Method and device for outputting medical information about subject by using machine-learning model

Abstract

A computing device according to one aspect comprises: at least one memory for storing at least one instruction; and at least one processor operating according to the at least one instruction. The at least one processor may: predict a probability score indicating the possibility of mutation of a specific gene by analyzing a pathological slide image of a subject by using a machine learning model; and output information related to a targeted therapy for the subject on the basis of the probability score.

Description

Method and apparatus for outputting medical information about a subject using a machine learning model

[0001] The present disclosure relates to a method and apparatus for outputting medical information about a subject using a machine learning model. More specifically, the present disclosure relates to an apparatus and method for predicting responsiveness to targeted therapy using a machine learning model that predicts the likelihood of mutation of a specific gene.

[0002] To make clinical decisions regarding targeted therapy, cancer patients (e.g., patients with non-small-cell lung cancer (NSCLC)) may undergo molecular or genetic tests to identify major driver mutations.

[0003] Previously, deep learning-based predictive models have been developed for patients who have not yet undergone testing to identify major genetic mutations (e.g., EGFR (epidermal growth factor receptor) mutations). For example, conventional predictive models are utilized as pre-screening tools to identify patients likely to have driver mutations by analyzing pathology slide images. In other words, conventional predictive models are used to screen patients who require additional confirmatory testing or to filter out patients who do not require additional molecular testing.

[0004] However, conventional predictive models are focused on pre-screening to predict the presence or absence of mutations, so they are not used to predict responsiveness to targeted treatment in patients who have already been confirmed to have a specific mutation (e.g., EGFR mutation).

[0005] The present disclosure aims to provide an apparatus and method capable of more precisely predicting responsiveness to a targeted therapy for patients who have already been confirmed to have a mutation in a specific gene (e.g., EGFR) and have become candidates for targeted therapy.

[0006] In addition, the present disclosure aims to provide an apparatus and method for utilizing the output (e.g., probability score) of a machine learning model trained to predict the mutation status of a gene as a biomarker for predicting a patient's responsiveness to treatment.

[0007] The technical challenges to be solved are not limited to those mentioned above, and other technical challenges may exist.

[0008] A computing device according to one aspect comprises: at least one memory in which at least one instruction is stored; and at least one processor that operates according to the at least one instruction. The at least one processor can predict a probability score indicating the possibility of a mutation of a specific gene by analyzing a pathology slide image of a subject using a machine learning model, and output information related to targeted treatment for the subject based on the probability score.

[0009] A method for outputting medical information about a subject according to another aspect may include: receiving a pathology slide image of the subject; predicting a probability score indicating the possibility of a specific gene mutation by analyzing the pathology slide image using a machine learning model; and outputting information related to targeted therapy for the subject based on the probability score.

[0010] A computer-readable recording medium according to another aspect may include a recording medium that records a program for executing the above-described method on a computer.

[0011] FIG. 1 is a diagram illustrating an example of a computing device outputting medical information according to one embodiment.

[0012] FIG. 2a is a configuration diagram illustrating an example of a user terminal according to one embodiment.

[0013] FIG. 2b is a configuration diagram illustrating an example of a server according to one embodiment.

[0014] FIG. 3 is a flowchart illustrating an example of a method for outputting medical information according to one embodiment.

[0015] FIG. 4 is a diagram illustrating an example in which a machine learning model is trained according to one embodiment.

[0016] FIG. 5 is a flowchart illustrating an example in which a processor according to one embodiment outputs information related to targeted treatment based on a probability score.

[0017] FIG. 6 is a diagram illustrating an example in which a processor according to one embodiment classifies a subject into one of a plurality of groups.

[0018] FIG. 7 is a diagram illustrating an example in which a processor according to one embodiment classifies a subject into one of a plurality of groups by considering variables corresponding to tumor heterogeneity.

[0019] FIG. 8 is a diagram illustrating examples of information related to targeted therapy according to one embodiment.

[0020] FIGS. 9 and FIGS. 10 are drawings illustrating an example of an image and information output to a display device according to one embodiment.

[0021] FIG. 11 is a drawing for illustrating an example of a system for outputting medical information according to one embodiment.

[0022] A computing device according to one aspect comprises: at least one memory in which at least one instruction is stored; and at least one processor that operates according to the at least one instruction. The at least one processor can predict a probability score indicating the possibility of a mutation of a specific gene by analyzing a pathology slide image of a subject using a machine learning model, and output information related to targeted treatment for the subject based on the probability score.

[0023] The terms used in the embodiments have been selected to be as close as possible to currently widely used general terms; however, these may vary depending on the intent of those skilled in the art, case law, the emergence of new technologies, etc. Additionally, in specific cases, terms have been selected at the applicant's discretion, and in such cases, their meanings will be described in detail in the relevant description section. Therefore, terms used in the specification must be defined not merely by their names, but based on their meanings and the content throughout the specification.

[0024] When a part of the specification is described as "comprising" a certain component, this means that, unless specifically stated otherwise, it does not exclude other components but may include additional components. Furthermore, terms such as "unit" and "module" as used in the specification refer to a unit that performs at least one function or operation, and this may be implemented in hardware or software, or as a combination of hardware and software.

[0025] Additionally, terms including ordinal numbers, such as "first" or "second," used in the specification may be used to describe various components, but said components shall not be limited by said terms. Such terms may be used for the purpose of distinguishing one component from another.

[0026] In the following, "medical information" may refer to any medically meaningful information or clinical information of a patient that can be extracted from medical images. Medical images may include not only pathology slide images but also radiographic images (X-ray, CT, MRI, PET, etc.). For example, medical information may include at least one of an immune phenotype, genotype, expression type, biomarker, tumor purity, information regarding RNA, tumor microenvironment, cancer regimen expressed in the pathology slide image, survival information, treatment response, treatment outcome, genetic characteristics, and medical records.

[0027] In addition, medical information may also include anatomical structural information extracted from medical images, types of lesions, locations and sizes of lesions, morphological features of lesions (e.g., boundaries, texture, density), functional indicators (e.g., blood flow, metabolic activity), abnormal findings of organs, indicators related to treatment prognosis obtained from medical images, information regarding findings obtained by analyzing medical images using artificial intelligence models, abnormality scores of said findings, reliability of said findings, and image biomarkers (radiomic features).

[0028] In addition, medical information may include findings such as the presence or absence of nodules in the medical image, signs of pneumonia, the presence of pneumothorax, the location and type of fractures, the location, size, shape, and boundary characteristics of masses, the distribution of microcalcifications, asymmetry, and breast tissue density and structural distortion. Such findings may be calculated along with, but are not limited to, an abnormality score or risk score for the relevant image.

[0029] Additionally, medical information may include, but is not limited to, the area, location, and size of specific tissues (e.g., cancer tissue, cancer stromal tissue, etc.) and / or specific cells (e.g., tumor cells, lymphocytes, macrophages, endothelial cells, fibroblasts, etc.) within the medical image, diagnostic information of cancer, information related to the patient's probability of developing cancer, and / or medical conclusions related to cancer treatment.

[0030] In addition, medical information may include not only quantified values ​​obtainable from medical images but also information visualizing the values, predictive information based on the values, image information, statistical information, etc. For example, medical information may be provided to a user terminal or output through a display device.

[0031] Embodiments are described in detail below with reference to the attached drawings. However, embodiments may be implemented in various different forms and are not limited to the examples described herein.

[0032] FIG. 1 is a diagram illustrating an example of a computing device outputting medical information according to one embodiment.

[0033] Referring to FIG. 1, the computing device (20) can analyze a pathology slide image (10) and output medical information (30).

[0034] For example, a computing device (20) may receive a pathology slide image (10) of a subject as input. The computing device (20) may analyze the pathology slide image (10) using a machine learning model and generate medical information (30) based on the analysis results. Here, the medical information (30) may include, but is not limited to, information related to targeted therapy for the subject.

[0035] A machine learning model refers to a statistical learning algorithm implemented based on the structure of a biological neural network, or a structure that executes such an algorithm. For example, a machine learning model may represent a model capable of problem-solving, in which nodes—artificial neurons that form a network through synaptic connections as in biological neural networks—learn by repeatedly adjusting the weights of the synapses to reduce the error between the correct output corresponding to a specific input and the inferred output. For example, a machine learning model may include arbitrary probability models, neural network models, etc., used in artificial intelligence learning methods such as deep learning.

[0036] For example, a machine learning model can be implemented as a multilayer perceptron (MLP) composed of multiple layers of nodes and connections between them. The machine learning model according to the present embodiment can be implemented using one of various artificial neural network model structures including an MLP. For example, the machine learning model may be composed of an input layer that receives an input signal or data from the outside, an output layer that outputs an output signal or data corresponding to the input data, and at least one hidden layer located between the input layer and the output layer, which receives a signal from the input layer, extracts a feature, and transmits it to the output layer. The output layer receives a signal or data from the hidden layer and outputs it to the outside.

[0037] Accordingly, the machine learning model can be trained to extract information about one or more objects (e.g., cells, tissues, structures, etc.) included in the pathology slide image (10).

[0038] Existing genotype predictor (GP) models have been used as tools to identify patients with a high likelihood of specific mutations by analyzing pathology slide images (10). However, existing models are not used to predict the responsiveness to targeted therapy in patients in whom specific mutations have already been confirmed.

[0039] Meanwhile, according to conventional technology, there are limitations to training machine learning models to predict a patient's responsiveness to targeted therapy. Specifically, obtaining pathology slide images paired with treatment outcome data for training a machine learning model is much more difficult than obtaining images paired with mutation status data. Furthermore, since treatment outcomes are influenced by many other factors not present in the tissue (e.g., patient demographics, hospital environment, etc.), the machine learning model training process becomes complex, and the potential for generalization is limited.

[0040] A computing device (20) according to one embodiment can provide a new biomarker that can more precisely predict responsiveness to the targeted therapy for subjects who have already been confirmed to have a mutation in a specific gene (e.g., EGFR) and have become candidates for targeted therapy. In addition, the computing device (20) can generate various medical information (30) using the new biomarker.

[0041] For example, the computing device (20) can predict a probability score indicating the possibility of a specific gene mutation by analyzing the pathology slide image (10) of the subject using a machine learning model. Then, the computing device (20) can output medical information (30) about the subject based on the probability score.

[0042] Additionally, the computing device (20) can utilize the output (e.g., probability score) of a machine learning model trained to predict the mutation state of a specific gene as a biomarker for predicting the treatment response.

[0043] Accordingly, the computing device (20) can predict responsiveness to targeted therapy using pathology slide images (10) for subjects in whom a specific gene mutation has already been confirmed. Thus, the computing device (20) can generate important information for establishing a treatment strategy for the subject without additional complex gene tests.

[0044] Additionally, the computing device (20) can classify subjects based on the predicted results of treatment responsiveness (e.g., good responder or poor responder) and establish a customized treatment plan optimized for the classified group.

[0045] Additionally, the computing device (20) can contribute to improving the prognosis of the subject and reducing unnecessary treatment by outputting medical information (30) for various clinical scenarios (e.g., palliative setting, adjuvant setting, neoadjuvant setting, etc.).

[0046] Hereinafter, with reference to FIGS. 2a to 11, examples are described in which a computing device (20) analyzes a pathology slide image (10) and generates medical information (30).

[0047] For example, the computing device (20) may be a user terminal or a server. In other words, the operations performed by the computing device (20) may be performed by a user terminal or a server. Alternatively, some of the operations performed by the computing device (20) may be performed by a user terminal, and the remainder may be performed by a server.

[0048] A user terminal may be an electronic device comprising a display device and a device for receiving user input (e.g., a keyboard, a mouse, etc.), and including memory and a processor. Additionally, the display device may be implemented as a touch screen to perform the function of receiving user input. For example, the user terminal may include, but is not limited to, notebook PCs, desktop PCs, laptops, tablet computers, smartphones, etc.

[0049] A server may be a device that communicates with external devices (e.g., user terminals). For example, a server may be a device that stores various data, including medical information and information about machine learning models. Alternatively, a server may be an electronic device that includes memory and a processor and possesses its own computing capabilities. For example, a server may be a cloud server or an on-premise server.

[0050] Hereinafter, examples of a user terminal and a server will be described with reference to FIGS. 2a and 2b.

[0051] FIG. 2a is a configuration diagram illustrating an example of a user terminal according to one embodiment.

[0052] Referring to FIG. 2a, the user terminal (100) includes a processor (110), memory (120), an input / output interface (130), and a communication module (140). For convenience of explanation, FIG. 2a only illustrates components related to the present invention. Accordingly, other general-purpose components may be included in the user terminal (100) in addition to the components illustrated in FIG. 2a. Furthermore, it is obvious to those skilled in the art that the processor (110), memory (120), input / output interface (130), and communication module (140) illustrated in FIG. 2a may be implemented as independent devices.

[0053] The processor (110) can process instructions of a computer program by performing basic arithmetic, logic, and input / output operations. Here, instructions may be provided from memory (120) or an external device (e.g., a server (200), etc.). Additionally, the processor (110) can control the overall operation of other components included in the user terminal (100).

[0054] The processor (110) can receive the pathology slide image of the subject.

[0055] The pathology slide image may be a whole slide image or any one of the patches into which the whole slide image is divided. For example, the processor (110) may analyze the whole slide image to classify cells and / or tissues. Or, the processor (110) may analyze the patches to classify cells and / or tissues. Hereinafter, the pathology slide image may refer to the whole slide image or the patches.

[0056] The processor (110) can predict a probability score indicating the possibility of a specific gene mutation by analyzing a pathology slide image using a machine learning model.

[0057] For example, a machine learning model may be trained based on at least one pathology slide image corresponding to a tissue sample obtained from at least one patient in whom a specific gene mutation has been confirmed, or at least one of information regarding the responsiveness to targeted treatment of at least one patient. Thus, the processor (110) can predict the responsiveness to targeted treatment of a subject in whom a specific mutation (e.g., EGFR mutation) has already been confirmed.

[0058] An example of the processor (110) analyzing a pathology slide image and predicting a probability score is described later with reference to steps 310 and 320 of FIG. 3.

[0059] The processor (110) can output information related to targeted treatment for the subject based on probability scores.

[0060] For example, the processor (110) can classify the subject into one of a plurality of groups based on a probability score. And, the processor (110) can output information related to targeted treatment based on the classification result.

[0061] At this time, the processor (110) may classify the subject into one of a plurality of groups by considering at least one additional variable corresponding to the analysis result of the subject's tumor, in addition to the probability score. For example, the processor (110) may classify the subject into one of a plurality of groups by combining the probability score or categorical information based on the probability score with at least one variable. Here, the at least one variable may be expressed as at least one of a continuous score or categorical information.

[0062] For example, information related to targeted therapy may include information regarding at least one of a palliative setting, an adjuvant setting, or a neoadjuvant setting. Specifically, information related to targeted therapy may include at least one of information regarding a treatment method, information regarding a drug dosage, or additional information regarding the treatment regimen.

[0063] Additionally, the processor (110) can output a heatmap image corresponding to a probability score by overlaying it on the pathology slide image. Additionally, the processor (110) can output information about at least one tissue or at least one cell represented on the pathology slide image.

[0064] An example of the processor (110) outputting information related to targeted treatment for a subject or other various medical information is described later with reference to step 330 of FIG. 3.

[0065] The processor (110) may be implemented as an array of multiple logic gates, or as a combination of a general-purpose microprocessor and memory storing a program that can be executed on the microprocessor. For example, the processor (110) 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 (110) may include an application-specific integrated circuit (ASIC), a programmable logic device (PLD), a field programmable gate array (FPGA), etc. For example, the processor (110) may refer to a combination of processing devices such as a combination of a digital signal processor (DSP) and a microprocessor, a combination of multiple microprocessors, a combination of one or more microprocessors combined with a digital signal processor (DSP) core, or any other combination of such configurations.

[0066] The memory (120) may include any non-transient computer-readable recording medium. As an example, the memory (120) may include a permanent mass storage device such as a random access memory (RAM), read-only memory (ROM), disk drive, solid state drive (SSD), or flash memory. As another example, a permanent mass storage device such as a ROM, SSD, flash memory, or disk drive may be a separate permanent storage device distinct from the memory. Additionally, the memory (120) may store an operating system (OS) and at least one program code (e.g., code for the processor (110) to perform an operation described later with reference to FIGS. 3 through 11).

[0067] These software components may be loaded from a computer-readable recording medium separate from the memory (120). This separate computer-readable recording medium may be a recording medium that can be directly connected to a user terminal (100), and may include, for example, a computer-readable recording medium such as a floppy drive, disk, tape, DVD / CD-ROM drive, memory card, etc. Alternatively, the software components may be loaded into the memory (120) through a communication module (140) that is not a computer-readable recording medium. For example, at least one program may be loaded into the memory (120) based on a computer program (e.g., a computer program for a processor (110) to perform an operation described later with reference to FIGS. 3 to 11) which is installed by files provided through the communication module (140) by developers or a file distribution system that distributes installation files for an application.

[0068] The input / output interface (130) may be a means for interfacing with a device for input or output (e.g., keyboard, mouse, etc.) that may be connected to or included in the user terminal (100). In FIG. 2a, the input / output interface (130) is shown as an element configured separately from the processor (110), but is not limited thereto, and the input / output interface (130) may be configured to be included in the processor (110).

[0069] The communication module (140) may provide a configuration or function for the server (200) and the user terminal (100) to communicate with each other via a network. Additionally, the communication module (140) may provide a configuration or function for the user terminal (100) to communicate with other external devices. For example, control signals, commands, data, etc. provided under the control of the processor (110) may be transmitted to the server (200) and / or external devices via the communication module (140) and the network.

[0070] Meanwhile, although not illustrated in FIG. 2a, the user terminal (100) may further include a display device. Alternatively, the user terminal (100) may be connected to an independent display device via wired or wireless communication to transmit and receive data to and from each other. For example, medical images, medical information, information related to machine learning models, etc., may be provided to the user through the display device.

[0071] FIG. 2b is a configuration diagram illustrating an example of a server according to one embodiment.

[0072] Referring to FIG. 2b, the server (200) includes a processor (210), memory (220), and a communication module (230). For convenience of explanation, FIG. 2b shows only the components related to the present invention. Accordingly, other general-purpose components may be included in the server (200) in addition to the components shown in FIG. 2b. Furthermore, it is obvious to those skilled in the art that the processor (210), memory (220), and communication module (230) shown in FIG. 2b may be implemented as independent devices.

[0073] The processor (210) can control the communication module (230) to transmit the pathology slide image to the user terminal (100). Alternatively, the server (200) can receive the pathology slide image from the user terminal (100).

[0074] Alternatively, the processor (210) can predict a probability score indicating the possibility of a specific gene mutation by analyzing a pathology slide image using a machine learning model. Then, the processor (210) can control the communication module (230) to transmit information about the probability score to the user terminal (100).

[0075] Alternatively, the processor (210) may generate information related to targeted treatment for the subject or various other medical information based on a probability score. And, the processor (210) may control the communication module (230) to transmit the generated information to the user terminal (100).

[0076] In other words, at least one of the operations of the processor (110) described above with reference to FIG. 2a can be performed by the processor (210). In this case, the user terminal (100) can output information transmitted from the server (200) through a display device.

[0077] Meanwhile, since the implementation example of the processor (210) is the same as the implementation example of the processor (110) described above with reference to FIG. 2a, a detailed description is omitted.

[0078] Various data, such as data generated according to the operation of the processor (210), can be stored in the memory (220). Additionally, an operating system (OS) and at least one program (e.g., a program required for the processor (210) to operate) can be stored in the memory (220).

[0079] Meanwhile, since the implementation example of the memory (220) is the same as the implementation example of the memory (120) described above with reference to FIG. 2a, a detailed description is omitted.

[0080] The communication module (230) may provide a configuration or function for the server (200) and the user terminal (100) to communicate with each other via a network. Additionally, the communication module (230) may provide a configuration or function for the server (200) to communicate with other external devices. For example, control signals, commands, data, etc. provided under the control of the processor (210) may be transmitted to the user terminal (100) and / or external devices via the communication module (230) and the network.

[0081] FIG. 3 is a flowchart illustrating an example of a method for outputting medical information according to one embodiment.

[0082] The method illustrated in FIG. 3 consists of steps processed chronologically in the computing device (20, 100, 200) or processor (110, 210) illustrated in FIG. 1 to 2b. Therefore, even if details are omitted below, the details described above regarding the computing device (20, 100, 200) or processor (110, 210) may also be applied to the method illustrated in FIG. 3. Additionally, as described above with reference to FIG. 2b, at least one of the steps performed by the processor (110) below may be processed by the processor (210).

[0083] In addition, below, the processor (110) outputting information or an image, etc. includes the processor (110) controlling a display device so that information or an image, etc. is output.

[0084] In step 310, the processor (110) can receive a pathology slide image of the subject.

[0085] For example, the pathology slide image may be a whole slide image (WSI) stained with H&E (hematoxylin and eosin), but is not limited thereto. At least one object included in the slide may be represented in the pathology slide image. Here, the object may include at least one of the type of cell represented on the pathology slide image, the shape of the cell, the shape of each of the components included in the cell, or the type of tissue represented on the pathology slide image (10).

[0086] For example, objects may include objects at the cell level and objects at the tissue level. Objects at the cell level may include types of cells (e.g., tumor cells, lymphocytes, fibroblasts, endothelial cells, etc.), morphology of cells (e.g., cell size, size of the nucleus, irregularity of the shape of the nucleus, etc.), and morphology of cell components (e.g., cell membrane, cytoplasm, shape of the cell nucleus). Additionally, objects at the tissue level may include types of tissues (e.g., cancer area, cancer stroma area, necrosis area, tertiary lymphoid structure (TLS), etc.).

[0087] In some cases, the processor (110) can identify objects in a pathology slide image by analyzing the pathology slide image. For example, the processor (110) can use a machine learning model to analyze the pathology slide image or the patches into which the pathology slide image is divided. For example, the processor (110) can generate patches by dividing the pathology slide image into a predetermined size (e.g., 1216*1216 pixels). Then, the processor (110) can analyze the patches.

[0088] Accordingly, the processor (110) can detect and classify cells expressed in the pathology slide image through a machine learning model. In addition, the processor (110) can detect and classify tissues included in the pathology slide image through a machine learning model.

[0089] In step 320, the processor (110) can predict a probability score indicating the possibility of a specific gene mutation by analyzing a pathology slide image using a machine learning model.

[0090] If the subject has non-small cell lung cancer (NSCLC), the specific gene may be EGFR. For example, if the pathology slide image is of a patient with non-small cell lung cancer (NSCLC), the processor (110) may calculate a probability score indicating the likelihood of an EGFR mutation. For example, the probability score may be any one of a series of values ​​included in the range of 0 to 1, but is not limited thereto.

[0091] In the example described above, cases with high probability scores may exhibit a typical EGFR-like histology. Conversely, cases with low probability scores may exhibit relatively diverse histological forms. That is, a typical EGFR histology may lead to a higher probability score.

[0092] In addition, cases with high probability scores in the example described above may show a superior therapeutic response to EGFR-TKI (i.e., super responders). Therefore, the probability score predicted by the processor (110) can also be used as a biomarker to predict the response to targeted therapy in patients with confirmed EGFR mutations. However, the above description is not limited to EGFR mutations and can be applied to other types of cancer or major gene mutations (driver mutations).

[0093] Meanwhile, morphological information itself may contain information that can serve as a predictive biomarker for targeted therapies (e.g., TKIs). For instance, such information may relate to altered histology due to resistance-related concurrent mutations, tumor heterogeneity, and the tumor microenvironment.

[0094] However, the carcinoma types and causative mutations applicable to the present disclosure are not limited to non-small cell lung cancer and EGFR, respectively. In other words, the present disclosure may be applicable to TKI treatment for non-small cell lung cancer patients as well as other targeted therapies for non-small cell lung cancer. Additionally, the present disclosure may be applicable to the treatment of patients with other mutations such as ALK (anaplastic lymphoma kinase), ROS1 (proto-oncogene tyrosine-protein kinase ROS), RET (rearranged during transfection) proto-oncogene, c-Met proto-oncogene, KRAS (kirsten rat sarcoma virus), or BRAF (B-Raf proto-oncogene).

[0095] For example, if the carcinoma is non-small cell lung cancer, the triggering mutation may be an ALK (anaplastic lymphoma kinase), ROS1 (proto-oncogene tyrosine-protein kinase ROS), RET (rearranged during transfection) proto-oncogene, c-Met proto-oncogene, KRAS (kirsten rat sarcoma virus), or BRAF (B-Raf proto-oncogene) mutation. Or, if the carcinoma is anaplastic thyroid cancer (ATC) or melanoma, the triggering mutation may be BRAF V600E. Or, if the carcinoma is colorectal cancer (CRC), the triggering mutation may be a mutation in genes such as KRAS and BRAF.

[0096] Meanwhile, the machine learning model for predicting the probability score may be the same model as the machine learning model described above with reference to step 310. Alternatively, the machine learning model for predicting the probability score may be a different model from the machine learning model described above with reference to step 310.

[0097] Hereinafter, with reference to FIG. 4, an example of training a machine learning model for predicting probability scores will be described.

[0098] FIG. 4 is a diagram illustrating an example in which a machine learning model is trained according to one embodiment.

[0099] Referring to FIG. 4, the machine learning model (410) is trained based on training data (420), and a trained machine learning model (430) can be generated.

[0100] Here, the training data (420) may include at least one pathology slide image obtained from at least one patient in whom a specific gene mutation has been confirmed, or at least one of information regarding the patient's responsiveness to targeted treatment.

[0101] When the machine learning model (410) is trained and the trained machine learning model (430) is generated, the trained machine learning model (430) can perform an inference step. That is, when a pathology slide image (440) of a new subject is input into the trained machine learning model (430), the model (430) can analyze the pathology slide image (440) and predict (output) a probability score (450) indicating the possibility of mutation of a specific gene.

[0102] Referring again to FIG. 3, at step 330, the processor (110) can output information related to targeted treatment for the subject based on a probability score.

[0103] For example, the processor (110) may classify the subject into one of a plurality of groups based on a probability score and output information related to targeted therapy based on the classification result. Here, the information related to targeted therapy may include information regarding at least one of a palliative setting, an adjuvant setting, or a neoadjuvant setting. Additionally, the information related to targeted therapy for each of the aforementioned therapies may include at least one of information regarding the treatment method, information regarding the dosage of the drug, or additional information regarding the treatment method.

[0104] Hereinafter, with reference to FIGS. 5 to 7, an example is described in which a processor (110) classifies a subject into one of a plurality of groups based on a probability score. Additionally, with reference to FIG. 8, an example is described in which the processor (110) outputs information related to targeted treatment based on the classification result.

[0105] FIG. 5 is a flowchart illustrating an example in which a processor according to one embodiment outputs information related to targeted treatment based on a probability score.

[0106] In step 510, the processor (110) can classify the subject into one of a plurality of groups based on the probability score.

[0107] The processor (110) can classify a subject into one of a plurality of groups based on the magnitude of the probability score. The probability score may be a specific value between 0 and 1. Thus, in order for a subject to be classified by the probability score, the processor (110) can determine a cutoff value for classifying the subject.

[0108] As an example, the processor (110) may classify the subject into either a first group (high score group or high response probability group) and a second group (low score group or low response probability group) according to one reference value. As another example, the processor (110) may classify the subject into either a first group (high score group or high response probability group), a second group (medium group), and a third group (low score group or low response probability group) according to two reference values. Alternatively, the processor (110) may classify the subject into any one of a plurality of groups using at least one reference value.

[0109] Here, the threshold value may be pre-set or modified by the user. As an example, the threshold value may be set based on basic statistics of a validation cohort distribution, such as the mean, median, quartiles, or deciles of probability scores. As another example, the threshold value may be set based on a validation cohort containing treatment response data. For instance, a threshold value that minimizes the hazard ratio of the respondent group compared to the non-responder group may be selected. As yet another example, the threshold value may be set according to criteria established in other existing clinical studies, guidelines, or expert consensus.

[0110] Meanwhile, the processor (110) can perform classification of the subject by considering at least one additional variable in addition to the probability score. Hereinafter, with reference to FIG. 6, an example in which the processor (110) performs classification of the subject by considering at least one additional variable is described.

[0111] FIG. 6 is a diagram illustrating an example in which a processor according to one embodiment classifies a subject into one of a plurality of groups.

[0112] Referring to FIG. 6, the processor (110) can output a classification result (630) of a subject based on a probability score (610) and at least one variable (620). For example, at least one variable (620) may be information corresponding to an analysis result of the subject's tumor.

[0113] For example, the processor (110) can generate a classification result (630) by combining a probability score (610) or categorical information based on the probability score (610) (e.g., High or Low) with at least one variable (620). In this case, at least one variable (620) may be expressed as at least one of a continuous score or categorical information.

[0114] The classification of high or low probability scores (610), as described below, can be determined by the processor (110) using the aforementioned reference value. The processor (110) can compare the probability score (610) with the reference value and determine that the probability score (610) is high if it is greater than the reference value, and determine that the probability score (610) is low if it is less than or equal to the reference value. Additionally, the processor (110) can generate categorical information based on the probability score (610) using at least one of the aforementioned reference values.

[0115] Classification of at least one variable (620) as high or low, large or small, etc., as described below, can be determined by the processor (110) using a reference value set for at least one variable (620). At this time, the reference value can be determined as a value based on data distribution, a value based on statistical rules, or a value that optimizes clinical response, and one or multiple reference values ​​can be set. The processor (110) can compare at least one variable (620) with the reference value and determine that if the at least one variable (620) is greater than the reference value, the at least one variable (620) is a large value or a high value, and if the at least one variable (620) is less than or equal to the reference value, the at least one variable (620) is a small value or a low value. Additionally, the processor (110) can generate categorical information based on at least one variable (620) using at least one reference value.

[0116] Here, examples of at least one variable (620) may be at least one of a variable corresponding to tumor heterogeneity, a variable corresponding to the tumor microenvironment (TME), a variable corresponding to immunohistochemistry (IHC) expression, a variable corresponding to molecular diagnostic data, or a variable corresponding to Radiology and Imaging-Based Biomarkers. Each of the above-mentioned variables will be described below.

[0117] Tumor heterogeneity refers to the diversity of cellular, molecular, or structural characteristics within a tumor. For example, a tumor can be evaluated based on nuclear pleomorphism, cellular morphology, glandular architecture, stromal composition, immune cell infiltration, vascular structure, etc. Heterogeneity can be assessed for any of the above items, regardless of their number.

[0118] The processor (110) can generate a classification result (630) of a subject by considering variables (e.g., categorical or continuous indicators) representing tumor heterogeneity derived from tissue morphology. Tumors with high heterogeneity are generally associated with treatment resistance. As described above, by utilizing variables corresponding to tumor heterogeneity, the accuracy of the prediction of the subject's treatment responsiveness can be improved.

[0119] Additionally, the processor (110) may also use variables (e.g., categorical or continuous indicators) representing a specific histomorphological type to generate the subject's classification result (630). For example, variables related to carcinomas likely to exhibit a more aggressive histological phenotype, such as neuroendocrine carcinoma or dedifferentiated carcinoma, or tumors that are morphologically similar but exhibit transcriptomic or epigenomic features correlated with aggressiveness or treatment resistance, may also be used in the analysis of the processor (110).

[0120] The tumor microenvironment refers to the complex and dynamic structure of a tumor and its surrounding environment, and includes cancer cells, blood vessels, immune cells, the extracellular matrix, and signaling molecules. The tumor microenvironment plays a key role in tumor growth, progression, metastasis, immune evasion, and treatment resistance.

[0121] For example, the tumor microenvironment can be expressed in various aspects such as immune phenotype, tumor infiltrating lymphocytes (TILs), endothelial cell densities, stromal composition, and fibroblast activation. The processor (110) can generate a classification result (630) of the subject using a continuous score (e.g., TIL density) or categorical information (e.g., immune phenotype) for the tumor microenvironment.

[0122] For example, a machine learning model analyzing pathology slide images can quantify the density of various cell types, such as lymphocytes, plasma cells, neutrophils, dendritic cells, macrophages, endothelial cells, and fibroblasts, in cancer cell regions or stromal regions within a tissue sample. For example, tumors exhibiting high vascular density (i.e., increased endothelial cell density) or immune cell density are likely to respond more effectively to specific treatments. Thus, by utilizing variables corresponding to the tumor microenvironment, the processor (110) can make more precise and robust predictions regarding the treatment outcome of the subject.

[0123] As an example, the processor (110) can generate a classification result (630) of a subject using a categorical variable such as an immune phenotype (inflamed or non-inflamed). For example, the processor (110) can classify a subject with a high probability score (610) and a non-inflamed immune phenotype as a 'highly likely to respond group'. Conversely, the processor (110) can classify a subject with a low probability score (610) and an inflammatory immune phenotype as a 'lowly likely to respond group'.

[0124] As another example, the processor (110) may generate a classification result (630) of a subject by combining a continuous score, such as cell density of lymphocytes, fibroblasts, etc., with a probability score (610). For example, the processor (110) may classify a subject with a high probability score (610) and a low density of specific cells (e.g., fibroblasts) as a ‘group with a high probability of response.’ Here, the criterion value defining whether the cell density is high or low may be determined according to the distribution of the dataset (e.g., the mean, median, 25th percentile, etc. of the available data), but is not limited thereto.

[0125] Immunohistochemistry (IHC) can provide information about the biological characteristics of tumor cells and the tumor microenvironment. Therefore, by using variables corresponding to immunohistochemistry, the predictive power of the subject's treatment outcome can be improved. For example, variables corresponding to immunohistochemistry can be expressed as continuous scores (e.g., TPS, CPS) or categorical information (e.g., HER2 positive, HER2 negative, etc.).

[0126] For example, in the case of a dual-target therapeutic agent (e.g., a bispecific antibody) that simultaneously targets EGFR and another target (e.g., MET), a quantitative value and a probability score (610) corresponding to the expression level of MET can be used together to generate a classification result (630). In this case, it can be predicted that a subject with a high probability score (610) and a high MET expression level will show a better response to the therapeutic agent.

[0127] Furthermore, the spatial distribution (e.g., density, proximity) of specific cells (e.g., CD3 T cells, FOXP3 T cells) relative to tumor cells can be quantified through immunohistochemistry (IHC) techniques. Since biological differences that induce changes in histomorphology can also affect the tumor microenvironment, IHC indicators describing the spatial relationships between specific cells and tumor cells can be used to accurately predict a subject's responsiveness to treatment.

[0128] As an example, categorical information (e.g., positive or negative) may be used as a variable corresponding to immunohistochemistry (IHC). For example, in the case of MET IHC, the processor (110) may classify subjects with a high probability score (610) and a negative MET IHC as a ‘group with a high probability of reaction.’

[0129] As another example, a series of scores (e.g., TPS, CPS, H-score) may be used as a variable corresponding to immunohistochemistry (IHC). For example, the processor (110) may classify subjects with high probability scores (610) and low MET IHC expression scores as a ‘group with high probability of response.’ Here, the threshold value for defining high and low MET IHC expression scores may be determined according to the distribution of the dataset (e.g., mean, median, 25th percentile, etc. of available data), but is not limited thereto.

[0130] As described above with reference to FIG. 1, the subject subject according to the present disclosure is a patient who has undergone a confirmatory molecular test such as PCR, NGS, or FISH. Accordingly, variables corresponding to molecular diagnostic data (e.g., the presence or absence of specific gene mutations, amplification, or fusion) can be used as indicators to improve the accuracy of predicting the subject's treatment response.

[0131] For example, NGS testing can provide information on the presence of subtypes of EGFR mutations (e.g., exon 19 deletion, L858R, etc.) and other co-mutations. In this case, certain co-mutations are known to induce resistance to targeted therapy, and this information can be used to predict the prognosis of the subject along with the probability score (610). This can be applied equally to tumors induced by gene fusion (e.g., sarcoma, hematological cancer, etc.). Additionally, information known to be related to treatment responsiveness (e.g., microsatellite instability (MSI), tumor mutation burden (TMB)) or information whose clinical significance has not yet been established can also be used to predict the prognosis of the subject along with the probability score (610).

[0132] Additionally, genetic variations detected through advanced sequencing technology can be used to predict the prognosis of the subject along with the probability score (610). For example, extrachromosomal DNA (ecDNA) amplification enables dynamic regulation of oncogenes and is associated with treatment resistance and tumor heterogeneity. Likewise, epigenetic modifications such as DNA methylation, histone modifications, and chromatin accessibility, or non-coding RNAs such as miRNA, lncRNA, and circRNA, affect gene expression and tumor progression. These biomarkers can also be used to predict the prognosis of the subject along with the probability score (610).

[0133] For example, molecular test results can be considered as binary information. For example, T790M is a subtype of EGFR variant known to cause resistance to EGFR targeted therapy. Therefore, the processor (110) can use the status of T790M (i.e., variant or wild type) along with a probability score (610) to generate a classification result (630) for the subject. For example, a subject with a high probability score (610) and a wild-type T790M can be classified as a 'highly responsive group'. Additionally, the presence or absence of ecDNA can also be used along with the probability score (610) to predict the subject's prognosis.

[0134] Information obtained from various medical imaging technologies such as X-ray, CT, MRI, and mammography (MMG) can be used as meaningful information about tumors and as biomarkers to predict a patient's responsiveness to treatment. For example, radiomic features describing tissue texture and border characteristics are correlated with a patient's responsiveness to treatment. Therefore, variables corresponding to Radiology and Imaging-Based Biomarkers can be used to quantify the probability related to a patient's responsiveness to treatment, along with a probability score (610).

[0135] For example, the processor (110) can calculate an imaging-based response score (e.g., a specific value between 0 and 1) and convert the calculated score into categorical information. The calculated score can be used in combination with a probability score (610) as a predictive biomarker in itself. The processor (110) can classify subjects with high probability scores (610) and high imaging-based response scores as a ‘group with a high probability of response.’

[0136] As described above with reference to Step 510 and Figure 6, the processor (110) classifies the subject into one of a plurality of groups, so that clinical recommendations for the subject may vary depending on the clinical setting (i.e., palliative setting, adjuvant setting, or neoadjuvant setting).

[0137] FIG. 7 is a diagram illustrating an example in which a processor according to one embodiment classifies a subject into one of a plurality of groups by considering variables corresponding to tumor heterogeneity.

[0138] Referring to FIG. 7, the x-axis (720) of the graph represents the Tumor Heterogeneity Score, and the y-axis (710) represents the Mutation Likelihood, i.e., the probability score (610).

[0139] Similar to the probability score (610), a distribution statistics-based or response optimization-based threshold value may also be used for the continuous score representing tumor heterogeneity (i.e., the value on the x-axis (720)). For example, the graph in FIG. 7 corresponds to a case where both the probability score (610) (i.e., the value on the y-axis (710)) and the tumor heterogeneity score (i.e., the value on the x-axis (720)) are divided into three categories.

[0140] The processor (110) can classify a subject into one of a plurality of groups by combining information on the x-axis (720) and information on the y-axis (710). For example, the processor (110) can classify subjects exhibiting a high probability score (610) and a low tumor heterogeneity score into a good responder group (i.e., a group with a high probability of response) (730). On the other hand, the processor (110) can classify subjects exhibiting a high tumor heterogeneity score and a low probability score (610) into a poor responder group (i.e., a group with a low probability of response) (740). Additionally, the processor (110) can classify subjects not included in the group (730) and the group (740) into an intermediate group.

[0141] Referring again to FIG. 5, at step 520, the processor (110) can output information related to targeted treatment based on the classification result.

[0142] The processor (110) can generate information related to targeted therapy for the subject (e.g., clinical recommendations) based on the subject's classification results. In this case, the information related to targeted therapy may vary depending on the subject's clinical setting (i.e., palliative setting, adjuvant setting, or neoadjuvant setting). In this case, the clinical setting refers to an environment where a therapy is provided, which is classified according to the purpose of treatment and the stage of treatment. For example, the clinical setting may include an environment where a therapy performed before surgery is provided, an environment where a therapy performed to remove minimal residual disease or prevent recurrence after surgery is provided, and an environment where a therapy performed when surgery is impossible or curative treatment is difficult is provided. Additionally, the information related to targeted therapy for each clinical setting may include at least one of information regarding the treatment method, information regarding the dosage of the drug, or additional information regarding the therapy.

[0143] Hereinafter, information related to targeted therapy will be explained in detail with reference to Fig. 8.

[0144] FIG. 8 is a diagram illustrating examples of information related to targeted therapy according to one embodiment.

[0145] Referring to FIG. 8, information (810) related to targeted therapy may include information regarding at least one of palliative therapy (820), adjuvant therapy (830), or prior adjuvant therapy (840). In other words, information (810) related to targeted therapy may be generated differently depending on the clinical environment of the subject. Additionally, within each therapy, the information (810) may include at least one of information regarding a treatment method (821, 831, 841), information regarding a dosage of a drug (822, 832), or additional information regarding a therapy (823, 833, 842).

[0146] That is, the processor (110) can determine a recommended treatment method based on the degree of response of the subject to targeted therapy. For example, the processor (110) can classify the subject into at least two categories (e.g., good responders, poor responders, etc.) based on the predicted degree of response to targeted therapy (i.e., probability score). Then, the processor (110) can recommend an appropriate treatment method according to the category.

[0147] First, an example is described in which the processor (110) provides information related to a palliative setting (820). A palliative setting (820) refers to treatment performed for the purpose of suppressing tumor progression, alleviating symptoms, and maintaining or improving the patient's quality of life in cases where curative removal of the tumor is difficult or where a cure is unlikely due to systemic metastasis. In the following, it is assumed that the subject is a patient with non-small cell lung cancer and that the targeted therapy is a TKI.

[0148] As an example, if a poor response to TKI monotherapy is predicted for a subject, the processor (110) may suggest considering TKI-based combination therapy as a treatment (821). Specifically, the treatment (821) comprises: i) a combination of a TKI and cytotoxic chemotherapy (e.g., osimertinib + chemotherapy); ii) a combination of a TKI and a monoclonal antibody (e.g., including immune checkpoint inhibitors); iii) a combination of a TKI and a bispecific antibody (e.g., lazertinib + amivantamab) or a multispecific antibody; iv) a combination of a TKI and cell therapy (e.g., CAR-T, TCR-T, etc.); v) a combination of a TKI and an allosteric inhibitor; vi) a combination of a TKI and a degrader (e.g., PROTACs, etc.); vii) a combination of a TKI and a radiopharmaceutical or other radiation therapy; viii) a triplet or quadruplet combination therapy of the above combinations; ix) a mutation-based treatment that does not include a TKI (e.g., a cancer vaccine or TCR therapy targeting a mutation-derived neoantigen). It may include the back.

[0149] As another example, if the subject is predicted to have a high response to treatment, the processor (110) may suggest a dose reduction or adjustment of the dosing schedule in relation to the dosage (822) of the drug.

[0150] As another example, the processor (110) may suggest additional testing as additional information (823). For example, if the subject is predicted to be a poor responder, the processor (110) may generate information recommending a biopsy of the liver metastasis site to identify alternative targets or confirm the resistance profile.

[0151] Next, an example is described in which the processor (110) provides information related to an adjuvant setting (830). The adjuvant setting (830) refers to a treatment regimen designed to eliminate minor residual disease after surgery and prevent recurrence. In this case, the information related to the adjuvant setting (830) generated by the processor (110) (e.g., a guide) can help balance the intensity of treatment and the necessity of treatment. The result of classifying the subject into one group by the processor (110) can be used as a reference to determine how much treatment should be provided to each subject.

[0152] For example, if the subject is predicted to be a poor responder, the processor (110) may generate a treatment method (831) suggesting the simultaneous or sequential administration of cytotoxic chemotherapy or chemoradiation. Conversely, if the subject is predicted to be a good responder, the processor (110) may generate a treatment method (831) suggesting avoiding chemoradiation or other systemic treatments.

[0153] As another example, the processor (110) may suggest adjusting the dosage (832) of the drug simultaneously or sequentially according to the predicted responsiveness of the subject.

[0154] As another example, the processor (110) may provide information regarding the timing of when the adjuvant therapy (830) is performed as additional information (833). For example, depending on the predicted responsiveness of the subject, the processor (110) may provide a proposal for the start time of the adjuvant therapy (830) or the adjustment period during which the adjuvant therapy (830) is performed.

[0155] Next, an example is described in which the processor (110) provides information related to neoadjuvant therapy (840). Neoadjuvant therapy (840) refers to treatment intended to shrink the tumor and remove micrometastasis early before surgery. The result of classifying the subject into one group by the processor (110) can have a direct impact on determining the subject's preoperative treatment strategy and timing. For example, neoadjuvant therapy (840) can generally be utilized to improve surgical outcomes in locally advanced cancer (e.g., cases involving large tumors or lymph node involvement) and to check for a treatment response early.

[0156] For example, if the subject is predicted to be a poor responder, the processor (110) may generate a treatment (841) suggesting a change in the preceding adjuvant therapy (840). Conversely, if the subject is predicted to be a good responder, the processor may generate a treatment (841) suggesting a reduction in the dose of chemotherapy or omission.

[0157] As another example, the processor (110) may generate information related to adjusting the timing of surgery and treatment as additional information (842). For example, if the subject is predicted to be a poor responder, the processor (110) may generate additional information (842), such as i) changing the pre-adjuvant therapy (840), ii) adding chemotherapy or alternative drugs, or iii) proceeding with early surgery to avoid ineffective treatment. Conversely, if the subject is predicted to be a good responder, additional information (842) may be generated, such as recommending shortening the duration of pre-adjuvant therapy (840) and proceeding with surgery.

[0158] Additionally, although not illustrated in FIG. 8, the processor (110) may suggest considering a change in monitoring schedule for good or poor responders through imaging examinations or blood-based measurements (e.g., nucleic acid-based liquid biopsy). Alternatively, the processor (110) may suggest considering a rebiopsy to re-evaluate the prognosis for next-generation targeted therapy strategies during treatment or at the point of progression after treatment.

[0159] Meanwhile, the processor (110) may output pathology slide images or various other information in addition to the information described above with reference to FIG. 8. Hereinafter, with reference to FIG. 9 and FIG. 10, examples of images or various information output by the processor (110) will be described.

[0160] FIGS. 9 and FIGS. 10 are drawings illustrating an example of an image and information output to a display device according to one embodiment.

[0161] Referring to FIG. 9, the processor (110) can output a heatmap image corresponding to probability scores by overlaying it on a pathology slide image. For example, on the screen (900), areas with high probability scores (e.g., red) and areas with low probability scores (e.g., blue) can be visually distinguished and displayed. Additionally, on the screen (900), summary information regarding the variations appearing on the pathology slide image (e.g., the ratio of each variation, etc.) and classification results for the subject (e.g., good responders or poor responders) can be output together.

[0162] Referring to FIG. 10, the processor (110) can output information about at least one tissue or at least one cell represented in a pathology slide image. For example, quantitative information about the tissue and cell analyzed by a machine learning model can be output on the screen (1000). This information may include 'Tissue Area', 'Cell Counts', 'Cell Densities on Tissues', etc., and may be based on the results of tissue segmentation and cell detection by the machine learning model.

[0163] FIG. 11 is a drawing for illustrating an example of a system for outputting medical information according to one embodiment.

[0164] Referring to FIG. 11, the system (1100) is an example of a system and network for analyzing biomarkers using a machine learning model.

[0165] A scanner (1121), a user terminal (1122, 1123), an image management system (1130), an AI-based biomarker analysis system (1140), a laboratory information management system (1150), and / or a hospital or laboratory server (1160) can each be connected to a network (1170), such as the internet, via one or more computers, servers, and / or mobile devices, or can communicate with a user (1112) via one or more computers and / or mobile devices.

[0166] According to various embodiments of the present disclosure, the method described above with reference to FIGS. 2a through 10 may be performed by at least one of a user terminal (1122, 1123), an image management system (1130), an AI-based biomarker analysis system (1140), a laboratory information management system (1150), and a hospital or laboratory server (1160) or a combination thereof.

[0167] If the medical image is a pathology slide image, the scanner (1121) can obtain a digitized image from a tissue sample slide (pathology slide) created using a tissue sample of the subject (1111).

[0168] User terminals (1122, 1123), image management systems (1130), AI-based biomarker analysis systems (1140), laboratory information management systems (1150) and / or hospital or laboratory servers (1160) may generate or obtain from other devices tissue samples of one or more subjects (1111), tissue sample slides (pathology slides), digitized images of tissue sample slides (pathology slides), various types of medical images of the subject, or any combination thereof. Additionally, user terminals (1122, 1123), image management systems (1130), AI-based biomarker analysis systems (1140), laboratory information management systems (1150) and / or hospital or laboratory servers (1160) may obtain any combination of subject-specific information, such as the age, medical history, cancer treatment history, family history, past biopsy records, or disease information of one or more subjects (1111).

[0169] A scanner (1121), a user terminal (1122, 1123), an AI-based biomarker analysis system (1140), a laboratory information management system (1150), and / or a hospital or laboratory server (1160) can transmit medical images, specific information of a subject (1111), and / or results of analyzing medical images to an image management system (1130) via a network (1170). The image management system (1130) may include a storage for storing received images and a storage device for storing analysis results.

[0170] In addition, according to various embodiments of the present disclosure, a machine learning model learned and trained to predict at least one of information regarding at least one cell, information regarding at least one region, and medical information (e.g., information related to biomarkers, medical diagnostic information, medical treatment information, etc.) from a medical image of a subject (1111) may be stored and operated in a user terminal (1122, 1123), an image management system (1130), etc.

[0171] As described above, the computing device (20) can predict responsiveness to targeted therapy using pathology slide images (10) for subjects in whom a specific gene mutation has already been confirmed. Thus, the computing device (20) can generate important information for establishing a treatment strategy for the subject without additional complex gene tests.

[0172] Additionally, the computing device (20) can classify subjects based on the predicted results of treatment responsiveness (e.g., good responder or poor responder) and establish a customized treatment plan optimized for the classified group.

[0173] Additionally, the computing device (20) can contribute to improving the prognosis of the subject and reducing unnecessary treatment by outputting medical information (30) for various clinical scenarios (e.g., palliative setting, adjuvant setting, neoadjuvant setting, etc.).

[0174] Meanwhile, the above-described method can be written as a program executable on a computer and can be implemented on a general-purpose digital computer that operates the program using a computer-readable recording medium. In addition, the structure of the data used in the above-described method can be recorded on a computer-readable recording medium through various means. The computer-readable recording medium includes storage media such as magnetic storage media (e.g., ROM, RAM, USB, floppy disk, hard disk, etc.) and optical reading media (e.g., CD-ROM, DVD, etc.).

[0175] A person skilled in the art related to the present embodiment will understand that it may be implemented in modified forms without departing from the essential characteristics of the description above. Therefore, the disclosed methods should be considered in an illustrative rather than a restrictive sense, and the scope of rights is defined in the claims rather than the description above, and should be interpreted to include all differences within the scope of equivalence.

Claims

1. At least one memory in which at least one instruction is stored; and It includes at least one processor that operates according to the above at least one instruction; and The above at least one processor is, A computing device that predicts a probability score indicating the possibility of a specific gene mutation by analyzing a pathology slide image of a subject using a machine learning model, and outputs information related to targeted therapy for the subject based on the probability score.

2. In Paragraph 1, The above machine learning model is, A computing device that learns based on at least one of at least one pathology slide image corresponding to a tissue sample obtained from at least one patient in whom a mutation of the specific gene is confirmed, or information regarding the responsiveness to targeted treatment of the at least one patient.

3. In Paragraph 1, The above at least one processor is, A computing device that classifies the subject into one of a plurality of groups based on the above probability score and outputs information related to the above target treatment based on the classification result.

4. In Paragraph 3, The above at least one processor is, A computing device that classifies a subject into any one of the plurality of groups by further considering at least one variable corresponding to the analysis result of the tumor of the subject.

5. In Paragraph 4, The above at least one processor is, Classifying the subject into one of the plurality of groups by combining the above probability score or categorical information based on the above probability score with the above at least one variable, and The above at least one variable is, A computing device represented by at least one of a continuous score or categorical information.

6. In Paragraph 1, The information related to the above targeted therapy is, A computing device comprising information regarding at least one of a palliative setting, an adjuvant setting, or a neoadjuvant setting.

7. In Paragraph 1, The information related to the above targeted therapy is, A computing device comprising at least one of information regarding a treatment method, information regarding a dosage of a drug, or additional information regarding a treatment regimen.

8. In Paragraph 1, The above at least one processor is, A computing device that outputs a heatmap image corresponding to the probability score by overlaying it on the pathology slide image.

9. In Paragraph 1, The above at least one processor is, A computing device that outputs information about at least one tissue or at least one cell represented in the pathology slide image above.

10. Step of receiving the pathology slide image of the subject; A step of predicting a probability score indicating the possibility of a specific gene mutation by analyzing the pathology slide image using a machine learning model; and A method for outputting medical information about a subject, comprising the step of outputting information related to targeted therapy for the subject based on the probability score above.

11. In Paragraph 10, The above machine learning model is, A method of learning based on at least one pathology slide image corresponding to a tissue sample obtained from at least one patient in whom a mutation of the specific gene is confirmed, or information regarding the responsiveness to targeted treatment of at least one patient.

12. In Paragraph 10, The above outputting step is, A step of classifying the subject into one of a plurality of groups based on the above probability score; and A method comprising the step of outputting information related to the target treatment based on the above classification result.

13. In Paragraph 12, The above classification step is, A method for classifying a subject into any one of the plurality of groups by further considering at least one variable corresponding to the analysis result of the tumor of the subject.

14. In Paragraph 13, The above classification step is, Classifying the subject into one of the plurality of groups by combining the above probability score or categorical information based on the above probability score with the above at least one variable, and The above at least one variable is, A method expressed as at least one of a continuous score or categorical information.

15. In Paragraph 10, The information related to the above targeted therapy is, A method comprising information regarding at least one of a palliative setting, an adjuvant setting, or a neoadjuvant setting.

16. In Paragraph 10, The information related to the above targeted therapy is, A method comprising at least one of information regarding a treatment method, information regarding a dosage of a drug, or additional information regarding a treatment regimen.

17. In Paragraph 10, The above outputting step is, A method of overlaying and outputting a heatmap image corresponding to the probability score on the pathology slide image.

18. In Paragraph 10, The above outputting step is, A method for outputting information about at least one tissue or at least one cell expressed in the above pathology slide image.

19. A computer-readable recording medium storing a program for executing the method of claim 1 on a computer.

Images