Diagnostic methods and devices for kidney disease

Abstract

A method and apparatus for diagnosing kidney disease are disclosed. A control method for a diagnostic apparatus according to one embodiment includes the steps of acquiring a retinal image of a subject and acquiring kidney disease diagnostic information for the subject using a machine learning model based on the retinal image, wherein the machine learning model includes a first model and a second model, the first model being a neural network model and the second model being a regression-based machine learning model.

Description

[Technical Field]

[0001] This invention relates to a method and apparatus for diagnosing kidney disease. [Background technology]

[0002] Retinal examinations are frequently used in ophthalmology as diagnostic tools because they allow for the observation of abnormalities in the retina, optic nerve, and macula, and the results can be confirmed relatively easily through imaging. Meanwhile, with the rapid advancements in artificial intelligence technology in recent years, the development of diagnostic AI is actively progressing in the field of medical diagnosis, particularly in image-based diagnosis. Global companies are investing heavily in the development of AI for the analysis of diverse medical imaging data, including large-scale data input through collaborations with the medical community, and some companies have succeeded in developing AI diagnostic tools that produce excellent diagnostic results.

[0003] Because retinal imaging allows for non-invasive observation of blood vessels within the body, there is a growing demand to expand the application of retinal imaging for diagnosis not only to eye diseases but also to kidney diseases. [Overview of the project] [Problems that the invention aims to solve]

[0004] The technical problem addressed by this application is to provide a method for diagnosing kidney disease that can acquire information about kidney disease with high accuracy using retinal images and machine learning models.

[0005] The technical problems addressed in this application are not limited to those described above, and any problems not mentioned can be clearly understood by a person with ordinary skill in the art to which this application pertains from this specification and the accompanying drawings. [Means for solving the problem]

[0006] According to one embodiment, a control method for a diagnostic device according to one embodiment includes the steps of acquiring a retinal image of a subject and acquiring kidney disease diagnostic information for the subject using a machine learning model based on the retinal image, wherein the machine learning model includes a first model and a second model, the first model being a neural network model and the second model being a regression-based machine learning model.

[0007] The technical solutions are not limited to those described above, and any technical solutions not mentioned herein will be clearly understood by a person with ordinary skill in the art to which this application pertains, based on this specification and the accompanying drawings. [Effects of the Invention]

[0008] According to this application, information regarding kidney disease can be obtained with high accuracy using retinal images and machine learning models.

[0009] The effects of the invention described herein are not limited to those described above, and any effects not mentioned herein will be clearly understood by a person with ordinary skill in the art to which this application pertains from this specification and the accompanying drawings. [Brief explanation of the drawing]

[0010] [Figure 1] Figure 1 shows a diagnostic system according to one embodiment. [Figure 2] Figure 2 is a block diagram illustrating a learning device according to one embodiment. [Figure 3] Figure 3 is a block diagram illustrating a diagnostic device according to one embodiment. [Figure 4] Figure 4 shows a diagnostic system according to one embodiment. [Figure 5] Figure 5 is a block diagram illustrating a client device according to one embodiment of the present invention. [Figure 6] Figure 6 is a diagram illustrating a diagnostic process according to one embodiment of the present invention. [Figure 7] Figure 7 is a diagram for explaining the configuration of a learning unit according to an embodiment of the present invention. [Figure 8] Figure 8 is a conceptual diagram for explaining an image dataset according to an embodiment of the present invention. [Figure 9] Figure 9 is a block diagram for explaining the learning process of a diagnosis model according to an embodiment of the present invention. [Figure 10] Figure 10 is a diagram for explaining the configuration of a diagnosis unit according to an embodiment of the present invention. [Figure 11] Figure 11 is a diagram for explaining the diagnosis process according to an embodiment of the present invention. [Figure 12] Figure 12 is a block diagram for explaining a diagnosis unit according to an embodiment of the present invention. [Figure 13] Figure 13 is a diagram for explaining the diagnosis process according to an embodiment of the present invention. [Figure 14] Figure 14 is a diagram for explaining a diagnosis system according to an embodiment of the present invention. [Figure 15] Figure 15 is a diagram for explaining a serial diagnosis model according to an embodiment. [Figure 16] Figure 16 is a diagram for explaining a serial diagnosis model according to another embodiment. [Figure 17] Figure 17 is a diagram for explaining a serial diagnosis model according to still another embodiment. [Figure 18] Figure 18 is a diagram for explaining a diagnosis method using a diagnosis model according to an embodiment. [Figure 19] Figure 19 is a diagram for explaining a kidney disease diagnosis method according to an embodiment. [Figure 20] Figure 20 illustrates a diagnosis model for obtaining kidney disease diagnosis information according to an embodiment. [Figure 21] Figure 21 shows the clinical characteristics of the subject according to Example 1. [Figure 22a] Figure 22a is a diagram for explaining the incidence rate of kidney disease-related events based on the kidney disease diagnosis information according to Example 1. [Figure 22b] Figure 22b is a diagram illustrating the incidence rate of kidney disease-related events based on kidney disease diagnostic information related to Example 1. [Figure 23] Figure 23 illustrates the predictive performance of kidney disease diagnostic information within a five-year period for kidney disease-related events in Example 1. [Figure 24] Figure 24 is a diagram illustrating a method for diagnosing kidney disease according to another embodiment. [Figure 25] Figure 25 is a diagram illustrating a method for providing guide information for kidney disease diagnostic information according to one embodiment. [Figure 26] Figure 26 is a diagram illustrating a method for providing guide information using existing prescription information and kidney disease diagnostic information according to one embodiment. [Figure 27] Figure 27 is a diagram illustrating a method for predicting the progression rate of renal disease using renal disease diagnostic information according to one embodiment. [Figure 28] Figure 28 is a diagram illustrating the clinical characteristics of the subjects in Example 2. [Figure 29a] Figure 29a is a diagram illustrating the KDIGO grade and the incidence of kidney disease-related events based on kidney disease diagnostic information in Example 2. [Figure 29b] Figure 29b is a diagram illustrating the KDIGO grade and the incidence of kidney disease-related events based on kidney disease diagnostic information in Example 2. [Modes for carrying out the invention]

[0011] A control method for a diagnostic device according to one embodiment includes the steps of acquiring a retinal image of a subject and acquiring kidney disease diagnostic information for the subject using a machine learning model based on the retinal image, wherein the machine learning model includes a first model and a second model, the first model being a neural network model and the second model being a regression-based machine learning model.

[0012] The above-mentioned objectives, features, and advantages of the present invention will become clearer through the following detailed description in conjunction with the accompanying drawings. However, since the present invention can be modified in various ways and may have several embodiments, specific embodiments will be illustrated in the drawings and described in detail below.

[0013] In the drawings, the thicknesses of layers and regions are exaggerated for clarity, and the term "on" or "on" an element or layer includes not only the immediate vicinity of the other element or layer, but also all cases where other layers or elements are interposed between them. Throughout the specification, the same reference numeral generally indicates the same element. Furthermore, elements with the same function within the same conceptual scope shown in the drawings of each embodiment are described using the same reference numeral.

[0014] If a specific description of a known function or configuration related to the present invention is deemed to unnecessarily obscure the gist of the invention, such detailed description will be omitted. Furthermore, the numbers used in the description of this specification (e.g., 1st, 2nd, etc.) are merely identifiers to distinguish one component from another.

[0015] Furthermore, the suffixes "module" and "part" used in the following description for the constituent elements are added or used interchangeably solely for the sake of ease of specification preparation, and do not have any distinct meaning or role in themselves.

[0016] The methods according to the embodiment can be embodied in the form of program instructions that can be performed through various computer means and recorded on a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, etc., individually or in combination. The program instructions recorded on the medium may be specifically designed and configured for the embodiment, or may be publicly known and available to those skilled in the art of computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes; optical media such as CD-ROMs and DVDs; magneto-optical media such as floptical disks; and hardware devices specifically configured to store and execute program instructions, such as ROM (Read Only Memory), RAM (Random Access Memory), and flash memory. Examples of program instructions include not only machine code, such as that produced by a compiler, but also high-level language code that can be executed by a computer using an interpreter or the like. The hardware devices described above may be configured to operate as one or more software modules to perform the operations of the embodiment, and vice versa.

[0017] 1. Diagnosis using retinal imaging

[0018] 1.1. Diagnostic System and Process

[0019] 1.1.1. Purpose and Definitions

[0020] The following describes diagnostic systems and methods to assist medical professionals in determining the presence or absence of disease or abnormalities that may be the basis for such a determination, based on eye images. In this specification, the term "diagnosis" may mean diagnostic assistance rather than directly diagnosing a disease. For the sake of clarity, the term "diagnosis" will be used below, but it may refer to diagnostic assistance.

[0021] In particular, this document describes a diagnostic method that uses deep learning techniques to construct a machine learning model for diagnosing diseases and uses the constructed model to assist in detecting the presence or absence of disease or abnormal findings. In this specification, an eyeball image is an image including the subject's eye and may include various images such as retinal images and / or fundus images. For the sake of explanation, this specification will focus on retinal images, but is not limited to retinal images, and the explanation in this specification can of course be applied to other eyeball images as well.

[0022] The machine learning models described herein may be designed based on a variety of machine learning libraries. For example, machine learning models can refer to various forms of models designed based on guided, unguided, semi-guided, or reinforcement learning artificial intelligence algorithms such as decision trees, random forest algorithms, stochastic gradient descent algorithms, neural network algorithms, k-nearest neighbors algorithms, linear regression, logistic regression, Cox proportional hazards model (survival model) (regression-based), support vector machines, k-means, hierarchical cluster analysis (HCA), expectation maximization, principal component analysis (PCA), kernel PCA, locally-linear embedding (LLE), t-distributed stochastic neighbor embedding (t-SNE), association rule mining (Apriori), and Eclat.

[0023] In the following, unless otherwise specified, machine learning models will primarily be described as neural network models for convenience. However, this does not necessarily mean that only models based on neural network algorithms are permitted, and it is obvious that models based on other algorithms may be substituted within the scope of the functions and objectives of the invention described herein.

[0024] According to one embodiment of the present invention, a diagnostic system or method may be provided that assists in the diagnosis of at least one of the following based on retinal images: eye disease, cardiovascular disease (and / or cardiocerebrovascular disease), kidney disease, or other systemic disease.

[0025] Exemplary examples, as used herein, eye diseases may include at least one of the following: cataract, glaucoma, macular degeneration, diabetic retinopathy, epiretinal membrane, macular hole, high / degenerative myopia, melanoma, retinal detachment, dry eye syndrome, presbyopia, and astigmatism.

[0026] Furthermore, cardiovascular disease may include at least one of the following: coronary artery disease (CAD), aortic stenosis, hypertension, arterial hypertension, heart failure, arrhythmia, atrial fibrillation, valvular heart disease, cardiomyopathy, peripheral vascular disease, peripheral artery disease (PAD), heart attack, aneurysm, aortic aneurysm (e.g., abdominal aortic aneurysm, thoracic aortic aneurysm), thrombotic disease (e.g., deep vein thrombosis, pulmonary embolism), myocardial infarction, and cerebrovascular disease. Furthermore, cardiovascular disease may include at least one of the following: stroke, ischemic stroke, cerebral infarction, cerebral hemorrhage, subarachnoid hemorrhage, transient ischemic attack, or death due to cardiovascular disease. Cardiovascular disease may also include complications. Complications may also include aspiration pneumonia, dysphagia, impaired motor function, impaired language function, impaired cognitive function, sleep disorders, emotional disorders, cranial neuralgia, urinary tract infection, malnutrition, deep vein thrombosis, pressure ulcers, falls, pain, seizures, and depression.

[0027] Furthermore, kidney disease may include at least one of the following: chronic kidney disease (CKD), acute kidney injury (AKI), kidney stones, nephrotic syndrome, glomerulonephritis, polycystic kidney disease (PKD), kidney cancer, and pyelonephritis. Kidney disease may also include complications. These complications may include side effects from dialysis, hypotension, muscle spasms, nausea and vomiting, headache, dialysis disequilibrium syndrome, itching, and arrhythmias.

[0028] Furthermore, other systemic diseases may include at least one of the following: diabetes, hypertension, hypotension, Alzheimer's disease, giant cell virus, and arteriosclerosis.

[0029] Furthermore, according to other embodiments of the present invention, various parameters of a subject can be predicted based on retinal images. For example, these parameters may include at least one of the parameters representing the subject's physical information, such as biological age, sex, height, weight, BMI index, body mass, body fat percentage, and body muscle mass. Furthermore, the aforementioned parameters may include at least one of the following diagnostic numerical parameters: hematocrit, red blood cell count, white blood cell count, hemoglobin level, platelet count, total iron-binding capacity (TIBC), iron level, ferritin (iron-storing protein) level, total protein level, albumin level, aspartate transferase level (AST), aminotransferase level, γ-GTP, γ-GT, alkaline phosphatase (ALP) level, globulin level, hepatitis antigen level, hepatitis antibody level, glycated hemoglobin level (HbA1c), blood urea nitrogen (BUN) level, creatinine level, uric acid level, total cholesterol level, high-density lipoprotein (HDL Cholesterol) level, low-density lipoprotein (LDL Cholesterol) level, triglyceride (TG) level, bicarbonate level, systolic blood pressure (SBP), and diastolic blood pressure (DBP).

[0030] According to yet another embodiment of the present invention, a diagnostic system or method for detecting abnormal retinal findings that can be used in the diagnosis of an eye disease or other disease may be provided. For example, color abnormalities of the entire retina, lens opacity, abnormalities in the cup-to-disc ratio (C / D ratio), macular abnormalities (e.g., macular hole), abnormalities in the diameter and course of blood vessels, abnormalities in the diameter of retinal arteries, retinal hemorrhage, microaneurysms, hard exudate, epiretinal membrane, myelinated nerve fiber, chorioretinal atrophy, retinal nerve fiber layer defect (RNFL defect), exudation, drusen, cataract, glaucoma, diabetic retinopathy, tessellated fundus, large optic cup, retinal vein occlusion (RVO), branch retinal vein occlusion (BRVO), central retinal vein occlusion (CRVO), retinal artery occlusion (RAO), rhegmatogenous retinal detachment RD), posterior serous / exudative RD, central serous chorioretinopathy (CSCR), VKH disease, maculopathy, retinal epithelial proliferation (ERM), macular hole (MH,Macular Hole, Pathological Myopia, Optic Nerve Degeneration, Optic Atrophy, Severe Hypertensive Retinopathy, Disc Swelling and Elevation, Dragged Disc, Pigmentary Degeneration, Congenital Disc Abnormality, Retinitis Pigmentosa, Bietti Crystalline Dystrophy, Peripheral Retinal Degeneration and Break, Myelinated Nerve Fiber, Vitreous Particles, Fundus Neoplasm, Massive Hard Exudates, Yellow-White Spots / Patches Diagnostic systems or methods may be provided to acquire findings such as spots / flecks, cotton-wool spots, vascular tortuosity, chorioretinal atrophy / coloboma, preretinal hemorrhage, fibrosis, laser spots, silicone oil in the eye, blurred fundus, blurred fundus without PDR, and blurred fundus with suspected PDR.

[0031] In this specification, diagnostic information may be understood to include diagnostic information based on the determination of the presence or absence of a disease, or the findings that underlie such diagnosis.

[0032] 1.1.2. Diagnostic System Configuration

[0033] According to one embodiment of the present invention, a diagnostic system may be provided.

[0034] Figure 1 is a diagram showing a diagnostic system according to one embodiment of the present invention. Referring to Figure 1, the diagnostic system 1 may include a learning device 10 for training a diagnostic model, a diagnostic device 20 for performing a diagnosis using the diagnostic model, and a client device 30 for acquiring a diagnostic request. The diagnostic system 1 may include multiple learning devices, multiple diagnostic devices, or multiple client devices.

[0035] The learning device 10 may include a learning unit 100. The learning unit 100 can train a diagnostic model. For example, the learning unit 100 can acquire a retinal image dataset and train a diagnostic model that detects diseases or abnormal findings from retinal images. The learning unit 100 may be included in the processor of the learning device 10, which will be described later, and may represent a functional representation of the processor's processing for learning.

[0036] The diagnostic device 20 may include a diagnostic unit 200. The diagnostic unit 200 can diagnose a disease or acquire auxiliary information used for diagnosis using a diagnostic model. For example, the diagnostic unit 200 can acquire diagnostic information using a diagnostic model trained by a learning unit. The diagnostic unit 200 may be included in the processor of the diagnostic device 20, which will be described later, and may represent a functional representation of the processor's processing for diagnosis.

[0037] The client device 30 may include an imaging unit 300. The imaging unit 300 can capture retinal images. The client device may be an ophthalmic retinal imaging device. Alternatively, the client device 30 may be a handheld device such as a smartphone or tablet PC.

[0038] In the diagnostic system 1 according to this embodiment, the learning device 10 acquires a dataset and trains the diagnostic model to determine a diagnostic model for use in diagnosis. When the diagnostic device receives an information request from a client, it uses the determined diagnostic model to acquire diagnostic information corresponding to the image to be diagnosed. The client device can request information from the diagnostic device and acquire the transmitted diagnostic information in response.

[0039] A diagnostic system according to another embodiment may include a diagnostic device and a client device that learn a diagnostic model and perform diagnostics using the same. A diagnostic system according to yet another embodiment may include a diagnostic device that learns a diagnostic model, obtains a diagnostic request and performs a diagnostic. A diagnostic system according to yet another embodiment may include a learning device that learns a diagnostic model and a diagnostic device that obtains a diagnostic request and performs a diagnostic.

[0040] The diagnostic system disclosed herein is not limited to the embodiments described above, but can be embodied in any form including a learning unit for learning a model, a diagnostic unit for acquiring diagnostic information according to the learned model, and an imaging unit for acquiring images to be diagnosed.

[0041] The following describes several embodiments of each device that constitutes the system.

[0042] 1.1.2.1. Learning Device

[0043] A learning device according to one embodiment of the present invention can train a diagnostic model that assists in diagnosis.

[0044] Figure 2 is a block diagram illustrating a learning device according to one embodiment of the present invention. Referring to Figure 2, the learning device 10 may include a processor 12 and a storage module 11.

[0045] The learning device 10 may include a processor 12. The processor 12 can control the operation of the learning device 10.

[0046] The processor 12 may include one or more of the following: a CPU (Central Processing Unit), RAM (Random Access Memory), a GPU (Graphics Processing Unit), one or more microprocessors, and other electronic components capable of processing input data according to a predetermined logic. Furthermore, the processor 12 may consist of at least one of these components.

[0047] The processor 12 can read system programs and various processing programs stored in the storage module 11. For example, the processor 12 can load processes and methods for performing diagnostics (described later) onto RAM and perform various processing according to the loaded programs. The processor 12 can also learn diagnostic models (described later).

[0048] The learning device 10 may include a storage module 11. The storage module 11 can store the data and learning models necessary for learning.

[0049] The storage module 11 can be embodied in non-volatile semiconductor memory, hard disk, flash memory, RAM, ROM (Read Only Memory), EEPROM (Electrically Erasable Programmable Read-Only Memory), or other tangible non-volatile recording media.

[0050] The storage module 11 can store various processing programs, parameters for processing these programs, or processing result data. For example, the storage module 11 can store data processing programs for performing diagnostics (described later), diagnostic process programs, parameters for executing each program, and data obtained in response to the execution of such programs (e.g., processed data or diagnostic result values). The storage module 11 can also store various diagnostic models (described later).

[0051] The learning device 10 may include a separate learning unit. The learning unit can perform training on a diagnostic model.

[0052] The learning unit may be included in the processor 12 described above. The learning unit may be stored in the storage module 11 described above. The learning unit may be realized by a part of the configuration of the processor 12 and the storage module 11 described above. For example, the learning unit may be stored in the storage module 11 and driven by the processor 12.

[0053] The learning device 10 may further include a communication module 13. The communication module 13 can communicate with external devices. For example, the communication module 13 can communicate with a diagnostic device, server device, or client device, which will be described later. The communication module 13 can communicate via wired or wireless connection. The communication module 13 can communicate bidirectionally or unidirectionally.

[0054] 1.1.2.2. Diagnostic equipment

[0055] The diagnostic device can acquire diagnostic information using a diagnostic model.

[0056] Figure 3 is a block diagram illustrating a diagnostic device according to one embodiment of the present invention. Referring to Figure 3, the diagnostic device 20 may include a processor 22 and a storage module 21.

[0057] The processor 22 may include one or more of the following: a CPU (Central Processing Unit), RAM (Random Access Memory), a GPU (Graphics Processing Unit), one or more microprocessors, and other electronic components capable of processing input data according to a predetermined logic. Furthermore, the processor 22 may consist of at least one or more such components.

[0058] The processor 22 can read system programs and various processing programs stored in the storage module 21. For example, the processor 22 can load processes and methods for performing the diagnosis described later onto RAM and perform various processing according to the loaded programs. The processor 22 can generate diagnostic information using a diagnostic model. The processor 22 can acquire diagnostic data for diagnosis (e.g., retinal data of the subject) and acquire diagnostic information predicted by the diagnostic data using a learned diagnostic model.

[0059] The storage module 21 can store diagnostic models. The storage module 21 can store parameters, variables, etc., of the diagnostic models.

[0060] The storage module 21 can be embodied in non-volatile semiconductor memory, hard disk, flash memory, RAM, ROM (Read Only Memory), EEPROM (Electrically Erasable Programmable Read-Only Memory), or other tangible non-volatile recording media.

[0061] The storage module 21 can store various processing programs, parameters for processing these programs, or processing result data. For example, the storage module 21 can store data processing program programs for performing diagnostics (described later), diagnostic process programs, parameters for executing each program, and data obtained in response to the execution of such programs (e.g., processed data or diagnostic result values). The storage module 21 can also store various diagnostic models (described later).

[0062] Although not shown in the diagram, the diagnostic device 20 may further include an input module. The input module can acquire user input. For example, the input module can acquire user input requesting diagnostic information. The input module can also acquire physical information of the subject (at least one of the following: height, weight, age, sex, race, smoking status, blood pressure (e.g., blood pressure value, presence or absence of hypertension), presence or absence of diabetes (or blood glucose level), cholesterol level). The processor 22 can acquire the information input through the input module.

[0063] The diagnostic device 20 may further include a communication module 23. The communication module 23 can communicate with a learning device and / or a client device. For example, the diagnostic device 20 may be configured as a server that communicates with a client device. This will be explained in more detail below.

[0064] 1.1.2.3. Server Equipment

[0065] According to one embodiment of the present invention, the diagnostic system may include a server device. The diagnostic system according to one embodiment of the present invention may also include a plurality of server devices.

[0066] The server device can store and / or run diagnostic models. The server device can store the weight values ​​that make up the learned diagnostic models. The server device can collect or store data used for diagnosis.

[0067] The server device can output the results of the diagnostic process using the diagnostic model to the client device. The server device can receive feedback from the client device. The server device can operate in the same way as the diagnostic device described above.

[0068] Figure 4 shows a diagnostic system according to one embodiment of the present invention. Referring to Figure 4, the diagnostic system 20 according to one embodiment of the present invention may include a diagnostic server 40, a learning device, and a client device.

[0069] The diagnostic server 40, i.e., the server device, can communicate with multiple learning devices or multiple diagnostic devices. Referring to Figure 6, the diagnostic server 40 can communicate with the first learning device 10a and the second learning device 10b. Referring to Figure 4, the diagnostic server 40 can communicate with the first client device 30a and the second client device 30b.

[0070] For example, the diagnostic server 40 can communicate with a first learning device 10a that trains a first diagnostic model for acquiring first diagnostic information, and a second learning device 10b that trains a second diagnostic model for acquiring second diagnostic information.

[0071] The diagnostic server 40 stores a first diagnostic model for acquiring first diagnostic information and a second diagnostic model for acquiring second diagnostic information, and can acquire diagnostic information in response to a request for diagnostic information acquisition from the first client device 30a or the second client device 30b, and transmit the acquired diagnostic information to the first client device 30a or the second client device 30b.

[0072] Alternatively, the diagnostic server 40 can communicate with a first client device 30a requesting first diagnostic information and a second client device 30b requesting second diagnostic information.

[0073] 1.1.2.4. Client Devices

[0074] The client device can request diagnostic information from the diagnostic device or server device. The client device can acquire the data necessary for diagnosis and transmit the acquired data to the diagnostic device.

[0075] Figure 5 is a block diagram illustrating a client device according to one embodiment of the present invention. Referring to Figure 5, the client device 30 according to one embodiment of the present invention may include an imaging module 31, a processor 32, and a communication module 33.

[0076] The imaging module 31 can acquire image or video data. The imaging module 31 can acquire retinal images. However, the client device 30 can be replaced with another form of data acquisition unit other than the imaging module 31.

[0077] The communication module 33 can communicate with external devices, such as diagnostic devices or server devices. The communication module 33 can communicate via wired or wireless communication.

[0078] The processor 32 can control the imaging module 31 to acquire images or data. The processor 32 can control the imaging module 31 to acquire retinal images. The processor 32 can transmit the acquired retinal images to a diagnostic device. The processor 32 can transmit the images acquired through the imaging module 31 to a server device via the communication module 33 and obtain diagnostic information generated based on these images.

[0079] Although not shown in the diagram, the client device may further include an output module. The output module may include a display that outputs video or images, or a speaker that outputs sound. The output module can output video or image data acquired by the imaging unit. The output module can output diagnostic information acquired from the diagnostic device.

[0080] Although not shown in the diagram, the client device may further include an input module. The input module can acquire user input. For example, the input module can acquire user input requesting diagnostic information. The input module can acquire user information to evaluate the diagnostic information acquired from the diagnostic device. The input module can also acquire physical information of the subject (at least one of the following: height, weight, age, sex, race, smoking status, blood pressure (e.g., blood pressure value, presence or absence of hypertension), presence or absence of diabetes (or blood glucose level), cholesterol level).

[0081] Although not shown in the diagram, the client device may further include a storage module. The storage module can store images acquired by the imaging unit.

[0082] 1.1.3. Overview of the Diagnostic Process

[0083] A diagnostic process can be performed by a diagnostic system or diagnostic device disclosed herein. The diagnostic process can be broadly considered as a learning process for learning a diagnostic model to be used for diagnosis, and a diagnostic process using the diagnostic model.

[0084] Figure 6 is a diagram illustrating a diagnostic process according to one embodiment of the present invention. Referring to Figure 6, the diagnostic process according to one embodiment of the present invention may include a learning process of acquiring and processing data (S11), learning a diagnostic model (S12), and acquiring parameters of the learned diagnostic model (S13), and a diagnostic process of acquiring data to be diagnosed (S21), and acquiring diagnostic information (S23) using a diagnostic model (S22) learned based on the data to be diagnosed.

[0085] More specifically, the learning process may include a data processing process that transforms the input learning image data into a state that can be used for training the model, and a learning process that trains the model using the transformed data. The learning process can be performed by the learning device described above.

[0086] The diagnostic process may include a data processing process that processes the input image data of the subject of examination to prepare it for diagnosis using a diagnostic model, and a diagnostic process that performs a diagnosis using the processed data. The diagnostic process may be performed by the diagnostic device or server device described above.

[0087] The following sections will explain each process.

[0088] 1.2. Learning Process

[0089] According to one embodiment of the present invention, a process for training a diagnostic model may be provided. Specifically, a process for training a diagnostic model that makes or assists in diagnosis based on retinal images may be initiated.

[0090] The learning process described below can be carried out by the learning device mentioned above.

[0091] 1.2.1. Learning Department

[0092] According to one embodiment of the present invention, the learning process may be carried out by a learning unit. The learning unit may be provided within the learning device described above.

[0093] Figure 7 is a diagram illustrating the configuration of a learning unit according to one embodiment of the present invention. Referring to Figure 7, the learning unit 100 may include a data processing module 110, a queue module 130, a learning module 150, and a learning result acquisition module 170. Each module can perform individual stages of the data processing process and the learning process, as will be described later. However, not all of the components and functions of each element described in Figure 7 are mandatory, and some elements may be added or omitted depending on the learning method.

[0094] 1.2.2. Data Processing Process

[0095] 1.2.2.1. Image Data Acquisition

[0096] According to one embodiment of the present invention, a dataset can be obtained. According to one embodiment of the present invention, a data processing module can obtain a dataset.

[0097] The dataset may also be an image dataset.

[0098] For example, a retinal image dataset may be used. A retinal image dataset can be acquired using a general non-mydriatic retinal camera, etc. The retinal images may be panoramic or wide retinal images. The retinal images may be red-free images. The retinal images may be infrared images. The retinal images may be autofluorescence images. The image data may be acquired in one of the following formats: JPG, PNG, DCM (DICOM), BMP, GIF, or TIFF.

[0099] As another example, the dataset may be an image dataset containing one of the following: OCT (Optical Coherence Tomography) images, OCT angiography images, or retinal angiography images. In this case, a diagnostic model trained using a dataset containing OCT images, OCT angiography images, or retinal angiography images can predict or output diagnostic information (or labels) based on the target OCT image, target OCT angiography image, or target retinal angiography image.

[0100] The dataset may include a training dataset. The dataset may include a test dataset. The dataset may include a validation dataset. In other words, the dataset can be assigned to at least one of the training dataset, test dataset, and validation dataset.

[0101] The dataset may be determined by considering the diagnostic information that the diagnostic model to be trained on through that dataset aims to acquire. For example, if the goal is to train a diagnostic model to acquire diagnostic information related to cataracts, the dataset to be acquired may be determined to be an infrared retinal image dataset. Alternatively, if the goal is to train a diagnostic model to acquire diagnostic information related to macular degeneration, the dataset to be acquired may be an autofluorescence retinal image dataset.

[0102] Individual data points in a dataset may include labels. There may be multiple labels. In other words, individual data points in a dataset may be labeled for at least one feature. For example, the dataset may be a retinal image dataset containing multiple retinal image data points, each of which may include diagnostic information labels (e.g., presence or absence of a specific disease) and / or findings information labels (e.g., presence or absence of an abnormality in a specific area) corresponding to the image.

[0103] As another example, the dataset may be a retinal image dataset, where each retinal image data may include peripheral information labels for that image. For example, each retinal image data may include peripheral information labels such as left / right eye information indicating whether the retinal image is from the left or right eye, gender information indicating whether it is from a female or male, and age information indicating the age of the subject from whom the retinal image was taken.

[0104] Figure 8 is a conceptual diagram illustrating an image dataset according to one embodiment of the present invention. Referring to Figure 8, the image dataset DS according to one embodiment of the present invention may include a plurality of image data IDs. Each image data ID may include an image I and a label L assigned to the image. Referring to Figure 10, the image dataset DS may include a first image data ID1 and a second image data ID2. The first image data ID1 may include a first image I1 and a first label L1 corresponding to the first image.

[0105] Figure 8 illustrates the case where one image data contains one label, but as mentioned above, one image data may contain multiple labels.

[0106] 1.2.2.2. Image Preprocessing

[0107] According to one embodiment of the present invention, image preprocessing can be performed. When images are used as input for training without any modifications, overfitting may occur as a result of learning unnecessary characteristics, and the training efficiency may decrease.

[0108] To prevent this, the data processing module can improve learning efficiency and performance by appropriately pre-processing image data to match the learning objectives.

[0109] In one embodiment, the data processing module can perform image preprocessing using or appropriately combining a variety of techniques, such as image resizing, grayscale conversion, histogram flattening, normalization, feature enhancement, image augmentation, noise reduction, boundary detection, segmentation, morphological operations, and color space conversion. The following describes image preprocessing using several techniques.

[0110] 1.2.2.2.1. Resizing Images

[0111] According to one embodiment of the present invention, the size of the acquired image data can be adjusted. That is, the image can be resized. According to one embodiment of the present invention, the image can be resized by the data processing module of the learning unit described above.

[0112] The size or aspect ratio of an image may be adjusted. Multiple acquired images may be resized to have a fixed size. Alternatively, an image may be resized to have a fixed aspect ratio. Resizing an image may also involve applying an image transformation filter to the image.

[0113] If the size or capacity of individual images acquired is excessively large or small, the image size or capacity can be adjusted to an appropriate size. Alternatively, if the size or capacity of individual images varies, the size or capacity can be standardized through resizing.

[0114] According to one embodiment, the image size can be adjusted. For example, if the image size exceeds an appropriate range, the image can be reduced in size through downsampling. Alternatively, if the image size falls below an appropriate range, the image can be enlarged through upsampling or interpolation.

[0115] In other embodiments, the size or aspect ratio of an image can be adjusted by cropping the image or by adding pixels to the acquired image. For example, if an image contains parts that are not needed for training, a portion of the image can be cropped to remove them. Alternatively, if a portion of the image is cropped and the aspect ratio is incorrect, a column or row can be added to adjust the aspect ratio of the image. In other words, the aspect ratio can be adjusted by adding margins or padding to the image.

[0116] In further embodiments, the image capacity and size or aspect ratio can be adjusted together. For example, if the image capacity is large, the image can be downsampled to reduce its capacity, and any unnecessary parts of the reduced image can be cropped to convert it into appropriate image data.

[0117] Furthermore, according to another embodiment of the present invention, the orientation of the image data can also be changed.

[0118] As a specific example, if a retinal image dataset is used as the dataset, each retinal image may have its capacity or size adjusted. Cropping can be performed to remove the margins of the retinal image excluding the retina, or padding can be performed to adjust the aspect ratio by filling in the cropped portion of the retinal image.

[0119] 1.2.2.2.2. Feature Emphasis

[0120] According to one embodiment of the present invention, the data processing module can perform preprocessing to enhance features of retinal images. For example, the data processing module can perform preprocessing on retinal images to facilitate the detection of abnormal signs of eye diseases or to enhance retinal blood vessels or blood flow changes.

[0121] As an example, image preprocessing may be performed on the resized image as described above. However, the invention disclosed herein is not limited to this, and preprocessing may be performed on the image without resizing. Adding image preprocessing may involve applying a preprocessing filter to the image.

[0122] According to one embodiment, a blur filter may be applied to the image. A Gaussian filter may be applied to the image. A Gaussian blur filter may also be applied to the image. Alternatively, a deblur filter may be applied to the image to sharpen it.

[0123] In other embodiments, filters may be applied to adjust or modulate the colors of an image. For example, filters may be applied to change the values ​​of some of the RGB values ​​that make up the image, or to binarize the image.

[0124] In further embodiments, filters may be applied to the image to emphasize specific elements. For example, retinal image data may be preprocessed to emphasize vascular elements from each image. In this case, the preprocessing to emphasize vascular elements may involve applying one or more filters sequentially or in combination.

[0125] Furthermore, according to one embodiment of the present invention, image preprocessing may be performed taking into consideration the characteristics of the diagnostic information to be acquired. For example, when acquiring diagnostic information related to findings such as retinal hemorrhage, drusen, microaneurysms, and exudates, preprocessing can be performed to convert the acquired retinal image into a red-free retinal image format.

[0126] 1.2.2.2.3. Image Augmentation

[0127] According to one embodiment of the present invention, images can be enhanced or augmented. Image enhancement can be performed by the data processing module of the learning unit described above.

[0128] Augmented images can be used to improve the training performance of diagnostic models. For example, if there is insufficient data for training a diagnostic model, the amount of training data can be increased by modulating existing training image data and using the modulated (or altered) images together with the original images. This suppresses overfitting, allows for deeper model layers, and improves prediction accuracy.

[0129] For example, image data can be augmented by flipping the image horizontally, cropping parts of the image, correcting the color values ​​of the image, or adding artificial noise. Specifically, cropping parts of an image can be done by cutting out a portion of the elements that make up the image, or by randomly cropping parts of the image. More examples include augmenting image data by flipping it horizontally, flipping it vertically, resizing it by a fixed ratio, cropping, padding, adjusting the color, or adjusting the brightness.

[0130] Furthermore, in one embodiment, the data processing module can rotate the retinal image to enhance it. Because the retina is circular in shape, there may be no data loss even when the retina rotates. As a result, when enhancing the image by rotating the retinal image, multiple retinal images can be acquired without data loss. For example, the data processing module can rotate the retinal image based on a predetermined angle (e.g., 15 degrees, 30 degrees, 60 degrees, etc.) to acquire multiple retinal images.

[0131] Furthermore, in one embodiment, the image data enhancement or expansion described above may generally be applied to the training dataset. However, it may also be applied to other datasets, such as the test dataset, i.e., a dataset for testing a model that has been trained using the training data and validated using the validation data.

[0132] As a concrete example, when a retinal image dataset is used as the dataset, an augmented retinal image dataset can be obtained by randomly applying one or more processes such as inverting, cropping, adding noise, or changing the colors of the images in order to increase the number of data points.

[0133] 1.2.2.3. Image serialization

[0134] According to one embodiment of the present invention, image data can be linearized. The image can be linearized by the data processing module of the learning unit described above. The serialization module can serialize the preprocessed image data and transmit it to the queue module.

[0135] When image data is used directly for training, decoding is necessary because the image data exists in image file formats such as JPG, PNB, and DCM. However, decoding every time training is performed can degrade the model training performance. Therefore, instead of using the image files directly for training, they can be serialized before training. Thus, image data can be serialized to improve training performance and speed. The image data to be serialized may be image data to which one or more of the above-mentioned image resizing and image preprocessing steps have been applied, or it may be image data that has not been processed in either way.

[0136] Each image data file in an image dataset can be converted to a string format. The image data can also be converted to a binarized data format. In particular, the image data can be converted to a data format suitable for use in training diagnostic models. For example, the image data can be converted to a TFRecord format for use in training diagnostic models using TensorFlow.

[0137] As a concrete example, if a set of retinal images is used as the dataset, the acquired retinal image set can be converted into TFRecord format and used to train a diagnostic model.

[0138] 1.2.2.4. Queue

[0139] A queue can be used to resolve data bottlenecks. The queue module in the learning unit described above can store image data in the queue and transmit it to the learning model module.

[0140] In particular, when using both a CPU (Central Processing Unit) and a GPU (Graphics Processing Unit) to advance the learning process, utilizing queues can minimize bottlenecks between the CPU and GPU, streamline database access, and improve memory usage efficiency.

[0141] The queue can store data used to train a diagnostic model. The queue can store image data. The image data stored in the queue may be image data that has been processed by at least one of the data processing processes described above (i.e., resizing, preprocessing, and augmentation), or it may be the image in its original state as acquired.

[0142] The queue can store image data, preferably serialized image data as described above. The queue can store image data and supply it to a diagnostic model. The queue can deliver image data to the diagnostic model in batch sizes.

[0143] The queue can provide image data. The queue can also provide data to the learning module, which will be described later. As data is extracted by the learning module, the amount of data stored in the queue may decrease.

[0144] As the diagnostic model's learning progresses, if the number of data points stored in the queue falls below a certain threshold, the queue can request data replenishment. The queue can request replenishment of specific types of data. When the learning unit receives a request for data replenishment, it can replenish the queue with data.

[0145] Queues can be located in the system memory of a learning device. For example, a queue can be formed in the RAM (Random Access Memory) of a central processing unit (CPU). In this case, the size, or capacity, of the queue can be determined according to the RAM capacity of the CPU. As the queue, a First In First Out (FIFO) queue, a Primary Queue, or a random queue may be used.

[0146] 1.2.3. Learning Process

[0147] According to one embodiment of the present invention, the learning process of the diagnostic model can be initiated.

[0148] According to one embodiment of the present invention, the learning of the diagnostic model may be performed by the learning device described above. The learning process may be performed by the processor of the learning device described above. The learning process may be performed by the learning module of the learning unit described above.

[0149] Figure 9 is a block diagram illustrating the learning process of a diagnostic model according to one embodiment of the present invention. Referring to Figure 9, the learning process of a diagnostic model according to one embodiment of the present invention may be carried out by acquiring data (S31), learning a diagnostic model (S32), verifying the learned model (S33), and acquiring the variables of the learned model (S34).

[0150] 1.2.3.1. Data Entry

[0151] A dataset for training a diagnostic model can be obtained.

[0152] The acquired data may be an image dataset processed by the data processing process described above. For example, the dataset may include retinal image data that has been resized, preprocessed with filters, augmented, and then serialized.

[0153] During the training phase of the diagnostic model, a training dataset may be acquired and used. During the validation phase of the diagnostic model, a validation dataset may be acquired and used. During the testing phase of the diagnostic model, a test dataset may be acquired and used. Each dataset may include retinal images and labels.

[0154] Datasets can be retrieved from a queue. Datasets can be retrieved from the queue in batch sizes. For example, if a batch size of 60 is specified, datasets can be extracted from the queue in batches of 60. The batch size may be limited by the GPU's RAM capacity.

[0155] Datasets can be randomly selected from a queue and used for training modules. Alternatively, datasets can be selected in the order they were stored in the queue.

[0156] The training module can extract datasets from a queue by specifying their composition. For example, the training module can extract retinal images with left eye labels and retinal image data with right eye labels from a particular subject so that they are used together for training.

[0157] The learning module can retrieve datasets with specific labels from a queue. For example, the learning module can retrieve retinal image datasets with abnormal diagnostic information labels from the queue. The learning module can also retrieve datasets from a queue by specifying the ratio of data points to each label. For example, the learning module can retrieve retinal image datasets from the queue such that the number of retinal image data points with abnormal diagnostic information labels is one-to-one with the number of retinal image data points with normal diagnostic information labels.

[0158] 1.2.3.2. Model Design

[0159] Diagnostic models can be designed as a variety of models, such as neural network models and machine learning models. In one embodiment, if the diagnostic model includes a neural network model, the neural network model may include multiple layers or hierarchies.

[0160] Neural network models can be implemented in the form of classifiers that generate diagnostic information. These classifiers can perform binary or multi-classification. For example, a neural network model may be a binary classification model that classifies input data into normal or abnormal classes based on target diagnostic information such as a specific disease or abnormal symptom. Alternatively, a neural network model may be a multi-classification model that classifies input data into multiple grade classes based on specific characteristics (e.g., disease progression). Or, a neural network model can be implemented as a regression model that generates specific numerical values ​​associated with a particular disease.

[0161] The neural network model may include a Convolutional Neural Network (CNN). At least one of the following CNN structures may be used: AlexNet, LENET, NIN, VGGNet, ResNet, WideResNet, GoogleNet, FractaNet, DenseNet, FitNet, RitResNet, HighwayNet, MobileNet, and DeeplySupervisedNet. The neural network model may be implemented using multiple CNN structures.

[0162] As an example, a neural network model can be implemented to include multiple VGGNet blocks. More specifically, a neural network model can be provided by combining a first structure in which a CNN layer with 64 filters of 3x3 size, a BN (Batch Normalization) layer, and a ReLU layer are sequentially connected, and a second block in which a CNN layer with 128 filters of 3x3 size, a ReLU layer, and a BN layer are sequentially connected.

[0163] The neural network model may include a max pooling layer following each CNN block, and may also include a GAP (Global Average pooling) layer, an FC (Fully Connected) layer, and an activation layer (e.g., sigmoid, softmax, etc.) at the end.

[0164] In other embodiments, if the diagnostic model includes a machine learning model, the machine learning model may include a linear regression model, a Cox proportional hazards model, and the like.

[0165] 1.2.3.3. Model Training

[0166] Diagnostic models can be trained using training datasets.

[0167] Diagnostic models can be trained using labeled datasets. However, the training process for diagnostic models described herein is not limited to this; diagnostic models can also be trained in an uninstructed manner using unlabeled data.

[0168] Diagnostic model training can be performed by obtaining result values ​​using a diagnostic model with arbitrary weight values ​​based on training image data, comparing the obtained result values ​​with the label values ​​of the training data, performing backpropagation according to the error, and optimizing the weight values. Furthermore, the training of the diagnostic model can be influenced by the model validation results, test results, and / or feedback from the diagnostic phase, as described later.

[0169] The diagnostic model described above can be trained using TensorFlow. However, the present invention is not limited to this, and frameworks such as Theano, Keras, Caffe, Torch, and CNTK (Microsoft Cognitive Toolkit) can also be used to train the diagnostic model.

[0170] 1.2.3.4. Model Validation

[0171] A diagnostic model can be validated using a validation dataset. Validation of the diagnostic model can be performed by obtaining result values ​​for the validation dataset from the trained diagnostic model and comparing these result values ​​with the labels on the validation dataset. Validation can also be performed by measuring the accuracy of the result values. Depending on the validation results, the parameters (e.g., weights and / or biases) or hyperparameters (e.g., learning rate) of the diagnostic model may be adjusted.

[0172] As an example, a learning device according to one embodiment of the present invention can train a diagnostic model that predicts diagnostic information based on retinal images, and can validate the learned model by comparing the diagnostic information for the retinal image with a validation label corresponding to the retinal image.

[0173] To validate a diagnostic model, an external validation set—that is, a dataset containing distinguishable factors not included in the training dataset—may be used. For example, the external validation set may be a dataset in which factors such as race, environment, age, and sex are distinguished from the training dataset.

[0174] 1.2.3.5. Model Testing

[0175] Diagnostic models can be tested using test datasets.

[0176] According to a learning process in one embodiment of the present invention, a diagnostic model can be tested using a test dataset distinct from a training dataset and a validation dataset. Depending on the test results, the parameters (e.g., weights and / or biases) or hyperparameters (e.g., learning rate) of the diagnostic model may be adjusted.

[0177] As an example, a learning device according to one embodiment of the present invention can test a learned and validated diagnostic model by obtaining result values ​​from a diagnostic model that has been trained to predict diagnostic information based on retinal images, using test retinal image data that has not been used for training and validation as input.

[0178] Testing a diagnostic model may utilize an external validation set, i.e., a dataset containing factors distinct from the training and / or validation data.

[0179] 1.2.3.6. Output of Results

[0180] As a result of training the diagnostic model, optimized model parameter values ​​can be obtained. As described above, more appropriate parameter (or variable) values ​​can be obtained by iteratively training the model using the test dataset. Once training has progressed sufficiently, optimized values ​​for weights and / or biases can be obtained.

[0181] According to one embodiment of the present invention, a learned diagnostic model and / or the parameters or variables of the learned diagnostic model may be stored in a learning device and / or a diagnostic device (or server). The learned diagnostic model may be used by the diagnostic device and / or client device, etc., to predict diagnostic information. Furthermore, the parameters or variables of the learned diagnostic model may also be updated by feedback obtained from the diagnostic device or client device.

[0182] 1.2.3.7. Model Ensemble

[0183] According to one embodiment of the present invention, multiple submodels can be learned simultaneously during the process of learning one diagnostic model. The multiple submodels may have different hierarchical structures from one another.

[0184] In this case, the diagnostic model according to one embodiment of the present invention can be realized by combining multiple sub-diagnostic models. In other words, the diagnostic model can be trained using an ensemble technique that combines multiple sub-diagnoses.

[0185] When forming an ensemble to construct a diagnostic model, predictions can be made by integrating the results predicted from various forms of sub-diagnostic models, thereby improving the accuracy of result predictions.

[0186] 1.3. Diagnostic Process

[0187] According to one embodiment of the present invention, a diagnostic process (or diagnostic process) may be provided that acquires diagnostic information using a diagnostic model. Specifically, the diagnostic process may utilize retinal images and predict diagnostic information (e.g., diagnostic information or findings information) through a learned diagnostic model.

[0188] The diagnostic process described below can be performed by a diagnostic device.

[0189] 1.3.1. Diagnostic Department

[0190] According to one embodiment of the present invention, the diagnostic process may be performed by the processor of the diagnostic device described above. The processor may be provided within the diagnostic device described above.

[0191] Figure 10 is a diagram illustrating the configuration of a diagnostic unit according to one embodiment of the present invention. Referring to Figure 10, the diagnostic unit 200 may include a diagnostic request acquisition module 210, a data processing module 230, a diagnostic module 250, and an output module 270.

[0192] Each module can perform individual stages of the data processing and learning processes, as described later. However, not all of the components and functions described in Figure 10 are mandatory; some components may be added, and some may be omitted, depending on the nature of the diagnosis.

[0193] 1.3.2. Data Acquisition and Diagnostic Request

[0194] A diagnostic device according to one embodiment of the present invention can acquire data to be diagnosed and acquire diagnostic information based on this data. The data to be diagnosed may be image data. Data acquisition and acquisition of diagnostic requests can be performed by the diagnostic request acquisition module of the diagnostic unit described above.

[0195] For example, diagnostic data (TD) may include diagnostic images (TI) and diagnostic subject information (PI; patient information).

[0196] The diagnostic image (TI) may be an image used to obtain diagnostic information about the subject being diagnosed. For example, the diagnostic image may be a retinal image and / or a retinal image. The diagnostic image (TI) may be in one of the following formats: JPG, PNG, DCM (DICOM), BMP, GIF, or TIFF.

[0197] Diagnostic object information (PI) may be information for identifying the object to be diagnosed. Alternatively, diagnostic object information (PI) may be characteristic information of the object to be diagnosed or an image. For example, diagnostic object information (PI) may include information such as the date and time the image to be diagnosed was taken, the equipment used for taking the image, the identification number, ID, name, gender, age, weight, race, smoking status, blood pressure (presence or absence of hypertension), and presence or absence of diabetes of the subject to be diagnosed. If the image to be diagnosed is a retinal image, the diagnostic object information (PI) may further include eye-related information such as binocular information indicating whether it is the left eye or the right eye.

[0198] The diagnostic device can acquire diagnostic requests. Along with the diagnostic request, the diagnostic device can acquire data to be diagnosed. Upon receiving a diagnostic request, the diagnostic device can acquire diagnostic information using a learned diagnostic model. The diagnostic device can acquire diagnostic requests from client devices. Alternatively, the diagnostic device can acquire diagnostic requests from users through separately provided input means.

[0199] 1.3.3. Data Processing Process

[0200] The acquired data can be processed. Data processing can be performed by the data processing module of the diagnostic unit described above.

[0201] The data processing process can generally be carried out in the same way as the data processing process in the learning process described above. Below, we will explain the data processing process in the diagnostic process, focusing on the differences from the data processing process in the learning process.

[0202] In the diagnostic process, the diagnostic device can acquire data in the same way as in the learning process. The acquired data may be in the same format as the data acquired in the learning process. For example, if the learning device trained a diagnostic model using DCM-formatted image data in the learning process, the diagnostic device can acquire DCM images and use the trained diagnostic model to obtain diagnostic information.

[0203] In the diagnostic process, the acquired diagnostic images may be resized, similar to the image data used in the training process. The diagnostic images may be reshaped to have an appropriate capacity, size, and / or aspect ratio in order to efficiently predict diagnostic information through the trained diagnostic model.

[0204] For example, if the image being diagnosed is a retinal image, resizing may be performed, such as cropping unnecessary parts of the image or reducing its size, in order to predict diagnostic information based on the retinal image.

[0205] In the diagnostic process, preprocessing filters may be applied to the acquired diagnostic images, similar to the image data used in the training process. Appropriate filters may be applied to the diagnostic images to further improve the accuracy of diagnostic information prediction through the trained diagnostic model.

[0206] For example, if the image to be diagnosed is a retinal image, pre-processing to facilitate the prediction of diagnostic information may be applied to the image, such as image pre-processing to enhance blood vessels or image pre-processing to enhance or weaken specific colors.

[0207] In the diagnostic process, the acquired diagnostic images can be serialized, similar to the image data used in the learning process. The diagnostic images can be converted or serialized into a format that facilitates the driving of the diagnostic model within a specific workframe.

[0208] Serialization of the images to be diagnosed may be omitted. This is because, unlike the learning phase, the amount of data the processor processes at once is not large during the diagnostic phase, and therefore the burden on data processing speed is relatively low.

[0209] In the diagnostic process, the acquired images to be diagnosed can be stored in a queue, similar to the image data used in the learning process. However, since the amount of data processed in the diagnostic process is less than in the learning process, the step of storing the data in a queue can be omitted.

[0210] On the other hand, since the diagnostic process does not require an increase in the amount of data, unlike the learning process, data augmentation or image enhancement procedures do not need to be used to obtain accurate diagnostic information.

[0211] 1.3.4. Diagnostic Process

[0212] According to one embodiment of the present invention, a diagnostic process using a learned diagnostic model can be initiated. The diagnostic process can be performed by the diagnostic device described above. The diagnostic process can be performed by the diagnostic server described above. The diagnostic process can be performed by the control unit of the diagnostic device described above. The diagnostic process can be performed by the diagnostic module of the diagnostic unit described above.

[0213] Figure 11 is a diagram illustrating a diagnostic process according to one embodiment of the present invention. Referring to Figure 11, the diagnostic process may be performed by acquiring data to be diagnosed (S31), using a learned diagnostic model (S42), and obtaining results corresponding to the acquired data to be diagnosed (S43). However, data processing may be performed selectively.

[0214] The following describes each stage of the diagnostic process with reference to Figure 11.

[0215] 1.3.4.1. Data Entry

[0216] According to one embodiment of the present invention, a diagnostic module can acquire diagnostic data. The acquired data may be processed as described above. For example, the acquired data may be retinal image data of a subject to which preprocessing has been applied to adjust the size and enhance the blood vessels. According to one embodiment of the present invention, left eye and right eye images of a single subject may be input together as diagnostic data.

[0217] 1.3.4.2. Data Classification

[0218] A diagnostic model configured as a classifier can classify input diagnostic images into positive or negative classes based on predetermined labels.

[0219] A trained diagnostic model can take data to be diagnosed as input and output predicted labels. The trained diagnostic model can output predicted values ​​for diagnostic information. Diagnostic information can be obtained using the trained diagnostic model. Diagnostic information can be determined based on the predicted labels.

[0220] For example, a diagnostic model can predict diagnostic information (i.e., information about the presence or absence of disease) or findings (i.e., information about the presence or absence of abnormal findings) for an eye disease or systemic disease in a subject. In this case, the diagnostic information or findings may be output in probabilistic form. For example, the probability that the subject has a specific disease or the probability that there is a specific abnormal finding in the subject's retinal image may be output. When using a diagnostic model provided in classifier form, the predicted label may be determined by considering whether the output probability value (or predicted score) exceeds a threshold.

[0221] As a concrete example, a diagnostic model can use the subject's retinal image as the diagnostic target image and output a probability value indicating whether or not the subject has diabetic retinopathy. When using a diagnostic model with a classifier where 1 is considered normal, the subject's retinal photograph is input into the diagnostic model, and the probability values ​​of normal:abnormal for the presence or absence of diabetic retinopathy can be obtained in a format such as 0.74:0.26.

[0222] Here, the explanation has been based on the case where data is classified using a classifier-type diagnostic model, but the present invention is not limited to this, and it is also possible to predict specific diagnostic values ​​(for example, blood pressure) using a diagnostic model embodied in a regression model form.

[0223] According to another embodiment of the present invention, image suitability information can be obtained. The suitability information can indicate whether the image to be diagnosed is suitable for obtaining diagnostic information using a diagnostic model.

[0224] Image suitability information may also be image quality information. Quality information or suitability information can indicate whether the image being diagnosed meets a standard level.

[0225] For example, if an image to be diagnosed has defects due to a fault in the imaging equipment or the influence of lighting during imaging, a non-conformity result may be output as conformity information for that image. If an image to be diagnosed contains noise above a certain level, that image may be judged as non-conformity.

[0226] The suitability information may be values ​​predicted using a diagnostic model. Alternatively, the suitability information may be information obtained through a separate image analysis process.

[0227] According to one embodiment, even if an image is classified as unsuitable, diagnostic information obtained based on the unsuitable image can still be acquired.

[0228] According to one embodiment, images classified as unsuitable can be re-examined by a diagnostic model.

[0229] In this case, the diagnostic model used for re-examination may be different from the diagnostic model used for the initial examination. For example, the diagnostic device may store a first diagnostic model and a second diagnostic model, and images classified as unsuitable through the first diagnostic model may be examined through the second diagnostic model.

[0230] In yet another embodiment of the present invention, a map may be obtained from a learned diagnostic model. Diagnostic information may include a map. The map may be obtained together with other diagnostic information. For example, the map may include a saliency map, a class activation map (CAM), a heat map, etc. In the case of a CAM, it may be obtained selectively. For example, in the case of a CAM, the CAM may be extracted and / or output if the diagnostic information or findings obtained by the diagnostic model are classified into an abnormal class.

[0231] 1.3.5. Output of diagnostic information

[0232] Diagnostic information may be determined based on the results output by the diagnostic model. The output of diagnostic information may be performed by the output module of the diagnostic unit described above. Diagnostic information may be output from the diagnostic device to the client device. Diagnostic information may be output from the diagnostic device to the server device. Diagnostic information may be stored in the diagnostic device or diagnostic server. Diagnostic information may be stored in a separately provided server device, etc.

[0233] Diagnostic information can be stored and managed in a database. For example, acquired diagnostic information can be stored and managed together with the diagnostic images of the subject, according to the subject's identification number. In this case, the diagnostic images and diagnostic information of the subject can be managed in chronological order. Managing diagnostic information and diagnostic images chronologically can facilitate the tracking and historical management of individual diagnostic information.

[0234] Diagnostic information may be provided to the user. Diagnostic information may be provided to the user through output means of the diagnostic device or client device. Diagnostic information may be output in a way that the user can perceive through visual or auditory output means provided on the diagnostic device or client device.

[0235] According to one embodiment of the present invention, an interface may be provided for effectively providing diagnostic information to the user.

[0236] Furthermore, if an image is classified as unsuitable, suitability information for that image may be provided along with it. For example, if an image is classified as unsuitable, diagnostic information and unsuitability determination information obtained in relation to that image may be provided together.

[0237] Images deemed unsuitable for diagnosis may be classified as images requiring re-imaging. In this case, guidance for re-imaging may be provided for the subject of the image classified as requiring re-imaging, along with suitability information. On the other hand, in response to providing diagnostic information acquired through the diagnostic model, feedback related to the learning of the diagnostic model may be obtained. For example, feedback may be obtained to adjust parameters or hyperparameters related to the learning of the diagnostic model. The feedback may be obtained through an input module provided in the diagnostic device or client device.

[0238] According to one embodiment of the present invention, the diagnostic information corresponding to the image to be diagnosed may include grade information. The grade information may be selected from among a plurality of grades. The grade information may be determined based on the diagnostic information and / or findings information obtained through the diagnostic model. The grade information may be determined by considering the suitability information or quality information of the image to be diagnosed. If the diagnostic model is a classifier model that performs multiple classifications, the grade information may be determined by considering the class into which the image to be diagnosed has been classified by the diagnostic model. If the diagnostic model is a regression model that generates numerical values ​​associated with a specific disease, the grade information may be determined by considering the output numerical values. In addition, the grade information may be determined by applying a predetermined cutoff value to the score corresponding to the diagnostic information.

[0239] For example, diagnostic information acquired in response to an image to be diagnosed may include either first-grade information or second-grade information, selected from the two. Grade information may be selected as first-grade information if abnormal findings or abnormal diagnostic information is acquired through the diagnostic model. Grade information may be selected as second-grade information if abnormal findings or abnormal diagnostic information is not acquired through the diagnostic model. Alternatively, grade information may be selected as first-grade information if the value acquired through the diagnostic model exceeds a standard value, and as second-grade information if the acquired value falls below the standard value. First-grade information can indicate the presence of stronger abnormal information in the image to be diagnosed compared to second-grade information.

[0240] On the other hand, grade information may be selected as third-grade information if, using image analysis or a diagnostic model, it is determined that the quality of the image to be diagnosed is below a certain standard. Alternatively, the diagnostic information may include third-grade information together with first or second-grade information.

[0241] 1.4. Diagnostic system using multiple diagnostic models

[0242] According to one embodiment of the present invention, diagnostic information can be output using the diagnostic model as described above. The diagnostic model described above may consist of one diagnostic model or multiple diagnostic models. When the diagnostic model consists of multiple diagnostic models, the multiple diagnostic models may be configured in parallel or in series. The parallel diagnostic model and the series diagnostic model will be described below.

[0243] 1.4.1.1. Configuration of the Parallel Diagnostic System

[0244] According to one embodiment of the present invention, a parallel diagnostic system for acquiring multiple diagnostic information can be provided. The parallel diagnostic system can train multiple diagnostic models for acquiring multiple diagnostic information, and acquire multiple diagnostic information using the trained multiple diagnostic models.

[0245] For example, a parallel diagnostic system can train a first diagnostic model that acquires first diagnostic information related to the presence or absence of eye disease in a subject based on retinal images, and a second diagnostic model that acquires second diagnostic information related to the presence or absence of systemic disease in a subject. Using the trained first and second diagnostic models, it can output diagnostic information regarding the presence or absence of eye disease and systemic disease in a subject.

[0246] Multiple diagnostic models can be trained in parallel and / or independently. By training multiple diagnostic models to predict different labels in this way, the prediction accuracy for each label can be improved, and the efficiency of the prediction operation can be increased. Since the previously described concepts apply to training multiple diagnostic models, a detailed explanation will be omitted.

[0247] 1.4.1.2. Parallel Diagnostic Process

[0248] According to one embodiment of the present invention, a diagnostic process for acquiring multiple diagnostic pieces of information may be provided. The diagnostic process for acquiring multiple diagnostic pieces of information may be embodied in the form of a parallel diagnostic process that includes multiple independent diagnostic processes.

[0249] According to one embodiment of the present invention, the diagnostic process may be performed by multiple diagnostic modules. Each diagnostic process may be performed independently.

[0250] Figure 12 is a block diagram illustrating a diagnostic unit according to one embodiment of the present invention.

[0251] Referring to Figure 12, the diagnostic unit 200 according to one embodiment of the present invention may include a diagnostic request acquisition module 211, a data processing module 231, a first diagnostic module 251, a second diagnostic module 253, and an output module 271. Unless otherwise specified, each module of the diagnostic unit 200 can operate in the same manner as the diagnostic modules of the diagnostic unit shown in Figure 10.

[0252] In Figure 12, the diagnostic request acquisition module 211, data processing module 231, and output module 271 are shown as common even when the diagnostic unit 200 includes multiple diagnostic modules. However, the present invention is not limited to this configuration, and multiple diagnostic request acquisition modules, data processing modules, and / or output modules can also be provided. Multiple diagnostic request acquisition modules, data processing modules, and / or output modules can also operate in parallel.

[0253] For example, the diagnostic unit 200 includes a first data processing module that performs a first processing on the input image to be diagnosed and a second processing module that performs a second data processing on the image to be diagnosed. The first diagnostic module can acquire first diagnostic information based on the first processed image to be diagnosed, and the second diagnostic module can acquire second diagnostic information based on the second processed image to be diagnosed. The first processing and / or second processing may be any one selected from image resizing, image color modulation, blur filter application, blood vessel enhancement processing, red-free conversion, cropping of a portion of an area, or extraction of a portion of an element.

[0254] Multiple diagnostic modules can acquire different diagnostic information from each other. Multiple diagnostic modules can acquire diagnostic information using different diagnostic models from each other. For example, the first diagnostic module can acquire first diagnostic information related to whether or not the subject has an eye disease using a first diagnostic model that predicts whether or not the subject has an eye disease, and the second diagnostic module can acquire second diagnostic information related to whether or not the subject has a systemic disease using a second diagnostic model that predicts whether or not the subject has a systemic disease.

[0255] As a more specific example, the first diagnostic module can obtain first diagnostic information regarding whether or not a subject has diabetic retinopathy using a first diagnostic model that predicts whether or not the subject has diabetic retinopathy based on retinal images, and the second diagnostic module can obtain second diagnostic information related to whether or not the subject has hypertension using a second diagnostic model that predicts whether or not the subject has hypertension based on retinal images.

[0256] Furthermore, the diagnostic process according to one embodiment of the present invention may include a plurality of sub-diagnostic processes. Each sub-diagnostic process may be performed using a different diagnostic model. Each sub-diagnostic process may be performed using a different diagnostic process. For example, the first diagnostic module may perform a first sub-diagnostic process to acquire first diagnostic information through the first diagnostic model. Alternatively, the second diagnostic module may perform a second sub-diagnostic process to acquire second diagnostic information through the second diagnostic model.

[0257] Multiple trained diagnostic models can take the data to be diagnosed as input and output a predicted label or probability. Each diagnostic model is configured as a classifier and can classify the input data to be diagnosed according to a predetermined label. In this case, multiple diagnostic models may be configured as classifiers trained on different characteristics from one another.

[0258] On the other hand, maps can be obtained from each diagnostic model. Maps can be obtained selectively. Maps can be extracted when predetermined conditions are met. For example, a first map can be obtained from a first diagnostic model if the first diagnostic information indicates that the subject is abnormal for a first characteristic.

[0259] Figure 13 is a diagram illustrating a diagnostic process according to one embodiment of the present invention.

[0260] Referring to Figure 13, the diagnostic process according to one embodiment of the present invention may include acquiring data to be diagnosed (S51), and acquiring diagnostic information corresponding to the data to be diagnosed using a first diagnostic model and a second diagnostic model (S51a, S51b) (S53). The data to be diagnosed may be processed data.

[0261] A diagnostic process according to one embodiment of the present invention may include acquiring first diagnostic information through a learned first diagnostic model and acquiring second diagnostic information through a learned second diagnostic model. The first and second diagnostic models can each acquire first and second diagnostic information based on the same diagnostic target data.

[0262] For example, the first diagnostic model and the second diagnostic model can obtain first diagnostic information regarding the subject's eye disease and second diagnostic information regarding the presence or absence of the subject's kidney disease, respectively, based on the retinal image being diagnosed.

[0263] Unless otherwise specified, the diagnostic process described in relation to FIG. 13 can be implemented in the same manner as the diagnostic process described above in relation to FIG. 11.

[0264] 1.4.1.3. Output of Diagnostic Information

[0265] According to an embodiment of the present invention, diagnostic information obtained by a parallel diagnostic process can be acquired. The acquired diagnostic information can be stored in a diagnostic device, a server device, and / or a client device. The acquired diagnostic information can be transmitted to an external device.

[0266] The plurality of diagnostic information can respectively indicate a plurality of labels predicted by a plurality of diagnostic models. The plurality of diagnostic information can respectively correspond to a plurality of labels predicted by a plurality of diagnostic models. Alternatively, the diagnostic information may be information determined based on a plurality of labels predicted by a plurality of diagnostic models. The diagnostic information can correspond to a plurality of labels predicted by a plurality of diagnostic models.

[0267] In other words, the first diagnostic information may be diagnostic information corresponding to the first label predicted through the first diagnostic model. Alternatively, the first diagnostic information may be diagnostic information determined by considering together the first label predicted through the first diagnostic model and the second label predicted through the second diagnostic model.

[0268] On the other hand, an image of a map obtained from a plurality of diagnostic models can be output. The map image can be output when a predetermined condition is satisfied. For example, when the first diagnostic information indicates that the subject is abnormal with respect to the first characteristic or when the second diagnostic information indicates that the subject is abnormal with respect to the second characteristic, a map image obtained from the diagnostic model that outputs the diagnostic information indicated as abnormal can be output.

[0269] Multiple diagnostic information and / or map images may be provided to the user. These diagnostic information items may be provided to the user through the output means of the diagnostic device or client device.

[0270] According to one embodiment of the present invention, the diagnostic information corresponding to the image to be diagnosed may include grade information. The grade information may be selected from among a plurality of grades. The grade information may be determined based on a plurality of diagnostic information and / or findings information obtained through a diagnostic model. The grade information may be determined by considering the suitability information or quality information of the image to be diagnosed. The grade information may be determined by considering the class into which the image to be diagnosed is classified by a plurality of diagnostic models. The grade information may be determined by considering numerical values ​​output from a plurality of diagnostic models.

[0271] For example, diagnostic information acquired in response to an image being diagnosed may include either first-grade information or second-grade information, selected from the available options. Grade information may be selected as first-grade information if at least one abnormal finding or abnormal diagnosis is obtained from the diagnostic information acquired through multiple diagnostic models. Grade information may be selected as second-grade information if no abnormal finding or abnormal diagnosis is obtained from the diagnostic information acquired through the diagnostic models.

[0272] Grade information can be selected as Grade 1 if at least one of the values ​​obtained through the diagnostic model exceeds the reference value, or as Grade 2 if all of the obtained values ​​fall below the reference value. Grade 1 information indicates that there is stronger abnormality information in the diagnostic image compared to Grade 2 information.

[0273] Grade information may be selected as third-grade information if, using image analysis or a diagnostic model, it is determined that the quality of the image being diagnosed is below a certain standard. Alternatively, the diagnostic information may include third-grade information together with first or second-grade information.

[0274] Furthermore, the diagnostic system according to one embodiment of the present invention may include a retinal image acquisition unit, a first processing unit, a second processing unit, a third processing unit, and a diagnostic information output unit.

[0275] According to one embodiment of the present invention, the diagnostic system may include a diagnostic device. The diagnostic device may include a retinal image acquisition unit, a first processing unit, a second processing unit, a third processing unit, and / or a diagnostic information output unit. However, the present invention is not limited thereto, and each part included in the diagnostic system may be located in an appropriate position on the learning device, the diagnostic device, the learning diagnostic server, and / or the client device. For convenience, the following description will be based on the case where the diagnostic device of the diagnostic system includes a retinal image acquisition unit, a first processing unit, a second processing unit, a third processing unit, and a diagnostic information output unit.

[0276] Figure 14 is a diagram illustrating a diagnostic system according to one embodiment of the present invention. Referring to Figure 14, the diagnostic system includes a diagnostic device, which may include a retinal image acquisition unit, a first processing unit, a second processing unit, a third processing unit, and a diagnostic information output unit.

[0277] According to one embodiment of the present invention, a diagnostic system that assists in diagnosing multiple diseases based on retinal images may include: a retinal image acquisition unit that acquires target retinal images which form the basis for acquiring diagnostic information for a subject; a first processing unit that acquires a first result for a subject using a first diagnostic model (the first diagnostic model is machine-learned based on a first set of retinal images) on the target retinal images; a second processing unit that acquires a second result for a subject using a second diagnostic model (the second diagnostic model is machine-learned based on a second set of retinal images which differs from the first set of retinal images in at least part of the way) on the target retinal images; a third processing unit that determines diagnostic information for a subject based on the first and second results; and a diagnostic information output unit that provides the determined diagnostic information to the user.

[0278] The third processing unit can consider the first and second results together to determine whether the diagnostic information corresponding to the target retinal image is normal or abnormal.

[0279] The third processing unit can determine diagnostic information for the subject by assigning priority to abnormal results so as to improve diagnostic accuracy.

[0280] When the first result is normal and the second result is normal, the third processing unit can determine the diagnostic information as normal, and when the first result is not normal or the second result is not normal, the third processing unit can determine the diagnostic information as abnormal.

[0281] The first result and the second result may be for the same disease or for different diseases.

[0282] The first processing unit and / or at least one map related to the first / second result is obtained through the first / second diagnostic model, and the diagnostic information output unit can output an image of at least one map.

[0283] When the diagnostic information obtained by the third processing unit is abnormal diagnostic information, the diagnostic information output unit can output an image of at least one map.

[0284] The diagnostic system further includes a fourth processing unit that obtains quality information of the target retinal image, and the diagnostic information output unit can output the quality information of the target retinal image obtained by the fourth processing unit.

[0285] [[ID=~]]When it is determined by the fourth processing unit that the quality information of the target retinal image is below a predetermined quality level, the diagnostic information output unit can provide, together with the diagnostic information determined by the user, information indicating that the quality information of the target retinal image is below the predetermined quality level.

[0286] 1.4.2.1. Configuration of the serial diagnostic system

[0287] According to one embodiment, a serial diagnostic system can be provided in which multiple diagnostic models are connected in series. Several embodiments of the serial-type diagnostic model will be described below.

[0288] Figure 15 is a diagram illustrating a series diagnostic model according to one embodiment. Referring to Figure 15, the diagnostic model 1000 may include a first series model 1100 and a second series model 1200.

[0289] The first diagnostic model 1100 can acquire input data including retinal images and obtain a first output (or intermediate output).

[0290] The first diagnostic model 1100 can acquire input data including retinal images and / or other medical diagnostic images and / or non-visual diagnostic materials. The input data may include retinal images, OCT images, iris images, angiography images, lung CT images, cardiac CT images, lung X-ray images, cardiac X-ray images, kidney X-ray images, other tomographic images, MRI images, or X-ray images. The input data may also include data indicating the subject's age, height, sex, smoking status, family history, etc.

[0291] The first output may be a value obtained by the output layer of the first diagnostic model 1100. For example, the first diagnostic model 1100 may be a classifier model, and the first output may include output values ​​at multiple nodes in the output layer of the first diagnostic model 1100. Alternatively, for example, the first diagnostic model 1100 may be a regression model, and the first output may include a numerical value obtained by the first diagnostic model 1100.

[0292] The first output may be a value provided by some layer of the first diagnostic model 1100. For example, the first output may be a value obtained based on the values ​​of the output layer of the first diagnostic model 1100. Alternatively, the first output may be a value obtained based on the values ​​of the hidden layer of the first diagnostic model 1100.

[0293] According to one embodiment, the first output may be a value obtained by an activation function in the output layer of the first diagnostic model 1100. If the output layer of the first diagnostic model 1100 includes multiple nodes (or neurons), the first output may include an output value corresponding to each of the multiple nodes or a value obtained through a predetermined function (e.g., sum) based on each output value.

[0294] The activation function may be any one of the following: sigmoid function, hyperbolic tangent function, ReLU (Rectified Linear Unit) function, PReLU, Leaky ReLU function, identity function, ELU (Exponential Linear Unit) function, or Maxout function.

[0295] The first output may be a feature map or feature value associated with the target disease. The first output may also be a probability map, sampling map, heat map, etc., associated with the target disease. The second diagnostic model 1200 may be configured to acquire diagnostic information based on the feature map or feature value associated with the target disease.

[0296] The first output may be a probabilistic phenotype associated with the target disease. For example, if the target disease is coronary artery disease and the diagnostic information is numerical information associated with the target coronary artery disease, the first output may be the probability of the subject having the target coronary artery disease obtained based on an eye image. The second diagnostic model 1200 may be configured to obtain diagnostic information for the target disease based on a probabilistic expression associated with the target disease.

[0297] The second diagnostic model 1200 can acquire a second output (or diagnostic information) based on the first output. The second diagnostic model 1200 may be a diagnostic model 1000 that has been trained to acquire a second output using the first output as input. The second output may be the various forms of diagnostic information described herein.

[0298] Figure 16 is a diagram illustrating a serial diagnostic model according to another embodiment. Referring to Figure 16, the diagnostic model 1000 may include a first diagnostic model 1100 and a second diagnostic model 1200.

[0299] The first diagnostic model 1100 can acquire input data including eyeball images and obtain a first output (intermediate output). The second diagnostic model 1200 can acquire diagnostic information based on the first output and the second input.

[0300] The second input may be the same input data as the first input. For example, the first diagnostic model 1100 can acquire a first output based on the eyeball image, and the second diagnostic model 1200 can acquire diagnostic information based on the first output and the eyeball image.

[0301] The second input may be input data obtained based on the first input. For example, the second input may be input data obtained by performing image processing on an eyeball image. For example, the second input may be a black and white processed eyeball image, an eyeball image with blood vessels emphasized, a blood vessel image extracted from an eyeball image, or an eyeball image from which blood vessels have been removed.

[0302] The second input may be input data that differs from the first input in at least part.

[0303] The second input may be image data different from the first input. The second input may include retinal images, OCT images, iris images, angiography images, lung CT images, cardiac CT images, lung X-ray images, cardiac X-ray images, kidney X-ray images, other tomographic images, MRI images, or X-ray images.

[0304] The second input may include non-visual information about the subject. For example, the first diagnostic model 1100 may obtain a first output regarding the target disease based on an eye image, and the second diagnostic model 1200 may obtain diagnostic information based on the first output and the second input (e.g., physical information of the subject (e.g., age, sex, race, smoking status, blood pressure (e.g., blood pressure value, presence or absence of hypertension), presence or absence of diabetes (or blood glucose level), cholesterol level, family history, etc.)).

[0305] Figure 17 is a diagram illustrating a serial diagnostic model according to yet another embodiment. Referring to Figure 17, the diagnostic model 1000 may include a first diagnostic model 1100 and a second diagnostic model 1200.

[0306] In comparison with Figure 16, the diagnostic model 1000 can further acquire the first diagnostic information obtained by the first diagnostic model 1100. The diagnostic model 1000 can acquire intermediate diagnostic information and secondary diagnostic information obtained based on the intermediate diagnostic information.

[0307] Diagnostic model 1000 can acquire first diagnostic information and second diagnostic information. First diagnostic model 1100 can acquire input data including an eyeball image and acquire a first output (first diagnostic information or intermediate output). Diagnostic model 1000 can acquire first diagnostic information based on the first output. Second diagnostic model 1200 can acquire second diagnostic information based at least in part on the first output. Diagnostic model 1000 can acquire second diagnostic information based at least in part on the first output, taking into account other information extracted from the eyeball image.

[0308] The first and second diagnostic information may be diagnostic information for the same target disease. The first diagnostic information may be diagnostic information that can be obtained based on eye images (clinically or through machine learning models), such as a probabilistic representation of the presence or absence of eye disease, vascular abnormalities, kidney disease, etc.

[0309] The second diagnostic information may provide more detailed diagnostic information than the first diagnostic information. For example, the second diagnostic information may include grade information indicating the risk level for the target disease, or score information indicating a score associated with the target disease, for the same target disease as the first diagnostic information.

[0310] The second diagnostic information is related to the first diagnostic information, but may also include diagnostic information that can be obtained by considering information other than imaging, such as disease progression rate and guide information related to the diagnostic information.

[0311] On the other hand, the first and second diagnostic information may be diagnostic information for different diseases. The first and second diagnostic information may be diagnostic information for different diseases belonging to the same group. For example, the first diagnostic information may be diagnostic information related to glaucoma belonging to the eye disease group, and the second diagnostic information may be diagnostic information related to macular degeneration belonging to the eye disease group. For example, the first diagnostic information may be diagnostic information related to drusen belonging to the eye disease group, and the second diagnostic information may be diagnostic information related to diabetic retinopathy belonging to the eye disease group.

[0312] The first and second diagnostic information may be diagnostic information for different diseases belonging to different groups. For example, the first diagnostic information may be diagnostic information related to macular degeneration or drusen belonging to the eye disease group, and the second diagnostic information may be diagnostic information related to hyperlipidemia belonging to the cardiovascular disease group.

[0313] In the above embodiment, the case where the diagnostic model 1000 has a first diagnostic model 1100 and a second diagnostic model 1200 was described as a baseline, but the diagnostic model 1000 may also include a larger number of diagnostic models. Furthermore, each diagnostic model may be connected through the parallel or series connection described above.

[0314] 1.4.2.2. Diagnosis through a serial diagnostic model

[0315] According to one embodiment of the invention described herein, a diagnostic method using a diagnostic model including series-connected submodels may be provided.

[0316] Figure 18 is a diagram illustrating a diagnostic method using a diagnostic model according to one embodiment.

[0317] Referring to Figure 18, the diagnostic method according to one embodiment may include the steps of acquiring input data (S61), acquiring first diagnostic information (S62), and acquiring second diagnostic information (S63).

[0318] The step of acquiring input data (S61) may include acquiring a retinal image of the subject. The step of acquiring input data (S61) may further include acquiring medical images of body parts other than the subject's eyes. The step of acquiring input data (S61) may further include acquiring non-visual information about the subject (e.g., the subject's physical information). The step of acquiring input data (S61) may further include performing preprocessing necessary for acquiring diagnostic information on the subject's eye image.

[0319] The step of obtaining first diagnostic information (S62) may include obtaining first diagnostic information for the subject through a first diagnostic model based on eye images. Obtaining first diagnostic information may also include obtaining diagnostic information related to a first disease. For example, obtaining first diagnostic information may include obtaining first diagnostic information indicating the probability that the subject's coronary artery calcium score is 0 or greater, based on eye images.

[0320] The step of obtaining the second diagnostic information (S63) may include obtaining the second diagnostic information for the subject through a second diagnostic model based on the first diagnostic information.

[0321] Obtaining a second diagnostic information may include obtaining diagnostic information that is related to the first disease and different from the first diagnostic information. For example, obtaining the first diagnostic information may include obtaining first diagnostic information indicating the probability that the subject's coronary artery calcium score is 0 or greater, and obtaining the second diagnostic information may include obtaining second diagnostic information indicating a score (e.g., the probability of a cardiovascular disease-related event occurring within the last 10 years) related to whether or not the subject has the target cardiovascular disease, based on the subject's first diagnostic information.

[0322] Alternatively, obtaining secondary diagnostic information may include obtaining diagnostic information related to a secondary disease different from the primary disease. For example, obtaining primary diagnostic information may include obtaining primary diagnostic information indicating whether or not the subject has a target eye disease, and obtaining secondary diagnostic information may include obtaining secondary diagnostic information indicating whether or not the subject has a cerebrovascular or cardiovascular disease.

[0323] The diagnostic model described above will be explained in detail below with specific examples of its implementation.

[0324] 2. Methods for diagnosing kidney disease

[0325] 2.1.1. Biomarkers for kidney disease

[0326] In one embodiment, renal disease biomarkers for predicting, measuring, and confirming the presence, progression, etc., of renal disease may be used in a renal disease diagnostic method. In this specification, renal disease biomarkers may include renal disease risk assessment tools.

[0327] For example, kidney disease biomarkers may include eGFR (Estimated Glomerular Filtration Rate), albuminuria levels, cystatin C levels, and KDIGO (Kidney Disease: Improving Global Outcomes) grades.

[0328] eGFR values ​​are calculated based on blood creatinine concentration, age, sex, and race, and can be used as an indicator to assess renal filtration function. For example, based on eGFR values, a low risk of kidney disease (90 mL / min / 1.73 m) is considered to be the standard. 2 (The above) Moderate risk (60-89 mL / min / 1.73 m) 2 ), moderate to high risk (45-59 mL / min / 1.73 m 2 ), high-to-high risk (30-44 mL / min / 1.73 m 2 ), maximum risk (29 mL / min / 1.73 m 2 It can be classified as follows:

[0329] Albuminuria is a measure of the ratio of albumin in the urine and can be used as a primary diagnostic indicator for kidney disease. Albuminuria may also be the ratio of albumin to creatinine. For example, based on albuminuria, the risk of kidney disease can be classified into low risk (less than 30 mg / g), moderate risk (30-300 mg / g), and high risk (300 mg / g or more).

[0330] Cystatin C levels may be a biomarker of renal function measured in the blood. A higher cystatin C level may indicate a higher risk of renal disease. For example, cystatin C levels can be used to classify the risk of renal disease into low-risk (1.0 mg / L) and high-risk (greater than 1.0 mg / L) groups. In other cases, cystatin C levels can be used to classify the risk of renal disease into low-risk (0.62 to 1.15 mg / L) and high-risk (greater than 1.15 mg / L) groups. Furthermore, in yet another case, cystatin C levels can be used to classify the risk of renal disease into low-risk (0.9 mg / L or less), low-to-intermediate risk (0.9 to 1.2 mg / L), intermediate risk (1.2 to 1.9 mg / L), intermediate-to-high risk (1.9 to 3.0 mg / L), and high-risk (greater than 3.0 mg / L) groups.

[0331] The KDIGO grade may be a biomarker of renal function determined by integrating eGFR value (or grade) and albuminuria value (or grade). For example, the KDIGO grade can be used to classify individuals into low-risk, intermediate-risk, and high-risk groups by combining eGFR value (or grade) and albuminuria value (or grade).

[0332] In addition to these, a variety of other kidney disease biomarkers can be used in kidney disease diagnostic methods.

[0333] 2.1.2. Diagnostic Methods for Renal Diseases

[0334] A renal disease diagnostic method according to one embodiment of this specification may be performed using at least one of the single diagnostic model, parallel diagnostic model, or serial diagnostic model described above. For the sake of clarity, the following description will focus on performing the renal disease diagnostic method using the serial diagnostic model.

[0335] Figure 19 is a diagram illustrating a method for diagnosing kidney disease according to one embodiment.

[0336] Referring to Figure 19, the processor of the diagnostic device may include a step of acquiring a retinal image (S100) and a step of acquiring kidney disease diagnostic information (S200).

[0337] In step S100, the diagnostic device's processor can acquire a retinal image. Depending on the embodiment, the diagnostic device's processor can also perform preprocessing, enhancement, serialization, etc., on the acquired retinal image. Since the above explanation can be applied to this, a detailed explanation will be omitted.

[0338] Furthermore, in step S200, the diagnostic device's processor can acquire renal disease diagnostic information. In this specification, renal disease diagnostic information may be expressed as Reti-CKD. The renal disease diagnostic method will be explained with reference to Figure 20.

[0339] Figure 20 illustrates a diagnostic model for obtaining kidney disease diagnostic information according to one embodiment.

[0340] Referring to Figure 20, the diagnostic model 1000 may be included in the processor and / or storage module of the diagnostic device. Furthermore, the descriptions of the diagnostic models in Figures 15 to 18 may apply to the diagnostic model 1000.

[0341] The diagnostic model 1000 may include a first diagnostic model 1100 and a second diagnostic model 1200. The first diagnostic model 1100 and the second diagnostic model 1200 may be machine learning models. Furthermore, the first diagnostic model 1100 and the second diagnostic model 1200 may be models based on the same algorithm or models based on different algorithms. For example, the first diagnostic model 1100 and the second diagnostic model 1200 may be neural network models. Alternatively, the first diagnostic model 1100 may be a neural network model, and the second diagnostic model 1200 may be a machine learning model that is not a neural network model. In the following explanation, for the sake of clarity, we will mainly describe the case where the first diagnostic model 1100 is a neural network model and the second diagnostic model 1200 is a machine learning model that is not a neural network model, but the explanation in this specification is not limited to this case.

[0342] The diagnostic device's processor can input retinal images (or pre-processed retinal images, serialized retinal images) into the first diagnostic model 1100. The diagnostic device's processor can then obtain the probability value and / or grade that the subject in the retinal image is currently at high risk of renal disease from the first diagnostic model 1100 into which the retinal image has been input. For example, the diagnostic device's processor can obtain from the first diagnostic model 1100 the eGFR value of the subject in the current retinal image of 60 mL / min / 1.73 m². 2A probability value and / or a grade for the probability that follows and / or the probability that albuminuria is present in the urine of the current subject (or the probability that the albuminuria value is 300 mg / g or more) can be obtained. Here, the probability value may be a number between 0 and 1.

[0343] Specifically, the first diagnosis model 1100 may be a neural network model with a CNN structure.

[0344] In one embodiment, the training target retinal images in the training process of the first diagnosis model 1100 may include retinal images of kidney disease patients and retinal images of healthy subjects who are not kidney disease patients.

[0345] Also, single images of the left and right eyes of the training target retinal images in the training process of the first diagnosis model 1100 can be separately trained. In the training process, the ground truth may be the presence or absence of kidney disease. Also, predetermined information can be labeled for each retinal image. As an example, for each retinal image, the eGFR value is 60 mL / min / 1.73m 2 Whether it is below or not and / or whether albuminuria is present in the urine of the subject (for example, whether the albuminuria value is 300 mg / g or more) can be labeled as a binary variable. For example, the eGFR value is 60 mL / min / 1.73m 2 If it is below or albuminuria is present in the urine of the subject, the subject of the retinal image is labeled as having kidney disease, and the eGFR value is 60 mL / min / 1.73m 2 If it is above and albuminuria is not present in the urine of the subject, the subject of the retinal image can be labeled as not having kidney disease. Of course, the exemplified eGFR value and albuminuria value above can be changed according to the embodiment. Also, as another example, the eGFR value and albuminuria value of the subject can be labeled for each retinal image.

[0346] Furthermore, during the evaluation process of the first diagnostic model 1100, probability values ​​are assigned to the left eye retinal image and the right eye retinal image, and the average of these probability values ​​can be considered as the output of the test result.

[0347] Furthermore, the final fully connected layer of the first diagnostic model 1100 can perform one logit probability prediction. The logit from the final fully connected layer can be converted into a probability using a sigmoid function. The first diagnostic model 1100 can also be trained to minimize target and prediction losses. In one embodiment, the first diagnostic model 1100 can also be trained using an AdamW Optimizer for 50 epochs with a learning rate of 0.0002 and a cosine learning rate schedule.

[0348] Furthermore, at least one of the following can be used to enhance the training data of the retinal images: mixup, cutmix, random augmentation, enhancing contrast module, and random crop. Focal loss and exponential moving average can also be used to train the first diagnostic model 1100, and the size of the training retinal images can be set to 384x384.

[0349] Furthermore, the diagnostic device's processor can input the output values ​​of the first diagnostic model 1100 and the subject's physical information (at least one of the following: age, sex, race, smoking status, blood pressure (e.g., blood pressure value, presence or absence of hypertension), presence or absence of diabetes (or blood glucose value), and cholesterol value) into the second diagnostic model 1200. The diagnostic device's processor can then obtain probability values ​​and / or grades from the second diagnostic model 1200 for the probability of a kidney disease-related event occurring within five years for the subject. Here, kidney disease-related events within five years may include the occurrence of fatal kidney disease, non-fatal kidney disease, and various events caused by kidney disease (hospitalization, treatment, surgery, death, etc.) within five years from the time of retinal image acquisition. In addition, the diagnostic device's processor can obtain probability values ​​from the second diagnostic model 1200 for the probability of a kidney disease-related event occurring within five years for the subject, apply a predetermined cutoff value to the obtained probability values, and obtain a grade corresponding to the obtained probability values. Furthermore, while this specification primarily describes kidney disease-related events within a five-year period, it is not limited to this. Depending on the embodiment, kidney disease-related events over various periods, such as within three years or ten years, can be applied to the descriptions herein.

[0350] As a result, the diagnostic device's processor can output probability values ​​(scores) and / or grades for the probability of a kidney disease-related event occurring within the subject's five-year period, as part of the kidney disease diagnostic information.

[0351] Furthermore, depending on the embodiment, the processor of the diagnostic device can obtain probability values ​​and / or grades for the risk of present kidney disease from the second diagnostic model 1200. Also, depending on the embodiment, the processor of the diagnostic device can obtain probability values ​​and / or grades for the probability of a subject experiencing kidney disease-related events within the past five years from the second diagnostic model 1200, and obtain probability values ​​and / or grades for the risk of present kidney disease based on the obtained probability values ​​and / or grades for the subject experiencing kidney disease-related events within the past five years. For example, the processor of the diagnostic device can compare the probability values ​​and / or grades for the probability of a subject experiencing kidney disease-related events within the past five years obtained from the second diagnostic model 1200 with at least one predetermined reference probability value and / or grade, and obtain probability values ​​and / or grades for the risk of present kidney disease based on the comparison results.

[0352] As an example, a high risk of existing biomarkers (for example, an eGFR value of 44 mL / min / 1.73 m²) 2 The following can be predetermined: a first score and / or first grade of renal disease diagnostic information corresponding to an albuminuria level of 300 mg / g or higher and / or a KDIGO grade of high risk or higher. The diagnostic device processor can then set the first score and / or first grade to a first threshold and / or first threshold grade. The diagnostic device processor can determine that the subject is currently at high risk of renal disease if the probability value and / or grade for the probability of renal disease-related events occurring within 5 years, obtained from the second diagnostic model 1200, is equal to or higher than the predetermined first threshold and / or first threshold grade. In the above example, one threshold score and / or threshold grade was described as one, but this is not limited to this, and multiple threshold scores and / or threshold grades can be set. For example, the threshold scores and / or threshold grades may include a first threshold and / or first threshold grade corresponding to a high risk for the existing biomarker and a second threshold and / or second threshold grade corresponding to a moderate risk for the existing biomarker.

[0353] In one embodiment, the second diagnostic model 1200 may be trained on training data. Here, training of the second diagnostic model 1200 may include fitting the second diagnostic model 1200. The second diagnostic model 1200 may also be constructed using a linear regression-based Cox proportional hazards model.

[0354] As an example, the training data may include the physical information of the subject of the training target retinal image of the first diagnostic model 1100, the results of follow-up observations of the subject's kidney disease events over a 5-year period, and the probability values ​​of the first diagnostic model 1100 for the training target retinal image. Furthermore, depending on the embodiment, the subject of the training target retinal image of the first diagnostic model 1100 and the subject of the training data of the second diagnostic model 1200 may be the same, or at least partially different. For example, among the training data of subjects of the training target retinal image of the first diagnostic model 1100, the training data of subjects who have already experienced a kidney disease event or are suffering from kidney disease at the reference time may be excluded from training the second diagnostic model 1200. This is because training the second diagnostic model 1200 with training data from healthy individuals who have not experienced a kidney disease event and are not suffering from kidney disease may improve the accuracy of the probability of kidney disease events occurring within 5 years output by the second diagnostic model 1200.

[0355] Furthermore, in one embodiment, during the training process of the second diagnostic model 1200, the Cox proportional hazards model can be fitted to the UK Biobank cohort to set the coefficients of the covariates of the Cox proportional hazards model. The probability of survival for kidney disease within 5 years in the UK Biobank cohort may be 0.9980896. The second diagnostic model 1200 can then be modeled as a probability of failure within 5 years. Although Example 1 uses the UK Biobank cohort as an example, the description herein is not limited thereto, and the second diagnostic model can be fitted or calibrated based on other cohorts besides the UK Biobank cohort.

[0356] Furthermore, in one embodiment, the second diagnostic model 1200 may include as variables age, sex, presence or absence of hypertension, presence or absence of diabetes, and probability values ​​for the current risk of kidney disease output from the first diagnostic model 1100. Since such variables can be obtained from the subject in a non-invasive manner, they can be used in the second diagnostic model 1200.

[0357] For example, the second diagnostic model 1200 can obtain probability values ​​for the probability of a subject experiencing a kidney disease-related event within five years based on the following formula 1. For example, the Cox proportional hazards model of the second diagnostic model 1200 can be constructed based on formulas 1 and 2.

[0358] [Formula 1] Predicted risk= 1-SP^(Exp[LP])

[0359] Here, Predicted risk is the probability value for the occurrence of a kidney disease-related event within 5 years, and SP (Survival Probability) is the probability of survival against kidney disease within 5 years, which may be 0.9980896 in the UK Biobank example mentioned above. LP (Linear Predictor) is explained by the following formula 2.

[0360] [Formula 2] LP=a1*RPS*100+a2*Age+a3*Female+a4*Hypertension+a5*Diabetes

[0361] Here, RPS (Retinal photograph-based Prediction Score) is the probability value for the risk of current kidney disease output from the first diagnostic model 1100, Age is the age, and Female indicates the subject's sex. For example, Female may be an indicator function that reflects a value of 1 if the subject is female and a value of 0 if the subject is male. Hypertension can reflect a value of 1 if the subject has high blood pressure and a value of 0 if they do not. For example, a user taking antihypertensive drugs can be set to have high blood pressure. Diabetes can reflect a value of 1 if the subject has diabetes (or is in the pre-diabetic stage) and a value of 0 if they do not have diabetes (or are in the pre-diabetic stage).

[0362] Furthermore, a1 to a5 are the coefficients of the covariates of each variable, and for example, through fitting the second diagnostic model 1200, a1 to a5 can be set to 0.0546351, 0.0357466, -0.3317242, -0.1207903, and 0.4679712, respectively. In addition, the processor of the diagnostic device can present the risk of kidney disease within 3, 5, or 10 years based on (or by applying) formulas 1 and 2. For example, the processor of the diagnostic device can add or substitute race, smoking status, cholesterol levels, etc., to the covariates of formula 2 and set the coefficients of each covariate through fitting the second diagnostic model 1200.

[0363] Furthermore, the diagnostic device's processor can obtain probability values ​​for the probability of renal disease-related events occurring within five years from the second diagnostic model 1200, and apply predetermined cutoff values ​​to these probability values ​​to obtain grades corresponding to those probability values. Of course, depending on the circumstances, predetermined cutoff values ​​may be applied to the second diagnostic model 1200 in advance, and grades corresponding to the probability values ​​may be obtained from the second diagnostic model 1200. In this case, there may be one or more cutoff values. For example, if there is one cutoff value, there may be two grades corresponding to the probability values, and if there are two cutoff values, there may be three grades corresponding to the probability values. The description herein can be applied to a variety of cutoff values ​​and a variety of cutoff values.

[0364] In one embodiment, the processor of the diagnostic device can set predetermined cutoff values ​​based on the population distribution of each group, which is divided into multiple grades (i.e., the incidence rate in each group), according to the follow-up observation results of a predetermined population. For example, the cutoff values ​​may be determined based on the results of follow-up observations of a predetermined population. For example, if follow-up observation results are obtained for the entire group of subjects followed up, with 0-n1% being the low-risk group, n1-n2% being the medium-risk group, n2-n3% being the high-risk group, and n3-100% being the highest-risk group, then n1%, n2%, and n3% may be set as the cutoff values, respectively. Alternatively, even if n1%, n2%, and n3% are not the cutoff values, the cutoff values ​​may be set so that the population distribution in each risk group of renal disease diagnostic information is similar to the follow-up observation results of the predetermined population.

[0365] For example, if the given population is the UK Biobank Cohort and the Korean Diabetic Cohort, cutoff values ​​may be set so that the population distribution in each risk group of renal disease diagnostic information is similar to the population distribution in each risk group of the UK Biobank Cohort and the Korean Diabetic Cohort, depending on the population distribution of low-risk, intermediate-risk, high-risk, and highest-risk groups in the UK Biobank Cohort and the Korean Diabetic Cohort.

[0366] Furthermore, the diagnostic device's processor can set cutoff values ​​based on the incidence rate of renal disease biomarkers. For example, in the case of eGFR, the risk of renal disease can be determined based on the eGFR value, specifically when the risk is low (90 mL / min / 1.73 m³). 2 (The above) Moderate risk (60-89 mL / min / 1.73 m) 2 ), moderate to high risk (45-59 mL / min / 1.73 m 2 ), high-to-high risk (30-44 mL / min / 1.73 m 2 ), maximum risk (29 mL / min / 1.73 m 2 The following classifications are possible. If the incidence rates based on eGFR values, i.e., low risk, moderate risk, moderate-high risk, high-high risk, and highest risk, are x1%, x2%, x3%, x4%, and x5% respectively, then cutoff values ​​can be set so that the proportion of people included in each grade of renal disease diagnostic information output by the diagnostic device's processor is x1%, x2%, x3%, x4%, and x5%. This can be applied not only to eGFR values ​​but also to biomarkers such as albuminuria levels, cystatin C levels, and KDIGO grade.

[0367] Furthermore, the incidence rate based on the Kidney Failure Risk Equation and the incidence rate based on the Risk Prediction Equation can also be used to set the cutoff values ​​mentioned above.

[0368] Furthermore, the cutoff value may be optimized to best detect the group corresponding to the high risk level for each biomarker.

[0369] As a specific example, the diagnostic device's processor can set a cutoff value using the ratio of high-risk and low-risk groups classified by the 5-year kidney risk calculator according to race and the 5-year kidney risk calculation method of the applicable clinical guidelines. This improves the performance of predicting high-risk groups in kidney disease diagnostic information. For example, to maximize the performance of kidney disease diagnostic information that diagnoses high-risk groups based on eGFR values, the cutoff value for kidney disease diagnostic information can be set so that the ratio of subjects in the low-risk group and the ratio of subjects in the high-risk group are similar to the ratio of subjects in the low-risk group and the ratio of subjects in the high-risk group in kidney disease diagnostic information. For example, the diagnostic device's processor can set the cutoff value for kidney disease diagnostic information so that the ratio of subjects in the low-risk group and the ratio of subjects in the high-risk group in kidney disease diagnostic information fall within a predetermined range of the ratio of subjects in the low-risk group and the ratio of subjects in the high-risk group based on eGFR values. In this case, the diagnostic device's processor can accurately determine which subjects belong to the low-risk and high-risk groups based on eGFR values.

[0370] Furthermore, the biomarkers used for training the first diagnostic model 1100 and the biomarkers used for setting the cutoff value of the second diagnostic model 1200 may be the same or different.

[0371] Furthermore, when cutoff values ​​are set based on each biomarker, the grade output as kidney disease diagnostic information by the diagnostic device's processor may be similar to the actual results of the biomarker for which the cutoff value has been set. For example, if a cutoff value corresponding to the eGFR value biomarker is applied to the probability value for the probability of a kidney disease-related event occurring within 5 years for a subject, output from the second diagnostic model 1200, and the subject's grade is output by the diagnostic device's processor, the subject's grade output by the diagnostic device's processor may be matched with the grade determined when the subject actually undergoes an eGFR-based examination. When actual examinations are performed based on eGFR, albuminuria, and cystatin C, inconveniences such as the subject having to undergo blood tests or urine tests may occur. However, in the kidney disease diagnostic method according to this specification, highly accurate kidney disease diagnostic information is obtained using only the subject's retinal image, thus improving user convenience.

[0372] Furthermore, in one embodiment, when renal disease diagnostic information is applied to the scores and / or grades of existing biomarkers, the incidence rate of renal disease-related events corresponding to the scores and / or grades of existing biomarkers can be categorized or stratified according to each group of renal disease diagnostic information (e.g., low-risk group, medium-risk group, high-risk group, and highest-risk group). The cutoff for renal disease diagnostic information can then be set so that, when renal disease diagnostic information is applied to the scores and / or grades of existing biomarkers, the incidence rate of renal disease-related events corresponding to the scores and / or grades of existing biomarkers is clearly stratified according to each group of renal disease diagnostic information.

[0373] The following describes in detail embodiments of a method for diagnosing kidney disease according to this specification.

[0374] 2.1.3. Example 1

[0375] Example 1 describes experimental results of a method for diagnosing renal disease according to this specification, using clinical data and retinal images from the UK Biobank cohort and the Korean diabetes cohort.

[0376] First, in Example 1, the first diagnostic model 1100 used 158,216 retinal images and clinical data (79,108 individuals) from the Korea Health Examination Center (each retinal image had an eGFR value of 60 mL / min / 1.73 m²). 2 Training may be based on whether or not the following conditions are met and / or whether or not albuminuria is present in the subject's urine.

[0377] Furthermore, in Example 1, the second diagnostic model 1200 can be trained (or fitted) using the UK Biobank cohort (30,477 subjects) and the Korean diabetes cohort (5,014 subjects). Here, participants with chronic kidney disease and an eGFR value of 90 mL / min / 1.73 m² are included. 2 Data from participants whose albuminuria is below a certain threshold or who have albuminuria detected (or whose albuminuria level is 30 mg / Cr or higher) may be excluded from training the second diagnostic model 1200. In this case, the diagnostic device processor can provide a highly accurate probability value and / or grade for the probability of a kidney disease-related event occurring within five years for healthy individuals who do not currently have kidney disease.

[0378] In addition, the contents described in Figure 20 may be applied to the first diagnostic model 1100 and the second diagnostic model 1200 in Example 1.

[0379] The results from Example 1 will be described below. In the following diagram, the Reti-CKD score represents the probability value for the occurrence of kidney disease-related events within 5 years as kidney disease diagnostic information, and each grade of the Reti-CKD score can represent the grade corresponding to each of the aforementioned probability values ​​as kidney disease diagnostic information.

[0380] Figure 21 shows the clinical characteristics of subjects according to the renal disease diagnostic information obtained in Example 1.

[0381] Referring to Figure 21, the data from the Korea Health Screening Center used to train the first diagnostic model 1100 consisted of 79,108 individuals, with a mean age of 49.5 years (standard deviation (SD): 11.8) and a mean eGFR value of 100.3 (standard deviation (SD): 14.3) mL / min / 1.73m². 2 This may also be the case. Furthermore, the data from the UK Biobank cohort used for training the second diagnostic model 1200 included 30,477 participants, of which 720 (2.4%) could be diagnosed with chronic kidney disease during an average follow-up period of 10.8 years (interquartile range (IQR): 10.7–11.0). Additionally, the data from the Korean diabetes cohort used for training the second diagnostic model 1200 included 5,014 participants, of which 206 (4.1%) could be diagnosed with chronic kidney disease during an average follow-up period of 6.1 years (interquartile range (IQR): 4.0–8.4). The mean eGFR value for the UK Biobank cohort was 99.4 (standard deviation (SD): 6.6) mL / min / 1.73m². 2 The mean eGFR value for the Korean diabetes cohort was 102.5 mL / min / 1.73 m² (standard deviation (SD): 9.1). 2 That's fine.

[0382] Figures 22a and 22b illustrate the incidence rate of kidney disease-related events according to the kidney disease diagnostic information obtained in Example 1.

[0383] Referring to Figures 22a and 22b, in the graphs of Figures 22a and 22b, the x-axis represents time (years), and the y-axis represents the incidence rate of chronic kidney disease-related events.

[0384] Figure 22a shows Kaplan-Meier curves representing the results of a Kaplan-Meier survival analysis conducted on four groups of subjects in the UK Biobank cohort, corresponding to four grades of renal disease diagnostic information. Figure 22b shows Kaplan-Meier curves representing the results of a Kaplan-Meier survival analysis conducted on four groups of subjects in the Korean diabetes cohort, corresponding to four grades of renal disease diagnostic information. In the UK Biobank cohort shown in Figure 22a, 321,317 person-years were studied during an average follow-up period of 10.8 years, while in the Korean diabetes cohort shown in Figure 22b, 30,122 person-years were studied during an average follow-up period of 6.1 years.

[0385] As shown in Figures 22a and 22b, renal disease diagnostic information can clearly differentiate renal disease risk based on four groups in the UK Biobank cohort and the Korean diabetes cohort. In Figures 22a and 22b, the hazard ratio (HR) for the development of chronic kidney disease can show dose-dependent associations according to the four grades. The adjusted hazard ratio per 1 SD increment of renal disease diagnostic information is 1.34 (95% confidence interval: 1.27-1.41) for the UK Biobank cohort in Figure 22a and may be 1.94 (95% confidence interval: 1.63-2.31) for the Korean diabetes cohort in Figure 22b.

[0386] Furthermore, the diagnostic information for kidney disease allows for the clear differentiation of kidney disease risk with significant hazard ratios even within subgroups based on sex, age, presence or absence of hypertension, and presence or absence of diabetes. For example, in the UK Biobank cohort shown in Figure 22a, the trend in hazard ratios for the development of chronic kidney disease was 1.52 overall (95% confidence interval: 1.42~1.63), 1.47 for men (95% confidence interval: 1.31~1.65), 1.51 for women (95% confidence interval: 1.36~1.67), 1.35 for those under 55 years old (95% confidence interval: 1.20~1.51), 1.46 for those 55 years and older (95% confidence interval: 1.29~1.65), and 1 for those without hypertension. 0.59 (95% confidence interval: 1.47~1.74), 1.32 (95% confidence interval: 1.12~1.56) with hypertension, 50 (95% confidence interval: 1.39~1.61) without diabetes, 1.35 (95% confidence interval: 1.07~1.72) with diabetes, 1.53 (95% confidence interval: 1.34~1.73) with eGFR value greater than 100, and 1.41 (95% confidence interval: 1.30~1.53) with eGFR value less than or equal to 100 are also acceptable. Furthermore, in the Korean diabetes cohort (b), the trend in the hazard ratio for the development of chronic kidney disease was 1.99 overall (95% confidence interval: 1.72-2.31), 2.17 for men (95% confidence interval: 1.73-2.70), 1.85 for women (95% confidence interval: 1.51-2.28), 1.75 for those under 55 years old (95% confidence interval: 1.28-2.39), and 1.97 for those 55 years and older. (95% confidence interval: 1.58~2.46), no hypertension 2.13 (95% confidence interval: 1.76~2.57), hypertension present 1.71 (95% confidence interval: 1.37~2.12), eGFR value greater than 100 2.10 (95% confidence interval: 1.62~2.72), eGFR value ≤100 1.49 (95% confidence interval: 1.28~1.78).

[0387] Figure 23 illustrates the predictive performance of kidney disease diagnostic information within five years based on Example 1.

[0388] Referring to Figure 23, the table in Figure 23 compares the performance of the Reti-CKD score (Reti-CKD) and the eGFR-based renal disease score (eGFR-CKD score) in the UK Biobank cohort and the Korean diabetes cohort. The eGFR-based renal disease score can be used as a biomarker to assess the risk of pre-existing renal disease.

[0389] In the table in Figure 23, the Net Reclassification Index (NRI) can be used to measure how much better the new model is than previous models. For the NRI, the UK Biobank cohort showed a similar result of 0.109 (95% confidence interval (CI): 0.44~0.156), and the Korean diabetes cohort showed a similar result of 0.179 (95% confidence interval: 0.017~0.292). Furthermore, when comparing the score based on renal disease diagnostic information with the eGFR-based renal disease score, the difference in C-statistics was 0.020 (95% confidence interval (CI): 0.011~0.029) for the UK Biobank cohort and 0.024 (95% confidence interval: 0.002~0.046) for the Korean diabetes cohort, suggesting that the score based on renal disease diagnostic information may be significantly larger than the eGFR-based renal disease score. As a result, scores based on kidney disease diagnostic information can demonstrate superior performance and higher accuracy when compared to existing eGFR-based kidney disease scores.

[0390] Also, although not shown in the table in Figure 23, the eGFR value is 90 mL / min / 1.73 m². 2When performing a sensitivity analysis on the entire population, excluding only the above-mentioned subjects, the difference in C-statistics between the renal disease diagnostic information-based score and the eGFR-based renal disease score was 0.008 (95% confidence interval (CI): 0.001~0.016) in the UK Biobank cohort and 0.057 (95% confidence interval: 0.048~0.067) in the Korean diabetes cohort. This suggests that the renal disease diagnostic information-based score may be significantly larger than the eGFR-based renal disease score. Therefore, the renal disease diagnostic information-based score can demonstrate superior sensitivity and performance when compared to existing eGFR-based renal disease scores. Consequently, even without prior eGFR testing via blood tests (for example, if the eGFR value is 90 mL / min / 1.73 m²), 2 Even if it's unclear whether the result is above or below the threshold, kidney disease diagnostic information, compared to eGFR, can better stratify the risk of future kidney disease in both the general population and diabetic patients.

[0391] Furthermore, although not shown in the table in Figure 23, sensitivity analysis can be performed by dividing the subjects into two groups: one with hypertension and / or diabetes, and the other without hypertension and diabetes.

[0392] When comparing a score based on renal disease diagnostic information with an eGFR-based renal disease score in a group without hypertension or diabetes, the difference in C-statistics was 0.025 (95% confidence interval (CI): 0.002~0.048) in the UK Biobank cohort, indicating that the score based on renal disease diagnostic information may be significantly larger than the eGFR-based renal disease score. This suggests that the score based on renal disease diagnostic information can demonstrate superior performance with even greater sensitivity than existing eGFR-based renal disease scores in healthy renal individuals without hypertension or diabetes.

[0393] Furthermore, in groups with hypertension and / or diabetes, when comparing the score based on renal disease diagnostic information with the eGFR-based renal disease score, the difference in C-statistics was 0.019 (95% confidence interval (CI): 0.008~0.030) in the UK Biobank cohort, indicating that the score based on renal disease diagnostic information may be significantly larger than the eGFR-based renal disease score. This suggests that the score based on renal disease diagnostic information can demonstrate superior performance with even greater sensitivity than existing eGFR-based renal disease scores, regardless of the presence or absence of hypertension and / or diabetes. Therefore, it can be seen that the score is applicable not only to the group of patients with hypertension and diabetes, which are the most important risk factors for renal disease, but also to renal disease that occurs due to problems with the kidneys themselves, regardless of hypertension / diabetes, and that risk prediction is possible.

[0394] 2.1.4. Obtaining information on kidney disease risk based on existing biomarkers

[0395] Figure 24 is a diagram illustrating a method for diagnosing kidney disease according to another embodiment.

[0396] Referring to Figure 24, the kidney disease diagnostic method according to another embodiment can acquire a retinal image (S300) and information regarding the risk of kidney disease corresponding to existing biomarkers as kidney disease diagnostic information (S400).

[0397] In step S300, the diagnostic device's processor can acquire a retinal image. Since Figure 19 and the explanation above apply to this process, a detailed explanation is omitted.

[0398] Furthermore, in step S400, the diagnostic device's processor can acquire information regarding the risk of renal disease corresponding to existing biomarkers as renal disease diagnostic information. The content described in the diagnostic model above can be applied to step S400.

[0399] Here, information regarding the risk of renal disease according to existing biomarkers means the risk of renal disease calculated according to existing biomarkers, and the diagnostic device's processor can obtain information regarding the risk of renal disease calculated according to existing biomarkers using retinal images. For example, according to step S400, the diagnostic device's processor can use a diagnostic model based on retinal images to obtain information regarding eGFR value and / or grade according to eGFR value, albuminuria value and / or grade according to albuminuria value, cystatin C value and / or grade according to cystatin C value, and output the obtained information, or output information regarding the final score and / or grade according to the final score based on the obtained information.

[0400] In one embodiment, the processor of the diagnostic device can use a diagnostic model to acquire information regarding the risk of kidney disease corresponding to existing biomarkers as kidney disease diagnostic information.

[0401] For example, the model may be trained using retinal images and information on the risk of renal disease corresponding to existing biomarkers labeled on the retinal images. For instance, each retinal image may be labeled with at least one of the following: eGFR value (and / or corresponding grade), albuminuria value (and / or corresponding grade), or cystatin C value (and / or corresponding grade), and the diagnostic model may be trained using the retinal images and their corresponding labels. The labels on the retinal images may also include information on at least one of the following of the subject of the retinal image: age, sex, race, smoking status, blood pressure (presence or absence of hypertension), and presence or absence of diabetes.

[0402] The diagnostic device's processor can input retinal images of the subject into a learned single diagnostic model to obtain information about the risk of kidney disease based on existing biomarkers. For example, if the diagnostic model is learned based on eGFR values, the diagnostic device's processor can input retinal images into the diagnostic model and obtain information from the model regarding eGFR values ​​and / or grades corresponding to those eGFR values.

[0403] Furthermore, depending on the embodiment, the processor of the diagnostic device can input information on the subject's age, sex, race, smoking status, blood pressure (presence or absence of hypertension), presence or absence of diabetes, or cholesterol level, along with the retinal image, into the diagnostic model, and obtain information on the eGFR value and / or grade corresponding to the eGFR value from the diagnostic model.

[0404] Furthermore, in one embodiment, the diagnostic models may be configured in parallel. For example, the descriptions of the diagnostic models in sections 1.4.1.1 to 1.4.1.3 may be applied to the diagnostic models.

[0405] Specifically, a diagnostic model may include multiple diagnostic models. For example, a diagnostic model may include a first diagnostic model and a second diagnostic model. The first and second diagnostic models may be trained using information on kidney disease risk corresponding to different biomarkers. For example, the first diagnostic model may be trained using retinal images labeled with eGFR values ​​(and / or corresponding grades), and the second diagnostic model may be trained using retinal images labeled with albuminuria values ​​(and / or corresponding grades). Of course, this is not limited to the first and second diagnostic models, and the diagnostic model may include additional diagnostic models such as a third and fourth diagnostic model. Furthermore, the labels on the retinal images may also include information on at least one of the following: age, sex, race, smoking status, blood pressure (presence or absence of hypertension), and presence or absence of diabetes of the subject of the retinal image.

[0406] Furthermore, the diagnostic device's processor can input the subject's retinal image into the first and second diagnostic models to obtain information on the risk of kidney disease corresponding to different biomarkers. For example, the diagnostic device's processor can obtain the subject's eGFR value (and / or corresponding grade) from the first diagnostic model and the subject's albuminuria value (and / or corresponding grade) from the second diagnostic model, and output the obtained information. The diagnostic device's processor can also output information on the final score and / or the grade corresponding to the final score based on the information obtained from each diagnostic model. For example, it can obtain and output information on the final score and / or the grade corresponding to the final score based on the subject's eGFR value (and / or corresponding grade) obtained from the first diagnostic model and the subject's albuminuria value (and / or corresponding grade) obtained from the second diagnostic model. As a result, the diagnostic device's processor can non-invasively obtain information on the risk of kidney disease corresponding to existing biomarkers as kidney disease diagnostic information using retinal images, without using invasive methods such as blood tests.

[0407] Furthermore, in one embodiment, the diagnostic models may be configured in series. For example, the descriptions of the diagnostic models in 1.4.2.1, 1.4.2.2, and 2.1.2 may apply to the diagnostic models.

[0408] As an example, the diagnostic model may include a first diagnostic model and a second diagnostic model, as shown in Figure 20. For example, the processor of the diagnostic device inputs the subject's retinal image into the first diagnostic model and calculates the probability that the subject is currently at high risk of kidney disease from the first diagnostic model (e.g., the subject's eGFR value is 60 mL / min / 1.73 m²). 2The probability of the following and / or the probability of albuminuria being present in the subject's urine can be obtained. The diagnostic device's processor can also take the output value of the first diagnostic model and the subject's physical information (at least one of the following: age, sex, race, smoking status, blood pressure (e.g., blood pressure value, presence or absence of hypertension), presence or absence of diabetes (or blood glucose level), and cholesterol level) as input. At this time, the second diagnostic model may be set with cutoff values ​​based on various biomarkers such as eGFR value, albuminuria value, and cystatin C value. For example, the cutoff value for the second diagnostic model may be 45-59 mL / min / 1.73m², which is the standard for the intermediate-to-high risk group of eGFR values. 2 A cutoff value may be set to be similar to the proportion of people who fall into the specified category and / or similar to the proportion of people who fall into the intermediate-risk group for albuminuria levels of 30-300 mg / g. This allows renal disease diagnostic information to predict intermediate-to-high-risk groups (or intermediate-risk groups) with high accuracy based on a variety of biomarkers.

[0409] 2.1.5. Provision of guide information for renal disease diagnostic information

[0410] Figure 25 is a diagram illustrating a method for providing guide information for kidney disease diagnostic information according to one embodiment.

[0411] Referring to Figure 25, the guide information provision method according to one embodiment may include the steps of acquiring kidney disease diagnostic information (S500) and providing guide information for the diagnostic information (S600).

[0412] In step S500, the diagnostic device's processor can acquire diagnostic information. Since the above explanation applies to step S500, a detailed explanation is omitted.

[0413] Furthermore, in step S600, the diagnostic device's processor can provide guide information for the diagnostic information. In steps S500 and S600, for the sake of explanation, the focus will be on kidney diseases, but the guide information described herein is not limited to these, and can of course be applied to a variety of diseases, including eye diseases and cardiovascular diseases. Accordingly, in steps S500 and S600, the focus will be on explaining guide information corresponding to kidney diagnostic information.

[0414] In one embodiment, guide information may refer to information regarding medical / non-medical treatments recommended to the subject in accordance with the renal disease diagnosis. For example, guide information may include prescription information, treatment information, and management information.

[0415] Prescription information may refer to information about medications (e.g., specialty drugs) recommended to a subject to maintain or improve their risk of renal disease in accordance with their renal disease diagnosis. Prescription information may also include information about the prescribed medication, timing of administration, and dosage. For example, prescription information may include information about prescriptions for one or more of the following: ACE inhibitors (Angiotensin-Converting Enzyme Inhibitors) (e.g., lisinopril, enalapril, hydrochlorothiazide), cholesterol-lowering agents (e.g., statins, atorvastatin, rosuvastatin), phosphate binders (e.g., calcitriol, sevelamer), and anticoagulants (e.g., warfarin). The prescription information may also include SGLT2 inhibitors (Sodium-Glucose Co-Transporter 2 Inhibitors) used as combination agents for diabetes prescriptions (e.g., dapagliflozin, empagliflozin, canagliflozin), ARBs (Angiotensin II Receptor Blockers) used as combination agents for hypertension prescriptions (e.g., losartan, valsartan), and GLP1 (Glucagon-Like Peptide-1) agonists and / or diuretics (e.g., furosemide) used as combination agents for diabetes and obesity prescriptions.

[0416] Furthermore, the diagnostic device's processor can provide prescription information taking into account medications that require caution when administered to patients with renal disease. For example, medications that require caution when administered to patients with the aforementioned disease include analgesics and antipyretics (non-steroidal anti-inflammatory drugs (NSAIDs), high-dose aspirin), antibiotics (aminoglycosides, amphotericin B, cephalosporins, penicillins), and beta-lactamase inhibitors. (inhibitors), quinolones, rifampin, sulfonamides, vancomycin, antiviral agents (acyclovir, adefovir, ganciclovir, atazanavir, indinavir, tenofovir), bisphosphonates (pamidronate, zoledronic acid), calcineurin inhibitors (cyclosporine, tacrolimus), anticancer agents (alkylating agents) Agents, cisplatin, methotrexate, mitomycin, interferon-alpha, proteasome inhibitors, vascular endothelial growth factor (VEGF) inhibitors, checkpoint inhibitors, iodine contrast agents used in CT and angiography, diuretics (loop diuretics, thiazides, triamterene), proton pump inhibitorsInhibitors (Dexlansoprazole, Esomeprazole, Lansoprazole, Omeprazole, Pantoprazole, Rabeprazole), Other drugs (Allopurinol, Gold sodium thiomalate, Lithium, Quinine, Sodium phosphate), Traditional Chinese medicine (Aristolochic acid, Cat's claw, Licorice root)

[0417] Furthermore, action information may refer to information about future actions recommended to the subject to maintain or improve their risk of kidney disease in accordance with their kidney disease diagnosis. For example, additional screening information may include information about secondary diagnoses or medical procedures the subject may undergo. As an example, additional screening information may include information about additional tests that may be required, information about hospitals / medical staff that can perform those tests, and information about recommended procedures / surgeries.

[0418] Furthermore, management information may include information on non-medical measures recommended to the subject to maintain or improve the risk of renal disease in accordance with the renal disease diagnosis information. For example, management information may include information on lifestyle habits, dietary habits, exercise, and non-specialized medications such as nutritional supplements to lower the risk of renal disease.

[0419] In one embodiment, the diagnostic device may be linked to an external monitoring device. Here, the monitoring device may mean a device that monitors the lifestyle or behavior of a subject. For example, the monitoring device may include portable devices, wearable devices, wellness measuring devices, etc. Alternatively, the monitoring device may be the client device described above. For example, the monitoring device may include an imaging unit that can capture images of the inside and outside of the eyeball through the imaging unit to acquire images of the inside and outside of the eyeball.

[0420] Furthermore, the monitoring device can monitor a wide range of information, including the subject's activity level, exercise method, exercise duration, food and drink intake, intake amount, health supplement intake information, sleep duration, sleep habits, heart rate, blood pressure, blood glucose level, body water content, oxygen level, body temperature, oxygen saturation, pulse wave, whether or not they have visited a hospital, whether or not they have undergone an examination, whether or not they have undergone a procedure / surgery, eye images, eGFR value, albuminuria value, and cystatin C value.

[0421] The diagnostic device's processor can communicate with the monitoring device via a communication module, either wired or wirelessly.

[0422] The diagnostic device's processor can provide guide information to the monitoring device. The monitoring device can then provide the subject with a variety of information based on the guide information and the monitored information. For example, the monitoring device can acquire management information (e.g., lifestyle information, dietary information, exercise information) as guide information from the diagnostic device, compare the management information with the monitored information to determine whether the monitored information matches the management information, and provide the determination result and / or additional information accordingly.

[0423] For example, if the monitored exercise time is less than the exercise time in the management information, the monitoring device can provide the subject with information to exercise in accordance with the management information. Also, if the monitored information on food and drink intake corresponds to the food and drink intake in the management information, the monitoring device can provide the subject with information that they are consuming food and drink in accordance with the management information.

[0424] Furthermore, the diagnostic device's processor can acquire monitoring information received from the monitoring device. Based on the guide information and the monitored information, the diagnostic device's processor can provide diverse information to the subject. For example, the diagnostic device's processor can compare the monitored information with the guide information to determine whether the monitored information matches the management information, and provide the determination result and / or additional information accordingly. The above-described examples of monitoring devices can be applied to the operation of the diagnostic device's processor.

[0425] Furthermore, the diagnostic device's processor can generate guide information by reflecting the monitoring information received from the monitoring device. For example, the diagnostic device's processor can acquire subject status information (exercise status, lifestyle, eating habits, etc.) based on the monitoring information, and modify the guide information determined as renal disease diagnostic information based on the subject's status information to suit the subject.

[0426] In one embodiment, the processor of the diagnostic device can provide guide information using a predetermined database. For example, the diagnostic device may include a database that matches guide information with scores and / or grades of renal disease diagnostic information. For example, if renal disease diagnostic information is expressed in three grades, the database may include guide information matched with the low-risk grade (e.g., prescription information - none, intervention information - information on the next consultation date, management information - dietary habit information provision, exercise information provision), guide information matched with the moderate-risk grade (e.g., prescription information - none, intervention information - additional test information provision, management information - dietary habit information provision, exercise information provision, non-specialized drug information provision), and guide information matched with the high-risk grade (e.g., prescription information - statin prescription information provision, intervention information - additional test information, recommended treatment / surgery information provision, management information - dietary habit information provision, exercise information provision, non-specialized drug information provision). The processor of the diagnostic device can provide guide information matched with renal disease diagnostic information based on the database.

[0427] In other embodiments, the diagnostic device's processor may provide guide information using a machine learning model. For example, the diagnostic device may include a guide information model based on a machine learning model or a neural network model. The guide information model may be learned based on the score and / or grade of renal disease diagnostic information and the guide information. Additionally, the guide information model may also be learned for various physical information of the subject (e.g., at least one of height, weight, sex, race, smoking status, blood pressure (e.g., blood pressure value, presence or absence of hypertension), presence or absence of diabetes (or blood glucose level), cholesterol level). This allows the diagnostic device's processor to input the score and / or grade of renal disease diagnostic information and the subject's physical information into the guide information model to obtain guide information for the subject. Depending on the embodiment, the guide information model may be included in the diagnostic model or configured independently of the diagnostic model.

[0428] 2.1.6. Provision of guide information using existing prescription information and renal disease support information

[0429] Figure 26 is a diagram illustrating a method for providing guide information using existing prescription information and kidney disease diagnostic information according to one embodiment.

[0430] Referring to Figure 26, the guide information provision method according to one embodiment may include the steps of acquiring the subject's existing prescription information (S710), acquiring renal disease diagnostic information (S720), and providing guide information based on the existing prescription information and renal disease diagnostic information (S730).

[0431] In some cases, kidney disease progresses in conjunction with underlying conditions such as diabetes, hypertension, and / or obesity. Furthermore, some prescription medications for diabetes, hypertension, and / or obesity often have therapeutic effects on kidney disease as well. For example, SGLT2 inhibitors are prescription medications for diabetes but can also be combination drugs that treat kidney disease, and ARBs are prescription medications for hypertension but can also be combination drugs that treat kidney disease. Similarly, GLP-1 (Glucagon-Like Peptide-1) agonists are prescription medications for diabetes and / or obesity but can also be combination drugs that treat kidney disease.

[0432] If a subject is taking such a combination drug, it may be necessary to provide prescription information to the subject that either does not require a separate prescription for kidney disease, depending on the subject's risk of kidney disease, or excludes drugs that would reduce the subject's therapeutic effect when taken together with the combination drug. For example, some studies have shown that a combination prescription of an ARB (a drug prescribed for high blood pressure) and an ACE inhibitor (a drug prescribed for kidney disease) is not recommended. In this way, when providing guidance information to subjects, especially prescription information, it is possible to provide accurate prescription information by taking into account the medications the subject is already taking.

[0433] This allows the diagnostic device's processor to provide guidance information, particularly prescription information, based on the subject's existing prescription information and renal disease diagnostic information.

[0434] According to step S710, the diagnostic device's processor can acquire the subject's existing prescription information. For example, the diagnostic device's processor can acquire existing prescription information from an external device or the diagnostic device's input module. The diagnostic device's processor can also acquire information about the subject's underlying medical conditions (e.g., presence or absence of hypertension, blood pressure value, presence or absence of diabetes, blood glucose level, obesity status, BMI value, cholesterol level, etc.) either as a substitute for or along with the existing prescription information.

[0435] Furthermore, in step S720, the diagnostic device's processor can acquire renal disease diagnostic information. Here, renal disease diagnostic information may refer to renal disease diagnostic information based on retinal images of the same subject as the subject for whom the existing prescription information (and / or information on the underlying disease) acquired in step S710 was obtained. Since the above explanation applies to step S720, a detailed explanation is omitted.

[0436] Furthermore, in step S730, the diagnostic device's processor can provide guidance information based on existing prescription information and renal disease diagnostic information.

[0437] The diagnostic device's processor can provide guidance information based on renal disease diagnostic information, as described in 2.1.5 above. However, in step S730, guidance information can be provided by considering not only renal disease diagnostic information but also existing prescription information (and / or information on the underlying disease).

[0438] In one embodiment, if the score and / or grade of the renal disease diagnostic information falls into the low-risk group, the diagnostic device processor may not provide prescription information among the guide information. The diagnostic device processor can then provide information useful for the treatment or management of the underlying disease when providing intervention information and / or management information if the subject receives a prescription for the treatment of the underlying disease through existing prescription information (and / or information on the underlying disease) or if the risk of the underlying disease is high.

[0439] Furthermore, if the renal disease diagnostic information score and / or grade falls into the intermediate-risk and / or high-risk group, the diagnostic device processor can provide prescription information from the guide information. If the existing prescription information includes combination drugs that can treat both the underlying disease and renal disease, such as SGLT2 inhibitors, ARBs, and GLP1 agonists, the processor can provide prescription information considering the risk of the renal disease diagnostic information. For example, if the risk of renal disease diagnostic support is moderate, the diagnostic device processor can provide prescription information advising the patient to continue taking the existing combination drug without prescribing a new renal disease medication. Also, if the risk of renal disease diagnostic support is high, the diagnostic device processor can provide prescription information for renal disease medications that can be taken together with the existing combination drug in the prescription information.

[0440] Furthermore, if existing prescription information does not include a treatment for the underlying disease, such as a combination drug, or if the risk of the underlying disease is low, and the risk of assisting in the diagnosis of renal disease is moderate or high, the diagnostic device's processor can provide prescription information for the renal disease treatment drug (or combination drug).

[0441] However, such examples are merely illustrative of the various embodiments described herein. Not limited to such examples, the processor of the diagnostic device may provide guide information using a predetermined database and / or guide information model, as described in 2.1.5. In this case, the database may store guide information matched with existing prescription information and renal disease diagnosis information, and the guide information model may be learned using guide information corresponding to existing prescription information and renal disease diagnosis information. Thus, the processor of the diagnostic device may obtain guide information from the database using existing prescription information and renal disease diagnosis information, or obtain guide information by inputting existing prescription information (or existing prescription information and renal disease diagnosis information) into the guide information model.

[0442] 2.1.7. Prediction of Renal Disease Progression Rate Using Renal Disease Diagnostic Information

[0443] Figure 27 is a diagram illustrating a method for predicting the progression rate of renal disease using renal disease diagnostic information according to one embodiment.

[0444] Referring to Figure 27, the method for predicting the progression rate of renal disease according to one embodiment may include the steps of obtaining result values ​​corresponding to a biomarker (S810), obtaining renal disease diagnostic information (S820), and determining the progression rate of renal disease using the result values ​​corresponding to the biomarker and the renal disease diagnostic information (S830).

[0445] In step S810, the diagnostic device's processor can acquire result values ​​corresponding to biomarkers. For example, the diagnostic device's processor can acquire biomarker scores and / or grades from an external device or the diagnostic device's input module. Here, the biomarkers may include the eGFR value (and / or corresponding grade), albuminuria value (and / or corresponding grade), or cystatin C value (and / or corresponding grade) mentioned above. In addition, the biomarkers may include other renal disease biomarkers or renal disease risk assessment tools.

[0446] Furthermore, in step S820, the diagnostic device's processor can acquire renal disease diagnostic information. Here, renal disease diagnostic information may refer to renal disease diagnostic information based on the retinal image of the same subject as the subject whose result value corresponds to the biomarker acquired in step S810. Since the above explanation applies to step S820, a detailed explanation is omitted.

[0447] Furthermore, in step S830, the rate of progression of kidney disease can be determined using result values ​​corresponding to biomarkers and kidney disease diagnostic information.

[0448] Specifically, even if the result values ​​corresponding to biomarkers are the same, the rate of progression of kidney disease may differ. For example, an eGFR value of 50 mL / min / 1.73 m² 2Even within the moderate-to-high risk group, the scores for renal disease diagnostic information may differ. However, even with the same eGFR value, individuals with higher renal disease diagnostic information scores may experience faster progression of renal disease than those with lower scores.

[0449] This will be explained in detail through Example 2.

[0450] 2.1.7.1. Example 2

[0451] Example 2 is for predicting the rate of kidney disease progression using KDIGO grade and kidney disease diagnostic information.

[0452] In Example 2, data from 5,346 diabetic patients at two tertiary hospitals in South Korea may be used. This includes patients with eGFR < 90 mL / min / 1.73 m². 2 Alternatively, data from individuals with albuminuria may be included. Data from individuals for whom information regarding retinal imaging, serum creatinine, or albuminuria is missing may be excluded.

[0453] Figure 28 is a diagram illustrating the clinical characteristics of the subjects in Example 2.

[0454] Referring to Figure 28, the KDIGO grade can be divided into three grades (low-risk, intermediate-risk, and high-risk) that take into account eGFR value (or grade) and albuminuria value (or grade). In the table in Figure 28, based on albuminuria value, the risk of renal disease can be divided into low-risk group (A1, less than 30 mg / g), intermediate-risk group (A2, 30-300 mg / g), and high-risk group (A3, ≥300 mg / g). Furthermore, based on eGFR value, the risk of renal disease can be divided into low-risk group (G1, 90 mL / min / 1.73 m²). 2 (The above) Intermediate-risk group (G2, 60-89 mL / min / 1.73 m²) 2 ), moderate-to-high risk group (G3a, 45-59 mL / min / 1.73 m²) 2 ), high-to-high risk group (G3b, 30-44 mL / min / 1.73 m²) 2), highest risk group (G4, 29 mL / min / 1.73 m²) 2 It can be classified as follows:

[0455] In the KDIGO grade, the low-risk group includes 3,135 individuals, comprising the low-risk group based on albuminuria levels, the low-risk group based on eGFR levels, and a portion of the intermediate-risk group; the intermediate-risk group includes 1,814 individuals, comprising the low-risk and intermediate-risk groups based on albuminuria levels, the low-risk group based on eGFR levels, the intermediate-risk group, and a portion of the intermediate-high-risk group; and the high-risk group may include 397 individuals, comprising all risk groups based on albuminuria levels and a portion of all risk groups based on eGFR levels.

[0456] Furthermore, the average age of the subjects was 62.4 (+ / -11.4) years, and 60.6% were male. The average eGFR of the subjects was 86.6 (+ / -15.3 mL / min per 1.73 m²). 2 ) and albuminuria may occur in 46.9% of participants. During a follow-up period of 5.0 years (interquartile range: 2.5–7.8), renal disease-related events may occur in 1,379 participants (25.8%).

[0457] Figures 29a and 29b illustrate the incidence of kidney disease-related events according to the KDIGO grade and kidney disease diagnostic information in Example 2.

[0458] Referring to Figures 29a and 29b, in the graphs of Figures 29a and 29b, the x-axis represents time (years), and the y-axis represents the incidence rate of chronic kidney disease-related events.

[0459] The graph in Figure 29a shows the incidence of kidney disease-related events according to the KDIGO grade (low-risk group (a1), intermediate-risk group (a2), and high-risk group (a3)), indicating that the incidence of kidney-related events can be higher in the high-risk group (a3).

[0460] The graph in Figure 29b shows the incidence of kidney disease-related events when the KDIGO grade is adjusted to reflect a score corresponding to kidney disease diagnostic information.

[0461] Group 1 (b1) represents the low-risk group of the KDIGO grade with a score of less than 20 according to the renal disease diagnostic information; Group 2 (b2) represents the low-risk group of the KDIGO grade with a score of 20 or higher according to the renal disease diagnostic information; Group 3 (b3) represents the intermediate-risk group of the KDIGO grade with a score of less than 20 according to the renal disease diagnostic information; Group 4 (b4) represents the intermediate-risk group of the KDIGO grade with a score of 20 or higher according to the renal disease diagnostic information; and Group 5 (b1) represents the high-risk group of the KDIGO grade.

[0462] Specifically, the incidence rate of kidney disease-related events in the low-risk group (a1) in the graph of Figure 29a can be categorized or stratified, as can be seen in the incidence rates of kidney disease-related events in the first group (b1) and the second group (b2) in (b). Here, the incidence rate of kidney disease-related events in the first group (b1) in (b) may be lower than that of the low-risk group (a1) in the graph of Figure 29a, and the incidence rate of kidney disease-related events in the second group (b2) in the graph of Figure 29b may be higher than that of the low-risk group (a1) in the graph of Figure 29a. In other words, even among subjects who fall under the same KDIGO grade low-risk group, the incidence rate of kidney disease-related events can be categorized or stratified according to the kidney disease diagnostic information score. To put it another way, even among subjects who fall under the same KDIGO grade low-risk group, subjects with a high kidney disease diagnostic information score may have a higher probability of experiencing kidney disease-related events than subjects with a low kidney disease diagnostic information score. This could mean that even among individuals who fall into the same low-risk group according to the KDIGO grade, those with a higher renal disease diagnostic information score may experience a faster progression of renal disease than those with a lower renal disease diagnostic information score.

[0463] Furthermore, the incidence rate of kidney disease-related events in the intermediate-risk group (a2) in the graph of Figure 29a can be categorized or stratified, similar to the incidence rates of kidney disease-related events in the third group (b3) and fourth group (b4) in the graph of Figure 29b. Here, the incidence rate of kidney disease-related events in the third group (b3) in the graph of Figure 29b may be lower than that of the intermediate-risk group (a2) in the graph of Figure 29a, and the incidence rate of kidney disease-related events in the fourth group (b4) in the graph of Figure 29b may be higher than that of the intermediate-risk group (a2) in the graph of Figure 29a. These results suggest that, as explained above, even among subjects who fall into the same intermediate-risk group in the KDIGO grade, those with a high kidney disease diagnostic information score may experience a faster rate of kidney disease progression than those with a low kidney disease diagnostic information score.

[0464] Furthermore, as shown in the graph in Figure 29a, when considering only the results corresponding to existing biomarkers, the incidence of renal disease-related events in the KDIGO-grade low-risk group (a1) may be higher than that in the KDIGO-grade intermediate-risk group (a2).

[0465] However, as shown in the graph in Figure 29b, the incidence of kidney disease-related events in Group 2 (b2), which consists of subjects with a KDIGO grade low risk group and a score of 20 or higher according to kidney disease diagnostic information, is sometimes higher than that in Group 3 (b3), which consists of subjects with a KDIGO grade intermediate risk group and a score of less than 20 according to kidney disease diagnostic information. This means that even among subjects with the same low KDIGO grade, those with a higher score for kidney disease diagnostic information may have a higher risk of kidney disease and a faster rate of kidney disease progression than those with a high KDIGO grade.

[0466] Furthermore, the C-statistic when reflecting a score based on renal disease diagnostic information in the KDIGO grade is 0.04 (95% confidence interval (CI): 0.02~0.04) compared to when using only the KDIGO grade. This indicates that when reflecting a score based on renal disease diagnostic information in the KDIGO grade, the risk of renal disease can be predicted with higher accuracy than when using only the KDIGO grade.

[0467] To explain Figure 27 again, in one embodiment, the processor of the diagnostic device can determine whether the score of the kidney disease diagnostic information is above a predetermined threshold.

[0468] Furthermore, the diagnostic device's processor can determine that the progression of kidney disease is rapid if the kidney disease diagnostic information score is above a predetermined threshold, and that the progression of kidney disease is slow if the kidney disease diagnostic information score is below a predetermined threshold.

[0469] For example, as in Example 2, even if the grades corresponding to the KDIGO grade, eGFR value, albuminuria value, and cystatin C value are the same for the first and second subjects, if the kidney disease diagnostic information score of the first subject is above a predetermined threshold (for example, a kidney disease diagnostic information score of 20) and the kidney disease diagnostic information score of the second subject is below the predetermined threshold, the processor of the diagnostic device can determine that the rate of kidney disease progression in the first subject is faster than that of the second subject.

[0470] Furthermore, in one embodiment, in step S830, the diagnostic device's processor can predict acute kidney injury (AKI) after cardiac surgery. Specifically, after cardiac surgery, kidney damage may or may not occur depending on the patient. For example, after cardiac surgery, 5% of patients develop chronic kidney disease, and 1-2% of patients may have complete kidney damage requiring dialysis or kidney transplantation.

[0471] While the existing biomarkers mentioned above can be used as biomarkers to predict acute kidney injury after cardiac surgery, their accuracy is sometimes low. The kidney disease diagnostic information described herein can predict the risk of kidney damage more accurately than existing biomarkers, and can also predict the risk of acute kidney injury after cardiac surgery, which existing biomarkers cannot predict.

[0472] For example, the processor of the diagnostic device can determine whether the score of the renal disease diagnostic information is above a predetermined threshold. The predetermined threshold may be the same as or different from the threshold used to predict the rate of progression of renal disease. The processor of the diagnostic device can then determine that there is a high probability of acute kidney injury after cardiac surgery if the score of the renal disease diagnostic information is above the predetermined threshold, and that there is a low probability of acute kidney injury after cardiac surgery if the score of the renal disease diagnostic information is below the predetermined threshold.

[0473] The diagnostic device's processor can then provide information regarding the rate of kidney disease progression (or information regarding acute kidney injury after cardiac surgery) and / or corresponding guidance information. In this case, because the rate of kidney disease progression (or the likelihood of acute kidney injury after cardiac surgery) differs between the first and second subjects, the diagnostic device's processor can provide different guidance information to the first and second subjects.

[0474] For example, even if the first and second subjects belong to the same risk group based on their existing biomarkers and / or kidney disease diagnostic information, different guidance information can be provided depending on the rate of kidney disease progression in the first and second subjects. For instance, if the first subject is predicted to have a rapid rate of kidney disease progression and belongs to a low-risk group, the diagnostic device's processor can provide guidance information to the effect that blood pressure, diabetes, etc., must be strictly managed. As an example, the diagnostic device's processor can provide guidance information including action information such as information on additional tests (e.g., explanation of additional tests, dates of additional tests, information on hospitals / medical staff that can perform additional tests, etc.) and / or management information such as recommended lifestyle correction goals and recommended dietary habits.

[0475] Furthermore, if the first subject is predicted to have a rapid progression of renal disease and is in a high-risk group, the diagnostic device's processor can increase the prescribed dosage for the first subject as part of the prescription information.

[0476] Furthermore, the diagnostic device's processor can communicate with the monitoring device described above. The diagnostic device's processor can provide the monitoring device with guide information determined according to the rate of kidney disease progression. For example, if the diagnostic device's processor determines that the rate of kidney disease progression is rapid in the first subject, who is in the low-risk group of kidney disease diagnostic information, asymptomatic and does not possess kidney risk factors, it can determine appropriate action information (e.g., additional test information) and / or management information (e.g., adjustment of food intake and drink, target exercise amount) as guide information and provide this guide information to the monitoring device. The monitoring device can provide the subject with the guide information obtained from the diagnostic device and determine whether the monitored information and the guide information match. If a match is found, the monitoring device can provide information indicating that the subject is complying well with the guide information; if a match is not found, the monitoring device can provide a warning to the subject to comply with the guide information.

[0477] Furthermore, the diagnostic device's processor can provide guidance information based on the likelihood of acute kidney injury after cardiac surgery. For example, if the subject is judged to have a high probability of acute kidney injury after cardiac surgery, the diagnostic device's processor can provide guidance information that recommends reducing the likelihood of such injury.

[0478] Furthermore, as described in 2.1.5, the diagnostic device's processor can provide guide information using a predetermined database and / or guide information model. In this case, the database can store guide information matched with renal disease diagnostic information and the rate of renal disease progression (or information regarding acute kidney injury after cardiac surgery), and the guide information model can be learned using renal disease diagnostic information and guide information corresponding to the rate of renal disease progression (or the possibility of acute kidney injury after cardiac surgery). As a result, the diagnostic device's processor can also acquire guide information from the database and / or guide information model using renal disease diagnostic information and the rate of renal disease progression (or the possibility of acute kidney injury after cardiac surgery).

[0479] For example, if a patient is in the low-risk group based on eGFR values ​​and their kidney disease diagnostic information score is above a predetermined threshold, the diagnostic device's processor can determine that the patient has a low risk to their kidneys, but their kidney disease progression rate is faster than other low-risk groups. This allows the diagnostic device's processor to provide information indicating a rapid progression rate of kidney disease and / or corresponding guidance information (for example, guidance information provided when the kidney disease diagnostic information is determined to be of moderate risk).

[0480] Furthermore, if the eGFR value indicates a low-risk group and the kidney disease diagnostic information score is below a predetermined threshold, the diagnostic device's processor can determine that the subject's kidney risk is also low and that the rate of kidney disease progression in the subject is slower compared to other low-risk groups. As a result, the diagnostic device's processor can provide information that the rate of kidney disease progression is slow and / or corresponding guidance information (for example, guidance information provided when the kidney disease diagnostic information is determined to be low-risk).

[0481] Furthermore, depending on the embodiment, different thresholds may be applied to the renal disease diagnostic information depending on the result value corresponding to the biomarker. For example, the diagnostic device may be set so that the threshold is different when the KDIGO grade is low risk, when it is medium risk, and when it is high risk.

[0482] Furthermore, depending on the embodiment, the threshold can be set using a variety of criteria. For example, the threshold can be set so that when kidney disease diagnostic information is applied to result values ​​corresponding to biomarkers, the incidence rate of kidney disease events is clearly stratified.

[0483] Various embodiments of this specification may be embodied as software containing instructions stored in a machine-readable storage medium (e.g., a computer). The machine may include electronic devices according to the disclosed embodiments, which are devices capable of calling instructions stored from the storage medium and operating in accordance with the called instructions. When the instructions are executed by a processor, the processor may perform the function corresponding to the instructions directly or, under the control of the processor, using other components. The instructions may include code generated or executed by a compiler or interpreter. The machine-readable storage medium may be provided in the form of a non-transitory storage medium. Here, “non-transitory storage medium” simply means that it is tangible and does not contain signals, and does not distinguish whether data is stored semi-permanently or temporarily on the storage medium. For example, “non-transitory storage medium” may include a buffer on which data is temporarily stored.

[0484] According to one embodiment, the methods according to the various embodiments disclosed herein may be provided in a Computer Program Product. The Computer Program Product may be traded as a commodity between sellers and buyers. The Computer Program Product may be distributed in the form of a device-readable storage medium (e.g., Compact Disc Read Only Memory, CD-ROM) or online through an application store (e.g., Play Store™). In the case of online distribution, at least a portion of the Computer Program Product (e.g., a Downloadable App) may be at least temporarily stored or temporarily generated in a storage medium such as the memory of the manufacturer's server, the application store's server, or an intermediary server.

[0485] As described above, although the embodiments have been described by limited embodiments and drawings, a person with ordinary skill in the art can make various modifications and variations from the above description. For example, the described technique may be performed in a different order than described, and / or the components of the described system, structure, apparatus, circuit, etc. may be combined or assembled in a different manner than described, or substituted or replaced by other components or equivalents, and the appropriate results may be achieved.

[0486] Therefore, other embodiments, other embodiments, and claims equivalent to those described below also fall within the scope of the claims.

Claims

1. A method for controlling a diagnostic device, The stage of acquiring a retinal image of the subject, The step includes obtaining renal disease diagnostic information for the subject using a machine learning model based on the retinal image, The aforementioned machine learning model includes a first model and a second model, The first model described above is a neural network model, The second model described above is a regression-based machine learning model for controlling a diagnostic device.

2. The first model is trained based on first training data (the first training data includes multiple retinal images) and result values ​​of a first biomarker corresponding to the first training data. The step of obtaining renal disease diagnostic information for the subject is: The retinal image is input to the first model and a first result value is obtained from the first model. The first score and the subject's physical information are input into the second model, and a second result value is obtained from the second model. A control method for a diagnostic device according to claim 1, comprising acquiring the kidney disease diagnostic information based on the second result value.

3. The result value from the first biomarker is at least one of the following: the probability that the eGFR (Estimated Glomerular Filtration Rate) value is less than or equal to the first value, or information regarding whether or not albuminuria is present in the urine of the subject. The control method for a diagnostic device according to claim 2, wherein the first result value represents at least one of the following: the probability that the subject's eGFR value is less than or equal to the first value, or the probability that albuminuria is present in the subject's urine.

4. The control method for a diagnostic device according to claim 3, wherein the second model is trained on second training data (the second training data includes tracking results of kidney disease events for subjects in a predetermined population).

5. A control method for a diagnostic device according to claim 4, wherein data relating to subjects from the predetermined population whose eGFR value is less than a second value (the second value being higher than the first value) is excluded from the second training data.

6. The first result value includes information regarding the subject's current risk of kidney disease, The control method for a diagnostic device according to claim 1, wherein the second result value includes information regarding the probability of future kidney disease-related events occurring in the subject.

7. The control method for a diagnostic device according to claim 6, wherein the second result value includes information regarding the probability of a kidney disease-related event occurring within five years for the subject.

8. The step of obtaining renal disease diagnostic information for the subject is: A control method for a diagnostic device according to claim 1, comprising applying a predetermined cutoff value to the second result value to obtain one grade from a plurality of grades corresponding to the subject as renal disease diagnostic information.

9. The aforementioned cutoff value is, A control method for a diagnostic device according to claim 8, which is set based on the population distribution in groups corresponding to multiple grades that are divided according to the results of follow-up observation of a predetermined population.

10. A method for controlling a diagnostic device according to claim 1, further comprising the step of providing guidance information to the subject based on the kidney disease diagnostic information.

11. The step of providing guidance information to the subject based on the aforementioned kidney disease diagnostic information is: A control method for a diagnostic device according to claim 10, wherein guide information corresponding to the kidney disease diagnostic information is provided using a pre-stored database or machine learning model.

12. The process further includes determining the rate of progression of the kidney disease in the subject based on the aforementioned kidney disease diagnostic information, The step of determining the rate of progression of the kidney disease in the subject based on the aforementioned kidney disease diagnostic information is: Obtain the result value from the biomarker of the subject, A control method for a diagnostic device according to claim 10, comprising determining the rate of progression of the kidney disease of the subject using the result value obtained from the biomarker and the second result value.

13. If the second result value is greater than or equal to a predetermined value, it is determined that the kidney disease is progressing faster than the rate of kidney disease progression expected according to the result value of the biomarker. A control method for a diagnostic device according to claim 12, wherein if the second result value is less than a predetermined value, it is determined that the kidney disease progresses more slowly than the rate of kidney disease progression expected according to the result value of the biomarker.

14. The step of providing guidance information to the subject based on the aforementioned kidney disease diagnostic information is: A control method for a diagnostic device according to claim 12, wherein guidance information for the subject is provided using the rate of progression of the subject's kidney disease and the kidney disease diagnostic information.

15. A recording medium on which a program for performing the method described in any one of claims 1 to 14 is recorded.

16. A diagnostic device, Storage module and Includes at least one processor, The aforementioned at least one processor acquires a retinal image of the subject, Based on the retinal image, a machine learning model stored in the storage module is used to obtain renal disease diagnostic information for the subject. The aforementioned machine learning model includes a first model and a second model, The first model described above is a neural network model, The second model described above is a diagnostic device, which is a regression-based machine learning model.

Images