Immune microenvironment characteristic-based hepatocellular carcinoma treatment effect evaluation method and system

Abstract

The method and system for evaluating the curative effect of hepatocellular carcinoma based on immune microenvironment characteristics are high in accuracy and reliability of evaluating the curative effect of hepatocellular carcinoma. The method comprises the following steps: (1) preparing a sample; (2) carrying out multiple immunohistochemical staining; (3) image acquisition and analysis; (4) analyzing spatial distribution characteristics; (5) multiplying the input image voxel signal numerical value by the weight vector of the convolution kernel, transmitting the multilayer convolution kernel, and finally obtaining a calculation result output by each network; and (6) establishing a prediction model: based on the spatial distribution characteristics and clinical characteristics of immune cells, constructing the prediction model by using the positive rate of CD8 + PD-L1 + and CD103 + CD8 + cells in the total area and the number of CD103 + cells within 30 microns around tumor cells so as to predict the immunotherapy curative effect of the HCC patient.

Description

Technical Field

[0001] The present invention relates to the technical field of biomedicine, and in particular to a method for evaluating the therapeutic efficacy of hepatocellular carcinoma based on immune microenvironment characteristics, and a system for evaluating the therapeutic efficacy of hepatocellular carcinoma based on immune microenvironment characteristics. Background Art

[0002] Liver cancer is the sixth most common malignant tumor globally and the third leading cause of cancer-related deaths. Hepatocellular carcinoma (HCC) predominates in liver cancer cases, accounting for approximately 90% of all cases. The preferred treatment methods for early HCC patients include surgical resection, transplantation, and local ablation techniques, while systemic treatment is usually the main treatment method for advanced patients. Targeted drugs such as sorafenib and lenvatinib have significantly prolonged the survival of advanced liver cancer patients. Although immunotherapy has good efficacy and relatively good safety, the effectiveness of immunotherapy is still limited, and immune-related toxicity cannot be ignored. Therefore, there is an urgent need to discover biomarkers that can predict the efficacy of immunotherapy and distinguish patients who can benefit from immunotherapy, so as to better guide the clinical treatment of HCC.

[0003] The tumor immune microenvironment helps to understand the processes of tumorigenesis, development, and metastasis, and has an important impact on the efficacy of immunotherapy. The liver is a key organ of immune tolerance. Hepatocellular carcinoma (HCC) often originates from chronic inflammations such as viral hepatitis and non-alcoholic fatty hepatitis, and its microenvironment composition is very complex and highly heterogeneous. By studying the numbers of tumor cells and various inflammatory cells, the expression of inflammation-related factors, and their spatial distributions, the immune microenvironment of HCC can generally be divided into immune-activated type, inflammatory type, excluded type, and desert-like type. In addition to lymphocytes and neutrophils, the numbers and spatial distributions of tertiary lymphoid structures, macrophages, B cells, and natural killer cells in cancer tissues also affect the evaluation of the immune microenvironment of HCC. With the development of HCC immunotherapy, the relationship between immune efficacy and the immune microenvironment has been attracting increasing attention. For example, the use of antibodies against PD-1, TIM-3, and LAG-3 can significantly restore the functions of tumor-infiltrating lymphocytes (TILs) in the tumor microenvironment. This restoration enhances the ability of TILs to fight against tumor cells, thereby improving the overall anti-tumor activity. The presence of inflammatory cells such as myeloid-derived suppressor cells and certain macrophages usually indicates poor prognosis. These results suggest that studying the tumor immune microenvironment of HCC is crucial for predicting the prognosis of HCC immunotherapy. Although the expression of PD-L1 in tumor cells has been regarded as a potential biomarker for predicting the responses of many types of cancers to immune checkpoint inhibitors (ICIs), its correlation with the prognosis of HCC remains unclear. The relationship between the tumor mutation burden and the response rate or overall survival (OS) in HCC has also not been clarified. Therefore, there is still a lack of biomarkers related to the tumor immune microenvironment of HCC to predict the efficacy of immunotherapy.

[0004] Due to the complexity of the tumor immune microenvironment, there are still some limitations in existing detection technologies: Immunohistochemistry (IHC) can only detect the expression of single or a small number of protein markers, cannot accurately identify specific immune cell subsets and functional states, and the results are easily affected by tissue heterogeneity, sampling bias, and subjective factors such as staining intensity and background noise; Flow cytometry requires fresh tumor tissue samples, which are difficult to obtain, and can only detect cell surface markers, cannot detect intracellular markers, and cannot obtain the spatial distribution information of cells; Single-cell sequencing is costly and technically difficult, and cannot obtain the morphological characteristics of cells; Spatial transcriptomics technology is not yet mature, difficult to apply on a large scale, and cannot detect protein expression.

[0005] Existing protocols for predicting and evaluating the efficacy of tumor immunotherapy have the following problems:

[0006] (1) Limited predictive ability of single indicators: For indicators such as PD-L1 expression and tumour mutation burden, their predictive value is not yet clear, and they cannot comprehensively reflect the complexity and dynamic changes of the TIME (tumour immune microenvironment).

[0007] (2) Lack of spatial information: Most of the existing predictive indicators only focus on the number or proportion of immune cells, while ignoring their spatial distribution characteristics in tumour tissues, which may affect the accuracy of prediction.

[0008] (3) Lack of individualized analysis: Existing predictive methods are difficult to take into account the individual differences of liver cancer patients, such as tumour heterogeneity, clinical characteristics and immune status, etc., so they cannot provide personalized treatment plans for patients. Summary of the Invention

[0009] To overcome the defects of the prior art, the technical problem to be solved by the present invention is to provide a method for evaluating the therapeutic efficacy of hepatocellular carcinoma based on immune microenvironment characteristics, which has high accuracy and strong reliability in evaluating the therapeutic efficacy of hepatocellular carcinoma.

[0010] The technical solution of the present invention is: A method for evaluating the therapeutic efficacy of hepatocellular carcinoma based on immune microenvironment characteristics, which includes the following steps:

[0011] (1) Sample preparation: Collect tumour tissue samples of hepatocellular carcinoma patients and perform pre-treatment before sample staining such as fixation and sectioning.

[0012] (2) Multiplex immunohistochemical staining: Use a specific antibody combination to perform multiplex immunohistochemical staining on the sections to simultaneously detect multiple immune cell types and molecular markers.

[0013] (3) Image acquisition and analysis: Use digital pathology equipment to perform high-precision image acquisition on the stained sections, and use image analysis software to identify and count immune cells.

[0014] (4) Analysis of spatial distribution characteristics: Use machine learning algorithms to quantitatively analyze the spatial distribution characteristics of immune cells, including cell density and distance from tumour cells.

[0015] (5) Multiply the input image voxel signal values by the weight vectors of the convolutional kernels and perform the transfer of multiple convolutional kernels, and finally obtain the calculation results output by each network.

[0016] (6) Prediction model establishment: Based on the spatial distribution characteristics and clinical characteristics of immune cells, use the total area of CD8 + PD-L1 + and CD103 + CD8+ The positive rate of cells and CD103 within 30 μm around tumor cells + The number of cells is used to construct a prediction model for predicting the efficacy of immunotherapy in HCC patients.

[0017]

[0018] The present invention collects MR images of patients and control groups from multiple medical centers, performs data preprocessing, constructs five deep learning network models including CNN, LeNet, VGG16, ResNet, and DenseNet, divides the data into training data sets, internal validation data sets, and external validation data sets, enhances the data in the training data set, retains the weight coefficients of different layers of each network in the neural learning network model, multiplies the input image voxel signal values by the weight vectors of the convolution kernels, and performs the transfer of multiple convolution kernels. Finally, the calculation results output by each network are obtained, and the optimal model is evaluated. Therefore, multiple models are tested in multi-center data to detect MMA in normal controls and non-MMA vascular diseases, increase the sample size of training, have strong generality, prevent overfitting, and have high detection accuracy.

[0019] It also provides a hepatocellular carcinoma treatment efficacy evaluation system based on immune microenvironment characteristics, which includes:

[0020] A sample preparation module configured to collect tumor tissue samples of hepatocellular carcinoma patients and perform preprocessing before fixing, sectioning, and sample staining;

[0021] A multiplex immunohistochemical staining module configured to perform multiplex immunohistochemical staining on the sections using an antibody combination to simultaneously detect multiple immune cell types and molecular markers;

[0022] An image acquisition and analysis module configured to use a digital pathology device to perform high-precision image acquisition on the stained sections and use image analysis software to identify and count immune cells;

[0023] A spatial distribution feature analysis module configured to quantitatively analyze the spatial distribution features of immune cells using machine learning algorithms, including cell density and distance from tumor cells;

[0024] A calculation module configured to multiply the input image voxel signal values by the weight vectors of the convolution kernels and perform the transfer of multiple convolution kernels, and finally obtain the calculation results output by each network;

[0025] A prediction model establishment module configured to, based on the spatial distribution features and clinical features of immune cells, use the total area of CD8 + PD-L1 + and CD103 + CD8 +The positive rate of cells and CD103 within 30 μm around tumor cells + The number of cells is used to construct a prediction model to predict the efficacy of immunotherapy for HCC patients. Brief Description of the Drawings

[0026] Figure 1 Shows a flowchart of a method for evaluating the therapeutic efficacy of hepatocellular carcinoma based on immune microenvironment characteristics according to the present invention.

[0027] Figure 2 Shows a technical roadmap of a method for evaluating the therapeutic efficacy of hepatocellular carcinoma based on immune microenvironment characteristics according to the present invention.

[0028] Figure 3 Shows H&E staining and panel mIHC staining diagrams.

[0029] Figure 4 Shows the survival curve of the positive rate of CD8 in the total area + PD-1 + cells positive rate survival curve.

[0030] Figure 5 Shows the survival curve of the positive rate of CD103 in the total area + CD8 + cells positive rate survival curve.

[0031] Figure 6 Shows the survival curve of spatial characteristics.

[0032] Figure 7 Shows the TIS survival curve. Detailed Description of the Invention

[0033] As Figure 1 、 Figure 2 shown, a method for evaluating the therapeutic efficacy of hepatocellular carcinoma based on immune microenvironment characteristics includes the following steps:

[0034] (1) Sample preparation: Collect tumor tissue samples from hepatocellular carcinoma patients and perform preprocessing before sample staining such as fixation and sectioning;

[0035] (2) Multiplex immunohistochemical staining: Use a specific antibody combination to perform multiplex immunohistochemical staining on the sections to simultaneously detect multiple types of immune cells and molecular markers;

[0036] (3) Image acquisition and analysis: Use digital pathology equipment to perform high-precision image acquisition on the stained sections, and use image analysis software to identify and count immune cells;

[0037] (4) Spatial distribution feature analysis: Use machine learning algorithms to quantitatively analyze the spatial distribution features of immune cells, including cell density and distance from tumor cells;

[0038] (5) Multiply the input image voxel signal values by the weight vectors of the convolutional kernels, and perform the transfer of multiple convolutional kernels, finally obtaining the calculation results output by each network.

[0039] (6) Prediction model establishment: Based on the spatial distribution characteristics and clinical characteristics of immune cells, use the total regional CD8 + PD-L1 + and CD103 + CD8 + positive cell rate, and the number of CD103 + cells within 30 μm around tumor cells to construct a prediction model for predicting

[0040] the efficacy of immunotherapy for HCC patients.

[0041] The present invention collects MR images of patients and control groups from multiple medical centers, performs data preprocessing, constructs five deep learning network models including CNN, LeNet, VGG16, ResNet, and DenseNet, divides the data into training data sets, internal validation data sets, and external validation data sets, enhances the data in the training data set, retains the weight coefficients of different layers of each network in the neural learning network model, multiplies the input image voxel signal values by the weight vectors of the convolutional kernels, and performs the transfer of multiple convolutional kernels, finally obtaining the calculation results output by each network, and evaluates to obtain the optimal model. Therefore, multiple models are tested in multi-center data to detect MMA in normal controls and non-MMA vascular diseases, increase the sample size of training, have strong versatility, prevent overfitting, and have high detection accuracy.

[0042] Preferably, in the step (1), use formalin-fixed paraffin-embedded pre-treatment tumor tissue samples. The tissue samples are collected within 30 minutes after sampling, stored in 10% formalin for 24 to 48 hours, and subjected to tissue dehydration, paraffin embedding, and sectioning within 2 working days. Five consecutive sections are cut, and 1 section is used for hematoxylin and eosin staining. The sections are required to be complete, without contamination or wrinkles, and have a thickness of 4 μm. Select a suitable patient population and collect high-quality tumor tissue samples to provide a basis for subsequent implementation. The patient population included in this step is representative, and the sample collection process is standardized, providing a reliable basis for subsequent implementation.

[0043] Preferably, in the step (2), the markers form 4 panels. The first template includes: CD8, LAG3, PD1, TIGIT, TIM3; the second template includes: CD8, PDL1, CD103, CLAUDIN182, PANCK; the third template includes: CD4, CD68, CD163, FOXP3, GPC3; the fourth template includes: HLADR, PANCK, CD66B, CD14, CD11C.

[0044] Biomarkers in the tumor immune microenvironment that may affect the efficacy of immunotherapy in patients with hepatocellular carcinoma:

[0045] (1) Tumor cell-specific markers: GPC3, Claudin18.2;

[0046] (2) Immunocyte subtype markers: CD8 (Cytotoxic T cell), CD4 (THelper cell), Foxp3 (Regulatory T cell), CD103 (Tissue-resident Memory T), CD68 (Macrophages), CD163 (Macrophages);

[0047] (3) Immune regulatory molecule markers: PD-1, PD-L1, LAG-3, TIM-3, TIGIT;

[0048] (4) Inflammatory cell molecule markers: CD66B, CD11C, CD14, HLA-DR.

[0049] The designs of panels 1-4 are as follows:

[0050] Table 1

[0051]

[0052]

[0053] Table 2

[0054]

[0055] Table 3

[0056]

[0057] Table 4

[0058]

[0059] Preferably, the multiplex immunohistochemical staining in step (2) includes the following steps:

[0060] (2.1) Dehydration: The FFPE tumor section is placed at 60 °C and heated for 12 h;

[0061] (2.2) Deparaffinization: Deparaffinize with xylene, 10 minutes / time, repeat 3 times;

[0062] (2.3) Hydration: 100% ethanol for 5 min, 95% ethanol for 5 min, 70% ethanol for 2 min;

[0063] (2.4) Antigen retrieval: Heat in EDTA buffer at pH 9.0 or citrate buffer at pH 6.0 for FoxP3;

[0064] (2.5) Primary antibody incubation: Incubate the primary antibody at room temperature for 10 min;

[0065] (2.6) Secondary antibody incubation: Dropwise add the HRP secondary antibody working solution and incubate at room temperature for 10 minutes;

[0066] (2.7) Fluorescent staining signal amplification: Enhance the fluorescent signal by tyramide signal amplification (TSA) technology;

[0067] (2.8) Repeat staining: After each antibody staining, repeat antigen retrieval and perform the next round of antibody staining until all markers in a single panel are stained;

[0068] (2.9) DAPI staining: After the section is air-dried, dropwise add the DAPI staining solution and incubate at room temperature in the dark for 10 min;

[0069] (2.10) Mounting: Rinse with 1×TBST by shaking for 3 times × 3 min / time, mount with anti-fluorescence quenching agent and store in the dark;

[0070] (2.11) Reading slides: Read the staining results through a high-throughput panoramic scanner;

[0071] (2.12) Quantitative statistical analysis: Perform image and statistical analysis through a panoramic pathology workstation.

[0072] Examples of staining results are as Figure 3 shown.

[0073] Through summary and data mining, tumor microenvironment biomarkers closely related to HCC immunotherapy were summarized and refined;

[0074] By carefully designing the detection panel, the uniqueness of the panel was achieved, and at the same time, the co-localization of multiple biomarkers and the spatial feature analysis within the panel became possible, comprehensively covering the biomarkers that can be used to predict the efficacy of hepatocellular carcinoma immunotherapy.

[0075] Preferably, step (3) includes the following steps:

[0076] (3.1) Take a whole-slide fluorescence image using an Olympus VS200 system equipped with an Olympus UPLXAPO 20× objective lens;

[0077] (3.2) Region division: Analyze the whole-slide fluorescence image on a PanoATLAS workstation using QuPath image analysis software. Determine the region of interest for analysis (ROA) based on the H&E-stained tissue section. Use the algorithms embedded in QuPath to perform accurate tissue classification through a labeling, training, and validation process, classifying it into tumor regions, stromal regions, and

[0078] normal tissue regions;

[0079] (3.3) Cell splitting: Use the algorithms in QuPath software to split cells into individual cells and identify nuclear, cytoplasmic, and cell membrane structures; perform splitting through DAPI staining and PANCK staining;

[0080] (3.4) Cell phenotype analysis: Analyze the phenotypes of cells based on the positive status and relative intensity of markers;

[0081]

[0082] (3.5) Positive rate calculation and analysis: Calculate the positive rate of each type of cell in each ROA. Calculate separately for different subtypes of positive cells. The cell positive rate is the ratio of the number of positive cells in the field of view to the total number of cells in the field of view, calculated according to the following formula:

[0083] Cell positive rate = (ΣC_P) / (ΣC)

[0084] where cells refer to cells of different subtypes, calculate the proportion separately for different subtypes, C is the cell count in the field of view; C_P is the count of phenotypic cells stained positively in C;

[0085] (3.6) Spatial analysis: Locate positive cells, calculate the distance between positive cells, and analyze the characteristics of the spatial organization of cells. Locate positive cells as two-dimensional Cartesian coordinate system data. The default data is pixel coordinates based on the rectangular image field of view. After converting to length coordinate data according to the magnification, calculate the distance between positive cells. The distance is calculated using the following Euclidean distance calculation

[0086] method:

[0087]

[0088] where d(x, y) represents the Euclidean distance between cell x and cell y, x 1 、x 2 represent cells​

[0089] The horizontal and vertical coordinates of x, y 1 ,y 2 Represents the horizontal and vertical coordinates of the cell y.

[0090] Preferably, in step (3.2), the ROA delineation rule / quality control rule is:

[0091] Areas with no or little DAPI nuclear staining: No DAPI staining, regardless of whether other channels are stained, are all excluded;

[0092] A small amount / sporadic DAPI staining, if it is blood and muscle tissue, all will be removed;

[0093] There was a DAPI-stained exclusion of non-cellular components;

[0094] Folded area: Folded areas caused by tissue detachment in the stained scan image are removed;

[0095] The scattered areas outside the overall sample are eliminated;

[0096] Normal epithelial tissue was removed.

[0097] By utilizing artificial intelligence and machine interpretation in the analysis process, the stability of the analysis results is ensured and the subjectivity of human interpretation is avoided. The tumor immune microenvironment characteristic data of 78 liver cancer patients were successfully obtained, including the number, phenotype and spatial distribution of cells. Combined with the corresponding clinical data and statistical analysis, it was revealed that the composition and spatial characteristics of TIICs are of great significance to the prognosis of liver cancer patients, and can be used as a potential biomarker to predict the efficacy of immunotherapy and patient survival.

[0098] Preferably, the step (6) comprises the following steps:

[0099] (6.1) Randomly divide the samples into training set and validation set;

[0100] (6.2) Use three machine learning algorithms, Logistic regression, random forest and gradient boosting classifier, to build a prediction model based on TIME feature TIS; use five-fold cross validation to improve the robustness of the model; use CD8 + PD-L1 + and CD103 + CD8 + The positive rate of cells and CD103 within 30 μm around tumor cells + The number of cells is used to define TIS.

[0101] Preferably, step (6) further includes step (6.3): constructing a multivariate COX proportional hazards model based on TIS, dividing patients into a high TIS group and a low TIS group, and calculating the difference in survival probabilities of PFS and OS between the two groups in the model constructed with TIS by the Kaplan-Meier method.

[0102] It shows that the established TIS model can effectively predict the efficacy of immunotherapy and has significant prognostic value. The machine learning model provides a more accurate tool for clinical applications, which can help screen HCC patients suitable for immunotherapy and contribute to the development of personalized immunotherapy regimens.

[0103] Those of ordinary skill in the art can understand that all or part of the steps in implementing the method of the above embodiments can be completed by instructing relevant hardware through a program. The program can be stored in a computer-readable storage medium. When the program is executed, it includes the steps of the method of the above embodiments, and the storage medium can be: ROM / RAM, magnetic disk, optical disk, memory card, etc. Therefore, corresponding to the method of the present invention, the present invention also simultaneously includes a hepatocellular carcinoma treatment efficacy evaluation system based on immune microenvironment characteristics, and this system is usually represented in the form of functional modules corresponding to the steps of the method. This system includes:

[0104] A sample preparation module configured to collect tumor tissue samples of hepatocellular carcinoma patients and perform preprocessing before sample staining such as fixation and sectioning;

[0105] A multiplex immunohistochemistry staining module configured to perform multiplex immunohistochemistry staining on the sections using an antibody combination to simultaneously detect multiple immune cell types and molecular markers;

[0106] An image acquisition and analysis module configured to use a digital pathology device to perform high-precision image acquisition on the stained sections and use image analysis software to identify and count immune cells;

[0107] A spatial distribution feature analysis module configured to quantitatively analyze the spatial distribution features of immune cells using machine learning algorithms, including cell density and distance from tumor cells;

[0108] A calculation module configured to multiply the input image voxel signal values by the weight vectors of the convolutional kernels and perform the transfer of multiple convolutional kernels, and finally obtain the calculation results output by each network;

[0109] A prediction model establishment module configured to, based on the spatial distribution features and clinical features of immune cells, adopt the total area of CD8 + PD-L1 + and CD103 + CD8 +The positive rate of cells and CD103 within 30 μm around tumor cells + The number of cells is used to construct a prediction model to predict the efficacy of immunotherapy in HCC patients.

[0110] Preferably, the image acquisition and analysis module performs the following steps:

[0111] (3.1) Take a whole-slide fluorescence image using an Olympus VS200 system equipped with an Olympus UPLXAPO 20× objective lens;

[0112] (3.2) Region division: Analyze the whole-slide fluorescence image on a PanoATLAS workstation using QuPath image analysis software. Determine the region of interest for analysis (ROA) based on the H&E-stained tissue section. Use the algorithms embedded in QuPath to perform accurate tissue classification through a labeling, training, and validation process, classifying it into tumor regions, stromal regions, and normal tissue regions;

[0113] (3.3) Cell splitting: Use the algorithms in QuPath software to split cells into individual cells and identify the nucleus, cytoplasm, and cell membrane structures; perform splitting through DAPI staining and PANCK staining;

[0114] (3.4) Cell phenotype analysis: Perform phenotype analysis on cells based on the positive status and relative intensity of markers;

[0115] (3.5) Positive rate calculation and analysis: Calculate the positive rate of each type of cell in each ROA. Calculate separately for different subtypes of positive cells. The cell positive rate is the ratio of the number of positive cells in the field of view to the total number of cells in the field of view, calculated according to the following formula:

[0116] Cell positive rate = (ΣC_P) / (ΣC)

[0117] Where cells refer to different subtypes of cells, and the ratio is calculated separately for different subtypes. C is the cell count in the field of view; C_P is the count of phenotypic cells stained positively in C;

[0118] (3.6) Spatial analysis: Locate positive cells, calculate the distance between positive cells, and analyze the spatial tissue distribution characteristics of cells. The positive cell location is data in a two-dimensional Cartesian coordinate system. The default data is pixel coordinates based on the rectangular image field of view, which is converted to length coordinate data according to the magnification factor and then the distance between positive cells is calculated. The distance is calculated using the following Euclidean distance calculation method:

[0119]

[0120] Where d(x, y) represents the Euclidean distance between cell x and cell y, x 1 、x2 represent the abscissa and ordinate of cell x, y 1 and y 2 represent the abscissa and ordinate of cell y.

[0121] Compared with the prior art, the present invention has the following advantages:

[0122] (1) Through the mIHC technology, the present invention can not only simultaneously detect the positive rates of multiple immune cells and their subsets, but also analyze their spatial distribution characteristics. The combination of the two can more comprehensively reflect the complexity of the HCC tumor immune microenvironment.

[0123] (2) The targets involved in the present invention are derived from recent literature reports, with a wide range, and more comprehensively cover the immune cell subsets of the HCC tumor immune microenvironment.

[0124] (3) The present invention combines the method of machine learning to analyze the positive rates of immune cell subsets and the spatial distribution information in situ in the HCC tumor immune microenvironment, and constructs a prediction model, which can significantly distinguish the survival status of HCC patients.

[0125] The above are only the preferred embodiments of the present invention, and do not impose any form of limitation on the present invention. Any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention still fall within the protection scope of the technical solution of the present invention.

Claims

1. A method for evaluating the therapeutic efficacy of hepatocellular carcinoma based on immune microenvironment characteristics, characterized by: It includes the following steps: (1) Sample preparation: Tumor tissue samples from patients with hepatocellular carcinoma were collected and fixed and sectioned before sample staining; (2) Multiplex immunohistochemical staining: Multiplex immunohistochemical staining of sections using specific antibody combinations to simultaneously detect multiple immune cell types and molecular markers; (3) Image acquisition and analysis: Use digital pathology equipment to acquire high-precision images of stained sections, and use image analysis software to identify and count immune cells; (4) Spatial distribution characteristics analysis: Use machine learning algorithms to quantitatively analyze the spatial distribution characteristics of immune cells, including cell density and distance from tumor cells; (5) Multiply the input image voxel signal value and the weight vector of the convolution kernel, and transfer the multi-layer convolution kernel to finally obtain the calculation result of each network output; (6) Prediction model establishment: Based on the spatial distribution characteristics and clinical characteristics of immune cells, the total regional CD8 + PD-L1 + and CD103 + CD8 + Cell positive rate, CD103 within 30μm around tumor cells + Cell number to build a prediction model to predict HCC Immunotherapy efficacy in patients.

2. The method for evaluating the therapeutic efficacy of hepatocellular carcinoma based on immune microenvironment characteristics according to claim 1, characterized in that: In the step (1), a formalin-fixed paraffin-embedded tumor tissue sample before treatment is used. The tissue sample is collected within 30 minutes after sampling, stored in 10% formalin for 24 to 48 hours, and the tissue is dehydrated, paraffin-embedded and sliced ​​within 2 working days. Five slices are cut continuously, one of which is used for hematoxylin and eosin staining. The slices are required to be complete, free of contamination and wrinkles, and have a thickness of 4 μm.

3. The method for evaluating the therapeutic efficacy of hepatocellular carcinoma based on immune microenvironment characteristics according to claim 2, characterized in that: In the step (2), the markers consist of 4 panels, the first template includes: CD8, LAG3, PD1, TIGIT, TIM3; the second template includes: CD8, PDL1, CD103, CLAUDIN182, PANCK; the third template includes: CD4, CD68, CD163, FOXP3, GPC3; the fourth template includes: HLADR, PANCK, CD66b, CD14, CD11c.

4. The method for evaluating the therapeutic efficacy of hepatocellular carcinoma based on immune microenvironment characteristics according to claim 3, characterized in that: The multiple immunohistochemical staining in step (2) comprises the following steps: (2.1) Dehydration: FFPE tumor sections were heated at 60°C for 12 h; (2.2) Dewaxing: xylene dewaxing, 10 min / time, repeated 3 times; (2.3) Hydration: 100% ethanol for 5 min, 95% ethanol for 5 min, 70% ethanol for 2 min; (2.4) Antigen retrieval: heating in EDTA buffer, pH 9.0 or citrate buffer, pH 6.0 for FoxP3; (2.5) Primary antibody incubation: Incubate with primary antibody at room temperature for 10 min; (2.6) Secondary antibody incubation: add HRP secondary antibody working solution and incubate at room temperature for 10 minutes; (2.7) Fluorescence staining signal amplification: enhance the fluorescence signal through tyramide signal TSA technology; (2.8) Repeat staining: After each antibody staining, repeat antigen retrieval and perform the next round of antibody staining until all markers on a single panel are stained; (2.9) DAPI staining: After the sections were dried, DAPI staining solution was added and incubated at room temperature in the dark for 10 min. (2.10) Sealing: Wash with 1×TBST 3 times for 3 min each time, seal with anti-fluorescence quencher, and store in dark place; (2.11) Reading: Read the staining results using a high-throughput panoramic scanner; (2.12) Quantitative statistical analysis: Image and statistical analysis were performed using a panoramic pathology workstation.

5. The method for evaluating the therapeutic efficacy of hepatocellular carcinoma based on immune microenvironment characteristics according to claim 4, characterized in that: The step (3) comprises the following steps: (3.1) Whole-mount fluorescence images were captured using an Olympus VS200 system equipped with an Olympus UPLXAPO 20× objective lens; (3.2) Regional division: The whole-slice fluorescence images were analyzed on the PanoATLAS workstation using QuPath image analysis software. The regions of interest (ROA) were determined based on the H&E-stained tissue sections. The algorithm embedded in QuPath was used to accurately classify tissues into tumor regions, stromal regions, and normal tissue regions through a process of labeling, training, and validation. (3.3) Cell segmentation: Use the algorithm in QuPath software to split cells into single cells and identify the nucleus, cytoplasm and cell membrane structure; split by DAPI staining and PANCK staining; (3.4) Cell phenotypic analysis: Perform phenotypic analysis on cells based on the positivity and relative intensity of the markers; (3.5) Calculation and analysis of positive rate: Calculate the positive rate of each cell in each ROA. Calculate the positive cells of different subtypes separately. The cell positive rate is the ratio of the number of positive cells in the field of view to the total number of cells in the field of view, calculated according to the following formula: Cell positive rate = (ΣC_P) / (ΣC) Among them, cells refer to cells of different subtypes, and the proportions of different subtypes are calculated separately. C is the cell count under the field of view; C_P is the phenotypic cell count that is positively stained in C; (3.6) Spatial analysis: locate positive cells, calculate the distance between positive cells, and analyze the spatial distribution characteristics of cells. The positive cells are located in a two-dimensional Cartesian coordinate system. The default data is the pixel coordinates based on the rectangular image field of view. The distance between positive cells is calculated after converting it into length coordinate data according to the magnification. The distance is calculated using the following Euclidean distance calculation method: Wherein, d(x, y) represents the Euclidean distance between cell x and cell y, x1 and x2 represent the horizontal coordinate and vertical coordinate of cell x, and y1 and y2 represent the horizontal coordinate and vertical coordinate of cell y.

6. The method for evaluating the therapeutic efficacy of hepatocellular carcinoma based on immune microenvironment characteristics according to claim 5, characterized in that: In step (3.2), the ROA delineation rules / quality control rules are: Areas with no or little DAPI nuclear staining: No DAPI staining, regardless of whether other channels are stained, are all excluded; A small amount / sporadic DAPI staining, if it is blood and muscle tissue, all will be removed; There was a DAPI-stained exclusion of non-cellular components; Folded area: Folded areas caused by tissue detachment in the stained scan image are removed; The scattered areas outside the overall sample are eliminated; Normal epithelial tissue was removed.

7. The method for evaluating the therapeutic efficacy of hepatocellular carcinoma based on immune microenvironment characteristics according to claim 6, characterized in that: The step (6) comprises the following steps: (6.1) Randomly divide the samples into training set and validation set; (6.2) Use three machine learning algorithms, Logistic regression, random forest and gradient boosting classifier, to build a prediction model based on TIME feature TIS; use five-fold cross validation to improve the robustness of the model; use CD8 + PD-L1 + and CD103 + CD8 + The positive rate of cells and CD103 within 30 μm around tumor cells + The number of cells is used to define TIS.

8. The method for evaluating the therapeutic efficacy of hepatocellular carcinoma based on immune microenvironment characteristics according to claim 7, characterized in that: The step (6) also includes step (6.3): constructing a multivariate COX proportional hazard model based on TIS, dividing the patients into a high TIS group and a low TIS group, and calculating the survival probability difference of PFS and OS between the two groups in the model constructed with TIS by the Kaplan-Meier method.

9. A hepatocellular carcinoma treatment efficacy evaluation system based on immune microenvironment characteristics, characterized by: It includes: A sample preparation module configured to collect tumor tissue samples from patients with hepatocellular carcinoma and perform pretreatment of the fixed and sectioned samples before staining; A multiplex immunohistochemical staining module configured to perform multiple immunohistochemical staining of sections using a combination of antibodies to simultaneously detect multiple immune cell types and molecular markers; An image acquisition and analysis module configured to acquire high-precision images of stained sections using digital pathology equipment and to identify and count immune cells using image analysis software; A spatial distribution feature analysis module, which is configured to use machine learning algorithms to quantitatively analyze the spatial distribution features of immune cells, including cell density and distance from tumor cells; A calculation module is configured to multiply the input image voxel signal value and the weight vector of the convolution kernel, and transfer multiple layers of convolution kernels to finally obtain the calculation result of each network output; The prediction model building module is configured based on the spatial distribution characteristics of immune cells and clinical characteristics, using total regional CD8 + PD-L1 + and CD103 + CD8 + Cell positive rate, CD103 within 30μm around tumor cells + The number of cells was used to construct a prediction model to predict the efficacy of immunotherapy in HCC patients.

10. The hepatocellular carcinoma treatment efficacy evaluation system based on immune microenvironment characteristics according to claim 9, characterized in that: The image acquisition and analysis module performs the following steps: (3.1) Whole-mount fluorescence images were captured using an Olympus VS200 system equipped with an Olympus UPLXAPO 20× objective lens; (3.2) Regional division: The whole-slice fluorescence images were analyzed on the PanoATLAS workstation using QuPath image analysis software. The regions of interest (ROA) were determined based on the H&E-stained tissue sections. The algorithm embedded in QuPath was used to accurately classify tissues into tumor regions, stromal regions, and normal tissue regions through a process of labeling, training, and validation. (3.3) Cell segmentation: Use the algorithm in QuPath software to split cells into single cells and identify the nucleus, cytoplasm and cell membrane structure; split by DAPI staining and PANCK staining; (3.4) Cell phenotypic analysis: Perform phenotypic analysis on cells based on the positivity and relative intensity of the markers; (3.5) Calculation and analysis of positive rate: Calculate the positive rate of each cell in each ROA. Calculate the positive cells of different subtypes separately. The cell positive rate is the ratio of the number of positive cells in the field of view to the total number of cells in the field of view, calculated according to the following formula: Cell positive rate = (ΣC_P) / (ΣC) Among them, cells refer to cells of different subtypes, and the proportions of different subtypes are calculated separately. C is the cell count under the field of view; C_P is the phenotypic cell count that is positively stained in C; (3.6) Spatial analysis: locate positive cells, calculate the distance between positive cells, and analyze the spatial distribution characteristics of cells. The positive cells are located in a two-dimensional Cartesian coordinate system. The default data is the pixel coordinates based on the rectangular image field of view. The distance between positive cells is calculated after converting it into length coordinate data according to the magnification. The distance is calculated using the following Euclidean distance calculation method: Wherein, d(x, y) represents the Euclidean distance between cell x and cell y, x1 and x2 represent the horizontal coordinate and vertical coordinate of cell x, and y1 and y2 represent the horizontal coordinate and vertical coordinate of cell y.

Images