A method for constructing a prognosis survival prediction model and a prediction method

Abstract

The present invention relates to the field of data processing technology, and discloses a method for constructing a prognosis survival prediction model and a prediction method. The construction method includes: obtaining a one-to-one corresponding fluorescence sample image of a sample tissue before treatment and the prognosis information of the patient corresponding to the sample tissue; segmenting myeloid cells and determining the fluorescence expression amount of each myeloid cell on different markers; determining the spatial distance between each myeloid cell in each sample tissue and other different types of myeloid cells to obtain cell spatial distance characteristics; determining the spatial interaction relationship between each myeloid cell in each sample tissue and surrounding cells to obtain cell interaction characteristics; dividing the training set and the test set; and constructing a prognosis survival prediction model based on the cell spatial distance characteristics and cell interaction characteristics in the training set and the test set. The present invention constructs a prognosis survival prediction model based on the spatial position relationship of myeloid cells, which provides a powerful reference basis for clinical judgment of tumor prognosis and survival.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a prognosis survival prediction model construction method and a prediction method. Background Art

[0002] Myeloid cells are cells derived from hematopoietic stem cells in the bone marrow, including granulocytes, red blood cells, macrophages, and monocytes. Myeloid cells are hematopoietic cells in a narrow sense, while lymphocytes are considered non-hematopoietic cells. Different types of cells have different physiological functions in the body, but all play a role in normal life activities.

[0003] Myeloid checkpoints refer to molecular signaling pathways and regulatory factors that play a key role in the bone marrow hematopoietic system. Myeloid checkpoints ensure the balance and stability of the hematopoietic process and maintain the normal function of the blood system. Disturbances in myeloid checkpoints may lead to abnormal hematopoietic function and the development of blood diseases.

[0004] Current immunotherapy, especially for predicting treatment outcomes in cancer patients, primarily focuses on the research and application of lymphoid immune checkpoints (such as PD-1, PD-L1, and CTLA4). These checkpoints primarily target T cells and other lymphocytes. However, relatively little research and testing methods are available for myeloid immune checkpoints (such as CD47, LILRB1, and CD171). Furthermore, existing PD-1 and PD-L1 assays primarily assess lymphoid immune responses and are unable to effectively predict or evaluate the efficacy of myeloid immune checkpoint drugs.

[0005] Furthermore, current clinical assessments of the efficacy of myeloid immune checkpoint drug treatments are primarily based on subjective and inaccurate assessments by clinicians based on imaging and laboratory results. Currently, there is a lack of a machine learning model based on spatial relationships between cells to inform clinical assessments of tumor prognosis and survival. Summary of the Invention

[0006] In view of this, the present invention provides a method for constructing a prognosis survival prediction model and a prediction method to provide a reference basis for clinical prognosis survival of tumors.

[0007] In a first aspect, the present invention provides a method for constructing a prognosis survival prediction model, the method comprising:

[0008] Obtaining a one-to-one corresponding fluorescent sample image of the sample tissue before treatment, wherein the fluorescent sample image is obtained by staining with antibodies based on myeloid cell markers;

[0009] Obtaining prognostic information for the patient corresponding to the sample tissue, including the patient's survival status after receiving different predetermined treatments;

[0010] Segmenting the myeloid cells in each sample tissue based on the fluorescent sample image to determine the fluorescence expression level of each myeloid cell on different markers; this includes dividing each fluorescent sample image into regions to obtain multiple regional images corresponding to each fluorescent sample image; determining the position of the nucleus of the myeloid cell based on the fluorescence expression level corresponding to the regional image; and segmenting the complete myeloid cell corresponding to the nucleus based on the position of the nucleus;

[0011] Determine the myeloid cell type annotation for each myeloid cell based on the fluorescence expression of different markers in each myeloid cell and the cell type annotation algorithm;

[0012] Based on the myeloid cell type annotations, the spatial distance between each myeloid cell and other different types of myeloid cells in each sample tissue is determined to obtain the cell spatial distance characteristics;

[0013] Based on myeloid cell type annotations, the spatial interaction relationship between each myeloid cell and surrounding cells in each sample tissue was determined to obtain cell interaction characteristics;

[0014] Divide the sample tissue into training set and test set according to the preset ratio and prognostic information;

[0015] A prognostic survival prediction model was constructed based on the cell spatial distance characteristics and cell interaction characteristics in the training set and the test set. The prognostic survival prediction model was used to predict the survival of pathological samples after intervention.

[0016] The present invention provides a method for constructing a prognosis prediction model for myeloid cells. By extracting the spatial characteristics of cells in fluorescent sample images, including cell segmentation, cell annotation, distance calculation, and interaction analysis, the characteristics between cells are extracted at the cellular level. Based on these spatial characteristics, a prognosis survival prediction model based on the spatial position relationship of myeloid cells is constructed, providing a strong reference basis for clinical judgment of tumor prognosis and survival.

[0017] In an optional embodiment, the marker antibody is selected from one or more of the following: CD45, CD45RA, CD45RO, CD11b, CD11c, CD14, CD15, CD16, CD33, CD68, CD86, CD163, CD206, SPP1, FOLR2, C1QC, CD1c, HLA-DR, CD141, S100A9, S100A8, CD117, CD123, MPO.

[0018] In an optional embodiment, each regional image includes a cancerous region and a paracancerous region, and determining the spatial distance between each myeloid cell and other myeloid cells of different types in each sample tissue to obtain a cell spatial distance feature includes:

[0019] determining a first Euclidean distance between each myeloid cell type and other different types of myeloid cells in the cancerous region included in each regional image;

[0020] determining the second Euclidean distance between each myeloid cell type and other different types of myeloid cells in the adjacent cancer region included in each regional image;

[0021] Based on all first Euclidean distances, determining a first spatial distance between each myeloid cell and other different types of myeloid cells in all cancerous areas of the sample tissue, where the first spatial distance includes: a first minimum spatial distance, a first maximum spatial distance, and a first average spatial distance;

[0022] Based on all the second Euclidean distances, determining a second spatial distance between each myeloid cell and other different types of myeloid cells in all adjacent cancer regions of the sample tissue, where the second spatial distances include: a second minimum spatial distance, a second maximum spatial distance, and a second average spatial distance;

[0023] The first spatial distance and the second spatial distance are used as cell spatial distance features;

[0024] After obtaining the cell space distance features, including:

[0025] Data processing and feature screening are performed on the distance features between the same myeloid cell types in the cell space distance features to obtain processed cell space distance features, which are used to construct a prognosis survival prediction model.

[0026] In this example, calculating the spatial distances between different types of myeloid cells provides a deeper understanding of the interactions and spatial distribution of cells within the tumor microenvironment. Furthermore, this example analyzes the differences in spatial distances between myeloid cells in the cancerous and adjacent areas, taking into account cellular changes during tumor progression. This helps to improve the comprehensiveness of the analysis and enhance the predictive accuracy of the constructed model.

[0027] In this embodiment, data processing and feature screening of distance features between the same myeloid cell types can improve the accuracy and robustness of the model, help better capture the characteristic relationships between cells, and provide stronger support for prognosis and survival prediction.

[0028] In an optional embodiment, each regional image includes a cancerous region and a paracancerous region, and determining the spatial interaction relationship between each myeloid cell and surrounding cells in each sample tissue to obtain cell interaction characteristics includes:

[0029] determining a first interaction feature between each myeloid cell and surrounding interacting cells in an interacting relationship in the cancer region included in each regional image;

[0030] Determining a second interaction feature between each myeloid cell and surrounding interacting cells in the peritumoral region included in each regional image; wherein the first interaction feature and the second interaction feature both include the proportion of each myeloid cell type in the interacting cells;

[0031] The first interaction feature and the second interaction feature are used as cell interaction features.

[0032] In this embodiment, using the spatial interaction relationship between two cells as an important reference indicator for disease prognosis can not only improve the accuracy and reliability of the prognosis survival prediction model, but also provide important information support for clinical decision-making.

[0033] In an optional embodiment, the features used to construct the prognostic survival prediction model include one or more of the following: T_MonoToM2_max, T_M1ToM2_mean, T_MonoToM2_mean, P_M2ToHLADRhicells_max, P_MonoToHLADRhicells_max, P_HLADRhicellsToMono_min P_MonoToHLADRhicells_min P_M2ToMono_mean, T_HLADRhicells_M2, T_M1_M1, T_M1_M2, T_M2_M1, T_M2_M2, T_Mono_M2, T_Neutro_M2, P_M2_Mono.

[0034] In a second aspect, the present invention provides a method for predicting prognosis and survival, comprising:

[0035] Acquiring a fluorescent sample image of a pathological sample to be predicted;

[0036] The fluorescence sample image is input into the prognosis survival prediction model constructed according to the prognosis survival prediction model construction method described in any of the above embodiments, and the survival status of the pathological sample to be predicted after intervention is output.

[0037] In a third aspect, the present invention provides a device for constructing a prognosis survival prediction model, the device comprising:

[0038] An acquisition module is used to obtain a fluorescent sample image corresponding to the sample tissue before treatment, the fluorescent sample image being obtained by staining with antibodies for myeloid cell markers; and is also used to obtain prognostic information about the patient corresponding to the sample tissue, including the patient's survival status after receiving different predetermined treatment methods;

[0039] a segmentation module for segmenting myeloid cells in each sample tissue based on the fluorescent sample image and determining the fluorescence expression level of each myeloid cell on different markers; wherein the module comprises dividing each fluorescent sample image into regions to obtain multiple regional images corresponding to each fluorescent sample image; determining the position of the nucleus of the myeloid cell based on the fluorescence expression level corresponding to the regional image; and segmenting the complete myeloid cell corresponding to the nucleus based on the position of the nucleus;

[0040] An annotation module, configured to determine the myeloid cell type annotation of each myeloid cell based on the fluorescence expression of different markers of each myeloid cell and a cell type annotation algorithm;

[0041] A distance feature extraction module is used to determine the spatial distance between each myeloid cell in each sample tissue and other myeloid cells of different types based on myeloid cell type annotations, and obtain cell spatial distance features;

[0042] The interaction feature extraction module is used to determine the spatial interaction relationship between each myeloid cell and surrounding cells in each sample tissue based on myeloid cell type annotations to obtain cell interaction features;

[0043] A partitioning module is used to divide the sample tissue into a training set and a test set according to a preset ratio and prognostic information;

[0044] A construction module is used to build a prognostic survival prediction model based on the cell spatial distance characteristics and cell interaction characteristics in the training set and the test set. The prognostic survival prediction model is used to predict the survival of pathological samples after intervention.

[0045] In a fourth aspect, the present invention provides a computer device comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the computer instructions to execute the method for constructing a prognosis survival prediction model according to the first aspect or any corresponding embodiment thereof.

[0046] In a fifth aspect, the present invention provides a computer-readable storage medium storing computer instructions, which are used to enable a computer to execute the method for constructing a prognosis survival prediction model according to the first aspect or any corresponding embodiment thereof.

[0047] In a sixth aspect, the present invention provides a computer program product comprising computer instructions for enabling a computer to execute the method for constructing a prognosis survival prediction model according to the first aspect or any corresponding embodiment thereof.

[0048] It should be noted that the prognosis survival prediction model construction device, computer device, and computer-readable storage medium provided by the present invention correspond to the above-mentioned prognosis survival prediction model construction method. Therefore, regarding the beneficial effects of the prognosis survival prediction model construction device, computer device, and computer-readable storage medium, please refer to the description of the corresponding beneficial effects of the prognosis survival prediction model construction method above, and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific 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 any creative work.

[0050] Figure 1 1 is a flow chart of a method for constructing a prognosis survival prediction model according to an embodiment of the present invention;

[0051] Figure 2 is a schematic diagram of sample information according to an embodiment of the present invention;

[0052] Figure 3 is a schematic diagram of a fluorescent sample image scanning result according to an embodiment of the present invention;

[0053] Figure 4 is a schematic diagram of a segmented fluorescence sample image according to an embodiment of the present invention;

[0054] Figure 5 is a schematic diagram of an annotated fluorescence sample image according to an embodiment of the present invention;

[0055] Figure 6 is a schematic diagram of a heat map after normalization according to an embodiment of the present invention;

[0056] Figure 7 is a schematic diagram of an ROC curve of a training set according to an embodiment of the present invention;

[0057] Figure 8 is a visualization diagram of other evaluation indicators of the training set according to an embodiment of the present invention;

[0058] Figure 9 is a schematic diagram of an ROC curve of a validation set according to an embodiment of the present invention;

[0059] Figure 10 is a visualization diagram of other evaluation indicators of the validation set according to an embodiment of the present invention;

[0060] Figure 11 is a schematic diagram of the importance of features in a model according to an embodiment of the present invention;

[0061] Figure 12 is a schematic diagram of a feature importance score broken line according to an embodiment of the present invention;

[0062] Figure 13 is a schematic diagram of the impact of 16 features on model performance according to an embodiment of the present invention;

[0063] Figure 14 is a structural block diagram of a device for constructing a prognosis survival prediction model according to an embodiment of the present invention;

[0064] Figure 15 Schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0065] To make the purpose, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying 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 those skilled in the art without making creative efforts shall fall within the scope of protection of the present invention.

[0066] Existing treatment efficacy prediction methods primarily target lymphoid immune checkpoints, such as PD-1, PD-L1, and CTLA4. However, there are also immune checkpoint drugs targeting myeloid cells on the market, such as CD47, LILRB1, and CD171. Existing PD-1 and PD-L1 assays are unable to effectively predict or evaluate the efficacy of these myeloid immune checkpoint drugs. Therefore, a model or method for prognosis prediction based on myeloid cells is urgently needed.

[0067] According to an embodiment of the present invention, an embodiment of a method for constructing a prognostic survival prediction model is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0068] In this embodiment, a method for constructing a prognosis survival prediction model is provided, which can be executed by a server, a terminal, a mobile terminal, and other devices. Figure 1FIG. 1 is a flow chart of a method for constructing a prognosis survival prediction model according to an embodiment of the present invention. Figure 1 As shown, the process includes the following steps:

[0069] Step S101: Obtain a fluorescent sample image corresponding to the pre-treatment sample tissue. The fluorescent sample image is obtained by staining with antibodies for myeloid cell markers. Prognostic information for the patient corresponding to the sample tissue is obtained. This prognostic information may include the patient's survival after receiving different predetermined treatments. These treatments may include surgical resection, chemoradiotherapy, immunotherapy, and targeted therapy. Survival information may include survival time, i.e., the prognostic information may include the patient's survival time after receiving a specific treatment.

[0070] This embodiment is specifically described with an example. In this embodiment, 90 sample tissues are provided as an example. For specific sample information, refer to Figure 2 As shown in the figure, T stage represents tumor stage, with "T" followed by a number (e.g., T1, T2, T3, T4). A larger number indicates a larger or more severe tumor. N stage represents lymph node stage, with "N" followed by a number (e.g., N0, N1, N2, N3). Increasing numbers generally indicate an increase in the number or degree of metastasis to lymph nodes. Different letters after "N" indicate different sizes or types of metastasis to lymph nodes. Furthermore, the survival time in this embodiment can refer to the survival time of myeloid cells after treatment, and can range from 1 month, 6 months, 12 months, 24 months, 60 months, 80 months, etc., without specific limitation.

[0071] The following describes in detail the process of acquiring a fluorescent sample image of one of the sample tissues.

[0072] Traditional immunohistochemical staining can only stain one marker (in this example, a marker antigen) per slide. If six markers are needed, six consecutive slides are required, and the final results may not be consistent at the cellular level. Especially for clinical samples, where tissue is precious, staining six markers on a single slide, at the same cellular level, would greatly facilitate and enhance subsequent analysis. In this example, multiplex immunohistochemical fluorescence (mIHC) staining is employed. mIHC staining allows for staining six markers on a single slide, enabling accurate characterization of the status of tumor and immune cells within the tumor microenvironment. This provides crucial information for accurate tumor diagnosis and is also crucial for tumor immunotherapy. The proportions and distribution of immune cell subsets, particularly the spatial relationship between immune cells and tumor cells, are crucial components of tumor prognosis and prediction research. This approach can help clinicians, who lack pathology expertise, easily, quickly, and effectively predict the prognosis and treatment efficacy of myeloid-targeted drugs.

[0073] It's important to note that myeloid cells refer to a series of white blood cells produced in the bone marrow, including granulocytes (Neutro), monocytes (Mono), and macrophages (Macro). These cells originate from hematopoietic stem cells and undergo a series of differentiation and maturation processes to become mature white blood cells, responsible for the body's immune defense and inflammation regulation.

[0074] In some optional embodiments, markers for identifying myeloid cells include one or more combinations of the following: CD45, CD45RA, CD45RO, CD11b, CD11c, CD14, CD15, CD16, CD33, CD68, CD86, CD163, CD206, SPP1, FOLR2, C1QC, CD1c, HLA-DR, CD141, S100A9, S100A8, CD117, CD123, MPO.

[0075] In this embodiment, the following marker combination examples for identifying myeloid cells are provided: CD68, CD11b, CD14, MPO, HLA-DR, and CD163. The specific meanings of the markers are as follows.

[0076] CD68: Commonly used to detect macrophages. Macrophages are important immune cells that play a crucial role in inflammation, infection, tumors, and tissue repair. CD68 is a macrophage marker. Testing for CD68 can help pathologists determine the location, number, and activity of macrophages in tissues, thereby providing insights into disease diagnosis and pathophysiology.

[0077] CD11b: An integrin, also known as Mac-1 or ITGAM, is a surface marker for monocytes, macrophages, and neutrophils. CD11b detection can help identify and quantify the presence and distribution of monocytes, macrophages, and neutrophils in tissues. These cells play a vital role in processes such as inflammation, infection, immune response, and tissue repair. Therefore, CD11b detection can help pathologists understand the pathophysiology of diseases and guide diagnosis and treatment planning.

[0078] CD14: A protein that plays a key role in innate immunity, it is a key surface marker of monocytes. In pathological diagnosis, CD14 detection can be used to identify immune cell infiltration patterns in specific diseases, such as autoimmune diseases, infectious diseases, and tumors. Localizing and counting CD14-positive cells can help physicians further understand the pathophysiology of the disease, thereby guiding treatment selection and assessing prognosis.

[0079] MPO: A granulocyte-specific enzyme, MPO is commonly used to detect granulocytes, including neutrophils (neutrophils) and eosinophils. MPO is a key enzyme within granulocytes, and its detection can help determine the presence and activity of granulocytes in tissues. Granulocytes play a crucial role in diseases such as inflammation, infection, and tumors. MPO detection helps assess the degree of tissue inflammation, granulocyte infiltration, and the pathophysiology of disease. Therefore, detecting MPO-positive cells can better understand the status of granulocytes in tissues, providing important information for disease diagnosis and treatment.

[0080] HLA-DR is an antigen presentation-associated protein primarily expressed on the surface of immune cells. HLA-DR is commonly used to identify professional antigen-presenting cells, primarily mature and activated monocytes, which possess antigen-presenting and immunomodulatory functions. HLA-DR-positive cells play a crucial role in tissues, particularly in conditions such as inflammation, infection, autoimmune diseases, and tumors. These cells, including macrophages, dendritic cells, and B lymphocytes, play a role in immune responses and antigen presentation. By detecting HLA-DR-positive cells, pathologists can assess immune cell activity, the extent of immune responses, and the immunopathological characteristics of diseases, thereby better understanding disease pathogenesis and progression.

[0081] CD163: A type I membrane protein, also known as the M130 antigen, primarily expressed in the mononuclear cell and macrophage systems. It is a marker for monocytes and macrophages, such as splenic dendritic cells and alveolar macrophages. It is primarily used to detect monocytes and tissue cells in tumors and reactive lesions.

[0082] In this embodiment, taking intestinal tissue as an example, a fluorescent sample image can be obtained through the following experimental steps. It includes: dewaxing of the baked slice, fixation, repair, blocking, blocking, primary antibody incubation, secondary antibody incubation, fluorescence incubation, repair, restarting blocking, primary antibody incubation, secondary antibody incubation, fluorescence incubation, repair, and stopping the repetitive steps after all 5 targets are marked. Continue fluorescence incubation to complete Opal Polaris (a multiple fluorescence staining technology, commonly used for tissue section analysis in biomedical research) labeling. Finally, nuclear staining, sealing, scanning to obtain a fluorescent sample image. Among them, the scanning results refer to Figure 3 shown.

[0083] Step S102 : segmenting the myeloid cells in each sample tissue based on the fluorescent sample image, and determining the fluorescence expression level of each myeloid cell on different markers.

[0084] Specifically, scanning involves placing the processed tissue sample under a fluorescence microscope, illuminating the sample with a laser of a specific wavelength, causing the labeled fluorescent dye to fluoresce, and then capturing the emitted fluorescence signal with a camera system to obtain a fluorescent sample image in this embodiment. This fluorescent sample image allows observation of the overall morphology of cells, understanding the fluorescence expression of different markers, and determining the cell type of each myeloid cell in the sample.

[0085] Furthermore, cell segmentation can first locate the position of each cell nucleus through the fluorescence expression of the DAPI (4', 6-diamidino-2-phenylindole) channel, and then find the cell boundary with the cell nucleus as the cell center, and finally obtain a complete cell information. The fluorescent sample image after segmentation is referenced Figure 4 The average fluorescence expression of each marker in each cell can also be used as the fluorescence expression of each marker in the cell, and this value can be used for subsequent cell clustering and cell type determination.

[0086] Step S103 : determining the myeloid cell type annotation of each myeloid cell based on the fluorescence expression levels of each myeloid cell on different markers and a cell type annotation algorithm.

[0087] After the cell type is preliminarily determined based on the fluorescence expression of different markers of myeloid cells, it can be further displayed on the image. Since the multiplex immunohistochemistry fluorescence scheme (mIHC) has a high background signal intensity, and the background signal intensity varies to varying degrees between different markers and even different samples, it is difficult to determine an overall threshold to remove the background signal intensity to achieve the purpose of accurate display. In this embodiment, a set of annotation tables suitable for this embodiment is designed, as shown in Table 1. This annotation table can be used to accurately annotate cell types using the CELESTA (v0.0.0.9000) algorithm. After verification, the results of the cell type classification annotations in this annotation table are consistent with the actual image. In this embodiment, annotations for Macro (macrophages), Mono (monocytes), Neutro (neutrophils), HLADR+ cells (HLA-DR positive cells), M1 (M1 macrophages or classically activated macrophages), M2 (M2 macrophages or alternatively activated macrophages), and Unknow cells (cells of unclear cell type) are completed. The annotated fluorescence sample image is referenced. Figure 5 shown.

[0088] Table 1

[0089]

[0090] After the annotation is completed, the heat map of the expression of different markers of each cell type is visualized and normalized. Figure 6 shown.

[0091] Step S104 : Based on the myeloid cell type annotations, the spatial distance between each myeloid cell in each sample tissue and other myeloid cells of different types is determined to obtain a cell spatial distance feature.

[0092] Specifically, in this embodiment, the Euclidean distance from each cell to the nearest cell of each cell type can be calculated one by one, including the maximum, minimum, and average distances. Continuing with the aforementioned sample tissue as an example, 216 dimensions of spatial distance features were ultimately obtained for the sample tissue, including T_HLADR+cellsToM1_mean, T_M1ToMacro_mean, T_MonoToNeutro_max, P_MonoToM1_max, T_NeutroToMono_min, and P_MacroToM2_min. The T at the beginning of the feature name represents cancer, and P represents paracancerous, thereby completing the extraction of cellular spatial distance features.

[0093] Step S105 : Based on the myeloid cell type annotations, the spatial interaction relationship between each myeloid cell and surrounding cells in each sample tissue is determined to obtain cell interaction characteristics.

[0094] In this example, each cell is defined as having a cell interaction relationship with its 10 nearest surrounding cells. The proportion of each cell type in all interacting cells is calculated, and this proportion is used as the feature of each cell in spatial interaction. Continuing with the above sample tissue as an example, 72-dimensional features at the spatial interaction level are finally obtained, including T_HLADR+cells_HLADR+cells, T_Macro_M1, etc., thereby realizing the extraction of spatial interaction features.

[0095] Step S106: Divide the sample tissue into a training set and a test set according to a preset ratio and prognostic information.

[0096] In this embodiment, we continue to take the above-mentioned sample tissue as an example, that is, taking 90 sample tissues as an example, the 90 samples can be split into a data set in a ratio of 7:3, of which 64 are used as training sets and 26 are used as test sets for testing. In the training set and the test set, long survival and short survival each account for half. In this embodiment, samples with a survival time greater than or equal to 70 months in the prognostic information can be defined as long survival, numbered 1, and the others can be defined as short survival, numbered 0. In the final sample, long survival and short survival each account for 50%. The independent test set has a total of 26 samples, including 13 long survivals and 13 short survivals. The test set is independent of the modeling process, does not participate in any feature selection and model building process, and is only used for the final test to ensure the accuracy of the test.

[0097] In some optional embodiments, the 16 features selected for building the model are as follows: T_MonoToM2_max, T_M1ToM2_mean, T_MonoToM2_mean, P_M2ToHLADRhicells_max, P_MonoToHLADRhicells_max, P_HLADRhicellsToMono_min P_MonoToHLADRhicells_min P_M2ToMono_mean, T_HLADRhicells_M2, T_M1_M1, T_M1_M2, T_M2_M1, T_M2_M2, T_Mono_M2, T_Neutro_M2, P_M2_Mono.

[0098] Where, T_MonoToM2_max: indicates the maximum distance between monocytes in the cancerous region and the nearest M2 macrophages. T_M1ToM2_mean: indicates the average distance between M1 macrophages in the cancerous region and the nearest M2 macrophages. T_MonoToM2_mean: indicates the average distance between monocytes in the cancerous region and the nearest M2 macrophages. P_M2ToHLADRhicells_max: indicates the maximum distance between M2 macrophages in the adjacent tumor region and the nearest HLADR-positive cells. P_MonoToHLADRhicells_max: indicates the maximum distance between monocytes in the adjacent tumor region and the nearest HLADR-positive cells. P_HLADRhicellsToMono_min: indicates the minimum distance between HLADR-positive cells in the adjacent tumor region and the nearest monocytes. P_MonoToHLADRhicells_min: indicates the minimum distance between monocytes in the adjacent tumor region and the nearest HLADR-positive cells. P_M2ToMono_mean: indicates the average distance between M2 macrophages in the adjacent tumor region and the nearest monocytes. T_HLADRhicells_M2: Indicates the proportion of M2 macrophages interacting with HLADR-positive cells in the cancerous region. T_M1_M1: Indicates the proportion of M1 macrophages interacting with M1 macrophages in the cancerous region. T_M1_M2: Indicates the proportion of M2 macrophages interacting with cancer cells in the M1 macrophages. T_M2_M1: Indicates the proportion of M1 macrophages interacting with M2 macrophages in the cancerous region. T_M2_M2: Indicates the proportion of M2 macrophages interacting with M2 macrophages in the cancerous region. T_Mono_M2: Indicates the proportion of M2 macrophages interacting with monocytes in the cancerous region. T_Neutro_M2: Indicates the proportion of M2 macrophages interacting with neutrophils in the cancerous region. P_M2_Mono: Indicates the proportion of monocytes interacting with M2 macrophages in the adjacent tumor region.

[0099] Constructing a prognosis survival prediction model based on the above characteristics can effectively improve the accuracy of model construction.

[0100] Step S107: constructing a prognosis survival prediction model based on the cell spatial distance features and cell interaction features in the training set and the test set. The prognosis survival prediction model is used to predict the survival of the pathological sample after intervention.

[0101] Specifically, in this embodiment, a random forest model can be used to input the corresponding sample information, cell spatial distance characteristics, cell interaction characteristics and other information in the training set into the random forest model for training to construct an initial survival prediction model. The AUC value of the model on the training set is 0.834, which has good prediction performance. Figure 7 This is a diagram of the ROC (Receiver Operating Characteristic) curve of the training set. In this embodiment, other evaluation indicators of the model training set are calculated and visualized. Figure 8 As shown in the figure, both showed good performance. Furthermore, the constructed prediction model was verified on the test set, and the AUC of the model on the test set was 0.787, which verified that it still maintained good prediction performance in the independent test set. Figure 9 This is a schematic diagram of the ROC curve of the test set. In this embodiment, other evaluation indicators of the model test set are calculated and visualized. Figure 10 In this embodiment, good performance is shown in both the training set and the test set. The prediction model has the advantages of convenient feature extraction and high accuracy.

[0102] Currently, common treatment efficacy prediction methods primarily target lymphoid immune checkpoints, such as PD-1, PD-L1, and CTLA4. However, these existing lymphoid immune checkpoints are unable to effectively predict or evaluate the efficacy of drugs targeting these myeloid immune checkpoints. Furthermore, current multiplex immunohistochemical fluorescence analysis methods often rely solely on visual observation of experimental results or simple counting of the percentage of cells positive for a specific marker or the percentage of cells positive for both positive and negative markers, lacking deeper analysis.

[0103] The present invention provides a method for constructing a prognosis prediction model based on myeloid cells. By extracting the spatial characteristics of cells in fluorescent sample images, including cell segmentation, cell annotation, distance calculation, and interaction analysis, the characteristics between cells are extracted at the cellular level. Based on these spatial characteristics, a prognosis survival prediction model based on the spatial position relationship of myeloid cells is constructed, providing a strong reference basis for clinical judgment of tumor prognosis and survival.

[0104] In some optional embodiments, the above step S102, i.e., segmenting the myeloid cells in each sample tissue based on the fluorescent sample image and determining the fluorescence expression level of each myeloid cell on different markers, includes:

[0105] Step S1021 : dividing each fluorescence sample image into regions to obtain a plurality of region images corresponding to each fluorescence sample image.

[0106] Step S1022 : determining the position of the nucleus of the myeloid cell according to the fluorescence expression level corresponding to the regional image.

[0107] Step S1023 : Based on the position of the cell nucleus, the complete myeloid cells corresponding to the cell nucleus are segmented.

[0108] Step S1024 , determining all fluorescence expression levels corresponding to each marker in intact myeloid cells, and calculating the average fluorescence expression level corresponding to each marker.

[0109] Step S1025 , taking the average fluorescence expression level as the fluorescence expression level of the corresponding marker.

[0110] For region segmentation, Qupath software can be used to select an appropriate number of regions of interest (ROIs) within each fluorescent sample image. Specifically, Qupath software can be used to open each tissue's qptiff image and, based on the tissue staining, select appropriate regions within the fluorescent sample image as ROIs for bioinformatics analysis. When conducting experimental analysis using tissue microarrays, each microarray spot can be considered a ROI. Qupath software can then be used to perform cell segmentation within the selected ROIs, extract the fluorescence expression levels of various markers for each segmented cell, and export the results as CSV data. This data can then be exported and a SpatialExperiment-type data structure object (spe) (SpatialExperiment v1.10.0, SingleCellExperiment v1.22.0, imcRtools v1.6.5) constructed to store cell information, expression levels, and prognostic information for each ROI. For each ROI, the position of each cell nucleus can be determined by fluorescence expression in the DAPI channel. Cell boundaries can then be determined using the nucleus as the cell center, ultimately identifying a complete myeloid cell population. In this embodiment, the fluorescent sample image is divided into regions, and cell segmentation is performed on each region image, which can more accurately segment myeloid cells and increase the accuracy of cell segmentation.

[0111] Furthermore, the average of all expression levels of each marker in each cell is used as the fluorescence expression level of that marker in that cell. That is, the average of the fluorescence expression levels of each marker in intact myeloid cells is used as the fluorescence expression level of that marker in that intact myeloid cell. Using the average to represent the overall expression level of these markers in a cell facilitates further cell classification and analysis.

[0112] In some optional embodiments, step S104, i.e., each regional image includes a cancerous region and a paracancerous region, and determining the spatial distance between each myeloid cell in each sample tissue and other different types of myeloid cells to obtain the cell spatial distance feature, includes:

[0113] Step S1041 : determining a first Euclidean distance between each myeloid cell type and other different types of myeloid cells in the cancerous region included in each regional image.

[0114] Step S1042 : determining the second Euclidean distance between each type of myeloid cell and other different types of myeloid cells in the adjacent cancerous region included in each regional image.

[0115] Step S1043 : Based on all first Euclidean distances, determine a first spatial distance between each myeloid cell and other different types of myeloid cells in all cancerous regions of the sample tissue. The first spatial distances include: a first minimum spatial distance, a first maximum spatial distance, and a first average spatial distance.

[0116] Step S1044 : Based on all the second Euclidean distances, determine the second spatial distances between each myeloid cell and other different types of myeloid cells in all adjacent adjacent cancerous regions of the sample tissue. The second spatial distances include: a second minimum spatial distance, a second maximum spatial distance, and a second average spatial distance.

[0117] Step S1045: Using the first spatial distance and the second spatial distance as cell spatial distance features.

[0118] Specifically, in this embodiment, the Euclidean distance between two different types of cells can be calculated. For example, if the coordinates of one cell in the regional image are (x1, y1) and the coordinates of the other cell in the regional image are (x2, y2), then the Euclidean distance can be calculated using the formula: distance^2 = (x1-x2)^2+(y1-y2)^2. Thus, the distance from each cell type to different cell types is obtained in each ROI, and by calculating the maximum, minimum, and mean values ​​of these distances in each sample tissue, a more comprehensive distance feature is obtained. Among them, since each sample tissue includes two regions, cancer and adjacent areas, when extracting features, the cell spatial distance features can be calculated and extracted for cancer and adjacent areas respectively, and finally the two parts of the features are combined. Still taking the above example, a 216-dimensional feature at the spatial distance level can be obtained.

[0119] In this example, calculating the spatial distances between different types of myeloid cells provides a deeper understanding of the interactions and spatial distribution of cells within the tumor microenvironment. Furthermore, this example analyzes the differences in spatial distances between myeloid cells in the cancerous and adjacent areas, taking into account cellular changes during tumor progression. This helps to improve the comprehensiveness of the analysis and enhance the predictive accuracy of the constructed model.

[0120] In some optional embodiments, after obtaining the cell space distance feature, the method includes:

[0121] Data processing and feature screening are performed on the distance features between the same myeloid cell types in the cell space distance features to obtain processed cell space distance features, which are used to construct a prognosis survival prediction model.

[0122] In this embodiment, after obtaining the cell space distance features and the cell interaction features, the cell space distance features and the cell interaction features need to be fused. Still taking the above example, 288-dimensional features corresponding to each sample tissue can be obtained. It is worth noting that in the calculation of these features, the final features do not include the cell type Unknow. Furthermore, in this embodiment, the distance features between the same cell types in the cell space distance features are also filtered and not included in the subsequent processing. All features are also standardized to eliminate the influence of the absolute numerical size differences between different features on the model construction. Still taking the above example, a total of 250 features are finally entered into the feature selection process.

[0123] Then, feature selection is performed. The 216+72-dimensional features extracted in the training set can be selected using the Boruta algorithm. The Boruta algorithm is a feature selection method based on random forests. Its main goal is to find the truly important features from a given feature set and distinguish them from irrelevant features. Figure 11 As shown, it is a schematic diagram of the importance of features in the model, refer to Figure 12 As shown, it is a line chart of feature importance score, where green is the confirmed important feature and yellow is the tentative feature. Finally, 16 features of two levels, important and tentative, were obtained, which were considered by the Boruta algorithm, including 10 confirmed important features, 6 tentative features and other unimportant features. The confirmed important features are: T_MonoToM2_mean, P_MonoToHLADRhicells_max, P_HLADRhicellsToMono_min, P_MonoToHLADRhicells_min, P_M2ToMono_mean, T_HLADRhicells_M2, T_M2_M1, T_Mono_M2, T_Neutro_M2, P_M2_Mono. The tentative features are: T_MonoToM2_max, T_M1ToM2_mean, P_M2ToHLADRhicells_max, T_M1_M1, T_M1_M2, T_M2_M2. For details, please refer to Figure 13 As shown, this is a schematic diagram of the impact of the 16 features on model performance.

[0124] In this embodiment, data processing and feature screening of distance features between the same myeloid cell types can improve the accuracy and robustness of the model, help better capture the characteristic relationships between cells, and provide stronger support for prognosis and survival prediction.

[0125] In some optional embodiments, step S105, i.e., each regional image includes a cancerous region and a paracancerous region, and determining the spatial interaction relationship between each myeloid cell and surrounding cells in each sample tissue to obtain cell interaction characteristics, includes:

[0126] Step S1051: determining a first interaction feature between each myeloid cell in the cancerous region included in each regional image and its surrounding interacting cells. Interacting cells are cells that have an interactive relationship with each other.

[0127] Step S1052: Determine a second interaction feature between each myeloid cell and surrounding interacting cells in the per-cancer region included in each regional image; wherein the first interaction feature and the second interaction feature both include the proportion of each myeloid cell type in the interacting cells.

[0128] Step S1053: Using the first interaction feature and the second interaction feature as cell interaction features.

[0129] In this embodiment, it is defined that each cell has a cell interaction relationship with its 10 nearest cells. The interaction calculation is performed on each cell to obtain the cell types of the 10 cells around it. The proportion of these cell types in the interacting cells is calculated to obtain the spatial interaction characteristics from the central cell to different cell types. The statistical average is performed on each sample tissue to obtain the spatial interaction characteristics between each cell type and other different cell types in each regional image. The cell spatial interaction characteristics are calculated for the cancer and adjacent areas of the same sample tissue respectively, and these features are combined. Still taking the above example as an example, 72-dimensional features at the cell interaction level can be obtained.

[0130] In this embodiment, using the spatial interaction relationship between two cells as an important reference indicator for disease prognosis can not only improve the accuracy and reliability of the prognosis survival prediction model, but also provide important information support for clinical decision-making.

[0131] In this embodiment, a method for predicting survival prognosis is provided, which can be executed by a device such as a server, a terminal, or a mobile terminal. The method includes the following steps:

[0132] Step S201, obtaining a fluorescent sample image of a pathological sample to be predicted;

[0133] Step S202 : inputting the fluorescence sample image into the prognosis survival prediction model constructed according to the prognosis survival prediction model construction method described in any of the above embodiments, and outputting the predicted survival time of the pathological sample to be predicted after intervention.

[0134] The present invention utilizes a prognosis survival prediction model constructed based on the spatial characteristics of myeloid cells, which can better predict prognosis survival and provide an important reference basis for clinical judgment of tumor prognosis survival.

[0135] The detailed introduction of the prognosis survival prediction model is shown in the above embodiment and will not be repeated here.

[0136] In this embodiment, a prognosis survival prediction model construction device is also provided, which is used to implement the above-mentioned embodiments and preferred embodiments, and the details already described will not be repeated. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, implementation using hardware, or a combination of software and hardware, is also possible and contemplated.

[0137] This embodiment provides a prognosis survival prediction model construction device, such as Figure 14 As shown, the device includes:

[0138] Acquisition module 301 acquires a fluorescent sample image corresponding to the sample tissue before treatment, the fluorescent sample image being obtained by staining with antibodies for myeloid cell markers; and is also used to acquire prognostic information of the patient corresponding to the sample tissue, including the patient's survival status after receiving different predetermined treatments;

[0139] The segmentation module 302 is configured to segment the myeloid cells in each sample tissue based on the fluorescent sample image and determine the fluorescence expression level of each myeloid cell on different markers. This includes dividing each fluorescent sample image into regions to obtain multiple regional images corresponding to each fluorescent sample image; determining the position of the nucleus of the myeloid cell based on the fluorescence expression level corresponding to the regional image; and segmenting the complete myeloid cell corresponding to the nucleus based on the position of the nucleus.

[0140] An annotation module 303 is used to determine the myeloid cell type annotation of each myeloid cell based on the fluorescence expression amount of each myeloid cell on different markers and the cell type annotation algorithm;

[0141] A distance feature extraction module 304 is configured to determine the spatial distance between each myeloid cell and other myeloid cells of different types in each sample tissue based on the myeloid cell type annotations, thereby obtaining a cell spatial distance feature;

[0142] An interaction feature extraction module 305 is used to determine the spatial interaction relationship between each myeloid cell and surrounding cells in each sample tissue based on the myeloid cell type annotation to obtain cell interaction features;

[0143] A division module 306 is used to divide the sample tissue into a training set and a test set according to a preset ratio and prognostic information;

[0144] The construction module 307 is used to construct a prognosis survival prediction model based on the cell spatial distance characteristics and cell interaction characteristics in the training set and the test set. The prognosis survival prediction model is used to predict the survival of the pathological sample after intervention.

[0145] In an optional embodiment, the segmentation module 302 includes:

[0146] The segmentation unit is used to divide each fluorescent sample image into regions to obtain multiple regional images corresponding to each fluorescent sample image; determine the position of the nucleus of the myeloid cell based on the fluorescence expression level corresponding to the regional image; segment the complete myeloid cell corresponding to the nucleus based on the position of the nucleus; determine all fluorescence expression levels corresponding to each marker in the complete myeloid cell, and calculate the average fluorescence expression level of each marker; and use the average fluorescence expression level as the fluorescence expression level of the corresponding marker.

[0147] In an optional embodiment, each regional image includes a cancerous region and a para-cancer region, and the distance feature extraction module 304 includes:

[0148] The distance feature extraction unit is used to determine the first Euclidean distance between each myeloid cell type in the cancerous area included in each regional image and other different types of myeloid cells; determine the second Euclidean distance between each myeloid cell type in the adjacent cancerous area included in each regional image and other different types of myeloid cells; based on all the first Euclidean distances, determine the first spatial distance between each myeloid cell in all cancerous areas of the sample tissue and other different types of myeloid cells, the first spatial distance including: a first minimum spatial distance, a first maximum spatial distance, and a first average spatial distance; based on all the second Euclidean distances, determine the second spatial distance between each myeloid cell in all adjacent cancerous areas of the sample tissue and other different types of myeloid cells, the second spatial distance including: a second minimum spatial distance, a second maximum spatial distance, and a second average spatial distance; and use the first spatial distance and the second spatial distance as cell spatial distance features.

[0149] In an optional embodiment, each regional image includes a cancerous region and a para-cancerous region, and the interaction feature extraction module 305 includes:

[0150] The interaction feature extraction unit is used to determine a first interaction feature of the interaction relationship between each myeloid cell in the cancer area included in each regional image and the surrounding cells with which the interaction relationship exists; and to determine a second interaction feature of the interaction relationship between each myeloid cell in the paracancerous area included in each regional image and the surrounding cells with which the interaction relationship exists; wherein the first interaction feature and the second interaction feature both include the proportion of each myeloid cell type in the interacting cells; and the first interaction feature and the second interaction feature are used as cell interaction features.

[0151] In an optional embodiment, the device further comprises:

[0152] The processing module is used to perform data processing and feature screening on the distance features between the same myeloid cell types in the cell space distance features to obtain the processed cell space distance features. The processed cell space distance features are used to construct a prognosis survival prediction model.

[0153] The prognosis survival prediction model construction device in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.

[0154] The further functional description of each of the above modules and units is the same as that of the above corresponding embodiments and will not be repeated here.

[0155] The embodiment of the present invention also provides a computer device having the above Figure 14 The prognostic survival prediction model construction device shown.

[0156] See also Figure 15 , Figure 15 is a structural diagram of a computer device provided by an optional embodiment of the present invention, such as Figure 15 As shown, the computer device includes: one or more processors 10, memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components utilize different buses to communicate with each other and can be installed on a common mainboard or installed in other ways as needed. The processor can process the instructions executed in the computer device, including instructions stored in the memory or on the memory to display the graphical information of the GUI on an external input / output device (such as, a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Equally, multiple computer devices can be connected, and each device provides part of the necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 15 A processor 10 is taken as an example.

[0157] The processor 10 may be a central processing unit, a network processor, or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit, a programmable logic device, or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic, or any combination thereof.

[0158] The memory 20 stores instructions that can be executed by at least one processor 10, so that the at least one processor 10 executes the method shown in the above embodiment.

[0159] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system and application programs required for at least one function; the data storage area may store data created based on the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely located relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0160] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid-state drive; the memory 20 may also include a combination of the above types of memory.

[0161] The computer device further includes a communication interface 30 for the computer device to communicate with other devices or a communication network.

[0162] The embodiment of the present invention also provides a computer-readable storage medium. The above-mentioned method according to the embodiment of the present invention can be implemented in hardware, firmware, or implemented as a computer code that can be recorded in a storage medium, or implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and downloaded through a network and will be stored in a local storage medium, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state drive, etc.; further, the storage medium can also include a combination of the above-mentioned types of memory. It can be understood that a computer, a processor, a microprocessor controller or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor or hardware, the method shown in the above embodiment is implemented.

[0163] A portion of the present invention may be applied as a computer program product, such as a computer program instruction, which, when executed by a computer, can call or provide the method and / or technical solution according to the present invention through the operation of the computer. Those skilled in the art should understand that the form in which the computer program instruction exists in a computer-readable medium includes, but is not limited to, a source file, an executable file, an installation package file, etc. Accordingly, the way in which the computer program instruction is executed by the computer includes, but is not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. Here, the computer-readable medium may be any available computer-readable storage medium or communication medium that can be accessed by the computer.

[0164] Although the embodiments of the present invention have been described with reference to the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention. Such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A method for constructing a prognosis survival prediction model, characterized in that: The method comprises: Obtaining a one-to-one corresponding fluorescent sample image of the sample tissue before treatment, wherein the fluorescent sample image is obtained by staining with an antibody for a marker of myeloid cells; Obtaining prognostic information for patients corresponding to the sample tissue, including the patient's survival after receiving different pre-set treatments; defining samples with a survival time of 70 months or longer as long-term survivors; Myeloid cells in each sample tissue are segmented based on the fluorescent sample image, and the fluorescence expression amount of each myeloid cell on different markers is determined; wherein, each fluorescent sample image is divided into regions to obtain multiple regional images corresponding to each fluorescent sample image; the position of the cell nucleus of the myeloid cell is determined based on the fluorescence expression amount corresponding to the regional image; based on the cell nucleus position, the complete myeloid cell corresponding to the cell nucleus is segmented; the markers are composed of CD14, CD68, CD163, CD11b, MPO and HLA-DR; based on the fluorescence expression amount of each myeloid cell on different markers and a cell type annotation algorithm, the myeloid cell type annotation of each myeloid cell is determined; Based on the myeloid cell type annotations, determining the spatial distance between each myeloid cell in each sample tissue and other myeloid cells of different types to obtain a cell spatial distance feature; the cell spatial distance feature is composed of the minimum spatial distance, the maximum spatial distance, and the average spatial distance between each myeloid cell and other myeloid cells of different types; The method comprises determining a first Euclidean distance between each myeloid cell type and other different types of myeloid cells in the cancerous region included in each of the regional images; and determining a first spatial distance between each myeloid cell and other different types of myeloid cells in all cancerous regions of the sample tissue based on all the first Euclidean distances, wherein the first spatial distance is composed of a first minimum spatial distance, a first maximum spatial distance, and a first average spatial distance. Based on the myeloid cell type annotations, the spatial interaction relationship between each myeloid cell in each sample tissue and the surrounding cells is determined to obtain cell interaction characteristics; each cell is defined to have a cell interaction relationship with its 10 nearest surrounding cells, and the proportion of each cell type in all interacting cells is calculated, and this proportion is used as the spatial interaction characteristic of each cell; Dividing the sample tissue into a training set and a test set according to a preset ratio and the prognostic information; A prognosis survival prediction model is constructed based on the cell spatial distance features and the cell interaction features in the training set and the test set, and the prognosis survival prediction model is used to predict the survival of pathological samples after intervention.

2. The method according to claim 1, characterized in that Each of the regional images includes a cancerous area and a paracancerous area. Determining the spatial distance between each myeloid cell and other different types of myeloid cells in each sample tissue to obtain a cell spatial distance feature includes: determining a first Euclidean distance between each myeloid cell type and other different types of myeloid cells in the cancerous region included in each of the region images; determining a second Euclidean distance between each myeloid cell type and other different types of myeloid cells in the adjacent cancer region included in each of the regional images; determining, based on all the first Euclidean distances, a first spatial distance between each myeloid cell and other different types of myeloid cells in all cancerous regions of the sample tissue, wherein the first spatial distance consists of a first minimum spatial distance, a first maximum spatial distance, and a first average spatial distance; determining, based on all the second Euclidean distances, a second spatial distance between each myeloid cell and other different types of myeloid cells in all adjacent paracancerous regions of the sample tissue, where the second spatial distance consists of a second minimum spatial distance, a second maximum spatial distance, and a second average spatial distance; using the first spatial distance and the second spatial distance as the cell spatial distance features; After obtaining the cell space distance feature, the method includes: Data processing and feature screening are performed on the distance features between the same myeloid cell types in the cell space distance features to obtain processed cell space distance features, and the processed cell space distance features are used to construct the prognosis survival prediction model.

3. The method according to claim 1, characterized in that Each of the regional images includes a cancerous area and a paracancerous area. Determining the spatial interaction relationship between each myeloid cell and surrounding cells in each sample tissue to obtain cell interaction characteristics includes: determining a first interaction feature between each myeloid cell in the cancer region included in each of the regional images and surrounding interacting cells having an interacting relationship; determining a second interaction feature between each myeloid cell and surrounding interacting cells in the per-cancer region included in each of the regional images; wherein the first interaction feature and the second interaction feature both include a proportion of each myeloid cell type in the interacting cells; The first interaction feature and the second interaction feature are used as the cell interaction feature.

4. The method according to claim 1, wherein The features used to construct the prognostic survival prediction model include one or more of the following: T_MonoToM2_max, T_M1ToM2_mean, T_MonoToM2_mean, P_M2ToHLADRhicells_max, P_MonoToHLADRhicells_max, P_HLADRhicellsToMono_min, P_ MonoToHLADRhicells_min, P_M2ToMono_mean, T_HLADRhicells_M2, T_M1_M1, T_M1_M2, T_M2_M1, T_M2_M2, T_Mono_M2, T_Neutro_M2, P_M2_Mono; Among them, T_MonoToM2_max: represents the maximum distance between monocytes in the cancer area and the nearest M2 macrophages; T_M1ToM2_mean: represents the average distance between M1 macrophages in the cancer area and the nearest M2 macrophages; T_MonoToM2_mean: represents the average distance between monocytes in the cancer area and the nearest M2 macrophages; P_M2ToHLADRhicells_max: represents the maximum distance between M2 macrophages in the adjacent cancer area and the nearest HLADR-positive cells; P_MonoToHLADRhicells_max: represents the maximum distance between monocytes in the adjacent cancer area and the nearest HLADR-positive cells; P_HLADRhicellsToMono_min: represents the minimum distance between HLADR-positive cells in the adjacent cancer area and the nearest monocytes; P_MonoToHLADRhicells_min: represents the minimum distance between monocytes in the adjacent cancer area and the nearest HLADR-positive cells; P_M 2ToMono_mean: indicates the average distance between M2 macrophages in the adjacent cancerous area and the nearest monocytes; T_HLADRhicells_M2: indicates the proportion of M2 macrophages that interact with HLADR-positive cells in the cancerous area; T_M1_M1: indicates the proportion of M1 macrophages that interact with M1 macrophages in the cancerous area; T_M1_M2: indicates the proportion of M2 macrophages that interact with cancer cells and M1 macrophages; T_M2_M1: indicates the proportion of M2 macrophages that interact with cancer cells and M1 macrophages. The percentage of M1 macrophages interacting with M2 macrophages in the cancerous area; T_M2_M2: the percentage of M2 macrophages interacting with M2 macrophages in the cancerous area; T_Mono_M2: the percentage of M2 macrophages interacting with monocytes in the cancerous area; T_Neutro_M2: the percentage of M2 macrophages interacting with neutrophils in the cancerous area; P_M2_Mono: the percentage of monocytes interacting with M2 macrophages in the adjacent cancerous area.

5. A prognosis survival prediction method, characterized in that: The method comprises: Acquiring a fluorescent sample image of a pathological sample to be predicted; The fluorescent sample image is input into a prognosis survival prediction model constructed according to the prognosis survival prediction model construction method according to any one of claims 1 to 4, and the survival status of the pathological sample to be predicted after intervention is output.

6. A device for constructing a prognosis survival prediction model, characterized in that: The device comprises: An acquisition module is configured to obtain a one-to-one fluorescent sample image corresponding to the sample tissue before treatment, the fluorescent sample image being obtained by staining with antibodies for myeloid cell markers; and further configured to obtain prognostic information for the patient corresponding to the sample tissue, the prognostic information including the patient's survival status after receiving different predetermined treatments; samples with a survival time of 70 months or longer are defined as long-term survivors; a segmentation module for segmenting myeloid cells in each sample tissue based on the fluorescent sample image and determining the fluorescence expression level of each myeloid cell on different markers; wherein the module comprises dividing each fluorescent sample image into regions to obtain multiple regional images corresponding to each fluorescent sample image; determining the position of the nucleus of the myeloid cell based on the fluorescence expression level corresponding to the regional image; and segmenting the complete myeloid cell corresponding to the nucleus based on the position of the nucleus; the markers are composed of CD14, CD68, CD163, CD11b, MPO and HLA-DR; An annotation module, configured to determine the myeloid cell type annotation of each myeloid cell based on the fluorescence expression of different markers of each myeloid cell and a cell type annotation algorithm; a distance feature extraction module, configured to determine the spatial distance between each myeloid cell in each sample tissue and other myeloid cells of different types based on the myeloid cell type annotations, and obtain a cell spatial distance feature; An interaction feature extraction module is used to determine the spatial interaction relationship between each myeloid cell in each sample tissue and the surrounding cells based on the myeloid cell type annotations to obtain cell interaction features; define the cell interaction relationship between each cell and its 10 nearest surrounding cells, calculate the proportion of each cell type in all interacting cells, and use this proportion as the spatial interaction feature of each cell; The cell space distance feature and the spatial interaction feature include: T_MonoToM2_max, T_M1ToM2_mean, T_MonoToM2_mean, P_M2ToHLADRhicells_max, P_MonoToHLADRhicells_max, P_HLADRhicellsToMono_min, P_MonoToHLADRhicells_min, P_M2ToMono_mean, T_HLADRhicells_M2, T_M1_M1, T_M1_M2, T_M2_M1, T_M2_M2, T_Mono_M2, T_Neutro_M2, P_M2_Mono; Among them, T_MonoToM2_max: represents the maximum distance between monocytes in the cancer area and the nearest M2 macrophages; T_M1ToM2_mean: represents the average distance between M1 macrophages in the cancer area and the nearest M2 macrophages; T_MonoToM2_mean: represents the average distance between monocytes in the cancer area and the nearest M2 macrophages; P_M2ToHLADRhicells_max: represents the maximum distance between M2 macrophages in the adjacent cancer area and the nearest HLADR-positive cells; P_MonoToHLADRhicells_max: represents the maximum distance between monocytes in the adjacent cancer area and the nearest HLADR-positive cells; P_HLADRhicellsToMono_min: represents the minimum distance between HLADR-positive cells in the adjacent cancer area and the nearest monocytes; P_MonoToHLADRhicells_min: represents the minimum distance between monocytes in the adjacent cancer area and the nearest HLADR-positive cells; P_M 2ToMono_mean: indicates the average distance between M2 macrophages in the adjacent cancerous area and the nearest monocytes; T_HLADRhicells_M2: indicates the proportion of M2 macrophages that interact with HLADR-positive cells in the cancerous area; T_M1_M1: indicates the proportion of M1 macrophages that interact with M1 macrophages in the cancerous area; T_M1_M2: indicates the proportion of M2 macrophages that interact with cancer cells and M1 macrophages; T_M2_M1: indicates the proportion of M2 macrophages that interact with cancer cells and M1 macrophages. =The percentage of M1 macrophages interacting with M2 macrophages in the cancerous area; T_M2_M2: The percentage of M2 macrophages interacting with M2 macrophages in the cancerous area; T_Mono_M2: The percentage of M2 macrophages interacting with monocytes in the cancerous area; T_Neutro_M2: The percentage of M2 macrophages interacting with neutrophils in the cancerous area; P_M2_Mono: The percentage of monocytes interacting with M2 macrophages in the adjacent cancerous area; A division module, configured to divide the sample tissue into a training set and a test set according to a preset ratio and the prognostic information; A construction module is used to construct a prognosis survival prediction model based on the cell spatial distance characteristics and the cell interaction characteristics in the training set and the test set, and the prognosis survival prediction model is used to predict the survival of pathological samples after intervention.

7. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the method for constructing a prognosis survival prediction model according to any one of claims 1 to 4 by executing the computer instructions.

8. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable the computer to execute the method for constructing a prognosis survival prediction model according to any one of claims 1 to 4.

9. A computer program product, characterized in that The method comprises computer instructions for causing a computer to execute the method for constructing a prognosis survival prediction model according to any one of claims 1 to 4.

Images