Method and apparatus for predicting treatment response to immune checkpoint inhibitors
By employing a machine learning model to assess TIL density changes, the method predicts treatment response to immune checkpoint inhibitors, addressing the inefficacy in current PD-1 or PD-L1 targeting therapies and enhancing treatment effectiveness for cancer patients.
Patent Information
- Application Number
- PCT/KR2024/006848
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2023-05-24
- Filing Date
- 2024-05-21
- Publication Date
- 2024-11-28
AI Technical Summary
Current immune checkpoint inhibitors targeting PD-1 or PD-L1 are only effective for cancer patients with PD-L1 expressing cancers, and more than 50% of such patients do not respond to these treatments, with the reason for non-response remaining unknown, necessitating a method to predict treatment response effectively.
A method using a machine learning model to determine the density of tumor-infiltrating lymphocytes (TILs) at different time points and calculate the fold change in TIL density to predict a cancer patient's response to immune checkpoint inhibitors, thereby identifying likely responders.
This approach allows for more accurate prediction of cancer patients' responsiveness to immune checkpoint inhibitors, enabling informed treatment decisions and improving treatment outcomes.
Smart Images

Figure KR2024006848_28112024_PF_FP_ABST
Abstract
Description
METHOD AND APPARATUS FOR PREDICTING TREATMENT RESPONSE TO IMMUNE CHECKPOINT INHIBITORS
[0001] The present disclosure relates to a method and apparatus for predicting treatment response to immune checkpoint inhibitors.
[0002] Tumor cells act on the host immune system in several ways to evade immune defenses in the tumor microenvironment. This phenomenon is generally referred to as "cancer immune escape." One of the most important components in this system is the immune checkpoint co-inhibitory signal mediated by programmed cell death protein 1 (PD-1) receptor and a ligand thereof, which is a programmed death ligand (PD-L1).
[0003] Recently, as cancer treatment by activating the body's immune system to kill cancer cells, cancer immunotherapy, in particular, immune checkpoint inhibitors, have been developed and used. One of the most widely used immune checkpoint inhibitors are inhibitors using monoclonal antibodies against PD-1 or PD-L1. These inhibitors block the binding of PD-1 on the surface of T cells to PD-L1 on the surface of cancer cells, thereby activating immune cells such as T cells to eliminate cancer cells. Immune checkpoint inhibitors, such as monoclonal antibodies targeting PD-1 or PD-L1 using this mechanism of action, are considered a new standard treatment for various cancer treatments.
[0004] However, inhibitors using monoclonal antibodies against PD-1 or PD-L1 are only effective against cancer cells expressing PD-L1. Nevertheless, it has been reported that more than 50% of cancer patients with PD-L1 expressing cancers do not respond to PD-1 or PD-L1 inhibitors, and the reason for this remains unknown. Therefore, in order to effectively treat cancer using immune checkpoint inhibitors targeting PD-1 or PD-L1, there is a need to select patients who show a high treatment effect with the corresponding immune checkpoint inhibitor.
[0005] In light of this, the inventors of the present application have developed a method to predict treatment response to immune checkpoint inhibitors by a computing device.
[0006] An objective of the present disclosure is to provide a method and apparatus for predicting treatment response to immune checkpoint inhibitors. Another objective of the present disclosure is to provide a computer-readable recording medium storing a program to cause the method to execute on a computer. The technical problems to be solved are not limited to the aforementioned technical problems, and other technical problems may exist.
[0007] One aspect provides a method for predicting treatment response to an immune checkpoint inhibitor, including determining a density of tumor-infiltrating lymphocytes (TILs) for a pathological image at a first time point using a machine learning model, determining a density of TILs for a pathological image at a second time point using the machine learning model, determining a fold change in TIL density by using the TIL density for the pathological image at the first time point and the TIL density for the pathological image at the second time point, and predicting a treatment response of a cancer patient to the immune checkpoint inhibitor based on the fold change in TIL density.
[0008] Another aspect provides a method for providing information for predicting treatment response to an immune checkpoint inhibitor, including determining a density of TILs for a pathological image at a first time point using a machine learning model, determining a density of TILs for a pathological image at a second time point using the machine learning model; determining a fold change in TIL density by using the TIL density for the pathological image at the first time point and the TIL density for the pathological image at the second time point; and predicting a treatment response of a cancer patient to the immune checkpoint inhibitor based on the fold change in TIL density.
[0009] Yet another aspect includes a computer-readable recording medium storing a program to cause the method to execute on a computer.
[0010] Another aspect provides a computing device including at least one memory, and at least one processor connected to the memory and configured to execute at least one computer-readable program included in the memory, wherein the at least one program includes instructions for: determining a density of TILs for a pathological image at a first time point using a machine learning model, determining a density of TILs for a pathological image at a second time point using the machine learning model, determining a fold change in TIL density by using the TIL density for the pathological image at the first time point and the TIL density for the pathological image at the second time point, and predicting a treatment response of a cancer patient to an immune checkpoint inhibitor based on the fold change in TIL density.
[0011] Another aspect provides a method of treating cancer, including determining a density of TILs for a pathological image at a first time point using a machine learning model,
[0012] determining a density of TILs for a pathological image at a second time point using the machine learning model, determining a fold change in TIL density by using the TIL density for the pathological image at the first time point and the TIL density for the pathological image at the second time point, and predicting a treatment response of a cancer patient to an immune checkpoint inhibitor based on the fold change in TIL density, and administering the immune checkpoint inhibitor to the patient when the predicted treatment response indicates that the cancer patient is likely to respond to the immune checkpoint inhibitor.
[0013] By using a method and apparatus for predicting treatment response to immune checkpoint inhibitors according to one aspect, the responsiveness of a cancer patient to immune checkpoint inhibitors can be predicted more accurately, and a treatment method can be determined.
[0014] FIG. 1 shows a diagram for explaining an example of a computing system that generates an analysis result for a pathological image according to an embodiment.
[0015] FIG. 2 shows a configuration diagram showing an example of a computing system for predicting treatment response to an immune checkpoint inhibitor according to an embodiment.
[0016] FIG. 3 shows a diagram for explaining examples of a processor analyzing a pathological image according to an embodiment.
[0017] FIG. 4 shows a flowchart for explaining an example of a method for predicting treatment response to an immune checkpoint inhibitor according to an embodiment.
[0018] FIG. 5 shows a schematic diagram explaining the operations of: collecting tumor cells from patients with locally advanced rectal cancer (LARC) during surgery before chemoradiotherapy (pre-CRT, B1), after CRT / before nivolumab administration (B2), after three doses of nivolumab (B3), and after five doses of nivolumab; and analyzing the same using an AI-based pathological slide image analyzer.
[0019] FIG. 6 shows a graph showing the density of intratumoral tumor-infiltrating lymphocytes (iTILs) and the fold change in density measured using an AI-based pathological slide image analyzer before CRT and after CRT / before nivolumab administration.
[0020] FIG. 7 shows a graph showing the density of stromal tumor-infiltrating lymphocytes (sTILs) and the fold change in density measured using an AI-based pathological slide image analyzer before CRT and after CRT / before nivolumab administration.
[0021] In embodiments, terms that are widely used in the relevant field were chosen as much as possible, but they may change depending on the intention of a person skilled in the art, precedents, or the emergence of new technologies. Also, in some cases, terms arbitrarily selected by the applicant are used, and in this case, their meanings will be described in detail in the corresponding description sections. Therefore, the terms used herein should be defined based on the meaning of the terms and the overall content of the specification, not just the literal names of the terms.
[0022] When, throughout the specification, the wording "a portion includes a component," this means that it may further include other components unless specifically stated otherwise. Also, terms "unit," "module," "part," etc. used herein refer to units that process at least one function or operation, which can be implemented by hardware or software, or a combination of hardware and software.
[0023] According to an embodiment of this specification, a "module" or "part" may be implemented by a processor and memory. The term "processor" should be broadly interpreted to include general-purpose processors, central processing units (CPUs), graphics processing units (GPUs), microprocessors, digital signal processors (DSPs), controllers, microcontrollers, state machines, etc. In some environments, the "processor" may refer to application-specific integrated circuits (ASICs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), etc.
[0024] The wording "each of a plurality of A" as used herein may refer to each of all components included in the plurality of A, or each of some components included in the plurality of A. For example, "each of plurality of tumor cells" may refer to each of all tumor cells included in the plurality of tumor cells, or each of some tumor cells included in the plurality of tumor cells. Similarly, each of plurality of immune cells may refer to each of all immune cells included in the plurality of immune cells, or each of some immune cells included in the plurality of immune cells.
[0025] The wording "similar" as used herein may include both the meaning of being identical and being similar. For example, two pieces of information being "similar" may indicate that the two pieces of information are identical or similar.
[0026] The wording "instruction" as used herein may refer to a component of a computer program that is a series of commands grouped by function and executed by a processor.
[0027] The following describes embodiments in detail with reference to the attached drawings. However, these examples can be implemented in various different forms and are not limited to the examples described herein.
[0028] FIG. 1 shows a diagram illustrating an example of a computing system 10 that generates an analysis result 30 for a pathological image 20 according to an embodiment.
[0029] Referring to FIG. 1, the computing system 10 may receive a pathological image 20 and generate an analysis result 30 for the pathological image 20. In this regard, the analysis result 30 and / or medical information generated based on the analysis result 30 may be used to predict a treatment response to an immune checkpoint inhibitor for a cancer patient.
[0030] Although FIG. 1 illustrates the computing system 10 as a single computing device, it is not limited thereto, and the computing system 10 may be configured to distributively process information and / or data through plurality of computing devices. Additionally, although a storage system communicable with the computing system 10 is not illustrated in FIG. 1, the computing system 10 may be connected or configured to communicate with one or more storage systems. The computing system 10 may be any computing device utilized to generate an analysis result 30 for the pathological image 20. In this regard, the computing device may refer to any kind of device equipped with computing functions, for example, but not limited to, a notebook, a desktop, a laptop, a server, a cloud system, etc.
[0031] The storage system configured to communicate with the computing system 10 may be a device or cloud system that stores and manages various data associated with pathological image analysis tasks. For efficient data management, the storage system may store and manage various data using a database. In this regard, the various data may include any data associated with pathological image analysis, and for example, the various data may include the pathological image 20 and histological information on the types, locations, and states of cells, tissues, and / or structures included in the pathological image 20. Additionally, the various data may further include clinical information such as the patient's age, menopausal status, clinical T stage (Clinical_T), breast imaging reporting and data system (BIRADS), number of tumors, tumor size, lymph node enlargement (e.g., Node_Enlargement), result information from a tissue test performed to check the status of the estrogen receptor in the cancer tissue (e.g., Biopsy_ER), result information from a tissue test performed to check the status of the progesterone receptor in the cancer tissue (e.g., Biopsy_PR), result information from a tissue test performed to determine whether the tumor is HER2 positive (e.g., Biopsy_HER2), result information evaluating whether the tumor was completely removed after the given cancer treatment surgery (e.g., pCR_final), pathology type, homologous recombination deficiency (HRD), etc.
[0032] The computing system 10 may receive the pathological image 20 obtained from the body tissue of a patient who is the subject of which a treatment response to an immune checkpoint inhibitor is to be predicted. This pathological image 20 may be received through a communicable storage medium (e.g., a hospital system, local / cloud storage system, etc.). The computing system 10 may analyze the pathological image 20 received and generate the analysis result 30 for the pathological image 20. In this regard, the pathological image 20 may include histological information on at least one patch included in the image.
[0033] The term "patch" as used herein may refer to a small area within a pathological image. For example, a patch may include an area corresponding to a semantic object extracted by performing segmentation on a pathological image. In some embodiments, the patch may refer to a combination of pixels associated with histological information generated by analyzing a pathological image.
[0034] The term "histological information" may include characteristics or information about cells, tissues, and / or structures within the body included in a pathological image. In this regard, characteristics of cells may include cytologic features such as a nucleus and a cell membrane. The histological information may refer to histological information on at least one patch included in a pathological image that is inferred through a machine learning model. On the other hand, the histological information may be obtained as a result of an annotator's annotation work. The term 'annotation' means the task of tagging histological information to a data sample, or the tagged information (i.e., notes) itself. The term "annotation" may be used interchangeably with terms such as tagging and labeling in the relevant technical field.
[0035] According to an embodiment, the computing system 10 may extract histological information, which is characteristics of cells, tissues, and / or structures within the target patient's body, by analyzing the pathological image 20. In detail, the computing system 10 may extract histological information on at least one patch included in the pathological image 20 by analyzing (e.g., inferring) the pathological image 20 using a machine learning model. For example, the histological information may include, but is not limited to, information about cells (e.g., tumor cells, lymphocytes, macrophages, dendritic cells, fibroblasts, endothelial cells, etc.) within the patch (e.g., the number of specific cells, information about the tissue in which specific cells are arranged). In this regard, a tumor cell may refer to a cell that ignores the cell growth cycle and continues excessive proliferation, and in particular, a malignant tumor cell that invades (infiltrates) surrounding tissue and spreads and grows (metastasizes) to distant tissues may be referred to as a cancer cell.
[0036] According to an embodiment, the computing system 10 may detect the expression of biomarkers in cells included in the pathological image 20 or extract information associated with the expression of biomarkers from the pathological image 20. For example, the pathological image 20 may include an H&E stained image. Additionally, the computing system 10 may extract information about the expression level of biomarkers of interest from the stained pathological image 20. For example, the computing system 10 may derive the density of tumor-infiltrating lymphocytes (TILs) from the pathological image. Also, the computing system 10 may calculate the fold change in density of TILs by using the density of TILs, and may predict the treatment response of the cancer patient to the immune checkpoint inhibitor based on the fold change of the density of TILs.
[0037] In some embodiments of histological information extracted from a patch by the computing system 10, the histological information may include information about tissues within the patch (e.g., cancer area, cancer epithelial, cancer stroma, normal epithelial, normal stroma, necrosis, fat, background, etc.). As another example of Information extracted from a patch by the computing system 10, the information extracted from the patch may include, but is not limited to, the density of TILs, the density of macrophage (MP), the density of fibroblast (FB), the density of endothelial cell (EC), the density of microsatellite instability (MSI) status, tumor mutational burden (TMB), tumor proportion score (TPS), combined positive score (CPS), etc. The information extracted from the patch is not limited to the above examples and may include any histological information that may be quantified in the patch, such as cell instability, cell cycle, and biological function.
[0038] As described above, the analysis result 30 generated by the computing system 10 and the medical information generated and / or output based on the computing system 10 may be used to predict and / or output the treatment response result to the immune checkpoint inhibitor. Additionally, by utilizing clinical information associated with the pathological slide received from an accessible external system as additional input data, the treatment response prediction result for the cancer patient to the immune checkpoint inhibitor may be inferred and / or output.
[0039] Hereinafter, an embodiment of how the computing system 10 analyzes a pathological image will be described with reference to FIGS. 2 to 4.
[0040] FIG. 2 is a configuration diagram illustrating an example of a computing system for predicting a treatment response to an immune checkpoint inhibitor according to an embodiment.
[0041] The computing system 10 may correspond to a computing system 200. Referring to FIG. 2, the computing system 200 may include a processor 210 and a memory 220. For convenience of explanation, only components related to the present disclosure are illustrated in FIG. 2. Therefore, 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 at least one of a server device and a cloud device, but is not limited thereto. In some embodiments, the computing system 200 may include one or more server devices. In some embodiments, the computing system 200 may include one or more cloud devices. In some embodiments, the computing system 200 may be configured and operated with a combination of a server device and a cloud device. Additionally, it would be obvious to those skilled in the art related to the present disclosure that the processor 210 and memory 220 shown in FIG. 2 may be implemented as separate devices.
[0042] The processor 210 may process instructions of a computer program by performing basic arithmetic, logic, and input / output operations. In this regard, commands may be provided from the memory 220 or an external device (e.g., a server, etc.). In some embodiments, the processor 210 may generally control the operation of other components included in the computing system 200.
[0043] The processor 210 may be implemented as an array of plurality of logic gates, or a combination of a general-purpose microprocessor and a memory storing a program executable by the microprocessor. For example, the processor 210 may include a general-purpose processor, a central processing unit (CPU), a microprocessor, a digital signal processor (DSP), a controller, a microcontroller, a state machine, etc. In some environments, the processor 210 may include application-specific integrated circuits (ASICs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), etc. For example, the processor 210 may refer to a combination of processing devices such as a combination of a digital signal processor (DSP) and a microprocessor, a combination of a plurality of microprocessors, a combination of one or more microprocessors coupled with a digital signal processor (DSP) core, or any other such combinations.
[0044] The processor 210 analyzes the pathological image. Analyzing the pathological image may include segmenting or detecting a specific tissue and / or cell from various tissues and cells expressed on a pathological slide, and then converting the same into information that can assist in making a medical judgment. In this regard, the operation of converting the same into information that can assist in making a medical judgment may refer to extracting features that can classify the pathology of a patient's disease, diagnose cancer, establish a cancer treatment plan, prescribe anticancer drugs, or predict the possibility of occurrence, and in particular, may include predicting a treatment response of a cancer patient to an immune checkpoint inhibitor.
[0045] For example, the processor 210 may derive the density of TILs from the pathological image using at least one machine learning model.
[0046] A tissue sample of a cancer patient may be collected from the cancer patient at any point during the treatment process including chemotherapy and radiation therapy, and the pathological image may be generated from the tissue sample. For example, the tissue sample may be obtained from the patient at least one time point of a time point before chemotherapy and radiation therapy, a time point after chemotherapy and radiation therapy / before immune checkpoint inhibitor administration, a time point after 3 doses of immune checkpoint inhibitor, or a time point of surgery after 5 doses of immune checkpoint inhibitor, and the pathological image may be generated based on the tissue sample.
[0047] Additionally, the processor 210 may calculate the fold change in density of TILs using the density of TILs by using at least one machine learning model.
[0048] As an analysis result, when the fold change in density of TILs is equal to or greater than a preset value, the result of classifying the pathological slide image as a case where the cancer patient responds to the immune checkpoint inhibitor may be output. In detail, the cancer patient responding to the immune checkpoint inhibitor may refer to that the cancer patient has good drug sensitivity and treatment prognosis for the immune checkpoint inhibitor.
[0049] Meanwhile, the processor 210 may derive the proportion of PD-L1 positive tumors from the pathological image using at least one machine learning model. The proportion of PD-L1 positive tumors may be calculated as a tumor proportion score (TPS) or a combined positive score (CPS). The TPS may be defined as the number of PD-L1 positive tumor cells divided by the total number of tumor cells, and the CPS may be defined as the number of PD-L1 positive tumor cells and immune cells divided by the total number of tumor cells.
[0050] As an analysis result, when the proportion of PD-L1 positive tumors is equal to or greater than a preset value, the processor 210 may output the result of classifying the pathological slide image as a case where the cancer patient has good drug sensitivity and treatment prognosis for the immune checkpoint inhibitor.
[0051] The memory 220 may include any non-transitory computer-readable recording medium. For example, the memory 220 may include a non-volatile mass storage device such as a random access memory (RAM), a read only memory (ROM), a disk drive, a solid state drive (SSD), a flash memory, etc. In some embodiments, a non-volatile mass storage device such as a ROM, SSD, a flash memory, a disk drive, etc. may be a separate permanent storage device distinct from the memory. Additionally, an operating system (OS) and at least one program code may be stored in the memory 220.
[0052] These software components may be loaded from a separate computer-readable recording medium distinct from the memory 220. Such a separate computer-readable recording medium may be a recording medium directly connectable to the computing system 200, and may include, for example, a computer-readable recording medium such as a floppy drive, disk, tape, DVD / CD-ROM drive, memory card, etc.
[0053] Although not shown 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 method to transmit and receive data between them. For example, through the display device, a pathological image, a pathological slide image, analysis information of the pathological slide image, medical information, additional information based on the medical information, etc. may be provided to a user.
[0054] FIG. 3 shows a diagram explaining examples of a processor analyzing pathological images according to an embodiment. Referring to FIG. 3 below, examples of how the processor 210 analyses pathological images are described. The embodiments to be described later with reference to FIG. 3 may be applied to deriving the density of TILs, calculating the fold change in density of TILs 330, and / or deriving the proportion of PD-L1 positive tumors which are to be performed by the processor 210.
[0055] Referring to FIG. 3, the processor 210 may analyze the pathological image 310 using a machine learning model 320. For example, the result of analyzing the pathological image 310 may be derived as the fold change in density of TILs 330.
[0056] In this regard, the fold change in density of TILs 330 may be a change rate calculated by comparing the density measured from a first time point pathology slide obtained from a patient before CRT and the density measured from a second time point pathology slide obtained from the patient after CRT.
[0057] For example, the processor 210 may output the detection result in the form of layers representing tissues in the pathological image 310 using the machine learning model 320. In this case, the machine learning model 320 may be trained to detect areas corresponding to tissues in the reference pathology slide images within the pathological image 310 using training data including a plurality of reference pathology slide images and a plurality of reference label information.
[0058] The processor 210 may perform classification on the plurality of tissues represented in the pathological image 310. In detail, the processor 210 may classify the pathological image 310 into at least one of a cancer area, a cancer stroma area, a necrosis area, and a background area.
[0059] However, examples of classifying at least a portion of the area represented in the pathological image 310 by the processor 210 are not limited to these. In other words, without being limited to the four types of areas (cancer area, cancer stroma area, necrosis area, and background area), the processor 210 may classify at least one area represented in the pathological image 310 into a plurality of categories according to various criteria. At least one area represented in the pathological image 310 may be classified into plurality of categories according to predefined criteria or criteria set by the user.
[0060] The processor 210 may perform classification on a plurality of cells represented in the pathological image 310 by analyzing the pathological image 310.
[0061] First, the processor 210 may detect cells from the pathological image 310 by analyzing the pathological image 310, and output the detection result in the form of layers representing the cells.
[0062] The processor 210 may output the detection result in the form of layers representing cells in the pathological image 310 using the machine learning model 320. In this case, the machine learning model 320 may be trained to detect the locations and types of cells in the reference pathology slide images within the pathological image 310 by using training data including a plurality of reference pathology slide images and a plurality of reference label information.
[0063] The processor 210 may perform classification on the plurality of cells represented in the pathological image 310. For example, the processor 210 may classify a plurality of cells represented in the pathological image 310 into at least one of a tumor cell, lymphocytes, macrophages, fibroblasts, endothelial cells, and other cells.
[0064] However, examples of the processor 210 classifying cells represented in the pathological image 310 are not limited thereto. In other words, without being limited to the six types of cells (i.e., tumor cells, lymphocytes, macrophages, fibroblasts, endothelial cells, and other cells), the processor 210 may classify cells represented in the pathological image 310 into plurality of categories according to various criteria. The cells represented in the pathological image 310 may be grouped into plurality of categories according to preset criteria or criteria set by the user.
[0065] In an embodiment, the processor 210 may determine at least one value of the density of iTILs, the density of sTILs, the density of TILs in the tumor microenvironment (TME), or the sTIL score, for the pathological image 310.
[0066] The term "the density of iTIL" as used this specification refers to the amount of immune cells within the tumor. The density of sTILs refers to the density of TILs in the connective tissue or stroma surrounding the tumor. The density of TME TIL refers to the overall density of TILs in the TME, indicating how many immune cells, lymphocytes, are distributed throughout the tumor tissue. As used herein, the density of TME TIL may be used interchangeably with total TIL (tTIL).
[0067] In an embodiment, the processor 210 may detect the cancer area and cancer stroma area in the pathological image using the machine learning model. The density of TME TIL may be determined based on the area of the TME including the cancer area and cancer stroma area, and the number of TILs included in the TME.
[0068] In an embodiment, the processor 210 may divide the pathological image 310 into a grid of a pre-set size. The processor 210 may determine the immune phenotype (IP) of each grid. Without limitation, examples of the immune phenotype may include an immune-inflamed phenotype, an immune-excluded phenotype, and an immune-desert phenotype. The processor 210 may determine the immune phenotype of a grid as the immune-inflamed phenotype when the density of TILs in the cancer area is greater than or equal to a first threshold value. The processor 210 may determine the immune phenotype of a grid as the immune-excluded phenotype when the density of TILs in the cancer area is less than the first threshold value and the density of TILs in the cancer stroma area is greater than or equal to a second threshold value. The processor 210 may determine the immune phenotype of a grid as the immune-desert phenotype when the density of TILs in the cancer area is less than the first threshold value and the density of TILs in the cancer stroma area is less than the second threshold value.
[0069] In an embodiment, the processor 210 may determine a score for the immune-inflamed phenotype of the pathological image 310 based on the immune phenotype of each grid. The processor 210 may output the responsiveness to an immune checkpoint inhibitor based on the score for the immune-inflamed phenotype. The processor 210 may determine that the responsiveness to the immune checkpoint inhibitor is negative when the score for the immune-inflamed phenotype is less than a pre-set cutoff value. The processor 210 may determine that the responsiveness to the immune checkpoint inhibitor is positive when the score for the immune-inflamed phenotype is greater than or equal to the pre-set cutoff value.
[0070] In an embodiment, the processor 210 may derive the proportion of PD-L1 positive tumors from the pathological image at a specific time point. The processor 210 may derive the proportion of PD-L1 positive tumors as a tumor proportion score (TPS) value or a combined positive score (CPS) value. The TPS may be defined as the number of PD-L1 positive tumor cells divided by the total number of tumor cells, and the CPS may be defined as the number of PD-L1 positive tumor cells and immune cells divided by the total number of tumor cells. The processor 210 may determine that the responsiveness to the immune checkpoint inhibitor is positive when the proportion of PD-L1 positive tumor cells among the total tumor cells derived from the pathological image at the specific time point is greater than or equal to a pre-set cutoff value.
[0071] In this regard, the machine learning model 320 refers to a statistical learning algorithm implemented based on the structure of a biological neural network or a structure that executes the algorithm.
[0072] For example, the machine learning model 320 may represent a model that has problem-solving ability by forming a network with nodes that are artificial neurons like synapses in a biological neural network, repeatedly adjusting the weights of the synapses so that the error between the correct output and the inferred output corresponding to a specific input decreases, similar to learning in biological neural networks. For example, the machine learning model 320 may include any probabilistic model, neural network model, etc. used in artificial intelligence learning methods such as deep learning.
[0073] For example, the machine learning model 320 can be implemented as a multilayer perceptron (MLP) consisting of multiple layers of nodes and connections therebetween. The machine learning model 320 according to this embodiment can be implemented using one of various artificial neural network model structures including MLP. For example, the machine learning model 320 may include an input layer that receives input signals or data from the outside, an output layer that outputs output signals or data corresponding to the input data, and at least one hidden layer which is located between the input layer and the output layer and receives signals from the input layer, extracts features, and transfers the result to the output layer. The output layer receives signals or data from the hidden layer and outputs the same to the outside.
[0074] Therefore, the machine learning model 320 may be trained to receive one or more pathological images 310 and extract information about one or more targets (e.g., cells, tissues, structures, etc.) included in the pathological image 310 and / or biomarker expression information.
[0075] FIG. 4 shows a flowchart illustrating one example of a method for analyzing a pathological image according to an embodiment. The method shown in FIG. 4 includes processes processed sequentially in the computing system 10 or 200 or the processor 210 shown in FIGS. 1 and 2. Therefore, even if it is omitted below, the contents described above regarding the computing system 10 or 200 or the processor 210 illustrated in FIGS. 1 and 2 may also be applied to the method illustrated in FIG. 4.
[0076] Referring to FIG. 4, the processor 210 determines the density of TILs for a pathological image at a first time point using a machine learning model in operation 410, determines the density of TILs for a pathological image at a second time point using the machine learning model in operation 420, determines the fold change in density of TILs by using the density of the TILs for the pathological image at the first time point and the density of the TILs for the pathological image at the second time point in operation 430, and predicts the treatment response of a cancer patient to an immune checkpoint inhibitor based on the fold change in density of TILs in operation 440. The treatment response of the cancer patient to the immune checkpoint inhibitor may vary depending on the fold change in density of TILs.
[0077] A more detailed description of the method for predicting the treatment response to the immune checkpoint inhibitor will be provided below.
[0078] One aspect provides a method of predicting a treatment response to an immune checkpoint inhibitor (ICI), including: determining the density of TILs for a pathological image at a first time point using a machine learning model; determining the density of TILs for a pathological image using the machine learning model at a second time point; determining a fold change in density of TILs using the density of the TILs for the pathological image at the first time point and the density of the TILs for the pathological image at the second time point; and predicting the treatment response of a cancer patient to an immune checkpoint inhibitor based on the fold change in density of TILs.
[0079] In an embodiment, the method of predicting the treatment response to the immune checkpoint inhibitor (ICI) may include: deriving the density of TILs from a pathological image; calculating the fold change in density of TILs by using the density of TILs; and predicting the treatment response of a cancer patient to the immune checkpoint inhibitor based on the fold change in density of TILs.
[0080] The term "pathological image" used herein refers to an image obtained by scanning a pathology slide that has undergone a series of chemical processing operations for microscopic observation of tissues, etc. removed from the human body and fixed and stained.
[0081] The pathological image may include a whole slide image (WSI) containing a high-resolution image of the entire slide, and may include at least one of an Hematoxylin & Eosin (H&E) stained slide or an immunohistochemistry (IHC) stained slide. The pathological image may also refer to a portion of the high-resolution whole slide image, for example, one or more patches. Additionally, the pathological image may refer to a digital image obtained by scanning a pathology slide using a digital scanner, and may include information about cells, tissues, and / or structures within the human body. Meanwhile, the pathological image may include one or more patches, and histological information may be applied (e.g., tagged) to one or more patches through annotation work. The term "pathological image" as used herein may be used interchangeably with "pathological image," "pathology slide image," "tissue slide image," "whole slide image (WSI)," etc. Also, the term "pathological image" as used herein may refer to at least a portion of the area included in the pathological image.
[0082] The term "tumor microenvironment (TME)" as used herein may refer to the environment including tissues and cells surrounding a tumor. For example, the TME may include not only the tumor cells themselves, but also the surrounding blood vessels of the tumor cells, extracellular stroma (surrounding tissue), immune cells, inflammatory mediators, etc., but is not limited thereto.
[0083] The term "tumor-infiltrating lymphocyte (TIL)" refers to lymphocytes that have gathered inside or around tumor cells, which have the specificity to attack tumors than peripheral blood mononuclear cells (PBMCs) present more effectively in the blood. In an embodiment, TILs may be iTILs and / or sTILs. In detail, TILs may include at least one of iTILs or sTILs.
[0084] The term "tumor cell (TC)" as used herein may refer to a cell that ignores the cell growth cycle and continues excessive proliferation. In particular, a malignant tumor cell that invades (infiltrates) surrounding tissue and spreads and grows (metastasizes) to distant tissues may be referred to as a cancer cell.
[0085] The term "immune checkpoint inhibitor" as used herein may refer to a substance that inhibits the mechanism of cancer cells that interferes with the activation of T cells. The immune checkpoint inhibitor may be any substance that can inhibit the function of immune checkpoint proteins, and may be a protein, a compound, a natural product, DNA, RNA, a peptide, etc., or an antibody, or a monoclonal antibody, and or a human antibody, a humanized antibody, or a chimeric antibody.
[0086] In an embodiment, the immune checkpoint inhibitor may include nivolumab. For example, the immune checkpoint inhibitor may be a protein, compound, natural product, DNA, RNA, peptide, or a combination thereof that may inhibit any one function from the group consisting of PD-L1, PD-1, CTLA4, PD-L2, LTF2, LAG3, A2aR, TIGIT, TIM-3, B7-H3, B7-H4, VISTA, CD47, BTLA, KIR, and IDO, but is not limited thereto. In detail, the immune checkpoint inhibitor may be any one selected from the group consisting of anti-PD-L1 antibody, anti-PD-1 antibody, anti-CTLA4 antibody, anti-PD-L2 antibody, LTF2 regulatory 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 combinations thereof, but is not limited thereto.
[0087] The term "treatment response of a cancer patient" as used herein may include pathological complete response (pCR), responsiveness of the cancer patient to the immune checkpoint inhibitor, etc. For example, "responding to the immune checkpoint inhibitor" may refer to "having high drug sensitivity to the immune checkpoint inhibitor" or "having a good treatment prognosis with administration of the immune checkpoint inhibitor."
[0088] The term "pathological complete response (pCR)" as used herein refers to a state in which a tumor mass is present but no cancer cells are found at all through tissue examination after surgery. For example, the pathological complete response may refer to the absence of invasive cancer in the body tissue according to chemotherapy or radiotherapy, but is not limited thereto. In detail, the pathological complete response may refer to a state in which all or at least a portion of the tumor cells present in the body tissue have been removed as a result of anticancer treatment, and generally, the better the pCR, the longer the patient's survival period can be.
[0089] The term "biomarker" as used herein may refer to a marker that can objectively measure normal or pathological conditions, degree of response to drugs, etc.
[0090] The term "machine learning model" as used herein may refer to the structure of a computer algorithm that learns from data to find patterns and make predictions or decisions. The machine learning model generally performs learning on training data, and may then perform prediction or classification on new data. For example, the machine learning model may include any model used to infer answers to a given input.
[0091] In an embodiment, the machine learning model may include an artificial neural network model including an input layer, multiple hidden layers, and an output layer. In this regard, each layer may include one or more nodes. For example, the machine learning model may be trained to infer histological information about pathological images and / or at least one patch included in the pathological images. In this case, histological information generated through annotation work may be used to train the machine learning model. In an embodiment, the machine learning model may be trained to infer responsiveness to cancer patient treatment based on an interaction score, characteristics about at least one of cells, tissues or structures in a pathological image, and / or clinical information about the patient. In addition, the machine learning model may include weights associated with a plurality of nodes included in the machine learning model. The weights may include any parameters associated with the machine learning model.
[0092] The machine learning model may refer to an artificial neural network model, and the artificial neural network model may refer to a machine learning model. The machine learning model according to this specification may be a model trained using various learning methods. For example, various learning methods such as supervised learning, unsupervised learning, reinforcement learning, etc. may be used in this disclosure, but are not limited thereto.
[0093] The term "learning" as used herein may refer to any operation of changing 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 an operation of changing or updating weights associated with a machine learning model through one or more forward propagations and backward propagations using at least one patch and histological information.
[0094] In an embodiment, the method of predicting a treatment response to an immune checkpoint inhibitor (ICI) may further include, using a machine learning model, detecting a cancer region and a cancer stroma region in a pathological image, wherein the density of TILs may be determined based on the area of the TME including the cancer region and the cancer stroma region and the number of TILs included in the TME.
[0095] In an embodiment, the method of predicting a treatment response to an immune checkpoint inhibitor (ICI), the pathological image at the first time point may include a whole slide image (WSI) of the cancer patient obtained before CRT, and the pathological image at the second time point may include a WSI of the cancer patient obtained after CRT and before administration of the immune checkpoint inhibitor.
[0096] In an embodiment, the operation of predicting the treatment response of the cancer patient to the immune checkpoint inhibitor may include determining that the cancer patient responds to the immune checkpoint inhibitor when the fold change in density of TILs is 1. 8 or more. In detail, the response of the patient with the cancer to the immune checkpoint inhibitor may indicate that the patient has a good drug sensitivity to the immune checkpoint inhibitor and a good prognosis for treatment.
[0097] The term "good treatment prognosis" as used herein may refer to the case where the cancer patient's chance of survival is high after diagnosis of cancer, treatment and follow-up observation. In detail, the term may indicate the absence or reduction of invasive cancer in the patient's body tissue according to cancer treatment such as chemotherapy, radiotherapy, administration of an immune checkpoint inhibitor, CRT, etc., and may indicate the case where the likelihood of tumor recurrence is low or absent.
[0098] In an embodiment, the method may further include: deriving a ratio of PD-L1 positive tumor cells to total tumor cells from the pathological image at the first time point; and predicting a treatment response to the immune checkpoint inhibitor using the ratio of PD-L1 positive tumor cells.
[0099] In an embodiment, the ratio of PD-L1 positive tumor cells may be calculated as a TPS value or a CPS value.
[0100] The term "tumor proportion score (TPS)" as used herein is a useful indicator for predicting the effect of a PD-1 inhibitor, and is defined as the number of PD-L1 positive tumor cells divided by the total number of tumor cells. The TPS may reflect the expression level and distribution of PD-L1.
[0101] The term "CPS" as used herein is a useful indicator for predicting the effect of a PD-L1 inhibitor, and is defined as the number of PD-L1 positive tumor cells and immune cells divided by the total number of tumor cells. The combined positive score (CPS) may consider the number of PD-L1 positive immune cells.
[0102] In an embodiment, the predicting operation may include determining that the cancer patient responds to the immune checkpoint inhibitor when the ratio of PD-L1 positive tumor cells to total tumor cells is greater than 1%. In detail, a treatment response of the cancer patient to an immune checkpoint inhibitor may indicate a determination that the cancer patient's drug sensitivity to the immune checkpoint inhibitor and treatment prognosis are considered good.
[0103] The term "cancer" as used herein typically refers to a physiological condition in an animal characterized by abnormal or uncontrolled cell growth. Cancer may be associated with, for example, metastasis, interference with normally functioning surrounding cells, secretion of cytokines or other secretory products at abnormal levels, suppression or enhancement of inflammatory or immune responses, neoplasia, premalignancy, malignancy, or invasion of surrounding or distant tissues or organs such as lymph node invasion.
[0104] In an embodiment, the cancer may be a solid tumor. In detail, the solid tumor may be any one or more selected from the group consisting of lung cancer, skin cancer, stomach cancer, gastrointestinal cancer, intestinal cancer, colorectal cancer, colon cancer, pancreatic cancer, liver cancer, thyroid cancer, uterine cancer, cervical cancer, ovarian cancer, testicular cancer, prostate cancer, breast cancer, and oral cancer, but is not limited thereto.
[0105] In an embodiment, the cancer may have microsatellite stability (MSS) or microsatellite instability (MSI-High).
[0106] The term "microsatellite" as used herein may refer to a state in which short DNA sequences of 1-10 bases are repeatedly arranged across the entire gene, with a tendency to vary between subjects.
[0107] The term "microsatellite instability (MSI)" as used herein may refer to a phenomenon in which mutations occurring in microsatellites are not repaired due to a defect in the mismatch repair (MMR) protein gene, resulting in a change in the number of repeats, and the length differs from that of normal cells. Microsatellite instability is characterized by a high mutation rate and the generation of frameshift-peptide neoantigens, increasing the number of lymphocytes infiltrating the tumor and the surrounding tumor to create a highly immunogenic environment.
[0108] The term "mismatch repair (MMR)" as used herein may refer to the function of specific proteins recognizing and repairing mismatches that occur due to DNA replication errors.
[0109] The term "mismatch repair deficient (dMMR)" as used herein may refer to a state where the mismatch repair (MMR) function does not work properly, causing continuous errors in the replicated DNA sequence. Mismatch repair deficiency (dMMR) occurs due to germ line or somatic mutations that inactivate the mismatch repair protein (MMR protein) or promoter hypermethylation. Mismatch repair deficiency (dMMR) causes numerous mutations, leading to carcinogenesis, while also increasing tumor antigen expression, resulting in an active immune response, and can be used as a marker to predict response to immunotherapy.
[0110] An aspect provides a method for treating cancer, including: determining the density of TILs in a pathological image at a first time point, by using a machine learning model; determining the density of TILs in a pathological image at a second time point, by using the machine learning model; determining the fold change in TILs density by using the density of TILs for the first pathological image at the first time point and the density of TILs for the second pathological image at the second time point; predicting the treatment response of a cancer patient to an immune checkpoint inhibitor based on the fold change in the TIL density; and administering, based on a treatment response prediction result, the immune checkpoint inhibitor to the patient in the case where the cancer patient responds to the immune checkpoint inhibitor.
[0111] The invention will be described in more detail through the following examples. However, these examples are provided for illustrative purposes only and the scope of the invention is not limited to these examples.
[0112]
[0113] Example 1. Development of an AI-powered Pathology Slide Image Analyzer
[0114] The computing system 10 or 200 may correspond to an AI-powered pathology slide image analyzer. To analyze the treatment response to immune checkpoint inhibitors in solid cancers based on AI, an AI-powered pathology slide image analyzer was used. The AI-powered pathology slide image analyzer may perform tissue segmentation into at least one cancer area or cancer stroma. The AI-powered pathology slide image analyzer may acquire information related to cell detection in the pathological images. Based on the obtained information, the AI-powered pathology slide image analyzer may estimate the spatial location of TILs within a TME.
[0115] The AI-powered pathology slide image analyzer was trained and validated by using the datasets shown in Tables 1 and 2 below.
[0116] LymphocytesTumor cellsTraining466,0001,640,000Tuning150,000568,000
[0117] Area (mm2)Cancer AreaCancer StromaBackground non-tumor tissueTraining7,8615,69318,505Tuning600421983
[0118]
[0119] Example 2. Analysis Subjects and Methods
[0120] 2-1. Analysis Subjects
[0121] Using the AI-powered pathology slide image analyzer, tumor samples from the VOLTAGE study, which is a multi-center phase 1 / 2 study, evaluating the efficacy of 5 doses of nivolumab following CRT and surgery in patients with locally advanced rectal cancer (LARC), were analyzed.
[0122] In detail, a total of 86 hematoxylin and eosin (H&E) stained whole slide images (WSIs) collected from all patients enrolled in the VOLTAGE study at pre- CRT and post-CRT and pre-nivolumab time points, were analyzed. In this analysis, the density of TILs in TME was quantified using the AI-powered pathology slide image analyzer from Example 1 based on 16,443 WSIs to which pathologist annotations were added.
[0123] 2-2. Analysis Methods
[0124] From locally advanced rectal cancer (LARC) patients, tumor cells were collected during surgery at pre-CRT (B1), post-CRT / pre-nivolumab (B2), after 3 doses of nivolumab (B3), and after 5 doses of nivolumab, and the tumor cells were analyzed using the AI-powered pathology slide image analyzer from Example 1. Analysis results are schematically shown in FIG. 5.
[0125] FIG. 5 shows a schematic diagram to explain the operation of collecting tumor cells from LARC patients during surgery at pre-CRT (B1), post-CRT / pre-nivolumab (B2), after 3 doses of nivolumab (B3), and after 5 doses of nivolumab, and analyzing the same using the AI pathology slide image analyzer.
[0126] For the AI-powered pathology slide image analysis, a total of 86 H&E whole slide images (WSIs) collected from all patients enrolled in the VOLTAGE study at the pre-CRT (B1) and post-CRT / pre-nivolumab (B2) time points were used.
[0127]
[0128] Example 3. Analysis of Clinical Results Prior to CRT
[0129] 3-1. Correlation between pCR Rate and Proportion of PD-L1 Positive Tumors
[0130] From the analysis subjects in Example 2, the correlation between the pCR rate and the proportion of PD-L1 positive tumors prior to CRT was analyzed.
[0131] In detail, the correlation between PD-L1, which shows enhanced immune activity in both tumor and immune cells, and the pCR rate was analyzed. To compare the expression level of PD-L1 within the tumor, a tumor proportion score (TPS) was defined and TPS of microsatellite stable (MSS) patients and microsatellite instability-high (MSI-H) patients was calculated. The TPS is expressed by the following Mathematical Formula 1:
[0132] Mathematical Formula 1
[0133] TPS = Number of PD-L1 positive tumor cells / Number of all viable tumor cells x 100
[0134] The results of analyzing the correlation between the pCR rate and the proportion of PD-L1 positive tumors in MSS patients and MSI-H patients are shown in Table 3.
[0135] pCR rate in MSS patients(N= 38)pCR rate in MSI-H patients(N= 5)PD-L1 (TPS) ≥1%75% (6 / 8)-PD-L1 (TPS) <1%17% (5 / 30)60% (3 / 5)
[0136] As shown in Table 3, when the proportion of PD-L1 positive tumors was 1% or more (TPS ≥ 1%), the pCR rate was 75% (6 / 8) in MSS patients. On the other hand, when the proportion of PD-L1 positive tumors was less than 1%, the pCR rate was 17% (5 / 30) in MSS patients and 60% (3 / 5) in MSI-H patients.
[0137] This indicates that MMS patients of which TPS is equal to or greater than 1% prior to CRT have good drug sensitivity to immune checkpoint inhibitors and a good prognosis for anticancer treatment.
[0138] 3-2. Correlation between pCR Rate and CD8 / eTreg Ratio
[0139] From the analysis subjects in Example 2, the correlation between the pCR rate prior to CRT and the CD8 / eTreg ratio in TILs was analyzed.
[0140] In detail, the results of analyzing the correlation between the pCR rate and the CD8+ T cell / effector regulatory T cell ratio (CD8 / eTreg ratio) in microsatellite stable (MSS) patients and microsatellite instability-high (MSI-H) patients are shown in Table 4.
[0141] pCR rate in MSS patients(N= 24)pCR rate in MSI-H patients(N= 5)CD8 / eTreg ratio before CRT ≥ 2.578% (7 / 9)67% (2 / 3)CD8 / eTreg ratio before CRT < 2.513% (2 / 15)50% (1 / 2)
[0142] As shown in Table 4, when the CD8 / eTreg ratio was 2.5 or higher, the pCR rate was 78% (7 / 9) in MSS patients and 67% (2 / 3) in MSI-H patients. On the other hand, when the CD8 / eTreg ratio was less than 2.5, the pCR rate was 13% (2 / 15) in MSS patients and 50% (1 / 2) in MSI-H patients.
[0143] These results indicate that MSS patients with a pre-CRT CD8 / eTreg ratio of 2.5 or higher have good drug sensitivity to immune checkpoint inhibitors and a good prognosis for anticancer treatment.
[0144] 3-3. Correlation between pCR Rate, Proportion of PD-L1 Positive Tumors, and CD8 / eTreg Ratio in MSS Patients
[0145] From the analysis subjects in Example 2, the correlation between the pCR rate prior to CRT and the CD8+ T cell / effector regulatory T cell ratio (CD8 / eTreg ratio) in TILs was analyzed.
[0146] In detail, the results of the analysis of the pathological complete remission (pCR) rate of the tumor proportion score (TPS) of Example 3-1 and the CD8 / eTreg ratio of Example 3-2 in MMS patients are shown in Table 5.
[0147] CD8 / eTreg > 2.5CD8 / eTreg < 2.5TotalPD-L1 (TPS) ≥ 1%100% (5 / 5)33% (1 / 3)75% (6 / 8)PD-L1 (TPS) < 1%50% (2 / 4)8% (1 / 12)19% (3 / 16)
[0148] As shown in Table 5, when the proportion of PD-L1 positive tumors was 1% or more and the CD8 / eTreg ratio was 2.5 or higher, the pCR rate was 100% (5 / 5). On the other hand, when the proportion of PD-L1 positive tumors was less than 1% and the CD8 / eTreg ratio was 2.5 or higher, the pCR rate was 50% (2 / 4). When the proportion of PD-L1 positive tumors was 1% or more and the CD8 / eTreg ratio was less than 2.5, the pCR rate was 75% (6 / 8). When the proportion of PD-L1 positive tumors was less than 1% and the CD8 / eTreg ratio was less than 2.5, the pCR rate was 19% (3 / 16).
[0149] These results indicate that when the proportion of PD-L1 positive tumors before CRT is 1% or more and a CD8 / eTreg ratio of 2.5 or higher, MMS patients show good drug sensitivity to immune checkpoint inhibitors and good prognosis for anti-cancer treatment.
[0150] 3-4. Correlation between pCR Rate and Changes in Tumor-Infiltrating Lymphocyte (TIL) Density within TME
[0151] From the analysis subjects in Example 2, the correlation between the pCR rate prior to CRT and changes in TIL density within the TME was analyzed.
[0152] In detail, the TIL density within the TME was measured at pre-CRT (B1) and post-CRT / pre-nivolumab (B2), and the change in density was calculated. The calculated values of the changes in TIL density within the TME are shown in Table 6.
[0153] AllpCR rate in MSSpCR rate in MSI-HtTIL on B2 / B1 ≥ 1.8x75.0% (6 / 8)33.3% (1 / 3)tTIL on B2 / B1 < 1.8x16.7% (5 / 30)100% (1 / 1)
[0154] As shown in Table 6, when the TIL density within TME at post-CRT / pre-nivolumab (B2) was 1.8 times or greater compared to the TIL density at pre-CRT (B1), the pCR rate was 75% (6 / 8) in MSS patients and 33.3% (1 / 3) in MSI-H patients. On the other hand, in the case where the TME TIL density at post-CRT / pre-nivolumab (B2) was less than 1.8 times of the TME TIL density at pre-CRT (B1), the pCR rate was 16.7% (5 / 30) in MSS patients and 100% (1 / 1) in MSI-H patients.
[0155] These results indicate that with respect to MSS patients of which the TME TIL density of post-CRT / pre-nivolumab (B2) was 1.8 times or greater compared to the TME TIL density at pre-CRT (B1), the drug sensitivity to immune checkpoint inhibitors and prognosis for anti-cancer treatment are favorable.
[0156] 3-5. Univariate Analysis of PCR Rate and Each Biomarker
[0157] The pCR rates for each biomarker measured in Examples 3-1 to 3-4 are summarized and results thereof are shown in Table 7. In detail, the pCR rates were calculated according to MSI status, TMB status, TPS, CPS, TIL density within the TME, and changes in TIL density. Results thereof are shown in Table 7. TPS is defined by Mathematical Formula 1, and combined positive score (CPS) is defined by Mathematical Formula 2:
[0158] Mathematical Formula 2
[0159] CPS = Number of PD-L1 positive (tumor cells + lymphocytes + macrophages) / Total number of viable tumor cells x 100
[0160] UnivariateNpCR rateMSI statusMSS3828.9% (11 / 38}MSI-H560% (3 / 5)TMB status< 102133.3% (7 / 21}≥ 10560% (3 / 5)TPS< 1%3423.5% (8 / 34}≥ 1%966.7% (6 / 9)CPS< 13627.8% (10 / 36}≥ 1757.1% (4 / 7)iTIL on B1< 37.7 / mm22425% (6 / 24}(55% quantile}≥ 37.7 / mm21942.1% (8 / 19)iTIL on B2< 798.8 / mm23625% (9 / 36}(86% quantile)≥798.8 / mm2666.7% (4 / 6)tTIL on B2 / B1< 1.8x3016.7% (5 / 30}in MSS≥ 1.8x875.0% (6 / 8)tTIL on B2 / B1< 1.8x3119.4% (6 / 31}in MSS+MSI-H≥ 1.8x1163.6% (7 / 11)
[0161] As shown in Table 7, the pCR rates were: 28.9% (11 / 38) in MSS patients and 60% (3 / 5) in MSI-H patients; 33.3% (7 / 21) when TMB is less than 10 and 60% (3 / 5) when TMB is equal to or greater than 10; 23.5% (8 / 34) when TPS is less than 1% and 66.7% (6 / 9) when TPS is equal to or greater than 1%.
[0162] Also, the pCR rates were: 27.8% (10 / 36) when CPS is less than 1 and 57.1% (4 / 7) when CPS is equal to or greater than 1;25% (6 / 24) when pre-CRT (B1) iTIL is less than 37.7 / mm2and 42.1% (8 / 19) when pre-CRT iTIL is equal to or greater than 37.7 / mm2; with respect to MSS patients, 16.7% (5 / 30) when the fold change between the density of pre-CRT (B1) TME TIL and the density of post-CRT / pre-nivolumab (B2) TME TIL (tTIL on B2 / B1) is less than 1.8x and 75.0% (6 / 8) when tTIL on B2 / B1 is equal to or greater than 1.8x; and with respect to MSS and MSI-H patients combined, 19.4% (6 / 31) when tTIL on B2 / B1 is less than 1.8x and 63.6% (7 / 11) when tTIL on B2 / B1 is equal to or greater than 1.8x.
[0163]
[0164] Example 4 Analysis of Fold Change in TIL Density
[0165] 4-1. Analysis of Fold Change in ITIL Density between Pre-CRT and Post-CRT / Pre-Nivolumab
[0166] The fold change in iTIL density between the pre-CRT and post-CRT / pre-nivolumab time points was analyzed by using the AI-powered pathology slide image analyzer from Example 1.
[0167] In detail, the iTIL density and fold change in iTIL density were measured in all samples (n=43), samples showing pCR (n=14), and samples not showing pCR (n=29). Results thereof are shown in FIG. 6.
[0168] FIG. 6 shows a graph showing the measurement of the iTIL density and density fold change, using the AI-powered pathology slide image analyzer at pre-CRT and at post-CRT / pre-nivolumab. This shows that while the fold change in iTIL density is increased in both patients that show pCR and does not show pCR between pre-CRT and post-CRT / pre-nivolumab, the degree of increase was greater in the patients showing pCR.
[0169] 4-2. Analysis of Fold Change in sTIL Density between Pre-CRT and Post-CRT / Pre-Nivolumab
[0170] The fold change between the density of the pre-CRT sTIL and the density of post-CRT / pre-nivolumab sTIL was analyzed by using the AI-powered pathology slide image analyzer from Example 1.
[0171] In detail, the sTIL density and fold change in sTIL density were measured in all samples (n=43), samples showing pCR (n=14), and samples not showing pCR (n=29). Results are shown in FIG. 7.
[0172] FIG. 7 shows a graph showing the measurement of the sTIL density and density fold change by using the AI-powered pathology slide image analyzer at pre-CRT and at post-CRT / pre-nivolumab. Unlike iTILs, there was no notable increase in sTIL density in the pCR samples, and in fact, the sTIL density was decreased in the non-pCR samples.
[0173]
[0174] Example 5Correlation among pCR Rate and tTIL Density Fold Change, TPS, and CPS in MSS-LARC Patients
[0175] The correlation among the pCR rate and the fold change in TME TIL density, TPS, and CPS in MSS-LARC patients was analyzed.
[0176] Table 8 shows the pCR rate according to the fold change between the TME TIL density of pre-CRT (B1) and the TME TIL density of post-CRT / pre-nivolumab (B2), the pCR rate according to pre-CRT (B1) MSI status, the pCR rate according to TPS values, and the pCR rate according to CPS values, of MSS-LARC patients.
[0177] VariableNpCR rate (%)MSI statusMSS3828.9% (11 / 38)TPS≥ 1%875% (6 / 8)< 1%3017% (5 / 30)CPS≥ 12832% (9 / 28)< 11020% (2 / 10)tTIL increased≥ 1.8x875% (6 / 8)< 1.8x3017% (5 / 30)
[0178] On the other hand, Table 9 shows the pCR rate measured by considering the fold change between the TME TIL density at pre-CRT (B1) and the TME TIL density at post-CRT / pre-nivolumab (B2) and the pre-CRT TPS value, of MSS-LARC patients.
[0179] VariabletTIL increased≥ 1.8x< 1.8xTPS≥ 1%100%(2 / 2)50%(3 / 6)< 1%67%(4 / 6)8%(2 / 24)
[0180] As shown in Tables 8 and 9, among patients whose the fold change in TME TIL was 1.8 times or more, the pCR rate was 100% for patients with TPS greater than or equal to 1%, and 67% for patients with TPS less than 1%. On the other hand, among patients whose the fold change in TME TIL was less than 1.8 times, the pCR rate was 50% for patients with TPS greater than or equal to 1%, and 8% for patients with TPS less than 1%.
[0181] This indicates that among patients whose the fold change between the density of pre-CRT TME TIL and the density of post-CRT / pre-nivolumab TME TIL was 1.8 times or more, patients with TPS greater than 1% have better drug sensitivity to immune checkpoint inhibitors and better prognosis for anti-cancer treatment.
[0182]
[0183] Example 6. Quantification and Statistical Analysis
[0184] The F1 score and IoU metric were used to evaluate the performance of cell and tissue models, respectively. Agreement or Cohen's kappa value was used to evaluate the agreement between the pathologist's interpretation and the AI analyzer's interpretation. The chi-squared test or Fisher's exact test was used for group comparisons of categorical variables. The Student's t-test or nonparametric Mann-Whitney U test was used for group comparisons of continuous variables.
[0185] All statistical analyses were performed using Python (version 3.7) and R software (version 4.2.3) (R Foundation for Statistical Computing, Vienna, Austria).
Claims
1.A method of predicting treatment response to an immune checkpoint inhibitor (ICI), the method comprising:determining a density of tumor-infiltrating lymphocytes (TILs) for a pathological image at a first time point using a machine learning model;determining a density of tumor-infiltrating lymphocytes for a pathological image at a second time point using the machine learning model;determining a fold change in density of tumor-infiltrating lymphocytes by using the density of tumor-infiltrating lymphocytes for the pathological image at the first time point and the density of tumor-infiltrating lymphocytes for the pathological image at the second time point; andpredicting a treatment response to the immune checkpoint inhibitor in a cancer patient based on the fold change in density of the tumor-infiltrating lymphocytes.2.The method of claim 1, further comprisingdetecting a cancer area and a cancer stroma area in a pathological image using the machine learning model, whereinthe density of the tumor-infiltrating lymphocytes is determined based on an area of a tumor microenvironment including the cancer area and the cancer stroma area, and the number of tumor-infiltrating lymphocytes included in the tumor microenvironment.3.The method of claim 1, whereinthe pathological image at the first time point includes a whole slide image (WSI) of the cancer patient obtained before chemoradiotherapy (CRT), andthe pathological image at the second time point includes a WSI of the cancer patient obtained after chemoradiotherapy and before administration of the immune checkpoint inhibitor.4.The method of claim 1, wherein the tumor-infiltrating lymphocytes include at least one of intratumoral tumor-infiltrating lymphocytes and stromal tumor-infiltrating lymphocytes.5.The method of claim 1, wherein the immune checkpoint inhibitor includes nivolumab.6.The method of claim 1, wherein the predicting of the treatment response of the cancer patient to the immune checkpoint inhibitor includes determining that the cancer patient responds to the immune checkpoint inhibitor when the fold change in density of tumor-infiltrating lymphocytes is 1. 8 or more.7.The method of claim 1, further comprising: deriving a ratio of PD-L1 positive tumor cells to total tumor cells from the pathological image at the first time point; andpredicting a treatment response to the immune checkpoint inhibitor using the ratio of PD-L1 positive tumor cells.8.The method of claim 7, wherein the ratio of PD-L1 positive tumor cells is calculated as a tumor proportion score (TPS) value or a combined positive score (CPS) value.9.The method of claim 7, wherein the predicting of the treatment response to the immune checkpoint inhibitor using the ratio of PD-L1 positive tumor cells includes determining that the tumor cells respond to the immune checkpoint inhibitor when the ratio of PD-L1 positive tumor cells to total tumor cells is greater than 1%.10.The method of claim 1, wherein the cancer is a solid tumor.11.The method of claim 10, wherein the cancer have microsatellite stability (MSS) or microsatellite instability (MSI-High).12.The method of claim 10, wherein the solid tumor is any 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.13.A computer program stored on a computer-readable recording medium for causing the method of claim 1 to execute on a computer.14.A computing system comprising:at least one memory; andat least one processor connected to the at least one memory and configured to execute at least one computer-readable program included in the at least one memory, whereinthe at least one processor is configured todetermine a density of tumor-infiltrating lymphocytes (TILs) for a pathological image at a first time point using a machine learning model;determine a density of tumor-infiltrating lymphocytes for a pathological image at a second time point using the machine learning model;determine a fold change in density of tumor-infiltrating lymphocytes by using the density of tumor-infiltrating lymphocytes for the pathological image at the first time point and the density of tumor-infiltrating lymphocytes for the pathological image at the second time point; andpredict a treatment response to the immune checkpoint inhibitor in a cancer patient based on the fold change in density of the tumor-infiltrating lymphocytes.15.The computing system of claim 14, whereinthe pathological image at the first time point includes a whole slide image (WSI) of the cancer patient obtained before chemoradiotherapy (CRT), andthe pathological image at the second time point includes a WSI of the cancer patient obtained after chemoradiotherapy and before administration of the immune checkpoint inhibitor.16.The computing system of claim 14, wherein the at least one processor is configured todetermine that the cancer patient responds to the immune checkpoint inhibitor when the fold change in density of tumor-infiltrating lymphocytes is 1. 8 or more.17.A cancer treatment method comprising:determining a density of tumor-infiltrating lymphocytes (TILs) for a pathological image at a first time point using a machine learning model;determining a density of tumor-infiltrating lymphocytes for a pathological image at a second time point using the machine learning model;determining a fold change in density of tumor-infiltrating lymphocytes by using the density of tumor-infiltrating lymphocytes for the pathological image at the first time point and the density of tumor-infiltrating lymphocytes for the pathological image at the second time point;predicting a treatment response to the immune checkpoint inhibitor in a cancer patient based on the fold change in density of the tumor-infiltrating lymphocytes; andadministering the immune checkpoint inhibitor to the cancer patient when it is determined that the cancer patient is responsive to the immune checkpoint inhibitor based on treatment response prediction results.
Citation Information
Patent Citations
Put-on type mounted work platform of main tower for bridge, and mounting method of the work platform
KR102537075B1
System and methods for quantitatively describing biophysical markers
US20170172450A1
A Radiomics-Based Imaging Tool to Monitor Tumor-Lymphocyte Infiltration and Outcome in Cancer Patients Treated by Anti-PD-1 / PD-L1
US20210003555A1
Method and system for predicting response to immune anticancer drugs
US20220036971A1
Methods and systems for evaluation of immune cell infiltrate in tumor samples
US20220270242A1
Cited By
Neural network prediction method for intestinal cancer immune response map, medium and equipment
CN120878186A
