Auxiliary identification method and system for tongue image of patient with diabetic nephropathy and medium

Abstract

The invention relates to the technical field of image processing, in particular to a tongue picture auxiliary identification method and system for a diabetic nephropathy patient and a medium. The method comprises the steps of obtaining tongue picture region images, clustering the tongue picture region images based on preset different numbers of clustering clusters, and screening an optimal clustering result according to a clustering result; obtaining the attribution degree of each pixel point in each cluster in the optimal clustering result by combining the gray scale gradient rule of the pixel points in different directions in the neighborhood of each pixel point and the relative position relationship between the pixel points and each cluster in the optimal clustering result; according to the difference of the attribution degrees of each pixel point and the neighborhood pixel points thereof in the same cluster in the optimal clustering result, acquiring the attribution expression degree of each pixel point and each cluster in the optimal clustering result; and according to the affiliation representation degree and the initial membership degree of each pixel point and each cluster in the optimal clustering result, the tongue picture region image is divided again. According to the invention, the accuracy of the tongue picture region image division result is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a method, system and medium for auxiliary identification of tongue images of patients with diabetic nephropathy. Background Art

[0002] Diabetic nephropathy affects the quality of life of patients. Tongue image, as a non-invasive biomarker, can reflect the physiological and pathological state of the human body. However, traditional tongue image recognition methods mainly rely on the doctor's experience and subjective judgment, requiring the doctor to observe every detail repeatedly and carefully, resulting in a large workload for the doctor.

[0003] During the development of diabetic nephropathy, changes in the human body will be reflected on the tongue, with changes in color and hyperplasia of texture. Image processing is used to assist tongue diagnosis, and clustering algorithms are usually used to divide the tongue image area. For example, the segmentation and quantification of areas such as abnormal tongue coating color, ecchymosis and cracks are used as auxiliary standards for identifying the development stage of diabetic nephropathy. Therefore, clustering is often performed on the images of the tongue area to extract features related to diabetic nephropathy, thereby reducing the workload of doctors. However, areas with more obvious changes in tongue coating color are easy to segment, while other abnormal manifestations are covered by adjacent areas during the FCM clustering process because of their small range in the entire tongue distribution space. Therefore, the clustering effect is not good, which in turn affects the subsequent analysis results. Summary of the invention

[0004] In order to solve the problem of poor clustering effect in the existing method when clustering pixels in the tongue image area image, the purpose of the present invention is to provide a tongue image auxiliary identification method, system and medium for diabetic nephropathy patients. The technical solution adopted is as follows: In a first aspect, the present invention provides a method for assisting in identifying tongue images of patients with diabetic nephropathy, the method comprising the following steps: Acquire tongue image of the patient with diabetic nephropathy to be tested; Based on the preset number of different clustering clusters, the pixel points in the tongue image region are clustered respectively, and the optimal clustering result is selected according to the grayscale discreteness of the pixel points in each clustering result; the grayscale gradient law of the pixel points in different directions in the neighborhood of each pixel point in the tongue image region and the relative position relationship between the pixel point and each cluster in the optimal clustering result are combined to obtain the degree of belonging of each pixel point to each cluster in the optimal clustering result; according to the difference between the degree of belonging of each pixel point and its neighboring pixel points in the same cluster in the optimal clustering result, the degree of belonging of each pixel point to each cluster in the optimal clustering result is obtained; According to the attribution expression and the initial membership of each pixel point to each cluster in the optimal clustering result, the tongue image region image is divided again to obtain different regions.

[0005] Preferably, the screening of the optimal clustering result according to the grayscale discreteness of the pixel points in each clustering result includes: Calculate the grayscale dispersion degree of all pixels in each cluster of the candidate result, and obtain the grayscale consistency index of the candidate result according to the grayscale dispersion degree of all pixels in all clusters of the candidate result, wherein the grayscale dispersion degree is negatively correlated with the grayscale consistency index; obtain the segmentation evaluation value of the candidate result according to the grayscale consistency index of the candidate result and the dispersion degree of the average grayscale value of the pixels in all clusters; the candidate result is any clustering result; The clustering result corresponding to the maximum value of the segmentation evaluation value is taken as the optimal clustering result.

[0006] Preferably, the grayscale gradient law of pixels in different directions in the neighborhood of each pixel in the tongue image region and the relative position relationship between the pixel and each cluster in the optimal clustering result are combined to obtain the degree of belonging of each pixel in each cluster in the optimal clustering result, including: Perform curve fitting on the grayscale values ​​of the pixels in the neighborhood of the candidate pixel in the feature direction and the grayscale value of the candidate pixel to obtain the fitting curve corresponding to the candidate pixel in the feature direction; According to the slope difference between each pixel point and its adjacent pixel points on the fitting curve, the grayscale gradient consistency of the candidate pixel point in the feature direction is obtained, and the slope difference is negatively correlated with the grayscale gradient consistency; The direction from the center point of each cluster in the optimal clustering result to the candidate pixel point is determined as the grayscale change direction of the candidate pixel point relative to each cluster; According to the grayscale gradient consistency of the candidate pixel points relative to each cluster in the grayscale change direction, the degree of belonging of the candidate pixel points to each cluster in the optimal clustering result is obtained; The candidate pixel point is any pixel point in the tongue image area image, and the characteristic direction is any direction.

[0007] Preferably, obtaining the degree of belonging of the candidate pixel to each cluster in the optimal clustering result according to the grayscale gradient consistency of the candidate pixel relative to each cluster in the grayscale change direction includes: The degree of belonging of the candidate pixel to the cluster to be analyzed is obtained according to the difference between the grayscale gradient consistency of the candidate pixel in the grayscale change direction relative to the cluster to be analyzed and the grayscale gradient consistency of the other clusters except the cluster to be analyzed, and the difference between the grayscale gradient consistency and the degree of belonging are negatively correlated; The cluster to be analyzed is any cluster in the optimal clustering result.

[0008] Preferably, obtaining the degree of belonging of each pixel to each cluster in the optimal clustering result according to the difference between the degrees of belonging of each pixel and its neighboring pixels in the same cluster in the optimal clustering result comprises: According to the difference between the degree of attribution of the candidate pixel in the cluster to be analyzed and the degree of attribution of the pixels in the neighborhood of the candidate pixel in the cluster to be analyzed, the attribution expression degree of the candidate pixel and the cluster to be analyzed is obtained, and the difference between the attribution degrees is negatively correlated with the attribution expression degree.

[0009] Preferably, the tongue image region image is divided again to obtain different regions according to the attribution expression and the initial membership of each pixel point to each cluster in the optimal clustering result, including: For any pixel point in the tongue image area: the product of the attribution expression degree of the pixel point to the cluster to be analyzed and the initial membership degree of the pixel point to the cluster to be analyzed is used as the final membership degree of the pixel point to the cluster to be analyzed; Each pixel point is divided into the cluster with the largest final membership to obtain different areas.

[0010] Preferably, the FCM clustering algorithm is used to cluster all the pixels in the tongue image area image.

[0011] Preferably, when clustering the pixels in the tongue image region image based on the preset number of different clustering clusters, it also includes: obtaining the initial membership degree of each pixel in the tongue image region image and each cluster in the clustering result.

[0012] In a second aspect, the present invention provides a tongue image auxiliary recognition system for patients with diabetic nephropathy, the system comprising: An image acquisition module, used for acquiring tongue image of the diabetic nephropathy patient to be detected; The attribution expression calculation module is used to cluster the pixels in the tongue image region image based on the number of preset different clustering clusters, and select the optimal clustering result according to the grayscale discreteness of the pixels in each clustering result; combine the grayscale gradient of the pixels in different directions in the neighborhood of each pixel in the tongue image region image and the relative position relationship between the pixel and each cluster in the optimal clustering result to obtain the attribution degree of each pixel in each cluster in the optimal clustering result; obtain the attribution expression degree of each pixel and each cluster in the optimal clustering result according to the difference between the attribution degree of each pixel and its neighboring pixels in the same cluster in the optimal clustering result; The region division module is used to divide the tongue image again to obtain different regions according to the attribution expression and the initial membership of each pixel point to each cluster in the optimal clustering result.

[0013] In a third aspect, the present invention provides a medium storing at least one program executable by a computer, wherein when the at least one program is executed by a computer, the computer executes the steps in the above-mentioned method for auxiliary identification of tongue images of patients with diabetic nephropathy.

[0014] The present invention has at least the following beneficial effects: The present invention first collects tongue image of a patient with diabetic nephropathy to be detected, then sets the number of multiple different clustering clusters, clusters the pixels in the tongue image region multiple times, analyzes the grayscale discreteness of the pixels in each clustering result, and selects the optimal clustering result; for patients with diabetic nephropathy, their tongue images will show special manifestations, and the pixels in the tongue image region image will show a certain grayscale change law, and further combines the grayscale gradient law of pixels in different directions in the neighborhood of each pixel in the tongue image region image and the relative position relationship between the pixel and each cluster in the optimal clustering result, and adjusts the initial membership of each pixel and each cluster in the optimal clustering result, so that the pixels in the tongue image region image are not only divided according to color differences, but also combined with the special texture change law of the tongue of patients with diabetic nephropathy, so that the division of each pixel is more accurate, and the regional division result is more complete and accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] 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.

[0016] Figure 1 A flow chart of a method for assisting in identifying tongue images of patients with diabetic nephropathy provided by an embodiment of the present invention; Figure 2 The optimal clustering result diagram of the tongue image region provided by the embodiment of the present invention; Figure 3 A flowchart of a method for obtaining attribution expressivity provided by an embodiment of the present invention; Figure 4 This is a structural block diagram of a tongue image auxiliary identification system for diabetic nephropathy patients provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0017] In order to further illustrate the technical means and effects adopted by the present invention to achieve the predetermined purpose of the invention, the following is a detailed description of a tongue image auxiliary identification method, system and medium for diabetic nephropathy patients proposed in accordance with the present invention in combination with the accompanying drawings and preferred embodiments.

[0018] 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.

[0019] The following is a detailed description of a method, system and medium for assisting in identifying tongue images of patients with diabetic nephropathy provided by the present invention in conjunction with the accompanying drawings.

[0020] An embodiment of a method for assisting in identifying tongue images of patients with diabetic nephropathy: The specific scenario targeted by this embodiment is as follows: when analyzing the symptoms of patients with diabetic nephropathy, first collect the tongue area image of the diabetic nephropathy patients, then set different clustering parameters, cluster the pixels in the tongue area image of the diabetic nephropathy patients, and select the optimal clustering result. Then, based on the optimal clustering result, divide the pixels again to improve the accuracy of the division results of the pixels in the tongue area image, which is convenient for subsequent analysis.

[0021] This embodiment proposes a tongue image auxiliary identification method for patients with diabetic nephropathy. Figure 1 As shown, a tongue image auxiliary identification method for diabetic nephropathy patients in this embodiment includes the following steps: Step S1, obtaining a tongue region image of a diabetic nephropathy patient to be detected.

[0022] In the early morning, on an empty stomach, the patient with diabetic nephropathy to be tested cleans his mouth with clean water, disinfects the outer ring of the sampler with an alcohol cotton ball, and the patient with diabetic nephropathy to be tested sits upright, places the mandible on the tongue surface fixing frame, fully fits the face with the fixing frame, relaxes the muscles naturally, opens the mouth naturally, extends the tongue naturally out of the mouth, the tongue surface is flat and not curled, and the tip of the tongue is naturally downward. Maintain this fixed posture for a few seconds to complete the tongue image shooting, filter and denoise, color correction and grayscale processing are performed on the collected tongue image, and the processed image is recorded as the tongue image of the patient with diabetic nephropathy to be tested for subsequent analysis and processing. Filtering and denoising, color correction and grayscale processing are all existing technologies, and will not be described in detail here.

[0023] At this point, the tongue area image of the diabetic nephropathy patient to be detected is obtained.

[0024] Step S2, clustering the pixels in the tongue image area image based on the preset number of different clustering clusters, and selecting the optimal clustering result according to the grayscale discreteness of the pixels in each clustering result; combining the grayscale gradient law of the pixels in different directions in the neighborhood of each pixel in the tongue image area image and the relative position relationship between the pixel and each cluster in the optimal clustering result, obtaining the degree of belonging of each pixel to each cluster in the optimal clustering result; obtaining the degree of belonging of each pixel to each cluster in the optimal clustering result according to the difference between the degree of belonging of each pixel and its neighboring pixels in the same cluster in the optimal clustering result.

[0025] Tongue image is one of the important indicators that reflect the function and health status of human internal organs. Diabetic nephropathy is a common complication of diabetic patients. During its development, the body function will change accordingly, which will be reflected in the tongue image, mainly manifested as changes in tongue quality and tongue coating color, as well as the generation of abnormal manifestations such as ecchymosis and cracks. In the process of doctors diagnosing diseases through tongue images, relying on the doctor's subjective experience and observation, lighting conditions or weak abnormal symptoms may lead to missed diagnosis, while images can objectively record and save tongue images, repeatedly view and perform various operations, and observe the details of tongue images more carefully. At the same time, through a large number of case matching, it can provide doctors with a more comprehensive and accurate basis for diagnosis and treatment.

[0026] In the process of tongue image-assisted identification of diabetic nephropathy, the actual tongue diagnosis steps of doctors are simulated. First, the color of the tongue quality and tongue coating is observed, followed by abnormal manifestations such as cracks. Different areas of the tongue of diseased patients have different manifestations. In order to better observe and analyze the tongue, these areas are segmented in the tongue image to assist doctors in analysis.

[0027] When analyzing the tongue appearance of a diabetic nephropathy patient to be detected, this embodiment first clusters the pixels in the tongue image area image of the diabetic nephropathy patient to be detected, and then analyzes the grayscale change pattern of the pixels in each cluster based on the clustering results to obtain the possibility that the pixel belongs to the area, confirm its degree of belonging, adjust the degree of belonging of the clustering output, and divide the pixel again, thereby improving the accuracy of the tongue segmentation area and reducing errors.

[0028] Step S21, clustering the pixels in the tongue image region image multiple times based on the preset number of different clusters, and selecting the best clustering result according to the grayscale discreteness of the pixels in each clustering result.

[0029] The boundaries between different areas of the tongue are generally fuzzy. For example, the tongue coating and tongue quality have a gradual transition. The FCM clustering algorithm can better handle fuzziness and segment areas more flexibly. Therefore, the FCM clustering algorithm is used to perform regional segmentation on tongue area images with fuzzy boundary features.

[0030] The tongue image of a normal person is manifested by the color difference between the tongue coating and the tongue body, and the clustering cluster is 2. However, the tongue image of a patient with diabetic nephropathy may have various abnormal manifestations, which can be mainly divided into the tongue body, tongue quality, cracks and small plaque areas of petechiae. The number of clustering clusters is selected to be 2 to 5, that is, the number of clustering clusters is 2, 3, 4 and 5 respectively. The FCM clustering algorithm is used to cluster the pixels in the tongue image area image respectively, and the clustering results of each time and the membership of each pixel in each clustering result to each cluster in the clustering result are obtained. The FCM clustering algorithm and the process of obtaining the membership are both existing technologies and will not be described in detail here. In specific applications, the implementer can set the number of clustering clusters according to the specific situation.

[0031] When performing FCM clustering on tongue image region images of different patients, the number of clusters should be selected according to the performance of the tongue image. More or less clusters will lead to improper regional segmentation. The color difference of each pixel within the class should be small, while there is a large color difference between classes. Therefore, all clustering results are analyzed and the optimal clustering result is selected for subsequent analysis and division.

[0032] Specifically, the clustering result of a clustering is taken as an example for analysis, and other clustering results can be processed by the method provided in this embodiment.

[0033] The result of any clustering is recorded as a candidate result, and the grayscale discreteness of all pixels in each cluster of the candidate result is calculated. According to the grayscale discreteness of all pixels in all clusters of the candidate result, the grayscale consistency index of the candidate result is obtained, and the grayscale discreteness is negatively correlated with the grayscale consistency index. According to the grayscale values ​​of all pixels in each cluster of the candidate result, the average grayscale value of the pixels in each cluster of the candidate result is calculated. According to the grayscale consistency index of the candidate result and the discreteness of the average grayscale value of the pixels in all clusters, the segmentation evaluation value of the candidate result is obtained.

[0034] The present embodiment uses the variance of the grayscale values ​​of all pixels in each cluster to characterize the grayscale dispersion of all pixels in each cluster, and uses the variance of the average grayscale value of all pixels in the cluster to characterize the dispersion of the average grayscale value of all pixels in the cluster.

[0035] A negative correlation means that the dependent variable will decrease as the independent variable increases, and the dependent variable will increase as the independent variable decreases. It can be a subtraction relationship, a division relationship, etc., which is determined by actual application.

[0036] In this embodiment, a specific calculation formula for the segmentation evaluation value is given. The segmentation evaluation value of the clustering result of the b-th clustering can be expressed as: ; in, represents the segmentation evaluation value of the clustering result of the b-th clustering, N represents the number of clusters in the clustering result of the b-th clustering, It represents the variance of the grayscale values ​​of all pixels in the nth cluster in the clustering result of the bth clustering. It represents the variance of the average gray value of all pixels in the cluster in the clustering result of the b-th clustering. Represents an exponential function with a natural constant as base.

[0037] The variance of the grayscale values ​​of all pixels in the nth cluster in the clustering result of the bth clustering is used to reflect the grayscale discreteness of all pixels in the nth cluster. The larger the value, the more discrete the grayscale distribution of the pixels in the cluster. It represents the gray consistency index of the clustering result of the b-th clustering. The larger the value, the greater the color consistency of the pixels in the same cluster in the clustering result of the b-th clustering, and the better the segmentation effect. The larger the variance of the average gray value of all pixels in the cluster in the clustering result of the b-th clustering, the greater the gray difference of pixels in different clusters, the better the clustering segmentation effect, and the larger the segmentation evaluation value.

[0038] By using the above method, the segmentation evaluation value of the clustering result of each clustering can be obtained. The larger the segmentation evaluation value, the more consistent the grayscale of the pixels in the same cluster, and the greater the grayscale difference of the pixels in different clusters, that is, the better the clustering effect. Therefore, the clustering result corresponding to the maximum value of the segmentation evaluation value is taken as the optimal clustering result, such as Figure 2 As shown, this figure is the optimal clustering result diagram of the tongue area image.

[0039] Step S22, combining the grayscale gradient rules of pixels in different directions in the neighborhood of each pixel in the tongue image area and the relative position relationship between the pixel and each cluster in the optimal clustering result, obtains the degree of belonging of each pixel in each cluster in the optimal clustering result.

[0040] After selecting the optimal clustering results, there are still some inaccurate pixel segmentation problems in the optimal clustering results. This is because the boundaries between the tongue image regions are fuzzy, and the regions cannot be accurately clustered and segmented based on color differences. It is also necessary to adjust the FCM clustering results according to the abnormal tongue image characteristics of diabetic nephropathy patients to accurately segment the regions according to the actual tongue image performance. FCM clustering allows each pixel to belong to multiple clusters with a certain degree of membership. Through the characteristics of tongue image performance, the degree of membership of each pixel belonging to different clusters is adjusted, so that the pixel is segmented into the correct area.

[0041] Abnormal manifestations in patients with diabetic nephropathy at various stages of development include changes in tongue color, an increase in petechiae or ecchymoses, changes in tongue coating color, an increase in cracks, etc. As diabetic nephropathy develops, the tongue coating color gradually becomes more different from the tongue color, changing from the center of the tongue to the edge of the tongue. The boundary between the tongue coating and the tongue is relatively blurred, but the grayscale value gradient direction from the center of the tongue coating to the edge of the tongue is consistent. There is a gradient rule for the color of pixels in different regions, which causes the color of pixels at the boundary of the region to change to a high degree of similarity with the color of another region, and thus is mistakenly segmented into other regions. Therefore, when a pixel point conforms to the grayscale gradient rule in the direction from the center of the region to the pixel point, it means that the pixel point has been segmented into the correct region.

[0042] Based on this, this embodiment analyzes the grayscale gradient rules of pixels in different directions within the neighborhood of the pixel point to determine the degree of belonging of each pixel point to each cluster in the optimal clustering result.

[0043] Taking a pixel point in the tongue image region image as an example for analysis, other pixel points in the tongue image region image can be processed using the method provided in this embodiment.

[0044] Specifically, any pixel point in the tongue image area image is recorded as a candidate pixel point, and any direction is recorded as a feature direction. Curve fitting is performed on the grayscale values ​​of the pixels in the neighborhood of the candidate pixel point in the feature direction and the grayscale value of the candidate pixel point, respectively, to obtain the fitting curve corresponding to the candidate pixel point in the feature direction, and obtain the slope corresponding to each pixel point on the fitting curve; the horizontal coordinate of the fitting curve is the order of the pixels, and the vertical coordinate is the grayscale value of the pixels. The curve fitting and slope acquisition methods are both existing technologies and will not be described in detail here.

[0045] According to the slope difference between each pixel point and its adjacent pixel points on the fitting curve, the grayscale gradient consistency of the candidate pixel point in the feature direction is obtained, and the slope difference is negatively correlated with the grayscale gradient consistency.

[0046] Among them, the negative correlation relationship means that the dependent variable will decrease as the independent variable increases, and the dependent variable will increase as the independent variable decreases. It can be a subtraction relationship, a division relationship, etc., which is determined by practical applications.

[0047] In this embodiment, a specific calculation formula for grayscale gradient consistency is given. The grayscale gradient consistency of the i-th pixel point in the tongue image region in the f-th direction can be expressed as: ; in, Indicates the grayscale gradient consistency of the i-th pixel in the tongue image in the f-th direction, represents the total number of pixels in the i-th pixel and its neighborhood in the f-th direction, It represents the slope of the rth pixel on the fitting curve corresponding to the i-th pixel in the f-th direction; Indicates the position of the i-th pixel on the fitting curve corresponding to the f-th direction. The slope corresponding to the pixel point, Indicates the absolute value sign.

[0048] The slope corresponding to the pixel point on the fitting curve is used to reflect the grayscale change. It is used to reflect the relationship between the rth pixel and the fth pixel on the fitting curve corresponding to the i-th pixel in the f-th direction. The difference between the slopes corresponding to the pixels, It represents the average difference in the slopes between adjacent pixels on the fitting curve corresponding to the ith pixel in the fth direction. It is used to reflect the consistency of the grayscale change direction between all pixels in the neighborhood of the ith pixel in the fth direction. The smaller the value, the higher the consistency of the grayscale value change direction in this direction, the stronger the grayscale gradient rule, that is, the greater the grayscale gradient consistency.

[0049] By adopting the above method, the grayscale gradient consistency of each pixel in the tongue image area in each direction can be obtained.

[0050] Next, we will still take the candidate pixel as an example to explain that the direction from the center point of each cluster in the optimal clustering result to the candidate pixel is determined as the grayscale change direction of the candidate pixel relative to each cluster. Next, based on the grayscale gradient consistency of the candidate pixel in the grayscale change direction relative to each cluster, the degree of belonging of the candidate pixel to each cluster in the optimal clustering result is obtained.

[0051] Specifically, any cluster in the optimal clustering result is recorded as the cluster to be analyzed, and the degree of belonging of the candidate pixel point in the cluster to be analyzed is obtained according to the difference between the grayscale gradient consistency of the candidate pixel point in the grayscale change direction relative to the cluster to be analyzed and the grayscale gradient consistency of other clusters except the cluster to be analyzed. The difference between the grayscale gradient consistencies is negatively correlated with the degree of belonging.

[0052] Among them, the negative correlation relationship means that the dependent variable will decrease as the independent variable increases, and the dependent variable will increase as the independent variable decreases. It can be a subtraction relationship, a division relationship, etc., which is determined by practical applications.

[0053] In this embodiment, a specific calculation formula for the degree of attribution is given. The degree of attribution of the i-th pixel point in the tongue image region to the j-th cluster in the optimal clustering result can be expressed as: ; in, represents the degree of belonging of the i-th pixel in the tongue image region to the j-th cluster in the optimal clustering result, M represents the number of clusters in the optimal clustering result, It represents the grayscale gradient consistency of the i-th pixel in the tongue image area in the grayscale change direction relative to the j-th cluster. It represents the grayscale gradient consistency of the i-th pixel in the tongue image area in the grayscale change direction of the m-th cluster except the j-th cluster.

[0054] It represents the difference between the grayscale gradient consistency of the i-th pixel in the grayscale change direction relative to the j-th cluster and the grayscale gradient consistency of the i-th pixel in the grayscale change direction of the m-th cluster except the j-th cluster. It indicates the difference between the grayscale gradient consistency of the i-th pixel in the grayscale change direction relative to the j-th cluster and the grayscale gradient consistency of the i-th pixel in the grayscale change direction of other clusters except the j-th cluster. When the value is larger, it means that the grayscale gradient law of the i-th pixel in the change direction of the j-th cluster is greater than the grayscale gradient law in other clusters, which means that the i-th pixel has a higher degree of belonging to the j-th cluster in the optimal clustering result.

[0055] By adopting the above method, the degree of belonging of each pixel point in the tongue image region to each cluster in the optimal clustering result can be obtained.

[0056] Step S23, obtaining the belonging expression degree of each pixel point to each cluster in the optimal clustering result according to the difference between the belonging degrees of each pixel point and its neighboring pixels in the same cluster in the optimal clustering result.

[0057] Considering that the degree of belonging to each cluster of each pixel in the optimal clustering result is prone to errors or accidents, it is necessary to further confirm the degree of belonging of each pixel to each cluster based on the similarity of the degrees of belonging of other pixels in the neighborhood of each pixel. For any pixel, the more similar the degrees of belonging of other pixels in the neighborhood of the pixel are to the degree of belonging of the pixel to a certain area, the greater the degree of belonging of the pixel to the area.

[0058] Based on this, this embodiment still takes the candidate pixel point as an example for analysis, and obtains the attribution expression degree of the candidate pixel point and the cluster to be analyzed according to the difference between the attribution degree of the candidate pixel point in the cluster to be analyzed and the attribution degree of the pixel points in the neighborhood of the candidate pixel point in the cluster to be analyzed. The difference between the attribution degrees is negatively correlated with the attribution expression degree.

[0059] Among them, the negative correlation relationship means that the dependent variable will decrease as the independent variable increases, and the dependent variable will increase as the independent variable decreases. It can be a subtraction relationship, a division relationship, etc., which is determined by practical applications.

[0060] In this embodiment, a specific calculation formula for the attribution expression is given. The attribution expression of the i-th pixel point in the tongue image region in the j-th cluster in the optimal clustering result can be expressed as: ; in, It represents the degree of belonging of the i-th pixel point in the tongue image area to the j-th cluster in the optimal clustering result. It indicates the degree of belonging of the i-th pixel point in the tongue image area to the j-th cluster in the optimal clustering result. It represents the mean of the degree of belonging of all pixels in the neighborhood of the i-th pixel in the tongue image to the j-th cluster in the optimal clustering result.

[0061] It represents the difference between the degree of belonging of the ith pixel in the jth cluster in the optimal clustering result and the mean degree of belonging of other pixels in its neighborhood in the jth cluster in the optimal clustering result. The smaller the value, the more likely the ith pixel is to belong to the jth cluster, and the greater its degree of belonging expression.

[0062] In this embodiment, for the i-th pixel point, the pixel points in its neighborhood are the pixel points in its eight neighborhoods. In other embodiments, the implementer may set it according to specific circumstances.

[0063] By using the above method, the degree of belonging of each pixel point to each cluster in the optimal clustering result can be obtained, such as Figure 3 As shown, this figure is a flow chart of a method for obtaining attribution expressivity.

[0064] Step S3, according to the attribution expression degree and the initial membership degree of each pixel point to each cluster in the optimal clustering result, the tongue image region image is divided again to obtain different regions.

[0065] Next, this embodiment adjusts the membership obtained in the FCM clustering according to the degree of belonging of each pixel point to each cluster in the optimal clustering result. The greater the degree of belonging of each pixel point to each region, the more likely the pixel point belongs to the region. The membership of the pixel point in the region is increased so that it can be segmented into the region during the clustering process.

[0066] For any pixel in the tongue image area image: the product of the attribution expression of the pixel to the cluster to be analyzed and the initial membership of the pixel to the cluster to be analyzed is used as the final membership of the pixel to the cluster to be analyzed. By adopting this method, the final membership of each pixel to each cluster in the optimal clustering result can be obtained. Among them, the initial membership is the membership of each pixel to each cluster in the optimal clustering result when obtaining the optimal clustering result. The process of obtaining the membership is a prior art and will not be described in detail here.

[0067] Next, each pixel point is divided into the corresponding cluster with the largest final membership degree to obtain multiple regions, that is, the pixel points in the tongue image region are divided into multiple regions.

[0068] The method provided in this embodiment obtains the adjusted membership of each pixel point to each area according to the tongue manifestation pattern of patients with diabetic nephropathy, and assigns each pixel point to the cluster area with the largest final membership to obtain the final image segmentation result, thereby improving the accuracy of the segmentation result. Doctors can subsequently analyze each area separately to improve the analysis efficiency and the accuracy of the evaluation results.

[0069] This embodiment first collects the tongue image of the diabetic nephropathy patient to be detected, and then sets the number of multiple different clustering clusters, and clusters the pixels in the tongue image area image multiple times, and analyzes the grayscale discreteness of the pixels in each clustering result to screen the optimal clustering result; for patients with diabetic nephropathy, their tongue image will show special manifestations, and the pixels in the tongue image area image will show a certain grayscale change law. Further combined with the grayscale gradient law of pixels in different directions in the neighborhood of each pixel in the tongue image area image and the relative position relationship between the pixel and each cluster in the optimal clustering result, the initial membership of each pixel and each cluster in the optimal clustering result is adjusted, so that the pixels in the tongue image area image are not only divided according to color differences, but also combined with the special texture change law of the tongue of patients with diabetic nephropathy, so that the division of each pixel is more accurate and the regional division result is more complete and accurate.

[0070] An embodiment of a tongue image auxiliary identification system for patients with diabetic nephropathy: like Figure 4 As shown, the figure shows a structural block diagram of a tongue image auxiliary recognition system for patients with diabetic nephropathy, the system comprising an image acquisition module, an attribution expression calculation module and a region division module; An image acquisition module, used for acquiring tongue image of the diabetic nephropathy patient to be detected; The attribution expression calculation module is used to cluster the pixels in the tongue image region image based on the number of preset different clustering clusters, and select the optimal clustering result according to the grayscale discreteness of the pixels in each clustering result; combine the grayscale gradient of the pixels in different directions in the neighborhood of each pixel in the tongue image region image and the relative position relationship between the pixel and each cluster in the optimal clustering result to obtain the attribution degree of each pixel in each cluster in the optimal clustering result; obtain the attribution expression degree of each pixel and each cluster in the optimal clustering result according to the difference between the attribution degree of each pixel and its neighboring pixels in the same cluster in the optimal clustering result; The region division module is used to divide the tongue image again to obtain different regions according to the attribution expression and the initial membership of each pixel point to each cluster in the optimal clustering result.

[0071] It should be understood that Figure 4 The structural block diagram of a tongue image auxiliary identification system for diabetic nephropathy patients and its modules can be implemented in various ways. For example, in some embodiments, the system and its modules can be implemented by hardware, software, or a combination of software and hardware. Among them, the hardware part can be implemented using dedicated logic; the software part can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or dedicated design hardware. Those skilled in the art can understand that the above-mentioned method and system can be implemented using computer executable instructions and / or included in a processor control code, for example, such as a carrier medium such as a disk, CD or DVD-ROM, a programmable memory such as a read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. Such code is provided on the carrier medium such as a disk, CD or DVD-ROM, a programmable memory such as a read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. The system and its modules of this specification can not only be implemented by hardware circuits such as ultra-large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, etc., or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc., but can also be implemented by software executed by various types of processors, and can also be implemented by a combination of the above-mentioned hardware circuits and software (for example, firmware).

[0072] For more details about the above modules, please refer to other places in this manual and will not be repeated here.

[0073] In other embodiments, a medium is also provided, which stores at least one program executable by a computer. When the at least one program is executed by a computer, the computer executes the steps of the tongue image auxiliary identification method for patients with diabetic nephropathy in the above embodiment. The medium can be a computer-readable storage medium.

[0074] Among them, the provided systems and media are used to execute the corresponding methods provided above. Therefore, the beneficial effects that can be achieved can refer to the beneficial effects in the corresponding methods provided above, and will not be repeated here.

[0075] It should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for assisting in identifying tongue images of patients with diabetic nephropathy, characterized in that: The method comprises the following steps: Acquire a tongue region image of a patient with diabetic nephropathy to be detected; Based on the preset number of different clustering clusters, the pixel points in the tongue image region are clustered respectively, and the optimal clustering result is selected according to the grayscale discreteness of the pixel points in each clustering result; the grayscale gradient law of the pixel points in different directions in the neighborhood of each pixel point in the tongue image region and the relative position relationship between the pixel point and each cluster in the optimal clustering result are combined to obtain the degree of belonging of each pixel point to each cluster in the optimal clustering result; according to the difference between the degree of belonging of each pixel point and its neighboring pixel points in the same cluster in the optimal clustering result, the degree of belonging of each pixel point to each cluster in the optimal clustering result is obtained; According to the attribution expression and the initial membership of each pixel point to each cluster in the optimal clustering result, the tongue image region image is divided again to obtain different regions.

2. The method for assisting in identifying tongue images of patients with diabetic nephropathy according to claim 1, characterized in that: The screening of the optimal clustering result according to the grayscale discreteness of the pixel points in each clustering result includes: Calculate the grayscale dispersion degree of all pixels in each cluster of the candidate result, and obtain the grayscale consistency index of the candidate result according to the grayscale dispersion degree of all pixels in all clusters of the candidate result, wherein the grayscale dispersion degree is negatively correlated with the grayscale consistency index; obtain the segmentation evaluation value of the candidate result according to the grayscale consistency index of the candidate result and the dispersion degree of the average grayscale value of the pixels in all clusters; the candidate result is any clustering result; The clustering result corresponding to the maximum value of the segmentation evaluation value is taken as the optimal clustering result.

3. The method for assisting in identifying tongue images of patients with diabetic nephropathy according to claim 1, characterized in that: The method combines the grayscale gradient law of pixels in different directions in the neighborhood of each pixel in the tongue image region and the relative position relationship between the pixel and each cluster in the optimal clustering result to obtain the degree of belonging of each pixel in each cluster in the optimal clustering result, including: Performing curve fitting on the grayscale values ​​of the pixels in the neighborhood of the candidate pixel in the feature direction and the grayscale value of the candidate pixel, respectively, to obtain a fitting curve corresponding to the candidate pixel in the feature direction; According to the slope difference between each pixel point and its adjacent pixel points on the fitting curve, the grayscale gradient consistency of the candidate pixel point in the feature direction is obtained, and the slope difference is negatively correlated with the grayscale gradient consistency; The direction from the center point of each cluster in the optimal clustering result to the candidate pixel point is determined as the grayscale change direction of the candidate pixel point relative to each cluster; According to the grayscale gradient consistency of the candidate pixel points relative to each cluster in the grayscale change direction, the degree of belonging of the candidate pixel points to each cluster in the optimal clustering result is obtained; The candidate pixel point is any pixel point in the tongue image area image, and the characteristic direction is any direction.

4. The method for assisting in identifying tongue images of patients with diabetic nephropathy according to claim 3, characterized in that: The step of obtaining the degree of belonging of the candidate pixel to each cluster in the optimal clustering result according to the grayscale gradient consistency of the candidate pixel relative to the grayscale change direction of each cluster includes: The degree of belonging of the candidate pixel to the cluster to be analyzed is obtained according to the difference between the grayscale gradient consistency of the candidate pixel in the grayscale change direction relative to the cluster to be analyzed and the grayscale gradient consistency of the other clusters except the cluster to be analyzed, and the difference between the grayscale gradient consistency and the degree of belonging are negatively correlated; The cluster to be analyzed is any cluster in the optimal clustering result.

5. The method for assisting in identifying tongue images of patients with diabetic nephropathy according to claim 3, characterized in that: The obtaining the belonging expression degree of each pixel point to each cluster in the optimal clustering result according to the difference between the belonging degrees of each pixel point and its neighboring pixel points in the same cluster in the optimal clustering result includes: According to the difference between the degree of attribution of the candidate pixel in the cluster to be analyzed and the degree of attribution of the pixels in the neighborhood of the candidate pixel in the cluster to be analyzed, the attribution expression degree of the candidate pixel and the cluster to be analyzed is obtained, and the difference between the attribution degrees is negatively correlated with the attribution expression degree.

6. The method for assisting in identifying tongue images of patients with diabetic nephropathy according to claim 4, characterized in that: The tongue image is divided again to obtain different regions according to the attribution expression and the initial membership of each pixel point to each cluster in the optimal clustering result, including: For any pixel point in the tongue image area: the product of the attribution expression degree of the pixel point to the cluster to be analyzed and the initial membership degree of the pixel point to the cluster to be analyzed is used as the final membership degree of the pixel point to the cluster to be analyzed; Each pixel point is divided into the cluster with the largest final membership to obtain different areas.

7. The method for assisting in identifying tongue images of patients with diabetic nephropathy according to claim 1, characterized in that: The FCM clustering algorithm is used to cluster all the pixels in the tongue area image.

8. The method for assisting in identifying tongue images of patients with diabetic nephropathy according to claim 1, characterized in that: When the pixels in the tongue image region image are clustered based on the preset number of different clusters, the method further includes: obtaining an initial membership degree between each pixel in the tongue image region image and each cluster in the clustering result.

9. A tongue image auxiliary recognition system for patients with diabetic nephropathy, characterized in that: The system comprises: An image acquisition module, used for acquiring tongue image of the diabetic nephropathy patient to be detected; The attribution expression calculation module is used to cluster the pixels in the tongue image region image based on the number of preset different clustering clusters, and select the optimal clustering result according to the grayscale discreteness of the pixels in each clustering result; combine the grayscale gradient of the pixels in different directions in the neighborhood of each pixel in the tongue image region image and the relative position relationship between the pixel and each cluster in the optimal clustering result to obtain the attribution degree of each pixel in each cluster in the optimal clustering result; obtain the attribution expression degree of each pixel and each cluster in the optimal clustering result according to the difference between the attribution degree of each pixel and its neighboring pixels in the same cluster in the optimal clustering result; The region division module is used to divide the tongue image again to obtain different regions according to the attribution expression and the initial membership of each pixel point to each cluster in the optimal clustering result.

10. A medium storing at least one program executable by a computer, characterized in that: When the at least one program is executed by a computer, the computer executes the steps of the method for assisting in identifying tongue images of patients with diabetic nephropathy as described in any one of claims 1 to 8.

Images