Immune cell classification method and system based on polarization image and immune environment evaluation method and system

Abstract

The invention discloses an immune cell classification method and system and an immune environment evaluation method and system based on polarization images, and the immune cell classification method comprises the following steps: a polarization image acquisition step: carrying out data collection on a pathological sample to obtain polarization data, and processing the polarization data to obtain at least one polarization image; a cell nucleus feature acquisition step: performing cell nucleus segmentation on the at least one polarization image, and performing calculation to obtain at least one cell nucleus feature; a cell characteristic acquisition step: performing cell segmentation on the at least one polarization image, and calculating to obtain at least one cell characteristic; and a cell identification step: obtaining the type of each immune cell in the polarization image through an immune cell classification model based on the cell nucleus characteristics and the cell characteristics.

Description

Technical Field

[0001] The present invention relates to the technical field of full-vector polarization imaging, and in particular to an immune cell classification method and system based on polarization images, and an immune environment assessment method and system. Background Art

[0002] Immune cells refer to cells that participate in or are related to immune responses, including lymphocytes, dendritic cells, monocytes / macrophages, granulocytes, mast cells, etc. Immune cells can be divided into many types, and various immune cells play important roles in the human body. Different types of immune cells have different functions in the tumor microenvironment and play a key role in the tumor environment. Immune cells can identify and destroy new tumor cells during cancer immune monitoring. Different types of immune cells may ultimately determine the different behaviors or activity levels of the tumor microenvironment. Therefore, a means is needed to determine the types and functional activities of different immune cells, and then infer the prognosis of the tumor for medication guidance.

[0003] The common immune cell classification methods currently include flow cytometry, immunohistochemistry and other methods for differentiation, but the use of flow cytometry must be measured in a single cell suspension, the operation of immunohistochemistry is complicated, the reproducibility of the results is poor, and the tissue is highly processed, which may lead to the loss of natural state information. Although these methods can distinguish different types of immune cells, fluorescent labeling for specific antigens in pathological tissues still faces great challenges. The operation of fluorescent labeling is complicated, and it is difficult to ensure the specificity and consistency of the labeling. Therefore, it is difficult for existing technologies to achieve accurate quantitative detection of immune cell immune activity in pathological tissues.

[0004] In the existing technology, the state of immune cells in the tumor microenvironment is usually observed on HE-stained pathological sections under a pathological microscope to evaluate the immune environment. However, in traditional HE pathological sections, due to the poor contrast of the image, it is difficult to effectively distinguish some immune cells. At the same time, the immunohistochemical method requires the sample to be labeled, the operation process is complicated, and the tissue is highly processed, which may lead to the loss of natural state information. The above detection methods cannot achieve effective classification and evaluation of the immune microenvironment.

[0005] Therefore, in order to solve the above-mentioned defects, it is urgent to develop an immune cell classification method and system based on polarization images, and an immune environment assessment method and system that overcome the above-mentioned defects. Summary of the invention

[0006] In view of the above problems, the present invention provides an immune cell classification method based on polarization images, which includes:

[0007] Polarization image acquisition step: collecting data on the pathological sample to obtain polarization data, and processing the polarization data to obtain at least one polarization image;

[0008] The cell nucleus feature acquisition step: after performing cell nucleus segmentation on at least one of the polarized images, at least one cell nucleus feature is calculated and obtained;

[0009] Cell feature acquisition step: after performing cell segmentation on at least one of the polarized images, at least one cell feature is calculated and obtained;

[0010] Cell identification step: obtaining the type of each immune cell in the polarization image through an immune cell classification model based on the cell nucleus characteristics and the cell characteristics.

[0011] In the above immune cell classification method, the cell nucleus feature acquisition step comprises:

[0012] The first segmentation step: performing global threshold segmentation on the polarization image to obtain at least one cell nucleus region;

[0013] The first calculation step: calculating each of the cell nucleus regions to obtain at least one cell nucleus feature.

[0014] In the above immune cell classification method, the cell feature acquisition step comprises:

[0015] The second segmentation step: performing global threshold segmentation on the polarization image to obtain at least one cell region;

[0016] The second calculation step: calculating each of the cell regions to obtain at least one cell feature.

[0017] The above-mentioned immune cell classification method, wherein the polarization image acquisition step includes processing at least one polarization parameter of the polarization data, and visualizing at least one polarization parameter to obtain at least one polarization image;

[0018] Wherein, the polarization parameters include at least one of depolarization, linear phase delay and fast axis azimuth, the polarization image includes at least one of depolarization image, linear phase delay image and fast axis azimuth image, the cell nuclear characteristics in the first calculation step include: at least one of cell nuclear circumference, cell nuclear area and cell nuclear roundness, and the cell characteristics in the second calculation step include: at least one of cell circumference, cell area, cell membrane roundness, cell nucleus-cytoplasm ratio, diffusion distance, cell cytoplasm concentration and cell membrane permeability.

[0019] In the above immune cell classification method, the first segmentation step includes:

[0020] Performing weighted integration on at least one of the polarization images using a first weight factor to obtain a first weighted boundary characteristic map;

[0021] Performing global threshold segmentation on the first weighted boundary characteristic map by Otsu method to obtain at least one cell nucleus region;

[0022] The second segmentation step includes:

[0023] Performing weighted integration on at least one of the polarization images using a second weight factor to obtain a second weighted boundary characteristic map;

[0024] The second weighted boundary characteristic map is subjected to global threshold segmentation by the Otsu method to obtain at least one cell region.

[0025] In the above immune cell classification method, the cell identification step comprises:

[0026] Feature processing step: integrating the cell nucleus feature and the cell feature to obtain an immune cell feature vector;

[0027] A clustering result obtaining step: clustering the immune cell feature vectors using the immune cell classification model to obtain a clustering result, wherein the immune cell classification model is obtained by training an unsupervised learning model;

[0028] Classification step: Identify the morphology of the immune cell population in each cluster in the clustering result to obtain the type of each immune cell.

[0029] The present invention further provides an immune cell classification system based on polarization images, wherein the immune cell classification method described above is applied, and the immune cell classification system comprises:

[0030] A polarization image acquisition unit collects data from the pathological sample to obtain polarization data, and processes the polarization data to obtain at least one polarization image;

[0031] A cell nucleus feature acquisition unit, which performs cell nucleus segmentation on at least one of the polarization images and calculates and obtains at least one cell nucleus feature;

[0032] A cell feature acquisition unit, which performs cell segmentation on at least one of the polarization images and calculates and obtains cell features;

[0033] The cell recognition unit obtains the type of each immune cell in the polarization image through an immune cell classification model based on the cell nucleus feature and the cell feature.

[0034] The present invention also provides an immune environment assessment method based on polarization images, wherein the immune cell classification method described above is applied, and the immune environment assessment method comprises:

[0035] Polarization image acquisition step: collecting data on the pathological sample to obtain polarization data, and processing the polarization data to obtain at least one polarization image;

[0036] The cell nucleus feature acquisition step: after performing cell nucleus segmentation on at least one of the polarized images, at least one cell nucleus feature is calculated and obtained;

[0037] Cell feature acquisition step: after performing cell segmentation on at least one of the polarization images, calculate and obtain cell features;

[0038] Cell identification step: obtaining the type of each immune cell in the polarization image through an immune cell classification model based on the cell nucleus characteristics and the cell characteristics;

[0039] Evaluation step: obtaining an immune environment evaluation result corresponding to the pathological sample based on the type of the immune cells, the characteristics of the cell nucleus and the characteristics of the cells.

[0040] The above-mentioned immune environment assessment method, wherein the assessment step comprises:

[0041] Index calculation step: calculating the index of each type of the immune cells based on the type of the immune cells, the characteristics of the cell nucleus and the characteristics of the cells;

[0042] Immuno-score acquisition step: obtaining an immune score corresponding to the pathological sample through an immune environment comprehensive evaluation model based on the indicators of the immune cells;

[0043] Evaluation result acquisition step: obtaining the immune environment evaluation result corresponding to the pathological sample according to the immune score based on the preset classification rules.

[0044] The present invention also provides an immune environment assessment system based on polarization images, wherein the immune environment assessment method described above is applied, and the immune environment assessment system comprises:

[0045] A polarization image acquisition unit collects data from the pathological sample to obtain polarization data, and processes the polarization data to obtain at least one polarization image;

[0046] A cell nucleus feature acquisition unit, which performs cell nucleus segmentation on at least one of the polarization images and calculates and obtains at least one cell nucleus feature;

[0047] A cell feature acquisition unit, which performs cell segmentation on at least one of the polarization images and calculates and obtains cell features;

[0048] A cell recognition unit, which obtains the type of each immune cell in the polarization image through an immune cell classification model based on the cell nucleus feature and the cell feature;

[0049] An evaluation unit is configured to obtain an immune environment evaluation result corresponding to the pathological sample based on the type of the immune cell, the cell nuclear characteristics and the cell characteristics.

[0050] Compared with the prior art, the present invention is effective in that: the present invention provides an immune cell classification method and system, an immune environment assessment method and system through an imaging method based on polarization images, which can perform immune cell typing without labeling, thereby providing a beneficial tool for the assessment of the immune environment.

[0051] Other features and advantages of the present invention will be described in the following description, and partly become obvious from the description, or be understood by implementing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures pointed out in the description and the drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0053] Figure 1 is a flow chart of the immune cell classification method of the present invention;

[0054] Figure 2 for Figure 1 The step-by-step flow chart of step S2;

[0055] Figure 3 for Figure 1 A step-by-step flow chart of step S3;

[0056] Figure 4 for Figure 1 The step-by-step flow chart of step S4;

[0057] Figure 5 is a schematic diagram of the structure of the immune cell classification system of the present invention;

[0058] Figure 6 is a flow chart of the immune environment assessment method of the present invention;

[0059] Figure 7 for Figure 6 The step-by-step flow chart of step S5;

[0060] Figure 8 It is a schematic diagram of the structure of the immune environment assessment system of the present invention. DETAILED DESCRIPTION

[0061] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0062] The exemplary embodiments and descriptions of the present invention are used to explain the present invention, but are not intended to limit the present invention. In addition, elements / components with the same or similar reference numerals used in the drawings and embodiments are used to represent the same or similar parts.

[0063] The terms “first”, “second”, “S1”, “S2”, etc. used in this document do not specifically refer to an order or sequence, nor are they used to limit the present invention. They are only used to distinguish elements or operations described with the same technical terms.

[0064] The words “include,” “including,” “have,” “contain,” etc. used in this document are open-ended terms, meaning including but not limited to.

[0065] As used herein, “plurality” includes “two” and “more than two”; as used herein, “plurality groups” includes “two groups” and “more than two groups”.

[0066] See also Figure 1 , Figure 1 FIG. 4 is a flow chart of the immune cell classification method of the present invention. Figure 1 As shown, the present invention proposes an immune cell classification method based on polarization images: by collecting and calculating the polarization data of the pathological sample, the cell nucleus is segmented and the characteristics of the cell nucleus are calculated; the cell membrane and the immune cells are segmented and the overall characteristics of the cell are calculated; then the cell nucleus and immune cell characteristics are integrated to construct an immune cell feature vector; and then the immune cells are clustered based on the immune cell feature vector, thereby classifying the identified immune cells based on unsupervised learning of the cell nucleus and cell characteristics.

[0067] Specifically, the immune cell classification method includes:

[0068] Polarization image acquisition step S1: collecting data on the pathological sample to obtain polarization data, and processing the polarization data to obtain at least one polarization image;

[0069] Cell nucleus feature acquisition step S2: after performing cell nucleus segmentation on at least one of the polarized images, at least one cell nucleus feature is calculated and obtained;

[0070] Cell feature acquisition step S3: after performing cell segmentation on at least one of the polarized images, at least one cell feature is calculated and obtained;

[0071] Cell identification step S4: obtaining the type of each immune cell in the polarization image through an immune cell classification model based on the cell nucleus characteristics and the cell characteristics.

[0072] Therefore, the present invention can perform label-free typing and activity detection of immune cells through an imaging method based on polarization images, and further provide a favorable tool for the evaluation of the immune environment, which will be described in detail later.

[0073] In this embodiment, the polarization image acquisition step S1 includes: collecting data on the pathological sample to obtain multiple sets of polarization data; performing calculation processing on the multiple sets of polarization data to obtain the multiple polarization parameters; and visualizing the multiple polarization parameters to obtain multiple polarization images corresponding to the multiple polarization parameters.

[0074] In this embodiment, after data acquisition is performed on the pathological sample to obtain 16 sets of polarization data, the 16 sets of polarization data are calculated and processed to obtain three polarization parameters, namely, depolarization, linear phase delay and fast axis azimuth, and then the three polarization parameters are visualized to obtain three polarization images, namely, depolarization image, linear phase delay image and fast axis azimuth image, wherein the calculation and visualization of depolarization, linear phase delay and fast axis azimuth are specifically described as follows:

[0075] 1) Deflection

[0076] Depolarization (D(x,y)) is a parameter that describes the degree of light polarization and is used to measure the depolarization effect of light in a sample. Through the Mueller matrix, depolarization can be expressed as:

[0077]

[0078] Where: M 11 (x,y) is the first row and first column element of the Mueller matrix; M 12 (x,y) is the first row and second column element of the Mueller matrix, M 13 (x,y) is the first row and third column element of the Mueller matrix, M 14 (x,y) is the 1st row and 4th column element of the Mueller matrix.

[0079] 2) Linear phase delay

[0080] The linear phase delay Δφ(x,y) is used to reflect the phase delay of light in the birefringent sample. The calculation formula is:

[0081]

[0082] Where: M 23 (x,y) and M 33 (x,y) are the 2nd row and 3rd column elements and the 3rd row and 3rd column elements of the Mueller matrix respectively.

[0083] 3) Fast axis azimuth

[0084] Fast axis azimuth θ p (x, y) describes the polarization direction of each pixel, and the calculation formula is:

[0085] D(x,y),Δφ(x,y),θ p (x, y) is converted into visual images, namely, depolarization image, linear phase delay image and fast axis azimuth image, which respectively represent the distribution of depolarization, linear phase delay and fast axis azimuth mode. These images will be used as input for subsequent image processing.

[0086] Please refer to Figure 2 , Figure 2 for Figure 1 The step-by-step flow chart of step S2 in FIG. Figure 2 The cell nucleus feature acquisition step S2 shown includes:

[0087] First segmentation step S21: performing global threshold segmentation on the polarization image to obtain at least one cell nucleus region;

[0088] The first calculation step S22: Calculate each of the cell nucleus regions to obtain at least one cell nucleus feature.

[0089] Cell nucleus segmentation is a key step in immune cell analysis. By utilizing the acquired polarization images, the cell nucleus region can be effectively segmented.

[0090] Specifically, the first segmentation step S21 includes: weighting and integrating at least one polarization image by a first weight factor to obtain a first weighted boundary characteristic map; performing global threshold segmentation on the first weighted boundary characteristic map by the Otsu method to obtain at least one cell nucleus region. Specifically, in this embodiment, after obtaining the depolarization image y D (x, y), linear phase delay image G φ (x,y) and fast axis azimuth image G θAfter the gradient map of (x, y) is obtained, the three are combined for weighted segmentation to extract the cell nucleus boundary. It should be noted that in this embodiment, the deskewing degree image G D (x, y), linear phase delay image G φ (x,y) and fast axis azimuth image G θ (x, y) is a gradient map as a preferred implementation method. The acquisition of the gradient map will be described in detail later. The specific description is as follows:

[0091] 1) Weighted combination

[0092] Combining three modalities (deflection image G D (x, y), linear phase delay image G φ (x,y) and fast axis azimuth image G θ (x, y)) to calculate the first weighted boundary characteristic graph G(x, y). The weighted formula is:

[0093] G(x,y)=w D ·G D (x,y)+w φ ·G φ (w,y)+w θ ·G θ (x,y);

[0094] Where: w D , w φ and w θ is the first weight factor, which controls the contribution of depolarization, linear phase delay and fast axis azimuth to the weighted boundary characteristic map respectively; G(x,y) is the first weighted boundary characteristic map, which is used to highlight the boundary features of the cell nucleus.

[0095] 2) Otsu method for global threshold segmentation

[0096] By using the Otsu method, the global optimal segmentation threshold t is automatically determined in the distribution of G(x,y) * , the boundary characteristics Figure 2 The specific formula is as follows:

[0097]

[0098] Where: ω1(t) and ω2(t) are the probabilities of pixels on both sides of the threshold t:

[0099]

[0100] μ1(t) and μ2(t) are the means of the pixels on either side of the threshold t:

[0101]

[0102] Determine the optimal threshold t * :

[0103]

[0104] Binarization segmentation result:

[0105]

[0106] Among them: B nucleus (x, y) is the binary segmentation result of the cell nucleus, that is, the cell nucleus area, where pixels with a value of 1 represent the cell nucleus boundary area and pixels with a value of 0 represent the background.

[0107] Among them, in the first calculation step S22, at least one of the cell nucleus features includes: at least one of the cell nucleus circumference, the cell nucleus area and the cell nucleus roundness. Specifically, in this embodiment, it includes:

[0108] 1) The circumference of the nucleus P e :

[0109] The perimeter of the nucleus e It refers to the length of the cell nucleus boundary after segmentation. It can be obtained by obtaining the boundary pixel set through the Canny edge detection algorithm and calculating the number of boundary pixels. The calculation formula of the perimeter is:

[0110]

[0111] Among them, B nucleus is the set of pixels on the boundary of the cell nucleus; (x, y) represents the position of each pixel on the boundary.

[0112] 2) The area of ​​the cell nucleus A r :

[0113] The area of ​​the cell nucleus A r Refers to the total number of pixels covered by the segmented cell nucleus area. The area calculation formula is as follows:

[0114]

[0115] Where N is the set of pixels in the cell nucleus; (x, y) represents the position of each pixel in the cell nucleus. This means that the area A r It is the sum of all the pixels belonging to the cell nucleus in the cell nucleus segmentation image. This feature can reflect the size of the cell nucleus.

[0116] 3) Roundness of the nucleus C:

[0117] The roundness C of the cell nucleus is a geometric feature that measures the degree of closeness between the shape of the cell nucleus and the ideal circle. The closer the roundness is to 1, the closer the shape of the cell nucleus is to a circle. The definition formula of roundness is:

[0118]

[0119] In this embodiment, after the polarization image acquisition step S1 and before the cell nucleus feature acquisition step S2, a preprocessing step may also be included, that is, before performing cell nucleus segmentation, each polarization image needs to be preprocessed to improve the accuracy and stability of segmentation. The preprocessing step includes the following steps: denoising step: smoothing the polarization image obtained by visualization through Gaussian filtering; equalization step: performing contrast processing on the smoothed polarization image through histogram equalization; gradient calculation step: extracting the boundary characteristics of the equalized polarization figure through gradient calculation to obtain a gradient map. The specific description is as follows:

[0120] 1) Gaussian filtering

[0121] Gaussian filtering is performed on the three modal images to remove noise and smooth the boundaries. The formula is:

[0122]

[0123] Where: g D (x,y),g φ (x,y) and g θ (x, y) represents the filtering results of the depolarization image, linear phase delay image and fast axis azimuth image respectively; D(x, y), Δφ(x, y) and θ p (x, y) is the depolarization image, linear phase delay image and fast axis azimuth image before filtering; G(u, v) is the Gaussian filter kernel used to smooth the image, and its formula is:

[0124]

[0125] Where σ is the standard deviation of the filter kernel, which determines the smoothness of the filter.

[0126] 2) Histogram equalization

[0127] For the filtered modal image g D (x,y),g φ (x,y) and g θ (x,y) is histogram equalized to enhance the contrast and improve the clarity of the boundary. The following are the specific steps:

[0128] Histogram of pixel values: Calculate the number of pixels h(i) at each gray level in the image:

[0129]

[0130] Where: δ(g(x,y)=i) is the indicator function, when g(x,y)=i, the value is 1, otherwise it is 0; h(i) is the number of pixels of gray level i;

[0131] Calculate the cumulative distribution function (CDF): The cumulative distribution function cdf(i) represents the cumulative frequency of pixels at gray level i and below:

[0132]

[0133] Normalize the cumulative distribution function:

[0134]

[0135] Where N is the total number of pixels in the image.

[0136] Remapping: Map the original value i to a new value g′(x,y) according to the cumulative distribution function cdf(i):

[0137]

[0138] Where: L is the total number of gray levels, usually 256; cdf min is the smallest non-zero value in the cumulative distribution function; Indicates rounding down.

[0139] The modal images after equalization are:

[0140] g′ D (x,y),g′ φ (x,y),g′ θ (x,y);

[0141] They represent the depolarization image, linear phase delay image and fast axis azimuth image after histogram equalization.

[0142] 3) Gradient calculation

[0143] In order to extract the boundary characteristics, the modal image g′ after equalization is D (x,y),g′ φ (x,y),g′ θ (x,y) calculates the gradient. The gradient calculation formula is as follows:

[0144] Deskewing gradient plot:

[0145]

[0146] Linear phase delay gradient plot:

[0147]

[0148] Fast axis azimuth gradient map:

[0149]

[0150] Among them: through G D (x,y),G φ (x,y) and G θ (x, y) represent the boundary characteristics of depolarization, linear phase delay and fast axis azimuth respectively; the gradient calculation is realized through partial derivatives, and the rate of change in the x and y directions is calculated respectively.

[0151] Through the above steps, we obtain the three modal images G after Gaussian filtering, histogram equalization and gradient calculation. D (x,y),G φ (x,y),G θ (x,y) lays the foundation for subsequent weighted segmentation.

[0152] Please refer to Figure 3 , Figure 3 for Figure 1 The step-by-step flow chart of step S3 in FIG. Figure 3 As shown, the cell feature acquisition step S3 includes:

[0153] The second segmentation step: performing global threshold segmentation on the polarization image to obtain at least one cell region;

[0154] The second calculation step: calculating each of the cell regions to obtain at least one cell feature.

[0155] Among them, the polarization image can effectively help distinguish cells from background areas and accurately segment the outline of cells, that is, the optical characteristics of cell membrane, cytoplasm and other subcellular structures can be reflected through phase information and intensity information. In this embodiment, the second segmentation step S31 includes: weighting and integrating at least one of the polarization images through a second weight factor to obtain a second weighted boundary characteristic map; performing global threshold segmentation on the second weighted boundary characteristic map through the Otsu method to obtain at least one cell area. In order to segment cells more accurately, after obtaining the depolarization image G D (x,y), linear phase delay image G φ (x, y) and fast axis azimuth image G θ After the gradient map of (x, y) is obtained, the three are combined for weighted segmentation to extract the cell boundary. It should be noted that in this embodiment, the deskewing degree image G D (x,y), linear phase delay image G φ (x, y) and fast axis azimuth image Gθ (x, y) is a gradient map as a preferred implementation, which is specifically described as follows:

[0156] 1) Weighted combination

[0157] Combining three modalities (deflection image G D (x,y), linear phase delay image G φ (x, y) and fast axis azimuth image G θ (x, y)) to calculate the second weighted boundary characteristic graph G′(x, y). The weighted formula is:

[0158] G′(x,y)=w DCell ·G D (x,y)+w φCell ·G φ (x,y)+w θCell ·G θ (x,y);

[0159] Where: w DCell , w φCell and w θCell is the second weight factor, which controls the contribution of depolarization, linear phase delay and fast axis azimuth to the weighted boundary characteristic map respectively; G′(x, y) is the second weighted boundary characteristic map, which is used to highlight the boundary characteristics of the cell.

[0160] 2) Otsu method for global threshold segmentation

[0161] The global optimal segmentation threshold t is automatically determined in the distribution of G′(x,y) by Otsu’s method. Ce * , the boundary characteristics Figure 2 The specific formula is as follows:

[0162]

[0163] Where: ω1(t) and ω2(t) are the probabilities of pixels on both sides of the threshold t:

[0164]

[0165] μ1(t) and μ2(t) are the means of the pixels on either side of the threshold t:

[0166]

[0167] Determine the optimal threshold t Cell * :

[0168]

[0169] Binarization segmentation result:

[0170]

[0171] Among them: B Cell (x, y) is the binary segmentation result of the cell, that is, the cell area, where pixels with a value of 1 represent the cell boundary area and pixels with a value of 0 represent the background.

[0172] Among them, the at least one cell feature in the second calculation step S32 includes: at least one of: cell perimeter, cell area, cell membrane roundness, cell nucleus-cytoplasm ratio, diffusion distance, cell cytoplasm concentration and cell membrane permeability. Specifically, in this embodiment, it includes:

[0173] In this embodiment, calculating cell features includes:

[0174] 1) The perimeter of the cell P ec

[0175] The cell perimeter P ec It refers to the length of the cell boundary after segmentation. It can be obtained by counting the number of boundary pixels. The formula is as follows:

[0176]

[0177] Among them, BCell is the set of cell boundary pixels, and (x, y) represents the position of each pixel on the boundary.

[0178] 2) The area of ​​the cell A re

[0179] The area of ​​the cell A re It refers to the total number of pixels covered by the cell area after segmentation. It reflects the overall size of the cell. The area calculation formula is as follows:

[0180]

[0181] Among them, C is the set of pixels in the cell area, and (x, y) represents the position of each pixel in the cell area.

[0182] 3) Roundness of cell membrane C c :

[0183] The circularity of the cell membrane C c It is used to measure the closeness between the cell shape and the ideal circle. c The closer it is to 1, the closer the shape of the cell is to the ideal circle. The smaller the value, the more irregular the cell shape. It is defined as:

[0184]

[0185] Among them, A re is the area of ​​the cell, P ec is the circumference of the cell.

[0186] 4) The nuclear-cytoplasmic ratio of the cell uc :

[0187] N-N uc It is the ratio of the nuclear area to the total cell area. It is an important characteristic to measure the size of the nucleus relative to the cytoplasm. Especially in some cell types, changes in the nuclear-to-cytoplasmic ratio are related to the functional activity of the cell. It is defined as:

[0188]

[0189] Among them, A nuc is the area of ​​the cell nucleus; A re is the total area of ​​the cell.

[0190] 5) Diffusion distance m

[0191] The diffusion distance m is used to describe the maximum distance between the cell membrane and the center of the cell nucleus. It reflects the degree of cell expansion. The diffusion distance is defined as:

[0192]

[0193] Among them, (x nuc ,y nuc ) is the coordinate of the center of mass of the cell nucleus; d((x,y),(x nuc ,y nuc )) represents the Euclidean distance between any point (x, y) on the cell membrane and the center of mass of the cell nucleus.

[0194] 6) Cell clathrate concentration Con(x,y)

[0195] The cladding concentration Con(x,y) is used to describe the average brightness and polarization parameters of the cytoplasm region within the cell, reflecting the density, staining depth, and optical activity of the cytoplasm. The brightness value of all pixels in the cell region and the polarization parameter are added and divided by the cell area A. re , can quantify the staining intensity and density of the cytoplasm. Its mathematical definition is as follows:

[0196]

[0197] I(x,y) represents the pixel brightness value at position (x,y), specifically the intensity component after image preprocessing (such as Gaussian filtering and histogram equalization). This brightness value is used to reflect the light intensity of the cytoplasm region.

[0198] P(x,y) represents the polarization parameter at position (x,y), which can be the optical polarization parameter calculated by the Mueller matrix, such as the linear phase delay Δφ(x,y) or the fast axis azimuth θ f (x,y).

[0199] A re is the area of ​​the cell region, that is, the number of pixels in C (cell pixel set).

[0200] Through Con(x,y), the optical and physical properties of the cytoplasm, such as density, staining depth, and polarization response, can be reflected.

[0201] Activity A based on cladding concentration i Defined as:

[0202]

[0203] Among them, Cell i represents the pixel set of the i-th type of immune cells, and Con(x,y) represents the patina concentration at the position (x,y). In this general formula, different values ​​of i can represent the number, proportion and activity of lymphocytes, plasma cells and other immune cells respectively.

[0204] 7) Cell membrane permeability

[0205] The permeability of the cell membrane is used to describe the permeability of the cell membrane, which is mainly related to the diffusion capacity of the cell membrane and is particularly important when evaluating the activity of immune cells. It is defined as:

[0206]

[0207] Where D is the diffusion coefficient. The average diffusion coefficient of proteins in the cytoplasm is usually 5-15 μm 2 / s, the diffusion coefficient of small molecule metabolites in water is higher, about 500μm 2 / s; Con(x,y) is the clathrate concentration; m is the maximum distance between the cell membrane and the center of the cell nucleus, that is, the diffusion distance.

[0208] Please refer to Figure 4 , Figure 4 for Figure 1 The step-by-step flow chart of step S4 is as follows: Figure 4 As shown, the cell identification step S4 includes:

[0209] Feature processing step S41: integrating the cell nucleus feature and the cell feature to obtain an immune cell feature vector;

[0210] Clustering result obtaining step S42: clustering the immune cell feature vectors using the immune cell classification model to obtain clustering results;

[0211] Classification step S43: Identify the morphology of the immune cell population in each cluster in the clustering result to obtain the type of each immune cell.

[0212] The feature processing step S41 includes: integrating the cell nucleus feature and the cell feature to obtain an initial immune cell feature vector; and performing dimensionality reduction processing on the initial immune cell feature vector to obtain a final immune cell feature vector.

[0213] The clustering result obtaining step S42 includes: clustering the final immune cell feature vector using an immune cell classification model obtained through unsupervised learning model training to obtain a clustering result.

[0214] It should be noted that the immune cell classification model of the present invention is obtained through training of an unsupervised learning model. Specifically, first, historical cell nuclear features and historical cell features are obtained through steps S1, S2, and S3 based on multiple historical pathological samples; secondly, multiple historical immune cell feature vectors are obtained through feature processing step S41, which are divided into training sets, test sets, and validation sets; finally, the unsupervised learning model is trained with the training set, test set, and validation set to obtain the final immune cell classification model.

[0215] In this embodiment, after completing the segmentation and feature extraction of the cell nucleus and the entire cell area, the next step is to integrate these features to form a complete feature vector. The construction of the feature vector is the basis for subsequent classification, analysis and model training, and the purpose is to uniformly quantify the morphological characteristics of the cell nucleus and the entire cell, and further identify the cell type through a multidimensional feature vector.

[0216] 1) For each cell, the following features are extracted:

[0217] Geometric characteristics: perimeter p of the nucleus e , the area of ​​the cell nucleus A r , the roundness of the nucleus C, the perimeter of the cell p ec , the area of ​​the cell A re , cell roundness C c , diffusion distance m;

[0218] Physical characteristics: cell clathrate concentration Con(y), cell membrane permeability J;

[0219] Biological characteristics: nuclear-cytoplasmic ratio of cells uc ;

[0220] After integrating these features, a high-dimensional feature vector F is formed:

[0221] F=[P e ,Ar ,C,P ec ,A re ,C c ,N uc ,m,Con(y),J];

[0222] 2) Perform feature dimensionality reduction

[0223] The goal of feature dimensionality reduction is to find the principal component that best explains the data variance from the high-dimensional feature vector F and project the data into a low-dimensional space. Through principal component analysis, the reduced-dimensional data can not only retain most of the useful information, but also reduce redundancy and simplify model training.

[0224] The high-dimensional feature vector F can be expressed as:

[0225] F=[p e ,A r ,C,P ec ,A re ,C c ,N uc ,m,Con(x,y),J];

[0226] The goal is to convert the feature vector F into a low-dimensional feature vector F′ through principal component analysis, retaining as much information as possible.

[0227] First, the original feature vector F is decentralized to eliminate the influence of different feature dimensions. The purpose of decentralization is to make the mean of each feature zero.

[0228] Given a feature matrix X (each row is a sample and each column is a feature), the decentralization process is:

[0229] X centered =X-μ;

[0230] Where μ is the mean vector of the feature matrix X. The mean μ of each feature j The calculation formula is:

[0231]

[0232] X ij is the jth eigenvalue of the ith sample; n is the number of samples.

[0233] 3) Calculate the covariance matrix

[0234] After decentralization, the covariance matrix Σ of the morphological features is calculated:

[0235]

[0236] The covariance matrix Σ is a p×p matrix, where p is the number of features, describing the linear relationship between each pair of features.

[0237] 4) Eigenvalue decomposition

[0238] Perform eigenvalue decomposition on the covariance matrix Σ to obtain the eigenvalue λ and the corresponding eigenvector v:

[0239] Σv=λv;

[0240] Among them, λ is the eigenvalue of the covariance matrix, which reflects the variance in each direction; v is the eigenvector, which indicates the corresponding principal component direction.

[0241] 5) Select the number of principal components

[0242] The eigenvalues ​​are arranged in descending order, representing the variance of each principal component. In order to select the appropriate number of principal components k, we can use the cumulative variance contribution rate as a criterion. The calculation formula for the cumulative variance contribution rate is:

[0243]

[0244] The number k of principal components whose cumulative variance contribution rate reaches 90% or 95% is usually selected to ensure that the feature vector after dimensionality reduction can retain the main information in the data while ignoring noise and minor features.

[0245] 6) Generate the feature vector after dimensionality reduction

[0246] After determining the number of principal components k, the original high-dimensional feature matrix X centered Projected into a low-dimensional space. By selecting the eigenvectors v1, v2, ..., v corresponding to the first k largest eigenvalues k As new basis vectors, the high-dimensional feature vectors are projected onto these principal components. The calculation formula for the reduced-dimensional feature vector X′ is:

[0247] X′=X centered ·V k ;

[0248] Among them, X′ is the feature matrix after dimensionality reduction, with dimension n×k; V k is a matrix consisting of the first k eigenvectors, with dimension p×k.

[0249] Through this operation, the original high-dimensional feature vector F is projected into a low-dimensional space, generating a new reduced-dimensional feature vector F′. The reduced-dimensional feature vector F′ can retain most of the variance information in the original feature while reducing the redundancy and noise of the data.

[0250] 9) Feature clustering of cells based on polarization parameters

[0251] The unsupervised learning algorithm of the present invention uses K-means clustering to classify immune cells into different categories based on the morphological characteristics of the cells. Through K-means clustering, we can identify and classify immune cell populations, especially in complex tumor microenvironments, where different types of immune cells have different functional characteristics. This process will use the morphological characteristics of the cells to help automatically classify immune cell populations such as plasma cells and lymphocytes.

[0252] 9.1 Feature Clustering Process

[0253] Feature preparation: The immune cell feature vector F′ after dimensionality reduction is used as input. The feature matrix composed of the input feature vector can be expressed as:

[0254]

[0255] Among them, F′ i is the feature vector of the ith cell, and n is the total number of cells.

[0256] 9.2 Steps of the K-means Algorithm

[0257] Initialize the centroid: randomly select K centroids τ1, τ2, ..., τ K , each centroid represents the center of a cluster.

[0258] Assign samples to the nearest centroid: According to the Euclidean distance, each immune cell feature vector F′ i Assigned to the cluster to which the nearest centroid belongs. The calculation formula is:

[0259]

[0260] Among them, z i is the cluster to which the ith cell belongs, and the centroid τ with the smallest distance to it is selected j .

[0261] Update centroid: Recalculate the centroid τ of each cluster j , which is the mean of all sample feature vectors in the cluster:

[0262]

[0263] Among them, C j is the set of cells belonging to cluster j, |C j | is the number of cells in the cluster.

[0264] Repeat the steps of assigning samples and updating the centroid until the centroid no longer changes significantly or the specified number of iterations is reached, and finally the cluster label z to which each cell belongs is obtained. i, and output the centroid of each cluster. At this point, each cluster corresponds to a different cell population. Based on its morphological characteristics, we can further analyze the immune cell type represented by each cluster.

[0265] It should be noted that in this embodiment, K-means clustering is a preferred implementation method.

[0266] The classification step S43 includes: identifying the characteristic differences among different clusters based on the centroid and characteristics of each cluster in the clustering result; and naming the clusters and dividing the cell types based on the characteristic differences and biological characteristics.

[0267] Specifically, after completing K-means clustering, the immune cell populations in each cluster will have similar morphology. By analyzing these clusters, different immune cell types can be identified, such as plasma cells, lymphocytes, etc. The specific analysis steps include:

[0268] 1) Analyze the characteristics of each cluster: Analyze the centroid and characteristics of each cluster after clustering to identify the differences in characteristics among different clusters. For example, immune cells in a cluster may show a high nuclear-to-cytoplasmic ratio and strong polarization parameters, which may represent active lymphocytes; another cluster may have lower cell membrane permeability and larger cell area, which represents a plasma cell population.

[0269] 2) Cell classification: Combining the biological characteristics of immune cells and clustering results, cell clusters are classified to obtain corresponding cell types.

[0270] Based on the classification method of the present invention, it is not necessary to stain, immunohistochemically label the pathological samples in the early stage of data collection, so that the immune cells can be classified not only by obtaining the polarization image of the pathological samples. On the other hand, based on the use of unsupervised classification algorithms, the labeling of training data is omitted, thereby reducing the complexity of the operation process and avoiding the loss of natural state information due to the high degree of tissue processing.

[0271] Please refer to Figure 6 , Figure 6 Flow chart of the immune environment assessment method of the present invention. Figure 6 As shown, the present invention also provides an immune environment assessment method based on polarization images, including the immune cell classification method described above, after counting the immune cells in the existing pathological image; based on the classification of the immune cells, the number, proportion and activity of each type of immune cells are calculated; different weights are assigned to each type of immune cells, and the immune environment of the existing sample is evaluated.

[0272] Specifically, the immune environment assessment method includes:

[0273] Polarization image acquisition step S1: collecting data on the pathological sample to obtain polarization data, and processing the polarization data to obtain at least one polarization image;

[0274] Cell nucleus feature acquisition step S2: after performing cell nucleus segmentation on at least one of the polarized images, at least one cell nucleus feature is calculated and obtained;

[0275] Cell feature acquisition step S3: after performing cell segmentation on at least one of the polarized images, at least one cell feature is calculated and obtained;

[0276] Cell identification step S4: obtaining the type of each immune cell in the polarization image through an immune cell classification model based on the cell nucleus feature and the cell feature;

[0277] Evaluation step S5: obtaining an immune environment evaluation result corresponding to the pathological sample based on the type of the immune cells, the cell nuclear characteristics and the cell characteristics.

[0278] Please refer to Figure 7 , Figure 7 for Figure 6 The step-by-step flow chart of step S5 in FIG. Figure 7 As shown, the evaluation step S5 includes:

[0279] Index calculation step S51: calculating an index of each type of the immune cells based on the type of the immune cells, the cell nucleus characteristics and the cell characteristics;

[0280] Immune score acquisition step S52: obtaining an immune score corresponding to the pathological sample through an immune environment comprehensive evaluation model based on the indicators of the immune cells;

[0281] Evaluation result acquisition step S53: obtaining the immune environment evaluation result corresponding to the pathological sample based on the immune score and the preset classification rules.

[0282] Among them, the indicators of immune cells include quantity, proportion and activity.

[0283] In this embodiment, the total number of cells cell_count is calculated based on the polarization image by automatically selecting a threshold using the Otsu method, identifying cells and background, and using the connected region labeling algorithm to assign a unique label L to each connected region. area , and calculate the statistical information S of each connected area area and centroid τ. N is the number of connected regions. L area (x,y) is the label of the pixel at position (x,y). S area (i) = {x min ,y min,w,h,A} is the statistical information of the i-th connected region, where: x min and min is the coordinate of the upper left corner of the connected region. w and h are the width and height of the connected region. A is the area of ​​the connected region. μ i =(u x ,u y ) are the centroid coordinates of the i-th connected region.

[0284] N,L area ,S area ,μ

[0285] =cv2.connectedComponentsWithStats(B,connectivity=8);

[0286] Filter small noise points:

[0287] V={i|S(i)[cv2.CC_STAT_AREA]>min_area,1≤i <N};

[0288] Count the number of connected areas that meet the conditions:

[0289] cell_count=|V|.

[0290] Specifically, after the immune cells are classified according to the aforementioned classification method, the number, proportion, and activity of lymphocytes, plasma cells, and other immune cells in the current region are counted. The proportion is P L ,P P ,P O , the number is N L ,N P ,N O The activities are A L ,A P ,A O .

[0291]

[0292] A L = δF′;

[0293]

[0294] A P = δF′;

[0295]

[0296] Among them, δ is a nonlinear change.

[0297] Establish a comprehensive immune environment assessment model based on the interactions of immune cells;

[0298]

[0299] Among them, N i is the number of the i-th type of immune cells, P i is the proportion of the i-th type of immune cells, A i is the activity of the i-th type of immune cells;

[0300] Define features to capture nonlinear relationships and interaction effects, f(P i ) is the nonlinear transformation of the proportion, f(P i )=log(P i +1). k(N i ,A j ) is the interaction term between quantity and activity, for example, k(N i ,A j )=N i ·A j ;

[0301] The objective function is to minimize the residual sum of squares (RSS) to determine the weights.

[0302]

[0303] According to the determined weight w i and u ij , calculate the immune score S of each sample score , and classify the samples into different immune states according to the scores. The upper and lower limits of the classification interval can be determined according to the scores of the immune scores of multiple samples, and the classification interval is determined based on the scores and the upper and lower limits. That is, in this embodiment, the immune environment is divided into seven classification intervals: immune failure, moderate suppression, mild suppression, stable, mildly active, moderately active, and immune attack. For example, among tens of thousands of samples from n patients, the statistically obtained immune score is between 0.1 and 0.7, thereby determining the upper limit of 0.7 and the lower limit of 0.1, the immune score of 0.1 is immune failure, the immune score of 0.1-0.2 is moderate suppression, the immune score of 0.2-0.3 is mild suppression, the immune score of 0.3-0.4 is stable, the immune score of 0.4-0.5 is mildly active, the immune score of 0.5-0.6 is moderately active, and the immune score of 0.6-0.7 is immune attack. It should be noted that the above immune score values ​​and the interval setting method of the classification rules are only for illustrative purposes. Users can determine the classification interval through different calculation methods according to the actual immune score values.

[0304] Please refer to Figure 5 , Figure 5 FIG. 1 is a schematic diagram of the structure of the immune cell classification system of the present invention. Figure 5As shown, an immune cell classification system based on polarization images of the present invention applies the immune cell classification method described above, and the immune cell classification system includes:

[0305] A polarization image acquisition unit 11 collects data from the pathological sample to obtain polarization data, and processes the polarization data to obtain at least one polarization image;

[0306] A cell nucleus feature acquisition unit 12, which performs cell nucleus segmentation on at least one of the polarization images and calculates and obtains at least one cell nucleus feature;

[0307] A cell feature acquisition unit 13, which performs cell segmentation on at least one of the polarization images and calculates and obtains at least one cell feature;

[0308] The cell identification unit 14 obtains the type of each immune cell in the polarization image through an immune cell classification model based on the cell nucleus feature and the cell feature.

[0309] Wherein, after the polarization image acquisition unit 11 collects data on the pathological sample to obtain multiple groups of polarization data, the multiple groups of polarization data are calculated and processed to obtain the polarization parameters; finally, the polarization parameters are visualized to obtain the polarization image.

[0310] Furthermore, the cell nucleus feature acquisition unit 12 includes:

[0311] The first segmentation module 121 performs global threshold segmentation on the polarization image to obtain at least one cell nucleus region;

[0312] The first calculation module 122 calculates each of the cell nucleus regions to obtain at least one cell nucleus feature.

[0313] The cell feature acquisition unit 13 comprises:

[0314] The second segmentation module 131 performs global threshold segmentation on the polarization image to obtain at least one cell region;

[0315] The second calculation module 132 calculates each of the cell regions to obtain at least one cell feature.

[0316] The first segmentation module 121 performs weighted integration of at least one of the polarization images by a first weight factor to obtain a first weighted boundary characteristic map; the first weighted boundary characteristic map is subjected to global threshold segmentation by the Otsu method to obtain at least one of the cell nucleus regions;

[0317] The second segmentation module 131 performs weighted integration on at least one of the polarization images using a second weight factor to obtain a second weighted boundary characteristic map; and performs global threshold segmentation on the second weighted boundary characteristic map using the Otsu method to obtain at least one cell region.

[0318] Furthermore, the cell identification unit 14 includes:

[0319] A feature processing module 141 integrates the cell nucleus feature and the cell feature to obtain an immune cell feature vector;

[0320] A clustering result obtaining module 142, clustering the immune cell feature vectors by using the immune cell classification model to obtain a clustering result, wherein the immune cell classification model is obtained by training an unsupervised learning model;

[0321] The classification module 143 identifies the morphology of the immune cell population in each cluster in the clustering result to obtain the type of each immune cell.

[0322] Please refer to Figure 8 , Figure 8 FIG. 1 is a schematic diagram of the structure of the immune environment assessment system of the present invention. Figure 8 As shown, an immune environment assessment system based on polarization images of the present invention, wherein the immune environment assessment method described above is applied, and the immune environment assessment system comprises:

[0323] A polarization image acquisition unit 11 collects data from the pathological sample to obtain polarization data, and processes the polarization data to obtain at least one polarization image;

[0324] A cell nucleus feature acquisition unit 12, which performs cell nucleus segmentation on at least one of the polarization images and calculates and obtains at least one cell nucleus feature;

[0325] A cell feature acquisition unit 13, which performs cell segmentation on at least one of the polarization images and calculates and obtains at least one cell feature;

[0326] A cell identification unit 14, which obtains the type of each immune cell in the polarization image through a preset immune cell classification model based on the cell nucleus feature and the cell feature;

[0327] The evaluation unit 15 obtains an immune environment evaluation result corresponding to the pathological sample based on the type of the immune cell, the cell nucleus feature and the cell feature.

[0328] Furthermore, the evaluation unit 15 comprises:

[0329] An index calculation module 151 calculates an index of each type of the immune cells based on the type of the immune cells, the cell nuclear characteristics and the cell characteristics;

[0330] An immune score acquisition module 152, which obtains an immune score corresponding to the pathological sample through an immune environment comprehensive evaluation model based on the indicators of the immune cells;

[0331] The evaluation result acquisition module 153 obtains the immune environment evaluation result corresponding to the pathological sample according to the immune score based on a preset classification rule.

[0332] In summary, for pathologists, the present invention provides the functions of immune cell classification and activity detection, simplifies the user operation process, and performs label-free imaging of samples to ensure the integrity of information. In addition, the contrast and resolution of pathological samples are improved, which is convenient for users to analyze microstructures, improve diagnostic efficiency and accuracy, determine the types and functional activities of different immune cells, evaluate the immune environment, and then infer the prognosis of tumors for medication guidance.

[0333] Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent substitutions for some of the technical features therein; and these modifications or substitutions do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for classifying immune cells based on polarization images, characterized in that: include: Polarization image acquisition step: collecting data on the pathological sample to obtain polarization data, and processing the polarization data to obtain at least one polarization image; The cell nucleus feature acquisition step: after performing cell nucleus segmentation on at least one of the polarized images, at least one cell nucleus feature is calculated and obtained; Cell feature acquisition step: after performing cell segmentation on at least one of the polarized images, at least one cell feature is calculated and obtained; Cell identification step: obtaining the type of each immune cell in the polarization image through an immune cell classification model based on the cell nucleus characteristics and the cell characteristics.

2. The immune cell classification method according to claim 1, characterized in that: The cell nucleus feature acquisition step comprises: The first segmentation step: performing global threshold segmentation on the polarization image to obtain at least one cell nucleus region; The first calculation step: calculating each of the cell nucleus regions to obtain at least one cell nucleus feature.

3. The immune cell classification method according to claim 2, characterized in that: The cell feature acquisition step comprises: The second segmentation step: performing global threshold segmentation on the polarization image to obtain at least one cell region; The second calculation step: calculating each of the cell regions to obtain at least one cell feature.

4. The immune cell classification method according to claim 3, characterized in that: The polarization image acquisition step includes processing the polarization data to obtain at least one polarization parameter, and visualizing at least one polarization parameter to obtain at least one polarization image; Wherein, the polarization parameters include at least one of depolarization, linear phase delay and fast axis azimuth, the polarization image includes at least one of depolarization image, linear phase delay image and fast axis azimuth image, the cell nuclear characteristics in the first calculation step include: at least one of cell nuclear circumference, cell nuclear area and cell nuclear roundness, and the cell characteristics in the second calculation step include: at least one of cell circumference, cell area, cell membrane roundness, cell nucleus-cytoplasm ratio, diffusion distance, cell cytoplasm concentration and cell membrane permeability.

5. The immune cell classification method according to claim 4, characterized in that: The first segmentation step includes: Performing weighted integration on at least one of the polarization images using a first weight factor to obtain a first weighted boundary characteristic map; Performing global threshold segmentation on the first weighted boundary characteristic map by Otsu method to obtain at least one cell nucleus region; The second segmentation step includes: Performing weighted integration on at least one of the polarization images using a second weight factor to obtain a second weighted boundary characteristic map; The second weighted boundary characteristic map is subjected to global threshold segmentation by the Otsu method to obtain at least one cell region.

6. The immune cell classification method according to claim 1, characterized in that: The cell identification step comprises: Feature processing step: integrating the cell nucleus feature and the cell feature to obtain an immune cell feature vector; A clustering result obtaining step: clustering the immune cell feature vectors using the immune cell classification model to obtain a clustering result, wherein the immune cell classification model is obtained by training an unsupervised learning model; Classification step: Identify the morphology of the immune cell population in each cluster in the clustering result to obtain the type of each immune cell.

7. An immune cell classification system based on polarization images, characterized in that: Using the immune cell classification method described in any one of claims 1 to 6, the immune cell classification system comprises: A polarization image acquisition unit collects data from the pathological sample to obtain polarization data, and processes the polarization data to obtain at least one polarization image; A cell nucleus feature acquisition unit, which performs cell nucleus segmentation on at least one of the polarization images and calculates and obtains at least one cell nucleus feature; A cell feature acquisition unit, which performs cell segmentation on at least one of the polarization images and calculates and obtains at least one cell feature; The cell recognition unit obtains the type of each immune cell in the polarization image through an immune cell classification model based on the cell nucleus feature and the cell feature.

8. An immune environment assessment method based on polarization images, characterized in that: Using the immune cell classification method described in any one of claims 1 to 6, the immune environment assessment method comprises: Polarization image acquisition step: collecting data on the pathological sample to obtain polarization data, and processing the polarization data to obtain at least one polarization image; The cell nucleus feature acquisition step: after performing cell nucleus segmentation on at least one of the polarized images, at least one cell nucleus feature is calculated and obtained; Cell feature acquisition step: after performing cell segmentation on at least one of the polarized images, at least one cell feature is calculated and obtained; Cell identification step: obtaining the type of each immune cell in the polarization image through an immune cell classification model based on the cell nucleus characteristics and the cell characteristics; Evaluation step: obtaining an immune environment evaluation result corresponding to the pathological sample based on the type of the immune cells, the characteristics of the cell nucleus and the characteristics of the cells.

9. The immune environment assessment method according to claim 8, characterized in that: The evaluation steps include: Index calculation step: calculating the index of each type of the immune cells based on the type of the immune cells, the characteristics of the cell nucleus and the characteristics of the cells; Immuno-score acquisition step: obtaining an immune score corresponding to the pathological sample through an immune environment comprehensive evaluation model based on the indicators of the immune cells; Evaluation result acquisition step: obtaining the immune environment evaluation result corresponding to the pathological sample according to the immune score based on the preset classification rules.

10. An immune environment assessment system based on polarization images, characterized in that: The immune environment assessment method according to any one of claims 8 to 9 is applied, wherein the immune environment assessment system comprises: A polarization image acquisition unit collects data from the pathological sample to obtain polarization data, and processes the polarization data to obtain at least one polarization image; A cell nucleus feature acquisition unit, which performs cell nucleus segmentation on at least one of the polarization images and calculates and obtains at least one cell nucleus feature; A cell feature acquisition unit, which performs cell segmentation on at least one of the polarization images and calculates and obtains at least one cell feature; A cell recognition unit, which obtains the type of each immune cell in the polarization image through an immune cell classification model based on the cell nucleus feature and the cell feature; An evaluation unit is configured to obtain an immune environment evaluation result corresponding to the pathological sample based on the type of the immune cell, the cell nuclear characteristics and the cell characteristics.

Images