Adhesion focus positioning method of endometriosis rat model

Abstract

The invention relates to the technical field of image recognition, in particular to an adhesion focus positioning method for an endometriosis rat model, which comprises the following steps of: firstly, adjusting a clustering distance according to color difference between saturation connected domains and a position inertial system, and clustering to obtain a connected domain cluster; due to the fact that the connected domain clusters contain the ectopic glands, the clusters need to be screened to obtain the ectopic glands, the gland probability of each connected domain cluster is obtained according to the characteristic that columnar epithelial cells are distributed on the inner walls of the ectopic glands, and the possibility that each connected domain cluster belongs to the ectopic glands is judged; due to the fact that the ectopic glands are distributed in a concentrated mode, the abnormal probability of the connected domain clusters is obtained through the aggregation of the connected domain clusters with the high gland probability, the possibility that the connected domain clusters belong to the abnormal ectopic glands is further judged, and the focus positioning accuracy of the endometriosis rat model is higher.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition, and particularly to a method for locating adhesion lesions in a rat model of endometriosis. Background Art

[0002] A rat model of endometriosis is an experimental animal model established by simulating the pathological characteristics of human endometriosis, mainly used for studying the pathogenesis and treatment methods of this disease. Common construction methods include autologous transplantation and allogeneic transplantation, that is, implanting the endometrial tissue of healthy rats into their abdominal cavity or other ectopic sites to form a pathological structure similar to human endometriotic lesions. This model can exhibit typical characteristics of endometriosis, such as the formation of ectopic lesions, inflammatory responses, pain behaviors, and reproductive function disorders, providing a reliable platform for in-depth understanding of the disease mechanism and evaluating the efficacy of new drugs.

[0003] Locating the lesions in a rat model of endometriosis can more intuitively evaluate the distribution, size, and quantity of the lesions, thereby better conducting relevant teaching on the pathological mechanism and progression of the disease. In addition, accurately locating the lesions is also of great significance for carrying out histological analysis, molecular biology detection, and evaluating the sensitivity and specificity of new diagnostic tools, providing key data support for the basic research and clinical transformation of endometriosis.

[0004] The hematoxylin-eosin (HE) staining method is widely used in section staining. Existing technologies usually observe HE-stained sections based on an optical microscope and identify and locate them according to the typical histological characteristics of ectopic lesions. However, since the lesions in a rat model of endometriosis coexist with surrounding tissues in the stained image, the lesion area is not obvious, making it difficult to distinguish some lesion areas from normal tissues, and thus resulting in inaccurate lesion location in the rat model of endometriosis. Summary of the Invention

[0005] In order to solve the technical problem that the lesions in a rat model of endometriosis coexist with surrounding tissues in the stained image, resulting in an unclear lesion area, making it difficult to distinguish some lesion areas from normal tissues, and thus leading to inaccurate lesion location in the rat model of endometriosis, the purpose of this application is to provide a method for locating adhesion lesions in a rat model of endometriosis, and the specific technical solutions adopted are as follows:

[0006] This application proposes a method for locating adhesion lesions in a rat model of endometriosis, and the method includes:

[0007] Collect stained section images of a rat model of endometriosis; determine all saturation connected regions in each saturation interval according to the saturation distribution of the stained section images; obtain all gray connected regions after graying the stained section images;

[0008] Determine the boundary possibility of each saturation connected region according to the distance between saturation connected regions and the saturation deviation of the corresponding saturation interval; screen out at least two connected region clustering clusters according to the boundary possibility and the positional distribution and relative saturation deviation between each saturation connected region;

[0009] Perform region synthesis according to the distribution adjacency of each gray connected region and the approximate columnar feature in morphology to determine all suspected columnar epithelial regions; determine the corresponding epithelial probability based on the number of gray connected regions and columnar approximation in the suspected columnar epithelial regions;

[0010] Determine the corresponding ectopic gland probability according to the boundary possibility distribution in each connected region clustering cluster and the epithelial probability of adjacent suspected columnar epithelial regions; determine the abnormality probability of each connected region clustering cluster based on the ectopic gland probability and the concentrated distribution of the connected region clustering cluster;

[0011] Screen out abnormal connected region clustering clusters according to the abnormality probability; locate the adhesion lesions of the rat model of endometriosis according to the abnormal connected region clustering clusters.

[0012] Further, the obtaining process of the saturation connected region includes:

[0013] Convert the stained section image into an HSV image; obtain all minimum values in the distribution histogram of the S channel corresponding to the HSV image; divide the distribution histogram of the S channel into at least two saturation intervals at intervals of all minimum values; use all connected regions composed of pixel points in each saturation interval as saturation connected regions.

[0014] Further, the obtaining process of the boundary possibility includes:

[0015] Successively use each saturation connected region as the target connected region; use the saturation interval where the target connected region is located as the target interval; use other saturation intervals outside the target interval as reference intervals;

[0016] Perform negative correlation mapping on the average value of the distances between the centroids of all saturation connected regions in each reference interval and the centroid of the target connected region to determine the boundary contribution index of each reference interval to the target connected region;

[0017] The difference between the maximum value of the S-channel value in each reference interval and the maximum value of the S-channel value in the target interval is used as the saturation interval deviation corresponding to each reference interval.

[0018] According to the product of the boundary contribution index and the saturation interval deviation, determine the boundary influence parameter of each reference interval on the target connected domain; normalize the accumulated value of the boundary influence parameters of all reference intervals corresponding to the target connected domain to determine the boundary possibility of the target connected domain.

[0019] Furthermore, the process of obtaining the connected domain clustering clusters includes:

[0020] According to the extension situation of the minimum value of the boundary possibility in the adjacent area of the target connected domain, determine the central connected domain of the target connected domain; use the other saturation connected domains outside the target connected domain as the comparison connected domains.

[0021] According to the difference between the mean value of the S-channel values of all pixel points in each comparison connected domain and the mean value of the S-channel values of all pixel points in the target connected domain, determine the saturation deviation value of each comparison connected domain.

[0022] The distance between the centroid of the target connected domain and the centroid of each comparison connected domain is used as the first comparison distance of each comparison connected domain; the distance between the centroid of the central connected domain and the centroid of each comparison connected domain is used as the second comparison distance of each comparison connected domain.

[0023] Calculate the difference between the second comparison distance and the first comparison distance, and use the product of the difference and the boundary possibility as the boundary weighting coefficient; calculate the sum value of the boundary weighting coefficient and the first comparison distance; use the product of the sum value and the saturation deviation value as the clustering distance between the target connected domain and each comparison connected domain.

[0024] Statistical clustering distances between every two saturation connected domains; perform clustering analysis on all saturation connected domains through the DBSCAN clustering algorithm according to all clustering distances to obtain at least two connected domain clustering clusters.

[0025] Furthermore, the process of obtaining the central connected domain includes:

[0026] Take the reference connected domain with the minimum boundary possibility in each saturation connected domain and all its adjacent saturation connected domains as the corresponding minimum value connected domain.

[0027] Taking the target connected domain as the seed point, using the acquisition of the minimum value connected domain as the growth condition and taking the minimum value connected domain as the new seed point for region growth. After the obtained growth region no longer changes, take the last obtained minimum value connected domain in the region growth process as the central connected domain of the target connected domain.

[0028] Further, the process of obtaining the suspected columnar epithelial region includes:

[0029] Performing a negative correlation mapping on the ratio between the width and the length of the minimum circumscribed rectangle of each gray-scale connected domain to determine the columnar degree of each gray-scale connected domain; taking the gray-scale connected domains with a columnar degree greater than a preset columnar threshold as columnar connected domains;

[0030] Merging all adjacent columnar connected domains to obtain all suspected columnar epithelial regions.

[0031] Further, the process of obtaining the epithelial probability includes:

[0032] Normalizing the product of the mean of the columnar degrees of all columnar connected domains in the suspected columnar epithelial region and the total number of columnar connected domains to determine the corresponding epithelial probability.

[0033] Further, the process of obtaining the probability of ectopic glands includes:

[0034] Taking the epithelial probability of the suspected columnar epithelial region closest to each saturation connected domain as the adjacent reference probability of each saturation connected domain;

[0035] Taking the product of the normalized value of the boundary possibility of each saturation connected domain and the adjacent reference probability as the local reference probability;

[0036] Determining the corresponding probability of ectopic glands according to the accumulated value of the local reference probabilities of all saturation connected domains in each connected domain clustering cluster.

[0037] Further, the process of obtaining the abnormal probability includes:

[0038] Successively taking each connected domain clustering cluster as the target clustering cluster; taking the other connected domain clustering clusters outside the target clustering cluster as the comparison clustering clusters;

[0039] Taking the minimum distance between all pixel points in each comparison clustering cluster and all pixel points in the target clustering cluster as the inter-cluster distance of each comparison clustering cluster;

[0040] Taking the product of the negative correlation normalized value of the inter-cluster distance and the probability of ectopic glands of the corresponding comparison clustering cluster as the reference concentration degree of each comparison clustering cluster;

[0041] Taking the product of the mean of the reference concentration degrees of all comparison clustering clusters and the probability of ectopic glands of the target clustering cluster as the abnormal probability of the target clustering cluster.

[0042] Further, the process of obtaining the abnormal connected domain clustering cluster includes:

[0043] Cluster the connected domain clusters with abnormal probabilities greater than the preset abnormal threshold as abnormal connected domain clusters.

[0044] The present application has the following beneficial effects:

[0045] During the process of locating the endometriosis rat model, since the gland is the most specific landmark structure in the lesion, the present invention obtains the position of the gland structure. Due to the uneven staining of the ectopic gland during HE staining, a single gland structure cannot be completely segmented. The present invention first obtains a number of saturation connected domains through the S channel, adjusts the clustering distance according to the color difference and position inertial system between the saturation connected domains, and clusters to obtain connected domain clusters; since the connected domain clusters contain ectopic glands, it is necessary to screen the clusters to obtain ectopic glands. By the characteristic that columnar epithelial cells are distributed on the inner wall of the ectopic gland, obtain the gland probability of each connected domain cluster, and judge the possibility of each connected domain cluster belonging to an ectopic gland; since the ectopic glands show a concentrated distribution pattern, through the aggregation of the connected domain clusters with higher gland probabilities, obtain the abnormal probability of the connected domain clusters, and further judge the possibility of the connected domain clusters belonging to abnormal ectopic glands. So far, the present invention marks the possible lesion positions in the stained section image through the abnormal probability of the connected domain clusters, thereby improving the accuracy of the lesion location of the endometriosis rat model. Description of the Drawings

[0046] 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 described 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 based on these drawings.

[0047] Figure 1 It is a flowchart of a method for locating adhesion lesions of an endometriosis rat model provided by an embodiment of the present invention;

[0048] Figure 2 It is a schematic diagram of the morphological structure of columnar epithelial cells provided by an embodiment of the present invention. Detailed Embodiments

[0049] To further elaborate on the technical means and effects adopted by the present invention to achieve the intended invention purpose, the following specifically describes, in conjunction with the accompanying drawings and preferred embodiments, a method for locating adhesion lesions in a rat model of endometriosis according to the present invention, including its specific implementation manner, structure, characteristics, and effects. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment, and specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form. In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be construed as implying or suggesting relative importance or implicitly indicating the quantity of the indicated technical features. Thus, features defined with "first" and "second" may explicitly or implicitly include one or more of such features.

[0050] 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 this invention belongs.

[0051] The following specifically describes, in conjunction with the accompanying drawings, the specific solution of a method for locating adhesion lesions in a rat model of endometriosis provided by the present invention.

[0052] The embodiments of the present application provide a method for locating adhesion lesions in a rat model of endometriosis. Please refer to Figure 1 , which shows a flowchart of a method for locating adhesion lesions in a rat model of endometriosis provided by an embodiment of the present invention. The method includes:

[0053] Step S101: Collect stained section images of a rat model of endometriosis; determine all saturation-connected regions in each saturation interval according to the saturation distribution of the stained section images; obtain all grayscale-connected regions after graying the stained section images.

[0054] In a specific implementation manner of the embodiments of the present invention, first, a rat model of endometriosis is made. Specifically, the cut endometrial tissue is attached to the rich vascular area of the subcutaneous fascia, and the rats are normally raised after the operation until it is confirmed that the rat model is established, thus obtaining a rat model of endometriosis. Then, HE stained sections are made, and the sections are observed and the stained section images of the rat model of endometriosis are collected by using a microscope camera.

[0055] Endometriosis is caused by the shedding of endometrial tissue under certain pathological conditions and its migration to extrauterine sites such as the abdominal cavity and ovaries through menstrual reflux, lymphatic or blood circulation. As a result, endometrial tissue grows in locations outside the uterine cavity and myometrium. Therefore, the lesions of endometriosis are similar in structure to normal endometrial tissue, and endometrial tissue contains glandular structures and stromal tissue. When localizing the lesions in a rat model of endometriosis, the glandular structure should be localized because glands are the most specific marker structures in the lesions and can directly reflect the presence of endometrial tissue. Thus, the position of the glandular structure is obtained. Under the influence of HE staining, ectopic glands show light pink cytoplasm, and the ectopic glands are usually surrounded by fibrotic stromal tissue, which appears as dark pink fibers. There is a contrast in saturation between the two, and the ectopic glands and ectopic stroma can be distinguished based on the color. Therefore, preferably, in some possible implementation manners of the embodiments of the present invention, the process of obtaining the saturation connected components includes:

[0056] Convert the stained section image into an HSV image; obtain all the minima in the distribution histogram of the S channel corresponding to the HSV image; divide the distribution histogram of the S channel into at least two saturation intervals at intervals of all the minima; use all the connected components formed by the pixel points in each saturation interval as the saturation connected components. Since the existence of the minima can clearly observe different saturation characteristics, the saturation intervals are divided by the minima. It should be noted that the S channel is the channel corresponding to saturation, and the subsequent S channel values correspond to saturation values, which are all conventional concepts of HSV images and will not be elaborated further here.

[0057] In a specific implementation manner of the embodiments of the present invention, the stained section image is grayscale-converted to obtain a grayscale section image; connected component analysis is performed on the grayscale section image to obtain all the grayscale connected components obtained in the embodiments of the present invention; the grayscale connected components are used for subsequent determination of the columnar epithelial region and will not be elaborated further here.

[0058] Step S102: Determine the boundary possibility of each saturation connected component according to the distance between the saturation connected components and the saturation deviation situation of the corresponding saturation intervals; screen out at least two connected component clustering clusters according to the boundary possibility and the positional distribution and saturation relative deviation situation among the saturation connected components.

[0059] Since the components in the ectopic glands have different affinities for dyes, the area of the ectopic glands will be divided into block areas with relatively small differences in saturation, that is, the inside of the ectopic glands will be divided into several saturation connected components. Therefore, it is necessary to cluster the saturation connected components to obtain the position of the entire ectopic gland.

[0060] When clustering saturation connected regions, the clustering distance of saturation connected regions is mainly adjusted according to the color of saturation connected regions. However, when clustering, the saturation connected regions located at the boundary of ectopic glands and the saturation connected regions located at the center of ectopic glands have the same or similar saturation differences, that is, color differences, compared with the saturation connected regions of the same interstitial tissue. However, the Euclidean distances between them and the saturation connected regions of the same interstitial tissue are very different. Therefore, the clustering distance cannot be adjusted only according to the color difference. In particular, it is necessary to distinguish the saturation connected regions at the boundary of ectopic glands from the interstitial tissue, so that the saturation connected regions belonging to ectopic glands can be clustered into one category, making the obtained ectopic glands more complete. Therefore, the boundary possibility of the saturation connected region is calculated.

[0061] Preferably, in some possible implementation manners of the embodiments of the present invention, the process of obtaining the boundary possibility includes:

[0062] Each saturation connected region is sequentially used as the target connected region; the saturation interval where the target connected region is located is used as the target interval; the other saturation intervals outside the target interval are used as the reference intervals; the average value of the distances between the centroids of all saturation connected regions in each reference interval and the centroid of the target connected region is negatively correlated and mapped to determine the boundary contribution index of each reference interval to the target connected region.

[0063] For each reference interval, the relatively smaller the distances between the respective saturation connected regions and the target connected region are, the more it indicates that the saturation connected regions closer to the target connected region are mostly located in the corresponding reference interval. And because the saturation difference between ectopic glands and ectopic stroma is relatively large, on the basis of a relatively large boundary contribution index. If the saturation characteristics of the corresponding reference region are more different from the saturation characteristics of the target connected region, the higher the possibility that the target region is located at the boundary; based on this characteristic, further, the difference between the maximum value of the S-channel value of each reference interval and the maximum value of the S-channel value of the target interval is used as the saturation interval deviation corresponding to each reference interval; thus, according to the product of the boundary contribution index and the saturation interval deviation, the boundary influence parameter of each reference interval on the target connected region is determined; the accumulated value of the boundary influence parameters of all reference intervals corresponding to the target connected region is normalized to determine the boundary possibility of the target connected region.

[0064] In a specific implementation manner of the embodiments of the present invention, the process of obtaining the boundary possibility is represented by the formula: where, D k is the boundary possibility of the target connected region k; A k is the number of reference intervals corresponding to the target connected region k; S k is the maximum value of the S-channel value in the saturation interval corresponding to the target connected region k; S k,ais the maximum value of the S-channel value in the a-th reference interval corresponding to the target connected region k; || is the absolute value symbol; B a is the number of saturation connected regions in the a-th reference interval corresponding to the target connected region k; d k,a,b is the distance between the centroid of the target connected region k and the centroid of the b-th saturation connected region in the corresponding a-th reference interval; exp() is the exponential function with the natural constant as the base; is the boundary contribution index of the corresponding a-th reference interval to the target connected region k; is the boundary influence parameter of the corresponding a-th reference interval to the target connected region k; Norm() is the linear normalization function; Further, the boundary possibilities of all saturation connected regions are determined according to the process of obtaining the boundary possibilities of the target connected region.

[0065] Further, the clustering distance between saturation connected regions is corrected according to the boundary possibilities, that is, the boundary possibilities are used as the adjustment degree. Preferably, in some possible implementation manners of the embodiments of the present invention, the process of obtaining the connected region clustering clusters includes:

[0066] According to the extension situation of the minimum value of the boundary possibilities in the adjacent region of the target connected region, the central connected region of the target connected region is determined; the process of obtaining the central connected region includes: taking the reference connected region with the smallest boundary possibility among each saturation connected region and all its adjacent saturation connected regions as the corresponding minimum value connected region; using the target connected region as the seed point, taking the obtaining of the minimum value connected region as the growth condition and using the minimum value connected region as the new seed point for region growth, until the obtained growth region does not change, and taking the minimum value connected region obtained finally in the region growth process as the central connected region of the target connected region. That is, for the target connected region, obtain the target connected region itself and the saturation connected region with the smallest boundary possibility among all its adjacent saturation connected regions, and continue to iterate to obtain the saturation connected region with the smallest boundary possibility among the smallest saturation connected region and all its adjacent saturation connected regions, until the smallest saturation connected region does not change, stop the iteration, and record the smallest saturation connected region as the central connected region of the target connected region; by determining the central connected region, it is possible to avoid the saturation connected regions at the boundary being divided into other clustering clusters when determining the clustering distance subsequently.

[0067] Further, the other saturation connected regions outside the target connected region are used as comparison connected regions; according to the difference between the mean value of the S-channel values of all pixel points in each comparison connected region and the mean value of the S-channel values of all pixel points in the target connected region, the saturation deviation value of each comparison connected region is determined. The smaller the saturation deviation value, the more likely it belongs to the same region, and the corresponding clustering distance should be smaller.

[0068] The distance between the centroid of the target connected component and the centroid of each comparison connected component is used as the first comparison distance for each comparison connected component; the distance between the centroid of the central connected component and the centroid of each comparison connected component is used as the second comparison distance for each comparison connected component; calculate the difference between the second comparison distance and the first comparison distance, and take the product of the difference and the boundary possibility as the boundary weighting coefficient. For the target connected component, if the target connected component is located at the boundary and the comparison connected component does not belong to the connected component of the same region, then theoretically the corresponding clustering distance should be larger; conversely, if the comparison connected component belongs to the connected component of the same region, then the corresponding clustering distance should be relatively smaller; therefore, for a target connected component with a large boundary possibility, the distance from the comparison connected component of the boundary of other types of regions is usually small. So at this time, in order to avoid being divided into other connected components, it is necessary to correct the distance by combining the difference between the second comparison distance and the first comparison distance; the larger the corresponding second comparison distance is relative to the first comparison distance, the greater the target boundary possibility, that is, the greater the boundary weighting coefficient, which means it is more in line with this situation. At this time, it is necessary to increase the physical distance to a certain extent, so that the comparison connected component and the target connected component are divided into different clustering clusters. Therefore, further calculate the sum value between the boundary weighting coefficient and the first comparison distance; take the product of the sum value and the saturation deviation value as the clustering distance between the target connected component and each comparison connected component; after adding the boundary weighted distance and the first comparison distance, the possibility that the connected component at the boundary position is divided into other types of regions is reduced, making the obtained clustering distance more reasonable.

[0069] In a specific implementation manner of the embodiment of the present invention, the process of obtaining the clustering distance is expressed by the formula: F k,h =|G k -G k，h |×(H k,h +(H′ k,h -H k,h )×D k );wherein, F k,h is the clustering distance between the target connected component k and the corresponding h-th comparison connected component; G k is the average value of the S-channel values of all pixel points in the target connected component k; G k,h is the corresponding h-th comparison connected component of the target connected component k; D k is the boundary possibility of the target connected component k; H k，h is the distance between the centroid of the target connected component k and the centroid of the corresponding h-th comparison connected component, that is, the first comparison distance; H′ kis the distance between the centroid of the central connected domain of the target connected domain k and the centroid of the corresponding h-th comparison connected domain, that is, the second comparison distance; it should be noted that in the embodiments of the present invention, the Euclidean distance calculation method is used for all distance calculations; (H′ k，h -H k，h )×D k is the boundary weighting coefficient between the target connected domain k and the corresponding h-th comparison connected domain.

[0070] Furthermore, the clustering distance between every two saturation connected domains is statistically calculated; all saturation connected domains are subjected to clustering analysis by the DBSCAN clustering algorithm according to all clustering distances to obtain at least two connected domain clustering clusters; it should be noted that the DBSCAN clustering algorithm is a well-known technical means in the art, and other clustering methods can also be adopted according to the specific implementation environment, such as the k-means clustering algorithm, which will not be further limited and elaborated here.

[0071] Step S103: Perform region synthesis according to the distribution adjacency of each gray-scale connected domain and the approximation of the columnar features in terms of morphology to determine all suspected columnar epithelial regions; based on the number of gray-scale connected domains in the suspected columnar epithelial regions and the columnar approximation, determine the corresponding epithelial probability.

[0072] Each connected domain clustering cluster represents a complete regional tissue structure. Some connected domain clustering clusters belong to ectopic glands, so it is necessary to screen out ectopic glands from the connected domain clustering clusters, that is, calculate the gland probability of each connected domain clustering cluster. Since ectopic glands are usually surrounded by columnar epithelial cells and are distributed along the inner wall of the glandular lumen, and the columnar epithelial cells are columnar or rectangular, the aspect ratio of the minimum circumscribed rectangle of the columnar epithelial cells is relatively large. And because ectopic glands are usually surrounded by columnar epithelial cells and the epithelial cells have the characteristic of continuous arrangement, so the area that may be columnar epithelial cells is obtained in the image, and then the gland probability of the area is judged and analyzed according to the surrounding property of the ectopic glands by the columnar epithelial cells. Please refer to Figure 2 , which shows a schematic diagram of the morphological structure of columnar epithelial cells provided by an embodiment of the present invention. In Figure 2 , the columnar epithelial cells all present slender columns and are densely distributed in continuous arrangement when closely attached.

[0073] Preferably, in some possible implementation manners of the embodiments of the present invention, the process of obtaining the suspected columnar epithelial region includes:

[0074] Perform a negative correlation mapping on the ratio between the width and length of the minimum bounding rectangle of each grayscale connected region to determine the columnar degree of each grayscale connected region; regard the grayscale connected regions with a columnar degree greater than a preset columnar threshold as columnar connected regions; merge all adjacent columnar connected regions to obtain all suspected columnar epithelial regions. First, columnar epithelial cells present a columnar shape and are approximately rectangular with a longer long side. Therefore, after making the minimum bounding rectangle for them, the length is longer relative to the width. Thus, the greater the calculated columnar degree, the greater the probability that the corresponding grayscale connected region belongs to columnar epithelial cells. And columnar epithelial cells have the characteristic of continuous arrangement. Therefore, further merge all adjacent columnar connected regions to determine the suspected columnar epithelial regions that may belong to the region where epithelial cells are located. It should be noted that the width is the length of the short side of the minimum bounding rectangle, and the length is the length of the long side of the minimum bounding rectangle.

[0075] In a specific implementation manner of the embodiment of the present invention, the process of obtaining the columnar degree is expressed by the formula: where Z r is the columnar degree of the r-th grayscale connected region; L1 r is the width of the minimum bounding rectangle of the r-th grayscale connected region; L2 r is the length of the minimum bounding rectangle of the r-th grayscale connected region. In a specific implementation manner of the embodiment of the present invention, the preset columnar threshold is set to 0.7, which can be adjusted according to the specific implementation environment.

[0076] Furthermore, based on the characteristic that columnar epithelial cells have continuous arrangement, measure the credibility of the epithelial region for the obtained suspected columnar epithelial regions, that is, verify the possibility that it belongs to the real columnar epithelial region, so as to better assign weights for subsequent calculation of gland probabilities. Preferably, in some possible implementation manners of the embodiment of the present invention, the process of obtaining the epithelial probability is expressed by the formula: The process of obtaining the epithelial probability includes:

[0077] Normalize the product of the mean of the columnar degrees of all columnar connected regions in the suspected columnar epithelial region and the total number of columnar connected regions to determine the corresponding epithelial probability. For each suspected columnar epithelial region, the greater the overall columnar degree of the columnar connected regions therein, the more the corresponding local shape features conform to the shape of epithelial cells; and the more the number of columnar connected regions, the denser the epithelial cells and the greater the continuity, and the more likely it is to conform to the characteristic of continuous arrangement of epithelial cells. Therefore, combine the mean of the columnar degrees of all columnar connected regions and the total number of columnar connected regions in the form of a product to determine the epithelial probability representing belonging to the real columnar epithelial region. The purpose of normalization is to enable the epithelial probability to better play the role of weight in the follow-up.

[0078] In a specific implementation manner of the embodiment of the present invention, the acquisition process of the epithelial probability is expressed by the formula:

[0079] where Q q is the epithelial probability of the q-th suspected columnar epithelial region; is the average value of the columnar degrees of all columnar connected components in the q-th suspected columnar epithelial region; M q is the total number of columnar connected components in the q-th suspected columnar epithelial region; Norm() is a linear normalization function.

[0080] Step S104: Determine the corresponding probability of ectopic glands according to the boundary possibility distribution in each connected component clustering cluster and the epithelial probability of the adjacent suspected columnar epithelial regions; determine the abnormality probability of each connected component clustering cluster based on the probability of ectopic glands and the concentrated distribution of the connected component clustering clusters.

[0081] Since ectopic glands are usually surrounded by columnar epithelial cells and the epithelial cells are distributed along the inner wall of the glandular cavity, the distance between the boundary of the ectopic gland and the epithelial cells is very close. Therefore, the distance between the saturation connected component with a higher boundary possibility and the epithelial cells is very close. Thus, the greater the epithelial probability of the suspected columnar epithelial region closest to the saturation connected component with a higher boundary possibility within the connected component clustering cluster, the greater the gland probability of the connected component clustering cluster. Based on this logic, the acquisition process of the probability of ectopic glands for each connected component clustering cluster includes:

[0082] Take the epithelial probability of the suspected columnar epithelial region closest to each saturation connected component as the adjacent reference probability of each saturation connected component; take the product between the normalized value of the boundary possibility of each saturation connected component and the adjacent reference probability as the local reference probability; determine the corresponding probability of ectopic glands according to the accumulated value of the local reference probabilities of all saturation connected components in each connected component clustering cluster. Based on the fact that the greater the epithelial probability of the suspected columnar epithelial region closest to the saturation connected component with a higher boundary possibility within the connected component clustering cluster, the greater the gland probability of the connected component clustering cluster, it can be known that for each connected component clustering cluster, the greater the overall local reference probability of its corresponding saturation connected components, the more likely it is that columnar epithelial cells are distributed on the edge of its corresponding position feature, and the more in line with the characteristics of ectopic glands, that is, the greater the probability of ectopic glands.

[0083] In a specific implementation manner of the embodiment of the present invention, the acquisition process of the probability of ectopic glands is expressed by the formula: where P c is the probability of ectopic glands of the c-th connected component clustering cluster; U c is the number of saturation connected components of the c-th connected component clustering cluster; Dc，u is the boundary possibility of the u-th saturation-connected domain in the c-th connected domain cluster; softmax() is the activation normalization function, which makes the sum of all normalized values equal to 1, that is, the sum of all boundary possibilities after normalization is 1. Combined with the summation symbol, it makes the calculated probability of ectopic glands more robust; Q' c，u is the epithelial probability of the suspected columnar epithelial area closest to the u-th saturation-connected domain in the c-th connected domain cluster.

[0084] In the case of endometriosis, the endometrial tissue sheds under certain pathological conditions and migrates to extrauterine sites such as the abdominal cavity and ovaries through menstruation reflux, lymphatic or blood circulation. Ectopic glands usually concentrate in certain specific areas, so the ectopic glands show a concentrated distribution phenomenon. Therefore, it is possible to judge whether there are ectopic glands according to whether they conform to the characteristics of concentrated distribution, so as to determine whether there is an abnormality. Preferably, in some possible implementation manners of the embodiments of the present invention, the process of obtaining the abnormality probability includes:

[0085] Successively take each connected domain cluster as the target cluster; take the other connected domain clusters outside the target cluster as the comparison clusters; take the minimum distance between all pixel points in each comparison cluster and all pixel points in the target cluster as the inter-cluster distance of each comparison cluster; for each connected domain cluster, if the overall inter-cluster distance between it and each comparison cluster is smaller, it means that the connected domain cluster conforms more to the concentrated distribution phenomenon of ectopic glands, and the corresponding possibility of abnormality is greater; therefore, further multiply the negative correlation normalization value of the inter-cluster distance by the ectopic gland probability of the corresponding comparison cluster as the reference concentration degree of each comparison cluster; the greater the reference concentration degree, the more likely it conforms to the characteristics of ectopic glands; while the ectopic gland probability represents the possibility of belonging to ectopic glands in another dimension, that is, the degree of abnormality. Therefore, combined with the overall size of the reference concentration degree with the ectopic gland probability as the weight, multiply the mean value of the reference concentration degrees of all comparison clusters by the ectopic gland probability of the target cluster as the abnormality probability of the target cluster; the greater the abnormality probability, the greater the probability of belonging to ectopic glands, and the more it conforms to the corresponding gland lesion characteristics.

[0086] In a specific implementation manner of the embodiments of the present invention, the process of obtaining the abnormality probability is expressed by the formula: where, T c is the abnormality probability of the c-th connected domain cluster; P c is the ectopic gland probability of the c-th connected domain cluster; V c is the number of comparison clusters of the c-th connected domain cluster; X c,vis the inter-cluster distance between the c-th connected component cluster and the v-th comparison cluster; P c,v is the probability of ectopic glands of the v-th comparison cluster of the c-th connected component cluster; softmax(exp(-X c，v )) × P c，v is the degree of concentration in the reference set of the v-th comparison cluster of the C-th connected component cluster; exp() is the exponential function with the natural constant as the base; softmax() is the activation normalization function.

[0087] Step S105: Screen out the abnormal connected component clusters according to the abnormal probability; locate the adhesion lesions of the endometriosis rat model according to the abnormal connected component clusters.

[0088] Since the greater the abnormal probability, the greater the probability of belonging to ectopic glands and the more in line with the characteristics of the corresponding glandular lesions; therefore, for the purpose of locating the adhesion lesions of the endometriosis rat model, it is further necessary to screen out the connected component clusters belonging to abnormalities. Specifically: The connected component clusters with an abnormal probability greater than the preset abnormal threshold are used as the abnormal connected component clusters. In a specific implementation manner of the embodiment of the present invention, the preset abnormal threshold is set to 0.67 and can be adjusted according to the specific implementation environment. After screening out the connected component clusters, the minimum bounding rectangle of the connected component clusters is obtained to obtain the corresponding abnormal annotation border, and the abnormal annotation border is annotated or segmented to locate the adhesion lesions of the endometriosis rat model.

[0089] In summary, the present application first adjusts the clustering distance according to the color difference and position inertial system between the saturation connected components, and clusters to obtain the connected component clusters; since the connected component clusters contain ectopic glands, it is necessary to screen the clusters to obtain ectopic glands. By the characteristic that columnar epithelial cells are distributed on the inner wall of the ectopic glands, the gland probability of each connected component cluster is obtained, and the possibility of each connected component cluster belonging to ectopic glands is judged; since the ectopic glands show a concentrated distribution pattern, through the aggregation of the connected component clusters with a higher gland probability, the abnormal probability of the connected component clusters is obtained, and further the possibility of the connected component clusters belonging to abnormal ectopic glands is judged. So far, the present invention annotates the possible lesion positions in the stained section image through the abnormal probability of the connected component clusters, thereby improving the accuracy of lesion location of the endometriosis rat model.

[0090] It should be noted that: the above sequence of the embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the 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.

[0091] 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, and the differences between each embodiment and other embodiments are emphasized.

Claims

1. A method for locating adhesion lesions of an endometriosis rat model, characterized in that The method includes: Collecting stained section images of a rat model of endometriosis; determining all saturation-connected regions under each saturation interval according to the saturation distribution of the stained section images; obtaining all gray-connected regions after graying the stained section images; Determining the boundary possibility of each saturation-connected region according to the distance between saturation-connected regions and the saturation deviation of the corresponding saturation interval; screening out at least two connected region clustering clusters according to the boundary possibility and the positional distribution and relative saturation deviation between each saturation-connected region; Performing region synthesis according to the distribution adjacency of each gray-connected region and the approximate columnar feature in morphology, and determining all suspected columnar epithelial regions; determining the corresponding epithelial probability based on the number of gray-connected regions and columnar approximation in the suspected columnar epithelial regions; Determining the corresponding ectopic gland probability according to the boundary possibility distribution in each connected region clustering cluster and the epithelial probability of adjacent suspected columnar epithelial regions; determining the abnormality probability of each connected region clustering cluster based on the ectopic gland probability and the concentrated distribution of the connected region clustering cluster; Screening out abnormal connected region clustering clusters according to the abnormality probability; localizing the adhesion lesions of the rat model of endometriosis according to the abnormal connected region clustering clusters.

2. The adhesion lesion localization method of an endometriosis rat model according to claim 1, characterized in that, The obtaining process of the saturation-connected region includes: Converting the stained section image into an HSV image; obtaining all minimum values in the distribution histogram of the S channel corresponding to the HSV image; dividing the distribution histogram of the S channel into at least two saturation intervals at intervals of all minimum values; taking all connected regions composed of pixel points under each saturation interval as saturation-connected regions.

3. A method for localizing adhesion lesions of an endometriosis rat model according to claim 2, characterized in that, The obtaining process of the boundary possibility includes: Successively taking each saturation-connected region as the target connected region; taking the saturation interval where the target connected region is located as the target interval; taking other saturation intervals outside the target interval as reference intervals; Performing negative correlation mapping on the average value of the distances between the centroids of all saturation-connected regions in each reference interval and the centroid of the target connected region to determine the boundary contribution index of each reference interval to the target connected region; Taking the difference between the maximum value of the S channel value of each reference interval and the maximum value of the S channel value of the target interval as the saturation interval deviation corresponding to each reference interval; Determining the boundary influence parameter of each reference interval on the target connected region according to the product of the boundary contribution index and the saturation interval deviation; normalizing the accumulated value of the boundary influence parameters of all reference intervals corresponding to the target connected region to determine the boundary possibility of the target connected region.

4. The method for locating adhesion lesions of an endometriosis rat model according to claim 3, characterized in that, The obtaining process of the connected region clustering cluster includes: Determining the central connected region of the target connected region according to the extension of the minimum value of the boundary possibility in the adjacent region of the target connected region; taking other saturation-connected regions outside the target connected region as comparison connected regions; Determining the saturation deviation value of each comparison connected region according to the difference between the mean value of the S channel values of all pixel points in each comparison connected region and the mean value of the S channel values of all pixel points in the target connected region; Take the distance between the centroid of the target connected domain and the centroid of each comparison connected domain as the first comparison distance of each comparison connected domain; take the distance between the centroid of the central connected domain and the centroid of each comparison connected domain as the second comparison distance of each comparison connected domain. Calculate the difference between the second comparison distance and the first comparison distance, and take the product of the difference and the boundary possibility as the boundary weighting coefficient; calculate the sum value between the boundary weighting coefficient and the first comparison distance; take the product of the sum value and the saturation deviation value as the clustering distance between the target connected domain and each comparison connected domain. Statistically calculate the clustering distances between every two saturation connected domains; perform clustering analysis on all saturation connected domains through the DBSCAN clustering algorithm based on all clustering distances to obtain at least two connected domain clustering clusters.

5. The adhesion lesion localization method of an endometriosis rat model according to claim 4, characterized in that, The process of obtaining the central connected domain includes: Take the reference connected domain with the minimum boundary possibility among each saturation connected domain and all its adjacent saturation connected domains as the corresponding minimum value connected domain. Using the target connected domain as the seed point, taking the acquisition of the minimum value connected domain as the growth condition and using the minimum value connected domain as the new seed point for region growth. After the obtained growth region no longer changes, take the last obtained minimum value connected domain in the region growth process as the central connected domain of the target connected domain.

6. The method for localizing adhesion lesions of an endometriosis rat model according to claim 1, characterized in that The process of obtaining the suspected columnar epithelial region includes: Perform a negative correlation mapping on the ratio of the width to the length of the minimum circumscribed rectangle of each grayscale connected domain to determine the columnar degree of each grayscale connected domain; take the grayscale connected domain with a columnar degree greater than the preset columnar threshold as the columnar connected domain. Merge all adjacent columnar connected domains to obtain all suspected columnar epithelial regions.

7. A method for localizing adhesion lesions of an endometriosis rat model according to claim 6, characterized in that, The process of obtaining the epithelial probability includes: Normalize the product of the mean of the columnar degrees of all columnar connected domains in the suspected columnar epithelial region and the total number of columnar connected domains to determine the corresponding epithelial probability.

8. A method for locating adhesion lesions of an endometriosis rat model according to claim 1, characterized in that The process of obtaining the probability of ectopic glands includes: Take the epithelial probability of the suspected columnar epithelial region closest to each saturation connected domain as the adjacent reference probability of each saturation connected domain. Take the product of the normalized value of the boundary possibility of each saturation connected domain and the adjacent reference probability as the local reference probability. Determine the corresponding probability of ectopic glands according to the cumulative value of the local reference probabilities of all saturation connected domains in each connected domain clustering cluster.

9. A method for locating adhesion lesions of an endometriosis rat model according to claim 1, characterized in that, The process of obtaining the abnormal probability includes: Successively take each connected domain clustering cluster as the target clustering cluster; take the other connected domain clustering clusters outside the target clustering cluster as the comparison clustering clusters. Take the minimum distance between all pixel points in each comparison clustering cluster and all pixel points in the target clustering cluster as the inter-cluster distance of each comparison clustering cluster. Take the product of the negative correlation normalized value of the inter-cluster distance and the probability of ectopic glands of the corresponding comparison clustering cluster as the reference concentration degree of each comparison clustering cluster. Take the product of the mean of the reference concentration degrees of all comparison clustering clusters and the probability of ectopic glands of the target clustering cluster as the abnormal probability of the target clustering cluster.

10. The method for localizing adhesion lesions of an endometriosis rat model according to claim 1, wherein, The process of obtaining the abnormal connected component clustering clusters includes: Taking the connected component clustering clusters with abnormal probabilities greater than the preset abnormal threshold as the abnormal connected component clustering clusters.

Images