Vagina wall plumpness evaluation system and evaluation method thereof

Abstract

The invention discloses a vaginal wall plumpness evaluation system and an evaluation method thereof. The system comprises a filtering module, a superpixel segmentation module, a superpixel feature calculation module, a similarity calculation module, a region segmentation module, a region merging module and an evaluation module. Fusing the minimum Euclidean distance of the RGB three channels of the similarity adjacent superpixels and chi-square test of the brightness histogram of the bright single channel; according to the part distribution rule of the image shot by the colposcope, the number of the areas is limited, the threshold value is set in a self-adaptive mode, the situation that the fixed threshold value setting is not suitable for various image data is avoided, and the accuracy of finally segmenting the image data into the N areas is improved; and rating the fullness by calculating the distance between the vaginal walls in the segmented region.

Description

Technical Field

[0001] The invention relates to the technical field of medical data processing. Background Art

[0002] In the prior art, there is no technical solution for the evaluation or data processing of vaginal wall fullness data that can be realized based on image data processing methods. In practice, it is mainly determined by doctors' subjective judgment based on experience, which cannot serve as an authoritative scientific basis.

[0003] Among the existing technologies for evaluating the state of the vaginal wall, there is a technical solution with publication number CN118430825A, which is a method for constructing a vaginal relaxation assessment model based on multi-source data. In this technical solution, a real-time active three-dimensional dynamic vaginal model is formed by collecting and processing electromyographic signals, and a vaginal relaxation assessment model is formed by tension adjustment degree data. There is no technical solution for performing image or data analysis on the fullness of the vaginal wall.

[0004] Therefore, it is urgent to seek a new technical solution to solve the above problems. Summary of the invention

[0005] Purpose of the invention: The object of the present invention is to provide a vaginal wall fullness assessment system that can improve the accuracy of vaginal wall image data region segmentation by image data processing means, and rate the fullness by calculating the distance between the vaginal walls in the segmented areas.

[0006] The present invention also provides a vaginal wall fullness evaluation method corresponding to the evaluation system.

[0007] Technical solution: The vaginal wall fullness assessment system provided by the present invention can adopt the following technical solution:

[0008] A vaginal wall fullness assessment system, comprising:

[0009] A filtering module is used to perform longitudinal mean filtering on the three channels of R, G, and B of the collected color vaginal wall image to retain the boundary area of ​​the fullness of the vaginal wall;

[0010] The superpixel segmentation module is used to segment the vaginal wall image into multiple superpixel blocks.

[0011] The superpixel feature calculation module is used to calculate the color feature and brightness feature of the superpixel block feature; the color feature is the mean of the RGB three channels in the superpixel block, and the brightness feature is the brightness histogram distribution in the superpixel block;

[0012] A similarity calculation module is used to calculate the similarity of two adjacent super-pixel blocks, including the calculation of color similarity and brightness similarity; color similarity is the Euclidean distance of the RGB three-channel mean values ​​of two adjacent super-pixel blocks, and brightness similarity is to convert the RGB color image into a brightness image; the similarity value between two adjacent super-pixel blocks is obtained by combining the color similarity and the brightness similarity; the similarity of every two adjacent super-pixel blocks in the entire vaginal wall image is obtained by this calculation method;

[0013] A region segmentation module is used to merge the similarities of adjacent superpixel blocks of the entire vaginal wall image by a minimum spanning tree merging method, so as to divide the entire vaginal wall image into a number of regions;

[0014] A region merging module is used to sort the regions by region size, take the N main regions with the largest number of regional pixels, merge the other regions into the main regions, and finally divide the colposcopic image into N regions to obtain an updated colposcopic image, where N is a positive integer greater than 1;

[0015] The evaluation module is used to evaluate the fullness of the vaginal wall; the central area judgment is performed on the updated colposcopy image; the central area width is the distance from the left vaginal wall to the right vaginal wall in the updated vaginal wall image. By calculating the pixel width of the central area, the distance between the vaginal walls is obtained as the basis for the fullness of the vaginal wall.

[0016] Corresponding to the above-mentioned vaginal wall fullness assessment system, the vaginal wall fullness assessment method provided by the present invention can adopt the following technical solutions:

[0017] A method for evaluating vaginal wall fullness comprises the following steps:

[0018] (1) performing longitudinal mean filtering on the R, G, and B channels of the collected color vaginal wall image to retain the boundary area of ​​the fullness of the vaginal wall;

[0019] (2) Segment the vaginal wall image into multiple superpixel blocks,

[0020] (3) Calculate the color feature and brightness feature of the superpixel block; the color feature is the mean of the RGB three channels in the superpixel block, and the brightness feature is the brightness histogram distribution in the superpixel block;

[0021] (4) Calculating the similarity between two adjacent super-pixel blocks, including the calculation of color similarity and brightness similarity; color similarity is the Euclidean distance between the RGB three-channel means of two adjacent super-pixel blocks, and brightness similarity is the conversion of the RGB color image into a brightness image; combining the color similarity and the brightness similarity to obtain the similarity value between the two adjacent super-pixel blocks; the similarity of every two adjacent super-pixel blocks in the entire vaginal wall image is obtained by this calculation method;

[0022] (5) merging the similarities of adjacent superpixel blocks of the entire vaginal wall image by a minimum spanning tree merging method to divide the entire vaginal wall image into several regions;

[0023] (6) sorting the regions by region size, taking N main regions with the largest number of pixels, merging the other regions into the main regions, and finally dividing the colposcopy image into N regions, where N is a positive integer greater than 1;

[0024] (7) Assessing the fullness of the vaginal wall; determining the central area of ​​the updated colposcopy image; the width of the central area is the distance from the left vaginal wall to the right vaginal wall in the updated vaginal wall image. By calculating the pixel width of the central area, the distance between the vaginal walls is obtained as the basis for the fullness of the vaginal wall.

[0025] Beneficial effects: In the present invention, by fusing the minimum Euclidean distance of the RGB three channels of similarity adjacent superpixels and the chi-square test of the brightness histogram of the bright single channel, the vaginal wall segmentation is not only affected by color changes, but also by brightness, and the segmentation effect is better; according to the distribution law of the parts of the image taken by the colposcopy, the method limits the number of regions and sets the threshold adaptively to avoid the fixed threshold setting being unsuitable for diverse image data, so as to improve the accuracy of the final segmentation of the image data into N regions; the fullness is rated by calculating the distance between the vaginal walls in the segmented areas, reducing the influence of human factors, so that the evaluation results obtained through data analysis are more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 The figure is a flow chart of the evaluation method in the present invention. DETAILED DESCRIPTION

[0027] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0028] Please combine Figure 1 As shown, this embodiment proposes a filtering module for performing longitudinal mean filtering on the three channels R, G, and B of the collected color vaginal wall image to retain the boundary area of ​​the fullness of the vaginal wall.

[0029] The superpixel segmentation module is used to segment the vaginal wall image into multiple superpixel blocks. In this functional module, segmenting multiple pixel blocks includes:

[0030] Evenly distribute the seed points in the image, calculate the gradient values ​​of all pixels in the neighborhood of the seed point (for example, the 3*3 neighborhood of the seed point), and move the seed point to the position with the minimum gradient; the number of seed points designed by this method is 5000;

[0031] The color colposcopy image is converted into a Lab image, and the XY coordinates are included, so that a total of 5-dimensional feature vectors are used to construct the distance metric; the distance metric between each pixel point and the seed point is searched and calculated respectively;

[0032] The color distance in Lab color space is calculated as follows:

[0033]

[0034] Among them, L j 、a j 、b j is the value of the three Lab channels of the pixel point, L i 、a i 、b i is the value of the three Lab channels of the cluster center; d c is the color distance between the pixel and the cluster center;

[0035] The spatial distance between a pixel and a cluster center is calculated as follows:

[0036]

[0037] Among them, X j , Y j is the coordinate of the pixel point, X i , Y i are the coordinates of the cluster center, d s is the spatial distance between the pixel and the cluster center.

[0038] The distance calculation metric combining spatial distance and color distance is as follows:

[0039]

[0040] Where m is d s The weight of , D is the distance metric;

[0041] Take the minimum value of D, and the corresponding seed point is the cluster center of the pixel;

[0042] Update the cluster center, average the features of all points, represent the features of the center point of the region, and calculate the position of the center point of the region.

[0043] Repeat the above steps until the cluster center no longer changes and multiple superpixel blocks are formed.

[0044] The superpixel feature calculation module is used to calculate the color features and brightness features of the superpixel block features; the color feature is the mean of the RGB three channels in the superpixel block, and the brightness feature is the brightness histogram distribution in the superpixel block.

[0045] The similarity calculation module is used to calculate the similarity between two adjacent super-pixel blocks, including the calculation of color similarity and brightness similarity; color similarity is the Euclidean distance between the RGB three-channel means of two adjacent super-pixel blocks, and brightness similarity is to convert the RGB color image into a brightness image; the similarity value between two adjacent super-pixel blocks is obtained by combining color similarity and brightness similarity; the similarity between every two adjacent super-pixel blocks in the entire vaginal wall image is obtained by this calculation method. In this functional module, the similarity calculation includes color similarity and brightness similarity;

[0046] Color similarity is calculated as follows:

[0047]

[0048] Among them, RGBSimilarity represents color similarity, α represents the weight of color similarity, and PatchAvg1 r PatchAvg2 represents the mean value of the red channel in superpixel block 1. r PatchAvg1 represents the mean value of the red channel in superpixel block 2. g PatchAvg2 represents the mean value of the green channel in superpixel block 1. g PatchAvg1 shows the mean value of the green channel in superpixel block 2. b PatchAvg2 represents the mean of the blue channel in superpixel block 1. g shows the mean value of the blue channel in superpixel block 2.

[0049] The conversion of RGB color image into brightness image is calculated as follows:

[0050] Y=0.299*R+0.587*G+0.114*B

[0051] Where R represents the value of the R channel of the image pixel, G represents the value of the G channel of the image pixel, and B represents the value of the B channel of the image pixel;

[0052] The brightness similarity is calculated as follows:

[0053]

[0054] Among them, BrightSimilarity represents the brightness similarity, the chi-square metric of the brightness histogram of two superpixel blocks; PatchHistList1 represents the brightness histogram in superpixel block 1, PatchHistList1 i PatchHistList2 represents the number of brightness values ​​i in super pixel block 1, where i is 0-255; PatchHistList2 represents the brightness histogram in super pixel block 2, irepresents the number of brightness values ​​i in superpixel block 2, where i is 0-255;

[0055] The calculation method combining color similarity and brightness similarity is as follows:

[0056] Similarity=α*RGBSimilarity+(1-α)*BrightSimilarity

[0057] Among them, α is the weight of color similarity; Similarity is the similarity value between two super pixel blocks.

[0058] The region segmentation module is used to merge the similarities of adjacent superpixel blocks of the entire vaginal wall image by merging the minimum spanning tree, so as to divide the entire vaginal wall image into several regions. In this functional module, region merging specifically includes:

[0059] The similarities of adjacent superpixel blocks of the entire image are sorted from small to large to form a similarity priority list; the two superpixel blocks with the lowest similarity values ​​are taken out from the similarity priority list, and when the similarity of the two superpixel blocks is lower than a preset threshold, the two superpixel blocks are regionally merged, and then the similarity values ​​of the superpixel blocks and the adjacent superpixel blocks are updated, and the similarity list is reconstructed; the two superpixel blocks with the lowest similarity values ​​are taken out from the priority queue of the reconstructed similarity list for regional merging, and the regional merging steps are repeated until the superpixel blocks cannot be merged, and the merging of the minimum spanning tree is terminated.

[0060] Since there are many final segmented areas in the area segmentation module, and the vaginal wall image of the colposcope is segmented into five areas at most, including the left vaginal wall, the right vaginal wall, the upper edge of the upper dilator, the lower edge of the lower dilator, and the middle cervical opening, it is necessary to merge the above segmentation results into small areas. Therefore, a regional merging module is used to sort the area sizes of several areas, take the N main areas (for example, five main areas) with the largest number of regional pixels, and merge other areas into the main areas, and finally divide the image of the colposcope into N areas, where N is a positive integer greater than 1. The area size is set by adaptive threshold through regional merging, such as a preferred method of merging into five main areas, the threshold is the number of pixels of the fifth largest area segmented, and the area less than the threshold is merged into the largest area that is most similar to it, which solves the problem of difficulty in setting a fixed threshold due to differences in the quality of the vaginal wall image. The colposcope image after merging other areas into the main area is used as the updated colposcope image.

[0061] The evaluation module is used to evaluate the fullness of the vaginal wall. Since the default center point of the colposcopy image is aligned between the left and right vaginal walls, the width of the central area of ​​the captured image is the distance from the left vaginal wall to the right vaginal wall. A full vaginal wall image may cover the cervix, so there is no cervical opening area. The center area only has the left and right vaginal wall areas. The left and right vaginal wall areas are close together. At this time, the vaginal wall distance is set to 0, and the fullness is good. The left vaginal wall area, the middle cervical area, and the right vaginal wall area exist near the center of the vaginal wall image with average fullness. By calculating the pixel width of the central cervical area in the updated colposcopy image, the distance between the vaginal walls is evaluated to reflect its fullness. The pixel width of the central cervical area is calculated as follows:

[0062]

[0063] In this paper, we take the ImgH*0.2 lines above and below the center of the image to calculate the pixel width of the central cervical area, totaling ImgH*0.2*2+1 lines, where ImgH represents the pixel height of the colposcopy image; Width i Represents the pixel width of the central cervical area of ​​the i-th row, MidLine represents the row number where the center point of the image is located, and then the average of the pixel widths of the cervical area of ​​the ImgH*0.2*2+1 row is taken. N represents the magnification of the colposcope when taking the image. Divide by N to unify the image magnification factor to one size to obtain the final pixel width of the cervical area Width Avg In this embodiment, it is preferably set to set the pixel width Width Avg If the value is lower than 200, the fullness is good; if the value is higher than 200, the fullness is average; in the vaginal wall image with poor fullness, there is only one area near the center with poor fullness.

[0064] Embodiment 2

[0065] Corresponding to the evaluation system of the above embodiment, this embodiment provides an evaluation method. Figure 1 shown.

[0066] The following steps are involved:

[0067] (1) The three channels R, G, and B of the collected color vaginal wall image are all subjected to longitudinal mean filtering to retain the boundary area of ​​the fullness of the vaginal wall.

[0068] (2) Segmenting the vaginal wall image into multiple super pixel blocks. Segmenting multiple pixel blocks includes:

[0069] Evenly distribute the seed points in the image, calculate the gradient values ​​of all pixels in the neighborhood of the seed points, and move the seed points to the position with the minimum gradient;

[0070] The color colposcopy image is converted into a Lab image, and the XY coordinates are included, so that a total of 5-dimensional feature vectors are used to construct the distance metric; the distance metric between each pixel point and the seed point is searched and calculated respectively;

[0071] The color distance in Lab color space is calculated as follows:

[0072]

[0073] Among them, L j 、a j 、b j is the value of the three Lab channels of the pixel point, L i 、a i 、b i is the value of the three Lab channels of the cluster center; d c is the color distance between the pixel and the cluster center;

[0074] The spatial distance between a pixel and a cluster center is calculated as follows:

[0075]

[0076] Among them, X j , Y j is the coordinate of the pixel point, X i , Y i are the coordinates of the cluster center, d s is the spatial distance between the pixel and the cluster center.

[0077] The distance calculation metric combining spatial distance and color distance is as follows:

[0078]

[0079] Where m is d s The weight of , D is the distance metric;

[0080] Take the minimum value of D, and the corresponding seed point is the cluster center of the pixel;

[0081] Update the cluster center, average the features of all points, and represent the features of the center point of the region, and calculate the position of the center point of the region;

[0082] Repeat the above steps until the cluster center no longer changes and multiple superpixel blocks are formed.

[0083] (3) Calculate the color features and brightness features of the superpixel block; the color feature is the mean of the RGB three channels in the superpixel block, and the brightness feature is the brightness histogram distribution in the superpixel block.

[0084] (4) Calculating the similarity of two adjacent super-pixel blocks, including the calculation of color similarity and brightness similarity; color similarity is the Euclidean distance between the RGB three-channel means of two adjacent super-pixel blocks, and brightness similarity is the conversion of the RGB color image into a brightness image; combining the color similarity and the brightness similarity to obtain the similarity value between the two adjacent super-pixel blocks; through this calculation method, the similarity of every two adjacent super-pixel blocks in the entire vaginal wall image is obtained; the similarity calculation includes color similarity and brightness similarity;

[0085] Color similarity is calculated as follows:

[0086]

[0087] Among them, RGBSimilarity represents color similarity, α represents the weight of color similarity, and PatchAvg1 r PatchAvg2 represents the mean value of the red channel in superpixel block 1. r PatchAvg1 represents the mean value of the red channel in superpixel block 2. g PatchAvg2 represents the mean value of the green channel in superpixel block 1. g PatchAvg1 shows the mean value of the green channel in superpixel block 2. b PatchAvg2 represents the mean of the blue channel in superpixel block 1. g shows the mean value of the blue channel in superpixel block 2;

[0088] The conversion of RGB color image into brightness image is calculated as follows:

[0089] Y=0.299*R+0.587*G+0.114*B

[0090] Where R represents the value of the R channel of the image pixel, G represents the value of the G channel of the image pixel, and B represents the value of the B channel of the image pixel;

[0091] The brightness similarity is calculated as follows:

[0092]

[0093] Among them, BrightSimilarity represents the brightness similarity, the chi-square metric of the brightness histogram of two superpixel blocks; PatchHistList1 represents the brightness histogram in superpixel block 1, PatchHistList1 i PatchHistList2 represents the number of brightness values ​​i in super pixel block 1, where i is 0-255; PatchHistList2 represents the brightness histogram in super pixel block 2, irepresents the number of brightness values ​​i in superpixel block 2, where i is 0-255;

[0094] The calculation method combining color similarity and brightness similarity is as follows:

[0095] Similarity=α*RGBSimilarity+(1-α)*BrightSimilarity

[0096] Among them, α is the weight of color similarity; Similarity is the similarity value between two super pixel blocks.

[0097] (5) The similarities of adjacent superpixel blocks of the entire vaginal wall image are merged by using a minimum spanning tree to divide the entire vaginal wall image into several regions.

[0098] (6) Sort the regions by size, select N main regions with the largest number of pixels, merge the other regions into the main regions, and finally divide the colposcopy image into N regions, where N is a positive integer greater than 1.

[0099] (7) Evaluate the fullness of the vaginal wall; since the default center point of the colposcopy image is aligned between the left and right vaginal walls, the width of the central area of ​​the captured image is the distance from the left vaginal wall to the right vaginal wall; a full vaginal wall image may cover the cervix, so there is no cervical opening area, and the center area only has the left and right vaginal wall areas, and the left and right vaginal wall areas are close together. At this time, the vaginal wall distance is set to 0, and the fullness is good; a vaginal wall image with average fullness has a left vaginal wall area, a middle cervical area, and a right vaginal wall area near the center. By calculating the pixel width of the central cervical area in the updated colposcopy image, the distance between the vaginal walls is evaluated to reflect its fullness. The pixel width of the central cervical area is calculated as follows:

[0100]

[0101] In this paper, we take the ImgH*0.2 lines above and below the center of the image to calculate the pixel width of the central cervical area, totaling ImgH*0.2*2+1 lines, where ImgH represents the pixel height of the colposcopy image; Width i Represents the pixel width of the central cervical area of ​​the i-th row, MidLine represents the row number where the center point of the image is located, and then the average of the pixel widths of the cervical area of ​​the ImgH*0.2*2+1 row is taken. N represents the magnification of the colposcope when taking the image. Divide by N to unify the image magnification factor to one size to obtain the final pixel width of the cervical area Width Avg; In this paper, when the pixel width is less than 200, the fullness is set to be good, and when it is greater than 200, the fullness is set to be average; for vaginal wall images with poor fullness, there is only one area near the center with poor fullness.

[0102] In addition, there are many specific implementation methods and approaches of the present invention, and the above is only a preferred implementation of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be regarded as the protection scope of the present invention.

Claims

1. A vaginal wall fullness assessment system, characterized in that: include: A filtering module is used to perform longitudinal mean filtering on the three channels of R, G, and B of the collected color vaginal wall image to retain the boundary area of ​​the fullness of the vaginal wall; The superpixel segmentation module is used to segment the vaginal wall image into multiple superpixel blocks. The superpixel feature calculation module is used to calculate the color feature and brightness feature of the superpixel block feature; the color feature is the mean of the RGB three channels in the superpixel block, and the brightness feature is the brightness histogram distribution in the superpixel block; A similarity calculation module is used to calculate the similarity of two adjacent super-pixel blocks, including the calculation of color similarity and brightness similarity; color similarity is the Euclidean distance of the RGB three-channel mean values ​​of two adjacent super-pixel blocks, and brightness similarity is to convert the RGB color image into a brightness image; the similarity value between two adjacent super-pixel blocks is obtained by combining the color similarity and the brightness similarity; the similarity of every two adjacent super-pixel blocks in the entire vaginal wall image is obtained by this calculation method; A region segmentation module is used to merge the similarities of adjacent superpixel blocks of the entire vaginal wall image by a minimum spanning tree merging method, so as to divide the entire vaginal wall image into a number of regions; A region merging module is used to sort the regions by region size, take the N main regions with the largest number of regional pixels, merge the other regions into the main regions, and finally divide the colposcopic image into N regions to obtain an updated colposcopic image, where N is a positive integer greater than 1; The evaluation module is used to evaluate the fullness of the vaginal wall; the central area judgment is performed on the updated colposcopy image; the central area width is the distance from the left vaginal wall to the right vaginal wall in the updated vaginal wall image. By calculating the pixel width of the central area, the distance between the vaginal walls is obtained as the basis for the fullness of the vaginal wall.

2. The vaginal wall fullness assessment system according to claim 1, characterized in that: In the superpixel segmentation module, segmenting multiple pixel blocks includes: Evenly distribute the seed points in the image, calculate the gradient values ​​of all pixels in the neighborhood of the seed points, and move the seed points to the position with the minimum gradient; The color colposcopy image is converted into a Lab image, and the XY coordinates are included, so that a total of 5-dimensional feature vectors are used to construct the distance metric; the distance metric between each pixel point and the seed point is searched and calculated respectively; The color distance in Lab color space is calculated as follows: Among them, L j 、a j 、b j is the value of the three Lab channels of the pixel point, L i 、a i 、b i is the value of the three Lab channels of the cluster center; d c is the color distance between the pixel and the cluster center; The spatial distance between a pixel and a cluster center is calculated as follows: Among them, X j , Y j is the coordinate of the pixel point, X i , Y i are the coordinates of the cluster center, d s is the spatial distance between the pixel and the cluster center; The distance calculation metric combining spatial distance and color distance is as follows: Where m is d s The weight of , D is the distance metric; Take the minimum value of D, and the corresponding seed point is the cluster center of the pixel; Update the cluster center, average the features of all points, and represent the features of the center point of the region, and calculate the position of the center point of the region; Repeat the above steps until the cluster center no longer changes and multiple superpixel blocks are formed.

3. The vaginal wall fullness assessment system according to claim 1, characterized in that: In the similarity calculation module, the similarity calculation includes color similarity and brightness similarity; Color similarity is calculated as follows: Among them, RGBSimilarity represents color similarity, α represents the weight of color similarity, and PatchAvg1 r PatchAvg2 represents the mean value of the red channel in superpixel block 1. r PatchAvg1 represents the mean value of the red channel in superpixel block 2. g PatchAvg2 represents the mean value of the green channel in superpixel block 1. g PatchAvg1 shows the mean value of the green channel in superpixel block 2. b PatchAvg2 represents the mean of the blue channel in superpixel block 1. g shows the mean value of the blue channel in superpixel block 2; The conversion of RGB color image into brightness image is calculated as follows: Y=0.299*R+0.587*G+0.114*B Where R represents the value of the R channel of the image pixel, G represents the value of the G channel of the image pixel, and B represents the value of the B channel of the image pixel; The brightness similarity is calculated as follows: Among them, BrightSimilarity represents the brightness similarity, the chi-square metric of the brightness histogram of two superpixel blocks; PatchHistList1 represents the brightness histogram in superpixel block 1, PatchHistList1 i PatchHistList2 represents the number of brightness values ​​i in super pixel block 1, where i is 0-255; PatchHistList2 represents the brightness histogram in super pixel block 2, i represents the number of brightness values ​​i in superpixel block 2, where i is 0-255; The calculation method combining color similarity and brightness similarity is as follows: Similarity=α*RGBSimilarity+(1-α)*BrightSimilarity Among them, α is the weight of color similarity; Similarity is the similarity value between two super pixel blocks.

4. The vaginal wall fullness assessment system according to claim 1 or 3, characterized in that: In the region segmentation module, region merging specifically includes: The similarities of adjacent superpixel blocks of the entire image are sorted from small to large to form a similarity priority list; the two superpixel blocks with the lowest similarity values ​​are taken out from the similarity priority list, and when the similarity of the two superpixel blocks is lower than a preset threshold, the two superpixel blocks are regionally merged, and then the similarity values ​​of the superpixel blocks and the adjacent superpixel blocks are updated, and the similarity list is reconstructed; the two superpixel blocks with the lowest similarity values ​​are taken out from the priority queue of the reconstructed similarity list for regional merging, and the regional merging steps are repeated until the superpixel blocks cannot be merged, and the merging of the minimum spanning tree is terminated.

5. A method for evaluating vaginal wall fullness, characterized in that: The following steps are involved: (1) performing longitudinal mean filtering on the R, G, and B channels of the collected color vaginal wall image to retain the boundary area of ​​the fullness of the vaginal wall; (2) Segment the vaginal wall image into multiple superpixel blocks, (3) Calculate the color feature and brightness feature of the superpixel block; the color feature is the mean of the RGB three channels in the superpixel block, and the brightness feature is the brightness histogram distribution in the superpixel block; (4) Calculating the similarity between two adjacent super-pixel blocks, including the calculation of color similarity and brightness similarity; color similarity is the Euclidean distance between the RGB three-channel means of two adjacent super-pixel blocks, and brightness similarity is the conversion of the RGB color image into a brightness image; combining the color similarity and the brightness similarity to obtain the similarity value between the two adjacent super-pixel blocks; the similarity of every two adjacent super-pixel blocks in the entire vaginal wall image is obtained by this calculation method; (5) merging the similarities of adjacent superpixel blocks of the entire vaginal wall image by a minimum spanning tree merging method to divide the entire vaginal wall image into several regions; (6) sorting the regions by region size, taking N main regions with the largest number of pixels, merging the other regions into the main regions, and finally dividing the colposcopy image into N regions, where N is a positive integer greater than 1; (7) Assessing the fullness of the vaginal wall; determining the central area of ​​the updated colposcopy image; the width of the central area is the distance from the left vaginal wall to the right vaginal wall in the updated vaginal wall image. By calculating the pixel width of the central area, the distance between the vaginal walls is obtained as the basis for the fullness of the vaginal wall.

6. The method for evaluating vaginal wall fullness according to claim 5, characterized in that: In step (2), segmenting the plurality of pixel blocks comprises: Evenly distribute the seed points in the image, calculate the gradient values ​​of all pixels in the neighborhood of the seed points, and move the seed points to the position with the minimum gradient; The color colposcopy image is converted into a Lab image, and the XY coordinates are included, so that a total of 5-dimensional feature vectors are used to construct the distance metric; the distance metric between each pixel point and the seed point is searched and calculated respectively; The color distance in Lab color space is calculated as follows: Among them, L j 、a j 、b j is the value of the three Lab channels of the pixel point, L i 、a i 、b i is the value of the three Lab channels of the cluster center; d c is the color distance between the pixel and the cluster center; The spatial distance between a pixel and a cluster center is calculated as follows: Among them, X j , Y j is the coordinate of the pixel point, X i , Y i are the coordinates of the cluster center, d s is the spatial distance between the pixel and the cluster center; The distance calculation metric combining spatial distance and color distance is as follows: Where m is d s The weight of , D is the distance metric; Take the minimum value of D, and the corresponding seed point is the cluster center of the pixel; Update the cluster center, average the features of all points, and represent the features of the center point of the region, and calculate the position of the center point of the region; Repeat the above steps until the cluster center no longer changes and multiple superpixel blocks are formed.

7. The method for evaluating vaginal wall fullness according to claim 5, characterized in that: In step (4), the similarity calculation includes color similarity and brightness similarity; Color similarity is calculated as follows: Among them, RGBSimilarity represents color similarity, α represents the weight of color similarity, and PatchAvg1 r PatchAvg2 represents the mean value of the red channel in superpixel block 1. r PatchAvg1 represents the mean value of the red channel in superpixel block 2. g PatchAvg2 represents the mean value of the green channel in superpixel block 1. g PatchAvg1 shows the mean value of the green channel in superpixel block 2. b PatchAvg2 represents the mean of the blue channel in superpixel block 1. g shows the mean value of the blue channel in superpixel block 2; The conversion of RGB color image into brightness image is calculated as follows: Y=0.299*R+0.587*G+0.114*B Where R represents the value of the R channel of the image pixel, G represents the value of the G channel of the image pixel, and B represents the value of the B channel of the image pixel; The brightness similarity is calculated as follows: Among them, BrightSimilarity represents the brightness similarity, the chi-square metric of the brightness histogram of two superpixel blocks; PatchHistList1 represents the brightness histogram in superpixel block 1, PatchHistList1 i PatchHistList2 represents the number of brightness values ​​i in super pixel block 1, where i is 0-255; PatchHistList2 represents the brightness histogram in super pixel block 2, i represents the number of brightness values ​​i in superpixel block 2, where i is 0-255; The calculation method combining color similarity and brightness similarity is as follows: Similarity=α*RGBSimilarity+(1-α)*BrightSimilarity Among them, α is the weight of color similarity; Similarity is the similarity value between two super pixel blocks.

8. An electronic device 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 steps of the method according to any one of claims 5 to 7 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 5 to 7 are implemented.

Images