texture area.

[0012] Furthermore, the method for obtaining the sports injury index includes:

[0013] taking the average of the edge directions of all the edge lines in each texture region as the corresponding texture direction; extending a straight line through the centroid of any texture region in the texture direction, and obtaining the other texture region whose centroid is closest to the extended straight line as the corresponding nearest texture region;

[0014] For any of the texture areas, the motion impairment index is obtained according to the grayscale value mean and grayscale value variance in each corresponding non-texture area, combined with the similarity between the texture direction and the texture direction of the corresponding nearest texture area.

[0015] Furthermore, the method for obtaining the skeletal muscle damage index includes:

[0016] Obtaining a foreground pixel of the amplitude spectrum from the binary image of the amplitude spectrum; establishing a two-dimensional coordinate system with the center point of the amplitude spectrum as the origin, the horizontal axis as the X-axis, and the vertical axis as the Y-axis; and using the angle between the line connecting the foreground pixel and the origin and the positive direction of the X-axis as the azimuth angle of the foreground pixel;

[0017] The skeletal muscle injury index is obtained based on the distance between each foreground pixel point and the coordinate origin in the amplitude spectrum, the difference in the azimuth angle between the foreground pixel point and the foreground pixel points in the same quadrant, and the grayscale value of the foreground pixel point in the amplitude spectrum.

[0018] Furthermore, the method for acquiring the pixel sequence to be analyzed includes:

[0019] In the texture region, the texture region is divided by all straight lines in the edge direction passing through the center point of the edge line to obtain texture sub-regions; and muscle fiber sub-regions are screened out based on the grayscale non-uniformity in the texture sub-regions;

[0020] Traversing the ultrasound image in the run direction to obtain all pixel sequences to be divided; performing a first division on the pixel sequence to be divided according to whether it belongs to the texture area; performing a second division on the pixel sequence obtained by the first division according to whether it belongs to the muscle fiber sub-area; and performing a third division on the pixel sequence obtained by the second division according to the muscle fiber sub-area to which it belongs;

[0021] The pixel sequence finally obtained by division is used as the pixel sequence to be analyzed.

[0022] Furthermore, the method for obtaining the muscle fiber sub-region includes:

[0023] The grayscale value variances of all the texture sub-regions are sorted from small to large, the sequence is divided at the point where the difference is the largest in the sorted sequence, and the texture sub-region of the partial sequence where the minimum grayscale value variance is located is marked as a muscle fiber sub-region.

[0024] Furthermore, the method for obtaining the muscle fiber significance index includes:

[0025] The grayscale value variance of the pixel sequence to be analyzed and the variance of the grayscale value mean within the pixel sequence to be analyzed are combined to obtain the muscle fiber significance index of the ultrasound image.

[0026] Furthermore, the method of adjusting the grayscale run-length matrix to extract the skeletal muscle texture features of the ultrasound image includes:

[0027] The skeletal muscle injury index and the muscle fiber significance index are integrated to obtain a quantity adjustment factor; the preset number of grayscale intervals is adjusted based on the quantity adjustment factor to obtain an appropriate number of intervals; a grayscale run matrix is ​​set according to the run direction and the appropriate number of intervals to extract the skeletal muscle texture features of the ultrasound image.

[0028] Furthermore, the method for obtaining the run direction includes:

[0029] The motion impairment index of all the texture regions is normalized using a softmax function, the normalized motion impairment index is used as the weight of the weighted summation corresponding to the texture direction, and the obtained weighted summation result is used as the run direction of the ultrasound image.

[0030] Furthermore, the method for obtaining the edge direction includes:

[0031] The principal axis of the edge line is obtained by the PCA algorithm, and the direction corresponding to the minimum angle between the principal axis and the X-axis of the coordinate system is used as the edge direction of the edge line.

[0032] The present invention has the following beneficial effects:

[0033] The present invention first extracts edge lines and their edge directions and texture areas in an ultrasonic image to facilitate analysis of the texture structure of local muscle fibers and calculation of the run direction; further obtains a motion injury index for each texture area to characterize the possibility of motion injury corresponding to each texture area, thereby preparing for subsequent adjustment of the run direction and accurate extraction of skeletal muscle texture features of motion injury; further weights the texture directions of all texture areas using the motion injury index as a weight to obtain the run direction required for extracting accurate and complete skeletal muscle motion injury texture features; further obtains the skeletal muscle injury index of the ultrasonic image from two perspectives, namely the grayscale value and spatial distribution of foreground pixels in the amplitude spectrum, characterizes the degree of muscle injury, and provides a basis for adjusting the grayscale run matrix; further obtains the muscle fiber significance index of the ultrasonic image based on the grayscale fluctuation of the pixel points in the run direction, characterizes the texture clarity of the pixel sequence, and provides more basis for adjusting the grayscale run matrix; finally, adjusts the grayscale run matrix based on the run direction, the skeletal muscle injury index and the muscle fiber significance index to extract the skeletal muscle texture features of the ultrasonic image. This solution analyzes the motion damage in the texture area of ​​the image to obtain the most suitable run direction. Then, based on the pixel distribution in the amplitude spectrum and the grayscale fluctuation in the run direction, the grayscale run matrix is ​​adjusted to adaptively extract skeletal muscle texture features, thereby improving the accuracy of skeletal muscle texture feature extraction for motion-injured skeletal muscles. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0035] Figure 1 A flowchart of a skeletal muscle texture feature extraction method for sports injury assessment provided by one embodiment of the present invention;

[0036] Figure 2 A two-dimensional ultrasound image of the muscle tissue of the upper limb on the healthy side provided by one embodiment of the present invention;

[0037] Figure 3 A two-dimensional ultrasonic image of the injured upper limb muscle tissue provided by one embodiment of the present invention;

[0038] Figure 4 A diagram showing the result of edge detection of an ultrasound image provided by one embodiment of the present invention;

[0039] Figure 5 A binary image of an amplitude spectrum provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0040] In order to further illustrate the technical means and effects adopted by the present invention to achieve the predetermined purpose of the invention, the following, in conjunction with the accompanying drawings and preferred embodiments, describes in detail the specific implementation, structure, characteristics and effects of a skeletal muscle texture feature extraction method for sports injury assessment proposed by the present invention. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics of one or more embodiments may be combined in any suitable form.

[0041] Unless defined otherwise, 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 invention belongs.

[0042] The following describes in detail a specific solution of a skeletal muscle texture feature extraction method for sports injury assessment provided by the present invention with reference to the accompanying drawings.

[0043] See also Figure 1 , which shows a flow chart of a skeletal muscle texture feature extraction method for sports injury assessment provided by one embodiment of the present invention, specifically comprising:

[0044] Step S1: Acquire an ultrasound image to be analyzed and establish a coordinate system; extract edge lines and their edge directions in the ultrasound image, and acquire the texture area of ​​the ultrasound image based on the spatial distribution of the edge lines and the difference in edge directions.

[0045] Sports and physical labor can easily lead to muscle damage. Magnetic resonance imaging (MRI) has long been the leading imaging method for evaluating muscle pathology. With technological advancements in ultrasound equipment and continuous improvements in ultrasound resolution, ultrasound has gradually become a key imaging method for diagnosing and evaluating muscle damage.

[0046] In one embodiment of the present invention, a high-frequency linear array probe (12-18 MHz) is used, and the axial resolution needs to be ≤ 0.1 mm to identify muscle fiber rupture;

[0047] The injured muscle group is placed in the anatomically neutral position (e.g., hip flexion 30°, knee extension 0° for rectus femoris assessment), and a fixed brace is used to eliminate interference from voluntary contraction.

[0048] The scanning direction of the probe should be parallel to the direction of muscle fiber running to ensure clear observation of muscle fiber arrangement and damage;

[0049] An ultrasound image of the target area (the area to be analyzed) of the current patient is acquired, and the image is pre-processed by filtering and noise reduction. In this example, the commonly used non-local mean filtering method is used to filter the ultrasound image.

[0050] See also Figure 2 , which shows a two-dimensional ultrasound image of the healthy upper limb muscle tissue provided by an embodiment of the present invention, please refer to Figure 3 , which shows a two-dimensional ultrasound image of the injured upper limb muscle tissue provided by an embodiment of the present invention. Figure 2 and Figure 3 a and b are the identification numbers of the ultrasound images. Figure 2 and Figure 3 It can be seen that the skeletal muscle texture on the healthy side is fibrous strip-like, the muscle fiber texture is consistent in direction, and the texture is clearer.

[0051] A two-dimensional rectangular coordinate system is established with the lower left corner of the ultrasound image as the origin for subsequent analysis, where the horizontal axis is the X-axis.

[0052] Since skeletal muscle has fiber bundle-like texture features in ultrasound images, we first extract the edge lines and edge directions in the ultrasound image. Figure 4 , which shows a result diagram of edge detection of an ultrasound image provided by an embodiment of the present invention; Figure 4 The Canny operator is used to perform edge detection on ultrasound images.

[0053] Taking into account the differences in muscle morphology in different areas, the distribution of muscle fibers is also different. Therefore, obtaining the texture area of ​​the ultrasound image based on the spatial distribution of the edge line and the difference in edge direction can more accurately fit the real muscle fiber structure, which is more conducive to analyzing the texture structure of local muscle fibers and calculating the stroke direction.

[0054] Preferably, in one embodiment of the present invention, considering that the principal axis direction of each edge line represents the local texture direction of the structure, the principal axis of the edge line is obtained by using the PCA principal component analysis algorithm, and the direction corresponding to the minimum angle between the principal axis and the X-axis of the coordinate system is taken as the edge direction of the edge line.

[0055] Considering that the farther the distance between edge lines is and the greater the difference in edge directions is, the less likely it is that they are the same local muscle area, the cluster distance is obtained based on the distance between the midpoints of the edge lines and the difference in edge directions;

[0056] Considering that the convex hull is the smallest convex polygon that surrounds all edge line point sets, it can contain all edge lines in the cluster, and the edges are closed and coherent, simplifying the regional shape. Therefore, all edge lines are clustered and classified based on the cluster distance, and the convex hull area of ​​each cluster is used as a texture area.

[0057] As an example, the point corresponding to the average value of the horizontal coordinate and the average value of the vertical coordinate of all pixels on the edge line is taken as the midpoint, and the Euclidean distance between the midpoints of two edge lines is taken as the distance between the midpoints of the edge lines, thereby expressing the spatial distribution of the edge lines; the absolute value of the difference in edge directions is taken as the direction difference value, thereby expressing the difference in edge directions; the arithmetic square root of the sum of the square of the distance between the midpoints of two edge lines and the square of the direction difference value is taken as the cluster distance of the two edge lines;

[0058] The clustering parameter K is obtained by the elbow method, and the K-means clustering algorithm is used to obtain clusters of multiple edge lines, and then a group of line segments that are concentrated in space and similar in direction is obtained, and then the convex hull area is constructed to obtain the texture area.

[0059] It should be noted that the unit of angle can be radians. Before calculating the cluster distance, the distance and direction difference values ​​between the midpoints of the two edge lines can be linearly normalized to eliminate the unit effect and reduce the impact of data magnitude. In other embodiments of the present invention, degrees can also be used as the angle unit, and the distance and direction difference values ​​between the midpoints of the two edge lines can be weighted and summed to obtain the cluster distance.

[0060] It should be noted that the linear normalization is performed under the corresponding data dimension; Euclidean distance, edge detection using the Canny operator, convex hull area construction method, PCA algorithm, K-means clustering algorithm and elbow method are all technical means well known to those skilled in the art and will not be repeated here.

[0061] The processing process for each ultrasound image is the same, and only one example is described here.

[0062] Step S2: Based on the directional similarity between each texture region and the nearest texture region, combined with the grayscale features in the non-texture region between the texture region and its minimum circumscribed circle, the motion impairment index of each texture region is obtained; the texture directions of all texture regions are weighted with the motion impairment index as the weight to obtain the run direction of the ultrasound image.

[0063] Considering that sports injuries are mainly acute injuries, mainly manifested as muscle fiber rupture, the directions of similar texture areas still have a high degree of similarity, and there are local hematomas or edema areas around the muscles, which affect the grayscale characteristics of adjacent non-texture areas; other injuries are usually manifested as disordered muscle fiber arrangement, without muscle fiber rupture characteristics, and usually without local hematomas or edema areas. Therefore, based on the directional similarity between each texture area and the nearest texture area, combined with the grayscale characteristics of the non-texture area between the texture area and its minimum circumscribed circle, the sports injury index of each texture area is obtained to characterize the possibility of each texture area corresponding to sports injury, preparing for the subsequent adjustment of the stroke direction and accurately extracting the skeletal muscle texture characteristics of sports injuries.

[0064] Preferably, in one embodiment of the present invention, the nearest other texture regions corresponding to each texture region are first determined. Considering that normal muscle fibers are arranged along a relatively uniform direction, even if they are segmented into multiple texture regions in an ultrasound image, these regions are usually structurally related by extension rather than by pure proximity. Finding the nearest region by extending the entire texture direction of the texture region is more consistent with the extension and propagation path of the muscle fibers.

[0065] Considering that in ultrasound images, edema or hematoma appears anechoic or uniformly hypoechoic, the corresponding grayscale shows a small mean grayscale value, and the grayscale is uniform and the grayscale value variance is small;

[0066] Based on this, the mean of the edge directions of all edge lines in each texture region is taken as the corresponding texture direction; a straight line is extended in the texture direction through the centroid of any texture region, and the other texture regions with the closest centroid to the extended straight line are obtained as the corresponding nearest texture regions;

[0067] For any textured area, the motion impairment index is obtained according to the grayscale value mean and grayscale value variance in each corresponding non-textured area, combined with the similarity of the texture direction with the texture direction of the corresponding nearest textured area.

[0068] Select any texture area as an example, and take the inverse of the sum of the absolute value of the difference in texture direction between the selected texture area and the corresponding nearest texture area and a preset positive parameter divided by zero, such as 0.01, as the direction similarity index; take the inverse of the product of the grayscale value mean and grayscale value variance of each non-texture area as the liquid area index of the non-texture area, showing the grayscale characteristics in the non-texture area;

[0069] The product of the directional similarity index of the selected texture area and the maximum liquid area index of all corresponding non-texture areas is used as the motion damage index of the selected texture area.

[0070] The difference in texture direction between different texture areas is expressed by the absolute value of the difference, and then the similarity of texture direction is expressed by taking the reciprocal for negative correlation adjustment. At the same time, to prevent the denominator from being zero, a preset positive parameter of zero division is incorporated before taking the reciprocal to obtain the direction similarity index, which represents the direction similarity between the texture area and the nearest texture area. The larger the direction similarity index, the more the selected texture area and the corresponding nearest texture area conform to the similarity characteristics of the texture direction of muscle fibers in the case of sports injury, and the larger the sports injury index.

[0071] The smaller the gray value mean and gray value variance in the non-texture area, the more likely the corresponding area is anechoic or uniformly hypoechoic, indicating that the non-texture area around the selected texture area is more likely to be a local hematoma or edema area. The larger the fluid area index, the greater the motion injury index.

[0072] It should be noted that the method of obtaining the minimum circumscribed circle is an existing technology. If the selected texture area does not have a nearest texture area or a non-texture area, its motion impairment index is set to 0.

[0073] In another embodiment of the present invention, when searching for the nearest texture region, the distance between the centroids of the texture regions may be added as an additional constraint. For example, a straight line is extended in the texture direction through the centroid of one texture region, and the distance from the centroid of the other texture region to the extended straight line is obtained. The distance between the centroids of the two texture regions is also obtained. The two distances are then fused with weights of 0.7 and 0.3 to obtain a matching distance. Since the texture direction needs to be given more attention when analyzing motion injuries, 0.7 corresponds to the distance from the centroid to the extended straight line, and 0.3 corresponds to the distance between the centroids.

[0074] The other texture region with the smallest matching distance is selected as the corresponding nearest texture region.

[0075] The motion impairment index of each texture region is obtained in the same way.

[0076] To obtain more accurate and complete texture features of skeletal muscle motion injury, it is necessary to extract the run angle of the grayscale run matrix based on the motion injury index of each texture region. Therefore, the texture directions of all texture regions are further weighted using the motion injury index to obtain the run direction of the ultrasound image.

[0077] Preferably, in one embodiment of the present invention, considering that the larger the motion injury index is, the more likely the corresponding texture area is to exhibit motion injury, then the texture direction of the corresponding texture area needs more attention, and the run direction should be more inclined to the texture direction, so the softmax function is used to normalize the motion injury index of all texture areas, and the normalized motion injury index is used as the weight of the weighted summation of the corresponding texture direction, and the weighted summation result obtained is used as the run direction of the ultrasound image.

[0078] It should be noted that the softmax function can normalize the data, and the overall sum after normalization is 1. This is a technical means well known to those skilled in the art and will not be described in detail.

[0079] Step S3: Obtain the amplitude spectrum of the ultrasound image; obtain the skeletal muscle damage index of the ultrasound image based on the grayscale value and spatial distribution of the foreground pixels in the amplitude spectrum; divide the pixels according to the distribution characteristics of the pixels in the run direction to obtain the pixel sequence to be analyzed; obtain the muscle fiber significance index of the ultrasound image based on the grayscale fluctuation of the pixel sequence to be analyzed.

[0080] In ultrasound images of normal muscle areas, muscle fibers are regularly arranged and the muscle bundles are clearly layered. However, in ultrasound images of muscle damage, muscle fibers may appear interrupted or disordered. Muscle texture is a fiber bundle structure with a certain degree of directionality. When extracting muscle texture features, the higher the degree of skeletal muscle damage, the more complex the muscle texture features (for example, hematoma edema from sports injuries, muscle atrophy from chronic injuries, etc.). During the grayscale interval division process, frequency domain analysis can be used to assess muscle texture damage based on the degree of texture disorder in the image and adjust the division accuracy.

[0081] Ultrasound images are processed using discrete Fourier transform to obtain the frequency domain matrix of the image; the frequency domain matrix is ​​a matrix of the same size as the image, and the value of each point in the matrix is ​​an imaginary number. The modulus of this imaginary number represents the amplitude of the waveform surface corresponding to the point, and the direction of the imaginary number represents the phase angle of the waveform surface corresponding to the point.

[0082] According to the frequency domain matrix, the amplitude spectrum of the image is obtained. Each point in the amplitude spectrum describes a plane wave, the distance from each point to the center point describes the frequency characteristics, the direction from the center point to this point is the direction of the plane wave, and the grayscale value of the point describes the amplitude of the plane wave.

[0083] Since the foreground pixels in the amplitude spectrum usually correspond to frequency components with significant texture features in the ultrasound image, analyzing such pixels can effectively eliminate background noise and irrelevant frequencies, highlighting the main direction and structural change characteristics of the muscle texture. At the same time, considering that the image texture of healthy muscle is regular and clear, the energy distribution of the amplitude spectrum is regular: the energy of the amplitude spectrum is concentrated in the low-frequency region and presents a centrally symmetric Gaussian distribution; while the amplitude spectrum of muscle damage presents a relatively dispersed energy distribution. Therefore, the skeletal muscle damage index of the ultrasound image is obtained based on the grayscale value and spatial distribution of the foreground pixels in the amplitude spectrum, which characterizes the degree of muscle damage and provides a basis for adjusting the grayscale run matrix.

[0084] Preferably, in one embodiment of the present invention, the amplitude spectrum is subjected to OTSU binary segmentation processing to extract foreground pixels with a pixel value of 1; see Figure 5 , which shows a binary image of an amplitude spectrum provided by an embodiment of the present invention.

[0085] Obtain foreground pixels of the amplitude spectrum from the binary image of the amplitude spectrum; to facilitate analysis of the spatial distribution of the foreground pixels, establish a two-dimensional coordinate system with the center point of the amplitude spectrum as the origin, the horizontal axis as the X-axis, and the vertical axis as the Y-axis; and use the angle between the line connecting the foreground pixel and the origin and the positive direction of the X-axis as the azimuth of the foreground pixel;

[0086] Considering that the farther the foreground pixel is from the coordinate origin, the higher the frequency is, and the less it conforms to the frequency characteristics of healthy muscle; the greater the angular difference between the foreground pixel and the foreground pixels in the same quadrant, the less concentrated the distribution is, and the more likely muscle damage is; and considering that the larger the grayscale value of the foreground pixel in the amplitude spectrum, the more significant the frequency component appears in the original image, the higher the corresponding image texture intensity, and the higher the degree of attention should be paid;

[0087] Based on this, the skeletal muscle damage index is obtained according to the distance between each foreground pixel and the coordinate origin in the amplitude spectrum, the difference in azimuth between the foreground pixel and the foreground pixels in the same quadrant, and the grayscale value of the foreground pixel in the amplitude spectrum.

[0088] As an example, to eliminate the influence of the absolute number of foreground pixels, the grayscale values ​​of all foreground pixels in the amplitude spectrum are normalized by the softmax function as grayscale weights; the average value of the azimuth angles of the foreground pixels in each quadrant is used as the reference azimuth angle of each quadrant; the product of the distance between the foreground pixel and the coordinate origin and the absolute value of the difference between the foreground pixel and the corresponding reference azimuth angle is used as the damage factor of the foreground pixel; the grayscale weight is used as the weighted summation weight, and the damage factors of all foreground pixels are weighted summed, and the weighted summation result is used as the skeletal muscle damage index.

[0089] Among them, the softmax function will amplify the weights of foreground pixels with significantly larger grayscale values, while avoiding the influence of the number of foreground pixels caused by factors such as ultrasound image resolution, and use the grayscale weight to represent the grayscale value of the foreground pixels for weighted summation; the spatial distribution of foreground pixels is analyzed from two perspectives: azimuth angle difference and distance from the origin. The larger the damage factor of the foreground pixel, the more discrete the spatial distribution, the more likely it is that muscle damage occurs, and the larger the skeletal muscle damage index, the more detailed the grayscale interval distinction of the texture extracted by the grayscale run matrix should be.

[0090] In addition to analyzing the degree of skeletal muscle damage and adjusting the grayscale interval division, it is also necessary to adjust the grayscale interval division based on the grayscale value difference characteristics of the pixel sequence in the actual stroke direction, especially the grayscale value distinction between the pixels representing texture information and other pixels.

[0091] The pixels representing texture information in ultrasound images are muscle fiber pixels. These pixels are mainly distributed in the texture area of ​​the image. In addition to edge pixels, the pixels between edges are also pixels that make up muscle fibers. The echo intensity of the same muscle fiber bundle is usually similar. In order to accurately analyze the significant features of muscle fibers and further provide a basis for adjusting the grayscale run matrix, the pixels are first divided according to the distribution characteristics of the pixels in the run direction to obtain the pixel sequence to be analyzed; then, based on the grayscale fluctuations of the pixel sequence to be analyzed, the texture clarity of the pixel sequence is analyzed to obtain the muscle fiber significance index of the ultrasound image.

[0092] Preferably, in one embodiment of the present invention, to facilitate analysis of pixel points between edge lines, considering that muscle fibers have a consistent arrangement direction, the texture area is divided based on edge direction lines of the center points of all edge lines. This can accurately segment the texture area into several texture sub-regions with consistent directions, which is beneficial for subsequent acquisition of a pixel sequence with clear grayscale texture fluctuations.

[0093] Therefore, first, within the texture area, the texture area is divided by all straight lines in the edge direction passing through the center point of the edge line to obtain texture sub-areas.

[0094] Considering that not all texture sub-regions are muscle fiber regions and the grayscale of muscle fibers is more uniform, the muscle fiber sub-regions are screened out according to the grayscale inhomogeneity within the texture sub-regions, which facilitates the detailed division of pixels in the run direction.

[0095] As an example, considering that the variance is smaller when the grayscale of muscle fibers is more uniform, the grayscale value variance is used to represent the grayscale inhomogeneity within the texture sub-region. Therefore, the grayscale value variances of all texture sub-regions are sorted from small to large, and the sequence is divided at the point where the difference is the largest in the sorted sequence. The texture sub-region of the part of the sequence where the minimum grayscale value variance is located is marked as a muscle fiber sub-region.

[0096] For example, there are 20 texture sub-regions, corresponding to 20 grayscale value variances. The grayscale value variances are sorted from small to large. The difference between the 12th and 13th grayscale value variances in the sorted sequence is the largest. The sequence is divided here into 1-12 and 13-20. The part of the sequence with the smallest grayscale value variance corresponds to 1-12. The texture sub-regions numbered 1-12 are marked as muscle fiber sub-regions.

[0097] The ultrasound image is further traversed in the run direction to obtain all pixel sequences to be divided to adapt to the texture direction of the muscle fiber. For example, when the run direction is horizontal, the pixel sequence to be divided is a sequence of pixel points in rows.

[0098] After obtaining the pixel sequence to be divided, it can be divided: first, the pixel sequence to be divided is divided for the first time according to whether it belongs to the texture area, and the pixel sequence of the texture segment and the non-texture segment is distinguished; further, the pixel sequence of the first division is divided for the second time according to whether it belongs to the muscle fiber sub-region, and is finely divided into muscle fiber-related segments and other texture segments; finally, the pixel sequence of the second division is divided for the third time according to the muscle fiber sub-region to which it belongs, and is independently segmented according to its corresponding muscle fiber sub-region, ensuring the consistency of the grayscale texture within the sequence and avoiding the interference of different structural information on the same sequence analysis;

[0099] The pixel sequence finally obtained by division is used as the pixel sequence to be analyzed, and the structural alignment of the pixel sequence is completed to correspond to the physiological area.

[0100] It should be noted that all original pixel information is retained, and the pixel sequences obtained by the three divisions are all used as the pixel sequences to be analyzed; the straight line in the edge direction passing through the center point of the edge line is called the division line; considering that some texture areas have overlapping parts, the texture areas with overlapping parts are merged before dividing the sub-areas. For non-overlapping texture areas, only the division lines within the corresponding texture area are used for division; for overlapping parts, the division lines of all texture areas including the overlapping parts are used for division;

[0101] In another embodiment of the present invention, texture areas with overlapping parts can also be merged and regarded as a larger texture area, and divided by dividing lines within the merged texture area; a variance threshold can also be set to mark texture sub-areas with grayscale value variance less than the variance threshold as muscle fiber sub-areas.

[0102] Preferably, in one embodiment of the present invention, considering that the muscle fiber texture is clear and not affected by other damaged tissues, the grayscale features of each pixel sequence in the current run direction have strong regularity, which is manifested as: the pixels of the same muscle fiber have similar grayscale values ​​and are divided into the same pixel sequence to be analyzed;

[0103] The echo intensity of pixels within the same muscle fiber is similar, and the grayscale value variance is small. However, different damaged tissues or muscle fibers may have different echo intensities and grayscale values. In this case, the grayscale value variance within the pixel sequence to be analyzed is small, and the grayscale value variance between different pixel sequences to be analyzed is large. This indicates that the muscle fiber texture of the ultrasound image is clearer and the muscle fiber significance index is larger.

[0104] Based on this, the variance of the grayscale value of the pixel sequence to be analyzed and the variance of the mean grayscale value within the pixel sequence to be analyzed are fused to obtain the muscle fiber significance index of the ultrasound image.

[0105] As an example, the variance of the mean grayscale values ​​in all pixel sequences to be analyzed is used as the numerator, the grayscale value variance of each pixel sequence to be analyzed is used as the denominator, and the fractional ratio is used as the texture saliency index of each pixel sequence to be analyzed, which represents the grayscale fluctuation of the pixel sequence to be analyzed; the mean of the texture saliency index of all pixel sequences to be analyzed is used as the muscle fiber saliency index of the ultrasound image. The larger the muscle fiber saliency index is, the more it indicates that the skeletal muscle texture features can be extracted without very detailed grayscale interval division.

[0106] Step S4: adjusting the grayscale run matrix according to the run direction, skeletal muscle damage index and muscle fiber significance index to extract skeletal muscle texture features of the ultrasound image.

[0107] The most suitable run direction of the ultrasound image is determined through step S2, and then the skeletal muscle damage index and the muscle fiber significance index are obtained through step S3. The degree of skeletal muscle damage and the degree of muscle fiber significance are analyzed to prepare for adjusting the grayscale run matrix. Therefore, the grayscale run matrix is ​​finally adjusted according to the run direction, skeletal muscle damage index and muscle fiber significance index to extract the skeletal muscle texture features of the ultrasound image.

[0108] Preferably, in one embodiment of the present invention, considering that the larger the skeletal muscle damage index is, the more detailed the grayscale interval distinction of the grayscale run matrix for extracting the texture should be, and the more grayscale intervals should be present; the larger the muscle fiber significance index is, the more it indicates that skeletal muscle texture features can be extracted without very detailed grayscale interval division, and the fewer the number of grayscale intervals should be, therefore, the skeletal muscle damage index and the muscle fiber significance index are fused to obtain a quantity adjustment factor, providing a basis for adjusting the grayscale intervals;

[0109] Then, the preset number of grayscale intervals is adjusted based on the number adjustment factor to obtain the appropriate number of intervals, thereby obtaining the number of grayscale intervals suitable for the skeletal muscle texture features in the ultrasound image.

[0110] As an example, the preset number of grayscale intervals is 16, the ratio of the skeletal muscle damage index to the muscle fiber significance index is used as the independent variable, and after mapping through the tanh(x) function, the sum of the mapped value and the constant 1 is used as the quantity adjustment factor; the product of the quantity adjustment factor and the preset number of grayscale intervals is rounded up, and the rounded result is used as the appropriate interval number; x represents the independent variable.

[0111] When the number of suitable intervals cannot be divided evenly by 256, the number of gray levels contained in the gray interval of the highest gray level is adjusted to the remainder of 256 divided by the number of suitable intervals.

[0112] Finally, the grayscale run matrix is ​​set according to the run direction and the appropriate number of intervals to extract the skeletal muscle texture features of the ultrasound image, thereby increasing the grayscale run matrix's ability to express the texture features of skeletal muscle sports injuries. This scheme does not involve the diagnosis and treatment of diseases, but is only used to better assist clinicians in analyzing the degree of injury.

[0113] In summary, in order to solve the technical problem that the existing texture extraction is easily interfered by other areas or textures, which affects the extraction accuracy of skeletal muscle texture features, the present invention proposes a skeletal muscle texture feature extraction method for motion injury assessment. The present invention first extracts the edge lines and their edge directions and texture areas in the ultrasound image; further analyzes the motion injury situation of each texture area, and weights the texture directions of all texture areas with the motion injury index as the weight to obtain the run direction of the ultrasound image; further, according to the grayscale value and spatial distribution of the foreground pixel points in the amplitude spectrum, combined with the grayscale fluctuation of the pixel sequence to be analyzed in the run direction, and the run direction, adjusts the grayscale run matrix to extract the skeletal muscle texture features of the ultrasound image. This scheme obtains the most suitable run direction by analyzing the motion injury situation of the texture area in the image, and then combines the pixel point distribution in the amplitude spectrum with the grayscale fluctuation in the run direction to construct a multi-dimensional index to adjust the grayscale run matrix, adaptively extract skeletal muscle texture features, and improve the accuracy of skeletal muscle texture feature extraction for motion injuries.

[0114] It should be noted that the order in which the embodiments of the present invention are described above is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0115] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments.

Claims

1. A skeletal muscle texture feature extraction method for sports injury assessment, which does not involve disease diagnosis and treatment, is characterized by: The method comprises: Acquire an ultrasound image to be analyzed and establish a coordinate system; extract edge lines and their edge directions in the ultrasound image, and obtain the texture area of ​​the ultrasound image based on the spatial distribution of the edge lines and the difference in edge directions; Based on the directional similarity between each texture region and the nearest texture region, combined with the grayscale characteristics of the non-texture region between the texture region and its minimum circumscribed circle, the motion impairment index of each texture region is obtained; the texture directions of all texture regions are weighted using the motion impairment index to obtain the run direction of the ultrasound image; Obtain the amplitude spectrum of the ultrasound image; obtain the skeletal muscle damage index of the ultrasound image based on the grayscale value and spatial distribution of the foreground pixels in the amplitude spectrum; divide the pixels according to the distribution characteristics of the pixels in the run direction to obtain a pixel sequence to be analyzed; obtain the muscle fiber significance index of the ultrasound image based on the grayscale fluctuation of the pixel sequence to be analyzed; The grayscale run matrix is ​​adjusted according to the run direction, skeletal muscle damage index and muscle fiber significance index to extract the skeletal muscle texture features of the ultrasound image; The method for obtaining the sports injury index includes: taking the average of the edge directions of all the edge lines in each texture region as the corresponding texture direction; extending a straight line through the centroid of any texture region in the texture direction, and obtaining the other texture region whose centroid is closest to the extended straight line as the corresponding nearest texture region; For any of the texture areas, the motion impairment index is obtained according to the grayscale value mean and grayscale value variance in each corresponding non-texture area, combined with the similarity between the texture direction and the texture direction of the corresponding nearest texture area; The method for obtaining the skeletal muscle damage index comprises: Obtaining a foreground pixel of the amplitude spectrum from the binary image of the amplitude spectrum; establishing a two-dimensional coordinate system with the center point of the amplitude spectrum as the origin, the horizontal axis as the X-axis, and the vertical axis as the Y-axis; and using the angle between the line connecting the foreground pixel and the origin and the positive direction of the X-axis as the azimuth angle of the foreground pixel; The skeletal muscle injury index is obtained based on the distance between each foreground pixel point and the coordinate origin in the amplitude spectrum, the difference in the azimuth angle between the foreground pixel point and the foreground pixel points in the same quadrant, and the grayscale value of the foreground pixel point in the amplitude spectrum.

2. The skeletal muscle texture feature extraction method for sports injury assessment according to claim 1, characterized in that: The method for obtaining the texture area includes: Obtaining a cluster distance based on the distance between the midpoints of the edge lines and the difference in the edge directions; All the edge lines are clustered and classified based on the cluster distance, and the convex hull area of ​​each cluster is used as a texture area.

3. The skeletal muscle texture feature extraction method for sports injury assessment according to claim 1, characterized in that: The method for acquiring the pixel sequence to be analyzed includes: In the texture region, the texture region is divided by all straight lines in the edge direction passing through the center point of the edge line to obtain texture sub-regions; and muscle fiber sub-regions are screened out based on the grayscale non-uniformity in the texture sub-regions; Traversing the ultrasound image in the run direction to obtain all pixel sequences to be divided; performing a first division on the pixel sequence to be divided according to whether it belongs to the texture area; performing a second division on the pixel sequence obtained by the first division according to whether it belongs to the muscle fiber sub-area; and performing a third division on the pixel sequence obtained by the second division according to the muscle fiber sub-area to which it belongs; The pixel sequence finally obtained by division is used as the pixel sequence to be analyzed.

4. The method for extracting skeletal muscle texture features for sports injury assessment according to claim 3, characterized in that: The method for obtaining the muscle fiber sub-region comprises: The grayscale value variances of all the texture sub-regions are sorted from small to large, the sequence is divided at the point where the difference is the largest in the sorted sequence, and the texture sub-region of the partial sequence where the minimum grayscale value variance is located is marked as a muscle fiber sub-region.

5. The method for extracting skeletal muscle texture features for sports injury assessment according to claim 1, characterized in that: The method for obtaining the muscle fiber significance index comprises: The grayscale value variance of the pixel sequence to be analyzed and the variance of the grayscale value mean within the pixel sequence to be analyzed are combined to obtain the muscle fiber significance index of the ultrasound image.

6. The method for extracting skeletal muscle texture features for sports injury assessment according to claim 1, characterized in that: The method of adjusting the grayscale run-length matrix to extract the skeletal muscle texture features of the ultrasound image includes: The skeletal muscle injury index and the muscle fiber significance index are integrated to obtain a quantity adjustment factor; the preset number of grayscale intervals is adjusted based on the quantity adjustment factor to obtain an appropriate number of intervals; a grayscale run matrix is ​​set according to the run direction and the appropriate number of intervals to extract the skeletal muscle texture features of the ultrasound image.

7. The method for extracting skeletal muscle texture features for sports injury assessment according to claim 1, characterized in that: The method for obtaining the run direction includes: The motion impairment index of all the texture regions is normalized using a softmax function, the normalized motion impairment index is used as the weight of the weighted summation corresponding to the texture direction, and the obtained weighted summation result is used as the run direction of the ultrasound image.

8. The method for extracting skeletal muscle texture features for sports injury assessment according to claim 1, characterized in that: The method for obtaining the edge direction includes: The principal axis of the edge line is obtained by the PCA algorithm, and the direction corresponding to the minimum angle between the principal axis and the X-axis of the coordinate system is used as the edge direction of the edge line.

Images