Quantitative analysis method and system for spatial distribution characteristics of tumor microenvironment cells

Abstract

The invention provides a tumor microenvironment cell spatial distribution characteristic quantitative analysis method and system, and the method comprises the steps: obtaining a pathological section image, and obtaining cell segmentation and classification information in the pathological section image through a pre-trained cell recognition model; obtaining cell space distribution information based on the cell segmentation and classification information; constructing a cell topology network according to the cell space distribution information, and obtaining a first cell space distribution characteristic; constructing a spatial weight matrix, and obtaining a second cell spatial distribution feature; and judging a tumor state by combining the first cell spatial distribution characteristic and the second cell spatial distribution characteristic so as to realize typing of the target object. The method has good cell segmentation and classification precision, fully pays attention to the adjacency relation, the action intensity and the space structure between cells, integrates multi-dimensional space distribution characteristic quantitative analysis, realizes a good tumor microenvironment cell analysis result, is used for accurately realizing the typing of a target object, and has good universality.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image processing, and in particular to a method and system for quantitatively analyzing the spatial distribution characteristics of tumor microenvironment cells. Background Art

[0002] The tumor microenvironment refers to the area between tumor cells and surrounding normal tissues, which is composed of tumor cells, tumor extracellular matrix, tumor-associated fibroblasts, immune cells, etc. The tumor microenvironment is closely related to tumor growth, metastasis, and immune escape.

[0003] Tumor microenvironment is usually used in the diagnosis and treatment of tumors. Traditional tumor microenvironment analysis mainly relies on manual evaluation, but tumor tissue is complex and heterogeneous, and is easily affected by subjectivity, resulting in low efficiency and large errors. The existing technology mostly uses whole-slice image scanning technology, which can improve more accurate cell positioning information, but the analysis of tumor microenvironment is not comprehensive enough, and it is easy to ignore the spatial relationship between cells. It is also difficult to capture the dynamic interactions of cells in the tumor microenvironment in advance, and there is still a problem of insufficient recognition accuracy when cell nuclei are densely stacked, which ultimately affects the accuracy of tumor microenvironment analysis.

[0004] Therefore, there is a need for a solution to solve at least one of the above problems. Summary of the invention

[0005] In response to the defects in the prior art, the present application proposes a method and system for quantitatively analyzing the spatial distribution characteristics of cells in a tumor microenvironment. In order to solve at least one of the above technical problems, the technical solution of the present application is: a method for quantitatively analyzing the spatial distribution characteristics of cells in a tumor microenvironment, comprising: obtaining a pathological section image, and obtaining cell segmentation and classification information in the pathological section image through a pre-trained cell recognition model; based on the cell segmentation and classification information, obtaining the cell spatial distribution information of the pathological section image; based on the cell spatial distribution information, constructing a cell topological network to obtain a first cell spatial distribution characteristic of the tumor microenvironment; based on the cell spatial distribution information, constructing a spatial weight matrix to obtain a second cell spatial distribution characteristic of the tumor microenvironment; based on the first cell spatial distribution characteristic and the second cell spatial distribution characteristic, judging the tumor state to achieve typing of the target object.

[0006] In a specific embodiment, the cell topological network includes: a first cell space topological network and / or a second cell space topological network; the first cell space topological network is constructed based on the relevant principles of Delaunay triangulation network, and the second cell space topological network is constructed by Voronoi diagram; The first cell space distribution characteristics obtained through the first cell space topological network include: one or any multiple of the link proportion, link edge distance and link frequency of each type of cells; the first cell space distribution characteristics obtained through the second cell space topological network include: one or any multiple of the area, area variance, mean, median and hybridity of each cell pixel.

[0007] In a specific embodiment, the first cell space topology network includes simplices, the union of all simplices forms the convex hull of the cell distribution, and the triangulation of any cell position contains no more than triangles; among them, Indicates the number of cells, represents the spatial dimension, Represents the number of cells in the convex hull; In the second cell space topology network, the Voronoi unit associated with the cell center is defined as: ; in, is a metric space, d is the distance function, K represents the index set that defines the center of a group of cells, that is, the index set of the site, represents a set of non-empty cells in space X, represents the cell center, Represents the Voronoi unit associated with the cell center.

[0008] In a specific embodiment, acquiring the second cell spatial distribution characteristics specifically includes: Based on the spatial weight matrix, the grid adjacency weights after gridding the tumor microenvironment are obtained; the number of cells in each grid is obtained as grid cell number information; and the spatial autocorrelation of the same type of cells in the tumor microenvironment and the spatial autocorrelation of different types of cells in the tumor microenvironment are analyzed by combining the grid adjacency weights with the grid cell number information.

[0009] In a specific embodiment, the spatial autocorrelation of the same type of cells in the tumor microenvironment is analyzed based on the univariate Moran's index; the calculation formula of the univariate Moran's index is: ; in, represents the number of grids after the tumor microenvironment is gridded. represents the adjacency weight between grid i and grid j, and Respectively represent the grid and Grid The number of cells of a certain type in It represents the mean number of cells of a certain type in all grids; The spatial autocorrelation between different types of cells in the tumor microenvironment was analyzed based on the bivariate Moran's index; the calculation formula of the bivariate Moran's index is: ; in, represents the number of grids after the tumor microenvironment is gridded. Indicates The number of cells of the first type in a grid, Indicates The number of cells of the second type in a grid, is an element in the weight matrix, representing the adjacency weight between grids. represents the mean of the first type of cells, ; represents the mean of the second type of cells, ; represents the spatial hysteresis value, .

[0010] In a specific embodiment, based on the first cell spatial distribution feature and the second cell spatial distribution feature, the tumor status is judged to classify the target object, specifically including: selecting preset classification feature parameters from the first cell spatial distribution feature and the second cell spatial distribution feature respectively; the preset classification feature parameters include: the chain edge distance between tumor cells and immune cells, the standard deviation of the chain edge distance between matrix cells and matrix cells, the variance of the chain edge distance between matrix cells and matrix cells, the number of chain edges between tumor cells and immune cells, the number of triangles in the triangular network, the number of polygons, the standard deviation of the chain edge distance between tumor cells and immune cells, and the variance of the chain edge distance between tumor cells and immune cells, one or any more; obtaining clinical information of the target object; combining the clinical information of the target object, the association established between the classification feature parameters and the life cycle of the target object, to judge the tumor status to achieve classification of the target object; the tumor status includes: one or any more of being in situ, locally extended and metastatic.

[0011] In a specific embodiment, the selection of preset parameter features specifically includes: performing KM survival analysis and Cox risk ratio analysis based on the overall survival and non-progression survival of the target object; judging whether each characteristic parameter in the first cell spatial distribution feature and the second cell spatial distribution feature meets the preset conditions; the preset conditions include one or more of the logrank-p value and Cox-p value of each characteristic parameter corresponding to the overall survival and the non-progression survival; if the preset conditions are met, it is used as the typing characteristic parameter.

[0012] In a specific embodiment, the acquisition of cell segmentation and classification information specifically includes: segmenting the pathological slice image according to preset specifications to obtain pathological slice sub-images; identifying cells in each of the pathological slice sub-images through the pre-trained cell recognition model to obtain the cell segmentation and classification information; the acquisition of cell spatial distribution information specifically includes: restoring the pathological slice image by in situ backtracking each of the pathological slice sub-images; and obtaining the cell spatial distribution information of the pathological slice image based on the restored pathological slice image and the cell segmentation and classification information.

[0013] In a specific embodiment, the cell recognition model is trained based on the STARDIST model, specifically including: obtaining and preprocessing a training sample set, the sample set including multiple training samples; the input data including: training images in the multiple training samples; the output data including: the number of each cell classification and each type of cell in the training image; using the input data and the output data to train a deep learning model and optimize the parameter settings of the deep learning model. If the each cell classification and the number of each type of cell in the training image output by the deep learning model meet the preset goals, the training is terminated to obtain the cell recognition model.

[0014] A quantitative analysis system for the spatial distribution characteristics of cells in a tumor microenvironment is applied to any of the aforementioned quantitative analysis methods for the spatial distribution characteristics of cells in a tumor microenvironment, comprising: a cell recognition and segmentation module: used to obtain a pathological section image, and obtain information about cell segmentation and classification in the pathological section image through a pre-trained cell recognition model; a distribution information acquisition module: used to obtain the cell spatial distribution information of the pathological section image based on the cell segmentation and classification information; a first feature acquisition module: used to construct a cell topology network based on the cell spatial distribution information, and obtain a first cell spatial distribution characteristic of the tumor microenvironment; a second feature acquisition module: used to construct a spatial weight matrix based on the cell spatial distribution information, and obtain a second cell spatial distribution characteristic of the tumor microenvironment; a judgment and typing module: used to judge the tumor state based on the first cell spatial distribution characteristic and the second cell spatial distribution characteristic, so as to achieve typing of the target object.

[0015] Beneficial effects: The present invention provides a method for quantitatively analyzing the spatial distribution characteristics of tumor microenvironment cells, which has good cell segmentation and classification accuracy, fully pays attention to the adjacency relationship, interaction intensity and spatial structure between cells, and comprehensively analyzes the spatial distribution characteristics of multiple dimensions, thereby achieving good tumor microenvironment cell analysis results, which can be used to accurately realize the typing of target objects and has good versatility; specifically, identifying, segmenting and classifying cells through a preset cell recognition model can effectively handle the overlap between cells, accurately locate the position of each cell, and provide more accurate cell morphology, thereby improving the cell segmentation accuracy; constructing a cell topological network, obtaining the first cell spatial distribution characteristics, and being able to accurately classify the target objects at the cell scale The adjacency relationship and interaction between cells in the tumor microenvironment are accurately modeled, thereby comprehensively revealing the spatial structure of the tumor microenvironment and increasing the spatial dimension of the analysis of the tumor microenvironment; a spatial weight matrix is ​​constructed, and the univariate Moran's index and the bivariate Moran's index are combined to obtain the second cell spatial distribution characteristics, quantify the spatial distribution patterns between different cell types in the tumor microenvironment, and reveal the spatial aggregation or dispersion trends between cell types. The univariate Moran's index can be used to analyze the spatial autocorrelation of similar cells, while the bivariate Moran's index is helpful to analyze the spatial interactions between different cell types, providing a new perspective for the functional study of the tumor microenvironment and thus helping to capture the dynamic changes of cells in advance. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0017] Figure 1 It is a flow chart of the method for quantitatively analyzing the spatial distribution characteristics of tumor microenvironment cells in an embodiment; Figure 2 It is a composition diagram of the quantitative analysis system of the spatial distribution characteristics of tumor microenvironment cells in the embodiment; Figure 3 is an example diagram of a first cell space topology network of an embodiment; Figure 4 is an example diagram of a second cell space topology network of an embodiment; Figure 5 The KM curve diagram of the standard deviation of the chain edge distance between tumor cells and immune cells in the embodiment; Figure 6 is a KM curve diagram of the standard deviation of the chain edge distance between the matrix cells and the matrix cells in the embodiment; Figure 7The recognition effect of the cell recognition model of the embodiment under different thresholds Figure 1 ; Figure 8 The recognition effect of the cell recognition model of the embodiment at different thresholds Figure 2 ; Fig. 9 This is a graph showing the results of the multi-factor Cox regression analysis of this embodiment; Fig.10 is an example diagram of training samples of the input cell recognition model of this embodiment; Fig.11 This is a recognition effect diagram of the cell recognition model cell output in this embodiment.

[0018] Reference numerals: 1-Cell recognition and segmentation module; 2-Distribution information acquisition module; 3-First feature acquisition module; 4-Second feature acquisition module; 5-Judgment and classification module. DETAILED DESCRIPTION

[0019] Hereinafter, various embodiments of the present disclosure will be described more fully. The present disclosure may have various embodiments, and adjustments and changes may be made therein. However, it should be understood that there is no intention to limit the various embodiments of the present disclosure to the specific embodiments disclosed herein, but the present disclosure should be understood to cover all adjustments, equivalents and / or alternatives falling within the spirit and scope of the various embodiments of the present disclosure. Hereinafter, the terms "including" or "may include" that can be used in various embodiments of the present disclosure indicate the existence of the disclosed functions, operations or elements, and do not limit the increase of one or more functions, operations or elements. In addition, as used in various embodiments of the present disclosure, the terms "including", "having" and their cognates are only intended to represent a specific feature, number, step, operation, element, assembly or combination of the foregoing items, and should not be understood as first excluding the existence of one or more other features, numbers, steps, operations, elements, assemblies or combinations of the foregoing items or increasing the possibility of one or more features, numbers, steps, operations, elements, assemblies or combinations of the foregoing items. In various embodiments of the present disclosure, the expression "or" or "at least one of A or / and B" includes any combination or all combinations of the words listed at the same time. For example, the expression "A or B" or "at least one of A or / and B" may include A, may include B, or may include both A and B. The expressions used in the various embodiments of the present disclosure (such as "first", "second", etc.) may modify the various constituent elements in the various embodiments, but may not limit the corresponding constituent elements. For example, the above expressions do not limit the order and / or importance of the elements. The above expressions are only used for the purpose of distinguishing one element from other elements. For example, the first user device and the second user device indicate different user devices, although both are user devices. For example, without departing from the scope of the various embodiments of the present disclosure, the first element may be referred to as the second element, and similarly, the second element may also be referred to as the first element. It should be noted that if it is described that one constituent element is "connected" to another constituent element, the first constituent element may be directly connected to the second constituent element, and a third constituent element may be "connected" between the first constituent element and the second constituent element. Conversely, when one constituent element is "directly connected" to another constituent element, it can be understood that there is no third constituent element between the first constituent element and the second constituent element. The term "user" used in various embodiments of the present disclosure may indicate a person using an electronic device or a device using an electronic device (e.g., an artificial intelligence electronic device). The terms used in various embodiments of the present disclosure are only used for the purpose of describing specific embodiments and are not intended to limit the various embodiments of the present disclosure. As used herein, the singular form is intended to also include the plural form, unless the context clearly indicates otherwise. Unless otherwise defined, all terms used herein (including technical terms and scientific terms) have the same meaning as those generally understood by a person of ordinary skill in the art to which the various embodiments of the present disclosure belong.The terms, such as those defined in generally used dictionaries, will be interpreted as having the same meaning as the contextual meaning in the relevant technical field and will not be interpreted as having an idealized meaning or an overly formal meaning unless clearly defined in various embodiments of the present disclosure.

[0020] Example The present application embodiment provides a method for quantitatively analyzing the spatial distribution characteristics of tumor microenvironment cells, such as Figure 1 Shown include: S100: Acquire a pathological section image, and obtain cell segmentation and classification information in the pathological section image through a pre-trained cell recognition model; Optionally, the cell recognition model can be based on the VGGNet model architecture, the InceptionNet model architecture, the HD-Yolo model architecture, the Mask-RCNN model architecture or the Transformer model architecture, or can be trained based on one or any combination of the FCN model architecture or the RetinaNet model architecture; of course, there is no restriction on the specific model architecture, as long as it meets the requirements of cell segmentation and classification recognition; specifically, in some embodiments of the present application, the cell recognition model is trained based on the STARDIST deep learning model, abandoning the traditional bounding box-based segmentation method, and using a star-shaped convex polygon parameterized by radial distance to represent the shape of the cell nucleus, thereby improving the accuracy of cell classification and segmentation.

[0021] S200: Based on the cell segmentation and classification information, the cell spatial distribution information of the pathological slice image is obtained; in some embodiments of the present application, the cell segmentation and classification information mainly includes the morphology and range division or segmentation of each cell after the cell recognition pattern recognition, and the category judgment of the divided or segmented cells, such as tumor cells, matrix cells, etc.

[0022] S300: Based on the cell spatial distribution information, a cell topological network is constructed to obtain a first cell spatial distribution characteristic of the tumor microenvironment; in some embodiments of the present application, the cell spatial distribution information mainly refers to the number, distribution position, relative distance, etc. of cells in space.

[0023] S400: Based on the cell spatial distribution information, a spatial weight matrix is ​​constructed to obtain the second cell spatial distribution characteristics of the tumor microenvironment; Among them, the first cell spatial distribution feature is obtained based on the cell topological network, focusing on the adjacency relationship, interaction strength and spatial topological structure between local individual cells, and extracting the cell spatial distribution in the tumor microenvironment as the cell spatial distribution feature in a numerical quantitative way, revealing the spatial organization pattern between cells, quantifying the geometric relationship and interaction strength between cells, and making up for the shortcomings of the existing technology in the analysis of spatial relationships between cells; the second cell spatial distribution feature is obtained based on the spatial weight matrix, focusing on the local overall cell spatial position distribution relationship in the tumor microenvironment, such as the spatial aggregation or dispersion trend between the same type of cells, the spatial interaction relationship between different types of cells, and the cell spatial position distribution relationship is embodied in a numerical form, realizing the effective characterization of the spatial distribution pattern of different cell types in the tumor microenvironment, and revealing the relationship between cell spatial distribution and tumor progression and treatment response. Cell types can specifically include: tumor cells, immune cells and matrix cells.

[0024] S500: Based on the first cell spatial distribution feature and the second cell spatial distribution feature, determine the tumor state to achieve typing of the target object.

[0025] The first cell spatial distribution feature and the second cell spatial distribution feature are combined to enrich the dimension of the analysis. On the basis of the morphology, number and position of the original cells, the adjacency relationship, interaction intensity and distribution trend between cells are quantitatively processed to accurately quantify the geometric relationship and spatial interaction between cells, thereby realizing a multi-dimensional quantitative analysis of the spatial distribution of cells in the tumor microenvironment, which helps to obtain more comprehensive cell distribution information, enhance the understanding of the tumor microenvironment, and facilitate the understanding of the role of cell-cell interactions in the tumor microenvironment in tumor development, immune escape and metastasis, so as to improve the accuracy of the analysis results, help to predict the dynamic effects and development of cells in the tumor microenvironment, and thus improve the accuracy of the target object classification. Specifically, in some embodiments of the present application, the pathological slice image is derived from the target object, and the target object can be a patient.

[0026] Further, the cell topological network includes: a first cell space topological network and / or a second cell space topological network; The first cell space topology network is constructed based on the relevant principles of Delaunay triangulation network, and the second cell space topology network is constructed through Voronoi diagram; specifically, the relevant principles of Delaunay triangulation network include: Delaunay criterion, using geometric conditions to determine whether a pair of adjacent triangles represents the optimal connection option. If a pair of triangles does not meet the criterion, the connection method will be readjusted; Delaunay triangulation, dividing its convex hull into several triangles, requiring that the circumscribed circles of these triangles do not contain any other points. This process maximizes the size of the minimum angle in each triangle and tends to avoid triangles that are too oblong; Delaunay triangulation is not unique, and the two possible divisions of the quadrilateral into two triangles both meet the "Delaunay condition", that is, the interior of the circumscribed circles of all triangles is empty. More specifically, the specific process of constructing the first cell space topology network includes: by obtaining the centroid of each cell nucleus, the spatial topology network is constructed using the Delaunay triangulation algorithm, in which each cell nucleus is connected to its adjacent cell nucleus through edges; through the relevant principles of Delaunay triangulation network, an optimal network structure is provided for the interaction and transmission between cells, such as Figure 3 shown.

[0027] Specifically, the Voronoi diagram is a partition of a plane into regions close to a given set of objects, which can also be classified as a tessellation; these objects are just a finite number of points in the plane (called seeds, sites, or generators); for each seed, there is a corresponding region, called a Voronoi cell, containing all points in the plane that are closer to the seed than to other seeds. That is, the Voronoi diagram divides the space into multiple regions according to the nearest property of the object, and the distance from any point of a convex polygon to an object point in the region contained by the polygon is less than the distance to any other object point; by treating cells as discrete points in space, the Voronoi diagram can effectively partition the cell space so that the proximity and relative position of each cell are clear, such as Figure 4 shown.

[0028] The first cell space distribution characteristics obtained through the first cell space topological network include: one or any multiple of the proportion of each type of cell links, link edge distance and link frequency; The first cell space distribution characteristics obtained through the second cell space topological network include: one or any multiple of the area, area variance, mean, median and hybridity of each cell pixel.

[0029] By constructing the cell topological network based on the relevant principles of Delaunay triangular network and Voronoi diagram, the differences between the characteristics of Delaunay triangular network and Voronoi diagram are fully utilized to obtain different characteristic parameters such as cell adjacency, interaction strength and spatial topological structure. For example, the Delaunay triangulation network accurately represents the neighboring relationship of cells, optimizes the connection between cells, provides an efficient communication structure and mechanical framework, simplifies the complex spatial structure of cells, and provides a basis for in-depth research on cell group behavior, cell communication, cell mechanics, etc. It is suitable for the analysis of the pathways of intercellular material diffusion and signal propagation links. Another example is the Voronoi diagram, which can help understand the distribution of cells in space and can also be used to calculate the range or influence area of ​​cell interaction, which is suitable for the analysis of tumor immune escape mechanisms; on the basis of identifying the aggregation and dispersion patterns of cells, the functional role and ecological niche of cells are inferred, and the spatial characteristic information such as the area, variance, mean, median, etc. of each cell pixel polygon is calculated, among which the area fluctuation is the difference value of the polygon area.

[0030] Furthermore, the first cell space topology network includes simplices, the union of all simplices forms the convex hull of the cell distribution, and the triangulation of any cell position contains no more than triangles; in, Indicates the number of cells, represents the spatial dimension, represents the number of cells in the convex hull; Simplex is a geometric and mathematical concept, usually used to describe a geometric shape formed by a number of linearly independent points; convex hull refers to the smallest convex polygon that contains a set of points. Specifically, if the obtained is a two-dimensional plane ( ) and the convex hull has cells, then the triangulation of any cell location contains at most triangles, and contains an additional external face. If the cell positions are distributed according to a Poisson process with constant strength, then each cell has, on average, six neighboring triangles; more generally, In the dimensional space, the same Poisson process makes the average number of neighbors of each cell depend only on .

[0031] In cell networks, neighboring cells tend to have a higher probability of interaction, such as signal transmission or material exchange; at the same time, the first cell space topological network follows the principle of "maximizing the minimum angle", avoiding long and narrow triangles, thus forming a geometrically optimal connection structure; in the cell space network, this structure minimizes unnatural long-distance connections, and to a large extent ensures that the connections between cells reflect the true topological relationship.

[0032] In the second cell space topology network, the Voronoi unit associated with the cell center is defined as: ; in, is a metric space, d is the distance function, K represents the index set that defines the center of a group of cells, that is, the index set of the site, represents a set of non-empty cells in space X, represents the cell center, Represents the Voronoi unit associated with the cell center.

[0033] The Voronoi unit contains all those points that are closest to the center of the cell, that is, all points whose distance to the cell on the plane is less than or equal to the distance to the center of other cells. The Voronoi unit of each cell defines a unique region in space, and all points in this region tend to correspond to this cell type; this structure of the second cell spatial topological network can not only help understand the distribution of cells in space, but also be used to calculate the range or influence area of ​​cell interactions, thereby clearly characterizing the interactions and spatial distribution between cells.

[0034] Furthermore, the acquisition of the second cell spatial distribution feature specifically includes: Based on the spatial weight matrix, the grid adjacency weights of the tumor microenvironment after gridding are obtained; Specifically, in some embodiments of the present application, the steps of constructing the spatial weight matrix include the following: grid division, dividing each 2560×2560 pixel pathological image block into 40×40 uniform grids, each grid is 64×64 pixels. Adjacency definition, if two grids share an edge or corner, that is, the Queen adjacency rule, they are defined as spatially adjacent, and the weight =1, otherwise = 0. Row normalization: Normalize the weight matrix so that the sum of the adjacent weights of each grid is 1, that is, =1.

[0035] The number of cells in each grid is obtained as grid cell number information; Combining grid adjacency weights with grid cell number information, the spatial autocorrelation of the same type of cells in the tumor microenvironment and the spatial autocorrelation of different types of cells in the tumor microenvironment were analyzed to quantify the spatial interactions between cell types.

[0036] Through quantitative analysis of the spatial autocorrelation of cells of the same type and between cells of different types, the spatial distribution trend of cells in the tumor microenvironment can be revealed, thereby analyzing the spatial interactions between different types of cells, such as the relationship between tumor cells and immune cells and stromal cells, thereby improving the accuracy of the description of the spatial heterogeneity of the tumor microenvironment and the cell-cell interaction pattern.

[0037] Furthermore, the spatial autocorrelation of the same type of cells in the tumor microenvironment was analyzed based on the univariate Moran's index; the calculation formula of the univariate Moran's index is: ; in, represents the number of grids after the tumor microenvironment is gridded. represents the adjacency weight between grid i and grid j, and Respectively represent the grid and Grid The number of cells of a certain type in represents the mean value of the number of cells of a certain type in all grids; specifically, in some embodiments of the present application, The values ​​of include "0" and "1", "0" means non-adjacent, and "1" means adjacent.

[0038] This index reflects the degree of spatial aggregation or dispersion of a certain type of cells, including the aforementioned tumor cells, matrix cells and immune cells. If the index is close to 1, it means that the cells of this type are highly clustered in space. If the index is close to -1, it means that the cells of this type are highly dispersed in space; A value close to 0 indicates that the distribution of cells of this type is random.

[0039] The spatial autocorrelation between different types of cells in the tumor microenvironment was analyzed based on the bivariate Moran's index; the calculation formula of the bivariate Moran's index is: ; in, represents the number of grids after the tumor microenvironment is gridded. Indicates The number of cells of the first type in a grid, Indicates The number of cells of the second type in a grid, is an element in the weight matrix, representing the adjacency weight between grids. represents the mean of the first type of cells, ; represents the mean of the second type of cells, ; represents the spatial hysteresis value, .

[0040] like The value of is equal to 1, then the grid and Grid If If the value of is equal to 0, the grid and Grid The bivariate Moran index is used to measure the degree of aggregation or dispersion of two types of cells in the tumor microenvironment space. The spatial lag value represents the weighted average of the number of neighboring cell types of a certain grid. For the cross-correlation between the first type of cells (which can be tumor cells) and the second type of cells (which can be immune cells), the spatial lag value is the weighted average of the number of the third type of cells (which can be stromal cells); the weighting is based on the spatial adjacency relationship between the grids, using the Queen adjacency method, that is, when two grids share an edge or corner, they are considered adjacent.

[0041] When calculating the hysteresis value, the weight matrix weights the neighborhood of each grid, the weight of the adjacent grid in the weight matrix is ​​1, and the weight of the non-adjacent grid is 0; based on the above settings, for each grid, the spatial hysteresis value of the third type of cells is calculated, that is, the weighted average of the number of matrix cells of all grids around the grid. Specifically, in some embodiments of the present application, the Moran's index (autocorrelation of similar and heterogeneous cells) of each grid is obtained by grid division; and the results of all valid grids are aggregated to calculate the average Moran's index, that is, the Moran's index of each grid is averaged, and invalid values, that is, cell-free grids or background grids, are skipped.

[0042] If the bivariate Moran index The value of > 0 and close to 1 indicates that the first type of cells and the second type of cells are spatially clustered, that is, these cell types tend to be concentrated together in certain areas, and then there may be an interaction between the first type of cells and the second type of cells; if the bivariate Moran index The value of <0 and close to -1 indicates that the first type of cells and the second type of cells are spatially dispersed, that is, these cell types tend to be far away from each other; if the bivariate Moran index The value of is close to 0: it means that the first type of cells and the second type of cells are randomly distributed in space without significant spatial patterns.

[0043] Based on the above conclusions, it can be seen that, for example, when tumor cells and stromal cells are clustered in space, it may be due to the matrix support required for tumor growth; for another example, when tumor cells and stromal cells are dispersed in space, it may be due to immune response and tumor suppression; thus, it can be used for preliminary judgment of tumor status and tumor microenvironment.

[0044] Further, based on the first cell spatial distribution feature and the second cell spatial distribution feature, the tumor state is judged to classify the target object, specifically including: Selecting preset typing feature parameters from the first cell spatial distribution feature and the second cell spatial distribution feature respectively; Specifically, in some embodiments of the present application, the first cell spatial distribution feature and the second cell spatial distribution feature are set with a total of 115 feature parameters to capture the spatial characteristics of the spatial arrangement and structural characteristics of the tumor microenvironment; the first cell spatial distribution feature is 109 in total, for example, the margins of the topological network obtained based on the spatial positions (i.e., cell positions) of the graph nodes of the first cell space topological network and the second cell space topological network in the graph measurement are 35 in total, which are used to describe the cell chain edge distances of the first cell space topological network; the frequency of the edges is 6 in total, which are used to describe the proportion of cell chain edges in the first cell space topological network; the absolute number of cells is 11 in total , which are used to describe the absolute number of cell chain edges and the absolute number of cells; there are 4 coefficients of variation of cell polygon area (CellCV), which are used to describe the coefficient of variation of the polygonal area of ​​the second cell space topological network; there are 16 cell neighborhood ranges, which are used to describe the average distance (mean, median, variance, standard deviation) between a specific type of cell and all its neighbors; there are 18 cell polygon areas, which are used to describe the polygonal area of ​​the second cell space topological network diagram; there are 3 cell density, which are used to describe cell density; and there are 16 cell diversity (CellDiv) around cells, which are used to describe the cell diversity of all types of cells.

[0045] The second cell spatial distribution feature includes a total of 6 slice Moran indexes; the slices are randomly selected from the WSI (full-view digital slice) of each patient according to the preset number, so the topological features of each patient are calculated based on the corresponding values ​​of all slices selected from the same WSI. Of course, there is no restriction on the selection of feature parameters.

[0046] As another example, the calculation formula for the degree of cell diversity is as follows: ; The target cell of The number of cell types of the nearest neighbor cells, The target cell The number of nearest neighbor cells.

[0047] When the target cells Nearest neighbor cells When they are of different types, = 1, otherwise 0. A cell diversity heterogeneity of 0 (minimum value) indicates that there are no other types of cells as neighbors, while a cell diversity heterogeneity of 1 (maximum value) indicates that there are no cells of the same type as neighbors; this method is used to calculate the cell diversity of each cell type.

[0048] The calculation formula for the variance of cell influence is as follows: ; in, represents the standard deviation of the areas of the Voronoi diagram polygons, Represents the average area of ​​the Voronoi diagram polygons for a particular cell type.

[0049] When the spatial distribution of cells and their neighbors is relatively consistent, the area change of the Voronoi diagram polygon is small and the CellCV value is low, otherwise it is high. When the point set is randomly distributed, the CellCV value is between 33% and 64%, indicating a random distribution; when the point set is clustered, the CellCV value is greater than 64%, indicating a clustered distribution; when the CellCV value is less than 33%, the point set is uniformly distributed; this method was used to calculate the cell influence variance of each cell type.

[0050] The preset classification characteristic parameters include: one or any multiple of the chain edge distance between tumor cells and immune cells, the standard deviation of the chain edge distance between matrix cells and matrix cells, the variance of the chain edge distance between matrix cells and matrix cells, the number of chain edges between tumor cells and immune cells, the number of triangles in the triangular network, the number of polygons, the standard deviation of the chain edge distance between tumor cells and immune cells, and the variance of the chain edge distance between tumor cells and immune cells; The clinical information of the target object is obtained. The clinical information of the target object may include age group information and overall cancer staging information. Of course, there is no limitation on the specific content of the clinical information of the target object. For example, it may also include the target object's emotions, physical condition, etc. The overall cancer staging includes: Stage I, early stage cancer, usually localized, and not spread; Stage II and Stage III, cancer is relatively localized or has spread to adjacent lymph nodes or tissues; Stage IV: cancer has already metastasized to distant sites, usually indicating advanced cancer.

[0051] Combining the association between the target object's clinical information, the classification characteristic parameters and the target object's life cycle, the tumor status is judged to achieve classification of the target object; illustratively, the tumor status is judged to achieve classification of the target object; this can be achieved by constructing a target object classification model; Specifically, based on the influence of the classification characteristic parameters and the clinical information of the target object on the survival period of the target object, the relationship with the survival period of the target object and the tumor status can be established respectively; for example, the characteristic "standard deviation of the chain edge distance of tumor-immunity" is almost positively correlated with the survival period of the target object, and the survival period of the target object with a high standard deviation is significantly higher than that of the target object with a low standard deviation; for example, the characteristic "standard deviation of the matrix-matrix chain edge distance" is also almost positively correlated with the survival period of the target object; at the same time, a large number of training samples are given, and the influence weights and scores of the classification characteristic parameters are set and optimized, such as a positive correlation can be set to a positive value, and a negative correlation can be set to a negative value; More specifically, the influence weights and scores of the classification feature parameters can be adjusted through a large number of training samples until the target accuracy is met. The disease score is calculated based on all the selected classification feature parameters combined with the clinical information of the target object, and is set at the optimized score threshold. The connection between the score threshold and the target object's life cycle and tumor status is established, and then based on the tumor status, the target object is classified; further, based on the historical changes in the target object's disease score, the dynamic state changes of the tumor are judged; the dynamic changes of the tumor include: one or more of tumor development, deterioration and metastasis, and static stability; illustratively, the connection between the disease score change and the dynamic change of the tumor can be established. If the target object's disease score changes slightly, it means that the tumor state is relatively stable; if the target object's disease score drops sharply, it means that the tumor state has deteriorated.

[0052] Furthermore, the selection of preset parameter features specifically includes: KM survival analysis and Cox hazard ratio analysis were performed based on the overall survival and non-progression survival of the target subjects. Among them, KM survival analysis (Kaplan-Meier survival analysis) is a statistical method commonly used in biomedical and clinical research, mainly used to estimate and compare the survival rate or survival time of different groups; Cox hazard ratio model (Cox Proportional Hazards Model) is a regression analysis method commonly used in medicine and biostatistics, mainly used to evaluate the impact of different variables (such as treatment methods, clinical characteristics, gene mutations, etc.) on the survival of cancer patients; Cox model is particularly suitable for survival data analysis, and can model the relationship between patient survival time and a series of predictive factors without assuming data distribution.

[0053] Specifically, in some embodiments of the present application, the analysis of the relationship between 115 spatial parameter features and the typing relationship of the target object is achieved based on the following method: all values ​​of a certain type of spatial characteristic parameters are sorted according to the numerical value, and a cutoff value is selected, and the patients corresponding to each value are grouped, specifically into a high group and a low group, and the clinical follow-up information of TCGA patients (including survival period, survival status, etc.) is combined to perform KM survival analysis and Cox risk ratio analysis to obtain the relationship between the spatial characteristic parameters and patient analysis or clinical prognosis prediction evaluation, thereby facilitating risk prediction and survival outcome prediction.

[0054] Determine whether each characteristic parameter in the first cell spatial distribution feature and the second cell spatial distribution feature meets the preset conditions; the preset conditions include one or more of the logrank-p value and Cox-p value corresponding to the overall survival and non-progression survival of each characteristic parameter; Log-rank test is a test method used to compare whether there are significant differences between two or more groups of survival curves. It is mainly used to compare whether there are significant differences in survival rates between different groups (such as treatment group and control group). The smaller the p value, the more significant the difference in survival curves between groups; Cox regression model (i.e., Cox proportional hazard model) is a regression analysis method used to analyze the impact of multiple factors on survival time. It can evaluate how one or more covariates affect survival time. The smaller the p value, the more significant the impact of the variable on survival time; in this way, it is determined whether the characteristic parameters have an impact on the target object and the size of the impact.

[0055] If it meets the preset conditions, it will be used as a typing characteristic parameter to obtain a more significant analysis effect in KM survival analysis and Cox risk ratio analysis after grouping the target objects under the optimal cutoff. After the p-values ​​obtained by KM and Cox are corrected by FDR correction, a relatively good significance can still be obtained.

[0056] Specifically, in some embodiments of the present application, there are four preset conditions: corresponding to the overall survival analysis: log-rank p value is less than 0.05, Cox-p value is less than 0.05, corresponding to the non-progression survival analysis: log-rank p value is less than 0.05, Cox-p value is less than 0.05; of course, there is no restriction on the setting of the preset conditions. After the FDR correction of each characteristic parameter, as shown in Table 1, the characteristic parameters that meet the preset conditions are the chain edge distance between tumor cells and immune cells, the standard deviation of the chain edge distance between matrix cells and matrix cells, the variance of the chain edge distance between matrix cells and matrix cells, the number of chain edges between tumor cells and immune cells, the number of triangles in the triangular network, the number of polygons, the standard deviation of the chain edge distance between tumor cells and immune cells, and the variance of the chain edge distance between tumor cells and immune cells. Among them, FDR correction is a method for controlling the false positive rate (i.e., the proportion of false rejection of the null hypothesis) in multiple hypothesis testing, which aims to reduce the number of false positive errors when multiple tests are performed simultaneously.

[0057] Table 1 KM and Cox analysis results of typing characteristic parameters

[0058] The close proximity of immune cells to cancer cells may be a signal of high levels of immune cell infiltration in the tumor microenvironment, which means that there are a large number of immune cells infiltrating around the tumor, which usually indicates that the interaction between the tumor and the immune system is relatively active. High levels of immune infiltration are associated with good anti-tumor response and better prognosis, especially in patients who respond to immunotherapy.

[0059] For example, Figure 5 As shown, Figure 5 The content is a survival curve of the "Standard Deviation of Tumor-Immune Chain Distance" based on Kapan-Meier survival analysis. In the survival curve of the "Standard Deviation of Tumor-Immune Chain Distance", patients with high standard deviations tend to have a higher probability of long-term survival. High standard deviations mean that there is a large difference in the distance between immune cells and tumor cells, which may reflect a more complex and heterogeneous immune response in the tumor microenvironment, such as multi-level immune infiltration patterns; low standard deviations indicate that immune cells are more concentrated and evenly distributed, which may indicate that immune cells are concentrated in a specific area of ​​the tumor and cannot effectively penetrate the entire tumor tissue.

[0060] For example, Figure 6 As shown, Figure 6The content is a survival curve of the "standard deviation of the matrix-matrix chain edge distance" based on Kapan-Meier survival analysis. In the survival curve of the "standard deviation of the matrix-matrix chain edge distance", it is shown that the survival probability of the high standard deviation group is significantly higher than that of the low standard deviation group; the high standard deviation means that the distance difference between matrix cells is large and the distribution is uneven. The cell density in some areas is high, while in other areas it is sparse. This high spatial heterogeneity of matrix cells limits the overall progression of the tumor. In contrast, the low standard deviation indicates that the distance difference between matrix cells is small and the cell distribution is more uniform. This uniform cell distribution may provide a more stable support environment for the tumor, promote its progression, and lead to a lower survival probability. Among them, the standard deviation of the distance between matrix cells is closely related to the risk of death of the target object. The distribution and behavior of matrix cells can affect the progression of the tumor and the prognosis of the target object. In addition, the applicant also systematically evaluated whether the typing characteristic parameters can be used independently for the typing of the target object based on multivariate Cox regression analysis. During the experiment, the "gender", "age", "tumor stage", "tumor type" and other spatial features corresponding to the typing characteristic parameters in the clinical follow-up texts of 158 TCGA patients were combined and multivariate analysis was performed. Among them, the p value of the variance (standard deviation) of the matrix-matrix chain edge distance is less than 0.1, which has a relatively good effect for the typing of the target object. The experimental results are as follows Fig. 9 shown.

[0061] In addition, the applicant also used the cutoff values ​​of spatial characteristics corresponding to the classification characteristic parameters calculated and screened by TCGA data to group 72 patients from a university's cancer prevention and treatment center. The experimental results also showed that the above classification characteristic parameters had a significant effect on the patient's prognosis.

[0062] Furthermore, the acquisition of cell segmentation and classification information specifically includes: segmenting the pathological slice image according to preset specifications to obtain pathological slice sub-images; specifically, in some embodiments of the present application, the size of the pathological slice image is 2560*2560, and the size of the pathological slice sub-image is 256*256; of course, there is no restriction on the size of the image rule. Through the pre-trained cell recognition model, the cells in each pathological slice sub-image are identified to obtain cell segmentation and classification information; the acquisition of cell spatial distribution information specifically includes: restoring the pathological slice image by tracing back each pathological slice sub-image in situ; based on the restored pathological slice image and the cell segmentation and classification information, the cell spatial distribution information of the pathological slice image is obtained.

[0063] Specifically, in some embodiments of the present application, some patients whose last follow-up time was more than 3 years were selected as analysis objects, totaling 158 cases; and the pathological sections of each patient were segmented and cut into several 2560*2560 pathological section images, and then 10 2560*2560 pathological section images of each patient were randomly selected for image segmentation and recognition, and segmented into 100 256*256 pathological section sub-images, and each pathological section sub-image was identified using the STARDIST model, and then the spatial information of the cells in each pathological section image was sorted out by the in-situ backtracking method, so as to obtain the cell spatial distribution information of the pathological section image; in this way, a cell topological network was constructed, and finally the cell spatial distribution information of the 10 randomly selected pathological section images was averaged to obtain the final cell spatial distribution information calculation value of the target object.

[0064] Furthermore, the cell recognition model is trained based on the STARDIST model, including: Acquire and preprocess a training sample set, the sample set including multiple training samples; specifically, in some embodiments of the present application, use the Lizard dataset, a total of 4981 non-overlapping image blocks of size 256×256, including 535063 labeled cell nuclei, divide 90% of the data into a training set, and divide 10% of the data into a validation set, the training set includes 4482 images, and the validation set includes 499 images; Since the number of different cell types varies greatly, for example, epithelial cells account for 64.35%, while neutrophils account for only 0.79%, this means that if the model is not balanced, it may tend to predict the majority class (such as epithelial cells) during training, thereby ignoring the minority class (such as neutrophils and eosinophils). Therefore, the applicant adopted the oversampling method to deal with the imbalance problem in the data set, and the optimized training cell distribution is shown in Table 2. Of course, there is no restriction on the specific data set type and data sample size.

[0065] Table 2 Distribution of cells in the training set after oversampling

[0066] The input data includes: training images in multiple training samples; The output data includes: the classification of each cell and the number of each type of cells in the corresponding training image; The input data and output data are used to train the deep learning model and optimize the parameter settings of the deep learning model. If the classification of each cell and the number of each type of cell in the training image output by the deep learning model meet the preset goals, the training is terminated to obtain the cell recognition model.

[0067] Specifically, in some embodiments of the present application, the cell recognition model includes 4 encoder and decoder layers, each layer applies two convolution operations, the convolution kernel size is 3×3, the initial number of convolution kernels is 64, and each layer is downsampled by a 2×2 pooling layer; the activation function uses ReLU, and a convolution layer with 256 convolution kernels is added after UNet for further processing. Specifically, in some embodiments of the present application, the parameter setting for optimizing the deep learning model can be a threshold, that is, a decision boundary, which is used to convert the predicted output of the model into a final classification label or decision. The applicant trained the deep learning model for 80 iterations, and optimized the threshold of the model based on the metrics commonly used in machine learning model evaluation and performance analysis, such as total loss, probability loss, classification loss, mean square error, etc. The candidate thresholds were 0.1, 0.2, and 0.3. The values ​​in the probability map were compared with the current threshold to generate a binary mask. For each binary mask, non-maximum suppression was applied to remove overlapping predictions; this can reduce duplicate detections and keep the most likely bounding boxes or segmentation areas; the optimal threshold configuration returned is prob_thresh=0.499797 and nms_thresh=0.2. When the model's predicted probability for a pixel or target is greater than or equal to 0.499797, the pixel or target will be regarded as a positive class (for example, the target exists), otherwise it will be regarded as a negative class (for example, the target does not exist); when the IoU of two bounding boxes is greater than this threshold, they are considered to overlap and one of them is retained.

[0068] More specifically, the cell recognition effect of the cell recognition model refers to Fig.10 and Fig.11 shown.

[0069] In addition, the applicant also selected 400 256*256 CRC images from other datasets as the test set, and paired the ground truth of the 400 image sets with the prediction results to visualize the relationship between the IoU threshold and the performance index ( Figure 2-3), it can be found that on the test set, when the IoU threshold (nms_thresh) is set to 0.2, the model can still perform relatively well in recognition. Among them, TP (True Positive) means that loU is greater than the threshold, and the prediction box overlaps with the ground truth box, indicating that the target is correctly detected; FP (False Positive) means that loU is less than or equal to the threshold, and the prediction box exists, but there is no corresponding ground truth, indicating an incorrect detection (the target is detected but there is actually no target). FN (False Negative) means that loU is less than the threshold, and there is a ground truth box, but there is no corresponding prediction box, indicating a missed detection (the target actually exists but is not detected). The recognition effect of the specific cell recognition model at different thresholds is shown in the figure below. Figure 7 and Figure 8 shown.

[0070] The cell recognition model is trained based on the STARDIST model architecture. It uses star-shaped convex polygons parameterized by radial distance to represent the shape of the cell nucleus. It can effectively handle the overlap between cells, accurately locate the position of each cell, and provide more accurate cell morphology. It overcomes the problem of large segmentation errors when dealing with high-density cell populations, making the cell segmentation results in complex tumor microenvironments more accurate. It can also avoid the influence of subjective factors to a certain extent, effectively improving the reliability, consistency and objectivity of the recognition results, while also significantly improving the recognition speed and efficiency.

[0071] The embodiments of the present application have at least the following beneficial effects: The embodiment of the present application provides a method for quantitatively analyzing the spatial distribution characteristics of tumor microenvironment cells, which has good cell segmentation and classification accuracy, pays full attention to the adjacency relationship, interaction intensity and spatial structure between cells, and comprehensively analyzes the spatial distribution characteristics of multiple dimensions to achieve good tumor microenvironment cell analysis results, which is used to accurately realize the typing of target objects and has good versatility; specifically, identifying, segmenting and classifying cells through a preset cell recognition model can effectively handle the overlap between cells, accurately locate the position of each cell, and provide more accurate cell morphology, thereby improving the cell segmentation accuracy; constructing a cell topological network, obtaining the first cell spatial distribution characteristics, and being able to accurately identify, segment and classify cells in the cell scale The adjacency and interaction between cells in the tumor microenvironment are accurately modeled, thereby comprehensively revealing the spatial structure of the tumor microenvironment and increasing the spatial dimension of the analysis of the tumor microenvironment; a spatial weight matrix is ​​constructed, and the univariate Moran's index and the bivariate Moran's index are combined to obtain the second cell spatial distribution characteristics, quantify the spatial distribution patterns between different cell types in the tumor microenvironment, and reveal the spatial aggregation or dispersion trends between cell types. The univariate Moran's index can be used to analyze the spatial autocorrelation of similar cells, while the bivariate Moran's index is helpful to analyze the spatial interactions between different cell types, providing a new perspective for the functional study of the tumor microenvironment and thus helping to capture the dynamic changes of cells in advance.

[0072] Example 2 The present application embodiment provides a system for quantitatively analyzing the spatial distribution characteristics of tumor microenvironment cells, which is applied to a method for quantitatively analyzing the spatial distribution characteristics of tumor microenvironment cells in Example 1, such as Figure 2 As shown, including: Cell recognition and segmentation module 1: used to obtain pathological slice images and obtain information about cell segmentation and classification in pathological slice images through a pre-trained cell recognition model; Distribution information acquisition module 2: used to obtain the cell spatial distribution information of the pathological section image based on the cell segmentation and classification information; First feature acquisition module 3: used to construct a cell topology network based on cell spatial distribution information and obtain the first cell spatial distribution feature of the tumor microenvironment; Second feature acquisition module 4: used to construct a spatial weight matrix based on the cell spatial distribution information to obtain the second cell spatial distribution feature of the tumor microenvironment; The classification module 5 is used to judge the tumor status based on the first cell spatial distribution feature and the second cell spatial distribution feature to achieve classification of the target object.

[0073] Furthermore, the cell topological network includes: a first cell space topological network and / or a second cell space topological network; the first cell space topological network is constructed based on the relevant principles of the Delaunay triangulation network, and the second cell space topological network is constructed through the Voronoi diagram; the first cell space distribution characteristics obtained through the first cell space topological network include: one or any more of the proportion of links of each type of cells, the link edge distance and the link frequency; the first cell space distribution characteristics obtained through the second cell space topological network include: one or any more of the area, area variance, mean, median and hybridity of each cell pixel.

[0074] Furthermore, the first cell space topology network includes simplices, the union of all simplices forms the convex hull of the cell distribution, and the triangulation of any cell position contains no more than triangles; among them, Indicates the number of cells, represents the spatial dimension, represents the number of cells in the convex hull; in the second cell space topology network, the Voronoi unit associated with the cell center is defined as: ;in, is a metric space, d is the distance function, K represents the index set that defines the center of a group of cells, that is, the index set of the site, represents a set of non-empty cells in space X, represents the cell center, Represents the Voronoi unit associated with the cell center.

[0075] Furthermore, the acquisition of the second cell spatial distribution characteristics specifically includes: based on the spatial weight matrix, obtaining the grid adjacency weights after gridding the tumor microenvironment; obtaining the number of cells in each grid as grid cell number information; combining the grid adjacency weights with the grid cell number information to analyze the spatial autocorrelation of the same type of cells in the tumor microenvironment and the spatial autocorrelation between different types of cells in the tumor microenvironment.

[0076] Furthermore, the spatial autocorrelation of the same type of cells in the tumor microenvironment was analyzed based on the univariate Moran's index; the calculation formula of the univariate Moran's index is: ; in, represents the number of grids after the tumor microenvironment is gridded. represents the adjacency weight between grid i and grid j, and Respectively represent the grid and Grid The number of cells of a certain type in It represents the mean number of cells of a certain type in all grids; The spatial autocorrelation between different types of cells in the tumor microenvironment was analyzed based on the bivariate Moran's index; the calculation formula of the bivariate Moran's index is: ; in, represents the number of grids after the tumor microenvironment is gridded. Indicates The number of cells of the first type in a grid, Indicates The number of cells of the second type in a grid, is an element in the weight matrix, representing the adjacency weight between grids. represents the mean of the first type of cells, ; represents the mean of the second type of cells, ; represents the spatial hysteresis value, .

[0077] Furthermore, based on the first cell spatial distribution feature and the second cell spatial distribution feature, the tumor status is judged to classify the target object, specifically including: selecting preset classification feature parameters from the first cell spatial distribution feature and the second cell spatial distribution feature respectively; the preset classification feature parameters include: the chain edge distance between tumor cells and immune cells, the standard deviation of the chain edge distance between matrix cells and matrix cells, the variance of the chain edge distance between matrix cells and matrix cells, the number of chain edges between tumor cells and immune cells, the number of triangles in the triangular network, the number of polygons, the standard deviation of the chain edge distance between tumor cells and immune cells, and the variance of the chain edge distance between tumor cells and immune cells, one or any multiple of them; obtaining clinical information of the target object; combining the clinical information of the target object, the classification feature parameters and the association established with the life cycle of the target object, to judge the tumor status to achieve classification of the target object; the tumor status includes: one or any multiple of being in situ, locally extended and metastatic.

[0078] Furthermore, the selection of preset parameter features specifically includes: performing KM survival analysis and Cox hazard ratio analysis based on the overall survival and non-progression survival of the target object; judging whether each characteristic parameter in the first cell spatial distribution feature and the second cell spatial distribution feature meets the preset conditions; the preset conditions include one or more of the logrank-p value and Cox-p value of each characteristic parameter corresponding to the overall survival and non-progression survival; if the preset conditions are met, it is used as the typing characteristic parameter.

[0079] Furthermore, the acquisition of cell segmentation and classification information specifically includes: segmenting the pathological slice image according to preset specifications to obtain pathological slice sub-images; identifying cells in each pathological slice sub-image through a pre-trained cell recognition model to obtain cell segmentation and classification information; the acquisition of cell spatial distribution information specifically includes: restoring the pathological slice image by in situ backtracking each pathological slice sub-image; and obtaining the cell spatial distribution information of the pathological slice image based on the restored pathological slice image and the cell segmentation and classification information.

[0080] Furthermore, the cell recognition model is trained based on the STARDIST model, including: Acquire and preprocess a training sample set, the sample set includes multiple training samples; input data includes: training images in the multiple training samples; output data includes: each cell classification and the number of each type of cells in the corresponding training image; use the input data and output data to train a deep learning model and optimize the parameter setting of the deep learning model. If the each cell classification and the number of each type of cells in the training image output by the deep learning model meet the preset goals, then end the training to obtain a cell recognition model.

[0081] Since the technical problem to be solved by this embodiment, the technical solution used to solve the technical problem, and the technical effect achieved are the same as the method described in Example 1, the technical details of the system in this embodiment can be found in Example 1 and will not be repeated here.

[0082] Those skilled in the art can understand that the modules in the device in the implementation scenario can be distributed in the device of the implementation scenario according to the description of the implementation scenario, or can be changed accordingly and located in one or more devices different from the present implementation scenario.

[0083] The modules of the above implementation scenarios can be combined into one module, or further divided into multiple sub-modules. The above serial numbers of the present invention are only for description and do not represent the advantages and disadvantages of the implementation scenarios.

[0084] The above disclosure is only a few specific implementation scenarios of the present invention, but the present invention is not limited thereto, and any changes that can be conceived by those skilled in the art should fall within the protection scope of the present invention.

Claims

1. A method for quantitatively analyzing the spatial distribution characteristics of tumor microenvironment cells, characterized in that: include: Acquire a pathological section image, and obtain cell segmentation and classification information in the pathological section image through a pre-trained cell recognition model; Based on the cell segmentation and classification information, obtaining the cell spatial distribution information of the pathological section image; Based on the cell spatial distribution information, a cell topological network is constructed to obtain a first cell spatial distribution feature of the tumor microenvironment; According to the cell spatial distribution information, a spatial weight matrix is ​​constructed to obtain the second cell spatial distribution characteristics of the tumor microenvironment; Based on the first cell spatial distribution characteristics and the second cell spatial distribution characteristics, the tumor status is judged to achieve typing of the target object.

2. A method for quantitatively analyzing the spatial distribution characteristics of tumor microenvironment cells according to claim 1, characterized in that: The cell topological network includes: a first cell space topological network and / or a second cell space topological network; The first cell space topology network is constructed based on the relevant principles of Delaunay triangulation network, and the second cell space topology network is constructed by Voronoi diagram; The first cell space distribution characteristics obtained through the first cell space topological network include: one or any multiple of: the proportion of each type of cell links, the link edge distance and the link frequency; The first cell space distribution characteristics obtained through the second cell space topological network include: one or any multiple of the area, area variance, mean, median and promiscuity of each cell pixel.

3. A method for quantitatively analyzing the spatial distribution characteristics of tumor microenvironment cells according to claim 2, characterized in that: The first cell space topology network includes simplices, the union of all simplices forms the convex hull of the cell distribution, and the triangulation of any cell position contains no more than triangles; in, Indicates the number of cells, represents the spatial dimension, represents the number of cells in the convex hull; In the second cell space topology network, the Voronoi unit associated with the cell center is defined as: ； in, is a metric space, d is the distance function, K represents the index set that defines the center of a group of cells, that is, the index set of the site, represents a set of non-empty cells in space X, represents the cell center, Represents the Voronoi unit associated with the cell center.

4. The method for quantitatively analyzing the spatial distribution characteristics of tumor microenvironment cells according to claim 1, characterized in that: The acquisition of the second cell spatial distribution characteristics specifically includes: Based on the spatial weight matrix, obtaining a grid adjacency weight after gridding the tumor microenvironment; The number of cells in each grid is obtained as grid cell number information; The grid adjacency weight and the grid cell quantity information are combined to analyze the spatial autocorrelation of the same type of cells in the tumor microenvironment and the spatial autocorrelation of different types of cells in the tumor microenvironment.

5. A method for quantitatively analyzing the spatial distribution characteristics of tumor microenvironment cells according to claim 4, characterized in that: Analyze the spatial autocorrelation of the same type of cells in the tumor microenvironment based on the univariate Moran's index; The calculation formula of the univariate Moran index is: ； in, represents the number of grids after the tumor microenvironment is gridded. represents the adjacency weight between grid i and grid j, and Respectively represent the grid and Grid The number of cells of a certain type in It represents the mean number of cells of a certain type in all grids; Analyze the spatial autocorrelation between different cell types in the tumor microenvironment based on the bivariate Moran's index; The calculation formula of the bivariate Moran index is: ； in, represents the number of grids after the tumor microenvironment is gridded. Indicates The number of cells of the first type in a grid, Indicates The number of cells of the second type in a grid, is an element in the weight matrix, representing the adjacency weight between grids. represents the mean of the first type of cells, ; represents the mean of the second type of cells, ; represents the spatial hysteresis value, .

6. The method for quantitatively analyzing the spatial distribution characteristics of tumor microenvironment cells according to claim 1, characterized in that: Based on the first cell spatial distribution feature and the second cell spatial distribution feature, determining the tumor state to achieve the classification of the target object specifically includes: Selecting preset typing characteristic parameters from the first cell spatial distribution characteristic and the second cell spatial distribution characteristic respectively; The preset classification characteristic parameters include: one or any multiple of the chain edge distance between tumor cells and immune cells, the standard deviation of the chain edge distance between matrix cells and matrix cells, the variance of the chain edge distance between matrix cells and matrix cells, the number of chain edges between tumor cells and immune cells, the number of triangles in the triangular network, the number of polygons, the standard deviation of the chain edge distance between tumor cells and immune cells, and the variance of the chain edge distance between tumor cells and immune cells; Obtain clinical information of target subjects; Combining the clinical information of the target object, the association between the classification characteristic parameters and the life cycle of the target object, determining the tumor status to achieve classification of the target object; The tumor status includes: one or any multiple of being in situ, locally extended, and metastatic.

7. A method for quantitatively analyzing the spatial distribution characteristics of tumor microenvironment cells according to claim 6, characterized in that: The selection of preset parameter features specifically includes: KM survival analysis and Cox hazard ratio analysis were performed based on the overall survival and progression-free survival of the target subjects; Determining whether each characteristic parameter in the first cell spatial distribution characteristic and the second cell spatial distribution characteristic meets a preset condition; The preset conditions include one or more value ranges of the logrank-p value and the Cox-p value of each characteristic parameter corresponding to the overall survival period and the non-progression survival period; If the preset conditions are met, it will be used as the typing feature parameter.

8. The method for quantitatively analyzing the spatial distribution characteristics of tumor microenvironment cells according to claim 1, characterized in that: Acquisition of cell segmentation and classification information, including: Segment the pathological slice image according to preset specifications to obtain pathological slice sub-images; Using the pre-trained cell recognition model, identifying cells in each of the pathological slice sub-images, and obtaining the cell segmentation and classification information; Acquisition of cell spatial distribution information, including: Restoring the pathological slice image by in-situ tracing back each of the pathological slice sub-images; Based on the restored pathological section image and the cell segmentation and classification information, the cell spatial distribution information of the pathological section image is obtained.

9. A method for quantitatively analyzing the spatial distribution characteristics of tumor microenvironment cells according to claim 8, characterized in that: The cell recognition model is trained based on the STARDIST model, specifically including: Acquire and preprocess a training sample set, wherein the sample set includes a plurality of training samples; The input data includes: a plurality of training images in the training samples; The output data includes: the number of each cell classification and each type of cell in the training image; The input data and the output data are used to train a deep learning model and optimize the parameter settings of the deep learning model. If the classification of each cell and the number of each type of cell in the training image output by the deep learning model meet the preset goals, the training is terminated to obtain the cell recognition model.

10. A system for quantitatively analyzing the spatial distribution characteristics of tumor microenvironment cells, applied to a method for quantitatively analyzing the spatial distribution characteristics of tumor microenvironment cells according to any one of claims 1 to 9, characterized in that: include: Cell recognition and segmentation module: used to obtain pathological slice images and obtain information about cell segmentation and classification in the pathological slice images through a pre-trained cell recognition model; Distribution information acquisition module: used to acquire the cell spatial distribution information of the pathological section image based on the cell segmentation and classification information; A first feature acquisition module: used to construct a cell topology network based on the cell spatial distribution information to obtain a first cell spatial distribution feature of the tumor microenvironment; A second feature acquisition module is used to construct a spatial weight matrix based on the cell spatial distribution information to obtain the second cell spatial distribution feature of the tumor microenvironment; A classification module is used to judge the tumor status based on the first cell spatial distribution characteristics and the second cell spatial distribution characteristics to achieve classification of the target object.

Images