Method and system for measuring changes in size of target lesion in x-ray image

Abstract

To provide a method for measuring changes in the size of a target lesion in an X-ray image.SOLUTION: The present disclosure comprises the steps of: receiving a first X-ray image including a target lesion and a second X-ray image including the target lesion; calculating an occupancy rate of a region corresponding to the target lesion in each of the reference regions in the first X-ray image and the second X-ray image; and measuring changes in the size of the target lesion based on the calculated occupancy rate.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present disclosure relates to a method and system for measuring changes in the size of a target lesion in an X-ray image. More specifically, among the reference regions in the X-ray image, the occupancy rate of the region corresponding to the target lesion is calculated, and based on the calculated occupancy rate, a method and system for determining changes in the size of the target lesion are provided. The present disclosure relates to a method and system for measuring changes in the size of a target lesion in an X-ray image. More specifically, among the reference regions in the X-ray image, the occupancy rate of the region corresponding to the target lesion is calculated, and based on the calculated occupancy rate, a method and system for determining changes in the size of the target lesion are provided. The present disclosure relates to a method and system for measuring changes in the size of a target lesion in an X-ray image. More specifically, among the reference regions in the X-ray image, the occupancy rate of the region corresponding to the target lesion is calculated, and based on the calculated occupancy rate, a method and system for determining changes in the size of the target lesion are provided. The present disclosure relates to a method and system for measuring changes in the size of a target lesion in an X-ray image. More specifically, among the reference regions in the X-ray image, the occupancy rate of the region corresponding to the target lesion is calculated, and based on the calculated occupancy rate, a method and system for determining changes in the size of the target lesion are provided.

Background Art

[0002] When a patient takes a number of X-ray photographs, changes in the lesions in the photographs can correspond to important information. For example, in the case of a patient with pneumothorax, X-rays can be continuously taken to observe changes in the size of the pneumothorax after surgery. When the size of the pneumothorax decreases, it means that the course of treatment is good, and when the size of the pneumothorax increases, it may correspond to an emergency situation and other measures may be required. Also, when a specific lesion is observed from an X-ray image, information regarding whether the lesion has existed continuously in the same size in the past, or whether its size has increased / decreased, can be necessary information for the treatment of the specific lesion. Therefore, information regarding changes in the size of a lesion over time can correspond to clinically very important information. When a patient takes a number of X-ray photographs, changes in the lesions in the photographs can correspond to important information. For example, in the case of a patient with pneumothorax, X-rays can be continuously taken to observe changes in the size of the pneumothorax after surgery. When the size of the pneumothorax decreases, it means that the course of treatment is good, and when the size of the pneumothorax increases, it may correspond to an emergency situation and other measures may be required. Also, when a specific lesion is observed from an X-ray image, information regarding whether the lesion has existed continuously in the same size in the past, or whether its size has increased / decreased, can be necessary information for the treatment of the specific lesion. Therefore, information regarding changes in the size of a lesion over time can correspond to clinically very important information. When a patient takes a number of X-ray photographs, changes in the lesions in the photographs can correspond to important information. For example, in the case of a patient with pneumothorax, X-rays can be continuously taken to observe changes in the size of the pneumothorax after surgery. When the size of the pneumothorax decreases, it means that the course of treatment is good, and when the size of the pneumothorax increases, it may correspond to an emergency situation and other measures may be required. Also, when a specific lesion is observed from an X-ray image, information regarding whether the lesion has existed continuously in the same size in the past, or whether its size has increased / decreased, can be necessary information for the treatment of the specific lesion. Therefore, information regarding changes in the size of a lesion over time can correspond to clinically very important information. When a patient takes a number of X-ray photographs, changes in the lesions in the photographs can correspond to important information. For example, in the case of a patient with pneumothorax, X-rays can be continuously taken to observe changes in the size of the pneumothorax after surgery. When the size of the pneumothorax decreases, it means that the course of treatment is good, and when the size of the pneumothorax increases, it may correspond to an emergency situation and other measures may be required. Also, when a specific lesion is observed from an X-ray image, information regarding whether the lesion has existed continuously in the same size in the past, or whether its size has increased / decreased, can be necessary information for the treatment of the specific lesion. Therefore, information regarding changes in the size of a lesion over time can correspond to clinically very important information. When a patient takes a number of X-ray photographs, changes in the lesions in the photographs can correspond to important information. For example, in the case of a patient with pneumothorax, X-rays can be continuously taken to observe changes in the size of the pneumothorax after surgery. When the size of the pneumothorax decreases, it means that the course of treatment is good, and when the size of the pneumothorax increases, it may correspond to an emergency situation and other measures may be required. Also, when a specific lesion is observed from an X-ray image, information regarding whether the lesion has existed continuously in the same size in the past, or whether its size has increased / decreased, can be necessary information for the treatment of the specific lesion. Therefore, information regarding changes in the size of a lesion over time can correspond to clinically very important information. When a patient takes a number of X-ray photographs, changes in the lesions in the photographs can correspond to important information. For example, in the case of a patient with pneumothorax, X-rays can be continuously taken to observe changes in the size of the pneumothorax after surgery. When the size of the pneumothorax decreases, it means that the course of treatment is good, and when the size of the pneumothorax increases, it may correspond to an emergency situation and other measures may be required. Also, when a specific lesion is observed from an X-ray image, information regarding whether the lesion has existed continuously in the same size in the past, or whether its size has increased / decreased, can be necessary information for the treatment of the specific lesion. Therefore, information regarding changes in the size of a lesion over time can correspond to clinically very important information. When a patient takes a number of X-ray photographs, changes in the lesions in the photographs can correspond to important information. For example, in the case of a patient with pneumothorax, X-rays can be continuously taken to observe changes in the size of the pneumothorax after surgery. When the size of the pneumothorax decreases, it means that the course of treatment is good, and when the size of the pneumothorax increases, it may correspond to an emergency situation and other measures may be required. Also, when a specific lesion is observed from an X-ray image, information regarding whether the lesion has existed continuously in the same size in the past, or whether its size has increased / decreased, can be necessary information for the treatment of the specific lesion. Therefore, information regarding changes in the size of a lesion over time can correspond to clinically very important information. When a patient takes a number of X-ray photographs, changes in the lesions in the photographs can correspond to important information. For example, in the case of a patient with pneumothorax, X-rays can be continuously taken to observe changes in the size of the pneumothorax after surgery. When the size of the pneumothorax decreases, it means that the course of treatment is good, and when the size of the pneumothorax increases, it may correspond to an emergency situation and other measures may be required. Also, when a specific lesion is observed from an X-ray image, information regarding whether the lesion has existed continuously in the same size in the past, or whether its size has increased / decreased, can be necessary information for the treatment of the specific lesion. Therefore, information regarding changes in the size of a lesion over time can correspond to clinically very important information.

[0003] For a doctor and a patient to accurately understand information regarding changes in the size of a lesion, the actual size of the lesion must be recognized. Existing CAD (Computer Aided Detection) methods notify the presence or absence of a lesion, enabling a doctor to read the patient's image more accurately and quickly. The region of the lesion (e.g., the lung, etc.) imaged depends on the patient's position (imaging environment) and is inconsistent. For a doctor and a patient to accurately understand information regarding changes in the size of a lesion, the actual size of the lesion must be recognized. Existing CAD (Computer Aided Detection) methods notify the presence or absence of a lesion, enabling a doctor to read the patient's image more accurately and quickly. The region of the lesion (e.g., the lung, etc.) imaged depends on the patient's position (imaging environment) and is inconsistent. For a doctor and a patient to accurately understand information regarding changes in the size of a lesion, the actual size of the lesion must be recognized. Existing CAD (Computer Aided Detection) methods notify the presence or absence of a lesion, enabling a doctor to read the patient's image more accurately and quickly. The region of the lesion (e.g., the lung, etc.) imaged depends on the patient's position (imaging environment) and is inconsistent. For a doctor and a patient to accurately understand information regarding changes in the size of a lesion, the actual size of the lesion must be recognized. Existing CAD (Computer Aided Detection) methods notify the presence or absence of a lesion, enabling a doctor to read the patient's image more accurately and quickly. The region of the lesion (e.g., the lung, etc.) imaged depends on the patient's position (imaging environment) and is inconsistent. is fixed, and it becomes difficult to normalize the image when only the pixel array is input Since X-rays have no reference line, it is difficult to calculate the actual size of a lesion using only X-ray images Thus, errors can occur in the change amount (or presence or absence of size change) measured by calculating the size of the lesion from two X-ray images

[0004] As a method for determining the presence or absence of a change in the size of a lesion using an existing CAD method, there are a method of absolutely comparing the sizes of lesions from two X-ray images and a method of comparing the sizes of lesions by registering X-ray images in pixel units According to the first method of absolutely comparing the sizes of lesions from two X-ray photographs, it may be difficult to solve the problem that the size of the lesion is displayed differently in the image due to various variables such as the patient's posture and the distance from the imaging device That is, the change in the size of the lesion may be inaccurate. Also According to the second method of registering X-ray images in pixel units and comparing the sizes of lesions, assuming that the registration in pixel units is successfully performed the difference in the size of the lesion can be predicted more accurately. However, the second method requires a large amount of computation for the registration itself in pixel units, and due to specific situations (for example, situations where the two X-ray images of the patient are very different and difficult to register), or the state of the images (for example the resolution of the image, the state according to the storage form of the image), the registration of the two X-ray images may be substantially impossible. Also, the registration in pixel units corresponds to a very complex algorithm, so the computing power required for this is required in specific situations (for example, situations where the two X-ray images of the patient are very different and difficult to register), or the state of the images (for example the resolution of the image, the state according to the storage form of the image), the registration of the two X-ray images may be substantially impossible. Also, the registration in pixel units corresponds to a very complex algorithm, so the computing power required for this is substantially impossible. Also, since the registration in pixel units corresponds to a very complex algorithm, the computing power required for this is substantially impossible. Also, since the registration in pixel units corresponds to a very complex algorithm, the computing power required for this is ​​​​It is a level that is difficult to commercialize, and a huge cost can be incurred in the development of algorithms.

Prior Art Documents

Patent Documents

[0005]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0006] The present disclosure provides a method and a system for measuring a change in the size of a target lesion for solving the above problems.

Means for Solving the Problems

[0007] The present disclosure can be embodied in various ways including a method, an apparatus (system), or a computer-readable storage medium for storing instructions, a computer program.

[0008] A method for measuring a change in the size of a target lesion in an X-ray image, performed by at least one computing device according to an embodiment of the present disclosure, includes steps of receiving a first X-ray image including the target lesion and a second X-ray image including the target lesion, calculating an occupancy rate occupied by a region corresponding to the target lesion among reference regions in each of the first X-ray image and the second X-ray image, and measuring a change in the size of the target lesion based on the calculated occupancy rate.

[0009] In an embodiment of the present disclosure, the calculating step is Determining a first reference region and a second reference region from each of the images, and a first Identifying a target lesion from each of the first Of the reference regions, calculating a first occupancy rate occupied by the identified target lesion, and of the second reference region Calculating a second occupancy rate occupied by the identified target lesion.

[0010] In one embodiment of the present disclosure, the step of determining the first reference region and the second reference region is Inputting the first X-ray image into a reference region extraction model and outputting the first reference region Inputting the second X-ray image into the reference region extraction model and outputting the second reference region The reference region extraction model is trained using a plurality of reference X-ray images and label information regarding the reference reference regions. Here, the second reference region corresponds to the first reference region.

[0011] In one embodiment of the present disclosure, the calculating step is to divide the number of pixels in the region occupied by the target lesion within the first reference region by the number of pixels corresponding to the first reference region, Calculating the first occupancy rate, and dividing the number of pixels in the region occupied by the target lesion within the second reference region by the number of pixels corresponding to the second reference region, Calculating the second occupancy rate.

[0012] In one embodiment of the present disclosure, the step of determining the first reference region and the second reference region is Determining a score for the first reference region and a score for the second reference region, and the step of identifying the target lesion is to determine a score for the target lesion within the first reference region And determining a score for the target lesion within the second reference region ​ including a step of determining a score, and the calculating step includes calculating a first occupancy rate based on the score for a first reference region and the score for a target lesion within the first reference region; and calculating a second occupancy rate based on the score for a second reference region and the score for the target lesion within the second reference region. In one embodiment of the present disclosure, the measuring step includes determining whether there is a change in the size of the target lesion based on the first occupancy rate and the second occupancy rate.

[0013] In one embodiment of the present disclosure, the step of determining whether there is a change in the size of the target lesion based on the first occupancy rate and the second occupancy rate includes calculating a change amount of the occupancy rate for the target lesion based on the first occupancy rate and the second occupancy rate; and comparing the calculated change amount of the occupancy rate for the target lesion with a reference value to determine whether there is a change in the size of the target lesion.

[0014] In one embodiment of the present disclosure, the reference value is determined based on a value related to a target metric calculated for a test set or a value related to prediction accuracy. In one embodiment of the present disclosure, the reference value includes a first reference value and a second reference value. When the calculated change amount of the occupancy rate for the target lesion is equal to or greater than the first reference value, it is determined that the size of the target lesion has increased. When the calculated change amount of the occupancy rate for the target lesion is less than the first reference value and equal to or greater than the second reference value, it is determined that there is no change in the size of the target lesion. When the calculated change amount of the occupancy rate for the target lesion is less than the second reference value,

[0015] In one embodiment of the present disclosure, the reference value is determined based on a value related to a target metric calculated for a test set or a value related to prediction accuracy.

[0016] In one embodiment of the present disclosure, the reference value includes a first reference value and a second reference value. When the calculated change amount of the occupancy rate for the target lesion is equal to or greater than the first reference value, it is determined that the size of the target lesion has increased. When the calculated change amount of the occupancy rate for the target lesion is less than the first reference value and equal to or greater than the second reference value, it is determined that there is no change in the size of the target lesion. When the calculated change amount of the occupancy rate for the target lesion is less than the second reference value, it is determined that the size of the target lesion has decreased. it is determined that the size of the target lesion has decreased. When the amount is smaller than the second reference value, it is determined that the size of the target lesion has decreased.

[0017] In one embodiment of the present disclosure, based on the first occupancy rate and the second occupancy rate, the step of determining whether there is a change in the size of the target lesion includes calculating the change amount of the occupancy rate for the target lesion based on the first occupancy rate and the second occupancy rate, and inputting the calculated change amount of the occupancy rate for the target lesion into the presence / absence of change determination model. Based on the output determination result, determining whether there is a change in the size of the target lesion. The presence / absence of change determination model includes a machine learning model that is trained to output a determination result regarding the presence / absence of a change in the size of the reference target lesion based on an input value related to the change amount of the reference occupancy rate for the target lesion. For the target lesion For the target lesion For the target lesion For the target lesion For the target lesion

[0018] In one embodiment of the present disclosure, the reference region is a region determined by dividing the entire region of each of the first X-ray image and the second X-ray image into a plurality of regions. For the target lesion

[0019] There is provided a computer program for executing, by a computer, the method for measuring the change in the size of the target lesion described above according to one embodiment of the present disclosure. For the target lesion

[0020] An information processing system according to one embodiment of the present disclosure includes a memory that stores one or more instructions, and executes the one or more stored instructions to receive a first X-ray image including a target lesion and a second X-ray image including the target lesion, calculate the occupancy rate occupied by the region corresponding to the target lesion among the reference regions in each of the first X-ray image and the second X-ray image, and based on the calculated occupancy rate, for the target lesion For the target lesion For the target lesion For the target lesion For the target lesion It includes a processor configured to measure a change in size.

Advantages of the Invention

[0021] In some embodiments of the present disclosure, based on the change in the occupancy rate of the target lesion relative to the reference region whether there is a change in the size of the target lesion is measured, so that even with a small amount of computation, the change in the size of the lesion can be accurately measured.

[0022] In some embodiments of the present disclosure, based on the change in the occupancy rate of the target lesion relative to the reference region whether there is a change in the size of the target lesion is measured, so that when measuring the change in the size of the target lesion it is not much affected by various variables such as the patient's posture and the distance from the imaging device. It is not affected much.

[0023] In some embodiments of the present disclosure, based on the change in the occupancy rate of the target lesion relative to the reference region whether there is a change in the size of the target lesion is determined, so that separate image alignment is not required, and excessive development costs are not required.

[0024] In some embodiments of the present disclosure, in the case of the lungs, there may be a change in size due to inhalation or exhalation. At this time, since the size of the target lesion in the lungs also changes together, the ratio of the size of the target lesion to the size of the lungs can be maintained. Therefore, such a change in the ratio can indicate the change in the target lesion. It can indicate the change in the target lesion.

[0025] In some embodiments of the present disclosure, a plurality of lesions included in the entire region (for example, the lungs, etc.) from an X-ray image can be detected separately, and the occupancy rate and / or the amount of change for each lesion can be calculated, so that the change in the size of each lesion can be measured. It can measure the change in the size of each lesion.

[0026] The effects of the present disclosure are not limited thereto, and other effects not mentioned, etc., should be clearly understood by those having ordinary knowledge in the technical field to which the present disclosure pertains (hereinafter referred to as "persons skilled in the art"). Persons having ordinary knowledge in the technical field to which the present disclosure pertains (hereinafter referred to as "persons skilled in the art") should clearly understand it.

Brief Description of the Drawings

[0027] Examples of the present disclosure will be described based on the following attached drawings. Here, similar reference numerals indicate similar elements, but are not limited thereto.

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Mode for Carrying Out the Invention

[0028] Hereinafter, specific contents for carrying out the present disclosure will be described in detail based on the accompanying drawings. However in the following description, if there is a risk of unnecessarily obscuring the gist of the present disclosure, specific descriptions of well-known functions and configurations will be omitted.

[0029] In the accompanying drawings, the same or corresponding components are given the same reference numerals. Also in the description of the following embodiments, duplicate descriptions of the same or corresponding components will be omitted where possible. However, even if the description of a component is omitted, such a component should not be intended to be excluded from a certain embodiment.

[0030] The advantages and features of the disclosed embodiments, and the methods for achieving them, will become clear based on the accompanying drawings by referring to the embodiments described below. However, the present disclosure is not limited to the embodiments disclosed below and can be embodied in various different forms. However, this embodiment is provided only to make the present disclosure complete and to enable those skilled in the art to accurately recognize the category of the invention.

[0031] The terms used in the present disclosure will be briefly explained, and the embodiments of the disclosure will be specifically described herein. The terms used in this specification are, as much as possible, general terms that are currently widely used, taking into account the functions in the present disclosure. However, this may change due to the intentions or precedents of those skilled in the relevant fields , the emergence of new technologies, etc. Also, in certain cases, terms arbitrarily selected by the applicant are used ​​There may also be words, but their meanings will be described in detail in the description part of the invention. Therefore , the terms used in this disclosure should not be mere names of simple terms, but should be defined based on the meaning the term has and the overall content of this disclosure .

[0032] In this disclosure, unless specifically identified in the context, singular expressions include plural expressions , and plural expressions can include singular expressions. Throughout the specification, if a certain part "includes" a certain component, this means that, unless otherwise stated to the contrary, it does not exclude other components , and can further include other components.

[0033] Also, the terms "module" or "part" used in the specification mean software or hardware components, and the "module" or "part" performs a certain role. However , the "module" or "part" is not limited in meaning to software or hardware. The "module" or "part" may be configured to be in an addressable storage medium , or may be configured to cause one or more processors to reproduce. Therefore , as an example, the "module" or "part" can include at least one of components such as software components, object-oriented software components, class components, task components, and processes, functions, attributes, procedures, subroutines, segments of program code, drivers, firmware, microcode, circuits, data, databases, data structures, tables, arrays, or variables. Components and the "module" or "part" can further include a smaller number of components and "modules" within, and the functions provided internally are coupled by a "module" or "section", or further separable into additional components and a "module" or "section".

[0034] According to an embodiment of the present disclosure, the "module" or "section" can be implemented by a processor and a memory. The "processor" should be broadly interpreted to include general-purpose processors, central processing units (CPUs), microprocessors, digital signal processors (DSPs), controllers, microcontrollers, state machines, etc. In some environments, the "processor" can also refer to application-specific integrated circuits (ASICs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), etc. The "processor" can also refer to a combination of processing devices, such as a combination of a DSP and a microprocessor, a combination of multiple microprocessors, a combination of one or more microprocessors coupled to a DSP core, or any other such configuration combination. Also, the "memory" should be broadly interpreted to include any electronic component capable of storing electronic information. The "memory" includes RAM (Random Access Memory), ROM (Read Only Memory), NVRAM (Non-Volatile Random Access Memory), PROM (Programmable Read-Only Memory), EPROM (Erasable Programmable Read-Only Memory), EEPROM (Electrically Erasable Programmable Read-Only Memory), flash memory, magnetic or optical data storage devices, registers, etc., which can be read by a processor. "processor" may include general-purpose processors, central processing units (CPUs), micro processors, digital signal processors (DSPs), controllers, microcontrollers, state machines and the like. In some environments, the "processor" may refer to application-specific integrated circuits (ASICs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), etc. The "processor" may also refer to a combination of processing devices, such as a combination of a DSP and a microprocessor, a combination of multiple microprocessors, a combination of one or more microprocessors coupled to a DSP core, or any other such combination of configurations. Also, the "memory" should be broadly interpreted to include any electronic component capable of storing electronic information. The "memory" includes RAM (Random Access Memory), ROM (Read Only Memory), NVRAM (Non-Volatile Random Access Memory), PROM (Programmable Read-Only Memory), EPROM (Erasable Programmable Read-Only Memory) EEPROM (Electrically Erasable Programmable Read-Only Memory), flash memory, magnetic or optical data storage devices, registers, etc., which can be read by a It can also refer to various types of possible media. When a processor can read information from memory / read or record information in memory, the memory is said to be in an electronic communication state with the processor. Memory integrated into a processor is in an electronic communication state with the processor.

[0035] In the present disclosure, an "X-ray image" can refer to any image taken by any inspection equipment that allows X-rays to penetrate at least a part of the human body using X-rays for imaging. For example, the inspection equipment includes not only general X-ray imagers but also special X-ray imagers (such as mammography units) adapted to specific parts of the human body, but is not limited thereto.

[0036] In the present disclosure, a "target lesion" can refer to data / information, an image area, an object, etc. that is the target of measurement of size change. For example, a "target lesion" can include an object to be detected through an X-ray image such as cancer or pneumothorax.

[0037] In the present disclosure, a "pixel" can refer to a pixel included in an X-ray image. For example, the "number of pixels" can refer to the number of pixels corresponding to a specific area in an X-ray image. At this time, when the X-ray images have the same resolution, the larger the number of pixels, the larger the size of the specific area in the X-ray image, and the smaller the number of pixels, the smaller the size of the specific area in the X-ray image.

[0038] In the present disclosure, an "artificial neural network model" is an example of a machine learning model and can include any model used to infer an answer for a given input. In one embodiment ​ According to this, an artificial neural network model can include an input layer, a plurality of hidden layers, and an output layer. Here, each layer can include one or more nodes. For example, the artificial neural network model can be trained to determine, identify, and / or detect a reference region and / or a region of a target lesion from an X-ray image. As another example, the artificial neural network model can be trained to output information regarding a change in the size of a target lesion based on a change amount in the occupancy rate with respect to the target lesion (e.g., a difference between a first occupancy rate and a second occupancy rate with respect to the target lesion, or a numerical value obtained by subtracting the first occupancy rate from the second occupancy rate). Also, the artificial neural network model can include weight values associated with a plurality of nodes included in the artificial neural network model. Here, the weight values can include any parameter associated with the artificial neural network model. In the present disclosure, "each of A and B" can refer to the components (e.g., regions) included in A and the components included in B. For example, "the reference region in each of the first X-ray image and the second X-ray image" can refer to the reference region in the first X-ray image and the reference region in the second X-ray image. As another example, "the region corresponding to the target lesion among the reference regions in each of the first X-ray image and the second X-ray image" can refer to the region corresponding to the target lesion among the reference regions in the first X-ray image and the region corresponding to the target lesion among the reference regions in the second X-ray image. In the present disclosure, "instruction" is gathered based on a function.

[0039]

[0040] One or more instruction words, which are components of a computer program and can be executed by a processor can be referred to as such.

[0041] In the present disclosure, "user" can refer to a person who uses a user terminal. For example, the user can include an annotator who performs annotation work. As another example, the user can include a doctor or a patient to whom measurement results of the change in the size of a target lesion are provided. Also, the user can refer to the user terminal, and conversely, the user terminal can refer to the user. That is, the user and the user terminal can be used interchangeably in this specification.

[0042] In the present disclosure, "annotation" can refer to annotation work and / or annotation information determined by the performance of annotation work (for example, label information). In the present disclosure, "annotation information" can refer to information for annotation work and / or information generated by annotation work (for example, label information).

[0043] In the present disclosure, "overall region" can refer to the region of the imaging target included in the X-ray image. For example, the overall region can refer to the region where an individual (for example, a patient) excluding the background region is imaged from the X-ray image. Alternatively, the overall region can refer to the region where a target tissue, organ, organ system, etc. to be observed through the X-ray image is imaged. Also, in the present disclosure, "reference region" can refer to at least a partial region of the overall region.

[0044] In this disclosure, "occupancy rate" refers to the ratio of the occupancy of a target region to a reference region. For example, the "occupancy rate" is the ratio of the size of the target lesion area to the size of the reference area. As another example, the "occupancy rate" can be calculated as a ratio of the number of pixels in the reference area to the number of pixels in the reference area. Additionally or alternatively, the "occupancy rate" can be calculated as the ratio of the number of pixels in the area of the lesion to the number of pixels in the area of the lesion. , can be calculated based on the predicted score (e.g., probability value) of the region, as well as the region size. For example, the "occupancy rate" is the predicted score of the area of the target lesion relative to the size of the reference area (e.g. For example, the ratio of the sum or average of the prediction scores of multiple pixels included in the target region. As another example, the "occupancy rate" can be calculated by the predicted score of the reference region (e.g., The area of the target lesion for each of the multiple pixels included in The predicted score (for example, the sum or average of the predicted scores of each of the multiple pixels included in the target region) Additionally or alternatively, the "occupancy rate" can be calculated by the ratio of the size of the target lesion to the total area of the target lesion. In addition to the size, a probability map (e.g., heat map) of the target lesion and / or the condition of the target lesion may be generated. can be calculated based on the condition.

[0045] FIG. 1 shows an example of a method for measuring changes in the size of a target lesion using an information processing system 100 according to an embodiment of the present disclosure. FIG. 1 is a diagram showing an example of providing a measurement result 130. The information processing system 100 Any computing device that is used to measure changes in A computing device may refer to any type of device that has computing capabilities. For example, notebook, desktop, laptop , which can be a tablet computer, a server, a cloud system, a user terminal, etc., but is not limited thereto. In FIG. 1, the information processing system 1 00 is shown as one computing device, but is not limited thereto. The information processing syst em 100 can be configured to perform distributed processing of information and / or data via a plurality of computing devices.

[0046] The information processing system 100 can be configured to be communicably connected to each of an image capturing device (e.g., an X-ray imaging device), a user terminal, and / or a storage system (or device). Here, the storage system can be a device or a cloud system that stores and manages various data related to a machine learning model for measuring changes in the size of a target lesion. For efficient management of data, the storage system can use a database to store and manage various data. Here, the various data can include any data related to the machine learning model, and can include, for example, X-ray images, label information regarding a reference region, test sets, a machine learning model, etc., but is not limited thereto. To measure changes in the size of a target lesion, the information processing system 100 can calculate the occupancy rate of the target lesion with respect to a reference region by comparing the reference region (e.g., a lung region) and the target lesion region observed within the reference region in an X-ray image (or video). That is,

[0047] the information processing system 100 can calculate the occupancy rate of the target lesion with respect to the reference region for each of a plurality of X-ray images having a difference in the imaging time point. The information processing system 100 can Changes in the occupancy ratio can measure changes in the size of the target lesion in the reference region.

[0048] In one embodiment, the information processing system 100 includes an imaging device, a user terminal, and / or or a storage system (or device) can receive a first X-ray image 110 including a target lesion and a second X-ray image 120 including the target lesion. For example, the information processing system 100 can sequentially receive the first X-ray image 110 and the second X-ray image 120. In another example the information processing system 100 can receive the first X-ray image 110 and the second X-ray image 120 simultaneously. Here, the first X-ray image 110 and the second X-ray image 120 can correspond to images of the same solid / target taken at different times. Additionally or alternatively, the first X-ray image 110 and / or the second X-ray image 120 received by the information processing system 100 may not include the target lesion.

[0049] Based on the received first X-ray image 110 and second X-ray image 120, the information processing system 100 can generate and / or output a measurement result 130 of the change in the size of the target lesion. Here, the measurement result 130 of the change in the size of the target lesion can include the absolute size value, occupancy ratio, presence or absence of a change in the size of the target lesion, degree of change, etc. of the target lesion included in each of the first X-ray image 110 and the second X-ray image 120. For this purpose, the information processing system 100 can calculate the occupancy ratio occupied by the region corresponding to the target lesion among the reference regions in each of the first X-ray image 110 and the second X-ray image 120.

[0050] The information processing system 100 can calculate the occupancy ratio occupied by the region corresponding to the target lesion among the reference regions in each of the first X-ray image 110 and the second X-ray image 120. For this reason, the information processing system 100 can calculate the occupancy ratio occupied by the region corresponding to the target lesion among the reference regions in each of the first X-ray image 110 and the second X-ray image 120. For this purpose, the information processing system 100 can calculate the occupancy ratio occupied by the region corresponding to the target lesion among the reference regions in each of the first X-ray image 110 and the second X-ray image Determine a first reference region and a second reference region from each of the first X-ray images 110 and the second X-ray images 120, and a target lesion can be identified and / or detected from each of the first X-ray images 110 and the second X-ray images 120. Here, the reference region can refer to a region determined by dividing the entire region (e.g., the lung region) of each of the first X-ray image 110 and the second X-ray image 120 into a plurality of regions (e.g., the left lung and the right lung, etc.). For example, the information processing system 100 inputs the first X-ray image 110 into a reference region extraction model to output a first reference region (e.g., the left lung region in the first X-ray image), and inputs the second X-ray image 120 into the reference region extraction model to output a second reference region corresponding to the first reference region (e.g., the left lung region in the second X-ray image). At this time, the information processing system 100 and / or the storage system can include a reference region extraction model learned using label information regarding a plurality of reference X-ray images and reference reference regions. Here, the reference region can refer to a region determined by dividing the entire region (e.g., the lung region) of each of the first X-ray image 110 and the second X-ray image 120 into a plurality of regions (e.g., the left lung and the right lung, etc.). For example, the information processing system 100 inputs the first X-ray image 110 into a reference region extraction model to output a first reference region (e.g., the left lung region in the first X-ray image), and inputs the second X-ray image 120 into the reference region extraction model to output a second reference region corresponding to the first reference region (e.g., the left lung region in the second X-ray image). At this time, the information processing system 100 and / or the storage system can include a reference region extraction model learned using label information regarding a plurality of reference X-ray images and reference reference regions. Here, the reference region can refer to a region determined by dividing the entire region (e.g., the lung region) of each of the first X-ray image 110 and the second X-ray image 120 into a plurality of regions (e.g., the left lung and the right lung, etc.). For example, the information processing system 100 inputs the first X-ray image 110 into a reference region extraction model to output a first reference region (e.g., the left lung region in the first X-ray image), and inputs the second X-ray image 120 into the reference region extraction model to output a second reference region corresponding to the first reference region (e.g., the left lung region in the second X-ray image). At this time, the information processing system 100 and / or the storage system can include a reference region extraction model learned using label information regarding a plurality of reference X-ray images and reference reference regions. Here, the reference region can refer to a region determined by dividing the entire region (e.g., the lung region) of each of the first X-ray image 110 and the second X-ray image 120 into a plurality of regions (e.g., the left lung and the right lung, etc.). For example, the information processing system 100 inputs the first X-ray image 110 into a reference region extraction model to output a first reference region (e.g., the left lung region in the first X-ray image), and inputs the second X-ray image 120 into the reference region extraction model to output a second reference region corresponding to the first reference region (e.g., the left lung region in the second X-ray image). At this time, the information processing system 100 and / or the storage system can include a reference region extraction model learned using label information regarding a plurality of reference X-ray images and reference reference regions.

[0051] Thereafter, the information processing system 100 can calculate a first occupancy rate occupied by the identified target lesion in the first reference region and a second occupancy rate occupied by the identified target lesion in the second reference region. For example, the information processing system 100 divides the number of pixels of the region occupied by the target lesion in the first reference region by the number of pixels corresponding to the first reference region to calculate the first occupancy rate, and divides the number of pixels of the region occupied by the target lesion in the second reference region by the number of pixels corresponding to the second reference region to calculate the second occupancy rate. As another example, the information processing system 100 can calculate a score for the first reference region and the target lesion Thereafter, the information processing system 100 can calculate a first occupancy rate occupied by the identified target lesion in the first reference region and a second occupancy rate occupied by the identified target lesion in the second reference region. For example, the information processing system 100 divides the number of pixels of the region occupied by the target lesion in the first reference region by the number of pixels corresponding to the first reference region to calculate the first occupancy rate, and divides the number of pixels of the region occupied by the target lesion in the second reference region by the number of pixels corresponding to the second reference region to calculate the second occupancy rate. As another example, the information processing system 100 can calculate a score for the first reference region and the target lesion Thereafter, the information processing system 100 can calculate a first occupancy rate occupied by the identified target lesion in the first reference region and a second occupancy rate occupied by the identified target lesion in the second reference region. For example, the information processing system 100 divides the number of pixels of the region occupied by the target lesion in the first reference region by the number of pixels corresponding to the first reference region to calculate the first occupancy rate, and divides the number of pixels of the region occupied by the target lesion in the second reference region by the number of pixels corresponding to the second reference region to calculate the second occupancy rate. As another example, the information processing system 100 can calculate a score for the first reference region and the target lesion Based on the score for [the first reference region], a first occupancy rate is calculated, and based on the score for the second reference region and the score for the target lesion, a second occupancy rate can be calculated. The information processing system 10 0 can measure the change in the size of the target lesion based on the calculated occupancy rate. For example, the information processing system 100 can, based on the first occupancy rate of the first reference region occupied by the identified target lesion, and the second occupancy rate of the second reference region occupied by the identified target lesion, determine whether there is a change in the size of the target lesion. That is, the information processing system 100 can measure the change in the size of the target lesion based on the change in the occupancy rate, rather than the absolute size of the target lesion included in the X-ray image.

[0052] In one embodiment, the information processing system 100 calculates the amount of change in the occupancy rate for the target lesion based on the first occupancy rate and the second occupancy rate, and compares the calculated amount of change in the occupancy rate for the target lesion with a reference value to determine whether there is a change in the size of the target lesion. For example, the reference value can include a first reference value and a second reference value. At this time, if the calculated amount of change in the occupancy rate for the target lesion is the same as or greater than the first reference value, it can be determined that the size of the target lesion has increased. If the calculated amount of change in the occupancy rate for the target lesion is less than the first reference value and the same as or greater than the second reference value, it can be determined that there is no change in the size of the target lesion. If the calculated amount of change in the occupancy rate for the target lesion is less than the second reference value, it can be determined that the size of the target lesion has decreased. Additionally or alternatively, the reference value can be determined based on a value related to the target metric calculated for the test set, or a value related to the prediction accuracy. ​​​​​​​​​​​​

[0053] In other embodiments, the information processing system 100 calculates the change amount of the occupancy rate for the target lesion based on the first occupancy rate and the second occupancy rate, and inputs the calculated change amount of the occupancy rate for the target lesion into the change presence / absence determination model, so that the presence / absence of a change in the size of the target lesion can be determined based on the output determination result. For this purpose, the information processing system 100 and / or the storage system can include a change presence / absence determination model, and the change presence / absence determination model can include a machine learning model trained to output a determination result regarding the presence / absence of a change in the size of the reference target lesion based on an input value for the change amount of the reference occupancy rate for the target lesion.

[0054] FIG. 2 is a block diagram showing the internal configuration of the information processing system 100 according to an embodiment of the present disclosure. The information processing system 100 can include a memory 210, a processor 220, a communication module 230, and an input / output interface 240. As shown in FIG. 2, the information processing system 100 can be configured to communicate information and / or data via a network using the communication module 230.

[0055] The memory 210 can include any non-transitory computer-readable recording medium. According to one embodiment, the memory 210 can include a RAM (random access memory), a ROM ( read only memory), a disk drive, an SSD (solid state drive), and a permanent mass storage device such as a flash memory (flash memory). As another example, the ROM, SSD, flash memory, and ​​​​​​​​​​A permanent large-capacity storage device such as a disk drive is a separate one that is classified from the memory. It can be included in the information processing system 100 as a permanent storage device. Also, in the memory 21 0, at least one program code (for example, an application for measuring the change in the size of the target lesion, a program for determining a reference region from an X-ray image , a program for identifying a target lesion from an X-ray image, a program for calculating the occupancy rate of the target lesion, code for such as a program, etc.) can be stored. , a program for identifying a target lesion from an X-ray image, a program for calculating the occupancy rate of the target lesion, a program for such as a program, etc.) can be stored.

[0056] Such software components can be loaded from a computer-readable recording medium separate from the memory 210. Such a separate computer-readable recording medium can include a recording medium directly connectable to such an information processing system 100 and can include, for example, a computer-readable recording medium such as a floppy drive, a disk, a tape, a DVD / CD-ROM drive, and a memory card. As another example, software components, etc., can also be loaded into the memory 210 via the communication module 230. For example , at least one program is installed by a file provided by a file distribution system that distributes developer or application installation files via the communication module 230 . For example, a computer program (for example, an application for measuring the change in the size of the target lesion, a program for determining a reference region from an X-ray image, a program for identifying a target lesion from an X-ray image, a program for calculating the occupancy rate of the target lesion, etc.) based on the file can be loaded into the memory 210. , a program for determining a reference region from an X-ray image, a program for identifying a target lesion from an X-ray image, a program for calculating the occupancy rate of the target lesion, etc.) based on the file can be loaded into the memory 210.

[0057] The processor 220 can be configured to process instructions of a computer program by performing basic arithmetic, logic, and input / output operations. The instructions can be provided by a user terminal (not shown) or another external system via the memory 210 or the communication module 230. For example, the processor 220 can be configured to execute instructions received by a program code stored in a recording device such as the memory 210.

[0058] The communication module 230 can provide a configuration and function for the information processing system 100 and the user terminal and / or the imaging device to communicate with each other via a network, and can provide a configuration and function for the information processing system 100 to communicate with a storage device and / or another system (for example, a separate cloud system, etc.). As an example, control signals, instructions, data, etc. provided under the control of the processor 220 of the information processing system 100 can pass through the communication module 230 and the network and be provided to the user terminal via the communication module of the user terminal. For example, the information processing system 100 can receive an X-ray image (for example, a first X-ray image including a target lesion and a second X-ray image including the target lesion, etc.) including the target lesion from an external device (for example, a storage device, an imaging device, an external system, etc.) via the communication module 230. Additionally, the information processing system 100 can provide the measurement result of the change in the size of the target lesion to the user terminal via the communication module 230.

[0059] Also, the input / output interface 240 of the information processing system 100 either connected to 00 or can be means for an interface with a device (not shown) for input or output that the information processing system 100 can include. In FIG. 2, the input / output interface 240 is shown as a separately configured element from the processor 220, but this is not limited thereto, and the input / output interface 240 can be configured to be included in the processor 220. The information processing system 100 can include more constituent elements than those shown in FIG. 2. However, it is not necessary to clearly show most of the conventional constituent elements. The processor 220 of the information processing system 100 can be configured to manage, process, and / or store information and / or data received from a plurality of user terminals and / or a plurality of external systems. According to one embodiment, the processor 220 can store, process, and transfer a received first X-ray image, a second X-ray image, etc. For example, the processor 220 can calculate the occupancy rate of the area corresponding to the target lesion among the reference areas in each of the first X-ray image and the second X-ray image. Further, the processor 220 can measure the change in the size of the target lesion based on the calculated

[0060] occupancy rate and transfer the measurement result information to the user terminal. occupancy rate and transfer the measurement result information to the user terminal. 220 can calculate the occupancy rate of the area corresponding to the target lesion among the reference areas in each of the first X-ray image and the second X-ray image. Further, the processor 220 can measure the change in the size of the target lesion based on the calculated occupancy rate and transfer the measurement result information to the user terminal. occupancy rate and transfer the measurement result information to the user terminal. occupancy rate and transfer the measurement result information to the user terminal.

[0061] FIG. 3 is a flowchart showing a method 300 for measuring the change in the size of a target lesion in an X-ray image according to an embodiment of the present disclosure. In one embodiment, the method 300 for measuring the change in the size of a target lesion can be performed by a processor (e.g., at least one processor of the information processing system). The method 300 for measuring the change in the size of a target lesion can be performed by the processor (e.g., at least one processor of the information processing system). The method 300 for measuring the change in the size of a target lesion can be performed by the processor (e.g., at least one processor of the information processing system). The method 300 for measuring the change in the size of a target lesion can be performed by the processor ​The processor can start by receiving a first X-ray image including the target lesion and a second X-ray image including the target lesion (S310). It can start by receiving a first X-ray image including the target lesion and a second X-ray image including the target lesion (S310).

[0062] The processor can calculate the occupancy rate of the area corresponding to the target lesion among the reference areas in each of the first X-ray image and the second X-ray image (S320). Here, the reference area can refer to the area determined by dividing the entire area of each of the first X-ray image and the second X-ray image into a plurality of areas. In one embodiment, the processor determines a first reference area and a second reference area from each of the first X-ray image and the second X-ray image, identifies the target lesion from each of the first X-ray image and the second X-ray image, and calculates a first occupancy rate of the identified target lesion in the first reference area and a second occupancy rate of the identified target lesion in the second reference area. For this purpose, the processor can input the first X-ray image into a reference area extraction model to output the first reference area, and input the second X-ray image into the reference area extraction model to output the second reference area. Here, the second reference area corresponds to the first reference area, and the reference area extraction model can be learned using a plurality of reference X-ray images and label information regarding the reference areas. It can calculate the occupancy rate of the area corresponding to the target lesion among the reference areas in each of the first X-ray image and the second X-ray image (S320). Here, the reference area can refer to the area determined by dividing the entire area of each of the first X-ray image and the second X-ray image into a plurality of areas. In one embodiment, the processor determines a first reference area and a second reference area from each of the first X-ray image and the second X-ray image, identifies the target lesion from each of the first X-ray image and the second X-ray image, and calculates a first occupancy rate of the identified target lesion in the first reference area and a second occupancy rate of the identified target lesion in the second reference area. For this purpose, the processor can input the first X-ray image into a reference area extraction model to output the first reference area, and input the second X-ray image into the reference area extraction model to output the second reference area. Here, the second reference area corresponds to the first reference area, and the reference area extraction model can be learned using a plurality of reference X-ray images and label information regarding the reference areas. It can calculate the occupancy rate of the area corresponding to the target lesion among the reference areas in each of the first X-ray image and the second X-ray image (S320). Here, the reference area can refer to the area determined by dividing the entire area of each of the first X-ray image and the second X-ray image into a plurality of areas. In one embodiment, the processor determines a first reference area and a second reference area from each of the first X-ray image and the second X-ray image, identifies the target lesion from each of the first X-ray image and the second X-ray image, and calculates a first occupancy rate of the identified target lesion in the first reference area and a second occupancy rate of the identified target lesion in the second reference area. For this purpose, the processor can input the first X-ray image into a reference area extraction model to output the first reference area, and input the second X-ray image into the reference area extraction model to output the second reference area. Here, the second reference area corresponds to the first reference area, and the reference area extraction model can be learned using a plurality of reference X-ray images and label information regarding the reference areas. It can calculate the occupancy rate of the area corresponding to the target lesion among the reference areas in each of the first X-ray image and the second X-ray image (S320). Here, the reference area can refer to the area determined by dividing the entire area of each of the first X-ray image and the second X-ray image into a plurality of areas. In one embodiment, the processor determines a first reference area and a second reference area from each of the first X-ray image and the second X-ray image, identifies the target lesion from each of the first X-ray image and the second X-ray image, and calculates a first occupancy rate of the identified target lesion in the first reference area and a second occupancy rate of the identified target lesion in the second reference area. For this purpose, the processor can input the first X-ray image into a reference area extraction model to output the first reference area, and input the second X-ray image into the reference area extraction model to output the second reference area. Here, the second reference area corresponds to the first reference area, and the reference area extraction model can be learned using a plurality of reference X-ray images and label information regarding the reference areas. It can calculate the occupancy rate of the area corresponding to the target lesion among the reference areas in each of the first X-ray image and the second X-ray image (S320). Here, the reference area can refer to the area determined by dividing the entire area of each of the first X-ray image and the second X-ray image into a plurality of areas. In one embodiment, the processor determines a first reference area and a second reference area from each of the first X-ray image and the second X-ray image, identifies the target lesion from each of the first X-ray image and the second X-ray image, and calculates a first occupancy rate of the identified target lesion in the first reference area and a second occupancy rate of the identified target lesion in the second reference area. For this purpose, the processor can input the first X-ray image into a reference area extraction model to output the first reference area, and input the second X-ray image into the reference area extraction model to output the second reference area. Here, the second reference area corresponds to the first reference area, and the reference area extraction model can be learned using a plurality of reference X-ray images and label information regarding the reference areas. It can calculate the occupancy rate of the area corresponding to the target lesion among the reference areas in each of the first X-ray image and the second X-ray image (S320). Here, the reference area can refer to the area determined by dividing the entire area of each of the first X-ray image and the second X-ray image into a plurality of areas. In one embodiment, the processor determines a first reference area and a second reference area from each of the first X-ray image and the second X-ray image, identifies the target lesion from each of the first X-ray image and the second X-ray image, and calculates a first occupancy rate of the identified target lesion in the first reference area and a second occupancy rate of the identified target lesion in the second reference area. For this purpose, the processor can input the first X-ray image into a reference area extraction model to output the first reference area, and input the second X-ray image into the reference area extraction model to output the second reference area. Here, the second reference area corresponds to the first reference area, and the reference area extraction model can be learned using a plurality of reference X-ray images and label information regarding the reference areas. It can calculate the occupancy rate of the area corresponding to the target lesion among the reference areas in each of the first X-ray image and the second X-ray image (S320). Here, the reference area can refer to the area determined by dividing the entire area of each of the first X-ray image and the second X-ray image into a plurality of areas. In one embodiment, the processor determines a first reference area and a second reference area from each of the first X-ray image and the second X-ray image, identifies the target lesion from each of the first X-ray image and the second X-ray image, and calculates a first occupancy rate of the identified target lesion in the first reference area and a second occupancy rate of the identified target lesion in the second reference area. For this purpose, the processor can input the first X-ray image into a reference area extraction model to output the first reference area, and input the second X-ray image into the reference area extraction model to output the second reference area. Here, the second reference area corresponds to the first reference area, and the reference area extraction model can be learned using a plurality of reference X-ray images and label information regarding the reference areas. It can calculate the occupancy rate of the area corresponding to the target lesion among the reference areas in each of the first X-ray image and the second X-ray image (S320). Here, the reference area can refer to the area determined by dividing the entire area of each of the first X-ray image and the second X-ray image into a plurality of areas. In one embodiment, the processor determines a first reference area and a second reference area from each of the first X-ray image and the second X-ray image, identifies the target lesion from each of the first X-ray image and the second X-ray image, and calculates a first occupancy rate of the identified target lesion in the first reference area and a second occupancy rate of the identified target lesion in the second reference area. For this purpose, the processor can input the first X-ray image into a reference area extraction model to output the first reference area, and input the second X-ray image into the reference area extraction model to output the second reference area. Here, the second reference area corresponds to the first reference area, and the reference area extraction model can be learned using a plurality of reference X-ray images and label information regarding the reference areas. It can calculate the occupancy rate of the area corresponding to the target lesion among the reference areas in each of the first X-ray image and the second X-ray image (S320). Here, the reference area can refer to the area determined by dividing the entire area of each of the first X-ray image and the second X-ray image into a plurality of areas. In one embodiment, the processor determines a first reference area and a second reference area from each of the first X-ray image and the second X-ray image, identifies the target lesion from each of the first X-ray image and the second X-ray image, and calculates a first occupancy rate of the identified target lesion in the first reference area and a second occupancy rate of the identified target lesion in the second reference area. For this purpose, the processor can input the first X-ray image into a reference area extraction model to output the first reference area, and input the second X-ray image into the reference area extraction model to output the second reference area. Here, the second reference area corresponds to the first reference area, and the reference area extraction model can be learned using a plurality of reference X-ray images and label information regarding the reference areas. It can calculate the occupancy rate of the area corresponding to the target lesion among the reference areas in each of the first X-ray image and the second X-ray image (S320). Here, the reference area can refer to the area determined by dividing the entire area of each of the first X-ray image and the second X-ray image into a plurality of areas. In one embodiment, the processor determines a first reference area and a second reference area from each of the first X-ray image and the second X-ray image, identifies the target lesion from each of the first X-ray image and the second X-ray image, and calculates a first occupancy rate of the identified target lesion in the first reference area and a second occupancy rate of the identified target lesion in the second reference area. For this purpose, the processor can input the first X-ray image into a reference area extraction model to output the first reference area, and input the second X-ray image into the reference area extraction model to output the second reference area. Here, the second reference area corresponds to the first reference area, and the reference area extraction model can be learned using a plurality of reference X-ray images and label information regarding the reference areas. Here, the second reference area corresponds to the first reference area, and the reference area extraction model can be learned using a plurality of reference X-ray images and label information regarding the reference areas. Here, the second reference area corresponds to the first reference area, and the reference area extraction model can be learned using a plurality of reference X-ray images and label information regarding the reference areas. For example, the processor can calculate the first occupancy rate by dividing the number of pixels of the area occupied by the target lesion within the first reference area by the number of pixels corresponding to the first reference area, and calculate the second occupancy rate by dividing the number of pixels of the area occupied by the target lesion within the second reference area by the number of pixels corresponding to the second reference area. For example, the processor can calculate the first occupancy rate by dividing the number of pixels of the area occupied by the target lesion within the first reference area by the number of pixels corresponding to the first reference area, and calculate the second occupancy rate by dividing the number of pixels of the area occupied by the target lesion within the second reference area by the number of pixels corresponding to the second reference area. For example, the processor can calculate the first occupancy rate by dividing the number of pixels of the area occupied by the target lesion within the first reference area by the number of pixels corresponding to the first reference area, and calculate the second occupancy rate by dividing the number of pixels of the area occupied by the target lesion within the second reference area by the number of pixels corresponding to the second reference area. For example, the processor can calculate the first occupancy rate by dividing the number of pixels of the area occupied by the target lesion within the first reference area by the number of pixels corresponding to the first reference area, and calculate the second occupancy rate by dividing the number of pixels of the area occupied by the target lesion within the second reference area by the number of pixels corresponding to the second reference area.

[0063] The processor can measure the change in size of the target lesion based on the calculated occupancy rate (S 330). In one embodiment, the processor determines based on the first occupancy rate and the second occupancy rate: For example, the processor may determine whether the size of the target lesion has changed. The change in the occupancy rate relative to the change (e.g., the value obtained by subtracting the first occupancy rate from the second occupancy rate) By comparing with the reference value, it is possible to determine whether or not there has been a change in the size of the target lesion. includes a first reference value and a second reference value, and subtracts the first occupancy rate from the second occupancy rate. If the value obtained is the same as or larger than the first reference value, the size of the target lesion is considered to have increased. The value obtained by subtracting the first occupancy rate from the second occupancy rate can be determined based on the first reference value. If it is small and the same or larger than the second standard value, there is no change in the size of the target lesion. It can be determined that the value obtained by subtracting the first occupancy rate from the second occupancy rate is equal to the second reference value. If the value is smaller than the reference value, it can be determined that the size of the target lesion has decreased. A numerical value associated with the target metric calculated for the set, or a value associated with the prediction accuracy. As another example, the processor may determine the first occupancy rate and the second occupancy rate based on a related value. By inputting the difference between the occupancy rates into the change judgment model, the output judgment result is , it can determine whether the size of the target lesion has changed. Here, the change determination model is based on the reference occupancy Based on the input value for the rate difference, a judgment result is output as to whether or not there is a change in the size of the reference lesion. The data may include a machine learning model trained to

[0064] FIG. 4 illustrates a process of training a reference region extraction model via a user terminal 420 according to an embodiment of the present disclosure. It is a diagram showing an example in which label information for [a certain purpose] is provided to the information processing system 100. X-ray Images can include regions that are commonly included depending on the purpose (for example, target tissues, organs, organ systems, viscera, etc.). According to one embodiment, in the case of a chest X-ray image, the position of the patient or the imaging equipment is adjusted so that the lung region is included in the image (or video), and then the image is taken. To measure the change in the size of the target lesion, a region corresponding to the reference region can be set in a region that can commonly exist in X-ray images (for example, past X-ray images, current X-ray images). Here, the reference region can be a region determined by dividing the entire region of each X-ray image into a plurality of regions. For example, at least one region of the six independent regions (「Upper Right region」, 「Upp er Left region」, 「Mid Right region」, 「Mid Left region」, 「L ower Right region」, 「Lower Left region」) used to read out the lung region from a chest X-ray image can be set as the reference region. That is, each of the six independent regions, or a region formed by fusing a part of the six independent regions (e.g., the entire lung region, the left / right region of the lung, the upper / middle / lower region of the lung) can correspond to the reference region. To calculate the occupancy rate of the target lesion with respect to the reference region, the information processing system (for example, at least one processor of the information processing system) 100 can determine the reference region from the X-ray image through the reference region extraction model. For this purpose, the information processing system 100 can generate / learn the reference region extraction model. In one embodiment, the information processing system 100 can learn an algorithm for searching for the reference region from each X-ray image by a machine learning method. er Left region」, 「Mid Right region」, 「Mid Left region」, 「L ower Right region」, 「Lower Left region」) of the reference region including at least one region can be set. That is, each of the six independent regions, or a region in which a part of the six independent regions is fused (e.g., the entire lung region, the left / right region of the lung, the upper / middle / lower region of the lung) can correspond to the reference region. region of the lung) can correspond to the reference region.

[0065] To calculate the occupancy rate of the target lesion with respect to the reference region, the information processing system (for example, at least one processor of the information processing system) 100 can determine the reference region from the X-ray image through the reference region extraction model. For this reason, the information processing system 100 can generate / learn the reference region extraction model. In one embodiment, the information processing system 100 can learn an algorithm for searching for the reference region from each X-ray image by a machine learning method. For this purpose, the information processing system 100 can generate / learn the reference region extraction model. In one embodiment, the information processing system 100 can learn an algorithm for searching for the reference region from each X-ray image by a machine learning method. For this purpose, the information processing system 100 can generate / learn the reference region extraction model. In one embodiment, the information processing system 100 can learn an algorithm for searching for the reference region from each X-ray image by a machine learning method. For this purpose, the information processing system 100 can generate / learn the reference region extraction model. In one embodiment, the information processing system 100 can learn an algorithm for searching for the reference region from each X-ray image by a machine learning method. For this purpose, the information processing system 100 can generate / learn the reference region extraction model. In one embodiment, the information processing system 100 can learn an algorithm for searching for the reference region from each X-ray image by a machine learning method. For example, the reference region extraction model can be referred to as a segmentation artificial neural network model. The reference region extraction model generated / learned in this way can be stored in an information processing system and / or a storage system. To generate / learn the reference region extraction model, the information processing system 100 can be configured to be communicably connected to the user terminal 420 and / or the storage system 410.

[0066] In one embodiment, to learn the reference region extraction model, the information processing system 100 can output a reference X-ray image to be annotated via the user terminal 420. Here, the reference X-ray image to be annotated can be received from the storage system 410. Then, the user (e.g., an annotator, etc.) can divide the entire region (e.g., the lung region) included in the reference X-ray image output via the user terminal 420 into a plurality of regions (e.g., Upper Right region, Mid Right region, Lower Right region, Upper Left region, Mid Left region, Lower Left region, etc.), determine label information (e.g., annotation information) for each of the regions, and provide it to the information processing system 100. For example, the user (e.g., an annotator, etc.) can provide the information processing system 100 with label information regarding the reference reference region via the user terminal 420. Here, the reference reference region can include at least one region among the plurality of regions for which the change in the size of the target lesion is to be measured. The information processing system 100 can generate and / or learn the reference region extraction model using the received reference X-ray image and the label information regarding the reference reference region. (e.g., Upper Right region, Mid Right region, Lower Right region, Upper Left region, Mid Left region, Lower Left region, etc.) and provide it to the information processing system 100. For example, the user (e.g., an annotator, etc.) can provide the information processing system 100 with label information regarding the reference reference region via the user terminal 420. Here, the reference reference region can include at least one region among the plurality of regions for which the change in the size of the target lesion is to be measured. The information processing system 100 can generate and / or learn the reference region extraction model using the received reference X-ray image and the label information regarding the reference reference region. ​​​​​​​​​​​​​​​​

[0067] For use as learning materials for a reference region extraction model, the information processing system 100 can provide a plurality of chest X-ray images to the user terminal 420 as learning images. The user can perform annotations on six regions of the lungs included in the plurality of chest X-ray images via the user terminal 420, and as an annotation result, provide the information processing system 100 with a plurality of learning images including label information regarding the six regions of the lungs. For example, the plurality of learning images can include a learning image 430 including label information regarding the six regions of the lungs. The information processing system 100 can generate / learn a model that determines each of the six regions of the lungs from a chest X-ray image based on the plurality of learning images including the learning image 430 including label information. Additionally or alternatively, the information processing system 100 can generate and / or learn a model that determines a reference region from a chest X-ray image based on the learning images and label information. Here, the reference region can include at least one of the six regions of the lungs.

[0068] In one embodiment, the information processing system 100 learns an algorithm into which one X-ray image is input, and can be learned in such a way as to minimize the loss with respect to the label information (i.e., annotation information) regarding the six regions of the lungs. For example, the information processing system 10 0 can calculate the probability that each pixel of the image corresponds to each region using the label information regarding the six regions of the lungs in the X-ray image. At this time, in order to distinguish between the left / right of each region, it can be assumed that the image is not flipped. Also, in order to remove noise ​​​​​​​​​​​​​For this purpose, a reference value (threshold) can be set, and values lower than the said value can be clipped. )

[0069] In FIG. 4, one user terminal 420 is shown, but it is not limited thereto, and a plurality of user terminals 420 can be configured to be connected to and communicate with the information processing system 100. Also , in FIG. 4, the storage system 410 is shown as one device, but it is not limited thereto , and it can be composed of a plurality of storage devices or composed of a system that supports a cloud. Also, in FIG. 4, each component of the system that generates / learns the reference region extraction model indicates a functional element that is functionally divided, and a plurality of components can be embodied in a form integrated with each other in an actual physical environment. In FIG. 4, the information processing system 10 0 and the storage system 410 are shown as separate systems, but it is not limited thereto, and they can also be configured to be integrated as one system.

[0070] FIG. 5 is a diagram showing an example in which the information processing system 100 according to an embodiment of the present disclosure determines a reference region from an X-ray image and outputs it via the user terminal 520. To measure the change in the size of a target lesion, the information processing system (for example, at least one processor of the information processing system) 100 can determine a reference region from the X-ray image. Here, the reference region can be determined by dividing the entire region of the X-ray image into a plurality of regions. In one embodiment , the information processing system 100 can determine a first reference region and a second reference region from each of the first X-ray image and the second X-ray image. In one embodiment, the information processing system ​​​​​100 inputs the first X-ray image into the reference region extraction model to output the first reference region and can input the second X-ray image into the reference region extraction model to output the second reference region. Here, the second reference region can correspond to the first reference region. For example, if the region corresponding to the left lung in the first X-ray image corresponds to the first reference region, then in the second X-ray image, the region corresponding to the left lung can also correspond to the second reference region.

[0071] The information processing system 100 can receive the target X-ray image (for example, the first X-ray image and the second X-ray image) for measuring the change in the size of the target lesion. For example, the information processing system 100 can receive the target X-ray image from the storage system 510, the user terminal 520, and / or the imaging device. The information processing system 100 can determine the reference region from the received target X-ray image. For example, the information processing system 100 can determine the reference region by dividing the entire region included in the target X-ray image into a plurality of regions. As shown in the figure, the information processing system 100 can divide the lung region included in the target X-ray image into six regions (for example, "Upper Right region", "Upper Left region", "Mid Ri ght region", "Mid Left region", "Lower Right region", "Low er Left region") to determine the reference region. Here, the reference region can include at least a part of the plurality of regions. According to one embodiment, the tissue (for example, lung) region included in the target X-ray image can be divided into a plurality of regions, and any combination of the divided plurality of regions can be determined as the reference region. At this time, such a combination of the plurality of regions can be divided into a plurality of regions, and any combination of the divided plurality of regions can be determined as the reference region. At this time, such a combination of the plurality of regions The occupancy rate can be calculated by calculating the ratio of the lesion size within the reference region. For example, the reference area can be the "Upper Right area", the "Mid Right area", The right lung region can be determined as the right lung region including the "Lower Right Region."

[0072] In FIG. 5, the information processing system 100 displays an x-ray image 53 showing multiple regions of the lungs. 0 is output via the user terminal 520, but this is not limiting. For example, The information processing system 100 divides the entire area into a plurality of areas to determine a reference area, and After the entire process of measuring the change in size of the object is completed, only the measurement results of the change in size are available to the user. The information processing system 100 can output the information via the terminal 520. The X-ray image in which the region or the reference region is displayed is not output via the user terminal 520. It is also possible.

[0073] Although one user terminal 520 is shown in FIG. 5, the present invention is not limited to this. The user terminal 520 may be configured to be connected to and communicate with the information processing system 100. Also, in FIG. 5, the storage system 510 is shown as a single device, but is not limited to this. It is not a system that consists of multiple storage devices or supports cloud computing. In addition, Fig. 5 shows a system for generating / learning a reference region extraction model. Each component of indicates a functional component that is functionally divided, and a plurality of components In Figure 5, the information processing system Although the system 100 and the storage system 510 are shown as separate systems, the present invention is not limited thereto. They can also be integrated into a single system.

[0074] FIG. 6 shows an example of calculating a first occupancy rate 628 from a first X-ray image 620 and a second occupancy rate 638 from a second X-ray image 630 according to an embodiment of the present disclosure. The processor (e.g., at least one processor of the information processing system) 610 can receive a first X-ray image 620 including a target lesion and a second X-ray image 630 including the target lesion, and calculate the occupancy rate of the region corresponding to the target lesion among the reference regions in each of the first X-ray image 620 and the second X-ray image 630. Here, the occupancy rate can refer to the size (e.g., the number of pixels, etc.) of the lesion located within the reference region with respect to the size of the reference region (e.g., the number of pixels, etc.). For example, when the left lung corresponds to the reference region, the occupancy rate of the target lesion with respect to the left lung can refer to the size (e.g., the number of pixels, etc.) of the target lesion located in the left lung with respect to the size (e.g., the number of pixels, etc.) of the left lung in the X-ray image. In one embodiment, the processor 610 can determine a first reference region 626 and a second reference region 636 from each of the first X-ray image 620 and the second X-ray image 630, and identify the target lesion from each of the first X-ray image 620 and the second X-ray image 630. Thereafter, the processor 610 can calculate a first occupancy rate 628 occupied by the identified target lesion among the first reference region 626 and a second occupancy rate 638 occupied by the identified target lesion among the second reference region 636. For example, as shown in the following equations 1 and 2, the processor 610 can calculate the number of pixels of the region 622 occupied by the target lesion in the first X-ray image 620. The processor (e.g., at least one processor of the information processing system) 610 can receive a first X-ray image 620 including a target lesion and a second X-ray image 630 including the target lesion, and calculate the occupancy rate of the region corresponding to the target lesion among the reference regions in each of the first X-ray image 620 and the second X-ray image 630. Here, the occupancy rate can refer to the size (e.g., the number of pixels, etc.) of the lesion located within the reference region with respect to the size of the reference region (e.g., the number of pixels, etc.). For example, when the left lung corresponds to the reference region, the occupancy rate of the target lesion with respect to the left lung can refer to the size (e.g., the number of pixels, etc.) of the target lesion located in the left lung with respect to the size (e.g., the number of pixels, etc.) of the left lung in the X-ray image. The processor (e.g., at least one processor of the information processing system) 610 can receive a first X-ray image 620 including a target lesion and a second X-ray image 630 including the target lesion, and calculate the occupancy rate of the region corresponding to the target lesion among the reference regions in each of the first X-ray image 620 and the second X-ray image 630. Here, the occupancy rate can refer to the size (e.g., the number of pixels, etc.) of the lesion located within the reference region with respect to the size of the reference region (e.g., the number of pixels, etc.). For example, when the left lung corresponds to the reference region, the occupancy rate of the target lesion with respect to the left lung can refer to the size (e.g., the number of pixels, etc.) of the target lesion located in the left lung with respect to the size (e.g., the number of pixels, etc.) of the left lung in the X-ray image. The occupancy rate can refer to the size (e.g., the number of pixels, etc.) of the lesion located within the reference region with respect to the size of the reference region (e.g., the number of pixels, etc.). For example, when the left lung corresponds to the reference region, the occupancy rate of the target lesion with respect to the left lung can refer to the size (e.g., the number of pixels, etc.) of the target lesion located in the left lung with respect to the size (e.g., the number of pixels, etc.) of the left lung in the X-ray image. The occupancy rate can refer to the size (e.g., the number of pixels, etc.) of the lesion located within the reference region with respect to the size of the reference region (e.g., the number of pixels, etc.). For example, when the left lung corresponds to the reference region, the occupancy rate of the target lesion with respect to the left lung can refer to the size (e.g., the number of pixels, etc.) of the target lesion located in the left lung with respect to the size (e.g., the number of pixels, etc.) of the left lung in the X-ray image. For example, when the left lung corresponds to the reference region, the occupancy rate of the target lesion with respect to the left lung can refer to the size (e.g., the number of pixels, etc.) of the target lesion located in the left lung with respect to the size (e.g., the number of pixels, etc.) of the left lung in the X-ray image. For example, when the left lung corresponds to the reference region, the occupancy rate of the target lesion with respect to the left lung can refer to the size (e.g., the number of pixels, etc.) of the target lesion located in the left lung with respect to the size (e.g., the number of pixels, etc.) of the left lung in the X-ray image. For example, when the left lung corresponds to the reference region, the occupancy rate of the target lesion with respect to the left lung can refer to the size (e.g., the number of pixels, etc.) of the target lesion located in the left lung with respect to the size (e.g., the number of pixels, etc.) of the left lung in the X-ray image.

[0075] In one embodiment, the processor 610 can determine a first reference region 626 and a second reference region 636 from each of the first X-ray image 620 and the second X-ray image 630, and identify the target lesion from each of the first X-ray image 620 and the second X-ray image 630. Thereafter, the processor 610 can calculate a first occupancy rate 628 occupied by the identified target lesion among the first reference region 626 and a second occupancy rate 638 occupied by the identified target lesion among the second reference region 636. For example, as shown in the following equations 1 and 2, the processor 610 can calculate the number of pixels of the region 622 occupied by the target lesion in the first X-ray image 620. Thereafter, the processor 610 can calculate a first occupancy rate 628 occupied by the identified target lesion among the first reference region 626 and a second occupancy rate 638 occupied by the identified target lesion among the second reference region 636. For example, as shown in the following equations 1 and 2, the processor 610 can calculate the number of pixels of the region 622 occupied by the target lesion in the first X-ray image 620. Thereafter, the processor 610 can calculate a first occupancy rate 628 occupied by the identified target lesion among the first reference region 626 and a second occupancy rate 638 occupied by the identified target lesion among the second reference region 636. For example, as shown in the following equations 1 and 2, the processor 610 can calculate the number of pixels of the region 622 occupied by the target lesion in the first X-ray image 620. Thereafter, the processor 610 can calculate a first occupancy rate 628 occupied by the identified target lesion among the first reference region 626 and a second occupancy rate 638 occupied by the identified target lesion among the second reference region 636. For example, as shown in the following equations 1 and 2, the processor 610 can calculate the number of pixels of the region 622 occupied by the target lesion in the first X-ray image 620. Thereafter, the processor 610 can calculate a first occupancy rate 628 occupied by the identified target lesion among the first reference region 626 and a second occupancy rate 638 occupied by the identified target lesion among the second reference region 636. For example, as shown in the following equations 1 and 2, the processor 610 can calculate the number of pixels of the region 622 occupied by the target lesion in the first X-ray image 620. For example, as shown in the following equations 1 and 2, the processor 610 can calculate the number of pixels of the region 622 occupied by the target lesion in the first X-ray image 620. By dividing the number of pixels in the target lesion by the number of pixels corresponding to the first reference region 626, the first occupancy rate 628 can be calculated. The number of pixels in the region 632 occupied by the target lesion in the second X-ray image 630 can be divided by the number of pixels corresponding to the second reference region 636 to calculate the second occupancy rate 638.

[0076]

Number

[0077]

Number

[0078] Based on the occupancy rate calculated as described above, the processor can compare the sizes of the target lesions in the two X-ray images regardless of the size of the X-ray images or the patient's position, and can calculate the change in the size of the target lesion more accurately.

[0079] As shown in the figure, the processor receives the first chest X-ray image 620, determines the lung region (i.e., the entire region) 624 included in the first chest X-ray image 620, and can determine the first reference region 626 corresponding to the left lung region among the determined lung region 624. Also the processor can identify the target lesion region 622 included in the first chest X-ray image 620. Here, the target lesion region can refer to the region occupied by the target lesion in the reference region. After that, the processor can calculate the first occupancy rate 628 by dividing the number of pixels in the target lesion region 622 by the number of pixels in the first reference region 6 26. Similarly, the processor receives the second chest X-ray image 630 and includes the second chest X-ray image 630 ​​​Determine the rare lung region 634, and among the determined lung regions 634, a second reference region 636 corresponding to the left lung region can be determined. Also, the processor can identify the target lesion region 632 included in the second chest X-ray image 630. Then, the processor divides the number of pixels of the target lesion region 632 by the number of pixels of the second reference region 636 to calculate a second occupancy rate 638.

[0080] In FIG. 6, an example is shown in which the processor receives the first X-ray image 620 and the second X-ray image 630 respectively and calculates the occupancy rate, but it is not limited thereto. For example, the processor can receive the first X-ray image 620 and the second X-ray image 630 simultaneously and calculate the occupancy rate. Alternatively, the processor can receive the first X-ray image 620 and the second X-ray image 630 sequentially and calculate the occupancy rate.

[0081] FIG. 7 is a diagram showing an example of generating a measurement result 720 of the change in the size of a target lesion based on a first occupancy rate 628 and a second occupancy rate 638 according to an embodiment of the present disclosure. The processor ( for example, at least one processor of the information processing system) can measure the change in the size of the target lesion based on the calculated occupancy rate. For example, as shown in the figure, a change measurement unit 710 included in the processor receives the first occupancy rate 628 and the second occupancy rate 638 calculated as described above, measures the change in the size of the target lesion, and can generate a measurement result 720 of the change in the size of the target lesion. The measurement result 720 of the change in the size of the target lesion includes first occupancy rate information, second occupancy rate information, change amount information of the occupancy rate of the target lesion, and / or information on the change in the size of the target lesion The information may include information about the target lesion (e.g., increase, decrease, no change, etc.). The information on the amount of change in the occupancy rate can be calculated using the following formula 3.

[0082]

number

[0083] The user can not only view the change in the occupancy rate of the target lesion, which is expressed numerically, but also the Information about whether there is a change (e.g., whether there is an increase, decrease, or no change) is required. In one embodiment, the change measurement unit 710 calculates the first occupancy rate 628 and the second occupancy rate 638 For example, the change measurement unit 710 can determine whether or not there has been a change in the size of the target lesion based on the above. The value obtained by subtracting the first occupancy rate 628 from the second occupancy rate 638 is compared with the reference value. In other words, the change measurement unit 710 can determine whether or not there is a change in the size of the lesion. Based on heuristic reference values, it is possible to determine whether or not there has been a change in the size of the target lesion. As an example, the reference values include a first reference value (t1) and a second reference value (t2). , the value obtained by subtracting the first occupancy rate 628 from the second occupancy rate 638 (i.e., the occupancy rate of the target lesion) If the change in the incidence rate is the same as or greater than the first reference value (t1), the target lesion is considered to be large. It can be determined that the size has increased. The number obtained by subtracting the first occupancy rate 628 from the second occupancy rate 638 The value is smaller than the first reference value (t1) and the same as the second reference value (t2) If the second occupancy rate is 638 or larger, it can be determined that there is no change in the size of the target lesion. If the value obtained by subtracting the first occupancy rate 628 is smaller than the second reference value (t2), It can be judged that the size of the lesion has decreased. In other words, some changes can be assumed to be no change. This is because the X-ray image may not exactly correspond to accurate information.

[0084] In one embodiment, the reference value (threshold) is a value related to the target metric calculated for the test set, or a value related to the prediction accuracy. Based on this, it can be determined. That is, to set the first reference value (t1) and the second reference value (t2), the processor receives the test set and can obtain a reference value with a high specific metric (such as auc, accuracy, etc.) for the received test set. For example, the processor can arbitrarily search for the reference value. Alternatively, since there is a curve with an operating point in the AUC (Area Under Curve) metric, the sensitivity and / or specificity of a specific reference value can be calculated. In view of such a point, the processor can search for the point where the sensitivity and / or specificity is maximized and set it as the reference value. In this way, the processor can set the first reference value between increase and no change and the second reference value between decrease and no change. For example, the processor can arbitrarily search for the reference value. Alternatively, since there is a curve with an operating point in the AUC (Area Under Curve) metric, the sensitivity and / or specificity of a specific reference value can be calculated. In view of such a point, the processor can search for the point where the sensitivity and / or specificity is maximized and set it as the reference value. In this way, the processor can set the first reference value between increase and no change and the second reference value between decrease and no change. For example, the processor can arbitrarily search for the reference value. Alternatively, since there is a curve with an operating point in the AUC (Area Under Curve) metric, the sensitivity and / or specificity of a specific reference value can be calculated. In view of such a point, the processor can search for the point where the sensitivity and / or specificity is maximized and set it as the reference value. In this way, the processor can set the first reference value between increase and no change and the second reference value between decrease and no change. example, the processor can arbitrarily search for the reference value. Alternatively, since there is a curve with an operating point in the AUC (Area Under Curve) metric, the sensitivity and / or specificity of a specific reference value can be calculated. In view of such a point, the processor can search for the point where the sensitivity and / or specificity is maximized and set it as the reference value. In this way, the processor can set the first reference value between increase and no change and the second reference value between decrease and no change. For example, the processor can arbitrarily search for the reference value. Alternatively, since there is a curve with an operating point in the AUC (Area Under Curve) metric, the sensitivity and / or specificity of a specific reference value can be calculated. In view of such a point, the processor can search for the point where the sensitivity and / or specificity is maximized and set it as the reference value. In this way, the processor can set the first reference value between increase and no change and the second reference value between decrease and no change. For example, the processor can arbitrarily search for the reference value. Alternatively, since there is a curve with an operating point in the AUC (Area Under Curve) metric, the sensitivity and / or specificity of a specific reference value can be calculated. In view of such a point, the processor can search for the point where the sensitivity and / or specificity is maximized and set it as the reference value. In this way, the processor can set the first reference value between increase and no change and the second reference value between decrease and no change. For example, the processor can arbitrarily search for the reference value. Alternatively, since there is a curve with an operating point in the AUC (Area Under Curve) metric, the sensitivity and / or specificity of a specific reference value can be calculated. In view of such a point, the processor can search for the point where the sensitivity and / or specificity is maximized and set it as the reference value. In this way, the processor can set the first reference value between increase and no change and the second reference value between decrease and no change. For example, the processor can arbitrarily search for the reference value. Alternatively, since there is a curve with an operating point in the AUC (Area Under Curve) metric, the sensitivity and / or specificity of a specific reference value can be calculated. In view of such a point, the processor can search for the point where the sensitivity and / or specificity is maximized and set it as the reference value. In this way, the processor can set the first reference value between increase and no change and the second reference value between decrease and no change. For example, the processor can arbitrarily search for the reference value. Alternatively, since there is a curve with an operating point in the AUC (Area Under Curve) metric, the sensitivity and / or specificity of a specific reference value can be calculated. In view of such a point, the processor can search for the point where the sensitivity and / or specificity is maximized and set it as the reference value. In this way, the processor can set the first reference value between increase and no change and the second reference value between decrease and no change.

[0085] As another example, the change measurement unit 710 can input the difference between the first occupancy rate 628 and the second occupancy rate 638 into the change presence / absence determination model, and based on the output determination result, determine whether there is a change in the size of the target lesion. Here, the difference between the first occupancy rate 628 and the second occupancy rate 638 can be referred to as the value obtained by subtracting the first occupancy rate 628 from the second occupancy rate 638 (that is, the change amount of the occupancy rate for the target lesion). Also, the change presence / absence determination model is a reference occupancy As another example, the change measurement unit 710 can input the difference between the first occupancy rate 628 and the second occupancy rate 638 into the change presence / absence determination model, and based on the output determination result, determine whether there is a change in the size of the target lesion. Here, the difference between the first occupancy rate 628 and the second occupancy rate 638 can be referred to as the value obtained by subtracting the first occupancy rate 628 from the second occupancy rate 638 (that is, the change amount of the occupancy rate for the target lesion). Also, the change presence / absence determination model is a reference occupancy As another example, the change measurement unit 710 can input the difference between the first occupancy rate 628 and the second occupancy rate 638 into the change presence / absence determination model, and based on the output determination result, determine whether there is a change in the size of the target lesion. Here, the difference between the first occupancy rate 628 and the second occupancy rate 638 can be referred to as the value obtained by subtracting the first occupancy rate 628 from the second occupancy rate 638 (that is, the change amount of the occupancy rate for the target lesion). Also, the change presence / absence determination model is a reference occupancy As another example, the change measurement unit 710 can input the difference between the first occupancy rate 628 and the second occupancy rate 638 into the change presence / absence determination model, and based on the output determination result, determine whether there is a change in the size of the target lesion. Here, the difference between the first occupancy rate 628 and the second occupancy rate 638 can be referred to as the value obtained by subtracting the first occupancy rate 628 from the second occupancy rate 638 (that is, the change amount of the occupancy rate for the target lesion). Also, the change presence / absence determination model is a reference occupancy As another example, the change measurement unit 710 can input the difference between the first occupancy rate 628 and the second occupancy rate 638 into the change presence / absence determination model, and based on the output determination result, determine whether there is a change in the size of the target lesion. Here, the difference between the first occupancy rate 628 and the second occupancy rate 638 can be referred to as the value obtained by subtracting the first occupancy rate 628 from the second occupancy rate 638 (that is, the change amount of the occupancy rate for the target lesion). Also, the change presence / absence determination model is a reference occupancy Based on the input value for the rate difference (or the amount of change in the reference occupancy rate for the target lesion), the machine learning model is learned to output a determination result regarding the presence or absence of a change in the size of the reference target lesion. To generate / learn the presence / absence of change determination model, user annotation work may be required. For example, the processor receives the user's label information (increase, decrease, no change) regarding the amount of change in the reference occupancy rate and the presence or absence of a change in the size of the target lesion, and can generate / learn a machine learning model that outputs one of increase, decrease, and no change of the target lesion by inputting the amount of change in the occupancy rate. The change measurement unit 710 can determine the presence or absence of a change in the size of the target lesion using the generated / learned machine learning model (for example, the presence / absence of change determination model). It can include a machine learning model learned to output a determination result regarding the presence or absence of a change in the size of the reference target lesion based on the input value for the rate difference (or the amount of change in the reference occupancy rate for the target lesion). To generate / learn the presence / absence of change determination model, user annotation work may be required. For example, the processor receives the user's label information (increase, decrease, no change) regarding the amount of change in the reference occupancy rate and the presence or absence of a change in the size of the target lesion, and can generate / learn a machine learning model that outputs one of increase, decrease, and no change of the target lesion by inputting the amount of change in the occupancy rate. The change measurement unit 710 can determine the presence or absence of a change in the size of the target lesion using the generated / learned machine learning model (for example, the presence / absence of change determination model). Based on the input value for the rate difference (or the amount of change in the reference occupancy rate for the target lesion), the machine learning model is learned to output a determination result regarding the presence or absence of a change in the size of the reference target lesion. To generate / learn the presence / absence of change determination model, user annotation work may be required. For example, the processor receives the user's label information (increase, decrease, no change) regarding the amount of change in the reference occupancy rate and the presence or absence of a change in the size of the target lesion, and can generate / learn a machine learning model that outputs one of increase, decrease, and no change of the target lesion by inputting the amount of change in the occupancy rate. The change measurement unit 710 can determine the presence or absence of a change in the size of the target lesion using the generated / learned machine learning model (for example, the presence / absence of change determination model). Based on the input value for the rate difference (or the amount of change in the reference occupancy rate for the target lesion), the machine learning model is learned to output a determination result regarding the presence or absence of a change in the size of the reference target lesion. To generate / learn the presence / absence of change determination model, user annotation work may be required. For example, the processor receives the user's label information (increase, decrease, no change) regarding the amount of change in the reference occupancy rate and the presence or absence of a change in the size of the target lesion, and can generate / learn a machine learning model that outputs one of increase, decrease, and no change of the target lesion by inputting the amount of change in the occupancy rate. The change measurement unit 710 can determine the presence or absence of a change in the size of the target lesion using the generated / learned machine learning model (for example, the presence / absence of change determination model). Based on the input value for the rate difference (or the amount of change in the reference occupancy rate for the target lesion), the machine learning model is learned to output a determination result regarding the presence or absence of a change in the size of the reference target lesion. To generate / learn the presence / absence of change determination model, user annotation work may be required. For example, the processor receives the user's label information (increase, decrease, no change) regarding the amount of change in the reference occupancy rate and the presence or absence of a change in the size of the target lesion, and can generate / learn a machine learning model that outputs one of increase, decrease, and no change of the target lesion by inputting the amount of change in the occupancy rate. The change measurement unit 710 can determine the presence or absence of a change in the size of the target lesion using the generated / learned machine learning model (for example, the presence / absence of change determination model). Based on the input value for the rate difference (or the amount of change in the reference occupancy rate for the target lesion), the machine learning model is learned to output a determination result regarding the presence or absence of a change in the size of the reference target lesion. To generate / learn the presence / absence of change determination model, user annotation work may be required. For example, the processor receives the user's label information (increase, decrease, no change) regarding the amount of change in the reference occupancy rate and the presence or absence of a change in the size of the target lesion, and can generate / learn a machine learning model that outputs one of increase, decrease, and no change of the target lesion by inputting the amount of change in the occupancy rate. The change measurement unit 710 can determine the presence or absence of a change in the size of the target lesion using the generated / learned machine learning model (for example, the presence / absence of change determination model). Based on the input value for the rate difference (or the amount of change in the reference occupancy rate for the target lesion), the machine learning model is learned to output a determination result regarding the presence or absence of a change in the size of the reference target lesion. To generate / learn the presence / absence of change determination model, user annotation work may be required. For example, the processor receives the user's label information (increase, decrease, no change) regarding the amount of change in the reference occupancy rate and the presence or absence of a change in the size of the target lesion, and can generate / learn a machine learning model that outputs one of increase, decrease, and no change of the target lesion by inputting the amount of change in the occupancy rate. The change measurement unit 710 can determine the presence or absence of a change in the size of the target lesion using the generated / learned machine learning model (for example, the presence / absence of change determination model). Based on the input value for the rate difference (or the amount of change in the reference occupancy rate for the target lesion), the machine learning model is learned to output a determination result regarding the presence or absence of a change in the size of the reference target lesion. To generate / learn the presence / absence of change determination model, user annotation work may be required. For example, the processor receives the user's label information (increase, decrease, no change) regarding the amount of change in the reference occupancy rate and the presence or absence of a change in the size of the target lesion, and can generate / learn a machine learning model that outputs one of increase, decrease, and no change of the target lesion by inputting the amount of change in the occupancy rate. The change measurement unit 710 can determine the presence or absence of a change in the size of the target lesion using the generated / learned machine learning model (for example, the presence / absence of change determination model).

[0086] FIG. 8 is a diagram showing an example of measuring the change in the size of the target lesion in the chest X-ray images 810 and 820 according to an embodiment of the present disclosure. The processor (for example, at least one processor of the information processing system) can measure the size of the entire region (for example, the lung) and / or the size of the reference region (for example, the right lung) in the X-ray image. Also, the processor can measure the size and position of the target lesion by the CAD (Computer Aided Detection) method or an existing algorithm. Based on the input value for the rate difference (or the amount of change in the reference occupancy rate for the target lesion), the machine learning model is learned to output a determination result regarding the presence or absence of a change in the size of the reference target lesion. To generate / learn the presence / absence of change determination model, user annotation work may be required. For example, the processor receives the user's label information (increase, decrease, no change) regarding the amount of change in the reference occupancy rate and the presence or absence of a change in the size of the target lesion, and can generate / learn a machine learning model that outputs one of increase, decrease, and no change of the target lesion by inputting the amount of change in the occupancy rate. The change measurement unit 710 can determine the presence or absence of a change in the size of the target lesion using the generated / learned machine learning model (for example, the presence / absence of change determination model). Based on the input value for the rate difference (or the amount of change in the reference occupancy rate for the target lesion), the machine learning model is learned to output a determination result regarding the presence or absence of a change in the size of the reference target lesion. To generate / learn the presence / absence of change determination model, user annotation work may be required. For example, the processor receives the user's label information (increase, decrease, no change) regarding the amount of change in the reference occupancy rate and the presence or absence of a change in the size of the target lesion, and can generate / learn a machine learning model that outputs one of increase, decrease, and no change of the target lesion by inputting the amount of change in the occupancy rate. The change measurement unit 710 can determine the presence or absence of a change in the size of the target lesion using the generated / learned machine learning model (for example, the presence / absence of change determination model). Based on the input value for the rate difference (or the amount of change in the reference occupancy rate for the target lesion), the machine learning model is learned to output a determination result regarding the presence or absence of a change in the size of the reference target lesion. To generate / learn the presence / absence of change determination model, user annotation work may be required. For example, the processor receives the user's label information (increase, decrease, no change) regarding the amount of change in the reference occupancy rate and the presence or absence of a change in the size of the target lesion, and can generate / learn a machine learning model that outputs one of increase, decrease, and no change of the target lesion by inputting the amount of change in the occupancy rate. The change measurement unit 710 can determine the presence or absence of a change in the size of the target lesion using the generated / learned machine learning model (for example, the presence / absence of change determination model). Based on the input value for the rate difference (or the amount of change in the reference occupancy rate for the target lesion), the machine learning model is learned to output a determination result regarding the presence or absence of a change in the size of the reference target lesion. To generate / learn the presence / absence of change determination model, user annotation work may be required. For example, the processor receives the user's label information (increase, decrease, no change) regarding the amount of change in the reference occupancy rate and the presence or absence of a change in the size of the target lesion, and can generate / learn a machine learning model that outputs one of increase, decrease, and no change of the target lesion by inputting the amount of change in the occupancy rate. The change measurement unit 710 can determine the presence or absence of a change in the size of the target lesion using the generated / learned machine learning model (for example, the presence / absence of change determination model). Based on the input value for the rate difference (or the amount of change in the reference occupancy rate for the target lesion), the machine learning model is learned to output a determination result regarding the presence or absence of a change in the size of the reference target lesion. To generate / learn the presence / absence of change determination model, user annotation work may be required. For example, the processor receives the user's label information (increase, decrease, no change) regarding the amount of change in the reference occupancy rate and the presence or absence of a change in the size of the target lesion, and can generate / learn a machine learning model that outputs one of increase, decrease, and no change of the target lesion by inputting the amount of change in the occupancy rate. The change measurement unit 710 can determine the presence or absence of a change in the size of the target lesion using the generated / learned machine learning model (for example, the presence / absence of change determination model).

[0087] As shown in Table 830 of FIG. 8, the size of the right lung region (lung_area1) on the first chest X-ray image 810 is measured as 11389.0, and the size of the right lung region (lung_area2) on the second chest X-ray image 820 is measured as 12076.0. That is, even if it corresponds to X-ray images of the lungs of the same individual / target, the size of the images Based on the input value for the rate difference (or the amount of change in the reference occupancy rate for the target lesion), the machine learning model is learned to output a determination result regarding the presence or absence of a change in the size of the reference target lesion. To generate / learn the presence / absence of change determination model, user annotation work may be required. For example, the processor receives the user's label information (increase, decrease, no change) regarding the amount of change in the reference occupancy rate and the presence or absence of a change in the size of the target lesion, and can generate / learn a machine learning model that outputs one of increase, decrease, and no change of the target lesion by inputting the amount of change in the occupancy rate. The change measurement unit 710 can determine the presence or absence of a change in the size of the target lesion using the generated / learned machine learning model (for example, the presence / absence of change determination model). Based on the input value for the rate difference (or the amount of change in the reference occupancy rate for the target lesion), the machine learning model is learned to output a determination result regarding the presence or absence of a change in the size of the reference target lesion. To generate / learn the presence / absence of change determination model, user annotation work may be required. For example, the processor receives the user's label information (increase, decrease, no change) regarding the amount of change in the reference occupancy rate and the presence or absence of a change in the size of the target lesion, and can generate / learn a machine learning model that outputs one of increase, decrease, and no change of the target lesion by inputting the amount of change in the occupancy rate. The change measurement unit 710 can determine the presence or absence of a change in the size of the target lesion using the generated / learned machine learning model (for example, the presence / absence of change determination model). Based on the input value for the rate difference (or the amount of change in the reference occupancy rate for the target lesion), the machine learning model is learned to output a determination result regarding the presence or absence of a change in the size of the reference target lesion. To generate / learn the presence / absence of change determination model, user annotation work may be required. For example, the processor receives the user's label information (increase, decrease, no change) regarding the amount of change in the reference occupancy rate and the presence or absence of a change in the size of the target lesion, and can generate / learn a machine learning model that outputs one of increase, decrease, and no change of the target lesion by inputting the amount of change in the occupancy rate. The change measurement unit 710 can determine the presence or absence of a change in the size of the target lesion using the generated / learned machine learning model (for example, the presence / absence of change determination model). Further, it can be measured such that the size of the lung region varies depending on the position and / or state of the individual / subject. Also, on the first chest X-ray image 810, the size of the target region (area 1) within the right lung region is measured to be approximately 1739.8578, and on the second chest X-ray image 820 the size of the target region (area2) within the right lung region is measured to be approximately 1553.8666. Accordingly, the first occupancy rate is 1739.8578 / 11389.0, which is calculated to be approximately 0.15277, and the second occupancy rate is 1553.8666 / 12076 which is calculated to be approximately 0.12867. The change amount (change _ratio) of the occupancy rate of the target lesion is 0.12867 - 0.15277, which is calculated to be approximately -0.0241.

[0088] In one embodiment, the processor can determine whether there is a change in the size of the target lesion by comparing the calculated change amount of the occupancy rate of approximately -0.0241 with a reference value. For example, when the calculated change amount of the occupancy rate is small compared to a reference value (e.g., a second reference value), the processor can determine that the size of the target lesion has decreased. Alternatively, when the calculated change amount of the occupancy rate is large compared to a first reference value and small compared to a second reference value, the processor can determine that there is no change in the size of the target lesion. As another example, the processor can input the calculated change amount of the occupancy rate (e.g., the difference between the first occupancy rate and the second occupancy rate) into a change determination model and determine whether there is a change in the size of the target lesion based on the output determination result. When the calculated change amount of the occupancy rate is large compared to a first reference value and small compared to a second reference value, the processor can determine that there is no change in the size of the target lesion. As another example, the processor can input the calculated change amount of the occupancy rate (e.g., the difference between the first occupancy rate and the second occupancy rate) into a change determination model and determine whether there is a change in the size of the target lesion based on the output determination result. By inputting the calculated change amount of the occupancy rate (e.g., the difference between the first occupancy rate and the second occupancy rate) into a change determination model the processor can determine whether there is a change in the size of the target lesion based on the output determination result.

[0089] In the case of the lungs, the size can change on X-ray images due to inhalation or exhalation. At this time, Since the size of the target lesion in the lung also changes accordingly, the ratio (i.e., occupancy rate) of the size of the target lesion to the size of the lung can be maintained. Therefore, such a change in the ratio can indicate a substantial change in the target lesion. In addition, multiple lesions contained in the lung can be detected separately, and the occupancy rate and / or the amount of change in the occupancy rate for each lesion can be calculated, so that the change in size for each lesion can be measured. Moreover, since the ratio change can indicate a substantial change in the target lesion, and multiple lesions contained in the lung can be detected separately, and the occupancy rate and / or the amount of change in the occupancy rate for each lesion can be calculated, the change in size for each lesion can be measured. Figure 9 is a diagram showing an example of determining the occupancy rate of a target lesion in a reference region according to an embodiment of the present disclosure and measuring the change in the size of the target lesion. The processor can determine (or calculate) the score for the first reference region 910 (e.g., the size of the first reference region, the number of pixels, etc.), the score for the second reference region 920 (e.g., the size of the second reference region, the number of pixels, etc.), the score for the target lesion 912 within the first reference region, and / or the score for the target lesion 922 within the second reference region. Then, the processor can calculate a change score based on the score for the first reference region 910, the score for the second reference region 920, the score for the target lesion 912 within the first reference region, and the score for the target lesion 922 within the second reference region using the Change Algorithm (900). For example, the processor can use a reference region extraction model to determine the reference regions 910 and 920 for each of the first X-ray image (taken at time t1) and the second X-ray image (taken at time t2, different from t1), and determine the scores for the reference regions.

[0090] Figure 9 is a diagram showing an example of determining the occupancy rate of a target lesion in a reference region according to an embodiment of the present disclosure and measuring the change in the size of the target lesion. The processor can determine (or calculate) the score for the first reference region 910 (e.g., the size of the first reference region, the number of pixels, etc.), the score for the second reference region 920 (e.g., the size of the second reference region, the number of pixels, etc.), the score for the target lesion 912 within the first reference region, and / or the score for the target lesion 922 within the second reference region. Then, the processor can calculate a change score based on the score for the first reference region 910, the score for the second reference region 920, the score for the target lesion 912 within the first reference region, and the score for the target lesion 922 within the second reference region using the Change Algorithm (900). For example, the processor can use a reference region extraction model to determine the reference regions 910 and 920 for each of the first X-ray image (taken at time t1) and the second X-ray image (taken at time t2, different from t1), and determine the scores for the reference regions. Figure 9 is a diagram showing an example of determining the occupancy rate of a target lesion in a reference region according to an embodiment of the present disclosure and measuring the change in the size of the target lesion. The processor can determine (or calculate) the score for the first reference region 910 (e.g., the size of the first reference region, the number of pixels, etc.), the score for the second reference region 920 (e.g., the size of the second reference region, the number of pixels, etc.), the score for the target lesion 912 within the first reference region, and / or the score for the target lesion 922 within the second reference region. Then, the processor can calculate a change score based on the score for the first reference region 910, the score for the second reference region 920, the score for the target lesion 912 within the first reference region, and the score for the target lesion 922 within the second reference region using the Change Algorithm (900). For example, the processor can use a reference region extraction model to determine the reference regions 910 and 920 for each of the first X-ray image (taken at time t1) and the second X-ray image (taken at time t2, different from t1), and determine the scores for the reference regions. Figure 9 is a diagram showing an example of determining the occupancy rate of a target lesion in a reference region according to an embodiment of the present disclosure and measuring the change in the size of the target lesion. The processor can determine (or calculate) the score for the first reference region 910 (e.g., the size of the first reference region, the number of pixels, etc.), the score for the second reference region 920 (e.g., the size of the second reference region, the number of pixels, etc.), the score for the target lesion 912 within the first reference region, and / or the score for the target lesion 922 within the second reference region. Then, the processor can calculate a change score based on the score for the first reference region 910, the score for the second reference region 920, the score for the target lesion 912 within the first reference region, and the score for the target lesion 922 within the second reference region using the Change Algorithm (900). For example, the processor can use a reference region extraction model to determine the reference regions 910 and 920 for each of the first X-ray image (taken at time t1) and the second X-ray image (taken at time t2, different from t1), and determine the scores for the reference regions. Figure 9 is a diagram showing an example of determining the occupancy rate of a target lesion in a reference region according to an embodiment of the present disclosure and measuring the change in the size of the target lesion. The processor can determine (or calculate) the score for the first reference region 910 (e.g., the size of the first reference region, the number of pixels, etc.), the score for the second reference region 920 (e.g., the size of the second reference region, the number of pixels, etc.), the score for the target lesion 912 within the first reference region, and / or the score for the target lesion 922 within the second reference region.

[0091] For example, the processor can use a reference region extraction model to determine the reference regions 910 and 920 for each of the first X-ray image (taken at time t1) and the second X-ray image (taken at time t2, different from t1), and determine the scores for the reference regions. Figure 9 is a diagram showing an example of determining the occupancy rate of a target lesion in a reference region according to an embodiment of the present disclosure and measuring the change in the size of the target lesion. The processor can determine (or calculate) the score for the first reference region 910 (e.g., the size of the first reference region, the number of pixels, etc.), the score for the second reference region 920 (e.g., the size of the second reference region, the number of pixels, etc.), the score for the target lesion 912 within the first reference region, and / or the score for the target lesion 922 within the second reference region. Then, the processor can calculate a change score based on the score for the first reference region 910, the score for the second reference region 920, the score for the target lesion 912 within the first reference region, and the score for the target lesion 922 within the second reference region using the Change Algorithm (900). In addition, the processor can determine a score for the target lesion 912 within the first reference region by using a target lesion detection model, a segmentation model, or the like. For this purpose, the processor can calculate a target lesion prediction score (or heat map value) for each of a plurality of pixels included in the first reference region. Similarly, the processor can calculate a target lesion prediction score for each of a plurality of pixels included in the second reference region by using a target lesion detection model, a segmentation model, or the like, thereby determining a score for the target lesion 922 within the second reference region.

[0092] In one embodiment, the processor determines a heat map value (e.g., a heat map value determined based on the target lesion prediction score) for each of a plurality of pixels included in the target lesion region within the reference region, and can calculate a score for the target lesion within each reference region as shown in Equation 4 below.

[0093]

Equation

[0094] Here, heat map represents the heat map value for the target lesion region, f(x) represents an arbitrary function for the x value, and output_score represents the score for the target lesion. For example, f(he at map) represents the sum of the heat map values for each of a plurality of pixels included in the target lesion region. As another example, f(heat map) represents the average value of the heat map values for each of a plurality of pixels included in the target lesion region.

[0095] Also, based on the calculated score, the processor can calculate a change score, such as in the following Equation 5. The processor can determine the presence or absence of a change in the size of the target lesion and / or the amount of change in size based on the calculated score. Here, output_score_t1 is the score for the target lesion within the first reference region, area_t1 is the score for the first reference region, output_score_t2 is the score for the target lesion within the second reference region, area_t2 is the score for the second reference region, and g(x, y, z, r) can represent an arbitrary function that calculates the first occupancy rate and the second occupancy rate based on x, y, z, r and calculates the change score of the size of the target lesion.

[0096]

Equation

[0097]

[0098] In one embodiment, the processor can calculate the first occupancy rate based on the score for the first reference region 910 and the score for the target lesion 912 within the first reference region, and calculate the second occupancy rate based on the score for the second reference region 920 and the score for the target lesion 922 within the second reference region. For example, the processor can calculate the first occupancy rate and the second occupancy rate using the following Equation 6 and Equation 7.

[0099]

Equation

[0100]

Equation

[0101] JPEG2025113359000009.jpg30162

[0102] Thereafter, based on the first occupancy rate and the second occupancy rate, the processor can determine the size of the target lesion. That is, based on the first occupancy rate and the second occupancy rate, the processor calculates a change score (e.g., a change score, a change amount of the occupancy rate, etc.), and based on the calculated change score, the processor can determine the presence or absence and / or the degree of change in the size of the target lesion. For example, the processor can calculate the change score using Equation 8 below.

[0103]

Equation

[0104] As another example, the processor can calculate the change score using Equation 9 below.

[0105]

Equation

[0106] As still another example, the processor can calculate the change score using Equation 10 below.

[0107]

Equation

[0108] JPEG2025113359000013.jpg13165

[0109] In FIG. 9, as an example of calculating a score for the target lesion, the target lesion within the reference region​ Although a heat map for a change is shown, it is not limited thereto. For example, the target To calculate a score for a target lesion, each target of a plurality of pixels included in the reference region Predicted values, probability maps, etc. for the target lesion are used.

[0110] FIG. 10 is a diagram showing an example of an artificial neural network model 1000 according to an embodiment of the present disclosure. The artificial neural network model 1000 is, as an example of a machine learning model, a statistical learning Algorithm embodied based on the structure of a biological neural network in machine learning (Machine Learning ) technology and cognitive science, or a structure that executes that algorithm.

[0111] According to one embodiment, the artificial neural network model 1000, like a biological neural network, has nodes such as artificial neurons that form a network by synaptic connections, and repeatedly adjusts the weighted values of the synapses so that the error between the correct output corresponding to a specific input and the inferred output decreases, and can show a machine learning model with problem-solving ability by learning. For example, the artificial neural network model 1000 can include any probability model, neural network model, etc. used in artificial intelligence learning methods such as machine learning and deep learning. For example, the artificial neural network model 1000 can include any probability model, neural network model, etc. used in artificial intelligence learning methods such as machine learning and deep learning. For example, the artificial neural network model 1000 can include any probability model, neural network model, etc. used in artificial intelligence learning methods such as machine learning and deep learning. can be.

[0112] According to one embodiment, the artificial neural network model 1000 can include an artificial neural network model configured to determine a reference region from an input X-ray image. Additionally or alternatively The artificial neural network model 1000 can include an artificial neural network model that identifies a target lesion from an input X-ray image. Additionally or alternatively, a first occupancy rate and a second occupancy rate The artificial neural network model 1000 can include an artificial neural network model that identifies a target lesion from an input X-ray image. Additionally or alternatively, a first occupancy rate and a second occupancy rate The artificial neural network model 1000 can include an artificial neural network model that identifies a target lesion from an input X-ray image. Additionally or alternatively, a first occupancy rate and a second occupancy rate Enter the difference between the two (e.g., the change in the occupancy rate for the target lesion) to measure the change in the size of the target lesion. The results of the judgment on whether or not the target lesion has changed (for example, whether or not the target lesion has increased, decreased, or remained unchanged) are output. The model may include an artificial neural network model configured as follows:

[0113] The artificial neural network model 1000 is a multi-layered network consisting of nodes and connections between them. The artificial neural network according to this embodiment is implemented as a multilayer perceptron (MLP). The neural network model 1000 is implemented using one of a variety of artificial neural network model structures, including MLP. As shown in FIG. 10, an artificial neural network model 1000 receives an input signal or data from the outside. an input layer 1020 that receives data 1010 and an output signal or data 1021 that corresponds to the input data; 050, and an output layer 1040 located between the input layer 1020 and the output layer 1040. n (herein) signals are received from the input layer 1020, characteristics are extracted, and the extracted characteristics are transmitted to the output layer 1040. The output layer consists of hidden layers 1030_1 to 1030_n (where n is a positive integer). 1040 receives signals from the hidden layers 1030_1 to 1030_n and outputs them to the outside.

[0114] The learning method of the artificial neural network model 110 involves inputting a teacher signal (correct answer) to solve a problem. There are two methods: a supervised learning method that learns to optimize the In one embodiment, there is an unsupervised learning method that does not require The information processing system outputs the reference region from the X-ray image using an artificial neural network model. 1000 supervised and / or unsupervised learning can be performed. For example, information processing The system can perform supervised learning of the artificial neural network model 1000 to output a reference region from an X-ray image using a plurality of reference X-ray images and label information regarding the reference region. Additionally or alternatively, the information processing system can perform supervised learning and / or unsupervised learning of the artificial neural network model 1000 to determine and output a reference region by dividing the entire region of the X-ray image (e.g., the lung region) into a plurality of regions. In other embodiments, the information processing system can perform supervised learning and / or unsupervised learning of the artificial neural network model 1000 to output a determination result regarding the presence or absence of a change in the size of a target lesion by inputting the difference between a first occupancy rate and a second occupancy rate into a change presence or absence determination model. For example, the information processing system can perform supervised learning of the artificial neural network model 1000 to output a determination result regarding the presence or absence of a change in the size of a reference target lesion based on an input value for a reference occupancy rate difference (e.g., the amount of change in the reference occupancy rate for a target lesion). The supervised learning of the artificial neural network model 1000 can be performed so as to output a reference region from the X-ray image using a plurality of reference X-ray images and label information regarding the reference region. Additionally or alternatively, the information processing system can perform supervised learning and / or unsupervised learning of the artificial neural network model 1000 to determine and output a reference region by dividing the entire region of the X-ray image (e.g., the lung region) into a plurality of regions. (For example, the lung region) into a plurality of regions so as to determine and output a reference region. The supervised learning and / or unsupervised learning of the artificial neural network model 1000 can be performed. In other embodiments, the information processing system can input the difference between the first occupancy rate and the second occupancy rate into a change presence or absence determination model to output a determination result regarding the presence or absence of a change in the size of the target lesion. By inputting the difference between the first occupancy rate and the second occupancy rate into a change presence or absence determination model, a determination result regarding the presence or absence of a change in the size of the target lesion can be output. The supervised learning and / or unsupervised learning of the artificial neural network model 1000 can be performed to output a determination result regarding the presence or absence of a change in the size of the target lesion. For example, the information processing system can output a determination result regarding the presence or absence of a change in the size of the reference target lesion based on an input value for the reference occupancy rate difference (e.g., the amount of change in the reference occupancy rate for the target lesion). Based on the input value for the reference occupancy rate difference (e.g., the amount of change in the reference occupancy rate for the target lesion), a determination result regarding the presence or absence of a change in the size of the reference target lesion can be output. The supervised learning of the artificial neural network model 1000 can be performed to output a determination result regarding the presence or absence of a change in the size of the reference target lesion. It is possible.

[0115] The artificial neural network model 1000 learned in this way can be stored in the memory of the information processing system (not shown), and in response to the input of data received from the communication module and / or the memory, it can determine a reference region and / or a target lesion region from the X-ray image and output the reference region and / or the target lesion region. Additionally or alternatively, the artificial neural network model 1000 can determine the presence or absence of a change in the size of the target lesion in response to an input regarding the occupancy rate difference of the target lesion (e.g., the amount of change in the occupancy rate for the target lesion) in a plurality of X-ray images, and output a determination result regarding the presence or absence of a change in the size of the target lesion. The supervised learning of the artificial neural network model 1000 can be performed so as to output a reference region from the X-ray image using a plurality of reference X-ray images and label information regarding the reference region. Additionally or alternatively, the information processing system can perform supervised learning and / or unsupervised learning of the artificial neural network model 1000 to determine and output a reference region by dividing the entire region of the X-ray image (e.g., the lung region) into a plurality of regions. In response to the input of data received from the communication module and / or the memory, it can determine a reference region and / or a target lesion region from the X-ray image and output the reference region and / or the target lesion region. Additionally or alternatively, the artificial neural network model 1000 can determine the presence or absence of a change in the size of the target lesion in response to an input regarding the occupancy rate difference of the target lesion (e.g., the amount of change in the occupancy rate for the target lesion) in a plurality of X-ray images, and output a determination result regarding the presence or absence of a change in the size of the target lesion. (For example, the amount of change in the occupancy rate for the target lesion) in a plurality of X-ray images. In response to the input regarding the occupancy rate difference of the target lesion (e.g., the amount of change in the occupancy rate for the target lesion) in a plurality of X-ray images, the presence or absence of a change in the size of the target lesion can be determined, and a determination result regarding the presence or absence of a change in the size of the target lesion can be output. It is possible to output a determination result regarding the presence or absence of a change in the size of the target lesion.

[0116] According to one embodiment, a machine learning model for determining a reference region, that is, an artificial neural network model The 1000 input variables can be one or more X-ray images. For example, the artificial neural network model The input variables input to the input layer 1020 of the 1000 can be an image vector 1010 formed by configuring one or more X-ray images as one vector data element. Depending on the input of the image The output variable output from the output layer 1040 of the artificial neural network model 1000 can be a vector 1050 indicating and / or characterizing the reference region and / or the target lesion region in the X-ray image image. Additionally, the output layer 1040 of the artificial neural network model 1000 can be configured to output a vector indicating and / or characterizing a plurality of regions obtained by dividing the entire region of the X-ray image At this time, the reference region can be a region including at least one of the plurality of regions. In the present disclosure, the output variable of the artificial neural network model 1000 is not limited to the types described above, and can include any information / data indicating the reference region and / or the target lesion region in the X-ray image In another embodiment, for a machine learning model for determining the presence or absence of a change in the size of a target lesion, that is, the input variables of the artificial neural network model 1000 can be the difference between the first occupancy rate and the second occupancy rate for the target lesion (for example, the numerical value obtained by subtracting the first occupancy rate from the second occupancy rate, the change amount of the occupancy rate for the target lesion ). For example, the input variables input to the input layer 1020 of the artificial neural network model 1000 can be a numerical vector 1010 formed by configuring the difference between the first occupancy rate and the second occupancy rate as one vector data element. Depending on the input of the difference between the first occupancy rate and the second occupancy rate The output variable output from the output layer 1040 of the artificial neural network model 1000 In the present disclosure, the output variable of the artificial neural network model 1000 is not limited to the types described above, and can include any information / data indicating the reference region and / or the target lesion region in the X-ray image In the present disclosure, the output variable of the artificial neural network model 1000 is not limited to the types described above, and can include any information / data indicating the reference region and / or the target lesion region in the X-ray image information / data.

[0117] In other embodiments, for a machine learning model for determining the presence or absence of a change in the size of a target lesion, that is, the input variables of the artificial neural network model 1000 can be the difference between the first occupancy rate and the second occupancy rate for the target lesion (for example, the numerical value obtained by subtracting the first occupancy rate from the second occupancy rate, the change amount of the occupancy rate for the target lesion ). For example, the input variables input to the input layer 1020 of the artificial neural network model 1000 can be the difference between the first occupancy rate and the second occupancy rate, configured as one vector data element to form a numerical vector 1010. Depending on the input of the difference between the first occupancy rate and the second occupancy rate for the target lesion, the output variable output from the output layer 1040 of the artificial neural network model 1000 can be a vector data element formed by configuring the difference between the first occupancy rate and the second occupancy rate as one vector data element to form a numerical vector 1010. Depending on the input of the difference between the first occupancy rate and the second occupancy rate The output variable output from the output layer 1040 of the artificial neural network model 1000 is a vector that indicates or characterizes the determination result regarding the presence or absence of a change in the size of the target lesion can be 1050. In the present disclosure, the output variable of the artificial neural network model 1000 is as described above is not limited to the types described above, and can include any information / data indicating the determination result regarding the presence or absence of a change in the size of the target lesion. Further, the output layer 104 0 of the artificial neural network model 1000 can be configured to output a vector indicating the reliability and / or accuracy regarding the output reference region, target lesion region, and the determination result regarding the presence or absence of a change in the size of the target lesion As described above, a plurality of input variables in the input layer 1020 and output variables corresponding thereto in the output layer 1040 of the artificial neural network model 1000 are each matched, and the synaptic values among the nodes included in the input layer 1020, hidden layers 10

[0118] 30_1 to 1030_n, and output layer 1040 are adjusted so that correct outputs corresponding to specific inputs can be learned and extracted. Through such a learning process, the hidden characteristics of the input variables of the artificial neural network model 1000 can be grasped, and the synaptic values (or weighted values) among the nodes of the artificial neural network model 1000 can be adjusted so that the error between the output variable calculated based on the input variable and the target output is reduced. Using the thus learned artificial neural network model 1000, information (e.g., position, size, pixel count, etc.) regarding the reference region and / or target lesion region can be output from the input X-ray image according to the input X-ray image. Additionally, using the artificial neural network model 1000, according to the change amount of the occupancy rate with respect to the input target lesion (e.g., the difference between the first occupancy rate and the second occupancy rate with respect to the target lesion) the determination result regarding the presence or absence of a change in the size of the target lesion can be output variable and the target output is reduced. Using the thus learned artificial neural network model 1000, information (e.g., position, size, pixel count, etc.) regarding the reference region and / or target lesion region can be output from the input X-ray image according to the input X-ray image. Additionally, using the artificial neural network model 1000, according to the change amount of the occupancy rate with respect to the input target lesion (e.g., the difference between the first occupancy rate and the second occupancy rate with respect to the target lesion) the determination result regarding the presence or absence of a change in the size of the target lesion can be output count, etc.) regarding the reference region and / or target lesion region can be output from the input X-ray image according to the input X-ray image. Additionally, using the artificial neural network model 1000, according to the change amount of the occupancy rate with respect to the input target lesion (e.g., the difference between the first occupancy rate and the second occupancy rate with respect to the target lesion) count, etc.) regarding the reference region and / or target lesion region can be output from the input X-ray image according to the input X-ray image. Additionally, using the artificial neural network model 1000, according to the change amount of the occupancy rate with respect to the input target lesion (e.g., the difference between the first occupancy rate and the second occupancy rate with respect to the target lesion) the determination result regarding the presence or absence of a change in the size of the target lesion can be output the determination result regarding the presence or absence of a change in the size of the target lesion can be output

[0119] The foregoing description of the disclosure is provided to enable those skilled in the art to practice or use the disclosure. Various modifications to the disclosure will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to various variations without departing from the spirit or scope of the disclosure. Accordingly, the disclosure is not intended to be limited to the examples described herein, but is intended to be accorded the broadest scope consistent with the principles and novel features disclosed herein. Thus, the disclosure is not intended to be limited to the examples described herein, but is intended to be accorded the broadest scope consistent with the principles and novel features disclosed herein. Thus, the disclosure is not intended to be limited to the examples described herein, but is intended to be accorded the broadest scope consistent with the principles and novel features disclosed herein. be possible.

[0120] Although the foregoing embodiments have been described as utilizing aspects of the presently disclosed subject matter in one or more stand-alone computer systems, the disclosure is not so limited and may be implemented in connection with any computing environment, such as a network or distributed computing environment. Accordingly, aspects of the subject matter of the disclosure may be implemented on multiple processing chips or devices, and storage may be similarly affected across multiple devices. These devices may include, for example, PCs, network servers, and portable devices. These devices may include, for example, PCs, network servers, and portable devices. These devices may include, for example, PCs, network servers, and portable devices. These devices may include, for example, PCs, network servers, and portable devices.

[0121] Although the disclosure has been described in connection with certain embodiments, various modifications and variations can be made without departing from the scope of the disclosure as will be understood by those of ordinary skill in the art to which the disclosure pertains. Moreover, such modifications and variations are to be considered within the scope of the claims appended hereto. Moreover, such modifications and variations are to be considered within the scope of the claims appended hereto. Moreover, such modifications and variations are to be considered within the scope of the claims appended hereto.

Claims

【Claim 1】 A method for measuring a change in the size of a target lesion in an X-ray image, which is performed by at least one computing device, comprising: receiving a first X-ray image including the target lesion and a second X-ray image including the target lesion; when dividing the entire area in each of the first X-ray image and the second X-ray image into a plurality of areas, calculating an occupancy rate of the area corresponding to the target lesion among the reference areas in each of the first X-ray image and the second X-ray image based on a prediction score for a determined reference area and a prediction score for the target lesion within the reference area; measuring whether there is a change in the size of the target lesion based on the calculated occupancy rate; outputting a determination result regarding whether there is a change in the size of the target lesion; The step of calculating the occupancy rate includes: determining a first reference area and a second reference area from each of the first X-ray image and the second X-ray image; identifying the target lesion from each of the first X-ray image and the second X-ray image; calculating a first occupancy rate based on a prediction score for the first reference area and a prediction score for the target lesion within the first reference area; calculating a second occupancy rate based on a prediction score for the second reference area and a prediction score for the target lesion within the second reference area; The step of measuring whether there is a change in the size of the target lesion includes: subtracting the first occupancy rate calculated from the first X-ray image from the second occupancy rate calculated from the second X-ray image to calculate a change amount of the occupancy rate; when the change amount of the occupancy rate is equal to or greater than a first reference value, determining that the size of the target lesion has increased; when the change amount of the occupancy rate is less than the first reference value and equal to or greater than a second reference value, determining that there is no change in the size of the target lesion; and when the change amount of the occupancy rate is less than the second reference value, determining that the size of the target lesion has decreased. A method for measuring a change in the size of a target lesion.

Images