Intelligent detection method of meniscus injury based on image processing

Abstract

The present invention relates to the field of meniscus injury detection, and specifically to an intelligent detection method for meniscus injury based on image processing. The method first obtains slice images of different layers of the meniscus of a patient's knee joint, extracts the tissue area of ​​the slice image, obtains the noise interference degree of the tissue area according to the distribution of the grayscale value of the pixel points in the tissue area, divides the tissue area into a meniscus area and a tissue synovial fluid area, and screens out the damaged pixels in the meniscus area, obtains the damage factor of each layer of the slice image according to the position distribution of the damaged pixels, the difference in the number of pixels between the meniscus area and the tissue synovial fluid area, and the noise interference degree of the tissue area, obtains the damage factor of each layer of the slice image according to the difference in the edge shape of the meniscus area between adjacent layers of the slice image, and detects the patient's meniscus injury based on the damage coefficient and the damage factor of each layer of the slice image. The present invention can improve the accuracy of meniscus injury detection.

Description

Technical Field

[0001] The invention relates to the field of meniscus injury detection, and in particular to an intelligent meniscus injury detection method based on image processing. Background Art

[0002] The meniscus is an important structure in the knee joint, mainly composed of cartilage tissue and shaped like a half moon. It has a cushioning effect and maintains the stability of the knee joint. It can evenly distribute the stability of the knee joint to reduce wear and tear, and help to evenly distribute the synovial fluid, thereby maintaining joint lubrication. When the patient engages in strenuous exercise, twisting or weight-bearing, it can cause the meniscus to tear. Therefore, it is necessary to detect the meniscus injury in order to treat it.

[0003] In the related art, multi-layer magnetic resonance slice images are usually obtained by using magnetic resonance imaging technology, and the meniscus injury characteristics of each layer of slice images are identified to detect whether the patient has meniscus injury. However, since magnetic resonance slice images are easily interfered by factors such as noise, and the inter-layer spacing between each layer of slice images is improperly set, the meniscus injury characteristics cannot be clearly and completely displayed in each layer of slice images, thereby reducing the accuracy of meniscus injury detection. Summary of the invention

[0004] In order to solve the technical problem that the characteristics of meniscus injury cannot be clearly and completely displayed in each layer of slice image, which reduces the accuracy of meniscus injury detection, the purpose of the present invention is to provide an intelligent detection method for meniscus injury based on image processing, and the technical solution adopted is as follows:

[0005] The present invention proposes an intelligent detection method for meniscus injury based on image processing, the method comprising:

[0006] An intelligent detection method for meniscus injury based on image processing, the method comprising:

[0007] Obtain slice images of different layers of the meniscus at the patient's knee joint;

[0008] A slice image of any layer is used as a target slice image, and edge detection is performed on the target slice image to obtain a tissue region in the target slice image; the noise interference degree of the tissue region is obtained according to the distribution of the grayscale values ​​of the pixels in the tissue region; threshold segmentation is performed on the tissue region to obtain a meniscus region and a tissue synovial region in the target slice image; damaged pixels are screened out from all pixels in the meniscus region according to the grayscale values ​​and gradient values ​​of the pixels in the meniscus region; a damage factor of the meniscus region of the target slice image is obtained according to the position distribution of damaged pixels in the meniscus region, the difference in the number of pixels between the meniscus region and the tissue synovial region, and the noise interference degree of the tissue region;

[0009] According to the position trend of the pixel points on the edge of the meniscus region in the target slice image, the edge smoothness of the meniscus region of the target slice image is obtained; two other slice images adjacent to the target slice image are used as reference slice images, and according to the difference in the edge smoothness of the meniscus region between the target slice image and the reference slice image, and the edge smoothness of the meniscus region of all slice images, the patient's injury coefficient is obtained; according to the injury coefficient and the injury factor of the meniscus region of all slice images, the patient's injury possibility is obtained;

[0010] Based on the injury possibility, the patient's meniscus injury is detected.

[0011] Furthermore, obtaining the noise interference degree of the tissue region includes:

[0012] Extracting a plurality of maximum pixel points from the tissue region in the target slice image, wherein the grayscale value of the maximum pixel point is greater than the grayscale values ​​of all other pixel points except the maximum pixel point in a preset neighborhood centered on the maximum pixel point;

[0013] Take any two maximum pixel points as a maximum point group, take the average of the grayscale values ​​of the two maximum pixel points in each maximum point group as the first grayscale eigenvalue of each maximum point group; take the average of the grayscale values ​​of all the pixel points on the line connecting the two maximum pixel points in each maximum point group as the second grayscale eigenvalue of each maximum point group; take the absolute value of the difference between the first grayscale eigenvalue and the second grayscale eigenvalue as the initial grayscale change degree of each maximum point group;

[0014] Performing negative correlation mapping on the distance between two maximum pixel points in each maximum point group to obtain a first weight parameter of each maximum point group; taking the product value of the first weight parameter and the initial grayscale change degree as the actual grayscale change degree of each maximum point group;

[0015] The average values ​​of the true grayscale variation degrees of all maximum point groups are normalized to obtain the noise interference degree of the tissue area.

[0016] Furthermore, screening out damaged pixels from all pixels in the meniscus region includes:

[0017] In the meniscus region of the target slice image, the average value of the gradient values ​​of all the pixels is taken as the overall gradient value of the meniscus region; the average value of the grayscale values ​​of all the pixels is taken as the overall grayscale value of the meniscus region;

[0018] Pixels in the meniscus region whose gradient values ​​are smaller than the overall gradient value and whose grayscale values ​​are larger than the overall grayscale value are regarded as damaged pixels.

[0019] Furthermore, the damage factors of the meniscus region of the target slice image are obtained including:

[0020] The connected domains formed by the damaged pixels in the meniscus region of the target slice image are used as the damaged regions in the meniscus region; the number of pixels in each damaged region is used as the numerator, the number of pixels in the minimum circumscribed rectangle of each damaged region is used as the denominator, and the ratio is used as the aggregation degree of each damaged region; the average value of the aggregation degrees of all damaged regions in the meniscus region is used as the first damage parameter of the meniscus region of the target slice image;

[0021] The number of pixels in the tissue synovial region is used as a numerator, the number of pixels in the meniscus region is used as a denominator, and the ratio is used as a second injury parameter of the meniscus region of the target slice image;

[0022] integrating the first injury parameter and the second injury parameter to obtain a comprehensive injury parameter of the meniscus region of the target slice image;

[0023] The inverse of the sum of the noise interference degree of the tissue area of ​​the target slice image and the preset first adjustment parameter is used as the second weight parameter of the meniscus area of ​​the target slice image;

[0024] The product value of the second weight parameter and the comprehensive damage parameter is normalized to obtain the damage factor of the meniscus area of ​​the target slice image.

[0025] Further, the step of obtaining the edge smoothness of the meniscus region of the target slice image includes:

[0026] An eight-chain code sequence of the edge of the meniscus region in the target slice image is obtained, and in the eight-chain code sequence, the absolute value of the difference between each element and the next adjacent element is used as the variation of each element;

[0027] The variance of the variation of all elements in the eight-chain code sequence is negatively correlated to obtain the edge smoothness of the meniscus region of the target slice image.

[0028] Further, obtaining the patient's injury coefficient includes:

[0029] Taking the average of the edge smoothness of the meniscus region of the two reference slice images of the target slice image as the reference edge smoothness of the target slice image;

[0030] The absolute value of the difference between the edge smoothness of the meniscus region of the target slice image and the reference edge smoothness is used as an edge smoothness difference parameter of the meniscus region of the target slice image;

[0031] The average value of the edge smoothness difference parameters of the meniscus region of the slice images of all layers is taken as the overall edge smoothness difference of the patient;

[0032] an average value of the edge smoothness of the meniscus region of the slice images of all layers as the third weight parameter of the patient;

[0033] The product value of the third weight parameter and the overall edge smoothness difference is normalized to obtain the patient's injury coefficient.

[0034] Further, obtaining the patient's injury probability includes:

[0035] The average value of the damage factors of the meniscus region of the slice images of all layers is taken as the overall damage factor of the patient;

[0036] The overall injury factor and the injury coefficient are integrated and normalized to obtain the patient's injury possibility.

[0037] Further, the detecting of the patient's meniscus injury based on the injury possibility includes:

[0038] If the injury possibility of the patient is greater than a preset injury threshold, then the patient's meniscus is injured; otherwise, the patient's meniscus is not injured.

[0039] Further, obtaining the tissue region in the target slice image includes:

[0040] The Canny edge detection algorithm is used to perform edge detection on the target slice image and extract the tissue area from the target slice image.

[0041] Further, obtaining the meniscus region and the synovial fluid region in the target slice image includes:

[0042] The Otsu threshold segmentation algorithm is used to perform threshold segmentation on the tissue area in the target slice image to obtain the meniscus area and tissue synovial fluid area in the target slice image.

[0043] The present invention has the following beneficial effects:

[0044] The present invention takes into account that the meniscus injury characteristics cannot be clearly and completely displayed in each layer of slice image, which will reduce the accuracy of meniscus injury detection. Therefore, slice images of different layers of the meniscus at the patient's knee joint are obtained. Considering that the meniscus injury characteristics in different slice images are different, the tissue area can be preliminarily extracted from the target slice image, wherein the tissue area includes the meniscus area and the synovial cartilage area around the meniscus. Considering that the noise existing in the tissue area will affect the accuracy of subsequent meniscus injury detection, and the noise has the characteristics of mutation and randomness, which will cause the grayscale changes of pixels in the tissue area to be more obvious and drastic, therefore, the degree of noise interference on the tissue area of ​​the target slice image can be first reflected by the degree of noise interference, so as to improve the accuracy of subsequent injury analysis, and then the tissue area is divided into a meniscus area and a tissue synovial area. Considering that the damaged area in the meniscus area has the characteristics of small grayscale change and high grayscale, the damage image is extracted from the meniscus area based on this feature. Primer points, when there is damage in the meniscus area, inflammatory effusion will appear around the meniscus, causing the surrounding tissue synovial area to invade the meniscus area, and the position distribution of the damaged pixel points in the meniscus area is relatively dense, so the damage factor is used to reflect the possibility of damage in the meniscus area of ​​the target slice image. Considering that when the interlayer spacing between each layer of slice image is improperly set, the meniscus damage characteristics cannot be fully displayed in each layer of slice image, affecting the detection of meniscus damage characteristics, but when the patient's meniscus is damaged, the tissue effusion caused by inflammation around the meniscus will invade the compressed meniscus, resulting in irregular edges of the meniscus area of ​​the slice image, and the shape characteristics of the edges between the meniscus areas of the slice images of adjacent layers are quite different. Therefore, based on this feature, the damage coefficient is used to reflect the possibility of damage in the patient's meniscus area, and then the patient's meniscus damage is accurately detected through the obtained damage possibility, thereby improving the accuracy of meniscus damage detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. 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 creative work.

[0046] Figure 1 A flow chart of an intelligent detection method for meniscus injury based on image processing provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0047] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the specific implementation method, structure, features and effects of an intelligent detection method for meniscus injury based on image processing proposed by the present invention in combination with the accompanying drawings and preferred embodiments. 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 in one or more embodiments may be combined in any suitable form.

[0048] 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.

[0049] The specific scheme of the intelligent detection method for meniscus injury based on image processing provided by the present invention is described in detail below with reference to the accompanying drawings.

[0050] See also Figure 1 , which shows a flow chart of an intelligent detection method for meniscus injury based on image processing provided by an embodiment of the present invention, the method comprising:

[0051] Step S1: Acquire slice images of different layers of the meniscus at the patient's knee joint.

[0052] The meniscus is an important structure in the human knee joint. Testing the meniscus can detect meniscus damage and provide timely treatment. Therefore, the embodiment of the present invention first uses a magnetic resonance device to scan slice images of different layers of the meniscus in the patient's knee joint, wherein the slice images collected by the magnetic resonance device are grayscale images, and the interlayer spacing between each layer of slice images is set to 5 mm. The specific value of the interlayer spacing can also be set by the implementer according to the specific implementation scenario, and is not limited here.

[0053] At this point, multi-slice images of the patient's meniscus have been acquired.

[0054] Step S2: taking a slice image of any layer as a target slice image, performing edge detection on the target slice image to obtain a tissue region in the target slice image; obtaining the degree of noise interference in the tissue region according to the distribution of grayscale values ​​of pixels in the tissue region; performing threshold segmentation on the tissue region to obtain a meniscus region and a tissue synovial region in the target slice image; screening out damaged pixels from all pixels in the meniscus region according to the grayscale values ​​and gradient values ​​of pixels in the meniscus region; obtaining a damage factor of the meniscus region of the target slice image according to the position distribution of damaged pixels in the meniscus region, the difference in the number of pixels between the meniscus region and the tissue synovial region, and the degree of noise interference in the tissue region.

[0055] Since the meniscus injury features in different slice images are different, it is necessary to analyze the slice images of any layer. First, the slice image of any layer is used as the target slice image. In the subsequent steps, the slice image of each layer is analyzed in the same way as the target slice image. Considering that the collected slice image contains not only the tissue area containing the meniscus but also the surrounding bone area, but the boundary between the two is relatively obvious, the embodiment of the present invention first uses the Canny edge detection algorithm to perform edge detection on the target slice image, and extracts the tissue area from the target slice image. The shape of the edge line between the tissue area where the meniscus is located and the bone area is relatively obvious. The edge of the tissue area where the meniscus is located is semi-lunar, wherein the tissue area includes the meniscus area and the synovial cartilage and other areas around the meniscus.

[0056] It should be noted that in other embodiments of the present invention, existing algorithms such as the Sobel algorithm or the Roberts algorithm may also be used for edge detection, which is not limited here.

[0057] Since the noise in the tissue area will affect the accuracy of subsequent meniscus injury detection, and the noise has the characteristics of mutation and randomness, which will cause the grayscale changes of pixels in the tissue area to be more obvious and drastic, the distribution of the grayscale values ​​of the pixels in the tissue area can be analyzed. The degree of noise interference obtained can reflect the degree of noise interference on the tissue area of ​​the target slice image, thereby improving the accuracy of subsequent injury analysis.

[0058] Preferably, in one embodiment of the present invention, the method for acquiring the noise interference degree of a tissue region specifically includes:

[0059] First, multiple maximum pixels are extracted from the tissue area in the target slice image. The grayscale value of the maximum pixel is greater than the grayscale values ​​of all other pixels except the maximum pixel in the preset neighborhood centered on the maximum pixel. That is to say, for a certain pixel in the tissue area, if the grayscale value of the pixel is greater than the grayscale values ​​of all other pixels in the preset neighborhood centered on the pixel, the pixel is considered to be a maximum pixel. The size of the preset neighborhood is set to The specific size of the preset neighborhood can also be set by the implementer according to the specific implementation scenario and is not limited here.

[0060] Then take any two maximum pixel points as a maximum point group, take the average grayscale values ​​of the two maximum pixel points in each maximum point group as the first grayscale eigenvalue of each maximum point group; take the average grayscale values ​​of all pixel points on the line connecting the two maximum pixel points in each maximum point group as the second grayscale eigenvalue of each maximum point group, wherein all pixel points on the line connecting the two maximum pixel points in the maximum point group do not include the two maximum pixel points; take the absolute value of the difference between the first grayscale eigenvalue and the second grayscale eigenvalue as the initial grayscale change degree of each maximum point group, the greater the initial grayscale change degree, the greater the difference between the grayscale values ​​of the two maximum pixel points in the maximum point group and the grayscale values ​​of the pixels between them, and thus the more obvious the grayscale change is.

[0061] Moreover, the smaller the distance between the two maximum pixel points in the maximum point group, the more drastic the grayscale change between them. Therefore, a negative correlation mapping is performed on the distance between the two maximum pixel points in each maximum point group to obtain the first weight parameter of each maximum point group. The product value of the first weight parameter and the initial grayscale change degree is used as the true grayscale change degree of each maximum point group. The greater the true grayscale change degree, the more obvious and drastic the grayscale change between the two maximum pixel points in the maximum point group, and thus the stronger the noise interference in the tissue area. Here, the first weight parameter has the effect of weighting the initial grayscale change degree, thereby improving the accuracy of subsequent calculations.

[0062] In one embodiment of the present invention, the Euclidean distance may be used to measure the distance between two maximum pixel points in each maximum point group.

[0063] Then, the average value of the true grayscale variation of all maximum point groups can be normalized to obtain the noise interference degree of the tissue area.

[0064] As an example, in one embodiment of the present invention, the expression of the noise interference degree of the tissue area can be specifically, for example, as follows:

[0065]

[0066] in, Indicates the degree of noise interference in the tissue area of ​​the target slice image; and Respectively represent The first grayscale eigenvalue and the second grayscale eigenvalue of the maximum point group; Indicates The initial grayscale change degree of each maximum point group; Indicates The Euclidean distance between the two maximum pixels in a maximum point group. Since the two maximum pixels in a maximum point group cannot be the same pixel, ; Indicates The first weight parameter of the maximum point group; Indicates the number of maximum point groups; Indicates The actual grayscale change degree of each maximum point group; Represents the hyperbolic tangent function, which is used for normalization.

[0067] It should be noted that, in other embodiments of the present invention, negative correlation mapping may also be achieved through other basic mathematical operations, which are not limited here.

[0068] Since the tissue area includes two areas, namely the meniscus area and the synovial cartilage area around the meniscus, and the grayscale difference between the two areas is large, the grayscale value of the pixels in the meniscus area is small, while the grayscale value of the pixels in the synovial cartilage area around the meniscus is large, the existing Otsu threshold segmentation algorithm can be first used to perform threshold segmentation on the tissue area in the target slice image to obtain the meniscus area and the tissue synovial area in the target slice image, wherein the tissue synovial area is the synovial cartilage area wrapped around the meniscus, and subsequently the possibility of meniscus damage can be analyzed based on the difference in the number of pixels between the meniscus area and the tissue synovial area. In other embodiments of the present invention, the tissue area can also be threshold segmented using, for example, an adaptive threshold segmentation algorithm, which is not limited here.

[0069] Compared with the mutation of the grayscale values ​​of pixels affected by noise in the meniscus area, the pixels in the damaged area in the meniscus area have the characteristics of small grayscale change and high grayscale. Therefore, the damaged pixels can be screened out from all the pixels in the meniscus area according to the grayscale value and gradient value of the pixels in the meniscus area. Subsequently, based on the distribution of damaged pixels in the meniscus area, the possibility of damage in the meniscus area of ​​the target slice image can be analyzed, wherein the gradient value of the pixels in the meniscus area can be calculated and obtained by the existing Sobel algorithm or Roberts algorithm, which is not limited here.

[0070] Preferably, in one embodiment of the present invention, the method for acquiring damaged pixel points in the meniscus region specifically includes:

[0071] In the meniscus area of ​​the target slice image, the average value of the gradient values ​​of all pixels is taken as the overall gradient value of the meniscus area; the average value of the grayscale values ​​of all pixels is taken as the overall grayscale value of the meniscus area, and the overall gradient value and the overall grayscale value are used as boundaries, and the pixels in the meniscus area with gradient values ​​less than the overall gradient value and grayscale values ​​greater than the overall grayscale value are taken as damaged pixels.

[0072] When there is damage in the meniscus area, inflammatory effusion will occur around the meniscus, causing the surrounding synovial tissue area to invade the meniscus area, and the position distribution of the damaged pixels in the meniscus area is relatively dense. Therefore, the position distribution of the damaged pixels in the meniscus area and the difference in the number of pixels between the meniscus area and the synovial tissue area are analyzed. At the same time, in order to reduce the interference of noise on the damage analysis, combined with the degree of noise interference in the tissue area, the possibility of damage in the meniscus area of ​​the target slice image is reflected by the acquired damage factor. The larger the damage factor, the greater the possibility of damage in the meniscus area of ​​the target slice image. Subsequently, based on the damage factor of the meniscus area of ​​each slice image, an accurate assessment can be made as to whether the patient has meniscus damage.

[0073] Preferably, in one embodiment of the present invention, the method for acquiring the damage factor of the meniscus region of the target slice image specifically includes:

[0074] First, each connected domain composed of damaged pixel points in the meniscus area of ​​the target slice image is taken as the damaged area in the meniscus area; the number of pixels in each damaged area is taken as the numerator, the number of pixels in the minimum circumscribed rectangle of each damaged area is taken as the denominator, and the ratio is taken as the degree of aggregation of each damaged area. The greater the degree of aggregation, the denser the position distribution of damaged pixel points in each damaged area. Then, the average value of the degree of aggregation of all damaged areas in the meniscus area can be taken as the first damage parameter of the meniscus area of ​​the target slice image. The larger the first damage parameter, the more likely it is that damage will occur in the meniscus area of ​​the target slice image.

[0075] The number of pixels in the synovial fluid area is taken as the numerator, the number of pixels in the meniscus area is taken as the denominator, and the ratio is taken as the second injury parameter of the meniscus area of ​​the target slice image. The larger the second injury parameter is, the more serious the invasion of the meniscus by the synovial fluid of the joint tissue around the meniscus is, which further indicates that the meniscus area is more likely to have injury and inflammation symptoms.

[0076] The first injury parameter and the second injury parameter may then be integrated to obtain the integrated injury parameter of the meniscus region of the target slice image.

[0077] In the embodiment of the present invention, the sum or product of the first injury parameter and the second injury parameter may be used as a comprehensive injury parameter of the meniscus region of the target slice image to achieve integration of the two, which is not limited herein.

[0078] As an example, in one embodiment of the present invention, the expression of the comprehensive injury parameter of the meniscus region of the target slice image can be specifically, for example, as follows:

[0079]

[0080]

[0081]

[0082] in, Indicates the comprehensive injury parameter of the meniscus region of the target slice image; A first injury parameter representing a meniscus region of a target slice image; a second injury parameter representing a meniscus region of the target slice image; Indicates the meniscus area The number of pixels in the damaged area; Indicates the meniscus area The number of pixels in the minimum bounding rectangle of the damage area; Indicates the meniscus area The degree of clustering of damaged areas; represents the number of damaged areas in the meniscus region of the target slice image; Represents the number of pixels in the synovial area of ​​the tissue; Represents the number of pixels in the meniscus area;

[0083] At the same time, the smaller the noise interference degree is, the smaller the degree to which the tissue area is affected by the noise is, and the more accurate the analysis of meniscus injury is. Therefore, the inverse of the sum of the noise interference degree of the tissue area of ​​the target slice image and the preset first adjustment parameter can be used as the second weight parameter of the meniscus area of ​​the target slice image, and then the product value of the second weight parameter and the comprehensive injury parameter is normalized to obtain the injury factor of the meniscus area of ​​the target slice image, wherein the second weight parameter is used to weight the comprehensive injury parameter, thereby improving the calculation accuracy of the injury factor and improving the accuracy of subsequent further analysis of the patient's meniscus injury.

[0084] As an example, in one embodiment of the present invention, the expression of the damage factor of the meniscus region of the target slice image may be specifically, for example, as follows:

[0085]

[0086] in, Indicates the damage factor of the meniscus region of the target slice image; Indicates the degree of noise interference in the tissue area of ​​the target slice image; A second weight parameter representing a meniscus region of a target slice image; Indicates the comprehensive injury parameter of the meniscus region of the target slice image; represents the hyperbolic tangent function, which is used for normalization; Indicates the preset first adjustment parameter, which is used to prevent the denominator from being 0. The value range is In one embodiment of the present invention, Set to 0.01, The specific value of can also be set by the implementer according to the specific implementation scenario and is not limited here.

[0087] The damage factor of the meniscus area in each slice image can be obtained by the same method as above.

[0088] Step S3: Obtain the edge smoothness of the meniscus region of the target slice image according to the position direction of the pixel points on the edge of the meniscus region in the target slice image; use two other slice images adjacent to the target slice image as reference slice images, and obtain the patient's injury coefficient according to the difference in edge smoothness of the meniscus region between the target slice image and the reference slice image, as well as the edge smoothness of the meniscus region of all layers of slice images; obtain the patient's injury possibility according to the injury coefficient and the injury factor of the meniscus region of all layers of slice images.

[0089] When the inter-layer spacing between slice images is improperly set, the image features of meniscus injury will be located between the slice images of each layer, thereby missing the acquisition of the image features of meniscus injury when acquiring the image, resulting in the meniscus injury features not being fully displayed in the slice images of each layer, affecting the detection of the meniscus injury features. However, when the patient's meniscus is damaged, the tissue effusion caused by inflammation around the meniscus will invade and compress the meniscus, resulting in irregular edges of the meniscus area in the slice image, and the shape features of the edges of the meniscus area between the slice images of adjacent layers are quite different. Therefore, the edge shape of the meniscus area can be first analyzed based on the position trend features of the pixel points on the edge of the meniscus area in the target slice image, and the smoothness of the edge can be obtained to reflect the smoothness or regularity of the edge of the meniscus area in the target slice image. Subsequently, the difference in edge smoothness of the meniscus area between the slice images of adjacent layers can be analyzed to accurately analyze the possibility of meniscus injury in the patient.

[0090] Preferably, in one embodiment of the present invention, the method for acquiring the edge smoothness of the meniscus region of the target slice image specifically includes:

[0091] An eight-chain code sequence of the edge of the meniscus region in the target slice image is obtained, wherein the eight-chain code is a technical means well known to those skilled in the art and will not be elaborated here. In the eight-chain code sequence, the absolute value of the difference between each element and the next adjacent element is used as the variation of each element. The more consistent the variation of each element is, the smoother the meniscus edge is. Therefore, the variance of the variation of all elements in the eight-chain code sequence can be negatively correlated mapped to obtain the edge smoothness of the meniscus region of the target slice image.

[0092] It should be noted that in the eight-chain code sequence, the last element does not have an adjacent next element, but since the edge of the meniscus region is generally a closed curve, the absolute value of the difference between the last element and the first element can be used as the change of the last element.

[0093] In other embodiments of the present invention, negative correlation mapping may also be performed on the standard deviation of the variation of all elements in the eight-chain code sequence to obtain the edge smoothness of the meniscus region of the target slice image, which is not limited here.

[0094] As an example, in one embodiment of the present invention, the expression of the edge smoothness of the meniscus region of the target slice image can be specifically, for example, as follows:

[0095]

[0096] in, Indicates the edge smoothness of the meniscus area of ​​the target slice image; The variance of the variation of all elements in the eight-chain code sequence representing the meniscus region of the target slice image; Indicates the preset second adjustment parameter, which is used to prevent the denominator from being 0. The value range is In one embodiment of the present invention, Set to 0.01, The specific value of can also be set by the implementer according to the specific implementation scenario and is not limited here.

[0097] It should be noted that, in other embodiments of the present invention, negative correlation mapping may also be achieved through other basic mathematical operations, which are not limited here.

[0098] The edge smoothness of the meniscus region of each slice image can be obtained by the same method as above.

[0099] From the above analysis, it can be seen that when a patient has meniscus injury, the shape characteristics of the edge of the meniscus region between the slice images of adjacent layers are relatively different, that is, the greater the difference in the edge smoothness of the meniscus region between the slice images of adjacent layers, the greater the difference in the edge smoothness of the meniscus region. Therefore, firstly, the two other slice images adjacent to the target slice image are used as reference slice images. It should be noted that for the first and last slice images, there is only one adjacent slice image. At this time, for the first slice image, the second and third slice images can be used as reference slice images for the first slice image, and for the last slice image, the second to last and third to last slice images can be used as reference slice images for the last slice image. Then, according to the difference in the edge smoothness of the meniscus region between the target slice image and the reference slice image, combined with the edge smoothness of the meniscus region of all slice images, the patient's injury coefficient can be obtained. The possibility of the patient's meniscus injury is reflected by the injury coefficient. Subsequently, the possibility of the patient's injury can be accurately calculated and analyzed by combining the injury coefficient and the injury factor of the meniscus region of each slice image.

[0100] Preferably, in one embodiment of the present invention, the method for obtaining the patient's injury coefficient specifically includes:

[0101] First, the average value of the edge smoothness of the meniscus area of ​​the target slice image and the two reference slice images is taken as the reference edge smoothness of the target slice image, and the absolute value of the difference between the edge smoothness of the meniscus area of ​​the target slice image and the reference edge smoothness is taken as the edge smoothness difference parameter of the meniscus area of ​​the target slice image. The larger the edge smoothness difference parameter is, the greater the difference in the meniscus edge shape characteristics between the target slice image and the reference slice image is.

[0102] Then, the average value of the edge smoothness difference parameters of the meniscus area of ​​all layers of the slice images is taken as the patient's overall edge smoothness difference. The larger the overall edge smoothness difference, the greater the difference in the meniscus edge shape characteristics between the slice images of adjacent layers, which means that the patient's meniscus is more likely to be damaged.

[0103] The average value of the edge smoothness of the meniscus area of ​​all layers of slice images is taken as the third weight parameter of the patient. The larger the third weight parameter is, the smoother and more regular the edges of the meniscus area of ​​each layer of slice images are. At this time, the greater the difference in the shape characteristics of the meniscus edge between the slice images of adjacent layers, the more it indicates that the patient's meniscus may be damaged. Therefore, the product value of the third weight parameter and the overall edge smoothness difference can be normalized to obtain the patient's injury coefficient, wherein the third weight parameter is used to weight the overall edge smoothness difference, thereby improving the calculation accuracy of the injury coefficient.

[0104] As an example, in one embodiment of the present invention, the expression of the patient's injury coefficient may be specifically, for example, as follows:

[0105]

[0106] in, represents the patient's injury factor; Indicates Edge smoothness of the meniscus region in the slice image; Indicates The average edge smoothness of the meniscus area of ​​the two reference slice images of the slice image, that is, Reference edge smoothness of layer slice images; Indicates Edge smoothing difference parameters of the meniscus region in the slice image; Indicates the number of slice image layers, that is, the number of slice images; Indicates the overall edge smoothness difference of the patient; represents the average value of the edge smoothness of the meniscus area of ​​the slice images of all layers, i.e., the third weight parameter of the patient; Represents the hyperbolic tangent function, which is used for normalization.

[0107] The larger the patient's injury coefficient is and the larger the damage factor of the meniscus area in each layer of slice image is, the greater the possibility that the patient's meniscus is damaged. Therefore, the patient's injury possibility can be obtained based on the injury coefficient and the damage factor of the meniscus area in all layers of slice images. Subsequently, whether the patient's meniscus is damaged can be accurately detected based on the injury possibility.

[0108] Preferably, in one embodiment of the present invention, the method for obtaining the possibility of injury of a patient specifically includes:

[0109] The average value of the damage factors of the meniscus area in all slice images is taken as the overall damage factor of the patient. The overall damage factor and the damage coefficient are then integrated and normalized to obtain the patient's injury possibility.

[0110] In the embodiment of the present invention, the integration of the overall damage factor and the damage coefficient can be achieved by calculating the sum or product of the two, which is not limited here.

[0111] As an example, in one embodiment of the present invention, the expression of the patient's injury probability may be specifically, for example, as follows:

[0112]

[0113] in, Indicates the patient's potential for injury; represents the patient's injury factor; Indicates Damage factor of the meniscus region in the slice image; Indicates the number of sliced ​​images; The average value of the damage factor of the meniscus area in the slice images of all layers, that is, the overall damage factor of the patient; Represents the hyperbolic tangent function, which is used for normalization.

[0114] Step S4: Based on the possibility of injury, the patient's meniscus injury is detected.

[0115] The greater the possibility of injury, the more likely the patient's meniscus is to be damaged. Therefore, based on the possibility of injury, the patient's meniscus injury can be accurately detected.

[0116] Preferably, in one embodiment of the present invention, the method for detecting meniscus injury of a patient specifically comprises:

[0117] If the possibility of injury of the patient is greater than the preset injury threshold, it means that the patient's meniscus is damaged. Otherwise, it means that the patient's meniscus is not damaged. The preset injury threshold is set to 0.6. The specific value of the preset injury threshold can also be set by the implementer according to the specific implementation scenario, and is not limited here.

[0118] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0119] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

Claims

1. An intelligent detection method for meniscus injury based on image processing, characterized in that: The method comprises: Obtain slice images of different layers of the meniscus at the patient's knee joint; A slice image of any layer is used as a target slice image, and edge detection is performed on the target slice image to obtain a tissue region in the target slice image; the degree of noise interference in the tissue region is obtained according to the distribution of grayscale values ​​of pixels in the tissue region; threshold segmentation is performed on the tissue region to obtain a meniscus region and a tissue synovial region in the target slice image; damaged pixels are screened out from all pixels in the meniscus region according to the grayscale values ​​and gradient values ​​of pixels in the meniscus region; a damage factor of the meniscus region of the target slice image is obtained according to the position distribution of damaged pixels in the meniscus region, the difference in the number of pixels between the meniscus region and the tissue synovial region, and the degree of noise interference in the tissue region, and the damage factor is used to reflect the possibility of damage in the meniscus region of the target slice image; According to the position trend of the pixel points on the edge of the meniscus region in the target slice image, the edge smoothness of the meniscus region of the target slice image is obtained; two other slice images adjacent to the target slice image are used as reference slice images, and according to the difference in the edge smoothness of the meniscus region between the target slice image and the reference slice image, and the edge smoothness of the meniscus region of all slice images, the patient's injury coefficient is obtained; according to the injury coefficient and the injury factor of the meniscus region of all slice images, the patient's injury possibility is obtained, and the injury coefficient is used to reflect the possibility of injury in the patient's meniscus region; Based on the injury possibility, the patient's meniscus injury is detected.

2. The intelligent detection method for meniscus injury based on image processing according to claim 1 is characterized in that: The noise interference degree of the tissue region is obtained by: Extracting a plurality of maximum pixel points from the tissue region in the target slice image, wherein the grayscale value of the maximum pixel point is greater than the grayscale values ​​of all other pixel points except the maximum pixel point in a preset neighborhood centered on the maximum pixel point; Take any two maximum pixel points as a maximum point group, take the average of the grayscale values ​​of the two maximum pixel points in each maximum point group as the first grayscale eigenvalue of each maximum point group; take the average of the grayscale values ​​of all the pixel points on the line connecting the two maximum pixel points in each maximum point group as the second grayscale eigenvalue of each maximum point group; take the absolute value of the difference between the first grayscale eigenvalue and the second grayscale eigenvalue as the initial grayscale change degree of each maximum point group; Performing negative correlation mapping on the distance between two maximum pixel points in each maximum point group to obtain a first weight parameter of each maximum point group; taking the product value of the first weight parameter and the initial grayscale change degree as the actual grayscale change degree of each maximum point group; The average values ​​of the true grayscale variation degrees of all maximum point groups are normalized to obtain the noise interference degree of the tissue area.

3. The intelligent detection method for meniscus injury based on image processing according to claim 1 is characterized in that: The step of selecting damaged pixels from all pixels in the meniscus region includes: In the meniscus region of the target slice image, the average value of the gradient values ​​of all the pixels is taken as the overall gradient value of the meniscus region; the average value of the grayscale values ​​of all the pixels is taken as the overall grayscale value of the meniscus region; Pixels in the meniscus region whose gradient values ​​are smaller than the overall gradient value and whose grayscale values ​​are larger than the overall grayscale value are regarded as damaged pixels.

4. The intelligent detection method for meniscus injury based on image processing according to claim 1 is characterized in that: The damage factors of the meniscus region of the target slice image are as follows: The connected domains formed by the damaged pixels in the meniscus region of the target slice image are used as the damaged regions in the meniscus region; the number of pixels in each damaged region is used as the numerator, the number of pixels in the minimum circumscribed rectangle of each damaged region is used as the denominator, and the ratio is used as the aggregation degree of each damaged region; the average value of the aggregation degrees of all damaged regions in the meniscus region is used as the first damage parameter of the meniscus region of the target slice image; The number of pixels in the tissue synovial region is used as a numerator, the number of pixels in the meniscus region is used as a denominator, and the ratio is used as a second injury parameter of the meniscus region of the target slice image; integrating the first injury parameter and the second injury parameter to obtain a comprehensive injury parameter of the meniscus region of the target slice image; The inverse of the sum of the noise interference degree of the tissue area of ​​the target slice image and the preset first adjustment parameter is used as the second weight parameter of the meniscus area of ​​the target slice image; The product value of the second weight parameter and the comprehensive damage parameter is normalized to obtain the damage factor of the meniscus area of ​​the target slice image.

5. The intelligent detection method for meniscus injury based on image processing according to claim 1 is characterized in that: The step of obtaining the edge smoothness of the meniscus region of the target slice image comprises: An eight-chain code sequence of the edge of the meniscus region in the target slice image is obtained, and in the eight-chain code sequence, the absolute value of the difference between each element and the next adjacent element is used as the variation of each element; The variance of the variation of all elements in the eight-chain code sequence is negatively correlated to obtain the edge smoothness of the meniscus region of the target slice image.

6. The intelligent detection method for meniscus injury based on image processing according to claim 1 is characterized in that: The obtaining of the patient's injury coefficient comprises: Taking the average of the edge smoothness of the meniscus region of the two reference slice images of the target slice image as the reference edge smoothness of the target slice image; The absolute value of the difference between the edge smoothness of the meniscus region of the target slice image and the reference edge smoothness is used as an edge smoothness difference parameter of the meniscus region of the target slice image; The average value of the edge smoothness difference parameters of the meniscus region of the slice images of all layers is taken as the overall edge smoothness difference of the patient; an average value of the edge smoothness of the meniscus region of the slice images of all layers as the third weight parameter of the patient; The product value of the third weight parameter and the overall edge smoothness difference is normalized to obtain the patient's injury coefficient.

7. The intelligent detection method for meniscus injury based on image processing according to claim 1 is characterized in that: The patient's injury probability is obtained by: The average value of the damage factors of the meniscus region of the slice images of all layers is taken as the overall damage factor of the patient; The overall injury factor and the injury coefficient are integrated and normalized to obtain the patient's injury possibility.

8. The intelligent detection method for meniscus injury based on image processing according to claim 1 is characterized in that: The detecting of the patient's meniscus injury based on the injury possibility includes: If the injury possibility of the patient is greater than a preset injury threshold, then the patient's meniscus is injured; otherwise, the patient's meniscus is not injured.

9. The intelligent detection method for meniscus injury based on image processing according to claim 1 is characterized in that: The obtaining of the tissue region in the target slice image comprises: The Canny edge detection algorithm is used to perform edge detection on the target slice image and extract the tissue area from the target slice image.

10. The intelligent detection method for meniscus injury based on image processing according to claim 1, characterized in that: The obtaining of the meniscus region and the synovial fluid region in the target slice image comprises: The Otsu threshold segmentation algorithm is used to perform threshold segmentation on the tissue area in the target slice image to obtain the meniscus area and tissue synovial fluid area in the target slice image.

Images