WO2025234833A1 - Method and device for predicting cancer treatment response to immune checkpoint inhibitor - Google Patents

Abstract

Provided is a method or a device for predicting a cancer treatment response to an immune checkpoint inhibitor, wherein the method, according to an aspect of the present disclosure, enables accurate prediction of the treatment response of a cancer patient to an immune checkpoint inhibitor and facilitates the selection of an appropriate treatment strategy.

Description

METHOD AND DEVICE FOR PREDICTING CANCER TREATMENT RESPONSE TO IMMUNE CHECKPOINT INHIBITOR

The present disclosure relates to a method and device for predicting a cancer treatment response to an immune checkpoint inhibitor.

Tumor cells employ various immune evasion strategies within the tumor microenvironment (TME) to escape immune surveillance and attack. This phenomenon is generally referred to as 'cancer immune escape' and plays a critical role in tumor survival and progression.

In recent years, cancer immunotherapy has been actively developed as a treatment strategy to overcome such immune evasion mechanisms. Among various cancer immunotherapy approaches, immune checkpoint inhibitors (ICIs) have demonstrated meaningful clinical efficacy across various solid tumors and have emerged as a key anticancer treatment modality. Immune checkpoint inhibitors function by blocking inhibitory signals to T cells, thereby restoring tumor-specific immune responses.

However, the therapeutic efficacy of immune checkpoint inhibitors varies significantly between patients. For example, in locally advanced rectal cancer (LARC), treatment responsiveness varies significantly depending on the patient's molecular subtype. In LARC patients with microsatellite instability-high (MSI-H) or mismatch repair-deficient (dMMR) profiles, the therapeutic efficacy of immune checkpoint inhibitors has been demonstrated in numerous studies. In contrast, in LARC patients with microsatellite stable (MSS) or mismatch repair-proficient (pMMR) profiles, the reported effects of immune checkpoint inhibitors have been inconsistent, and treatment responses have generally been unclear.

Accordingly, to maximize the therapeutic benefit of immune checkpoint inhibitors while minimizing unnecessary treatment, there is a growing need to develop biomarkers and predictive techniques capable of identifying not only MSI-H or dMMR patients, but also MSS or pMMR patients who are more likely to respond to immune checkpoint inhibitor therapy.

Therefore, the inventors have developed a novel biomarker capable of effectively predicting treatment responsiveness to immune checkpoint inhibitors by quantitatively analyzing spatial location information among immune cells within the tumor microenvironment of cancer patients.

One aspect provides a method and a device for predicting a treatment response to an immune checkpoint inhibitor. Another aspect provides a computer-readable recording medium storing a program for executing the method on a computer.

The technical issues to be addressed by the present disclosure are not limited to those mentioned above. Other issues and advantages not explicitly stated will be understood from the following description and will become more apparent through the embodiments of the present disclosure. It will also be understood that the problems and advantages to be solved by the present disclosure may be achieved by the means and a combination thereof described in the claims.

A first aspect provides a method of predicting a cancer treatment response to an immune checkpoint inhibitor (ICI), the method being performed by a device including at least one processor, the method including: detecting cells in a pathological image of a patient; deriving information about a spatial location between tumor-associated macrophages (TAMs) and immune cells based on the detected cells; and predicting the cancer treatment response of the patient to the ICI based on the derived spatial location information.

A second aspect provides a computing system including: at least one memory; and at least one processor connected to the memory and configured to execute at least one computer-readable program stored in the memory, wherein the processor is configured to: detect cells in a pathological image of a patient; derive information about a spatial location between tumor-associated macrophages (TAMs) and immune cells based on the detected cells; and predict the cancer treatment response of the patient to an immune checkpoint inhibitor (ICI) based on the derived spatial location information.

A third aspect provides a computer-readable recording medium storing a program for executing the method according to the first aspect of the present disclosure.

A fourth aspect provides a method of treating cancer, the method being performed using a device including at least one processor, the method including: detecting cells in a pathological image of a patient; deriving information about a spatial location between tumor-associated macrophages (TAMs) and immune cells based on the detected cells; calculating a spatial proximity score between the TAMs and the immune cells by comparing the derived spatial location information with a proportion of immune cells distributed in a tumor microenvironment (TME); performing an operation of predicting a cancer treatment response to an immune checkpoint inhibitor based on the spatial proximity score between the TAMs and the immune cells; and administering the immune checkpoint inhibitor to the patient when the patient is determined to be responsive to the immune checkpoint inhibitor based on the predicted treatment response.

Other aspects, features, and advantages not described above will become apparent from the following drawings, claims, and detailed description of the invention.

According to an aspect, the method and device for predicting a cancer treatment response to an immune checkpoint inhibitor enable more accurate prediction of a patient's responsiveness to an immune checkpoint inhibitor and determination of an appropriate treatment strategy.

FIG. 1 is a diagram illustrating an example of a computing system that generates analysis results for a pathological image according to an embodiment.

FIG. 2 is a block diagram illustrating an example configuration of a computing system for predicting treatment response to an immune checkpoint inhibitor according to an embodiment.

FIG. 3 is a diagram illustrating an example of a processor analyzing a pathological image according to an embodiment.

FIG. 4 is a flowchart illustrating an example method of predicting treatment response to an immune checkpoint inhibitor according to an embodiment.

FIG. 5 shows multiplex immunofluorescence (mIF) images analyzed using antibodies for six markers specifically expressed in T cells, tumor-associated macrophages (TAMs), myeloid-derived suppressor cells (MDSCs), and dendritic cells (DCs), along with DAPI.

FIG. 6 is a schematic diagram illustrating a method of analyzing information about a spatial location between tumor-associated macrophages (TAMs) and PD-L1 positive cells using an AI-based pathology slide image analyzer according to an embodiment.

FIG. 7 is a schematic diagram illustrating a method of analyzing the proportion of PD-L1 positive cells within a defined radius (50μm) centered on a tumor-associated macrophage (TAM) according to an embodiment

FIG. 8 is a graph showing TAM-PD-L1 proximity scores for responders and non-responders to an immune checkpoint inhibitor at a pre-chemoradiotherapy (CRT) time point in patients with MSS LARC, and pathological complete response (pCR) rates between a high-score group and a low-score group classified based on a cut-off value.

FIG. 9A is a graph showing the predictive performance of TAM-PD-L1 proximity score for pathological complete response (pCR), represented by area under receiver operating characteristic curve (AUROC), and FIG. 9B is a graph comparing pCR rates between high-score group and low-score group classified based on cut-off value of TAM-PD-L1 proximity score.

FIG. 10A shows, for each patient, distribution ratios of PD-L1 positive (PD-L1⁺) cells by cell type as well as total proportions of each PD-L1⁺ cell type, as measured using CD68⁺ marker. FIG. 10B shows, for each patient, distribution ratios of PD-L1⁺ cells by cell type as well as total proportions of each PD-L1⁺ cell type, as measured using CD14⁺ marker. FIG. 10C shows, for each patient, distribution ratios of PD-L1⁺ cells by cell type as well as total proportions of each PD-L1⁺ cell type, as measured using CD80⁺ marker. FIG. 10D shows, for each patient, distribution ratios of PD-L1⁺ cells by cell type as well as total proportions of each PD-L1⁺ cell type, as measured using CD163⁺ marker. FIG. 10E shows, for each patient, distribution ratios of PD-L1⁺ cells by cell type as well as total proportions of each PD-L1⁺ cell type, as measured using CD206⁺ marker.

FIG. 11A is a graph showing TAM-PD-L1 proximity scores derived using CD68⁺, CD14⁺, CD80⁺, CD163⁺, and CD206⁺ markers from multiplex immunofluorescence (mIF) images collected from patients prior to chemoradiotherapy (CRT). FIG. 11B is a graph showing predictive performance for pathological complete response (pCR), represented as AUROC values, of TAM-PD-L1 proximity scores derived using each of markers in FIG. 11A.

FIG. 12A is a graph showing TAM-PD-L1 proximity scores derived by applying proportions of PD-L1 positive (PD-L1⁺) cells located within radii of 30μm, 50μm, and 100μm centered on tumor-associated macrophages (TAMs). FIG. 12B is a graph showing predictive performance for pathological complete response (pCR), evaluated as AUROC values, of TAM-PD-L1 proximity scores derived by applying proportions of PD-L1⁺ cells located within each of radii described in FIG. 12A.

FIG. 13 is a graph illustrating tumor regression grade (TRG) values defined by American Joint Committee on Cancer (AJCC) for high and low TAM-PD-L1 proximity score groups, which were classified based on cut-off value, where proximity scores were derived using AI-based pathology slide image analyzer from multiplex immunofluorescence (mIF) images collected from patients prior to chemoradiotherapy (CRT).

The terminology used herein is selected to align with terms commonly used in the art at present, although such terminology may vary depending on the understanding of those skilled in the art, judicial interpretations, or the emergence of new technologies. However, in some instances, terms may have been selected at the discretion of the applicant for purposes of explanation, in which case their meaning will be clarified in the relevant portions of the specification. Accordingly, the terms used herein should be interpreted not merely by their names alone, but in light of their meaning and the overall context of the specification.

Throughout the present specification, when a component is described as "including" or "comprising" another element, it should be understood that, unless explicitly stated otherwise, the component may include other elements and does not exclude the presence of other elements. Furthermore, terms such as "unit," "module," or "part" as used herein refer to units that process at least one function or operation. Such units may be implemented in hardware, software, or a combination thereof.

In an embodiment according to the present specification, a "module" or "part" may be implemented using a processor and memory. The term "processor" should be interpreted broadly to include a general-purpose processor, central processing unit (CPU), graphics processing unit (GPU), microprocessor, digital signal processor (DSP), controller, microcontroller, state machine, or the like. In certain environments, the processor may also refer to application-specific integrated circuits (ASICs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), or other specialized hardware components.

The present disclosure may be embodied in various forms and should not be construed as limited to the embodiments set forth herein. Various embodiments are illustrated in the drawings and described in detail below. The effects and features of the present disclosure, as well as the means for achieving them, will be clearly understood with reference to the embodiments described in conjunction with the accompanying drawings. However, the present disclosure is not limited to the embodiments set forth herein and may be implemented in various other forms.

In the following description, embodiments of the present disclosure will be explained in detail with reference to the accompanying drawings. Where appropriate, the same or corresponding reference numerals are used for identical or equivalent components across the drawings, and redundant descriptions thereof may be omitted.

FIG. 1 illustrates an example of a computing system 10 configured to generate an analysis result for a pathological image 20 according to an embodiment.

Referring to FIG. 1, the computing system 10 may receive the pathological image 20 and generate an analysis result 30 for the pathological image 20. Here, the analysis result 30 and/or medical information derived from the analysis result 30 may be used to predict a treatment response to an immune checkpoint inhibitor in a cancer patient.

In FIG. 1, the computing system 10 is illustrated as a single computing device; however, without being limited thereto, the system 10 may be configured to distribute information and/or data processing across multiple computing devices. Although a storage system communicable with the computing system 10 is not shown in FIG. 1, the computing system 10 may be connected to or configured to communicate with one or more storage systems.

The computing system 10 may be any computing device used to generate the analysis result 30 for the pathological image 20. Here, the computing device may refer to any type of device equipped with computing functionality, including, but not limited to, a notebook, desktop, laptop, server, or cloud system.

A storage system configured to communicate with the computing system 10 may be a device or cloud-based system for storing and managing various types of data associated with pathological image analysis. For efficient data management, the storage system may use a database to store and manage the data. Here, the data may include any information associated with pathological image analysis, such as the pathological image 20 itself and histological components including types, locations, and conditions of cells, tissues, and/or structures contained in the image. The data may further include clinical information such as the patient's age, menopausal status, clinical T stage (Clinical_T), number of tumors, tumor size, lymph node enlargement (e.g., Node_Enlargement), results of biopsy tests conducted to assess the status of estrogen receptors in tumor tissue (e.g., Biopsy_ER), results of biopsy tests for progesterone receptor status (e.g., Biopsy_PR), evaluation results indicating whether the tumor was completely removed after a given cancer treatment (e.g., pCR_final), pathology type, and homologous recombination deficiency (HRD).

The computing system 10 may receive a pathological image 20 acquired from the tissue of a patient who is the subject for predicting treatment response to an immune checkpoint inhibitor. The pathological image 20 may be received via a communication-enabled storage medium, such as a hospital system or local/cloud-based storage system. The computing system 10 may analyze the received pathological image 20 to generate an analysis result 30. The pathological image 20 may include histological components corresponding to at least one patch contained in the image.

As used in the present specification, the term 'patch' may refer to a small region within a pathological image. For example, a patch may correspond to a region extracted by performing segmentation on a pathological image to isolate semantic objects. In another example, a patch may refer to a set of pixels associated with histological components generated through analysis of the pathological image.

As used in the present specification, the term 'histological components' may include characteristics or information regarding cells, tissues, and/or structures of the human body contained in the pathological image. Such characteristics of cells may include cytologic features such as nuclei and cell membranes. The histological components may refer to information for at least one patch within the pathological image, inferred by a machine learning model. In some cases, histological components may be obtained as a result of manual annotation performed by an annotator. The term 'annotation' refers to either the process of tagging histological information to data samples or the tagged information itself (i.e., the annotation or comment data). In the relevant technical field, the term 'annotation' may be used interchangeably with terms such as 'tagging' or 'labeling'.

According to an embodiment, the computing system 10 may extract histological components―i.e., characteristics of cells, tissues, and/or structures in the human body―by analyzing the pathological image 20. Specifically, the computing system 10 may analyze the pathological image 20 using a machine learning model (e.g., through inference) to extract histological components for at least one patch included in the pathological image 20. For example, the histological components may include information on the cells within a patch―such as tumor cells, lymphocytes, T cells, myeloid-derived suppressor cells (MDSCs), macrophages, tumor-associated macrophages (TAMs), dendritic cells, fibroblasts, and endothelial cells―such as the number of specific cells or the tissue context in which particular cells are located. However, the components are not limited thereto.

Here, tumor cells may refer to cells that disregard the normal cell growth cycle and continue to proliferate excessively. In particular, tumor cells that invade surrounding tissue and spread to distant sites to grow may be referred to as cancer cells.

According to an embodiment, the computing system 10 may detect the expression of biomarkers in the cells included in the pathological image 20, or may extract information related to biomarker expression from the pathological image 20. Specifically, the computing system 10 may analyze the pathological image 20 to detect the spatial locations of various immune cells within the tumor microenvironment (TME), including tumor-associated macrophages (TAMs) and T lymphocytes expressing specific markers such as PD-L1, PD-1, CTLA-4, CD8, and FOXP3. Here, the term 'marker' refers to a protein specifically expressed in each immune cell type, and may also be referred to as a biomarker.

For example, the pathological image may include multiplex immunofluorescence (mIF) image data. Since mIF images contain cell-level protein expression information and spatial location data (e.g., x and y coordinates), the computing system 10 may analyze the pathological image 20 to measure distances between various immune cells, including TAMs and T lymphocytes expressing PD-L1.

As another example of information extractable from a patch, the computing system 10 may use information about a spatial location between TAMs and T lymphocytes expressing PD-L1―along with other immune cells―to calculate the proportion of immune cells located within a defined radius centered on each TAM. In addition, the computing system 10 may calculate a spatial proximity score between tumor-associated macrophages (TAMs) and immune cells by comparing the derived spatial location information with the proportion of immune cells distributed in the tumor microenvironment (TME). Based on the calculated spatial proximity score between TAMs and immune cells, the system may predict the patient's treatment response to an immune checkpoint inhibitor (ICI).

Examples of histological components extractable from a patch by the computing system 10 may include morphological or functional characteristics of the tissue within the patch. For example, the computing system 10 may extract histological information such as tissue density, cell distribution, nuclear size and shape, and staining intensity.

Another example of information extractable from a patch may include molecular pathological features such as microsatellite instability (MSI) status. The information extractable from a patch is not limited to the examples above and may include any histological features that are quantifiable from the patch, such as cellular instability, cell cycle characteristics, or biological functions.

As described above, the analysis result 30 generated by the computing system 10, and/or medical information generated and output based on the analysis result 30, may be used to predict and/or output the treatment response to an immune checkpoint inhibitor. In addition, clinical information associated with the pathological image and received from an accessible external system may be used as additional input data, allowing the computing system to infer and/or output the predicted treatment response to an immune checkpoint inhibitor for a cancer patient.

The following describes an example of how the computing system 10 analyzes a pathological image with reference to FIGS. 2 to 4.

FIG. 2 is a block diagram illustrating an example configuration of a computing system for predicting a treatment response to an immune checkpoint inhibitor, according to an embodiment.

The computing system 10 described above may correspond to computing system 200. Referring to FIG. 2, the computing system 200 includes a processor 210 and a memory 220. For convenience of explanation, FIG. 2 illustrates only the components relevant to the present invention. Accordingly, in addition to the components shown in FIG. 2, other general-purpose components may be further included in the computing system 200. For example, the computing system 200 may include, but is not limited to, at least one of a server device and a cloud-based device. In another example, the computing system 200 may be composed of one or more server devices. In another example, the computing system 200 may be composed of one or more cloud-based devices. In another example, the computing system 200 may be implemented as a combination of server and cloud-based devices operating together. It will also be apparent to those skilled in the art that the processor 210 and memory 220 shown in FIG. 2 may be implemented as separate physical devices.

The processor 210 may perform basic arithmetic, logic, and input/output operations to process instructions of a computer program. The instructions may be provided from the memory 220 or from an external device, such as a server. The processor 210 may also control the overall operation of other components included in the computing system 200.

The processor 210 may be implemented as an array of multiple logic gates or as a combination of a general-purpose microprocessor and memory in which a program executable by the microprocessor is stored. The processor 210 may include a general-purpose processor, central processing unit (CPU), microprocessor, digital signal processor (DSP), controller, microcontroller, state machine, or the like. In certain environments, the processor 210 may include an application-specific integrated circuit (ASIC), programmable logic device (PLD), or field-programmable gate array (FPGA). For example, the processor 210 may also refer to combinations of such components―for example, a combination of a DSP and a microprocessor, multiple microprocessors, one or more microprocessors coupled with a DSP core, or any other similar configuration.

The processor 210 analyzes the pathological image. The analysis of the pathological image includes a process of segmenting or detecting specific tissues and/or cells among various tissues and cells expressed on a pathology slide, followed by conversion of the segmented or detected elements into information that can assist in making medical decisions. The process of converting into information that can assist in making medical decisions may involve extracting features that enable tasks such as classifying the pathological state of a patient's disease, diagnosing cancer, formulating a treatment plan for cancer, prescribing anticancer agents, or predicting disease onset. In particular, the process may include predicting a treatment response to an immune checkpoint inhibitor in a cancer patient.

For example, the processor 210 may derive information about a spatial location between tumor-associated macrophages (TAMs) and immune cells using at least one machine learning model.

Tissue samples from cancer patients may be collected at any point during the course of cancer treatment, including chemoradiotherapy (CRT), and the pathological image may be generated from such a tissue sample. For example, the tissue sample may be obtained at one or more time points, including: prior to chemoradiotherapy (CRT); after CRT but before administration of an immune checkpoint inhibitor; after the third administration of the immune checkpoint inhibitor; or at the time of surgery following the fifth administration of the immune checkpoint inhibitor. Preferably, the sample may be collected prior to CRT. However, the timing of collection is not limited to the time points listed above. The pathological image may be generated based on the corresponding tissue sample.

The processor 210 may calculate the proportion of immune cells located within a defined radius centered on each TAM by analyzing the pathological image using at least one machine learning model. The defined radius may range from 20μm to 120μm. For example, the radius may be 20μm, 30μm, 40μm, 50μm, 60μm, 70μm, 80μm, 90μm, 100μm, 110μm, or 120μm. More specifically, the radius may be 30μm, 50μm, or 100μm, and preferably 50μm, but is not limited thereto.

By analyzing the pathological image using at least one machine learning model, the processor 210 may also calculate a spatial proximity score between TAMs and immune cells by comparing the proportion of immune cells located within the defined radius centered on each TAM to the proportion of immune cells distributed throughout the tumor microenvironment (TME). Specifically, the spatial proximity score between the TAMs and immune cells may be expressed as the TAM-PD-L1 proximity score.

The TAM-PD-L1 proximity score refers to the relative ratio of PD-L1⁺/DAPI cells within a defined radius centered on each TAM to the overall PD-L1⁺/DAPI cell ratio in the TME within the region of interest (ROI).

Based on the analysis of the pathological image, the processor 210 may output a classification result indicating that the pathological slide image corresponds to a case in which the cancer patient is responsive to an immune checkpoint inhibitor, when the spatial proximity score between tumor- associated macrophages (TAMs) and immune cells exceeds a predetermined threshold. Specifically, the cancer patient being responsive to an immune checkpoint inhibitor may indicate both a high level of drug sensitivity to the immune checkpoint inhibitor and a favorable treatment outcome.

Meanwhile, the processor 210 may derive a spatial proximity score between tumor-associated macrophages (TAMs) and immune cells using various TAM markers, by analyzing the pathological image with at least one machine learning model. The TAM markers may include one or more selected from the group consisting of CD68⁺, CD14⁺, CD80⁺, CD163⁺, and CD206⁺.

Based on the analysis of the pathological image, the processor 210 may output a classification result indicating that the pathological image corresponds to a case in which the cancer patient has both high drug sensitivity to the immune checkpoint inhibitor and a favorable treatment outcome, when the spatial proximity score between tumor- associated macrophages (TAMs) and immune cells―derived using the various TAM markers―exceeds a predetermined threshold.

Meanwhile, the processor 210 may derive a spatial proximity score between TAMs and immune cells using the proportion of PD-L1⁺ cells distributed within various radii centered on each TAM, by analyzing the pathological image with at least one machine learning model.

Based on the analysis of the pathological image, the processor 210 may output a classification result indicating that the pathological image corresponds to a case in which the cancer patient has both high drug sensitivity to the immune checkpoint inhibitor and a favorable treatment outcome, when the spatial proximity score between TAMs and immune cells―derived using the proportion of PD-L1⁺ cells distributed within various radii centered on each TAM―exceeds a predetermined threshold.

Meanwhile, the processor 210 may derive a spatial proximity score between tumor-associated macrophages (TAMs) and immune cells based on the type of PD-L1⁺ cell, by analyzing the pathological image using at least one machine learning model. Specifically, the processor 210 may classify the PD-L1⁺ cells as follows: cells co-expressing CTLA-4⁺, CD4⁺, CD8⁺, FoxP3⁺, or PD-1⁺ may be classified as PD-L1⁺ T cells; cells co-expressing CK⁺ may be classified as PD-L1⁺ tumor cells; and the remaining PD-L1⁺ cells may be classified as PD-L1⁺ other cells. Based on the analysis result, when the spatial proximity score between TAMs and immune cells―derived according to the type of PD-L1⁺ cell―exceeds a predetermined threshold, the processor 210 may output a classification result indicating that the pathological image corresponds to a case in which the cancer patient has both high drug sensitivity to the immune checkpoint inhibitor and a favorable treatment outcome.

The memory 220 may include a non-transitory computer-readable recording medium. In an example, the memory 220 may include a permanent mass storage device such as random access memory (RAM), read-only memory (ROM), a disk drive, a solid-state drive (SSD), or flash memory. In another example, non-volatile mass storage devices such as ROM, SSD, flash memory, or disk drives may be implemented as a separate permanent storage device distinct from the memory. The memory 220 may store an operating system (OS) and at least one program code.

These software components may be loaded from a computer-readable recording medium that is separate from the memory 220. Such a separate computer-readable recording medium may be a storage medium directly connectable to the computing system 200, and may include, for example, a floppy drive, disk, tape, DVD/CD-ROM drive, or memory card.

Although not illustrated in FIG. 2, the computing system 200 may further include a display device. Alternatively, the computing system 200 may be connected to an independent display device via a wired or wireless communication interface to transmit and receive data. Through the display device, various types of information may be provided to the user, such as the pathological image, the pathology slide image, analysis information derived from the pathology slide image, medical information, and additional information based on the medical information.

FIG. 3 is a diagram illustrating examples of how the processor analyzes a pathological image according to an embodiment. With reference to FIG. 3, examples of how the processor 210 analyzes a pathological image will be described below. The examples described with reference to FIG. 3 are applicable to the derivation of information about a spatial location between tumor-associated macrophages (TAMs) and immune cells, the calculation of the proportion of immune cells located within a defined radius centered on each TAM, and/or the calculation of a spatial proximity score between TAMs and immune cells.

Referring to FIG. 3, the processor 210 may analyze a pathological image 310 using a machine learning model 320. For example, based on the analysis of the pathological image 310, the processor 210 may calculate a spatial proximity score 330 between tumor-associated macrophages (TAMs) and immune cells.

The spatial proximity score 330 between tumor-associated macrophages (TAMs) and immune cells may be a score calculated by detecting and classifying TAMs and immune cells using markers specifically expressed in TAMs and in immune cells, respectively, based on a pathological image acquired from a patient prior to chemoradiotherapy (CRT), and by deriving information about a spatial location between the identified cells. Specifically, the immune cells may be PD-L1⁺ cells, and the spatial proximity score 330 between tumor- associated macrophages (TAMs) and immune cells may be a proximity score between TAMs and PD-L1⁺ cells.

The spatial proximity score 330 between tumor-associated macrophages (TAMs) and PD-L1⁺ cells may refer to the relative ratio of PD-L1⁺/DAPI⁺ cells located within a defined radius centered on each TAM to the overall PD-L1⁺/DAPI⁺ cell ratio in the tumor microenvironment (TME) within a region of interest (ROI).

Meanwhile, the processor 210 may output detection results in the form of multiple layers representing tissue structures in the pathological image 310 by using the machine learning model 320. In this case, the machine learning model 320 may be trained based on training data including a plurality of reference pathology slide images and corresponding reference label information, and may be configured to detect regions within the pathological image 310 that correspond to tissues in the reference pathology images. Additionally, the processor 210 may analyze the pathological image 310 and perform classification of the plurality of cells represented in the pathological image 310.

First, the processor 210 may analyze the pathological image 310, detect cells from the pathological image 310, and output the detection results in the form of layers representing the cells.

The processor 210 may output the detection results in the form of layers representing the cells in the pathological image 310 using the machine learning model 320. In this case, the machine learning model 320 may be trained using training data that includes a plurality of reference pathology slide images and corresponding reference label information, so as to detect the locations and types of cells in the pathological image 310 that correspond to those in the reference pathology slide images.

For example, the processor 210 may classify the plurality of cells included in the pathological image 310 into at least one of the following categories: T cells, myeloid-derived suppressor cells (MDSCs), dendritic cells (DCs), macrophages, natural killer (NK) cells, mast cells, and neutrophils. However, the classification of cells in the pathological image 310 by the processor 210 is not limited to the examples described above. In other words, the processor 210 may classify the cells in the pathological image 310 into multiple categories based on a variety of criteria, without being limited to the cell types described above. The cells included in the pathological image 310 may be grouped into multiple categories according to predefined criteria or user-defined criteria.

In one example, the processor 210 may classify PD-L1⁺ cells included in the pathological image 310 as follows: cells co-expressing CTLA-4⁺, CD4⁺, CD8⁺, FoxP3⁺, or PD-1⁺ may be classified as PD-L1⁺ T cells; cells co-expressing CK⁺ may be classified as PD-L1⁺ tumor cells; and the remaining PD-L1⁺ cells may be classified as PD-L1⁺ other cells.

In another example, the processor 210 may use a machine learning model to derive the proportion of PD-L1⁺ cells located within a defined radius centered on each TAM in the pathological image. The proportion of PD-L1⁺ cells within a defined radius centered on a TAM may be determined by drawing a circle with a radius of 30μm to 120μm centered on the TAM and calculating the number of PD-L1⁺ cells and the number of total cells (DAPI⁺ cells) within that circle.

Here, the machine learning model 320 refers to a statistical learning algorithm implemented based on the structure of a biological neural network, or to a structure that executes such an algorithm.

For example, the machine learning model 320 may be a model composed of artificial neurons, or nodes, that form a network through synaptic connections―similar to biological neural networks―and that learn by iteratively adjusting synaptic weights to reduce the error between the correct output corresponding to a specific input and the inferred output. Through this process, the model acquires problem-solving capabilities. The machine learning model 320 may include, for example, any probabilistic model or neural network model used in artificial intelligence learning methods such as deep learning.

For example, the machine learning model 320 may be implemented as a multilayer perceptron (MLP) composed of multiple layers of nodes and the connections between the layers. The machine learning model 320 according to this embodiment may be implemented using one of various artificial neural network architectures that include an MLP. For example, the machine learning model 320 may include an input layer that receives input signals or data from an external source, an output layer that outputs signals or data corresponding to the input, and at least one hidden layer located between the input and output layers, which receives signals from the input layer, extracts features, and transmits the extracted features to the output layer. The output layer receives signals or data from the hidden layer and outputs the information to the outside.

Accordingly, the machine learning model 320 may be trained to receive one or more pathological images 310 and extract information related to one or more target objects (e.g., cells, tissues, structures, etc.) included in the pathological images 310 and/or biomarker expression information.

In one example, the processor 210 may determine a spatial proximity score between tumor-associated macrophages (TAMs) and PD-L1⁺ cells based on the ratio of PD-L1⁺ cells within a 50μm radius centered on each TAM and the ratio of PD-L1⁺ cells within the entire tumor microenvironment (TME) in the region of interest (ROI). This spatial proximity score refers to the relative ratio of PD-L1⁺/DAPI⁺ cells within a defined radius centered on each TAM compared to the PD-L1⁺/DAPI⁺ cell ratio across the entire TME in the ROI, and may be expressed as the TAM-PD-L1 proximity score.

In addition, the processor 210 may output the predicted responsiveness to an immune checkpoint inhibitor based on the TAM-PD-L1 proximity score. If the TAM-PD-L1 proximity score is below a predetermined cut-off value, the processor 210 may determine that the responsiveness to the immune checkpoint inhibitor is negative. Conversely, if the TAM-PD-L1 proximity score is equal to or greater than the predetermined cut-off value, the processor 210 may determine that the responsiveness to the immune checkpoint inhibitor is positive.

FIG. 4 is a flowchart illustrating an example method of analyzing a pathological image according to an embodiment. The method illustrated in FIG. 4 includes a series of steps that may be sequentially processed by the computing system 10 or 200 or the processor 210, as illustrated in FIGS. 1 and 2. Accordingly, even if certain details are omitted below, the descriptions provided above with respect to the computing system 10 or 200 and the processor 210 shown in FIGS. 1 and 2 are equally applicable to the method illustrated in FIG. 4.

Referring to FIG. 4, at step 410, the processor 210 detects cells from a pathological image of a patient. At step 420, the processor derives information about a spatial location between tumor-associated macrophages (TAMs) and immune cells based on the detected cells. At step 430, the processor predicts the patient's cancer treatment response to an immune checkpoint inhibitor (ICI) based on the derived spatial location information.

The spatial location information may refer to the proportion of immune cells located within a defined radius centered on each tumor-associated macrophage (TAM), and the cancer patient's treatment response to an immune checkpoint inhibitor may vary depending on this spatial location information.

The defined radius may range from 20μm to 120μm. For example, the radius may be 20μm, 30μm, 40μm, 50μm, 60μm, 70μm, 80μm, 90μm, 100μm, 110μm, or 120μm. More specifically, the radius may be 30μm, 50μm, or 100μm, and preferably 50μm, but is not limited thereto.

A more detailed description of the method for predicting a treatment response to an immune checkpoint inhibitor is provided below.

According to one aspect, there is provided a method of predicting a cancer treatment response to an immune checkpoint inhibitor (ICI), the method being performed by a device including at least one processor, the method including: detecting cells in a pathological image of a patient;

deriving information about a spatial location between tumor-associated macrophages (TAMs) and immune cells based on the detected cells; and

predicting the cancer treatment response of the patient to the ICI based on the derived spatial location information.

As used in the present specification, the term "pathological image" refers to an image obtained by scanning a pathology slide that has been chemically processed, fixed, and stained in order to microscopically observe tissue or other material extracted from the human body. The pathological image may include a high-resolution digitized whole slide image (WSI), which may be obtained from a hematoxylin and eosin (H&E) stained slide, an immunohistochemistry (IHC) stained slide, or a multiplex immunofluorescence (mIF) slide. Preferably, the image may be obtained from an mIF slide, but is not limited thereto.

The pathological image may also refer to a portion of the high-resolution whole slide image, such as one or more patches. The pathological image may refer to a digital image obtained by scanning a pathology slide using a digital scanner, and may include information on cells, tissues, and/or structures of the human body. The pathological image may contain one or more patches, and each patch may be associated with histological information applied through an annotation process (e.g., tagging). In the present specification, the term 'pathological image' may be used interchangeably with terms such as 'pathology image,' 'pathology slide image,' 'tissue slide image,' or 'whole slide image (WSI)'. Also, in the present specification, the term 'pathological image' may refer to 'at least a portion of a pathology image'.

In an embodiment, the pathological image may be generated based on a tissue sample collected from a cancer patient at any point during the course of cancer treatment, including chemoradiotherapy. For example, the tissue sample may be obtained from the patient at one or more time points selected from before chemoradiotherapy, after chemoradiotherapy but before administration of an immune checkpoint inhibitor, after the third administration of an immune checkpoint inhibitor, or at the time of surgery following the fifth administration of the immune checkpoint inhibitor. Preferably, the sample may be obtained prior to chemoradiotherapy (CRT), but is not limited thereto.

In an embodiment, the pathological image may be generated based on a tissue sample collected from a patient with locally advanced cancer. The term "locally advanced cancer" refers to a condition in which the tumor has spread to nearby tissues or lymph nodes but has not metastasized to distant organs.

In the present specification, the term "tumor-associated macrophage (TAM)" refers to a representative type of immune cell present within the tumor microenvironment (TME), and denotes macrophages formed by the differentiation of monocytes that have infiltrated tumor tissue. Unlike M1-type macrophages typically observed in general inflammatory responses, tumor-associated macrophages (TAMs) predominantly exhibit an immunosuppressive M2-type phenotype and play a key role in establishing a tumor-promoting environment, including tumor progression, angiogenesis, metastasis promotion, and immune evasion. These TAMs contribute to creating a favorable environment for tumor survival and dissemination within the body by suppressing antitumor immune responses. Accordingly, regulation of TAM activation or polarization can induce antitumor immune responses or suppress tumor growth, making TAMs a key therapeutic target in cancer immunotherapy.

In the present specification, the term "immune cells" refers to cells that function to protect the body from pathogens, abnormal cells, external antigens, and the like. These immune cells are involved in both innate and adaptive immunity, each contributing to the regulation and execution of immune responses through their respective functions. Representative immune cells include innate immune cells such as macrophages, dendritic cells, natural killer (NK) cells, and neutrophils, as well as adaptive immune cells such as T cells and B cells.

In an embodiment, the immune cells may be PD-L1⁺ cells. For example, the PD-L1⁺ cells may include, but are not limited to, one or more selected from the group consisting of T cells, myeloid-derived suppressor cells (MDSCs), dendritic cells (DCs), macrophages, NK cells, mast cells, and neutrophils.

In an embodiment, the step 410 of detecting cells in the pathological image may include classifying tumor-associated macrophages (TAMs) and immune cells using markers specifically expressed in TAMs and markers specifically expressed in immune cells.

As used in the present specification, the term "marker" refers to a biological indicator used to identify or distinguish the type, condition, or functional characteristics of a specific cell, and may be a protein expressed on the cell surface (e.g., CD4, CD8, PD-1, etc.).

Specifically, the marker specifically expressed in TAMs may include, but is not limited to, one or more selected from the group consisting of CD14, CD68, CD80, CD163, and CD206. The marker specifically expressed in immune cells may include, but is not limited to, one or more selected from the group consisting of PD-L1; and CTLA-4, CD4, CD8, FoxP3, and PD-1. For example, the marker specifically expressed in immune cells may be a combination of PD-L1 and CD4, PD-L1 and CD8, PD-L1 and FoxP3, or PD-L1 and PD-1.

As used in the present specification, the term "spatial location information" refers to the physical position or arrangement of a specific cell within a space, and includes information indicating the relative location of that cell in relation to surrounding cells. The spatial location information may be expressed in two-dimensional or three-dimensional spatial coordinates and may include information on the distribution, density, and positional interactions of specific cells within a tissue section.

In an embodiment, the spatial location information may include the proportion of immune cells located within a radius of 20μm to 120μm centered on each tumor-associated macrophage (TAM), and the immune cells may be PD-L1⁺ cells.

In an embodiment, the method of predicting a cancer treatment response to an immune checkpoint inhibitor (ICI) may further include: calculating a spatial proximity score between tumor-associated macrophages (TAMs) and immune cells by comparing the derived spatial location information with the proportion of immune cells distributed in the tumor microenvironment (TME); and predicting the cancer treatment response to the ICI based on the spatial proximity score between the TAMs and immune cells.

As used in the present specification, the term "tumor microenvironment (TME)" may refer to the environment composed of tissues and cells surrounding the tumor. For example, the tumor microenvironment may include not only the tumor cells themselves, but also the surrounding vasculature, extracellular matrix (adjacent tissue), immune cells, inflammatory mediators, and the like, without being limited thereto.

As used in the present specification, the term "spatial proximity score between tumor-associated macrophages (TAMs) and immune cells" refers to a metric that quantitatively represents the spatial distance or positional relationship between TAMs and other immune cells within tissue. The "spatial proximity score between tumor-associated macrophages (TAMs) and immune cells" may be calculated based on the physical distance, relative distribution, or degree of spatial clustering between TAMs and immune cells, and may be derived using coordinate information extracted from two-dimensional or three-dimensional tissue images. Specifically, the "spatial proximity score between tumor-associated macrophages (TAMs) and immune cells" may be a "spatial proximity score between tumor-associated macrophages (TAMs) and PD-L1⁺ cells," and may be expressed as the TAM-PD-L1 proximity score.

As used in the present specification, the term "TAM-PD-L1 proximity score" refers to a metric derived by comparing the proportion of PD-L1⁺ cells located within a defined radius centered on a tumor-associated macrophage (TAM) to the proportion of PD-L1⁺ cells across the entire tumor microenvironment (TME) within a region of interest (ROI). Specifically, this term refers to the relative ratio of PD-L1⁺/DAPI⁺ cells within a defined radius centered on each TAM to the PD-L1⁺/DAPI⁺ cell ratio across the entire TME in the ROI. The TAM-PD-L1 proximity score may be expressed by the following equation.

[Equation 1]

In an embodiment, the method of predicting a cancer treatment response to an immune checkpoint inhibitor (ICI) may further include determining that the patient is cancer treatment-responsive to the immune checkpoint inhibitor is present when the spatial proximity score between tumor-associated macrophages (TAMs) and immune cells is equal to or greater than a cut-off value.

Here, the cut-off value is used as a criterion for distinguishing between a high and a low TAM-PD-L1 proximity score group. Specifically, when the TAM-PD-L1 proximity score is equal to or greater than 1.67, the case is classified into the high-score group (High); when the score is less than 1.67, the case is classified into the low-score group (Low). Additionally, a TAM-PD-L1 proximity score equal to or greater than 1.67 may indicate that the patient is cancer treatment-responsive to the immune checkpoint inhibitor.

As used in the present specification, the term "cancer treatment response" may include pathological complete response (pCR) or responsiveness to an immune checkpoint inhibitor. For example, "being responsive to an immune checkpoint inhibitor" may refer to a case in which the patient exhibits high sensitivity to the immune checkpoint inhibitor or demonstrates improved treatment outcome following its administration. Here, "improved treatment outcome" refers to a case in which, following cancer diagnosis, the patient exhibits a high likelihood of survival based on treatment and follow-up observation, and more specifically, may include cases in which the absence or reduction of invasive cancer in body tissues is confirmed following treatment such as chemotherapy, radiotherapy, administration of an immune checkpoint inhibitor, or chemoradiotherapy, or cases in which the risk of tumor recurrence is low.

As used in the present specification, the term 'pathological complete response (pCR)' refers to a state in which, following surgical tissue examination, no cancer cells are detected even if a tumor mass is present. For example, pathological complete response may refer to, but is not limited to, the absence of invasive cancer within human tissue as a result of treatment such as chemotherapy, radiotherapy, or immunotherapy including immune checkpoint inhibitors. Specifically, pathological complete response may refer to a state in which all or at least some of the tumor cells previously present in human tissue have been eliminated as a result of anticancer treatment. In general, when pathological complete response (pCR) is achieved, the patient may experience a longer survival period.

As used in the present specification, the term "cancer" refers to a physiological condition in an animal characterized typically by abnormal or uncontrolled cell growth. The cancer may be associated with, for example, metastasis; interference with normally functioning surrounding cells; the release of cytokines or other secretory products at abnormal levels; suppression or enhancement of inflammatory or immunological responses; neoplasia; premalignant or malignant conditions; or invasion of adjacent or distant tissues or organs, such as lymph node involvement.

In an embodiment, the cancer may be a solid tumor. Specifically, the solid tumor may include, but is not limited to, one or more selected from the group consisting of lung cancer, skin cancer, stomach cancer, gastrointestinal cancer, intestinal cancer, colorectal cancer, colon cancer, rectal cancer, pancreatic cancer, liver cancer, thyroid cancer, uterine cancer, cervical cancer, ovarian cancer, testicular cancer, prostate cancer, breast cancer, and oral cancer.

In an embodiment, the cancer may be microsatellite stable (MSS).

As used in the present specification, the term "microsatellite stable (MSS)" refers to a genetic characteristic in which short repetitive sequences known as microsatellites remain relatively stable, with minimal mutations occurring during the process of DNA replication within a cell. MSS is typically observed when the DNA mismatch repair (MMR) system is functioning normally. Tumors with MSS generally have a lower mutational burden in the genome and are less likely to induce an immune response, which often results in reduced responsiveness to immunotherapy.

As used in the present specification, the term "mismatch repair (MMR)" may refer to a function in which specific proteins recognize and correct base-pair mismatches that occur during DNA replication.

As used in the present specification, the term "microsatellite instability (MSI)" may refer to a phenomenon in which mutations occurring in microsatellites are not repaired due to genetic defects in MMR proteins, resulting in changes in the number of repeats and deviations in microsatellite length compared to normal cells. Microsatellite instability is characterized by a high mutation rate and the generation of frameshift-peptide neoantigens, and creates a highly immunogenic environment by increasing the infiltration of lymphocytes into and around the tumor.

As used in the present specification, the term "immune checkpoint inhibitor (ICI)" may refer to a substance that restores the antitumor activity of immune cells, including T cells, by inhibiting the activity of immune checkpoint proteins. The immune checkpoint inhibitor may be any substance capable of inhibiting the function of immune checkpoint proteins without limitation, and may include, for example, a protein, compound, natural substance, DNA, RNA, or peptide. Preferably, the immune checkpoint inhibitor may be an antibody; more preferably, a monoclonal antibody; and even more preferably, a human antibody, a humanized antibody, or a chimeric antibody.

For example, the immune checkpoint inhibitor may include, but is not limited to, a protein, compound, natural substance, DNA, RNA, peptide, or a combination thereof that inhibits the function of any one selected from the group consisting of PD-L1, PD-1, CTLA-4, PD-L2, LTF2, LAG3, A2aR, TIGIT, TIM-3, B7-H3, B7-H4, VISTA, CD47, BTLA, KIR, and IDO.

Specifically, the immune checkpoint inhibitor may include, but is not limited to, any one selected from the group consisting of an anti-PD-L1 antibody, anti-PD-1 antibody, anti-CTLA-4 antibody, anti-PD-L2 antibody, LTF2-modulating antibody, anti-LAG3 antibody, anti-A2aR antibody, anti-TIGIT antibody, anti-TIM-3 antibody, anti-B7-H3 antibody, anti-B7-H4 antibody, anti-VISTA antibody, anti-CD47 antibody, anti-BTLA antibody, anti-KIR antibody, anti-IDO antibody, and a combination thereof.

In an embodiment, the immune checkpoint inhibitor may include, but is not limited to, one or more selected from the group consisting of pembrolizumab, nivolumab, atezolizumab, durvalumab, avelumab, ipilimumab, tremelimumab, camrelizumab, ciplizumab, toligliumab, relatlimab, dostarlimab, Bavencio, emplilimab, cemiplimab, batisimab, LBL-007, and BAT1308. For example, the immune checkpoint inhibitor may include, but is not limited to, nivolumab.

As used in the present specification, the term "device including at least one processor" refers to an electronic component or system that includes one or more processors capable of performing functions such as computation, data processing, and generation of control signals. The processor may include, for example, a central processing unit (CPU), graphics processing unit (GPU), digital signal processor (DSP), artificial intelligence processor, microcontroller (MCU), or other devices capable of performing computational and control functions. Specifically, the processor may be configured to execute the machine learning model or operate based on the machine learning model.

As used in the present specification, the term "machine learning model" or "learning model" may refer to the structure of a computer algorithm that learns from data to discover patterns and make predictions or decisions. The machine learning model generally learns from training data and subsequently performs prediction or classification on new data. For example, the machine learning model may include any model used to infer an answer based on a given input.

In an embodiment, the machine learning model may include an artificial neural network model including an input layer, a plurality of hidden layers, and an output layer. Here, each layer may include one or more nodes. For example, the machine learning model may be trained to infer histological components from a pathology image and/or from at least one patch contained in the pathology image. In this case, the histological components generated through annotation may be used to train the machine learning model. In another example, the machine learning model may be trained to infer a cancer patient's treatment responsiveness based on interaction scores, characteristics of at least one of cells, tissues, or structures in the pathology image, and/or clinical information about the patient. Further, the machine learning model may also include weights associated with the multiple nodes contained in the model. The weights may include any parameters associated with the machine learning model.

In the present specification, the term "machine learning model" may refer to an artificial neural network model, and vice versa. The machine learning model according to the present specification may be trained using various learning methods. For example, the present disclosure may utilize, but is not limited to, supervised learning, unsupervised learning, or reinforcement learning.

As used in the present specification, the term 'learning' may refer to any process of modifying the weights included in a machine learning model using at least one patch, interaction score, histological information, and/or clinical information. In an embodiment, learning may refer to a process of modifying or updating the weights associated with the machine learning model by performing one or more forward propagations and backward propagations using at least one patch and histological information.

According to one aspect, a computer-readable recording medium storing a computer program is provided for executing the method of predicting a treatment response to an immune checkpoint inhibitor (ICI) on a computer.

According to another aspect, provided is a computing system including: at least one memory; and at least one processor connected to the memory and configured to execute at least one computer-readable program stored in the memory, wherein the processor is configured to detect cells in a pathological image of a patient, derive information about a spatial location between tumor-associated macrophages (TAMs) and immune cells based on the detected cells, and predict the cancer treatment response of the patient to the immune checkpoint inhibitor (ICI) based on the derived spatial location information.

According to one aspect, there is provided a method of treating cancer, the method being performed using a device including at least one processor, the method including: detecting cells in a pathological image of a patient; acquiring information about a spatial location between tumor-associated macrophages (TAMs) and immune cells based on the detected cells; calculating a spatial proximity score between the TAMs and immune cells by comparing the acquired spatial location information with the proportion of immune cells distributed in the tumor microenvironment (TME); performing an operation of predicting a cancer treatment response to an immune checkpoint inhibitor based on the spatial proximity score between the TAMs and immune cells; and administering the immune checkpoint inhibitor to the patient when it is determined, based on the predicted treatment response, that the patient is cancer treatment-responsive to the immune checkpoint inhibitor.

The following examples are provided to describe the present disclosure in greater detail. However, these examples are illustrative only and should not be construed as limiting the scope of the present disclosure in any way.

Example 1. Development of AI-powered pathology slide image analyzer

The computing system 10, 200 may correspond to an AI-powered pathology slide image analyzer. An AI-powered pathology slide image analyzer was used to analyze the treatment response to an immune checkpoint inhibitor in solid cancer based on AI.

The AI-powered pathology slide image analyzer is capable of extracting characteristics or information on cells, tissues, and/or structures of the human body contained in a pathology image. Specifically, the AI-powered pathology slide image analyzer may use a computing device or an image analyzer that analyzes multiplex immunofluorescence (mIF) images to estimate the spatial locations of various immune cells―including tumor-associated macrophages (TAMs) and T lymphocytes expressing specific markers such as PD-L1, PD-1, CTLA-4, CD8, and FOXP3―within a tumor microenvironment (TME).

Since the mIF image data include protein expression information at the cellular level and spatial location information (e.g., x and y coordinates), the AI-powered pathology slide image analyzer can quantitatively analyze the distances between various immune cells―including TAMs and T lymphocytes expressing PD-L1―based on the acquired information.

Example 2. Study Population

Tumor samples from the VOLTAGE clinical trial (ClinicalTrials.gov identifier: NCT02948348) were analyzed using the AI-powered pathology slide image analyzer.

The VOLTAGE clinical trial is a multicenter phase 1/2 study conducted to evaluate the efficacy of five administrations of nivolumab followed by surgery in patients with locally advanced rectal cancer (LARC) who underwent chemoradiotherapy (CRT). The patients enrolled in the VOLTAGE clinical trial completed CRT including capecitabine and radiotherapy, and then received five cycles of nivolumab monotherapy. Subsequently, curative resection was performed, including abdominoperineal resection with either sphincter-preserving surgery or total mesorectal excision (TME). Administration of nivolumab began within 14 days after completion of CRT and prior to surgery, and TME surgery was performed 12 weeks later. Among the patients in the VOLTAGE clinical trial, 38 had MSS LARC.

Example 3. Spatial Analysis of Tumor-Associated Macrophages (TAMs) and Immune Cells

3.1 Detection and Classification of Immune Cells Using Multiplex Immunofluorescence (mIF) Images

To evaluate whether an increase in specific immune cell populations around tumor-associated macrophages (TAMs) is correlated with pathological complete response (pCR) to nivolumab in patients with locally advanced rectal cancer (LARC), spatial analysis was performed between TAMs and immune cells. For this analysis, multiplex immunofluorescence (mIF) images collected prior to chemoradiotherapy (CRT) from 38 patients enrolled in the VOLTAGE study were analyzed using the AI-powered pathology slide image analyzer described in Example 1.

The mIF images collected prior to CRT from the 38 patients were obtained by scanning slides that had been stained using multiplex immunofluorescence. Regions of interest (ROIs) were selected by a pathologist based on tissue size and specific criteria.

The mIF images visually display a total of 35 markers specifically expressed in T cells, TAMs, myeloid-derived suppressor cells (MDSCs), and dendritic cells (DCs). The markers associated with T cells (T-cell ①, T-cell ②), TAMs, MDSCs, and DCs are shown in Table 1.

Table 1 lists the antibodies, clones, and secondary antibodies corresponding to the markers for each cell type. (DAPI: 4'6-diamidino-2-phenylindole; DC: dendritic cell; MDSC: myeloid-derived suppressor cell; M-MDSC: monocytic MDSC; PMN-MDSC: polymorphonuclear MDSC; TAM: tumor-associated macrophage) Representative mIF images for five panels incorporating these markers are shown in FIG. 5.

FIG. 5 shows mIF images analyzed using antibodies for six markers specifically expressed in each of T cells, TAMs, MDSCs, and DCs, along with DAPI. Each of the five panels corresponding to T cells, TAMs, MDSCs, and DCs includes antibodies for six markers specifically expressed in the respective cell type and DAPI information, and represents images analyzing a total of 35 variables.

The number of PD-L1 positive cells in the mIF images was analyzed using QuPath, an open-source software for digital image analysis. Cell detection was performed using the DAPI channel, and single-channel fluorescence detection was conducted using the positive cell detection function with a predefined threshold for each marker in each panel.

The number of PD-L1⁺ cells in the mIF images was determined based on the number of cells co-expressing PD-L1⁺ with CTLA-4⁺, CD4⁺, CD8⁺, FoxP3⁺, or PD-1⁺. The number of TAMs was determined based on the number of CD14⁺, CD68⁺, CD80⁺, CD163⁺, or CD206⁺ cells. The number of PD-L1⁺ T cells was determined based on the number of cells co-expressing CD4⁺/PD-L1⁺, CD8⁺/PD-L1⁺, or FoxP3⁺/PD-L1⁺.

3.2 Distance Analysis Between Tumor-Associated Macrophages (TAMs) and Immune Cells

Fluorescence images of markers specifically expressed in each cell type were analyzed to obtain the location information of cells expressing the respective markers within regions of interest (ROIs). Based on this information, the distances between tumor-associated macrophages (TAMs) and various immune cells, including T lymphocytes, were analyzed within the tumor microenvironment (TME). In addition, the proportions of various immune cells, including T lymphocytes, located within a defined radius centered on each TAM were analyzed.

Using the method described in Example 3.1, cells were detected and classified using multiplex immunofluorescence (mIF) panels constructed based on markers specifically expressed in each cell type. Specifically, among PD-L1⁺ cells, those co-expressing CTLA-4⁺, CD4⁺, CD8⁺, FoxP3⁺, or PD-1⁺ were classified as PD-L1⁺ T cells, and those co-expressing CK⁺ were classified as PD-L1⁺ tumor cells. The remaining PD-L1⁺ cells were classified as PD-L1⁺ other cells.

For each classified CD68⁺ TAM, a circular area with a radius of 50μm was defined, and the number of PD-L1⁺ cells within that area was measured. In addition, the proportion of each PD-L1⁺ cell type relative to the total number of DAPI⁺ cells within the defined circular area was calculated and compared to the corresponding proportion in the entire TME within the ROI. In other words, the ratio of PD-L1⁺ to DAPI⁺ cells within a defined radius centered on each TAM was calculated and compared to the PD-L1⁺/DAPI cell ratio in the entire TME within the ROI, and the TAM-PD-L1 proximity score was derived based on this comparison.

[Equation 1]

More specifically, FIGS. 6 and 7 schematically illustrate a method of analyzing information about a spatial location between TAMs and PD-L1⁺ cells using the AI-powered pathology slide image analyzer described in Example 1.

FIG. 6 is a schematic diagram illustrating a method of analyzing information about a spatial location between TAMs and PD-L1⁺ cells using the AI-powered pathology slide image analyzer according to an embodiment.

FIG. 7 is a schematic diagram illustrating a method of analyzing the proportion of PD-L1⁺ cells located within a defined radius (50μm) centered on each tumor-associated macrophage (TAM).

As shown in FIGS. 6 and 7, the AI-powered pathology slide image analyzer described in Example 1 can be used to derive a spatial proximity score between tumor-associated macrophages (TAMs) and immune cells.

Example 4. Prediction of Pathological Complete Response (pCR) Using TAM-PD-L1 Proximity Score

To evaluate the effect of spatial proximity between tumor-associated macrophages (TAMs) and immune cells on the prediction of pathological complete response (pCR) following immune checkpoint inhibitor therapy, TAM-PD-L1 proximity scores were calculated from multiplex immunofluorescence (mIF) images collected prior to chemoradiotherapy (CRT) for each MSS LARC patient, using the method described in Example 3.2, and were compared between the immune checkpoint inhibitor responder group and non-responder group. In addition, the pCR rates were calculated for the high and low TAM-PD-L1 proximity score groups.

Specifically, the cut-off value of the TAM-PD-L1 proximity score was gradually varied, and for each cut-off value, patients were classified into a high-score group (High) and a low-score group (Low), and the pCR rate was calculated for each group. Based on these results, the optimal cut-off value was determined.

Table 2 shows the pCR rates and statistical significance for the groups classified as High or Low based on gradually adjusted cut-off values of the TAM-PD-L1 proximity score.

As shown in Table 2, the optimal cut-off value for classifying patients into high and low TAM-PD-L1 proximity score groups was determined to be 1.67.

Meanwhile, using the cut-off value of 1.67 to classify patients into high and low TAM-PD-L1 proximity score groups, the pCR rates for each group were calculated. The results are presented in FIGS. 8 and 9.

FIG. 8 is a graph showing the TAM-PD-L1 proximity scores calculated for the responder and non-responder groups to immune checkpoint inhibitor therapy at the pre-CRT time point in MSS LARC patients, along with the pCR rates for the high-score and low-score groups classified using the cut-off value of 1.67. Responders are shown in blue and non-responders in red.

FIG. 9A is a graph illustrating the predictive performance of the TAM-PD-L1 proximity score for pCR, represented by the area under the receiver operating characteristic curve (AUROC). FIG. 9B compares the pCR rates between the high-score and low-score groups classified using the cut-off value of 1.67.

As shown in FIGS. 8 to 9B and Table 2, analysis of the spatial proximity between TAMs and PD-L1⁺ cells revealed that the optimal cut-off value for the TAM-PD-L1 proximity score was 1.67. Based on this cut-off, the patients were divided into two groups: 25 patients were classified into the high-score group (≥1.67), and 13 patients were classified into the low-score group (<1.67) (2.19 vs. 1.69, p=0.011).

As shown in FIGS. 8 and 9B, the overall pCR rate among all 38 patients was 28.9 %. In the group with a TAM-PD-L1 proximity score of 1.67 or higher, the pCR rate was 44 % (11 out of 25 patients), whereas in the group with a score below the cut-off, no pCR was observed (0 %, 0 out of 13 patients). The difference in pCR rate between the two groups was statistically significant (44.0 % vs. 0.0 %, p=0.006).

In addition, as shown in FIG. 9A, the area under the ROC curve (AUROC) for evaluating the predictive performance of the TAM-PD-L1 proximity score with respect to pCR was 0.768 (p=0.005).

These results indicate that the spatial proximity between TAMs and PD-L1⁺ cells serves as an important indicator for predicting pCR following concurrent chemoradiotherapy and immune checkpoint inhibitor therapy administered prior to surgery.

Example 5. Evaluation of Predictive Performance of TAM-PD-L1 Proximity Score for Pathological Complete Response (pCR) According to PD-L1+ Cell Type

For the mIF images of 38 patients, the classification criteria defined in Example 3.2 were applied to determine the proportion of PD-L1⁺ cells by cell type.

Specifically, among PD-L1⁺ cells, those co-expressing CTLA-4⁺, CD4⁺, CD8⁺, FoxP3⁺, or PD-1⁺ were classified as PD-L1⁺ T cells, and those co-expressing CK⁺ were classified as PD-L1⁺ tumor cells. The remaining PD-L1⁺ cells were classified as PD-L1⁺ other cells. Using various TAM markers, the proportions of PD-L1⁺ tumor cells, PD-L1⁺ T cells, and PD-L1⁺ other cells were measured for each patient based on their mIF image, and the results are shown in FIG. 10. The proportions of PD-L1⁺ tumor cells, PD-L1⁺ T cells, and PD-L1⁺ other cells are shown in the bar graphs (on the left), and the total proportions of these cell types are presented in the pie charts (on the right).

FIG. 10 shows bar graphs representing the distribution ratios of PD-L1⁺ cells by cell type using various tumor-associated macrophage (TAM) markers and pie charts indicating the total proportions of PD-L1⁺ cell types for each patient. In two patients, no PD-L1⁺ cells were detected.

FIG. 10A shows the distribution ratios of PD-L1⁺ cells by cell type for each patient and the total proportions of each PD-L1⁺ cell type, measured using the CD68⁺ marker. FIG. 10B shows the distribution ratios of PD-L1⁺ cells by cell type for each patient and the total proportions of each PD-L1⁺ cell type, measured using the CD14⁺ marker. FIG. 10C shows the distribution ratios of PD-L1⁺ cells by cell type for each patient and the total proportions of each PD-L1⁺ cell type, measured using the CD80⁺ marker. FIG. 10D shows the distribution ratios of PD-L1⁺ cells by cell type for each patient and the total proportions of each PD-L1⁺ cell type, measured using the CD163⁺ marker. FIG. 10E shows the distribution ratios of PD-L1⁺ cells by cell type for each patient and the total proportions of each PD-L1⁺ cell type, measured using the CD206⁺ marker.

As shown in FIGS. 10A to 10E, T cells accounted for the highest proportion among the PD-L1⁺ cell populations.

Meanwhile, the predictive performance for pCR was evaluated based on TAM-PD-L1 proximity scores stratified by PD-L1⁺ cell type, using the area under the receiver operating characteristic curve (AUROC).

The AUROC value for evaluating the predictive performance of the proximity score between TAMs and PD-L1⁺ T cells for pCR was 0.768 (p=0.005), identical to the AUROC value of the TAM-PD-L1 proximity score. In contrast, the AUROC value for the proximity score between TAMs and PD-L1⁺ tumor cells in predicting pCR was 0.497 (p=0.519), and the AUROC value for PD-L1⁺ other cells was 0.566 (p=0.270).

In addition, the predictive performance for pCR based on the spatial distribution of other immune cells that do not express PD-L1 and are located near TAMs was found to be generally lower than the predictive performance for pCR based on the proximity score between TAMs and PD-L1⁺ T cells.

Specifically, the predictive performance of CD4(+) T cells and CD8(+) T cells was represented by AUROC values of 0.687 (p=0.038) and 0.690 (p=0.036), respectively. The predictive performance of CTLA-4(+) T cells and PD-1(+) T cells was represented by AUROC values of 0.589 (p=0.201) and 0.582 (p=0.220), respectively. The predictive performance of FOXP3(+) T cells was represented by an AUROC value of 0.731 (p=0.014). These values were lower than the predictive performance based on the proximity score between TAMs and PD-L1⁺ T cells.

These results indicate that, in predicting pCR to immune checkpoint inhibitor therapy following concurrent chemoradiotherapy (CRT) in locally advanced rectal cancer (LARC), the spatial distribution of PD-L1⁺ T cells is the most critical factor, and that the proximity score between TAMs and PD-L1⁺ T cells may serve as a biomarker for predicting pCR.

Example 6. Derivation of TAM-PD-L1 Proximity Score Using Various Tumor-Associated Macrophage (TAM) Markers

The TAM-PD-L1 proximity score was calculated using various tumor-associated macrophage (TAM) markers, and the predictive performance for pathological complete response (pCR) of the proximity score derived from each marker was evaluated based on the area under the ROC curve (AUROC).

Specifically, CD68⁺, CD14⁺, CD80⁺, CD163⁺, and CD206⁺ markers were used to detect TAMs in the multiplex immunofluorescence (mIF) images, and the proximity scores between TAMs and PD-L1⁺ cells were derived using the methods described in Examples 1 to 3.

FIG. 11 presents the proportions of PD-L1⁺ cells by cell type for each of the TAM markers CD68⁺, CD14⁺, CD80⁺, CD163⁺, and CD206⁺, as well as the TAM-PD-L1 proximity scores derived using these various TAM markers and the corresponding predictive performance for pCR evaluated as AUROC values.

FIG. 11 presents the evaluation results of the predictive performance (AUROC) of the TAM-PD-L1 proximity scores derived using various TAM markers for pCR. FIG. 11A is a graph showing the TAM-PD-L1 proximity scores derived using CD68⁺, CD14⁺, CD80⁺, CD163⁺, and CD206⁺ markers from mIF images collected prior to CRT. FIG. 11B shows the AUROC values representing the predictive performance for pCR of the TAM-PD-L1 proximity scores derived using each of these markers.

As shown in FIG. 11A, the TAM-PD-L1 proximity scores calculated using CD68⁺, CD14⁺, CD80⁺, CD163⁺, and CD206⁺ markers were consistently and significantly higher in the immune checkpoint inhibitor responder group (blue) than in the non-responder group (red). In addition, the group with high TAM-PD-L1 proximity scores―calculated using the various TAM markers―showed relatively higher pCR rates, whereas the group with low TAM-PD-L1 proximity scores showed lower pCR rates. This trend was observed consistently across all TAM markers.

As shown in FIG. 11B, the predictive performance (AUROC) for pCR of the TAM-PD-L1 proximity scores derived using the various TAM markers ranged from 0.764 to 0.781, demonstrating consistent predictive performance.

These results indicate that TAM-PD-L1 proximity scores derived using various TAM markers may be utilized to predict treatment responsiveness to immune checkpoint inhibitors.

Example 7. Analysis of TAM-PD-L1 Proximity Scores Derived Using Proportion of PD-L1+ Cells Distributed Within Various Radii Centered on Tumor-Associated Macrophages (TAMs)

In defining the TAM-PD-L1 proximity score, the proportion of immune cells located within various radii centered on each tumor-associated macrophage (TAM) was applied. Specifically, the TAM-PD-L1 proximity score was derived by applying the proportion of PD-L1⁺ cells located within radii of 30μm, 50μm, and 100μm centered on each TAM. The predictive performance for pathological complete response (pCR) of the derived TAM-PD-L1 proximity scores was evaluated as AUROC values, and the results are presented in FIG. 12.

FIG. 12 shows the results of deriving the TAM-PD-L1 proximity scores by applying the proportion of PD-L1⁺ cells located within radii of 30μm, 50μm, and 100μm centered on each TAM, and evaluating the predictive performance of the derived scores for pCR as AUROC values.

FIG. 12A is a graph showing the results of TAM-PD-L1 proximity scores derived by applying the proportion of PD-L1⁺ cells located within radii of 30μm, 50μm, and 100μm centered on each TAM, and FIG. 12B shows the results of evaluating the predictive performance for pCR, as AUROC values, of the TAM-PD-L1 proximity scores derived using the proportion of PD-L1⁺ cells located within each of the radii described above.

As shown in FIG. 12A, the TAM-PD-L1 proximity scores derived by applying the proportion of PD-L1⁺ cells located within radii of 30μm, 50μm, and 100μm centered on each TAM were consistently and significantly higher in the immune checkpoint inhibitor responder group (blue) than in the non-responder group (red). In the group with high TAM-PD-L1 proximity scores―derived by applying the proportion of PD-L1⁺ cells within each radius centered on TAMs―the pCR rate was relatively high, whereas in the group with low TAM-PD-L1 proximity scores, the pCR rate was low. This trend was consistently observed across all TAM-PD-L1 proximity scores derived using the proportions of PD-L1⁺ cells located within 30μm, 50μm, and 100μm radii centered on TAMs.

As shown in FIG. 12B, the predictive performance (AUROC) for pCR of the TAM-PD-L1 proximity scores derived using the proportion of PD-L1⁺ cells located within 30μm, 50μm, and 100μm radii centered on TAMs ranged from 0.694 to 0.768.

These results indicate that TAM-PD-L1 proximity scores derived using the proportion of PD-L1⁺ cells located within various radii centered on TAMs may be utilized to predict treatment responsiveness to immune checkpoint inhibitors.

Example 8. Quantification and Statistical Analysis

Group comparisons for categorical variables were conducted using the chi-squared test or Fisher's exact test.

The predictive performance of the models was evaluated using the area under the receiver operating characteristic curve (AUROC).

All statistical analyses were performed using R version and QuPath version.

Experimental Example 1. Assessment of Tumor Regression Grade (TRG) Values Defined by American Joint Committee on Cancer (AJCC) in High and Low TAM-PD-L1 Proximity Score Groups

To assess the potential utility of the TAM-PD-L1 proximity score as a biomarker for predicting prognosis following immune checkpoint inhibitor (ICI) therapy, the correlation between the TAM-PD-L1 proximity scores and tumor regression grades (TRG) was analyzed in the patients from Example 2. Specifically, the TAM-PD-L1 proximity scores derived for each patient using the method described in Example 3 were compared with the distribution of corresponding TRG values (0 to 3). The results are shown in FIG. 13.

FIG. 13 is a graph illustrating TRG values defined by the AJCC for each group, where the groups were stratified based on a cut-off value of 1.67 applied to the TAM-PD-L1 proximity scores. The proximity scores were derived from multiplex immunofluorescence (mIF) images collected prior to CRT using the AI-powered pathology slide image analyzer described in Example 1.

AJCC tumor regression grades (TRG) values refer to tumor regression grades defined by the American Joint Committee on Cancer (AJCC) and are used as a criterion to evaluate the pathological treatment response of tumors following anticancer therapy. In particular, TRG is used to assess the extent of tumor regression in patients who have received neoadjuvant treatment, and grading (TRG 0-3) is based on the proportion of tumor and fibrotic tissue observed in pathological examination.

Specifically, TRG 0 indicates a complete response, where the tumor has completely disappeared and corresponds to pathological complete response (pCR). TRG 1 indicates a near-complete response, in which only a minimal number of viable tumor cells remain. TRG 2 indicates a partial response, where the tumor has decreased in size but residual tumor remains. TRG 3 indicates a poor or no response, meaning there is little or no therapeutic effect.

As shown in FIG. 13, patients in the high TAM-PD-L1 proximity score group predominantly exhibited lower TRG values as defined by the AJCC. Specifically, TRG 0 was observed in 11 patients (44.0 %), TRG 1 in 2 patients (8.0 %), TRG 2 in 7 patients (28.0 %), and TRG 3 in 5 patients (20.0 %).

In contrast, patients in the low TAM-PD-L1 proximity score group tended to have relatively higher TRG values. Specifically, TRG 0 was observed in 0 patients (0.0 %), TRG 1 in 2 patients (18.2 %), TRG 2 in 6 patients (54.5 %), and TRG 3 in 3 patients (27.3 %).

In summary, patients in the high TAM-PD-L1 proximity score group were more likely to respond to immune checkpoint inhibitor therapy and achieve pathological complete response (pCR), whereas patients in the low-score group showed limited responsiveness and rarely exhibited pCR. In addition, patients with high TAM-PD-L1 proximity scores tended to exhibit lower TRG values, reflecting an effective response to immune checkpoint inhibitor therapy, whereas patients with low TAM-PD-L1 proximity scores tended to retain high TRG values due to minimal therapeutic effect.

These results suggest that the TAM-PD-L1 proximity score may serve as a predictive biomarker for response to immune checkpoint inhibitor therapy and that patients with high TAM-PD-L1 proximity scores are more likely to benefit from preoperative immune checkpoint inhibitor treatment.

Claims (20)

1. A method of predicting a cancer treatment response to an immune checkpoint inhibitor (ICI), the method being performed by a device comprising at least one processor, the method comprising:

   detecting cells in a pathological image of a patient;

   deriving information about a spatial location between tumor-associated macrophages (TAMs) and immune cells based on the detected cells; and

   predicting the cancer treatment response of the patient to the ICI based on the derived spatial location information.
2. The method of claim 1,

   wherein the immune cells are PD-L1 positive cells.
3. The method of claim 2,

   wherein the PD-L1 positive cells are one or more selected from the group consisting of T cells, myeloid-derived suppressor cells (MDSCs), dendritic cells (DCs), macrophages, natural killer (NK) cells, mast cells, and neutrophils.
4. The method of claim 1,

   wherein the detecting of cells in the pathological image comprises classifying the TAMs and the immune cells by using a marker specifically expressed in the TAMs and a marker specifically expressed in the immune cells.
5. The method of claim 4,

   wherein the marker specifically expressed in the TAMs is at least one selected from the group consisting of CD14, CD68, CD80, CD163, and CD206.
6. The method of claim 4,

   wherein the marker specifically expressed in the immune cells comprises PD-L1 and at least one selected from the group consisting of CTLA-4, CD4, CD8, FoxP3, and PD-1.
7. The method of claim 1,

   wherein the spatial location information comprises a proportion of immune cells present within a radius of 20 μm to 120 μm centered on the TAMs.
8. The method of claim 1, further comprising:

   calculating a spatial proximity score between the TAMs and the immune cells by comparing the derived spatial location information with a proportion of immune cells distributed in a tumor microenvironment (TME); and

   predicting a cancer treatment response to an immune checkpoint inhibitor based on the spatial proximity score between the TAMs and the immune cells.
9. The method of claim 8, further comprising:

   determining that the patient is cancer treatment-responsive to the immune checkpoint inhibitor when the spatial proximity score between the TAMs and the immune cells is equal to or greater than a cut-off value.
10. The method of claim 1,

    wherein the cancer is a solid cancer.
11. The method of claim 10,

    wherein the cancer is microsatellite stable (MSS).
12. The method of claim 10,

    wherein the solid cancer is at least one selected from the group consisting of lung cancer, skin cancer, stomach cancer, gastrointestinal cancer, intestinal cancer, colorectal cancer, colon cancer, rectal cancer, pancreatic cancer, liver cancer, thyroid cancer, uterine cancer, cervical cancer, ovarian cancer, testicular cancer, prostate cancer, breast cancer, and oral cancer.
13. The method of claim 1,

    wherein the immune checkpoint inhibitor (ICI) is at least one selected from the group consisting of pembrolizumab, nivolumab, atezolizumab, durvalumab, avelumab, ipilimumab, tremelimumab, camrelizumab, ciplizumab, toligliumab, relatlimab, dostarlimab, Bavencio, emplilimab, cemiplimab, batisimab, LBL-007, and BAT1308.
14. The method of claim 1,

    wherein the pathological image is obtained prior to chemoradiotherapy (CRT).
15. The method of claim 1,

    wherein the pathological image comprises a multiplex immunofluorescence (mIF) image.
16. A computer program stored in a computer-readable recording medium for executing the method of claim 1 on a computer.
17. A computing system comprising:

    at least one memory; and

    at least one processor connected to the memory and configured to execute at least one computer-readable program stored in the memory,

    wherein the at least one processor is configured to:

    detect cells in a pathological image of a patient;

    derive information about a spatial location between tumor-associated macrophages (TAMs) and immune cells based on the detected cells; and

    predict a cancer treatment response of the patient to an immune checkpoint inhibitor (ICI) based on the derived spatial location information.
18. The computing system of claim 17,

    wherein the at least one processor is further configured to:

    calculate a spatial proximity score between the TAMs and the immune cells by comparing the derived spatial location information with a proportion of immune cells distributed in a tumor microenvironment (TME); and

    predict a cancer treatment response of the patient to an immune checkpoint inhibitor based on the spatial proximity score between the TAMs and the immune cells.
19. The computing system of claim 18,

    wherein the at least one processor is further configured to

    determine that the patient is cancer treatment-responsive to the immune checkpoint inhibitor when the spatial proximity score between the TAMs and the immune cells is equal to or greater than a cut-off value.
20. A method of treating cancer, the method being performed using a device comprising at least one processor, the method comprising:

    detecting cells in a pathological image of a patient;

    deriving information about a spatial location between tumor-associated macrophages (TAMs) and immune cells based on the detected cells;

    calculating a spatial proximity score between TAMs and the immune cells by comparing the derived spatial location information with a proportion of immune cells distributed in a tumor microenvironment (TME);

    performing an operation of predicting a treatment response to an immune checkpoint inhibitor based on the spatial proximity score between the TAMs and the immune cells; and

    administering the immune checkpoint inhibitor to the patient when the patient is determined to be responsive to the immune checkpoint inhibitor based on the predicted treatment response.