Automatic identification method for arthropathy in motion medical image

Abstract

The invention relates to the technical field of image processing, in particular to a method for automatically identifying arthropathy in a sports medical image, which comprises the following steps of: acquiring the sports medical image and acquiring an edge in the sports medical image; according to the distribution positions of pixel points on the edges in the motion medical image, the possibility that end points in different disconnected edges are located on the same skeleton edge is obtained, and the disconnected edges are connected according to the possibility to obtain a complete skeleton edge; obtaining a plurality of joint gaps according to the distribution of pixel points in the complete skeleton edge; and according to the width of each position in the same joint gap and the width difference of different joint gaps, obtaining the lesion characteristics of the joint gaps and lesion joints. According to the method, the extension trend of the skeleton is analyzed, the incomplete skeleton edges are connected into the complete skeleton edge, and the joint position in the motion medical image is accurately obtained by further analyzing the characteristics of the joint edges, so that the diseased joint is accurately detected and recognized.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and particularly relates to an automatic recognition method for joint lesions in sports medicine images. Background Art

[0002] In the field of sports medicine, the early and accurate diagnosis of joint lesions is crucial for the prevention, treatment, and rehabilitation of sports injuries. Sports medicine images refer to medical images used for diagnosing, evaluating, and monitoring diseases and lesions related to sports injuries, joint diseases, and the musculoskeletal system. However, due to the influence of soft tissues wrapping the bones during the acquisition of sports medicine images, the obtained bone edges are incomplete, which in turn affects the accuracy of detecting and identifying joint lesions. That is, traditional sports medicine images directly collected cannot accurately identify joint lesions. Summary of the Invention

[0003] The present invention provides an automatic recognition method for joint lesions in sports medicine images to solve the existing problem that joint lesions cannot be accurately identified directly from the collected sports medicine images.

[0004] The automatic recognition method for joint lesions in sports medicine images of the present invention adopts the following technical solutions: It includes the following steps: Collect sports medicine images and obtain the edges in the sports medicine images; According to the distribution positions of pixel points on the edges in the sports medicine images, obtain the discontinuous edges and the endpoints in the discontinuous edges; according to the pixel point distribution around the endpoints in the discontinuous edges, obtain the extension trends of the endpoints in the discontinuous edges; according to the differences in the extension trends of the endpoints in different discontinuous edges, obtain the possibility that the endpoints in different discontinuous edges are located on the same bone edge; according to the possibility that the endpoints in different discontinuous edges are located on the same bone edge, connect the discontinuous edges to obtain a complete bone edge; According to the distribution of pixel points in the complete bone edge, obtain the possibility that the pixel points in the complete bone edge are located at the joint; according to the possibility that the pixel points in the complete bone edge are located at the joint, obtain several joint spaces; According to the widths at various positions in the same joint space and the differences in widths among different joint spaces, obtain the lesion characteristics of the joint space; according to the lesion characteristics of the joint space, obtain the diseased joints.

[0005] Preferably, the specific method for obtaining the discontinuous edges and the endpoints in the discontinuous edges according to the distribution positions of pixel points on the edges in the sports medicine images is as follows: All unclosed edges in the sports medicine images are recorded as discontinuous edges; for any discontinuous edge, the two edge pixel points with the farthest distance in the discontinuous edge are used as the endpoints of the discontinuous edge.

[0006] Preferably, the method for obtaining the extension trend of the endpoint in the disconnection edge according to the distribution of the pixel points around the endpoint in the disconnection edge includes the following specific steps: For any endpoint in the disconnection edge, a preset number of neighborhood pixel points is set. ; The edge pixel points closest to the endpoint in the disconnection edge are used as the neighborhood pixel points of the endpoint. According to the position coordinates of the neighborhood pixel points of the endpoint in the sports medicine image, the extension trend of the endpoint in the disconnection edge is obtained. The specific calculation formula is: ; In the formula, represents the extension trend of the endpoint in the disconnection edge; represents the number of neighborhood pixel points of the endpoint; represents the th ordinate of the th neighborhood pixel point of the endpoint in the sports medicine image; represents the ordinate of the th neighborhood pixel point of the endpoint in the sports medicine image; represents the abscissa of the th neighborhood pixel point of the endpoint in the sports medicine image; represents the abscissa of the

[0007] th neighborhood pixel point of the endpoint in the sports medicine image. Preferably, the method for obtaining the possibility that the endpoints in different disconnection edges are located on the same bone edge according to the difference in the extension trends of the endpoints in different disconnection edges includes the following specific steps: For any endpoint in any disconnection edge, a local window is constructed with the endpoint as the center. The is the preset side length of the local window; The endpoints in the local window except the endpoint are denoted as target points, and several target points are obtained; For any target point, according to the extension trends of the endpoint and the target point, the possibility that the endpoint and the target point are located on the same bone edge is obtained. The specific calculation formula is: ; In the formula, represents the possibility that the endpoint and the target point are located on the same bone edge; represents the extension trend of the endpoint in the disconnection edge; represents the extension trend of the target point.

[0008] Preferably, the method for connecting the disconnected edges to obtain a complete bone edge according to the possibility that the end points in different disconnected edges are located on the same bone edge includes the following specific steps: Preset a possibility threshold ; for any end point in any disconnected edge, obtain the possibility that the end point and all target points are located on the same bone edge, and mark the target points with a possibility greater than as connection points, and connect the connection point closest to the end point to obtain a complete bone edge.

[0009] Preferably, the method for obtaining the possibility that the pixel points in the complete bone edge are located at the joint according to the distribution of the pixel points in the complete bone edge includes the following specific steps: For any pixel point in the complete bone edge, preset a number of adjacent pixel points , and take the pixel points closest to the pixel point in the complete bone edge as the adjacent pixel points of the pixel point. According to the position coordinates of the adjacent pixel points of the pixel point, obtain the possibility that the pixel point is located on the arc edge. The specific calculation formula is: ; In the formula, represents the possibility that the pixel point is located on the arc edge; represents the number of adjacent pixel points of the pixel point; represents the ordinate of the th adjacent pixel point of the pixel point in the sports medicine image; represents the ordinate of the th adjacent pixel point of the pixel point in the sports medicine image; represents the abscissa of the th adjacent pixel point of the pixel point in the sports medicine image; represents the abscissa of the th adjacent pixel point of the pixel point in the sports medicine image; represents the absolute value function; represents the linear normalization function; According to the possibility that the pixel point is located on the arc edge, and in combination with the position coordinates of the first and last adjacent pixel points of the pixel point, obtain the possibility that the pixel point is located at the joint.

[0010] Preferably, the specific calculation formula for obtaining the possibility that the pixel point is located at the joint is: ; In the formula, Indicates the possibility that the pixel point is located at a joint; Indicates the possibility that the pixel point is located at the edge of an arc; Indicates the ordinate of the first neighboring pixel point of the pixel point in the sports medicine image; Indicates the ordinate of the last neighboring pixel point of the pixel point in the sports medicine image; Indicates the abscissa of the first neighboring pixel point of the pixel point in the sports medicine image; Indicates the abscissa of the last neighboring pixel point of the pixel point in the sports medicine image; Indicates the absolute value function; Indicates the linear normalization function.

[0011] Preferably, the specific method for obtaining several joint spaces according to the possibility that the pixel points in the complete bone edge are located at joints includes: Preset a possibility threshold for being located at a joint ; Mark the pixel points whose possibility of being located at a joint is greater than or equal to as joint pixel points to obtain all joint pixel points; Obtain the number of joints in the sports medicine image and denote it as , use the Euclidean distance between all joint pixel points as the metric distance, perform k-means clustering on all joint pixel points, and set the value of K in the k-means distance algorithm to , to obtain cluster classes in the sports medicine image, denoted as mother cluster classes; For any mother cluster class, use the Euclidean distance between all joint pixel points in the mother cluster class as the metric distance, perform k-means clustering on all joint pixel points in the mother cluster class to obtain 4 sub-cluster classes in the mother cluster class; For any sub-cluster class in the mother cluster class, take the sub-cluster class in the mother cluster class that is closest to the sub-cluster class as the corresponding sub-cluster class of the sub-cluster class; Obtain the corresponding sub-cluster classes of all sub-cluster classes in the mother cluster class, and connect the cluster centers of each sub-cluster class in the mother cluster class and its corresponding sub-cluster class to obtain two joint edge lines, and take the gap between the two joint lines as the joint space.

[0012] Preferably, the specific calculation formula for obtaining the lesion characteristics of the joint space according to the width at each position in the same joint space and the difference in width between different joint spaces is: ; In the formula, represents the lesion characteristics of the joint space, Represents the number of joint spaces in the sports medicine image; Represents the average width at all positions in the nth joint space in the sports medicine image; Represents the average width at all positions in the joint space; Represents the standard deviation of the width at all positions in the joint space;

[0013] Preferably, the method for obtaining the diseased joint according to the lesion characteristics of the joint space specifically includes: Preset a lesion threshold , and for any joint space, if the lesion characteristics of the joint space are greater than or equal to , the joint corresponding to the joint space is diseased.

[0014] The beneficial effect of the technical solution of the present invention is that: by analyzing the distribution positions of the pixel points on the edge in the sports medicine image, the present application obtains the possibility that the endpoints in different discontinuous edges are located on the same bone edge, and connects the discontinuous edges based on this to obtain a complete bone edge; since the soft tissue wrapping the bone will produce shadows when collecting the sports medicine image, which leads to the edges in the sports medicine image being incomplete and intermittent, and in order to accurately detect and identify the lesion characteristics of the patient's joint, a complete bone edge needs to be obtained. Also, because the extension trend of the bone is similar within a local range, the incomplete bone edges can be connected according to the similarity of the extension trends of the incomplete bone edges to obtain a complete bone edge; To detect and identify the joint part with a complete bone edge required for the lesion characteristics of the patient's joint, several joint spaces are obtained by analyzing the distribution of the pixel points in the complete bone edge; also, in the case of normal without lesions, the joint space widths of the corresponding joint spaces of different joints should be similar, and the widths at different positions of the same joint space should be consistent; therefore, according to the widths at each position in the same joint space and the differences in widths between different joint spaces, the lesion characteristics of the joint space and the diseased joint are obtained. The present application connects the incomplete bone edges into a complete bone edge by analyzing the extension trend of the bone, and further accurately obtains the joint positions in the sports medicine image by analyzing the characteristics of the edges of the joints, so as to accurately detect and identify the diseased joints. Description of the Drawings

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0016] Figure 1 It is a flowchart of the steps of an automatic joint lesion recognition method in sports medicine images of the present invention; Figure 2 It is an example diagram of a sports medicine edge image. Detailed implementation manners

[0017] To further elaborate on the technical means and effects adopted by the present invention to achieve the intended invention purpose, the following, in combination with the drawings and preferred embodiments, details the specific implementation manners, structures, features, and effects of an automatic joint lesion recognition method in sports medicine images proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present invention belongs.

[0019] The following specifically describes the specific solution of an automatic joint lesion recognition method in sports medicine images provided by the present invention in combination with the drawings.

[0020] Please refer to Figure 1 , which shows a flowchart of the steps of an automatic joint lesion recognition method in sports medicine images provided by an embodiment of the present invention. The method includes the following steps: Step S001: Collect sports medicine images and obtain the edges in the sports medicine images.

[0021] It should be noted that sports medicine images refer to medical images used for diagnosing, evaluating, and monitoring diseases and lesions related to sports injuries, joint diseases, and the musculoskeletal system. And as an automatic joint lesion recognition method in sports medicine images in this embodiment, specifically, through the bones and joints in the patient's sports medicine images, the lesion characteristics of the patient's joints are detected and recognized. Therefore, first, the bone edges in the patient's sports medicine images need to be collected. In this embodiment, the sports medicine images of the palm are used as an example for description.

[0022] Specifically, an X-ray image of the patient's palm is collected by an X-ray imaging device, and the X-ray image of the patient's palm is used as a sports medicine image. An edge detection algorithm is used to obtain the edges in the sports medicine image, such as Figure 2 shown. Figure 2 This is an example diagram of a sports medicine edge image. Since the edge detection algorithm is a well-known existing technology, it will not be elaborated in this embodiment, and this embodiment does not make a rigid requirement for the edge detection algorithm. In this embodiment, the sobel edge detection algorithm is used as an example for description.

[0023] So far, the edges in the sports medicine image are obtained.

[0024] Step S002: According to the distribution positions of the pixel points on the edges in the sports medicine image, obtain the discontinuous edges and the endpoints in the discontinuous edges; according to the pixel point distribution around the endpoints in the discontinuous edges, obtain the extension trends of the endpoints in the discontinuous edges; according to the differences in the extension trends of the endpoints in different discontinuous edges, obtain the possibility that the endpoints in different discontinuous edges are located on the same bone edge; according to the possibility that the endpoints in different discontinuous edges are located on the same bone edge, connect the discontinuous edges to obtain a complete bone edge.

[0025] It should be noted that the edges in the sports medicine image collected in step S001 are the bone edges in the sports medicine image of the palm; when collecting the sports medicine image, due to the shadows generated by the soft tissues wrapping the bones, the edges in the sports medicine image are not complete and are intermittent. In order to accurately detect and identify the lesion characteristics of the patient's joints, a complete bone edge needs to be obtained. Since the extension trends of the bones are similar within a local range, the incomplete bone edges can be connected according to the similarity degree of the extension trends of the incomplete bone edges to obtain a complete bone edge for accurately detecting and identifying the lesion characteristics of the patient's joints.

[0026] Specifically, all the unclosed edges in the sports medicine image are recorded as discontinuous edges; for any discontinuous edge, the two edge pixel points with the farthest distance in the discontinuous edge are used as the endpoints of the discontinuous edge; For any endpoint in the discontinuous edge, a preset number of neighborhood pixel points , the specific value of which can be set according to the actual situation, and this embodiment does not make a rigid requirement. In this embodiment, is used as an example for description; the edge pixel points (including the endpoint) closest to the endpoint in the discontinuous edge are used as the neighborhood pixel points of the endpoint. According to the position coordinates of the neighborhood pixel points of the endpoint in the sports medicine image, the extension trend of the endpoint in the discontinuous edge is obtained, and its specific calculation formula is: ； In the formula, represents the extension trend of the endpoint in the discontinuous edge; represents the number of neighborhood pixel points of the endpoint; represents the th ordinate of the neighborhood pixel point of the endpoint in the sports medicine image; represents the th ordinate of the neighborhood pixel point of the endpoint in the sports medicine image; represents the th abscissa of the neighborhood pixel point of the endpoint in the sports medicine image; represents the th abscissa of the neighborhood pixel point of the endpoint in the sports medicine image.

[0027] Furthermore, for any endpoint in any discontinuous edge, a local window is constructed with the endpoint as the center, and the is the preset side length of the local window. The specific value can be set according to the actual situation and is not strictly required in this embodiment. In this embodiment, is taken as an example for description; the endpoints in the local window except the endpoint are denoted as target points, and several target points are obtained; For any target point, according to the extension trend between the endpoint and the target point, the possibility that the endpoint and the target point are located on the same bone edge is obtained. The specific calculation formula is: ; In the formula, represents the possibility that the endpoint and the target point are located on the same bone edge; represents the extension trend of the endpoint in the discontinuous edge; represents the extension trend of the target point.

[0028] It should be noted that the extension trend represents the extension trend of the incomplete bone edge. The smaller the value of , the more similar the extension trends of the incomplete edges corresponding to the endpoint and the target point are. When the distance between two incomplete edges is close and their extension trends are similar, the two incomplete edges are more likely to be located on the same bone edge. Therefore,

[0029] Specifically, a possibility threshold is preset, and the The specific value can be set according to the actual situation, and there is no strict requirement in this embodiment. In this embodiment, is taken as an example for description; for any endpoint in any disconnected edge, the target point with a possibility greater than of being located on the same bone edge as the endpoint is recorded as the connection point, and the connection point closest to the endpoint is connected to obtain the complete bone edge.

[0030] Step S003: According to the distribution of pixel points in the complete bone edge, obtain the possibility that the pixel points in the complete bone edge are located at the joints; according to the possibility that the pixel points in the complete bone edge are located at the joints, obtain several joint gaps.

[0031] It should be noted that an automatic recognition method for joint lesions in sports medicine images specifically detects and recognizes the lesion characteristics of the patient's joints through the bone joints in the patient's sports medicine images. Therefore, after obtaining the complete bone edge in step S002, it is also necessary to obtain the joint part of the complete bone edge; and since the bone edge at the joint is arc-shaped, the possibility that each pixel point in the complete edge is located at the arc edge can be obtained according to each pixel point in the complete edge and its surrounding pixel points, and then the possibility that the pixel point is located at the joint can be obtained to obtain the joint position in the complete bone.

[0032] Specifically, for any pixel point in the complete bone edge, a preset number of adjacent pixel points is set. The specific value of can be set according to the actual situation, and there is no strict requirement in this embodiment. In this embodiment, is taken as an example for description. The pixel points closest to the pixel point in the complete bone edge (including the pixel point) are used as the adjacent pixel points of the pixel point. According to the position coordinates of the adjacent pixel points of the pixel point, the possibility that the pixel point is located at the arc edge is obtained. The specific calculation formula is: ; In the formula, represents the possibility that the pixel point is located at the arc edge; represents the number of adjacent pixel points of the pixel point; represents the ordinate of the th adjacent pixel point of the pixel point in the sports medicine image; represents the ordinate of the th adjacent pixel point of the pixel point in the sports medicine image; represents the ordinate of the th adjacent pixel point of the pixel point in the sports medicine image; represents the The abscissa of a neighboring pixel point in a sports medicine image; Represents the absolute value function; Represents the linear normalization function, and its specific normalization range is for all pixel points in the complete bone edge 。

[0033] It should be noted that represents the extension trend of adjacent neighboring pixel points, and when the pixel point is located on the arc edge, the sum of the extension trends of all adjacent neighboring pixel points of the pixel point approaches 0. Therefore the smaller the value of

[0034] the more likely the pixel point is to be located on the arc edge; however, in the complete bone edge, in addition to the edge under the joint position, there is also an edge under the fingertip position with an arc feature. However, since the arc feature of the edge under the fingertip position is horizontally distributed, while the arc feature of the edge under the joint position is vertically distributed; therefore, the possibility of the pixel point being located at the joint can be further obtained by combining the distribution direction of its arc. ; In the formula, represents the possibility that the pixel point is located at the joint; represents the possibility that the pixel point is located on the arc edge; represents the ordinate of the first neighboring pixel point of the pixel point in the sports medicine image; represents the ordinate of the last neighboring pixel point of the pixel point in the sports medicine image; represents the abscissa of the first neighboring pixel point of the pixel point in the sports medicine image; represents the abscissa of the last neighboring pixel point of the pixel point in the sports medicine image; represents the absolute value function; represents the linear normalization function, and its specific normalization range is for all pixel points in the complete bone edge 。

[0035] It should be noted that when the arc is horizontally distributed, the horizontal distance between the two end pixel points of the arc is large and the vertical distance is small, and when the arc is vertically distributed, the horizontal distance between the two end pixel points of the arc is small and the vertical distance is large. Therefore the larger the value of The smaller the value, the more vertical the arc distribution becomes. Further, considering that the pixel points are more likely to be located on the edge of the arc, the possibility of the pixel points being located on the vertical arc can be obtained. And when the possibility of the pixel points being located on the vertical arc is greater, the possibility of the pixel points being located at the joint is higher. Based on this, the joint positions in the complete skeleton are obtained.

[0036] Specifically, a possibility threshold located at the joint is preset. , the specific value can be set according to the actual situation, and there is no strict requirement in this embodiment. In this embodiment, is taken as an example for description; the pixel points with a possibility greater than or equal to at the joint are recorded as joint pixel points, and all joint pixel points are obtained; the number of joints in the sports medicine image is recorded as . Taking the Euclidean distance between all joint pixel points as the metric distance, all joint pixel points are subjected to k-means clustering, and the K value in the k-means distance algorithm is set to , and cluster classes in the sports medicine image are obtained, denoted as parent cluster classes.

[0037] It should be noted that the joint pixel points represent the pixel points located at the edge of the joint, and each parent cluster class corresponds to each joint position. Since the joint pixel points are located around the joint, the number of joints in the collected sports medicine image can be obtained in advance, and combined with the distance between the joint pixel points, all joint pixel points are clustered to obtain the joint positions.

[0038] It should be further noted that as an automatic recognition method for joint lesions in sports medicine images, this embodiment specifically detects and recognizes joint lesion characteristics by analyzing the width of the joint space. Therefore, after obtaining the joint positions, the joint space needs to be further obtained to detect and recognize the diseased joints.

[0039] Furthermore, for any parent cluster class, taking the Euclidean distance between all joint pixel points in the parent cluster class as the metric distance, all joint pixel points in the parent cluster class are subjected to k-means clustering (the K value in the k-means distance algorithm is set to 4) to obtain 4 sub-cluster classes in the parent cluster class; For any sub-cluster class in the parent cluster class, the sub-cluster class in the parent cluster class that is closest to the sub-cluster class is used as the corresponding sub-cluster class of the sub-cluster class; The corresponding sub-cluster classes of all sub-cluster classes in the parent cluster class are obtained, and the cluster centers of each sub-cluster class in the parent cluster class and its corresponding sub-cluster class are connected to obtain two joint edge lines, and the gap between the two joint lines is used as the joint space.

[0040] Thus, the joint space is obtained.

[0041] Step S004: Obtain the lesion characteristics of the joint space according to the widths at different positions in the same joint space and the differences in widths between different joint spaces; obtain the diseased joint according to the lesion characteristics of the joint space.

[0042] It should be noted that under normal and non-lesioned conditions, the joint space widths of the corresponding joint spaces of different joints should be similar, and the widths at different positions of the same joint space should be consistent. When lesions such as rheumatoid arthritis occur in the joint, various problems such as narrowing of the finger joint space and joint deformity will be caused due to joint cartilage damage, inflammatory reactions, etc., resulting in a decrease in the similarity of the corresponding positions of the joint. Therefore, after obtaining all joint spaces through step S003, according to the similarity of the joint space widths at the corresponding positions of different joints and the consistency of the widths at different positions of the same joint space, the possibility of lesions at each position under each joint can be obtained, so as to detect and identify joint lesions.

[0043] Specifically, for any joint space, according to the widths at each position in the joint space and the widths at each position in all joint spaces, obtain the lesion characteristics of the joint space, and its specific calculation formula is: ; In the formula, represents the lesion characteristics of the joint space, represents the number of joint spaces in the sports medicine image; represents the average width of all positions in the th joint space in the sports medicine image; represents the average width of all positions in the joint space; represents the standard deviation of the widths of all positions in the joint space; represents the sigmoid function, which is used for normalization processing in this embodiment.

[0044] It should be noted that; under normal and non-lesioned conditions, the joint space widths of the corresponding joint spaces of different joints should be similar, and the widths at different positions of the same joint space are similar; while the larger the value of , the greater the difference in width between the joint space and other joint spaces, indicating that the widths at each position in the joint space are more inconsistent. Therefore,

[0045] Specifically, preset a lesion threshold , the The specific value can be set according to the actual situation, and there is no strict requirement in this embodiment. In this embodiment, is taken as an example for description. For any joint space, if the lesion characteristics of the joint space are greater than or equal to , the joint corresponding to the joint space is diseased. If the lesion characteristics of the joint space are less than , the joint corresponding to the joint space is normal.

[0046] So far, this embodiment is completed.

[0047] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the principles of the present invention shall be included in the protection scope of the present invention.

Claims

1. An automatic recognition method for joint lesions in sports medicine images, characterized in that, The method includes the following steps: Collect sports medicine images and obtain the edges in the sports medicine images; According to the distribution positions of the pixel points on the edges in the sports medicine images, obtain the discontinuous edges and the endpoints in the discontinuous edges; according to the pixel point distribution around the endpoints in the discontinuous edges, obtain the extension trends of the endpoints in the discontinuous edges; according to the differences in the extension trends of the endpoints in different discontinuous edges, obtain the possibility that the endpoints in different discontinuous edges are located on the same bone edge; according to the possibility that the endpoints in different discontinuous edges are located on the same bone edge, connect the discontinuous edges to obtain a complete bone edge; According to the pixel point distribution in the complete bone edge, obtain the possibility that the pixel points in the complete bone edge are located at joints; according to the possibility that the pixel points in the complete bone edge are located at joints, obtain several joint spaces; According to the widths at different positions in the same joint space and the differences in widths between different joint spaces, obtain the lesion characteristics of the joint space; according to the lesion characteristics of the joint space, obtain the diseased joints.

2. The automatic joint lesion recognition method in sports medicine images according to claim 1, characterized in that, The specific method for obtaining the discontinuous edges and the endpoints in the discontinuous edges according to the distribution positions of the pixel points on the edges in the sports medicine images is as follows: Record all the unclosed edges in the sports medicine image as discontinuous edges; for any discontinuous edge, take the two edge pixel points with the farthest distance in the discontinuous edge as the endpoints of the discontinuous edge.

3. The automatic recognition method for joint lesions in sports medicine images according to claim 1, characterized in that The specific method for obtaining the extension trend of the endpoints in the discontinuous edges according to the pixel point distribution around the endpoints in the discontinuous edges is as follows: For any endpoint in the discontinuous edge, a number of neighborhood pixel points is preset. ; Among the edge pixels closest to the endpoint in the discontinuous edge, are used as the neighborhood pixel points of the endpoint. According to the position coordinates of the neighborhood pixel points of the endpoint in the sports medicine image, the extension trend of the endpoint in the discontinuous edge is obtained. The specific calculation formula is as follows: ； In the formula, represents the extension trend of the endpoint in the disconnection edge; represents the number of neighborhood pixel points of the endpoint; represents the ordinate of the th neighborhood pixel point of the endpoint in the sports medicine image; ordinate of the th neighborhood pixel point of the endpoint in the sports medicine image; abscissa of the th neighborhood pixel point of the endpoint in the sports medicine image; abscissa of the th neighborhood pixel point of the endpoint in the sports medicine image.

4. The automatic recognition method for joint lesions in sports medicine images according to claim 1, characterized in that The specific method for obtaining the possibility that the endpoints in different discontinuous edges are located on the same bone edge according to the differences in the extension trends of the endpoints in different discontinuous edges is as follows: For any endpoint in any broken edge, construct a local window centered at the endpoint, where the is the preset side length of the local window; Denote the endpoints in the local window except the endpoint as target points to obtain a number of target points; For any target point, according to the extension trend between the endpoint and the target point, obtain the possibility that the endpoint and the target point are located on the same bone edge, and its specific calculation formula is: ； In the formula, represents the possibility that the endpoint and the target point are located on the same bone edge; represents the extension trend of the endpoint in the discontinuous edge; represents the extension trend of the target point.

5. The automatic recognition method for joint lesions in sports medicine images according to claim 4, wherein The specific method for connecting the discontinuous edges to obtain a complete bone edge according to the possibility that the endpoints in different discontinuous edges are located on the same bone edge is as follows: Preset a possibility threshold ; For any endpoint in any disconnected edge, obtain the possibility that the endpoint and all target points are on the same bone edge, and mark the target points with a possibility greater than as connection points, and connect the connection point closest to the endpoint to obtain a complete bone edge.

6. The automatic joint lesion recognition method in sports medicine images according to claim 1, wherein, The specific method for obtaining the possibility that the pixel points in the complete bone edge are located at joints according to the pixel point distribution in the complete bone edge is as follows: For any pixel point on the edge of the complete bone, a preset number of adjacent pixel points is set , and the pixel points on the edge of the complete bone that are closest to the pixel point are used as the adjacent pixel points of the pixel point. According to the position coordinates of the adjacent pixel points of the pixel point, the possibility that the pixel point is located on the circular arc edge is obtained. The specific calculation formula is as follows: ； In the formula, represents the possibility that the pixel point is located on the arc edge; represents the number of adjacent pixel points of the pixel point; represents the ordinate of the th adjacent pixel point of the pixel point in the sports medicine image; ordinate of the th adjacent pixel point of the pixel point in the sports medicine image; abscissa of the th adjacent pixel point of the pixel point in the sports medicine image; abscissa of the represents the absolute value function; represents the linear normalization function; According to the possibility that the pixel point is located on an arc edge, combine the position coordinates of the first adjacent pixel point and the last adjacent pixel point of the pixel point to obtain the possibility that the pixel point is located at a joint.

7. The automatic joint lesion recognition method in sports medicine images according to claim 6, characterized in that The specific calculation formula for obtaining the possibility that the pixel point is located at a joint is: ； In the formula, represents the possibility that the pixel point is located at the joint; represents the possibility that the pixel point is located at the arc edge; represents the ordinate of the first adjacent pixel point of the pixel point in the sports medicine image; represents the ordinate of the last adjacent pixel point of the pixel point in the sports medicine image; represents the abscissa of the first adjacent pixel point of the pixel point in the sports medicine image; represents the abscissa of the last adjacent pixel point of the pixel point in the sports medicine image; represents the absolute value function; represents the linear normalization function.

8. The automatic joint lesion recognition method in sports medicine images according to claim 1, wherein The specific method for obtaining several joint spaces according to the possibility that the pixel points in the complete bone edge are located at joints is as follows: Preset a possibility threshold at the joint ; Denote the pixel points with a possibility greater than or equal to at the joint as joint pixel points to obtain all joint pixel points; Obtain the number of joints in the sports medicine image and denote it as , use the Euclidean distance between all joint pixel points as the metric distance, perform k-means clustering on all joint pixel points, and set the K value in the k-means distance algorithm to , to obtain cluster classes in the sports medicine image, denoted as parent cluster classes; For any mother cluster class, use the Euclidean distance between all joint pixel points in the mother cluster class as the metric distance, and perform k-means clustering on all joint pixel points in the mother cluster class to obtain 4 sub-cluster classes in the mother cluster class; For any sub-cluster class in the mother cluster class, take the sub-cluster class in the mother cluster class that is closest to the sub-cluster class as the corresponding sub-cluster class of the sub-cluster class; Obtain the corresponding sub-cluster classes of all sub-cluster classes in the mother cluster class, and connect the cluster class centers of each sub-cluster class in the mother cluster class and its corresponding sub-cluster class to obtain two joint edge lines, and take the gap between the two joint lines as the joint gap.

9. The automatic recognition method for joint lesions in sports medicine images according to claim 1, wherein The specific calculation formula for obtaining the lesion characteristics of the joint gap according to the widths at each position in the same joint gap and the differences in widths among different joint gaps is: ； In the formula, represents the pathological features of the joint space, represents the number of joint spaces in the sports medicine image; represents the average width at all positions in the th joint space in the sports medicine image; represents the average width at all positions in the joint space; represents the sigmoid function.

10. The automatic recognition method for joint lesions in sports medicine images according to claim 1, wherein The specific method for obtaining the diseased joint according to the lesion characteristics of the joint gap includes: Preset a lesion threshold , for any joint space, if the lesion characteristics of the joint space are greater than or equal to , the joint corresponding to the joint space is diseased.

Images