Tumor recurrence risk prediction system combined with traditional Chinese medicine constitution identification

Abstract

The invention relates to the technical field of medical image processing, in particular to a tumor recurrence risk prediction system combined with traditional Chinese medicine constitution identification. According to the method, the tongue body color characterization degree of the tongue body image is obtained according to the color features of the tongue body image; according to the position distribution of the tongue body tooth mark points and the tongue body tooth mark sharp points in the tongue body image, the tongue body tooth mark characterization degree of the tongue body image is obtained; according to the color features of the pixel points in the blood stasis point area and the cracking area of the tongue body and the position distribution of the pixel points among different areas of the tongue body, the tongue quality characterization degree of the tongue body image is obtained; and obtaining a tumor recurrence risk probability according to the difference between the sample points of the tumor tongue body images in the to-be-detected period and different clinical periods and the difference between the sample points of the tumor tongue body images in the to-be-detected period and the sample points of the healthy tongue body images in the to-be-detected period. According to the method, the accuracy of tumor recurrence risk prediction is improved by accurately analyzing the color, tongue quality and tongue shape characteristics in the tongue body image.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image processing, and in particular to a tumor recurrence risk prediction system combined with traditional Chinese medicine constitution identification. Background Art

[0002] In the process of TCM dialectics, the tongue image can be used to obtain the sensitivity of the changes in the state of the body's internal organs. The tongue images of different types of tumors are specific. At the same time, the tongue image is also highly correlated with the clinical stage. Therefore, the risk of tumor recurrence can be predicted by analyzing the tongue images.

[0003] In the prior art, traditional Chinese medicine tongue identification is used to observe the tongue condition in three main aspects: tongue color, tongue quality, and tongue coating. However, in the process of analyzing tongue cracks, the edge fracture curve cannot accurately characterize the health status of the tongue surface of tumor subjects. There is a lack of targeted correlation analysis of the tongue quality characteristics of tumor subjects, and the prediction results of tumor recurrence risk are poor. Summary of the invention

[0004] In order to solve the technical problem of lack of targeted correlation with the tongue characteristics of the subject and poor prediction results of tumor recurrence risk, the purpose of the present invention is to provide a tumor recurrence risk prediction system combined with TCM constitution identification. The technical scheme adopted is as follows: The present invention proposes a tumor recurrence risk prediction system combined with TCM constitution identification, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the following steps are implemented: Obtaining tumor tongue images and healthy tongue images of tumor subjects at different stages, wherein the tongue images include tooth mark points, petechiae areas, cracked areas, and middle areas of the tongue; According to the color difference characteristics of the tumor tongue image and the healthy tongue image, the tongue color representation degree of the tongue image is obtained; for the tumor tongue image or the healthy tongue image, the tongue tooth mark cusp is obtained according to the morphological characteristics of the tongue image; according to the position distribution of the tongue tooth mark marking points and the tongue tooth mark cusp, the tongue tooth mark representation degree of the tongue image is obtained; The tongue quality representation degree of the tongue image is obtained according to the color characteristics of the pixels in the petechiae and fissure areas of the tongue and the position distribution of the pixels between different areas on the tongue image; The tongue color representation degree, tongue quality representation degree and tongue tooth mark representation degree of the tongue image constitute the sample points of the feature space. According to the difference in sample points of tumor tongue images between the test period and different clinical periods, as well as the difference in sample points between tumor tongue images in the test period and healthy tongue images, the tumor recurrence risk probability in the test period is obtained.

[0005] Furthermore, the method for obtaining the degree of tongue body color characterization includes: Obtain the differences in different channels between the tumor tongue body image and the healthy tongue body image in the RGB space, and obtain the differences in different channels between the tumor tongue body image and the healthy tongue body image in the HSV space; Normalize the differences in all channels and accumulate the normalization results as the degree of tongue body color characterization of the tumor tongue body image; For the healthy tongue body image, set the degree of tongue body color characterization to 0.

[0006] Furthermore, the method for obtaining the tip points of tongue body tooth marks includes: Obtain the edge curve of the tongue body image. Take the minimum value point on the edge curve as the origin, the horizontal direction from left to right as the horizontal axis, and the vertical direction along the minimum value point as the vertical axis to construct an edge curve coordinate system; Obtain the maximum value points in the horizontal direction in the edge curve coordinate system as the tip points of tongue body tooth marks.

[0007] Furthermore, the method for obtaining the degree of tongue body tooth mark characterization includes: Connect the tongue body tooth mark marked points and the tip points of tongue body tooth marks in sequence through straight line segments according to a preset order to construct a closed polygon area as the closed area of tongue body tooth marks; Obtain the ratio of the number of pixel points in the closed area of tongue body tooth marks to the number of all pixel points in the tongue body image as the degree of tongue body tooth mark characterization.

[0008] Furthermore, the method for obtaining the degree of tongue texture characterization includes: Obtain the degree of tongue body stasis point characterization of the tongue body image according to the color characteristics of the pixel points in the tongue body stasis point area and the position distribution of the pixel points between the tongue body stasis point area and the tongue body middle area; Remove the intersecting pixel points of the tongue body stasis point area and the tongue body crack area from the tongue body crack area to obtain an updated tongue body crack area; obtain the edge curve in the updated tongue body crack area, and obtain the edge fracture curve according to the direction characteristics of the endpoints between the edge curves; Obtain the vertical distance from the endpoint of each edge fracture curve to the edge curve, select a preset number of edge curves with the smallest vertical distance as the adjacent curves of each edge fracture curve, and use the corresponding endpoints and the pixel points on the adjacent curves as reference pixel points; obtain the best merging curve direction of the edge fracture curve according to the local direction similarity of the corresponding reference pixel points between the edge fracture curve and different adjacent curves; Fit adjacent edge fracture curves with the same optimal merging curve direction to obtain fracture fitting curves and a set of cracking curves; obtain the tongue quality characterization degree of the tongue image according to the tongue stasis dot characterization degree of the tongue image, the number of curves in the set of cracking curves, and the changing trend of the curves.

[0009] Furthermore, the method for obtaining the tongue stasis dot characterization degree includes: Obtain the ratio of the number of intersection pixels between the tongue stasis dot area and the tongue middle area in the tongue image to the number of pixels in the tongue middle area as the first stasis dot possibility. Obtain the binary image of the tongue stasis dot area, and obtain the ratio of the number of pixels with a value of 1 to the number of pixels with a value of 0 in the binary image as the second stasis dot possibility. Obtain the product between the first stasis dot possibility and the second stasis dot possibility as the tongue stasis dot characterization degree of the tongue image.

[0010] Furthermore, the method for obtaining the edge fracture curve includes: Take the derivative at each endpoint of the edge curve as the directionality of each endpoint. If there is a consistent directionality of the endpoints with the minimum relative distance between adjacent edge curves, take the corresponding adjacent edge curves as edge fracture curves.

[0011] Furthermore, the method for obtaining the optimal merging curve direction includes: Obtain the principal component direction within the neighborhood range of the corresponding reference pixels between the edge fracture curve and each adjacent curve, and calculate the cosine similarity of the corresponding principal component direction as the local direction similarity between the edge fracture curve and the corresponding reference pixels of each adjacent curve. Select the one with the largest local direction similarity between the edge fracture curve and different adjacent curves, and take the principal component direction of the corresponding adjacent curve as the optimal merging curve direction of the edge fracture curve.

[0012] Furthermore, the method for obtaining the tongue quality characterization degree includes: Obtain the average slope of different curves in the tongue cracking area as the overall slope level; obtain the cumulative value of the difference between the slopes of different curves in the tongue cracking area and the overall slope level as the cracking chaos degree. Obtain the product of the tongue stasis dot characterization degree, the number of curves, and the cracking chaos degree as the tongue quality characterization degree.

[0013] Furthermore, the method for obtaining the tumor recurrence risk probability includes: Based on the first relative distance between the sample points of the tumor tongue image corresponding to the period to be measured and different clinical periods, and the second relative distance between the sample points of the tumor tongue image and the healthy tongue image in the period to be measured, the tumor recurrence risk probability in the period to be measured is obtained. The first relative distance is negatively correlated with the tumor recurrence risk probability, and the second relative distance is positively correlated with the tumor recurrence risk probability.

[0014] The present invention has the following beneficial effects: In the present invention, since the tumor tongue may have abnormal color due to pathological changes, the degree of tongue color representation of the tongue image is obtained according to the color difference characteristics between the tumor tongue image and the healthy tongue image; for the tumor tongue image or the healthy tongue image, according to the morphological characteristics of the tongue image, the tongue tooth mark apex points can be obtained, which can more accurately describe the morphological characteristics of the tongue; according to the position distribution of the tongue tooth mark marking points and the tongue tooth mark apex points, the degree of tongue tooth mark representation of the tongue image is obtained, which can more comprehensively analyze the tooth mark characteristics of the tongue; the change of the tongue texture characteristics may be closely related to the occurrence and development of certain diseases. According to the color characteristics of the pixel points in the tongue stasis point area and the cracking area, and the position distribution of the pixel points between different areas on the tongue image, the degree of tongue texture representation of the tongue image is obtained; taking the degree of tongue color representation, the degree of tongue texture representation, and the degree of tongue tooth mark representation of the tongue image as the sample points of the feature space, the health status and lesion characteristics of the tongue are more comprehensively described. According to the differences between the sample points of the tumor tongue image between the period to be measured and different clinical periods, and the differences between the sample points of the tumor tongue image and the healthy tongue image in the period to be measured, the tumor recurrence risk probability in the period to be measured is obtained. By accurately analyzing the characteristics of color, tongue texture, and tongue shape in the tongue image, the present invention improves the accuracy of tumor recurrence risk. 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 following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained according to these drawings.

[0016] Figure 1 It is a flowchart of an implementation method of a tumor recurrence risk prediction system combining traditional Chinese medicine constitution identification provided by an embodiment of the present invention; Figure 2 It is a flowchart of a method for obtaining the degree of tongue texture representation provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] In order 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 accompanying drawings and preferred embodiments, details a tumor recurrence risk prediction system that combines traditional Chinese medicine constitution identification, including its specific implementation manner, structure, features, and effects, as follows. 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 skilled in the technical field to which the present invention belongs.

[0019] The following specifically describes the specific solution of a tumor recurrence risk prediction system that combines traditional Chinese medicine constitution identification provided by the present invention with reference to the accompanying drawings.

[0020] Please refer to Figure 1 , which shows the implementation method flowchart of a tumor recurrence risk prediction system that combines traditional Chinese medicine constitution identification provided by an embodiment of the present invention. The method specifically includes: Step S1: Obtain the tumor tongue images and healthy tongue images of tumor examinees at different times. The tongue images include tongue body dent mark points, petechial areas, cracking areas, and the middle area of the tongue.

[0021] In the embodiment of the present invention, in order to predict and evaluate the tumor recurrence risk, it is necessary to analyze tongue image features such as tongue color and tongue texture at different times. First, at each time, the examinee sits on the examination chair, places the mandible on the mandible support plate of the head fixing frame, and instructs the examinee to keep the tongue surface dry, naturally extend the tongue, expose the tongue surface, and maintain the posture for 5 seconds. The implementer clicks and saves the captured facial image including the tongue body area of the examinee through the computer connected to the digital camera. In order to make a comparative observation, the facial image of a healthy examinee is obtained.

[0022] It should be noted that since the captured facial images contain features of other parts of the face, in order to facilitate more targeted analysis of the tongue body of the examinee, the GraphCut technology is used to segment the facial images of tumor examinees and healthy examinees respectively to obtain the tumor tongue images and healthy tongue images. Specifically, GraphCut is a technical means well-known to those skilled in the art and will not be elaborated here.

[0023] In the process of analyzing the tongue images of gastric tumor subjects, due to the different health conditions of the subjects themselves, some tumor subjects also have other complications when suffering from tumors, resulting in interference in the obtained tongue images. When using computer software to analyze tongue images, large errors are likely to occur. It should be noted that in the embodiments of the present invention, by combining the relevant experience of the implementers, the tongue tooth mark marking points, the stasis point areas, the cracking areas, and the middle tongue areas of the tongue are obtained through computer interaction for analysis.

[0024] Step S2: According to the color difference characteristics between the tumor tongue image and the healthy tongue image, obtain the tongue color representation degree of the tongue image; for the tumor tongue image or the healthy tongue image, according to the morphological characteristics of the tongue image, obtain the tongue tooth mark cusp points; according to the position distribution of the tongue tooth mark marking points and the tongue tooth mark cusp points, obtain the tongue tooth mark representation degree of the tongue image.

[0025] The tongue color of tumor subjects is relatively dark, being dark red or crimson, while the tongue color of healthy subjects is mostly light red. Therefore, there is a certain color gap in the tongue color between the tumor tongue image and the healthy tongue image. The larger the gap, the more obvious the tumor symptoms are shown in the tumor tongue image, and the greater the tongue color representation degree; according to the color difference characteristics between the tumor tongue image and the healthy tongue image, obtain the tongue color representation degree of the tumor tongue image.

[0026] Preferably, in an embodiment of the present invention, the method for obtaining the tongue color representation degree includes: Obtain the differences in different channels between the tumor tongue image and the healthy tongue image in the RGB space, and obtain the differences in different channels between the tumor tongue image and the healthy tongue image in the HSV space; Normalize the differences of all channels, and accumulate the normalization results as the tongue color representation degree of the tumor tongue image; For the healthy tongue image, set the tongue color representation degree to 0.

[0027] In an embodiment of the present invention, the RGB space and the HSV space reveal the pathological characteristics of the tongue image from different angles, reflect the color characteristics of the pixel points and the illumination information, and more accurately reflect the tongue symptoms; the larger the difference in channels, the less relevant the characteristics shown between the tumor tongue image and the healthy tongue image, the more obvious the symptoms shown, the more likely the tongue color is to show the tongue symptom characteristics, and the greater the tongue color representation degree.

[0028] Due to abnormal physical reactions, the more obvious the tooth mark changes in the tongue shape of the subject are, and the physiological state is reflected through the morphological characteristics of the tongue image; for the tumor tongue image or the healthy tongue image, according to the morphological characteristics of the tongue image, obtain the tongue tooth mark cusp points.

[0029] Preferably, in an embodiment of the present invention, the method for obtaining the tip points of the tooth marks on the tongue body includes: Obtain the edge curve of the tongue body image. Taking the minimum point on the edge curve as the origin, the horizontal direction from left to right as the horizontal axis, and the vertical direction along the minimum point as the vertical axis, construct an edge curve coordinate system; Obtain the maximum points in the horizontal direction in the edge curve coordinate system as the tip points of the tooth marks on the tongue body.

[0030] It should be noted that in the embodiment of the present invention, when processing the image by using the GraphCut technology, the edge curve in the tongue body image is obtained by interactive box selection or the active contour model. The specific means are well-known technical means to those skilled in the art and will not be elaborated here.

[0031] The tongue body is in a swollen pathological state, and it is more likely to have tooth marks. The tooth mark marking points reflect the spatial distribution density of the tooth marks, and the tip points of the tooth marks reflect the depth and morphological sharpness of the tooth marks. The larger the position range of the tooth marks, the greater the degree of tooth mark characterization, and the more likely it is to show tumor characteristics; according to the position distribution of the tooth mark marking points and the tip points of the tooth marks on the tongue body, obtain the degree of tooth mark characterization of the tongue body image.

[0032] Preferably, in an embodiment of the present invention, the method for obtaining the degree of tooth mark characterization of the tongue body includes: Connect the tooth mark marking points and the tip points of the tooth marks on the tongue body in sequence through straight line segments according to a preset order to construct a closed polygon area as the closed area of the tooth marks on the tongue body; Obtain the ratio of the number of pixel points in the closed area of the tooth marks on the tongue body to the number of all pixel points in the tongue body image as the degree of tooth mark characterization of the tongue body.

[0033] It should be noted that in the embodiment of the present invention, the preset order is clockwise or counterclockwise, and the tooth mark marking points and the tip points of the tooth marks on the tongue body are connected in sequence, which will not be elaborated here.

[0034] Different parts of the tongue body correspond to the pathological conditions of the body's internal organs. The middle area of the tongue reflects the abnormal health status of tumor patients; the formation of stasis points on the tongue is mostly due to microcirculation disorders of the tongue. Due to the aggregation of red blood cells in the microvessels of the tongue body into clusters, visible stasis patches and stasis points are formed. The stasis patches and stasis points are black. By analyzing the color characteristics of the pixel points in the stasis point area of the tongue body and the distribution relative to the middle area of the tongue, quantify the degree of stasis point characterization of the tongue body. According to the color characteristics of the pixel points in the stasis point area of the tongue body and the position distribution of the pixel points between the stasis point area of the tongue body and the middle area of the tongue body, obtain the degree of stasis point characterization of the tongue body image.

[0035] Preferably, in an embodiment of the present invention, the method for obtaining the degree of stasis point characterization of the tongue body includes: Obtain the ratio of the number of intersecting pixel points between the tongue stasis point area and the middle area of the tongue body in the tongue body image to the number of pixel points in the middle area of the tongue body as the first stasis point possibility; Obtain the binary image of the tongue stasis point area on the tongue surface, and obtain the ratio of the number of pixel points with a value of 1 to the number of pixel points with a value of 0 in the binary image as the second stasis point possibility; Obtain the product between the first stasis point possibility and the second stasis point possibility as the degree of tongue stasis point characterization of the tongue body image.

[0036] In an embodiment of the present invention, the more the number of intersecting pixel points between the tongue stasis point area and the middle area of the tongue body, the more stasis points appear in the middle area of the tongue, the more the stasis point characteristics are reflected, and the greater the degree of tongue stasis point characterization; for binarization, the image of the tongue stasis point area is used as the input of the OTSU method. The darker-colored image pixel points on the tongue body image are mapped to 1, and the remaining pixel points are mapped to 0 to obtain binarization; the more the number of pixel points with a value of 1, the larger the range of darker-colored pixel points, the more it reflects, and the higher the possibility of the degree of tongue stasis point characterization, and the more likely there are tumor characteristics. The specific OTSU method is a well-known technical means for those skilled in the art and will not be elaborated here.

[0037] Step S3: Obtain the degree of tongue quality characterization of the tongue body image according to the color characteristics of the pixel points in the tongue stasis point area and the cracking area, and the position distribution of the pixel points between different areas on the tongue body image.

[0038] The middle part of the tongue surface represents the mapping reaction area of the stomach. In most cases, the formation of stasis points is caused by tongue microcirculation disorders. Since the red blood cells in the microvessels of the tongue body aggregate into groups, visible ecchymoses and stasis points are formed. The more stasis points in the middle of the tongue, the more obvious the characteristics of tumors are reflected; the tongue stasis point area of the tongue body may overlap with the cracking area, resulting in an impact on the generation of a confusion curve; and the more cracking curves in the cracking area, the greater the tongue quality characteristic performance index, the greater the tongue quality characteristic performance, and the more the characteristics of the tumor tongue body image can be reflected. Therefore, the degree of tongue quality characterization of the tongue body image is obtained according to the color characteristics of the pixel points in the tongue stasis point area and the cracking area, and the position distribution of the pixel points between different areas on the tongue body image.

[0039] Preferably, in an embodiment of the present invention, for the method of obtaining the degree of tongue quality characterization, please refer to Figure 2 which shows a flowchart of a method for obtaining the degree of tongue quality characterization, including: Step S201: Obtain the degree of tongue stasis point characterization of the tongue body image according to the color characteristics of the pixel points in the tongue stasis point area, and the position distribution of the pixel points between the tongue stasis point area and the middle area of the tongue body.

[0040] Preferably, in an embodiment of the present invention, the method for obtaining the characterization degree of tongue body stasis points includes: Obtaining the ratio of the number of intersection pixels between the tongue body stasis point area and the middle area of the tongue body in the tongue body image to the number of pixels in the middle area of the tongue body as the first stasis point possibility; Obtaining a binary image of the tongue body stasis point area, and obtaining the ratio of the number of pixels with a value of 1 to the number of pixels with a value of 0 in the binary image as the second stasis point possibility; Obtaining the product between the first stasis point possibility and the second stasis point possibility as the characterization degree of the tongue body stasis points in the tongue body image.

[0041] Among them, the more the number of intersection pixels between the tongue body stasis point area and the middle area of the tongue body, the more obvious the stasis points in the middle of the tongue, and the greater the characterization degree of the stasis points; the color of the pixels in the stasis point area is darker than that of the pixels in other areas. By analyzing the binary image, the image information can be simplified, and the distribution of the stasis points can be more intuitively understood. The stasis point area of the tongue body is darker in color, and in the binary image, the pixel value of the stasis point pixels is larger than that of the pixels on the surface of other tongue bodies. The more the number of pixels with a larger value, the larger the proportion of the stasis point area, and the greater the characterization degree of the stasis points.

[0042] It should be noted that, in an embodiment of the present invention, the image of the tongue body stasis point area can be used as the input of the OTSU method to obtain a binary image of the tongue body stasis point area. The specific means are well-known technical means to those skilled in the art and will not be elaborated here.

[0043] Step S202: Remove the intersection pixels of the tongue body stasis point area and the tongue body crack area from the tongue body crack area to obtain an updated tongue body crack area; obtain the edge curve in the updated tongue body crack area, and obtain an edge break curve according to the direction characteristics of the endpoints between the edge curves.

[0044] The tongue body stasis point area and the crack area may overlap, resulting in some confusing curves formed by stasis points. It is necessary to remove the intersection pixels. In other embodiments of the present invention, the pixel values in the binary image can also be inverted, and the product of the gray value of the pixels in the crack area and the pixel value of the corresponding position in the inverted binary image of the stasis point area is calculated as the updated gray value of the pixels in the crack area. Among them, the pixel value of the stasis point pixels in the inverted binary image becomes 0, and the updated gray value is 0, so as to eliminate the stasis points in the crack area.

[0045] It should be noted that, in an embodiment of the present invention, the method for obtaining the edge break curve includes: Taking the derivative at each endpoint of the edge curve as the directionality of each endpoint; If the directions of the endpoints corresponding to the minimum relative distances between adjacent edge curves are consistent, the corresponding adjacent edge curves are regarded as edge fracture curves.

[0046] Step S203: Obtain the perpendicular distances from the endpoints of each edge fracture curve to the edge curve, select a preset number of edge curves with the minimum perpendicular distances as the adjacent curves of each edge fracture curve, and use the corresponding endpoints and the pixel points on the adjacent curves as reference pixel points; according to the local direction similarity of the corresponding reference pixel points between the edge fracture curve and different adjacent curves, obtain the optimal merging curve direction of the edge fracture curve.

[0047] It should be noted that the distance between an endpoint and a curve is the perpendicular line drawn from this endpoint to the curve, and the distance from the position of this endpoint to the foot of the perpendicular is used as the perpendicular distance to quantify the relative position characteristics of each endpoint and the curve; the preset number can be specifically set according to specific situations and will not be limited and elaborated here.

[0048] Preferably, in an embodiment of the present invention, the method for obtaining the optimal merging curve direction includes: Obtain the principal component directions within the neighborhood range of the corresponding reference pixel points between the edge fracture curve and each adjacent curve, calculate the cosine similarity of the corresponding principal component directions as the local direction similarity of the corresponding reference pixel points between the edge fracture curve and each adjacent curve; Select the one with the largest local direction similarity among the edge fracture curve and different adjacent curves, and use the principal component direction of the corresponding adjacent curve as the optimal merging curve direction of the edge fracture curve.

[0049] It should be noted that, in an embodiment of the present invention, the neighborhood range is a 15×15 window area constructed with the reference pixel point as the center and adjacent pixel points. In other embodiments of the present invention, the neighborhood range can be specifically set according to specific situations and will not be limited and elaborated here.

[0050] It should be noted that the principal component direction reflects the main change direction of the curve within the local neighborhood range and can be obtained through the PCA algorithm. The specific means are well-known technical means to those skilled in the art and will not be elaborated here.

[0051] Step S204: Fit the adjacent edge fracture curves with the same optimal merging curve direction to obtain a fracture fitting curve and obtain a set of cracking curves; according to the tongue stasis dot characterization degree of the tongue image, the number of curves in the set of cracking curves, and the change trend of the curves, obtain the tongue texture characterization degree of the tongue image.

[0052] Based on this, existing fitting methods such as polynomial fitting or least squares method are used to fit the edge fracture curve to form a complete curve, which together with the remaining edge curves forms a set of cracking curves. The specific fitting method is a well-known technical means for those skilled in the art and will not be elaborated here.

[0053] Preferably, in an embodiment of the present invention, the method for obtaining the characterization degree of the tongue body and tongue texture includes: Obtain the average slope of different curves in the cracked area of the tongue body as the overall slope level; obtain the cumulative value of the difference between the slopes of different curves in the cracked area of the tongue body and the overall slope level as the degree of cracking disorder. Obtain the product of the characterization degree of tongue stasis points, the number of curves, and the degree of cracking disorder as the characterization degree of tongue texture.

[0054] It should be noted that the greater the difference between the slopes of different curves in the cracked area of the tongue body and the overall slope level, the more the slope of the curve deviates from the overall slope level, the greater the difference in the change of the cracked curves in the cracked area, the more disordered the distribution, the more tumor characteristics can be reflected, the greater the characterization degree of tongue stasis points, and the greater the characterization degree of tongue texture.

[0055] Step S4: Use the characterization degree of the tongue body color, the characterization degree of the tongue texture, and the characterization degree of tongue indentation of the tongue body image to form sample points in the feature space. According to the differences between the sample points of the tumor tongue body images between the test period and different clinical periods, and the differences between the sample points of the tumor tongue body image in the test period and the healthy tongue body image, obtain the tumor recurrence risk probability in the test period.

[0056] Preferably, the tongue textures of tumor subjects in different periods have different manifestations. By analyzing the correlation relationships of multiple features in the tongue body image to reflect the manifestation of the tumor. In an embodiment of the present invention, the method for obtaining the tumor recurrence risk probability includes: Obtain the tumor recurrence risk probability of the tumor subject according to the first relative distance between the sample points corresponding to the tumor tongue body images between the test period and different clinical periods, and the second relative distance between the sample points of the tumor tongue body image in the test period and the healthy tongue body image. The first relative distance is negatively correlated with the tumor recurrence risk probability, and the second relative distance is positively correlated with the tumor recurrence risk probability.

[0057] It should be noted that according to relevant professional knowledge, in clinical surgery, tumors are divided into four stages. Tumor tongue images in different stages have different characteristic manifestations. By analyzing the degree of correlation of characteristic representations between the tumor tongue image of the stage to be measured and other tongue images, the greater the degree of correlation, the smaller the relative distance, and the more similar the relevant characteristics of the tumor tongue. Clinically, the later the tumor of the tumor subject, the more obvious the characteristics of the tumor, and the first relative distance can better reflect the credibility of the relevant characteristics. Therefore, the greater the preset weight parameter set for the first relative distance of the tumor tongue image in the later stage, the greater the first relative distance, the farther away from the tumor characteristics, and the smaller the recurrence probability. The second relative distance of the sample points between the tumor tongue image and the healthy tongue image reflects the relevant characteristics with the healthy subject. The smaller the second relative distance, the smaller the difference between the tumor tongue image and the healthy tongue image, the more it shows healthy characteristics, and the smaller the recurrence probability. Among them, the preset weight parameters for the 4 stages are 0.15, 0.2, 0.3, and 0.35 in sequence.

[0058] In an embodiment of the present invention, for a tumor subject, calculate the cumulative value of the product of the first relative distance corresponding to all clinical stages and the corresponding preset weight parameter, obtain the ratio of the cumulative value to the second relative distance, and perform a negative correlation mapping through an exponential function with the natural constant as the base, as the tumor recurrence risk probability of the stage to be measured. That is, through the above basic mathematical operations, establish the correlation relationship between the first relative distance, the second relative distance, and the tumor recurrence risk probability. The greater the first relative distance and the smaller the second relative distance, the greater the recurrence risk probability.

[0059] It should be noted that in the embodiments of the present invention, the relative distance is obtained through existing distance calculation methods such as Euclidean distance or Manhattan distance. The specific Euclidean distance and Manhattan distance are well-known technical means to those skilled in the art and will not be elaborated here.

[0060] In summary, the present invention obtains the degree of tongue color representation of the tumor tongue image according to the color difference characteristics between the tumor tongue image and the healthy tongue image; obtains the degree of tongue indentation representation of the tongue image according to the position distribution of the tongue indentation marked points and the tongue indentation cusp points in the tongue image; obtains the degree of tongue quality representation of the tongue image according to the color characteristics of the pixel points in the tongue ecchymosis area and the cracked area, as well as the position distribution of the pixel points between different regions of the tongue; and then constitutes the sample points of the feature space. According to the differences in the sample points of the tumor tongue images between the stage to be measured and different clinical stages, and the differences in the sample points between the tumor tongue image of the stage to be measured and the healthy tongue image, obtain the tumor recurrence risk probability of the stage to be measured. The present invention improves the accuracy of tumor recurrence risk prediction by accurately analyzing the characteristics of color, tongue quality, and tongue shape in the tongue image.

[0061] It should be noted that the above order of the embodiments of the present invention is only for description and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0062] Each embodiment in this specification is described in a progressive manner. For the same or similar parts among the embodiments, reference can be made to each other. The key point of each embodiment is to illustrate the differences from other embodiments.

Claims

1. A tumor recurrence risk prediction system combined with traditional Chinese medicine constitution identification, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, the following steps are implemented: Obtaining tumor tongue images and healthy tongue images of tumor subjects at different stages, wherein the tongue images include tooth mark points, petechiae areas, cracked areas, and middle areas of the tongue; According to the color difference characteristics of the tumor tongue image and the healthy tongue image, the tongue color representation degree of the tongue image is obtained; for the tumor tongue image or the healthy tongue image, the tongue tooth mark cusp is obtained according to the morphological characteristics of the tongue image; according to the position distribution of the tongue tooth mark marking points and the tongue tooth mark cusp, the tongue tooth mark representation degree of the tongue image is obtained; The tongue quality representation degree of the tongue image is obtained according to the color characteristics of the pixels in the petechiae and fissure areas of the tongue and the position distribution of the pixels between different areas on the tongue image; The tongue color representation degree, tongue quality representation degree and tongue tooth mark representation degree of the tongue image constitute the sample points of the feature space. According to the difference in sample points of tumor tongue images between the test period and different clinical periods, as well as the difference in sample points between tumor tongue images in the test period and healthy tongue images, the tumor recurrence risk probability in the test period is obtained.

2. The tumor recurrence risk prediction system combined with traditional Chinese medicine constitution identification according to claim 1, characterized in that, The method for obtaining the degree of tongue color representation comprises: Obtain the difference between the tumor tongue image and the healthy tongue image in different channels in the RGB space, and obtain the difference between the tumor tongue image and the healthy tongue image in different channels in the HSV space; The differences of all channels are normalized, and the normalized results are accumulated as the tongue color representation degree of the tumor tongue image; For healthy tongue images, the tongue color representation level is set to 0.

3. The tumor recurrence risk prediction system combined with traditional Chinese medicine constitution identification according to claim 1, characterized in that, The method for obtaining the tip of the tongue tooth mark comprises: Obtain the edge curve of the tongue image, take the minimum value point in the edge curve as the origin, the horizontal direction from left to right as the horizontal axis, and the vertical direction along the minimum value point as the vertical axis to construct the edge curve coordinate system; The maximum point in the horizontal direction of the edge curve coordinate system is obtained as the tip point of the tongue tooth mark.

4. The tumor recurrence risk prediction system combined with traditional Chinese medicine constitution identification according to claim 1, characterized in that, The method for obtaining the degree of representation of the tooth marks on the tongue comprises: Connect the tongue tooth mark marking points and the tongue tooth mark tip points in sequence through straight line segments in a preset order to construct a closed polygonal area as the tongue tooth mark closed area; The ratio of the number of pixel points in the closed area of the tongue tooth marks to the number of all pixel points in the tongue image is obtained as the degree of characterization of the tongue tooth marks.

5. The tumor recurrence risk prediction system combined with traditional Chinese medicine constitution identification according to claim 1, characterized in that The method for obtaining the tongue quality characterization degree comprises: Obtaining the degree of tongue petechiae representation of the tongue image according to the color characteristics of the pixels in the tongue petechiae region and the position distribution of the pixels between the tongue petechiae region and the tongue mid-region; The intersecting pixels of the tongue petechiae area and the tongue crack area are removed from the tongue crack area to obtain the tongue crack update area; the edge curve in the tongue crack update area is obtained, and the edge fracture curve is obtained according to the direction characteristics of the endpoints between the edge curves; Obtain the perpendicular distance from the endpoints of each edge break curve to the edge curve, select a preset number of edge curves with the smallest perpendicular distance as the neighboring curves of each edge break curve, and use the corresponding endpoints and the pixel points on the neighboring curves as reference pixel points; according to the local direction similarity of the corresponding reference pixel points between the edge break curve and different neighboring curves, obtain the optimal merging curve direction of the edge break curve. Fit adjacent edge break curves with the same optimal merging curve direction to obtain a broken fitting curve and obtain a set of cracking curves; according to the tongue stasis dot characterization degree of the tongue image, the number of curves in the set of cracking curves, and the change trend of the curves, obtain the tongue texture characterization degree of the tongue image.

6. The tumor recurrence risk prediction system combined with traditional Chinese medicine constitution identification according to claim 5, characterized in that, The method for obtaining the tongue stasis dot characterization degree includes: Obtain the ratio of the number of intersecting pixel points between the tongue stasis dot area and the middle tongue area in the tongue image to the number of pixel points in the middle tongue area as the first stasis dot possibility. Obtain the binary image of the tongue stasis dot area, and obtain the ratio of the number of pixel points with a value of 1 to the number of pixel points with a value of 0 in the binary image as the second stasis dot possibility. Obtain the product between the first stasis dot possibility and the second stasis dot possibility as the tongue stasis dot characterization degree of the tongue image.

7. A tumor recurrence risk prediction system combined with traditional Chinese medicine constitution identification according to claim 5, characterized in that, The method for obtaining the edge break curve includes: Derive at each endpoint of the edge curve as the directionality of each endpoint. If there is a consistency in the directionality of the endpoints with the smallest relative distance between adjacent edge curves, the corresponding adjacent edge curves are used as edge break curves.

8. A tumor recurrence risk prediction system combined with traditional Chinese medicine constitution identification according to claim 5, characterized in that, The method for obtaining the optimal merging curve direction includes: Obtain the principal component direction within the neighborhood range of the corresponding reference pixel points between the edge break curve and each neighboring curve, and calculate the cosine similarity of the corresponding principal component direction as the local direction similarity of the corresponding reference pixel points between the edge break curve and each neighboring curve. Select the one with the largest local direction similarity between the edge break curve and different neighboring curves, and use the principal component direction of the corresponding neighboring curve as the optimal merging curve direction of the edge break curve.

9. The tumor recurrence risk prediction system combined with traditional Chinese medicine constitution identification according to claim 5, characterized in that The method for obtaining the tongue texture characterization degree includes: Obtain the average slope of different curves in the tongue cracking area as the overall slope level; obtain the cumulative value of the difference between the slopes of different curves in the tongue cracking area and the overall slope level as the cracking chaos degree. Obtain the product of the tongue stasis dot characterization degree, the number of curves, and the cracking chaos degree as the tongue texture characterization degree.

10. The tumor recurrence risk prediction system combined with traditional Chinese medicine constitution identification according to claim 1, characterized in that, The method for obtaining the tumor recurrence risk probability includes: According to the first relative distance of the corresponding sample points of the tumor tongue image between the period to be measured and different clinical periods, and the second relative distance between the sample points of the tumor tongue image in the period to be measured and the healthy tongue image, obtain the tumor recurrence risk probability in the period to be measured. The first relative distance is negatively correlated with the tumor recurrence risk probability, and the second relative distance is positively correlated with the tumor recurrence risk probability.

Images