KR102119057B1 - Method for learning fracture diagnosis model and apparatus for the same - Google Patents

Abstract

According to the present invention, a method for learning a fracture diagnosis model can be provided. The method for learning a fracture diagnosis model is to learn a learning model for diagnosing fractures by using a medical image. The method comprises the steps of: learning a general fracture learning model based on a fracture medical image obtained by photographing a fractured region in a body region, wherein a general fracture learning model is learned using the fracture medical image corresponding to various body regions; fixing the weight of an artificial neural network provided in the general fracture learning model, and constructing a part-unit fracture learning model of a structure in which a value output from the general fracture learning model is input into a characteristic learning model; and inputting a part-unit fracture medical image corresponding to a specific part of the body region into the general fracture learning model, and setting a corresponding fracture diagnosis result as a target variable of the characteristic learning model to perform learning on the part-unit fracture learning model. Therefore, high-performance learning models can be established using a small amount of labeling data.

Description

METHOD FOR LEARNING FRACTURE DIAGNOSIS MODEL AND APPARATUS FOR THE SAME}

The present disclosure relates to a deep learning model learning technique, and more particularly, to a method and apparatus for performing learning on fracture using medical images.

Deep learning is to learn a very large amount of data and, when new data is input, select the answer with the highest probability based on the learning result. Since deep learning can operate adaptively according to an image and automatically finds characteristic factors in the process of learning a model based on data, recent attempts to utilize it in the field of artificial intelligence are increasing.

However, in order for the trained model to derive accurate and accurate results, a large amount of data learning is required.

In particular, in order to apply artificial intelligence technology to the medical field, a large amount of data (ie, labeling data) verified by experts is required, but due to time and cost problems, large amounts of data verified by experts are built. There is a problem that is not easy to do.

In addition, in the region of the body region, where fractures or lesions are not frequently generated, there is a shortage of constructing a learning model using medical images due to a lack of medical images taken of the corresponding regions.

The technical problem of the present disclosure is to provide a learning method and apparatus capable of constructing a high-performance learning model using a small number of labeling data.

The technical problems to be achieved in the present disclosure are not limited to the technical problems mentioned above, and other technical problems that are not mentioned will be clearly understood by those skilled in the art from the description below. Will be able to.

According to an aspect of the present disclosure, a method for learning a fracture diagnostic model may be provided. The method is a method for learning a learning model for diagnosing a fracture using medical images, based on a fracture medical image photographing a region where a fracture occurs in a body region, but the fracture medical treatment corresponding to various body regions The process of learning a general fracture learning model using an image, fixing the weight of the artificial neural network provided in the general fracture learning model, and the value output from the general fracture learning model is a unit unit of the structure input to the feature learning model. A process of constructing a fracture learning model and a medical image of a unit-level fracture corresponding to a specific region in the body region are input to the general fracture learning model, and the corresponding fracture diagnosis result is set as a target variable of the feature learning model By doing so, it may include a process of performing the learning on the learning unit fracture model.

According to another aspect of the present disclosure, a fracture diagnosis model learning apparatus may be provided. The device is a device for learning a learning model for diagnosing fractures by using medical images. The device is based on a fracture medical image photographing an area where a fracture occurs in a body region, but the fracture medical image corresponding to various body regions Using a general fracture learning unit for learning a general fracture learning model using, and a general fracture learning model from the general fracture learning unit, fixing the weight of the artificial neural network provided in the general fracture learning model, and learning the general fracture It is equipped with a site-level fracture learning model management unit constituting a site-level fracture learning model having a structure in which values output from a model are input to a feature learning model, and the general fracture learning model and the feature learning model with fixed weights of the artificial neural network In the part-unit fracture learning model, a part-unit fracture medical image corresponding to a specific part of the body region is input, and a corresponding fracture diagnosis result is set as an objective variable of the feature learning model, and thus the part-unit fracture learning model It may include a unit-based fracture learning unit for performing learning.

The features briefly summarized above with respect to the present disclosure are merely illustrative aspects of the detailed description of the present disclosure described below, and do not limit the scope of the present disclosure.

According to the present disclosure, a method and apparatus for learning a lesion capable of constructing a high-performance learning model using a small number of labeling data may be provided.

The effects obtainable in the present disclosure are not limited to the above-mentioned effects, and other effects not mentioned may be clearly understood by those skilled in the art from the following description. will be.

1 is a block diagram showing the configuration of a fracture diagnosis model learning apparatus according to an embodiment of the present disclosure.
2A is a diagram illustrating a configuration of a data set used for learning a general fracture learning model according to an embodiment of the present disclosure.
2B to 2E are diagrams illustrating a medical image used as a set of learning data in FIG. 2A.
FIG. 3A is a diagram illustrating a configuration of a data set used for learning a part-unit fracture learning model according to an embodiment of the present disclosure.
3B to 3E illustrate a medical image used as a set of learning data in FIG. 3A.
FIG. 4 is a diagram illustrating a structure of a bone fracture learning model constructed by a fracture diagnosis model learning apparatus according to an embodiment of the present disclosure.
5 is a flowchart illustrating a sequence of a method for learning a fracture diagnosis model according to an embodiment of the present disclosure.
6 is a block diagram illustrating a computing system for executing a fracture diagnosis model learning apparatus and method according to an embodiment of the present disclosure.

Hereinafter, exemplary embodiments of the present disclosure will be described in detail with reference to the accompanying drawings so that those skilled in the art to which the present disclosure pertains can easily carry out the embodiments. However, the present disclosure can be implemented in many different forms and is not limited to the embodiments described herein.

In describing the embodiments of the present disclosure, when it is determined that a detailed description of known configurations or functions may obscure the subject matter of the present disclosure, detailed description thereof will be omitted. In the drawings, parts irrelevant to the description of the present disclosure are omitted, and similar reference numerals are used for similar parts.

In the present disclosure, when a component is said to be "connected", "coupled" or "connected" with another component, this is not only a direct connection relationship, but also an indirect connection relationship in which another component exists in the middle. It may also include. Also, when a component is said to "include" or "have" another component, this means that other components may be further included instead of excluding other components unless specifically stated to the contrary. .

In the present disclosure, terms such as first and second are used only for the purpose of distinguishing one component from other components, and do not limit the order, importance, or the like between components unless otherwise specified. Accordingly, within the scope of the present disclosure, the first component in one embodiment may be referred to as a second component in another embodiment, and likewise the second component in one embodiment may be the first component in another embodiment It can also be called.

In the present disclosure, components that are distinguished from each other are for clarity of each feature, and do not mean components are separated. That is, a plurality of components may be integrated to be composed of one hardware or software unit, or one component may be distributed to be composed of a plurality of hardware or software units. Accordingly, such integrated or distributed embodiments are included within the scope of the present disclosure, unless otherwise stated.

In the present disclosure, components described in various embodiments are not necessarily essential components, and some may be optional components. Accordingly, an embodiment consisting of a subset of components described in one embodiment is also included in the scope of the present disclosure. Also, embodiments including other elements in addition to the elements described in various embodiments are included in the scope of the present disclosure.

Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings.

1 is a block diagram showing the configuration of a fracture diagnosis model learning apparatus according to an embodiment of the present disclosure.

Referring to FIG. 1, the learning apparatus 10 of the fracture diagnosis model may include a general fracture learning unit 11, a site-based fracture learning model management unit 13, and a site-level fracture learning unit 15.

The general fracture learning unit 11 is a component that processes learning for the general fracture learning model 110, based on a convolutional neural network (CNN) technique or a pooling technique, related to the general fracture. The training for the general fracture learning model 110 is processed using the training data.

At this time, the general fracture learning model 110 is a learning model capable of diagnosing fractures occurring in various body regions using a fracture medical image (eg, an x-ray image) of the region where the fracture occurred in the body region. , It can be learned to detect global features related to fractures in fracture medical images. To this end, the general fracture learning unit 11 inputs a medical image (eg, an x-ray image) into the general fracture learning model 110 and sets an environment in which a task related to fracture in the medical image can be input as an objective variable. Can provide.

Furthermore, the tasks related to fracture can be variously performed, including a classification task to classify the state of a fracture, a fracture object detection task to detect an object having a fracture in a medical image, and a fracture to a medical image. It may include a segmentation (segmentation) task to extract the generated area.

The general fracture learning unit 11 may provide an environment in which the general fracture learning model 110 can be configured differently for each task. For example, the general fracture learning unit 11 may provide a menu or UI for selecting a type of task, and may set the general fracture learning model 110 according to the selected task through the menu or UI. In addition, the general fracture learning unit 11 may provide a menu or a UI through which target variables suitable for the selected task can be input, and the information input through the menu or UI is set as the target variable to enable the general fracture learning model ( 110).

The site-level fracture learning model management unit 13 combines the general fracture learning model 110 provided as a basic layer of the artificial neural network and the feature learning model 151 provided as an extension layer of the artificial neural network, thereby increasing the site-level fracture learning model 150 ), in which case, the general fracture learning model 110 may be set to precede the feature learning model 151.

Furthermore, since the unit-level fracture learning model management unit 13 uses the weight of the artificial neural network provided in the general fracture learning model 110 when the learning of the general fracture learning model 110 reaches a predetermined level, the general fracture is used. There is no need to proceed with learning of the learning model 110 any more. In consideration of this, the site-level fracture learning model management unit 13 may provide a control command so that the learning of the general fracture learning model 110 does not proceed to the general fracture learning unit 11.

As another example, the general fracture learning unit 11 continuously learns about the general fracture learning model 110 and periodically updates the artificial neural network provided in the general fracture learning model 110 to study the unit-level fracture learning model. It is also possible to configure 150.

The regional unit fracture learning model management unit 13 may provide the regional unit fracture learning model 150 configured as described above to the regional unit fracture learning unit 15.

The unit-level fracture learning unit 15 may perform learning on the unit-unit fracture learning model 150 configured in the unit-unit fracture learning model management unit 13, and the unit-unit fracture learning model 150 is a unit-unit medical image. It can be configured to receive the input and detect the fracture of the unit corresponding to the. To this end, the unit-unit fracture learning unit 15 receives a medical image (hereinafter referred to as a “unit-unit medical image”) of a specific region of the body region (eg, a region where the ribs are located), and a unit-unit medical treatment. It is possible to provide an environment in which an objective variable reflecting a task related to a fracture can be input in an image.

As described above, the tasks related to the fracture may be variously performed, a classification task to classify the state of the fracture, a fracture object detection task to detect an object having a fracture in a medical image, and a fracture to occur in a medical image It may include a segmentation task to extract the area.

The site-based fracture learning unit 15 may provide an environment in which the site-based fracture learning model 150 can be configured differently for each task. For example, the site-based fracture learning unit 15 may provide a menu or UI for selecting a type of task, and the site-based fracture learning model 150 may be set according to the selected task through the menu or the UI. . In addition, the site-based fracture learning unit 15 may provide a menu or UI through which a target variable suitable for a selected task can be input, and the information input through the menu or UI is set as a target variable to learn the site-level fracture. The model 150 can be trained. For example, the unit-level fracture learning unit 15 may input information that classifies the state of a fracture, information designating an object in which a fracture occurs, information designating an area in which a fracture occurs, as an objective variable.

The deeper the artificial neural network, the more specialized features can be extracted for a task. As described above, the general fracture learning model 110 provided as a basic layer of the artificial neural network is configured in a fixed form. Since the feature learning model 151 provided as an extension layer is learned on a densenet basis, the feature learning model 151 provides a feature specialized to fracture of a specific region (eg, a region where the ribs are located) in the body region. It is configured to be extracted.

Furthermore, the site-based fracture learning unit 15 may further include a loss function calculation unit 16 for calculating a loss function, and a site in consideration of the loss function provided by the loss function calculation unit 16 The unit fracture learning model 150 may be trained.

For example, the learning section 15 is part units fracture to calculate a first loss function (Loss function _step1) and features the second loss function for a learning model (151) (Loss function _step2) for common fractures learning model 110 The final loss function may be calculated by applying weights to the calculated first and second loss functions (Loss function _step1 and Loss function _step2 ). Final loss of function units fracture site learning section 15 calculates the (Loss function _end) can be achieved through the operation of Equation 1 below.

Here, W1 and W2 are weights applied to the first and second loss functions, respectively, and may be set to a value proportional to the number of layers provided in the general fracture learning model 110 and the feature learning model 151.

2A illustrates a configuration of a data set used for learning a general fracture learning model according to an embodiment of the present disclosure, and FIGS. 2B to 2E illustrate medical images used as the learning data set of FIG. 2A.

First, referring to FIG. 2A, the first learning data set 200 may include at least one fracture medical image 201, 202, 203, and at least one fracture medical image 201, 202, 203. May be a medical image (eg, X-ray) of a user's body with a fracture.

In addition, the first learning data set 200 may include a plurality of labeling data (211, 212, 213, ..., 221, 222, 223, ..., 231, 232, 233, ...) have. Labeling data (211, 212, 213, ..., 221, 222, 223, ..., 231, 232, 233, ...) can be classified according to each task, and a general fracture learning model ( 110) may be provided as an objective variable of the general fracture learning model 110 in accordance with the task selected during learning. For example, when a classification task is selected during learning of the general fracture learning model 110, the labeling data 211, 212 of the first task corresponding to the input of at least one fracture medical image 201, 202, 203 , 213) may be set and provided as an objective variable. When the fracture object detection task is selected when learning the general fracture learning model 110, the labeling data 221 of the second task corresponding to the input of the at least one fracture medical image 201, 202, 203, 222, 223) may be set and provided as an objective variable. Similarly, when a segmentation task is selected during learning of the general fracture learning model 110, the labeling data 231, 232 of the third task corresponding to the input of at least one fracture medical image 201, 202, 203 , 233) can be set and provided as an objective variable.

3A illustrates a configuration of a data set used for learning a region-based fracture learning model according to an embodiment of the present disclosure, and FIGS. 3B to 3E illustrate medical images used as the learning data set of FIG. 3A. .

Referring to FIG. 3A, the second learning data set 310 may include at least one partial unit fracture medical image 301,302, 303, and at least one partial unit fracture medical image 301,302, 303 is a user It may be a medical image (eg, X-ray) of a state in which a fracture occurs in a specific area of the body. For example, a specific region may be a region where ribs are present.

And, the second learning data set 300 may include a plurality of labeling data (311, 312, 313, ..., 321, 322, 323, ..., 331, 332, 333, ...) have. Labeling data (311, 312, 313, ..., 321, 322, 323, ..., 331, 332, 333, ...) can be classified in response to each task, and learning model of fractures in each unit In accordance with the task selected during the learning of 150, it may be provided as an objective variable of the learning unit fracture model 150 for each unit. For example, when a classification task is selected during learning of the unit-level fracture learning model 150, the labeling data of the first task corresponding to the input of at least one site-level fracture medical image 301, 302, 303 ( 311, 312, 313) can be set and provided as the target variable. When the fracture object detection task is selected, the labeling data 321, 322, 323 of the second task corresponding to the input of the at least one site-level fracture medical image 301, 302, 303 is the target variable. Can be set and provided. Similarly, when the segmentation task is selected, the labeling data 331, 332, 333 of the third task corresponding to the input of at least one site-level fracture medical image 301, 302, 303 is set as the target variable And can be provided.

FIG. 4 is a diagram illustrating a structure of a bone fracture learning model configured by a fracture diagnosis model learning apparatus according to an embodiment of the present disclosure.

First, among the body regions, the region where the ribs are located may not be accurately detected in the fracture medical image (eg, x-ray image), so it is necessary to learn the characteristics of the rib fracture to more accurately detect fractures in the region where the ribs are located. Do. However, the rib fracture medical image may be relatively small compared to the general fracture medical image, and when a learning model is trained using only a small amount of the rib fracture medical image, problems such as overfiiting may occur. In addition, although a large amount of data is needed for the learning model to more accurately detect rib fractures, it is difficult to establish a specialized learning model for rib fractures because it is difficult to secure large amounts of data with special characteristics.

In consideration of this, the site-based fracture learning model management unit according to an embodiment of the present disclosure learns fractures occurring in various areas of the body to build a basic layer 410 of an artificial neural network, and then, for example, a specific area, for example, a rib The extension layer 420 is constructed by additionally learning the located area.

Specifically, the site-level fracture learning model management unit 13 according to an embodiment of the present disclosure preferentially trains the general fracture learning model 410 capable of detecting a global feature related to the fracture. After performing, the general fracture learning model 410 is provided to construct the unit-level fracture learning model 400. That is, the site-level fracture learning model management unit 13 preferentially advances the general fracture learning model 410 to construct the general fracture learning model 410, and weights the artificial neural network provided in the general fracture learning model 410. It can be fixed to be provided at the front end of the part-unit fracture learning model 400.

Subsequently, the site-level fracture learning model management unit 13 may configure an extension layer of the artificial neural network by combining the feature learning model 420 at the rear end of the general fracture learning model 410 provided as a basic layer of the artificial neural network. . Based on this structure, the unit-level fracture learning model 400 receives an input of a unit-level medical image, detects additional features from information output preferentially through the general-fracture learning model 410, and detects the detected elements. The feature learning model 420 may be constructed by applying predetermined weights to the fields.

5 is a flowchart illustrating a sequence of a method for learning a fracture diagnosis model according to an embodiment of the present disclosure.

The fracture diagnostic model learning method according to an embodiment of the present disclosure may be performed by the fracture diagnostic model learning apparatus described above.

First, in step S501, the fracture diagnosis model learning apparatus processes learning for the general fracture learning model. For example, the fracture diagnostic model learning apparatus processes learning for the general fracture learning model 110 related to the general fracture based on a convolutional neural network (CNN) technique or a pooling technique. At this time, the fracture diagnostic model learning apparatus may use the training data set illustrated in FIGS. 2A to 2E as input data or a target variable of the general fracture learning model.

Furthermore, the tasks related to fracture can be variously performed, including a classification task to classify the state of a fracture, a fracture object detection task to detect an object having a fracture in a medical image, and a fracture to a medical image. It may include a segmentation (segmentation) task to extract the generated area.

The fracture diagnosis model learning apparatus may provide an environment in which a general fracture learning model can be configured differently for each task. For example, the fracture diagnosis model learning apparatus may provide a menu or a UI for selecting a type of task, and a general fracture learning model may be set according to the selected task through the menu or the UI. In addition, the fracture diagnosis model learning apparatus may provide a menu or a UI for inputting a target variable suitable for a selected task, and setting information input through such a menu or UI as a target variable to learn about a general fracture learning model. You can do

Through the step S501, the general fracture learning model learns information about fractures occurring in various body regions, and thus can be trained to detect global features related to fractures in fracture medical images.

Of the body regions, the region where the ribs are located may not be accurately detected in the fracture medical image (eg, x-ray image), so it is necessary to learn the characteristics of the rib fracture to more accurately detect fractures in the region where the ribs are located. However, the rib fracture medical image may be relatively small compared to the general fracture medical image, and when a learning model is trained using only a small amount of the rib fracture medical image, problems such as overfiiting may occur. In addition, although a large amount of data is needed for the learning model to more accurately detect rib fractures, it is difficult to establish a specialized learning model for rib fractures because it is difficult to secure large amounts of data with special characteristics.

In consideration of this, in step S502, the fracture diagnosis model learning apparatus can construct a fracture model for each unit by combining a general fracture learning model provided as a basic layer of the artificial neural network and a feature learning model provided as an extension layer of the artificial neural network. have. At this time, the fracture diagnosis model learning apparatus may be set such that the general fracture learning model is relatively advanced compared to the feature learning model.

Subsequently, in step S503, the fracture diagnosis model learning apparatus may perform learning on a fracture model for each unit. Based on the above-described structure, the fracture diagnosis model learning apparatus receives an input of a medical image of a unit of a unit, detects additional features from information preferentially output through the general fracture learning model, and weights the detected elements with a predetermined weight. You can build a feature learning model by applying.

As described above, a task related to fracture may be variously performed, a classification task to classify the state of a fracture, a fracture object detection task to detect an object having a fracture in a medical image, and an area where a fracture has occurred in a medical image. It may include a segmentation task to extract.

The fracture diagnosis model learning apparatus may construct a fracture model for each unit differently for each task. For example, the fracture diagnosis model learning apparatus may provide a menu or UI for selecting a type of task, and a fracture model for each unit may be set according to the selected task through the menu or UI. In addition, the fracture diagnosis model learning apparatus may provide a menu or a UI for inputting a target variable suitable for a selected task, and setting information input through the menu or UI as a target variable to set a fracture model for a unit-level fracture learning model. Learning can be performed. For example, the fracture diagnosis model learning apparatus may input, as an objective variable, information classifying the state of a fracture, information specifying an object in which a fracture occurred, information specifying an area in which a fracture occurred.

The deeper the artificial neural network, the more specialized features can be extracted for the task. As described above, the general fracture learning model provided as a basic layer of the artificial neural network is constructed in a fixed form and expanded. Since the feature learning model provided as a layer is learned based on a densenet, the feature learning model is configured to extract a feature specialized for fracture of a specific region (eg, a region where ribs are located) in the body region.

Furthermore, the fracture diagnosis model learning apparatus may calculate a loss function (S504), and update the fracture model for each unit in consideration of the calculated loss function (S505).

For example, in step S504, the fracture diagnosis model learning apparatus may calculate a first loss function for a general fracture learning model (Loss function _step1 ) and a second loss function for a feature learning model (Loss function _step2 ), and is calculated. The final loss function may be calculated by applying weights to the first and second loss functions (Loss function _step1 and Loss function _step2 ). The final loss function (Loss function _final ) may be calculated through the operation of Equation 1 described above.

Steps S503 to S505 may be repeatedly performed until the operation of the fracture diagnosis model learning apparatus is finished.

6 is a block diagram illustrating a computing system for executing a fracture diagnosis model learning apparatus and method according to an embodiment of the present disclosure.

Referring to FIG. 6, the computing system 1000 includes at least one processor 1100 connected through a bus 1200, a memory 1300, a user interface input device 1400, a user interface output device 1500, and storage 1600, and the network interface 1700.

The processor 1100 may be a central processing unit (CPU) or a semiconductor device that executes processing for instructions stored in the memory 1300 and/or storage 1600. The memory 1300 and the storage 1600 may include various types of volatile or nonvolatile storage media. For example, the memory 1300 may include read only memory (ROM) and random access memory (RAM).

Thus, steps of a method or algorithm described in connection with the embodiments disclosed herein may be directly implemented by hardware, software modules, or a combination of the two, executed by processor 1100. The software modules reside in storage media (ie, memory 1300 and/or storage 1600) such as RAM memory, flash memory, ROM memory, EPROM memory, EEPROM memory, registers, hard disk, removable disk, CD-ROM. You may. An exemplary storage medium is coupled to the processor 1100, which can read information from and write information to the storage medium. Alternatively, the storage medium may be integral with the processor 1100. Processors and storage media may reside within an application specific integrated circuit (ASIC). The ASIC may reside in a user terminal. Alternatively, the processor and storage medium may reside as separate components within the user terminal.

Exemplary methods of the present disclosure are expressed as a series of operations for clarity of description, but are not intended to limit the order in which the steps are performed, and each step may be performed simultaneously or in a different order if necessary. In order to implement the method according to the present disclosure, the steps illustrated may include other steps in addition, other steps may be included in addition to the other steps, or other additional steps may be included in addition to some steps.

The various embodiments of the present disclosure are not intended to list all possible combinations, but are intended to illustrate representative aspects of the present disclosure, and the details described in various embodiments may be applied independently or in combination of two or more.

Further, various embodiments of the present disclosure may be implemented by hardware, firmware, software, or a combination thereof. For implementation by hardware, one or more Application Specific Integrated Circuits (ASICs), Digital Signal Processors (DSPs), Digital Signal Processing Devices (DSPDs), Programmable Logic Devices (PLDs), Field Programmable Gate Arrays (FPGAs), Universal It can be implemented by a processor (general processor), a controller, a microcontroller, a microprocessor.

The scope of the present disclosure includes software or machine-executable instructions (eg, operating systems, applications, firmware, programs, etc.) that cause actions according to various embodiment methods to be executed on a device or computer, and such software or Instructions include a non-transitory computer-readable medium that is stored and executable on a device or computer.

Claims (14)

An electronic device including at least one processor, a user interface input device, and a user interface output device, in a method of learning a learning model for diagnosing a fracture using medical images,
The at least one processor is based on a fracture medical image of a region in which a fracture has occurred in a body region, and using the fracture medical image corresponding to various body regions to learn a general fracture learning model,
The process in which the at least one processor fixes the weight of the artificial neural network provided in the general fracture learning model, and constructs a partial unit fracture learning model having a structure in which a value output from the general fracture learning model is input to a feature learning model. and,
The at least one processor inputs a unit-level fracture medical image corresponding to a specific region of the body region into the general fracture learning model, and sets the corresponding fracture diagnosis result as a target variable of the feature learning model, Includes the process of performing the learning on the fracture model for each unit,
The process of performing the learning on the part-unit fracture learning model is
The first loss function is calculated through the operation of Equation 1 below, but is applied to the first loss function and the second loss function, respectively, based on the number of layers included in the general fracture learning model and the feature learning model. And determining a second weight value.
[Equation 1]

Here, Loss function _final represents the final loss function, Loss function _step1 represents the first loss function corresponding to the common fractures learning models, Loss function _step2 denotes the second loss function corresponding to the feature learning models , w1 represents the first weight value corresponding to the general fracture learning model, and w2 represents the second weight value corresponding to the feature learning model. According to claim 1,
The process of performing the learning on the part-unit fracture learning model is
A process in which the general fracture learning model provides a first result corresponding to the medical image of the unit-level fracture,
The learning method of a fracture diagnosis model, characterized in that the learning model comprises the step of learning the fracture diagnosis result corresponding to the first result. According to claim 1,
The process of performing the learning on the part-unit fracture learning model is
Calculating the first loss function for the general fracture learning model;
Calculating the second loss function for the feature learning model,
And calculating a final loss function by applying the first and second weights to the first loss function and the second loss function, respectively.
delete According to claim 1,
The process of determining the weight applied to the first loss function and the second loss function,
A method of learning a fracture diagnosis model, characterized in that the weight applied to the first loss function and the second loss function is determined in proportion to the number of layers included in the general fracture learning model and the feature learning model. According to claim 1,
The process of performing the learning on the part-unit fracture learning model is
The process of confirming the type of task for learning the learning unit fracture model, and
Setting the target variable according to the identified task type,
And learning the feature learning model by reflecting the determined target variable. The method of claim 6,
The task type is,
It includes the classification of the state of the fracture step by step (classification), a fracture object detection (object detection) to detect the object where the fracture is present, and a segmentation (segmentation) to extract the region where the fracture is present. A method for learning a fracture diagnostic model. In the apparatus for learning a learning model for diagnosing a fracture using medical imaging,
Based on a fracture medical image of a region where a fracture occurs in a body region, a general fracture learning unit for learning a general fracture learning model using the fracture medical image corresponding to various body regions,
A general fracture learning model is provided from the general fracture learning unit, a weight of an artificial neural network provided in the general fracture learning model is fixed, and a value output from the general fracture learning model is a unit of a structure input to the feature learning model. Fracture unit learning model management unit constituting the fracture learning model,
A partial unit fracture medical image corresponding to a specific portion of the body region is input to a partial unit fracture learning model provided with the general fracture learning model and the feature learning model with fixed weights of the artificial neural network, and corresponding thereto The fracture diagnosis unit includes a site-level fracture learning unit configured to set a fracture diagnosis result as an objective variable of the feature learning model to perform learning on the site-level fracture learning model,
The site unit fracture learning unit,
A final loss function is calculated through the operation of Equation 1 below, and the learning of the fracture model for each unit is performed by reflecting the final loss function, but the layer included in the general fracture learning model and the feature learning model And a loss function calculation unit for determining first and second weights applied to the first loss function and the second loss function, respectively, based on the number.
[Equation 1]

Here, Loss function _final represents the final loss function, Loss function _step1 represents the first loss function corresponding to the common fractures learning models, Loss function _step2 denotes the second loss function corresponding to the feature learning models , w1 represents the first weight value corresponding to the general fracture learning model, and w2 represents the second weight value corresponding to the feature learning model. The method of claim 8,
The site unit fracture learning unit,
Set the medical image of the fracture of the site unit as an input to the general fracture learning model,
Set the first result output through the general fracture learning model as an input to the feature learning model,
A device for learning a fracture diagnosis model, characterized in that the fracture diagnosis result is set as a target variable. delete delete The method of claim 8,
The loss function calculation unit,
A learning apparatus for a fracture diagnosis model, characterized in that the weight applied to the first loss function and the second loss function is determined in proportion to the number of layers included in the general fracture learning model and the feature learning model. The method of claim 8,
The site unit fracture learning unit,
Check the task type for learning the part-unit fracture learning model,
Set the target variable according to the identified task type,
A learning apparatus for a fracture diagnosis model, characterized by performing learning on the feature learning model by reflecting the determined target variable. The method of claim 13,
The task type is,
It includes the classification of the state of the fracture step by step (classification), a fracture object detection (object detection) to detect the object where the fracture is present, and a segmentation (segmentation) to extract the region where the fracture is present. A learning device for a fracture diagnostic model.

Images