CN118196789A - Artificial intelligence processing method of tumor pathological tissue image - Google Patents
Artificial intelligence processing method of tumor pathological tissue image Download PDFInfo
- Publication number
- CN118196789A CN118196789A CN202410598674.XA CN202410598674A CN118196789A CN 118196789 A CN118196789 A CN 118196789A CN 202410598674 A CN202410598674 A CN 202410598674A CN 118196789 A CN118196789 A CN 118196789A
- Authority
- CN
- China
- Prior art keywords
- region
- communication
- parameters
- areas
- parameter
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 206010028980 Neoplasm Diseases 0.000 title claims abstract description 59
- 230000001575 pathological effect Effects 0.000 title claims abstract description 54
- 238000003672 processing method Methods 0.000 title claims abstract description 21
- 238000013473 artificial intelligence Methods 0.000 title claims description 17
- 238000004891 communication Methods 0.000 claims abstract description 201
- 239000013598 vector Substances 0.000 claims abstract description 82
- 230000000877 morphologic effect Effects 0.000 claims abstract description 62
- 238000010586 diagram Methods 0.000 claims abstract description 15
- 238000012545 processing Methods 0.000 claims abstract description 11
- 230000000875 corresponding effect Effects 0.000 claims description 70
- 238000000034 method Methods 0.000 claims description 35
- 238000012937 correction Methods 0.000 claims description 23
- 210000000481 breast Anatomy 0.000 claims description 17
- 238000004364 calculation method Methods 0.000 claims description 15
- 238000013507 mapping Methods 0.000 claims description 14
- 238000010606 normalization Methods 0.000 claims description 11
- 238000011065 in-situ storage Methods 0.000 claims description 7
- 230000011218 segmentation Effects 0.000 claims description 7
- 201000010983 breast ductal carcinoma Diseases 0.000 claims description 6
- 230000002596 correlated effect Effects 0.000 claims description 2
- 238000004458 analytical method Methods 0.000 abstract description 11
- 230000002776 aggregation Effects 0.000 abstract description 8
- 238000004220 aggregation Methods 0.000 abstract description 8
- 238000010191 image analysis Methods 0.000 abstract description 3
- 210000004027 cell Anatomy 0.000 description 72
- 210000001519 tissue Anatomy 0.000 description 42
- 230000002159 abnormal effect Effects 0.000 description 16
- 208000009458 Carcinoma in Situ Diseases 0.000 description 7
- 230000003902 lesion Effects 0.000 description 7
- 201000011510 cancer Diseases 0.000 description 6
- 230000008569 process Effects 0.000 description 5
- 208000037396 Intraductal Noninfiltrating Carcinoma Diseases 0.000 description 4
- 208000028715 ductal breast carcinoma in situ Diseases 0.000 description 4
- 206010073094 Intraductal proliferative breast lesion Diseases 0.000 description 3
- 201000007273 ductal carcinoma in situ Diseases 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 239000003086 colorant Substances 0.000 description 2
- 230000006378 damage Effects 0.000 description 2
- 238000003745 diagnosis Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000002360 preparation method Methods 0.000 description 2
- 206010006187 Breast cancer Diseases 0.000 description 1
- 208000026310 Breast neoplasm Diseases 0.000 description 1
- 108010077544 Chromatin Proteins 0.000 description 1
- 239000008280 blood Substances 0.000 description 1
- 210000004369 blood Anatomy 0.000 description 1
- 210000004204 blood vessel Anatomy 0.000 description 1
- 230000005907 cancer growth Effects 0.000 description 1
- 210000003483 chromatin Anatomy 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 230000001086 cytosolic effect Effects 0.000 description 1
- 238000004043 dyeing Methods 0.000 description 1
- 239000012530 fluid Substances 0.000 description 1
- 229940088597 hormone Drugs 0.000 description 1
- 239000005556 hormone Substances 0.000 description 1
- 206010020718 hyperplasia Diseases 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 210000001165 lymph node Anatomy 0.000 description 1
- 210000005075 mammary gland Anatomy 0.000 description 1
- 230000036285 pathological change Effects 0.000 description 1
- 231100000915 pathological change Toxicity 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 230000035755 proliferation Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000028327 secretion Effects 0.000 description 1
- 238000010186 staining Methods 0.000 description 1
- 230000001502 supplementing effect Effects 0.000 description 1
Landscapes
- Image Processing (AREA)
Abstract
The invention relates to the technical field of image analysis and processing, in particular to an artificial intelligent processing method of tumor pathological tissue images. Firstly, acquiring a communication region in a tumor pathological tissue image; further analyzing the shape characteristics and the size characteristics of each connected region, and combining the discrete characteristics of the gray values of the pixels in each connected region to obtain an initial region characteristic vector; further acquiring a corrected region feature vector according to similar features of the initial region feature vectors of other communication regions around each communication region; further according to the difference characteristics of the corrected region characteristic vector, combining the spatial distribution characteristics to obtain a clustering distance; and finally obtaining a hierarchical clustering pedigree diagram. According to the embodiment of the invention, the accurate clustering distance is determined through the morphological characteristics, the gray level characteristics and the local similar characteristics of the region, the influence caused by normal cell aggregation is reduced, a more accurate hierarchical clustering lineage diagram is obtained, and the analysis of related personnel is facilitated.
Description
Technical Field
The invention relates to the technical field of image analysis and processing, in particular to an artificial intelligent processing method of tumor pathological tissue images.
Background
Ductal carcinoma in situ (Ductal Carcinoma In Situ, DCIS) is a form of early stage breast cancer, which refers to the growth of cancer cells within the ductal breast that have not spread to surrounding tissues or other sites. In this case, the cancer cells are localized to a small area within the breast, called the home site, without invading surrounding tissues or lymph nodes. Since cancer cells do not spread to other parts of other bodies, there is no direct harm to the body. If the cancer of the catheter invades into tissues around the mammary gland or other parts, the catheter breaks through and enters blood, so that cancer cells rapidly spread in the body, and huge damage is caused to the body. Therefore, the accuracy of the breast duct in-situ cancer pathological tissue image is ensured, and it is important to carry out timely treatment according to a reasonable and safe operation scheme formulated by the image.
The method is characterized in that the method comprises the steps of distinguishing diseased cells from normal cells, namely, the important step of image analysis of pathological tissue of tumors, distinguishing normal cells from diseased cells by adopting a hierarchical clustering algorithm in a clustering mode generally according to the color difference and the position relation of images, but when the colors of the cells are distinguished, the cells in a normal area and the cells in a diseased area are stained with colors, and gray values are overlapped, so that normal cells or diseased cells cannot be distinguished according to the difference of the gray values; meanwhile, the cell distance between the inside of the duct and the outside of the duct is smaller according to the distance clustering, so that the clustering result is abnormal, and finally, the in-situ breast duct cancer classification result is inaccurate in distinguishing, and the hierarchical clustering lineage diagram of the obtained tumor pathological tissue image is inaccurate.
Disclosure of Invention
In order to solve the technical problem that the traditional hierarchical clustering method is not ideal for processing the tumor pathological tissue image, the invention aims to provide an artificial intelligent processing method for the tumor pathological tissue image, and the adopted technical scheme is as follows:
Acquiring a tumor pathological tissue image of breast duct carcinoma in situ; threshold segmentation is carried out on the tumor pathological tissue image, and a communication area in the tumor pathological tissue image is obtained;
analyzing the shape characteristics and the size characteristics of each communication area to obtain the initial morphological normal parameters of each communication area; acquiring gray level normal degree parameters according to the discrete characteristics of gray level values of pixel points in each communication area; constructing an initial region feature vector according to the initial form normal parameter and the gray level normal degree parameter of each connected region;
Obtaining a corrected region feature vector of each connected region according to similar features of the initial region feature vectors of the other connected regions of the preset neighborhood parameters closest to each connected region; according to the difference characteristics of the characteristic vectors of the correction areas between any two communication areas, combining the spatial distribution characteristics of the characteristic vectors of the correction areas to obtain the clustering distance between the two corresponding communication areas;
and carrying out hierarchical clustering on all the connected areas according to the clustering distance to obtain the hierarchical clustering lineage diagram of the tumor pathological tissue image.
Further, the method for acquiring the initial morphological normal parameters comprises the following steps:
taking the number of pixel points in each communication area as the size characteristic parameter of each communication area;
taking the square of the number of boundary pixel points of each communication area as the perimeter characteristic parameter of each communication area; taking the ratio of the number of pixel points in each communication area to the corresponding perimeter characteristic parameter as the shape index of each communication area; the pixel points in the communication area comprise boundary pixel points;
Acquiring shape characteristic parameters of each communication region according to the difference characteristics of the shape index of each communication region and the shape index of the standard circle;
Combining the size characteristic parameters and the shape characteristic parameters of each communication area to obtain initial morphological normal parameters of each communication area; the size characteristic parameter is inversely related to the initial morphological normal parameter; the shape characteristic parameter is positively correlated with the initial morphology normal parameter; and normalizing the initial morphological normal parameters.
Further, the method for acquiring the shape characteristic parameter comprises the following steps:
Associating the shape index of each of the communication areas with And (3) carrying out negative correlation mapping and normalization on the absolute value of the difference value of the connected areas to serve as a shape characteristic parameter of each connected area.
Further, the method for acquiring the initial morphological normal parameters comprises the following steps:
And normalizing the ratio of the shape characteristic parameter to the size characteristic parameter of each communication region to obtain an initial morphological normal parameter of each communication region.
Further, the method for acquiring the gray level normal degree parameter comprises the following steps:
acquiring the variance of the gray value of the pixel point in each communication area as a first discrete characteristic parameter;
Acquiring a quarter bit distance of gray values of pixel points in each communication area as a second discrete characteristic parameter;
Acquiring gray level normal degree parameters corresponding to each communication region according to the first discrete characteristic parameters and the second discrete characteristic parameters corresponding to each communication region; the first discrete characteristic parameter and the second discrete characteristic parameter are inversely related to the gray level normal degree parameter; and the gray level normal degree parameters are subjected to normalization processing.
Further, the method for acquiring the gray level normal degree parameter comprises the following steps:
And carrying out negative correlation mapping and normalization on the product of the first discrete characteristic parameter and the second discrete characteristic parameter corresponding to each communication region, and then taking the product as a gray level normal degree parameter corresponding to each communication region.
Further, the method for acquiring the correction region feature vector comprises the following steps:
taking the minimum Euclidean distance between the pixel points between the two communication areas as a reference distance between the two corresponding communication areas; determining other communication areas with preset neighborhood parameters, the reference distance between the other communication areas and any communication area is the smallest, and using the other communication areas as adjacent communication areas of the corresponding communication areas;
According to the similar features between the initial region feature vectors of the adjacent communication regions corresponding to each communication region, combining the corresponding reference distances to obtain reference degree parameters of each adjacent communication region corresponding to each communication region;
Acquiring the reference degree weight of each adjacent communication area corresponding to each communication area according to the proportion of the reference degree parameter of each adjacent communication area corresponding to each communication area occupying all the reference degree parameters;
According to the reference degree weight of each adjacent communication region corresponding to each communication region, the initial morphological normal parameters of each adjacent communication region are combined, and the initial morphological normal parameters of each communication region are corrected to obtain corrected morphological normal parameters;
And replacing the corresponding initial morphological normal parameters by using the corrected morphological normal parameters to obtain corrected region feature vectors of each connected region.
Further, the method for acquiring the corrected morphological normal parameters comprises the following steps:
; wherein/> The serial number of the connected area; /(I)Represents the/>Initial morphological normal parameters of the individual connected regions; /(I)Represents the/>Correcting morphological normal parameters of the connected areas; /(I)Representing preset neighborhood parameters; /(I)A serial number indicating an adjacent communication area; /(I)Represents the/>First/>, corresponding to each communication regionInitial morphological normal parameters of the adjacent connected regions; /(I)Represents the/>First/>, corresponding to each communication regionReference degree weight of each adjacent connected region,/>，Represents the/>Average value of initial region feature vectors of all adjacent connected regions corresponding to each connected region,/>Represent the firstFirst/>, corresponding to each communication regionInitial region feature vectors of adjacent connected regions; /(I)A computation function representing a binary norm of the vector; /(I)Represents the/>The communicating region and the corresponding first/>Reference distance between adjacent connected areas.
Further, the method for acquiring the clustering distance comprises the following steps:
Taking a second norm of the difference of the correction region feature vectors between two connected regions as a first distance parameter between the two connected regions;
acquiring a second distance parameter between two connected areas by using a calculation formula of the second distance parameter;
Taking the product of the first distance parameter and the second distance parameter between any two connected areas as the clustering distance between the two corresponding connected areas.
Further, the calculation formula of the second distance parameter includes:
; wherein/> Represents the/>The communicating region and the/>Second distance parameters corresponding to the communication areas; /(I)For obtaining a minimum value in the set; /(I)A serial number indicating the connected region; /(I)A corrected region feature vector representing an h-th connected region; /(I)A computation function representing the two norms of the vector.
The invention has the following beneficial effects:
firstly, acquiring a tumor pathological tissue image of breast duct in-situ cancer, and providing a basis for the subsequent image processing steps; threshold segmentation is further carried out on the tumor pathological tissue image, a communication region in the tumor pathological tissue image is obtained, the interference of the background on the subsequent analysis process is removed, and the analysis of morphological characteristics of cells in different regions is facilitated; further analyzing the shape characteristics and the size characteristics of each communication region, acquiring the initial morphological normal parameters of each communication region, evaluating the cell characteristics in the communication region from the cell morphology angle of the communication region, and preparing for the subsequent analysis of the clustering distance characteristics among different communication regions and accurate clustering; further, according to the discrete features of the gray values of the pixel points in each connected region, gray level normal degree parameters are obtained, and from the gray level distribution angle of the connected region, the cell features in the connected region are evaluated, so that more basis is provided for the subsequent analysis of the clustering distance features among different connected regions; constructing an initial region feature vector according to the initial form normal parameters and the gray level normal degree parameters of each communication region, so that the form features and the gray level features are convenient to complement, and comprehensively analyzing the features of the communication regions; further correcting the initial region feature vector of each connected region according to the similar features of the initial region feature vectors of other connected regions with the nearest preset neighborhood parameters from each connected region, and acquiring the corrected region feature vector of each connected region, so as to eliminate errors caused by normal cell aggregation and prepare for subsequent accurate clustering; further correcting the difference characteristics of the regional characteristic vectors, simultaneously avoiding gathering the communication regions with unobvious cell characteristics into one type, and combining the spatial distribution characteristics of the corrected regional characteristic vectors to acquire the clustering distance between the two corresponding communication regions so as to provide accurate basis for subsequent clustering; and finally, hierarchical clustering is carried out on all the connected areas according to the clustering distance, and a hierarchical clustering pedigree diagram of the tumor pathological tissue image is obtained. According to the embodiment of the invention, the accurate clustering distance is determined through the morphological characteristics, the gray level characteristics and the local similar characteristics of the communication area, the influence caused by normal cell aggregation is reduced, a more accurate hierarchical clustering lineage diagram is obtained, and the analysis of related staff is facilitated.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions and advantages of the prior art, the following description will briefly explain the drawings used in the embodiments or the description of the prior art, and it is obvious that the drawings in the following description are only some embodiments of the invention, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flowchart of an artificial intelligence processing method for tumor pathological tissue image according to an embodiment of the present invention;
FIG. 2 is a view of an original tumor pathological tissue image of ductal carcinoma in situ, according to one embodiment of the present invention;
FIG. 3 is a hierarchical clustering lineage diagram obtained by a conventional hierarchical clustering method according to an embodiment of the present invention;
fig. 4 is a hierarchical clustering lineage diagram obtained by using an optimized hierarchical clustering method according to an embodiment of the present invention.
Detailed Description
In order to further describe the technical means and effects adopted by the invention to achieve the preset aim, the following is a detailed description of specific implementation, structure, characteristics and effects of an artificial intelligence processing method for tumor pathological tissue image according to the invention with reference to the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" means that the embodiments are not necessarily the same. Furthermore, the particular features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
The following specifically describes a specific scheme of an artificial intelligence processing method for tumor pathological tissue images provided by the invention with reference to the accompanying drawings.
Referring to fig. 1, a flowchart of an artificial intelligence processing method for a tumor pathological tissue image according to an embodiment of the present invention specifically includes:
step S1: firstly, obtaining a tumor pathological tissue image of breast duct carcinoma in situ; threshold segmentation is carried out on the tumor pathological tissue image, and a communication area in the tumor pathological tissue image is obtained.
In the embodiment of the invention, firstly, a tumor pathological tissue image of breast duct carcinoma in situ is acquired, and a basis is provided for the subsequent image processing step; referring to fig. 2, an original tumor pathological tissue image of breast duct carcinoma in situ provided by an embodiment of the present invention is shown, and it can be seen from fig. 2 that the original tumor pathological tissue image includes background areas such as tissue fluid, which is not beneficial to analyze morphological characteristics of cells in different areas, so that threshold segmentation is performed on the tumor pathological tissue image, a connected area in the tumor pathological tissue image is obtained, interference of the background on a subsequent analysis process is removed, calculation amount of subsequent processing is reduced, and processing efficiency is improved.
Preferably, in one embodiment of the present invention, the oxford method is adopted to perform threshold segmentation on the tumor pathological tissue image, in consideration of the advantages of strong adaptability, simple calculation, wide application range and the like.
It should be noted that, the oxford method is a technical means well known to those skilled in the art, and in other embodiments of the present invention, an implementer may also obtain a connected region in an image by using other threshold segmentation methods such as a watershed algorithm, which are all technical means well known to those skilled in the art, and will not be described again.
It should be noted that the embodiment of the invention is an artificial intelligence processing method of tumor pathological tissue images, which is used for assisting related personnel and does not belong to a direct diagnosis method or a treatment method.
Step S2: analyzing the shape characteristics and the size characteristics of each communication area to obtain the initial morphological normal parameters of each communication area; acquiring gray level normal degree parameters according to the discrete characteristics of gray level values of pixel points in each connected region; and constructing an initial region feature vector according to the initial form normal parameter and the gray level normal degree parameter of each connected region.
Experience shows that abnormal hyperplasia of cells in a lesion area of breast duct in-situ cancer can occur, so that morphological difference exists between normal cells and abnormal cells, and the abnormal cells in the lesion area have aggregation property due to excessive proliferation, so that a plurality of cells exist in the same communication area, normal cells are generally poor in viscosity and present certain isolated characteristics, so that the initial morphological normal parameters of each communication area can be obtained by analyzing the shape characteristics and the size characteristics of each communication area, and preparation is provided for subsequent analysis of clustering distance characteristics among different communication areas and accurate clustering.
Preferably, in the embodiment of the present invention, the larger the number of pixels in the connected region, the larger the size of the connected region is, the more likely that a plurality of cells are aggregated together, so the larger the number of pixels, the more likely that an abnormal connected region is, and the smaller the initial morphological normal parameter is; considering that normal cells are isolated, one normal cell corresponds to one communication area, the shape of the normal cell is round or nearly round, the abnormal cell is divided by aggregation and self, the shape of the communication area is various and has larger difference from the round, and the ratio of the area of the round to the square of the perimeter isTherefore, the ratio of the area of the connected region to the square of the perimeter can be obtained, and the ratio can be compared with/>Analyzing the initial morphological normal parameters of the connected region;
Based on the above, the number of pixel points in each communication area is taken as the size characteristic parameter of each communication area;
taking the square of the number of boundary pixel points of each communication area as the perimeter characteristic parameter of each communication area; taking the ratio of the number of pixel points in each communication area to the corresponding perimeter characteristic parameter as the shape index of each communication area; the pixel points in the communication area comprise boundary pixel points;
Correlating the shape index of each communication region with And (3) carrying out negative correlation mapping and normalization on the absolute value of the difference value of the connected regions to serve as a shape characteristic parameter of each connected region.
And normalizing the ratio of the shape characteristic parameter to the size characteristic parameter of each connected region to obtain the initial morphological normal parameter of each connected region.
The calculation formula of the initial morphological normal parameters comprises:
；
Wherein, The serial number of the connected area; /(I)Represents the/>Initial morphological normal parameters of the individual connected regions; /(I)Represents the/>Size characteristic parameters of the communicated areas; /(I)Represents the/>The number of boundary pixel points of each communication area; /(I)Represent the firstShape index of each communication region; /(I)Represents the/>Shape characteristic parameters of the connected areas; /(I)Expressed as natural constant/>An exponential function of the base;/(Representing a maximum-minimum normalization function.
In the calculation formula of the normal parameters of the initial form,The smaller the pixel point number of the communication area is, the more probably single cells are, and the larger the initial morphological normal parameters of the communication area are; shape index of connected region/>Shape index of standard circle/>The smaller the difference, the closer the shape of the connected region to a circle, the more likely it is a normal cell, and the larger the initial morphological normal parameter of the connected region.
It should be noted that, in other embodiments of the present invention, other basic mathematical operations or function mapping may be used to implement the related mapping, which are all technical means known to those skilled in the art, and are not described herein.
In the embodiment of the invention, the characteristics of hyperproliferation of cells in a lesion area are considered, nuclear division is increased, chromatin distribution is uneven, the gray scale difference of pixel points in a communication area is large, and the same color is displayed in an image during normal cell imaging, so that gray scale normal degree parameters can be obtained according to the discrete characteristics of gray scale values of the pixel points in each communication area, and more basis is provided for subsequent analysis of clustering distance characteristics among different communication areas.
Preferably, in one embodiment of the present invention, the larger the variance of the gray value of the pixel point in the connected region is considered, which means that the worse the staining uniformity of the cells in the connected region is, the more obvious the characteristic of the abnormal cells is, and the smaller the gray level normal degree parameter is; the fact that the number of the pixel points possibly subjected to abnormal communication areas is too large is considered, the corresponding gray value distribution is concentrated, so that the variance of the abnormal communication areas is smaller, the gray value in the communication areas is corrected by using the quarter bit distance of the gray value, the larger the quarter bit distance of the gray value in the areas is, the larger the span of the gray value of the pixel points is, the more scattered the gray value is, the smaller the concentration is, the greater the possibility of representing the abnormal areas is, the more concentrated the gray value distribution is, and the greater the dyeing uniformity is, and the higher the possibility of representing the normal areas is;
Based on the first discrete characteristic parameters, the variance of the gray value of the pixel point in each connected region is obtained;
Acquiring a quarter bit distance of gray values of pixel points in each connected region as a second discrete characteristic parameter;
And carrying out negative correlation mapping and normalization on the product of the first discrete characteristic parameter and the second discrete characteristic parameter corresponding to each communication region, and then taking the product as a gray level normal degree parameter corresponding to each communication region.
The calculation formula of the gray level normal degree parameter comprises:
；
Wherein, The serial number of the connected area; /(I)Represents the/>Gray level normal degree parameters of the connected areas; /(I)Represents the/>The pixel point gray value in each communication area is quarter bit distance; /(I)Represents the/>Variance of gray values of pixel points in the connected areas; /(I)Expressed as natural constant/>An exponential function of the base;/(Representing a maximum-minimum normalization function.
It should be noted that, in other embodiments of the present invention, other basic mathematical operations or function mapping may be used to implement the related mapping, which are all technical means known to those skilled in the art, and are not described herein.
In the embodiment of the invention, the cell characteristics in the communication area are considered to be represented by the initial morphological normal parameters and the gray level normal degree parameters, and the morphological characteristics and the gray level characteristics have certain complementarity, so that the two characteristics are combined to construct the characteristic vector, the characteristics of the cells in the communication area can be more comprehensively described, different communication areas can be conveniently and better distinguished, and preparation is made for obtaining accurate clustering distances subsequently.
Preferably, in one embodiment of the present invention, the initial region feature vector，/>The serial number of the connected area; /(I)Represents the/>Initial morphological normal parameters of the individual connected regions; /(I)Represents the/>And gray level normal degree parameters of the connected areas.
In other embodiments of the present invention, the practitioner may also constructAnd other forms of vectors.
Step S3: acquiring a corrected region feature vector of each connected region according to similar features of initial region feature vectors of other connected regions of preset neighborhood parameters which are nearest to each connected region; and according to the difference characteristics of the characteristic vectors of the correction areas between any two communication areas, combining the spatial distribution characteristics of the characteristic vectors of the correction areas, and obtaining the clustering distance between the two corresponding communication areas.
In the embodiment of the invention, considering that normal cells may be affected by cytoplasmic flow, a certain aggregation exists in the normal cells, so that a communicating region of the normal cells is also composed of a plurality of cells, the shape and the size characteristics of the communicating region are affected, and the normal parameters of the initial morphology have larger errors, and the characteristic vector of the initial region has certain errors, so that the characteristic vector of the initial region is required to be corrected.
The cells of breast duct in-situ cancer are influenced by secretion of factors, hormone and the like of lesion cells, so that the lesion cells have influence on surrounding cells, normal cells are converted into lesion cells, and the surrounding of abnormal cells is also abnormal cells; meanwhile, cancer cells of breast duct in-situ cancer proliferate in the breast duct, do not invade surrounding tissues or blood vessels, so that the mobility is poor, the cells are gathered in a certain area, so that the cells have similarity in surrounding cells, and the initial area feature vector of each communication area can be corrected according to the similar features of the initial area feature vectors of other communication areas around each communication area, and the corrected area feature vector of each communication area can be obtained.
Preferably, in one embodiment of the present invention, considering that the connected regions have respective shapes and sizes, it is difficult to obtain adjacent connected regions by establishing a neighborhood region, so by determining the number of neighborhood connected regions first, the nearest preset neighborhood parameter connected regions are selected as the adjacent connected regions; considering that the smaller the difference between the initial region feature vector of the adjacent communication region and the average value of the initial region feature vectors of all the adjacent communication regions corresponding to the same communication region, the smaller the deviation degree of the initial region feature vector of the adjacent communication region is, the higher the similarity degree is, and the stronger the reference property of the initial form normal parameters of the adjacent communication region is; meanwhile, the closer the distance between the communication area and the adjacent communication area is, the more likely the communication area is the similar cell area, and the stronger the reference is, so the correction is carried out by utilizing the distance characteristic, the normalization is carried out by utilizing the sum of all reference degree parameters, and the reference of the adjacent communication area is unified;
Based on this, the minimum euclidean distance between the pixel points between the two communication areas is taken as the reference distance between the two corresponding communication areas; determining other communication areas with preset neighborhood parameters, the reference distance between the other communication areas and any communication area is the smallest, and using the other communication areas as adjacent communication areas of the corresponding communication areas;
according to similar features between initial region feature vectors of adjacent communication regions corresponding to each communication region, combining corresponding reference distances to obtain reference degree parameters of each adjacent communication region corresponding to each communication region;
Acquiring the reference degree weight of each adjacent communication area corresponding to each communication area according to the proportion of the reference degree parameter of each adjacent communication area corresponding to each communication area occupying all the reference degree parameters;
According to the reference degree weight of each adjacent communication region corresponding to each communication region, the initial morphological normal parameters of each adjacent communication region are combined, and the initial morphological normal parameters of each communication region are corrected to obtain corrected morphological normal parameters;
And replacing the corresponding initial morphological normal parameters by using the corrected morphological normal parameters to obtain corrected region feature vectors of each connected region.
The calculation formula for correcting the morphological normal parameters comprises:
；
Wherein, The serial number of the connected area; /(I)Represents the/>Initial morphological normal parameters of the individual connected regions; /(I)Represents the/>Correcting morphological normal parameters of the connected areas; /(I)Representing preset neighborhood parameters; /(I)A serial number indicating an adjacent communication area; Represents the/> First/>, corresponding to each communication regionInitial morphological normal parameters of the adjacent connected regions; /(I)Represents the/>First/>, corresponding to each communication regionReference degree weight of each adjacent connected region,/>，Represents the/>Average value of initial region feature vectors of all adjacent connected regions corresponding to each connected region,/>Represents the/>First/>, corresponding to each communication regionInitial region feature vectors of adjacent connected regions; /(I)A computation function representing a binary norm of the vector; /(I)Represents the/>The communicating region and the corresponding first/>Reference distance between adjacent connected areas.
In the calculation formula for correcting the morphological normal parameters,The smaller the difference between the initial region feature vector of the adjacent connected region and the average value of the initial region feature vectors of all the adjacent connected regions corresponding to the same connected region is, the higher the similarity is, and the stronger the initial form normal parameter referential of the adjacent connected region is; /(I)The smaller the distance between the communicating region and the adjacent communicating region is, the more likely the communicating region is a similar cell region, and the stronger the reference property is; The larger the reference of the adjacent communication area is, the stronger the reference of the adjacent communication area is, the larger the occupied proportion is, and the weight is larger; /(I) The overall characteristics of the normal parameters of the initial morphology of other communication areas around the communication area are represented, when the overall normal parameters of the initial morphology of the surrounding communication areas are larger, the analysis from the aspect of morphological characteristics is shown that surrounding cells are more likely to be normal cells, the cells of the communication areas are also more likely to be time-normal cells, and the normal parameters of the initial morphology need to be increased; conversely, a reduction is required; and correcting the normal parameters of the initial form through the local similarity of the cell distribution to obtain a more accurate corrected region feature vector, so as to prepare for the subsequent accurate clustering distance.
It should be noted that the two norms of the vector are well known to those skilled in the art, and in other embodiments of the present invention, the practitioner may also employOther basic mathematical operations or function mapping may be used to implement the correlation mapping, which are all technical means known to those skilled in the art, and are not described herein.
In the embodiment of the invention, considering that the normal characteristic or the abnormal characteristic is not very obvious in the hierarchical clustering process, because the characteristic vectors of the correction region are closer in space, the communication regions with the characteristics not very obvious in the hierarchical clustering process are preferentially clustered, cells which are actually expressed as different categories are clustered into one category, and an inaccurate clustering result is obtained, so that the regions with the normal characteristic or the abnormal characteristic which are obvious in the characteristic vectors of the correction region are preferentially clustered, and when the clustering distance is calculated, the clustering distance between the two corresponding communication regions can be obtained by combining the spatial distribution characteristic of the characteristic vectors of the correction region besides the difference characteristic of the characteristic vectors of the correction region between any two communication regions.
Preferably, in the embodiment of the present invention, the larger the difference of the feature vectors of the correction regions between the two connected regions is considered, the less likely the cell categories in the two connected regions are the same, and the larger the clustering distance is; taking into account the correction region feature vectorOr/>The smaller the difference, the more obvious the abnormal or normal features of cells in the connected region, the more the preferential clustering is needed, so the correction region feature vector and/>, are utilizedOr/>Supplementing the minimum value of the difference of the two groups to obtain more accurate clustering distance;
Based on this, a second norm of the difference of the correction region feature vector between the two connected regions is taken as a first distance parameter between the two connected regions;
acquiring a second distance parameter between two connected areas by using a calculation formula of the second distance parameter;
taking the product of the first distance parameter and the second distance parameter between any two connected areas as the clustering distance between the two corresponding connected areas.
The calculation formula of the clustering distance comprises:
；
Wherein, Represents the/>The communicating region and the/>Cluster distance between the connected areas; /(I)Represents the/>The communicating region and the/>First distance parameter between connected regions,/>；/>Represents the/>The communicating region and the/>The calculation formula of the second distance parameter corresponding to each communication area is as follows:；/> For obtaining a minimum value in the set; /(I) A serial number indicating the connected region; /(I)A corrected region feature vector representing an h-th connected region; /(I)A computation function representing the two norms of the vector.
In a calculation formula of the clustering distance, the smaller the first distance parameter is, the smaller the difference of the feature vectors of the correction areas between the two connected areas is; the smaller the second distance parameter is, the more obvious the cell characteristics expressed by the characteristic vectors of the correction areas of the two connected areas are, the smaller the clustering distance is only when the first distance parameter and the second distance parameter are smaller at the same time, the two connected areas are stated to meet the two conditions of obvious area characteristics and similar area characteristics at the same time, and therefore clustering is more accurate.
Since the clustering distance between two connected regions is obtained, the clustering distance is calculated; In other embodiments of the present invention, other basic mathematical operations or function mapping may be used to implement the relevant mapping, which are all technical means well known to those skilled in the art, and are not described herein.
Step S4: and carrying out hierarchical clustering on all the connected areas according to the clustering distance to obtain a hierarchical clustering lineage diagram of the tumor pathological tissue image.
After the clustering distance between the connected areas is obtained, hierarchical clustering can be performed on all the connected areas, and the areas are mainly required to be divided into normal areas and lesion areas, so that the number of final categories of the clustering is 2, and the feature vector of the corrected area is closest toThe connected areas of the breast duct in situ cancer are marked to obtain hierarchical clustering pedigree diagrams of tumor pathological tissue images, a doctor is assisted to rapidly and accurately identify pathological change conditions of the breast duct in situ cancer, comprehensive judgment is carried out by combining clinical experience and other examination results, and the accuracy of diagnosis and operation is improved.
Referring to fig. 3 and fig. 4, fig. 3 is a hierarchical clustering lineage diagram obtained by a conventional hierarchical clustering method according to an embodiment of the present invention; fig. 4 is a hierarchical clustering lineage diagram obtained by an optimized hierarchical clustering method according to an embodiment of the present invention, and as can be seen by comparing fig. 3 and fig. 4, the result obtained by the conventional hierarchical clustering method is interfered by a connected region composed of normal cells, and a large number of normal cells with aggregation are determined as abnormal cells.
It should be noted that hierarchical clustering algorithms are well known to those skilled in the art, and will not be described herein.
In summary, in order to solve the technical problem that the traditional hierarchical clustering method is not ideal for processing the tumor pathological tissue image, the invention provides an artificial intelligent processing method for the tumor pathological tissue image. Firstly, acquiring a tumor pathological tissue image of breast duct carcinoma in situ and acquiring a communication region in the tumor pathological tissue image; further analyzing the shape characteristics and the size characteristics of each connected region, and combining the discrete characteristics of the gray values of the pixels in each connected region to obtain an initial region characteristic vector; correcting the initial region feature vector of each connected region according to similar features of the initial region feature vectors of other connected regions with the nearest preset neighborhood parameters to each connected region, and obtaining corrected region feature vectors of each connected region; further according to the difference characteristics of the characteristic vectors of the correction areas between any two communication areas, combining the spatial distribution characteristics of the characteristic vectors of the correction areas to obtain the clustering distance between the two corresponding communication areas; and finally, hierarchical clustering is carried out on all the connected areas according to the clustering distance, and a hierarchical clustering pedigree diagram of the tumor pathological tissue image is obtained. According to the embodiment of the invention, the accurate clustering distance is determined through the morphological characteristics, the gray level characteristics and the local similar characteristics of the region, the influence caused by normal cell aggregation is reduced, a more accurate hierarchical clustering lineage diagram is obtained, and the analysis of related personnel is facilitated.
It should be noted that: the 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 accompanying drawings do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
In this specification, each embodiment is described in a progressive manner, and identical and similar parts of each embodiment are all referred to each other, and each embodiment mainly describes differences from other embodiments.
Claims (10)
1. An artificial intelligence processing method of tumor pathological tissue images, which is characterized by comprising the following steps:
Acquiring a tumor pathological tissue image of breast duct carcinoma in situ; threshold segmentation is carried out on the tumor pathological tissue image, and a communication area in the tumor pathological tissue image is obtained;
analyzing the shape characteristics and the size characteristics of each communication area to obtain the initial morphological normal parameters of each communication area; acquiring gray level normal degree parameters according to the discrete characteristics of gray level values of pixel points in each communication area; constructing an initial region feature vector according to the initial form normal parameter and the gray level normal degree parameter of each connected region;
Obtaining a corrected region feature vector of each connected region according to similar features of the initial region feature vectors of the other connected regions of the preset neighborhood parameters closest to each connected region; according to the difference characteristics of the characteristic vectors of the correction areas between any two communication areas, combining the spatial distribution characteristics of the characteristic vectors of the correction areas to obtain the clustering distance between the two corresponding communication areas;
and carrying out hierarchical clustering on all the connected areas according to the clustering distance to obtain the hierarchical clustering lineage diagram of the tumor pathological tissue image.
2. The artificial intelligence processing method of tumor pathological tissue image according to claim 1, wherein the method for acquiring the initial morphological normal parameters comprises the following steps:
taking the number of pixel points in each communication area as the size characteristic parameter of each communication area;
taking the square of the number of boundary pixel points of each communication area as the perimeter characteristic parameter of each communication area; taking the ratio of the number of pixel points in each communication area to the corresponding perimeter characteristic parameter as the shape index of each communication area; the pixel points in the communication area comprise boundary pixel points;
Acquiring shape characteristic parameters of each communication region according to the difference characteristics of the shape index of each communication region and the shape index of the standard circle;
Combining the size characteristic parameters and the shape characteristic parameters of each communication area to obtain initial morphological normal parameters of each communication area; the size characteristic parameter is inversely related to the initial morphological normal parameter; the shape characteristic parameter is positively correlated with the initial morphology normal parameter; and normalizing the initial morphological normal parameters.
3. The artificial intelligence processing method of tumor pathological tissue image according to claim 2, wherein the method for obtaining the shape characteristic parameter comprises the following steps:
Associating the shape index of each of the communication areas with And (3) carrying out negative correlation mapping and normalization on the absolute value of the difference value of the connected areas to serve as a shape characteristic parameter of each connected area.
4. The artificial intelligence processing method of tumor pathological tissue image according to claim 2, wherein the method for acquiring the initial morphological normal parameters comprises the following steps:
And normalizing the ratio of the shape characteristic parameter to the size characteristic parameter of each communication region to obtain an initial morphological normal parameter of each communication region.
5. The artificial intelligence processing method of tumor pathological tissue image according to claim 1, wherein the method for acquiring the gray level normal degree parameter comprises the following steps:
acquiring the variance of the gray value of the pixel point in each communication area as a first discrete characteristic parameter;
Acquiring a quarter bit distance of gray values of pixel points in each communication area as a second discrete characteristic parameter;
Acquiring gray level normal degree parameters corresponding to each communication region according to the first discrete characteristic parameters and the second discrete characteristic parameters corresponding to each communication region; the first discrete characteristic parameter and the second discrete characteristic parameter are inversely related to the gray level normal degree parameter; and the gray level normal degree parameters are subjected to normalization processing.
6. The artificial intelligence processing method of tumor pathological tissue image according to claim 5, wherein the method for acquiring the gray scale normal degree parameter comprises the following steps:
And carrying out negative correlation mapping and normalization on the product of the first discrete characteristic parameter and the second discrete characteristic parameter corresponding to each communication region, and then taking the product as a gray level normal degree parameter corresponding to each communication region.
7. The artificial intelligence processing method of a tumor pathological tissue image according to claim 1, wherein the method for acquiring the feature vector of the correction area comprises the following steps:
taking the minimum Euclidean distance between the pixel points between the two communication areas as a reference distance between the two corresponding communication areas; determining other communication areas with preset neighborhood parameters, the reference distance between the other communication areas and any communication area is the smallest, and using the other communication areas as adjacent communication areas of the corresponding communication areas;
According to the similar features between the initial region feature vectors of the adjacent communication regions corresponding to each communication region, combining the corresponding reference distances to obtain reference degree parameters of each adjacent communication region corresponding to each communication region;
Acquiring the reference degree weight of each adjacent communication area corresponding to each communication area according to the proportion of the reference degree parameter of each adjacent communication area corresponding to each communication area occupying all the reference degree parameters;
According to the reference degree weight of each adjacent communication region corresponding to each communication region, the initial morphological normal parameters of each adjacent communication region are combined, and the initial morphological normal parameters of each communication region are corrected to obtain corrected morphological normal parameters;
And replacing the corresponding initial morphological normal parameters by using the corrected morphological normal parameters to obtain corrected region feature vectors of each connected region.
8. The artificial intelligence processing method of tumor pathological tissue image according to claim 7, wherein the method for obtaining the corrected morphological normal parameters comprises the following steps:
; wherein/> The serial number of the connected area; /(I)Represents the/>Initial morphological normal parameters of the individual connected regions; /(I)Represents the/>Correcting morphological normal parameters of the connected areas; /(I)Representing preset neighborhood parameters; /(I)A serial number indicating an adjacent communication area; /(I)Represents the/>First/>, corresponding to each communication regionInitial morphological normal parameters of the adjacent connected regions; /(I)Represents the/>First/>, corresponding to each communication regionThe reference degree weights of the adjacent connected regions,，/>Represents the/>Average value of initial region feature vectors of all adjacent connected regions corresponding to each connected region,/>Represents the/>First/>, corresponding to each communication regionInitial region feature vectors of adjacent connected regions; /(I)A computation function representing a binary norm of the vector; /(I)Represents the/>The communicating region and the corresponding first/>Reference distance between adjacent connected areas.
9. The artificial intelligence processing method of tumor pathological tissue image according to claim 1, wherein the clustering distance obtaining method comprises the following steps:
Taking a second norm of the difference of the correction region feature vectors between two connected regions as a first distance parameter between the two connected regions;
acquiring a second distance parameter between two connected areas by using a calculation formula of the second distance parameter;
Taking the product of the first distance parameter and the second distance parameter between any two connected areas as the clustering distance between the two corresponding connected areas.
10. The artificial intelligence processing method of tumor pathological tissue image according to claim 9, wherein the calculation formula of the second distance parameter comprises: ; wherein, Represents the/>The communicating region and the/>Second distance parameters corresponding to the communication areas; /(I)For obtaining a minimum value in the set; /(I)A serial number indicating the connected region; /(I)A corrected region feature vector representing an h-th connected region; /(I)A computation function representing the two norms of the vector.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410598674.XA CN118196789A (en) | 2024-05-15 | 2024-05-15 | Artificial intelligence processing method of tumor pathological tissue image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410598674.XA CN118196789A (en) | 2024-05-15 | 2024-05-15 | Artificial intelligence processing method of tumor pathological tissue image |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN118196789A true CN118196789A (en) | 2024-06-14 |
Family
ID=91399020
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410598674.XA Pending CN118196789A (en) | 2024-05-15 | 2024-05-15 | Artificial intelligence processing method of tumor pathological tissue image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN118196789A (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140072213A1 (en) * | 2012-09-13 | 2014-03-13 | Los Alamos National Security, Llc | Object detection approach using generative sparse, hierarchical networks with top-down and lateral connections for combining texture/color detection and shape/contour detection |
| US20190042826A1 (en) * | 2017-08-04 | 2019-02-07 | Oregon Health & Science University | Automatic nuclei segmentation in histopathology images |
| CN109685767A (en) * | 2018-11-26 | 2019-04-26 | 西北工业大学 | A kind of bimodal brain tumor MRI dividing method based on Cluster-Fusion algorithm |
| CN116523802A (en) * | 2023-07-04 | 2023-08-01 | 天津大学 | Enhancement optimization method for liver ultrasonic image |
-
2024
- 2024-05-15 CN CN202410598674.XA patent/CN118196789A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140072213A1 (en) * | 2012-09-13 | 2014-03-13 | Los Alamos National Security, Llc | Object detection approach using generative sparse, hierarchical networks with top-down and lateral connections for combining texture/color detection and shape/contour detection |
| US20190042826A1 (en) * | 2017-08-04 | 2019-02-07 | Oregon Health & Science University | Automatic nuclei segmentation in histopathology images |
| CN109685767A (en) * | 2018-11-26 | 2019-04-26 | 西北工业大学 | A kind of bimodal brain tumor MRI dividing method based on Cluster-Fusion algorithm |
| CN116523802A (en) * | 2023-07-04 | 2023-08-01 | 天津大学 | Enhancement optimization method for liver ultrasonic image |
Non-Patent Citations (1)
| Title |
|---|
| 刘腾飞等: "基于Matlab的肿瘤细胞识别系统", 《电子设计工程》, vol. 29, no. 6, 31 December 2021 (2021-12-31), pages 1 - 5 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106204587B (en) | Multiple organ dividing method based on depth convolutional neural networks and region-competitive model | |
| CN104732213B (en) | A kind of area of computer aided Mass detection method based on mammary gland magnetic resonance image | |
| Belsare et al. | Histopathological image analysis using image processing techniques: An overview | |
| Tosta et al. | Segmentation methods of H&E-stained histological images of lymphoma: A review | |
| CN109064470B (en) | Image segmentation method and device based on self-adaptive fuzzy clustering | |
| EP4075380B1 (en) | Medical image-based tumor detection and diagnostic device | |
| Giannini et al. | A fully automatic algorithm for segmentation of the breasts in DCE-MR images | |
| CN115345893B (en) | Ovarian tissue canceration region segmentation method based on image processing | |
| US20030169915A1 (en) | Abnormal shadow detecting system | |
| CN114820663B (en) | Assistant positioning method for determining radio frequency ablation therapy | |
| Peng et al. | Segmentation of prostatic glands in histology images | |
| CN108765411A (en) | A kind of tumor classification method based on image group | |
| Pashoutan et al. | Automatic breast tumor classification using a level set method and feature extraction in mammography | |
| CN118196789A (en) | Artificial intelligence processing method of tumor pathological tissue image | |
| CN102800090A (en) | Blood cell segmentation method | |
| CN111292285B (en) | Automatic screening method for diabetes mellitus based on naive Bayes and support vector machine | |
| Frackiewicz et al. | Breast lesion segmentation in DCE-MRI Imaging | |
| Bhuiyan et al. | An adaptive region growing segmentation for blood vessel detection from retinal images | |
| Koper et al. | Breast lesion segmentation in DCE-MRI imaging | |
| Kuo et al. | Automated classification of breast carcinoma cell based on image processing and support vector machine | |
| CN117974692B (en) | Ophthalmic medical image processing method based on region growing | |
| Liu et al. | Segmentation of Mammography Images Based on Spectrum Clustering Method | |
| Dutta et al. | A deep convolutional neural network based framework for breast cancer detection | |
| Tay | Algorithms for Tissue Image Analysis using Multifractal Techniques | |
| Selvy et al. | An Improved GA-MILSVM classification approach for diagnosis of breast lesions from stain images |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination |